Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Harmonizing water footprint assessments for agricultural production in Southern Amazonia Lathuillière, Michael Jacques 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2018_may_lathuilliere_michael.pdf [ 36.52MB ]
JSON: 24-1.0365279.json
JSON-LD: 24-1.0365279-ld.json
RDF/XML (Pretty): 24-1.0365279-rdf.xml
RDF/JSON: 24-1.0365279-rdf.json
Turtle: 24-1.0365279-turtle.txt
N-Triples: 24-1.0365279-rdf-ntriples.txt
Original Record: 24-1.0365279-source.json
Full Text

Full Text

Harmonizing water footprintassessments for agricultural productionin Southern AmazoniabyMichael Jacques LathuillièreB.Sc Honours, The University of British Columbia, 2002M.Sc, The University of British Columbia, 2011A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinThe Faculty of Graduate and Postdoctoral Studies(Resources, Environment and Sustainability)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)April 2018© Michael Jacques Lathuillière 2018AbstractSince its inception in 2002, the water footprint (WF) has brought new insight into the direct and indirect(or supply chain) uses of water in the production and consumption of goods and services. Today, thisemerging field is mainly represented by two distinct communities following distinct WF approaches: thewater resources management community which follows guidelines from the WF Network, and the lifecycle assessment (LCA) community which focuses on assessing impacts from water use. This thesisseeks to harmonize WF assessments by combining and contrasting methods and objectives from bothcommunities with the overarching goal of informing water decision-making by (1) considering limits towater resources within a river basin (the “Nature” domain), and (2) considering water use in productionsystems (the “Production” domain). Following this proposed framework to combine WF assessments(Chapter 2), I assess how each approach may address water management for agricultural productionin Southern Amazonia (Mato Grosso, Brazil), a region that has dramatically increased its soybean andcattle production through deforestation. In Chapters 3 and 4, I respectively measure andmodel theWF ofcropland and cattle to highlight on-farm water use strategies for agricultural production (the volumetricWF assessment phase). Chapter 5 focuses on the Xingu Basin of Mato Grosso for which I assesswater scarcity of current and future agricultural production (the volumetric WF sustainability assessmentphase). Finally, in Chapter 6, I integrate existing water use in LCA methods to highlight water useefficiencies through impact assessment (the WF impact assessment phase). Results show differentland and water management options for crops and cattle in Southern Amazonia, but also demonstratethat water use for future production could reach sustainable limits, should cropland irrigation and cattleconfinement become more widespread. Moreover, the role of water vapour supply to the atmospherethrough evapotranspiration is stressed as an important process that could affect future water availabilitydue to the importance of moisture recycling on regional precipitation. This research provides contexton the role of land management on water resources, while combining water decisions affecting bothproduction systems and resource limits imposed by the water cycle.iiLay SummaryAs an indicator of direct and indirect (or supply chain) freshwater use, the water footprint has been de-scribed and used differently by two research communities: as a tool to quantify the appropriation andexchange of freshwater resources through products and services, and as a step towards estimatingpotential impacts within a life cycle assessment. This thesis proposes a framework to combine bothperspectives before applying a new harmonized water footprint assessment to soybean and cattle pro-duction in Southern Amazonia, a region of historically high deforestation rates. While both researchcommunities may see the water footprint though different lenses, the combination of their respectivemethods provides valuable information on water resources decision-making. Results suggest strategiesto reduce water use at field and product levels to minimize water use and impacts to the water cycle,and provide guidance on land and water management for production to promote sustainable practices.iiiPrefaceChapter 1 and Chapter 7 present the Introduction and Conclusion of the work described in the thesis.They are the sole work of the author. Chapters 2 to 6 were written by the author with the help of co-authors and colleagues.Chapter 2 is a literature review in which I devised the details of the conceptual framework with inputfrom Drs Cécile Bulle (Université du Québec à Montréal, UQAM) and Mark Johnson (University of BritishColumbia, UBC).Chapter 3 constitutes the findings of field work which I carried out at Capuaba farm (Lucas do RioVerde, Mato Grosso) thanks to the kind collaboration of farm owner José Eduardo de Macedo Soares Jrand the farm staff. The field work was completed with help from Dr Higo Dalmagro of the Universidadede Cuiabá (UNIC, Cuiabá, Mato Grosso). Instrumental support was provided by Dr Paulo Arruda of theUniversidade Federal deMato Grosso (UFMT, Cuiabá, Mato Grosso) as well as Dr Iain Hawthorne (UBC,Vancouver). Drs Andrew Black and Mark Johnson provided key input on data analysis and revision ofthe chapter, while laboratory space was kindly provided by Dr Eduardo Couto (UFMT). All co-authorsprovided feedback on the chapter which was accepted for publication on 25 March 2018: Lathuillière, M.J., H. J. Dalmagro, T. A. Black, P. H. Z. de Arruda, I. Hawthorne, E. G. Couto, M. S. Johnson, Rain-fed andirrigated cropland-atmosphere water fluxes and their implications for agricultural production in SouthernAmazonia, Agricultural and Forest Meteorology, in press, doi: 10.1016/j.agrformet.2018.03.023, 2018.Chapter 4 combines work done in collaboration with Kylen Solvik and Dr Marcia Macedo of theWoods Hole Research Center (WHRC, USA), as well as Dr Jordan Graesser (Boston University, USA).These collaborators kindly provided GIS maps and support for determining reservoir size and locationin Mato Grosso as well as feedback on the interpretation of results. Moreover, Drs Eduardo Miranda,Eduardo Couto and Mark Johnson helped gather statistical data and provided feedback on the interpre-tation of results. All co-authors provided feedback on the chapter.Chapter 5 is the result of a collaboration with Drs Michael Coe, Andrea Castanho, Jordan Graesserand Mark Johnson. Drs Michael Coe and Andrea Castanho (WHRC) made available all the hydrologi-ivcal modeling results (Integrated BIosphere Simulator, IBIS), while Drs Michael Coe, Andrea Castanho(WHRC) and Mark Johnson gave input and feedback on interpretation of results. Dr Jordan Graesserprovided land use maps for the basin which were used for modeling. All co-authors provided feedbackon the chapter which was published on 21 March 2018: Lathuillière, M. J., M. T. Coe, A. Castanho, J.Graesser, and M. S. Johnson, Evaluating water use for agricultural intensification in Southern Amazo-nia using the Water Footprint Sustainability Assessment, Water, 10(4), 329, doi: 10.3390/w10040349,2018.Chapter 6 is the result of a collaboration with Drs Cécile Bulle and Mark Johnson who provided inputon data analysis and interpretation as well as feedback on the chapter.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviList of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiiList of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxivAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxixDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxxi1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Water use for agricultural production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 The water footprint: an emerging research field . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Objective and research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.4 Land and water management in Southern Amazonia . . . . . . . . . . . . . . . . . . . . . 51.5 Significance and outline of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 The Harmonized Water Footprint Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . 102.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.2 Nature, Society and Production domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.3 Linking Nature and Production through Society with water footprints . . . . . . . . . . . 15vi2.3.1 The volumetric water footprint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.3.2 The volumetric water footprint assessment . . . . . . . . . . . . . . . . . . . . . . 172.3.3 The water footprint impact assessment . . . . . . . . . . . . . . . . . . . . . . . . 182.3.4 The volumetric water footprint sustainability assessment . . . . . . . . . . . . . . 192.3.5 Implications of the proposed harmonized water footprint assessment . . . . . . 202.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253 Measuring the Volumetric Water Footprint of Crops through Water Productivity . . . . 263.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.2 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.2.1 Site description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.2.2 Crop and canopy monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.2.3 Eddy covariance data processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.2.4 Eddy covariance latent heat flux gap-filling . . . . . . . . . . . . . . . . . . . . . . 343.2.5 Soybean and maize crop coefficients and crop canopy conductance . . . . . . . 353.2.6 Water productivity assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383.3.1 Cropland evapotranspiration of rain-fed and irrigated fields . . . . . . . . . . . . 383.3.2 Water balance of rain-fed cropland . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.3.3 Soybean and maize development cycles . . . . . . . . . . . . . . . . . . . . . . . 443.3.4 Changes in crop transpiration and water productivity . . . . . . . . . . . . . . . . 453.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.4.1 Water vapour supply of rain-fed and irrigated cropland to the atmosphere . . . . 473.4.2 Crop evapotranspiration and water productivity . . . . . . . . . . . . . . . . . . . . 483.4.3 Regional implications for land and water management . . . . . . . . . . . . . . . 503.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524 Modeling the Volumetric Water Footprint of Cattle . . . . . . . . . . . . . . . . . . . . . . . 534.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 554.2.1 Cattle production in Mato Grosso . . . . . . . . . . . . . . . . . . . . . . . . . . . . 554.2.2 Volumetric water footprint of cattle production . . . . . . . . . . . . . . . . . . . . 574.2.2.1 Animal water consumptive use . . . . . . . . . . . . . . . . . . . . . . . . 57vii4.2.2.2 Feed water consumptive use . . . . . . . . . . . . . . . . . . . . . . . . . 584.2.2.3 Water consumption from evaporation of farm impoundments . . . . . . 594.2.3 Living herd population and annual water consumptive use . . . . . . . . . . . . . 604.2.4 Land footprint of cattle production . . . . . . . . . . . . . . . . . . . . . . . . . . . 614.2.5 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624.3.1 Cattle volumetric water, land and carbon footprints . . . . . . . . . . . . . . . . . 624.3.2 Evolution of land and water for cattle in Mato Grosso . . . . . . . . . . . . . . . . 654.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 684.4.1 On-farm land and water appropriation for cattle . . . . . . . . . . . . . . . . . . . 684.4.2 Land and water appropriation for feed . . . . . . . . . . . . . . . . . . . . . . . . . 704.4.3 Comparison to literature values and research limitations . . . . . . . . . . . . . . 714.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 725 Volumetric Water Footprint Sustainability Assessment in the Xingu Basin . . . . . . . . 745.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 745.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 755.2.1 The Xingu Basin of Mato Grosso . . . . . . . . . . . . . . . . . . . . . . . . . . . . 755.2.2 Integrated BIosphere Simulator (IBIS) . . . . . . . . . . . . . . . . . . . . . . . . . 765.2.3 Volumetric water footprint sustainability assessment . . . . . . . . . . . . . . . . 795.2.3.1 Goal and scope definition . . . . . . . . . . . . . . . . . . . . . . . . . . . 795.2.3.2 Water footprint accounting . . . . . . . . . . . . . . . . . . . . . . . . . . 795.2.3.3 Water scarcity calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . 825.2.3.4 Interpretation and response formulation through scenarios . . . . . . . 835.2.4 Data processing and sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . 845.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 865.3.1 Past and future volumetric water footprints . . . . . . . . . . . . . . . . . . . . . . 865.3.2 Blue and green water availability and scarcity . . . . . . . . . . . . . . . . . . . . . 875.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 895.4.1 Agricultural development and water resources . . . . . . . . . . . . . . . . . . . . 895.4.2 Changes in water scarcity with land and water management . . . . . . . . . . . 925.4.3 Response formulation and study limitations . . . . . . . . . . . . . . . . . . . . . . 94viii5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 986 Water Footprint Impact Assessment of Water Use for Cropland and Cattle . . . . . . . 996.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 996.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1006.2.1 Goal and scope definition, and functional units . . . . . . . . . . . . . . . . . . . . 1006.2.2 Life cycle inventory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1026.2.3 Life cycle impact assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1046.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1066.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1136.4.1 Crop and cattle intensification impacts on water partitioning . . . . . . . . . . . . 1136.4.2 Complementarity in mid-point impacts . . . . . . . . . . . . . . . . . . . . . . . . . 1156.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1167 Conclusions and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1187.1 Overall research significance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1187.2 Integrating results from the harmonized water footprint assessment . . . . . . . . . . . . 1197.2.1 Volumetric water footprint assessment . . . . . . . . . . . . . . . . . . . . . . . . . 1197.2.2 Water footprint impact assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . 1207.2.3 Volumetric water footprint sustainability assessment . . . . . . . . . . . . . . . . 1217.2.4 Policy decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1237.3 Perspectives of the water footprint and future work . . . . . . . . . . . . . . . . . . . . . . 125References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158A Chapter 3 supporting information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158A.1 The Soyflux station . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158A.2 Crop height monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158A.3 Quality control of eddy covariance data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163A.3.1 Energy balance closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163A.3.2 Eddy covariance latent heat flux gap filling . . . . . . . . . . . . . . . . . . . . . . 165A.4 Calculation of reference evapotranspiration . . . . . . . . . . . . . . . . . . . . . . . . . . 168ixA.5 AquaCrop settings used for crop modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . 169A.6 Energy partitioning in Rainfed-1 and Irrigated fields . . . . . . . . . . . . . . . . . . . . . 169A.7 Water potential, soil volumetric water content and percolation . . . . . . . . . . . . . . . 171A.8 NDVI measurements in the Rainfed-1 field . . . . . . . . . . . . . . . . . . . . . . . . . . . 173B Chapter 4 supporting information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177B.1 Carbon footprint of Brazilian cattle production . . . . . . . . . . . . . . . . . . . . . . . . . 177B.2 Volumetric water footprint of pasture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177B.3 Validation of remote sensing information . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179B.3.1 Validation of pasture area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179B.3.2 Validation of small farm impoundment area . . . . . . . . . . . . . . . . . . . . . . 180B.4 Distribution of small farm impoundments and reservoir cattle density . . . . . . . . . . . 183B.5 Changes in pasture cattle density in Mato Grosso . . . . . . . . . . . . . . . . . . . . . . 186B.6 Relationship between reservoir and pasture area . . . . . . . . . . . . . . . . . . . . . . . 188C Chapter 5 supporting information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189C.1 Integrated BIosphere Simulator (IBIS) model validation of discharge . . . . . . . . . . . 189C.2 Input data used for the volumetric water footprint accounting . . . . . . . . . . . . . . . . 191C.3 Determination of environmental flow requirements . . . . . . . . . . . . . . . . . . . . . . 193C.4 Land use cover for deforestation scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . 194C.5 Total blue volumetric water footprints and hydrologic conditions . . . . . . . . . . . . . . 194C.6 Land use evapotranspiration contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 197xList of Tables2.1 Main considerations for the Nature, Production and Society domains (Figure 2.1) withfocus on the natural sciences and engineering . . . . . . . . . . . . . . . . . . . . . . . . . 162.2 The harmonized water footprint (WF) assessment, step by step. . . . . . . . . . . . . . 222.3 Summary of considerations of the main stages in the harmonized water footprint (WF)assessment (Table 2.2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.1 Management and crop rotations at the Soyflux site of Capuaba farm (Figure 3.1). . . . 303.2 Soyflux station equipment and the respective fields (Rainfed-1, Rainfed-2, Irrigated) thatthey monitor (see Figure A.1, Appendix A). . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.3 Evapotranspiration (ET) and precipitation in both Rainfed-1 and Irrigated fields. ET resultsare provided with a confidence interval obtained using values of Priestley-Taylor α (αlow -αhigh). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383.4 Crop grain yield, crop evapotranspiration (ETC), reference evapotranspiration (ET0), wa-ter productivity (WP) and mean canopy conductance (gc) in the Rainfed-1 and Irrigatedfields. ETC values are provided with a confidence interval obtained using values ofPriestley-Taylor α (αlow -αhigh), and ET0 values are provided with a confidence interval ob-tained from the propagation of measurement errors . . . . . . . . . . . . . . . . . . . . . . 413.5 Crop coefficients (KC) obtained for soybean planted in 2015 (October 2015 to January2016), and 2016 (October 2016 to February 2017), and maize (February 2016 to July2016)measured in the Rainfed-1 field, with NormalizedDifference Vegetation Index (NDVI)measured in the Rainfed-2 field. KC values are presented with a confidence interval ob-tained from the confidence intervals from both evapotranspiration (ET) and reference ET(ET0). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46xi3.6 AquaCrop simulations of crop transpiration (Tr, mm (crop cycle)-1), grain yield (ton ha-1)and water productivity based on transpiration (WPTr, kg m-3) for soybean and maize plant-ing dates. For each simulation, soybean developed over 127 days with maize plantingoccurring one week after the soybean, with a development cycle of 151 days). . . . . . 473.7 Precipitation (P, mm y-1) and evapotranspiration (ET, mm y-1) measurements in MatoGrosso, Brazil. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494.1 Parameters used to model cattle growth in Mato Grosso, Brazil, from Cardoso et al.36. 564.2 Variables used to estimate cattle water consumptive use following the growth model de-scribed in Table 4.1 and shown in Figure 4.1. . . . . . . . . . . . . . . . . . . . . . . . . . 584.3 Volumetric water footprint (VWF) of cattle feed in Mato Grosso, Brazil. . . . . . . . . . . 594.4 Water consumption of Nelore (Bos taurus indicus) in Mato Grosso (L (kg LW)-1) repre-sented by the volumetric water footprint of the animal (VWFanimal), feed (VWF feed ), andsmall farm reservoir evaporation (W res). The value of VWFanimal is the sum of the wateringested by the animal through the feed (W feed ), milk (Wmilk), liquid water drunk (Wdrink),metabolic water (Wmet), and water mixed in the feed (Wmix) when confined in feedlots inthe finishing stage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644.5 Pasture and reservoir cattle density, volumetric water, land and carbon footprints (VWF,LF, CF) for cattle production in Mato Grosso for the 2001-2015 period. . . . . . . . . . . 654.6 Summary of changes in the 104 municipal units (MUs) of Mato Grosso related to changesin pasture area (Ap, ha) and reservoir area (Ares, ha) since 2000. The values of R2 aregiven to represent the correlation of Ap and Ares as a function of time (years) . . . . . . 675.1 Description of scenarios for 2030 and 2050 activities in the Xingu Basin of Mato Grosso(XBMT) following deforestation (business-as-usual (BAU) and governance (GOV)243),and climate change scenarios (Representative Concentration Pathways (RCP) 4.5 and8.5 W m-2). BAU and GOV scenarios also illustrate agricultural intensification optionsfocused respectively on green water (BAU) and blue water (GOV) appropriation. . . . . 855.2 Summary of effects and responses for two agricultural production options focused onproduction intensification in the Xingu Basin of Mato Grosso. An illustration of theseoptions is shown in Figure 1.2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97xii6.1 Summary of mid-point impacts of land occupation that consider the partitioning of precip-itation into blue and green water at the land surface through evapotranspiration (ET) fromnatural vegetation (with evapotranspiration ETNV ), current land use (ETLU), and environ-mental flow requirements (ETEFR) (see Section 6.2 for further description). Characteriza-tion factors for life cycle impact assessment (LCIA) from Quinteiro et al.204 are subject toconditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1016.2 Input parameters used in this study, following Lathuillière et al.143 . . . . . . . . . . . . . 1076.3 Life cycle inventory data for both crop (1 ha) and cattle (1 kg LW) production. . . . . . . 1096.4 Characterization factors for the consumption of blue water (CFw ) and the land occupationimpacts for natural vegetation (NV)-to-cropland (rain-fed and irrigated) and NV-to-pasturein both Amazon and Cerrado biomes: Groundwater Recharge Potential (CFGWRP), Pre-cipitation Reduction Potential (CFPRP), Runoff Reduction Potential (CFRRP), TerrestrialGreen Water Flow (CFTGWF ) and River Blue Water production (CFRBWP). . . . . . . . . 1116.5 Characterization factors for the consumption of blue water (CFw ) and the land occupa-tion impacts for pasture-to-cropland (rain-fed and irrigated) in both Amazon and Cerradobiomes: Groundwater Recharge Potential (CFGWRP), Precipitation Reduction Potential(CFPRP), Runoff Reduction Potential (CFRRP), Terrestrial Green Water Flow (CFTGWF )and River Blue Water production (CFRBWP). . . . . . . . . . . . . . . . . . . . . . . . . . . 1127.1 Results from the harmonized water footprint (WF) assessment for SAM. . . . . . . . . . 1247.2 Strengths and limitations identified in the individual steps of the harmonized water foot-print (WF) assessment: the volumetric water footprint (VWF) assessment, the WF impactassessment (WFIA), and the VWF sustainability assessment (VWFSA) as shown in Table2.2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128A.1 Models used to infer daily crop height following measurements made in the Rainfed-1field and inferred from camera pictures. The soybean crop growth period of 2015-2016did not benefit from the automatic camera installation. . . . . . . . . . . . . . . . . . . . . 160A.2 Models used to infer daily crop height following measurements made in the Irrigated field.The Irrigated field did not benefit from the automatic camera installation. . . . . . . . . . 161A.3 Occurrence of the half-hourly value of the atmospheric stability parameter (ζ) in the Rainfed-1 and Irrigated fields classified following Franssen et al.81. . . . . . . . . . . . . . . . . . 164xiiiA.4 Linear regression table of Priestley-Taylor α values measured in the Rainfed-1 field (αlowand αhigh) as a function of daily mean soil volumetric water content (θ) measured at dif-ferent soil depths in the Rainfed-2 field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166A.5 Linear regression results of modeled LE (LEmod ) using calibrated Priestley-Taylor α values(αlow -αhigh) as a function of measured LE (LEmeas) by eddy covariance. . . . . . . . . . 167A.6 Input data and parameters used in AquaCrop for both soybean and maize. . . . . . . . 170A.7 Comparison of measured evapotranspiration (ET), yield and water productivity (WP) ofsoybean and maize in Rainfed-1 compared to modeled values obtained from AquaCropafter validation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170A.8 Soil field capacity (θfc) and dry soil (θds) determined for 0.05-m, 0.10-m, 0.30-m, and0.60-m depths in the Rainfed-2 field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171B.1 Recent greenhouse gas emissions estimates for cattle herds in Brazil. None of thesestudies considered land use change in their estimates. . . . . . . . . . . . . . . . . . . . . 178B.2 Greenhouse gas emissions from land use change for cattle production in the Amazonbiome of Mato Grosso, Brazil. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179B.3 Volumetric water footprint (m3 (kg DM)-1) of feed from pasture following two productivityscenarios: 3 tons DM ha-1 (low productivity), and 5.3 tons DM ha-1 (high productivity). 180B.4 Correlation coefficients (m and b) when comparing pasture area estimates obtained usinganimal population from IBGE121 (Ap,IBGE), and remote sensing (Ap,RS)99. Comparisonswere made for all 104 municipal units. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181B.5 Comparison of reservoir evaporation (W res) with reservoir cattle density (RCD) consider-ing all Mato Grosso municipal units (MUs), and all MUs without those within the limits ofthe Pantanal wetland. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185B.6 Reservoir cattle density (RCD, cattle ha-1), and changes in values of small farm reservoirevaporation allocated to cattle production (W res, L (kg LW)-1) considering all farm im-poundments, and removing evaporation allocated to fish tanks. Values of RCD excludingfish tanks are provided as a range based on fish production yields (3.5 ton ha-1 and 7 tonha-1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185B.7 Summary of changes in pasture cattle density (PCD, cattle ha-1) over time. . . . . . . . 188xivC.1 Cropland and pasture evapotranspiration (ET) according to Lathuillière et al.138,141 andtheir respective areas estimated from agricultural production information121, and Landsatimagery99 (bottom-up approach) to determine total ET for agriculture (ETAG). . . . . . . 191C.2 Average live animal population in 2000 and 2014 hydrologic years with animal water de-mand and living condition assumptions. Populations were obtained from IBGE121, includeboth males and females and were allocated to the Xingu Basin of Mato Grosso based onarea of municipalities contained within the basin. Chicken and swine populations wererecalculated based on life expectancy described in equation 5.5. . . . . . . . . . . . . . . 192C.3 Urban, rural, industrial worker population and domestic and industrial blue water demandin the Xingu Basin of Mato Grosso. Note that blue water consumption was assumed tobe 50% of blue water demand. Data derived from IBGE121 and ANA4. . . . . . . . . . . 193C.4 Total forest cover as described by land use maps obtained by Soares-Filho et al.243 inthe Xingu Basin of Mato Grosso for business-as-usual (BAU) and governance (GOV)deforestation scenarios. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195C.5 Total blue volumetric water footprint for agricultural, industrial and domestic uses in theXingu Basin of Mato Grosso in 2000 and 2014 hydrologic years, as well as scenariosfor 2030 and 2050 (see Table C.2 and C.3 for input data, as well as Table 5.1 for thedescription of scenarios). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195C.6 Total annual and 3-month mean runoff in the Xingu Basin of Mato Grosso obtained fromIBIS simulations and land use (equation 5.1). . . . . . . . . . . . . . . . . . . . . . . . . . 196C.7 Individual land use contributions to evapotranspiration (ET) obtained in this study usingthe bottom-up approach between 2000 and 2010 compared to values obtained by Silvérioet al.238 using the MODIS ET product172. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197C.8 Values of total evapotranspiration (ETT ), evapotranspiration of the natural vegetation(ETNV ), potential natural vegetation (ETPNV ), and the combined ET of agriculture andresidual landscapes (ETAG + ETR) in the Xingu Basin of Mato Grosso between 2000 and2050 hydrologic years considering business-as-usual (BAU) and governance (GOV) de-forestation, and Representative Concentration Pathways (RCP 4.5 and 8.5 W m-2). Allvalues were obtained using the top-down approach (Figure C.4). . . . . . . . . . . . . . 198xvList of Figures1.1 Southern Amazonia (SAM) and the Brazilian state of Mato Grosso with the boundariesof the Amazon, Cerrado and Pantanal biomes143. Reprinted with permission of ElsevierLtd. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.2 Land and water management options available in Southern Amazonia (SAM) to increasefuture agricultural production, and their consequences on the partitioning of precipitation(P, assumed to be 2000 mm y-1 in the above example) into blue (blue arrows) and greenwater (shown as ET) following initial conditions141. Values shown in the panels are waterflows (in mm y-1) and the green arrow in panel D represents 300 mm y-1of harvestedrainwater. Reprinted with permission under Creative Commons Attribution 3.0. . . . . . 92.1 The relationships of water in Nature, Production and Society domains . . . . . . . . . . 162.2 Proposed harmonized water footprint (WF) assessment and terminology combining ap-proaches from the WF Network117 and life cycle assessment (LCA)126,127 within the Pro-duction and Nature domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.1 Location of the Soyflux site of Capuaba farm in Lucas do Rio Verde, Mato Grosso, Brazil.The inset shows the state of Mato Grosso with the location of the municipality of Lucasdo Rio Verde. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.2 Evapotranspiration (ET) measurements at the Soyflux site shown with precipitation (P,mm d-1) (a), 24-hour mean shortwave irradiance (Rs, W m-2) (b), reference evapotran-spiration (ET0, mm d-1) (c), and cropland ET (mm d-1) measured in the Rainfed-1 (d) andIrrigated (e) fields. Values of ET0 and ET are represented by their confidence intervalsbased on systematic and standard errors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39xvi3.3 Evapotranspiration-related variables over the study period at the Rainfed-1 field includingprecipitation (P, mm d-1) (a), 24-hour mean vapour pressure deficit (D, kPa) (b), evapo-transpiration (ET, mm d-1) (c), daytime average canopy conductance (gc, m s-1) (d), andsoil water potential (ψ) at the 0.10-m, 0.30-m, and 0.60-m depths (e). . . . . . . . . . . . 423.4 Evapotranspiration-related variables over the study period at the Irrigated field includingprecipitation (P, mm d-1) (a), 24-hour mean vapour pressure deficit (D, kPa) (b), evap-otranspiration (ET, mm d-1) (c), and daytime average canopy conductance (gc, m s-1)(d). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.1 Process flow for the cattle production system of Mato Grosso, Brazil. . . . . . . . . . . . 564.2 Average animal volumetric water footprint (VWFanimal , L (kg LW)-1), and land footprint(LF, m2 (kg LW)-1) for the state of Mato Grosso between 2001 and 2015 consideringpasture finishing. Values of VWFanimal are separated into water content of feed (W feed ),animal drinking water (Wdrink), metabolic water (Wmet) and water evaporated by smallfarm reservoirs (W res). Values of W res represents total small farm reservoir evaporationminus evaporation allocated to fish tanks, with fish tank area determined by mean fishproduction (3.5-7 ton ha-1 of water). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 634.3 Evolution of pasture (PCD) and reservoir (RCD) cattle densities (considering all small farmreservoirs) between 2001 and 2015 in Mato Grosso derived using 2001 median valuesof 0.54 cattle ha-1 and 510 cattle ha-1, respectively. This evolution is separated into fourgroups: Pasture Dense (PCD > 0.54 cattle ha-1, RCD < 510 cattle ha-1), High Density(PCD > 0.54 cattle ha-1 and RCD > 510 cattle ha-1), Low Density (PCD < 0.54 cattle ha-1and RCD < 510 cattle ha-1), and Water Dense (PCD < 0.54 cattle ha-1 and RCD > 510cattle ha-1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 685.1 The Xingu Basin of Mato Grosso (XBMT) and its sub-basins: the Upper Xingu Basin(yellow) and the Xingu Headwaters (green) with the main rivers and the location of thedischarge measurement station used for validation5. The inset shows the position ofXBMT (black) in relation to the Xingu River Basin (black outline) and the state of MatoGrosso (gray). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 775.2 Total blue volumetric water footprint (VWF) of agriculture in the Xingu Basin of MatoGrosso for the 2000 and 2014 hydrologic years. . . . . . . . . . . . . . . . . . . . . . . . . 87xvii5.3 Changes in contributions to evapotranspiration (ET) for natural vegetation (ETNV ), pasture(ETP), cropland (ETC) and residual landscapes (ETR) in the Xingu Basin of Mato Grossoin the 2000 and 2014 hydrologic years (September-August). Values obtained through thebottom-up approach as described in the text. . . . . . . . . . . . . . . . . . . . . . . . . . . 885.4 Annual blue (WSB) and green (WSG) water scarcities for the Xingu Basin of Mato Grossoin 2000 and 2014 hydrologic years, business-as-usual (BAU) and governance (GOV) de-forestation scenarios in 2030 and 2050 considering Representative Concentration Path-ways (RCP 4.5 and 8.5 W m-2). Values of WSG were obtained assuming that 35%, 50%and 80% natural vegetation cover in the basin was maintained (Table 5.1). . . . . . . . . 906.1 Scenarios for production systems considered in this study for estimating the mid-pointenvironmental impacts of cropland and cattle production systems (cradle-to-farm gate).Scenarios include cropland extensification on natural vegetation (NV) or pasture, anddifferences in pasture productivity for cattle. An illustrative example is shown in Figure1.2 (panels A, B and E). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1036.2 Mid-point impacts of water consumption and land occupation of cropland (m3 ha-1) fromAmazon natural vegetation (NV) and pasture. Impacts are Precipitation Reduction Po-tential (PRP), Terrestrial Green Water Flows (TGWF), Groundwater Recharge Potential(GWRP), Runoff Reduction Potential (RRP), River Blue Water Production (RBWP). . . 1106.3 Mid-point impacts of water consumption and land occupation of cropland (m3 ha-1) fromCerrado natural vegetation (NV) and pasture. Impacts are Precipitation Reduction Po-tential (PRP), Terrestrial Green Water Flows (TGWF), Groundwater Recharge Potential(GWRP), Runoff Reduction Potential (RRP), River Blue Water Production (RBWP). . . 1106.4 Mid-point impacts of water consumption and land occupation for cattle (m3 (kg LW)-1) inboth Amazon and Cerrado biomes. Impacts are Precipitation Reduction Potential (PRP),Terrestrial GreenWater Flows (TGWF), Groundwater Recharge Potential (GWRP), RunoffReduction Potential (RRP), River Blue Water Production (RBWP). . . . . . . . . . . . . 1126.5 Complementarity of mid-point impacts of Groundwater Recharge Potential (GWRP), RiverBlue Water Production (RBWP), Terrestrial Green Water Flows (TGWF), Precipitation Re-duction Potential (PRP), and Runoff Reduction Potential (RRP) in a natural vegetation(NV)-to-cropland or pasture-to-cropland land use transition. . . . . . . . . . . . . . . . . . 117xviiiA.1 The Soyflux station described in Chapter 3 following soybean (top) and maize (bottom)crop development cycles between September 2015 and June 2016. Note the location ofthe Rainfed-1 field (background, right) and Rainfed-2 field (foreground, left) shown in Fig-ure 3.1. The additional tower structure in the Rainfed-2 field contains the net radiometerand NDVI sensor described in Table 3.2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159A.2 Soybean crop height measurements taken in 2015 and 2016 in the Rainfed-1 andRainfed-2 fields and separated into initial (4-6 days), development (7-60 days), mid-season (61-100 days), and final phase (101-127 days). Only the Rainfed-2 measurements made in2016 took advantage of the automatic camera setup for height measurements. . . . . . 162A.3 Maize crop height measurements taken in 2016 in Rainfed-1 and Rainfed-2 and separatedinto initial (0-20 days), development (21-56 days), mid-season (57-97 days), and finalphase (98-151 days). Only the Rainfed-2 measurements took advantage of the automaticcamera setup for height measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162A.4 Energy balance closure as the sum of latent heat (LEmeas) and sensible heat (H) fluxesmeasured by eddy covariance in the Rainfed-1 field as a function of the difference betweennet radiation (Rn) and ground heat flux (G) measured in the Rainfed-2 field. The equationfor the regression line (blue line) is also shown. . . . . . . . . . . . . . . . . . . . . . . . . 163A.5 Energy balance closure as the sum of latent heat (LEmeas) and sensible heat (H) fluxesmeasured by eddy covariance in the Irrigated field as a function of the difference betweennet radiation (Rn) and ground heat flux (G) measured in the Rainfed-2 field. The equationfor the regression line (blue line) is also shown. . . . . . . . . . . . . . . . . . . . . . . . . 164A.6 Example of calculation of the Priestley-Taylor α obtained through linear regression for theRainfed-1 field on 9 December 2016 (a) and between 25 November 2016 and 1 December2016 (b). All plotted data represents half-hourly measurements. . . . . . . . . . . . . . . 165A.7 Daily mean Priestley-Taylor α values calculated for the Rainfed-1 and Irrigated fields be-tween 18 September 2015 and 4 February 2017. . . . . . . . . . . . . . . . . . . . . . . . 166A.8 Comparison of gap-filled values of LE (LE modeled) using the daily mean Priestley-Taylorα values with measurements of LE (LE measured, or LEmeas) obtained from eddy covari-ance in both the Rainfed-1 (left) and Irrigated (right) fields. The blue lines represent theregression lines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167xixA.9 Radiation partitioning described by linear regression of latent heat flux (LE, W m-2) as afunction of the difference between net radiation (Rn, W m-2) and ground heat flux (G, Wm-2) for Rainfed-1 in the wet (a) and dry (b) season, and the Irrigated field in the wet (c)and dry (d) seasons. Blue lines are the regression lines (equations also shown). . . . . 172A.10 Daily changes in soil water storage down to the 0.60-m depth (ΔSWS, mm d-1) in theRainfed-2 field, as a function of daily available water (P −ET, mm d-1) in the Rainfed-1field. The blue line is the regression line (equation also shown). . . . . . . . . . . . . . . 174A.11 Soil water balance derived frommeasurements of daily precipitation (P, mm d-1) (a), evap-otranspiration (ET, mm d-1), changes in daily average soil water storage (ΔSWS, mm d-1)(c), daily average drainage below the 0.60-m depth (ε, mm d-1) (d), and soil availablewater fraction (Awf, dimensionless) in the root zone (e) . . . . . . . . . . . . . . . . . . . . 175A.12 Environmental and crop characteristics in the Rainfed-1 field: precipitation (P, mm d-1)(a), 24-hour mean net radiation (Rn, W m-2) (b), Normalized Difference Vegetation Index(NDVI, dimensionless), evapotranspiration (ET, mm d-1), and the ratio of ET to referenceET (ET/ET0, dimensionless) (e). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176B.1 Comparison of estimates of total pasture area in Mato Grosso from IBGE121 (this study)following Lathuillière et al.138, and remote sensing including MODIS148 and Landsat99.Estimates fromMODIS are provided with and without the consideration of protected areas(PA). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182B.2 Distribution of water evaporated from reservoirs (W res, L (kg LW)-1) in 2015 in the 104municipal units of Mato Grosso. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183B.3 Distribution of reservoir cattle density (RCD, cattle ha-1) in 2015 in the 104 municipal unitsof Mato Grosso. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184B.4 Significant increases and decreases (p ≤ 0.05) in reservoir cattle density (RCD, cattleha-1) between 2001 and 2015. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184B.5 Distribution of pasture cattle density (PCD, cattle ha-1) in 2015 in the 104 municipal unitsof Mato Grosso. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186B.6 Significant increases and decreases (p≤ 0.05) in pasture cattle density (PCD, cattle ha-1)between 2001 and 2015. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187B.7 Relationship between small farm reservoir area (Ares) and pasture area (Ap) in the 104MUs of Mato Grosso in 2001 and 2015. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188xxC.1 Validation of the monthly discharge (R(t)) for the Xingu Headwaters in the 2000 (n = 12)and 2005 (n = 4) hydrologic years at station 18430000 located in Marcelândia (MatoGrosso)5 (Figure 5.1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190C.2 Modeled compared to observed 3-month mean discharge at station 18430000 located inMarcelândia (Mato Grosso)5 for the Xingu Headwaters in the 2000 (n = 12) and 2005 (n= 4) hydrologic years, and for the 1975-2005 (n = 120) period (Figure 5.1). . . . . . . . 190C.3 Exceedance probability curve for the Xingu Headwaters obtained from monthly observa-tions at Marcelândia, Mato Grosso (Passagem BR80, station 18430000, 10° 46’ 38” S,53° 5’ 44” W)5 for the 1975-2005 period (n = 363). . . . . . . . . . . . . . . . . . . . . . . 194C.4 Land contributions to evapotranspiration (ET) in the Xingu Basin of Mato Grosso between2000 and 2050 using the top-down approach and following business-as-usual (BAU) andgovernance (GOV) deforestation scenarios, and Representative Concentration Pathways(RCP 4.5 and 8.5 W m-2) as described in Table 5.1. . . . . . . . . . . . . . . . . . . . . . 199xxiList of AcronymsAcronym NameA.U. Animal UnitsAWARE Available WAter REmainingBAU Business-as-usual deforestation scenario243BAURCP4.5 BAU considering Representative Concentration Pathway 4.5 W m-2BAURCP8.5 BAU considering Representative Concentration Pathway 8.5 W m-2CF Carbon FootprintDM Dry MatterEFR Environmental Flow RequirementsET EvapotranspirationG C4 grass IBIS simulationGOV Governance deforestation scenario243GOVRCP4.5 GOV considering Representative Concentration Pathway 4.5 W m-2GOVRCP8.5 GOV considering Representative Concentration Pathway 8.5 W m-2GWRP Groundwater Recharge PotentialIBIS Integrated BIosphere SimulatorIWRM Integrated Water Resources ManagementLAI Leaf Area IndexLCA Life Cycle AssessmentLCI Life Cycle InventoryLCIA Life Cycle Impact AssessmentLF Land FootprintLU Land Use (current)LW Live weightMAR Mean Annual RunoffMDG(s) Millennium Development Goal(s)MU Municipal UnitNDVI Normalized Difference Vegetation IndexNV Natural VegetationPCD Pasture Cattle DensityPNV Potential Natural Vegetation IBIS simulationPRP Precipitation Reduction PotentialRBWP River Blue Water ProductionRCD Reservoir Cattle DensityRCP Representative Concentration PathwayRRP Runoff Reduction PotentialSAM Southern AmazoniaSDG(s) Sustainable Development Goal(s)TGWF Terrestrial Green Water FlowsUN LCI United Nations Life Cycle InitiativexxiiVWF Volumetric Water FootprintVWFSA Volumetric Water Footprint Sustainability AssessmentWP Water ProductivityWF Water FootprintWFIA Water Footprint Impact AssessmentWSF Water Scarcity FootprintWULCA Water Use in Life Cycle AssessmentXBMT Xingu Basin of Mato GrossoxxiiiList of SymbolsSymbol Name UnitsA Land occupation area haANV (t) Area of natural vegetation in the Xingu Basin ofMato Grossom2Ap Pasture area haAp,IBGE Pasture area determined using the censusinformation from IBGE121haAp,RS Pasture area determined from remote sensing99 haAres Reservoir area haAMDi Blue water availability minus demand for basin i m3 m-2 mo-1AMDworld World average blue water availability minusdemand0.0136 m3 m-2 mo-1Awf Soil available water fraction dimensionlessB Ratio of aerodynamic to surface resistance 0.24 (day), 0.96(night)b y intercept of the Ap,RS versus Ap,IBGE curve haCFGWRP Characterization factor of land occupation mid-pointimpact Groundwater Recharge Potentialm3 ha-1 y-1CFj Characterization factor of land occupation mid-pointimpact jm3 ha-1 y-1CFPRP Characterization factor of land occupation mid-pointimpact Precipitation Reduction Potentialm3 ha-1 y-1CFRBWP Characterization factor for the mid-point impactRiver Blue Water ProductiondimensionlessCFRRP Characterization factor of land occupation mid-pointimpact Runoff Reduction Potentialm3 ha-1 y-1CFTGWF Characterization factor for the mid-point impactTerrestrial Green Water FlowsdimensionlessCFw Characterization factor based on available waterremaining31dimensionlesscp Specific heat of air at constant pressure J kg °C-1D Vapour pressure deficit kPad Zero plane displacement mdays Total number of days in the animal developmentcycledDig Feed digestibility %DMI Dry matter intake kg d-1mo-1dr Inverse relative distance Earth-Sun radea Actual vapour pressure kPaEFR Environmental Flow Requirements m3 y-1ET Evapotranspiration mm d-1xxivETAG Evapotranspiration of agriculture in the Xingu Basinof Mato Grossom3 y-1ETC Evapotranspiration of cropland in the Xingu Basinof Mato Grossom3 y-1ETC,i Crop evapotranspiration for day i mm d-1ETEFR Evapotranspiration that maintains environmentalflow requirementsmm y-1ETLU Evapotranspiration of current land use mm y-1ETLU,eff Effective evapotranspiration of current land use m3 ha-1 y-1ETNV Evapotranspiration of natural vegetation in theXingu Basin of Mato Grossom3 y-1ETNV,eff Effective evapotranspiration of natural vegetation m3 ha-1 y-1ETP Pasture evapotranspiration mm y-1ETPNV Total evapotranspiration in the Xingu Basin of MatoGrosso under Potential Natural Vegetationmm y-1ETPNV,eff Effective evapotranspiration of Potential NaturalVegetationm3 ha-1 y-1ETR Evapotranspiration of residual landscapes in theXingu Basin of Mato Grossom3 y-1ETRNV Evapotranspiration reserved for natural vegetationin the Xingu Basin of Mato Grossom3 y-1ETT Total evapotranspiration in the Xingu Basin of MatoGrossomm y-1, m3 y-1ETt Daily total evapotranspiration of Rainfed-1 field onday tmmETUN Evapotranspiration of areas agriculturallyunproductive in the Xingu Basin of Mato Grossom3 y-1ET0 Reference evapotranspiration mm 30 min-1ET0,i Reference evapotranspiration for day i mm d-1er Basin internal evaporation recycling ratio (XinguRiver Basin)0.22 (dimensionless)fAU Animal unit factor A.U. animal-1F f (t) Fraction of forest cover in the pixel of interestlocated in the Xingu Basin of Mato GrossodimensionlessFNV (t) Fraction of forest cover in the Xingu Basin of MatoGrossodimensionlessG Ground heat flux W m-2or MJ m-2 30min-1Ggrass Ground heat flux under theoretical grass surface MJ m-2 30 min-1Gsc Solar constant 0.0820 MJ m-2min-1GWRLU Groundwater recharge of current land use mm y-1GWRNV Groundwater recharge of natural vegetation mm y-1g Gravitational constant 9.81 m s-2gc Bulk observed canopy conductance m s-1ga Aerodynamic conductance m s-1H Sensible heat flux W m-2H i,t Total cattle population in municipal unit i and year t cattleh Crop height mIocc Land occupation impacts on the water cycle m3j day of the year dimensionlessKC,i Crop coefficient for day i dimensionlessKS Hydraulic saturation mm d-1xxvk von Karman constant 0.4L Soil depth mmLCIw Blue water footprint inventory m3 ha-1Li,t Total living cattle population in municipal unit i andyear tcattleLm Longitude of the Soyflux site °Lz Longitude at the center of the time zone °LE Latent heat flux W m-2LE* Latent heat flux corrected for energy balanceclosureW m-2LEmeas Latent heat flux measured W m-2LF Land footprint of cattle m2 (kg LW)-1LSD Livestock density A.U. ha-1MC Moisture content of feed %m Slope of the AP,RS versus AP,IBGE curve dimensionlessN Total animal population animalsNGWeff Effective net green water m3 ha-1 y-1P Precipitation mm, mm d-1 or mm y-1Pk,i (t) Animal population in Municipal Unit i and calendaryear tanimalsPCD Pasture cattle density cattle per ha ofpasturePt Daily total precipitation on day t mmQ50 Runoff at the point of 50% exceedence probability m3 s-1 mo-1Q90 Runoff at the point of 90% exceedence probability m3 s-1 mo-1RCD Reservoir cattle density cattle per ha of waterR(t) Runoff in the Xingu Basin of Mato Grosso mm mo-1Ra Extraterrestrial radiation MJ m-2 30 min-1RG(t) Runoff in the Xingu Basin of Mato Grossoconsidering in the C4 grass IBIS simulationmm mo-1Rn Net radiation W m-2or MJ m-2 30min-1Rn,grass Net radiation above the theoretical grass surface MJ m-2 30 min-1Rnl Net outgoing longwave radiation MJ m-2 30 min-1RPNV (t) Runoff in the Xingu Basin of Mato Grossoconsidering in the Potential Natural Vegetation IBISsimulationmm mo-1Rs Shortwave irradiance W m-2or MJ m-2 30min-1Rso Clear-sky incoming shortwave irradiance MJ m-2 30 min-1Sc Solar time seasonal correctionT Temperature °CTLU Total livestock units A.U.Tr Transpiration mm period-1t time from s to dtocc Land occupation time yu Wind speed measured by the sonic anemometer m s-1u* Friction velocity m s-1u2 Wind speed, 2 m above the ground m s-1VWC Daily mean soil volumetric water content m3 m-3VWCt Daily mean water soil volumetric water content onday tm3 m-3xxviVWFanimal Volumetric water footprint of the animal L (kg LW)-1VWF feed Volumetric water footprint of the feed L (kg LW)-1VWF j (t) Volumetric water footprint of activity j m3 y-1W Total amount of water consumed by cattle L d-1mo-1Wdrink Water drunk by cattle L d-1mo-1Wevap Water lost by cattle to evaporation L d-1mo-1W feces Water content in cattle feces L d-1mo-1W feed Water contained in feed and ingested by cattle L d-1mo-1Wmix Water needed to mix feed for cattle L d-1mo-1Wmilk Milk ingested by cattle L d-1 mo-1W res Small farm reservoir evaporation L mo-1W tot Total daily water intake L d-1mo-1Wurine Water content in cattle urine L d-1mo-1WWG Water incorporated by cattle L d-1mo-1WA(t) Water availability in the basin m3 y-1WAB(t) Blue water availability in the basin m3 y-1WAG(t) Green water availability in the basin m3 y-1WP Water Productivity kg m-3WPTr Water productivity based on transpiration kg m-3WS(t) Water scarcity dimensionlessWSB(t) Blue water scarcity dimensionlessWSG(t) Green water scarcity dimensionlessz Measurement height mzoH Roughness length for heat mzoM Roughness length for momentum mα Priestley-Taylor coefficient dimensionlessα* Priestley-Taylor coefficient obtained with LE* dimensionlessαer Blue water basin internal evaporation recycling ratio(Xingu River Basin)0.07 (dimensionless)αhigh High estimate of the Priestley-Taylor coefficient dimensionlessαlow Low estimate of the Priestley-Taylor coefficient dimensionlessαmeas Priestley-Taylor coefficient obtained from LEmeas dimensionlessΔ Slope of the vapour pressure vs. temperature curve kPa °C-1ΔSWS Change in daily average soil water storage mm d-1δ Solar declination radε or εt Daily average drainage beyond 0.60 m in theRainfed-1 field (on day t)mm d-1ζ Atmospheric stability dimensionlessθ Soil volumetric water content m3 m-3θds Soil volumetric water content at dry soil (proxy forpermanent wilting point)m3 m-3θfc Soil volumetric water content at field capacity m3 m-3γ Psychrometric constant kPa °C-1λ Latent heat of vapourization MJ kg-1ρ Density of dry air kg m-3σv Stefan Boltzmann constant 4.903 MJ K4 m-2 30min-1φ Latitute of Soyflux station radχEFR Fraction of environmental flow requirements to thelong-term mean discharge (Xingu River Basin)0.42 (dimensionless)ΨH Integral diabatic correction factor for heat dimensionlessΨM Integral diabatic correction factor for momentum dimensionlessxxviiψ Soil water potential kPaω Solar time angle at mid-point of the 30-min window radω1 Solar time angle at the beginning of 30-min window radω2 Solar time angle at the end of the 30-min window radxxviiiAcknowledgmentsI am forever grateful to my supervisor, Dr Mark Johnson, for his support and encouragement duringmy graduate training and his continued optimism throughout my research, even during the most difficulttimes. The freedom he gave me through the years allowed me to fully grow as a researcher learningabout the fascinating field of ecohydrology. Much of the work presented in this thesis could not havebeen carried out without the essential collaboration of Dr Eduardo Guimarães Couto of the UniversidadeFederal de Mato Grosso (UFMT) in Cuiabá, Brazil, his academic support and friendship during my timein Brazil. Both Drs Johnson and Couto truly made my graduate studies an unforgettable experience.The research described in this thesis is a contribution to the project entitled “Integrating land use plan-ning and water governance in Amazonia: towards improving freshwater security in the agricultural fron-tier of Mato Grosso” which was supported by the Belmont Forum and G8 Research Councils FreshwaterSecurity Grant [G8PJ-437376-2012] through the Natural Sciences and Engineering Research Council(NSERC), as well as the Agricultural Water Innovations in the Tropics (AgWIT) project funded by the EUJoint Call for theWater Joint Programming Initiative (JPI) [WTWPJ 506082-2016] through NSERC. Addi-tional financial support was provided by the NSERC Vanier Graduate Scholarship [201411DVC-347484-257696], the NSERC Michael Smith Foreign Study Supplement [488016-2015], the Mitacs GlobalinkResearch Award [IT05435], the University of British Columbia (UBC) through the Izaak Walton KillamPre-Doctoral Fellowship and the Kruger Graduate Fellowship, as well as the Canadian GeophysicalUnion’s Don Gray Scholarship in Canadian Hydrology.I am grateful to my Ph.D committee members for their valuable guidance over the course of thisresearch: Dr T. Andrew Black (UBC) for his valuable input during our discussion of eddy covarianceresults, Dr Cécile Bulle (Université du Québec à Montréal, UQAM) for working closely with me on waterand land use in life cycle assessment, and inviting me to spend time at CIRAIG in Montreal, and DrMichael Coe (Woods Hole Research Center, WHRC) for his guidance and insight into Southern Ama-zonia’s regional hydrology. Moreover, I am thankful for additional collaborations for this work from DrHigo Dalmagro (Universidade de Cuiabá, UNIC) who accompanied me on all my field trips, Drs IainxxixHawthorne (UBC) and Paulo Arruda (UFMT) for technical support at the field site, Dr Jordan Graesser(Boston University), Kylen Solvik and Dr Marcia Macedo (WHRC) for their help with remote sensing inMato Grosso, and Dr Andrea Castanho (WHRC) for providing hydrological modeling support.The field work described in this thesis could not have been possible without the kindness of Capuabafarm owner José Eduardo de Macedo Soares Jr and support from the farm staff that made my graduateexperience extremely enjoyable and stimulating. Additional Brazilian support was kindly provided bySuzana Souza dos Santos, Drs Ricardo Santos Amorim, José de Souza Nogueira (UFMT), and OsvaldoBorges Pinto Jr (UNIC), with special thanks to Drs Francisco de Almeida Lobo and Carmen EugeniaRodríguez Ortíz for sharing their laboratory space and providing valuable feedback on photosyntheticprocesses. I would also like to thank Drs George Vourlitis (University of California San Marcos), Anne-Marie Boulay, Manuele Margni and Viêt Cao (CIRAIG) for insightful discussions about eddy covarianceand life cycle assessment. I am grateful for the amazing assistance from the staff at the Institute forResources, Environment and Sustainability, especially Linda Stewart, Bonnie Leung, Gillian Harris, LisaJohannesen, and Stefanie Ickert.Finally, I would like to thank my family for their support through these years, especially my amazingwife Verônica Laura de Campos Lathuillière, who has been by my side every step of the way.xxxDedicationIn loving memory of Lucien Lathuillière (1913-1999), who cared deeply about our water footprint.xxxiChapter 1Introduction1.1 Water use for agricultural productionWater resources and their management are of primary concern to society, particularly in the context offood production. The agricultural sector is the number one consumer of water, accounting for 70-80%of global withdrawals, primarily due to irrigation279. However, these withdrawals are accounted sep-arately from the global water use for crops which includes both irrigated and rain-fed cropland. Totalcrop water use was estimated to be over 7000 km3 y-1 of which close to 90% came from rain-fed sys-tems146, increasing the original allocation of water use for food. The already existing threats of droughtand water scarcity260,281, which are expected to worsen food production and security under climatechange72, deepen the challenge to eradicate hunger as established by the United Nations with the Mil-lennium Development Goals (MDGs)257 to 2015, and now the Sustainable Development Goals (SDGs)to 2030258.The magnitude and regional disparity in global crop water requirements suggest that water use andmanagement can play an important role in alleviating poverty and hunger. An ecohydrological paradigmpresented in the early 2000s by Falkenmark and Rockström71 provided a key framework for addressingwater use for agriculture by separating water into “green” and “blue” resources based on the partitioningof rainfall on the landscape71. Green water refers to soil moisture consumed by plants through evapo-transpiration (ET) and regenerated only via precipitation, while blue water is sourced from rivers, lakesand aquifers71. Through this framework, agricultural yields can be increased following strategies thatfocus on the use of green and blue water resources, respectively for rain-fed and irrigated cropland71.These strategies, however, carry trade-offs in the hydrological cycle128. For instance, the upstream useof rainwater harvesting or irrigation can deprive downstream ecosystems and other water users of bluewater, while rain-fed agricultural expansion into natural ecosystems can increase downstream blue waterresources due to increased runoff128. Strategies that seek to increase agricultural yields by favouringtranspiration over evaporation (or increasing water productivity93,128) were highlighted as having the1greatest global impact on rain-fed agricultural production, thereby raising the importance of green waterresources for food production222.The topic of water use for agricultural production has, more recently, been considered in the contextof the Earth’s Planetary Boundaries223 by focusing specifically on the use of irrigation to meet futureagricultural demand this century. As 2600 km3 y-1 of blue water is currently used for crops, an additional1700 km3 y-1 would be required to address food demand, and an extra 1550 km3 y-1 to meet climatechange mitigation requirements in 2050. This combined 5850 km3 y-1 approaches the current waterPlanetary Boundary of 6000 km3 y-1 224. The efficient use of green water for agricultural productioncan therefore play an important role in maintaining our planet within sustainable limits. On a moreregional level, increases in agricultural output have been obtained through both deforestation (e.g. SouthAmerica) and irrigation expansion (e.g. India) with effects on regional water cycles227. However, thesechanges come with potentially large negative feedbacks. Loss of natural forest cover reduces watervapour flows to the atmosphere, thereby reducing the potential regeneration of regional rainfall, while atthe same time, increasing surface temperatures66 with potential detrimental effects on crop yields. Theprevalence of groundwater depletion has been of major concern in key agricultural production centers,introducing resource constraints for people, ecosystems and future agricultural production output94.The relationship between production and consumption centers of agricultural products through tradeadds an extra dimension to freshwater resource use. Green and blue water resources connect producersand consumers of agricultural products through a complex global supply chain, which could extendgeographically within a country or across international boundaries. This so-called “virtual water” trade1represents the total amount of water used for agricultural production that is transferred from productionto consumption centers. Virtual water content is typically estimated as ET sourced from either green orblue water and generally exceeds the physical water content in the crop117. For instance, cereal cropsrequired on average about 1109 m3 of green water per tonne of crop, and 291 m3 blue water per tonneof crop235. This virtual water trade relationship between producers and consumers therefore, provides aglobal element to freshwater resources management linked specifically to products272, and can revealadditional strategies for improving water use efficiencies and resources management for agriculturalproduction.21.2 The water footprint: an emerging research fieldThe water footprint (WF) surfaced in 2002 as an indicator of freshwater use109,119 which, through theyears, has matured into an emerging research field113. In its original definition, the WF representsthe volume of freshwater consumed and polluted either directly by individuals or through a farm, plantor business operation, and indirectly via the consumption of goods and services by individuals, citiesor nations, or through supply chains117. Based on initial qualitative assessments of virtual water in-troduced by Allan1, the quantification of freshwater use for agricultural products through the WF109highlighted three important notions: (1) a significant amount of freshwater is consumed and pollutedthrough consumption and production processes, particularly in the case of agricultural products114; (2)given the extent of international trade, consumption patterns of products and services flow virtually fromproduction to consumption centers thereby allowing water stressed nations to conserve water within theirborders, leading to global water savings56,58; (3) there is a need for better understanding of indirect wa-ter consumption (water that is consumed to make products that are subsequently traded, also known assupply chain water consumption) in the context of sustainable water use116,231, particularly given thatglobal water withdrawals often take place in highly stressed watersheds217. The first WF assessmentsfollowed methods which were formalized in the “Water Footprint Assessment Manual” released by theWF Network in 2011117. This manual provides guidelines on calculating WF of products, a businessor national consumption, but also describes how to use the WF in a “Water Footprint Sustainability As-sessment”117 that aims to assess efficient, equitable and environmentally sustainable use of freshwaterby comparing total water consumption (represented by the total WF) to total water availability in a riverbasin117,118.Since 2008, the inclusion of water use in life cycle assessment (LCA) has provided additional con-text for the analysis of direct and indirect water uses with the objective of carrying out an impact as-sessment. LCA is a scientific method based on the life cycle approach, meaning that processes arestudied over entire life cycles, from “cradle to grave”105. Following the ISO 14044 standard126, a LCAprovides a quantitative assessment of potential impacts: (1) mid-point impacts identify problems (e.g.eutrophication, acidification, etc.), and (2) end-point impacts quantify consequences of these problemsto human health, environmental quality and natural resources105. The strength of this “life cycle think-ing” is the consideration of direct and indirect inputs (or supply chain inputs) to a production systemto highlight specific steps of greater concern in the life cycle (or hotspots) and is applicable to goodsand services, nations, organizations or even lifestyles105. Despite a late inclusion of freshwater use in3LCA134, methodological developments have grown rapidly in recent years22,205 with a formalized ISO14046 standard published in 2014 that outlines the WF “principles, requirements and guidelines”127.Recent modeling work has focused on mapping potential environmental impacts of freshwater use onhuman health28,169, competition over available water31 expressed as a water scarcity footprint follow-ing ISO 14046 terminology127, and partial mapping of impacts to ecosystem quality181, and naturalresources193. However, a complete and comprehensive set of models is still lacking in the academicliterature.There exists common ground between the WF approaches as described by the WF Network117and the LCA community29, namely the importance of “supply chain thinking” in water management andimpact assessment113. However, there has been mutual criticism between the respective communitiesthat has mainly focused on methodological steps, the meaning of WF results89,112,115,195, contextualiza-tion of meaningful decisions using the WF111, and other footprints more generally113,215. Thus far, theWF, as employed and promoted by the WF Network, has been used with the specific goal of informingwater management decisions within Integrated Water Resources Management (IWRM), but has alsoprovided guidance on improvements in the efficient use of water in processes, particularly in agricul-tural production113. Moreover, the WF has emphasized potential effects of consumption decisions onwater scarcity through demand-side management (e.g., the effect of diets on water resources267), butalso transparency of indirect water use of companies145, while highlighting European dependency oninternational water resources for imports69. In LCA, the WF has been used to assess potential impactsof water use in agricultural products such as mangoes214, or livestock218, with the main objective ofassessing the environmental performance of a product or process, or informing production decisionsduring the design phase of a food product (a process known as “eco-design”)189.Despite differences, the two WF communities do agree on some terminologies such as the designa-tion of water resources as either blue or green water as defined above. Both communities also value theuse of life cycle thinking for purposes geared towards water management in the case of theWF approachfrom the WF Network, and reduction of potential impacts in LCA29. There is also overlapping consen-sus around future academic work such as defining and accounting for green water scarcity117,180,233,and identifying environmental impacts linked to green water use, especially in the context of land usechange181,205. As such, agricultural products, which were the first products assessed for their WF119,remain at the forefront of WF research mainly because: (1) food represents an important source of indi-rect water use embodied in international trade56,114, (2) the majority of global agricultural production israin-fed, and therefore depends on green water114, and (3) a large amount of cropland expansion occurs4in the tropics162, specifically in regions dominated by rain-fed systems, which are therefore dependenton green water resources for production. Additionally, the quantification of water use for agricultureobtained through models commonly used in the WF literature carry large uncertainty189, therefore de-cisions that are based on the WF may change according to data availability and/or quality in commondatabases, or obtained using different models and approaches42,189.1.3 Objective and research questionsThe objective of this thesis is to explore and advance the emerging field of the WF by focusing onmulti-level decision-making resulting from the two main WF approaches, and by concentratingon agricultural products that (1) rely on green water resources, (2) have historically depended on landuse change to increase production, and (3) carry known environmental consequences in its region ofproduction. This objective is driven by three research questions:1. How complementary are the aspects of both WF Network and/or LCA approaches when appliedto agricultural products in a region both reliant on green water resources and land use change toexpand production?2. How do WF approaches address the partitioning of precipitation into green and blue water re-sources and what insights might these approaches bring to land and water management?3. What policy decisions may be informed by both WF Network and LCA approaches in the contextof expanding tropical agricultural production, and how may these decisions be complementary inthe management of land and water resources?Within this context, I seek to also investigate the strengths and limitations of WF approaches in address-ing water management for agricultural products in a region that has experienced agricultural expansion.To answer these questions, I will focus on Southern Amazonia (SAM), more specifically the Brazilianstate of Mato Grosso (Figure 1.1) which has been a global center for the production of soybean and beefcommodities exported nation- and worldwide.1.4 Land and water management in Southern AmazoniaFollowing initial opening of land for pasture in the Amazon and Cerrado (or savanna) biomes (Figure1.1), additional cropland expansion began in the 1970s starting in southern Brazil and moving northward5Figure 1.1: Southern Amazonia (SAM) and the Brazilian state of Mato Grosso with the boundaries ofthe Amazon, Cerrado and Pantanal biomes143. Reprinted with permission of Elsevier Ltd.closer to the Amazon63,239. In the 1990s, the advance of soybean into the Cerrado replaced pasture-land, thereby pushing new pasture further north into the Amazon biome by the mid-2000s18,148 withfurther expansion in central Brazilian states246. Despite historical expansion, deforestation in the regiondropped considerably in the 2000s suggesting a decoupling of land use change with agricultural pro-duction148. This drop was attributed to increased law enforcement for illegal deforestation, limitations inthe access to credit for producers located in regions of greater deforestation rates, as well a “SoybeanMoratorium” (2006) and a “Cattle Agreement” (2009) which imposed restrictions on exports of soybeanand beef produced on previously deforested land175. Today, Brazil is the second largest producer ofsoybean in the world261 with the state of Mato Grosso (Figure 1.1) leading national production with 26Mtons of soybean harvested on 9 Mha of land in 2015 (or close to 30% of Brazilian total production)121.Mato Grosso is also home to over 30 million cattle herded on about 23 Mha of pasture121 making it oneof the largest centers of meat production in Brazil.Agricultural expansion in both Amazon and Cerrado biomes has sparked environmental concernsover the ecological integrity of both biomes as a result of deforestation for cropland and pasture expan-sion. Land use change in the region has impacted biodiversity41, led to greenhouse gas emissions83,179,as well as affected energy partitioning on the land141 thereby altering the water cycle45,188 with po-6tential consequences on terrestrial and aquatic ecosystems38,141. Agricultural intensification since themid-2000s triggered additional concerns regarding the use of natural resources for cropland and cattleintensification. Agricultural production in Mato Grosso is almost entirely rain-fed141 and its productivityhas been projected to decline as a result of climate change and reduced regional precipitation8,184. Assuch, the use of irrigation in SAM remains a viable option as insurance against future uncertainty inprecipitation regimes141.The predicted decline in precipitation is partly due to an imbalance in the atmospheric water bal-ance as a result of regional deforestation184, with evidence now available from extensive modelingwork15,247,248 and precipitation measurements across the Amazon biome32,107. Tropical forest and sa-vanna ET are typically greater than cropland and pasture141 such that widespread deforestation canreduce total water vapour transfer to the atmosphere138,238,246. This reduction can then affect precip-itation in the Amazon Basin and the greater South American continent132. A decline in regional ETcan also delay the onset of the wet season51,278 with possible feedbacks on terrestrial ecosystems inthe region59 thereby triggering a “savannization” of the Amazon forest exacerbated by drought and fireevents237.SAM is an agricultural production center for soybean and beef commodities which are traded withinternational partners who may indirectly affect the use of water resources through their consumption(or supply chain). The application of the WF in the region, and more generally Brazil, has been lim-ited55,88,114,139,160,163,186,187, and mainly focused on quantifying volumes of freshwater needed for agri-cultural products. At the same time, impacts of deforestation and climate change on both carbon andwater cycles have been well documented141, with models predicting increasing local temperatures inSAM regardless of greenhouse gas forcing238. As such, applying the WF to agricultural products inSAM has the potential to extend important gaps in knowledge in the application of the indicator, but alsoinform decision-making for water resources at regional and product levels.1.5 Significance and outline of this thesisSo far, the WF academic literature has been embodied by two separate research communities with littleor no integration of approaches and interpretation of results in the context of water resources. Guidedby the objective and research questions listed above, this thesis seeks to eliminate the current confu-sion originating from the multiple interpretations given to the WF (water productivity, water resourcesmanagement, impact assessment) by proposing a framework for harmonizing current perspectives and7methods. Despite the global importance of green water resources in agricultural production, there hasbeen little advancement in considering green water in the WF literature beyond volumetric estimates(e.g., 1109 m3 of green water per tonne of crop235), especially in SAM. Most, if not all, published resultsrely on crop modeling approaches, based on assumptions that may not be relevant to regional produc-tion. Moreover, the concept of green water scarcity, although well defined in a recent review233, hasyet to be applied extensively with estimates of green water availability117, especially within the contextof land use and land use change. To date, only one study tackling green water scarcity in Amazoniahas been published in the academic literature163. Finally, the consideration of the potential impacts ofland use and land use change on the water cycle in LCA is still in its infancy with several methods beingdeveloped independently and little progress towards integration205.SAM, and specifically the state of Mato Grosso, has all the attributes of a region of interest for applyingWF approaches to answer the above research questions: (1) agricultural production is almost entirelyrain-fed and therefore reliant on green water resources141, (2) the increase in agricultural productionhas mainly benefited from land use change in both Amazon and Cerrado biomes141, (3) the region is aglobal center for agricultural production exported internationally139. To increase production, the region isfaced with a suite of land and water management options141 (Figure 1.2) which could be informed byWFapproaches: agricultural expansion into natural ecosystems (Figure 1.2, panel A), cropland expansioninto current pasturelands (Figure 1.2, panel B), a more intensive use of green water resources on currentagricultural land through promotion of transpiration over evaporation, or “vapour shift” (Figure 1.2, panelC), rainwater harvesting or irrigation (Figure 1.2, panels D and E, respectively).This thesis describes original research aimed at advancing the emerging field of the WF in five chap-ters. Chapter 2 outlines the framework used in this thesis to harmonize WF assessments to be appliedto agricultural production and expansion in SAM. Chapters 3 and 4 seek to estimate the volumetric WFof soybean and cattle through field measurements and modeling. Chapter 3 describes field measure-ments of ET using eddy covariance in Mato Grosso and used to obtain the volumetric WF of soybean aswell as as maize, rice and bean crops. Chapter 4 provides an estimate of the volumetric WF of cattleusing production system modeling, and remote sensing. Chapter 5 adopts a river basin view in theapplication of the WF to inform decision-making for water resources management in the Xingu Basin ofMato Grosso considering current and future land and water resources use for agricultural production.Chapter 6 addresses environmental impacts of cropland and cattle production using new and existinglife cycle impact assessment models. All results are then integrated in Chapter 7 for policy formulationat the Brazilian agricultural frontier, and an analysis of strengths and limitations of individual WF assess-8ments. Appendices A, B and C provide supplemental information respectively for Chapters 3, 4 and5.Figure 1.2: Land and water management options available in Southern Amazonia (SAM) to increasefuture agricultural production, and their consequences on the partitioning of precipitation (P, assumedto be 2000mm y-1 in the above example) into blue (blue arrows) and green water (shown as ET) followinginitial conditions141. Values shown in the panels are water flows (in mm y-1) and the green arrow in panelD represents 300 mm y-1of harvested rainwater. Reprinted with permission under Creative CommonsAttribution 3.0.9Chapter 2The Harmonized Water FootprintAssessment2.1 IntroductionThe efficient, equitable and sustainable management of our planet’s water resources is one of the mainchallenges humanity is currently facing. For example, 1.2 billion people experience physical waterscarcity260, while close to four billion people worldwide live under extreme water scarcity at least somemonths of the year159. By 2050, the number of people living under medium and severe water stresscould reach 5 billion, with water demand more than doubling for domestic, livestock and electricity281. In2015, the United Nations launched the Sustainable Development Goals (SDGs) with a roadmap to 2030which includes objectives for clean water and sanitation (SDG 6) to “ensure availability and sustainablemanagement of water and sanitation for all”280. While SDG 6 is specific to water and sanitation, othergoals also include water, either directly (e.g. SDG 14: Life under water, SDG 15: Life on land) or indi-rectly (SDG 2: Zero hunger, SDG 7: Affordable and clean energy, SDG 12: Responsible consumptionand production), therefore requiring a wide range of governance strategies, measures and indicators toensure that these goals are met.The relation between human-beings and nature in the context of water management has evolvedthrough the centuries, but recently this interaction has been embodied by the IWRM and Water Securityconcepts, which themselves have evolved over time. The introduction of IWRM in the 1992 DublinInternational Conference onWater and Development outlined the necessity to integrate knowledge aboutthe complex physical interactions over the full water cycle (e.g., by considering surface and groundwaterinteractions), as well as between the water cycle and society231. Savenije et al.231 describe an evolutionin IWRM towards coordinated action among stakeholders sharing water resources in a given geographicspace (e.g., a river basin) and a period of time (e.g., the hydrologic year) with trade-offs to be weighted soas to “maximize the resultant economic and social welfare in an equitable manner without compromising10the sustainability of vital ecosystems” following the definition from the Global Water Partnership95.Over the past decade, the concept of Water Security has been gaining attention as an importantparadigm for water resources management and, in some cases, may be considered an extension ofIWRM17. In 2013, the United Nations Water Task Force on Water Security defined it as “the capac-ity of a population to safeguard sustainable access to adequate quantities of acceptable quality waterfor sustaining livelihoods, human well-being, and socio-economic development, for ensuring protectionagainst water pollution and water related disasters, and for preserving ecosystems in a climate of peaceand political stability”259. Many other definitions have been proposed and used in specific contexts285,but more generally, Water Security implies a cross-sectoral influence of water in social, economic, eco-logical and political layers affecting individuals and society. While there are many overlapping concernsbetween IWRM and Water Security perspectives, Bakker and Morinville17 describe that Water Securityfurther implies: (1) the protection of water resources, (2) the idea of a threshold that may affect socio-ecological resilience, and (3) the necessity to respond to risks given imperfect information about waterresources with an emphasis on adaptive management.Along with the evolution of thought regarding the relationships between humans, society and waterresources is the notion of scale of action and the increasing importance of global structures affect-ing the water cycle. Water resources management is particular in that local management has globalimpacts, while at the same time, local and global forces can constrain present and future local waterresources272. Increases in extreme precipitation events and localized droughts resulting from globalclimate change103 can affect local water availability. Local flood or physical water scarcity can affectlocal food production with consequences on global food prices. Likewise, inter-basin transfers, the ef-fects of the global economy on water quantity and quality impose additional stresses on water resourcesby actors that are not using local water resources directly. For instance, production and consumptionactivities represent a large portion of hidden water use for trade, requiring an additional considerationof water use efficiencies in distant watersheds110. This indirect water use (supply chain use, or wateruse crossing a production to consumption boundary) has important consequences in consumption andproduction activities, especially given that water withdrawals typically occur in stressed watersheds217.The introduction of the WF concept in 2002119 brought to light an important connection betweenproduction and consumption activities, and water resources. In its original definition, the WF quantifiesvolumetric freshwater use of a product or a service by summing direct (or operational use) and indirect(or supply chain use) water consumption117, thereby highlighting the link between the consumption ofproducts and the global water cycle114. The WF can address various aspects of SDG 12 such as: 12.211“achieve sustainable management and efficient use of natural resources”, 12.4 “achieve the environ-mentally sound management of chemicals and wastes throughout their life cycle (. . . )”, 12.6 “encouragecompanies (. . . ) to adopt sustainable practices (. . . )”, and 12.7 “promote public procurement practicesthat are sustainable (. . . )258. When dealing with water specifically, the above goals then have repercus-sions for other water related SDGs (e.g., SDG 6, SDG 14) depending on how the WF is determined.There are currently two distinct and complementary approaches to the WF, each of which followsspecific steps with a focus either on water resources management or impact assessment29. The WFhas been described as a freshwater volume which can be compared to total sustainable limits within aboundary following steps published by the WF Network117. In addition, the WF has also been referredto as a freshwater volume or an environmental impact to be compared to a benchmark to illustratewhether a product or activity is more or less sustainable111. When impact assessment is of concern,the WF then follows the ISO 14046 standard127. Despite these differences in perspectives, these twoWF approaches make an important connection between the physical boundary of the natural resourceand the boundary of production systems that compose an integral part of the economy. Conclusionsfrom these WF approaches can then highlight actions that Society can take with respect to water uses.However, most WF assessments are typically carried out following one of the two mentioned ap-proaches without individually addressing the full scope of water resource decision-making. This chapterseeks to combine the two main WF perspectives into one harmonized WF assessment. Rather thanfocusing on parallel approaches as described by Boulay et al.29, I highlight the type of decision-makingthat follows each assessment. I propose to associate WF decision-making into two groups associatedto two boundaries linked to distinct “domains”, based on the level of intervention that each decision car-ries on the water cycle: (1) a physical boundary represented by the “Nature” domain with a focus onwater resources management, and (2) a production system boundary represented by the “Production”domain, that is nested within the Nature domain but with a focus on water use and impact assessment inproduction processes. I argue for a more integrated discussion around water resources decision-makingas they relate to different boundaries in the domains to differentiate actors and their specific actions inthe water cycle. The importance of complementarity between the main two WF approaches has alreadybeen highlighted in the literature29; I extend this existing complementarity through a proposed harmo-nized WF assessment. I first provide a description of the Nature and Production domains and theirconsiderations of water resources (Section 2.2) before describing how each domain is addressed bydistinct aspects of current WF approaches (Section 2.3). The proposed harmonized WF assessmentconstitutes the framework for work presented in the remainder of the thesis.122.2 Nature, Society and Production domainsWater resources connect Nature and Production through Society. Society’s actions on water resourcesaffect the environment with feedbacks that are better identified than those of climate change166. Natureis what is designated as containing the global hydrological cycle and water in all of its states (liquid,solid and gaseous) and storage locations in the environment, whether natural (e.g., lake, aquifers) oranthropogenic (e.g., reservoirs). Production is nested within Nature and encompasses activities thatare specific to humans, and are connected to the local and global economies. Production specificallyembodies agricultural and industrial production processes with the objective of making products thatare traded and consumed by Society. Society represents all human activities which connect Nature andProduction domains (Figure 2.1).Society uses water resources directly (e.g., for drinking, cleaning, etc.), but also indirectly by con-suming products and services through Production116 (Figure 2.1). For instance, while cotton irrigationconstitutes a direct water use in Production, this use can be embedded in a cotton t-shirt consumedby Society to represent an indirect water use. Water that is withdrawn and not returned to the water-shed due to evaporation during use, product integration, inter-basin transfers, or direct release into thesea is considered consumed20. Water consumption is differentiated from water withdrawals which in-clude return flows. These return flows typically accompany water use by Society with releases of waterback into Nature, often at a different quality than what was abstracted. Inflows of water into Societyare typically labeled as domestic water use, but also agricultural and industrial water uses when thoseinflows enter Production, although detailed water statistics among sectors of the economy are often notavailable108. Moreover, many strategies to reduce water use are promoted through campaigns that aimto raise awareness on direct domestic uses, thereby leaving out important water saving strategies thatcould be implemented in other areas of Society, such as Production. For instance, the global averageper capita WF of consumption was 1385 m3 y-1 in 1996-2005, of which 3.8% represented domesticwater uses114. Similarly, while companies and utilities may report their direct water use, water that isindirectly consumed is often unreported145.Water resources within the purviews of Nature, Production and Society domains (Figure 2.1) are thefocus of different academic fields of natural and social sciences, and engineering, each of which con-sider distinct boundaries and require different skills for the study of water resources and water resourcesmanagement (Table 2.1). When exclusively considering the natural sciences and engineering, the studyof water in Nature typically means focusing on the natural water cycle traditionally represented by the13connection of a variety of flows in the landscape. These flows have a range of magnitudes and residencetimes that differ based on natural processes and physical and chemical states of water in the environ-ment. As such, groundwater is only differentiated by its residence time, where “fossil” groundwater canbe thousands of years old compared to shallower aquifers containing “modern” groundwater94. Hydrol-ogy, and more recently ecohydrology, is a primary field of study for water in Nature, which has beenexpanding into fields within ecology or biology as implications of water resources on ecosystem quality,functions and services require the consideration of incremental overlaps of hydrological processes withother processes within Nature. Physiographic boundaries for the study of water in Nature typically relateto a watershed within a river basin that includes monitoring stations, with water flows evaluated usinghydrological models and, more recently, remote sensing, to provide a detailed description of the watercycle in space and time166. Typical research questions are often bound to the quantification of waterquantity and quality through in-situ data monitoring or field sampling to gather information for hydro-logic models, or to test for effects of human activities such as agriculture, forestry and mining on thelandscape.Water resource considerations change slightly when moving from Nature into Production. Waterflows are considered as inputs to and outputs from production systems that are designed by engineerswithin the typical boundaries of a production facility, several interconnected factories and/or businesseslinked through the global supply chain116. The technical nature of processes in the Production domainis dominated by the general field of engineering, where principal water concerns relate to efficiencies inoverall processes, as well as the output of production. Water may be viewed as an input into the produc-tion system for the direct integration into a product, or as an input for the production of such a product(e.g., cooling systems). Typical research questions are related to the improvement of the efficiency ofwater use in the production system (e.g., water productivity of cropland irrigation) with the objective ofhaving greater production output per unit of water input in relation to management decisions or techno-logical solutions. Improvements in efficiency can also include economic and environmental efficienciesconsidering economic output and environmental impact per unit water input. Similarly, techniques re-ducing the amount of waste released during production processes or treatment of effluent returning toNature are of concern. Typical research questions relate to engineering efficiencies of systems, as wellas recycling and reuse of water effluents, and the water quality of any return flows to Nature.Society links both Nature and Production through water resources management and water gover-nance. Water resources are often viewed as stocks (e.g., reservoirs, aquifers, rivers) connected throughflows which are managed to secure water availability for Society. While the natural sciences may focus14on temporal water supply and demand, engineering focuses on water works for supply in relation tolocal demand, with treatment plants or storm drainage management systems as a technological supportfor meeting water quality objectives. Information gathered in Nature provides important data for waterresources management and the physical boundaries that might limit Production in a given geographiclocation. Water use by Society is dictated by water management and governance structures which couldbe limited by national or sub-national borders, municipal boundaries or hydrological units defined by thewater resource itself (e.g., watershed, aquifer). Typical research questions relate to future supply anddemand of water resources in a region considering scenarios that could include climate change effectson the local hydrological cycle, or changes in local demand through population growth and economicactivity. These questions depend on data monitoring networks as well as hydrological models with afocus on management.The above described Nature, Production and Society domains are only a guide for what can beconsidered to be typical descriptions of the water cycle for water resources management. With Produc-tion being nested into Society, itself nested into Nature (Figure 2.1), we see a multi-layered structurefor water decision making focused on the Nature and Production domains according to their respectivesystem boundaries. In the Nature domain, boundaries are physical, while those of Production dependon the production system and may involve several farm fields, factories and products located in differentgeographic locations (Table 2.1). Water resources decision-making in Production therefore carries con-sequences in Nature that operate through the more complex connections of the global economy. Wecan use the nested relationship of Nature and Production domains to harmonize existing views aboutthe WF, described next.2.3 Linking Nature and Production through Society with waterfootprints2.3.1 The volumetric water footprintSince 2002, the WF has grown as a new research field by filling an important gap in knowledge andconnecting the boundaries between Nature and Production domains through the quantification of wateruse for production and consumption processes113. The quantification of the WF in volumetric terms,which I refer to here as the volumetric WF (VWF), represents the amount of freshwater consumed fora production or consumption activity, focused specifically on liquid water (blue water) and soil moisture15Figure 2.1: The relationships of water in Nature, Production and Society domainsTable 2.1: Main considerations for the Nature, Production and Society domains (Figure 2.1) with focuson the natural sciences and engineeringDomain Fields Water cycleinterpretationSystemboundaryWater useNature Hydrology,ecohydrology,ecology, biology,etc.Flows and residencetimesRiver basin,watershed,landscapesEnvironmental quality,ecosystem servicesSociety Water resourcesmanagement,water works,engineeringIn- and out-flowsbetween water stocksand outflows back toNatureCountry, state,city, etc.Direct agriculture,industry and domesticusesProduction Engineering Input to a productionsystem and output toNatureField, factory,global supplychainDirect use inoperations andindirect use in thesupply chain16regenerated by precipitation (green water), as well as water required to dilute a chemical or thermal pol-lution load to background levels of water quality standards (gray water)117. Following initial discussionsabout virtual water trade by Allan1, the VWF quickly gained traction in research. In particular, the VWFhas highlighted important trade connections of water intensive products through elucidation of the virtualwater trade network114, its evolution over time56 and economic aspects183.The VWF is synonymous to a WF Inventory in the ISO 14046 standard127. This quantification steprelies on the analysis of a unit process considering the entire life cycle of the products and activitiesentering the unit process, from resource use all the way to disposal or recycling105. This focus thereforerequires detailed knowledge about production systems in their entirety, which often means involvingseveral sub-processes (e.g., cooling, transportation, packaging), and carefully selecting what should beincluded in the system under study and what could be considered a background process (e.g., energygeneration for the production process). The focus on production systems requires detailed databasesof product systems according to production processes and geographic locations such as the Ecoinventdatabase ( whose recent update includes detailed water information194. As such,the VWF (or WF Inventory) focuses exclusively on the Production domain. Water is often treated as aninput to the production system (e.g., irrigation use for crop production) with a VWF typically expressed asa volume per unit output (or known more generally as a functional unit following ISO 14044126), whichcould be represented by a mass or economic output (e.g., m3 of water per tonne of crop, m3 of waterper dollar of output). The VWF (or WF Inventory) is the starting point of the two main WF approaches,which I propose to harmonize here into one WF assessment considering the domain focus of each ofthe following steps defined in Table 2.2: (1) Goal and scope definition, (2) VWF accounting, (3) VWFassessment, (4) WF impact assessment, (5) VWF sustainability assessment, (6) policy decisions. Iemphasize here the differences in terminology used in this proposed assessment with previous stepsfrom Hoekstra et al.117 and ISO 14046127. These differences are described in detail below, as well asother potential uses of the WF, especially in LCA, which can be more general in scope105.2.3.2 The volumetric water footprint assessmentThe VWF assessment follows the quantification of the VWF (or WF Inventory), and parallels what hasbeen described as water productivity (as the inverse of the VWF), a commonly used metric which hasbeen employed to highlight water efficiencies in agricultural production, and should include economicefficiencies and multiple benefits of the production system93. The VWF assessment can highlight effi-ciencies in the production system when compared to individual product VWF benchmarks158, but also17global efficiencies in water consumption for production and consumption activities when considering vir-tual water trade56. Virtual water trade analysis has been studied globally56,114, intra-nationally57,266, orconsidering virtual water trade balances of countries68,232, and more recently, cities268 (for a full list ofstudies, see Ercin et al.69) to highlight global water efficiencies based on differences in VWF of tradedproducts, while recognizing that trade decisions should not be focused exclusively on VWF113.2.3.3 The water footprint impact assessmentThe WF impact assessment (WFIA) relates to the field of LCA whose goal is to quantify environmentalimpacts of production and consumption activities. The inclusion of water use in LCA emerged in 2008with the proposal that water consumption and degradation activities carry environmental impacts thatshould be quantified134. LCA is a scientific method which relies on the logical sequence of a cause-effectchain that connects resource use to potential impacts from a production or organizational standpoint105.Impacts are considered on a relative basis since real impacts often cannot be explicitly measured; rather,LCA models rely on previous work in fields such as ecotoxicology, water chemistry or epidemiology toderive models that quantify a level of impact with a resource use and emissions release. The WFIA isbut one step of a WF assessment as defined by ISO 14046127: (1) Goal and scope definition, (2) WFInventory, (3) WFIA, and (4) interpretation (following the terminology of the standard). In step 2, theWF Inventory serves as a building block for characterization of impacts (using characterization factors)which relies on models that can translate the volume of water consumed and pollution released during aproduction or consumption activity into an explicit impact quantified per functional unit. These impactsare expressed in terms of either mid-point impacts (e.g. eutrophication, acidification, etc.) or end-pointimpacts classified within human health, ecosystem quality and natural resources. One can thereforeimagine an exhaustive suite of impact assessment models given the complexity of the effects that waterconsumption and degradation may entail. For instance, Pfister et al.193 proposed three end-point impactassessment models based on the effects of potential deprivation of water consumption on human health,ecosystem quality and water resources. These models respectively express the effect of reduced wateravailability on crop irrigation leading to potential nutritional losses, declines in net primary productionleading to environmental degradation, and water resources more generally leading to a rise in energydemand for desalination193. Many other models have been proposed to describe impacts to humanhealth28,169 and environmental quality181 but overall model integration remains needed. Here, I useWFIA to attribute the quantification of impacts as they relate to a well-defined functional unit linked to aproduction system (e.g., 1 tonne of agricultural product, 10,000 hand dryings) representing the ultimate18use of the product or activity of study within a well-defined production system boundary.As such, and similarly to the VWF assessment, I focus theWFIA on the original product-based scopein LCA related to the Production domain despite wider emerging LCA scopes recently proposed105. Theguiding standard are ISO 14044126, and ISO 14046127. ISO 14046127 provides principles, requirementsand guidelines on how to conduct such an assessment when considering water quantity alone (termed“water scarcity footprint”), or when considering both water quality and quantity (termed “water availabilityfootprint”) The volumetric water footprint sustainability assessmentOver the years, research focus on the quantification of VWF of single processes and activities has beenslowly replaced by the VWF sustainability assessment (VWFSA) aimed at assessing the sustainableuse of water resources113. This assessment is one step of a four step process guided by the manualreleased by the WF Network117: (1) Goal and scope definition, (2) WF accounting, (3) WF sustainabilityassessment, and (4) policy recommendation117 (following the terminology of the manual). In step 2,the VWF serves as a building block to obtain the total water consumed within the system boundary,defined either geographically (river basin, country, etc.) or within the boundaries of a business, andinclude virtual water trade across the boundary113. However, the ultimate goal of the VWFSA is torelate water consumption to maximum sustainable limits116. Thus, in step 3 of the assessment, thesum of VWF of all processes and activities taking place within the study’s boundaries is compared towater availability, which, in the case of blue water is defined as the natural runoff minus environmentalflow requirements117. While step 2 requires intimate knowledge of the Production domain (describedabove), step 3 requires detailed knowledge from the Nature domain derived from the natural sciences fora detailed picture of the watershed or river basin system, its ecosystems and vulnerabilities as they relateto water quality and quantity in the region. As such, the VWFSA contextualizes what was focused onthe Production domain, but includes background information related to the Nature domain (Figure 2.1).The assessment is mainly guided by the premise that the main issues of concern are the sustainableand equitable use of water resources113 with implications on water resources management locally andglobally272. The VWFSA has provided information on the sustainability of water use in major basins ofthe world114,159, with some indication of demand side management, such as the implications of dietson water resources266–268. Solutions are aimed at guiding policy incentives to reduce VWF with theintention of improving global water efficiencies without consideration of potential differences in wateravailability among regions within the context of existing virtual water trade networks.192.3.5 Implications of the proposed harmonized water footprint assessmentFollowing the above descriptions, different WF assessments express different perspectives based onwhether the domain focus is on Production or contextualized within the Nature domain (Table 2.2). De-spite many overlapping goals29, the development of the WF field has involved two main communitiesrepresented by the WF Network (focused on the VWF assessment and the VWFSA), and the LCA com-munity (focused on WFIA), with important academic debates mainly focused on how to consider waterscarcity89,112,115,195. Combining the perspectives into each domain focus reveals how all assessmentsmentioned thus far could be combined into one harmonized WF assessment (Figure 2.2). First, theVWF assessment as well as the WFIA are exclusively focused on individual actions on the productionsystem (micro level decision-making) and rely on the definition of the functional unit in the Productiondomain. The information revealed by each assessment relates to individual process improvements inwater consumption and degradation or a reduction in potential impacts, meaning that solutions are in-herently focused on the unit process. Quantified results obtained from VWF assessment or WFIA aretypically compared to results obtained for the same functional unit produced in a different context (e.g.,soybean produced in one system compared to another) in order to highlight potential improvements inenvironmental performance to the production system105,111. When seeking to scale up VWF assess-ment or WFIA results to larger production quantities, values increase proportionally with the amount ofactivity or process under study (e.g., a volume or impact per tonne of product is 1000 times large thanfor 1 kg of product), and do not consider other activities or processes with different functional units inthe analysis.The VWFSA requires a scale-up of the unit processes or activities within a defined boundary (e.g.,river basin, country, business, etc.) to provide information on water use and degradation at a greaterscale (macro-level decision-making) and considering other processes111. In this context, which is out-side Production but still part of the Nature domain, the VWFSA parallels other assessments as part ofthe emerging Environmental Footprint Assessment field that is concerned with the human appropriationof resources translated into quantified resource indicators (e.g., m3 of water, CO2 emitted, m2 of land)to assess environmental pressure from human activities116. Sustainability limits have been expressedin terms of Planetary Boundaries223, which for water represents 4000-6000 km3 y-1 according to Rock-ström et al.224 with a limit of 1100-4500 km3 y-1 for blue water based on environmental flow requirementconsiderations90. Current levels of green and gray WF have been estimated at 6700 km3 y-1 and 1400km3 y-1 respectively, without however being associated with a sustainable limit116.At least two main issues need to be mentioned regarding the proposed harmonized WF assessment:20Figure 2.2: Proposed harmonized water footprint (WF) assessment and terminology combining ap-proaches from the WF Network117 and life cycle assessment (LCA)126,127 within the Production andNature domains.21Table 2.2: The harmonized water footprint (WF) assessment, step by step.Stage Name Step1 Goal and scopedefinitionDefine the objectives of the study, the functional unit andgeographic extent, the intended audience and how resultswill be used. This step follows step 1 of current WFapproaches117,1272 Volumetric WF, orWF InventoryCalculate the amount of water consumed (and polluted)for the unit process under consideration. Water resourcesare defined in terms of green and blue water which areaccounted separately.3 Volumetric WFAssessmentCompare results of stage 2 to a geographic benchmarkwith similar technology and identify possible water savingsto improve the efficiency of water use in the productionsystem.4 WF ImpactAssessmentCharacterize results from stage 2 using a factor thattranslates the water consumed (and polluted) into aquantifiable potential impact using characterizationfactors. Compare these impacts to a benchmark for asimilar product, or the same product in a differentproduction system or geographic region to identifyimprovements in the production system. This step followsthe ISO 14044126 and ISO 14046127 standards.5 Volumetric WFSustainabilityAssessmentAdd volumetric WFs of all unit processes (separatinggreen from blue) within a geographic extent or businessnetwork and compare to water availability (green and blue)to identify allocation of water resources within the studyboundary. This step follows the WF Network manual117.6 Policy decisions Integrate findings from stages 3 to 5 and providerecommendations for the production system (perfunctional unit, from stages 3 and 4) and the geographicextent (from stage 5). Conflicting decisions should behighlighted with potential cost and benefit analysis.22the normalization of language in both Nature and Production domains, and the consideration of waterscarcity. As it stands today, WF practitioners use different language based on whether their work fo-cuses primarily on the Nature or Production domains (Table 2.3), reflecting the main fields associatedwith each assessment (Table 2.1). The most common difference, which has led to miscommunicationbetween the WF communities, relates to the use of the term “flow”. In the Nature domain, this termis synonymous to what is used in hydrology and water resources management as a transfer of waterbetween hydrological stocks (Table 2.1). For instance, the consumption of soil moisture through ET forthe agricultural production of cotton represents a flow of water from the biosphere to the atmosphere.However, in the Production domain, a “flow” refers to inputs to and outputs from a production system.Water resources are seen therefore as an “input from nature” entering the production system with an“output to nature” represented by a reduction in water availability (or an amount of water released at adifferent quality than that which entered the production system), with potential consequences on humanhealth and the environment. Following the cotton example above, soil moisture therefore representsa flow into the cotton production system, with a release to the atmosphere. Knowledge of such differ-ences and terms are important to improve communication between WF communities when combiningassessments into one harmonized WF assessment.Secondly, the debate about water scarcity and how to represent it in either absolute or relative termsis a major point of disagreement between the WF Network and LCA communities, and whether waterscarcity should be “weighted”112,195, which also reflects the focus of the primary domain (Productionand Nature) in the respective analyses. The “weight” of water scarcity emerged with the premise thatwater availability should reflect local geographical realities of water stress in order to quantify differentenvironmental impacts with production activities. Pfister et al.193 used a water scarcity index (spanningfrom 0, or no stress, to 1, full stress) as a weighting factor to estimate end-point impacts to human health,ecosystem quality and natural resources. Similarly, the more recent water stress indicator for AvailableWAter REmaining (AWARE)31 presents an updated version of this index by providing a quantification forthe amount of water remaining in a basin (or a country) after human and ecosystem demands have beenmet. In fact, these so-called “stress indicators” or “weighting factors” are analogous to characterizationfactors in a WFIA. This approach is therefore fundamentally different than the VWFSA derived scarcityindex whose focus is on the sustainable limits in the Nature domain through a macro-analysis, with noreference to any functional unit.23Table 2.3: Summary of considerations of the main stages in the harmonized water footprint (WF) as-sessment (Table 2.2)Stage 2. Volumetric WFassessment3. WF impactassessment4. Volumetric WFsustainabilityassessmentGuiding approach inthis thesisNone specified ISO 14046127 WF Network manual117Basis for analysis Unit process or activity Unit process or activity Sum of unit processesand activities with aboundary defined in the“Goal and Scopedefinition” stageConsideration foranalysis of waterconsumedEfficient use of water inthe process or activityEffects of waterconsumption anddegradation on humanhealth and theenvironmentComparison of totalwater use compared tosustainable limitsexpressed by wateravailabilityPrimary domainfocusaProduction Production NatureMain objective Improve global or localwater efficiency per unitprocess or functionalunitReduce localenvironmental impactsper unit process orfunctional unitEnsure sustainable andequitable use of waterwithin global limitsPolicy directives Identify water useefficiencies inproduction systemsbased on comparativeassertionsIdentify environmentalimpact hotspots basedon local water scarcityand qualityIdentify water resourceuse efficiencies andimprove sustainablewater resourcesmanagementComparativeassessmentBenchmark of wateruse per unit process orfunctional unitComparison ofenvironmentalperformance per unitprocess or functionalunitComparison of waterconsumption with wateravailabilityLanguage Related to systemsanalysis and thedescription ofproduction systemmodelingRelated to systemsanalysis and thedescription ofproduction systemmodelingRelated to hydrologyand the description andmeasurements ofnatural processesaSociety is expected to be influential on both Nature and Production domains242.4 ConclusionThe harmonized WF assessment can now be used in SAM following the steps described above (Table2.2) represented by the individual chapters of this thesis. First, I carry out a VWF assessment followingVWF measurements of cropland (Chapter 3), and VWF modeling of cattle (Chapter 4) in SAM. Then, Ifocus on decision-making for water resources management in the Xingu Basin of Mato Grosso after car-rying out a VWFSA of the basin (Chapter 5), before finally returning to the Production domain by lookingat WFIA of both cropland and cattle in the Xingu Basin of Mato Grosso (Chapter 6). The conclusion ofthis thesis (Chapter 7) integrates all results and formulates policy responses based on results from theindividual phases of the harmonized WF assessment.25Chapter 3Measuring the Volumetric WaterFootprint of Crops through WaterProductivity3.1 IntroductionBrazil has been the center of international attention for its rapid increase in agricultural production. Be-tween 1990 and 2015, the total area planted to non-perennial crops increased from 46Mha to 71Mha121,mostly driven by commodities such as soybean, maize, and sugar cane which, together, represent 90%of cropland area63. As the leading crop, soybean production almost tripled from 20 Mtons produced on12 Mha of land in 1990121 to an estimated 104 Mtons produced on 34 Mha of land in 2017261, makingBrazil the second largest producer in the world closely behind the United States (117 Mtons in 2016)262.To increase its agricultural output, Brazil has historically relied on both land use change and increasesin yields63. Soybean cropping areas have increased from southern to northern Brazilian states into theCerrado and Amazon biomes18,63,239, and soybean yields almost doubled from a mean national yieldof 1.7 ton ha-1 in 1990 to 3.0 ton ha-1 in 2015121.SAM is the largest producing soybean region in Brazil, with production concentrated in the stateof Mato Grosso and its Amazon and Cerrado biomes (Figure 1.1) with a predominance of Oxisols inthe region150. Agricultural expansion has been more evident in this region with the rapid conversionof humid tropical forest and savanna landscapes into soybean and pasture, both of which have beenproduced almost exclusively under rain-fed conditions18,148. At the same time, the rapid growth of doublecropping systems (i.e., two crop cycles per year within the same field) has allowed further intensificationof agricultural output by planting maize, cotton or rice immediately after the soybean harvest to takeadvantage of the end of the wet season11. Maize cultivation as a second annual crop expanded rapidly26in Mato Grosso between 2001 and 2011, with total area increasing from 0.5 Mha to 2.9 Mha during thisperiod245.Research on the impacts of SAM’s changing land use and land cover has largely focused on dy-namics between agricultural production and deforestation18,97,148, indirect land use change dynamicsbetween soybean and pasture expansion10, as well as regional greenhouse gas emissions from de-forestation83,84 linked to agricultural output129,179,283. In addition, land use change effects on the localwater cycle have been described in relation to impacts to water quantity63 and quality173,220, stream tem-peratures149, regional scale effects on water yields38, and water vapour flows to the atmosphere141.Forest-to-cropland and forest-to-pasture transitions are typically accompanied by a drop in landscapeET, which, when accumulated across the landscape, can reduce water vapour transfers to the atmo-sphere141,238. This change in atmospheric feedback can in turn affect surface temperatures200,238, aswell as regional precipitation recycling15, with potential effects on natural ecosystems and rain-fed agri-culture59,184. Changes in regional ET with land use change have been quantified at the multi-state,state and river basin levels138,238,246. Lathuillière et al.138 found that tropical forest contributions to totalET in Mato Grosso dropped 10% between 2001 and 2009 (from 593 km3 y-1 to 474 km3 y-1), with totalcropland ET returning about 180 km3 y-1, or 15% of all water vapour flows to the atmosphere in 2009.Direct field measurements of ET are still lacking in SAM, especially for cropland and pasture. Re-search efforts in the 1990s led to an initial network of eddy covariance towers in Brazil to be installed tomeasure carbon and water fluxes in natural ecosystems of the Amazon biome53,54,98,120,130,151; this net-work has since expanded to other biomes in SAM226,274. Such direct ET measurements can elucidatethe effects of modeling assumptions on modeled ET already achieved for tropical forest vegetation43which should be repeated for cropland and pasture. However, no direct ET measurements have beenpublished to date to evaluate the magnitude of and controls on ET for typical agricultural systems in theregion, and potential differences between rain-fed and irrigated systems.In this study, we used eddy covariance to measure cropland ET with a micrometeorological tower lo-cated between two adjacent fields (rain-fed and irrigated) with three objectives: (1) to provide a detailedwater balance of SAM cropland containing soybean, (2) to measure crop characteristics and crop coef-ficients for crop modeling purposes, and (3) to explore differences in crop transpiration and productivitywith irrigation practices through crop modeling. In addition to providing key observations for future land-atmosphere and crop models, our results provide insight into differences between current agriculturalproduction based on rain-fed cropland with potential future production practices using irrigation.273.2 Materials and methods3.2.1 Site descriptionThe research site consists of a seven-meter tall micrometeorological tower installed at Capuaba farm(13° 17’ 15.036” S, 56° 05’ 17.354” W, 427 m altitude) in the municipality of Lucas do Rio Verde, MatoGrosso (Figure 3.1). The 1500 ha farm is located in the Cerrado biome and was established in thelate 1980s after clearing natural vegetation. The farm produces soybean (Glycine max) as the primarycrop, and maize (Zea mays) as the secondary (or double crop), but also produces rice (Oryza sativa)and bean (Phaseolus vulgaris) as well as other cover crops based on the time of year (e.g., Brachiararuziziensis). General farming practices consist of minimizing soil disturbance through direct seeding (orno-tillage) and the use of a wide variety of cover crops that change annually. Soil at the site is a red-yellow latosol (Haplustox) with a clay texture average of 59.4% clay, 29.4% silt and 11.5% sand over the0-0.20 m depth following Embrapa-CNPS67. The soil pH is 5.4 (in water, sampled in March 2017) withan average organic matter content of 26.38 g dm-3 (0-0.20 m depth, sampled in March 2017), and anaverage soil bulk density of 1.09 g cm-3 (0-0.18 m depth, sampled in April 2017). Data from the closestmeteorological station located in the municipality of Diamantino, Mato Grosso (14° 40’ S, 56° 27’ W)showed a 1999-2011 average precipitation of 2144 mm y-1 separated into a wet season (October-April,1982 mm) and dry season (May-September, 162 mm) with occasional cold fronts coming from the southin the austral winter when temperatures can temporarily drop to 16 °C, well below the mean annualtemperature of 26.7 °C124,139The micrometeorological tower (hencefourth the “Soyflux” station) was positioned on flat terrain adja-cent to three fields under different management and crop rotations (Figure 3.1, Table 3.1): “Rainfed-1” lo-cated north-east of the tower is the main rain-fed field where soybean is typically planted at the beginningof the rainy season and double cropped with maize following the soybean harvest; “Rainfed-2” is locatedimmediately north north-east of the tower and is closest to the station where soil and spectral sensorswere installed with similar rotations as in Rainfed-1; “Irrigated” is a 136 ha field equipped with a centralpivot irrigation system located south-west of the station in which soybean is typically planted ahead ofthe rainy season, followed by rice and bean in the dry season. Crops were generally planted 0.05-0.06 m(soybean), up to 0.30 m (maize) apart in 0.50 m separated rows resulting in about 67,000-400,000 plantsha-1depending on the crop. Crops planted in the Irrigated field were only provided with irrigation in thedry season and until the onset of the wet season (September-October), while during the rest of the year,crops in the Irrigated field were rain-fed as in the Rainfed-1 and Rainfed-2 fields. The Soyflux station (see28Figure 3.1: Location of the Soyflux site of Capuaba farm in Lucas do Rio Verde, Mato Grosso, Brazil.The inset shows the state of Mato Grosso with the location of the municipality of Lucas do Rio Verde.29Table 3.1: Management and crop rotations at the Soyflux site of Capuaba farm (Figure 3.1).Field Crop Planting date Harvest date Days VarietyRainfed-1 Soybean 28 Oct 2015 11 Feb 2016 106 TMG1180 RRMaize 13 Feb 2016 13 Jul 2016 151 MG652 PWBrachiara 14 Jul 2016 4 Oct 2016 82Soybean 5 Oct 2016 4 Feb 2017 122 M8372R IproRainfed-2 Soybean 8 Oct 2015 1 Feb 2016 116 NS7901 RRMaize 20 Feb 2016 15 Jul 2016 146 MG652Soybean 6 Oct 2016 17 Jan 2017 103 NS7901 RRIrrigated Soybean 29 Sep 2015 13 Jan 2016 106 M8210 IproRice 1 Feb 2016 30 Apr 2016 89 Ana 8001Bean 14 Jun 2016 22 Sep 2016 100 Anfc 9Stubble 23 Sep 2016 29 Sep 2016 6Soybean 30 Sep 2016 4 Feb 2017 127 M8372 IproFigure A.1, Appendix A) was equipped with soil, radiation and canopy sensors alongside an eddy covari-ance system with continuous data collection between 18 September 2015 and 4 February 2017 (includ-ing two soybean harvests). This time period comprises one El Niño (2015-2016) and one La Niña cycle(2016-2017). All sensors (Table 3.2) were connected to a CR1000 datalogger equipped with an AM416relay multiplexer (Campbell Scientific Inc., Logan, Utah, USA) with measurements taken every 30 s andaveraged on a half-hourly basis. Additional information on the Soyflux site can be found on the AmeriFluxwebsite (site BR-CMT): Crop and canopy monitoringCrop height was monitored for two purposes: (1) to mathematically adjust the displacement height forflux calculations, and (2) to complement the development cycle observations of soybean and maizecrops. Crop height was measured with a tape measure (± 0.01 m accuracy) during regular field sitevisits to derive linear models to infer daily displacement heights in the Rainfed-1 and Irrigated fields.To track the development cycle of both soybean and maize, the crop height in the Rainfed-2 field wasmonitored hourly (from sunrise to sunset) using an automated camera system consisting of a Hero 4Camera (GoPro Inc., San Matero, CA, USA) connected to an Arduino Nano microcontroller (Arduino, The camera was aimed at a 2 m tall pole positioned vertically in thecamera’s field view (Figure A.1). The pole was marked every 0.25 m starting from the ground to estimate30Table 3.2: Soyflux station equipment and the respective fields (Rainfed-1, Rainfed-2, Irrigated) that theymonitor (see Figure A.1, Appendix A).Parameter(Accuracy)Name/model Manufacturer Height/Depth(m)Field(s)CO2/H2O concentration(1%)LI-7500A LI-CORBiosciences,Lincoln, NB, USA3.70 Rainfed-1, IrrigatedWind speed (0.05 m s-1) 81000 R.M. Young,Traverse City, MI,USA3.70 Rainfed-1, IrrigatedAir temperature (0.3 °C),relative humidity (4%),precipitation (5%),atmospheric pressure(≤ 1 hPa), wind direction(0.3 °), wind speed(0.3 m s-1)WXT520 Vaisala Inc.,Helsinki, Finland3.40 AllNet radiation(5%, directional error)NR Lite-2 Kipp & Zonen,Delft, theNetherlands2.90 Rainfed-2NDVI (10%) SRS Decagon DevicesInc., Pullman, WA,USA2.90 Rainfed-2Incoming shortwaveradiation (5%)LI-200x-L LI-CORBiosciences,Lincoln, NB, USA2.20 AllPhotosynthetically activeradiation (photon flux)(5%)LI-190SAQuantumLI-CORBiosciences,Lincoln, NB, USA2.20 AllSoil heat flux (5-15%,self-calibrating)HFP03 Hukseflux, Delft,the Netherlands−0.08 Rainfed-2Soil heat flux TEM UBC Biomet Group −0.08, −0.08,−0.08Rainfed-2Water potential (25%),temperature (1 °C)MPS2 Decagon DevicesInc., Pullman, WA,USA−0.05, −0.10,−0.30, −0.60Rainfed-2Volumetric water content(0.03 m3 m-3),temperature, conductivity(1 °C)GS3 Decagon DevicesInc., Pullman, WA,USA−0.05, −0.10,−0.30, −0.60Rainfed-231crop height from the GoPro pictures. Daily captures of the crop height were then validated with a rulerduring field visits, with camera height estimates determined with a rough accuracy of ± 0.05 m. A generaldescription of the crop height measurements and models used to infer crop height are detailed in TableA.1 and A.2 and Figures A.2 and A.3 in Appendix A.The crop development cycle was separated into stages following phases proposed by Food and Agri-culture Organization (FAO) guidelines2: initial, development, mid-season and harvest. These phasesare characterized by both crop development stages and crop height: (1) the initial stage begins at plant-ing and extends to shoot emergence, (2) the development stage spans from shoot emergence to thecrop’s maximum height, prior to flowering, (3) the mid-season stage includes flowering, yield formationand ripening when the crop maintains maximum height, (4) the end period begins when the crop enterssenescence and ends on the day the crop is harvested. This information was complemented using Nor-malizedDifference Vegetation Index (NDVI). NDVImeasurements consisted of an upward and downwardpointing pair of spectral reflectance sensors (SRS-NDVI, Decagon Devices Inc., Pullman, Washington,USA) installed for the 2016 crop development cycles at a height of 2.90 m. The downward facing sensorwas pointed towards the soybean and maize canopies in the Rainfed-2 field. Additional leaf area index(LAI) measurements were made during field visits through differential photosynthetically active radiationmeasurements above and below the crop canopy using an AccurPAR ceptometer (Decagon DevicesInc., Pullman, Washington, USA). These data were measured exclusively on sunny field days and usedfor additional interpretation of the NDVI data.3.2.3 Eddy covariance data processingThemicrometeorological instrumentation for making eddy covariancemeasurements comprised aModel81000 ultrasonic anemometer (R.M. Young Company, Traverse City, Michigan, USA) installed on thetower alongside an open-path Model LI-7500A CO2/H2O infrared gas analyzer (LI-COR Biosciences,Lincoln, Nebraska, USA) connected to a Model LI-7550 analyzer interface unit (LI-COR Biosciences,Lincoln, Nebraska, USA) (Figure A.1, Appendix A). The ultrasonic anemometer and infrared gas ana-lyzer were installed 3.70 m above the ground and provided raw data measurements at half-hour intervalswith an acquisition frequency of 20 Hz (Table 3.2). The LI-7550 analyzer interface unit used embed-ded software (version 7.3.1, upgraded to 8.0 in 2016) to organize the raw data collection, prior to theflux calculations and initial data quality control steps in EddyPro® software (version 6.2.0) (LI-COR Bio-sciences, Lincoln, Nebraska, USA). Raw data processing in EddyPro® included flux calculations overhalf-hourly intervals with spike removals following Vickers and Mahrt271 before applying flux corrections32including coordinate rotation of the ultrasonic anemometer, correcting for the separation between theultrasonic anemometer and infrared gas analyzer measurements, and correcting for the density effectsdue to heat and water vapour transfer275. Given the rapid change in the height of the crop canopyduring a crop’s development cycle, a dynamic displacement height was input into EddyPro® followinglinear models based on field observations along with information on planting dates (Tables A.1 and A.2,Appendix A).The half-hourly calculated flux measurements were then selected based on quality control flags,precipitation events, and wind direction to provide a time series of ET for both Rainfed-1 and Irrigatedfields (no data was selected from the Rainfed-2 field). Quality control flags of 2 from Foken et al.78(derived in EddyPro®) were removed from the time series as well as flux measurements made duringprecipitation events detected by the meteorological station, and fluxes measured during low turbulence.The remaining measurements were then selected based on wind direction by assigning the 0-150°window to the Rainfed-1 field, and the 150-320° window to the Irrigated field. Low turbulence fluxeswere removed based on friction velocity (u* < 0.081 m s-1 in Rainfed-1 and u* < 0.072 m s-1 in theIrrigated field, and u* < 0.99728 m s-1for nighttime fluxes only); these thresholds were determined usingthe online tool for eddy covariance gap filling and flux partitioning available from the Max Planck Institute( Resulting data coveragefollowing exclusions related to quality control and turbulence below the u* threshold represented 36%of total measurements. Given the data selection based on wind direction, a flux measurement in theRainfed-1 field automatically generated a data gap in the Irrigated field, and vice versa, each of whichrequired filling (see Section 3.2.4). A total of 4103 (16.8%) and 4559 (18.8%) half-hourly measurementswere assigned to the Rainfed-1 and Irrigated fields, respectively.Data quality was assessed by determining the energy balance closure of the half-hourly measure-ments of sensible (H) and latent heat (LEmeas) fluxes obtained for both Rainfed-1 and Irrigated fieldsassuming similar net radiation (Rn) and ground heat flux (G) as measured in the Rainfed-2 field (seeFigures A.4 and A.5 in Appendix A). There, G obtained as the sum of flux measured by soil heat fluxplates at the 0.08-m depth and the rate of change in heat storage in the soil layer between the surfaceand the 0.08-m depth estimated using the soil temperature measured at the 0.05-m depth using theMPS2 (Decagon Devices Inc., Pullman, Washington, USA) (Table 3.2). Over the measurement period,energy balance closure (30-min data) was 61% in the Rainfed-1 field (LEmeas + H = 0.61(Rn − G) +32.14, R2 = 0.75) (Figure A.4) and 82% in the Irrigated field (LEmeas + H = 0.82(Rn − G) + 17.16, R2 =0.87) (Figure A.5). Final values of daily ET were calculated from gap-filled and energy balance closure33corrected latent heat flux (LE) (see Section 3.2.4).3.2.4 Eddy covariance latent heat flux gap-fillingMissing values of LEmeas were gap-filled with calibrated values of the Priestley-Taylor203 α followingVourlitis et al.273,274. Gap-filling was performed in three steps: (1) we derived a range of mean Priestley-Taylor α coefficients (αlow -αhigh) obtained by summing systematic and random errors, respectively fromthe variability in mean α and the lack of energy balance closure274, (2) we gap-filled αlow and αhighcoefficients according to periods of missing data or equipment failure, and (3) we used the range of αcoefficients to gap-fill LEmeas and ET by providing a range of values obtained from the range (αlow -αhigh)which we interpreted as a confidence interval. First, we derived a range of Priestley-Taylor α coefficientsas the sum of systematic and random errors from the eddy covariance measurements. The Priestley-Taylor equation relates measurements of LEmeas with those of Rn and G203LEmes = αmesΔΔ+ γ(Rn−G) (3.1)where LEmeas (W m-2) is the measured latent heat flux, Rn (W m-2) is the net radiation, G (W m-2) is theground heat flux, Δ (kPa °C-1) is the slope of the saturated vapour pressure versus temperature curve,γ (kPa °C-1) is the psychrometric constant, and αmeas (dimensionless) is the parameter obtained fromLEmeas (as either αlow or αhigh). Values of the α coefficient typically range from 0.40 to 1.2 based on wateravailability, and changes in vegetation surfaces (e.g., canopy conductance and surface roughness)273.We performed linear regressions of LEmeas against ΔΔ+γ (Rn−G) on the half-hourly data obtained fromthe Soyflux station (combining daytime and nighttime data) over daily (n = 48) and weekly (n = 336)periods to obtain mean values of αmeas and a corresponding 95% confidence interval (from regressionstatistics) for each period by forcing the regressions through the origin (see Figure A.6). Wind directionwas used to obtain separate α coefficients from LEmeas measured in Rainfed-1 and Irrigation fields. Val-ues of LEmeas obtained during and immediately following rainfall events (up to 1.5 hours) were removedto prevent bias in the values of αmeas during periods of combined surface saturation and high cloudcover. Values of αmeas derived from either too little data (n < 5 for daily data), too few daytime values (n< 3), or poor correlation (R2 < 0.50) were not retained as reliable coefficients. For both daily and weeklydata, single values of mean αmeas missing in the time series were interpolated with values before andafter the gap following Vourlitis et al.274. Greater gaps were filled using weekly data and the relationshipbetween α and soil volumetric water content (see below).34The low values of the daily α coefficients (αlow ) were obtained as αmeas minus half of the 95% confi-dence interval obtained from the linear regressions. The high values of daily α coefficients (αhigh) wereobtained from a sum of a correction to LEmeas for energy balance closure plus half of the 95% confidenceinterval. To correct LEmeas for lack of energy balance closure, we followed Barr et al.19LE∗ =LEmes(Rn−G)H+ LEmes(3.2)where LE* (Wm-2) is the latent heat flux after forcing energy balance closure. Coefficients obtained fromLE* (as α*) were obtained through linear regression following equation 3.1 and using LE* . The final rangeof α coefficients (αlow -αhigh) represents a confidence interval which was used to gap fill missing valuesof LEmeas.Values of daily mean α may be missing in the time series due to either missing values of H, or theselection process of a reliable daily α when applying equation 3.1. Missing values of daily α were gap-filled with weekly values of α. In cases where gaps were greater than one week (as in September andOctober 2015, and January 2017), we use a relationship of α as a function of daily mean soil volumetricwater content (θ) at 0.30 m to derive the (αlow -αhigh) range during those periods. Linear regressionswere αlow = 2.42θ (R2 = 0.49) and αhigh = 3.95θ + 0.02 (R2 = 0.46) (Table A.4). The remaining 19-pointgap in the Rainfed-1 field (September-October 2015) was filled using daily mean θ at 0.05 m to derive(αlow -αhigh) (as the only available sensor measurements in that time period) (Table A.4). The aboveprocedure allowed for a full time series of α in the Rainfed-1 field (Figure A.7). Data in the Irrigated fieldcould not be fully gap-filled in October 2015 due to differences in soil moisture between the Rainfed-2and the Irrigated field with irrigation schedules. Gap-filled values of LEmeas were in agreement with themeasurements in both fields (Figure A.8, Table A.5). Gap filled values of LEmeas using αlow and αhighwere then converted to ET and summed over the day (n = 48) to provide a confidence interval for dailyET (mm d-1).3.2.5 Soybean and maize crop coefficients and crop canopy conductanceDaily cropland ET measurements obtained from the Rainfed-1 and Irrigated fields were first used toobtain crop coefficients, prior to extracting crop canopy conductance in both fields. Crop coefficientshave been used extensively for crop water modeling, and specifically for the calculation of irrigationrequirements following FAO guidelines2 described in equation 3.3 under the conditions that crop water35requirements are fully met,KC, =ETC,ET0,(3.3)where KC,i (dimensionless) is the crop coefficient defined for day i in the crop development cycle, ETC,iis the crop ET (mm d-1) on day i, and ET0,i (mm d-1) is the reference ET on day i. Values of ET0 werecalculated based on the Penman-Monteith equation2 considering well-watered short grass of 0.12-mheight with an albedo of 0.23, in neutral stability conditions, and a surface conductance of 0.020 m s-1(or surface resistance of 50 s m-1) and 0.011 m s-1 at night (900 s m-1)3,191 as followsET0 =1λΔ(Rn−G)+ 18.60γDT+273.152Δ+ γ(1+B2)(3.4)where ET0 is expressed in mm per 30-min period, Rn− G are in MJ m-2 30 min-1, λ (MJ kg-1) is thelatent heat of vapourization, T (°C) is the air temperature, u2 (m s-1) is the wind speed measured at the2-m height, D (kPa) is the vapour pressure deficit, and B (dimensionless) is the ratio of aerodynamicto surface resistance (0.24 during the day; 0.96 at night)191. Values of ET0 were first calculated on ahalf-hourly basis using meteorological data from the Soyflux site, and then summed on a daily basis(daytime and nighttime). Given that ET0 is meant to represent the theoretical grass crop, the values ofRn and G cannot be the measured values above and below the crop canopy given the differences insurface albedo and crop cover throughout the year. We therefore model values of Rn and G followingrecommendations of Allen et al.2 through energy balance equations (see Section A.4).We also calculated canopy conductance from the Penman-Monteith equation as follows102,167gc = gΔ(Rn−G)+ ρcpDgγLE− Δγ− 1−1(3.5)where gc (m s-1) is the bulk observed canopy conductance, ρ (kg m-3) is the density of air, cp (J kg-1°C-1) is the specific heat of air at constant pressure, and ga (m s-1) is the aerodynamic conductance andis calculated using Malhi et al.151g =–2∗+1k∗lnzoMzoH+ΨM−ΨH™−1(3.6)where u (m s-1) is the wind speed measured by the sonic anemometer, u* (m s-1) is the friction velocity,k (0.4) is the von Karman constant, zoM and zoH (m) are the roughness lengths for momentum andheat, respectively, and are equal to 0.1h and 0.02h (with h, the crop height in m), and ΨM and ΨH36(dimensionless) are the integral diabatic correction factors for momentum and heat, respectively. Thesecorrection factors depend on the atmospheric stability ζ defined as follows35,81ς=−0.4g(z− d)Hρcp (T+ 273.15)3∗(3.7)forς < 1,ΨH = ΨM = 6ln(1+ ς) (3.8)forς > 1,ΨH = −2ln‚1+p1− 16ς2Œ,ΨM = 0.6ΨH (3.9)where g (9.81 m s-2) is the gravitational constant, z (m) is the measurement height, and d (m) is thezero plane displacement. In cases where H was not measured at the site due to equipment failure, wecalculated ga from the first term in equation 3.6 or€−2∗Š−1only. Mean values of gc were obtainedconsidering mean LE values obtained from gap-filling using the range (αlow -αhigh).All data filtering and calculation steps were performed using R Statistical Software207 (v.3.4.0) in RStudio (1.0.143) and packages openair 37, zoo284, Hmisc 101, and graphs generated with ggplot2 276,grid 207, gridExtra13, and scales277.3.2.6 Water productivity assessmentsSoybean and maize data obtained in the Rainfed-1 field (ET, yield), and the Rainfed-2 field (soil watercontent) were used to calibrate FAO’s AquaCrop model v.6.0 ( to separate crop evaporation and crop transpiration(measured together as ET in the field) and to explore changes in crop transpiration and water productivity(WP) with planting dates and irrigation schedules. Crop WP is defined as the ratio of harvested crop (kg)to the volume of ET (or grain yield (kg m-2) divided by ET (m)) over the course of the crop developmentcycle93, and was calculated for all crops from ET measurements and grain yield reported by the farmer.Crop modeling through AquaCrop is based on the relationship between water and harvested grain yieldto estimate ET through the crop development cycle using a soil water balance249. Input data required forAquaCrop include: meteorological data, crop parameters, soil information (e.g., permanent wilting point,field capacity), and field management practices (Table A.6). The AquaCrop calibration step requiresdefining the extent of the canopy cover, the crop development cycle from field observations for majorstages (shoot emergence, flowering, senescence), as well as soil characteristics. The model was thenvalidated using NDVI (as a proxy for canopy cover), soil water content, ET and yield results to match WPin Rainfed-1 for both soybean and maize (Table A.7). The validated models were then used to estimate37Table 3.3: Evapotranspiration (ET) and precipitation in both Rainfed-1 and Irrigated fields. ET resultsare provided with a confidence interval obtained using values of Priestley-Taylor α (αlow -αhigh).Period Rainfed-1 ET Irrigated ET Precipitationmm period-1 (mm d-1)18 Sep 2015 - 4 Feb 2017 (all data) 1265 ± 294(2.50 ± 0.58)1411 ± 181(2.79 ± 0.36)309918 Sep 2015 - 17 Sep 2016 (annual) 800 ± 187(2.19 ± 0.51)981 ± 119(2.69 ± 0.33)18391 Nov 2016 - 31 May 2016 (wet) 597 ± 138(3.88 ± 0.90)628 ± 83(4.08 ± 0.54)14631 Jun 2016 - 31 Aug 2016 (dry) 93 ± 23(1.02 ± 0.25)243 ± 22a(2.67 ± 0.24)1111 Oct 2015 - 31 Oct 2015 (dry month) 55 ± 15(1.83 ± 0.50)NAb 168aIrrigation estimated at 118 mm; bNot calculated due to unfilled gap in the datacrop transpiration and WP using different soybean planting dates and assuming maize was planted oneweek after the soybean harvest: 28 September 2015, 28 October 2015, 1 September 2016, 1 October2016 and 1 November 2016.3.3 Results3.3.1 Cropland evapotranspiration of rain-fed and irrigated fieldsTotal cropland ET was 1265 ± 294 mm and 1414 ± 181 mm for the Rainfed-1 and Irrigated fields, respec-tively, considering all crops and short periods between harvests and planting between 18 September2015 and 4 February 2017 (Table 3.3). During this period, average air temperature was 24.9 °C (sd =4.4), and total precipitation (P) was 3099 mm leading to values of ET/P equal to 0.41 in the Rainfed-1field. When considering one full year of cropland in both fields (18 September 2015 to 17 September2016), total ET was 800 ± 187 mm y-1 and 981 ± 119 mm y-1 in the Rainfed-1 and Irrigated fields, with1839 mm y-1 of precipitation (Table 3.3). The ET0 for the time period was 1830 ± 493 mm y-1 (Figure3.2), resulting in an annual ET/ET0 ratio of 0.44 and 0.54 , respectively in the Rainfed-1 and Irrigatedfields.38Figure 3.2: Evapotranspiration (ET) measurements at the Soyflux site shown with precipitation (P, mmd-1) (a), 24-hour mean shortwave irradiance (Rs, W m-2) (b), reference evapotranspiration (ET0, mmd-1) (c), and cropland ET (mm d-1) measured in the Rainfed-1 (d) and Irrigated (e) fields. Values of ET0and ET are represented by their confidence intervals based on systematic and standard errors.39We note that ET from Rainfed-1 and Irrigated fields were equal when considering confidence inter-vals for the entire time series (about 16.5 months), annual and wet season data (Table 3.3). However,differences in ET measurements for both fields were observed between 1 June 2016 and 31 August2016 when ET was 93 ± 23 mm (dry season)-1 in the Rainfed-1 field and 243 ± 22 mm (dry season)-1 inthe Irrigated field (Table 3.3, Figure 3.2d and 3.2e). This period corresponded to the dry season whenET0 almost doubled to 8 mm d-1 when compared to the wet season (November-May) (Figure 3.2c). Thisdifference between fields was also apparent for the partitioning of available energy in LE (Figure A.9),and the Priestley-Taylor α whose mean values were 0.36 ± 0.10 in the Rainfed-1 field compared to 0.91± 0.10 in the Irrigated field (Figure A.7) in the dry months. About 60% of available energy was usedfor LE in both fields in the wet season (Figure A.9a and A.9), a level which was maintained in the dryseason in the Irrigated field (68%) but not in the Rainfed-1 field (26%).The additional ET in the irrigated field coincided with planting of irrigated bean in the dry season(100-day crop cycle) while the Rainfed-1 field was planted to brachiara over the dry season. Irrigationwas applied to the bean crop at a rate of 6-7 mm every other night before the bean flowering stage(about 80 days), and 9 mm every other night after flowering (about 20 days) (pers. comm.). Fromthis irrigation schedule we estimate a total application of irrigation water of 118 mm between June andAugust, representing a volume of about 160,000m3 of irrigation over the course of the bean developmentcycle (the Irrigated field is 136 ha). A similar comparison could not bemade in October 2015 during whichthe farmer applied irrigation to soybean planted in the Irrigated field due to equipment malfunction. Inthat month, total precipitation was 168 mm mo-1 (Table 3.3) and contained a two-week period with noprecipitation (Figure 3.2a).Crop ET (ETC) showed differences due to crop type and development cycle (Table 3.4, Figure 3.3and 3.4) with soybean ETC being equal or greater than all other crops (Table 3.4). Based on crop yieldsfor each field, values of soybean WP in the Rainfed-1 field were 1.00-1.66 kg m-3 (2015-2016) and0.77-1.24 kg m-3 (2016-2017). In the Irrigated field, soybean WP was 0.80-1.08 kg m-3 (2016-2017)(see Table 3.4 for other crops). Mean daytime canopy conductance was obtained for each crop with thelowest value observed for maize with 5.08 mm s-1 (sd = 3.87) (Table 3.4).3.3.2 Water balance of rain-fed croplandOne dry season occurred over the measurement period between April and July 2016 during whichmonthly precipitation was < 100 mm mo-1. Given the differences in soil water conditions between wetand dry seasons and using daily average soil water matric potential (ψ) measurements between −3340Table 3.4: Crop grain yield, crop evapotranspiration (ETC), reference evapotranspiration (ET0), waterproductivity (WP) and mean canopy conductance (gc) in the Rainfed-1 and Irrigated fields. ETC valuesare provided with a confidence interval obtained using values of Priestley-Taylor α (αlow -αhigh), and ET0values are provided with a confidence interval obtained from the propagation of measurement errorsField Days Crop Grain yielda ETC ET0 WP gcton ha-1 mm mm kg m-3 mm s-1 (sd)(mm d-1) (mm d-1)Rainfed-1 106 Soybean 4.140 332 ± 82(3.1 ± 0.8)505 ± 151(4.8 ± 1.4)1.00-1.66 11.2 (22.7)151 Maize 7.620 313 ± 68(2.1 ± 0.5)700 ± 198(4.6 ± 1.3)2.00-3.11 5.08 (3.87)82 Brachiara NAb 141 ± 32(1.7 ± 0.4)471 ± 109(5.7 ± 1.3)NAb 6.01 (19.8)122 Soybean 4.020 423 ± 99(3.5 ± 0.8)547 ± 162(4.5 ± 1.3)0.77-1.24 31.7 (45.1)Irrigated 106 Soybean 3.714 271 ± 38c(2.6 ± 0.4)552 ± 156(5.2 ± 1.5)NAd 7.29 (6.01)89 Rice 3.300 277 ± 37(3.1 ± 0.4)416 ± 124(4.7 ± 1.4)1.05-1.38 10.9 (12.4)100 Bean 1.620 272 ± 25(2.7 ± 0.3)568 ± 131(5.7 ± 1.3)0.55-0.66 10.4 (11.6)127 Soybean 3.714 404 ± 59(3.2 ± 0.5)569 ± 167(4.5 ± 1.3)0.80-1.08 17.8 (22.7)aAt 14% moisture content; bNo yield data available; cIncludes a 13-day gap in October 2015; dNotcalculated due to data gap41Figure 3.3: Evapotranspiration-related variables over the study period at the Rainfed-1 field includingprecipitation (P, mm d-1) (a), 24-hour mean vapour pressure deficit (D, kPa) (b), evapotranspiration (ET,mm d-1) (c), daytime average canopy conductance (gc, m s-1) (d), and soil water potential (ψ) at the0.10-m, 0.30-m, and 0.60-m depths (e).42Figure 3.4: Evapotranspiration-related variables over the study period at the Irrigated field includingprecipitation (P, mm d-1) (a), 24-hour mean vapour pressure deficit (D, kPa) (b), evapotranspiration (ET,mm d-1) (c), and daytime average canopy conductance (gc, m s-1) (d).43kPa and −10 kPa, we determined field capacities of 0.305 m3 m-3 at the 0.05-m depth (sd = 0.056, n =292), 0.169 m3 m-3 at the 0.10-m depth (sd = 0.041, n = 326), 0.274 m3 m-3 at the 0.30-m depth (sd =0.043, n = 319), and 0.206 m3 m-3 at the 0.60-m depth (sd = 0.045, n = 349). Similarly, for ψ < −500kPa (sensor limit) we determined dry soil (as proxy for permanent wilting points) at 0.154 m3 m-3 forthe 0.05-m depth (sd = 0.040, n = 101), 0.097 m3 m-3 at the 0.10-m depth (sd = 0.003, n = 35), 0.128m3 m-3 at the 0.30-m (sd = 0.008, n = 92), and 0.125 m3 m-3 at the 0.60-m depth (sd = 0.004, n = 74)(Table A.8). Based on these values during the soybean and maize development cycles, the soil was ator above field capacity over 90% of the time for soybean in the 2015-2016 and 2016-2017 cycles, andover 66% of the time for maize in 2016 (about 60 days into the crop development cycle) (Figure 3.3).We used linear regression to compare changes in the daily total soil water storage at the 0.60-mdepth with P −ET, assuming no runoff and that capillary rise was negligible compared to P and ET (seeSection A.7, Appendix A). These assumptions were based on several factors: the deep groundwaterlevel at the site, observed puddling on the field in response to large rain events, and no surficial runoffobserved. Hence, we would expect water inputs to be captured either by the ET or the θ measurementfollowing infiltration. The relationship between the change in daily soil water storage at the 0.60-m depth(ΔSWS) and P −ET was expressed by ΔSWS = 0.47(P −ET)−1.75 (R2 = 0.33) (Figure A.10). Drainagebelow 0.60 m was derived using equation A.10 (see Appendix A). Average drainage at 0.60-m depthwas 4.85 mm d-1 (sd = 12.1) and 5.73 mm d-1 (sd = 11.9) for soybean in the 2015-2016 and 2016-2017 seasons respectively. Average drainage during the maize crop cycle was 2.76 mm d-1 (sd = 9.75)(Figure A.11).Beyond the April to July 2016 dry season where available water fraction approached 0 (Figure A.11),a short dry period was experienced at the beginning of the soybean development cycle in November2015 when the soil’s available water fraction at the 0.60-m depth dropped below 0.30 (Figure A.11) whenprecipitation was 127 mm mo-1 and ET was 74 mm mo-1. These values compare to November 2016when values of θ were close to or above field capacity at the 0.60-m depth when monthly precipitationand ET were 310 mm mo-1 and 116 mm mo-1, respectively. Unfortunately, without any soil sensors orinformation about additional water inputs in the Irrigated field, we were unable to perform a similar waterbalance analysis in that field.3.3.3 Soybean and maize development cyclesThe development of soybean and maize in the Rainfed-1 field carefully tracked crop development inthe Rainfed-2 field. This permitted us to utilize ET from the Rainfed-1 field to determine detailed crop44coefficient values throughout the respective crop cycles (see public videos of crop development in theRainfed-2 field for soybean (October 2016 to February 2017, 127 days),, and maize (February 2016 to July 2017, 151 days), Values of KC were separated into phases defined by theFAO as initial, development, mid-season, and end, based on observations in the crop development (e.g.,shoot emergence, flowering, etc.) and changes in crop height250 to which values ofKC were assigned forcrop modeling purposes2. Daily values of KC for soybean and maize crops were obtained consideringvalues of daily ET/ET0 (Figure A.12) when the soil was at or above field capacity at the 0.30-m depthfor the initial and development phases, and at or above field capacity at 0.60 m for the mid-season andend-phases (Table 3.5). In 2016, NDVI followed the evolving KC values of soybean starting at 0.19(initial phase), rising to 0.92 at the end of the development phase and 0.40 at harvest (Table 3.5, FigureA.12c). Values of NDVI for maize increased from 0.20 at the beginning of the development cycle to 0.87at the beginning of the mid-season phase before dropping to 0.27 at harvest (Table 3.5, Figure A.12c).LAI data was restricted to dates for which field visits were made; the highest values of LAI for maizewere 3.80 (16 April 2016) and 3.53 (13 May 2016) before dropping to 0.66 in the senescence phase (30June 2016).3.3.4 Changes in crop transpiration and water productivityAquaCrop simulations indicated that crop transpiration represented about 50% of soybean and maizeET, with marginal increases in soybean transpiration observed with the application of irrigation require-ments in the simulations (Table 3.6). The largest irrigation requirement was observed for soybeanplanted on 28 September 2015 (94.3 mm (crop cycle)-1), when irrigation increased the modeled yieldfrom 3.836 ton ha-1 to 4.007 ton ha-1 and transpiration by close to 10 mm (crop cycle)-1 for the season(Table 3.6). Similar increases in yield were not observed when soybean was planted on 1 September2016. Values of soybean WP based on transpiration (WPTr ) were lowest in September 2015 (1.50 kgm-3) and highest in October 2016 (1.80 kg m-3) (Table 3.6). Assuming maize planting in this double-cropped system occurs one week following soybean harvest, maize planted earlier (28 January 2016,following soybean planted on 28 September 2015) showed a slightly larger yield (7.620 ton ha-1) thanmaize planted one month later (7.587 ton ha-1) despite showing similar transpiration at 188.2 mm and189.1 mm for the crop cycle, respectively (49% of ET) (Table 3.6).45Table 3.5: Crop coefficients (KC) obtained for soybean planted in 2015 (October 2015 to January 2016),and 2016 (October 2016 to February 2017), and maize (February 2016 to July 2016) measured in theRainfed-1 field, with Normalized Difference Vegetation Index (NDVI) measured in the Rainfed-2 field.KC values are presented with a confidence interval obtained from the confidence intervals from bothevapotranspiration (ET) and reference ET (ET0).Crop Average development cycle meanNDVIKC at field capacityThis studyFAO KCvalues250phase days start-end start-end range (days)(mean)Soybean Initial (2015) 0-6 NAa 0.75 ± 0.06 - 0.66 ± 0.03(0.65 ± 0.04)0.3-0.4(20-25)Initial (2016) 0.19-0.17 0.59 ± 0.04 - 0.81 ± 0.02(0.43 ± 0.02)Development (2015) 7-60 NAa 0.66 ± 0.43 - 1.47 ± 0.10(0.60 ± 0.09)0.7-0.8(25-35)Development (2016) 0.20-0.92 0.65 ± 0.16 - 0.81 ± 0.22(0.79 ± 0.07)Mid-season (2015) 61-100 NAa 0.97 ± 0.16 - 1.30 ± 0.06(0.76 ± 0.08)1-1.15(45-65)Mid-season (2016) 0.92-0.47 0.98 ± 0.08 - 0.91 ± 0.11(0.8 ± 0.07)End (2015) 101-127 NAa 0.71 ± 0.06 - 0.66 ± 0.04(0.54 ± 0.04)0.7-0.8(20-30)End (2016) 0.45-0.40 0.95 ± 0.06 - 0.71 ± 0.01(0.77 ± 0.10)Harvest (2015) 127 NAa 0.66 ± 0.03 0.4-0.5Harvest (2016) 0.40 0.71 ± 0.01Maize Initial 0-20 NAa 0.30 ± 0.00 - 0.77 ± 0.01(0.52 ± 0.05)0.3-0.5(15-30)Development 21-56 0.20-0.86 0.67 ± 0.08 - NAb(0.71 ± 0.012)0.7-0.85(30-45)Mid-season 57-97 0.87-0.74 0.49 ± 0.21 - NAb(0.54 ± 0.08)1.05-1.2(30-45)End 98-151 0.70-0.27 NAb 0.8-0.9(10-30)Harvest 151 0.27 NAb 0.55-0.6aNo data available, sensor installation was on 8 March 2016; bNo data available due to ψ < −33 kPa46Table 3.6: AquaCrop simulations of crop transpiration (Tr, mm (crop cycle)-1), grain yield (ton ha-1) andwater productivity based on transpiration (WPTr, kg m-3) for soybean and maize planting dates. Foreach simulation, soybean developed over 127 days with maize planting occurring one week after thesoybean, with a development cycle of 151 days).Planting date Tr (%ET) Grain yield WPTr Irrigation(mm (cropcycle)-1)(ton ha-1) (kg m-3) (mm (cropcycle)-1)Soybean28 September 2015 256.1 (53) 3.836 1.50 028 September 2015 266.8 (55) 4.007 1.50 94.328 October 2015 232.8 (50) 4.008 1.72 028 October 2015 232.9 (51) 4.009 1.72 19.91 September 2016 248.3 (53) 4.054 1.63 01 September 2016 250.9 (54) 4.023 1.60 37.41 October 2016 223.3 (52) 4.026 1.80 01 October 2016 223.3 (52) 4.027 1.80 4.9Maize28 January 2016 188.2 (49) 7.620 4.05 027 February 2016 189.1 (49) 7.587 4.01 03.4 Discussion3.4.1 Water vapour supply of rain-fed and irrigated cropland to theatmosphereOur measurements confirmed a lower rain-fed cropland ET compared to natural vegetation in the region(Table 3.7). We conclude that even in the case of a farm practicing double cropping with an additionalcover crop in the dry season, a transition from a natural ecosystem to rain-fed cropland is expectedto generate a decrease in landscape ET in SAM. Annual cropland or pasture typically transpire lessthan forests due to shorter development cycles (< 150 days in this study), shorter vegetation height(0.40-2.40 m at the farm) and rooting depth, but also lower LAI and different photosynthetic pathwayswhich all affect how water is exchanged with the atmosphere during photosynthesis141. Regional stud-ies have quantified how land use transitions to rain-fed cropland have affected water vapour transfers tothe atmosphere138,238,246, with potential effects on regional precipitation recycling and surface temper-atures141,238.As a result of a lower rain-fed cropland ET compared to natural vegetation, we expect greater runoff,soil water content and/or drainage. Despite our estimate of drainage below the 0.60-m depth, we still47expect deeper soil water to be taken up by soybean or maize roots which can penetrate beyond onemeter in depth250 and contribute to ET. The values of P −ET accounted for half of the daily change insoil water storage down to the 0.60-m depth indicating additional losses through deep percolation. Soilresistivity measurements across a forest-soybean transect confirmed soil water content increasing upto 7-m depth in SAM173. Thus, soil water can percolate into the water table an increase river discharge,as observed in SAM where river discharge for soybean dominated catchments was four times greaterthan forest catchments104. Regionally, similar dynamics have shown greater discharge of major riverbasins in Amazonia as a result of land use change46,52.In contrast, mean annual irrigated cropland ET was similar to that of other natural ecosystems mea-sured by eddy covariance in Mato Grosso (Table 3.7). While we observed differences in ET between therain-fed and irrigated fields due to crop selection, the main difference in annual ET came from irrigationof bean between June and September 2016. The 243 ± 22 mm of ET from the irrigated field measuredin the dry season provided additional water vapour transfer to the atmosphere making it similar to othernatural ecosystems in Southern Amazonia that are typically maintained by deep roots that can accessadditional sources of water176. Priante-Filho et al.202 reported a June-August ET of 195 ± 8 mm forrain-fed pasture (derived from mean daily data ± 95% confidence interval) and 375 ± 8 mm for a tran-sition forest (i.e., a forest within the diffuse Cerrado-Amazon ecotone), while the December-Februaryperiod showed ET values of 290 ± 22 mm and 345 ± 33 mm, respectively. Our values were much lowerfor rain-fed brachiara, but greater for irrigated bean. Dry season irrigation also maintained the energypartitioning of cropland into LE observed in the wet season to levels greater than those observed byPriante-Filho et al.202 at 54% for rain-fed pasture in the dry season. Such differences confirm the impor-tance of the presence or absence of water in maintaining ET in the dry season, which, once adequatelysupplied with water, would mostly depend on radiation77.3.4.2 Crop evapotranspiration and water productivityOur results provide measured crop ET and KC values of soybean and maize as well as mean canopyconductance and NDVI in SAM for future use in regional modeling. Crop coefficients were within rangesproposed by the FAO250. The FAO KC values are typically used for determining crop water require-ments assuming perfect water management, a disease-free plant and ideal soil conditions2. Values ofKC reported here were lower than those previously used for calculating cropland ET in Mato Grosso138.Average Mato Grosso modeled ET for soybean and maize for the 2000-2009 period was 363-540 mmand 157-312 mm (crop cycle)-1, respectively, based on planting dates138. Dias et al.62 also reported48Table 3.7: Precipitation (P, mm y-1) and evapotranspiration (ET, mm y-1) measurements in Mato Grosso,Brazil.Site(period) Vegetation Location P ET References(mm y-1)Sinop(2000-2006) Transition forest11° 24.75’ S 2137 965 Vourlitis etal.27455° 19.50’ WCapuaba farm(2015-2016) Rain-fed cropland13° 17’ 15” S 1839 801 ± 187 This study56° 05’ 17” WCapuaba farm(2015-2016) Irrigated cropland13° 17’ 15” S 1839a 982 ± 119 This study56° 05’ 17” WFazenda Miranda(2011-2012)Cerradoforest-grassland mix15° 43’ 51” S 1030 927 Rodrigues etal.22656° 04’17” WaEstimated 118 mm of irrigation applied in the dry seasona simulated soybean ET in Southern Amazonia of 678.5 mm. These values were greater than whatwas measured in this study for soybean, but lower for maize likely due to assumptions in the devel-opment cycle of each crop (126 days for soybean; 100 days for maize), values of mid-season KC at1.50 and 1.40, respectively138, as well as precipitation. These differences along with the above aver-age grain yield reported by the farm (4.1 ton ha-1 for soybean, 7.6 ton ha-1 for maize) compared to theMato Grosso-wide average (3.10 ton ha-1 for soybean 5.98 ton ha-1 for maize according to IBGE121),provide measured values of WP and VWF that could be used as benchmarks for regional WP and VWFassessment purposes. Such benchmarks have been published for Mato Grosso for soybean (0.52 kgm-3, or 1923 m3 ton-1, respectively for WP and VWF), maize (0.74 kg m-3, or 1352 m3ton-1), rice (0.49kg m-3, or 2041 m3 ton-1) and bean (0.34 kg m-3, or 2941 m3 ton-1)156. All reported values of WP werelower (or VWF higher) than what was found in this study.The planting of rain-fed rice (Irrigated field) and maize (Rainfed-1 field) as a second crop followingthe soybean harvest showed that crops could still be harvested with high yields while taking advantageof rainfall and residual soil moisture at the end of the rainy season. The rice development cycle took fulladvantage of rainfall in 2016 due to earlier planting of soybean with irrigation, while the second half ofthe crop development cycle of maize coincided with the end of the rainy season. The use of irrigationfor early planted soybean was beneficial to yield without changing soybean WP, but also favoured maize49yield produced at a higher WP.Despite not being widely used in SAM, irrigation is one strategy that could be applied in Brazil toclose the yield gap estimated at as high as 1.6 ton ha-1 due to water deficit in SAM234. In 2006, irrigationwas used on 214,000 ha of cropland in Mato Grosso mostly for soybean (108,080 irrigated ha for 5.8Mha planted in Mato Grosso), maize (39,445 irrigated ha for 1.1 Mha planted), sugar cane (24,743irrigated ha for 0.2 Mha planted) and cotton (23,449 irrigated ha for 0.4 Mha planted)121. Irrigation forsoybean is one possible option for farmers to adapt to precipitation variability at the onset of the wetseason when soybean is typically planted as shown with the different strategies adopted at the farmbetween the El Niño (2015-2016) and La Niña (2016-2017) cycles. Similarly, irrigation may becomemore attractive in a warmer SAM climate whose dry season is expected to increase in duration withfurther deforestation51. Finally, irrigation allowed to increase crop frequency through a bean “triple-crop” that can bring additional income to farmers and represents an extension to the double-croppingsystem that is now widely practiced in the region245.Irrigation expansion, whether for soybean irrigation or for triple-cropping systems, would require ad-ditional investment in infrastructure. The state of Mato Grosso’s potential irrigation area is 10 Mha withexpansion depending on energy, water supply needs, as well social and environmental constraints76.Such an expansion would observe the constraints of the National Irrigation Plan which has the objectiveof “incentivizing the expansion of irrigated agricultural area and the increase of productivity based onenvironmental sustainability”33. For SAM, such an expansion would also require a region-wide assess-ment of the effects of land and water management for future agricultural production.3.4.3 Regional implications for land and water managementWater vapour transfers to the atmosphere from irrigated cropland, along with marginal improvements toyield with irrigation have regional implications for land and water management in SAM. Soybean produc-tivity could be increased by applying irrigation to soybean crops planted early in El Niño years, controllingfor a larger portion of ET being used productively through transpiration rather than lost through evapo-ration. While this so-called “vapour shift” is often promoted as a means to improve rain-fed agriculturethrough productive water use221, its application with additional irrigation has unknown region-wide ef-fects on the water cycle (see below). Additionally, the marginal improvement in yield through irrigation asshown in AquaCrop suggests that there is a limit to maximum yield that can be obtained solely throughirrigation in SAM, but which might be overcome through crop management or breeding to reach potentialyields closer to 4.6 ton ha-1 according to Sentelhas et al.234. However, these potential improvements50may be negatively affected by future climatic conditions. For instance, an increase of 1 °C of MatoGrosso temperatures impacts negatively not only maize yields (−2.6%) but also cropping frequency (assoybean/maize, −3.2%) and cropping area (−4.2%)50. Oliveira et al.184 reported a drop of up to 33%in future soybean yield in Amazonia based on changes in atmospheric CO2 concentrations, climatechange scenarios, and reduced rainfall related to deforestation. These declines were mostly attributedto carbon assimilation184, in which case irrigation would only serve to maintain a yield that is alreadypredicted to decline, and therefore could only provide additional agricultural resilience to precipitationvariability. This effect could potentially limit further intensification of agriculture in SAM, thereby favour-ing further extensification into Amazon and Cerrado vegetation, or existing pastureland141, as well asincreasing international pressures for additional agricultural land in Africa12.There are unknown environmental consequences to the widespread expansion of irrigated agricul-ture and its potential impacts on the water cycle. First, the additional supply of water vapour to theatmosphere through irrigation, particularly in the dry season, could shift the atmospheric water balanceby re-supplying water vapour previously lost through deforestation in the region141. As deforestationin Amazonia has been shown to affect regional precipitation recycling15,247, additional water vapoursupply could maintain this cycle and change the regional climate, but these effects have not yet beenaccounted for in the literature. Globally, land use change contributed to a 2.8% decrease in ET, while irri-gation contributed to a 1.9% increase in runoff227, but such effects can change based on regional scales.Widespread irrigation would likely affect surface and groundwater supplies as well as downstream wa-ter users. Mato Grosso is located upstream of main river systems (Amazon in the north and Paraguaiin the south) for which upstream (agriculture) and downstream uses (hydroelectric power) as well asecosystems (e.g., wetlands in the Amazon and the Pantanal region) should be considered38. Similarly,while double-cropping has increased in SAM245 with little impact to water quality and erosion174, thelong-term effects of additional fertilizer and pesticides on water quality are still unknown12 and are likelyto be more widespread with triple-cropping systems supported by irrigation expansion. Therefore, futureagricultural production planning, particularly as it relates to the intensification of production by a com-bined decrease in deforestation and potential use of irrigation, should consider the water quantity andquality trade-offs in these land and water management options141.513.5 ConclusionThis chapter provided insight into cropland ET in both rain-fed and irrigated systems in SAM. This studyalso provided context for future water resources trade-offs arising from different land and water manage-ment options for maintaining or increasing agricultural production. Direct ET measurements confirmedthat a rain-fed soybean-maize rotation as a double crop system in SAM had lower annual ET than thatreported for natural vegetation in the region, in contrast to a soybean-rice-bean rotation receiving irriga-tion. Irrigation can be used to maintain high WP for soybean planted at the end of the dry season, which,in turn, allows for an increased cropping frequency through a triple-crop. While the expansion of rain-fedagriculture in SAM is known to reduce water vapour supply to the atmosphere. This effect could slowdown or be reversed by an increase in water vapour supply to the atmosphere following widespreadirrigation, but not without consequences on surface or groundwater resources. Field measurementspresented in this study generated important parameters that can be used for modeling purposes. Fu-ture research should explore effects of the widespread use of irrigation and triple-cropping systems onthe water cycle, including quantifying land and water resources trade-offs in the context of agriculturalintensification in SAM.52Chapter 4Modeling the Volumetric WaterFootprint of Cattle4.1 IntroductionSouth America has been the largest agricultural frontier on the planet for almost two decades. Since2001, 96.9 Mha of pasture expansion took place on the continent, including large areas in Brazil andParaguay100. In 2012, Brazil was the largest producer of cattle globally with a population of 211 million,13% of which were raised in the central western state of Mato Grosso121. Recently, Mato Grosso’scattle production has been under scrutiny, particularly with respect to pasture expansion into both theCerrado (or savanna) and Amazon biomes (Figure 1.1). Between 2000 and 2009, total pasture areain Mato Grosso only varied marginally between 22 Mha and 24 Mha138, similar to the national trend63.This apparent stagnation masks indirect land use change dynamics that took place during the 2000s,particularly related to soybean expansion. Between 2000 and 2010, about half of the 5.5 Mha of forestclearings in Mato Grosso were utilized for pasture, while soybean expanded primarily into previouslyestablished pastures148. The wave of soybean expansion may have displaced pasturelands farther intothe Amazon agricultural frontier, indirectly causing more deforestation.The agricultural sector’s role in deforestation in both the Amazon and Cerrado biomes has beenthe subject of much research18,97,148,175,178, suggesting cattle intensification to avoid additional landuse change. For example, Brazilian pastureland was found to sustain 32-34% of the potential pasturecarrying capacity, which indicates opportunities to increase productivity nationwide252. Under a landsparing scenario, Strassburg et al.252 determined that the demand for cattle products in 2040 could bemet through a combined increase in pasture productivity (increased cattle density per unit area) and herdproductivity, achieved by an increase in carcass yield through breeding and feed improvements. Whilebreeding initiatives depend primarily on selection processes, feed quality, or research and development,further increases in cattle density on current pastureland would rely on a coordinated effort and a suite53of land use policies and economic factors. A pilot project has already established that investment inintensification could be financially sound with economic and environmental co-benefits48,253. Successfulintensification, however, is sensitive to price fluctuations in the beef market86,254.Livestock density in Brazil has been historically low, but increased steadily from 0.31 animal units(A.U., the equivalent of 450 kg of live weight of cattle) ha-1 in 1970 to 0.81 A.U. ha-1 in 1996209. Inthe Amazon and Cerrado, livestock densities at least doubled between 1990 and 201263. In MatoGrosso, livestock density increased from 0.74 A.U. ha-1 to 1.21 A.U. ha-1 between 1996 and 2006138,with beef production experiencing a drop post-2006 (along with the price of beef), before fully recoveringin 2012175. The increase in cattle density to 1.23 cattle ha-1 in the 2000s40 followed a drop in statewideAmazon deforestation. These changes were attributed, in part, to the introduction of economic restric-tions on farmers living in sensitive southern Amazonian municipalities; development of an initiative in2009 cattle herded on newly deforested land from the beef supply chain (the Cattle Agreement); andincreased law enforcement through monitoring and property registrations175.Research has also focused on greenhouse gas emissions of cattle production. One estimate from22 farms in Mato Grosso determined a range of 4.8-8.2 kg CO2-eq per kg live weight (LW) includingfarming operations, agricultural inputs for pasture management, and enteric fermentation from cattle40(Table B.1, Appendix B). However, these values increase considerably when including greenhouse gasemissions from land use change (Table B.2). Between 1990 and 2006, a total of 120 Tg CO2-eq accom-panied exports of beef produced in the Amazon region283. In the Amazon biome of Mato Grosso, totalgreenhouse gas emissions attributed to the beef production system from forest or crop conversions intopasture dropped by almost 50% decreasing from 201-209 Tg CO2-eq y-1 between 2001 and 200561,129to 115 Tg CO2-eq y-1 between 2006 and 2010129. Nationwide campaigns to reduce deforestation by80% in the Amazon and 40% in the Cerrado85, and encourage investments in pasture restoration orlivestock-forestry integration (e.g., Low Carbon Emission Agriculture Program known in Portuguese asplano Agricultura de Baixa Emissão de Carbono)153 are but a few of the carbon-focused initiatives tar-geting the cattle production system in recent years. However, while an increase in cattle density canhelp achieve deforestation targets, Mato Grosso would still need to answer questions regarding wateravailability and usage.Little information is available on water use for cattle in Mato Grosso and should be evaluated along-side land use and greenhouse gas emissions. Regular monitoring of the VWF, land footprint (LF), andcarbon footprint (CF) would provide amore complete picture of resource appropriation for cattle ranching.Research on the water consumption of production systems and supply chains has increased consider-54ably with the development of the VWF117. Many global and regional studies on animal products haveattempted to quantify resource appropriation88,157,267 or environmental impacts of water consumptionthrough LCAs218,219.In Mato Grosso, cattle typically drink from small farm impoundments that are either rain-fed or createdfrom damming (first or second-order) on-farm streams. This water source has to be carefully managed toguarantee year-long availability to cattle. Reduced precipitation and high evaporation rates can severelyreduce water availability in impoundments during the dry season (May-November), implying vulnerabilityof the production system to extended droughts. This study’s objective is to evaluate land and water re-source appropriation for cattle in Mato Grosso by quantifying the evolution of the LF and VWF from 2001to 2015, and to provide guidance on future strategies for land and water management. This informationis paramount to understand possible limits to future intensification of cattle in the region.4.2 Methodology4.2.1 Cattle production in Mato GrossoWe consider the cattle production system involving the Nelore breed from the species Bos taurus indicuswhich represents 90% of the cattle breeds in Mato Grosso70. We assume a state-wide average cradle-to-gate cattle production system following a 46-51-month cycle according to practices described for MatoGrosso36,40,70, during which females (cows, heifers) and males (steers, bulls) reach 430 kg and 520 kg,respectively. Calves are assumed to weigh 30 kg at birth and are weaned at 165 kg following Cardosoet al.36 (Table 4.1). In this system, cattle spend most of their lives on open pasturelands except in thefinishing stage when they can either remain in pasture or transfer to confinement (Figure 4.1). Pasturefinishing is a continuation of the adult development phase with pasture dry matter intakes continuing untilslaughter. Confinement may occur in the last 6-8 months of the animal’s life when cattle is generally feda diet of 70% feed and 30% silage.We analyzed land use and water consumption at the Mato Grosso municipality level for 2001, 2006,2011 and 2015. Given the change in size, shape, and number of municipalities in Mato Grosso duringthe study period, we combined the municipalities that had changed into greater municipal units (MUs)that remained constant over the study period, following methods outlined by Lathuillière et al.138. Thiscombination of political units resulted in 104 MUs for the entire state.55Table 4.1: Parameters used to model cattle growth in Mato Grosso, Brazil, from Cardoso et al.36.Stages Male FemaleWeight Weight gain Period Weight Weight gain Periodkg kg d-1 months kg kg d-1 monthsBirth 30 0 30 0Calf 30-165 0.30 0-15 30-148 0.24 0-19Adult 165-380 0.30 16-39 148-360 0.24 20-46Finishing 380-650 0.60 40-47 360-430 0.48 47-52Figure 4.1: Process flow for the cattle production system of Mato Grosso, Brazil.564.2.2 Volumetric water footprint of cattle production4.2.2.1 Animal water consumptive useWater for cattle is sourced from drinking water and moisture content in feed as well as additional watercreated by the animal’s digestion system (metabolic water)74. We apply the animal water balance ofRidoutt et al.218 to derive the amount of water consumed by the animal based on its development cycleand finishing stage (open pasture and feedlot). The total water consumed by the animal (W, L d-1mo-1)is expressed asWƒeed+Wdrnk+Wmet =Wep+WWG+Wƒeces+Wrne (4.1)On the water input side of equation 4.1, W feed is the water contained in the feed that depends on drymatter intake (DMI, kg d-1mo-1), the moisture content of the feed (MC, %), and the amount of milkconsumed by the calf until eight months of age (Wmilk , L d-1mo-1), and the amount of water needed tomix feed (Wmix), assumed to be 0.5 L (kg DMI)-1 according to Mekonnen and Hoekstra156 (equation4.2). Wdrink (L d-1mo-1) is the amount of water drunk by the animal, and Wmet is the water resultingfrom metabolic production as described in equation 4.3 following Ridoutt et al.218, and is a functionof feed digestibility (Dig, %). The sum W feed and Wdrink represents the total daily water intake of theanimal (W tot). On the water output side of equation 4.1, Wevap is the amount of water lost by theanimal to evaporation, WWG is the amount of water incorporated by the animal, W feces is the watercontent in animal feces, whileWurine is the water content in urine218. Following these definitions,W feed ,Wdrink ,Wmet ,W feces andWurine represent water flows into and out of the animal, whileWWG andWevaprepresent water consumptive uses.Wƒeed =DM×MC100−MC+Wmk+Wm (4.2)Wmet = 0.6DMDg100(4.3)As our study is exclusively focused on water quantity, we assume that water contained in urine andfeces are entirely evaporated; as such we consider the water uses of equation 4.1 as being exclusivelywater consumptive uses. This is a slight deviation from Ridoutt et al.218 who considered W feces lost toevaporation, but Wurine as a flow added to discharge with potential water quality impacts. We believeour assumptions to be reasonable for Mato Grosso considering the region’s high potential ET. On pas-tureland, moisture in cattle urine and feces can evaporate rapidly, as opposed to industrially confined57Table 4.2: Variables used to estimate cattle water consumptive use following the growth model describedin Table 4.1 and shown in Figure 4.1.Variable Symbol Value Unit ReferenceDry matter intake DMI 1.5% kgLWakg d-1 This studyWater drunk by animal Wdrink 11% kg LW L d-1 This studyFeed moisture content - pasture MCfeed 75 % This studyFeed moisture content - other MCfeed 12 % This studyFeed digestibility Dig 56.3 % Lima et al.144Milk consumptionb Wmilk 2% kg LW % This studyaLW = cattle live weight; bCalves are assumed to drink milk until 8 monthscattle whose waste is typically collected in pits for subsequent spraying into fields. While the loss ofmoisture through evaporation may occur at different stages in the production system, we still considertotal evaporation by accounting for water inputs on the left-hand side of equation 4.1. All parametersused to calculateW feed , Wdrink andWmet are listed in Table 4.2. We do not include water consumed fortransport between stages of development as the production system is generally confined to one singleproperty, nor do we include water consumed for the production of minerals, vitamins, and veterinarianservices administered to the herd. Feed water consumptive useWater consumption from pasture and crops used for feed were considered as additional consumptionindirectly attributed to cattle production. Diets depend on the animal, the production system considered,and whether confinement is involved at the finishing stage (Figure 4.1). The VWF of pasture was es-timated for each MU using spatial precipitation data from CHIRPS (v.2.0) from Funk et al.82 input intothe model from Zhang et al.287 (see equation B.1, Appendix B) (Table 4.3). All feed for feedlot finishing(61% maize, 10% sorghum, 8% soy meal, 8% cottonseed, 8% soybean grain165) was assumed to besourced in Mato Grosso, with VWF of crops obtained from previous research114,138 and derived usingcrop coefficients following FAO guidelines1 (Table 4.3). Crop modeling resulted in a total feed VWF of20.4 L (kg LW)-1 d-1, assuming the average feed composition. Given that there was minimal cropland orpastureland irrigation in Mato Grosso during the study period, all feed water use is exclusively sourcedfrom green water138.58Table 4.3: Volumetric water footprint (VWF) of cattle feed in Mato Grosso, Brazil.Feed VWF (L kg-1) ReferenceMaize and maize silage 590 Lathuillière et al.138Soybean grain 1640 Lathuillière et al.138Soybean meal 1924 Hoekstra and Mekonnen114Cottonseed 675 Lathuillière et al.138Sorghum 2089 Hoekstra and Mekonnen1144.2.2.3 Water consumption from evaporation of farm impoundmentsIn Mato Grosso, cattle mostly rely on open water from small reservoirs for drinking. Evaporation fromthese small reservoirs thus need to be accounted as a blue water (surface water) consumptive use218.Little information is available on small farm reservoirs in the region other than a recent estimate ofsize and depth by Rodrigues et al.225. We estimated reservoir area using a two-stage automated ma-chine learning classification procedure. In the first stage, Landsat composites were created using dryseason images from three-year intervals (1999-2001, 2004-2006, 2009-2011, and 2013-2015), cen-tered on our years of interest. We performed an initial image classification to separate water fromnon-water features, using a random forest classifier implemented in Google Earth Engine ( The resulting water/non-water predictions were then filtered usingimage morphological closing (dilation followed by erosion) to eliminate river fragments and reduce noisefrom misclassified pixels.In the second stage, we used an xgboost gradient boosted classifier to separate artificial reservoirsfrom natural water bodies (e.g., rivers and lakes) in the water/non-water predictions. We included shapeand spectral predictive features (calculated using Scikit Image264) for each water body which includedwater body perimeter, area, image moments, Hu Invariants, and the NDVI values within and aroundthe water object. The final artificial/natural water body dataset was then filtered to include only smallerreservoirs (< 1 km2) located within 200 m of pasture areas. We validated our results with reservoirdata from northern Mato Grosso obtained using a similar approach with a 2007 ASTER (15-m) imagemosaic149 (see Appendix B).We expect a small portion of total reservoir area to be allocated to aquaculture in Mato Grosso as99% of production occurs in either excavated tanks or impoundments122 (the remainder occurring innets within larger reservoirs). IMEA122 found that the largest tanks were 10.2 ha in size in south-central59Mato Grosso, and 2.6 ha for the rest of the state. Therefore, we expect that aquaculture should largely becaptured at the Landsat scale of 30 m. Data on aquaculture production for Mato Grosso is available from2008 to 2011 at the state level170,171 and at theMU level from 2013 to 2015 only121. We extrapolated fishproduction linearly and back in time for the 2000 to 2007 period assuming a constant annual decreasein 5000 tons y-1 until a minimum of 5000 tons y-1 was reached. This assumption led to an estimated5000 tons y-1 of fish in 2001, 15,505 tons y-1 in 2006, with reported values in 2011 of 48,748 tons y-1,and in 2015 of 47,438 tons y-1. To calculate the area of reservoirs for aquaculture, we assumed above(7 tons ha-1 water) and below (3.5 tons ha-1 water) average fish yields following values determined forthe state of Mato Grosso122. This production would represent a 6800-13,500 ha of reservoir area in2015 to be either removed from or shared with the cattle production system.To obtain the total area of small farm reservoirs within each MU, we calculated shape metrics foreach water body to be allocated to the live cattle population. The total impoundment area was multipliedby the average ET0 (mm y-1) obtained from Xavier et al.282 to be allocated to the total herd live weightin each MU. We then obtained an average Mato Grosso small reservoir evaporation (in L (kg LW)-1)from which we subtracted evaporation from fish tanks considering both the range of possible yieldsdescribed above, and average ET0 from all MUs in Mato Grosso. Since the Xavier et al.282 times seriesended in 2013, we assumed that ET0 in 2014/15 was equal to that in 2013. Total animal live weight wascalculated for each MU assuming a 50:50 ratio of males and females in each MU (see Section 4.2.3) andtheir average live weight in each of the calf (94.86 kg), mid-life (267.96 kg) and finishing stages (427.93kg). Lastly, we combined our estimates of impoundment area with data on the number of live animalsto calculate the reservoir cattle density (RCD, cattle per ha of water) per MU for each study year.4.2.3 Living herd population and annual water consumptive useTotal herd population was estimated in each MU for the 2000-2015 period using agricultural productiondata121 combined with information on relative population from Anualpec9 from 2006 to 2015. Accordingto Anualpec9, the relative cattle offtake rate in Mato Grosso was 17% (sd = 0.3), separated into 29%for females (sd = 1.6), and 71% for males (sd = 1.65). The calf and milking cow population represented27% (sd = 1.7%) and 2.8% (sd = 0.3 %) of the total population, respectively. Using this information,we deduced the living cattle population by applying the relative population information from Anualpec9to the agricultural production data from IBGE121 assuming constant proportions over the 2000-2015period. The total living herd Li,t (cattle) in each MU i and year t is the difference between the total herdH i,t reported by IBGE121 and slaughtered herd 0.17H i,t . Values of Li,t are then separated into each60animal development stage as described in equation 4.4L,t = 0.27H,t+ (L,t−0.27H,t−0.17H,t+1)+ 0.17H,t+1 (4.4)where each individual term represents the average cattle population respectively in the calf (0.27H i,t),adult and finishing stages (Figure 4.1) (0.17H i,t+1) as the number of animals slaughtered in year t+1.Given the development cycles for calf (15-19 months), mid-life (24-27 months) and finishing stages(6-8 months ), the total water consumption of the live herd in each municipality i and year t was thenderived based on the annualized water consumed at each development stage, averaged from cow andbull development cycles. Total animal live weight was calculated for each MU considering the averagevalues of Li,t , assuming a 50:50 ratio of cows and bulls in each MU and their average live weight in eachof the calf, mid-life and finishing stages.4.2.4 Land footprint of cattle productionWe combined remote sensing information with inverse yield information to combine direct and indirectland uses for cattle. First, total pasture area was obtained for each MU using Landsat imagery forthe 1.5-year composites of 2000/01 (January 2000 to August 2001), 2005/06, 2010/11 and 2014/15obtained fromGraesser and Ramankutty99. We then combined the total pasture area with the live animalpopulation in each MU to derive the pasture cattle density (PCD, cattle per ha of pasture) assuming thatthe pasture area was exclusively used by cattle rather than other animals (e.g., horses, donkeys, lambs,etc.). MU specific pasture areas obtained from remote sensing images were validated using pasturearea derived from animal population from IBGE121 following the method described in Lathuillière etal.138. Briefly, total animal livestock densities were derived for 1996 and 2006 from agricultural censusinformation which provided total animal population and pasture area. A regression of livestock densityversus pasture area was then used to derive pasture area in non-census years between 2000 and 2006.Validation of remote sensing data (Ap,RS) with the above estimate of pasture area in each of the 104 MUsof Mato Grosso (Ap,IBGE) gave Ap,RS = 1.41Ap,IBGE −3793 (R2 = 0.77) and Ap,RS = 1.40Ap,IBGE −2926(R2 = 0.74), respectively for 2000 and 2001, and Ap,RS = 1.36Ap,IBGE −5622 (R2 = 0.82) and Ap,RS =1.47Ap,IBGE −5934 (R2 = 0.82) respectively for 2005 and 2006 (see Table B.4, Appendix B). The directLF is then determined as the inverse of the PCD by allocating all pasture to living cattle. In addition, weestimated the amount of land required to grow inputs for the feed composition by considering the MatoGrosso average inverse yields in 2000, 2005 and 2010 for maize (3.08 × 10-4 ha kg-1), soybean (3.3661× 10-4 ha kg-1), cotton (2.78 × 10-4 ha kg-1), and sorghum (6.01 × 10-4 ha kg-1)121. When consideringthe average feed composition, we estimated the indirect LF of feed to be 8.02 × 10-6 ha d-1 (kg LW)- Sensitivity analysisWe performed a sensitivity analysis for the VWF calculation that took into account high and low pastureproductivity scenarios, as well as both male and female cattle development cycles (Table 4.1) in order toprovide a window of possible results considering average conditions in Mato Grosso. Furthermore, weprovide a VWF estimate that removes MUs overlapping the Pantanal wetland, as many impoundmentsin that biome may be natural or may not be used by cattle We also considered total reservoir evaporationwith and without the total Mato Grosso fish tank area obtained from fish production (see Section Results4.3.1 Cattle volumetric water, land and carbon footprintsMato Grosso’s average direct VWF for cattle at farm gate ranged from 236 L (kg LW)-1 to 348 L (kgLW)-1between 2001 and 2015, considering sex, finishing stage, meteorological conditions affectingreservoir evaporation (as W res), and the removal of evaporation allocated to fish tanks (Table 4.4). Forpasture finishing, the VWF was divided into 21-23% green water (from W feed ) and 76-80% blue water(from Wmilk , Wdrink , Wmet , and W res), whereas for feedlot finishing it was 18-20% green water and80-83% blue water (including Wmix).Reservoir evaporation was the largest contributor to the VWF representing 40-59% of total waterconsumed, followed by drinking (23-34% for Wdrink) and water contained in feed (14-23% for W feed )(Figure 4.2). When considering total small reservoir evaporation, values of W res varied with locationand meteorological conditions across the state, with values ranging from 6 L (kg LW)-1 to 1806 L (kgLW)-1 in 2001, and 16 L (kg LW)-1 to 1953 L (kg LW)-1 in 2015 (Figure B.2). The mean RCD (consideringtotal reservoir area in a MU) decreased from 872 cattle ha-1 (sd = 1352) in 2001 to 706 cattle ha-1 (sd= 653) in 2015 (Table B.6). Larger changes in RCD were observed between 2001 and 2015, with thehightest densities occurring in southeastern and northern Mato Grosso in 2015 (Figure B.3). Removal ofthe MUs located in the Pantanal wetland did not change the mean W res and RCD (Table B.5), howeverallocation of evaporation to fish tanks decreased the value of W res by as much as 46% in 2011 (TableB.6). In the same time period, the average LF rose steadily from 87 m2 (kg LW)-1 in 2001 to 305 m262Figure 4.2: Average animal volumetric water footprint (VWFanimal , L (kg LW)-1), and land footprint (LF,m2 (kg LW)-1) for the state of Mato Grosso between 2001 and 2015 considering pasture finishing. Valuesof VWFanimal are separated into water content of feed (W feed ), animal drinking water (Wdrink), metabolicwater (Wmet) and water evaporated by small farm reservoirs (W res). Values of W res represents totalsmall farm reservoir evaporationminus evaporation allocated to fish tanks, with fish tank area determinedby mean fish production (3.5-7 ton ha-1 of water).(kg LW)-1 in 2011 before reaching 299 m2 (kg LW)-1 in 2015 (Figure 4.2). These values were computedusing the statewide average PCD which increased from 0.72 cattle ha-1 to 0.91 cattle ha-1 between 2001and 2015 (Table 4.5).Confinement in the finishing stage only slightly decreased the cattle VWF from 132 L (kg LW)-1 to126 L (kg LW)-1 in males and 156 to 153 L (kg LW)-1 in females due to differences in moisture contentbetween pasture and feed, with only marginal increases in blue water frommixing water (3-7 L (kg LW)-1,Table 4.4). The VWF of feed ranged between 2.14 × 104 L (kg LW)-1 and 5.21 × 104 L (kg LW)-1, basedon the typical diet in the finishing stage and pasture productivity as identified within MUs (Table 4.4).The mean green VWF of pasture between 2001 and 2015 ranged from 1.48 m3 per kg dry matter (DM)-1to 2.62 m3 (kg DM)-1 for the high and low productivity scenarios, respectively. These values showedsmall variability across Mato Grosso between 2001 and 2015 with the lowest VWF estimated at 1.46 m3(kg DM)-1 (sd = 0.04) in 2015 and the highest at 2.66 m3 (kg DM)-1 (sd = 0.11) in 2006 (Table B.3). Theuse of feed in the finishing stage provided an additional 14.1 m2 (kg LW)-1 when considering agriculturalproducts going in to the feed (14.7 m2 (kg LW)-1 for males and 13.4 m2 (kg LW)-1 for females).By combining the above information with previously published results on the CF of cattle (Table63Table 4.4: Water consumption of Nelore (Bos taurus indicus) in Mato Grosso (L (kg LW)-1) representedby the volumetric water footprint of the animal (VWFanimal), feed (VWF feed ), and small farm reservoirevaporation (W res). The value of VWFanimal is the sum of the water ingested by the animal through thefeed (W feed ), milk (Wmilk), liquid water drunk (Wdrink), metabolic water (Wmet), and water mixed in thefeed (Wmix) when confined in feedlots in the finishing stage.Male FemaleL (kg LW)-1 Pasture Feedlot Pasture FeedlotW feed 51 43 61 52Wmilk 1 1 1 1Wdrink 74 74 88 88Wmet 6 6 7 7Wmix 0 3 0 7VWFanimal 132 126 156 153W res 2001All reservoirs (sd) 197 (267)without fish tanksa 158-161W res 2015All reservoirs (sd) 215 (316)without fish tanksa 192-217Pasture productivityb High Low High Low High Low High LowVWF feed (× 104) 2.49 4.41 2.14 3.45 2.94 5.21 2.66 4.45aFollowing removal of evaporation of small farm reservoirs allocated to fish tanks based on mean fishproduction (3.5-7 ton ha-1 of water); bHigh pasture productivity: 5.3 ton dry matter ha-1, low pastureproductivity: 3 ton dry matter ha-164Table 4.5: Pasture and reservoir cattle density, volumetric water, land and carbon footprints (VWF, LF,CF) for cattle production in Mato Grosso for the 2001-2015 period.Parameter/Footprint Component Ranges Total ReferenceCattle density (cattleha-1)Pasture 0.72 (0.54) - 0.91 (0.56) NA This studyReservoir(total area)1085 (1543) - 702 (653) NA This studyVWF (L (kg LW)-1)Animal 143 (15) 339-357 This studyReservoir 197 (267) - 215 (316) This studyFeed 2.14 × 104 - 5.21 × 104 NA This studyLF (m2 (kg LW)-1) Animal 87-299 100.4-313.7 This studyFeed 13.4-14.7 This studyCF(kg CO2-eq (kg LW)-1) Animal, inputs 4.8-8.2 1455-1458 Cerri et al.40Land usechange (2006)1450b Cederberget al.39aSee total emissions in the Amazon biome of Mato Grosso in Table B.2;bAssumed that carcass weight equivalent was 50% of live weight4.5) we obtained ranges of VWF, LF and CF for cattle in Mato Grosso for the time steps considered,including direct (animal) and indirect (feed) land and water appropriation. Reported values for the CFinclude emissions from the animal and inputs into the production system totaling 4.8-8.2 kg CO2-eq (kgLW)-1 according to Cerri et al.40. Land use change emissions attributed to cattle in 2006 was 1450 kgCO2-eq (kg LW)-1 according to Cederberg et al.39 (Table 4.5), while total emissions attributed to cattleproduction in 2001-2005 were estimated at 200.6-208.9 Tg CO2-eq y-1 following DeFries et al.61 andKarstensen et al.129, and 114.8 Tg CO2-eq y-1 for the 2006-2010 period129 (Table B.2).4.3.2 Evolution of land and water for cattle in Mato GrossoTotal pasture area in Mato Grosso increased from 31 Mha to 34 Mha between 2001 and 2011 beforedecreasing to 31 Mha in 2015. Total area for small farm impoundments was 47,515 ha (or 46,087-46,802 ha excluding fish tanks) in 2001 and 70,058 ha (or 56,624-57,401 ha) in 2015 (Table B.6). In2001 total reservoir evaporation was 6.70 × 1011 L, and increased to 9.15 × 1011 L in 2015 (Table B.5).When excluding the MUs within the Pantanal wetland (where natural water bodies may be confusedwith reservoirs), total evaporation was 5.03 × 1011 L y-1 in 2001 and 7.26 × 1011 L y-1 in 2015 (TableB.5). Pasture area increased linearly in 43 MUs over the same time period (mean R2 = 0.58, and65significantly (p ≤ 0.05) in 4 MUs), while reservoir area increased linearly in 96 MUs (mean R2 = 0.59,and significantly in 5 MUs) (Table 4.6). Dividing the results into two periods (2001-2011 and 2011-2015)revealed different rates of change over the study period. A total of 64 MUs showed linear increases inpasture area (mean R2 = 0.76, and significantly in 8 MUs) from 2001 to 2011, but only 10 MUs showedincreases from 2011 to 2015. A total of 72 MUs showed linear increases in total reservoir area (meanR2 = 0.47, and significantly in 1 MU) from 2001 to 2011, but 92 MUs showed increases from 2011 to2015 (table 3). In individual MUs, surface area of the reservoirs represented, at most, 1% of pasturearea between 2001 and 2015. We noted a strong correlation between impoundment area and pasturearea in the 104 MUs (R2 = 0.53) and a relationship expressed by logAres = 0.87logAp − 2.25 in 2001, and logAres 0.93AP − 2.37 in 2015 (R2 = 0.61) (Figure B.7). Changes in pasture and water areaswere accompanied by an increase in live cattle population from 16.5 million in 2001 to 24.3 million in2011, before stabilizing to 24.4 million in 2015. At the political boundary level, 95 MUs exhibited linearincreases in live cattle population (mean R2 = 0.70, significantly in 29 MUs) between 2001 and 2015,with about half of MUs showing increases in cattle population between 2011 and 2015 (Table 4.6).As a result of these changes, the trends for cattle production in Mato Grosso showed an increasein PCD and a decrease in RCD due to the increase in cattle population, slowdown in pasture areaexpansion, and slower increase in reservoir area. In 2015, greater values of PCD were observed innorthern and southwestern Mato Grosso and 6 MUs in the south east (Figure B.5), with values of RCDfollowing similar geographical trends as PCD (Figure B.3). We grouped individual MUs into 4 groupsaccording to their values of PCD and RCD between 2001 and 2015 compared to median 2001 valuesof 0.54 cattle ha-1 and 510 cattle ha-1, respectively: the “Pasture Dense” group contains MUs with PCD> 0.54 cattle ha-1 and RCD < 510 cattle ha-1, the “High Density” group has MUs with PCD > 0.54 cattleha-1 and RCD > 510 cattle ha-1, the “Low Density” group has MUs with PCD < 0.54 cattle ha-1 andRCD < 510 cattle ha-1, and the “Water Dense” group has MUs with PCD < 0.54 cattle ha-1 and RCD >510 cattle ha-1 (Figure 4.3). Between 2001 and 2015, the number of MUs with PCD > 0.54 cattle ha-1(Pasture Dense and High Density groups) increased from 53 to 70, while the number of MUs with RCD> 510 cattle ha-1 (Water Dense and High Density groups) increased from 69 to 74. Most changes inPCD and RCD occurred between 2001 and 2011, with only 2 MUs moving from Low Density to HighDensity groups between 2011 and 2015. Values of PCD and RCD that significantly changed between2001 and 2015 (p ≤ 0.05) are shown in Figures B.6 and B.4, respectively.66Table 4.6: Summary of changes in the 104 municipal units (MUs) of Mato Grosso related to changes inpasture area (Ap, ha) and reservoir area (Ares, ha) since 2000. The values of R2 are given to representthe correlation of Ap and Ares as a function of time (years)Year 2001-2015 2001-2011 2011-2015aTotal pasture area (Ap) (Mha) 31.3-31.3 31.3-34.2 34.2-31.3Total reservoir area (Ares) (ha)excluding fish tanksb46,444-54,013 46,444-39,287 39,287-54,013Total change in Ap (ha) −73,778 +2,910,495 –2,984,272Total change in Ares (ha) +16,663 +2217 +14,446Total change in Ares excluding fishtanks (ha)+4,538 to+10,600–10,282 to–4033+14,633 to+14,820Total live cattle population increase +7,837,714 +7,755,605 +81,609MU with increasing Ap (R2) 43 (0.58) 64 (0.76) 10MU with increasing Ap (R2), p ≤ 0.05 4 (0.92) 8 (1.0) 10MU with decreasing Ap (R2) 61 (0.57) 40 (0.70) 90MU with decreasing Ap (R2), p ≤ 0.05 9 (0.95) 3 (1.0) 90MU with increasing Ares (R2) 96 (0.59) 72 (0.47) 92MU with increasing Ares (R2), p ≤ 0.05 5 (0.97) 1 (1.0) 92MU with decreasing Ares (R2) 8 (0.21) 22 (0.27) 8MU with decreasing Ares (R2),p ≤ 0.050 0 8MU with increasing live cattlepopulation (R2)95 (0.70) 96 (0.81) 53MU with increasing live cattlepopulation (R2), p ≤ 0.0529 (0.95) 17 (1.0) 53MU with decreasing live cattlepopulation (R2)9 (0.40) 8 (0.54) 47MU with decreasing live cattlepopulation (R2), p ≤ 0.051 (0.93) 0 47aNo R2 values were reported because only 2 years were considered;bAres values are given as the mean obtained considering the range of mean fish production (3.5-7 tonha-1 of water)67Figure 4.3: Evolution of pasture (PCD) and reservoir (RCD) cattle densities (considering all small farmreservoirs) between 2001 and 2015 in Mato Grosso derived using 2001 median values of 0.54 cattleha-1 and 510 cattle ha-1, respectively. This evolution is separated into four groups: Pasture Dense (PCD> 0.54 cattle ha-1, RCD < 510 cattle ha-1), High Density (PCD > 0.54 cattle ha-1 and RCD > 510 cattleha-1), Low Density (PCD < 0.54 cattle ha-1 and RCD < 510 cattle ha-1), and Water Dense (PCD < 0.54cattle ha-1 and RCD > 510 cattle ha-1).4.4 Discussion4.4.1 On-farm land and water appropriation for cattleThe VWF, LF and CF are typically studied separately, but together, provide additional insight into on- andoff-farm resource management strategies. Remote sensing information reproduced the general trendof pasture expansion for Mato Grosso for the 2000-2010 period18,148 with a progression towards inten-sification of land use for cattle, which is expressed by the LF. The apparent plateau in cattle populationin 2014 together with a drop in pasture area (increase in PCD) suggests a transition into a new phaseof cattle intensification in Mato Grosso after 2011. This assumes that deforestation for pasture wasnot displaced into natural vegetation as seen in the 2000s10,18. Previous studies indicate that althoughthe pasture areas in Mato Grosso remained relatively constant between 2001 and 2015, the landscapechanged dramatically within this time period. Large expanses of forest were converted to pasture at arate of 400,000-600,000 ha y-1 between 2000 and 2005, but 800,000 ha of pastures were subsequentlyconverted into cropland between 2006 and 2010148. In the 2000s, the transition from pasture for crop-land increased opportunity costs of forested land farther north into the Amazon10,18, but also into the68Cerrado biome (affecting the price of land). The Cerrado is an area that has undergone significant landuse change resulting from agricultural expansion following a slowdown in deforestation in the Amazonbiome: between 2003 and 2013, cropland area doubled in the Cerrado region from 1.2 to 2.5 Mha with74% of land use change occurring in natural vegetation246, while in the same period conversions topasture affected 3.18 Mha of natural vegetation178.Land use change activities carry greenhouse gas emissions that are attributable to cattle throughthe CF. Emissions from land use change are typically the largest in the production process and caninclude legacy emissions (e.g., decomposition) with allocation schemes that extend decades followingdeforestation39,129,283. Intensification of cattle production on existing rangelands is a widely proposedstrategy for curbing deforestation and reducing the CF of cattle48,49,92,236. This approach has been metwith some skepticism given difficulties with enforcement and the fact that, according to the Brazilian For-est Code, large swaths of Cerrado can still be legally deforested for pasture or crop production14,196,241.Although further intensification on already cleared land (or through confinement) would reduce the CFattributed from land use change, the general increase in cattle herd would be followed by a rise in directcontributions from the cattle production system at values as high as 8.2 kg CO2-eq (kg LW)-1 accordingto Cerri et al.40, with some variability in estimates48.Identifying on-farm strategies and incentives to promote intensification has been challenging. Recentwork indicates that individual producers relying on production for income were less likely to implementnew practices to boost productivity, unless the property was part of the Rural Environment Registry(CAR, Portuguese acronym)137. Latawiec et al.137 report that, despite being considered a source ofwater on their property, 70% of farmers interviewed in northern Mato Grosso did not see any financialbenefit to forests, which, along with constraints from access to qualified labour and capacity building,constitute important barriers to intensifying cattle production. At the state and federal levels, enforcementof the Forest Code and the Cattle Agreement of 2009 have helped decrease deforestation for cattle92,175,whereas taxation or subsidies could help further increase cattle density and spare natural vegetation48.Unlike the LF, the VWF continued to increase past 2011 through the establishment of an additional20,000 ha of farm reservoirs, with different effects on RCD based on allocation of evaporation to fishtanks (see Section 4.4.3). Small farm impoundments are a key source of drinking water for cattle inMato Grosso and should be considered carefully in on-farm water balances. Reservoir evaporation isknown to be a major loss of water in Australia218, estimated nationally16 at 1.3-2.9 × 1012 L for anaverage volumetric density of 0.01 hm3 km-2 (maximum of 0.12 hm3 km-2)152. Small farm dams andreservoirs are also common in northeastern Brazil, where the area of small reservoirs has increased691.95% y-1 between 1970 and 2002152. In Central Brazil, Rodrigues et al.225 used remote sensing tomap 147 small reservoirs at a volumetric capacity of 1.8 × 10-3 hm3 km-2 in the Preto River basin. InMato Grosso, small dams are constructed in headwater streams, thereby creating up-stream and down-stream trade-offs for human and ecosystem water uses141. The increase in surface area created bythese impoundments can increase stream temperatures149 and evaporation141, while reducing runoffand stream connectivity34, as well as sediment loads168, all of which can alter stream habitats andnegatively impact stream biota. In this study, annual evaporative rates were fully allocated to cattle, butwere also expected to vary over the course of wet and dry seasons, thereby potentially decreasing thewater supply for cattle. Open water evaporation of a small reservoir in northeastern Brazil was 3593 mmy-1 with a maximum of 8.9 mm d-1 (December) and a minimum of 5.3 mm d-1 (July) with water levelsdropping by 1 m between September and December7. As such, we expect farmers to either supplyadditional water (surface or groundwater) to cattle or further concentrate animals near larger reservoirswhen necessary, particularly during drought periods or during the finishing stage when cattle is confined.From a water management perspective, a strategy focused on reducing reservoir evaporation wouldreduce water consumption of cattle from the impoundments. For instance, Baillie16 lists evaporationmitigation technologies that include floating covers or chemical barriers. Such on-farm water manage-ment would likely change once a given property no longer supports cattle, but maintains a network ofreservoirs on the property. For instance, the region that is now known as Tanguro ranch in easternMato Grosso104 was dedicated to cattle ranching for about 15 years prior to converting all of its pasturearea to cropland, yet many of its reservoirs remain intact. The fate of impoundments following the re-moval of cattle on farm properties is an important question for future water resources management anduse, given that these water bodies could be available for other agricultural purposes such as croplandirrigation or aquaculture production, which has increased dramatically in the state in recent years (seeSection 4.4.3).4.4.2 Land and water appropriation for feedSeveral studies have quantified the water used to produce feed as an indirect contributor to the VWFassociated with animal products88,157,187. Others have examined the competition between the use ofwater for crops or feed211, in addition to land resources and greenhouse gas emissions, as metrics ofenvironmental performance for livestock and animal products73. Since the production of animal feed isgenerally the largest contributor to the indirect VWF of cattle, reducing the water required to producefeed is often presented as a strategy for improving the efficiency of cattle production overall. Both the LF70and VWF can be reduced by increasing feed productivity while reducing water requirements, a strategyknown as increasing feed water productivity93. Since cropland and pasture in Mato Grosso are almostentirely rain-fed, reductions in the VWF of feed could be achieved through a water vapour shift favouringtranspiration over evaporation221 with potential savings of nearly 2 × 104 L (kg LW)-1 as shown from ourpasture productivity scenarios.Feed VWF can vary greatly based on the individual VWF of crops, pasture and roughage, and themix used in diets and production systems88,187. For instance, the VWF of Brazilian meat was foundto more than double when moving from a confined to a grazing system88. Similarly, Palhares et al.187show that the choice of feed composition can dramatically change the VWF based on the inputs used. Incontrast, the LF of cattle in the current system is much larger than the feed currently used in the finishingstage. Even in the case where cattle were confined their entire life cycle, the LF would reach 51-61 m2(kg LW)-1. However, with additional greenhouse gas emissions would also need to be considered if feedwere to be sourced from cropland that expanded into previously deforested areas148,246. For instance,the CF of soybean grown in Mato Grosso, which could be used as feed, was estimated at 12.2 tonCO2-eq ha-1 y-1 according to Novaes et al.179, equivalent to 4.0 ton CO2-eq ton-1 soybean.Feed type and efficiency can accelerate the cattle development cycle and reduce the time to slaugh-ter255. This strategy effectively aims to reduce inputs of the production system by reducing the amountof total feed VWF, LF and CF per kg LW. It should be noted that such a strategy also aims at increasingproduction output which favours an increase in animal population. Unlike direct LF, VWF and CF, whichare linked directly to land and water management practices in Mato Grosso, the use of feed affects in-dividual animals, which have increased throughout the study period, thereby necessarily increasing thefootprints associated with feed.4.4.3 Comparison to literature values and research limitationsOur values obtained for the LF and VWFwere comparable to previously published values in the literature.Our LF values were greater than those reported for cattle species in Australia (< 100 m2 (kg LW)-1),while our VWF results were comparable to the range of 24.7-234 L (kg LW)-1reported by Ridoutt etal.218. Moreover, reservoirs and cattle watering represented between 15% and 90% of total direct waterconsumption of cattle considering geographic areas218, confirming a strong link between water use andmeteorological conditions similar to this study. Our blue VWF of cattle in the finishing stage (1334-1587L (kg LW)-1) was within the 1407-1420 L (kg LW)-1 range reported by Palhares et al.187 in the stateof São Paulo. Moreover, our modeled estimates of pasture VWF was within the 20,000-50,000 L (kg71LW)-1 range reported by Gerbens-Leenes et al.88 for Brazilian grazing systems, while our feed VWF wasgreater than the 2000-16,300 L (kg LW)-1 range reported by Palhares et al.187 for the finishing stage.The largest uncertainty in the VWF value came from the Landsat estimate of reservoir area and theallocation of reservoirs to aquaculture. Our analysis excluded very large reservoirs (> 1 km2), as theyare unlikely to be used directly for cattle production. The detection of a small reservoir, on the other hand,was limited by the resolution of the remote sensing product (30-m Landsat; 0.09 ha). Comparison of theLandsat derived reservoir map with higher resolution data (15-m ASTER; 0.02 ha)149 (see Section B.3.2)suggests that the Landsat approach underestimates the total number of reservoirs by almost 200%,and is conservative in its estimate of reservoir area. Furthermore, our Landsat mosaics were biasedtowards cloud-free images obtained during the dry season, the time of year with the highest evaporativedemand and lowest reservoir volumes and areas. This effect may also explain the inter-annual changesin values of reservoir area detected using remote sensing, especially in 2006 and 2011, which followedtwo El Niño cycles known to have caused drought in SAM141. Given these factors, our results are likelyto be a conservative estimate of the total reservoir area in Mato Grosso, as well as the changes in theRCD over our study period. One additional source of uncertainty arises from the change from Landsat5 to Landsat 8 during the 2011-2015 period. This change in sensors adds to the uncertainty in the areachange estimates from 2011 and 2015, but we expect this effect to be small.Moreover, our allocation of reservoir evaporation relied on reference ET. Following one year of fieldmonitoring, Malveira et al.152 measured 3953 mm y-1 of evaporation compared to 2694 mm y-1 of refer-ence ET in northeastern Brazil suggesting that our allocation underrepresented the actual evaporationfrom reservoirs, especially during the driest times of the year. In cases of allocation of pasture VWF tocattle, studies have either allocated the entire pasture area ET to the cattle herd, or the amount eatenby each animal210. One could consider allocating reservoir evaporation proportionally to the amount ofwater drunk by cattle, which would further decrease the VWF of cattle that we have reported. In sum-mary, the combination of small reservoir areas detected in the dry season using remote sensing, alongwith an underestimation of evaporation using reference ET means that our VWF of cattle were overallconservative.4.5 ConclusionThis study combined new estimates of VWF and LF with published estimates of CF for cattle productionto show the evolution of land use, water appropriation, and greenhouse gas emissions in Mato Grosso.72The animal population grew by 47% between 2000 and 2015, requiring an increasing amount of waterfrom small farm impoundments and a total pasture area that stabilized at 31 Mha following 15 years ofpasture expansion, deforestation, and associated greenhouse gas emissions. The largest direct waterconsumption activity for cattle was from small farm reservoirs, whose evaporative losses should beconsidered in on-farm water management. The evolution of cattle production suggests that the demandfor water resources will increase as cattle ranching intensifies. This water will be sourced either throughfurther development of small farm reservoirs, or by pumping surface or groundwater. Moreover, the useof feed in future animal confinement will call for greater land and water appropriation with potential watersavings based on more efficient water use for cropland and pasture. Future research should explorefurther the details of the water balance of small farm reservoirs such as the inter-annual water availabilityand its relationship to cattle production, particularly considering a warmer and drier climate expected forthe region.73Chapter 5Volumetric Water FootprintSustainability Assessment in theXingu Basin5.1 IntroductionSAM has experienced significant development since the 1990s, with agricultural production expandingrapidly through land use change in both the Amazon and Cerrado (or savanna) biomes239. Naturalvegetation cover has been gradually replaced by pasture and soybean landscapes18, often through anatural vegetation -to pasture to- cropland transition148,246. This increase in agricultural production hashad important socio-economic and environmental implications. Socio-economic indicators suggest agrowth in tertiary sector up- and down-stream of soybean production, with evidence of local investmentof financial returns212. At the same time, deforestation has been shown to alter local climate and watercycles thereby pushing the Amazon towards a tipping point59 that could significantly alter the biome.Changes to above and belowground carbon stocks have implications for global climate change83, whileland use change can affect the water cycle by increasing river discharge45 and diminishing water vapoursupply to the atmosphere with implications for regional precipitation138,238,246. Changes to the watercycle, in particular, affect economic activity through hydropower generation and agriculture8,184,251, butcan also affect aquatic and terrestrial ecosystems38,141.Agricultural expansion in the region has been followed by infrastructure development such as roadnetworks59, population growth and land activities that trigger further deforestation. Between 1991 and2010, the population of Mato Grosso increased from 2 to 3 million, while the animal population increasedfrom 22 to 82 million, led mainly by cattle121. These increases put additional pressures on land useand local demand of natural resources, particularly water. Atmospheric feedbacks could negatively74affect agricultural production when considering changes to regional climate and precipitation regimes184,but could also trigger infrastructure investment in irrigation with additional effects on water withdrawalsand feedbacks on water resources141. Therefore, feedbacks between agricultural production, land usechange, and human and animal population growth need to be evaluated in order to estimate futuredevelopment scenarios in SAM.This study aims to quantify these changes by carrying out a VWFSA in the Xingu Basin of MatoGrosso (XBMT) located in SAM, an area that has experienced the land use change dynamics describedabove. When focusing on water quantity in a VWFSA, the blue and green VWF are compared to localwater availability to derive local water scarcity as a step towards formulating a policy recommendation117.Many studies have applied VWFSAs to derive blue water scarcity at a global scale114,159, but only onestudy to date has attempted to quantify green water scarcity163. More studies using the concept of greenwater scarcity are thus needed to verify the full extent of a VWFSA117,233.We build upon previous research results on the water cycle of XBMT to carry out a VWFSA followingHoekstra et al.117 for the period of 2000-2015. We also evaluate scenarios for 2030-2031 and 2050-2051with the objectives of formulating responses for water resources management based on past and futureland and water use decisions. The combination of land use change, climate change and agriculturalproduction scenarios within a blue and green VWFSA is informative to both water resources manage-ment, and the WF community seeking to apply this assessment regionally. The XBMT represents aunique basin for such a study, given its geographic location in the so-called “arc-of-deforestation” andthe importance in future land use change for agricultural production, but also because of agriculture’sreliance on precipitation in the region. The combination of land use and hydrologic data with informationon domestic and industrial water consumption remains mostly unexplored in SAM. Therefore, there isan opportunity to use such information in a VWFSA to provide a greater context for water resourcesmanagement and inform decision-making for regional production processes.5.2 Methodology5.2.1 The Xingu Basin of Mato GrossoThe XBMT (Figure 5.1) is a 170,000 km2 basin located in SAM, separated into the Xingu Headwaters(139,000 km2) that flow north into the Upper Xingu Basin (31,000 km2)4,269, through the state of Paráand into the Amazon River, to constitute the greater Xingu River Basin (510,000 km2)188. The XBMT is75located at the intersection of both Amazon (80% of the basin) and Cerrado (20%) biomes and had 50%(85,000 km2) of its forest cover in 2010, of which about 34,000 km2 was contained within conservationareas that include parts of the Xingu Indigenous Reserve238 (Figure 5.1). Between 2001 and 2010,the XBMT lost 18,838 km2 of forest to either cropland (3347 km2) or pasture (15,491 km2) with furtherevidence of conversion of 4962 km2 of pasture into cropland238. In 2015, agricultural production formunicipalities in the basin consisted of 1.3 Mha of soybean121, about 5.4 Mha of pasture138, and lessthan 12,000 ha of permanent crops (e.g., papayas, bananas, rubber trees, etc.)121. In addition, theXBMT contains close to 10,000 small farm reservoirs mainly used to supply drinking water for cattle149.In 2015, the cattle population reached about 3.5 million in the municipalities of the basin121.From a total of 199,015 people living in XBMT in 2007, 125,279 made up the urban population(63%), and 73,736 represented the rural population (37%), with the portion serviced by the generalwater network reaching 47% and 49%, respectively4. Most of the drinking water for communities inthe Xingu Headwaters is supplied by deep wells (60%), followed by surface water (20%), shallow wells(10%) and a mix of surface water and deep wells (10%), while 100% of the water in the Upper Xingu issupplied exclusively by deep wells4. Total domestic water demand was estimated at 0.0208 m3 s-1 inthe Xingu Headwaters and 0.1814 m3 s-1 in the Upper Xingu, while industrial demand (as transformationindustry) was 0.0023 m3 s-1 and 0.226 m3 s-1, respectively4. Given the importance of the agriculturalsector in the region, there is additional water demand for livestock, aquaculture with about 47.6 ha offish tanks, and a total irrigation demand of 1.447 m3 s-1in 20064.5.2.2 Integrated BIosphere Simulator (IBIS)Hydrology in the XBMT was modeled using the Integrated BIosphere Simulator (IBIS) (v.2.5) whichcombines ecological processes related to the water and carbon cycles with vegetation dynamics, cli-mate, canopy and vegetation physiology, and phenology on a monthly or annual basis79,80,136. IBISrepresents the soil-plant-atmosphere continuum to simulate soil moisture and evapotranspiration (ET)through six soil layers to 8 m depth (and soil temperatures), vegetation structure, stomatal conductanceand photosynthetic pathways, all forced with atmospheric conditions80,136. The model was previouslyvalidated by Panday et al.188 in a study of the water balance of the Xingu River Basin from 2001 to 2010using atmospheric forcing with data from the Climate Research Unit (CRU TS v.3.2.1). Surface runoffwas derived as the difference between ET and the balance of soil moisture, with the latter derived frominfiltration (from the Green-Ampt equation) and dynamics in the soil (from the Richards equation)188.Following Panday et al.188, we combine IBIS results with land use maps to derive the monthly water76Figure 5.1: The Xingu Basin of Mato Grosso (XBMT) and its sub-basins: the Upper Xingu Basin (yellow)and the Xingu Headwaters (green) with the main rivers and the location of the discharge measurementstation used for validation5. The inset shows the position of XBMT (black) in relation to the Xingu RiverBasin (black outline) and the state of Mato Grosso (gray).77balance of the XBMT for 2000-2001, 2014-2015, 2030-2031, and 2050-2051 (0.5° resolution, and hy-drologic years as September to August) following two simulations: (1) considering the basin’s potentialnatural vegetation (PNV) as defined by Ramankutty and Foley208, (2) considering the replacement ofall natural vegetation by C4 grass (G) as a representation of complete deforestation in the basin. Hy-drology for 2030-2031 and 2050-2051 was obtained from an average of 23 IPCC global climate modelsand considering two different Representative Concentration Pathways (RCP) of 4.5 and 8.5 W m-2.We derived total runoff in the basin by linearly associating runoff from PNV and G IBIS simulationsfor the basin in hydrologic year t following Panday et al.188R(t) = RPNV(t)Fƒ (t)+1− Fƒ (t)RG(t) (5.1)where R(t) (mm mo-1) is the monthly discharge in the basin, RPNV (t) (mm mo-1) is the total runoff inthe basin under a PNV simulation, F f (dimensionless) is the fraction of forest cover in the pixel of in-terest, and RG(t) (m3 mo-1) is the total runoff in the G simulation. The fraction F f was obtained fromland cover maps derived from Landsat imagery (30-m resolution)99, while future land use in 2030 and2050 was obtained by Soares-Filho et al.243 based on distinct deforestation scenarios: a business-as-usual scenario (BAU) in which 1997-2002 deforestation is maintained with planned transportation in-frastructure, and a governance scenario (GOV) which assumes similar deforestation rates as BAU, butin which a maximum deforested area representing 50% of each Amazonian sub-region is imposed243.When combining climate change with deforestation scenarios, we obtained four distinct scenarios for2030 and 2050 (BAURCP4.5, BAURCP8.5, GOVRCP4.5, GOVRCP8.5). Values of R(2000), R(2014), R(2030),R(2050)were obtained for the XBMT, andR(2000)was obtained for the Xingu Headwaters and validatedagainst monthly mean river discharge measured at Marcelândia, Mato Grosso (Passagem BR80, station18430000, 10° 46’ 38” S, 53° 5’ 44” W)5 with a Pearson correlation of 0.83 (Figure C.1, Appendix C).Values of R(t) were obtained annually and interannually with three-month averages for the years listedabove.Values of R(t) were then used to derive annual basin ET (ETT (t), mm y-1) using the water balanceequation shown in equation 5.2, and assuming a change in annual storage close to 0 following findingsfrom Panday et al.188,ETT(t) = P(t)−R(t) (5.2)where P(t) (mm y-1) is the precipitation input to the IBIS model. Similarly, we use equation 5.2 to deriveETPNV (t), or the annual ET of the basin under PNV, using RPNV (t) and the IBIS precipitation input. All78values of ET were obtained for 2000, 2014, 2030 and 2050 hydrologic years.5.2.3 Volumetric water footprint sustainability assessment5.2.3.1 Goal and scope definitionThe goal of this VWFSA is to determine changes in blue and green water scarcities from productionprocesses in the XBMT in recent history, and, considering deforestation and climate change scenariosfor 2030 and 2050, to: (1) provide a hotspot analysis of water use in the basin as guidance for futurewater allocation decisions, and (2) explore links between blue and green water scarcities in the basinconsidering land use change histories. This assessment focuses exclusively on water quantity andtherefore considers blue and green VWF separately, and does not address water quality as expressedby the gray VWF. Water footprint accountingThe accounting step includes the calculation of the blue and green VWFs of all processes occurring inthe basin for the 2000, 2014, 2030 and 2050 hydrologic years, representing production in recent years(2000, 2014) and defined following distinct scenarios for future conditions (2030, 2050, see Section5.2.3.4). The selection of the 2000 and 2014 hydrologic years was based on the intense land use changehistory in the basin within this time period as attested by land use maps99,188,238. Long-term runoffobservation in the Xingu River Basin at Marcelândia5 showed a change in runoff of −14% (February2001) and +23% (December 2000) compared to the mean 1975-2005 discharge. We focus exclusivelyon production processes, leaving out any local consumption of products that might be produced outsidethe basin. This assumption is reasonable given the regional focus on agricultural products for export139,with a majority of crops grown in the region supplied as input feed for livestock. Cropland and pasturein Mato Grosso have been nearly exclusively rain-fed138, and therefore only require green water whoseconsumption is estimated by ET.Green volumetric water footprint of agriculture in the context of the basin’s land use systemsWe obtain the green VWF of agriculture by combining top-down and bottom-up approaches to trackchanges in the green VWF from 2000 to 2050 hydrologic years (top-down approach) and 2000 to 2014hydrologic years (bottom-up approach). First, we propose that total annual ET of the XBMT is equal tothe sum of contributions from natural vegetation, agricultural land, and a residual term as described in79equation 5.3ETT(t) = ETNV(t)+ ETAG(t)+ ETR(t) (5.3)where ETT (t) (m3 y-1) in the annual ET in the basin in hydrologic year t obtained from equation 5.2,ETNV (t) (m3 y-1) is the annual ET from natural vegetation (as tropical humid or savanna forest, shrubland,etc.) in the basin, ETAG(t) (m3 y-1) is the annual ET from agricultural land (as cropland and pasturecombined), and ETR(t) (m3 y-1) is a residual ET term, which accounts for other land use systems (e.g.,forest clearance, urban areas) and water bodies (e.g., rivers, wetlands) that may or may not be includedin human consumption activity. In the top-down approach, we extract ETAG(t) + ETR(t) from a calculationof ETNV (t)ETNV(t) =∑jETPNV,j(t)ANV,j(t)FNV,j(t) (5.4)where ETNV (t) (m3 y-1) is the natural vegetation ET contribution in the basin, ETPNV,j (t)(m y-1) is the ETof the IBIS PNV simulation for each IBIS raster j of area ANV,j (t) (3080 × 106 m2) within the basin, andconsidering the fraction of forested land FNV,j (dimensionless). This approach allowed for the disaggre-gation of ETT into ETNV and (ETAG + ETR), which we use to analyze the hydrologic years between 2000and 2050.The bottom-up approach was applied for the 2000 and 2014 hydrologic years in which we used av-erage pasture and cropland ET estimates from Lathuillière et al.138,141 together with land use estimatesextracted from Landsat imagery99. We considered single- and double-cropped soybean (with rice ormaize) as the main crops in the region (Table C.1, Appendix C). This assumption is reasonable consid-ering that between 2000 and 2015, soybean represented 48-69% of total annual cropland in the basin,while maize and rice represented 12-23% and 33-3%, respectively121 with an ever increasing amount ofmaize double cropping in Mato Grosso245. During the same time period, perennial crops representedless than 1% of total agricultural land121 and were therefore not considered further in this green wa-ter accounting step. Residual ET (ETR) was then derived using equation 5.3 and, in this approach,may include ET that could be allocated to a production activity occurring in urban areas, or other landuse systems with no immediate productive activity (e.g., ET following forest clearance). Differences inETR between the top-down and bottom-up approaches may be interpreted as a systematic error in theallocation of ET to a particular landscape or human activity.Blue volumetric water footprint of agriculture The blue VWF of agriculture includes irrigation, butalso water consumption from livestock production systems. In 2006, about 3200 ha of perennial crops80were irrigated within XBMT and therefore we assume that the majority of irrigation in the XBMT wasnot applied to soybean or pasture between 2000 and 2014. Blue water use was estimated for animalproduction systems in pasture (ruminants), as well as confined facilities (chicken and swine), and in-cludes animal drinking water as well as water used for washing of animal housing. Feed for all animalproduction was assumed to be sourced from within the region, and is therefore already accounted for inthe agricultural green VWF (see Section Water consumption for cattle follows steps describedin Chapter 4 to allocate green and blue water per kg of LW based on sex, animal development stage anddiet. Here, we considered drinking water sourced from small farm reservoirs in the basin detected byremote sensing (see Section, Chapter 4). All other animals were assumed to have their drinkingwater sourced by the main water system. As described in Section 4.2.3 (Chapter 4), cattle population re-ported by agricultural production data121 is a total animal population which does not consider the annuallive population in their different stages of development. The annual live animal population for munici-pality i and calendar year t is the difference between the total herd population (Hi,t) and the number ofanimals slaughtered (0.17Hi,t). The live annual population Li,t can then be expressed by equation 4.4(Chapter 4). Sheep and goat annual offtake rates were assumed identical to that of cattle (17%), whilehorses, donkeys, and mules were not considered to be consumed and therefore their live populationwas equated to the total herd population reported by IBGE121.The swine and chicken development cycles were assumed to be 70 and 42 days respectively161,187,from which we derived average swine and chicken populations following125Pk,(t) = dysPk,(t)365(5.5)where Pk,(t) (animals) is the average population of animal k, in municipality i and calendar year t, daysis the total number of days of the animal’s development cycle, and Pk,i (t) is the population of animal kreported by national statistics121. To reflect animal population information available from IBGE121 forcalendar years into the hydrologic years used in this study, we take the average of the two consecutivecalendar years that overlap with each hydrologic year. Similar to crops, animal population for eachMU located inside the basin was scaled based on the percent area located inside XBMT (Table C.2,Appendix C).Animal water consumption was derived following ANA4 which provides water demand per animal,assuming an average adult consumption. For confined swine and chicken, production we assumed 3.4× 10-3 m3 animal-1 used to clean animal housing after slaughter following Palhares186 for swine, which81was also assumed for chicken housing. These volumes are assumed to be entirely consumed. Whileour drinking water consumption estimate is based on adult animal water demand and likely constitutesan overestimation of animal blue water consumption, our water consumption estimate for cleaning islikely an underestimate given the housing turnaround for both swine (70 days) and chicken (42 days)production (Table C.2). Large and small ruminants were not allocated any water for cleaning as theywere assumed to spend their lifetime in pasture.Domestic and industrial blue volumetric water footprints We estimated domestic water consump-tion based on urban and rural human populations within the basin and the total population receivingmunicipal services based on information from ANA4. We assumed a constant population growth in thebasin at a rate of 3.0% y-1 until 2014-2015 according to IBGE121. By assuming a total basin populationof 199,015 in 2010 (the same as 2007 information reported by ANA4), we derived total population in theremaining years, maintaining the same proportion of urban and rural population not serviced by the mu-nicipal system (47% and 49%, respectively) (Table C.3). Water consumption was calculated assuminga 50% return flow to surface water, and based on a rural water demand of 70 × 10-3 m3 d-1 cap-1 andan urban demand of 0.260 m3 d-1 cap-1. The 47% of the urban population that was not serviced by themunicipalities was assigned a water demand equal to rural demand4 (Table C.3).Industrial water consumption was based on the number of industrial workers in both extraction andtransformation industries assuming industrial demand of 3.5 m3 d-1 cap-1 according to ANA4. In 2010,the number of industrial workers was 4.1% of the total population within the basin121, which we as-sumed to be of constant proportion between 2000 and 2015. Similar to domestic water consumption,we assumed a 50% return flow of industrial water (Table C.3). Water scarcity calculationWe estimate water scarcity within the XBMT in hydrologic year t following equation 5.6117WS(t) =∑jVWFj(t)WA(t)(5.6)where WS(t) (dimensionless) is the water scarcity, VWF j (t) (m3 y-1) is the VWF of all activities j (deter-mined in Section, and WA(t) (m3 y-1) is the water available in the basin over time t. Values ofWS(t) are defined for both blue and green water and vary from 0 (no scarcity) to 1 (extreme scarcity) togauge how water use has evolved within the basin. For both blue and green water resources, WA(t) is82defined following equations 5.7 and 5.8117WAB(t) = R(t)−EFR (5.7)WAG(t) = ETT(t)−ETRNV(t)−ETUN(t) (5.8)where WAB(t) (m3 y-1) is the blue water availability, R(t) (m3 y-1) is the natural discharge (or dischargewithout human appropriation in the basin, defined in equation 5.1), and EFR (m3 y-1) is the environmentalflow requirement defined for the XBMT. When considering our top-down VWF accounting approach,the value of EFR was defined according to mean annual runoff following Smakhtin et al.240 with avalue of 45.9 km3 y-1 to keep natural ecosystems in a “fair” condition (see Section C.3, Appendix C).When considering the bottom-up VWF accounting approach, values of EFR were defined for each 3-month mean discharge between 2000 and 2014 hydrologic years as 0.20R(t) following Richter et al.213.Green water availability in hydrologic year t, WAG(t) (m3 y-1), was obtained by subtracting from theETT (t) (m3 y-1) the ET reserve to natural vegetation as ETRNV (t) (m3 y-1) and the ET of areas that areagriculturally unproductive, ETUN(t) (m3 y-1). We interpret ETUN(t) as the amount of small impoundmentevaporation for cattle production whose area we consider unavailable for agricultural expansion. Thevalue of ETRNV (t) is interpreted as a percentage of total basin ET (ETT (t)) as measured in the 2000hydrologic year and based on the Brazilian Federal Forest Codeminimum requirements for natural forestcover in both the Amazon (80%), Cerrado (35%), and transition (50%) zones201. As a result, WAG andWSG are calculated using these three minimum requirements expressed in ETRNV (t) of equation 5.8and equal to 0.80ETT (2000), 0.35ETT (2000) and 0.50ETT (2000). Interpretation and response formulation through scenariosBlue and green water scarcities were interpreted following previously defined benchmarks. Blue waterscarcity was “severe” when WSB > 2, “significant” when 1.5 < WSB < 2, “moderate” when 1 < WSB <1.5, and “low” when WSB < 1 following Hoekstra et al.118. Green water scarcity was “unsustainable”when WSG > 1, a “threat” when 0.5 < WSG <1, “within sustainable limits” when 0.25 < WSG < 0.5, andsustainable when WSG < 0.25 following Miguel Ayala et al.163. Results were then interpreted followingdeforestation (BAU, GOV) and climate change (RCP 4.5 and 8.5 W m-2) scenarios, and onto which weadded population growth and agricultural production scenarios (Table 5.1).First, we assumed that human population will continue to grow at current rates, or 3.0% y-1 until 2050,and assumed a similar breakdown in rural and urban population as in the 2000s, with industrial activity83assumed to be proportional to population growth. Primary sector growth was based on projections madeby the Brazilian Ministry of Agriculture for the 2025-2026 period focusing specifically on soybean, maize,cattle, swine and chicken production154, assuming continuous growth in the basin between 2030 and2050. We assumed a 35% increase in soybean production (or a 30% increase soybean area at currentyields) from 6.3 Mtons (or 2.1 Mha) in 2015 to 8.5 Mtons (or 2.8 Mha) in 2030 and an additional 35%increase (at current yields) to 11.5 Mtons (3.8 Mha) in 2050. When considered together, the total surfacearea for soybean and pasture were well within non-forested areas in the deforestation scenarios for 2030and 2050 of 13 Mha and 14 Mha (BAU), and 10 Mha and 11 Mha (GOV), respectively. Cattle, pig andchicken populations were assumed to increase respectively 3.0% y-1, 2.7% y-1 and 2.4% y-1 until 2050but with organizational differences in production systems based on two agricultural production options(Table 5.1).We considered two agricultural intensification options linked to increased production, but based onincreases in green water (the Green Option, Figure 1.2, panel A or B) and blue water (the Blue Option 2,Figure 1.2, panel E) resources appropriation. In the BAU scenario, average PCD for the XBMT in 2014(0.87 live cattle ha-1) was maintained to require a 0.4 Mha of additional pasture in 2030 (total of 4.4 Mha)and 3 Mha (total of 7.0 Mha) in 2050 (Green Option). Evaporation from small farm reservoirs in 2030 and2050 was scaled with cattle population on pasture based on 40 m3 cattle-1 y-1 of evaporation obtainedfor 2014-2015. In the GOV scenario, all additional cattle in 2030 were confined on 2014-2015 pasturearea to reach a livestock density of 1.3 cattle ha-1 (affecting 5.2 million animals). In 2050, additionalcattle were confined with a total population breakdown of 5.2 million cattle on pasture and 3.1 millionraised in confinement. We assumed that confined cattle did not use small farm reservoirs, but othersources that do not carry evaporation (e.g., groundwater). At the same time, we assumed that 90 mmof irrigation was applied in September-October to the entire soybean area (Blue Option).5.2.4 Data processing and sensitivity analysisData processing was carried out using statistical R Statistical Software207 (v.3.4.0) in RStudio (v.1.0.143)with packages: raster (v.2.5-8)106, sp26,190, rgdal (v.1.2-7)25,maptools (v.0.9-2)24, and ncdf4 (v.1.16)197.Our results are provided using a series of values to highlight the extent of water scarcity in the basinsuch as the use of both a bottom-up (2000, 2014) and top-down (2030, 2050) approach for allocatingET to vegetation. Our response formulation for green water resources was based on a suite of restric-tions following mandatory natural vegetation cover outlined in the Federal Forest Code (35%, 50%, and80%), which served as a sensitivity analysis for green water scarcity (WSG). Blue VWF values were84Table 5.1: Description of scenarios for 2030 and 2050 activities in the Xingu Basin of Mato Grosso(XBMT) following deforestation (business-as-usual (BAU) and governance (GOV)243), and climatechange scenarios (Representative Concentration Pathways (RCP) 4.5 and 8.5 W m-2). BAU and GOVscenarios also illustrate agricultural intensification options focused respectively on green water (BAU)and blue water (GOV) appropriation.Scenario Year Humanpopulation;industrialworkersLivestockpopulationDescriptionBAURCP4.5BAURCP8.52030 336,335;211,7225,233,040 cattle74,069 pigs792,674 chickenHuman population increases athistoric growth rate; industry growsproportionally to human settlement;soybean production requires 2.8 Mhaof land; cattle population requires 4.4Mha of pastureBAURCP4.5BAURCP8.52050 568,407;357,8098,372,864 cattle114,066 pigs1,173,157 chickenHuman population increases athistoric growth rate; industry growsproportionally to human settlement;soybean production requires 3.8 Mhaof land; cattle population requires 7.0Mha of pastureGOVRCP4.5GOVRCP8.52030 336,335;211.7225,233,040 cattle74,069 pigs792,674 chickenHuman population increases athistoric growth rate; industry growsproportionally to human settlement;soybean production requires 2.8 Mhaof land; cattle population is herded on4 Mha of pastureGOVRCP4.5GOVRCP8.52050 568,407;357,8098,372,864 cattle114,066 pigs1,173,157 chickenHuman population increases athistoric growth rate; industry growsproportionally to human settlement;soybean production requires 3.8 Mhaof land; cattle population is splitbetween pasture (5.2 million) andconfinement (3.1 million); soybean isirrigated 90 mm in September-October85considered to be conservative estimates, particularly for cattle production (see Chapter 4), as well asthe high return flows (50%) attributed to withdrawals.5.3 Results5.3.1 Past and future volumetric water footprintsBetween 2000 and 2014, the sum of cropland and pasture areas increased 31% from 4.7Mha to 6.2Mha.Changes in the consumption of blue water expressed by the total blue VWF increased from 0.153 km3 y-1in 2000 to 0.218 km3 y-1 in 2014 (Figure 5.2). The blue WF was dominated by agriculture representing97% of total water use, followed by domestic and industrial uses (Table C.5). Water evaporation fromsmall farm reservoirs represented 66% of total agricultural blue VWF in 2000, and 67% in 2014, followedby animal drinking (respectively 32% and 31%) and irrigation (2% of total consumption in both years)(Figure 5.2). Between 2000 and 2014, the total area of small farm reservoirs increased 37% from 6914ha to 9463 ha of water, leading to a total evaporation of 0.099 km3 y-1 and 0.141 km3 y-1, respectively.Domestic blue water consumption computed here was similar to values from ANA4, which reported3.47 × 10-3 km3 y-1 in 2007, while our industrial consumption estimates were three orders of magnitudesmaller than the 3.55 × 10-2 km3 y-1 reported for 20074. Differences in industrial uses are primarilyattributed to our separation of confined animals from industry, as well as our focus on extractive andtransformative industries. Combining animal and industrial water consumptive uses raised our computedvalues closer to those reported by ANA4. The total blue VWF increased with larger human and animalpopulations in 2030 and 2050. In 2030, agricultural water use nearly doubled to 0.258 km3 y-1, while thecombined industrial and domestic uses increased to 9.90 × 10-3 km3 y-1 (Table C.5). In 2050, agriculturalwater use increased to 0.517 km3 y-1 and 0.391 km3 y-1 in the BAU and GOV scenarios, respectively.In the case of cattle confinement and early season soybean irrigation (GOVRCP8.5), consumption roseto 3.81 km3 y-1.Agricultural expansion resulted in an increase in the total green VWF of agriculture (as ETAG) from40.6 km3 y-1 in 2000 to 49.9 km3 y-1 in 2014 (Table C.7, Appendix C). This change was led by croplandET which increased from 7% to 29% of ETAG, while pasture dropped from 93% to 71% of ETAG inthe same time period (Figure 5.3). The increase in green VWF occurred at the expense of the naturalvegetation whose contributions to ET dropped 11% between 2000 and 2014 due to a decrease in forestcover by roughly 1.4 Mha. Changes in ETAG and ETNV obtained through the bottom-up approach were86Figure 5.2: Total blue volumetric water footprint (VWF) of agriculture in the Xingu Basin of Mato Grossofor the 2000 and 2014 hydrologic years.similar to results from Silvério et al.238 (Table C.7). Further deforestation for agriculture in 2030 and2050 increased ET of non-forested areas to 188.6 km3 y-1 and 209.6 km3 y-1 for the BAU scenariosin 2030 and 2050 respectively, and 147.2 and 147.3 km3 y-1for the GOV scenarios (average climatechange scenarios) (Figure C.4, Table C.8).5.3.2 Blue and green water availability and scarcityAnnual runoff decreased from 74.9 km3 y-1 to 70.4 km3 y-1 between 2000 and 2014 (Table C.6), which,when considering environmental flow requirements, left 43.4 km3 y-1 (in 2000) and 40.8 km3 y-1 (in 2014)of blue water available in the basin. The decrease in annual runoff followed the decline in precipitationfrom 1999 mm y-1 in 2000 to 1934 mm y-1 in 2014 (Table C.6). When considering 3-month windows, thedecrease in runoff was more prominent in the December-February period where values decreased from20.7 km3 3-month-1 in 2000-2001 to 14.3 km3 3-month-1 in 2014-2015 (Table C.6), which we relate to areduction in September-November precipitation from 519 mm 3-months-1 in 2001 to 447 mm 3-months-187Figure 5.3: Changes in contributions to evapotranspiration (ET) for natural vegetation (ETNV ), pasture(ETP), cropland (ETC) and residual landscapes (ETR) in the Xingu Basin of Mato Grosso in the 2000and 2014 hydrologic years (September-August). Values obtained through the bottom-up approach asdescribed in the 2014.The combination of deforestation and climate change in the scenarios generally increased runoff by2% in 2030 when compared to 2000 (GOVRCP4.5), and by 8% in 2050 (BAURCP4.5) despite a reductionin precipitation (Table C.6). The GOVRCP8.5 scenario was the only exception with a decrease in runoff of1% in 2050 for a precipitation decline to 1952 mm y-1. Focusing on climate change effects alone, runoffwith potential natural vegetation cover in the basin decreased from 69.8 km3 y-1 in 2000 to 64.1 km3 y-1in 2014, 67.9-69.1 km3 y-1 in 2030 and 65.7-69.0 km3 y-1 in 2050 (Table C.6). Inter-annual changes inrunoff were apparent when considering 3-month windows: runoff generally increased at the beginningof the wet season (September-November, +13 to +20%), before decreasing at the end of the wet season(December-February, −62 to −71%). Dry season runoff increased between 22% and 52% in the Juneto August periods when compared to 2000 (Table C.6).Land contributions to ET in the basin were similar between 2000 (279.0 km3 y-1) and 2014 (272.0km3 y-1) (Table C.8). In 2000, forests represented 50-69% of contributions (bottom-up and top-down es-timates), while agriculture represented 15% (bottom-up estimates) (Table C.7 and C.8). In 2014, thesevalues changed to 46-63% and 18% for forests and agriculture, respectively. Total land contributions toET dropped by up to 4% in the GOVRCP4.5 scenario in 2030 and both BAURCP4.5 and GOVRCP4.5 sce-88narios in 2050 (Table C.8, Figure C.4) with differences in contributions based on forest cover. Forestsin both GOVRCP4.5 scenario provided 122.9 km3 y-1 and 122.5 km3 y-1 of ET in 2030 and 2050, respec-tively. In contrast, BAURCP4.5 showed a reduction of natural vegetation ET from 82.1 km3 y-1 in 2030 to60.0 km3 y-1 in 2050 as a result of reduced forest cover (Table C.8, Figure C.4).Annual blue water scarcity values were less than 0.10 (Figure 5.4) with the largest value recordedin the GOVRCP8.5 scenario in 2050 (0.09). Inter-annual values increased to 0.65 for the GOVRCP8.5scenario between September and November 2050 due to early soybean planting and irrigation (withinter-annual values ≤ 0.03 the rest of the year). Annual green water scarcity values changed accordingto deforestation scenarios, but also due to restrictions placed on the allocation of natural vegetation.Between 2000 and 2014, green water scarcity was at least “within sustainable limits” (WSG < 0.50)when considering the bottom-up approach, moving closer to “threat” conditions (0.5 < WSG < 1) in thetop-down approach (Figure 5.4). In 2030 and 2050, green water scarcity values increased closer to 1.1in the BAU-2050 scenarios considering 35% of natural vegetation allocated to the basin, and beyond 1.2when allocation increased to 50% and 80%. In the same time period, the GOV scenarios maintainedWSG < 1 with a 50% allocation to natural vegetation, but moved to “unsustainable conditions” in both2030 and 2050 when allocating 80% of the basin to natural vegetation (Figure 5.4).5.4 Discussion5.4.1 Agricultural development and water resourcesAgriculture was found to be the largest contributor to the total blue VWF in the basin, with animal waterconsumption for drinking and from reservoir evaporation representing the largest component. Waterallocated to animal production systems in 2014 was equivalent to the consumption of 2.3 million peopleconnected to the municipal system. Animal population in the basin was historically led by cattle, butpig and chicken production have increased in recent years121, effectively increasing water consumptionand the water supply needed for production. Chickens and pigs are typically raised in confined facilitiesin Mato Grosso and, therefore, rely on surface or groundwater pumped for drinking water. In contrast,cattle in Mato Grosso rely on small reservoirs whose evaporation constitutes more than half of agricul-tural blue water consumption. Some of these reservoirs are constructed from impoundments of smallstreams, which contribute to stream warming with potential effects on stream chemistry149 and hydro-logic connectivity34. Regional effects of these reservoirs on hydrology remain relatively unexplored in89Figure 5.4: Annual blue (WSB) and green (WSG) water scarcities for the Xingu Basin of Mato Grossoin 2000 and 2014 hydrologic years, business-as-usual (BAU) and governance (GOV) deforestation sce-narios in 2030 and 2050 considering Representative Concentration Pathways (RCP 4.5 and 8.5 W m-2).Values of WSG were obtained assuming that 35%, 50% and 80% natural vegetation cover in the basinwas maintained (Table 5.1).90SAM.The replacement of natural vegetation by cropland and pasture was illustrated by an increase in greenwater appropriation in the basin. We report a decline in pasture area in 2014 compared to 2000, which,when combined with increasing cattle population, led to an increase in cattle density (0.57 cattle ha-1in 2001 to 0.97 cattle ha-1 in 2015), following general trends in the state of Mato Grosso (see Chapter4). The replacement of deep rooted natural vegetation with shallow-rooted crops and pasture affectsradiation partitioning by decreasing latent heat and increasing sensible heat fluxes141. These changes inradiation partitioning have important consequences on surface temperatures. Silvério et al.238 showedthat cropland and pasture surface temperatures in the XBMTwere 6.4 °C and 4.3 °C greater than forests.As a result, deforestation between 2000 and 2010 led to an average basin temperature increase of 0.3°C on top of the 1.7 °C increase that had occurred because of deforestation prior to 2000238. The XinguIndigenous Park located in the heart of the basin (Figure 5.1) showed surface temperatures 3 °C lowerinside the protected area compared to the rest of the basin47. Such effects illustrate the importance ofmaintaining natural forest cover.Water consumption in future agricultural production varied substantially based on the conditions ofproduction, which include land expansion and intensification. Our evaluation of two agricultural expan-sion scenarios highlights the extent of future green water appropriation from rain-fed agriculture whichcarries consequences on carbon and water cycles139. Agricultural intensification for both crops andlivestock requires either more efficient use of green water on current land, a reallocation of green waterresources from crops to livestock, additional blue water consumption from irrigation, or a combinationof the above141. Under current production practices, the onset of the wet season dictates when (or if) asecond crop (typically maize) could be planted11,245. Farmers may plant soybean earlier in the season(e.g., in September) and irrigate fields until the onset of the wet season (e.g., approximately 16 October2007 in the basin11) to allow for earlier planting and harvesting of maize, and the potential success oftwo crops. Under this strategy, farmers could also add a third irrigated dry season crop (e.g., bean)leading to additional blue water consumption (see Chapter 3).Similarly, future cattle production may include additional confinement as a strategy to free pasturefor cropland expansion. A larger cattle population means greater appropriation of both green water(through feed) and blue water (through drinking, small farm reservoirs, cleaning of pens, etc.) (seeChapter 4). Confinement could also move towards the use of blue water sources other than thosestored in small farm reservoirs (e.g., groundwater), in which case the total blue VWF of cattle coulddrop. However, this apparent efficiency has to be assessed considering the use of reservoirs in the91long term, or their possible decommissioning or alternative use in other production systems (e.g., asirrigation). Potential water savings through efficiencies in the cattle production system (e.g., reservoirevaporation management) could also reduce the blue WF of cattle to allow greater water availabilitydownstream (Chapter 4).Since 2000, the state of Mato Grosso increased meat production for both domestic consumptionand international exports. The amount of water used for production is therefore virtually transferred toconsumers within and outside of Brazil55 (80% of Brazilian production is consumed within the countryaccording to FABOV70). Between 2000 and 2014, Mato Grosso meat exports rose from 27,000 tons to387,000 tons155, thereby increasing the amount of water consumed regionally for foreign export, alongwith soybean commodities139. For instance, 27% of Europe’s virtual water imports between 2006 and2015 came from soybean trade69.The selection of future production systems proposed through our scenarios can therefore changethe resource appropriation for regional production, which already carries nutrient and carbon footprintsthat can be allocated to consumers139. This connection between consumption and production centershas inspired demand-side management of water use through the supply chain. For instance, Van-ham et al.267 estimated the VWF of different European diets and their implications for water resources.Supply chain interventions in the region have been motivated by deforestation and climate change impli-cations though both the Soybean Moratorium or the Cattle Agreement175, but could also include waterresources given the close link between land and water resources management in agricultural productionsystems141.5.4.2 Changes in water scarcity with land and water managementActivities in the basin through present day found to be within blue water sustainable limits. Green waterresources, however, were within sustainable limits under specific conditions only. Inter-annual bluewater scarcity moved closer to “moderate” under irrigation expansion and cattle confinement, reflectingthe potential vulnerability of the basin to dry season agricultural water use. A total of 234 irrigation pivotscovering almost 28,000 ha were identified in the municipalities overlapping XBMT6 and expansion couldincrease given the 10 Mha irrigation capacity estimated for Mato Grosso76. Similarly, the developedreservoir capacity for cattle is a measure to ensure continued drinking water in the dry season whenanimals may need more water due to meteorological conditions187. Water consumed for agriculturalproduction is then unavailable for other human and ecosystem uses in the greater Xingu River Basin,and may affect wetlands or hydroelectric power production38,188. Water rationing has already taken92place as a result of drought (e.g., 2005) and the lack of infrastructure to cope with low water levels,particularly in the Xingu Headwaters4. We therefore expect future water use for irrigation and cattleto also come from additional sources (e.g., surface and groundwater sources) should water becomescarce in the dry season. Consequently, both intensification of soybean and cattle production shouldcarefully observe the effects on future water scarcity in the basin in agricultural management plans.While policies have mostly focused on maintaining forest cover to protect biodiversity and reducegreenhouse gas emissions, these policies can also play a role in maintaining sustainable water resourceuse. The sustainable limits that we calculated relied on our estimate of water availability (WAB, WAG)which depended in turn on the interaction of land and water management initiatives. We found thatnatural runoff (i.e., runoff without any consumption activity, affecting WAB) would change in 2030 and2050 as a result of deforestation and climate change, while total land ET (WAG) responded directly to theallocation of land to natural vegetation cover, with a feedback on natural runoff (see below). Changesin natural runoff resulting from deforestation and climate change have already been measured in theregion. For example, the 15% forest cover loss between 1971 and 2010 in the Xingu River Basin led toa 6% increase in runoff, while climate variability led to a 2% decrease in precipitation and 14% decreasein runoff188. Groundwater is known to act as a buffer in the basin, particularly in the dry season whenrunoff could diminish due to an extending dry season ET199. These changes can be exacerbated bythe amount of deforestation in the basin represented by the BAU and GOV scenarios also guided byBrazilian Federal Law.The determination of green water scarcity assumed an increasing amount of land allocated to naturalvegetation in the basin based on natural vegetation cover mandated by the Federal Forest Code201. Assuch, our interpretation of green water scarcity was based on the amount of vegetation cover lost inthe basin in relation to Federal thresholds, which vary by biome from 35% (Cerrado savanna) to 80%(Amazon forest). For instance, in 2014, green water scarcity was within “threat” conditions when allo-cating 80% of the basin to natural vegetation (based on ET in the 2000 hydrologic year as describedin Section These “threat” conditions mean that from the total amount of green water avail-able in the XBMT (represented by total ET, ETT ), the amount that could be put to use for agriculturalproduction approached the limits mandated by the retention of natural vegetation cover (80%). Evenin a restrictive deforestation scenario (GOV), green water appropriation would be unsustainable unlessthe policy goal for natural vegetation cover were reduced from 80% to 50%, in which case the basin’sgreen water scarcity changes from “unsustainable” to “threat” conditions. The XBMT is located withinthe Amazon and Cerrado biomes, which have different mandatory levels of natural vegetation cover93based on whether a property was within the Cerrado (35%), Amazon (80%), or the transition zone be-tween the two (50%). We therefore conclude, that future green water appropriation will, at best, remainunder “threat” conditions considering both a restrictive deforestation scenario (GOV) and a 50% naturalvegetation cover. This analysis does not include potential indirect land use change that might occur inother biomes18,178,246.The increase in green water appropriation by cropland and pasture from natural vegetation throughagricultural extensification, was previously observed in the basin238, at the Mato Grosso state level138,and the Cerrado246. These studies show that land use change can impact the water cycle by returningless water vapour to the atmosphere when compared to natural vegetation with a potential reduction onregional precipitation133,286. Regional precipitation is sourced from greenwater resources as opposed toocean evaporation278, such that land use change may, in turn, affect water availability within and outsidethe basin8,132,184,251. This so-called “moisture recycling”, however, is also expected to be affected bythe expansion of irrigation practices which could transfer additional water vapour to the atmosphere inthe dry season when regional evaporation recycling is enhanced15.5.4.3 Response formulation and study limitationsOur scenarios represent two possible agricultural production options141 considering agriculture remainsthe largest water consumer in the basin. These options reflected whether agricultural intensification re-lied on cropland expansion into pasture (the Green Option), or whether cropping frequency and livestockconfinement becomesmore widespread in the future (the Blue Option) (Table 5.2). Further appropriationof green water from either natural vegetation or pasture depends on land use policies and incentives(e.g. Federal Forest Code, Protective Areas, etc.), while blue water use depends on water manage-ment, which has generally focused on human rather than ecosystem requirements38. Both options haveconsequences for future water availability: continued reduction in natural vegetation cover, which is ac-companied by reduced water vapour supply to the atmosphere could also affect terrestrial ecosystemsthat rely on precipitation for ecosystem functioning141, while dry season water consumed in intensifiedlivestock and irrigation systems could harm aquatic ecosystems downstream.Regional water resources planning requires that connectivity of the water cycle among basins andbiomes be maintained in order to secure future water availability within the basin and beyond. Waterresources management options should consider upstream rain-fed agriculture and small farm reservoirsand their effects on downstream hydroelectric power. Currently, large hydropower dams (> 10 MW)require environmental licenses and impact assessment studies, while smaller dams do not38 suggesting94possible conflicts between up- and downstream water uses. As 22% and 48% of evaporation in theXingu and Amazon Basins, respectively, return to the same basins as precipitation23,263, land and watermanagement in a basin should go beyond its physical boundaries. So far, effects of land use changeon moisture recycling has been absent in water management, in part, due to the difficulty to connectprecipitation source and sink regions in governance131.Water management strategies should also include green and blue water resource use efficiencygains at the field level. For instance, small farm reservoir management should strive to reduce totalevaporation (see Section 4.4.1, Chapter 4), especially when combining livestock confinement with thewidespread use of irrigation for soybean planted at the end of the dry season (as described in our BlueOption). Moreover, green water use should attempt to improve transpiration over evaporation141, whileirrigation should be used efficiently. These actions depend on each individual farmer, their productionsystems and the available training for capacity building of such options. For instance, the recent increaseof cattle density on the current pastureland relied on increased pasture productivity with the potential toreduce the amount of water for feed (see Chapter 4). However, such an initiative has been difficult toimplement in the region137, and the financial returns of increased cattle density still depend strongly onthe price fluctuations in the beef market86.Our study focused on environmental aspects related to water quantity, not social nor economic im-plications of water consumption, nor the effects of water quality on scarcity through the gray WF. As thelargest water consuming sector in the basin, agriculture likely carries the greatest impacts both sociallyand economically. Some studies have made strong connections between agricultural development andhuman and economic development12,212. The effects on water quality resulting from widespread fertil-izer application in the XBMT have been inconclusive thus far with respect to eutrophication174, while fewstudies have investigated the effects of pesticides on water quality in SAM12. The increase in livestockconfinement for both swine and chicken production suggests additional on-farm waste managementwhich could also affect water quality and were not considered in this study.Results of this study relied on the accuracy of the IBIS model to represent the water cycle from landuse maps. Our bottom-up approach relied on maps obtained from Landsat imagery which were usedto infer runoff, and ET using average land use system values derived from previously published results.The derived runoff and ET results were used exclusively for the 2000-2001 and 2014-2015 period andwere close to observations (see Appendix C). Our top-down approach used for the 2030-2031 and2050-2051 periods relied on the assumption that cropland and pasture ET were equal. Cropland andpasture ET can differ by almost 100 mm per crop (see Table C.1) suggesting a potential overestimation95of agricultural land ET (Figure C.4). A reduction in agricultural ET would increase the estimated runoffand decrease agricultural green water consumption. These changes would have a small effect on ourannual blue water scarcity values, and limited effect on our green water scarcity values which were moresensitive to the allocation of ET to natural vegetation (ETRNV ).Our results used IBIS to infer natural runoff under deforestation and climate change scenarios, whichdo not include the feedbacks of water consumption activities. First, blue water scarcity values wereestimated based on the appropriation of runoff as the blue water source. The currently reported XBMTwater use is made up of only 20% of surface water with the remainder coming from deep and shallowwells4. We therefore expect future dry season blue water scarcity limits to take longer to reach as aresult of groundwater extraction in the case of soybean irrigation and cattle confinement. Groundwaterin southeastern Amazonia is deep and known to also feed streams in the Xingu Headwaters104,198.Therefore, the effects of extensive groundwater extraction could only partially contribute to blue waterscarcity. Our results, however, are still expected to represent a general trend towards greater waterscarcity given the large contribution of drinking water for cattle and evaporation from small farm reservoirswhich was entirely attributed to surface water. In this case we also expect groundwater storage to actas a blue water source available to alleviate agricultural water demand in cases of domestic, industrialconsumption and additional demand from confined livestock and irrigated agriculture which merit furtherinvestigation. It is important to note that the inter-annual water scarcity values were based on 3-monthmeans of natural runoff obtained from IBIS, which we found to be close to observed values betweenSeptember and November when blue water scarcity was its greatest in 2050.Moisture recycling feedbacks resulting from reduced vegetation cover and an expanding small farmreservoir network were not included in our estimate of both long-term green and blue water availabilityand, therefore, water scarcity indicators. A reduction in precipitation as a result of land use changewould reduce green water availability in the basin and therefore increase themagnitudes of our estimatestowardsmore unsustainable limits. Similarly, reduced precipitation in the basin can further affect runoff atthe regional scale38, thereby increasing blue water scarcity as estimated here. Both of these limitations,therefore suggest that our results represent mainly a conservative estimate of the effects described inthis study.96Table 5.2: Summary of effects and responses for two agricultural production options focused on pro-duction intensification in the Xingu Basin of Mato Grosso. An illustration of these options is shown inFigure 1.2.Option 1 (Green Option) Option 2 (Blue Option)DescriptionCrops Cattle Crops CattleStrategy Increaseproduction byincreasing croppedareaIntensifyproduction oncurrent landIncrease cropfrequency (triplecropping)Intensifyproduction throughconfinementLand useresponseExpansion of cropsinto pasturelandCattleconcentration oncurrent, moreproductivepasturesCroplandexpansion intopasturelandIncrease cattleconfinementWater useresponseReallocate greenwater from cattle tocroplandReduce water usefor moreproductive pasture;feed sourcedoff-farm (virtualwater transfer);Increase smallreservoir capacityUse irrigation forearly soybeanplanting; include adry seasonirrigated cropIncrease smallreservoir capacity;supplementaldrinking fromsurface andgroundwater inconfined systemsEffects onblue wateruse andscarcityBlue water consumption increaseswith animal population, reservoirevaporation and groundwater use, butremains within sustainable limitsBlue water consumption approachessustainable limits in the dry seasonwith potential effects on downstreamwater availabilityEffects ongreen wateruse andscarcityGreen water use increases for cropsand decreases for pasture (greenwater scarcity constant); long-termgreen water availability may changedue to local (land use) and global(CO2 emissions) climate change;additional evaporation from farmreservoirs increase water vapour flowsto the atmosphere; changes in rainfallaffect blue and green water availabilityin- and out of the basin.Green water use increases for cropsand decreases for pasture (greenwater scarcity constant); long-termgreen water availability may changedue to local (land use) and global(CO2 emissions) climate changes;additional ET from crop irrigation andevaporation from farm reservoirsincrease water vapour flows to theatmosphere; changes in rainfall affectblue and green water availability in-and out of the basin.Water man-agementconsidera-tionsImprove efficiency of blue water use,especially the reduction of evaporationfrom farm reservoirs; consider effectsof land use on water availability(precipitation, runoff) beyond thebasin; integrate land and waterpolicies.Improve efficiencies in blue water usefor irrigation and confined livestock;groundwater management or the useof old farm reservoirs could be usedwithout affecting runoff; considereffects of land use on water availability(e.g. from additional water vapoursupply to the atmosphere).975.5 ConclusionThe application of the VWFSA revealed the importance of the agricultural sector for future land and wa-ter management initiatives in the XBMT. Our study has also provided an important case for estimatingblue and green water scarcities in the context of land use change, climate change and agricultural pro-duction scenarios. Agricultural expansion between 2000 and 2015 led to conditions under which greenwater scarcity moved towards “threat” conditions, while blue water resources remained within sustain-able limits. The evaluation of two water resource use options for agricultural intensification confirmedthe importance of land use policies in further reducing deforestation activity as a driver for intensifyingagricultural production in the basin. Future cropland expansion can rely on further green water appropri-ation by expanding onto pasture, while cattle confinement and cropland irrigation for increased croppingfrequency have the potential of bringing the basin towards dry season sustainable limits. Future studiesshould consider the role of small farm reservoirs and irrigation in the water cycle to identify their im-portance for regional groundwater storage, downstream blue water availability, and also for large scalemoisture recycling and the atmospheric water balance.98Chapter 6Water Footprint Impact Assessmentof Water Use for Cropland and Cattle6.1 IntroductionThe landscape of Brazil’s central Western region has changed significantly since the 1990s followinga rapid rise in the production of agricultural commodities18,63,148,239. Today, the state of Mato Grosso(Figure 1.1) is the largest producer of both soybean and beef in Brazil, and has mostly relied on the ex-pansion of cropland and pasture in both Amazon and Cerrado biomes to reach national and internationalproduction rankings75. The appropriation of natural resources for this expansion has grown together withland use change139 with noted environmental impacts which include the loss of biodiversity41, changesin surface238 and stream temperatures149, as well as degradation of terrestrial ecosystems due to areduction in regional precipitation capable, in part, of tipping the Amazon biome into a “savannization”process59,237. Water resources in Amazonia are particularly at risk of further degradation from croplandand pasture expansion, but also dam construction and mining38. In parallel, additional disruptions to thewater cycle have affected regional evaporation recycling into precipitation141, which may affect futurerain-fed agricultural production and hydropower generation8,184,251.Lathuillière et al.141 defined five possible expansion options for the region which include agriculturalexpansion into natural ecosystems or current pastureland, and agricultural intensification using rainwa-ter harvesting, irrigation, or by improving water vapour flows through an increase in transpiration overevaporation (Figure 1.2). Each option carries distinct uses of water resources that closely follow landmanagement and the resulting partitioning of precipitation into blue and green water141. Differences inland use for agricultural products in SAM therefore entail different potential environmental impacts as aresult of precipitation partitioning which merits further attention in LCA.Recent methodological advances focusing on water use in LCA have addressed differences in thecause and effect impact pathways of the consumption of blue and green water, particularly as they99relate to ET140,164,180,204,216. Some methods have focused on the effects of water consumption onscarcity: Ridoutt and Pfister216 assessed changes in blue water flows as a result of changes in ET onthe land, Núñez et al.180 considered a ratio of water consumption to availability similar to what has beendefined by theWFNetwork117 (see equation 5.6 in Chapter 5). Other methods have highlighted potentialproblems (or mid-point impacts in LCA terminology) reflecting changes in precipitation partitioning andthe distribution of green and blue water at the land surface (Table 6.1): Quinteiro et al.204 introduced theTerrestrial Green Water Flows (TGWF) and River Blue Water Production (RBWP) mid-point impacts todescribe changes in the respective flows to the atmosphere and to liquid stocks as a result of land use.Similar to TGWF, Lathuillière et al.140 proposed the Precipitation Reduction Potential (PRP) impact asa land transformation and occupation impact following United Nations Environment Life Cycle Initiative(UN LCI) guidelines135, which could be considered complementary to Groundwater Recharge Potential(GWRP) described by Saad et al.230 (Table 6.1).This study focuses on water consumption and land occupation impacts of cropland (which includessoybean) and cattle production in SAM with the goal of comparing agricultural production options in theregion using current life cycle impact assessment (LCIA) methods. We follow the four phases of a LCAdescribed in ISO 14044126, and also consider the water scarcity footprint (WSF) following ISO 14046127to highlight competition over blue water resources31: (1) Goal and scope definition, (2) life cycle/WFInventory, (3) life cycle/WF Impact Assessment, (4) interpretation. Results are aimed at providing inputon the environmental performance of the two most common products in the region, while at the sametime comparing and contrasting the available LCIA methods that focus specifically on blue and greenwater partitioning on land (Table 6.1).6.2 Methodology6.2.1 Goal and scope definition, and functional unitsThe goal of the study is to compare agricultural production practices in SAM with the methods that focusspecifically on green and blue water partitioning on land (Table 6.1). We compare extensification andintensification production systems for cropland and cattle based on possible choices of land and waterresources which include the use of irrigation as well as an increase in pasture productivity in both Ama-zon and Cerrado biomes (Figure 6.1). We focus specifically on water quantity with mid-point impactslinked to water consumption (defined as a WSF following ISO 14046 terminology127), and land occupa-100Table 6.1: Summary of mid-point impacts of land occupation that consider the partitioning of precipitationinto blue and green water at the land surface through evapotranspiration (ET) from natural vegetation(with evapotranspiration ETNV ), current land use (ETLU), and environmental flow requirements (ETEFR)(see Section 6.2 for further description). Characterization factors for life cycle impact assessment (LCIA)from Quinteiro et al.204 are subject to conditions.Mid-point impact Description Life CycleInventory(LCI)Characterizationfactor for the LifeCycle ImpactAssessment(LCIA)Referenceto atmospherePrecipitationReductionPotential (PRP)Reduction inregionalprecipitationreturning to thesame river basinArea (A)(ETNV−ETLU)erLathuillière etal.140Terrestrial GreenWater Flows(TGWF)Reduction in ETnot returning to theriver basin;conditions forETLU,eff < ETPNV,effEffective netgreen water(NGWeff ) 1−ETLU,eƒ ƒETNV,eƒ ƒQuinteiro et al.204To landGroundwaterRecharge Potential(GWRP)Change ingroundwaterrechargeArea (A)GWRNV−GWRLUSaad et al.230Runoff ReductionPotential (RRP)Reduction in runoffgenerated byregionalprecipitationreturning to thesame river basinArea (A)(ETNV−ETLU)αerThis study, basedon Berger et al.23River Blue WaterProduction(RBWP)Changes insurface runoff fromincreases in landET; conditions forETPNV,eff < ETLU,eff< ETEFR,effNGWeffETLU,eƒ ƒETEFR,eƒ ƒQuinteiro et al.204101tion. According to the ISO 14046 standard127, a WSF is the result of an life cycle assessment focusedon potential impacts due to blue water consumption (see Section 2.3.3), and has been expressed as afunction of the level of water scarcity in a basin31.The geographical scope is limited by the boundaries ofthe XBMT (Figure 5.1) while considering production practices averaged for Mato Grosso’s Amazon andCerrado biomes (Figure 1.1). The system boundaries are the cradle-to-farm gate production of cropsand cattle in 2014-2015. For crops, we consider soybean in both rain-fed and irrigated systems as twooptions of interest for production in the region that respectively represent extensification and intensifi-cation options (Figure 6.1). These systems are an integral part of soybean production as the primarycrop of interest and the main driver of land use change in Mato Grosso18,63,148,246. We consider tworotations: a rain-fed soybean-maize rotation, and an irrigated soybean-rice-bean rotation (Chapter 3). Inrain-fed systems, soybean is typically planted at the beginning of the wet season (October-November)with maize immediately planted following the soybean harvest (February-March). In the irrigated sys-tem, soybean is planted at the end of the dry season (September) to allow for an earlier harvest tobenefit a rice harvest in the wet season (April), prior to planting a triple crop in the dry season (fullyirrigated, see Chapter 3). These options reflect the commonly used double cropping system245, and anirrigation option which allows for a crop to be planted in the dry season (bean in this case). For cattle,we consider the production system of the Nelore species (Bos taurus indicus, most common in MatoGrosso), focusing on the differences in pasture productivity as an indicator for an increase in pasturecattle density, and also consider water consumed by animals in small farm reservoirs (see Chapter 4).The functional units for cropland and cattle are respectively 1 ha of cropland (containing soybean) and1 kg LW at farm-gate.6.2.2 Life cycle inventoryWe consider three life cycle inventories (LCI) for each production system based blue water consumption,and changes in blue and green water from land occupation (Table 6.1): one LCI for the WSF (blue waterconsumed as described with the WF Inventory), one LCI based on land area (A), and one LCI basedon ET (as effective net green water, NGWeff , described in equation 6.1). Blue water uses for crops andcattle production are in competition with other human and ecosystem uses in the basin and are thereforesusceptible to deprive these users of water31, and expressed in LCI by theWF inventory. WF inventoriesinclude blue water consumed for irrigation (assuming all irrigation becomes ET), drinking water for cattleprovided by small farm impoundments, and the volume of water evaporated from these reservoirs. Bluewater consumption was based on previous results for both cropland (Chapter 3) and cattle (Chapter 4)102Figure 6.1: Scenarios for production systems considered in this study for estimating the mid-point envi-ronmental impacts of cropland and cattle production systems (cradle-to-farm gate). Scenarios includecropland extensification on natural vegetation (NV) or pasture, and differences in pasture productivityfor cattle. An illustrative example is shown in Figure 1.2 (panels A, B and E).(see Table 6.2). Blue water was allocated to dry season irrigation of bean (118 mm over the season, seeChapter 3), as well as water consumption by cattle which was estimated over the course of the animal’sdevelopment cycle in 48 months (Chapter 4) and divided by two to obtain mean annual consumption.Given the lack of information on small farm reservoir water balances in the region, we assumed thatboth cattle drinking and evaporation from the reservoirs diminished streamflow or groundwater rechargewith potential impacts on future water availability. In 2014, the XBMT contained 9463 ha of small farmimpoundments detected using remote sensing (Chapter 5). Drinking water for cattle was based on thedevelopment stage of the animal and averaged 40.5 × 10-3 m3 y-1 (kg LW)-1. There was 0.141 km3y-1 of small reservoir evaporation in the XBMT in 2014-2015 for a total live cattle population of about3.5 million (Chapter 5) which we attributed to the total cattle live weight based on mean cattle weight inrespective development phases (95 kg LW cattle-1 for calves, 266 kg LW cattle-1 for mid-life cattle, and429 kg LW cattle-1at the end-of-life) (Chapter 4). This calculation provided a mean allocation of smallfarm reservoir evaporation attributed to the live cattle herd of 0.16 m3 y-1 (kg LW)-1 .Current impact assessment methods involving water partitioning from land occupation (Table 6.1)either include land area (A) or ET (as NGWeff , see below) in the LCI calculation. Both methods forderiving GWRP and PRP use land occupation (A, ha) as the LCI, as well as the Runoff ReductionPotential (RRP) impact introduced in this study (see Section 6.2.3). Land occupation is based on the103annual occupation of land from the crop rotations containing soybean in both rain-fed and irrigatedsystems (see Chapter 3). We translate pasture consumed by cattle into hectares of land in Mato Grossobased on an averaged male and female 48-month animal development cycle (divided by two to obtainannual consumption), and considering low and high pasture productivity scenarios (see Chapter 4)(Table 6.3). In the case of TGWF and RBWP, Quinteiro et al.204 introduced ET in the LCI as defined byeffective net green water (NGWeff , m3 ha-1 y-1) shown in equation 6.1NGWeƒ ƒ = ETLU,eƒ ƒ − ETNV,eƒ ƒ (6.1)where ETLU,eff (m3 ha-1 y-1) is the effective ET of the current land use (cropland or pasture), ETNV,eff (m3ha-1 y-1) is the effective ET of the natural vegetation (Amazon or Cerrado). Both ETLU,eff and ETNV,effare calculated following Quinteiro et al.204 in equation 6.2ET,eƒ ƒ = ET− ETer (6.2)where ETi is the ET of a land use i, and er (0.22, dimensionless) is the basin internal evaporationrecycling ratio23 constrained to the Xingu River Basin. The physical meaning of NGWeff is the change inwater vapour returning to the atmosphere that is effectively lost from the basin (considered consumed),following the change in land occupation from NV to LU. In addition, we consider the possible effectsof small reservoir evaporation on the water cycle by comparing the amount of evaporation from theimpoundments to the NV. We therefore apply equation 6.1 by considering ETLU,eff as mean evaporationfrom reservoirs.6.2.3 Life cycle impact assessmentWe follow two frameworks from the UN LCI to determine mid-point impacts of water consumption andland occupation. Impacts of blue water consumption are assessed by estimating the amount of waterdeprived to human and ecosystems in the region using the AWARE method31. Briefly, this methodprovides a WSF127 through a characterization factor that represents the degree of competition in a riverbasin following equation 6.331. The WSF has also been referred to as a stressed weighted WF112,WSF= LCCF (6.3)104where LCIw (m3 ha-1) is the blue WF inventory (or all blue water consumed per ha: e.g., irrigation,water evaporated from reservoirs, etc.), CFw (dimensionless) is the characterization factor based onthe available water remaining in the basin, taking into account both human and ecosystem blue waterconsumption. Values of CFw were obtained following Boulay et al.31CF =AMDordAMD(6.4)where AMDworld (0.0136 m3 m-2 mo-1) is a global normalization factor representing global blue wateravailability minus demand, and AMDi represents blue water availability minus demand in river basini as the difference between available blue water, human and ecosystem consumptions divided by thearea of the basin. For the Xingu River Basin, the annual value of CFw is 1 (no irrigation) and 1.1 (withirrigation)31.Secondly, impacts of land occupation on the water cycle were determined following Koellner et al.135as shown in equation 6.5occ = ACFjtocc (6.5)where Iocc is the land occupation impact, calculated using a characterization factors of impact j (CFj ),area A (ha) and occupation time tocc (years). Characterization factors were calculated for threemid-pointimpacts affecting the land and atmospheric water cycles: GWRP following Saad et al.230, PRP followingLathuillière et al.140, and RRP which we propose in this study, all of which are shown in equations 6.6to 6.8CFGWRP =GWRNV−GWRLU (6.6)CFPRP = (ETNV−ETLU)er (6.7)CFRRP = (ETNV−ETLU)αer (6.8)where CFGWRP , CFPRP , and CFRRP (m3 ha-1 y-1) are the respective characterization factors of landoccupation for GWRP, PRP, and RRP,GWR and ET (m3 ha-1 y-1) are respectively groundwater rechargeand ET for natural vegetation (GWRNV , ETNV ) and the land use (GWRLU , ETLU), and αer is the basininternal evaporation recycling ratio multiplied by a runoff coefficient r which together equal to 0.07 (di-mensionless), and represent a recycling ratio of water vapour returning to the basin as blue water23.Values of GWR were obtained following the water balance equation described by Saad et al.230 and105shown in equation 6.9GWR=P− ETr(6.9)where P (mm y-1) is the annual precipitation, and r (dimensionless) is the runoff coefficient, both of whichare defined for NV and LU to derive GWRNV and GWRLU .Characterization factors for TGWF and RBWP (as CFTGWF and CFRBWP ,dimensionless) were ob-tained from equations 6.10 and 6.11 following Quinteiro et al.204CFTGWF = 1−ETLU,eƒ ƒETNV,eƒ ƒ(6.10)when ETLU,eff < ETNV,eff (both of which are obtained with equation 6.2), andCFRBWP =ETLU,eƒ ƒETEFR,eƒ ƒ(6.11)when ETPNV,eff < ETLU,eff < ETEFR,eff , where ETEFR,eff is the effective ET that maintains environmentalflow requirements in the Xingu River Basin and was defined by equation 6.12 following Quinteiro et al.204ETEFR,eƒ ƒ = P−χEFRP−ETNV,eƒ ƒ(6.12)where χEFR (0.42, dimensionless) is the fraction of environmental flow requirements to the long-termmean discharge of the Xingu River Basin (see Chapter 5). The conditions to apply equations 6.10 and6.11 are the characterization factors under specific land occupation scenarios described in this study,in which both characterization factors are mutually exclusive. Therefore, if ETLU,eff < ETPNV,eff , thenCFRBWP = 0, and if ETPNV,eff < ETLU,eff < ETEFR,eff , then CFTGWF = 0 according to Quinteiro et al.204.All characterization factors were derived using previously published input data from both the Amazonand Cerrado biomes as described in Lathuillière et al.143 and adapted for rain-fed cropland, irrigatedcropland, and pasture using input parameters shown in Table ResultsPotential impacts of cropland and cattle production obtained following calculations from the LCI (Table6.3) showed differences with respect to the biome and production system with both positive and negativeimpacts based on the impact category considered (Figures 6.2 to 6.4). Land occupation impacts from106Table 6.2: Input parameters used in this study, following Lathuillière et al.143Parameter Symbol Amazon Cerrado Unit ReferencePrecipitation P 2096 1369 mm y-1 Rodrigues et al.226ET of naturalvegetationETNV 1099 817 mm y-1 Lathuillière et al.138;Oliveira et al.185Cropland ET,rain-fedETLU 801 mm y-1 See Chapter 3Cropland ET,irrigatedETLU 982 mm y-1 See Chapter 3Pasture ET ETLU 794 mm y-1 See Chapter 3Pastureproductivity (low)5.3 ton DM ha-1 See Chapter 4Pastureproductivity (high)3.0 ton DM ha-1 See Chapter 4Dry seasonirrigation118 mm y-1 See Chapter 3Small farmreservoirevaporation1421 mm y-1 See Chapter 5Runoff coefficient r 2a dimensionless Lathuillière et al.143Basin internalevaporationrecyclingcoefficient(precipitation)er 0.22 dimensionless Berger et al.23Basin internalevaporationrecyclingcoefficient (bluewater)αer 0.07 dimensionless Berger et al.23Filtration distanceto groundwater0.8-1.5 m Beck et al.21;Lathuillière et al.143aAssumption107cropland replacing NV were greater in the Amazon biome than in the Cerrado with larger differencesin all categories, particularly with rain-fed cropland’s GWRP (−1490 m3 in the Amazon compared to−690 m3 in the Cerrado), PRP (656 m3 in the Amazon compared to 304 m3 in the Cerrado) and TGWF(−630 m3 in the Amazon compared to −158 m3 in the Cerrado). A change from irrigated to rain-fedcropland in both biomes carries a WSF (representing the degree of competition resulting from bluewater consumption at 1298 m3 world equivalents for irrigated cropland), and reductions in impacts tothe atmosphere, and to the land. In the Cerrado, the main difference between rain-fed and irrigatedcropland was the change in sign of the impacts of PRP (304 m3 to −95 m3), GWRP (−690 m3 to 215m3) or RRP (97 m3 to −30 m3), while the same shift replaced TGWF impacts (−158 m3) by RBWPimpacts (299 m3) (Figures 6.2 and 6.3). The replacement of pasture with cropland in both Amazon andCerrado biomes showed much smaller impacts compared to irrigated cropland. Impacts decreased forPRP (−426 m3) and RRP (−136 m3), and increased for GWRP (968 m3), and RBWP (1346 m3). Cattleproduction generally showed lower impacts in the Cerrado biome when compared to the Amazon biome,and in high productivity pasture when compared to low productivity (Figure 6.4 ). The impact categorieswith the largest magnitude were PRP (4.5 m3 (kg LW)-1 for a low productivity pasture system in theAmazon), GWRP (−5.1 m3 (kg LW)-1), TGWF (−2.2 10-2 m3 (kg LW)-1) and RRP (1.4 m3 (kg LW)-1)(Figure 6.4).The above impacts were obtained from the characterization factors shown in Tables 6.4 and 6.5.Characterization factors for changes in water vapour transfers to the atmosphere with land occupation(CFPRP, CFTGWF ) ranged from −426 m3 ha-1 y-1 for a pasture-to-irrigated cropland transition (Table6.5) to 683 m3 ha-1 y-1 from an Amazon NV-to-pasture transition (Table 6.4). These characterizationfactors had matching values of CFTGWF of 0 for CFPRP< 0, and up to 0.28 for CFTGWF > 0 (Tables 6.4and 6.5). Values of CFGWRP were negative when considering NV-to-rain-fed cropland transitions, butpositive in Cerrado NV-to-irrigated cropland (257 m3 ha-1 y-1), pasture-to-rain-fed cropland (63 m3 ha-1y-1) and pasture-to-irrigated cropland (968 m3 ha-1 y-1). Values of CFRRP were of opposite sign to thatof CFGWRP with values ranging from −136 m3 ha-1 y-1 (pasture-to-irrigated cropland) to 217 m3 ha-1y-1 (Amazon NV-to-pasture transition). A positive CFGWRP was also matched by a non-zero value ofCFRBWP in Cerrado NV-to-irrigated cropland (0.89), pasture-to-rain-fed cropland (0.73), and pasture-to-irrigated cropland (0.89) transitions (Table 6.4).108Table 6.3: Life cycle inventory data for both crop (1 ha) and cattle (1 kg LW) production.Product(functionalunit)Biome Blue (ET) Blue(drink) Green NGW eff Land use Referencesm3 y-1 ha y-1Cropland,rain-fed(1 ha)Amazon 0 0 8010 −2324 1 Chapter 3,IBGE121Cerrado −1076Cropland,irrigated(1 ha)Amazon 1180 0 8640 −913 1 Chapter 3,IBGE121Cerrado 335Cattle,low pro-ductivitypasture(1 kg LW)Amazon 0.16 (2.74× 10-2)a20 × 10-3 25.7 −8 3.26 × 10-3 Chapter 4,IBGE121Cerrado −4Cattle,high pro-ductivitypasture(1 kg LW)Amazon 0.16 (4.09× 10-2)a20 × 10-3 2.2 −4 1.85 × 10-3Chapter 4,IBGE121Cerrado −2aValue in brackets is the effective blue water evaporation from reservoirs109Figure 6.2: Mid-point impacts of water consumption and land occupation of cropland (m3 ha-1) fromAmazon natural vegetation (NV) and pasture. Impacts are Precipitation Reduction Potential (PRP),Terrestrial Green Water Flows (TGWF), Groundwater Recharge Potential (GWRP), Runoff ReductionPotential (RRP), River Blue Water Production (RBWP).Figure 6.3: Mid-point impacts of water consumption and land occupation of cropland (m3 ha-1) fromCerrado natural vegetation (NV) and pasture. Impacts are Precipitation Reduction Potential (PRP),Terrestrial Green Water Flows (TGWF), Groundwater Recharge Potential (GWRP), Runoff ReductionPotential (RRP), River Blue Water Production (RBWP).110Table 6.4: Characterization factors for the consumption of blue water (CFw ) and the land occupationimpacts for natural vegetation (NV)-to-cropland (rain-fed and irrigated) and NV-to-pasture in both Ama-zon and Cerrado biomes: Groundwater Recharge Potential (CFGWRP), Precipitation Reduction Potential(CFPRP), Runoff Reduction Potential (CFRRP), Terrestrial Green Water Flow (CFTGWF ) and River BlueWater production (CFRBWP).CharacterizationFactorCropland, rain-fed Cropland, irrigated Pasture ReservoirsAmazon Cerrado Amazon Cerrado Amazon Cerrado Amazon CerradotoatmosphereCFPRP(m3 ha-1 y-1)656 304 257 -95 683 331 −715 −1067CFTGWFa 0.27 0.15 0.11 0 0.28 0.16 0 0to landCFw 0 0 1.1 1.1 0 0 1.1 1.1CFGWRP(m3 ha-1 y-1)−1490 −690 −585 215 −1553 −753 NAb NAbCFRRP(m3 ha-1 y-1)209 97 82 −30 217 1025 −228 −340CFRBWPa 0 0 0 0.89 0 0 0.90 1aDimensionless; bNot available111Figure 6.4: Mid-point impacts of water consumption and land occupation for cattle (m3 (kg LW)-1) in bothAmazon and Cerrado biomes. Impacts are Precipitation Reduction Potential (PRP), Terrestrial GreenWater Flows (TGWF), Groundwater Recharge Potential (GWRP), Runoff Reduction Potential (RRP),River Blue Water Production (RBWP).Table 6.5: Characterization factors for the consumption of blue water (CFw ) and the land occupation im-pacts for pasture-to-cropland (rain-fed and irrigated) in both Amazon and Cerrado biomes: GroundwaterRecharge Potential (CFGWRP), Precipitation Reduction Potential (CFPRP), Runoff Reduction Potential(CFRRP), Terrestrial Green Water Flow (CFTGWF ) and River Blue Water production (CFRBWP).Characterization factor Cropland, rain-fed Cropland, irrigatedto atmosphereCFPRP(m3 ha-1 y-1) −27.7 −426CFTGWFa 0 0to landCFw 0 1.1CFGWRP(m3 ha-1 y-1) 63 968CFRRP(m3 ha-1 y-1) −8.8 −136CFRBWPa 0.73 0.89aDimensionless1126.4 Discussion6.4.1 Crop and cattle intensification impacts on water partitioningThe combined LCIA methods were able to provide additional information on the potential effects of cur-rent intensification at the field level within current trends in the region148,244. In SAM, pasture has histor-ically replaced NV in both Amazon and Cerrado biomes while cropland has replaced NV but also olderpasturelands, often leading to indirect land use change through additional deforestation of NV for pastureinto the Amazon biome10,18. Such deforestation activity has had noted impacts on biodiversity, aboveand belowground carbon as well as erosion as quantified in LCA143, which can be complemented by theeffects described by precipitation partitioning. A NV-to-rain-fed cropland transition was accompanied bya reduction in water vapour transfer to the atmosphere which translated into a loss of water vapour fromthe basin (as quantified with TGWF), a loss of precipitation recycled within the basin (from PRP), or re-turning to land as blue water (RRP), leading to additional local groundwater recharge (GWRP). Overall,this transition would increase blue water resources within the basin with a trade-off between upstreamgroundwater (expressed in GWRP) and downstream surface water resources (TGWF, RBWP). The lossof water from the basin (TGWF, RBWP) can affect water availability downstream and, consequently,increase water scarcity as in the case of cropland irrigation. Moreover, cropland irrigation transferredblue water resources to the atmosphere through ET, especially land occupation on Cerrado NV or pas-ture. The amount of precipitation recycled within the basin (PRP) actually increased when consideringimpacts of irrigated cropland in the Cerrado which could have potential benefits to ecosystems140.Impacts of cattle production were affected by low productivity and high productivity pasture mostlyfrom the amount of dry matter that cattle can consume per hectare of pasture as well as the choice ofNV. The change in land occupation impacts of cattle following a NV-to-pasture transition showed similarimpacts than the NV-to-cropland transitions with losses of water vapour returning to the atmosphere(PRP), as well as surface and groundwater (RRP, GWRP). Similar to irrigation, small farm impound-ments constructed for cattle drinking did not carry any losses of water vapour outside of the basin asexpressed through TGWF, rather they potentially reduced the production of blue water downstream.Indeed, extensive networks of small farm reservoirs can reduce stream connectivity34 and favour addi-tional evaporation with effects on downstream water availability (see Chapter 5).Cropland extensification into pasture overall had the lowest land occupation impacts while cattleintensification in the Cerrado had lower impacts than in the Amazon biome. The Brazilian Federal ForestCode currently places deforestation limits on properties located in the Amazon and Cerrado biomes113through the Legal Reserve which requires farmers to respectively maintain 80% and 20% of naturalforest cover (depending on the year of deforestation and farm size)33. This difference has historicallyled to more deforestation in the Cerrado compared to the Amazon253 with impacts on biodiversity andecosystem services142. There were distinct blue and green water trade-offs expressed in the croplandextensification and cropland irrigation impact assessments expressed through TGWF and RBWP. Bothimpacts quantify the effects of land use change on downstream water availability and could therefore belinked also to the WSF206 (see Section 6.4.2).We note uncertainties with the above interpretation of our results which are common in LCA studiesthat rely on crop water balances and can affect both LCI and the characterization factors189,205. Weexpect uncertainty in our LCI values as a result of geographic differences in water use for crops andcattle across the basin. Our values of ET came from field measurements in both rain-fed and irrigatedfields and assumed no field runoff, and little information on drainage (Chapter 3) which can change withfield declination and soil conditions. The amount of water consumed by cattle was based primarily on thetotal live weight of the animal and represented an average of male and female consumption in the stateof Mato Grosso (Chapter 4). Drinking water for cattle can vary greatly based on climate187,218, whilewe also expect geographic differences in the small farm reservoir evaporation across the basin. Thecharacterization factors used in this study represent regional averages for the state of Mato Grosso in thecase of GWRP143, while all characterization factors related to the internal processes of the basin (e.g.,PRP, WSF) can also know geographic and temporal variability. While our values of ET for NV in bothAmazon and Cerrado biomes have been estimated through remote sensing138,185, field measurementshave confirmed the difference in magnitude between vegetation spanning from 965 mm y-1 (Cerrado)to 1384 mm y-1 (Amazon) and follow the precipitation gradient across the two biomes141. This averagedifference in ET between the biomes therefore confirms the difference in impacts observed, despite thegeographic uncertainty in landscape ET from cropland and pasture.Furthermore, this study also focused on attributional LCA by allocating impacts to two products as-suming that their respective systems are mutually exclusive. In fact, the options proposed for agriculturalintensification overall are interconnected (Chapter 5). For instance, between 2001 and 2010, close to4962 km2 of pasture was converted to cropland in the XBMT238, while cattle animal population in-creased, thereby increasing cattle density from 0.57 cattle ha-1 to 0.97 cattle ha-1in the basin (Chapter5). Therefore, our proposed cropland extensification option could lead to cattle intensification, but alsocattle extensification in- or outside the basin (indirect land use change) which was unaccounted for inthis study.1146.4.2 Complementarity in mid-point impactsThe mid-point impacts used in this study represent changes in hydrological flows as a result of landoccupation and water consumption which can be synonymous from an ISO 14046 perspective127, butwere considered separately in this study. On the one hand, land occupation can change hydrologybased on precipitation partitioning with consequences on end-point impacts (resulting from mid-pointimpacts) to human health, ecosystems or water resources. The mid-point impacts used in this studycould be interpreted as so-called fate factors in LCA, which are used in LCIA to estimate changes inthe water cycle to derive end-point impacts (e.g., impacts to ecosystem quality181). For instance, inLathuillière et al.140, PRP represents the precipitation volume not returning to the river basin as a resultof a land occupation or transformation activity with potential impacts to terrestrial ecosystems as a resultof diminished soil moisture. Similarly, other fate factors have been conceptualized in the context ofgroundwater extraction265. The long-term effects of these fate factors also affect water availability. Forinstance, the land occupation impact GWRP showed increased recharge following a NV-to-croplandtransition, while a reduction in precipitation could also reduce long-term runoff within the basin. Theseeffects on water resources are also expected to change competition over remaining water resourceswhich were expressed with the WSF in the case of blue water consumption only. Land use, land usechange and dam operations are known as major contributors to changes in water availability and havealready contributed to moving water scarcity further downstream270, which could also be expressedmore explicitly in LCA.The WSF as expressed using the AWARE method31 aims to answer the question posed by Boulayet al.30,31 with a specific focus on water quantity: “What is the potential to deprive another freshwateruser (human or ecosystem) by consuming freshwater in this region?”. The method is based on theratio of water demand to availability where availability is defined as the amount of water remaining oncehuman and ecosystem demands have been met31. In this method, theWSF is considered a “proxy” mid-point, meaning that it isn’t linked to any particular end-point impact31. A similar “proxy” mid-point wasdeveloped by Berger et al.23 as the water depletion index which was used to evaluate a water depletionrisk with respect to an activity in a basin. The results from these impact indicators are therefore differentthan the interpretation of impacts as fate-factors used as a predictive impact assessment expressedthrough an end-point impact as seen with PRP, RRP, and GWRP. Impact categories TGWF and RBWPbest represent the changes in water from land occupation that could be considered “consumed” andcould be further represented in a WSF from land occupation206. The water depletion index proposed byBerger et al.23 was based on a ratio of water consumption to availability which considers the removal of115evaporative fraction returning to the basin as precipitation from water consumption through er (withoutconsidering ecosystem water demand). If we interpret TGWF and RBWP as water effectively consumedas a result of land occupation, then we can calculate a WSF using a water depletion of 0.01 m3 depletedper m3 consumed following Berger et al.23. This would provide a risk of freshwater depletion rangingfrom 0.71-13.5 m3 water depleted ha-1 for cropland and 2.4 × 10-4 to 2.2 × 10-2 m3 water depleted(kg LW)-1 for livestock (considering absolute values of TGWF and RBWP). Similarly, Quinteiro et al.206propose to use AWARE characterization factors (or CFw as described in this chapter, equation 6.4) asa means to derive a WSF from a change in river runoff (Figure 6.5).TheWater use in LCA (WULCA) working group has recommended a framework for including transfersof freshwater sources and sinks in LCIA by considering hydrological “compartments” in the water cycle,keeping in mind regional and global scale effects182. In our study’s context, the magnitude of thesetransfers depended primarily on the magnitude of er which was used either in the LCI (for TGWF, RBWP,see equation 6.1), or in the LCIA phase (for PRP and RRP, see equations 6.7 and 6.8). Our value ofer was constrained to the Xingu River Basin, but could be confined to the Amazon biome or even thecontinent140,263 as a means to represent differences between local and regional hydrological scales.For instance, Lathuillière et al.140 calculated PRP of soybean production in Amazonia in a small region(2.76 × 1010 m2), the Xingu River Basin (5.1 × 1011 m2), and the Amazon biome (7.0 × 1012 m2) andestimated respective PRP impacts of 86.5 m3 ton-1 soybean, 323 m3 ton-1, and 703 m3 ton-1 followingthe corresponding values of er for each area of influence. Similarly, impacts expressed in GWRP couldbe complemented with fate factors based on groundwater depletion as derived by van Zelm et al.265.Moreover, the recognition of impacts to the end-point impact to natural resources could respond to thelong-term effects of land occupation on water availability, rather than expressing the impact purely as aWSF. Within the development of these indicators in LCA, it is important to maintain focus on avoidingdouble counting on both water quantity and quality perspectives181.6.5 ConclusionThis study aimed to evaluate potential agricultural production options for cropland and cattle in SAM byobserving agricultural extensification and intensification using six distinct impact assessment methodsthat focus on the effects of land occupation and water consumption on water quantity. Our croplandextensification option relying on a pasture-to-cropland land use transition resulted in lower impacts ofproduction when compared to NV-to-cropland transition, while irrigation showed potential benefits when116Figure 6.5: Complementarity of mid-point impacts of Groundwater Recharge Potential (GWRP), RiverBlue Water Production (RBWP), Terrestrial Green Water Flows (TGWF), Precipitation Reduction Poten-tial (PRP), and Runoff Reduction Potential (RRP) in a natural vegetation (NV)-to-cropland or pasture-to-cropland land use transition.focusing specifically on land occupation impacts due to additional water vapour transfers to the atmo-sphere. The comparison of high and low productivity pasture for cattle revealed the importance ofpasture management in reducing the impacts of cattle production in the region, but also the effects ofland use on downstream water availability. While five of the impact assessment methods tested werespecifically linked to land occupation, further model integration is needed to assess the full extent of landoccupation on the water cycle. We have suggested a path forward to further integrate the link betweenland occupation impacts and the WSF, while future research should also consider longer term impactsto freshwater resources embodied in a natural resources end-point impact.117Chapter 7Conclusions and Outlook7.1 Overall research significanceTheWF has, so far, provided additional perspective on the importance of water resources for agriculturalproduction and food consumption with implications on the role of future water management to help feedthe world224,272. The emerging field embodied by the WF, has matured since 2002113 and has reacheda level where the integration of methods and perspectives is necessary for multilevel water resourcesdecision-making. The overall objective of this thesis was to advance the field of the WF by sheddingmore light on the perspectives behind the main WF approaches expressed within the harmonize WF as-sessment (Chapter 2). Moreover, the application of the individual phases of the harmonized assessmentto agricultural production in SAM (Chapters 3 to 6) has highlighted important strengths and limitationsof the current methods (described further in Section 7.3). In the preceding chapters, I explored the inter-pretations of the WF as employed by the water resources management community (following guidelinesfrom the WF Network117), and the LCA community (following ISO14044126 and ISO 14046127), pre-sented and applied a framework to harmonize existing methods into one assessment that can informboth micro- and macro-level decisions at the field, product and river basin levels. As such, the proposedframework can be followed in its entirety (Table 2.2), or considering specific phases of interest based onthe defined goals and scope of the assessment within the domain of interest defined as either Nature orProduction (Figure 2.2).The harmonized WF assessment was applied to soybean production and cattle ranching in SAM,considering field measurements, production system modeling, and river basin modeling of both naturalhydrological processes and consumption activities following distinct scenarios and agricultural produc-tion options for the future (e.g., as shown in Figure 1.2). Information obtained by the individual phasesof the harmonized WF assessment are outlined below and integrated to provide insight into policy de-cisions for water management in the region. Following the Goal and Scope defined in Chapter 2, theassessment called for an estimate of volumetric WFs for both soybean and cattle, followed by a VWF118assessment (Chapters 3, 4, Section 7.2.1), a WFIA (Chapter 6, Section 7.2.2), and a VWFSA (Chapter5, Section 7.2.3). Results are summarized below, prior to providing additional perspectives on the WFand future work (Section 7.3).7.2 Integrating results from the harmonized water footprintassessment7.2.1 Volumetric water footprint assessmentThere are opportunities to reduce the total volume of water for production in key agricultural products ofSAM, starting with the measurement of the VWF of one tonne of soybean, and cropland more generally(Chapter 3), as well as the modeling of the VWF of cattle (Chapter 4). The average green VWF was780-1182 m3 ton-1 soybean for the 2015-2017 period and considering both rain-fed and irrigated fields.These results were much lower than the modeled results for the 2000-2010 period of 1590 m3 ton-1soybean estimated by Lathuillière et al.139, and the 1553 m3 ton-1soybean for the best 10th percentilegreen-blue VWF obtained globally for the 1996-2005 period158, suggesting only limited opportunity forfurther reduction in water use for soybean production in the studied systems. Irrigation at the farm levelwas used primarily as a strategy for anticipating the soybean planting date and increasing the croppingfrequency with a triple crop of bean. This strategy is different from other strategies proposed, for in-stance, in sub-Saharan Africa where water can be a limiting factor in reaching high yields (known asclosing the yield gap)221,222. Further reduction in the VWF would require an increase in yield for thesamewater use, whichmay be achieved by increasing fertilizer and other inputs, technology (breeding orseed development), while either maintaining or reducing evaporation in favour of crop transpiration141.These strategies are field specific and therefore should consider local soil and climate conditions. Ad-ditional fertilizer may contaminate surface and groundwater, although little evidence to date suggeststhat soils in soybean-dominated watersheds are leaching considerable N or P in Mato Grosso’s waterbodies173,174,228. While health issues have been reported in nearby states following pesticide use, sim-ilar evidence on the effects of pesticides is scarce in Mato Grosso12. Similarly, rainwater harvesting orirrigation could increase water vapour flows to the atmosphere as an additional supply of water for agri-culture in the dry season (Chapter 3), but also either increase (through percolation) or decrease (throughwater abstraction) groundwater recharge93,141. Finally, technological advances that reduce the lengthof crop development cycles or improve drought resistance would also contribute to reducing the green119VWF, assuming similar yields are maintained.In 2015-2016, average operational costs of soybean production in Mato Grosso amounted to 921.18USD ha-1or 319 USD ton-1 soybean123 (for a Mato Grosso yield of 2.89 ton ha-1 according to IBGE121),more than half of which constituted the cost of inputs123. Given that the 2016 Brazilian producer priceof soybean was 343.90 USD ton-1 (or roughly 0.35 USD per m-3 of water)75, producers would have hadto carefully weigh costs and benefits to improve the economic return of soybean production. Maize wasthe fastest growing second crop, in terms of area expansion, between 2000 and 2011245, which allowsfor additional income in the same annual period and therefore is also an important consideration forfarm-level decision-making. Additional income could come from an irrigated triple cropping system con-sidering longer-term developments in the production system such as large investments in infrastructure.The average VWF of cattle for Mato Grosso was 324-373 L (kg LW)-1 of which about 58-59% wasattributed to small farm reservoir evaporation in 2015, with an additional 21,400-52,100 L (kg LW)-1of green water required for feed (Chapter 4). The VWF of cattle could be reduced through on-farmwater management strategies that reduce evaporation of reservoirs, but also the relationship betweencattle population and reservoir capacity. The large standard deviation in evaporation from small farmreservoirs suggested that regional benchmarks could be set for comparison of water consumption forcattle production across Mato Grosso. The VWF of the feed depended primarily on pasture productivitywhich is also linked to cattle population and density in relation to pasture area. Together, a decrease inreservoir evaporation and increase in pasture productivity would reduce the total VWF of cattle, whichcould be achieved through cattle intensification. Similarly, breeding and genetics may speed up thedevelopment cycle of cattle252 thereby reducing the amount of resources going into the productionsystem. There are still many barriers in the region to successfully increase cattle concentration in MatoGrosso137, some of which are the costs of implementation which were measured in a range of 800-2600USD ha-1 in a pilot project involving 13 farms in northern Mato Grosso86. While further intensificationof cattle has often been seen as a strategy to reduce greenhouse gas emissions from deforestation, acontinued increase in cattle population252 will require an increasing amount of water which, therefore,raises questions about the use of natural resources in cattle intensification.7.2.2 Water footprint impact assessmentOne hectare of cropland in the XBMT showed greater impacts to the water cycle in the Amazon whencompared to the Cerrado, expressed, for instance, by changes in groundwater recharge, regional pre-cipitation and downstream discharge (Chapter 6). These potential impacts can vary with the hydrological120scale, and accompany a series of additional impacts to biodiversity and ecosystem services that havealso been used for comparison of soybean production through different land uses142,143. In 2010, totalimpacts to ecosystem services (which include groundwater recharge) were also greater in the Ama-zon than the Cerrado biome with the greatest impacts calculated in the Climate Regulation Potential,Mechanical Water Purification Potential, and Biotic Production Potential impact categories, which, re-spectively, describe the amount of above and below ground carbon lost to the atmosphere, and the soil’sability to filter water and sustain biomass142,143. Land occupation impacts were significantly reducedwhen considering cropland produced on land converted from pasture in both the Amazon and Cerrado.Similarly, impacts of cattle production on the water cycle were greater in the Amazon when com-pared to the Cerrado with greater impacts estimated for lower productivity pasture. The transformationof natural vegetation into pasture carries impacts many years after a land use change activity with a con-ventional 20 years of allocation to the new land use into the future which could include both pasture andcropland143. Consequently, impacts from land use change can be allocated to cattle and/or croplandbased on the year of deforestation of natural vegetation. Cattle production has the potential to deprivedownstream users of water through its water consumption activities which include drinking water, butalso changes in water availability as a result of land use change, as shown by the WSF and its potentialrelationship with land occupation (Figure 6.5).Both cropland and cattle production impacts on the water cycle point to a more favourable intensifica-tion of production by avoiding any further regional deforestation. The intensification options also includecropland irrigation which further increase competition of water resources in the basin. Additionally, ir-rigation may carry water quality impacts which were not considered here, but whose impact pathwayshave been described in LCA189, which can also impact water resources overall. Nevertheless, bothcropland and cattle production are linked by the trajectories of their land use transitions and thereforeneed to be considered together to avoid indirect land use change impacts. A minimized impact would berepresented by the combined cropland extensification into pasture and a more intensive use of pasturefor cattle (as opposed to further expansion of pasture into natural vegetation).7.2.3 Volumetric water footprint sustainability assessmentBoth green and blue water scarcity indices obtained through the VWFSA of the XBMT showed landconstraints following two deforestation scenarios, and also highlighted that water is unlikely to be limitingproduction unless there is rapid expansion of irrigation for soybean and continued cattle confinement(Chapter 5). As production is expected to increase in the basin, so is the combined water use for121cropland and cattle production which was assessed considering extensification (green water focused)and intensification scenarios, including the use of irrigation for cropland and confinement for cattle (bluewater focused) (see illustration in Figure 1.2, panels A and E). Production was within sustainable limitsin the 2000s, but decisions on how to increase production could affect future water scarcity. On the onehand, greenwater resources were classified as “threatened” in the XBMT, approaching and going beyondsustainable limits in 2030 and 2050 scenarios. On the other hand, the widespread use of irrigation couldlead to blue water scarcity at the end of the dry season when water consumption is expected to increase.While green and blue water resources in Mato Grosso were used within sustainable limits, other foot-prints need to be considered such as the carbon, land and nutrient footprints. These additional footprintshave been reported for soybean produced in 2010139 when deforestation for soybean was estimated at97 m2 ton-1 for the 2006-2010 period while total greenhouse gas emissions were 1.55 ton CO2-eq ton-1and the amount of nutrients remaining in Mato Grosso fields were 3.8-5.8 kg P ton-1 and 0.9 kg K ton-1.Deforestation for pasture has been carefully tracked in the XBMT as well as the overall greenhousegas emissions, including from land use change (Table B.1 and B.2, Appendix B). Intensification optionsfor cattle will reduce the land use change contribution to greenhouse gas emissions (as the largestcontribution) but may also entail additional environmental consequences related to methane emissions(herd and manure management) and water quality. These indicators, together, need to be consideredin decision-making with respect to future crop and cattle production and its sustainable production in theregion.The above effects will also rely on international partners importing both agricultural products andcattle from the region as Brazil’s total VWF of production represented 41% of all of Latin America andthe Caribbean region’s VWF of national production (1162 km3 y-1)160. Brazil is a net virtual waterexporter (54.8 km3 y-1) with exports concentrated on Europe (as 41% of gross exports) and Asia (32%of gross exports)55. Mato Grosso is a net exporter of virtual water mostly concentrated on crops andlivestock, with 10.2 km3 exported to China and 4.0 km3 exported to Europe through the soybean cropalone in 2010139. Reliance on virtual water imports can increase risks to the supply chain due to waterdependency from other countries. Between 2006 and 2015, 27% of the total virtual water flow intoEurope was due to soybean with “very high dependency” on imports and external water resources, butwith low vulnerability with respect to central-western Brazil69.1227.2.4 Policy decisionsA summary of the harmonized WF assessment results presented for both crops and cattle in this thesisis shown in Table 7.1 for multi-dimensional decision-making at both micro- and macro-levels, and theregional scales of the hydrologic cycle (Figure 2.2). Agricultural expansion has already contributed togreenhouse gas emissions from land use change in SAM, as well as impacts to biodiversity and ecosys-tem services related to climate regulation and the ability of soil to sustain biomass and filter water143.While several on-farm decisions have been proposed to maintain carbon stocks below ground followingland use change (e.g., through the Low Carbon Agriculture Credit program153), additional impacts tothe water cycle are expected with growing production. Despite greater groundwater recharge, reducedprecipitation from deforestation can further affect the Amazon biome as well as agricultural productionthereby increasing drought vulnerability of both natural ecosystems and agroecosystems184. As a re-sult, adaptation measures to maintain agricultural productivity may involve the use of irrigation whichcan return water vapour to the atmosphere, increase regional precipitation and deplete groundwaterresources141.At state and Federal government levels, responses proposed by the harmonized WF assessmentecho previously proposed initiatives, such as continuous enforcement of the Brazilian Federal ForestCode and land conservation through protective areas which have both evolved since 2000175. Giventhe very high dependency of Europe on soybean imports from the region, supply chain decisions have arole to play in reducing environmental impacts of production, and some have already been implementedthrough initiatives such as the Soybean Moratorium91 and incentives to intensify production. At thefield level, water decisions relate to the possible implementation of irrigation technology to raise farmincome through amore favourable double cropping systemwith greater yields attained through additionalinputs. Additional research on practices is required, however, to understand trade-offs in the costs andbenefits of such initiatives. Additional initiatives to provide value to protected forests on private landshould be considered such as payments for ecosystem services242, increases in production of highvalue products coming from tropical forests177, as well as the consideration of forests to recycle ET intoregional precipitation65, of which Brazil is a main source in the South American continent131.Uncertainties still remain following the harmonized WF assessment presented in this thesis, whichhas focused exclusively on water quantity. The economic and environmental costs of farming practicespromoting agricultural intensification need to be better understood and compared to the benefits relatedto yield improvements through fertilizer application for cropland and pasture. The possible increase inprecipitation as a result of greater water vapour transfers to the atmosphere throughwidespread irrigation123Table 7.1: Results from the harmonized water footprint (WF) assessment for SAM.Stage 2. VWF assessment 3. WF impact assessment 4. VWFSAResults For soybean, 780-1182m3 ton-1; for cattle,324-373 L (kg LW)-1 (bluewater), with 58-59%attributed to reservoirevaporation (blue water)and 21,400-52,100 L (kgLW)-1 for feed (greenwater)Cropland and cattle affectthe water cycle due to landoccupation from PRP(257-656 m3 ha-1, 2.5-4.5m3 (kg LW)-1), or GWRP(−1490 to −585 m3 ha-1,−5.1 to −2.9 m3 (kg LW)-1),in the Amazon; WSF were1298 m3-eq ha-1 (irrigatedcropland) and 0.20 m3-eq(kg LW)-1In 2014, WSB was 0.01(annual), and WSG was0.42-0.84 when considering80% of natural forest coverretained in the basin;irrigation expansion anddeforestation increase WSBto 0.65 in the dry season(2050), and WSG to 1.25 in arestrictive deforestationscenarioComparativeassertionGlobal benchmark 1553m3 ton-1 soybean of best10% global VWF158;large standard deviationfor reservoir evaporationsuggesting geographicdifferences in VWFPotential impacts weregreater in the Amazon thanin the Cerrado, and lowestwhen cropland replacedpasture, and for greaterpasture productivity forcattle; irrigation and cattledrinking carry a WSF withadditional potential scarcityfrom land occupationComparison of water usewith availability; WSB withinsustainable limits and WSGapproaching “threat”conditions (2014); basinapproached sustainablelimits due to irrigationexpansion and deforestation,even in restrictivedeforestation scenarioShort termactionsIncrease fertilizer use toincrease yields; on- andoff-farm watermanagement focused onreservoir evaporation andfeedPromote croplandintensification andextensification on currentpasture with furtherintensification of cattle oncurrent pastureIncrease efficiency ofagricultural water use;trade-off in green water usefor pasture to cropland withagricultural intensificationLong termactionsShorten crop cyclethrough technology;invest in irrigation forearly soybean plantingand triple crop; shortencattle cycle throughgenetics; seek reservoirevaporation reductioninitiativesReduce deforestation inAmazon and Cerrado;allocate indirect land usechange to cattleReduce deforestation inAmazon and Cerrado;highlight connectionsbetween up- anddownstream, and in- andout-of-basin green and bluewater use and availability(reservoir evaporation,precipitation recycling)Uncertainty Costs and benefits ofadditional inputscompared toimprovements (e.g.yield); costs of increasedpasture productivity withcurrent land and waterEffects of agriculturalintensification on waterquality; irrigation trade-offson precipitation recycling;water quality implicationsEffects of livestockintensification on waterquality; demand of soybeanfrom trade partners; waterscarcity due to infrastructureand climate; groundwateravailabilityActors in thewater cycleFarmers Farmers, state and Federalgovernments, supply chaininitiativesState and Federalgovernments, supply chaininitiatives124has, to date, not been considered in modeling studies141, and therefore represents an important gapin knowledge for future land and water planning for agricultural production with regional implications.Similarly, the effect of the extensive network of farm impoundments on the water cycle needs to bequantified to better understand the aggregated effects of percolation and evaporation, respectively onriver runoff and regional precipitation. Such information could better guide the recommendations tomaintain or reduce evaporation in the region. While agriculture was the sector consuming the largestvolume of water in the XBMT, it was not, however, through the consumption of blue water for irrigation,rather for animal production which is likely to increase in the future. The VWFSA showed that bluewater resources were used within limits on an annual basis, while official reports describe occasionalphysical scarcity as a result of infrastructure4 in the basin. Furthermore, little information is availableon groundwater extraction and availability, as well as future foreign demand for regional commodities,which may all increase pressure on water resources in the region.7.3 Perspectives of the water footprint and future workThe harmonized WF assessment was proposed as a step to integrate the different approaches andmethods currently used in the academic literature. Perhaps the most important contribution of the WFliterature has been the application of life cycle thinking to water resources which has shed greater lighton the role of indirect water uses (or supply chain uses) on production and consumption processes.This consideration remains an important challenge to the implementation of policy responses for con-sumers and producers, particularly considering today’s complex global supply chains44. This challengewas highlighted in this thesis with a primary focus on SAM as a production center with decisions mostlycentered on farmers, state and Federal governments, and only loose directives with respect to supplychain interventions based on other studies focused on consumption centers (e.g., Europe as describedin Ercin et al.69). So far in SAM, supply chain interventions have been more apparent in the manage-ment of land rather than water through the Soybean Moratorium and the Cattle Agreement175. Theseexperiences may serve as an example for future supply chain initiatives in relation to water management.The use of a single indicator such as the VWF has been criticized in the literature, and it gener-ally has been recognized that it should not form the sole basis of decision-making113. In addition,WF assessments also typically separate water resources into green, blue, and sometimes gray wateras a means to highlight different responses based on different hydrological cycles and water quality.This work has demonstrated that overall decision-making for agricultural production in SAM can change125based on the consideration of the individual water resource “colours” (water quality through gray waterwas not considered in this thesis), while water resources on their own, can only provide one element fordecision-making. While the separation of green and blue water resources in WF assessments has beencriticized192, this work has shown that the inclusion of green water and green water scarcity enabledthe joint consideration of land and water resources together, which has not been apparent in other WFassessments (e.g. the VWFSA from Hoekstra et al.118). Both green and blue water perspectives pro-vided additional information about potential changes in precipitation recycling with deforestation, therebygiving additional roles to trees for fixing carbon, regulating local climate, while returning water vapour tothe atmosphere66. Global freshwater is one of several Planetary Boundaries which also include climatechange, biodiversity, chemical pollution, among others223, and while each boundary is set individually,they should be observed together44. This limitation was apparent in this work not only because of theclose relationship between land and water management in SAM, but also due to other effects discussedwith respect to carbon and nutrients evaluated for the region in previous studies.This work showed great value in combining micro- and macro-level decisions with strengths andlimitations identified at each individual phase of the harmonized WF assessment (Table 7.2). The VWFassessment focused on micro, or field decisions based on the volume of freshwater used in agriculturalproduction and offered solutions that could potentially improve efficiency of water use in productionsystems. Such an initiative has been part of a more general objective geared towards resource useefficiency which, in agriculture, is also known as water productivity93. The actors on water resourcesare those that interact with the production system either directly (producers) or indirectly (suppliers).As mentioned above, a clear connection about action between producers and consumers needs to bemade as a way to highlight additional steps that can improve water efficiency. For instance, a Europeandairy farmer importing feed made from the soybean grown at the farm described in Chapter 3 couldfavour conditions that reduce indirect VWF provided precise information on water use at the field level.Such initiatives would need to be based on large databases combining trade information with water useon individual farms (e.g., as detailed by Godar et al.96). Finally, responses are linked to a functionalunit that is generally directly linked to a freshwater volume and might promote a decision that could becontradicted by others related to resources and emissions44.The aggregation of VWF into the VWFSA to determine blue and green water scarcity in the XBMTwassimilar to what has been called a Water Productivity Analysis93 with the main difference coming from therecognition of indirect water uses from consumption which can enter and exit a basin through virtual watertrade117. The VWFSA had the merit to compare water consumption to local boundaries defined from the126basin’s hydrological cycle, while linking both land and water management objectives together throughthe green VWF. This link, however, could be considered a special case in the context of land use changein SAM in which deforestation scenarios were tested and related to green water availability, based on thefraction of natural vegetation cover mandated by Brazilian Federal law201. In the context of the WF andthe VWFSA, green water availability has been defined as the “part of the green water flow available forbiomass production for human purposes”233 and relies on estimates of the amount of land reserved forNature (also known as the “environmental green water requirement”233). Additional difficulties arose withthe differences between “bottom-up” and “top-down” approaches, which have already been highlightedin the literature42. These differences, particularly in the XBMT which has limited available data, canprove to be an important caveat in carrying out such an assessment. Making the clear link betweenproducers (Mato Grosso) and consumers (e.g., Europe) complicates the potential policy decisions apartfrom the evaluation of the dependency of a consumer on foreign resources69. While some supply chaininitiatives have shown to be successful in Mato Grosso in the case of deforestation175, a similar case forwater resources alone might be more difficult to make and therefore should likely be seen as a land useissue with multiple benefits (e.g., CO2 emissions, biodiversity, ecosystem services, etc.)66. As such, itis imperative to associate the WF with other footprints in the VWFSA. Similarly, the lack of informationand inclusion of groundwater in such an assessment constitutes a major limitation to the decisions aslimits to this resource are still difficult to estimate44.Finally, the WFIA, and more generally water use in LCA, is still in its infancy with a number of impactpathways that remain to be developed, particularly in the end-point impact categories affecting ecosys-tems and natural resources. In itself, LCA allows for a comprehensive assessment that can include awide variety of impacts linked to the Planetary Boundaries (climate change, impacts of chemical pollu-tion, etc.)223. However, decisions that may result from LCA are only as good as the comparisons thatare tested in the study. For instance, our comparison of production systems in Amazon and Cerradobiomes might mislead decision-makers to favour production in the Cerrado over the Amazon biome.To this effect, for a study to be comprehensive, it should be able to analyze all available options in theproduction system, including potential effects of indirect land use change within Brazil, or across interna-tional borders12. Similar to the VWF assessment, LCA is influenced by “the efficiency mindset”87 which,following the findings of this research, consistently promotes the use of fewer resources as a means toreduce environmental impacts (i.e., reduce the LCI values). Furthermore, these impacts are linked toa specific functional unit which may not only influence decisions, but whose connections to sustainablelimits are difficult to express in one single functional unit. The application of LCA goes beyond products127Table 7.2: Strengths and limitations identified in the individual steps of the harmonized water foot-print (WF) assessment: the volumetric water footprint (VWF) assessment, the WF impact assessment(WFIA), and the VWF sustainability assessment (VWFSA) as shown in Table 2.2.Step Strengths LimitationsVWF assessment(Chapters 3, 4)Micro-level decisions are identified;decisions are made at the producerlevel with potential influence fromconsumers (unclear at this time)Resources other than waterinfluence decision-making; findingsare based on comparative assertionsmostly focused on water volumesVWFSA(Chapter 5)Assessment considers localboundaries with local data; land andwater linked through green VWF andgreen water availabilityAssessment of green water scarcitydepends on interpretation ofenvironmental green waterrequirements; top-down andbottom-up approaches carry differentwater scarcity results; resourcesother than water influencedecision-making; lack ofgroundwater boundary due to limiteddata44WFIA(Chapter 6)Assessments are typicallycomprehensive (water quantity andquality); impact assessment fullyintegrated into the productionsystem; decisions are made at theproducer level with potentialinfluence from consumers (unclearat this time)Methodological advances are stillneeded to integrate changes in waterflows from land use, and theinteraction between land use andwater scarcity; decisions may beaffected by the defined functionalunit which is rarely comprehensive;findings are based on comparativeassertions which carry value choicesand may include organizations or lifestyles105. There are still limited studies that have applied LCA to aterritory, and this discrepancy is likely due to the translation of a multifunctional system into one singlefunctional unit147.WF assessments can greatly benefit from the combined analysis proposed by the harmonized WFassessment as each individual phase of the assessment can potentially be limited in scope, and hasspecific strengths and limitations (Table 7.2). Aside from increasing the number of pilot studies apply-ing individual components of the harmonized WF assessment, greater emphasis should be placed onlinking producers and consumers to make apparent decisions that affect the global supply chain in re-lation to water resources as well as for other important resources and emissions linked to the PlanetaryBoundaries223.128References[1] Allan, J., Virtual water: A strategic resource global solutions to regional deficits, Ground Water,34(4), 545–546, doi:10.1111/j.1745-6584.1998.tb02825.x, 1998.[2] Allen, R. G., D. Pereira, L. S. Raes, and M. Smith, Crop Evapotranspiration: Guidelines for Com-puting CropWater Requirements, Food and Agriculture Organization of the United Nations, Rome,1998.[3] Allen, R. G., L. S. Pereira, T. A. Howell, and M. E. Jensen, Evapotranspiration information report-ing: I. Factors governing measurement accuracy, Agricultural Water Management, 98(6), 899–920, doi:10.1016/j.agwat.2010.12.015, 2011.[4] ANA, Plano estratégico de recursos hídricos dos afluentes da margem direita do Rio Amazonas,Agência Nacional de Águas, Brasilia, DF, Brazil, 2013.[5] ANA, Hidroweb,, 2017.[6] ANA, and Embrapa/CNPMS, Levantamento da agricultura irrigada por pivôs centrais no Brasil -ano 2014, Brasilia, DF, Brazil, 2016.[7] Antonino, A. C. D., C. Hammecker, S. M. L. G. Montenegro, A. M. Netto, R. Angulo-Jaramillo,and C. A. B. O. Lira, Subirrigation of land bordering small reservoirs in the semi-arid region inthe Northeast of Brazil: Monitoring and water balance, Agricultural Water Management, 73(2),131–147, doi:10.1016/j.agwat.2004.10.001, 2005.[8] António Sumila, T. C., G. F. Pires, V. C. Fontes, and M. H. Costa, Sources of Water Vapor toEconomically Relevant Regions in Amazonia and the Effect of Deforestation, Journal of Hydrom-eteorology, 18(6), 1643–1655, doi:10.1175/jhm-d-16-0133.1, 2017.[9] Anualpec,, 2015.129[10] Arima, E. Y., P. Richards, R. Walker, and M. M. Caldas, Statistical confirmation of indirect landuse change in the Brazilian Amazon, Environmental Research Letters, 6(2), 024,010, doi:10.1088/1748-9326/6/2/024010, 2011.[11] Arvor, D., V. Dubreuil, J. Ronchail, M. Simões, and B. M. Funatsu, Spatial patterns of rainfallregimes related to levels of double cropping agriculture systems in Mato Grosso (Brazil), Interna-tional Journal of Climatology, 34(8), 2622–2633, doi:10.1002/joc.3863, 2014.[12] Arvor, D., I. Tritsch, C. Barcellos, N. Jegou, and V. Dubreuil, Land use sustainability on the South-Eastern Amazon agricultural frontier: Recent progress and the challenges ahead, Applied Geog-raphy, 80, 86–97, doi:10.1016/j.apgeog.2017.02.003, 2017.[13] Auguie, B., gridExtra: Miscellaneous Functions for "Grid" Graphics. R package version 2.2.1.,2016.[14] Azevedo, A. A., R. Rajão, M. A. Costa, M. C. C. Stabile, M. N. Macedo, T. N. P. dos Reis, A. Alen-car, B. S. Soares-Filho, and R. Pacheco, Limits of Brazil’s Forest Code as a means to end illegaldeforestation, Proceedings of the National Academy of Sciences of the U.S.A., 114(29), 7653–7658, doi:10.1073/pnas.1604768114, 2017.[15] Bagley, J. E., A. R. Desai, K. J. Harding, P. K. Snyder, and J. A. Foley, Drought and Deforestation:Has Land Cover Change Influenced Recent Precipitation Extremes in the Amazon?, Journal ofClimate, 27(1), 345–361, doi:10.1175/jcli-d-12-00369.1, 2014.[16] Baillie, C., Assessment of Evaporation Losses and Evaporation Mitigation Technologies for OnFarm Water Storages across Australia, (05), 1–52, 2008.[17] Bakker, K., and C. Morinville, The governance dimensions of water security: a review, Philo-sophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences,371(2002), 20130,116–20130,116, doi:10.1098/rsta.2013.0116, 2013.[18] Barona, E., N. Ramankutty, G. Hyman, and O. T. Coomes, The role of pasture and soybeanin deforestation of the Brazilian Amazon, Environmental Research Letters, 5(2), 024,002, doi:10.1088/1748-9326/5/2/024002, 2010.[19] Barr, A. G., G. van der Kamp, T. A. Black, J. H. McCaughey, and Z. Nesic, Energy balance closureat the BERMS flux towers in relation to the water balance of theWhite Gull Creek watershed 1999-1302009, Agricultural and Forest Meteorology, 153, 3–13, doi:10.1016/j.agrformet.2011.05.017,2012.[20] Bayart, J.-B., C. Bulle, L. Deschênes, M. Margni, S. Pfister, F. Vince, and A. Koehler, A frameworkfor assessing off-stream freshwater use in LCA, The International Journal of Life Cycle Assess-ment, 15(5), 439–453, doi:10.1007/s11367-010-0172-7, 2010.[21] Beck, T., U. Bos, B. Wittstock, M. Baitz, M. Fischer, and K. Seldbauer, Land use indicator valuecalculation in life cycle assessment: Method report, vol. 9, 67 pp., 2011.[22] Berger, M., and M. Finkbeiner, Water Footprinting: How to Address Water Use in Life Cycle As-sessment?, Sustainability, 2(4), 919–944, doi:10.3390/su2040919, 2010.[23] Berger, M., R. van der Ent, S. Eisner, V. Bach, and M. Finkbeiner, Water Accounting and Vul-nerability Evaluation (WAVE): Considering Atmospheric Evaporation Recycling and the Risk ofFreshwater Depletion in Water Footprinting, Environmental Science & Technology, 48(8), 4521–4528, doi:10.1021/es404994t, 2014.[24] Bivand, R., and N. Lewin-Koh, maptools: Tools for Reading and Handling Spatial Objects, Rpackage version 0.9-2, 2017.[25] Bivand, R., T. Keitt, and B. Rowlingson, rgdal: Bindings for the Geospatial Data Abstraction Li-brary, R package version 1.2-7, 2017.[26] Bivand, R. S., E. Pebesma, and V. Gomez-Rubio, Applied spatial data analysis with R, 2nd ed.,Springer, NY, 2013.[27] Bivand, R. and Keitt, T. and Rowlingson, B., rgdal: Bindings for the Geospatial Data AbstractionLibrary. R package version 1.2-7., 2017.[28] Boulay, A.-M., C. Bulle, J.-B. Bayart, L. Deschênes, and M. Margni, Regional Characterizationof Freshwater Use in LCA: Modeling Direct Impacts on Human Health, Environmental Science &Technology, 45(20), 8948–8957, doi:10.1021/es1030883, 2011.[29] Boulay, A.-M., A. Y. Hoekstra, and S. Vionnet, Complementarities of Water-Focused Life CycleAssessment and Water Footprint Assessment, Environmental Science & Technology, 47(21),11,926–11,927, doi:10.1021/es403928f, 2013.131[30] Boulay, A.-M., et al., Consensus building on the development of a stress-based indicator for LCA-based impact assessment of water consumption: outcome of the expert workshops, The Interna-tional Journal of Life Cycle Assessment, 20(5), 577–583, doi:10.1007/s11367-015-0869-8, 2015.[31] Boulay, A.-M., et al., The WULCA consensus characterization model for water scarcity footprints:assessing impacts of water consumption based on available water remaining (AWARE), The In-ternational Journal of Life Cycle Assessment, 23(2), 368–378, doi:10.1007/s11367-017-1333-8,2018.[32] Brando, P.M., S. J. Goetz, A. Baccini, D. C. Nepstad, P. S. A. Beck, andM. C. Christman, Seasonaland interannual variability of climate and vegetation indices across the Amazon, Proceedingsof the National Academy of Sciences of the U.S.A., 107(33), 14,685–14,690, doi:10.1073/pnas.0908741107, 2010.[33] Brasil, Lei Nº 12.787 de 11 de janeiro de 2013,, 2013.[34] Callow, J., and K. Smettem, The effect of farm dams and constructed banks on hydrologic con-nectivity and runoff estimation in agricultural landscapes, Environmental Modelling & Software,24(8), 959–968, doi:10.1016/j.envsoft.2009.02.003, 2009.[35] Campbell, G., and J. Norman, An Introduction to Environmental Biophysics, 2nd ed., 286 pp.,Springer, New York, NY, USA, 1998.[36] Cardoso, A. S., A. Berndt, A. Leytem, B. J. Alves, I. d. N. de Carvalho, L. H. de Barros Soares,S. Urquiaga, and R. M. Boddey, Impact of the intensification of beef production in Brazil on green-house gas emissions and land use, Agricultural Systems, 143, 86–96, doi:10.1016/j.agsy.2015.12.007, 2016.[37] Carslaw, D. C., and K. Ropkins, openair — an R package for air quality data analysis, Environ-mental Modelling & Software, 27-28, 52–61, 2012.[38] Castello, L., and M. N. Macedo, Large-scale degradation of Amazonian freshwater ecosystems,Global Change Biology, 22(3), 990–1007, doi:10.1111/gcb.13173, 2016.[39] Cederberg, C., U. M. Persson, K. Neovius, S. Molander, and R. Clift, Including Carbon EmissionsfromDeforestation in the Carbon Footprint of Brazilian Beef, Environmental Science & Technology,45(5), 1773–1779, doi:10.1021/es103240z, 2011.132[40] Cerri, C. C., C. S. Moreira, P. A. Alves, G. S. Raucci, B. D. A. Castigioni, F. F. C. Mello, D. G. P.Cerri, and C. E. P. Cerri, Assessing the carbon footprint of beef cattle in Brazil: A case studywith 22 farms in the State of Mato Grosso, Journal of Cleaner Production, 112(4), 2593–2600,doi:10.1016/j.jclepro.2015.10.072, 2016.[41] Chaplin-Kramer, R., et al., Spatial patterns of agricultural expansion determine impacts on bio-diversity and carbon storage, Proceedings of the National Academy of Sciences of the U.S.A.,112(24), 7402–7407, doi:10.1073/pnas.1406485112, 2015.[42] Chenoweth, J., M. Hadjikakou, and C. Zoumides, Quantifying the human impact on water re-sources: a critical review of the water footprint concept, Hydrology and Earth System Sciences,18(6), 2325–2342, doi:10.5194/hess-18-2325-2014, 2014.[43] Christoffersen, B. O., et al., Mechanisms of water supply and vegetation demand govern the sea-sonality and magnitude of evapotranspiration in Amazonia and Cerrado, Agricultural and ForestMeteorology, 191, 33–50, doi:10.1016/j.agrformet.2014.02.008, 2014.[44] Clift, R., et al., The Challenges of Applying Planetary Boundaries as a Basis for StrategicDecision-Making in Companies with Global Supply Chains, Sustainability, 9(2), 279, doi:10.3390/su9020279, 2017.[45] Coe, M. T., M. H. Costa, and B. S. Soares-Filho, The influence of historical and potential futuredeforestation on the stream flow of the Amazon River - Land surface processes and atmosphericfeedbacks, Journal of Hydrology, 369(1-2), 165–174, doi:10.1016/j.jhydrol.2009.02.043, 2009.[46] Coe, M. T., E. M. Latrubesse, M. E. Ferreira, and M. L. Amsler, The effects of deforestation andclimate variability on the streamflow of the Araguaia River, Brazil, Biogeochemistry, 105(1-3),119–131, doi:10.1007/s10533-011-9582-2, 2011.[47] Coe, M. T., P. M. Brando, L. A. Deegan, M. N. Macedo, C. Neill, and D. V. Silvério, The Forestsof the Amazon and Cerrado Moderate Regional Climate and Are the Key to the Future, TropicalConservation Science, 10(September), 1–6, doi:10.1177/1940082917720671, 2017.[48] Cohn, A. S., A. Mosnier, P. Havlík, H. Valin, M. Herrero, E. Schmid, M. O’Hare, andM. Obersteiner,Cattle ranching intensification in Brazil can reduce global greenhouse gas emissions by sparingland from deforestation., Proceedings of the National Academy of Sciences of the U.S.A., 111(20),7236–7241, doi:10.1073/pnas.1307163111, 2014.133[49] Cohn, A. S., J. Gil, T. Berger, H. Pellegrina, and C. Toledo, Patterns and processes of pastureto crop conversion in Brazil: Evidence from Mato Grosso state, Land Use Policy, 55, 108–120,doi:10.1016/j.landusepol.2016.03.005, 2016.[50] Cohn, A. S., L. K. VanWey, S. A. Spera, and J. F. Mustard, Cropping frequency and area responseto climate variability can exceed yield response, Nature Climate Change, 6(6), 601–604, doi:10.1038/nclimate2934, 2016.[51] Costa, M. H., and G. F. Pires, Effects of Amazon and Central Brazil deforestation scenarios on theduration of the dry season in the arc of deforestation, International Journal of Climatology, 30(13),1970–1979, doi:10.1002/joc.2048, 2010.[52] Costa, M. H., A. Botta, and J. A. Cardille, Effects of large-scale changes in land cover on thedischarge of the Tocantins River, Southeastern Amazonia, Journal of Hydrology, 283(1-4), 206–217, doi:10.1016/s0022-1694(03)00267-1, 2003.[53] da Rocha, H. R., M. L. Goulden, S. D. Miller, M. C. Menton, L. D. V. O. Pinto, H. C. De Freitas,and A. M. E Silva Figueira, Seasonality of water and heat fluxes over a tropical forest in easternAmazonia, Ecological Applications, 14, 22–32, doi:10.1890/02-6001, 2004.[54] da Rocha, H. R., et al., Patterns of water and heat flux across a biome gradient from tropical forestto savanna in Brazil, Journal of Geophysical Research, 114, G00B12, doi:10.1029/2007jg000640,2009.[55] da Silva, V., et al., Water Footprint and Virtual Water Trade of Brazil, Water, 8(11), 517, doi:10.3390/w8110517, 2016.[56] Dalin, C., M. Konar, N. Hanasaki, A. Rinaldo, and I. Rodriguez-Iturbe, Evolution of the global virtualwater trade network, Proceedings of the National Academy of Sciences of the U.S.A., 109(16),5989–5994, doi:10.1073/pnas.1203176109, 2012.[57] Dalin, C., N. Hanasaki, H. Qiu, D. L. Mauzerall, and I. Rodriguez-Iturbe, Water resources transfersthrough Chinese interprovincial and foreign food trade, Proceedings of the National Academy ofSciences of the U.S.A., 111(27), 9774–9779, doi:10.1073/pnas.1404749111, 2014.[58] Dalin, C., Y. Wada, T. Kastner, and M. J. Puma, Groundwater depletion embedded in internationalfood trade, Nature, 543(7647), 700–704, doi:10.1038/nature21403, 2017.134[59] Davidson, E. A., et al., The Amazon basin in transition, Nature, 481(7381), 321–328, doi:10.1038/nature10717, 2012.[60] de Almeida, C. A., A. C. Coutinho, J. C. D. M. Esquerdo, M. Adami, A. Venturieri, C. G. Diniz,N. Dessay, L. Durieux, and A. R. Gomes, High spatial resolution land use and land cover mappingof the Brazilian Legal Amazon in 2008 using Landsat-5/TM and MODIS data, Acta Amazonica,46(3), 291–302, doi:10.1590/1809-4392201505504, 2016.[61] Defries, R. S., D. C. Morton, G. R. Van Der Werf, L. Giglio, G. J. Collatz, J. T. Randerson,R. A. Houghton, P. K. Kasibhatla, and Y. Shimabukuro, Fire-related carbon emissions fromland use transitions in southern Amazonia, Geophysical Research Letters, 35(22), 1–5, doi:10.1029/2008gl035689, 2008.[62] Dias, L. C. P., M. N. Macedo, M. H. Costa, M. T. Coe, and C. Neill, Effects of land cover change onevapotranspiration and streamflow of small catchments in the Upper Xingu River Basin, CentralBrazil, Journal of Hydrology: Regional Studies, 4(PB), 108–122, doi:10.1016/j.ejrh.2015.05.010,2015.[63] Dias, L. C. P., F. M. Pimenta, A. B. Santos, M. H. Costa, and R. J. Ladle, Patterns of land use,extensification, and intensification of Brazilian agriculture, Global Change Biology, pp. 1–16, doi:10.1111/gcb.13314, 2016.[64] Dick, M., M. Abreu da Silva, and H. Dewes, Life cycle assessment of beef cattle production intwo typical grassland systems of southern Brazil, Journal of Cleaner Production, 96, 426–434,doi:10.1016/j.jclepro.2014.01.080, 2015.[65] Ellison, D., M. N. Futter, and K. Bishop, On the forest cover-water yield debate: From demand-to supply-side thinking, Global Change Biology, 18(3), 806–820, doi:10.1111/j.1365-2486.2011.02589.x, 2012.[66] Ellison, D., et al., Trees, forests and water: Cool insights for a hot world, Global EnvironmentalChange, 43, 51–61, doi:10.1016/j.gloenvcha.2017.01.002, 2017.[67] Embrapa-CNPS, Manual de métodos de análise de solo, 2nd ed., 212 pp., Embrapa-CNPS, Riode Janeiro, RJ, 1997.[68] Ercin, A., M. Mekonnen, and A. Hoekstra, Sustainability of national consumption from a water135resources perspective: The case study for France, Ecological Economics, 88, 133–147, doi:10.1016/j.ecolecon.2013.01.015, 2013.[69] Ercin, A., D. Chico, and A. K. Chapagain, Dependencies of Europe’ s economy on other parts ofthe world in terms of water resources, Techreport D12.1, Water Footprint Network, 2016.[70] FABOV, Diagnóstico da cadeia produtiva agroindustrial da bovinocultura de corte do estado deMato Grosso, Fundo de Apoio à Bovinocultura de Corte, Cuiabá, MT, Brazil, 2007.[71] Falkenmark, M., and J. Rockström, The New Blue and Green Water Paradigm: Breaking NewGround for Water Resources Planning and Management, Journal of Water Resources Planningand Management, 132(3), 129–132, doi:10.1061/(asce)0733-9496(2006)132:3(129), 2006.[72] FAO, In Brief: The state of food and agriculture, 2013: Climate Change, Agriculture and FoodSecurity, Food and Agriculture Organization, Rome, Italy, 2013.[73] FAO, Environmental performance of animal feeds supply chains, Livestock Environmental As-sessment and Performance Partnership, Food and Agriculture Organization, Rome, Italy, 2015.[74] FAO, Guidelines for water use assessment of livestock production systems and supply chains,Livestock Environmental Assessment and Performance Partnership, Food and Agriculture Orga-nization, Rome, Italy, draft ed., 2018.[75] FAOSTAT,, 2017.[76] FEALQ, Análise territorial para o desenvolvimento da agricultural irrigada no Brasil, Fundaçãode Estudos Agrários Luiz de Queiroz, Piracicaba, SP, Brazil, 2014.[77] Fisher, J. B., et al., The land-atmosphere water flux in the tropics, Global Change Biology, 15(11),2694–2714, doi:10.1111/j.1365-2486.2008.01813.x, 2009.[78] Foken, T., M. Gockede, M. Mauder, L. Mahrt, B. Amiro, and J. Munger, Post-field quality control,in Handbook of Micrometeorology: A Guide for Surface Flux Measurements, 81–108 pp., KluwerAcademic, Dordrecht, Germany, 2004.[79] Foley, J. A., I. C. Prentice, N. Ramankutty, S. Levis, D. Pollard, S. Sitch, and A. Haxeltine, Anintegrated biosphere model of land surface processes, terrestrial carbon balance, and vegetationdynamics, doi:10.1029/96gb02692, 1996.136[80] Foley, J.A., C.J. Kucharik, D. P., Integrated Biosphere Simulator Model (IBIS), Version 2.5., ORNLDAAC, Oak Ridge, TN, USA, doi:10.3334/ORNLDAAC/808, 2005.[81] Franssen, H. J. H., R. Stöckli, I. Lehner, E. Rotenberg, and S. I. Seneviratne, Energy balanceclosure of eddy-covariance data: Amultisite analysis for European FLUXNET stations, Agriculturaland Forest Meteorology, 150(12), 1553–1567, doi:10.1016/j.agrformet.2010.08.005, 2010.[82] Funk, C., et al., The climate hazards infrared precipitation with stations–a new environmentalrecord for monitoring extremes, Scientific Data, 2, 150,066, doi:10.1038/sdata.2015.66, 2015.[83] Galford, G. L., J. Melillo, J. F. Mustard, C. E. P. Cerri, and C. C. Cerri, The Amazon Frontier of Land-Use Change: Croplands and Consequences for Greenhouse Gas Emissions, Earth Interactions,14(15), 1–24, doi:10.1175/2010ei327.1, 2010.[84] Galford, G. L., J. M. Melillo, D. W. Kicklighter, J. F. Mustard, T. W. Cronin, C. E. P. Cerri, andC. C. Cerri, Historical carbon emissions and uptake from the agricultural frontier of the BrazilianAmazon, Ecological Applications, 21(3), 750–763, doi:10.1890/09-1957.1, 2011.[85] Galford, G. L., B. Soares-Filho, and C. E. P. Cerri, Prospects for land-use sustainability on theagricultural frontier of the Brazilian Amazon., Philosophical transactions of the Royal Society ofLondon. Series B, Biological sciences, 368(1619), 20120,171, doi:10.1098/rstb.2012.0171, 2013.[86] Garcia, E., F. Ramos Filho, G. Mallmann, and F. Fonseca, Costs, Benefits and Challenges ofSustainable Livestock Intensification in a Major Deforestation Frontier in the Brazilian Amazon,Sustainability, 9(1), 158, doi:10.3390/su9010158, 2017.[87] Garnett, T., Three perspectives on sustainable food security: Efficiency, demand restraint, foodsystem transformation. What role for life cycle assessment?, Journal of Cleaner Production, 73,10–18, doi:10.1016/j.jclepro.2013.07.045, 2014.[88] Gerbens-Leenes, P. W., M. Mekonnen, and A. Y. Hoekstra, The water footprint of poultry, porkand beef: A comparative study in different countries and production systems, Water Resourcesand Industry, 1-2, 25–36, doi:10.1016/j.wri.2013.03.001, 2013.[89] Gerbens-Leenes, W., A. Y. Hoekstra, and T. H. van der Meer, The water footprint of bioenergy,Proceedings of the National Academy of Sciences of the U.S.A., 106(25), 10,219–10,223, doi:10.1073/pnas.0812619106, 2009.137[90] Gerten, D., H. Hoff, J. Rockström, J. Jägermeyr, M. Kummu, and A. V. Pastor, Towards a revisedplanetary boundary for consumptive freshwater use: role of environmental flow requirements,Cur-rent Opinion in Environmental Sustainability, 5(6), 551–558, doi:10.1016/j.cosust.2013.11.001,2013.[91] Gibbs, B. H. K., et al., Brazil’s Soy Moratorium, Science, 347(6220), 377–378, doi:10.1126/science.aaa0181, 2015.[92] Gibbs, H. K., J. Munger, J. L'Roe, P. Barreto, R. Pereira, M. Christie, T. Amaral, and N. F. Walker,Did ranchers and slaughterhouses respond to zero-deforestation agreements in the BrazilianAmazon?, Conservation Letters, 9(1), 32–42, doi:10.1111/conl.12175, 2015.[93] Giordano, M., H. Turral, S. M. Scheierling, D. O. Tréguer, and P. G. McCornick, Beyond More"Crop per Drop": evolving thinking on agricultural water productivity., Tech. rep., InternationalWater Management Institute, Colombo, Sri Lanka, doi:10.5337/2017.202., 2017.[94] Gleeson, T., K. M. Befus, S. Jasechko, E. Luijendijk, and M. B. Cardenas, The global volume anddistribution of modern groundwater, Nature Geoscience, 9(2), 161–167, doi:10.1038/ngeo2590,2015.[95] Global Water Partnership, Integrated Water Resources Management, TAC Background papersNo 4, Tech. rep., Global Water Partnership, Stockholm, Sweden, 2000.[96] Godar, J., U. M. Persson, E. J. Tizado, and P. Meyfroidt, Towards more accurate and policy rel-evant footprint analyses: Tracing fine-scale socio-environmental impacts of production to con-sumption, Ecological Economics, 112, 25–35, doi:10.1016/j.ecolecon.2015.02.003, 2015.[97] Gollnow, F., and T. Lakes, Policy change, land use, and agriculture: The case of soy productionand cattle ranching in Brazil, 2001-2012, Applied Geography, 55, 203–211, doi:10.1016/j.apgeog.2014.09.003, 2014.[98] Grace, J., Y. Malhi, J. Lloyd, J. McIntyre, A. C. Miranda, P. Meir, and H. S. Miranda, The useof eddy covariance to infer the net carbon dioxide uptake of Brazilian rain forest, Global ChangeBiology, 2(3), 209–217, doi:10.1111/j.1365-2486.1996.tb00073.x, 1996.[99] Graesser, J., and N. Ramankutty, Detection of cropland field parcels from Landsat imagery, Re-mote Sensing of Environment, 201(April 2016), 165–180, doi:10.1016/j.rse.2017.08.027, 2017.138[100] Graesser, J., T. M. Aide, H. R. Grau, and N. Ramankutty, Cropland/pastureland dynamics andthe slowdown of deforestation in Latin America, Environmental Research Letters, 10(3), 034,017,doi:10.1088/1748-9326/10/3/034017, 2015.[101] Harrell Jr, F. E., and et al., Hmisc: Harrell Miscellaneous, R package version 4.0-3, 2017.[102] Harris, P. P., C. Huntingford, P. M. Cox, J. H. C. Gash, and Y. Malhi, Effect of soil moistureon canopy conductance of Amazonian rainforest, Agricultural and Forest Meteorology, 122(3-4),215–227, doi:10.1016/j.agrformet.2003.09.006, 2004.[103] Hartmann, D. L., et al., Observations: Atmosphere and Surface., in Climate Change 2013: ThePhysical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the In-tergovernmental Panel on Climate Change, edited by T. Stocker, D. Qin, G.-K. Plattner, M. Tignor,S. K. Allen, J. Boschung, A. Nauels, Y. Xia, B. V., and P. M. Midgley, Cambridge University Press,Cambridge, UK, 2013.[104] Hayhoe, S., C. Neill, S. Porder, R. McHorney, P. Lefebvre, M. Coe, H. Elsenbeer, and A. Krusche,Conversion to soy on the Amazonian agricultural frontier increases streamflow without affectingstormflow dynamics, Global Change Biology, 17(5), 1821–1833, doi:10.1111/j.1365-2486.2010.02392.x, 2011.[105] Hellweg, S., and L. Milà i Canals, Emerging approaches, challenges and opportunities in life cycleassessment, Science, 344(6188), 1109–1113, doi:10.1126/science.1248361, 2014.[106] Hijmans, R. J., raster: Geographic Data Analysis and Modeling, R package version 2.5-8, 2016.[107] Hilker, T., A. I. Lyapustin, C. J. Tucker, F. G. Hall, R. B. Myneni, Y. Wang, J. Bi, Y. Mendes deMoura, and P. J. Sellers, Vegetation dynamics and rainfall sensitivity of the Amazon, Proceedingsof the National Academy of Sciences of the U.S.A., 111(45), 16,041–16,046, doi:10.1073/pnas.1404870111, 2014.[108] Hoekstra, A., The water footprint of industry, in Assessing and measuring environmental impactand sustainability, edited by J. J. Klemes, pp. 221–254, Butterworth-Heinemann (Elsevier), Ox-ford, UK, 2015.[109] Hoekstra, A., and P. Hung, Globalisation of water resources: international virtual water flows inrelation to crop trade,Global Environmental Change, 15(1), 45–56, doi:10.1016/j.gloenvcha.2004.06.004, 2005.139[110] Hoekstra, A. Y., The Global Dimension of Water Governance: Why the River Basin Approach IsNo Longer Sufficient and Why Cooperative Action at Global Level Is Needed, Water, 3(1), 21–46,doi:10.3390/w3010021, 2010.[111] Hoekstra, A. Y., The sustainability of a single activity, production process or product, EcologicalIndicators, 57, 82–84, doi:10.1016/j.ecolind.2015.04.022, 2015.[112] Hoekstra, A. Y., A critique on the water-scarcity weighted water footprint in LCA, Ecological Indi-cators, 66, 564–573, doi:10.1016/j.ecolind.2016.02.026, 2016.[113] Hoekstra, A. Y., Water Footprint Assessment: Evolvement of a New Research Field, Water Re-sources Management, 31(10), 3061–3081, doi:10.1007/s11269-017-1618-5, 2017.[114] Hoekstra, A. Y., andM.M.Mekonnen, The water footprint of humanity, Proceedings of the NationalAcademy of Sciences of the U.S.A., 109(9), 3232–3237, doi:10.1073/pnas.1109936109, 2012.[115] Hoekstra, A. Y., and M. M. Mekonnen, Reply to Ridoutt and Huang: From water footprint as-sessment to policy, Proceedings of the National Academy of Sciences of the U.S.A., 109(22),E1425–E1425, doi:10.1073/pnas.1205186109, 2012.[116] Hoekstra, A. Y., and T. O. Wiedmann, Humanity’s unsustainable environmental footprint, Science,344(6188), 1114–1117, doi:10.1126/science.1248365, 2014.[117] Hoekstra, A. Y., A. K. Chapagain, M. M. Aldaya, and M. M. Mekonnen, The Water Footprint As-sessment Manual, 228 pp., Earthscan, London, UK, doi:978-1-84971-279-8, 2011.[118] Hoekstra, A. Y., M. M. Mekonnen, A. K. Chapagain, R. E. Mathews, and B. D. Richter, GlobalMonthly Water Scarcity: Blue Water Footprints versus Blue Water Availability, PLoS ONE, 7(2),e32,688, doi:10.1371/journal.pone.0032688, 2012.[119] Hoekstra, A.Y., Hung, P., Virtual water trade: A quantification of virtual water flows between na-tions in relation to international crop trade, Value of Water Research Report Series No.11., Tech.rep., UNESCO-IHE, Delft, the Netherlands, 2002.[120] Hutyra, L. R., J. W. Munger, C. A. Nobre, S. R. Saleska, S. A. Vieira, and S. C. Wofsy, Climaticvariability and vegetation vulnerability in Amazonia, Geophysical Research Letters, 32(24), 1–4,doi:10.1029/2005gl024981, 2005.[121] IBGE, Banco de dados agregados,, 2017.140[122] IMEA, Diagnóstico da Piscicultura em Mato Grosso, Instituto Mato-grossense de EconomiaAgropecuária, Cuiabá, MT, Brazil, 2014.[123] IMEA, Boletim da rentabilidade agrícola no Mato Grosso, Instituto Mato-grossense de EconomiaAgropecuária, Cuiabá, MT, Brazil, 2015.[124] INMET,, 2016.[125] IPCC, 2006 IPCC Guidelines for National Greenhouse Gas Inventories, Prepared by the NationalGreenhouse Gas Inventories Programme, 10.1–10.87 pp., IGES, Japan, 2006.[126] ISO, ISO 14044:2006, environmental management - life cycle assessment - requirements andguidelines, Geneva, Switzerland, 2006.[127] ISO, ISO 14046:2014, Water footprint: Principles, requirements and guidelines, Geneva, Switzer-land, 2015.[128] Karlberg, L., Rockström, J., Falkenmark, M., Water Resource Implications of Upgrading RainfedAgriculture - Focus on Green and Blue Water Trade-offs, in Rainfed Agriculture: Unlocking thePotential, edited by T. Wanu, S.P., Rockstrom, J., Oweis, chap. 3, pp. 44–53, CAB International,Oxford, UK, 2009.[129] Karstensen, J., G. P. Peters, and R. M. Andrew, Attribution of CO2 emissions from Brazilian de-forestation to consumers between 1990 and 2010, Environmental Research Letters, 8, 024,005,doi:10.1088/1748-9326/8/2/024005, 2013.[130] Keller, M., et al., Ecological Research in the Large-Scale Biosphere-Atmosphere Experiment inAmazonia: Early Results, Ecological Applications, 14(sp4), 3–16, doi:10.1890/03-6003, 2004.[131] Keys, P., L. Wang-Erlandsson, L. Gordon, V. Galaz, and J. Ebbesson, Approaching moisturerecycling governance, Global Environmental Change, 45(April), 15–23, doi:10.1016/j.gloenvcha.2017.04.007, 2017.[132] Keys, P. W., L. Wang-Erlandsson, and L. J. Gordon, Revealing Invisible Water: Moisture Recy-cling as an EcosystemService, PLoSONE, 11(3), e0151,993, doi:10.1371/journal.pone.0151993,2016.[133] Khanna, J., D. Medvigy, S. Fueglistaler, and R. Walko, Regional dry-season climate changesdue to three decades of Amazonian deforestation, Nature Climate Change, 7(3), 200–204, doi:10.1038/nclimate3226, 2017.141[134] Koehler, A., Water use in LCA: managing the planet’s freshwater resources, The InternationalJournal of Life Cycle Assessment, 13(6), 451–455, doi:10.1007/s11367-008-0028-6, 2008.[135] Koellner, T., et al., UNEP-SETAC guideline on global land use impact assessment on biodiversityand ecosystem services in LCA, The International Journal of Life Cycle Assessment, 18(6), 1188–1202, doi:10.1007/s11367-013-0579-z, 2013.[136] Kucharik, C. J., J. A. Foley, C. Delire, V. A. Fisher, M. T. Coe, J. D. Lenters, C. Young-Molling, andN. Ramankutty, Testing the performance of a dynamic global ecosystem model: Water balance,carbon balance, and vegetation structure, Global Biogeochemical Cycles, 14(3), 795–825, doi:10.1029/1999gb001138, 2000.[137] Latawiec, A. E., et al., Improving land management in Brazil: A perspective from producers, Agri-culture, Ecosystems & Environment, 240, 276–286, doi:10.1016/j.agee.2017.01.043, 2017.[138] Lathuillière, M. J., M. S. Johnson, and S. Donner, Water use by terrestrial ecosystems: temporalvariability in rainforest and agricultural contributions to evapotranspiration in Mato Grosso, Brazil,Environmental Research Letters, 7(2), 024,024, doi:10.1088/1748-9326/7/2/024024, 2012.[139] Lathuillière, M. J., M. S. Johnson, G. L. Galford, and E. G. Couto, Environmental footprints showChina and Europe’s evolving resource appropriation for soybean production in Mato Grosso,Brazil, Environmental Research Letters, 9(7), 074,001, doi:10.1088/1748-9326/9/7/074001,2014.[140] Lathuillière, M. J., C. Bulle, andM. S. Johnson, Land Use in LCA: Including Regionally Altered Pre-cipitation to Quantify Ecosystem Damage, Environmental Science & Technology, 50(21), 11,769–11,778, doi:10.1021/acs.est.6b02311, 2016.[141] Lathuillière, M. J., M. T. Coe, and M. S. Johnson, A review of green- and blue-water resourcesand their trade-offs for future agricultural production in the Amazon Basin: what could irrigatedagriculture mean for Amazonia?, Hydrology and Earth System Sciences, 20(6), 2179–2194, doi:10.5194/hess-20-2179-2016, 2016.[142] Lathuillière, M. J., E. J. Miranda, C. Bulle, E. G. Couto, and M. S. Johnson, Past and futuresoybean land transformation impacts in Mato Grosso , Brazil, in 10th International Conference onthe Life Cycle Assessment of Food, 1, pp. 1–8, Dublin, Ireland, 2016.142[143] Lathuillière, M. J., E. J. Miranda, C. Bulle, E. G. Couto, and M. S. Johnson, Land occupation andtransformation impacts of soybean production in Southern Amazonia, Brazil, Journal of CleanerProduction, 149, 680–689, doi:10.1016/j.jclepro.2017.02.120, 2017.[144] Lima, M.A., Pessoa, M.C.P.Y., Neves, M.C., Carvalho, E., Emissões de metano por fermentaçãoentérica e manejo de dejetos de animais. Segundo inventário brasileiro de emissões antrópicasde gases de efeito estufa, Brasilia, DF, Brazil, 2010.[145] Linneman, M., A. Hoekstra, and W. Berkhout, Ranking Water Transparency of Dutch Stock-ListedCompanies, Sustainability, 7(4), 4341–4359, doi:10.3390/su7044341, 2015.[146] Liu, J., and H. Yang, Spatially explicit assessment of global consumptive water uses in cropland:Green and blue water, Journal of Hydrology, 384(3-4), 187–197, doi:10.1016/j.jhydrol.2009.11.024, 2010.[147] Loiseau, E., G. Junqua, P. Roux, and V. Bellon-Maurel, Environmental assessment of a territory:An overview of existing tools and methods, Journal of Environmental Management, 112, 213–225,doi:10.1016/j.jenvman.2012.07.024, 2012.[148] Macedo, M. N., R. S. DeFries, D. C. Morton, C. M. Stickler, G. L. Galford, and Y. E. Shimabukuro,Decoupling of deforestation and soy production in the southern Amazon during the late 2000s,Proceedings of the National Academy of Sciences of the U.S.A., 109(4), 1341–1346, doi:10.1073/pnas.1111374109, 2012.[149] Macedo, M. N., M. T. Coe, R. DeFries, M. Uriarte, P. M. Brando, C. Neill, and W. S. Walker, Land-use-driven stream warming in southeastern Amazonia, Philosophical Transactions of the RoyalSociety B: Biological Sciences, 368(1619), 20120,153–20120,153, doi:10.1098/rstb.2012.0153,2013.[150] Maia, S. M. F., S. M. Ogle, C. E. P. Cerri, and C. C. Cerri, Soil organic carbon stock change due toland use activity along the agricultural frontier of the southwestern Amazon, Brazil, between 1970and 2002, Global Change Biology, 16(10), 2775–2788, doi:10.1111/j.1365-2486.2009.02105.x,2010.[151] Malhi, Y., E. Pegoraro, A. D. Nobre, M. G. P. Pereira, J. Grace, A. D. Culf, and R. Clement,Energy and water dynamics of a central Amazonian rain forest, Journal of Geophysical ResearchAtmospheres, 107(20), 1–17, doi:10.1029/2001jd000623, 2002.143[152] Malveira, V. T. C., J. C. de Araújo, and A. Güntner, Hydrological Impact of a High-Density ReservoirNetwork in Semiarid Northeastern Brazil, Journal of Hydrologic Engineering, 17(1), 109–117, doi:10.1061/(asce)he.1943-5584.0000404, 2012.[153] MAPA, Plano Setorial de Mitigação e de Adaptação às Mudanças Climáticas para a Consoli-dação de uma Economia de Baixa Emissão de Carbono na Agricultura, Ministério da Agricultura,Pecuária e Abastecimento, Brasilia, DF, Brazil, 2012.[154] MAPA, Projeções do agronegócio: Brasil 2015/16 a 2025/26 projeções de longo prazo, Ministérioda Agricultura, Pecuária e Abastecimento, Brasilia, DF, Brazil, 2016.[155] MAPA, Agrostat,, 2017.[156] Mekonnen, M. M., and A. Y. Hoekstra, The green, blue and grey water footprint of crops andderived crop products, Hydrology and Earth System Sciences, 15(5), 1577–1600, doi:10.5194/hessd-8-763-2011, 2011.[157] Mekonnen, M. M., and A. Y. Hoekstra, A Global Assessment of theWater Footprint of Farm AnimalProducts, Ecosystems, 15(3), 401–415, doi:10.1007/s10021-011-9517-8, 2012.[158] Mekonnen, M. M., and A. Y. Hoekstra, Water footprint benchmarks for crop production: A firstglobal assessment, Ecological Indicators, 46, 214–223, doi:10.1016/j.ecolind.2014.06.013, 2014.[159] Mekonnen, M. M., and A. Y. Hoekstra, Four billion people facing severe water scarcity, ScienceAdvances, 2(2), e1500,323–e1500,323, doi:10.1126/sciadv.1500323, 2016.[160] Mekonnen, M. M., M. Pahlow, M. M. Aldaya, E. Zarate, and A. Y. Hoekstra, Sustainability, effi-ciency and equitability of water consumption and pollution in Latin America and the Caribbean,Sustainability, 7(2), 2086–2112, doi:10.3390/su7022086, 2015.[161] Mesa, D., E. Muniz, A. Souza, and B. Geffroy, Broiler-Housing Conditions Affect the Performance,Revista Brasileira de Ciência Avícola, 19(2), 263–272, doi:10.1590/1806-9061-2016-0346, 2017.[162] Meyfroidt, P., et al., Multiple pathways of commodity crop expansion in tropical forest landscapes,Environmental Research Letters, 9(7), 074,012, doi:10.1088/1748-9326/9/7/074012, 2014.[163] Miguel Ayala, L., M. van Eupen, G. Zhang, M. Pérez-Soba, L. G. Martorano, L. S. Lisboa, and N. E.Beltrao, Impact of agricultural expansion on water footprint in the Amazon under climate changescenarios, Science of The Total Environment, 569-570, 1159–1173, doi:10.1016/j.scitotenv.2016.06.191, 2016.144[164] Milà i Canals, L., J. Chenoweth, A. Chapagain, S. Orr, A. Antón, and R. Clift, Assessingfreshwater use impacts in LCA: Part I–Inventory modelling and characterisation factors for themain impact pathways, The International Journal of Life Cycle Assessment, 14(1), 28–42, doi:10.1007/s11367-008-0030-z, 2009.[165] Millen, D. D., R. D. L. Pacheco, M. D. B. Arrigoni, M. L. Galyean, and J. T. Vasconcelos, A snapshotof management practices and nutritional recommendations used by feedlot nutritionists in Brazil,Journal of Animal Science, 87(10), 3427–3439, doi:10.2527/jas.2009-1880, 2009.[166] Montanari, A., et al., "Panta Rhei - Everything Flows": Change in hydrology and society–TheIAHS Scientific Decade 2013–2022, Hydrological Sciences Journal, 58(6), 1256–1275, doi:10.1080/02626667.2013.809088, 2013.[167] Monteith, J. L., Evaporation and surface temperature, Quarterly Journal of the Royal Meteorolog-ical Society, 107(451), 1–27, doi:10.1002/qj.49710745102, 1981.[168] Morais, J. O., and L. S. Pinheiro, The effect of semi-aridity and damming on sedimentary dynamicsin estuaries -Northeastern region of Brazil, in Journal of Coastal Research, vol. 64, pp. 1540–1544,Szczecin, Poland, 2011.[169] Motoshita, M., N. Itsubo, and A. Inaba, Development of impact factors on damage to health byinfectious diseases caused by domestic water scarcity, The International Journal of Life CycleAssessment, 16(1), 65–73, doi:10.1007/s11367-010-0236-8, 2011.[170] MPA, Boletim Estatístico da Pesca e Aquacultura, Ministério da Pesca e Aquicultura, Brasilia, DF,Brazil, 2011.[171] MPA, Boletim Estatístico da Pesca e Aquicultura - 2010, Ministério da Pesca e Aquicultura,Brasilia, DF, Brazil, doi:10.1007/s12032-011-9855-6, 2012.[172] Mu, Q., M. Zhao, and S. W. Running, Improvements to a MODIS global terrestrial evapotranspira-tion algorithm, Remote Sensing of Environment, 115(8), 1781–1800, doi:10.1016/j.rse.2011.02.019, 2011.[173] Neill, C., et al., Watershed responses to Amazon soya bean cropland expansion and intensifica-tion, Philosophical Transactions of the Royal Society of London. Series B, Biological sciences,368(1619), 2–7, doi:10.1098/rstb.2012.0425, 2013.145[174] Neill, C., et al., Surprisingly Modest Water Quality Impacts From Expansion and Intensification ofLarge-Sscale Commercial Agriculture in the Brazilian Amazon-Cerrado Region, Tropical Conser-vation Science, 10, 1–5, doi:10.1177/1940082917720669, 2017.[175] Nepstad, D., et al., Slowing Amazon deforestation through public policy and interventions in beefand soy supply chains., Science, 344(6188), 1118–23, doi:10.1126/science.1248525, 2014.[176] Nepstad, D. C., et al., The role of deep roots in the hydrological and carbon cycles of Amazonianforests and pastures, Nature, 372(6507), 666–669, doi:10.1038/372666a0, 1994.[177] Nobre, C. A., G. Sampaio, L. S. Borma, J. C. Castilla-Rubio, J. S. Silva, and M. Cardoso, Land-use and climate change risks in the Amazon and the need of a novel sustainable developmentparadigm, Proceedings of the National Academy of Sciences of the U.S.A., 113(39), 10,759–10,768, doi:10.1073/pnas.1605516113, 2016.[178] Noojipady, P., C. D. Morton, N. M. Macedo, C. D. Victoria, C. Huang, K. H. Gibbs, and L. E. Bolfe,Forest carbon emissions from cropland expansion in the Brazilian Cerrado biome, EnvironmentalResearch Letters, 12(2), 025,004, doi:10.1088/1748-9326/aa5986, 2017.[179] Novaes, R. M. L., R. A. A. Pazianotto, M. Brandão, B. J. R. Alves, A. May, and M. I. S. Folegatti-Matsuura, Estimating 20-year land-use change and derived CO2 emissions associated with crops,pasture and forestry in Brazil and each of its 27 states,Global Change Biology, 23(9), 3716–3728,doi:10.1111/gcb.13708, 2017.[180] Núñez, M., S. Pfister, A. Antón, P. Muñoz, S. Hellweg, A. Koehler, and J. Rieradevall, Assessingthe Environmental Impact of Water Consumption by Energy Crops Grown in Spain, Journal ofIndustrial Ecology, 17(1), 90–102, doi:10.1111/j.1530-9290.2011.00449.x, 2013.[181] Núñez, M., C. R. Bouchard, C. Bulle, A. M. Boulay, and M. Margni, Critical analysis of life cycleimpact assessment methods addressing consequences of freshwater use on ecosystems andrecommendations for future method development, International Journal of Life Cycle Assessment,21(12), 1799–1815, doi:10.1007/s11367-016-1127-4, 2016.[182] Núñez, M., R. K. Rosenbaum, S. Karimpour, A.-M. Boulay, M. J. Lathuillière, M.Margni, L. Scherer,F. Verones, and S. Pfister, A multimedia hydrological fate modeling framework to assess waterconsumption impacts in life cycle assessment, Environmental Science & Technology, doi:10.1021/acs.est.7b05207, 2018.146[183] Oki, T., S. Yano, and N. Hanasaki, Economic aspects of virtual water trade, Environmental Re-search Letters, 12(4), 044,002, doi:10.1088/1748-9326/aa625f, 2017.[184] Oliveira, L. J. C., M. H. Costa, B. S. Soares-Filho, and M. T. Coe, Large-scale expansion of agri-culture in Amazonia may be a no-win scenario, Environmental Research Letters, 8(2), 024,021,doi:10.1088/1748-9326/8/2/024021, 2013.[185] Oliveira, P. T. S., M. A. Nearing, M. S. Moran, D. C. Goodrich, E. Wendland, and H. V. Gupta,Trends in water balance components across the Brazilian Cerrado, Water Resources Research,50(9), 7100–7114, doi:10.1002/2013wr015202, 2014.[186] Palhares, J. C. P., Pegada hídrica dos suínos abatidos nos Estados da Região Centro-Sul doBrasil, Acta Scientiarum. Animal Sciences, 33(3), 309–314, doi:10.4025/actascianimsci.v33i3.9924, 2011.[187] Palhares, J. C. P., M. Morelli, and C. C. Junior, Impact of roughage-concentrate ratio on the waterfootprints of beef feedlots, Agricultural Systems, 155(December 2016), 126–135, doi:10.1016/j.agsy.2017.04.009, 2017.[188] Panday, P. K., M. T. Coe, M. N. Macedo, P. Lefebvre, and A. D. d. A. Castanho, Deforestationoffsets water balance changes due to climate variability in the Xingu River in eastern Amazonia,Journal of Hydrology, 523, 822–829, doi:10.1016/j.jhydrol.2015.02.018, 2015.[189] Payen, S., C. Basset-Mens, F. Colin, and P. Roignant, Inventory of field water flows for agri-foodLCA: critical review and recommendations of modelling options, The International Journal of LifeCycle Assessment, pp. 1–20, doi:10.1007/s11367-017-1353-4, 2017.[190] Pebesma, E. J., and R. S. Bivand, Classes and methods for spatial data in R, R News, 5(2), 9–13,2005.[191] Pereira, L. S., R. G. Allen, M. Smith, and D. Raes, Crop evapotranspiration estimation with FAO56:Past and future, Agricultural Water Management, 147(March), 4–20, doi:10.1016/j.agwat.2014.07.031, 2015.[192] Perry, C., Water footprints: Path to enlightenment, or false trail?, Agricultural Water Management,134, 119–125, doi:10.1016/j.agwat.2013.12.004, 2014.147[193] Pfister, S., A. Koehler, and S. Hellweg, Assessing the Environmental Impacts of FreshwaterConsumption in LCA, Environmental Science & Technology, 43(11), 4098–4104, doi:10.1021/es802423e, 2009.[194] Pfister, S., S. Vionnet, T. Levova, and S. Humbert, Ecoinvent 3: assessing water use in LCA andfacilitating water footprinting, The International Journal of Life Cycle Assessment, 21(9), 1349–1360, doi:10.1007/s11367-015-0937-0, 2016.[195] Pfister, S., et al., Understanding the LCA and ISO water footprint: A response to Hoekstra (2016)"A critique on the water-scarcity weighted water footprint in LCA", Ecological Indicators, 72, 352–359, doi:10.1016/j.ecolind.2016.07.051, 2017.[196] Phalan, B., W. J. Ripple, and P. Smith, Increasing beef production won’t reduce emissions, GlobalChange Biology, 22(10), 3255–3256, doi:10.1111/gcb.13436, 2016.[197] Pierce, D., ncdf4: Interface to Unidata netCDF (Version 4 or Earlier) Format Data Files, R packageversion 1.16, 2017.[198] Pokhrel, Y. N., Y. Fan, G. Miguez-Macho, P. J. F. Yeh, and S. C. Han, The role of groundwater inthe Amazon water cycle: 3. Influence on terrestrial water storage computations and comparisonwith GRACE, Journal of Geophysical Research Atmospheres, 118(8), 3233–3244, doi:10.1002/jgrd.50335, 2013.[199] Pokhrel, Y. N., Y. Fan, and G. Miguez-Macho, Potential hydrologic changes in the Amazon bythe end of the 21st century and the groundwater buffer, Environmental Research Letters, 9(8),084,004, doi:10.1088/1748-9326/9/8/084004, 2014.[200] Pongratz, J., L. Bounoua, R. S. DeFries, D. C. Morton, L. O. Anderson, W. Mauser, and C. A.Klink, The Impact of Land Cover Change on Surface Energy and Water Balance in Mato Grosso,Brazil, Earth Interactions, 10(19), 1–17, doi:10.1175/ei176.1, 2006.[201] Presidência da República, Lei Nº 12.651, de 25 de maio de 2012,, 2012.[202] Priante-Filho, N., et al., Comparison of the mass and energy exchange of a pasture and a ma-ture transitional tropical forest of the southern Amazon Basin during a seasonal transition, GlobalChange Biology, 10(5), 863–876, doi:10.1111/j.1529-8817.2003.00775.x, 2004.148[203] Priestley, C. H. B., and R. J. Taylor, On the Assessment of Surface Heat Flux and Evapo-ration Using Large-Scale Parameters, Monthly Weather Review, 100(2), 81–92, doi:10.1175/1520-0493(1972)100<0081:otaosh>;2, 1972.[204] Quinteiro, P., A. C. Dias, M. Silva, B. G. Ridoutt, and L. Arroja, A contribution to the environmentalimpact assessment of green water flows, Journal of Cleaner Production, 93, 318–329, doi:10.1016/j.jclepro.2015.01.022, 2015.[205] Quinteiro, P., B. G. Ridoutt, L. Arroja, and A. C. Dias, Identification of methodological challengesremaining in the assessment of a water scarcity footprint: a review, The International Journal ofLife Cycle Assessment, doi:10.1007/s11367-017-1304-0, 2017.[206] Quinteiro, P., S. Rafael, P. Villanueva-Rey, B. Ridoutt, M. Lopes, L. Arroja, and A. C. Dias, Acharacterisation model to address the environmental impact of green water flows for water scarcityfootprints, Science of The Total Environment, 626, 1210–1218, doi:10.1016/j.scitotenv.2018.01.201, 2018.[207] R Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Sta-tistical Computing, Vienna, Austria, 2017.[208] Ramankutty, N., and J. A. Foley, Estimating historical changes in global land cover: Croplandsfrom 1700 to 1992,Global Biogeochemical Cycles, 13(4), 997–1027, doi:10.1029/1999gb900046,1999.[209] Ramos, P., Indices de Rendimento da Agropecuária Brasileira, Ministério do DesenvolvimentoAgrário, Nucleo de Estudos Agrários e Desenvolvimento Rural (NEAD), Brasilia, DF, Brazil, 2005.[210] Ran, Y., M. Lannerstad, M. Herrero, C. Van Middelaar, and I. De Boer, Assessing water resourceuse in livestock production: A review of methods, Livestock Science, 187, 68–79, doi:10.1016/j.livsci.2016.02.012, 2016.[211] Ran, Y., C. E. van Middelaar, M. Lannerstad, M. Herrero, and I. J. de Boer, Freshwater use in live-stock production - To be used for food crops or livestock feed?, Agricultural Systems, 155(August2016), 1–8, doi:10.1016/j.agsy.2017.03.008, 2017.[212] Richards, P., H. Pellegrina, L. VanWey, and S. Spera, Soybean Development: The Impact of aDecade of Agricultural Change on Urban and Economic Growth in Mato Grosso, Brazil, PLoSONE, 10(4), e0122,510, doi:10.1371/journal.pone.0122510, 2015.149[213] Richter, B. D., M. M. Davis, C. Apse, and C. Konrad, A presumptive standard for environmentalflow protecttion, River Research and Applications, 28(8), 1312–1321, doi:10.1002/rra.1511, 2012.[214] Ridoutt, B., P. Juliano, P. Sanguansri, and J. Sellahewa, The water footprint of food waste: casestudy of fresh mango in Australia, Journal of Cleaner Production, 18(16-17), 1714–1721, doi:10.1016/j.jclepro.2010.07.011, 2010.[215] Ridoutt, B., et al., Making sense of the minefield of footprint indicators, Environmental Scienceand Technology, 49(5), 2601–2603, doi:10.1021/acs.est.5b00163, 2015.[216] Ridoutt, B. G., and S. Pfister, A revised approach to water footprinting to make transparent the im-pacts of consumption and production on global freshwater scarcity,Global Environmental Change,20(1), 113–120, doi:10.1016/j.gloenvcha.2009.08.003, 2010.[217] Ridoutt, B. G., and S. Pfister, Reducing humanity’s water footprint, Environmental Science & Tech-nology, 44(16), 6019–6021, doi:10.1021/es101907z, 2010.[218] Ridoutt, B. G., P. Sanguansri, M. Freer, and G. S. Harper, Water footprint of livestock: comparisonof six geographically defined beef production systems, The International Journal of Life CycleAssessment, 17(2), 165–175, doi:10.1007/s11367-011-0346-y, 2012.[219] Ridoutt, B. G., G. Page, K. Opie, J. Huang, and W. Bellotti, Carbon, water and land use footprintsof beef cattle production systems in southern Australia, Journal of Cleaner Production, 73, 24–30,doi:10.1016/j.jclepro.2013.08.012, 2014.[220] Riskin, S. H., C. Neill, K. Jankowski, A. V. Krusche, R. McHorney, H. Elsenbeer, M. N. Macedo,D. Nunes, and S. Porder, Solute and sediment export from Amazon forest and soybean headwaterstreams, Ecological Applications, 27(1), 193–207, doi:10.1002/eap.1428, 2017.[221] Rockström, J., Water for food and nature in drought-prone tropics: vapour shift in rain-fed agricul-ture, Philosophical Transactions of the Royal Society B: Biological Sciences, 358(1440), 1997–2009, doi:10.1098/rstb.2003.1400, 2003.[222] Rockström, J., M. Lannerstad, and M. Falkenmark, Assessing the water challenge of a newgreen revolution in developing countries., Proceedings of the National Academy of Sciences ofthe U.S.A., 104(15), 6253–60, doi:10.1073/pnas.0605739104, 2007.[223] Rockström, J., et al., A safe operating space for humanity, Nature, 461(7263), 472–475, doi:10.1038/461472a, 2009.150[224] Rockström, J., et al., The unfolding water drama in the Anthropocene: towards a resilience basedperspective on water for global sustainability, Ecohydrology, 7(5), 1249–1261, doi:10.1002/eco.1562, 2014.[225] Rodrigues, L. N., E. E. Sano, T. S. Steenhuis, and D. P. Passo, Estimation of small reservoirstorage capacities with remote sensing in the Brazilian Savannah region, Water Resources Man-agement, 26(4), 873–882, doi:10.1007/s11269-011-9941-8, 2012.[226] Rodrigues, T. R., G. L. Vourlitis, F. D. A. Lobo, R. G. De Oliveira, and J. D. S. Nogueira, Seasonalvariation in energy balance and canopy conductance for a tropical savanna ecosystem of southcentral Mato Grosso, Brazil, Journal of Geophysical Research: Biogeosciences, 119(1), 1–13,doi:10.1002/2013jg002472, 2014.[227] Rost, S., D. Gerten, A. Bondeau, W. Lucht, J. Rohwer, and S. Schaphoff, Agricultural green andblue water consumption and its influence on the global water system,Water Resources Research,44(9), 1–17, doi:10.1029/2007wr006331, 2008.[228] Roy, E. D., et al., The phosphorus cost of agricultural intensification in the tropics, Nature Plants,2(5), 16,043, doi:10.1038/nplants.2016.43, 2016.[229] Ruviaro, C. F., C. M. de Léis, V. d. N. Lampert, J. O. J. Barcellos, and H. Dewes, Carbon footprintin different beef production systems on a southern Brazilian farm: a case study, Journal of CleanerProduction, 96, 435–443, doi:10.1016/j.jclepro.2014.01.037, 2015.[230] Saad, R., T. Koellner, and M. Margni, Land use impacts on freshwater regulation, erosion regula-tion, and water purification: a spatial approach for a global scale level, The International Journalof Life Cycle Assessment, 18(6), 1253–1264, doi:10.1007/s11367-013-0577-1, 2013.[231] Savenije, H. H. G., A. Y. Hoekstra, and P. van der Zaag, Evolving water science in the Anthro-pocene, Hydrology and Earth System Sciences, 18(1), 319–332, doi:10.5194/hess-18-319-2014,2014.[232] Schyns, J. F., and A. Y. Hoekstra, The Added Value of Water Footprint Assessment for NationalWater Policy: A Case Study for Morocco, PLoS ONE, 9(6), e99,705, doi:10.1371/journal.pone.0099705, 2014.[233] Schyns, J. F., A. Y. Hoekstra, and M. J. Booij, Review and classification of indicators of green151water availability and scarcity, Hydrology and Earth System Sciences, 19(11), 4581–4608, doi:10.5194/hessd-12-5519-2015, 2015.[234] Sentelhas, P. C., R. Battisti, G. M. S. Câmara, J. R. B. Farias, A. C. Hampf, and C. Nendel, Thesoybean yield gap in Brazil–magnitude, causes and possible solutions for sustainable production,The Journal of Agricultural Science, 153(08), 1394–1411, doi:10.1017/S0021859615000313,2015.[235] Siebert, S., and P. Döll, Quantifying blue and green virtual water contents in global crop productionas well as potential production losses without irrigation, Journal of Hydrology, 384(3-4), 198–217,doi:10.1016/j.jhydrol.2009.07.031, 2010.[236] Silva, R. D. O., L. G. Barioni, J. A. J. Hall, M. F. Matsuura, T. Z. Albertini, F. A. Fernandes, andD. Moran, Increasing beef production could lower greenhouse gas emissions in Brazil if decoupledfrom deforestation, (January), 3–8, doi:10.1038/nclimate2916, 2016.[237] Silvério, D. V., P. M. Brando, J. K. Balch, F. E. Putz, D. C. Nepstad, C. Oliveira-Santos, andM. M. C. Bustamante, Testing the Amazon savannization hypothesis: fire effects on invasion of aneotropical forest by native cerrado and exotic pasture grasses, Philosophical Transactions of theRoyal Society B: Biological Sciences, 368(1619), 20120,427–20120,427, doi:10.1098/rstb.2012.0427, 2013.[238] Silvério, D. V., P. M. Brando, M. N. Macedo, P. S. A. Beck, M. Bustamante, and M. T. Coe, Agri-cultural expansion dominates climate changes in southeastern Amazonia: the overlooked non-GHG forcing, Environmental Research Letters, 10(10), 104,015, doi:10.1088/1748-9326/10/10/104015, 2015.[239] Simon, M. F., and F. L. Garagorry, The expansion of agriculture in the Brazilian Amazon, Environ-mental Conservation, 32(03), 203, doi:10.1017/s0376892905002201, 2006.[240] Smakhtin, V., C. Revenga, and P. Döll, A Pilot Global Assessment of Environmental Water Re-quirements and Scarcity, Water Int., 29(3), 307–317, doi:10.1080/02508060408691785, 2004.[241] Soares-filho, B., R. Rajão, M. Macedo, A. Carneiro, W. Costa, M. Coe, H. Rodrigues, and A. Alen-car, Cracking Brazil ’ s Forest Code, Science, 344(April), 363–364, doi:10.1126/science.124663,2014.152[242] Soares-Filho, B., et al., Brazil’s market for trading forest certificates, PLoS ONE, 11(4), 1–18,doi:10.1371/journal.pone.0152311, 2016.[243] Soares-Filho, B.S. and Nepstad, D.C. and Curran, L.M. and Voll, E. and Garcia, R.A. and Ramos,C.A. and McDonald, A.J. and Lefebvre, P.A. and Schlesinger, P., LBA-ECO LC-14 ModeledDeforestation Scenarios, Amazon Basin: 2002-2050, ORNL DAAC, Oak Ridge, TN, USA, doi:10.3334/ORNLDAAC/1153, 2013.[244] Spera, S., Agricultural Intensification Can Preserve the Brazilian Cerrado: Applying Lessons FromMato Grosso and Goiás to Brazil’s Last Agricultural Frontier, Tropical Conservation Science, 10,1–7, doi:10.1177/1940082917720662, 2017.[245] Spera, S. A., A. S. Cohn, L. K. VanWey, J. F. Mustard, B. F. Rudorff, J. Risso, and M. Adami,Recent cropping frequency, expansion, and abandonment in Mato Grosso, Brazil had selectiveland characteristics, Environmental Research Letters, 9(6), 064,010, doi:10.1088/1748-9326/9/6/064010, 2014.[246] Spera, S. A., G. L. Galford, M. T. Coe, M. N. Macedo, and J. F. Mustard, Land-use change affectswater recycling in Brazil’s last agricultural frontier, Global Change Biology, 22(10), 3405–3413,doi:10.1111/gcb.13298, 2016.[247] Spracklen, D. V., and L. Garcia-Carreras, The impact of Amazonian deforestation on Amazonbasin rainfall, Geophysical Research Letters, 42(21), 9546–9552, doi:10.1002/2015gl066063,2015.[248] Spracklen, D. V., S. R. Arnold, and C. M. Taylor, Observations of increased tropical rainfall pre-ceded by air passage over forests, Nature, 489(7415), 282–285, doi:10.1038/nature11390, 2012.[249] Steduto, P., D. Raes, T. Hsiao, and E. Fereres, AquaCrop: a new model for crop prediction underwater deficit conditions, Options Méditerranéennes, 33(80), 285–292, 2009.[250] Steduto, P., Hsiao, T. C., Fereres, E., Raes, D., FAO Irrigation and Drainage paper 66: Crop yieldresponse to water, Food and Agriculture Organization, Rome, Italy, 2012.[251] Stickler, C. M., M. T. Coe, M. H. Costa, D. C. Nepstad, D. G. McGrath, L. C. P. Dias, H. O.Rodrigues, and B. S. Soares-Filho, Dependence of hydropower energy generation on forests inthe Amazon Basin at local and regional scales, Proceedings of the National Academy of Sciencesof the U.S.A., 110(23), 9601–9606, doi:10.1073/pnas.1215331110, 2013.153[252] Strassburg, B. B., A. E. Latawiec, L. G. Barioni, C. A. Nobre, V. P. da Silva, J. F. Valentim,M. Vianna, and E. D. Assad, When enough should be enough: Improving the use of currentagricultural lands could meet production demands and spare natural habitats in Brazil, GlobalEnvironmental Change, 28(1), 84–97, doi:10.1016/j.gloenvcha.2014.06.001, 2014.[253] Strassburg, B. B. N., et al., Moment of truth for the Cerrado hotspot, Nature Ecology & Evolution,1(4), 0099, doi:10.1038/s41559-017-0099, 2017.[254] Strassburg, B., Micol, L., Ramos, F., Seroa da Motta, R., Latawiec, A., Lisauskas, F., IncreasingAgricultural Output While Avoiding Deforestation - A case study for Mato Grosso, Brazil, Rio deJaneiro, RJ, Brazil, 2012.[255] Talamini, E., C. F. Ruviaro, T. J. Florindo, and G. I. B. D. M. Florindo, Improving feed efficiency asa strategy to reduce beef carbon footprint in the Brazilian Midwest region, International Journal ofEnvironment and Sustainable Development, 16(4), 379, doi:10.1504/ijesd.2017.10007706, 2017.[256] Thiago, L. R. L. S. and Silva, J. M., Aspectos práticos da suplementação alimentar de bovinos decorte, Campo Grande, MS, Brazil, 2006.[257] UN, Millenium Development Goals and Beyond 2015,, 2017.[258] UN, Sustainable development knowledge platform,, 2017.[259] UN-Water, Water Security and the Global Water Agenda: a UN-Water Analytical Brief, UnitedNations University, Hamilton, ON, Canada, doi:10.1016/0022-1694(68)90080-2, 2013.[260] UN-Water/FAO, Coping with water scarcity: Challenge of the twenty-first century, United Nations,doi:10.3362/0262-8104.2005.038, 2007.[261] USDA,World Agricultural Production,,United States Department of Agriculture, 2017.[262] USDA, National Agricultural Statistics Services,, 2017.[263] van der Ent, R. J., H. H. G. Savenije, B. Schaefli, and S. C. Steele-Dunne, Origin and fate ofatmospheric moisture over continents, Water Resources Research, 46(9), 1–12, doi:10.1029/2010wr009127, 2010.154[264] van der Walt, S., J. L. Schönberger, J. Nunez-Iglesias, F. Boulogne, J. D. Warner, N. Yager,E. Gouillart, and T. Yu, scikit-image: image processing in python, PeerJ, 2, e453, doi:10.7717/peerj.453, 2014.[265] van Zelm, R., A. M. Schipper, M. Rombouts, J. Snepvangers, a. J. Huijbregts, and A. Root,Implementing groundwater extraction in life cycle impact assessment : characterization fac-tors based on plant species richness, Environmental Science & Technology, 45(2), 1–16, doi:10.1021/es102383v, 2011.[266] Vanham, D., and G. Bidoglio, The water footprint of agricultural products in European river basins,Environmental Research Letters, 9(6), 064,007, doi:10.1088/1748-9326/9/6/064007, 2014.[267] Vanham, D., M. Mekonnen, and A. Hoekstra, The water footprint of the EU for different diets,Ecological Indicators, 32, 1–8, doi:10.1016/j.ecolind.2013.02.020, 2013.[268] Vanham, D., T. Mak, and B. Gawlik, Urban food consumption and associated water resources:The example of Dutch cities, Science of The Total Environment, 565, 232–239, doi:10.1016/j.scitotenv.2016.04.172, 2016.[269] Velasquez, H.Q.C, Bernasconi, P., Fique por dentro : A bacia do Rio Xingu em Mato Grosso,Instituto Socioambiental, Instituto Centro de Vida, São Paulo, Brazil, 2010.[270] Veldkamp, T., et al., Water scarcity hotspots travel downstream due to human interventions in the20th and 21st century, Nature Communications, 8, 15,697, doi:10.1038/ncomms15697, 2017.[271] Vickers, D., and L. Mahrt, Quality Control and Flux Sampling Problems for Tower and Aircraft Data,Journal of Atmospheric and Oceanic Technology, 14(3), 512–526, doi:10.1175/1520-0426(1997)014<0512:qcafsp>;2, 1997.[272] Vörösmarty, C. J., A. Y. Hoekstra, and S. E. Bunn, What scale for water governance ?, Science,349(6247), 478–479, doi:10.1126/science.349.6247.478-a, 2015.[273] Vourlitis, G. L., N. Priante Filho, M. M. S. Hayashi, J. S. Nogueira, F. T. Caseiro, and J. H.Campelo Jr, Seasonal variations in the evapotranspiration of a transitional tropical forest of MatoGrosso, Brazil, Water Resources Research, 38(6), 1–11, doi:10.1029/2000wr000122, 2002.[274] Vourlitis, G. L., J. de Souza Nogueira, F. de Almeida Lobo, and O. B. Pinto, Variations inevapotranspiration and climate for an Amazonian semi-deciduous forest over seasonal, an-155nual, and El Niño cycles, International Journal of Biometeorology, 59(2), 217–230, doi:10.1007/s00484-014-0837-1, 2015.[275] Webb, E. K., G. I. Pearman, and R. Leuning, Correction of flux measurements for density effectsdue to heat and water vapour transfer, Quarterly Journal of the Royal Meteorological Society,106(447), 85–100, doi:10.1256/smsqj.44706, 1980.[276] Wickham, H., ggplot2: Elegant Graphics for Data Analysis, Springer-Verlag, New York, NY, USA,2009.[277] Wickham, H., scales: Scale Functions for Visualization. R package version 0.4.1., 2016.[278] Wright, J. S., R. Fu, J. R. Worden, S. Chakraborty, N. E. Clinton, C. Risi, Y. Sun, and L. Yin,Rainforest-initiated wet season onset over the southern Amazon, Proceedings of the NationalAcademy of Sciences of the U.S.A., 114(32), 8481–8486, doi:10.1073/pnas.1621516114, 2017.[279] WWAP, The United Nations World Water Development Report 3: Water in a Changing World,UNESCO, London, UK, doi:10.3390/w3020618, 2009.[280] WWAP, The United Nations World Water Development Report 2015: Water for a SustainableWorld, UNESCO, Paris, France, doi:978-92-3-100071-3, 2015.[281] WWAP, Facing the challenges: Case studies and indicators, UNESCO, Paris, France, 2015.[282] Xavier, A. C., C.W. King, and B. R. Scanlon, Daily griddedmeteorological variables in Brazil (1980-2013), International Journal of Climatology, 36(6), 2644–2659, doi:10.1002/joc.4518, 2016.[283] Zaks, D. P. M., C. C. Barford, N. Ramankutty, and J. A. Foley, Producer and consumer responsi-bility for greenhouse gas emissions from agricultural production - a perspective from the BrazilianAmazon, Environmental Research Letters, 4(4), 044,010, doi:10.1088/1748-9326/4/4/044010,2009.[284] Zeileis, A., and G. Grothendieck, zoo : S3 Infrastructure for Regular and Irregular Time Series,Journal of Statistical Software, 14(6), doi:10.18637/jss.v014.i06, 2005.[285] Zeitoun, M., Lankford, B., Bakker, K., Conway, D., Introduction: A battle of ideas for water security,in Water Security: Principles, perspectives and practices, edited by D. Lankford, B., Bakker, K.,Zeitoun, M., Conwa, chap. 1, pp. 3–25, Earthscan, New-York, NY, USA, 2014.156[286] Zemp, D. C., C.-F. Schleussner, H. M. J. Barbosa, M. Hirota, V. Montade, G. Sampaio, A. Staal,L. Wang-Erlandsson, and A. Rammig, Self-amplified Amazon forest loss due to vegetation-atmosphere feedbacks, Nature Communications, 8, 14,681, doi:10.1038/ncomms14681, 2017.[287] Zhang, L., W. R. Dawes, and G. R. Walker, Response of mean annual evapotranspiration tovegetation changes at catchment scale, Water Resources Research, 37(3), 701–708, doi:10.1029/2000wr900325, 2001.157AppendicesAppendix AChapter 3 supporting informationA.1 The Soyflux stationEquipment described in Table 3.2 is shown in Figure A.1 during the soybean and maize crop develop-ment cycle of the Rainfed-1 and Rainfed-2 fields.A.2 Crop height monitoringCrop height was monitored using field measurements performed during site visits as well as an auto-mated camera taking hourly pictures of the soybean and maize development cycles in Rainfed-2 asdescribed in Section 3.2.2. First, crop height was plotted as a function of time to derive models used toadjust the daily displacement height for flux calculation in EddyPro® shown in Tables A.1 and A.2. In ad-dition, camera pictures of the soybean and maize development cycles allowed for a detailed descriptionof the development cycle of both crops (Figures A.2 and A.3).Pictures taken in the Rainfed-2 field during the soybean and maize development cycles were usedto make two separate movies showing the different stages of Figures A.2 and A.3. The soybean devel-opment cycle from October 2016 to February 2017 is publicly available at, while the maize development cycle from February 2016 to July 2016is available at (please apply high definitionviewing for best quality).158Figure A.1: The Soyflux station described in Chapter 3 following soybean (top) and maize (bottom)crop development cycles between September 2015 and June 2016. Note the location of the Rainfed-1field (background, right) and Rainfed-2 field (foreground, left) shown in Figure 3.1. The additional towerstructure in the Rainfed-2 field contains the net radiometer and NDVI sensor described in Table 3.2.159Table A.1: Models used to infer daily crop height following measurements made in the Rainfed-1 fieldand inferred from camera pictures. The soybean crop growth period of 2015-2016 did not benefit fromthe automatic camera installation.Crop/year Height (cm) Start date End dateSoybean/20150 28 October 2015 2 November 20154 3 November 2015 3 November 20151.49days−9.13 (R2= 0.99) 4 November 2015 28 December 201580 29 December 2015 11 January 2016−1.13days + 165a 12 January 2016 11 February 2016Maize/20160 13 February 2016 19 February 20165 20 February 2016 25 February 20163.42days−38.9 (R2=0.97) 26 February 2016 1 May 2016230 2 May 2016 18 July 2016Brachiara/2016 50 19 July 2016 16 September 201630 17 September 2016 4 October 2016Soybean/20160 5 October 2016 11 October 20161.62days−11.1 (R2=0.96) 14 October 2016 21 December 2016115 22 December 2016 3 January 2017−0.86days + 192a 4 January 2017 16 January 2017103 17 January 2017 4 February 2017aCurve determined from straight line between the end of the mid-season and the crop height atharvest (2 points)160Table A.2: Models used to infer daily crop height following measurements made in the Irrigated field.The Irrigated field did not benefit from the automatic camera installation.Crop/year Height (cm) Start date End dateSoybean/20150 29 September 2015 4 October 20151.13days−4.79 (R2=0.94) 5 October 2015 29 November 201565 30 November 2015 3 December 2015−0.76days + 114a 4 December 2015 13 Jan 2016Rice/20160 1 February 2016 5 Feb 20162.90exp(0.06days) (R2=0.94) 6 February 2016 19 March 201658 20 March 2016 30 April 2016Bean/20160 14 June 2016 14 June 20160.50days + 2.30 (R2=0.98) 15 June 2016 16 September 201650 17 September 2016 22 September 2016Stubble/2016 10 23 September 2016 29 September 2016Soybean/20160 30 September 2016 4 October 20163 5 October 2016 7 October 20165 8 October 2016 10 October 20161.59days−12.3 (R2=0.92) 11 October 2016 29 November 201690 30 November 2016 4 February 2017aCurve determined from straight line between the end of the mid-season and the crop height atharvest (2 points)161Figure A.2: Soybean crop height measurements taken in 2015 and 2016 in the Rainfed-1 and Rainfed-2fields and separated into initial (4-6 days), development (7-60 days), mid-season (61-100 days), andfinal phase (101-127 days). Only the Rainfed-2 measurements made in 2016 took advantage of theautomatic camera setup for height measurements.Figure A.3: Maize crop height measurements taken in 2016 in Rainfed-1 and Rainfed-2 and separatedinto initial (0-20 days), development (21-56 days), mid-season (57-97 days), and final phase (98-151days). Only the Rainfed-2 measurements took advantage of the automatic camera setup for heightmeasurements.162Figure A.4: Energy balance closure as the sum of latent heat (LEmeas) and sensible heat (H) fluxesmeasured by eddy covariance in the Rainfed-1 field as a function of the difference between net radiation(Rn) and ground heat flux (G) measured in the Rainfed-2 field. The equation for the regression line (blueline) is also shown.A.3 Quality control of eddy covariance dataA.3.1 Energy balance closureEnergy balance closure was assessed for all half-hour measurements made between 18 September2015 and 4 February 2017 through linear regression of the sum of latent (LE) and sensible heat fluxes(H) obtained from the Soyflux station (separated using wind direction 0-150° for the Rainfed-1 field and150-320° for the Irrigated field, as described in Section 3.2.3), with the difference between net radiation(Rn) and ground heat flux (G) measured in the Rainfed-2 field. The energy balance closure was 61%(LEmeas + H = 0.61(Rn−G) + 32.14, R2 = 0.75) for Rainfed-1 (Figure A.4), and 82% (LEmeas + H =0.82(Rn−G) + 17.16, R2 = 0.87) for the Irrigated field (Figure A.5).Differences in the energy balance closure were attributed to possible advection from nearby fieldswith different crops than the Irrigated and Rainfed-1 fields, but also differences in atmospheric stabilitybased on the conditions of each field. The atmospheric stability ζ (dimensionless) is given by equation3.7 (Chapter 3). Results of values of ζ were classified following ranges defined by Franssen et al.81 whostudied 26 European FLUXNET sites (Table A.3).There were more measurements made in the Irrigated field (n = 4460) than in the Rainfed-1 field (n= 4211) with an overall greater occurrence of slightly unstable ζ in the Irrigated field. Energy balance163Figure A.5: Energy balance closure as the sum of latent heat (LEmeas) and sensible heat (H) fluxesmeasured by eddy covariance in the Irrigated field as a function of the difference between net radiation(Rn) and ground heat flux (G) measured in the Rainfed-2 field. The equation for the regression line (blueline) is also shown.Table A.3: Occurrence of the half-hourly value of the atmospheric stability parameter (ζ) in the Rainfed-1and Irrigated fields classified following Franssen et al.81.Field ζ ≥ 0.1 −0.1 < ζ < 0.1 −0.5 < ζ < −0.1 ζ < −0.5stable neutral slightly unstable very unstableRainfed-1 (n = 4211) 1269 (30%) 1576 (37%) 943 (22%) 423 (10%)Irrigated (n = 4460) 1164 (26%) 1591 (36%) 1194 (27%) 511 (11%)164Figure A.6: Example of calculation of the Priestley-Taylor α obtained through linear regression for theRainfed-1 field on 9 December 2016 (a) and between 25 November 2016 and 1 December 2016 (b). Allplotted data represents half-hourly measurements.closure was found to be affected by atmospheric stability conditions with worse closures reported whenζ ≥ 0.1 compared to ζ < −0.1 and a sharp drop in energy balance closure when ζ > 0 (from roughly80% closure to 55% closure)81 which corresponds to our 20% differences in closure between the twofields. Analysis on a monthly basis showed slight differences in stability between fields. The Rainfed-1field showed greater occurrence of a stable ζ in December 2015, April, June, July and October 2016.The Irrigated field showed greater occurrence of a neutral ζ in May, and August to November 2016 anda prevalence of slightly unstable conditions in September and October 2016.A.3.2 Eddy covariance latent heat flux gap fillingGap filling was performed using calibrated values of α obtained from the Priestley-Taylor equation203over daily and weekly periods as described in equation 3.1 (Chapter 3). Values of α were obtained byplotting linear regressions of LEmeas obtained from eddy covariance data against ΔΔ+γ (Rn−G) on thehalf-hourly measurements forced through the origin (Figure A.6). Values of α were derived as a range(αlow -αhigh) based on systematic and random errors as described in the main document.Daily values of αlow and αhigh obtained in Rainfed-1 were plotted as a function of soil volumetric watercontent (θ) at 0.05-m, 0.10-m, 0.30-m, and 0.60-m depths (measured in the Rainfed-2 field) to derivelinear regressions used to gap-fill values of α for gaps greater than one week (Table A.4). Following thegap filling steps described in Section 3.2.4, we obtained a full time series of calibrated daily Priestley-165Table A.4: Linear regression table of Priestley-Taylor α values measured in the Rainfed-1 field (αlow andαhigh) as a function of daily mean soil volumetric water content (θ) measured at different soil depths inthe Rainfed-2 field.Soil depth(m)0.05 0.10 0.30 0.60αlow2.36θ −0.06 3.62θ + 0.03 2.42θ 2.99θ −0.02R2=0.47 R2=0.38 R2=0.49 R2=0.33αhigh3.81θ −0.07 5.93θ + 0.06 3.95θ + 0.02 4.61θ + 0.06R2=0.44 R2=0.36 R2=0.46 R2=0.30Figure A.7: Daily mean Priestley-Taylor α values calculated for the Rainfed-1 and Irrigated fields between18 September 2015 and 4 February 2017.Taylor α values (Figures A.7).Mean values of α (obtained from αlow and αhigh) ranged from 0.16 to 1.97 (Rainfed-1) and 0.19 to 1.58(Irrigated) throughout the year according to the meteorological conditions and canopy development inboth Rainfed-1 and Irrigated fields (Figure A.7). Clear differences were observed between the Rainfed-1and Irrigated fields: the value of α typically increased with the crop development to 0.5-1.0 in the wetseason when soybean (October-February) and maize (February-July) were planted, and dropped toabout 0.20 in the dry season. In the Irrigated field, values of α were typically greater (Figure A.7) andincreased in the dry season with the prevalence of irrigation for bean (June-September) to values closerto 1.2. Both time series were similar to those described for natural vegetation in Sinop, Mato Grosso166Figure A.8: Comparison of gap-filled values of LE (LE modeled) using the daily mean Priestley-Taylor αvalues with measurements of LE (LE measured, or LEmeas) obtained from eddy covariance in both theRainfed-1 (left) and Irrigated (right) fields. The blue lines represent the regression lines.Table A.5: Linear regression results of modeled LE (LEmod ) using calibrated Priestley-Taylor α values(αlow -αhigh) as a function of measured LE (LEmeas) by eddy covariance.Gap-filling Rainfed-1 Irrigatedwith αlowLEmod = 0.91LEmeas−3.20 LEmod = 0.87LEmeas−2.19R2= 0.92 R2= 0.78with αhighLEmod = 1.46LEmeas−3.62 LEmod = 1.11LEmeas−2.56R2= 0.92 R2= 0.79(about 200 km north of the Soyflux tower) reported by Vourlitis et al.274 where a seven-year averagewas lowest in the dry season (0.5-0.6 in September) and highest near the end of the wet season (near1.0 in May).Values of LE, gap-filled using the calibrated values of the Priestley-Taylor α, were then comparedto measurements of LE (LEmeas) (Figure A.8). Values of LE obtained from αhigh and αlow (Table A.5)show how the range of Priestley-Taylor α values provides a range of LE values that contain the LEmeasurements.167A.4 Calculation of reference evapotranspirationReference ET (ET0) represents ET from a theoretical grass crop and was calculated following equation3.42 in the main document. Given that the surface considered is theoretical, we model half-hourly valuesof Rn andG following the steps and assumptions below. Net radiation was calculated following equationA.12Rn,grss = (1−0.23)Rs−Rn (A.1)where Rn,grass (MJ m-2 30-min-1) is the net radiation above the theoretical grass surface, Rs (MJ m-230-min-1) is the incoming shortwave radiation measured at the Soyflux site, and Rnl (MJ m-2 30-min-1)is the net outgoing longwave radiation. The value of 0.23 in equation A.1 is the albedo of the theoreticalgrass reference crop2. The values of Rnl are calculated as per Allen et al.2Rn = σ (T+ 273.13)0.34−0.14pe1.35RsRso− 0.35(A.2)where σv is the Stefan-Bolzmann constant (4.903 MJ K-4 m-2 30-min-1), T (°C) is the air temperature,ea (kPa) is the actual vapour pressure, and Rso (MJ m-2 30-min-1) is the clear-sky incoming shortwaveradiation. We estimatedRso as a function of extraterrestrial radiation (Ra) which was calculated for everyhalf-hour in the day following equation A.32, while the nighttime ratio of RsRso was assumed to be 0.5R = 1260piGscdr [(ω2−ω1)sin(φ)sin(δ)+ cos(φ)cos(δ)(sin(ω2)− sin(ω1))] (A.3)where Gsc is the solar constant (0.0820 MJ m-2 min-1), dr is the inverse relative distance between Earthand the Sun and calculated in equation A.42, ω1 and ω2 (rad) are the solar time angles, respectively atthe beginning and the end of each half-hourly estimate and are obtained by equations A.5 to A.7 belowfollowing Allen et al.2, φ (rad) is the Soyflux’s station latitude, and δ (rad) is the solar declination obtainedfrom equation A.8 below2dr = 1+ 0.033cos2pij365(A.4)ω=pi2[(t+ 0.06667(Lz− Lm)+ Sc)− 12] (A.5)ω1 =ω−0.5pi24(A.6)ω2 =ω+0.5pi24(A.7)168δ= 0.409sin2pij365−1.39(A.8)where j is the day of the year, ω is the solar time angle at the mid-point of the time windows (at 0.25and 0.75 hours), t is the clock time at the mid-point of the time window, Lz is the longitude of the centerof the time zone under consideration (60° for Amazon Standard Time), and Lm is the longitude of theSoyflux site, and Sc is the solar time seasonal correction obtained following Allen et al.2Sc = 0.1645sin(2b)−0.1255cos(b)−0.025sin(b) (A.9)where b is equal to 2pi(j−81)364 according to Allen et al.2.From the above steps we obtain Rso as a mean function of Ra with Rso = 0.77R (sd = 0.16)on half-hour intervals when Rn > 500 W m-2 as an indication of clear sky conditions. Moreover, soilheat flux of the grass surface (Ggrass) is assumed to equal 0.1Rn,grass during the day and 0.5Rn,grassat night according to Allen et al.2. All computations were carried out with R Statistical Software207(v.3.4.0) in R Studio (v. 1.0.143) with the openair package37. The source code is available online at AquaCrop settings used for crop modelingWe explored differences in evaporation, crop transpiration and crop water productivity (WP) as a functionof planting dates and irrigation practices using AquaCrop (v.6.0) from Steduto et al.249. Details on theinput data used are shown in Table A.6. The model validation step involved fine tuning the modelparameters to best represent the crop development cycle using NDVI measured in Rainfed-2 as a proxy(Figure A.12), and crop ET (Table 3.4) before increasing the crop’s harvest index to meet the yieldreported by the farmer in the Rainfed-1 field (Table 3.4). When running the soybean and maize modelsfor 2016 we obtained the results shown in Table A.7.A.6 Energy partitioning in Rainfed-1 and Irrigated fieldsWe plotted the mean LE (gap-filled values) as a function of the difference Rn−G in both wet (September-May) and dry (June-August) seasons for the Rainfed-1 and Irrigated fields (Figure A.9). A large portion ofavailable energy was used for LE in the wet season as shown by the high correlations between availableenergy and LE with slopes of 0.62 and 0.63 in both the Rainfed-1 and Irrigated fields (Figure A.9a and169Table A.6: Input data and parameters used in AquaCrop for both soybean and maize.Input data Parameters Values changed Default valuesClimate Rainfall, temperature,ET0Field measurements fromthe meteorologicalstation; ET0 calculatedfollowing equation 3.4.Atmospheric CO2concentration (403ppm)Crop Canopy cover anddevelopment, cropcoefficients from soybeanand maize planted in2016NDVI measurementsused as a proxy forcanopy cover; cropcoefficients (KC) fromTable 3.5Crop waterproductivity,rooting depth,responses tostressManagement Irrigation IrrigationrequirementSoil profile Field capacity, permanentwilting point, hydraulicconductivity at saturation(KS)Soil volumetric watercontent measurementsfrom Table A.8, curvenumber was set to lowest(10), KS assumed 65 mmd-1Conductivity of 0dS m-1, runoff is0.5 mmGroundwater Deep groundwater(no capillary rise)Soil initial conditions Volumetric water content Based on measurementsmade in Rainfed-2 on thechosen planting dateConductivity of 0dS m-1Table A.7: Comparison of measured evapotranspiration (ET), yield and water productivity (WP) of soy-bean and maize in Rainfed-1 compared to modeled values obtained from AquaCrop after validation.Crop (planting date) ET (mm) Yield (ton ha-1) WP (kg m-3)Soybean (5 October 2016)measured 423 ± 99 4.020 0.77-1.24modeled 431.6 4.026 1.20Maize (13 February 2016)measured 314 ± 67 7.620 2.00-3.09modeled 386.2 7.598 2.52170Table A.8: Soil field capacity (θfc) and dry soil (θds) determined for 0.05-m, 0.10-m, 0.30-m, and 0.60-mdepths in the Rainfed-2 field.Depth (m) θfc (m3 m-3) θds (m3 m-3)(sd, n) (sd, n)0.05 0.305 0.154(0.056, 292) (0.040, 101)0.10 0.169 0.097(0.041, 326) (0.003, 35)0.30 0.274 0.128(0.043, 319) (0.008, 92)0.60 0.206 0.125(0.045, 349) (0.004, 74)A.9b). However, in the dry season, energy partitioning depended on whether the field was suppliedwith irrigation. In the Irrigated field, the relationship between available energy and LE was maintainedto similar levels as those in the wet seasons (Figure A.9d), but changed without irrigation (Figure A.9b)with a drop in the slope relating available energy to LE from 0.68 to 0.26. The above slopes for theRainfed-1 field were similar to those observed for pasture in Sinop at 0.54 in the wet season, but muchlower than in the dry season (0.41 observed) from Priante-Filho et al.202.A.7 Water potential, soil volumetric water content andpercolationWe combined the soil volumetric water content (θ) and water potential (ψ) measurements to provide anestimate of field capacity (−33 kPa < ψ < −10 kPa) and dry soil as a proxy for permanent wilting point (ψ< −500 kPa) (Table A.8).The soil water balance equation for the Rainfed-1 field can be defined by equation A.102, assumingno runoff or sub-surface flowΔSWS= L(VWCt+1−VWCt) = Pt−ETt−ϵt (A.10)where ΔSWS (mm) is the change in soil water storage, L (mm) is the soil depth considered, VWCt+1 −171Figure A.9: Radiation partitioning described by linear regression of latent heat flux (LE, W m-2) as afunction of the difference between net radiation (Rn, W m-2) and ground heat flux (G, W m-2) for Rainfed-1 in the wet (a) and dry (b) season, and the Irrigated field in the wet (c) and dry (d) seasons. Blue linesare the regression lines (equations also shown).172VWCt (m3 m-3) is the difference in daily mean soil water storage between day t and day t+1, Pt (mm)is the daily total precipitation on day t, ETt (mm) is the daily total ET of the Rainfed-1 field on day t,and εt (mm) is the daily average drainage beyond depth L. Total soil water content VWC was calculatedfollowing equation A.11VWC= 0.07θ0.05+ 0.07θ0.10+ 0.25θ0.30+ 0.45θ0.60 (A.11)where VWC (m3 m-3) was obtained up to the 0.60-m depth considering representative layers: θ0.05 ,for the 0-0.07 m layer, θ0.10 for the 0.07-0.13 m layer, θ0.30 for the 0.20-0.45 m layer, and θ0.60 for the0.45-0.60 m layer. Drainage beyond the 0.60-m depth was obtained using equation A.10 and solvingfor εt. Finally, we obtained the average daily soil available water fraction (Awf )35Aƒ =θ− θdsθƒ c− θds (A.12)where θ (m3 m-3) is the average daily soil volumetric water content, θfc (m3 m-3) is the field capacity,and θds (m3 m-3) is the proxy for permanent wilting point.The daily water balance was plotted using linear regression (Figure A.10) without considering εt. Thedaily change in soil water storage up to 0.60-m depth as a function of P − ET was described by ΔSWS= 0.47(P −ET) −1.75 (R2 = 0.33) suggesting deeper drainage, beyond the 0.60-m depth.A.8 NDVI measurements in the Rainfed-1 fieldNDVI wasmeasured in the Rainfed-1 field over the course of the maize and soybean development cycles(Figure A.12c)173Figure A.10: Daily changes in soil water storage down to the 0.60-m depth (ΔSWS, mm d-1) in theRainfed-2 field, as a function of daily available water (P −ET, mm d-1) in the Rainfed-1 field. The blueline is the regression line (equation also shown).174Figure A.11: Soil water balance derived from measurements of daily precipitation (P, mm d-1) (a), evap-otranspiration (ET, mm d-1), changes in daily average soil water storage (ΔSWS, mm d-1) (c), dailyaverage drainage below the 0.60-m depth (ε, mm d-1) (d), and soil available water fraction (Awf, dimen-sionless) in the root zone (e) .175Figure A.12: Environmental and crop characteristics in the Rainfed-1 field: precipitation (P, mm d-1)(a), 24-hour mean net radiation (Rn, W m-2) (b), Normalized Difference Vegetation Index (NDVI, dimen-sionless), evapotranspiration (ET, mm d-1), and the ratio of ET to reference ET (ET/ET0, dimensionless)(e).176Appendix BChapter 4 supporting informationB.1 Carbon footprint of Brazilian cattle productionAlthough many studies have been published on the carbon footprint (CF) of Brazilian cattle (Table B.1),most do not include land use change in the calculation of the final result. For completeness, we in-clude recent results of greenhouse gas emissions from land use change that were attributed to cattleproduction in Mato Grosso (Table B.2).B.2 Volumetric water footprint of pastureThe VWF of pasture was determined by estimating pasture ET in each municipal unit (MU) of MatoGrosso before relating it to high and low pasture productivity scenarios. Pasture ET was estimated fromprecipitation following Zhang et al.287ETP =1+ 0.51100P1+ 0.51100P +P1100P (B.1)where ETP (mm y-1) is the pasture ET, and P (mm y-1) is the annual precipitation. Pasture ET wasestimated for each MU from spatial daily precipitation data in Mato Grosso using CHIRPS (v.2.0) fromFunk et al.82 available online ( Precipitation was obtained annually from 2000 to 2015 using statistical software R (v.3.4.0)207 in RStudio (v.1.0.143) equipped with the packages raster (v.2.5-8)106, sp26,190, rgdal (v.1.2-7)27, maptools(v.0.9-2)25, and ncdf4 (v.1.16)197. Annual values of ETP were then combined into two pasture produc-tivity scenarios. The high pasture productivity scenario assumed a production of 5.3 tons of dry matter(DM) per hectare based on a three year estimate from Thiago and Silva256 of 1.5 animal units (A.U.)ha-1 y-1 with a 300 kg ha-1 15-15-15 N-P-K fertilizer application. The low productivity pasture scenarioassumed production of 3 tons DM ha-1 y-1. The combination of the values of ETP with pasture produc-177Table B.1: Recent greenhouse gas emissions estimates for cattle herds in Brazil. None of these studiesconsidered land use change in their estimates.BrazilianstateSpecies Description and method Greenhouse gasemissions(kg CO2-eq (kgLW)-1)ReferenceMato Grosso No mention(likelyNelore)aDetailed farm and herdinformation; cattle herd <2000 heads; includes tier2 emissions (field, diesel,agricultural inputs, etc.)4.8-8.2 Cerri et al.40Mato Grosso No mention(likelyNelore)aDetailed farm and herdinformation; cattle herd >2000 heads; includes tier2 emissions5.0-7.2 Cerri et al.40Brazil Nelore orNelorecrossesDetailed herd informationfollowing five distinctscenarios of increasedconfinement; includes tier2 emissions59.0-117a Cardoso et al.36Rio Grandedo Sulno mention(likelyAngus)aFarm and databaseinformation, extensivesystem22.52 Dick et al.64Rio Grandedo Sulno mention(likelyAngus)aFarm and databaseinformation intensivesystem9.16 Dick et al.64Rio Grandedo SulAngus Rye and sorghumpasture; includes tier 2emissions18.3 Ruviaro et al.229Rio Grandedo SulAngus Natural grass system;includes tier 2 emissions42.6 Ruviaro et al.229aCould not obtain confirmation from authors at the time of writingbResults were reported in kg CO2-eq (carcass weight)-1 and converted here into CO2-eq (kg LW)-1assuming a 50% carcass weight178Table B.2: Greenhouse gas emissions from land use change for cattle production in the Amazon biomeof Mato Grosso, Brazil.Land usechange Description and method YearsGreenhousegas emissions(Tg CO2-eq y-1)ReferenceForest-to-pasturetransitionCarbon losses from landuse change and annualpasture maintenance2001-2005 200.6 DeFries et al.61Forest- andcropland-to-pasturetransitionIncludes emissions fromdecomposition 2001-2005 208.9 Karstensen et al.129Forest- andcropland-to-pasturetransitionIncludes emissions fromdecomposition 2006-2010 114.8 Karstensen et al.129tivity assumptions led to pasture VWF of 2610 L (kg DM)-1 (sd = 56) and 1480 L (kg DM)-1 (sd = 32) forlow and high pasture productivity respectively. Average results are shown in Table B.3.B.3 Validation of remote sensing informationB.3.1 Validation of pasture areaPasture area obtained from remote sensing as described in the main document, was compared to pas-ture area predicted from animal population, following the steps described in Lathuillière et al.18,138.Briefly, total livestock units (TLU, A.U.) were determined following equation B.2TLU(t, ) =∑kN(, t,k)ƒAU(t,k) (B.2)where values of TLU were calculated for year t and MU i following the total animal population (N) ofanimal k, and the respective animal unit factor fAU (A.U. animal-1)209. We then derive the livestockdensity (LSD, A.U. ha-1) for year t and municipal unit i asLSD(t, ) =TLU(t, )AP(t, )(B.3)179Table B.3: Volumetric water footprint (m3 (kg DM)-1) of feed from pasture following two productivityscenarios: 3 tons DM ha-1 (low productivity), and 5.3 tons DM ha-1 (high productivity).Year Pasture productivityLow High(m3 (kg DM)-1)2001 2.61 (0.13) 1.47 (0.07)2006 2.77 (0.11) 1.50 (0.07)2011 2.64 (0.14) 1.50 (0.08)2015 2.57 (0.13) 1.46 (0.07)where Ap (ha) is the pasture area for year t and municipal unit i. Values of Ap are known for 1996and 2006 (years when an agricultural census was carried out121), which allows for the construction ofa two-point regression curves for each MU i relating LSD(1996, i) and LSD(2006, i) with Ap(1996, i)and Ap(2006, i). These regressions were then used to derive Ap for each MU i in non-census yearswhen TLU(t, i) can be derived from agricultural production information121. For the 2007-2015 period,when no agricultural census information was available, we estimatedAp assuming a continuous increasein LSD(t, i) following the increase between 1996 and 2006, and assuming a constant LSD(t, i) equalto LSD(2006, i). Values of Ap obtained from the method above were then compared to pasture areadetermined using Landsat imagery99 for all 104 MUs of Mato Grosso. Given that Landsat imagerywas derived from images spanning 1.5 years, we validated pasture areas using agricultural productionareas for two consecutive years and found good agreement between the estimates (Table B.4). Wefound better agreement between the two datasets when we assumed a constant LSD for the IBGE121estimate, confirming a change in the statewide LSD following the 2006 agricultural census. The totalMato Grosso pasture areas estimated from the animal population and from Landsat imagery were lowerthan previous estimates using MODIS when including and excluding protected areas148, but still greaterthan the average 14 Mha estimated in 2008 and 2010 by Almeida et al.60 (Figure B.1).B.3.2 Validation of small farm impoundment areaReservoir area was calculated using a two-stage automated machine learning classification procedure.First, water pixels were identified. Second, water bodies were classified as reservoir or non-reservoir180Table B.4: Correlation coefficients (m and b) when comparing pasture area estimates obtained usinganimal population from IBGE121 (Ap,IBGE), and remote sensing (Ap,RS)99. Comparisons were made forall 104 municipal units.Validation: Ap,RS = mAp,IBGE + b (R2)Year of pasture area forremote sensing (Ap,RS)Year of pasture area ofIBGE121 (Ap,IBGE )m b R22000/01 2000 1.41 −3793 0.772000/01 2001 1.40 −2926 0.742005/06 2005 1.36 5622 0.822005/06 2006 1.47 5934 0.822010/11 2010increasing LSDa1.54 15120 0.802010/11 2011increasing LSDa1.55 21490 0.782010/11 2010constant LSDb1.38 8064 0.822010/11 2011constant LSDb1.22 2634 0.772014/15 2014increasing LSDa1.62 18542 0.782014/15 2015increasing LSDa1.78 11741 0.832014/15 2014constant LSDb1.40 13074 0.822014/15 2015constant LSDb1.41 9574 0.86aAssumes an increase in livestock density following the trend between 1996 and 2006bAssumes a steady livestock density after 2006 equal to LSD(2006, i)181Figure B.1: Comparison of estimates of total pasture area in Mato Grosso from IBGE121 (this study)following Lathuillière et al.138, and remote sensing including MODIS148 and Landsat99. Estimates fromMODIS are provided with and without the consideration of protected areas (PA).(e.g., natural stream or lake). Training data for reservoir locations was gathered by manually classifyingpoints as land, reservoir water, or natural water using very high-resolution imagery available on GoogleEarth. In each stage of the reservoir classification, 20% of the training data was set aside to be usedfor testing. The independent test accuracy for reservoir locations ranged from 97.1-98.0% for the firststage and 86.4-92.4% for the second. We compared our 2010 Landsat-based estimate of reservoir area(30-m resolution) for the Xingu Basin of Mato Grosso to a previously published estimate based on 2007ASTER data (15-m resolution)149. Results from this study identified a total of 3985 reservoirs coveringan estimated area of 7260 ha, compared with 9994 reservoirs and 20,760 ha identified by the ASTERdata. The discrepancy in the total number of reservoirs detected suggests that many reservoirs in thelandscape are smaller than the detection limit of Landsat data, which was 3 Landsat pixels (0.27 ha) forour classification. These very small reservoirs are widespread in the landscape and likely account formost of the nearly three-fold difference in reservoir area seen between the two datasets. Even in caseswhere reservoirs were detected by both datasets, the ASTER-based area estimates were consistentlyhigher than Landsat-based ones. In addition to differences in spatial resolution, this pattern could beattributed to differences in spectral resolution, object-oriented classification methodologies, or precipi-tation across the two study years. In all cases, the Landsat-based estimates were more conservative182Figure B.2: Distribution of water evaporated from reservoirs (W res, L (kg LW)-1) in 2015 in the 104municipal units of Mato Grosso.than ASTER, suggesting that the results presented here may be an underestimate of the total reservoirarea in Mato Grosso state.B.4 Distribution of small farm impoundments and reservoircattle densityWe compared average values of W res and reservoir cattle density by removing the MUs which containthe Pantanal (Caceres, Mirassol d’Oeste, Lambari d’Oeste, Curvelândia, Barão do Melgaço, Itiquira,Nossa Senhora do Livramento, Poconé, and Santo Antonio do Leverger) shown in Table B.5. Therewere no major differences between the means. By allocation a portion of small farm impoundmentevaporation to fish tanks, the average values of W res dropped by as much as 46% (Table B.6).183Figure B.3: Distribution of reservoir cattle density (RCD, cattle ha-1) in 2015 in the 104 municipal unitsof Mato Grosso.Figure B.4: Significant increases and decreases (p ≤ 0.05) in reservoir cattle density (RCD, cattle ha-1)between 2001 and 2015.184Table B.5: Comparison of reservoir evaporation (W res) with reservoir cattle density (RCD) considering allMato Grosso municipal units (MUs), and all MUs without those within the limits of the Pantanal wetland.Year W res (L (kg LW)-1) (sd) RCD (cattle ha-1) (sd) Total reservoirevaporation (L y-1)All MUs withoutPantanalAll MUs withoutPantanalAll MUs withoutPantanal2001 197 (267) 188 (262) 872 (1352) 874 (1350) 6.70 × 1011 5.03 × 10112006 157 (206) 149 (203) 1156 (1784) 1202 (1826) 6.00 × 1011 4.46 × 10112011 178 (245) 171 (247) 860 (922) 884 (935) 7.22 × 1011 5.67 × 10112015 215 (316) 2210 (318) 706 (653) 727 (664) 9.15 × 1011 7.26 × 1011Table B.6: Reservoir cattle density (RCD, cattle ha-1), and changes in values of small farm reservoirevaporation allocated to cattle production (W res, L (kg LW)-1) considering all farm impoundments, andremoving evaporation allocated to fish tanks. Values of RCD excluding fish tanks are provided as arange based on fish production yields (3.5 ton ha-1 and 7 ton ha-1).2001 2006 2011 2015Total farm reservoir area(ha)47,516 41,184 49,733 70,058Farm reservoir areasexcluding fish tanks (ha)46,087–46,802 36,754–38,969 35,805–42,769 56,624–57,401RCD (sd) (cattle ha-1) Allfarm reservoirs872 (1352) 1156 (1784) 860 (922) 706 (653)W res (sd) (L (kg LW)-1)All farm reservoirs197 (267) 157 (206) 178 (245) 215 (316)W res (L (kg LW)-1)Excluding fish tanks158-161 101-108 88-105 192-217Decrease in W res basedon allocation−19% −33% −46% −5%185Figure B.5: Distribution of pasture cattle density (PCD, cattle ha-1) in 2015 in the 104 municipal units ofMato Grosso.B.5 Changes in pasture cattle density in Mato GrossoMato Grosso’s pasture cattle density (PCD) increased from 0.75 cattle ha-1 (sd = 0.54) in 2001 to 0.93cattle ha-1 (sd = 0.56) in 2015. In 2015, the largest PCD values appeared along the Amazon biome dividein northern and southeastern Mato Grosso, as well as 6 MUs in the southwest (Figure B.5). MU specificvalues ranged from 0.05 to 2.64 cattle ha-1 in 2000/01 to 0.11 to 2.52 cattle ha-1 in 2015. Significantchanges in PCD between 2001 and 2015 (p ≤ 0.05) affected 25 of the 104 MUs (or 24%) with 23 MUsshowing positive correlations and 2 negative correlations (Alta Floresta, Marcelândia) in northern MatoGrosso (Figure B.6). The majority of MUs that experienced significant increases in PCD in the timeperiod, showed density < 1.0 cattle ha-1 in 2001 with the exception of Figeiropolis D’oeste (1.32 cattleha-1), and São José do Povo (1.46 cattle ha-1), while the twoMUs that experienced significant decreasesin PCD had 1.96 cattle ha-1 (Alta Floresta) and 0.88 cattle ha-1 (Marcelândia) in 2001 (Table B.7).The majority of the MUs with significant increases in PCD over time (n = 18) experienced slightdeclines in pasture area along with increases in animal population. In contrast, the two MUs that showedsignificant decreases in PCD over time experienced a combined increase in both pasture area andanimal population: Alta Floresta went from 0.23 Mha of pasture in 2001 to 0.37 Mha of pasture in 2015186Figure B.6: Significant increases and decreases (p ≤ 0.05) in pasture cattle density (PCD, cattle ha-1)between 2001 and sustain respectively 447,931 and 594,644 animals; Marcelândia went from 0.12 Mha of pasture in2001 to 0.20 Mha of pasture in 2015 to sustain respectively 106,338 and 156,579 animals.187Table B.7: Summary of changes in pasture cattle density (PCD, cattle ha-1) over time.Description MU (n) 2000/01 PCD(cattle ha-1) (sd)2014/15 PCD(cattle ha-1) (sd)All data 104 0.75 (0.54) 0.93 (0.56)Significant increase inPCD between 2001 and201523 0.51 (0.36) 0.84 (0.55)Significant decrease inPCD between 2001 and20152 1.42 (0.77) 1.21 (0.52)No significant change inPCD over time79 0.80 (0.56) 0.94 (0.56)B.6 Relationship between reservoir and pasture areaFigure B.7: Relationship between small farm reservoir area (Ares) and pasture area (Ap) in the 104 MUsof Mato Grosso in 2001 and 2015.188Appendix CChapter 5 supporting informationC.1 Integrated BIosphere Simulator (IBIS) model validation ofdischargeDespite successful validation of the IBIS model outputs (discharge, evapotranspiration (ET), total waterstorage) of the Xingu River Basin as per Panday et al.188, we validated the IBIS modeled runoff for asmall area of the Xingu Basin of Mato Grosso (XBMT) encompassing the Xingu Headwaters. Riverdischarge R(t) for the 2000 and 2005 hydrologic years were obtained following equation 5.1 comparedto station 18430000 located in Marcelândia (10° 46” 38’ S, 53° 5’ 44” W) (Figure 5.1) with data availablefrom 1975 to 2005 from ANA5. Monthly values of R(2000) (n = 12) and R(2005) (n = 4) compared wellto publicly available data (Figure C.1) showing a Pearson correlation value of r = 0.83 (compared to 0.89in the 2000s for the Xingu River Basin in Panday et al.188).We observed larger discrepancies between modeled and observed R(t) in the November-Januaryperiod and therefore analyzed interannual R(t) using 3-month averages to provide a magnitude of wateravailability in both dry and wet seasons (Figure C.2). The linear regression of modeled versus measured3-month average discharge for R(2000) (n = 4) gave R(t) modeled = 1.18R(t) measured – 561 (R2 =0.88).189Figure C.1: Validation of the monthly discharge (R(t)) for the Xingu Headwaters in the 2000 (n = 12) and2005 (n = 4) hydrologic years at station 18430000 located in Marcelândia (Mato Grosso)5 (Figure 5.1).Figure C.2: Modeled compared to observed 3-month mean discharge at station 18430000 located inMarcelândia (Mato Grosso)5 for the Xingu Headwaters in the 2000 (n = 12) and 2005 (n = 4) hydrologicyears, and for the 1975-2005 (n = 120) period (Figure 5.1).190C.2 Input data used for the volumetric water footprint accountingTable C.1: Cropland and pasture evapotranspiration (ET) according to Lathuillière et al.138,141 and theirrespective areas estimated from agricultural production information121, and Landsat imagery99 (bottom-up approach) to determine total ET for agriculture (ETAG).Land use ET Areaa: 2001, 2015121 Area: 2000, 201499mm y-1 Mha MhaForest 1099 NA 12.8, 11.4Pasture 822-889 3.4, 2.3 4.4, 4.2Soybean + fallow 608-688 0.020, 2.2Soybean + maize + fallow717-8080.095, 0.730.32, 2.1Soybean + rice + fallow 0.26, 0.081aAs data is available by municipal unit, these areas represent a percent of total production based onthe percent area of the political unit contained with the Xingu Basin of Mato Grosso. Maize and rice areassumed as double crops following soybean planting/harvest and are assumed to have similar totalcrop ET191Table C.2: Average live animal population in 2000 and 2014 hydrologic years with animal water demandand living condition assumptions. Populations were obtained from IBGE121, include both males andfemales and were allocated to the Xingu Basin of Mato Grosso based on area of municipalities containedwithin the basin. Chicken and swine populations were recalculated based on life expectancy describedin equation 5.5.Animal Conditions Population Blue waterconsumptiontotal live animals m3 d-1 animal-1Hydrologic year 2000 2014Cattle Pasture 2,534,975 3,535,838 50 × 10-3Horsesa Pasture 28,954 47,766 50 × 10-3Buffaloesa Pasture 4467 2781 50 × 10-3Donkeysa,b Pasture 578 633 50 × 10-3Mulesa,b Pasture 9124 12,908 50 × 10-3Swine Confined 16,358 51,724 0.125 × 10-3Goats Pasture 2388 2973 54.0 × 10-3Sheep Pasture 16,691 38,544 54.0 × 10-3Chicken/Roosters Confined 72,303 572,741 0.284 × 10-3aNo data available for 2015, the population was assumed constant in 2013 and 2014; bANA4192Table C.3: Urban, rural, industrial worker population and domestic and industrial blue water demand inthe Xingu Basin of Mato Grosso. Note that blue water consumption was assumed to be 50% of bluewater demand. Data derived from IBGE121 and ANA4.Connected to the water system? Population Blue water demandm3 d-1 cap-1Hydrologic year 2000 2014Total population 141,301 222,101domestic-urban Yes 41,806 65,711 260 × 10-3domestic-urban No 47,142 74,100 70 × 10-3domestic-rural Yes 25,653 40,322 70 × 10-3domestic-rural No 26,700 41,968 70 × 10-3industrial workers 88,948 139,811 3.5C.3 Determination of environmental flow requirementsWe followed the procedure described in Smakhtin et al.240 to derive annual environmental flow require-ments (EFR) to maintain ecosystems in “fair” conditions. From an ecological management perspective,these conditions are described as: “the dynamics of the biota have been disturbed. Some sensitivespecies are lost and/or reduced in extent. Alien species may occur”240 which is defined from the val-ues of Q50 and Q90 obtained from the long-term discharge data of the Xingu Headwaters observedbetween 1975 and 2005 at Marcelândia (Passagem BR80, station 18430000, 10° 46’ 38” S, 53° 5’ 44”W)5 (Figure C.3). Mean annual runoff (MAR) of the Xingu Headwaters was 1921 m3 s-1 mo-1 with aQ50 of 1455 m3 s-1 mo-1 (76% MAR) and a Q90 of 810 m3 s-1 mo-1 (42% MAR). Smakhtin et al.240then define EFR as the sum of low flow (Q50) and high flow (Q90) with the low flow set to zero in caseswhere Q90 exceeds 40% MAR (which is the case for the Xingu Headwaters). Our estimate of annualEFR was therefore 42% MAR which is slightly greater than the Amazon basin average of 31% MAR andthe average EFR for the Xingu Basin of 20-25% MAR from Smakhtin et al.240.193Figure C.3: Exceedance probability curve for the Xingu Headwaters obtained frommonthly observationsat Marcelândia, Mato Grosso (Passagem BR80, station 18430000, 10° 46’ 38” S, 53° 5’ 44” W)5 for the1975-2005 period (n = 363).C.4 Land use cover for deforestation scenariosFollowing deforestationmaps obtained fromSoares-Filho et al.243 we extracted forest cover from business-as-usual (BAU) and governance (GOV) scenarios (see Table 5.1) for 2030 and 2050 in the XBMT (TableC.4). The deforestation scenario maps were obtained at 1 km2 resolution and estimate a total XBMTsurface area of 159,256 km2 according to Soares-Filho et al.243 compared to 177,000 km2 obtainedfrom Landsat imagery from Graesser and Ramankutty99.C.5 Total blue volumetric water footprints and hydrologicconditionsWe obtained the total annual blue water consumed in the XBMT according to steps described in Sections5.2.3.2 (Table C.5) and compare to the annual estimated runoff in the basin (Table C.6) to obtain bluewater scarcity for 2000 and 2014, as well as the deforestation and climate scenarios described in Table5.1. We also divided annual runoff into 3-month sums to account for seasonal variability (Table C.6).194Table C.4: Total forest cover as described by land use maps obtained by Soares-Filho et al.243 in theXingu Basin of Mato Grosso for business-as-usual (BAU) and governance (GOV) deforestation scenar-ios.Deforestation scenario Total forest cover (km2) Total forest cover (% basin)BAU-2030 45,114 28.33BAU-2050 32,619 20.48GOV-2030 68,462 42.99GOV-2050 67,096 42.13Table C.5: Total blue volumetric water footprint for agricultural, industrial and domestic uses in the XinguBasin of Mato Grosso in 2000 and 2014 hydrologic years, as well as scenarios for 2030 and 2050 (seeTable C.2 and C.3 for input data, as well as Table 5.1 for the description of scenarios).Year Scenario Agricultural Industrial Domestickm3 y-12000-01 0.153 1.60 × 10-5 4.09 × 10-32014-15 0.218 2.56 × 10-5 6.54 × 10-32030-31 BAURCP4.5 0.255 3.86 × 10-5 9.86 × 10-3BAURCP8.5 0.255 3.86 × 10-5 9.86 × 10-3GOVRCP4.5 0.255 3.86 × 10-5 9.86 × 10-3GOVRCP8.5 0.255 3.86 × 10-5 9.86 × 10-32050-51 BAURCP4.5 0.517 6.53 × 10-5 1.67 × 10-2BAURCP8.5 0.517 6.53 × 10-5 1.67 × 10-2GOVRCP4.5 0.391 6.53 × 10-5 1.67 × 10-2GOVRCP8.5 3.81 6.53 × 10-5 1.67 × 10-2195Table C.6: Total annual and 3-month mean runoff in the Xingu Basin of Mato Grosso obtained from IBISsimulations and land use (equation 5.1).Year Scenario Precipitation PNVa Annual Sep-Nov Dec-Feb Mar-May Jun-Augmm y-1 km3 y-1 km3 3-months-1 (% change from 2000)2000-01 1999 69.8 74.9 5.8 20.7 36.9 11.52014-15 1934 64.1 70.4 5.9 14.3 38.3 12.02030-31 BAURCP4.5 1966 67.9 78.6 (+5) 6.9 (+20) 7.1 (–66) 47.1 (+28) 17.4 (+52)BAURCP8.5 1971 69.1 80.0 (+7) 6.6 (+14) 6.1 (–69) 49.5 (+34) 17.5 (+52)GOVRCP4.5 1966 67.9 76.3 (+2) 6.9 (+19) 6.6 (–68) 45.6 (+24) 17.3 (+50)GOVRCP8.5 1971 69.1 77.8 (+4) 6.5 (+13) 6.0 (–71) 47.9 (+30) 17.4 (+51)2050-51 BAURCP4.5 1969 69.0 80.8 (+8) 6.9 (+19) 8.0 (–62) 49.1 (+33) 16.9 (+47)BAURCP8.5 1952 65.7 77.7 (+4) 6.8 (+18) 6.3 (–69) 47.6 (+29) 17.0 (+48)GOVRCP4.5 1969 69.0 77.4 (+3) 6.8 (+18) 7.2 (–65) 46.8 (+27) 16.6 (+45)GOVRCP8.5 1952 65.7 74.3 (–1) 6.8 (+17) 5.9 (–71) 44.9 (+22) 16.7 (+45)aPotential Natural Vegetation following Ramankutty and Foley208196Table C.7: Individual land use contributions to evapotranspiration (ET) obtained in this study using thebottom-up approach between 2000 and 2010 compared to values obtained by Silvério et al.238 usingthe MODIS ET product172.Land use Study 2000 2005 2010 2014Forest, shrubland This study 141 129 125Forest, Cerrado Silvério et al.238 142 142 138Pasture This study 37.8 41.2 35.6Silvério et al.238 34.7 47.7 50.7Cropland This study 2.8 10.2 14.3Silvério et al.238 1.6 6.2 8.9Agriculture (Pasture +Cropland)This study 40.6 51.5 49.9Silvério et al.238 36.3 53.9 59.6Total ET This study 181 180 175Silvério et al.238 179 195 198Deviations in total ET Comparisonbetween this studyandSilvério et al.238+1% –8%C.6 Land use evapotranspiration contributionsWe used both top-down and bottom-up approaches to estimate the changes in land contributions toET. First, the bottom-up approach was used following steps described in the main document in order todevise changes between 2000 and 2014. Results were compared to land ET estimates derived by Sil-vério et al.238 using MODerate resolution Imaging Spectroradiometer ET product172 in the XBMT (TableC.7). Our results were close to those of Silvério et al.238 who report a decrease in ET of approximately35 km3 in the 2000s (considering land use transitions affecting natural vegetation). Silvério et al.238report that 12% of forests in the basin (18,838 km2) were either converted to cropland (3347 km2) orpasture (15,491 km2) between 2000 and 2010. The difference between our values obtained through thebottom-up approach and those of Silvério et al.238 was attributed to differences in resolution between theproducts used (1 km for MODIS compared to 30 m for IBIS) as well as the model steps used to obtainET with the Penman-Monteith equation in MOD16 from Mu et al.172 and our procedure (see Section5.2.3.2).197Table C.8: Values of total evapotranspiration (ETT ), evapotranspiration of the natural vegetation (ETNV ),potential natural vegetation (ETPNV ), and the combined ET of agriculture and residual landscapes(ETAG + ETR) in the Xingu Basin of Mato Grosso between 2000 and 2050 hydrologic years consider-ing business-as-usual (BAU) and governance (GOV) deforestation, and Representative ConcentrationPathways (RCP 4.5 and 8.5 W m-2). All values were obtained using the top-down approach (FigureC.4).Year Scenario ETT ETNV ETPNV ETAG + ETRkm3 y-12000-01 279.0 191.2 284.1 87.82014-15 272.0 172.2 278.4 99.82030-31 BAURCP4.5 269.6 82.1 280.3 187.5BAURCP8.5 271.8 82.0 279.7 189.8GOVRCP4.5 268.9 122.9 280.3 146.0GOVRCP8.5 271.1 122.7 279.7 148.42050-51 BAURCP4.5 267.9 60.0 279.9 207.9BAURCP8.5 271.3 60.0 280.0 211.3GOVRCP4.5 268.0 122.5 279.9 145.5GOVRCP8.5 271.4 122.3 280.0 149.1We then used the top-down approach using IBIS simulations to describe changes in ET land contri-butions following deforestation scenarios and RCPs (Figure C.4, Table C.8).198Figure C.4: Land contributions to evapotranspiration (ET) in the Xingu Basin of Mato Grosso between2000 and 2050 using the top-down approach and following business-as-usual (BAU) and governance(GOV) deforestation scenarios, and Representative Concentration Pathways (RCP 4.5 and 8.5 W m-2)as described in Table 5.1.199


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items