Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Riparian forest management and regeneration : effects on forest structure and stream ecological processes… Rossetti de Paula, Felipe 2018

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

Item Metadata

Download

Media
24-ubc_2018_september_rossettidepaula_felipe.pdf [ 4.89MB ]
Metadata
JSON: 24-1.0367004.json
JSON-LD: 24-1.0367004-ld.json
RDF/XML (Pretty): 24-1.0367004-rdf.xml
RDF/JSON: 24-1.0367004-rdf.json
Turtle: 24-1.0367004-turtle.txt
N-Triples: 24-1.0367004-rdf-ntriples.txt
Original Record: 24-1.0367004-source.json
Full Text
24-1.0367004-fulltext.txt
Citation
24-1.0367004.ris

Full Text

RIPARIAN FOREST MANAGEMENT AND REGENERATION: EFFECTS ON FOREST STRUCTURE AND STREAM ECOLOGICAL PROCESSES IN STREAMS OF EASTERN AMAZON, BRAZIL   by   Felipe Rossetti de Paula   B.Sc., University of São Paulo, 2007  M.Sc., São Paulo State University “Julio de Mesquita Filho”, 2010    A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF   DOCTOR OF PHILOSOPHY  in  THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES  (Forestry)    THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)    May 2018    © Felipe Rossetti de Paula, 2018     ii  The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled:  Riparian forest management and regeneration: effects on forest structure and stream ecological processes in streams of eastern Amazon, Brazil  submitted by Felipe Rossetti de Paula  in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Forestry  Examining Committee: John Richardson, Forestry Supervisor  Stephen Mitchell, Forestry Supervisory Committee Member  Sarah Gergel, Forestry University Examiner Sean Smukler, Land and Food Systems University Examiner   Additional Supervisory Committee Members: Silvio Frosini de Barros Ferraz, Professor of Forest Hydrology, University of São Paulo, Brazil Supervisory Committee Member Brett Eaton, Geography Supervisory Committee Member      iii  Abstract  In tropical areas, deforestation and forest degradation are major threats to forests and the surrounding ecosystems, such as riparian forests and streams. These threats to riparian forests and streams in agricultural areas can be reduced by the implementation of riparian buffers. However, disturbances from cleared areas may still impact the forests due to the high edge-to-area ratio of buffers. Selective logging inside the buffers also has the potential to degrade riparian forests and streams. Forest regeneration has the potential to restore ecosystems in degraded lands, but disturbances from ongoing agriculture are expected to arrest or delay the regeneration process in the buffers. I evaluated the efficacy of two riparian buffer management strategies: (1) land abandonment for natural regeneration and (2) the maintenance of mature forest. I also sampled additional sites of different ages to evaluate where buffer treatments fit after riparian forest alterations. I hypothesized that the riparian buffers resulting from land abandonment would have less forest canopy, simpler stand structure, less large wood, higher primary production, and higher decomposition rates than a regenerated riparian forest that was surrounded by mature forest. I expected the same outcomes when comparing the riparian buffers of mature forest versus mature riparian reference sites. I found that forest structure did not differ significantly between riparian buffer management treatments, however, my ordination analysis revealed signs of forest degradation after selective logging in the riparian buffers of mature forest. I found no significant effect of riparian buffer management in any stream variable studied. Large wood was related to channel width and stem density. Stream respiration increased and primary production decreased as the regeneration process advances. Decomposition differed among species, apparently by differences on leaf structural compounds. My results show that both buffer management strategies can be effective for protection of riparian and stream ecosystem in agricultural landscapes. Land abandonment is a viable and inexpensive restoration action where ongoing disturbances are mild and the propagules for regeneration are available. While the implementation of riparian buffers of mature forests is an effective strategy, selective logging should be excluded from these areas as disturbances may intensify in the future.   iv  Lay summary  Forest clearance, selective tree removal, and forest fires are major threats to tropical forests. When these threats affect forests bordering stream ecosystems (the riparian forests), the streams are also affected. These changes are particularly important in agricultural areas, where land management techniques may be incongruent with protecting riparian forests and streams. I examined how forest attributes (number and size of trees, canopy height and openness) and stream processes (input of coarse organic material, oxygen production and consumption, and organic material decomposition) are affected in riparian forests adjacent to agriculture and compared these characteristics with riparian forests surrounded by conserved forests. I found that forest attributes and stream processes in general did not differ between riparian forest surrounded by agriculture and riparian forests surrounded by conserved forests. I concluded that current agricultural practices at my sites were not affecting the riparian forests and the stream ecosystems.      v  Preface  Research chapters of this thesis were written as a series of manuscripts for publishing in peer-reviewed journals. I was responsible for conceiving the idea, developing research questions, selecting the field sites, conducting field data sampling, conducting statistical analysis, and thesis writing. I was helped in the field by several field assistants (see acknowledgements). Dr. John S. Richardson, my supervisor, helped with developing research questions, applying the appropriate statistical analysis, interpreting the results, and the writing of my thesis. Dr. Silvio Frosini de Barros Ferraz helped with developing research questions, funding the field work, and the writing of my thesis. Members of the supervisory committee (Dr. Stephen J. Mitchell, and Dr. Brett Eaton) assisted me during the development of my research proposal and the writing of my thesis.                     vi  Table of contents  Abstract………………………………………………………………………………….....iii Lay summary………………………………………………………………………………iv Preface………………………………………………………………………………………v Table of contents…………………………………………………………………………..vi List of tables………………………………………………………………………………..ix List of figures……………………………………………………………………………….x List of abbreviations……………………………………………………………………..xvi List of symbols……………………………………………………………………………xix Acknowledgements………………………………………………………………………..xx Dedication……………………………………………………………………………….xxiv Chapter 1: Introduction……………………………………………………………………1 1.1 Stream ecosystem in the tropics ............................................................................... 1 1.2 Land-use changes in the tropics and impacts on surrounding ecosystems .......... 2 1.3 Disturbances and the successional process .............................................................. 3 1.4 Forest succession in abandoned agricultural lands ................................................ 4 1.5 Forest degradation processes (edge effects, fires, and selective logging) and its impacts of forest ecosystems ........................................................................................... 6 1.6 Riparian forest functions for streams ...................................................................... 7 1.7 Forest degradation and regeneration and its impacts on riparian and stream ecosystems ......................................................................................................................... 9 1.8 Riparian buffers as management strategy to protect riparian and stream ecosystems: are buffers really effective? ...................................................................... 10 1.9 Natural regeneration in abandoned agricultural lands as a management strategy to recover riparian forests and protect streams ........................................... 13 1.10 Buffer from maintenance of mature forests as a management strategy to protect riparian forests and streams in agricultural areas ........................................ 14 1.11 Thesis objectives and overview ............................................................................. 15 vii  Chapter 2: Short-term assessment of riparian forest structure shows recovery in regenerated buffers and signs of degradation in selective-logging buffers in eastern Amazon, Brazil……………………………………………………………………………19 2.1 Introduction .............................................................................................................. 19 2.2 Methods..................................................................................................................... 20 2.2.1. Study area ........................................................................................................... 20 2.2.2 Site selection and experimental design ................................................................ 25 2.2.3 Riparian forest and canopy structure measurements ........................................... 31 2.2.4 Statistical analysis................................................................................................ 32 2.3 Results ....................................................................................................................... 33 2.4 Discussion ................................................................................................................. 40 2.4.1 Riparian forest regeneration ................................................................................ 40 2.4.2 Effects of riparian buffer management on riparian forest structure .................... 41 Chapter 3: Effects of riparian buffer management practices on large wood in streams of eastern Amazon, Pará State, Brazil…………………………………………………...49 3.1 Introduction .............................................................................................................. 49 3.2 Methods..................................................................................................................... 50 3.2.1 Study area, experimental design, and riparian structure measurements .............. 50 3.2.2 Stream channel measurements ............................................................................. 50 3.2.3 LW and LW habitat structure .............................................................................. 51 3.2.4 Statistical analysis................................................................................................ 52 3.3 Results ....................................................................................................................... 54 3.4 Discussion ................................................................................................................. 65 3.4.1 Environmental controls of in-stream LW ............................................................ 65 3.4.2 Effects of riparian buffers management on in-stream LW .................................. 67 Chapter 4: Young secondary regeneration in riparian buffers allows fast recovery of forest structure and ecosystem metabolism in agricultural streams of Amazonia, Brazil………………………………………………………………………………………72 4.1 Introduction .............................................................................................................. 72 4.2 Methods..................................................................................................................... 74 4.2.1 Study area, sampling design, and riparian forest measurements ......................... 74 viii  4.2.2 Stream metabolism measures .............................................................................. 74 4.2.3 Stream environmental variables .......................................................................... 76 4.2.4 Statistical analysis................................................................................................ 78 4.3 Results ....................................................................................................................... 78 4.4 Discussion ................................................................................................................. 92 4.4.1 Environmental controls of stream metabolism .................................................... 92 4.4.2 Effects of riparian management and regeneration on stream metabolism ........... 93 Chapter 5: Influence of riparian forest management practices, regeneration, and leaf type on decomposition in streams of Amazonia, Pará State, Brazil…………………...99 5.1 Introduction .............................................................................................................. 99 5.2 Methods................................................................................................................... 101 5.2.1 Study area, experimental design, and data collection ........................................ 101 5.2.2 Leaf decomposition experiment ........................................................................ 101 5.2.3 Leaf chemical analysis....................................................................................... 105 5.2.4 Statistical analysis.............................................................................................. 105 5.3 Results ..................................................................................................................... 106 5.4 Discussion ............................................................................................................... 114 5.4.1 Factors affecting decomposition rates ............................................................... 114 5.4.2 Effects of riparian management and regeneration on decomposition rates ....... 118 Chapter 6: Conclusions and Recommendations……………………………………….122 6.1 Overview ................................................................................................................. 122 6.2 Riparian buffer regeneration in abandoned agricultural lands ........................ 122 6.3 Riparian buffer from maintenance of mature forests in agricultural lands .... 125 6.4 Limitations and future perspectives to better manage riparian forests and protect streams in agricultural landscapes ................................................................ 130 6.5 Management implications ..................................................................................... 135 References………………………………………………………………………………..138 Appendices……………………………………………………………………………….162     ix  List of tables  Table 2.1 Management treatment and general characteristics of the sampled streams……30 Table 2.2 Results of pairwise comparison tests among riparian management treatments. A = YR. B = YRF. C = OR. D = REF. E = YRPB. F = MFPB………....................................37 Table 2.3 Variable scores on each axis retained in the analysis…………………………...37 Table 2.4 Revised stream classification based on forest stand development after sites ordination in the PCA. Original classification is on the left and the revised classes on the right………………………………………………………………………………………...38 Table 3.1 Results of pairwise comparison tests among riparian management treatments. A = YR. B = YRF. C = OR. D = REF. E = YRPB. F = MFPB. Variables related to forest and canopy structure were tested in Chapter 2…………………………………………………55 Table 3.2 Results of model selection for LW variables…………………………..……….60 Table 4.1 Results of pairwise comparison tests among riparian management treatments. A = YR. B = YRF. C = OR. D = REF. E = YRPB. F = MFPB………………………………79 Table 4.2 Results of correlation analysis between forest and stream environmental predictors and stream CR24 and GPP. Values in bold indicate r ≥ 0.30. 2ST – 2 station. DW – 1 station downstream. UP – 1 station upstream……………………………………..87 Table 4.3 Results of model selection for respiration and primary production…………….88 Table 5.1 Results of pairwise comparison tests among riparian management treatments and leaf species. Variables related to forest and canopy structure were tested in Chapter 2, LW and channel in Chapter 3, and water quality and respiration in Chapter 4……………….107 Table 5.2 Leaf cellulose, lignin, and nutrient content for four riparian tree species……..109 Table 5.3 Results of model selection for decomposition rates. SP – leaf species. SD – stem density. RFR – riparian forest regeneration. SED – suspended sediments. DIS – discharge. TEMP – temperature. CON – conductivity. DO – dissolved oxygen. PC1RF – PCA axis 1 for forest structure. CR24 – respiration. CWI – channel width…………………………..111     x  List of figures  Figure 1.1 Simplified conceptual diagram showing forest trajectories over the years during succession (Budowski (1965) proposed classification schemes and discussed in Chazdon (2014)). Panel A indicates expected pathways of succession after riparian buffer implementation in abandoned lands. Panel B indicates expected pathways of succession after selective logging in riparian buffers. Black arrows show forest recovery in the absence of disturbances. Gray arrows show forest trajectories after the influence of disturbances. Numbers in parenthesis refer to the treatments considered in this work and described in the text………………………………………………………………………………………….17 Figure 2.1 Location of the Paragominas municipality in the Pará State, Brazil (top-left). Dashed line represents the area where the sampling sites were located and are illustrated in more detail on Figure 2.2…………………………………………………………...……...21 Figure 2.2 Location of sampling sites in the study area. The black meandering line is the Capim River, which delimits the Paragominas municipality borders………………….......22 Figure 2.3 Forest cover maps showing an example each treatment of this study. Sites are shown in higher resolution than Figure 2.2. A. REF. B. YR. C. YRF. D. OR. E. YRPB. F. MFPB………………………………………………………………………………………27 Figure 2.4 Streams representing each treatment in this study. A. REF. B. YR. C. YRF. D. OR. E. YRPB. F. MFPB…………………………………………………………................28 Figure 2.5 Diameter distribution for stems in each management treatment. YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture……………………………………….....................34 Figure 2.6 Bar graphs showing means (± 1 S.D.) of the predictor variables for each management treatment. Basal area (A) and stem density (B). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture……………………………………………………………….35 Figure 2.7 Bar graphs showing means (± 1 S.D.) of the predictor variables for each management treatment. Canopy height (A), canopy density (B), and light intensity (C). YR xi  - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture………………………….36 Figure 2.8 Ordination graph showing the distribution of sites according to management (A) and regeneration (B) treatments. YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture. SI – Secondary initial. SL – Secondary late. MF – Mature forest………………...39 Figure 2.9 Pasture under fire management used to clean the field. A) Burned fields adjacent to riparian forests. B) Surface fire advancing toward the interior of riparian forest edge, burning fine litter and tree saplings………………………………………………….44 Figure 2.10 Pasture under low intensity management with lots of shrubs and saplings growing in the grass field…………………………………………………………………..45 Figure 2.11 Harvested big diameter tree left behind in riparian area……………………...47 Figure 3.1 Bar graphs showing means (± 1 S.D.) of the predictor variables for each management treatment. Channel width (A) and discharge (B). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture……………………………………………………………….55 Figure 3.2 Diameter distribution for LW in each management treatment. YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture……………………………………….....................57 Figure 3.3 Bar graphs showing means (± 1 S.D.) of the LW variables per channel length for each management treatment. LW density (A), LW volume (B), and LW pool density (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture………………………….58 Figure 3.4 Bar graphs showing means (± 1 S.D.) of the LW variables per channel area for each management treatment. LW density (A), LW volume (B), and LW pool density (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old xii  regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture………………………….59 Figure 3.5 Relationships between LW variables (density and volume per channel length) and channel width in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○)………………………………………………………………….61 Figure 3.6 Relationships between LW pool density per channel length and channel width (A) and discharge (B) in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○)…………………………………………………...62 Figure 3.7 Relationships between LW density per channel area and channel width (A) and discharge (B) in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○)………………………………………………………………….63 Figure 3.8 Relationships between LW volume per channel area and stem density in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○)...64 Figure 3.9 Relationships between LW pool density per channel area and channel width in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○)………………………………………………………………………………….65 Figure 3.10 Cut tree in the riparian area (A) and cut LW inside the stream (B). Note the cut ends…………………………………………………………………………………………71 Figure 4.1 Bar graphs showing means (± 1 S.D.) of the predictor variables for each management treatment. Litter flux (A), suspended sediments (B), and nitrate (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old xiii  regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture………………………….80 Figure 4.2 Bar graphs showing means (± 1 S.D.) of the predictor variables for each management treatment. Conductivity (A) and water velocity (B). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture……………………………………….....................................81 Figure 4.3 Bar graphs showing means (± 1 S.D.) of respiration (CR24) for each management treatment. Two-station (A), 1-station downstream (B), and 1-station upstream (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture………………………….83 Figure 4.4 Bar graphs showing means (± 1 S.D.) of primary production (GPP) for each management treatment. Two-station (A), 1-station downstream (B), and 1-station upstream (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture………………………….84 Figure 4.5 Bar graphs showing means (± 1 S.D.) of net ecosystem productivity (NEP) for each management treatment. Two-station (A), 1-station downstream (B), and 1-station upstream (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture………………………85 Figure 4.6 Bar graphs showing means (± 1 S.D.) of primary production/respiration ratio (GPP/CR24) for each management treatment. Two-station (A), 1-station downstream (B), and 1-station upstream (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture……………………………………………………………………………………...86 Figure 4.7 Relationships between respiration (CR24) and conductivity (A) and discharge (B) in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young xiv  regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○)………………………………………………………………………………….89 Figure 4.8 Relationships between respiration (CR24) and conductivity (A) and forest structure (B) in 24 streams. Codes and colors in A: SI – Secondary initial (Δ, red line). SL – Secondary late (+, black line). MF – Mature forest (○, green line).Codes in B: YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○)……………..90 Figure 4.9 A - Bar graphs showing means (± 1 S.D.) of gross primary production (GPP) for levels of the regeneration treatment. SI – Secondary initial. SL – Secondary late. MF – Mature forest. B - Relationship between GPP and forest structure in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○)……………..91 Figure 4.10 Riparian canopy opening in two selective logging sites by (A) fallen trees in the riparian area and (B) recent fallen trees spanning the stream channel…………………98 Figure 5.1 Leaf samples for the four tree species selected in this study. A. Macrolobium angustifolium (Fabaceae). B. Licania sp. (Chrysobalanaceae). C. Siparuna bifida (Siparunaceae). D. Henriettea succosa (Melastomataceae). Pen size (15 cm) is used for scale……………………………………………………………………………………….104 Figure 5.2 Bar graphs showing means (± 1 S.D.) of the decomposition rates for each management treatment. Species H. succosa (A), S. bifida (B), M. angustifolium (C), and Licania sp (D). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture……………………..108 Figure 5.3 Bar graphs showing means (± 1 S.D.) of decomposition rate for each leaf species…………………………………………………………………………………….109 Figure 5.4 Relationships between riparian forest structure and k - He (A), k – Ma (B), and k – Li (C) in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer xv  of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○)………………………………………………………………...112 Figure 5.5 Bar graphs showing means (± 1 S.D.) of the decomposition rates for each regeneration group. Species H. succosa (A), S. bifida (B), M. angustifolium (C), and Licania sp. (D). MF – Mature forest. SI – Secondary initial. SL – Secondary late………………………………………………………………………………………...113 Figure 5.6 Leaf colonization by leaf miners Chironomidae. A. Macrolobium angustifolium leaf with signs of colonization. B. Chironomidae larvae (in red) living inside of a M. angustifolium leaf. C. Chironomidae pupae living inside of a H. succosa leaf. D. Chironomidae larvae collected from the inside and surface of leaves. All pictures were taken after a 15-day incubation period in the stream from a field experiment in 2014…..116 Figure 5.7 Suspended sediments and FPOM in one YRPB stream. A. Stream reach. B. Bag smothering (bag location in the stream bottom is indicated by the red arrow). C. Leaf bag with FPOM adhered on its surface………………………………………………..............117 Figure 5.8 Leaves of M. angustifolium retrieved after normal conditions (A) and after bag smothering (B). Retrieval after 15 days from a field experiment in 2014. Tweezer size (11.5 cm) is used for scale………………………………………………………………………118 Figure 6.1 Revised and simplified conceptual diagram showing forest trajectories over the years after buffer implementation. A. The regeneration of forest structure in the buffer is similar to natural succession (Budowski (1965) proposed classification schemes and discussed in Chazdon (2014)) for the first 8-12 years of regeneration under mild disturbances. B. Riparian forests lost structure in the buffer after selective logging and have structural measures similar to older regenerated sites (18 – 22 years). Black arrows show forest recovery in the absence or under mild disturbances and riparian forest degradation after selective logging. Gray arrows show expected forest trajectories after the influence of disturbances. Letter X in red represents trajectories not observed in the buffers. Dotted arrows represent expected but undetermined future trajectories in this study. Numbers in parentheses refer to the treatments considered in this work and described in Chapter 1……………………………………………………………………………………...........128   xvi  List of abbreviations  df    Degrees of Freedom AFDM   Ash-Free Dry Mass AICc    Akaike Information Criteria Corrected for Small Sample Sizes ANOVA   Analysis of Variance BA    Basal Area CANO   Canopy Density CON    Conductivity CPOM   Coarse Particulate Organic Matter CR24    Community Respiration CWI    Channel Width DBH    Diameter at Breast Height DIS    Discharge. DO    Dissolved Oxygen DOC    Dissolved Organic Carbon  DOM    Dissolved Organic Matter DW    1-Station Downstream. FAO    Food and Agriculture Organization FPOC    Fine Particulate Organic Carbon FPOM   Fine Particulate Organic Matter GLM    Generalized Linear Models GLMM   Generalized Linear Mixed Models  GPP    Gross Primary Production He    Henriettea succosa HEIG    Canopy Height ICMS    Imposto Sobre Circulação de Mercadorias IPAM    Instituto de Pesquisa Ambiental da Amazônia IMAZON   Instituto do Homem e Meio Ambiente da Amazônia Li    Licania sp. LIGHT   Light Intensity xvii  LIT    Litter Ln    Natural Logarithm LW    Large Wood LWDE   Large Wood Density LWPDE   Large Wood Pool Density LWVO   Large Wood Volume Ma    Macrolobium angustifolium  MF    Mature Forest MFPB   Buffers of Mature Forests Surrounded by Pasture NEP    Net Ecosystem Productivity NIT    Nitrate OR    Old Regeneration PC1RF   PCA Axis 1 for Riparian Forest Structure PCA    Principal Components Analysis PCA 1   Principal Components Analysis Axis 1 PCA 2   Principal Components Analysis Axis 2 REDD   Reducing Emissions from Deforestation and Forest Degradation REF    Reference RFM    Riparian Forest Management RFR -    Riparian Forest Regeneration RIL    Reduced Impact Logging SD    Stem Density S.D.    Standard Deviation SED    Suspended Sediments Si    Siparuna bifida  SI    Secondary Initial SL    Secondary Late TEMP   Temperature UP    1-Station Upstream UN    United Nations USA    United States of America xviii  VEL    Water Velocity YR    Deforestation or Young Regeneration YRF    Young Regeneration Surrounded by Forest YRPB    Buffers of Young Forest Regeneration Surrounded by Pasture 2ST    Two Station    xix  List of symbols  Ca    Calcium k    Decomposition Rate K    Potassium KO2    Reaeration Coefficient Mg    Magnesium N    Nitrogen NaCl    Sodium Chloride O2    Oxygen P    Phosphorous R2adj    Adjusted Square-Root S    Sulfur     xx  Acknowledgements  I would like to thank my supervisor Dr. John S. Richardson by having accepted me in his lab and also by having embraced my research ideas. I am also very thankful for his guidance during the PhD, his support during the challenging times in the field, and especially, for his huge effort to review my thesis in a short period of time and his patience with me during the final steps of the PhD. I am really appreciative of all the experiences that John’s lab provided me at UBC. Thanks John!  I thank Dr. Silvio Frosini de Barros Ferraz by keeping his lab open to me when I decided to pursue a PhD abroad and by keeping a partnership with me during my PhD. I also thank Dr. Silvio for having embraced my ideas (since my undergraduate studies) and for submitting the funding proposal to The São Paulo Research Foundation (FAPESP) to conduct my research in the Amazon region.  I am very thankful to Dr. Pedro Gerhard for making all the necessary arrangements so I could carry out my work in the facilities of the Brazilian Agricultural Research Corporation (Embrapa Amazônia Oriental) and also the arrangements to give me access to the study area. I am also very thankful to he and his wife Naiane Sangaletti for providing me a home during my stay in Belém. I also thank Dr. Pedro for having embraced my ideas since we became friends during his PhD.  I thank my committee members Dr. Stephen J. Mitchell and Dr. Brett Eaton for their important comments and edits on my thesis.  I thank Silvia Sena and her daughters (Samia and Suelen) for providing me a home during my stay in Belém after Pedro and Naiane returned to São Paulo.  I thank Dr. Ademir Roberto Ruschel, Dr. Juliana Feitosa Felizzola, and Neusa Maria Ferreira (lab technician) for assisting my work in the Brazilian Agricultural Research Corporation (Embrapa Amazônia Oriental). xxi   I am very thankful to all the people in CKBV Florestal Ltda, especially Francisco Sanches, Karen Juliana dos Anjos, and Josue Evandro Ribeiro Ferreira, for giving me access to the field area, providing people to help me in the field, and providing housing in the field station during all my field work in Paragominas.  I thank the staff of the Tropical Forest Institute in Paragominas, especially Iran Paz Pires, for providing housing in the field station during some periods of my field work.  I thank all farmers in Paragominas that gave me access to work on their property.  I am very thankful to all my field assistants (Josinete Monteles, Jefferson Santos, Erick Manzano Macias, Ewerton Sena, Lucio Davi Moraes Brabo, and Wendel Ferreira Reis) for helping me in the field, and most importantly, for their friendship, for never have complained about the bologna sandwiches with juice in most of our lunches, and by cheering me up when the car broke in the field several times, when the car got stuck in the mud, and all the other issues that I had there.   I thank Ednaldo Augusto Pinheiro Nascimento and Miguel Pastana do Nascimento (field botanists) for helping me with the riparian forest inventory and providing great moments in the field. These two did complain about the bologna sandwiches with juice, but be honest guys, you miss that!  I thank Dr. Toby Gardner and people from Instituto do Homem e Meio Ambiente da Amazônia (IMAZON) for providing access to the land-use maps and other digital materials from the study area.  I thank all the people from the Forest Hydrology Laboratory / University of São Paulo, especially the PhD student Aline Aparecida Fransozi for her amazing work on managing the financial resources of our project. I also thank Dr. Ricardo Taniwaki for his help with materials while I was in the field, Dr. Maira Ometto Bezerra for her help with xxii  oxygen probes handling and other guidances, and also the M.Sc. Gláucia Regina Santos for exchanging ideas about the decomposition experiment.  I thank Dr. Hugo Henrique Saulino from the Federal University of São Carlos for doing the Chironomidae larvae identification for our project.  I thank the Brazilian National Council for Scientific and Technological Development (CNPq) for funding my studies at UBC and The São Paulo Research Foundation (FAPESP) for funding my field research in the Amazon. I also thank Dr. Gene Namkoong and his family for endowing the NAMKOONG Family Fellowship in Forest Sciences for a graduate student in the Department of Forest and Conservation Sciences, which I was a recipient in 2015-2016.  A special thanks to my Brazilian friends here in Vancouver: Marina Giacomin, Julia Varela, Catarina Wor, Juliana Magalhães, Vinicius Lube, Rafael Candido Ribeiro, Leilane Ronqui, Alexandre Bagdonas Henrique, Climerio Silva Neto, Paulo Guilherme Molin, Flavia Polati Ferreira, and others.  I thank all the people in my office with whom I shared great moments here at UBC over the years.  Finally, I would like to thank all the members of Stream and Riparian Research Laboratory over the years, especially Alex Yeung and Sean Naman. Thanks everyone for the friendship!  In all these years of PhD, I had the opportunity to meet several amazing people in the different cities that I lived during my activities (Vancouver, Piracicaba, Belém, Goianésia do Pará, and Paragominas). All these people contributed in some way for this work to have happened. From the respectful and friendly roommates that I shared houses with here in Vancouver, the cooks that happily served my meal every night in Paragominas, and the gas station employees in Goianésia that always treated me as a friend, I am very xxiii  thankful to all of them for having provided moments of happiness, which were important to keep me doing the hard work.       xxiv                       Dedication This thesis is dedicated to my parents Carlos and Margarida, my brother Gustavo, and my sister Julia who provided me with love and support during all important steps of my life.      1  Chapter 1: Introduction  1.1 Stream ecosystems in the tropics  Freshwater habitats in tropical regions are characterized by having high biodiversity, high productivity levels, and high habitat complexity. Freshwater habitats in the tropics are home of an enormous biodiversity, especially of fishes (Lowe-McConnell, 1975; Helfman, 1997; Berra, 2001; Moyle and Cech, 2004). The South American region has the richest fish fauna in the world (Bonetto and Wais, 1995; Lewis et al. 1995; Helfman, 1997; Moyle and Cech, 2004). Most of the documented fish species inhabit larger rivers while for stream ecosystems most of the species are still poorly documented (Lewis et al. 1995; Buckup et al. 2007).  Stream ecosystems in the tropics have warmer waters and are also fairly oligotrophic (Boulton et al. 2008). Thus, they are expected to have high decomposition rates due to increased metabolism of fungi and bacteria, the most important organisms responsible for leaf decomposition in some tropical streams (Rueda-Delgado et al. 2006; Ardón and Pringle, 2008; Wantzen et al. 2008; Boyero et al. 2009). However, leaf-mining Chironomidae larvae and large benthic omnivores (shrimps and fishes) are also important for leaf breakdown after leaf conditioning by microorganisms (Henderson and Walker, 1986; Wantzen et al. 2008; Boyero et al. 2009; Chará-Serna et al. 2012; Biasi et al. 2013; Tanaka et al. 2016). Tropical streams seem to have a large variety of organisms involved in leaf processing when compared to temperate streams, where leaves are mostly processed by detritivorous insects (Wantzen et al. 2008; Boyero et al. 2009; Lecerf and Richardson, 2010).  As in temperate streams, large wood is also an important structural element that increases habitat complexity in tropical streams. But in addition to forming habitat units (pools; Paula et al. 2011) that provide cover and food for fishes (Wright and Flecker, 2004; Lujan et al. 2011), LW is home for some specialist species of catfishes that live and complete their life cycles in the holes created in LW of hardwood tree species (Power, 2  2003). Although the high importance of LW for the aquatic community in the tropics, the dynamics of this material in the channel are still poorly understood and not very well studied compared to LW dynamics in temperate streams.  Tropical rivers and streams are very threatened ecosystems around the world, putting at high risk the rich biodiversity that inhabits these environments (Dudgeon, 2000; Ramírez et al. 2008). The main threat for these ecosystems is deforestation driven by land use changes, but water diversion for agriculture and human consumption is also a concern (Dudgeon, 2000; Ramírez et al. 2008; Leal et al. 2016; Paula et al. 2018; Taniwaki et al., 2018). These threats deteriorate stream habitat conditions and water quality and reduce water availability and stream connectivity, threatening the survivorship of the aquatic organisms (Ramírez et al. 2008; Brejão et al. 2017; Leitão et al. 2017; Leal et al. 2018; Paula et al. 2018). Although the increasing pressure on lotic ecosystems by humans, there is little work being done to document the ecology and biodiversity of tropical streams, making difficult the conservation efforts to protect streams (Boyero et al. 2000; Dudgeon, 2000), especially in landscapes with increased rates of land use changes like in the Amazon region (Ferraz et al. 2005; Gardner et al. 2013). Studies conducted in different regions of the Brazilian Amazon showed that even small changes of land use at different spatial and temporal scales are enough to degrade streams and harm the aquatic biodiversity (Brejão et al. 2017; Leitão et al. 2017; Brito et al. 2018; Leal et al. 2018).  1.2 Land-use changes in the tropics and impacts on surrounding ecosystems  Tropical forests have been extensively modified by anthropogenic activities such as deforestation (i.e. conversion of native forests to another land-use) and forest degradation (i.e. forest exposures to fires, logging, edge effects, woodfuel collection, hunting; Nepstad et al. 1991; Chazdon, 2014; Laurance et al. 2014; Barlow et al. 2016). In Brazil, the country with the greatest annual net loss of forest area (984 km2, 0.2%) in the world (FAO, 2016), the deforestation process starts with the logging of the most valuable tree species followed by forest clearance for pasture or slash-and-burn farm fields establishment (Nepstad et al. 1991; Uhl et al. 1991). Recently, these areas (especially pastures) are being 3  replaced by large soybean plantations (Gardner et al. 2013). The remaining forests in these agricultural areas are usually degraded by logging and also by edge effects (Uhl et al. 1991; Laurance et al. 2000; Laurance et al. 2006a; Barlow et al. 2016). Eventually, some agricultural areas will be abandoned and the regeneration process will take place at different rates depending on previous land-use intensity, soil nutrients, and propagule availability (Uhl, 1987; Uhl et al. 1988; Nepstad et al. 1991).  Deforestation and forest degradation have large impacts for riparian and stream ecosystems (Paula et al. 2011; Leal et al. 2016; Brejão et al. 2017; Leitão et al. 2017), and the most common management strategy used to reduce these impacts in managed areas is the implementation of riparian buffers (Richardson et al. 2012). However, riparian buffers may be susceptible to disturbances and degradation due to their great edge-to-area ratio (Grizzel and Wolff, 1998; Kiffney et al. 2003; Moore et al. 2005; Bahuguna et al. 2010; Braithwaite and Mallik, 2012). In addition to edge effects, selective logging in riparian areas also has the potential to degrade riparian forests and affect stream processes (Lecerf and Richardson, 2010; Mallik et al. 2014). Forest regeneration, on the other hand, has the potential to restore ecosystems in degraded lands, including riparian forests (Chazdon, 2008; González et al. 2016, 2017). However, the disturbances from ongoing agriculture are expected to arrest or delay the regeneration process in the riparian buffers and further increase degradation. In this thesis, I evaluated the efficacy of two riparian buffer management strategies (land abandonment for riparian forest regeneration and implementation of riparian buffer of mature forest) in agricultural landscapes intended to protect riparian and stream ecosystems.   1.3 Disturbances and the successional process  A disturbance is an event that disrupts the structure of ecosystems, communities, populations, and alters the availability of resources and the physical environment (White and Pickett, 1985). Early ecological studies considered undisturbed ecosystems as largely closed systems dominated by internal recycling of elements, self-regulation and deterministic dynamics, stable endpoints, and free of human influence (Clements, 1916; 4  Braun, 1935). Clements (1916) suggested that, in the absence of disturbance, plant community development was a predictable process that progressed to a single stable community-type characteristic of a particular climate (the climate climax also called monoclimax). Later in ecological science, this equilibrium view was replaced by the nonequilibrium or multiequilibrium perspective, which recognized that most ecosystems exhibit unbalanced inputs and losses, their dynamics are influenced by varying external and internal forces, they exhibit no single stable equilibrium, the disturbances are a natural component of their dynamics, and human activities exert a pervasive influence (Whittaker, 1953; White, 1979).  After disturbance, ecosystems undergo succession, a directional change in ecosystem composition, structure, and functioning (Budowski, 1965; Guariguata and Ostertag, 2001; Franklin et al. 2002; Chazdon, 2014). The ecosystem trajectory after a disturbance will depend both on characteristics of the disturbance (size, intensity, duration, and frequency) and the properties of the ecosystem, such as resilience and stability (Holling, 1973; Chapin et al. 2011; Clewell and Aronson, 2013). If the disturbance is not severe, ecosystems can return to their pre-disturbance conditions (Holling, 1973; Chapin et al. 2011; Hodgson et al. 2015). However, if a disturbance is too severe and goes beyond the ecosystem’s resilience, the ecosystem may shift to an alternative stable state that can persist for a long time even if the disturbance is no longer acting (Holling, 1973; Chapin et al. 2011). Usually, this alternative stable state sustains less biodiversity and diminished rates of ecological processes compared with the previous ecosystem (Clewell and Aronson, 2013). One example of this condition can be observed in tropical forests of the Amazon region, where the biodiversity rich primary forests are replaced by weedy grasses and shrubs in degraded lands after severe pasture management (Uhl et al. 1988; Nepstad et al. 1991).  1.4 Forest succession in abandoned agricultural lands  After deforestation followed by land abandonment, the regeneration process in forest ecosystems begins with the colonization of pioneer plants like grasses and shrubs. As time advances, trees start to colonize the area. In this forest ecosystem, the plant 5  community structure and composition is very simple, characterized by a single-layer canopy and low canopy height, low basal area, lower rates of ecological processes (low organic matter accumulation and decomposition, and low nutrient cycling), and a small number of species adapted to high light and poor soil conditions. As the regeneration process proceeds, this initial stage is replaced by second-stage vegetation composed primarily of tree species. At this stage, the vegetation structure still has high stem density (tree trunks) and low basal area, but presents some level of complexity (some high diameter trees and a tall, closed, and homogeneous canopy). The community composition has more tree species than the previous stage through the establishment of shade-tolerant species and further development of ecological processes. Finally, the regeneration process ends in a “climax” community with a highly complex forest structure (high basal area and high, dense, multi-layer canopy), a rich biodiversity (high diversity of tree and other life forms), and complex ecological interactions (Budowski, 1965; Guariguata and Ostertag, 2001; DeWalt et al. 2003; Chazdon, 2014). Theoretically, at this end point of succession, the community will reach a “climax” state that will be intermittently altered by gap dynamics or eventually reset by a larger disturbance (Budowski, 1965; Guariguata and Ostertag, 2001; Chazdon, 2014).   In tropical landscapes, forest regeneration is an important component of the land use dynamics (Fearnside, 1996; Houghton et al. 2000; Ferraz et al. 2005). However, forest regeneration may be hindered by ongoing disturbance in many agricultural landscapes (Gerwing, 2002; Griscom et al. 2009), and by land degradation (soil compaction, nutrient loss, and low propagule availability) caused by the duration and the intensity of past land use (Uhl et al. 1988; Nepstad et al. 1991). In the Brazilian Amazon, pasture management has intensified in some areas, with periodic additions of fertilizers, weed control, and site bulldozing (Uhl et al. 1988; Nepstad et al. 1991; Figueiredo et al. 2010). These pastures were later abandoned, but due to their historically intense management, the regeneration process was compromised and did not progress toward an advanced successional stage, generally stopping at initial successional stages. These areas are now called old fields and they are dominated by weedy grasses and shrubs. In these areas, there are numerous impediments to tree establishment (low soil nutrients and propagule availability, seed 6  predation, seedling predation, and root competition with old field vegetation), increasing the chance that forest recovery will not take place (Uhl et al. 1988; Nepstad et al. 1991). In another study, Zarin et al. (2005) found that repeated fires (1 to 10 cycles of burning and abandonment) decreased forest regeneration and carbon accumulation in slash-and-burn farms in the Amazon.   1.5 Forest degradation processes (edge effects, fires, and selective logging) and its impacts of forest ecosystems  Deforestation reduces forest cover in the landscape and the remaining forest areas become more subject to disturbances originating from the surrounding cleared areas (Laurance et al. 2000; Turner et al. 2001; Laurance et al. 2006a; Barlow et al. 2016). These disturbances include fire, light, and wind, and once they affect the forest patch, they change the environmental conditions inside the patch, especially at the edges. With the increase in local light incidence and wind, microclimate will become warmer and drier, causing the death of species more adapted to the cooler and more humid environment found in the forest interior (Murcia, 1995; Gehlhausen et al. 2000; Laurance et al. 2000; Cochrane and Laurance, 2002; Schwartz et al. 2017). These new conditions also accelerate the growth of early successional tree species and woody climbing plants (Chen et al. 1992; Murcia, 1995; Gehlhausen et al. 2000; Laurance et al. 2006a; Jones et al. 2017). In tropical areas, woody climbing plants are an additional factor in the death of remaining trees and tree fall during wind periods (Putz, 1990; Laurance et al. 2000). With the increased mortality rates of bigger trees and enhanced establishment rates of small-diameter, secondary tree species, these degraded forests often have increased stem density, reduced forest biomass, reduced canopy height and cover, and reduced tree species diversity (Laurance, 2000; Laurance, 2006a). In contrast to edge effects that are persistent disturbances (press disturbances), fires are more variable on the temporal scale of a year (pulse disturbances; usually occurs once in a year in agricultural lands of Brazil, during pasture renovation). Fires cause the same degradation processes described above, however, they may be exacerbated by the dryer environment and the excessive accumulation of litter material on the forest ground (Barlow and Peres, 2004; Laurance, 2006b). These changes of forest structure, composition, and 7  environmental conditions due to edge effects can also be observed further away from patch edges as different edge effects may penetrate 40 to 100 m towards forest interior areas (Laurance et al. 1998; Laurance et al. 2006b).  Selective logging is another source of forest degradation (Uhl et al. 1991; Chazdon, 2014; Barlow et al. 2016). The removal of targeted, timber-valuable tree individuals contributes to reduced tree biomass, species richness, and other changes at the stand level (Chazdon, 2014; Mallik et al. 2014; Martin et al. 2015; Richardson and Peres, 2016). It also creates canopy gaps and reduces canopy height and density, further increasing light incidence and also making the microclimate warmer and drier (Chazdon, 2014; Mallik et al. 2014). These changes to forest stands accelerate the growth of early successional tree species, altering the structure and composition of the plant community (Chazdon, 2014; Mallik et al. 2014). These changes in forest stands conditions created by selective logging may also exacerbate fire effects on these forests (Barlow and Peres, 2004; Laurance, 2006b; Barlow et al. 2016). The level of forest degradation will vary with the logging technique being employed, with the highest level in intensive logging operations (large amount of trees removed, damage on remaining trees during the operations, and soils disturbed by machinery) and lowest using Reduced Impact Logging (RIL; Uhl et al. 1991; Chazdon, 2014; Martin et al. 2015). In those operations using RIL, changes of forest structure and composition will resemble gap dynamics (canopy gaps created by tree fall followed by the growth of already established saplings and low colonization of secondary tree species), while those using intensive techniques will resemble early stages of succession (stand dominated by colonization of secondary tree species; Chazdon, 2014; Martin et al. 2015).  In this study, I employ the term forest degradation to refer to changes on forest structural attributes (basal area, stem density, canopy height and density, high light intensity) caused by logging and edge effects.  1.6 Riparian forest functions for streams  8  Riparian habitats are complex ecosystems that support high levels of biodiversity and perform several ecological functions for the surrounding ecosystems (Naiman et al. 2005; Kuglerová et al. 2014; Ramey and Richardson, 2017). In forested landscapes, the forest canopy covers entirely the small channels, controlling light incidence and primary productivity (Warren et al. 2016), and is also a control on the rates of organic matter inputs to streams (Naiman and Décamps, 1997; Richardson and Danehy, 2007; Kiffney and Richardson, 2010; Tank et al. 2010). Riparian forests also supply important food and structural resources for the aquatic environment, like coarse particulate organic matter (CPOM) and large pieces of wood (Naiman et al. 2005; Paula et al. 2011).   Leaves are the most important fraction of CPOM and in some studies it comprises 61-65% of the total CPOM delivered to the stream during the autumn in the temperate zone (Allan and Castillo, 2007). Once in the channel, leaves start to leach, releasing some dissolved organic matter (DOM: ≤ 0.7 μm) into the stream water, followed by the colonization of fungi and bacteria that start the decomposition process for further processing by shredders (Webster et al. 1999). Decomposition is a key process in streams and is affected by biological (leaf characteristics and invertebrate community; Gonçalves et al. 2007; Tanaka et al. 2016) and environmental factors (stream discharge, water temperature, and nutrients; Gulis and Suberkrop, 2003; Rueda-Delgado et al. 2006; Ferreira and Chauvet, 2011a and b; Martins et al. 2018).   The structure and composition of riparian forests also is important for the decomposition process because it can affect the amount and composition of leaves that enter the channel (Kiffney and Richardson, 2010; Hoover et al. 2011; Kominoski et al. 2011; Bilby and Heffner, 2016) and also the detritivore fauna that lives in streams (Tanaka et al. 2016). It may also influence the decomposition rate by controlling water temperature (Kiffney et al. 2003) since some experimental studies have shown that increased temperatures increased decomposition by microbial activity (Ferreira and Chauvet, 2011a and b; Piggott et al. 2015; Martins et al. 2018). As the amount of light can limit productivity of the primary producers and their support of the food chain, these streams are largely dependent on the processing of allochthonous organic material from the 9  surrounding forest to provide the energy necessary for the upper trophic levels (Naiman and Décamps, 1997; Richardson and Danehy, 2007; Tank et al. 2010).  Large wood (LW - pieces ≥ 10 cm in diameter and ≥ 1 m in length), such as trees, branches, and roots, is important for stream ecosystem structure and function (Andrus et al. 1988; Lamberti and Gregory, 1996; Bilby and Bisson, 1998). It affects channel width and depth, and creates channel habitat units (pools and cascades) (Bisson et al. 1987; Rosenfeld and Huato, 2003; Elosegi et al. 2017), which provides refuges against strong currents and predators, and an important place to settle for some organisms (Sedell et al. 1990; Bilby and Bisson, 1998; Boss and Richardson, 2002; Benke and Wallace, 2003). Large wood is an important food resource for some aquatic invertebrates (Anderson et al. 1978; Eggert and Wallace, 2007; Valente-Neto et al. 2015), which later will become food for higher trophic levels (Lujan et al. 2011). LW also acts as sediment barriers during periods of high flow, facilitating sediment deposition and substrate sorting which can provide feeding and reproduction sites for fishes and other organisms as well (Maser and Sedell, 1994; Bilby and Bisson, 1998).  1.7 Forest degradation and regeneration and its impacts on riparian and stream ecosystems  In degraded and regenerated areas, despite having residual canopy, the reduced canopy height and density may increase sunlight and also the amount and quality (high carbon/nitrogen ratio) of litter reaching the stream channel (Bunn et al. 1999; Kiffney et al. 2003; Kiffney and Richardson, 2010; Bilby and Heffner, 2016; Kaylor and Warren, 2017). The reduction of canopy structure increases sunlight and temperature in the most open stream reaches, increasing primary production and shifting the trophic status of the ecosystem from less heterotrophic to more autotrophic (Bunn et al. 1999; Bechtold et al. 2017). This shift will cause great changes in stream community structure and composition when primary production substantially exceeds respiration. For example, the primary producer communities are constrained by shading, have low abundance, and are mostly composed of palatable unicellular algae (diatoms). When the forest canopy is completely 10  removed or extremely degraded, light is no longer a limiting resource so the community may shift to high abundance of mostly less palatable filamentous green algae and macrophytes (Bunn et al. 1999; Mosisch et al. 2001; Kiffney et al. 2003). When coupled with other impacts from the deforested areas, like fertilization and nutrient export to the stream waters, this shift of stream trophic status can ultimately lead to serious deterioration of water resources if extreme events of eutrophication occurs, which can be extremely harmful for the aquatic and terrestrial fauna and also for humans (Carpenter et al. 1998; Allan and Castillo, 2007; Martinelli and Filoso, 2008). Also, the increase of water temperature and the input of fast-growing leaves may accelerate the decomposition rates and increase the energy and nutrient availability for the stream ecosystem (Mckie and Malmqvist, 2009; Kominoski et al. 2011; Kominoski et al. 2013). Therefore, it is important to protect riparian forests in agricultural landscapes in order to maintain the heterotrophic status of stream ecosystems and also conserve the stream biodiversity and water resources.  Changes in the environmental conditions in degraded and regenerated riparian areas increase the number of small secondary trees that later will fall into the streams, and therefore affect LW inputs (Andrus et al. 1988; Bilby and Ward, 1991; Paula et al. 2011). Small LW (10 to 40 cm in diameter; considering the probability of forming pools according to Rosenfeld and Huato (2003)) will still be supplied by these forests. However, the higher rates of decomposition and transport of this material in the channel, and the absence of future, larger LW sources from the riparian areas will decrease the amount of wood inside streams (Andrus et al. 1988; Warren and Kraft, 2008; Cadol and Wohl, 2010; Paula et al. 2013). These changes in LW will reduce the stream’s structural complexity as habitat and cover will be provided mostly by small wood and only temporarily available for channel structuring, disappearing from the reach when the wood is completely gone (Rosenfeld and Huato, 2003; Paula et al. 2011; Paula et al. 2013).  1.8 Riparian buffers as management strategy to protect riparian and stream ecosystems: are buffers really effective?  11  The most common riparian management strategy in logging and agricultural areas is the implementation of riparian buffers (Kiffney et al. 2003; Gomi et al. 2006a; Brasil, 2012; Richardson et al. 2012). The main objectives of these forested buffers are to protect water resources, the riparian and stream ecosystem, and their biodiversity (Richardson et al. 2012). Riparian forested buffers are designed to reduce the amount of sediments and nutrients input to the streams in agricultural and timber logging landscapes (Gomi et al. 2006a; Richardson et al. 2012). These buffers are also important to maintain the supply of LW and the fluxes of light and organic material to the stream, and hence, provide habitat structure and control primary production and keep the heterotrophic status of stream ecosystems, respectively (Kiffney et al. 2003; Gomi et al. 2006a; Bahuguna et al. 2010; Kiffney and Richardson, 2010; Giling et al. 2013). In general, buffer widths vary according to stream size, and for headwater streams ≤ 5 m wide, the buffer width can vary from 5 to 30 m in Brazil (Brasil, 2012). In other countries, riparian buffer widths vary according to different purposes (stream size, presence of economically valuable fish species), and may not even be mandatory (Lee et al. 2004; Cristan et al. 2016; Ring et al. 2017). Around the world, forested buffers can result from the maintenance of mature forest (Kiffney et al. 2003; Braithwaite and Mallik, 2012; Brasil, 2012), active planting of native tree species (Brasil, 2012; Giling et al. 2013; Brancalion et al. 2014), and land abandonment for forest regeneration (Brasil, 2012; Nunes et al. 2015).   Riparian buffers in Brazil were first implemented in 1965 with the approval of the Brazilian Forest Code (Brasil, 1965). Within this legislation, all streams ≤ 10 m wide located in agricultural areas were required to have 30 m wide riparian buffers (Permanent Preservation Areas - PPA) with native forest along the entire drainage network. Waterbodies larger than 10 m were required to have wider buffers. The mechanisms addressed in the legislation to implement the buffers were leaving a strip of native vegetation around the streams on newly cleared areas, while for those riparian areas that were deforested in the past, they should be restored by natural regeneration or active tree planting. Although the Forest Code is an important instrument to promote forest conservation in agricultural areas of Brazil, its implementation proved challenging to enforce (Soares-Filho et al. 2014). For example, riparian areas are still deforested even after 12  the legislation approval; riparian buffers left on stream margins are selective logged (selective logging is prohibited in these areas); deforested riparian areas are still used as agriculture/pastures; deforested riparian areas are abandoned for natural regeneration without any management plan that monitor the success of the regeneration process (Soares-Filho et al. 2014; Nunes et al. 2015). All these situations restrict the role of riparian buffers in protecting riparian and stream ecosystems in agricultural landscapes and to date, there are still few studies to evaluate the negative impacts that riparian alterations may have to the surrounding ecosystems and the effectiveness of different buffer management strategies to protect riparian and aquatic ecosystems in Brazil (Paula et al. 2011; Brancalion et al. 2014; Tanaka et al. 2016; Hunt et al. 2017; Taniwaki et al. 2017; Leal et al. 2018; Brito et al. 2018). Recently, modifications of the Brazilian Forest Code reduced the buffer width to be reforested (5 to 15 m depending of the size of the property; property size classes vary according to the regions in the country (Brasil, 2012)). This change in riparian buffers may have strong ecological implications for riparian and stream ecosystems because most Brazilian agricultural landscapes will have less protection provided by the riparian forests and may see their ecological functions decrease with smaller forested buffer width (Wenger, 1999; Kiffney et al. 2003).  Another concern is that in managed landscapes, forested riparian buffers are constantly exposed to the disturbances from the surrounding area (Grizzel and Wolff, 1998; Kiffney et al. 2003; Moore et al. 2005; Bahuguna et al. 2010; Braithwaite and Mallik, 2012). Considering that riparian forests in these areas are composed of small buffer strips (generally 30 m in width), these forests are subject to the same disturbances present in forest edges, and hence, the structure, species composition, and ecological processes of these riparian forests are expected to be similar to that in disturbed forest edges. These disturbances are particularly important in agricultural areas where buffers will be exposed for a long period of time compared to buffers in timber logging areas where the cleared area will be abandoned or even replanted for forest regeneration, and hence, reducing and further eliminating the impacts on the buffers (Franklin et al. 2002; Yeung et al. 2017). For example, pasture renovation using fire is a common and frequent action to improve the productivity of pastures in Amazonia, however, fire use is not well managed and the fire 13  can spreads to the adjacent forest fragments and riparian forests (Gerwing, 2002; Laurance, 2006b). In addition to fires, there is plant trampling by cattle that enter the forest, killing tree seedlings and affecting tree recruitment (Griscom et al. 2009). Hence, the negative impacts on the riparian forested buffers in are expected to affect the ecological processes of streams.   In this study, I evaluated the two most employed riparian buffer management strategies in Brazil: (1) land abandonment for natural regeneration and (2) the maintenance of mature forest.  1.9 Natural regeneration in abandoned agricultural lands as a management strategy to recover riparian forests and protect streams  Land abandonment for natural forest regeneration is one strategy to recover ecosystems (Chazdon, 2008; Clewel and Aronson, 2013), including the riparian forests in areas of agriculture in order to protect water resources and stream ecosystems (Carline and Walsh, 2007; Brasil, 2012; Forget et al. 2013; González et al. 2016, 2017). However, it is important to highlight that this type of restoration action does not perform well in all situations, especially in areas where soil fertility was exhausted and regeneration sources were depleted (Uhl et al. 1988; Zarin et al. 2005). Also, the ongoing agriculture can also be a source of disturbances that may delay or arrest the regeneration process in the surrounding abandoned areas, i.e. through edge effects (Laurance et al. 2000; Vasconcelos and Luizão, 2004; Laurance, 2006a). Depending on the disturbance intensity and frequency, and the ecological processes considered, edge effects can be observed further into the patch from the edges (Laurance et al. 1998; Laurance et al. 2006b), so riparian buffers may be very exposed to these disturbances.  Several studies have addressed the effects of past and present land use on the regeneration process, but all of them focused on upland forests (Uhl et al. 1988; Zarin et al. 2005; Griscom et al. 2009). More recently, studies have been addressing the regeneration of abandoned riparian lands in agricultural areas (Carline and Walsh, 2007; Thogmartin et al. 14  2009; Forget et al. 2013; González et al. 2016, 2017). For example, González et al. (2016, 2017) compared a chronosequence of stand ages and found that plant community was recovering. However, they compared sites that were previously used by several human activities (not only agricultural sites) and they also did not control for the external disturbances. Studies done in North America found that some of the ecological attributes of streams are recovering after riparian forest regeneration (Warren et al. 2007; Stovall et al. 2009; Warren et al. 2013; Bechtold et al. 2017). However, these regenerating forests were located primarily in previously logged areas, where large areas (including the upland forests) were abandoned or even replanted for regeneration. Riparian regeneration and stream recovery in this context is more likely to occur in comparison to riparian regeneration in agricultural areas because disturbances are reduced and later disappear as the upland forests regenerate (Franklin et al. 2002; Yeung et al. 2017).   Land abandonment for natural forest is inexpensive compared to active tree planting (Benini et al. 2017) and has a big potential to restore degraded lands in tropical areas, especially in poor and developing countries that do not have sufficient funds to plant trees. However, there is not enough information available about the effectiveness of this strategy in riparian buffers surrounded by agriculture.   1.10 Buffer from maintenance of mature forests as a management strategy to protect riparian forests and streams in agricultural areas  Leaving a strip of native vegetation around streams is the most common strategy of riparian buffers implementation (Richardson et al. 2012). In this riparian buffer strategy, tree removal varies across states in USA and provinces in Canada (Blinn and Kilgore, 2001; Lee et al. 2004). Recently, some specific logging operations in USA and Canada applied selective logging in riparian areas with the idea to emulate natural disturbances (Kreutzweiser et al. 2009a; Sibley et al. 2012; Mallik et al. 2014; Cristan et al. 2016). Some studies reported no changes to stream ecological processes and communities (Kreutzweiser et al. 2009a; Kreutzweiser et al. 2009b; Chizinski et al. 2010; Kreutzweiser et al. 2010), while others reported reduction to leaf decomposition rates (Lecerf and 15  Richardson, 2010) and changes to riparian forest structure (Palik et al. 2012; Zenner et al. 2012; Mallik et al. 2014). However, some studies also found that riparian and stream ecosystems started the recovery process some years later after partial harvest (Mallik et al. 2014; Yeung et al. 2017). In these cases, riparian recovery is more likely to happen because these logged riparian areas were located on large, logged areas that will be abandoned or even replanted for forest regeneration (Franklin et al. 2002). As mentioned previously, this is a different context than agriculture, where current disturbances from the cleared area may delay the regeneration process and further degrade the riparian forests, compromising their ecological functions for streams.   Although selective logging in riparian buffers located in agricultural areas is not recommended as a riparian management practice (Dr. David Kreutzweiser, personal communication) and is prohibited by the Brazilian legislation (Brasil, 1998), some farmers in Brazil perform illegal logging (tree removal without a permit) within riparian buffers trusting that they will not be discovered by the limited satellite imagery program and the ground enforcement operations. Hence, these riparian forests will lose structure, will be exposed to edge effects, and probably will be less effective at protecting stream ecosystems. At this time, selective logging in riparian areas is a problem with unknown consequences for riparian and stream ecosystems located in tropical agricultural areas.  1.11 Thesis objectives and overview  My objective in this study was to evaluate the effectiveness of riparian buffers management in agricultural areas using a landscape in the Brazilian Amazon. I considered the two most common riparian buffer management strategies in Brazil: (1) land abandonment for natural regeneration and (2) the maintenance of mature forest. The abandonment of large agricultural areas for forest regeneration that includes the riparian areas was also considered as a management strategy to compare with the buffer treatments. I hypothesized that the riparian buffers resulting from land abandonment or maintenance of mature forest would have less forest canopy and a simpler stand structure (low basal area, high stem density, low canopy height, low canopy density, and high light intensity) than 16  regenerated riparian forests that were surrounded by mature forests and mature riparian reference sites, respectively (Chapter 2). I hypothesized that the riparian buffers resulting from land abandonment or maintenance of mature forest would be less effective at protecting stream ecosystems by providing less LW (Chapter 3) and less shading, which would shift the stream to more autotrophy and lower heterotrophy (Chapter 4) and increase leaf decomposition rates (Chapter 5). To test my hypotheses, I selected 24 streams in six different conditions: 1. deforested or young regeneration (1 - 4 years old), 2. young regeneration surrounded by mature forest (8 - 12 years old), 3. old regeneration (18 to 22 years old), 4. reference (pristine forests), 5. buffers of young forest regeneration surrounded by pasture (8 - 12 years old), and 6. buffers of mature forests surrounded by pasture. I focused on differences between conditions 4 versus 6, and 2 versus 5. These comparisons are the most important because they address my buffers treatments. Treatments 1 and 3 were included to obtain a regeneration gradient, which I used to evaluate where buffer treatments fit after riparian forest alterations. For the two riparian buffer management strategies considered in this study, I predicted that the buffers resulting from land abandonment would take longer to advance through the regeneration process as they might be affected by disturbances (edge effects, fires) from the surrounding agricultural landscape (Figure 1.1a), while the buffers from maintenance of mature forests would be degraded by selective logging and other disturbances from the surrounding agriculture (Figure 1.1b).  In Chapter 2, I examined forest (basal area and stem density) and canopy (height, density, and light intensity), structural variables considered important (by providing shade and organic material that act as food (leaves) and structural elements (LW) to the stream ecosystem) to sustain ecological processes to streams. In Chapter 3, I examined in-stream LW (density, volume) and related stream pool habitats (LW pool density). I then analyzed the effects of the riparian buffers management plus the influence of basal area, stem density, channel wetted width, and discharge. In Chapter 4, I examined stream primary productivity and respiration. I then analyzed the effects of the riparian buffers management plus the influence of riparian vegetation structure (basal area, stem density, canopy height, canopy density, light intensity), organic material input (litter and LW), water quality (temperature, conductivity, nitrate, suspended sediments), stream flows (discharge and 17  velocity), and channel wetted width. In Chapter 5, I examined decomposition rates of leaf litter from four species of trees. I then analyzed the effects of the riparian buffers management and leaf type, plus the influence of riparian vegetation structure (basal area, stem density, canopy height, canopy density, light intensity), water quality (temperature, conductivity, nitrate, suspended sediments), stream respiration, stream flows (discharge and velocity), and channel wetted width. In Chapter 6, I integrated the ecological and management implications of my findings and make recommendations to improve riparian buffer research and management in agricultural landscapes.                Figure 1.1 Simplified conceptual diagram showing forest trajectories over the years during succession (Budowski (1965) proposed classification schemes and discussed in Chazdon (2014)). Panel A indicates expected pathways of succession after riparian buffer implementation in abandoned lands. Panel B indicates expected pathways of succession after selective logging in riparian buffers. Black arrows show forest recovery in the absence of disturbances. Gray arrows show forest trajectories after the influence of disturbances. Numbers in parenthesis refer to the treatments considered in this work and described in the text. 18       Forest and canopy structural complexityPioneer (1)Early secondary (2)Late secondary (3)Stable climax, not static (4)Disturbance ceasedDisturbance remainsDeforestation (1)Extreme disturbanceNo forest successionRegeneration under disturbance (5)Disturbance ceasedDisturbanceTime after succession begins (years)AForest and canopy structural complexityPioneer (1)Early secondary (2)Late secondary (3)Stable climax, not static (4)Climax after disturbance (6)Disturbance ceasedDisturbances remainDeforestation (1)Extreme disturbanceDisturbanceNo forest successionTime after succession begins (years)B19  Chapter 2: Short-term assessment of riparian forest structure shows recovery in regenerated buffers and signs of degradation in selective-logging buffers in eastern Amazon, Brazil  2.1 Introduction  Riparian habitats are complex ecosystems that support high levels of biodiversity and perform several ecological functions for the surrounding ecosystems (Naiman et al. 2005; Kuglerová et al. 2014; Ramey and Richardson, 2017). These areas are usually disturbed by overbank flow and also have moist soils due to groundwater input from upland areas, which together with forest canopy, creates a mosaic of microclimate and soil conditions (Naiman et al. 2005; Kuglerová et al. 2014; Ramey and Richardson, 2017). Riparian habitats are also natural corridors that allow organisms to disperse in the landscape, offering temporary or permanent habitats for these species (Naiman et al. 2005). Riparian habitats are also an important component for the life cycle of some aquatic organisms, like insects, which have an aquatic larval form, and after emergence will complete their adult phases in the riparian area and possibly become a food resource for riparian terrestrial fauna (Richardson and Sato, 2015). In forested landscapes, riparian forests are extremely important for the ecological processes of headwater streams. The forest canopy covers almost the entire channel, controlling light incidence and primary productivity, and is also a control on the rates of organic matter inputs to streams (Naiman and Décamps, 1997; Richardson and Danehy, 2007; Kiffney and Richardson, 2010; Bechtold et al. 2017; Kaylor et al. 2017). The roots of riparian trees are important to stabilize stream margins and reduce erosion (Allmendinger et al. 2005). Riparian forests also act as filters of sediments, nutrients, and pollutants from the upland areas (Naiman et al. 2005; Richardson et al. 2012), and supply important structural resources for the aquatic environment, like fine and large pieces of wood (Naiman et al. 2005; Paula et al. 2011). Hence, changes of riparian forest amount, structure, and composition in these forested landscapes due to land-use changes have an important impact in the ecological processes of streams and surrounding environments.  20  Anthropogenic activities such as deforestation and forest degradation reduce forest amount, structure, and composition at the stand level (Laurance et al. 2000; Laurance et al. 2006a; Mallik et al. 2014; Richardson and Peres, 2016). The removal of targeted timber-valuable tree individuals contributes to reduced forest biomass, reduced forest canopy, and altered tree composition in the area immediately after logging (Okuda et al. 2003; Villela et al. 2006; Richardson and Peres, 2016). In agricultural areas, remaining forests will be more susceptible to edge effects and a variety of disturbances (wind, fires, cattle, pollutants) that may degrade forest structure and composition by killing big diameter, old-growth trees and increasing the recruitment of low-diameter secondary trees (Putz, 1990; Harper et al. 2005; Laurance et al. 2000, 2006a). Changes to riparian forest stand structure after logging and edge effects create canopy gaps, further increasing light incidence and temperature inside the forest stand (Putz, 1990; Laurance et al. 2000; Harper et al. 2005; Jones et al. 2017).  The most common riparian management strategy in logging and agricultural areas is the implementation of riparian buffers (Kiffney et al. 2003; Gomi et al. 2006a; Brasil, 2012; Richardson et al. 2012). However, riparian buffers are also subject to disturbances originating from the surrounding cleared areas (Grizzel and Wolff, 1998; Kiffney et al. 2003; Carline and Walsh, 2007; Bahuguna et al. 2010; Braithwaite and Mallik, 2012). My objective in this study is to evaluate the effectiveness of riparian buffers management in agricultural landscapes of the Brazilian Amazon. The study design to address this objective was described in Chapter 1. Here, I hypothesize that the riparian buffers resulting from land abandonment and maintenance of mature forest will have less forest stand and canopy structure (low basal area, high stem density, low canopy height, low canopy density, and high light intensity) when compared to a regenerated riparian forest (also land abandonment) surrounded by mature forest and to mature riparian reference sites, respectively.   2.2 Methods  2.2.1. Study area  21   The study was conducted in the municipality of Paragominas, located in the northeastern part of Pará State, Brazil (Figures 2.1 and 2.2). The Paragominas region is located in the Ipixuna and Barreriras formation (predominant clay mineralogy is kaolinite). Deeply weathered Oxisol soils are dominant in upper landscape positions and clay-rich (40-60%) Plinthustult soils occur in lower positions, developed from both clay-rich colluvium from upslope and from the sandier Barreiras Formation (Figueiredo et al. 2010). This area is predominantly a flat to softly undulating topography belonging to the Região Geomorfológica Planalto Setentrional Pará-Maranhão.  Figure 2.1 Location of the Paragominas municipality in the Pará State, Brazil (top-left). Dashed line represents the area where the sampling sites were located and are illustrated in more detail on Figure 2.2.  22  Figure 2.2 Location of sampling sites in the study area. The black meandering line is the Capim River, which delimits the Paragominas municipality borders.  23  The climate is classified as tropical rainy (Aw) with distinct dry (June to December) and rain seasons (January to July) in Koeppen’s classification (Verissimo et al. 1992; Bastos et al. 2005). Mean annual rainfall was 1760 mm for the period of 1973-2003, ranging from 878 to 2766 mm with strong seasonality (Moraes et al. 2006). For the same period, maximum annual average temperature was 33°C with monthly variation between 30 and 34oC, and minimum annual average temperature was 22oC with monthly variation between 21 and 23oC (Bastos et al. 2005).   The forest of Paragominas is evergreen, lowland rainforest and species rich (Veríssimo et al. 1992; Moraes et al. 2006). Two physiognomic types of evergreen rain forest dominate the landscape: (i) a 25-35 m tall, closed canopy forest; and (ii) a 15-20 m tall, vine-laden, open canopy forest (Uhl et al. 1988). The mature forest has an average height canopy of 30 m, varying between 25 to 40 m and with some emergent trees reaching 45 m (Moraes et al. 2006; Francez et al. 2007). Tree density is 315 per ha and tree diameter varies from 5 to 105 cm (Souza et al. 2006). The predominant tree species belong to the families Violaceae (Rinorea flavescens), Lecythidaceae (Lecythis idatimon, Eschweilera grandiflora, Eschweilera pedicellata), and Fabaceae (Poecilanthe effuse; Francez et al. 2007). Currently, most of the forest is composed of degraded and secondary forests due to logging of the most timber-valuable tree species and by regenerated areas followed by pasture abandonment and forest regeneration, respectively (Nepstad et al. 1991; Uhl et al. 1991). In these secondary forests, canopy heights range from 7 to 14 m, basal area varies between 8.7 to 21.4 m2ha-1, and tree density varies from 4800 to 7100 trees per hectare (Uhl et al. 1988). The most abundant tree species belong to the families Urticaceae (Cecropia spp.), Boraginaceae (Cordia bicolor), Euphorbiaceae (Croton diasii), Fabaceae (Inga alba, Inga thibaudiana), Solanaceae (Solanum rugosum), Cannabaceae (Trema micrantha), and Hypericaceae (Vismia guianensis; Uhl et al. 1988).   Human settlements in Paragominas region started recently in the 1960s following the construction of the Belém-Brasília highway. In the 1950s and 1960s cattle ranchers from southern Brazilian states settled in Paragominas stimulated by both federal incentive programs (investment tax credits and subsidized rural credit) and escalating land values. In 24  the 1980s and 1990s there was a large increase in the timber industry due to the decrease of economic prospects for cattle ranching (Nepstad et al. 1991; Verissimo et al. 1992; Gardner et al. 2013). Land-use history in the last decades is characterized by selective logging and slash-and-burn conversion of many areas. Later, many of the selectively logged forests were harvested again and converted to pasture as the most desirable timber species became scarce (Uhl et al. 1991). Pasture areas are usually grazed for six to eight years and then are abandoned to fallow or cultivated again under more intensive pasture management (one cattle per hectare under proper periodic additions of fertilizers and weed control (Uhl et al. 1988; Figueiredo et al. 2010)).   The Paragominas municipality is located in an area called “Arco do Desmatamento” (Deforestation Arc). According to the Amazon Environmental Institute Research (IPAM, 2018), it is an area of 500000 km2 that encompasses the Brazilian states of Pará, Mato Grosso, Rondônia, and Acre. This is the region where the agriculture advances toward the Amazon forest and in which is observed the highest deforestation rates in the Amazon (IPAM, 2018). In the early 2000s, the region had decreasing rates of deforestation of primary vegetation and the establishment of large-scale mechanized agriculture. Mechanized agriculture increased rapidly in recent years over pasture and secondary forested areas, currently occupying 60000 ha of soybean, rice, and corn; Paragominas also had a rapid recent expansion of silviculture (Eucalyptus spp. (Myrtaceae) and Shizolobium amazonicum (Fabaceae)) (Figueiredo et al. 2010; Gardner et al. 2013). The majority of agricultural properties is less than 1000 ha and provides agricultural products for local and regional urban centers (Gardner et al. 2013).   Cattle-ranching and logging activities have greatly altered the original forested landscape in the area and today the landscape is composed of a mosaic of  regenerated forests, logged forests, old fields (grass and shrub) and pastures that are agriculturally and biologically impoverished (Uhl, 1987; Uhl et al. 1988; 1991; Nepstad et al. 1991; Barlow et al. 2016). However, Paragominas city has recently embarked upon high-visibility, multi-sectoral sustainability initiatives, especially the foundation of the Município Verde (Green County) initiative for promoting sustainable land-use systems. This process has strong 25  support from non-governmental organizations, the farmers’ union and local government, and the main goal is to promote sustainability and the conservation of ecosystems and biodiversity (Gardner et al. 2013).  2.2.2 Site selection and experimental design  The study sites were selected using a forest cover map (30 m resolution, year 2010). This map had several categories of forest regeneration and was elaborated by the Amazon Institute of People and the Environment (IMAZON, http://imazon.org.br/?lang=en) for the whole Paragominas municipality. The details of map elaboration are described in Nunes et al. (2015). My selected sites were located on the west side of Paragominas municipality (Figure 2.2). The sites were concentrated on this region due to logistic reasons (accessibility, infra-structure). A large portion of this region belongs to a timber company who exploit timber in a sustainable way (carefully planning of logging roads and tree fall direction, prescribed amount of wood removal by species according to the Brazilian legislation, no logging on riparian areas). Other sites were located on the surrounding farms. In my work, all forest regeneration was a result of land abandonment (see treatments description for specific detail). The characteristics of primary mature forests and secondary forests were described in the previous section. For this study, I considered 24 sites in six management treatments, each with 4 replicates:    Reference sites (REF, control) – streams located inside conserved areas of primary mature forest (Figures 2.3A and 2.4A). All sites located inside the timber company property, meaning that the legislation regarding riparian protection (no selective logging in the riparian area) was respected. At the time of sampling, there was no selective logging in the surrounding upland forests.  Deforestation or young regeneration (YR, 1 to 4 years old) – deforested streams or streams with young regenerated riparian forests (Figures 2.3B and 2.4B). Deforested streams do not comply with the legislation, while riparian regeneration is mostly to attend the current legislation regarding restoring illegally deforested riparian areas (Brasil, 2012). 26   Young regeneration surrounded by forest (YRF, 8 to 12 years old) – regenerated forests (riparian plus the upland forests) surrounded by mature forests (Figures 2.3C and 2.4C). All the previous deforested area is under regeneration. These lands were abandoned by becoming unproductive sites or to attend the legislation regarding restoring illegally large deforested areas, including the riparian area (Brasil, 2012).  Old regeneration (OR, 18 to 22 years old) – streams with regenerated riparian forests surrounded by mature forests (Figures 2.3D and 2.4D). All the previous deforested area is under regeneration. These lands were abandoned by becoming unproductive sites or to attend the legislation regarding restoring illegally large deforested areas, including the riparian area (Brasil, 2012).   Buffers of young forest regeneration surrounded by pasture (YRPB, 8 to 12 years) - streams with 30 m buffers of regenerated riparian forests surrounded by pasture (Figures 2.3E and 2.4E). Only the riparian buffer area is under regeneration mostly to attend the current legislation regarding restoring illegally deforested riparian areas (Brasil, 2012).  Buffers of mature forests surrounded by pasture (MFPB) - streams with 30 m buffers of primary (but recently selectively logged) forests surrounded by pasture (Figures 2.3F and 2.4F). Original forest (before selective logging) similar to reference forest. All sites were located outside the timber company property, meaning that the legislation regarding riparian protection (no selective logging in the riparian area) was not respected.    27  Figure 2.3 Forest cover maps showing an example of each treatment of this study. Sites are shown in higher resolution than Figure 2.2. A. REF. B. YR. C. YRF. D. OR. E. YRPB. F. MFPB.  28  Figure 2.4 Streams representing each treatment in this study. A. REF. B. YR. C. YRF. D. OR. E. YRPB. F. MFPB.   The forest cover map (illustrated in Figures 2.2 and 2.3) was initially composed of 22 categories (most of them were categories of forest regeneration). For this study, I reclassified the map by keeping the most important ‘no regeneration’ categories (mature forests, cleared areas) and by grouping the regeneration categories as described previously. To allocate streams to each treatment, I first considered the dominant regeneration category 29  present around the stream using the map as a starting point. In the field, I visited and checked each stream by examining the dominant vegetation along the studied reach. Streams were selected only if the vegetation on site matched the treatment I previously designated from the map. In addition to vegetation characteristics, I tried to limit stream sizes to 1 to 5 m in wetted channel width (only site MFPB4 (5.88 m) slightly exceeded this criterion). Stream dimensions and discharge are presented in Table 2.1.  All regeneration sites in YRF and OR were previously large pasture areas that were abandoned by unknown reasons. Riparian buffers in YRPB sites were previously pasture areas that were abandoned to comply with the Brazilian legislation (Brasil, 1965, 2012). They were only recently abandoned, probably because the Brazilian government has increased enforcement in the last decade.     30  Table 2.1 Management treatment and general characteristics of the sampled streams. Stream Management treatment Code Channel wetted width (m) Channel area (m2) Channel depth (m) Discharge (L/s) S1 Young regeneration YR1 2.42 362.34 0.36 155.37 S2 Young regeneration YR2 1.76 263.91 0.31 14.73 S3 Young regeneration YR3 2.78 416.25 0.26 53.23 S4 Young regeneration YR4 4.32 648.28 0.46 63.45 S5 Young regeneration surrounded by forest YRF1 2.07 309.84 0.17 26.72 S6 Young regeneration surrounded by forest YRF2 2.89 433.50 0.25 44.5 S7 Young regeneration surrounded by forest YRF3 1.64 245.63 0.13 15.46 S8 Young regeneration surrounded by forest YRF4 2.53 380.16 0.23 21.73 S9 Old regeneration OR1 3.48 522.66 0.28 169.54 S10 Old regeneration OR2 1.80 270.09 0.13 14.01 S11 Old regeneration OR3 2.64 396.19 0.17 65.01 S12 Old regeneration OR4 2.37 355.78 0.18 40.06 S13 Reference REF1 2.66 398.44 0.20 59.94 S14 Reference REF2 1.84 276.50 0.12 29.96 S15 Reference REF3 2.88 432.19 0.25 84.22 S16 Reference REF4 2.46 368.44 0.20 69.11 S17 Buffers (30 m) of young regeneration surrounded by pasture YRPB1 2.13 320.16 0.17 36.71 S18 Buffers (30 m) of young regeneration surrounded by pasture YRPB2 2.93 438.94 0.20 40.97 S19 Buffers (30 m) of young regeneration surrounded by pasture YRPB3 3.17 475.31 0.18 34.01 S20 Buffers (30 m) of young regeneration surrounded by pasture YRPB4 2.43 364.69 0.22 9.45 S21 Buffers (30 m) of mature logged forests surrounded by pasture MFPB1 4.83 724.50 0.26 35.65 S22 Buffers (30 m) of mature logged forests surrounded by pasture MFPB2 4.60 690.00 0.27 82.93 S23 Buffers (30 m) of mature logged forests surrounded by pasture MFPB3 3.90 585.47 0.20 26.34 S24 Buffers (30 m) of mature logged forests surrounded by pasture MFPB4 5.88 881.72 0.29 125.2 31   2.2.3 Riparian forest and canopy structure measurements  Riparian forest structural variables were collected in 10 plots of 10 x 10 m each (5 on each side of the stream) along the 150 m study reach of each stream (Figure A.1). To delimit the plots, I extended a tape 10 m adjacent and parallel to the stream margin, and at each extremity, I extended another tape toward the forest (following an angle of 90°). An effort was made to evenly space (20 m apart) the plots along the stream margin, but due to stream meandering in some locations, plots were allocated closer to each other in these places (10 m distant or even side by side). After delimitation, I started counting and measuring all trees with diameter at breast height ≥ 3 cm. I calculated stem density by dividing the number of stems by plot area and basal area for each stem following the equation below:  Basal area = (DBH/100)2 x 3.1415 / 4  where DBH is the diameter at breast height (cm). I then summed the basal area for each stem within the plot and divided this value by plot area (ha). I calculated mean stem density and mean basal area for each plot, then I obtained a mean value for the stream reach. I also constructed histograms of stem diameter distribution for each treatment. I also estimated shrub, sapling, and liana cover in small plots of 1 x 1 m at the center of each vegetation plot. Total area sampled was 1000 m2 and 10 m2 for plots and small plots, respectively.  Canopy measures (height, density, and light intensity) were collected in a 150 m stream reach at approximately one meter above the channel and at the center of each transect end point (16 transects, each starting 10 m apart) used to characterize channel width and depth. To estimate canopy height, I first estimated by the eye and then checked the estimated value by a laser distance meter (Model X2, PREXISO AG, Glattbrugg, Switzerland). Where the canopy was very homogeneous, I did this procedure on the three first measures and then followed with sampling only using estimation by eye. The digital 32  distance measurer was always used when in sections where the canopy was more heterogeneous. Canopy density above the channel was estimated using a convex densiometer (Model A, Forest Densiometers, Rapid City, USA) facing upstream, downstream, left and right. I then multiplied the total count of canopy opening points by 1.04 to obtain percent of overhead area not occupied by canopy and then subtracted this value from 100 to estimate canopy density in percent. Light intensity was measured with a digital light meter (Model MLM 1011, Minipa Electronics, Houston, USA). I used light intensity as a measure of channel shading. To reduce the error related to differences in light intensity as the day advances, I narrowed the sampling interval to the brightest period of the day (starting at 10:00 and ending at 16:00). I also took a measure in an open area in the surroundings, which was used to correct the values due to differences in sampling period among sites. I then averaged the measures collected in each transect to estimate mean canopy height, density, and shading for the entire stream reach. Extreme values of light intensity were considered as outliers and were dropped from the average calculation.  2.2.4 Statistical analysis  I started checking the distribution of variables to select the most appropriate statistical analysis. After this inspection, non-normal response continuous variables were log transformed and analyzed for differences among treatments using Analysis of Variance (ANOVA; Quinn and Keough, 2002). Proportion data were analyzed for differences among treatments using Generalized Linear Models (GLM) and the Binomial link function (Crawley, 2007). After running these analyses, residuals were checked for normality and homogeneity of variances in the ANOVA models and for overdispersion in the GLM models. In the presence of overdispersion (residual-scaled deviance much larger than the residual degrees of freedom), the model was fitted again using the quasi-binomial link function (Crawley, 2007). I then did pairwise comparison of group means (Tukey’s test) to assess differences among treatments for the significant ANOVAs (Quinn and Keough, 2002).   33  I used a multivariate analysis to see how sites in each treatment fit the pre-defined groups based on riparian forest and canopy structure variables. I constructed a biplot using Principal Components Analysis (PCA). As an unconstrained ordination technique, PCA is most suitable for quantitative variables; it preserves the Euclidean distance, and the relationships detected are linear (Borcard et al. 2011). I computed the PCA on the correlation matrix (standardized variables on the same measurement scale) and retained the first two PCA axes for interpretation. All the analyses were done in R (R Development Core Team, 2016).  2.3 Results  A total of 6618 stems were measured and 72.54 % of them had DBHs in the smallest, 3-10 cm size class (Figure 2.5). Treatments REF and YR had the highest (53.11 m2·ha-1) and the lowest (6.69 m2·ha-1) basal area, respectively. Treatments YRF and YRPB had similar values of basal area between them (27.35 and 28.24 m2·ha-1, respectively; Figure 6). Treatments OR and MFPB also had similar values of basal area between them (44.34 and 46.41 m2·ha-1, respectively). For stem density, treatment YRPB and YR had the highest (3625 n·ha-1) and the lowest (1325 n·ha-1) value, respectively; while for the other treatments stand density values were more similar (Figure 2.6).   34  Figure 2.5 Diameter distribution for stems in each management treatment. YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.   Diameter classes (cm)Frequency (%)0102030405060708090YR0102030405060708090YRF0102030405060708090OR01020304050607080903_10 10_20 20_30 30_40 40_50 ≥50RE01020304050607080903_10 10_20 20_30 30_40 40_50 ≥50YRPB01020304050607080903_10 10_20 20_30 30_40 40_50 ≥50MFPB35  Figure 2.6 Bar graphs showing means (± 1 S.D.) of the predictor variables for each management treatment. Basal area (A) and stem density (B). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.   Treatments YR and YRPB had the lowest values of canopy height (4.18 and 9.62 m, respectively), while treatments REF and MFPB had the highest values (16.52 and 15.96 m, respectively; Figure 2.7a). Treatment YRF had higher canopy height than treatment YRPB (12.84 and 9.62 m, respectively) and treatments REF and MFPB had similar values. Treatment YR had the lowest (50.84%) and the highest (4.38 lux) value of canopy closure and light intensity (Figures 2.7b and 2.7c), respectively, while the other treatments had similar values of canopy density among them. The lowest value of light intensity was observed for treatment REF (1.66 lux).         YR OR RE YRPBYRF MFPB YR OR RE YRPBYRF MFPBA B C D E FManagement groups Basal area (m2/ha)020406080A B C D E FManagement groups Stem density (n/ha)010002000300040005000A B36  Figure 2.7 Bar graphs showing means (± 1 S.D.) of the predictor variables for each management treatment. Canopy height (A), canopy density (B), and light intensity (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.    The ANOVA and pairwise comparison tests showed that basal area, canopy height, and light intensity were different among groups (Table 2.2), while stem density and canopy density had no statistical differences among groups. Considering the most important group comparisons raised in my hypothesis, no differences were detected between YRF and YRPB and between REF and MFPB for any of the variables tested.   YR OR RE YRPBYRF MFPBYR OR RE YRPBYRF MFPBYR OR RE YRPBYRF MFPBA B C D E FManagement groupsCanopy height (m)05101520A B C D E FManagement groupsCanopy density (%)020406080100A B C D E FManagement groupsLight intensity (lux)012345A BCLog light intensity (lux)37  Table 2.2 Results of pairwise comparison tests among riparian management treatments. A = YR. B = YRF. C = OR. D = REF. E = YRPB. F = MFPB. Response Model adjustment Statistical test Probability Tukey Means contrast Basal area No transformation F test <0.01 A<C,D,F / D>B,E Stem density No transformation F test 0.06 Not significant Canopy height Transformed F test <0.01 A<B,C,D,F / E<D,F Canopy density Binomial Chi-square test 0.57 Not significant Light intensity Transformed F test <0.01 A>B,C,D,E,F  The results of PCA showed that axes one and two explained 72.6% and 16.9% of the variance in the data, respectively. The variables basal area, canopy height, canopy density, and light intensity were those that most contributed to axis one, whereas stem density was the variable that contributed most to axis two (Table 2.3).   Table 2.3 Variable scores on each axis retained in the analysis. Variable PC1 PC2 Basal area 1.27 -0.32 Stem density 0.97 1.05 Canopy height 1.24 -0.69 Canopy density 1.33 0.3 Light intensity -1.36 0.1  The ordination biplot (Figure 2.8a) showed that axis one separated group YR from all other groups based on basal area and canopy characteristics. Axis two separated groups YRF and YRPB from groups OR, REF, and MFPB based on stem density. However, sites OR2 and MFPB4 were more similar to YRF and YRPB sites, whereas sites YRPB3 and YRF3 were more similar to REF, OR, and MFPB. Sites REF, OR, and MFPB have mature forests structural characteristics, but looking more carefully at the ordination plot, sites MFPB and OR are close to each other, whereas REF are slightly separated from the other two mature treatments. These results show that the variation in riparian forest structure in my sites is mostly related to the natural regeneration rather than riparian buffer management. Based on these results, I created a new classification for my sites based on natural forest succession stage (Table 2.4). These new classification categories were: SI – secondary initial, SL – secondary late, and MF – mature forest. I then plotted this new 38  classification in a new ordination plot (Figure 2.8b) and used it as a new categorical variable in my further analysis in this work.  Table 2.4 Revised stream classification based on forest stand development after sites ordination in the PCA. Original classification is on the left and the revised classes on the right. Stream Riparian management treatments Forest succession treatments Code Description Code Description S1 YR1 Young regeneration SI1 Secondary initial S2 YR2 Young regeneration SI2 Secondary initial S3 YR3 Young regeneration SI3 Secondary initial S4 YR4 Young regeneration SI4 Secondary initial S5 YRF1 Young regeneration surrounded by forest SL1 Secondary late S6 YRF2 Young regeneration surrounded by forest SL2 Secondary late S7 YRF3 Young regeneration surrounded by forest MF1 Mature forest S8 YRF4 Young regeneration surrounded by forest SL3 Secondary late S9 OR1 Old regeneration MF2 Mature forest S10 OR2 Old regeneration SL4 Secondary late S11 OR3 Old regeneration MF3 Mature forest S12 OR4 Old regeneration MF4 Mature forest S13 REF1 Reference MF5 Mature forest S14 REF2 Reference MF6 Mature forest S15 REF3 Reference MF7 Mature forest S16 REF4 Reference MF8 Mature forest S17 YRPB1 Buffers of young regeneration surrounded by pasture SL5 Secondary late S18 YRPB2 Buffers of young regeneration surrounded by pasture SL6 Secondary late S19 YRPB3 Buffers of young regeneration surrounded by pasture MF9 Mature forest S20 YRPB4 Buffers of young regeneration surrounded by pasture SL7 Secondary late S21 MFPB1 Buffers of mature forests surrounded by pasture MF10 Mature forest S22 MFPB2 Buffers of mature forests surrounded by pasture MF11 Mature forest S23 MFPB3 Buffers of mature forests surrounded by pasture MF12 Mature forest S24 MFPB4 Buffers of mature forests surrounded by pasture SL8 Secondary late       39  Figure 2.8 Ordination graph showing the distribution of sites according to management (A) and regeneration (B) treatments. YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture. SI – Secondary initial. SL – Secondary late. MF – Mature forest.  40   2.4 Discussion  2.4.1 Riparian forest regeneration  I found no differences in forest and canopy structure measures between treatments YRF and YRPB and treatments REF and MFPB. The differences I found among treatments represent the expected pattern in forest succession, not buffer management. These results were further confirmed by my ordination analysis, which grouped my sites into three groups representing the forest successional process and not my six initial groups based on the type of management. Therefore, I concluded that riparian forest and canopy structure in my sites reflect stand developmental processes. During forest succession, as the regeneration process advances, forests recover their structure by increasing basal area, reducing stem density, and increasing canopy height and density; the re-establishment of canopy cover reduces light incidence and increases shading (Guariguata and Ostertag, 2001; Franklin et al. 2002; Chazdon, 2014). If all factors important for the regeneration process are present in the area at the greatest conditions (abundant propagules, suitable seedbeds, and environmental conditions), the recovery of forest structure and composition tends to be fast (Uhl et al. 1988; Nepstad et al. 1991; Chazdon, 2014). For example, studies of tropical forest regeneration after previous low-intensity land use in different parts of the world found that secondary forests recovered floristic composition in approximately 30 years in Ivory Coast, Africa (Kassi N’Dja and Guillaume, 2008) and 40 years in Brazil, South America (Piotto et al. 2009). The recovery of stand and canopy structure tends to be even faster (15-20 years, Guariguata and Ostertag, 2001), as observed in this study. In cases where propagule availability and soil nutrients were severely diminished, or disturbances such as burning, animal trampling or browsing are frequent in the area, the recovery process will be slowed or arrested, with different possible trajectories depending on the nature of the initial and current conditions present during the succession process (Uhl et al. 1988; Nepstad et al. 1991; Griscom et al. 2009; Chazdon, 2014). In this study, I found evidence that the recovery of forest and canopy structure is proceeding well in the abandoned areas, even in the riparian buffers, suggesting that the factors important for the 41  regeneration process were not severely affected during forest clearing and subsequent agricultural use. The effects of these factors (biological legacies and the landscape context) on the regeneration process seem to be more important than buffer management (proximity to pastures) in this study.  2.4.2 Effects of riparian buffer management on riparian forest structure   Riparian buffer management practices had no effect on the forest structural measures studied here. Therefore, my hypothesis was not supported by my data. I expected to find low basal area, high stem density, low canopy height, low canopy density, and high light intensity in sites YRPB and MFPB when compared to YRF and REF, respectively. My expectations were based on the effects that adjacency to pasture could have on riparian forest and canopy structure. For example, I expected that riparian buffers with younger forests surrounded by pastures would be more affected by fire, trampling, grazing, and abundant woody vines. These factors slow down the regeneration process (Schnitzer et al. 2000; Zarin et al. 2005; Griscom et al. 2009), not allowing forests to recover stand and canopy structure (increase canopy height, decrease canopy openness, increase shading, increase basal area, and reduce stand density). On the other hand, the regenerated riparian forests surrounded by mature forest are not exposed to these disturbances from the adjacent pasture fields, which allows the forest to grow faster and possibly recover stand structure and canopy cover characteristics. In the case of riparian buffers with mature forest surrounded by pasture, I expected to find low basal area, high stem density, low canopy height, low canopy density, and high light intensity due to recent selective logging that occurred in these sites and windthrow of exposed trees. This could have decreased basal area by the removal of the biggest trees, which could have increased canopy gaps in the forest, decreasing canopy height and density and increasing light intensity. Hence, this would increase the number of low diameter stems of secondary tree species (Zener et al. 2012; Mallik et al. 2014). Edge effects could have the same impacts on these mature forest buffers (Putz, 1990; Murcia, 1995; Laurance et al. 2000, 2006a; Harper et al. 2005). What I found was similar forest and canopy structure in YRPB and MFPB when compared to YRF and REF, respectively.  42   The results for YRF and YRPB provide evidence that natural regeneration within riparian buffers allow for recovery of stand condition within these pasture landscapes in the Amazon. I hypothesize that low intensity disturbances (fires) are one of the most important factors to explain my results, in addition to biological legacies from past forest cover and seeds and dispersers from the surrounding forest under sustainable forest operations. Although YRPB sites had the highest amount of liana cover in the small plots compared with YRF (possibly because canopy height is still recovering in the YRPB), these lianas seem not to be affecting stand structure recovery in YRPB. Studies conducted in the Amazon showed that surface fires are becoming frequent in the region (Barlow and Peres, 2006; Laurance, 2006b; De Faria et al. 2017). These surface fires usually kill only the small diameter trees, but as the frequency of these fires increases, the largest trees show high mortality levels (Barlow and Peres, 2004). Also, repeated fires decreased forest regeneration and carbon accumulation in slash-and-burn farms in the Amazon (Zarin et al. 2005). In my study area, I observed farmers burning the fields for pasture maintenance, but fire did not advance too much (approximately 3 m) towards the interior riparian forest from the buffer edge (Figure 2.9). I concentrated my sampling in the 10 m adjacent to streams in these 30 m riparian buffers, so having a wider buffer seems to be important to reduce the effects of edge-related fires in this agricultural landscape.   Regarding biological legacies from past forest cover (like remnant roots, stems, and seed bank), they help to accelerate the regeneration process. The availability of these biological legacies is related with the pre-disturbance forest condition, and usually they are more abundant in lightly used areas than moderate to heavily used areas, with the lowest abundance in the heavy use (Uhl et al. 1988; Nepstad et al. 1991; Chazdon, 2014). Studies have shown that basal area, stem density, and species richness decrease as the number of fallow cycles increases due to the decrease of propagule availability (Lawrence, 2004; Sovu et al. 2009). However, in some situations, more intensive shifting cultivation did not cause a decline in the regeneration of non-pioneer species, which reflects the high resilience of resprouts to advance the forest regeneration even in areas with high use fallow cycles (McNamara et al. 2012). Two pastures surrounding my YRPB sites could be classified as 43  low intensity use as there were lots of shrub resprouts in the area (Figure 2.10). This might be a sign that the riparian areas of these sites were lightly to moderately used as pasture in the past, which might have helped with maintaining the regeneration trajectory in these buffers.  The proximity to remnant patches of native forest cover also accelerates the recovery process (Chazdon, 2014). These remnant patches are sources of seed rain and also faunal dispersers that bring new propagules for the cleared area (Chazdon, 2014). In this study, all cleared areas (both regenerating and agriculture) were within a landscape with abundant native forests as my study area was located in a large area of sustainably managed logging. In the field, I frequently saw toucans, monkeys, agoutis, and tapirs inside my buffers or moving in the surrounding landscape. I hypothesize these animals use the riparian forest buffers as corridors to move through agricultural areas, and by doing so, they are bringing propagules with them to the regenerating riparian forests.                  44  Figure 2.9 Pasture under fire management used to clean the field. A) Burned fields adjacent to riparian forests. B) Surface fire advancing toward the interior of riparian forest edge, burning fine litter and tree saplings.   Another possible explanation for my regenerated buffer results is that sites YRF, instead of developing faster as would be expected by being surrounded by mature forests, AB45  could have slowed down the regeneration because of previous land-use intensity in the sites. As mentioned earlier, biological legacies decrease with increasing past land-use intensity (Uhl et al. 1988; Nepstad et al. 1991; Chazdon, 2014). I hypothesize that past land-use is not slowing down forest recover in these sites as absence of current disturbance and the presence of native forest in the surrounding area probably can overcome this issue. Also, at least two pasture sites in this area are still under light/moderate use as observed in the field, which might be an indication that pastures in these areas had the same management practice in the past. Therefore, it seems more likely that sites YRPB are recovering structure at the same rate as sites YRF.  Figure 2.10 Pasture under low intensity management with lots of shrubs and saplings growing in the grass field.   In the case of mature buffers, my results show some evidence of ongoing forest degradation in these buffers as they seem to be losing structure by the removal of the biggest trees and also canopy alterations (Palik et al. 2012; Zenner et al. 2012). Qualitative 46  data on the small plots also showed the highest amount of saplings in these sites. I believe that these findings result from the type of management applied in these mature buffers. For example, it seems that only the biggest trees were targeted in these buffers, so the harvested sections were very disturbed as the direction of tree fall was not determined in advance to avoid damaging surrounding trees (as it is commonly done by the implementation of RIL guidelines in sustainable forest operations). This, of course, increased forest degradation and canopy opening in these partially-harvested sections. Buffer sections with unharvested big trees were not degraded to the same level as partially-harvested sections. This probably balanced the effects of forest alterations by partial-harvest in these buffers, which I hypothesize would be worse if more trees were removed by a prescribed high amount of basal area or in some uncontrolled and intensive way. Also, I hypothesize that harvesting in these sites was done at a slow rate and probably is still ongoing in some instances. This reflects the destination for this timber, which in this case are local uses in the property and not commercial use. In the field, I observed several cut trees that were left behind, probably because loggers targeted the tree by its size and not by wood characteristics for the desired application (Figure 2.11). It is possible that more wood will be removed in the future at a slow rate, depending on the local demand of the property. This means that these forests will be further degraded and probably more susceptible to other disturbances coming from the edges.   The harvesting of large trees decreases basal area and reduces canopy height and density, opening the canopy for more sunlight, further increasing the density of pioneer species. Palik et al. (2012) found lower basal area (after 1 and 9 years post-harvesting) and higher stem density (early successional trees, subcanopy trees, later successional trees, and shrubs after 1 year post-harvesting) in the two 30 m riparian buffer, partial harvest treatments when compared to reference sites and unlogged 30 m riparian forest buffers due to the prescribed tree removal and blowdown, respectively. Zenner et al. (2012) found lower basal area in the two riparian buffer, partial-harvest treatments (both 46 m width which had low and medium prescribed basal area removal) after 1 and 3 years post-harvesting when compared to reference sites. Zenner et al. (2012) also found an increase in light availability, and hence, an increase in understory shrub and sapling biomass in the 47  partially harvested treatments. Mallik et al. (2014) found that the abundance of juvenile trees of balsam fir, mountain ash, and white birch increased with the size of gaps created after partial harvesting in riparian buffers. Braithwaite and Mallik (2012) assessed riparian forest and canopy structure between reference and 40 m unlogged riparian buffers and found no differences in basal area and canopy cover in the 0 to 20 m width section of the buffer; however, basal area differed between treatments in the 20 to 40 m width section of the buffer while canopy cover just differed at the extreme end of the buffer (around 35 to 40 m). In Braithwaite and Mallik’s (2012) work, as the buffer treatment had no selective logging within it, the changes observed in basal area and canopy cover at the outer part of the buffer were affected explicitly by edge effects, supporting my statement that edge effect would have potential to affect my harvested buffers in the future.   Figure 2.11 Harvested big diameter tree left behind in riparian area.   My results have important ecological and management implications that will be discussed in Chapter 6 and also raise some important concerns for the long-term 48  conservation of these riparian forests. At the moment, it is unclear if the forests in the YRPB treatments will achieve forest conditions as observed in OR and REF sites (will they advance or keep the current forest structure?) and if the MFPB treatments will lose forest structure through accelerated loss of big trees. These future forest trajectories can have a strong impact on the long-term ecological processes provision for the adjacent stream ecosystem, and hence, on the conservation of streams and their biodiversity.    49  Chapter 3: Effects of riparian buffer management practices on large wood in streams of eastern Amazon, Pará State, Brazil  3.1 Introduction  Riparian forests are important for the ecological processes of streams, and particularly headwater streams. They control light incidence, temperature, and nutrients for primary production, reduce channel margin erosion, control the inputs of fine sediments and nutrients, and supply food and structural resources to the aquatic fauna, such as coarse particulate organic matter and large wood (Naiman and Décamps, 1997; Richardson and Danehy, 2007). Large wood increases stream habitat structure (pools and cascades; Bisson et al. 1987; Rosenfeld and Huato, 2003; Elosegi et al. 2017) and provides several ecological functions for the aquatic organisms, such as cover, food, and places to settle (Anderson et al. 1978; Boss and Richardson, 2002; Benke and Wallace, 2003; Eggert and Wallace, 2007; Valente-Neto et al. 2015).  Anthropogenic activities such as deforestation and forest degradation affect LW input to stream ecosystems, decreasing LW input to the channel and compromising the structural complexity of stream habitats (Paula et al. 2011; Leal et al. 2016). This decrease in LW happens because the source of LW is completely removed after land clearing (Wing and Skaugset, 2002; Paula et al. 2011), and it also decreases when forests are severely degraded or at early successional stages due to the predominance of small diameter and soft wood trees (Andrus et al. 1988; Bilby and Ward, 1991; Paula et al. 2011). When these small trees fall in the channel, they decompose and break-up faster than larger trees, and they can be easily transported to downstream reaches (Andrus et al. 1988; Warren and Kraft, 2008; Cadol and Wohl, 2010; Paula et al. 2013), which decreases their role in habitat formation (Rosenfeld and Huato, 2003). In logged riparian areas, the removal of targeted timber trees may reduce future sources of wood to the channel.   The most common riparian management strategy in timber production and agricultural areas is the implementation of riparian buffers (Kiffney et al. 2003; Gomi et al. 50  2006a; Brasil, 2012; Richardson et al. 2012). In agricultural areas, forested riparian buffers are surrounded by agriculture and continuously affected by edge effects. Therefore, I expect that these riparian buffers will be more disturbed when compared to reference sites and regenerating forests surrounded by mature forests, where the riparian forests would not be exposed to edge effects and disturbances from the agricultural land. Hence, they would be less effective at supplying more and bigger LW to the stream.   My objective in this study is to evaluate the effectiveness of riparian buffer management on protecting streams in agricultural landscapes of the Brazilian Amazon. The study design to address this objective was described in Chapter 1. Here, I hypothesize that the riparian buffers will provide less effective LW (smaller and less stable) for the channel, which means lesser amounts of LW and associated pools in these sites when compared with a regenerated riparian forest resulting from land abandonment that is surrounded by mature forest (for the regenerated buffer) and to mature riparian reference sites (for the mature logged buffer).  3.2 Methods  3.2.1 Study area, experimental design, and riparian structure measurements  The study area was described in Chapter 2 (section 2.2.1). Site selection and the experimental design were described in Chapter 2 (section 2.2.2). Riparian forest measures were described in Chapter 2 (section 2.2.3).   3.2.2 Stream channel measurements  Stream data were collected once at baseflow conditions in a 150 m stream reach during the dry season. A subset of streams were sampled in 2015 (August to September) and another in 2016 (June to August). The reach was divided at 16 transect points each 10 m apart (Figure A.1). At each transect point, I stretched a measuring tape over the channel to collect stream wetted width and five measures evenly spaced of depth and substrate 51  (Fitzpatrick et al. 1998). I estimated water discharge using the dilution gauging method (Gordon et al. 2004). Water discharge was estimated once in each stream on the same period of stream data collection. First, I used a stock solution of NaCl and a series of dilutions in the laboratory to produce a calibration curve by measuring the conductivity of these solutions with a portable YSI probe (Model 85, Yellow Springs Instruments, Yellow Springs, USA). I next applied a regression equation of the target concentration against the specific conductance measured to get the slope of the regression. This slope was used further to convert the conductivity (µS·cm-1) measured in the field to g·L-1 of NaCl. In the field, I placed the conductivity probe in the stream thalweg at the downstream station and released the solution in a slug at the upstream station in a fast velocity spot to ensure mixing. I then started the timer and moved downstream to the probe location to monitor and record conductivity changes at 20 second intervals when conductivity started to increase until it returned to background conditions. I then calculated discharge by integrating the area under the curve of NaCl concentrations.   3.2.3 LW and LW habitat structure  I counted and measured the length and middle diameter of all LW pieces that were inside the active channel (touching the water or hanging on stream margins). I obtained large wood density by dividing the number of pieces by reach length and also by reach area. The volume of each piece was calculated using the LW diameter and length assuming the piece as a cylindrical form (Schreuder et al. 1993). Total volume in the reach was obtained by summing the volume of each individual piece, then divided the total volume by reach length and also by reach area.  The channel habitat units associated with LW (pools and runs that had LW influencing habitat formation or habitat structure) were counted, measured, and classified according to its formation process (erosive or damming). I used three criteria (shape, depth, and water velocity) to classify a habitat as a pool according to Bisson et al. (1998). I classified runs as being homogeneous in width, depth, and water velocity (intermediate water depth and velocity when compared to pools and riffles). I excluded the habitat units 52  that spanned less than half of the channel width. I classified pools as LW habitat in cases where LW (an individual piece or more than one piece) were the main agent of formation or when they were increasing habitat structure by making the habitat unit deeper. For example, I observed that in my streams some runs were deeper because of a high abundance of LW on the stream bottom. As the stream substrate was mostly sandy, the areas close to the LW were eroded, making the habitat deeper than the surrounding area. In cases where more than one piece was present in the habitat, I tried to identify the most important one, i.e. if the deepest section of the habitat was closer to the biggest LW or if the biggest LW created the main structure by trapping the smallest pieces. I obtained LW pool density by dividing the number of wood-formed pools by reach length and also by reach area.  3.2.4 Statistical analysis  The response variables in this study were LW density (LWDE), LW volume (LWVO), and LW pool density (LWPDE). I started checking the distribution of all variables to select the most appropriate statistical analysis. After this inspection, non-normal continuous variables were log transformed and analyzed for differences among treatments using Analysis of Variance (ANOVA; Quinn and Keough, 2002). After running the analyses, residuals were checked for normality and homogeneity of variances in the ANOVA models. I then did pairwise comparison of group means (Tukey’s test) to assess differences among treatments for the significant ANOVAs (Quinn and Keough, 2002).   I used multiple regression models including categorical and continuous variables as predictors (Quinn and Keough, 2002). I included continuous variables considered to be important to LW dynamics in stream ecosystems based on the literature, including channel width, basal area, stem density, and channel discharge. I also included the riparian forest regeneration class based on the new regeneration groups defined in the previous chapter since riparian regeneration is expected to influence LW dynamics in small streams. Therefore, my categorical variables were Riparian Forest Management class (RFM, with 6 levels as previously described in Chapter 2) and Riparian Forest Regeneration class (RFR, 53  with 3 levels generated in Chapter 2; MF – Mature Forest, SI – Secondary Initial, and SL – Secondary Late). To include the categorical variables in the regression models, I converted them to indicator variables (Quinn and Keough, 2002). In this way, each level of the categorical variable becomes an indicator variable coded zero or one. For my management treatment, I generated five indicator variables and for my regeneration treatment I generated two indicator variables (i.e. number of levels – 1). In this analysis, the coefficients of indicator variables are compared to a reference level – in this case, the reference levels were Reference (REF) for the management treatment and Mature Forest (MF) for the regeneration treatment.   I fit regression models containing all predictors (categorical plus continuous - full model), a combination of one categorical and one continuous predictor, and a categorical or continuous predictor alone (reduced models). I excluded the reduced models resulting from the combination of the regeneration (categorical) and the riparian structural variables (continuous predictors) because they were co-linear (forest regeneration groups were generated from riparian structure variables after a PCA). I did not test all the combinations possible among predictors, including the interactions between factors and continuous, because I am more interested in testing the effects of the categories on the responses, but also accounting for the effect of a continuous predictor that is important for my responses based on the literature. It is not my main interest to evaluate which combination of predictors best explain my responses. In the end, I had a final candidate set of 13 models for LW density and LW volume and 19 for LW pool density (for LW pool density I also included LW density and volume as continuous predictors).   To fit my models, I applied Generalized Linear Models specifying the appropriate probability distribution. In this case, I used a Gaussian distribution because all of my responses were continuous data. The response LW volume was log transformed prior to analysis to improve data normality and model fit. I then selected the best supported models using the Akaike Information Criteria corrected for small sample sizes (AICc). Models with the lowest AICc are considered the most parsimonious in a candidate model set (Burnham and Anderson, 2002). I considered the models with a ΔAICc ≤ 2 as having substantial 54  support to be selected as best models (Burnham and Anderson, 2002). I also calculated Akaike weights, which can be interpreted as the probability that a model is the best supported of those in the candidate set (Burnham and Anderson, 2002). To avoid adding uninformative parameters (i.e., one that adds little improvement to the likelihood) in my models, I omitted any model where the addition of a second parameter did not result in a net improvement to the AICc score (Arnold, 2010). The addition of an uninformative parameter to a model with one or more informative parameters results in only a small penalty to the AICc score, and such overfitted models may erroneously be reported as valid alternatives (Burnham and Anderson 2002; Arnold, 2010). If a model selected had an uninformative parameter, that model was excluded and the analysis was done again. The analyses were done in R (R Development Core Team, 2016).  3.3 Results  The ANOVA results showed that channel width, channel area, LW density per channel length, and LW volume per channel length were significantly different among treatments (Table 3.1). However, when analysing LW density and volume per channel area, the ANOVA results were not significant. Treatment MFPB had the highest value for mean channel width (4.80 m) and channel area (720.42 m2), while the other treatments had lower values that were not different (Figure 3.1a). For stream discharge, treatments YR, OR, and MFPB had similar values among them (approximately 72 L/s approximately) and were higher than the other treatments; treatments YRF and YRPB had similar values between them (approximately 30 L/s approximately) and were lower than the others (Figure 3.1b). Treatments YR and OR had larger standard deviations in discharge compared with the other treatments.       55  Table 3.1 Results of pairwise comparison tests among riparian management treatments. A = YR. B = YRF. C = OR. D = REF. E = YRPB. F = MFPB. Variables related to forest structure were tested in Chapter 2. Variable Probability Tukey Means contrast Channel width (m) <0.01 F>A,B,C,D,E Channel area (m2) <0.01 F>A,B,C,D,E Channel discharge (L/s) 0.48 NS LW density (pieces/m) <0.01 F>A,B,C,D,E LW volume (m3/m) 0.01 F>B,C,E LW pool density (units/m) 0.39 NS LW density (pieces/m2) 0.48 NS LW volume (m3/m2) 0.17 NS LW pool density (units/m2) 0.1 NS  Figure 3.1 Bar graphs showing means (± 1 S.D.) of the predictor variables for each management treatment. Channel width (A) and discharge (B). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.   A total of 1545 LW pieces were measured and 68.3% of them had diameters in the smallest size class (Figure 3.2). When evaluating LW variables per channel length, treatment MFPB had the highest density (0.85 pieces/m) of LW and treatment REF had the lowest (0.26 pieces/m) density (Figure 3.3a). For LW volume, treatment MFPB also had the highest values (0.0017 m3/m) while the lowest values were observed for treatments YR OR REF YRPBYRF MFPB YR OR REF YRPBYRF MFPBA BA B C D E FManagement groupsChannel width (m)0123456A B C D E FManagement groups Discharge (L/s)05010015056  YRF, OR, and YRPB (Figure 3.3b, Table 3.1). When evaluating LW variables per channel area, treatment MFPB had the highest density (0.176 pieces/m2) of LW and treatment REF had the lowest (0.110 pieces/m2) density (Figure 3.4a). For LW volume, treatment YR had the highest values (0.00037 m3/m2) while the lowest values were observed for treatments OR and YRPB (Figure 3.4b). Treatment YR, despite having low values for basal area and stem density, had higher LW density and volume than REF sites.  A total of 243 LW habitats were measured (227 pools and 16 runs) and the majority of them were classified as erosive (only 16 were dammed pools). I found 617 LW pieces forming or increasing habitat structure. From this total, 117 pieces were acting alone and 500 pieces were acting in conjunction (more than one piece) on channel habitats. The mean diameter of the LW acting on habitat was 24.5 cm (considering only those LW identified as being the main agent of habitat formation or structure). Treatment OR had more LW pool density per length (0.09 units/m) and per area (0.03 units/m2) compared with the other groups (Figure 3.3c and Figure 3.4c). Although treatment MFPB had the higher values for LW abundance and volume, it had lower LW habitat (0.05 units/m and 0.01 units/m2) than the other groups.   57  Figure 3.2 Diameter distribution for LW in each management treatment. YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.  Diameter classes (cm)Frequency (%)0102030405060708010_20 20_30 30_40 40_50 ≥50REF0102030405060708010_20 20_30 30_40 40_50 ≥50BYRP0102030405060708010_20 20_30 30_40 40_50 ≥50MFPB01020304050607080OR01020304050607080YR01020304050607080YRF58  Figure 3.3 Bar graphs showing means (± 1 S.D.) of the LW variables per channel length for each management treatment. LW density (A), LW volume (B), and LW pool density (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.            YR OR REF YRPBYRF MFPBA BCYR OR REF YRPBYRF MFPBLW volume (m3/m)LW pool density (units/m)LW density (pieces/m)A B C D E FManagement groupsLW density (pieces/m)0.00.20.40.60.81.01.2A B C D E FManagement groupsLW volume (m3/m)0.00000.00050.00100.00150.00200.0025A B C D E FManagement groupsLW pool density (units/m)0.000.020.040.060.080.100.12YR OR REF YRPBYRF MFPB59  Figure 3.4 Bar graphs showing means (± 1 S.D.) of the LW variables per channel area for each management treatment. LW density (A), LW volume (B), and LW pool density (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.          YR OR REF YRPBYRF MFPBA BCYR OR REF YRPBYRF MFPBLW volume (m3/m2)LW pool density (units/m2)LW density (pieces/m2)YR OR REF YRPBYRF MFPBA B C D E FManagement groupsLW density (n/m2)0.000.050.100.150.200.250.30A B C D E FManagement groupsLW  pool density (n/m2)0.000.010.020.030.040.050.060.07A B C D E FManagement groupsLW volume (m3/m2)0e+001e-042e-043e-044e-045e-046e-040.00020.00040.0006060  The results of model selection showed that channel width was the main predictor of LW density and volume per length (Table 3.2, Appendix A.1). In both cases, these variables were positively related with channel width (Figure 3.5). For LW pool density, the predictors selected were discharge, channel width, LW volume, and basal area (Table 3.2). All these predictors were negatively related to LW pool density (Figure 3.6). However, the explanatory power of the relationships for pool density was low.   Table 3.2 Results of model selection for LW variables. Model df AICc Delta_AICc Weight R2adj LW variables per channel length      LW density       CWI 3 -27.12 0 0.87 0.73       LW volume      CWI 3 -3.28 0 0.79 0.45       LW pool density       DIS 3 -108.08 0 0.27 0.08 CWI 3 -107.88 0.2 0.24 0.07 LWVO 3 -106.36 1.72 0.11 0.01 BA 3 -106.12 1.92 0.1 0.01       LW variables per channel area      LW density      CWI 3 -76.86 0 0.3 0.06 DIS 3 -76.64 0.22 0.27 0.05 BA 3 -75.78 1.08 0.18 0.02 SD 3 -75.28 1.58 0.14 0.01       LW volume      SD 3 -398.33 0 0.47 0.14       LW pool density      CWI 3 -148.81 0 0.91 0.46 Predictors: CWI – Channel Width; BA – Basal Area; SD – Stem Density; DIS – Discharge.  61  When evaluating LW variables per channel area, the results of model selection showed that LW density was best explained by channel width (positive relationship), discharge (negative relationship), basal area (negative relationship), and stem density (positive relationship; Table 3.2 and Figure 3.7). For LW volume, the selected predictor was stem density (negative relationship; Table 3.2 and Figure 3.8). Channel width was the main predictor of LW pool density (negative relationship; Table 3.2 and Figure 3.9). Except for the relationship between LW pool density and channel width, the explanatory power of the relationships for LW density and volume were low. The riparian management treatments had no effect on any of the LW variables analyzed.   Figure 3.5 Relationships between LW variables (density and volume per channel length) and channel width in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○).        2 3 4 5 60.20.40.60.81.0Channel width (m)LW density (n/m)Channel width (m)LW volume (m3/m)LW density (pieces/m)2 3 4 5 60.00050.00100.00150.00200.0025Channel width (m)LW volume (m3/m)62  Figure 3.6 Relationships between LW pool density per channel length and channel width (A) and discharge (B) in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○).                 2 3 4 5 60.040.060.080.10Channel width (m)LW pool density (units/m)50 100 1500.040.060.080.10Discharge (L/s)LW pool density (units/m)2.4 2.6 2.8 3.0 3.2 3.40.040.060.080.10LW volume (m3/m)LW pool density (units/m)0 10 20 30 40 50 60 700.040.060.080.10Basal area (m2/ha)LW pool density (units/m)A BC D42 3 65Channel width (m)0.040.080.100.06LW pool density (units/m)15050 100Discharge (L/s)0.040.080.100.06LW pool density (units/m)2.82.4 2.6 3.23.0LW volume (m3/m)3.4 200 10 4030Basal area (m2/ha)50 60 7063  Figure 3.7 Relationships between LW density per channel area and channel width (A) and discharge (B) in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○).                  2 3 4 5 60.050.100.150.200.25Channel width (m)LW density (pieces/m2)50 100 1500.050.100.150.200.25Discharge (L/s)LW density (pieces/m2)0 10 20 30 40 50 60 700.050.100.150.200.25Basal area (m2/ha)LW density (pieces/m2)0 1000 2000 3000 4000 50000.050.100.150.200.25Stem density (n/ha)LW density (pieces/m2)A BC D64  Figure 3.8 Relationships between LW volume per channel area and stem density in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○).                 0 1000 2000 3000 4000 50001e-042e-043e-044e-045e-046e-04Stem density (n/ha) LW volume (m3/m2)LW volume (m3/m2)0.00020.00040.000665  Figure 3.9 Relationships between LW pool density per channel area and channel width in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○).   3.4 Discussion  3.4.1 Environmental controls of in-stream LW  The abundance and volume of LW in the studied streams increased as streams become wider, but LW abundance decreased as discharge increased. In general, LW abundance decreases and LW volume increases as stream channels become wider (Bilby and Ward, 1989, 1991; Chen et al. 2006; Ruiz-Villanueva et al. 2016; Wohl et al. 2017). This happens because larger channels have stronger hydraulic forces that move small pieces downstream, keeping only larger pieces that stay longer in the stream channel (Bilby and Ward, 1989, 1991; Chen et al. 2006; Ruiz-Villanueva et al. 2016; Wohl et al. 2017). However, if discharge does not increase with the increase on channel width, the abundance of LW tends to increase as more pieces accumulate in the reach (more storage area and less 2 3 4 5 60.010.020.030.040.050.06Channel width (m)LW pool density (units/m2)66  transport). This might be happening in my streams since LW abundance increased as channels became wider and channel width was not strong correlated with discharge (r = 0.43). Also, I observed in the field that the remaining wood from riparian areas was discarded in the channel after riparian deforestation. Therefore, larger channels seem to accumulate more wood by high input and less transport.  The abundance and volume of LW in the studied streams were also related to riparian forest structure. I found that LW density is negatively related to basal area and positively related to stem density, and LW volume is negatively related to stem density. In general, mature forests have high basal area and low stem density (Chazdon, 2014). In these sites, although LW recruitment is usualy low (small amount of wood added at frequent intervals, Bilby and Bisson, 1998; Benda et al. 2002), the density and volume of in-stream LW is high because of the recruitment of big trees that are slowly transported and decomposed (Hedman et al. 1996; Young et al. 2006; Warren and Kraft, 2008; Warren et al. 2009). In secondary forests, basal area is low and stem density is high (Chazdon, 2014). In these secondary sites, the density and volume of in-stream LW is usually lower than in streams with mature forests because of low tree recruitment and fast transport and decomposition of small and soft wood (Hedman et al. 1996; Young et al. 2006; Warren and Kraft, 2008; Warren et al. 2009; Paula et al. 2013). However, in-stream LW volume in secondary areas may also be similar to that observed in old-growth sites because of increased tree mortality and wood input as stand development advances and reaches the stem exclusion stage (Benda et al. 2002). Therefore, my findings may reflect LW input according to the riparian forest stand development. However, it is important to note that my regenerating sites are too young (stand initiation phase) to provide much wood compared to the secondary sites (50 years old) studied by Benda et al. (2002). In this way, the increased LW density and decreased LW volume in my secondary sites may be better explained by land management history in the regenerating sites. As I mentioned previously, I observed cut pieces of LW inside the streams, suggesting that this remaining material was discarded in the stream channel after riparian deforestation, increasing in-stream LW.  67  The abundance of channel habitat units decreases as discharge and channel width increases because the habitat units become smaller (sub-units) and less distinguishable due to high discharge (Bisson et al. 2007). In the case of LW pools, the abundance also decreases because of the smaller amount of LW in larger channels (Bilby and Ward, 1989, 1991; Ruiz-Villanueva et al. 2016). I found fewer LW pool in larger channels, but contrary to the literature, these channels had higher LW abundance. Apparently, this is due to the large amount of small diameter and less effective wood that accumulates in these channels as the low discharge seems not to transport them downstream. Regarding forest structure, some studies found a higher amount of LW pools in streams draining mature forest areas than secondary forest stands (Bilby and Ward, 1991; Montgomery et al. 1995; Jackson and Wohl, 2015). In general, more and larger wood pieces are supplied by mature forest than secondary forests (Andrus et al. 1988; Bilby and Ward, 1991; Jackson and Wohl, 2015), and riparian forests provide more and bigger LW as the secondary forest succeeds to later stages (Hedman et al. 1996; Wing and Skaugset, 2002; Warren et al. 2009). If there is more LW available in the streams, the probability of them forming a pool increases, especially if these LW are in the biggest diameter classes (Montgomery et al. 1995; Rosenfeld and Huato, 2003). Although weak, I found negative relationships between LWPDE with increasing basal area and LW volume, probably because the increasing basal area in my sites did not correspond with more LW in the channel. The negative response between LWPDE with LW volume does not follow the forest structure of the sites, but the fact of discarding wood in the channels, especially lots of small pieces that are less effective at improving habitat formation.  3.4.2 Effects of riparian buffers management on in-stream LW  My results showed that all managed sites had higher LW density than reference sites, supporting my interpretation that past deforestation is an important factor that explains LW in these sites, where right after deforestation trees and branches were actively deposited into the streams. This fact was also observed in the field by observing cut pieces inside these streams (some of them with the stump still in the riparian area), especially in the most recently disturbed ones, like the MFPB sites (Figure 3.10). However, riparian 68  buffer management practices had no significant effect on the LW measures studied here. Therefore, my hypothesis was not supported by my data.   I expected to find less LW in YRPB and MFPB sites when compared to YRF and REF, respectively. My expectations were based on the effects that riparian management could have on riparian forest structure and the ongoing supply of LW. Earlier in this work, I stated that riparian buffers might behave like forest edges, which are dominated by short-lived secondary tree species, have high abundance of woody vines, and they are also more exposed to edge effects (Murcia, 1995; Gehlhausen et al. 2000). These factors might increase LW input by increased tree mortality (Laurance et al. 2000). However, these smaller secondary trees decompose and are transported easier in the channel than bigger LW (Andrus et al. 1988; Warren and Kraft, 2008; Cadol and Wohl, 2010; Paula et al. 2013). Therefore, the amount of in-stream LW might be lower when compared to streams with younger forests surrounded by mature forest, where the riparian forests would not be exposed to edge effects, grow faster, and possibly supply more and bigger LW to the streams. In the case of riparian buffers with mature forest surrounded by pasture, I expected to find less LW in these streams due to recent selective logging that occurred in these sites, which could eliminate future sources of LW to these streams when compared to reference sites. What I found was a similar amount of LW between YRPB and YRF, and more LW in MFPB than REF.   Within my regenerated riparian buffers, the regenerated riparian forests appear to provide a lower amount of wood, which also has low diameter, decomposes faster, so these buffers would be less effective at providing LW and associated physical structure for the streams in the long term (Bilby and Ward, 1991; Montgomery et al. 1995; Jackson and Wohl, 2015). Paula et al. (2011) found a low amount of wood in streams of the Corumbataí River basin in São Paulo State, Brazil, which seemed to be due to land-use. The riparian forests of these streams are at secondary stages of succession, so the majority of LW supplied to the streams are small pieces and less efficient for channel structuring (Paula et al. 2011; Paula et al. 2013). These forests are recovering in a landscape that was intensively managed in the past and it is also experiencing ongoing disturbance effects associated with 69  agriculture, including burning and trampling by cattle (Rodrigues, 1999). Possibly, the riparian forests in Paula et al.’s studies have not recovered enough of their structure to provide bigger wood for the streams. I was expecting something similar to my regenerated riparian buffers in this study, but the larger LW that is a legacy from previous deforestation in the area seems to be the main contributor of current wood in my streams, not the current riparian buffer management practices. I will need a long-term study of wood dynamics to have a better understanding of the possible consequences of riparian buffer management in my sites. Although I found that riparian forests are recovering structure in this study, it is not clear if the regeneration process will advance through the successional stages, and hence, provide better wood for the streams.  Riparian buffers of mature forests are often left on stream margins to supply LW to streams. Gomi et al. (2006a) found in Malaysia that one of the 20 m-wide forested buffers had twice as much LW as another forested buffer of the same width. They attributed this result to the removal of trees for commercial timber in the buffer (that had twice as much LW), which degraded the forest and increased wood recruitment. They also found more LW in the managed sites (deforested riparian forest and the forested buffer with twice as much LW) than the reference site. They also found the highest amount of wood in the deforested site, where wood was also discarded to the channel. Another study found higher post-harvest windthrow in 10 m buffer treatment sites compared with 30 m buffers and unharvested controls along streams of coastal British Columbia, Canada (Bahuguna et al. 2010). In the Bahuguna’s study, the fallen trees usually spanned the stream channel, but as they decay, they will break up and fall into the streams, further contributing to habitat structure.   The results of my previous chapter showed that riparian forest structure results from the regeneration process, not buffer management. According to the literature, the regeneration process is important for explaining LW in streams as riparian forests recover biomass as they succeed to later stages, providing more and bigger LW to the channel (Hedman et al. 1996; Wing and Skaugset, 2002; Young et al. 2006; Warren et al. 2009), which also contributes to more pool habitats (Montgomery et al. 1995; Rosenfeld and 70  Huato, 2003). However, my regeneration categorical variable, although present in a model ranked as the second best one in my analysis for LW density and volume, had substantially less support to explain both LW responses in comparison to channel width.  In conclusion, although riparian buffer management increases susceptibility of riparian forests to disturbances and may affect LW supply to the streams, I did not find this effect in my study. Even the ongoing regeneration process in my sites seems not to be the main cause of variation in the measured LW variables. It seems that past deforestation in the area is the main explanation of the observed pattern of LW abundance in these streams, and larger channels retain more debris from the deforestation process. Apparently, this increase of LW abundance helped to increase LW volume in larger channels as discharge had no significant effect on LW. However, this increased abundance did not provide significant improvement to pool formation as most of these wood pieces were small and less effective to act on habitat structure.                 71  Figure 3.10 Cut tree in the riparian area (A) and cut LW inside the stream (B). Note the cut ends.    AB72  Chapter 4: Young secondary regeneration in riparian buffers allows fast recovery of forest structure and ecosystem metabolism in agricultural streams of Amazonia, Brazil  4.1 Introduction  Streams are small ecosystems that have their ecological processes largely influenced by the adjacent riparian forests in forested landscapes. Being narrow, forested streams are completely covered by the forest canopy, which reduces the amount of sunlight reaching the channel and provides lots of coarse organic matter for these ecosystems (Fisher and Likens, 1973; Likens and Bormann, 1974; Richardson and Danehy, 2007). The amount of light available to the primary producers to produce energy and support the basis of the food chain is limited by overhead canopy (Warren et al. 2016; Bechtold et al. 2017; Kaylor et al. 2017), so the system is largely dependent on the processing of allochthonous organic material (particulate and dissolved forms) from the surrounding forest to provide the energy necessary for the ecosystem (Fisher and Likens, 1973; Likens and Bormann, 1974; Richardson and Danehy, 2007).  Anthropogenic activities such as deforestation and forest degradation remove or change forest canopy along the stream margins or nearby (Kiffney et al. 2003; Paula et al. 2011; Zenner et al. 2012; Mallik et al. 2014). As a result, light incidence reaching the stream will increase (Brosofske et al. 1997; Mosisch et al. 2001; Kiffney et al. 2003; Bechtold et al. 2017). Complete or partial canopy removal after deforestation or selective logging, respectively, may reduce the input of organic material from the riparian forest to the stream but fast-growing shrubs will replace some of this input as revegetation proceeds (Kiffney and Richardson, 2010). These changes of canopy structure, light, and organic matter regimes usually increase primary production, possibly shifting the trophic status of the ecosystem from heterotrophic to more autotrophic (Bunn et al. 1999; Bechtold et al. 2017). This shift will cause changes in stream community structure and composition when primary production substantially exceeds respiration. In most cases, primary producer communities are constrained by shading, have low abundance, and are mostly composed of 73  palatable, unicellular algae (diatoms). When the forest canopy is completely removed or extremely degraded, light is no longer a limiting resource so the community shifts to high abundance of mostly filamentous green algae and macrophytes (Bunn et al. 1999; Mosisch et al. 2001; Kiffney et al. 2003).  The most common riparian management strategy in logging and agricultural areas is the implementation of riparian buffers (Kiffney et al. 2003; Gomi et al. 2006a; Brasil, 2012; Richardson et al. 2012). However, after land clearing for agriculture or forest harvesting, the remaining forest areas may become more subject to disturbances originating from the surrounding cleared areas. The attributes of patch edges increase tree mortality, creating large canopy gaps and reducing canopy cover, further increasing light incidence and temperature inside the forest stand (Putz, 1990; Laurance et al. 2000; Harper et al. 2005). These changes of light and temperature will also affect the adjacent stream ecosystem (Brosofske et al. 1997; Warren et al. 2016; Bechtold et al. 2017; Kaylor et al. 2017), possibly leading to an increase in primary productivity when compared to streams with undisturbed forests (Mosisch et al. 2001; Kiffney et al. 2003; Bechtold et al. 2017), where the riparian forests would not be exposed to edge effects and will not face abrupt changes of primary productivity levels. Therefore, changes in riparian forest and canopy structure in buffers may have a large impact on stream metabolism.  My objective in this study is to evaluate the effectiveness of riparian buffer management in agricultural landscapes of the Brazilian Amazon. The study design to address this objective was described in Chapter 1. Here, I hypothesize that the riparian buffers will provide less shading for the channel, which will result in increased stream primary production when compared with a regenerated riparian forest resulting from land abandonment that is surrounded by mature forest (for the regenerated buffer) and to mature riparian reference sites (for the mature logged buffer).     74  4.2 Methods  4.2.1 Study area, sampling design, and riparian forest measurements  The study area was presented in Chapter 2 (section 2.2.1). Sampling design was presented in Chapter 2 (section 2.2.2). Riparian forest measurements were described in Chapter 2 (section 2.2.3).  4.2.2 Stream metabolism measures  I measured whole-stream metabolism during the dry season in 2015 (August to November) using the open-system, 2-station and 1-station diurnal [O2] change methods (Odum, 1956; Marzolf et al. 1994). During the sampling period, each stream was sampled once for O2 concentration, temperature, and conductivity at 5-min intervals over 32 hours using YSI probes (600 OMS V2-1 ROX optical DO, Yellow Springs Instruments, Yellow Springs, USA). Before going to the field, I calibrated the two probes in water-saturated air and kept them recording in the same container for 24 h to check for any malfunctioning and differences between probes’ measures. In the field, I again tested the probes (only at the beginning of the study) by calibrating and placing them together in the stream for 60 minutes. At each stream, I calibrated the probes on site using air saturated for 15 min. by placing the probe in a sealed container following the manufacturer’s manual specifications (YSI 2012; the specifications say that both water-saturated air and air-saturated calibration methods perform well). I then installed the probes in the stream thalweg at approximately 5 cm above the stream bottom with the probes pointing into the current. The selected stream reaches had a relatively laminar flow and were homogeneous in terms of geomorphology, substrate, and land-cover (Grace and Imberger, 2006; Bott, 2007). I placed probes 150 m apart in each stream, one at the downstream and other at the upstream end of the study reach.  After deployment of probes, I dissolved 200 g of NaCl (conservative tracer) in a 1 L container with stream water, released the slug of NaCl at the upstream site, and recorded 75  water conductivity with a portable conductivity meter (Model 85, Yellow Springs Instruments, Yellow Springs, USA) at the downstream end to estimate travel time (the time for the solution to pass through the reach), stream discharge (described in Chapter 3, section 3.2.2), and water velocity. Before releasing the solution, I collected background water samples for water chemistry characterization (described in the next section). I estimated reaeration coefficients (KO2) from reach hydraulics (water velocity) and geomorphology (stream depth) according to the surface renewal model proposed by Owens et al. (1964) and described in Bott (2007). This choice of method was based on field limitations although it is not the most recommended. Grace and Imberger (2006) and Dr. Arturo Elosegi (personal communication) recommended KO2 estimation by the night-time regression method. Among several methods available, Aristegi et al. (2009) and Mulholland et al. (2001) recommended the night-time regression and the energy dissipation method to estimate KO2 when the most accurate tracer gas injection is not available. However, the night-time regression requires enough variation of [O2] overnight (Grace and Imberger, 2006; Acuña et al. 2009), and as I had very low productivity streams, we had poor estimates of KO2 using this method. For the energy dissipation method, it requires measures of stream gradient, which were not accurately measurable for my sites (I had very flat terrain in the study reach and only coarse resolution maps from the study area). Therefore, the only suitable method was the surface renewal method, which provides reasonable estimates considering the slow, relatively laminar flow, and the reach characteristics of my studied streams (Grace and Imberger, 2006; Both, 2007).  In the laboratory, I downloaded the field data into a computer using the software EcoWatch Lite version 1.0.3.4 (Yellow Springs Instruments, Yellow Springs, USA; available at www.ysi.com). I then inserted the [O2] and temperature data from upstream and downstream probes on digital spreadsheets following the models provided in Bott (2007). To calculate gross primary production (GPP), community respiration (CR24), and net ecosystem productivity (NEP), I followed the calculations described in Bott (2007) for both the 2-station and 1-station method. For the 2-station method, I adjusted upstream and downstream data by the time travel and obtained the [O2] saturation at the measured temperature at each sampling time by fitting a polynomial regression based on a table of 76  saturation values and then applying this regression model to my temperature data. I then calculated the [O2] saturation deficit or surplus (measured [O2] – saturated [O2]) and computed the downstream-upstream difference in [O2] in each travel time. For the next step, I adjusted the reaeration coefficient for the stream water temperature at each sampling time and calculated [O2] exchange by multiplying the saturation surplus or deficit by the temperature-adjusted reaeration coefficient. Using the dissolved [O2] exchange corrected by reaeration, I calculated GPP and CR24. I averaged the respiration rates (rates measured within the predawn and post-sunset period) to obtain a mean community respiration and multiplied this value by 24 to obtain total daily community respiration (CR24). To obtain GPP, I summed the rates of net [O2] change for the photoperiod, multiplied the average respiration rate by the length of the photoperiod, and added the absolute value to the sum of net changes to generate an estimate of GPP. The photoperiod was obtained by taking the time of sunrise and sunset in the field as the Photosynthetic Active Radiation sensor was not available. I then subtracted CR24 from GPP to obtain NEP and divided GPP by CR24 to obtain productivity/respiration ratio. If GPP/CR24 ratio is > 1, GPP exceeds CR24 and the system gains energy, i.e. autotrophic; if GPP/CR24 is < 1, CR24 exceeds GPP and the system uses more energy than is produced, i.e. heterotrophic (Bott, 2007). The calculations were similar for the 1-station method, excluding the adjustments for travel time.  4.2.3 Stream environmental variables  In addition to riparian forest and canopy structure variables, I measured stream variables that are important for the stream metabolism, like coarse organic material input (litter and LW), water quality (conductivity, temperature, nitrate, and suspended sediments), stream flow (discharge and velocity), and channel size (wetted width and depth). Stream data (wetted width and depth), LW, and channel discharge were collected in a 150 m stream reach at each site as described in Chapter 3. Water velocity was calculated dividing the upstream/downstream distance by the time of slug release and peak conductivity.   77  I estimated plant litter flux during the rising phase of leaf senescence in the study area (July to October, Barlow et al. 2007). Each stream reach was sampled only once in this period by placing a set of 32 buckets, each 10 m apart on each side of the channel over three days. Each bucket was placed in the stream margin up to 1 m from the water edge, tied by a rope onto the nearest two trees, and free of any understory obstruction. Each bucket had a circular opening of 0.0615 m2 (total area sampled at each stream was 1.97 m2). During each sampling period, two streams were measured at the same time. After the sampling period, I collected and stored the material for further analysis and placed the buckets at the next pair of sampling sites. In the laboratory, I sorted the material into the following items: leaves, branches, flowers, fruits, insects, and residue (small fragments of the previous items). The sorted material was dried at 60 °C for 24 h to determine dry mass. We then used the dry mass of all items to calculate the total plant litter flux (mass/m2/day).  Stream water conductivity and temperature were recorded for a period of 32 h by the same O2 probes. I then averaged the recorded values for a period of 24 h. Water samples for suspended sediments and ions (1 L and 60 ml polyethylene bottles, respectively) were collected close to the stream surface in the middle of the stream at the downstream site just before the release of the NaCl slug. I determined total suspended sediments by filtering the sample through precombusted and weighed 0.7 µm glass microfiber filters (Whatman Grade GF/F, Sigma-Aldrich, MO, USA) and drying filters to constant mass. I then subtracted the initial weight of the filter from the final weight after filtration and drying (APHA, 2005). For ions analysis, I followed the standards of the Laboratory of Sustainable Systems Analysis, Embrapa Amazônia Oriental, Belém. I added a preservative (thymol) to the sample immediately after sampling in the field and then kept it refrigerated at ~5 °C until filtration and further analysis. In the Embrapa laboratory, samples were filtered through 0.4 µm polycarbonate filters (Millipore, Billerica, MA, USA) and stored at 5 °C until analysis by ion chromatography (Dionex DX-120, Sunnyvale, CA, USA). Standard solutions (Shimadzu, Columbia, MD and Environmental Research Associates, Arvada, CO, USA) were used with all analyses for quality assurance.   78  4.2.4 Statistical analysis  For data analysis I followed the same statistical procedures described in Chapter 3. The response variables in this chapter were community respiration and gross primary production obtained from the 2-station method. I included 15 continuous variables considered to be important to stream metabolism based on the literature. These variables were grouped in different sets of variables: riparian vegetation structure (basal area, stem density, canopy height, canopy density, light intensity), organic matter input (litter and LW), water quality (temperature, conductivity, nitrate, suspended sediments), stream flow (discharge and velocity), and channel size (wetted width). My categorical variables were the same described in Chapter 3. To reduce the number of predictors and retain the most important ones in each set of variables, I first did a correlation analysis in order to select those that had an influence on my response variables (r ≥ 0.30). The correlation analysis also identified those predictors that were highly correlated with each other (r ≥ 0.70). To deal with correlated predictors in some set of variables (riparian forest structure and channel hydraulics), I used the PCA axis 1 for riparian forest structure (basal area and canopy measures; see previous Chapter 2) and retained only discharge in the set of channel flows.  After predictor selection and construction of models (see Chapter 3), I had a final candidate set of 16 models for respiration and 10 for primary productivity. I fitted my models using Generalized Linear Models specifying the Gaussian probability distribution (the responses were continuous). In order to improve data normality and model fit, the responses were square-root transformed. I then selected the best supported models using the Akaike Information Criteria corrected for small sample sizes (AICc) as described in Chapter 3. I considered the models with a ΔAICc ≤ 2 to be selected as the best models for further interpretation. The analyses were done in R (R Development Core Team, 2016).  4.3 Results  79  The results for basal area, stem density, and canopy variables were presented in Chapter 2. The results for LW volume, channel width, and discharge were presented in Chapter 3. The ANOVA results showed that only litter and primary production were different among management treatments (Table 4.1). For plant litter flux, treatment YRPB had the highest (4.46 ±0.50 g/m2/day; mean ±S.D.) value and YR the lowest (2.04 ±0.18 g/m2/day; Figure 4.1a). Treatment YRF had the lowest value of suspended sediments (0.81 ±0.50 mg/L) and treatments YRPB and MFPB had the highest values (2.16 ±0.75 and 2.04 ±0.50 mg/L, respectively; Figure 4.1b). Nitrate concentration was higher in treatment YRPB (0.40 ±0.23 mg/L) and lower in treatment OR (0.18 ±0.05 mg/L); treatments REF and MFPB had similar values (0.22 ±0.06 and 0.23 ± 0.08 mg/L, respectively; Figure 4.1c). All treatments had similar values of water temperature (around 25.7 ±0.93 oC). The lowest value of conductivity was observed for treatment OR (13.7 ±0.82 µS/cm) while treatments YRF and YRPB (16.5 ±2.29 and 16.2 ±3.11 µS/cm, respectively) and YR and MFPB had similar values (15.7 ±2.38 and 15.7 ±0.43 µS/cm, respectively; Figure 4.2a). Treatment REF had the highest value of water velocity (0.11 ±0.005 m/s) and treatment YRF the lowest (0.05 ±0.01 m/s); treatments YR and OR had similar values of water velocity (0.07 ±0.06 and 0.08 ±0.02 m/s, respectively; Figure 4.2b).   Table 4.1 Results of pairwise comparison tests among riparian management treatments. A = YR. B = YRF. C = OR. D = REF. E = YRPB. F = MFPB.  Variable Probability Tukey Means contrast Litter <0.01 E>A,B,C,D,F Suspended sediments 0.22 NS Nitrate 0.54 NS Temperature 0.42 NS Conductivity 0.59 NS Water velocity 0.29 NS Respiration 0.58 NS Primary production 0.02 A>F     80  Figure 4.1 Bar graphs showing means (± 1 S.D.) of the predictor variables for each management treatment. Litter flux (A), suspended sediments (B), and nitrate (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.          YR OR REF YRPBYRF MFPBA B C D E FManagement groupsLitter flux (g/m2/day)012345A B C D E FManagement groupsSuspended sediments (mg/L)0.00.51.01.52.02.53.03.5A B C D E FManagement groupsNitrate (mg/L)0.00.10.20.30.40.50.60.7A BCYR OR REF YRPBYRF MFPBYR OR REF YRPBYRF MFPB81  Figure 4.2 Bar graphs showing means (± 1 S.D.) of the predictor variables for each management treatment. Conductivity (A) and water velocity (B). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.   For the 2-station method, treatments YRPB and MFPB had the highest values of CR24 (10.8 ±7.66 and 10.56 ±6.54 g O2/m2/day, respectively) while treatment YR had the lowest (3.51 ±2.97 g O2/m2/day); treatments YRF and REF had similar values (7.50 ±2.60 and 7.51 ±6.36 g O2/m2/day, respectively; Figure 4.3a). This pattern was also observed for NEP (Figure 4.5a). Treatments YR and OR had the highest values and also the largest standard errors for GPP (0.60 ±0.37 and 0.30 ±0.41 g O2/m2/day, respectively) while the other treatments had similar values (Figure 4.4a). The GPP/CR24 ratio was lowest for treatments MFPB and YRF (0.007 ±0.01 and 0.01 ±0.01, respectively) and highest for treatments YR and REF (0.34 ±0.25 and 0.29 ±0.50, respectively; Figure 4.6a). The global average for CR24, GPP, and NEP were 7.76 (± 5.99), 0.20 (± 0.31), and -7.55 (±6.12) g O2/m2/day, respectively.  For the 1-station downstream method, treatment MFPB had the highest value of CR24 (18.66 ±20.33 g O2/m2/day) and treatment YR the lowest (4.12 ±2.45 g O2/m2/day); treatment YRF was lower than treatment YRPB (7.84 ±4.73 and 25.38 ±27.76 g O2/m2/day, respectively) and treatment REF was lower than MFPB (10.14 ±6.25 and 18.66 ±20.33 g YR OR REF YRPBYRF MFPBA B C D E FManagement groupsConductivity (uS/cm)05101520A B C D E FManagement groups Water velocity (m/s)0.000.050.100.15YR OR REF YRPBYRF MFPBA B82  O2/m2/day, respectively; Figure 4.3b). For GPP, treatment REF had the highest value (0.33 ±0.18 g O2/m2/day) and YRPB the lowest (0.09 ±0.10 g O2/m2/day; Figure 4.4b). For NEP, the pattern among groups was again similar to that observed for CR24 (Figure 4.5b). The GPP/CR24 ratio was lowest for treatment YRPB (0.01 ±0.01) and highest for treatment YR (0.08 ±0.06; Figure 4.6b). The global average for CR24, GPP, and NEP were 12.3 (± 16.29), 0.23 (± 0.20), and -12.15 (±16.34) g O2/m2/day, respectively.  For the 1-station upstream method, treatment YRPB had the highest value of CR24 (11.27 ±4.47 g O2/m2/day) and treatment YR the lowest (4.06 ±2.88 g O2/m2/day); treatment YRF was lower than treatment YRPB (7.67 ±4.42 and 11.27 ±4.47 g O2/m2/day, respectively) and treatment REF was lower than MFPB (10.86 ±6.45 and 17.46±17.91 g O2/m2/day, respectively; Figure 4.3c). For GPP, treatments REF and YR had the highest value (0.43 ±0.30 and 0.28 ±0.29 g O2/m2/day, respectively) and YRPB the lowest (0.05 ±0.09 g O2/m2/day; Figure 4.4c). For NEP, the pattern among groups was similar to that observed for CR24 (Figure 4.5c). The GPP/CR24 ratio was lowest for treatment YRPB (0.007 ±0.01) and highest for treatment YR (0.07 ±0.05; Figure 4.6c). The global average for CR24, GPP, and NEP were 9.77 (± 9.37), 0.22 (± 0.24), and -9.54 (9.40) g O2/m2/day, respectively.  Comparing the 2-station and the 1-station methods, the values for CR24 were similar between them, except for the values observed for treatment YRPB and MFPB in the 1-station downstream method. For GPP, the values were similar for both 1-station methods, and comparing the three methods, the greatest differences were observed for sites REF and OR in the 2-station method. For NEP, these comparisons are the same for CR24. For the GPP/CR24 ratio, the results for both 1-station methods were very similar, and comparing the three methods, the greatest differences were observed for sites YR, REF, and OR in the 2-station method.  The results showed that NEP was negative for all treatments and all methods, meaning that CR24 is higher than primary production (even for treatment YR). Also, there is an overall trend of increasing CR24 and NEP from deforested/young deforested to older 83  regeneration and reference sites, being the greatest for the riparian buffer sites. For GPP, the observable trend for both upstream and downstream 1-station method is a decrease of GPP from deforested/young deforested to older, being higher at reference sites. However, GPP values were very low, even for the YR in the 2-station method (the highest GPP value reported in this work). All sites had GPP/CR24 lower than 1. All these measures show that the systems are heterotrophic and are consuming more energy than they produce.  Figure 4.3 Bar graphs showing means (± 1 S.D.) of respiration (CR24) for each management treatment. Two-station (A), 1-station downstream (B), and 1-station upstream (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.     A B C D E FManagement groupsRespiration (gO2/m2/day)0102030405060A B C D E FManagement groupsRespiration (gO2/m2/day) – 1 station downstream0102030405060A B C D E FManagement groupsRespiration (gO2/m2/day) – 1 station upstream0102030405060AYR OR REF YRPBYRF MFPB YR OR REF YRPBYRF MFPBYR OR REF YRPBYRF MFPBRespiration (gO2/m2/day)Respiration (gO2/m2/day) –1 station downstreamRespiration (gO2/m2/day) –1 station upstreamBC84  Figure 4.4 Bar graphs showing means (± 1 S.D.) of primary production (GPP) for each management treatment. Two-station (A), 1-station downstream (B), and 1-station upstream (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.            A B C D E FManagement groupsPrimary Production (gO2/m2/day)0.00.20.40.60.81.01.2A B C D E FManagement groupsPrimary Production (gO2/m2/day) – 1 station downstream0.00.20.40.60.81.01.2A B C D E FManagement groupsPrimary Production (gO2/m2/day) – 1 station upstream0.00.20.40.60.81.01.2AYR OR REF YRPBYRF MFPBPrimary Production (gO2/m2/day)Primary Production (gO2/m2/day) –1 station downstreamPrimary Production (gO2/m2/day) –1station upstreamBCYR OR REF YRPBYRF MFPBYR OR REF YRPBYRF MFPB85  Figure 4.5 Bar graphs showing means (± 1 S.D.) of net ecosystem productivity (NEP) for each management treatment. Two-station (A), 1-station downstream (B), and 1-station upstream (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.            A B C D E FManagement groupsNet Ecosystem Productivity (gO2/m2/day)0-10-20-30-40-50-60A B C D E FManagement groupsNet Ecosystem Productivity (gO2/m2/day) – 1 station downstream0-10-20-30-40-50-60A B C D E FManagement groupsNet Ecosystem Productivity (gO2/m2/day) – 1 station upstream0-10-20-30-40-50-60AYR OR REF YRPBYRF MFPBNet Ecosystem Productivity (gO2/m2/day)Net Ecosystem Productivity (gO2/m2/day) –1 station downstreamNet Ecosystem Productivity (gO2/m2/day) –1 station upstreamBCYR OR REF YRPBYRF MFPBYR OR REF YRPBYRF MFPB86  Figure 4.6 Bar graphs showing means (± 1 S.D.) of primary production/respiration ratio (GPP/CR24) for each management treatment. Two-station (A), 1-station downstream (B), and 1-station upstream (C). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.    Considering only the 2-station method (Table 4.2), the correlation results showed that basal area, stem density, canopy height, canopy density, light intensity, litter flux, conductivity, discharge, and water velocity were correlated with CR24, and basal area, stem density, canopy height, canopy density, light intensity, and discharge, and water velocity were correlated with GPP. The results of model selection (Table 4.3, Appendix A.2) showed that conductivity, discharge, riparian forest regeneration, and riparian forest A B C D E FManagement groupsPrimary Production/Respiration ratio0.00.20.40.60.81.0A B C D E FManagement groupsPrimary Production/Respiration ratio – 1 station downstream0.00.20.40.60.81.0A B C D E FManagement groupsPrimary Production/Respiration ratio – 1 station upstream0.00.20.40.60.81.0AYR OR REF YRPBYRF MFPBPrimary Production/Respiration ratioPrimary Production/Respiration ratio –1 station downstreamBCPrimary Production/Respiration ratio –1 station upstreamYR OR REF YRPBYRF MFPBYR OR REF YRPBYRF MFPB87  structure are the main predictors that explain CR24. Riparian forest regeneration and riparian forest structure were the main predictors explaining GPP. Respiration increased with conductivity (Figure 4.7a) and was lower in secondary initial (SI) class than in mature forest (MF) and secondary late (SL) classes (Figure 4.8a). Discharge had a negative effect on CR24 while forest structure had a positive effect (Figures 4.7b and 4.8b). Primary production was higher in SI sites than MF and SL (Figure 4.9a). Forest structure had a negative effect on GPP (Figure 4.9b). The riparian buffer management treatments had no significant effect on CR24 and GPP.  Table 4.2 Results of correlation analysis between forest and stream environmental predictors and stream CR24 and GPP. Values in bold indicate r ≥ 0.30. 2ST – 2-station. DW – 1-station downstream. UP – 1-station upstream.   CR24    GPP    2ST DW UP  2ST DW UP BA 0.305 0.281 0.379  -0.352 -0.146 -0.043 SD 0.310 0.237 0.223  -0.445 -0.237 -0.093 HEIG 0.352 0.312 0.475  -0.355 -0.009 0.078 CANO 0.425 0.412 0.523  -0.411 -0.207 -0.066 LIGHT -0.300 -0.383 -0.521  0.520 0.147 0.021 LWDE 0.169 0.057 0.150  -0.069 0.125 0.180 LWVO -0.050 -0.066 -0.046  0.038 0.183 0.30 LIT 0.311 0.473 0.449  -0.047 -0.308 -0.382 SED 0.089 0.376 0.251  0.179 -0.051 -0.157 NIT 0.036 -0.135 -0.131  -0.013 0.201 0.215 TEMP -0.184 -0.112 -0.244  0.206 -0.137 -0.267 CON 0.465 0.288 0.063  -0.210 -0.114 -0.147 DIS -0.432 -0.491 -0.501  0.414 0.354 0.343 VEL -0.324 -0.215 -0.160  0.263 0.469 0.395 CWI 0.110 -0.021 0.028  -0.083 0.089 0.200 BA = basal area. SD = stem density. HEIG = canopy height. CANO = canopy density. LIGHT = light intensity. LWDE = LW density. LWVO = LW volume. LIT = litter. SED = suspended sediments. NIT = nitrate. TEMP = temperature. CON = conductivity. DIS = discharge. VEL = water velocity. CWI = channel width.    88  Table 4.3 Results of model selection for respiration and primary production. Model df AICc Delta_AICc Weight R2adj Respiration (CR24)         CON 3 75.35 0 0.33 0.18 DIS 3 76.25 0.89 0.21 0.15 RFR+CON 5 76.96 1.61 0.15 0.25 PC1RF 3 77.11 1.76 0.14 0.11       Primary production (GPP)    RFRa 4 9.73 0 0.51 0.31 PC1RF 3 11.44 1.72 0.22 0.2 aSI>MF and SL / MF=SL after Tukey multiple comparisons Predictors: RFR – Riparian Forest Regeneration. PC1RF – PCA axis 1 for riparian forest structure. CON – Conductivity. DIS – Discharge.                   89  Figure 4.7 Relationships between respiration (CR24) and conductivity (A) and discharge (B) in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○).  14 16 18 201234Conductivity (uS/cm)Respiration (gO2/m2/day)ARespiration (gO2/m2/day)0.05 0.10 0.151234Discharge (m3/s)Respiration (gO2/m2/day)B90  Figure 4.8 Relationships between respiration (CR24) and conductivity (A) and forest structure (B) in 24 streams. Codes and colors in A: SI – Secondary initial (Δ, red line). SL – Secondary late (+, black line). MF – Mature forest (○, green line).Codes in B: YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○).  -2.0 -1.5 -1.0 -0.5 0.0 0.51234PCA Axis 1 for forest structureRespiration (gO2/m2/day)BRespiration (gO2/m2/day)14 16 18 201234Conductivity (uS/cm)Respiration (gO2/m2/day)A91  Figure 4.9 A - Bar graphs showing means (± 1 S.D.) of gross primary production (GPP) for levels of the regeneration treatment. SI – Secondary initial. SL – Secondary late. MF – Mature forest. B - Relationship between GPP and forest and canopy structure in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○).  MF SI SLRegeneration groupsPrimary production (gO2/m2/day)0.00.20.40.60.81.01.2-2.0 -1.5 -1.0 -0.5 0.0 0.50.00.20.40.60.81.0PCA Axis 1 for forest structurePrimary production (gO2/m2/day)MF SI SLRegeneration groupsPrimary production (gO2/m2/day)0.00.20.40.60.81.01.2MF SI SLRegeneration groupsPrimary production (gO2/m2/day)0.00.20.40.60.81.01.2MF SI SLRegeneration groupsPrimary production (gO2/m2/day)0.00.20.40.60.81.01.2Regeneration groupsABPrimary Production (gO2/m2/day)92   4.4 Discussion  4.4.1 Environmental controls of stream metabolism  In this study, CR24 was explained by conductivity, discharge, forest regeneration, and forest structure. Primary productivity was best explained by forest regeneration and forest structure. Conductivity is a measure of ions dissolved in the water (Allan and Castillo, 2007), and possibly, respiration increases in areas with lots of nutrients. Several microcosm experiments reported an increase in litter processing rates by microorganisms after an increase in nutrient concentrations (Ferreira and Chauvet, 2011a; Kominoski et al. 2015; Piggott et al. 2015). However, I did not find an effect of nutrient (nitrate) in my study. Lovatt et al. (2014) found that decomposition of alder leaves increased the amount of ammonium, phosphate, and dissolved organic carbon (DOC) in the stream water, but not nitrate. In this study, if respiration is affected by nutrients generated from decomposition, perhaps other forms of nutrients might have affected my respiration measures. In this study, the concentrations of ammonium and phosphate were below the equipaments’ detection levels. Regarding the effect of discharge, high discharges can increase the concentration of DOC and fine particulate organic carbon (FPOC) in stream water due to organic materials carried from the terrestrial ecosystem and also due to re-suspension from deeper layers of sediments (Uehlinger and Naegeli, 1998; Acuña et al. 2011; O’Connor et al. 2012). High DOC and FPOC tend to increase CR24 (Jones and Lennon, 2015). My sites did not face high discharges from storm events during sampling as reported in other studies. It is possible that low discharges may facilitate microbial and invertebrate activity by reducing shear stress, and hence, increase metabolic rates and respiration. This is partially addressed by Giorgio and Ferreira (2000), which state that low flows increase heterotrophic activity. However, the authors also say that increased heterotrophic activity, coupled with low rates of gas exchange with the atmosphere, also increase daily anoxia and hence decrease CR24.  I found the highest respiration rates in the buffer sites YRPB and MFPB. These values were higher than those reported for the middle-age (6.22 g O2/m2/day) and old-93  growth (2.75 g O2/m2/day) streams of Bechtold et al. (2017). This finding can be due to the higher amount of organic material in these stream reaches, i.e. MFPB sites had the highest amount of LW inside the channel while YRPB had the highest amount of litter flux into the channel. Moreover, even in the young regenerated/deforested sites I observed that respiration was higher than primary productivity, probably due to the high amount of LW in these deforested reaches. Studies have found that respiration is higher when the amount of benthic organic matter is also high, especially during the fall season in temperate areas when trees lose their leaves (Acuña et al. 2004). This benthic organic material increases the concentration of dissolved organic carbon (DOC) by decomposition, and ultimately, increases respiration and decreases primary production by increasing DOC concentration for microbial respiration and reducing light availability (by light attenuation) for primary producers (Jones and Lennon, 2015). Therefore, my metabolism measures may also have been affected by DOC, a factor not measured in my study. Also, it is possible that the streams of this region, even when being deforested in the past, never lose their heterotrophic status despite experiencing an increase in primary productivity.   4.4.2 Effects of riparian management and regeneration on stream metabolism  Riparian forest regeneration and riparian forest structure were important for explaining both CR24 and GPP. As observed in Chapter 2, riparian forest structure is a result of forest regeneration in my study sites. The results showed that GPP was higher in areas with poor canopy cover (e.g. in the initial stages of forest regeneration) and lower in areas where the canopy is very dense and shades the stream (e.g. intermediate stages of forest regeneration). The opposite was observed for CR24. This result follows the light and litter dynamics reported in studies of forest succession. As stand development advances and forests recover their structure (density, biomass, and height), so canopy cover re-establishes, reducing light penetration to the ground and increasing the input of organic material (Guariguata and Ostertag, 2001; Franklin et al. 2002; Barlow et al. 2007; Kaylor et al. 2017). This results in increasing respiration and reduced primary productivity in the stream (McTammany et al. 2007; Giling et al. 2013; Bechtold et al. 2017). However, the GPP increased a bit in mature sites, and my estimates of GPP for my OR and REF sites 94  were higher than younger sites in some situations. Bechtold et al. (2017) found the greatest GPP values for old-growth riparian forests when comparing riparian forests with different developmental stages and canopy structure. They attributed this result to the small to moderate disturbances dynamic (canopy gaps) and increased light availability in these old-growth sites compared to mid-aged sites that have a uniform canopy and a very shaded understory. Warren et al. (2013) also found high levels of light reaching the channels in complex, old-growth riparian forest compared to stream reaches with simpler, second-growth riparian forests. In contrast to Warren’s finding, our measure of light intensity was the lowest for reference sites, and canopy height and openness were similar among forested sites in this study. My GPP values for REF (0.43 g O2/m2/day) and OR (0.33 g O2/m2/day) were lower compared to the values found by Bechtold et al. (2017) for middle-age (0.81 g O2/m2/day) and old-growth (0.77 g O2/m2/day) streams of New York and New Hampshire states in USA.   Riparian buffer management practices had no effect on the metabolism measures studied here. Therefore, my hypothesis was not supported by my data. I expected to find high primary production than respiration in sites YRPB and MFPB when compared to YRF and REF, respectively. My expectations were based on the effects that riparian management could have on riparian forest and canopy structure. For example, I expected that riparian buffers with younger forests surrounded by pastures would behave like forest patch edges. These patch edges are dominated by secondary tree species with short life cycles, which had a high abundance of woody vines growing on tree stems and the canopy, and they are also more exposed edge effects (Chen et al. 1992; Murcia, 1995; Gehlhausen et al. 2000; Laurance et al. 2000, 2006a; Schwartz et al. 2017). All these attributes of patch edges increase tree mortality, which in turn will decrease canopy height, increase canopy openness, and increase light incidence (Putz, 1990; Laurance et al. 2000; Harper et al. 2005). This possibly could increase primary production in the channel when compared to streams with younger forests surrounded by mature forest, where the riparian forests would not be exposed to edge effects, grow faster, and possibly recovering canopy cover characteristics. In the case of riparian buffers with mature forest surrounded by pasture, I expected to find high primary production in these streams due to recent selective logging 95  that occurred in these sites, which could have increased canopy gaps in the forest, increasing the amount of light reaching the stream. What I found was a higher respiration rate in YRPB and MFPB when compared to YRF and REF, respectively, but this result was not significant. Regarding primary productivity, I found that it was lower in YRPB and MFPB when compared to YRF and REF, respectively, when measured by the 1-station method (both downstream and upstream). When measured by the 2-station method, primary productivity was very similar when I make the same comparisons above.  For my regenerated buffers, I hypothesize that edge effects and the agricultural disturbances were not strong enough to delay forest regeneration in the buffers. Hence, riparian forests in these buffers are recovering canopy cover just like larger forested areas, which are also recovering the heterotrophic status of the streams. This might be true as I observed low intensity fires burning the edges of the riparian forests, but not advancing towards the stream margins.  Riparian buffers of mature forests can influence stream shading and light flux. Kiffney et al. (2003) found higher values of photosynthetically active radiation, water temperature, and periphyton biomass in their clear-cut and forested buffer treatments (10 and 30 m wide buffers) than at unharvested sites. They also found that these variables were negatively associated with buffer width. They mentioned that forest canopy removal and buffer narrowing increased the amount of sunlight reaching the channel through the canopy and through the sides of the buffer, which further increased stream water temperature and periphyton growth. Kiffney et al. (2003) concluded that a forested riparian buffer 30 m wide or more along both stream margins are necessary to minimize the changes associated with timber harvesting adjacent to stream ecosystems. A study conducted by Brosofske et al. (1997) also found changes in riparian microclimate (air temperature, soil temperature, surface air temperature, relative humidity, short-wave solar radiation, and wind speed) in riparian buffers surrounded by recent forest harvesting.   Studies in temperate regions evaluating the effects of partial harvest (similar to the MFPB buffers studied here) on riparian-buffer stand characteristics found that light 96  incidence increased according to the amount of basal area removed and also to the size of the gaps created in the canopy (Zenner et al. 2012; Mallik et al. 2014). Tree removal will create gaps in the canopy that are supposed to increase the amount of light incidence, and hence, increase primary production in the stream. In the study of Brosofske et al. (1997), increased light incidence within the buffer was the main factor that altered stream microclimate. However, they did not measure stream metabolism. In my study, no changes were observed in canopy structure, light incidence, and metabolism measures when comparing MFPB buffers and REF sites. Similar to my results, Kreutzweiser et al. (2009a) found that partial-harvest logging in riparian buffers of boreal mixed-wood forests sustained effective canopy cover (even at the highest average removal of the total basal area (28%) in their study), which did not increase surface stream light incidence and water temperature. For my buffers with mature forest, I hypothesize that the clearing of the adjacent area, the buffer implementation, and the selective logging are too recent, so edge effects and forest degradation by logging were still ongoing and might affect my metabolism measures only in the future – I observed canopy openings by several cut trees in the riparian area and also recently fallen trees spanning the channel in the most degraded stream reaches (Figure 4.10). As presented in Chapter 2, the removal of targeted trees is already showing some signs of forest structural degradation in these sites. If this degradation advances in the future, the stream metabolism might respond with a possible increase in GPP due to canopy gaps and increased sunlight.   There are other factors which may affect GPP in my streams. Suspended sediments and dissolved organic carbon (DOC) are factors that decrease light penetration in the water column, decreasing light availability for primary producers (Jones and Lennon, 2015). I found that sites MFPB had the second highest amount of suspended sediments. I also found that my deforested sites YR were still heterotrophic and had low primary production, possibly by the high amount of LW discarded on these sites. The LW decomposition, although slow, might release and increase DOC in the water. Sites MFPB also had the highest amount of LW in the streams. Therefore, even future light increase due to canopy degradation may not cause an overall increase in GPP in contrary to my predictions. I observed filamentous algae growing in the most canopy degraded sections, especially if the 97  water was shallow with sand substrate, so possibly the increase in primary production might be very local. Long-term monitoring studies are necessary to access the possible impacts of selective logging in these riparian and stream ecosystems.                             98  Figure 4.10 Riparian canopy opening in two selective logging sites by (A) fallen trees in the riparian area and (B) recent fallen trees spanning the stream channel.   AB99  Chapter 5: Influence of riparian forest management practices, regeneration, and leaf type on decomposition in streams of Amazonia, Pará State, Brazil  5.1 Introduction  Streams are small ecosystems that have their ecological processes largely influenced by the adjacent riparian forests. Being narrow, forested streams are mostly covered by the forest canopy, which reduces the amount of sunlight reaching the channel and supplies lots of coarse particulate organic material (CPOM) for these ecosystems (Fisher and Likens, 1973; Likens and Bormann, 1974; Richardson and Danehy, 2007). Once in the channel, leaves (the most important fraction of CPOM) start to leach, followed by the colonization of fungi and bacteria that start the decomposition process for further processing by shredders (Webster et al. 1999). Decomposition is a key process in streams and is affected by biological (e.g., leaf characteristics and invertebrate community; Gonçalves et al. 2007; Tanaka et al. 2016) and environmental factors (e.g., stream discharge (Rueda-Delgado et al. 2006), water temperature (Ferreira and Chauvet, 2011a and b; Piggott et al. 2015; Martins et al. 2018), and nutrients (Gulis and Suberkropp, 2003; Ferreira and Chauvet, 2011a; Kominoski et al. 2015). The structure and composition of riparian forests is also important for the decomposition process because it can affect the amount and composition of leaves that enter the channel (Canhoto and Graça, 1996; Kiffney and Richardson, 2010; Kominoski et al. 2011) and the detritivore fauna that lives in the streams (Tanaka et al. 2016). It may also influence the decomposition rate by controlling water temperature (Kiffney et al. 2003; Gomi et al. 2006b) as some experimental studies have shown that increased temperatures increased decomposition by microbial activity (Ferreira and Chauvet, 2011a and b; Piggott et al. 2015; Martins et al. 2018). Due to the importance of the decomposition process to stream ecosystems, it is considered an indicator of ecosystem functioning after land-use changes and riparian forest degradation (Gessner and Chauvet, 2002; Lecerf and Richardson, 2010; Chauvet et al. 2016).  Anthropogenic activities such as deforestation and forest degradation remove or change forest canopy along the stream margins or nearby (Kiffney et al. 2003; Paula et al. 100  2011; Zenner et al. 2012; Mallik et al. 2014). As a result, light incidence reaching the stream and stream water temperatures may increase (Brosofske et al. 1997; Mosisch et al. 2001; Kiffney et al. 2003; Gomi et al. 2006b; Bechtold et al. 2017). Also, the input of organic material from the riparian forest to the stream will be temporarily reduced and later altered by fast-growing shrubs (Kiffney and Richardson, 2010; Hoover et al. 2011; Kominoski et al. 2011). During the initial phases of regeneration, riparian forests will partially recover canopy characteristics and riparian tree composition will mostly be composed of fast-growth secondary species. Hence, these changes in canopy structure and composition may increase sunlight and organic material reaching the channel, further increasing microbial activity and decomposition (Kominoski et al. 2011). Protecting riparian forests in agricultural landscapes is essential to influence light incidence, water temperatures, leaf input, and maintain decomposition within natural ranges.  The most common riparian management strategy in logging and agricultural areas is the implementation of riparian buffers (Kiffney et al. 2003; Gomi et al. 2006a; Brasil, 2012; Richardson et al. 2012). However, these buffers may be exposed to disturbances originating from the surrounding cleared areas. Edge effects, for example, increase tree mortality, creating large canopy gaps and reducing canopy cover, further increasing light incidence and temperature inside the forest stand (Putz, 1990; Laurance et al. 2000; Harper et al. 2005), which together with wind, will increase litterfall (Sizer et al. 2000; Vasconcelos et al. 2004). This litter will be mostly composed of secondary tree species, which are a high quality resource (high nutrient, and low carbon and secondary compounds content; Coley, 1983; Dantas and Phillipson, 1989; Sizer et al. 2000; Kominoski et al. 2013). These changes will also affect the adjacent stream ecosystem (Brosofske et al. 1997; Kiffney et al. 2003; Bechtold et al. 2017), possibly leading to a change in leaf litter input (Kiffney and Richardson, 2010; Hoover et al. 2011; Kominoski et al. 2013), microbial activity, and organic matter decomposition (Ferreira and Chauvet, 2011a and b; Piggott et al. 2015; Martins et al. 2018) when compared to streams with riparian forests not exposed to edge effects.  101  My objective in this study is to evaluate the effectiveness of riparian buffer management in agricultural landscapes of the Brazilian Amazon. I also evaluated the effects of possible differences of riparian forest composition on stream decomposition rates by estimating leaf decomposition rates for two secondary and two primary tree species that were dominant in the riparian areas of the region. The study design to address this objective was described in Chapter 1. Here, I hypothesized that the riparian buffers will provide less shading for the channel, which may increase water temperature and leaf litter decomposition when compared with a regenerated riparian forest resulting from land abandonment that is surrounded by mature forest (for the regenerated buffer) and to mature riparian reference sites (for the logged buffer). I also predicted that litter decomposition from fast-growing trees will be faster in the buffer sites due to forest alterations.  5.2 Methods  5.2.1 Study area, experimental design, and data collection  The study area, site selection, and the experimental design were described in Chapter 2. Riparian forest measures were described in Chapter 2. Stream channel measures were described in Chapter 3. Stream metabolism measurements were described in Chapter 4.   5.2.2 Leaf decomposition experiment  I started the leaf decomposition experiment by searching for the most common riparian trees in the area. I used a previous list of species from the surrounding areas and also did field inspections with the assistance of a local botanist during site selection. After visiting streams in primary and secondary forests, I collected botanical material from the most common species on each site and took them to the Laboratory of Botany of Embrapa Amazônia Oriental for further identification. I selected four species (Figure 5.1), two from primary forests (Macrolobium angustifolium (Fabaceae) and Licania sp. (Chrysobalanaceae)) and two from secondary forests (Siparuna bifida (Siparunaceae) and 102  Henriettea succosa (Melastomataceae)). We used the following code to represent the species in this work: Ma - M. angustifolium, Li - Licania sp., Si - S. bifida, and He – H. succosa. The criteria used to select the species was whether they were the most abundant in at least two visited sites. I then collected approximately 1 kg of green leaves for each species from different tree individuals and at least in two different sites (to avoid home-field advantage effect on leaf decomposition (Jackrel and Wootton, 2014)); even if the site collected was in a stream to be sampled, the leaf collection was done further downstream of the sampling site. The leaves were taken to the laboratory, and for each species separately, I mixed together the leaves in order to increase variability of leaf characteristics. I placed leaves in paper bags and put them in the oven to dry at 65 °C degrees until they reached constant mass (approximately 24 hours), except for H. succosa leaves, which required more time. After drying, I prepared a total of 288 packs (regular plastic bags) each containing 10 g of leaves. The packs were carefully transported to the sampling sites inside boxes to avoid breakage.  I installed three sets of four leaf bags (three bags for each species, 12 in total) in each stream reach during September and October 2015. At the sampling site, I filled 12 buckets with stream water and placed the leaves into the water to wet and soften the material to facilitate leaf handling and to avoid breakage. I prepared the leaf bags (1 cm plastic mesh size) and inserted a numbered plastic tag to identify the species since during the decomposition process leaves may become undistinguishable. Due to differences in leaf size, bag dimensions were slightly different for each species: 17 x 16 cm for M. angustifolium and S. bifida, 20 x 18 cm for H. succosa, and 24 x 20 cm for Licania sp. Each bag in a set was tied with a small rope (around 30 cm distant from each other) and the set of bags was then tied to a nearby tree or root in the stream margin. Each set of bags was placed around two meters apart from each other in the reach. I placed the bags in slow-moving waters, such as pool and channel margins, backwater pools, and runs. The water depth where bags were placed varied according to stream depth, but all bags were completely submerged. Bags were incubated for a period of 60 days to ensure at least 50% of mass loss for the fast decomposition species based on previous studies in the area (unpublished data). During retrieval, I carefully removed each bag from the stream bottom 103  and immediately placed it inside an individual bucket with stream water for transport to the field station. Leaf bags were kept refrigerated and processed within the next 24 or 32 hours. In the field station, leaves were washed using tap water and placed in a bucket for air drying. I then placed the dried leaves in paper bags, placed in the oven at 65 °C for 24 hours, and stored for further analysis. I weighed the dry material to determine the remaining dry mass and then burned the material in a muffle furnace at 500 °C for 4 hours to determine ash mass. The percent ash-free dry mass (% AFDM) remaining was obtained based on Benfield (2007). I then regressed the natural log of % AFDM remaining on days of exposure, where the slope of the regression is the decomposition rate (k).                        104  Figure 5.1 Leaf samples for the four tree species selected in this study. A. Macrolobium angustifolium (Fabaceae). B. Licania sp. (Chrysobalanaceae). C. Siparuna bifida (Siparunaceae). D. Henriettea succosa (Melastomataceae). Pen size (15 cm) is used for scale.  105   5.2.3 Leaf chemical analysis  I collected leaves from different individuals for each species, placed in the stove at 65 °C for 24 hours, processed in a Wiley mill, and sent to the laboratory. Samples were digested using nitric-perchloric acid for phosphorous (P), potassium (K), calcium (Ca), magnesium (Mg), and sulfur (S) determination and sulfuric acid for nitrogen (N) determination. Phosphorous, K, Ca, Mg, and S were determined by atomic absorption spectrophotometer and N by the standard Kjeldahl technique. Lignin was extracted using alcohol-toluene and ethyl-alcohol solutions and determined by the Klason method. Hemicelluloses were determined by the oxidation of lignin using a solution of sodium chlorite and acetic acid. The lignin and hemicellulose analysis was done in the Laboratory of Bromatology of the “Luiz de Queiroz” College of Agriculture/University of São Paulo/Brazil. The macronutrient analysis was done in the Laboratory of Applied Ecology of the same institution.  5.2.4 Statistical analysis  I began with one-way ANOVAs for the decomposition variables and dissolved oxygen. The other predictors considered here were tested in previous chapters. I then did a correlation analysis between the decomposition rates for each species and the forest and stream environmental predictors. Later, I ran separate model selections for each species individually, then a single analysis with all species. However, in addition to the management and regeneration categorical variables, I was also interested in the effect of leaf species on decomposition. I tested the combinations of categorical plus continuous predictors, the latter of which were considered important to decomposition based on the literature. These variables were the same as considered in previous chapters and were grouped in different sets: riparian vegetation structure (basal area, stem density, canopy height, canopy density, light intensity), stream respiration, water quality (temperature, conductivity, nitrate, suspended sediments, dissolved oxygen), stream flow (discharge and velocity), and channel size (wetted width).  106   My final candidate set of models contained 31 models for each species individually and 65 models for overall decomposition. For each species, I fitted my models using Generalized Linear Models specifying the Gaussian probability distribution (the responses were continuous). For overall decomposition, I used Generalized Linear Mixed Models (GLMM) due to our data structure (all leaf types were present in each stream). My responses had a normal distribution so no transformations were used. I then selected the best supported models using the Akaike Information Criteria corrected for small sample sizes (AICc) as described in Chapter 3. After model selection, if a model with only a single categorical variable was selected, we used Tukey multiple comparisons of means to see which level differed from each other. The analyses were done in R (R Development Core Team, 2016).  5.3 Results  The results for all continuous predictors (except dissolved oxygen) considered here were presented in previous chapters. The results of correlation analysis for responses and predictors are presented in Appendix A.3. The ANOVA results showed that only decomposition rate by species was significant (Table 5.1). Riparian management treatments had no significant effect on the decomposition variables (Table 5.1). Dissolved oxygen (Figure A.2) had an overall mean of 6.01 mg/L ±0.65 (mean ± S.D.), and REF treatment had the highest values (6.45 ±0.42 mg/L) while treatment YRPB had the lowest values (5.76 ±0.66 mg/L).   For species He, the decomposition rate (as Ln(%AFDMloss/day)) was highest for treatment OR (0.011 ±0.0027), while treatments REF (0.0086 ±0.0021), YR (0.0087 ±0.001), and YRF (0.0089 ±0.0006) had the lowest values (Figure 5.2a). For species Si, treatments YR (0.0076 ±0.00066) and YRPB (0.0074 ±0.00066) had the lowest values, while REF (0.0086 ±0.00069), OR (0.0086 ±0.0015), and YRF (0.0084 ±0.0015) had the highest values (Figure 5.2b).   107  For species Ma, treatments YR (0.0039 ±0.0011) and YRPB (0.0042 ±0.00078) had the lowest values while REF (0.0059 ±0.00078), YRF (0.0055 ±0.0011), and OR (0.0055 ±0.00019) had the highest values (Figure 5.2c). For species Li, treatment YR (0.0011 ±0.00045) had the lowest value and MFPB (0.002 ±0.00087) had the highest value (Figure 5.2d).   Table 5.1 Results of pairwise comparison tests among riparian management treatments and leaf species. Variables related to forest and canopy structure were tested in Chapter 2, LW and channel in Chapter 3, and water quality and respiration in Chapter 4. Variable Probability Tukey Means contrast Dissolved oxygen 0.77 NS k <0.001 He>Si,Ma,Li; Si>Ma,Li; Ma>Li k - He 0.52 NS k - Se 0.57 NS k - Ma 0.06 NS k - Li 1.33 NS  Comparing our treatments YRF and YRPB, there is a fairly strong effect size of buffer treatment for all decomposition measures (11.1% higher for He, 12.3% lower for Si, 22.8% lower for Ma, and 19.9% lower for Li in comparison with YRF), and the variation between these treatments was large (except for species Ma). For treatments REF and MFPB, the effect size was also strong (15.5% higher for He, 5.1% lower for Si, 17.9% lower for Ma, and 27.4% higher for Li in comparison with REF), however, the variation was more constant between treatments (except for species He). Except for species He, there is an observable trend of decrease in decomposition rates in more altered sites, like YR and YRPB, while the decomposition rates tend to increase in treatments YRF and OR to levels observed in REF sites. The decomposition rates tended to be higher on MFPB sites compared to YR and YRPB sites, but lower than in YRF, OR, and REF sites (except for He and Li species).   Considering the decomposition rate among species (Figure 5.3), He had the highest decomposition rate (0.0096 ±0.0020), followed by Si (0.0081 ±0.0011), Ma (0.0050 ±0.0011), and Li (0.0017 ±0.00067).  108   Figure 5.2 Bar graphs showing means (± 1 S.D.) of the decomposition rates for each management treatment. Species H. succosa (A), S. bifida (B), M. angustifolium (C), and Licania sp (D). YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.        Ak(Ln (% AFDM/day))BCA B C D E FManagement groupsk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014DA B C D E FManagement groupsk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014A B C D E FManagement groupsk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014A B C D E FManagement groupsk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014YR OR REF YRPBYRF MFPB YR OR REF YRPBYRF M PB0.000.0020.0040.0060.0080.010.0120.0140.000.0020.0040.0060.0080.010.0120.014109  Figure 5.3 Bar graphs showing means (± 1 S.D.) of decomposition rate for each leaf species.    For cellulose and lignin content, species He and Si had the lowest values and species Li and Ma had the highest values (Table 5.2). For macronutrient content, the lowest values were observed for species He and Li (except Ca for He) and the highest values for Si and Ma.   Table 5.2 Leaf cellulose, lignin, and nutrient content for four riparian tree species.  Cellulose Lignin N P K Ca Mg S Species  %  g Kg -1 Henriettea succosa  21.5 7.57 13.2 0.9 3.7 7.3 1.5 0.8 Siparuna bifida  23.58 8.83 23.5 1.3 6.7 4.1 1.9 1.1 Licania sp.  30.26 33.65 11.7 1.0 4.2 2.4 1.5 0.9 Macrolobium angustifolium  31.64 33.06 19.3 1.3 6.7 4.3 1.6 1.1  The results of model selection showed that leaf species was the main predictor of decomposition rate (Table 5.3, Appendix A.4). Evaluating the species individually, forest regeneration and forest structure were selected among the first best models for all species (Table 5.3, Appendix A.5). In general, the decomposition rates increased with an increase in forest structure (high basal area, high canopy height, high canopy density, and low stem MA HE SILIJA MA ME NISpeciesk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014k(% AFDM per day)Ma He SiLiJA MA ME NISpeciesk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014k(Ln (% AFDM/day))0.0000.0020.0040.0060.0080.0100.012JA MA ME NISpeciesk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014110  density and light intensity; Figure 5.4). Also, the decomposition rates were lower in young regenerated sites (group SI), remained low or were slightly higher after 8-12 years of regeneration (SL), and reach the highest levels at mature forest (MF) sites (Figure 5.5). In addition to forest predictors, suspended sediments, flow, temperature, conductivity, dissolved oxygen, respiration, and channel width were also among the best selected models (Figures A.3 and A.4). Also, the management treatments had no effect on the decomposition rate for any species analyzed.                          111  Table 5.3 Results of model selection for decomposition rates. SP – leaf species. SD – stem density. RFR – riparian forest regeneration. SED – suspended sediments. DIS – discharge. TEMP – temperature. CON – conductivity. DO – dissolved oxygen. PC1RF – PCA axis 1 for forest structure. CR24 – respiration. CWI – channel width Model df AICc Delta_AICc Weight R2adj k           SP 6 -931.96 0 1 0.83       k - He      SD 3 -224.82 0 0.19 0.09 RFR 4 -223.6 1.21 0.1 0.11 SED 3 -222.84 1.91 0.07 0.01       k - Si      RFR* 4 -255.01 0 0.19 0.21 RFR+DIS 5 -254.73 0.27 0.16 0.26 RFR+TEMP 5 -253.6 1.4 0.09 0.23 RFR+CON 5 -253.41 1.6 0.08 0.22       k - Ma      RFR+DO 5 -257.15 0 0.2 0.3 PC1RF 3 -256.97 0.19 0.18 0.18 RFR** 4 -255.75 1.4 0.1 0.2 RFR+CR24 5 -255.56 1.59 0.09 0.26 DO 3 -255.37 1.78 0.08 0.12       k - Li      PC1RF 3 -277.88 0 0.17 0.07 CWI 3 -277.19 0.69 0.12 0.04 RFR 4 -276.57 1.31 0.09 0.09 *MF>SI,SL **MF>SI     112  Figure 5.4 Relationships between riparian forest structure and k - He (A), k – Ma (B), and k – Li (C) in 24 streams. YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○).          0 1000 2000 3000 4000 50000.0060.0080.0100.0120.014Stem density (individuals/ha)k (g/day)-2.0 -1.5 -1.0 -0.5 0.0 0.50.0030.0040.0050.0060.007PCA Axis 1 for forest structurek (g/day)-2.0 -1.5 -1.0 -0.5 0.0 0.50.00050.00100.00150.00200.00250.00300.0035PCA Axis 1 for forest structurek (g/day)A BCk(Ln (% AFDM/day))-1.-1.-2.0.0060.0080.0100.0120.0140.0030.0070.0040.0050.0060.0.-0.2010 4030 50Stem densi  ( ndividuals/ha) PCA Axis 1 for rest structure-1.-1.-2. 0.0.-0.PCA Axis 1 for forest structure0.00050.00250.00100.00150.00200.00300.0035k(Ln (% AFDM/day))113  Figure 5.5 Bar graphs showing means (± 1 S.D.) of the decomposition rates for each regeneration group. Species H. succosa (A), S. bifida (B), M. angustifolium (C), and Licania sp. (D). MF – Mature forest. SI – Secondary initial. SL – Secondary late.            MF SI SLRegeneration groupsk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014MF SI SLRegeneration groupsk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014MF SI SLRegeneration groupsk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014MF SI SLRegeneration groupsk (g/day) 0.0000.0020.0040.0060.0080.0100.0120.014A BC DMF SLSI F Ik(Ln (% AFDM/day))0.0000.0020.0040.0060.0080.0100.0120.0140.0000.0020.0040.0060.0080.0100.0120.014114  5.4 Discussion  5.4.1 Factors affecting decomposition rates  Leaf decomposition in this study was most influenced by leaf type. However, when considering decomposition rates by species, I found that forest structure and regeneration were also important. Plant species have different leaf structures and chemical compounds that result from traits related to the plant life cycles and ability to colonize and live in specific environments, such as in nutrient-poor soils, drought soils, or environments with high rates of herbivory (Coley, 1983; Dantas and Phillipson, 1989; Coq, 2010; Kominoski et al. 2013; García-Palacios et al. 2016). These traits can influence different ecological processes in terrestrial and aquatic ecosystems, including litter decomposition (Coq, 2010; Kominoski et al. 2013; García-Palacios et al. 2016). In terrestrial and stream ecosystems, the amount of structural or secondary compounds, and nutrients were found to be the main drivers of leaf decomposition as these leaf attributes affect leaf palatability to fungi and shredder invertebrates (Coq, 2010; Bruder et al. 2014; García-Palacios et al. 2016; Gonçalves Jr. et al. 2017). Ardón and Pringle (2008) and Gonçalves Jr. et al. (2017) found that the amount of structural compounds (cellulose, hemicellulose, and lignin) was an important factor for leaf decomposition in tropical streams, with the leaf species having lower amounts of structural compounds having the highest rates of decomposition. My results are consistent with their finding as leaf decomposition rates of my species match the gradient of cellulose and lignin content, but not nutrient concentrations.   The importance of riparian forest conditions to leaf decomposition and associated shredder fauna was observed by Tanaka et al. (2016) in a stream located in a sugar cane watershed in São Paulo State, Brazil. Tanaka et al. (2016) found that leaf decomposition rates and leaf miner Stenochironomus abundance was higher in sites located inside the forest remnant in comparison with deforested sites located upstream of the forest patch. They also found that the Stenochironomus abundance was positively related to tree density, the proportion of leaf mass loss, and DO concentrations in the water. They concluded that the presence of forest improved stream water quality, which increased leaf miner activity 115  and leaf decomposition rate. Based on macroinvertebrates sampling and identification from a previous study in this area, I found that the main taxa associated with leaf decomposition in my streams was the Stenochironomus (Hugo Saulino from the Federal University of São Carlos (personal communication)). On that previous sampling, I found Stenochironomus living inside the leaves (Figure 5.6), and also small shrimps and fishes foraging on the leaf bags (unpublished data). Leaf-miner chironomids of the genus Stenochironomus were classified as shredders by Chará-Serna et al. (2012) as they consume leaf material, so I hypothesize they are playing an important role in leaf decomposition in my sites. Ligeiro et al. (2010) and Biasi et al. (2013) also reported the importance of the family Chironomidae for leaf decomposition in streams of Brazil. Casotti et al. (2015) found that decomposition rates and shredders abundance were the lowest in streams with degraded riparian zones (deforestation, low tree cover, and exotic fruit plants), also demonstrating the importance of riparian forest conditions to leaf decomposition.  In my study, the removal of riparian trees on YR and the agricultural activities in the uplands of sites YRPB and MFPB might have changed the stream habitat and water conditions desirable for Stenochironomus. In the field, I observed some bag smothering in the YR and YRPB sites by dense fine sediments (Figure 5.7), which could have reduced access to the leaves. I also observed that some leaves became dark and stiffened after bag smothering (Figure 5.8). Studies found that leaf burying in the stream sediments leads to anaerobic conditions and chemical reactions in the leaf tissue that altered leaf chemistry and made the leaf less palatable to the organisms (Herbst, 1980; Danger et al. 2012). This observation is supported by my data since we found the highest concentration of suspended sediments in these sites. However, Sanpera-Calbet et al. (2012) did not find effects of fine sediments on leaf decomposition in experimental studies. Also, I found that decomposition rates were positively related to suspended sediments for the He species, which also had a similar decomposition rate among REF and YR sites and even higher in YRPB and MFPB when compared to REF sites. It is possible that species He, by having the fastest decomposition rate in this study, was promptly colonized and consumed by the chironomids, and when bags became smothered, most of the leaf mass was already decomposed in comparison with the other species that had lower mass loss than He. A 116  chironomid’s larval stage duration may last less than two weeks depending on species and environmental conditions, and warm water and high-quality food are factors responsible for shorter life cycles (Ferrington et al. 2008).   Figure 5.6 Leaf colonization by leaf miners Chironomidae. A. Macrolobium angustifolium leaf with signs of colonization. B. Chironomidae larvae (in red) living inside of a M. angustifolium leaf. C. Chironomidae pupae living inside of a H. succosa leaf. D. Chironomidae larvae collected from the inside and surface of leaves. All pictures were taken after a 15-day incubation period in the stream from a field experiment in 2014.   Although increases of water temperature have increased leaf decomposition in experimental studies, I found that higher temperature was associated with lower decomposition rate on the regenerating groups SL and MF for Si species, but not at deforested sites SI. In deforested sites located in agricultural areas, leaf decomposition was reported to be higher due to fungal responses to elevated temperature and nutrient inputs ADCB117  from the agricultural fields (Niyogi et al. 2003; Paul et al. 2006). In the experimental study of Martins et al. (2018), increased temperatures at intermediate levels (27.36 ±0.56 °C) increased fungal activity and leaf decomposition, but temperatures at higher levels (29.62 ±0.37 °C) decreased fungal activity and leaf decomposition.   Figure 5.7 Suspended sediments and FPOM in one YRPB stream. A. Stream reach. B. Bag smothering (bag location in the stream bottom is indicated by the red arrow). C. Leaf bag with FPOM adhered on its surface.         A BC118  Figure 5.8 Leaves of M. angustifolium retrieved after normal conditions (A) and after bag smothering (B). Retrieval after 15 days from a field experiment in 2014. Tweezer size (11.5 cm) is used for scale.   5.4.2 Effects of riparian management and regeneration on decomposition rates  Riparian buffer management practices had no significant effect on the decomposition rates. The effect size (as percent difference) of buffers was large for both my main pair of comparisons, but larger variation for the regeneration buffers indicate that large variability was more important to not finding a statistical difference. However, the variation was smaller for the logged buffer, so apparently I have more confidence about this result. Also, the fact that effect direction varies in different directions suggests something more complex in these systems. Therefore, my hypothesis was not supported by my data. I expected to find higher decomposition rates in sites YRPB and MFPB when compared to YRF and REF, respectively. My expectations were based on the effects that riparian management could have on riparian forest structure. Earlier in this work, I stated that riparian buffers might behave like forest edges, which are dominated by short-lived secondary tree species and they are also more exposed to edge effects (Murcia, 1995; Gehlhausen et al. 2000). Forest alterations reduce forest canopy height and density, further increasing light incidence and stream water temperature (Kiffney et al. 2003; Gomi et al. 2006b), which could increase biological activity on leaf decomposition, especially leaf decomposition of short-lived secondary trees that are more labile (Coley, 1983; Dantas and Phillipson, 1989). My expectations were also based on the information that increased A B119  temperature in experimental studies increased leaf decomposition by microbial activity (Ferreira and Chauvet, 2011a and b; Piggott et al. 2015; Martins et al. 2018). I found lower decomposition rates in sites YRPB and MFPB for species Si and Ma, and higher rates in sites YRPB for species He and in sites MFPB for species He and Li. The higher decomposition rates at sites YRPB and MFPB for fast-decomposing species (He) was predicted by my hypothesis, however, I did not find any statistical difference among treatments.  Studies of riparian buffers after harvesting in temperate regions found that leaf decomposition was significantly lower in completely harvested sites and also sites with large and narrow riparian buffers compared to reference sites (Kreutzweiser et al. 2008; Lecerf and Richardson, 2010). These lower rates of leaf decomposition were attributed mainly to reduction of shredder abundances and diversity due to disturbances to stream hydrology and bed morphology (Kreutzweiser et al. 2008; Lecerf and Richardson, 2010). This type of habitat degradation (sedimentation after high flows) after harvesting operations is commonly reported in the literature and riparian buffer prescriptions are expected to reduce these impacts (Gomi et al. 2006a). However, even the riparian buffer treatments were insufficient to restrain fine sediments from entering the streams in the Lecerf and Richardson (2010) study. Lecerf and Richardson (2010) also sampled leaf-decaying fungi, but they showed no response to forest harvesting. The input of fine sediment from upland harvest may also bury the leaf bags, not giving the shredders access to the leaves. Lecerf and Richardson (2010) also found lower decomposition rates in thinned sites (a condition similar to my selective logging sites) which was attributed to leaf smothering as they did not find effects on the shredder community. In contrast, Kreutzweiser et al. (2010) found no effects of riparian forest thinning on stream leaf decomposition and invertebrate communities.   The results of Chapter 2 showed that riparian forest structure results from the regeneration process, not buffer management. According to the literature, the regeneration process is important to explain leaf decomposition by recovering invertebrate community, including shredders. Stone and Wallace (1998) found the invertebrate community more 120  similar to reference sites after 16 years of forest regeneration; in this period, shredder abundance increased above reference conditions. In the same study area of Stone and Wallace (1998), Ely and Wallace (2010) found that shredder abundance was equal to reference conditions after 26 years of forest regeneration. Yeung et al. (2017) found that leaf decomposition and shredder abundance on previous harvested sites and sites with riparian buffers converged to values found at reference sites after 15 years of logging. In my study, decomposition rates followed the recovery of forest structure, showing a trajectory of recovery from slow decomposition in the YR sites to increasing decomposition in YRF, OR, and REF for all species analyzed. This trend was also observed when considering the regeneration groups. Decomposition was lower in deforested sites, and after around 10 years of forest regeneration (YRF sites), the process was already recovered to levels similar to those of REF sites.   The possible mechanisms by how regeneration is influencing decomposition is not very clear in my study. For example, it is possible that forest regeneration recovered DO levels, the leaf miner chironomids Stenochironomus, and leaf decomposition as found by Tanaka et al. (2016). I found a positive correlation between DO levels and decomposition rates for species Ma and Li, but DO levels decrease with increasing forest structure (the correlation is very weak though). It is possible that high respiration rates on forested sites reduced DO levels. In Chapter 4, I found evidence that the recovery of riparian forest structure was associated with higher respiration rates, and in this chapter I found that respiration rates were negatively related to DO levels. Decomposition showed contrasting responses to that of respiration rates (negative for Ma and positive for Si species) and the regressions were also weak. In addition to DO, it is also possible that the regeneration of larger areas including the riparian forests (YRF and OR sites) decreased suspended sediments entering the streams, also contributing to recovery of decomposition rates.  Although riparian buffer management increases riparian forests’ susceptibility to disturbances and may affect stream water temperature, I did not find a significant effect on leaf decomposition rates (although I have observed a trend of reduced decomposition in streams with riparian buffers for Si and Ma species when compared to YRF, OR, and REF 121  sites). It seems that upland agricultural activities may affect leaf decomposition by increasing sedimentation at these sites and smothering the leaves. This impact on the stream might affect leaf miners’ activities in deforested streams and in streams with degraded and regenerated riparian forests. Also, the ongoing regeneration process at my sites (except for YRPB) seems to be recovering the rates of leaf decomposition, possibly by recovering stream respiration and reducing suspended sediments.           122  Chapter 6: Conclusions and Recommendations  6.1 Overview  Riparian buffers are susceptible to disturbances, especially in agricultural areas where edge effects (press disturbance) and fires (pulse disturbance) have the potential to be a persistent disturbance for the adjacent ecosystems. Therefore, this thesis aimed to evaluate the effectiveness of two riparian management strategies, i.e. (1) land abandonment for natural regeneration and (2) maintenance of mature forest for protecting riparian forests and stream ecosystems in tropical agricultural areas. In Chapter 2, I showed that riparian forest structure in the regenerated buffer surrounded by agriculture was not different from a regenerating riparian forest surrounded by mature upland forest. I also showed that the forest structure of the selectively logged buffer was not different from the mature forest reference sites (although logged buffers showed signs of degradation – low basal area, low canopy height and density, high light intensity). I then evaluated how the characteristics of riparian forests in these buffers affected different stream ecological processes. In Chapter 3, I showed that LW in the streams was not related to current forest conditions. Rather, LW appeared to be a result of past deforestation in the area and also logging inside riparian buffers that discarded the residual wood in the channel. In Chapter 4, I showed that respiration and primary production were not different among treatments, but respiration had high values for both buffer treatments. In Chapter 5, I showed that leaf decomposition, in general, did not differ significantly among management treatments. However, decomposition rates were lower in both buffer treatments for two species and higher for one species. There were no significant differences considering the two main riparian buffer strategies in any of the measures I considered in this study.  6.2 Riparian buffer regeneration in abandoned agricultural lands  One of the objectives of this thesis was to evaluate if land abandonment for riparian forest regeneration was an effective strategy to restore riparian forests and the associated stream ecological processes in agricultural areas. Land abandonment for forest regeneration 123  is an inexpensive strategy to recover forest ecosystems (Chazdon, 2008; Shoo and Catterall, 2013; Benini et al. 2017). Currently, deforestation and land degradation is a major concern in poor and developing tropical countries (Chazdon, 2008; Shoo and Catterall, 2013; FAO, 2016), so natural regeneration in abandoned lands has a big appeal for recovery of these areas (Chazdon, 2008; Shoo and Catterall, 2013; Benini et al. 2017). However, it is important to highlight that this type of restoration action does not perform well in all situations, especially those where soil fertility was exhausted and regeneration sources were depleted (Uhl et al. 1988; Zarin et al. 2005). Also, ongoing agriculture activities can also be a source of disturbances that may delay or arrest the regeneration process in the surrounding abandoned areas (Laurance et al. 2000; Vasconcelos and Luizão, 2004; Laurance, 2006a).   Although several studies have addressed the effects of past and present land-use on the regeneration process, all these studies have been focused on upland forests (Uhl et al. 1988; Zarin et al. 2005; Griscom et al. 2009). More recently, studies have been addressing the regeneration of abandoned riparian lands in agricultural areas, studying how plant communities recovered to more natural communities present today in the area (González et al. 2016, 2017), the environmental controls of the regeneration (Thogmartin et al. 2009), the exclusion of local disturbances (Carline and Walsh, 2007; Forget et al. 2013), and the role of the surrounding landscape for tree recruitment (Forget et al. 2013). Although González et al. (2016, 2017) have found recovery of community attributes, their approach was more focused on comparing a chronosequence of stand ages, not evaluating the effectiveness of the regeneration by controlling for external disturbances as I did in my work. Also, they compared sites that were previously used by several human activities, not only agricultural sites. In my study, I grouped my sites in a gradient of stand age in order to contextualize where my regenerated buffers fall in this regeneration gradient, which allowed me to infer effectiveness specifically for an agricultural context.   Studies done in North America evaluated how riparian forests recovered their ecological attributes and their functions for streams (Warren et al. 2007; Stovall et al. 2009; Warren et al. 2013; Bechtold et al. 2017). However, these regenerating forests were located 124  primarily in previously logged areas, where large areas were abandoned or even replanted for regeneration, which also includes the upland forests. Although conditions after clearcutting differ greatly from those after natural disturbances, riparian regeneration in this condition is more likely to occur in comparison to riparian regeneration in agricultural areas because disturbances are reduced and later disappear as the upland forests regenerate (Franklin et al. 2002). Some studies evaluated the riparian recovery and its impacts on bank stabilization and in-stream features in agricultural areas by controlling for cattle disturbance at the local scale (Carline and Walsh, 2007; Forget et al. 2013). Although they have addressed a very short time scale (5 to 6 years), they were still able to find positive effects for the plant community and the streams. In my study, I addressed a longer time scale than Carline and Walsh (2007) and Forget et al. (2013), but still short compared to the time necessary for a forest community to reach a “climax” state (Budowski, 1965; Chazdon, 2014). In my study, my specific experimental design was limited in its ability to consider older ages of regeneration as older riparian forested buffers were not present in that region.  The studies done so far regarding riparian regeneration and stream recovery in an agricultural context have not provided detailed information about the effectiveness of regeneration in a forested versus agricultural context. McTammany et al. (2007) studied forest regeneration and stream recovery in abandoned agricultural lands, although they accounted for the upland regeneration, not only in the riparian buffer. I did studies about the importance of secondary riparian forests for stream ecosystems in São Paulo State, Brazil (Paula et al. 2011; 2013; 2018). I found that these secondary forests had mostly low diameter trees at the stream reach scale, and these forests were fragmented at the drainage and watershed scales, which affected their capacity to supply LW and retain fine sediments, respectively. I hypothesize that the past and current land-use in that area did not allow for an effective recovery of the riparian forests. I also sampled streams in 30 m riparian buffers, but I could not do the same comparison I did here (regeneration in a natural versus agricultural context) because there were no regenerated forests surrounded by mature forests (like my YRF treatment in this study) as old-growth forests in interior São Paulo State are very scarce (Rodrigues, 1999; Victor et al. 2005; Ribeiro et al. 2009).  125  In this study, I addressed a study design that excluded external disturbances at larger scales. Therefore, my study was the first to compare same-age riparian forests regenerating under natural versus agricultural contexts, where the surrounding disturbance effect for the regeneration is permanent and potentially detrimental to riparian regeneration. Although this comparison was done only for younger forests (8 to 12 years old), as older riparian forests in buffers still do not occur in my study area, these forests had developed sufficiently to recover structural attributes and stream functional processes.  By finding no differences among treatments and based on the PCA ordination results, I am assuming that recovery has occurred in my buffer sites, so riparian forest stand and canopy structure recovered according to the expected successional processes described in the literature (Budowski, 1965; Guariguata and Ostertag, 2001; Chazdon, 2014). From these results, I determined that within 8-12 years following agricultural abandonment in the riparian zone, riparian forests recovered structure to the same level of a regenerated riparian forest in a natural context, without the effect of persistent disturbances from surrounding agriculture as initially posed in my hypothesis (Figure 6.1a). Therefore, my results showed that riparian buffer implementation by land abandonment for forest regeneration is an effective way to restore riparian forest structure in this agricultural landscape.   6.3 Riparian buffer from maintenance of mature forests in agricultural lands  Another objective of this thesis was to evaluate if the maintenance of mature forest buffers with selective logging was an effective strategy to protect riparian forests and the associated stream ecological processes in agricultural areas. Selective logging at varied intensities is a widespread technique used in logging operations in tropical areas (Chazdon, 2014; Martin et al. 2015). However, the Brazilian legislation strictly prohibits harvesting within the riparian areas in logging concessions and also in riparian buffers in agricultural areas (Brasil, 1998). Regardless of this high level of protection, farmers take advantage of the limitations of the satellite imagery program and the ground enforcement operations to perform illegal logging within riparian buffers.   126  Although several studies already reported the effects of selective logging on the riparian plant community (Palik et al. 2012; Zenner et al. 2012; Mallik et al. 2014) and for stream ecological processes (Lecerf and Richardson, 2010; Kreutzweiser et al. 2009a and b; Yeung et al. 2017), most of them were done in temperate regions of North America and in a different context from mine. These studies compared partially harvested riparian forests versus unlogged riparian forests where the logged riparian area was surrounded by logged forests, not agriculture. They found alterations in forest community and canopy structure (Palik et al. 2012; Zenner et al. 2012; Mallik et al. 2014) and also in stream processes after logging (Lecerf and Richardson, 2010), but forest attributes began to recover (Palik et al. 2012; Mallik et al. 2014) and stream processes recovered after forest regeneration in the logged buffer (Yeung et al. 2017). Regeneration is more likely to occur in partially harvested riparian areas as the upland forests regenerate, the disturbances from the cleared area are reduced and eliminated, which contributes to a faster recovery of the logged riparian forest (Franklin et al. 2002). This is very different from logged riparian forests in agricultural areas where persistent disturbances may affect forest recovery and further increase forest degradation.   The literature provides several examples in which tropical forests lose structure and biodiversity after fragmentation (Laurance, 2000, 2006a), logging (Martin et al. 2015; Richardson and Peres, 2016), and due to a synergistic effect among disturbances acting on these forests (Laurance, 2006b; Barlow et al. 2016). Although these findings refer specifically to upland forests, the same is expected to happen in riparian forests as they have a high edge-area ratio like forest edges, but no studies were done before my study. A recent study in the Brazilian Amazon compared riparian buffers of mature forests surrounded by agriculture versus reference sites (Nagy et al. 2015). They did not find differences in woody plant biomass and carbon storage between these sites, however, these buffers were larger (75 to 325 m wide) than those required by the Brazilian legislation and they did not have selective logging within the buffers. Thus, they did not provide the information necessary to assess the effects of logging on riparian forests in agricultural areas. Therefore, my study fills this gap by comparing a buffer width closer (usually 30 m wide, but with some parts narrower or wider) to that required by the Brazilian legislation 127  (30 m for streams < 10 m wide), the presence of selective logging in the buffer, and the presence of agriculture surrounding the buffer.  I found that the structure of forest and canopy in the logged buffers were the same as reference sites, without significant effects of logging and other disturbances from the surrounding agriculture as initially expected. Therefore, my results showed that implementing riparian buffers of mature forests is an effective way to maintain riparian forest structure in this agricultural landscape. However, I found signs of forest degradation (low basal area, low canopy height and density, high light intensity) that I attributed mainly to the logging that occurred recently in these sites. I hypothesize that the logging and edge effects were minimal by the time of my sampling, but I highlight that this might change in the future if these effects get stronger by shifting the surrounding agriculture to a more intensive land-use, and if more tree removal happens in these buffers (Figure 6.1b). .  The effects of selective logging within buffers in agricultural areas in the tropics are an unexplored topic, for which I emphasize the need for more studies in other tropical regions to complement my findings. In addition to this, it is important to evaluate whether the most degraded sections of my logged buffers will recover after logging. Mallik et al. (2014) found that the plant community under different gap sizes was recovering after 7 years post-logging in riparian buffers located in a logging context, but this is not guaranteed in my sites due to the agricultural context, which may shift to other intensive land-use in the future.          128                      Figure 6.1 Revised and simplified conceptual diagram showing forest trajectories over the years after buffer implementation. A. The regeneration of forest structure in the buffer is similar to natural succession (Budowski (1965) proposed classification schemes and discussed in Chazdon (2014)) for the first 8-12 years of regeneration under mild disturbances. B. Riparian forests lost structure in the buffer after selective logging and have structural measures similar to older regenerated sites (18 – 22 years). Black arrows show forest recovery in the absence or under mild disturbances and riparian forest degradation after selective logging. Gray arrows show expected forest trajectories after the influence of disturbances. Letter X in red represents trajectories not observed in the buffers. Dotted arrows represent expected but undetermined future trajectories in this study. Numbers in parentheses refer to the treatments considered in this work and described in Chapter 1.  129      Forest and canopy structural complexityPioneer (1)Early secondary (2 and 5)Late secondary (3)Stable climax, not static (4)Strong disturbancesDeforestation (1)Regeneration under natural conditions and mild disturbancesDisturbance ceasedMild disturbancesTime after succession begins (years)AXXDisturbance remainsExtreme disturbanceNo forest successionXForest and canopy structural complexityPioneer (1)Early secondary (2)Late secondary (3)Stable climax, not static (4)Climax after disturbance (6)Disturbance ceasedDisturbance remainDeforestation (1)Extreme disturbanceDisturbanceNo forest successionTime after succession begins (years)BStrong disturbance130  6.4 Limitations and future perspectives to better manage riparian forests and protect streams in agricultural landscapes  From my results, I concluded that the two riparian buffer management strategies addressed here, i.e. (1) land abandonment for natural regeneration and (2) maintenance of mature forest, were effective for protecting riparian forest and stream ecosystems. However, there are some limitations in my study that require close attention when using my results for management purposes. First, my study area was located within a region where sustainable management forestry was the main commercial activity, that is, the landscape matrix was forest with large patches (1,400 to 5,000 ha) of agricultural activities distributed over this matrix. This means that seed sources are available from the surrounding forests. This is the opposite condition of what is more common in the world, where the landscape matrix is agricultural with patches of remaining degraded forests (Ribeiro et al. 2009; Forget et al. 2013; Chazdon, 2014; Barlow et al. 2016; Paula et al. 2018). This condition is what predominates in the deforestation arc in the Amazon and also in other regions of Brazil, like São Paulo State. In São Paulo State, the deforestation process is older and spans almost 200 years (Victor et al. 2005). Forest regeneration in São Paulo started mostly after the approval of the Forest Code in 1965. However, these regenerated and remaining forests are constantly exposed to the agricultural activities, which might compromise their recovery. My past studies in the Corumbatai River basin, São Paulo State, showed that most regenerated forests are at initial and intermediate stages of succession and they are less likely to provide ecological function for streams (Paula et al. 2011; 2013; 2018). These sites had lower stem density and basal area than my younger (YRF and YRPB) and older (OR) regenerated sites. Although I do not know the exact age for all these riparian forests, I assume that they are older than 10 years, and a maximum age of 40 years (the difference between my sampling year (2006) and the year that the Forest Code (1965) was implemented). The oldest site that I sampled there is approximately 28 years old (Paula et al. 2018), an older age than my OR sites (18 – 22 years), but it still had almost half basal area than my OR sites. Riparian forests of that region seem to be recovering slower than those I sampled in this study. Therefore, I hypothesize that land abandonment in that area is not as effective as what I observed in this study. 131   It is important to say that in areas where the deforestation process is very recent, my findings can be very informative for management and restoration programs as the land probably was not extensively used and the surrounding large patches of conserved forests can provide the regeneration sources. In my study area, the current disturbance seemed not to go deeper into the buffer, the past land-use seemed not to deplete the propagules (roots and stems to resprout), and the surrounding forests seem to provide seeds via frugivorous fauna that I observed moving around adjacent to the pasture fields and within the buffers. Regeneration success likely depends on the intensity of the persistent disturbances that these forests are subject to. Based on my earlier work, I expected a stronger effect from disturbances, but as I found here, the disturbances were apparently mild and the landscape seemed to favor rapid regeneration.  The second limitation is that the duration of my study was too short when considering that forests in the YRPB buffers are still in their initial stages of regenerating and forest in the MFPB buffers are possibly still being degraded by logging and edge effects. Also, my sampling considered only one year of weather conditions, which did not address potential inter-annual variation in the processes studied. Ecological processes in these forests and streams may follow different trajectories in the future from what I have observed here. For instance, it is unclear if the forests on the YRPB treatments will achieve a mature state (will it advance or keep the current forest structure?) and if the MFPB treatments will lose forest structure and change composition (by losing big trees over time and increasing the abundance of secondary species) or if the recovery process will succeed. Also, it is important to consider how these future trajectories will impact stream ecosystems. Therefore, I emphasize that my findings are limited to the current conditions I observed in the field (weak edge effects and low levels of riparian forest degradation after selective logging). I recommend long-term monitoring studies to assess how the structure of the riparian forests in these buffers will change through time and how this can affect the different stream ecological processes I studied here.   132  Monitoring the regeneration in these buffers is also important to ensure an adequate supply of LW in the future as the remaining in-stream LW will decompose. Riparian forest regeneration takes decades to recover the desired structure and composition to supply more effective wood to the channel. During this recovery time, the channel lacks habitat structure provided by LW if most pieces were lost during forestry or agricultural activities (Fausch and Northcote, 1992; Paula et al. 2011). In temperate zones this time is expected to be more than 50 years (Andrus et al. 1988). In my study, I found that the OR sites had a similar diameter distribution as REF sites, and they had no significant difference in basal area from REF sites. Apparently, the forest structure at these sites recovered more quickly to supply adequate LW than in temperate regions, although forest composition (another important aspect for LW dynamics) has not been quantified at this moment (LW from hardwood trees decompose slower than soft wood). Therefore, if the ongoing regeneration advances successfully at my regenerated buffers, the current LW in the channel might fill this gap between forest regeneration and the start of wood supply, without severely affecting stream ecological processes and its biodiversity.  The third limitation is that for my logged buffer, although I have found signs of degradation, I was not able to disentangle the effects of selective logging from edge effects. Apparently, selective logging was the main cause of this observed degradation, but to confirm this, I should have included in my study a riparian buffer of mature forest without logging, which I did not find in the nearby areas where I was working. Therefore, I recommend future studies address this issue in order to complement my findings and better understand the effectiveness of riparian buffers of mature forest in agricultural areas of the tropics.  In this study, I evaluated stream ecological processes that are expected to be more affected by alterations in forest cover at local scales (riparian forest in the study reach). However, these processes may also be affected by forest alterations at larger spatial scales. For example, an increase in fine sediments is usually associated with an increased runoff caused by deforestation in the catchment and also by forest cover fragmentation in the drainage network (Sutherland et al. 2002; Liébault et al. 2005; Paula et al. 2018). These 133  excessive fine sediments may smother the stream bottom or be suspended in the water column, reducing decomposition and primary production levels, respectively (Lecerf and Richardson, 2010; O’Connor et al. 2012). On the other hand, these same forest alterations in the landscape may increase nutrient concentrations in stream water (Johnson et al. 1997; Taniwaki et al. 2017), potentially increasing decomposition and primary production levels (Schiller et al. 2008; Clapcott and Barmuta, 2010). Even LW may be affected by these changes of forest cover in the landscape, as increased runoff increases stream discharge and downstream transport of small pieces of wood (Paula et al. 2013). These forest alterations at larger scales were not addressed here, but they may also be important for explaining the unexplained variance of my response variables, especially my functional measures (metabolism), as discharge seems to not affect wood transport in these streams.  Another important thing that I did not consider in my study is the previous land-use intensity of my sites. I know they were all pasture sites, but specific information about the previous management of these pasture sites was not available. As shown by Uhl et al. (1988), land-use history was an important factor that affected the regeneration in previous pasture sites in the Amazon. For further studies, I recommend trying to include this information in the analysis. For my future work on this topic, if I do not have access to this information, I will include the variable “time used as pasture/agriculture” estimated by satellite imagery as a proxy for land-use intensity.   In this study, I found that riparian forest structural variables are recovering on the 30 m buffers. However, recent modifications of the Brazilian environmental legislation reduced the buffer widths in agricultural areas established before 2008 (see Soares-Filho et al. 2014). By reducing buffer width, it is expected that these forests will become more vulnerable and less effective at protecting streams as found in some studies (Wenger, 1999; Kiffney et al. 2003; Bahuguna et al. 2010). I hypothesize that if my buffers were narrower than 30 m, some of the processes studied here would be more affected as I observed surface fires burning the edges of one of my buffers and the edges of nearby forests in that region. However, my study did not provide answers for this question, so I recommend further studies on this topic to ensure that riparian areas and streams are effectively being protected 134  in agricultural landscapes, especially about how the regeneration would be affected in these narrower buffers.  In general, I found that the two buffer management strategies were effective for protecting streams as I did not find differences between the main comparisons done in this work. However, some results suggested that buffer management might have an effect despite finding no significant differences. For example, LW abundance and volume, although they seem to result from past deforestation in the area, they had the highest values in the logged buffers, suggesting that selective logging was also responsible for the LW increase in the streams with logged buffers. Litterfall was greater in the regenerated buffer, suggesting that buffer exposure to wind and a dryer environment might have increased litterfall in these buffers which were dominated by fast-growing species. Respiration in both buffer treatments was higher than reference sites, suggesting that the buffer might have an unexpected effect on respiration, possibly by increasing organic matter input (LW and litterfall) in these buffers. Also, leaf decomposition had different responses in buffers among the species evaluated. These results showed that some stream responses to buffer management are complex, highlighting the need for future studies about riparian buffers and stream responses in agricultural landscapes.   Studies about riparian buffers have compared how forests in riparian buffers differed from unlogged old-growth forests in terms of forest (Braithwaite and Mallik, 2012) and stream protection (Gomi et al. 2006a). Also, they compared whether different buffer widths were effective for protecting riparian and stream ecosystems (Wenger, 1999; Kiffney et al. 2003; Clinton, 2011; Nagy et al. 2015; Yeung et al. 2017), and evaluated how partial harvesting in riparian areas were effective in comparison to unlogged buffers and reference sites (Kreutzweiser et al. 2009a, 2010; Palik et al. 2012; Zenner et al. 2012; Mallik et al. 2014). From these studies and others, there is a literature available that answers our main questions about what we can and cannot protect with each of these different riparian management strategies. However, in tropical agricultural areas, there is little information specifically for the two buffer management strategies I studied here. Therefore, I recommend more studies about these particular forms of riparian management, 135  especially focusing on land abandonment for riparian regeneration as forest regeneration is increasing around the globe (Fearnside, 1996; Houghton et al. 2000; FAO, 2016).   In Brazil, natural regeneration is one option of riparian buffer implementation specified in the legislation. However, it is necessary to consider that several ecosystem attributes and processes in both riparian areas and streams may not be protected in the same way by implementing riparian buffers from mature forests. Although forest structural attributes and stream functional attributes have a fast recovery (as I observed here), LW supply and forest composition take longer to recover - considering that regeneration will successfully advance in agricultural buffers. By establishing different management strategies for riparian buffers, we have the general goal of protecting water resources, ecosystems, and biodiversity (Richardson et al. 2012). However, each of these strategies may have different outcomes, and to ensure that riparian areas and streams are being best-protected, we need to know what we can actually protect through each of these different management strategies.   6.5 Management implications  My results showed that riparian buffers from natural regeneration and mature forests were effective at protecting riparian and stream ecosystems. In the Amazon, the estimated cost of riparian restoration by tree planting varies from R$7,430 to R$17,492 per hectare (US$2,300 to US$5,455), which involves soil site preparation, planting, and the maintenance of the seedlings which need extensive care in the first years after planting (Benini et al. 2017). By comparison, natural regeneration in areas with high potential for regeneration requires only ant control and is estimated to cost R$180 (US$56.15) per hectare (Benini et al. 2017), and leaving a buffer of mature forest has no cost. Not all farmers have the financial resources available to restore their deforested riparian areas, which makes land abandonment for natural regeneration an option for them to be in compliance with the current legislation. The results I found can also be informative for other tropical countries that face deforestation issues, have a similar landscape context, and have limited resources to restore their lands. However, it is extremely important that the 136  natural regeneration be continuously monitored in order to evaluate the effectiveness of the regeneration process. In situations where the area was not intensively used for agriculture and native forest patches are present in the landscape, the regeneration sources will be readily available for the succession process.   In the Amazon region, most abandoned-pasture areas that are currently undergoing natural forest regeneration were lightly to moderately used in the past, so they have a high potential for natural regeneration (Uhl et al. 1988; Fearnside, 1996). Estimates show that 30 to 47.6% of previously deforested areas in the Amazon are covered by secondary forests (Fearnside 1996; Houghton et al. 2000). In 2010, there was approximately 56,369 ha (44%) of riparian areas to be reforested by active or passive restoration in my study area (Nunes et al. 2015). However, forest fires are becoming more common in the Amazon region (Barlow and Peres, 2006; Laurance, 2006b; De Faria et al. 2017) and this has the potential to slow down the regeneration process, as repeated fires decreases carbon accumulation, increases tree mortality, and impacts soil nutrients and physical properties (Reich et al. 2001; Barlow and Peres, 2004; Zarin et al. 2005). Although the Brazilian legislation already recommends the controlled use of fire in agricultural areas, fire is not very well managed in the field. Therefore, improving the use of fire in these agricultural landscapes seems to be important to the dynamics of the forests in this region, avoiding forest degradation and allowing forest regeneration in targeted areas, such as the riparian areas. Also, conservation programs that highlight the impacts of unmanaged fires on the surrounding forests are also important.   Regarding riparian buffers of mature forest, it would be important to keep selective logging away from these sites. In agricultural areas, in addition to illegal selective logging, riparian buffers are susceptible to edge effects and other disturbances from the surrounding agricultural areas. My results showed signs of forest degradation after selective logging of the buffer, which I attributed to the logging activity. Also, I assume that this degradation could be even worse if the disturbances from the agricultural area were stronger. Therefore, logging in riparian areas has the potential to degrade riparian forests and should be avoided in agricultural areas. In my study, for example, I hypothesize that the illegal logging in the 137  buffers caused an increase in amount of in-stream LW by discarding this material into the channel.  To fight illegal logging in riparian buffers, it would be important to increase law enforcement operations and also create alternative means (like tax incentives or payment for environmental services provision) to compensate farmers that already comply with the environmental legislation. These alternative means would recognize farmers’ responsible practices, their important role in providing ecosystem services, and avoid selective logging in riparian buffers as a means to increase profits. Similar mechanisms already exist, such as the Ecological Tax Program in Brazil (ICMS Ecológico; The Nature Conservancy, 2018) and the Reducing Emissions from Deforestation and Forest Degradation in Developing Countries (REDD+; UN-REDD Programme, 2016), but the financial resources are directed to local governments and nations, respectively, and not directly to farmers, so they may or may not be attended. I consider that these alternative means would bring more economical and ecological benefits than allowing selective logging in the buffers. Therefore, I recommend more discussions and future studies on this topic.  138  References  Acuña V, Giorgi A, Muñoz I, Uehlinger U, Sabater S (2004) Flow extremes and benthic organic matter shape the metabolism of a headwater Mediterranean stream. Freshwater Biology 49:960-971.  Acuña V, Guasch H, Giorgi A, Izagirre O (2009) Flujo de energía en el ecosistema: metabolismo fluvial. In: Elosegi A, Sabater S (eds). Conceptos y técnicas en ecología fluvial. Fundación BBVA, Bilbao.  Acuña V, Vilches C, Giorgi A (2011) As productive and slow as a stream can be – the metabolism of a Pampean stream. Journal of the North American Benthological Society 30:71-83.  Anderson NH, Sedell JR, Roberts LM, Triska FJ (1978) The role of aquatic invertebrates in processing of wood debris in coniferous forest streams. The American Midland Naturalist 100:64-82.  Andrus CW, Long BA, Froehlich HA (1988) Woody debris and its contribution to pool formation in a coastal stream 50 years after logging. Canadian Journal of Fisheries and Aquatic Sciences 45:2080–2086.  Allan JD, Castillo MM (2007) Stream ecology: structure and function of running waters. 2nd edition. Springer, New York.  Allmendinger NE, Pizzuto, JE, Potter NJ, Johnson TE, Hession WC (2005) The influence of riparian vegetation on stream width, eastern Pennsylvania, USA. Geological Society of America Bulletin 117: 229-243.  APHA – American Public Health Association (2005) Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington.  Ardón M, Pringle CM (2008) Do secondary compounds inhibit microbial- and insect-mediated leaf breakdown in a tropical rainforest stream, Costa Rica? Oecologia 155:311-323.  Aristegi L, Izagirre O, Elosegi A (2009) Comparison of several methods to calculate reaeration in streams, and their effects on estimation of metabolism. Hydrobiologia 635:113-124.  139  Arnold TW (2010) Uninformative parameters and model selection using Akaike’sInformation Criterion. The Journal of Wildlife Management 74:1175-1178.  Bahuguna D, Mitchell ST, Miquelajauregui Y (2010) Windthrow and recruitment of large woody debris in riparian stands. Forest Ecology and Management 259:2048-2055.  Barlow J, Peres CA (2004) Ecological responses to El Niño-induced surface fires in Central Brazilian Amazonia: management implications for flammable tropical forests. Philosophical Transactions of the Royal Society of London B 359:367-380.  Barlow J, Gardner TA, Ferreira LV, Peres CA (2007) Litter fall and decomposition in primary, secondary and plantations forests in the Brazilian Amazon. Forest Ecology and Management 247:91-97.  Barlow J, Lennox GD, Ferreira J, Berenguer E, Less AC, Nally RM, Thomson JR, Ferraz SFB, Louzada J, Oliveira VHF, Parry L, Solar RRC, Vieira ICG, Aragão LEOC, Begotti RA, Braga RF, Cardoso TM, Oliveira Jr RC, Souza Jr CM, Moura NG, Nunes SS, Siqueira JV, Pardini R, Silveira JM, Vaz-de-Mello FZ, Veiga RCS, Venturieri A, Gardner TA (2016) Anthropogenic disturbance in tropical forests can double biodiversity loss from deforestation. Nature 535:144-147.  Bastos TX, Pacheco NA, Figueiredo RO, Silva GFG (2005) Características Agroclimáticas do Município de Paragominas. Documentos 228. Embrapa Amazônia Oriental, Belém.  Bechtold HA, Rosi EJ, Warren DR, Keeton WS (2017) Forest age influences in-stream ecosystem processes in Northeastern US. Ecosystems 20:1058-1071.  Benda LE, Bigelow P, Worsley TM (2002) Recruitment of wood to streams in old-growth and second-growth redwood forests, northern California, USA. Canadian Journal of Forest Research 32:1460-1477.  Benfield EF (2007) Decomposition of leaf material. In: Hauer R, Lamberti GA (eds). Methods in Stream Ecology. Academic Press, San Diego.  Benke AC, Wallace JB (2003) Influence of wood on invertebrate communities in streams and rivers. In: Gregory SV, Boyer KL, Gurnell AM (eds). The ecology and management of wood in world rivers. American Fisheries Society, Bethesda.  Benini RM, Lenti FEB, Tymus JRC, Silva APM, Isernhagen I (2017) The costs of restoring native vegetation in Brazil. In: Benini RM, Adeodato S (eds). Forest restoration economy. The Nature Conservancy, São Paulo. 140   Berra TM (2001) Freshwater fish distribution. Academic Press, San Diego.  Biasi C, Tonin AM, Restello RM, Hepp LU (2013) The colonisation of leaf litter by Chironomidae (Diptera): the influence of chemical quality and exposure duration in a subtropical stream. Limnologica 43:427-433.  Bilby RE, Ward JW (1989) Changes in characteristics and functions of woody debries with increasing size of streams in western Washington. Transactions of the North American Fisheries Society 118:368-378.  Bilby RE, Ward JW (1991) Characteristics and function of large woody debris in streams draining old-growth, clear-cut, and second-growth forests in southwestern Washington. Canadian Journal of Fisheries and Aquatic Sciences 48:2499–2508.  Bilby RE, Bisson PA (1998) Function and distribution of large woody debris. In: Naiman RJ, Bilby RE (eds). River ecology and management: lessons from the Pacific Coastal Ecoregion. Springer, New York.  Bilby RE, Heffner JT (2016) Factors influencing litter delivery to streams. Forest Ecology and Management 369:29-37.  Bisson PA, Bilby RE, Bryant MD, Dolloff CA, Grette GB, House RA, Murphy ML, Koski KV, Sedell JR (1987) Large Woody Debris in Forested Streams in the Pacific Northwest: Past, Present, and Future. In: Salo EO, Cundy TW (eds). Streamside Manegement: forestry and fisheries interactions. Institute of Forest Resources, Seattle.  Bisson PA, Montgomery DR, Buffington JM (2007) Valley segments, stream reaches, and channel units. In: Hauer R, Lamberti GA (eds) Methods in Stream Ecology. 2nd edition. Academic Press, San Diego.  Bonetto AA, Wais IR (1995) Southern South American streams and rivers. In: Cushing CE, Cummins KW, Minshall GW (eds) Ecosystems of the world: river and stream ecosystems. Elsevier Academic Press, London.  Blinn CR, Kilgore MA (2001) Riparian management practices: a summary of state guidelines. Journal of Forestry 99:11-17.  Borcard D, Gillet F, Legendre P (2011) Numerical Ecology with R. Springer, New York.   141  Boss SM, Richardson JS (2002) Effects of food and cover on the growth, survival, and movement of cutthroat trout (Oncorhynchus clarki) in coastal streams. Canadian Journal of Fisheries and Aquatic Sciences 59:1044-1053.  Bott TL (2007) Primary productivity and community respiration. In: Hauer R, Lamberti GA (eds). Methods in Stream Ecology. Academic Press, San Diego.  Boulton AJ, Boyero L, Covich AP, Dobson M, Lake S, Pearson R (2008) Are tropical streams ecologically different from temperate streams? In: Dudgeon D (ed). Tropical Stream Ecology. Academic Press, San Diego.  Boyero L (2000) Towards a global stream ecology. Trends in Ecology and Evolution 15:390-391.  Boyero L, Ramírez A, Dudgeon D, Pearson RG (2009) Are tropical streams really different? Journal of the North American Benthological Society 28:397-403.  Braithwaite NT, Mallik AU (2012) Edge effects of wildfire and riparian buffers along boreal forest streams. Journal of Applied Ecology 49:192-201.  Brancalion PHS, Cardozo IV, Camatta A, Aronson J, Rodrigues RR (2014) Cultural ecosystem services and popular perceptions of the benefits of an ecological restoration project in the Brazilian Atlantic Forest. Restoration Ecology 22: 65-71.  Brasil (1965) Lei no 4771 de 15 de setembro de 1965.  Brasil (1998) Lei no 9605 de 12 de fevereiro de 1998.  Brasil (2012) Medida Provisória no 571 de 25 de maio de 2012.  Braun EL (1935) The undifferentiated deciduous forest climax ant the association-segregate. Ecology 16: 514-519.  Brejão GL, Hoeinghaus DJ, Pérez-Mayorga MA, Ferraz SFB, Casatti L (2017) Threshold responses of Amazonian stream fishes to timing and extent of deforestation. Conservation Biology DOI:10.1111/cobi.13061  Brito JG, Martins RT, Oliveira VC, Hamada N, Nessimian JL, Hughes RM, Ferraz SFB, Paula FR (2018) Biological indicators of diversity in tropical streams: congruence in the similarity of invertebrate assemblages. Ecological Indicators 85:85-92.  142  Brosofske KD, Chen J, Naiman RJ, Franklin JF (1997) Harvesting effects on microclimate gradients from small streams to uplands in Western Washington. Ecological Applications 7:1188-1200.  Bruder A, Schindler MH, Moretti MS, Gessner MO (2014) Litter decomposition in a temperate and a tropical stream: the effects of species mixing, litter quality and shredders. Freshwater Biology 59:438-449.  Buckup PA, Menezes NA, Ghazzi MS (2007) Catálogo das espécies de peixes de água doce do Brasil. Museu Nacional, Rio de Janeiro.  Budowski G (1965) Distribution of tropical American rain forest species in the light of successional processes. Turrialba 15:40-42.  Bunn SE, Davies PM, Mosisch TD (1999) Ecosystem measures of river health and their response to riparian and catchment degradation. Freshwater Biology 41:333-345.  Burnham KP, Anderson DR (2002) Model selection and multimodel inference: a practical information-theoretic approach. Springer, New York.   Cadol D, Wohl E (2010) Wood retention and transport in tropical, headwater streams, La Selva Biological station, Costa Rica. Geomorphology 123:61-73.  Canhoto C, Graça MAS (1996) Decomposition of Eucalyptus globulus leaves and three native leaf species (Alnus glutinosa, Castanea sativia and Quercus faginea) in a Portuguese low order stream, Hydrobiologia 333:79-85.  Carline RF, Walsh MC (2007) Responses to riparian restoration in the Spring Creek watershed, Central Pennsylvania. Restoration Ecology 15:731-742.  Carpenter SR, Caraco NF, Correll DL, Howarth RW, Sharpley AN, Smith VH (1998) Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecological Applications 8:559-568.  Casotti CG, Kiffer Jr. WP, Costa LC, Rangel JV, Casagrande LC, Moretti MS (2015) Assessing the importance of riparian zones conservation for leaf decomposition in streams. Natureza & Conservação: Brazilian Journal of Nature Conservation 13:178-182.  Chapin FS III, Matson PA, Vitousek PM (2011) Principles of Terrestrial Ecosystem Ecology. 2nd edition. Springer, New York.  143  Chará-Serna AM, Chara JD, Zúñiga MDC, Pearson RG, Boyero L (2012) Diets of leaf litter-associated invertebrates in three tropical streams. Annales de Limnologie – International Journal of Limnology 48:139-144.  Chauvet E, Ferreira V, Giller PS, Mckie BG, Tiegs SD, Woodward G, Elosegi A, Dobson M, Fleituch T, Graça MAS, Gulis V, Hladyz S, Lacoursière JO, Lecerf A, Pozo J, Preda E, Riipinen M, Risnoveanu G, Vadineanu A, Vought LBM, Gessner MO (2016) Litter decomposition as an indicator of stream ecosystem functioning at local-to-continental scales: insights from the European RivFunction Project. Advances in Ecological Research 55:99-182.  Chazdon RL (2008) Beyond deforestation: restoring forests and ecosystem services on degraded lands. Science 320:1458-1460.  Chazdon RL (2014) Second Growth: the promise of tropical forest regeneration in an age of deforestation. The University of Chicago Press, Chicago.  Chen J, Franklin JF, Spies TA (1992) Vegetation responses to edge environments in old-growth Douglas-Fir forests. Ecological Applications 4:387-396.  Chen X, Wei X, Scherer R, Luider C, Darlington W (2006) A watershed scale assessment of in-stream large woody debris patterns in the southern interior of British Columbia. Forest Ecology and Management 229:50-62.  Chizinski CJ, Vondracek B, Blinn CR, Newman RM, Atuke DM, Fredricks K, Hemstad NA, Merten E, Schlesser N (2010) The influence of partial timber harvesting in riparian buffers on macroinvertebrate and fish communities in small streams in Minnesota, USA. Forest Ecology and Management 259:1946-1958.  Clapcott JE, Barmuta LA (2010) Forest clearance increases metabolism and organic matter processes in small headwater streams. Journal of North American Benthological Society 29:546-561.  Clements FE (1916) Plant Succession: An Analysis of the Development of Vegetation. Carnegie Institution of Washington Publication 242, Washington.  Clewel AF, Aronson J (2013) Ecological Restoration: Principles, Values, and Structure of an Emerging Profession. Island Press, Washington.  Clinton BD (2011) Stream water responses to timber harvest: riparian buffer width effectiveness. Forest Ecology and Management 261:979-988. 144   Cochrane MA, Laurance WF (2002) Fires as a large-scale edge effect in Amazonian forests. Journal of Tropical Ecology 18:311-325.  Coley PD (1983) Herbivory and defensive characteristics of tree species in a lowland tropical forest. Ecological Monographs 53:209-234.  Coq S, Souquet J, Meudec E, Cheynier V, Hattenschwiler S (2010) Interspecific variation in leaf litter tannins drives decomposition in a tropical rain forest of French Guiana. Ecology 91:2080-2091.  Crawley JC (2007) The R book. John Wiley & Sons Ltd., Chichester.  Cristan R, Aust WM, Bolding MC, Barrett SM, Munsell JF, Schilling E (2016) Effectiveness of forestry best management practices in the United States: literature review. Forest Ecology and Management 360:133-151.  Danger M, Cornut J, Elger A, Chauvet E (2012) Effects of burial on leaf litter quality, microbial conditioning and palatability to three shredder taxa. Freshwater Biology 57:1017-1030.  Dantas M, Phillipson J (1989) Litterfall and litter nutrient concentration content in primary and secondary Amazonian “terra firme” rain forest. Journal of Tropical Ecology 5:27-36.  De Faria BL, Brando PM, Macedo MN, Panday PK, Soares-Filho BS, Coe MT (2017) Current and future patterns of fire-induced forest degradation in Amazonia. Environmental Research Letters 12:095005.  DeWalt SJ, Maliakal SK, Denslow JS (2003) Changes in vegetation structure and composition along a tropical forest chronosequence: implications for wildlife. Forest Ecology and Management 182:139-151.  Dudgeon D (2000) The ecology of tropical asian rivers and streams in relation to biodiversity conservation. Annual Review of Ecology and Systematics 31:239-263.  Eggert SL, Wallace JB (2007) Wood biofilm as a food resource for stream detritivores. Limnology and Oceanography 52:1239-1245.  Elosegi A, Díez JR, Flores L, Molinero J (2017) Pools, channel form, and sediment storage in wood-restored streams: potential effects on downstream reservoirs. Geomorphology 279:165-175. 145   Ely DT, Wallace JB (2010) Long-term functional group recovery of lotic macroinvertebrates from logging disturbance. Canadian Journal of Fisheries and Aquatic Science 67:1126-1134.  FAO - Food and Agriculture Organization of the United Nations (2016) Global Forest Resources Assessment 2015. How are the world’s forests changing? FAO, Rome.  Fausch KD, Northcote TG (1992) Large woody debris and salmonid habitat in a small coastal Bristish Columbia stream. Canadian Journal of Fisheries and Aquatic Sciences 49:682-693.  Fearnside PM (1996) Amazonian deforestation and global warming: carbon stocks in vegetation replacing Brazil’s Amazon forest. Forest Ecology and Management 80:21-34.  Ferraz SFB, Vettorazzi CA, Theobald DM, Ballester MVR (2005) Landscape dynamics of Amazonian deforestation between 1984 and 2002 in central Rondônia, Brazil: assessment and future scenarios. Forest Ecology and Management 204:67-83.  Ferreira V, Chauvet E (2011a) Synergistic effects of water temperature and dissolved nutrients on litter decomposition and associated fungi. Global Change Biology 17:551-564.  Ferreira V, Chauvet E (2011b) Future increase in temperature more than decrease in litter quality can affect microbial litter decomposition in streams. Oecologia 167:279-291.  Ferrington Jr LC, Berg MB, Coffman WP (2008) Chironomidae. In: Merrit RW, Cummins KW, Berg MB (eds). An introduction to the aquatic insects of North America. Kendal/Hunt Publishing Company, Dubuque.  Figueiredo RO, Markewitz D, Davidson EA, Schuler AE, Watrin OS, Silva PS (2010) Land-use effects on the chemical attributes of low-order streams in the eastern Amazon. Journal of Geophysical Research 115:G04004.  Fisher SG, Likens GE (1973) Energy flow in Bear Brook, New Hampshire: an integrative approach to stream ecosystem metabolism. Ecological Monographs 43:421-439.  Fitzpatrick FA, Waite IR, D’Arconte PJ, Meador MR, Maupin MA, Gurtz ME (1998) Revised methods for characterizing stream habitat in the National Water-Quality Assessment Program: U.S. Geological Survey Water-Resources Investigations Report 98-4052. U.S. Geological Survey, Raleigh.   146  Forget G, Carreau C, Coeur DL, Bernez I (2013) Ecological restoration of headwaters in a rural landscape (Normandy, France): a passive approach taking hedge networks into account for riparian tree recruitment. Restoration Ecology 21:96-104.  Francez LMB, Carvalho JOP, Jardim FCS (2007) Changes on floristic composition after logging in a terra firme forest in the region of Paragominas, PA. Acta Amazonica 37:219-228.  Franklin JF, Spies TA, Van Pelt R, Carey AB, Thornburgh DA, Berg DR, Lindenmayer DB, Harmon ME, Keeton WS, Shaw DC, Bible K, Chen J (2002) Disturbances and structural development of natural forest ecosystems with silvicultural implications, using Douglas-fir forests as an example. Forest Ecology and Management 155:399-423.  García-Palacios P, McKie BG, Handa IT, Frainer A, Hattenschwiler S (2016) The importance of litter traits and decomposers for litter decomposition: a comparison of aquatic and terrestrial ecosystems within and across biomes. Functional Ecology 30:819-829.  Gardner TA, Ferreira J, Barlow J, Less A et al. (2013) A social and ecological assessment of tropical land uses at multiple scales: the Sustainable Amazon Network. Philosophical Transactions of the Royal Society B: Biological Sciences 368:20120166.  Gehlhausen SM, Schwartz MW, Augspurger CK (2000) Vegetation and microclimate edge effects in two mixed-mesophytic forest fragments. Plant Ecology 147:21-35.  Gerwing JJ (2002) Degradation of forests through logging and fire in the eastern Brazilian Amazon. Forest Ecology and Management 157:131-141.  Gessner MO, Chauvet E (2002) A case for using litter breakdown to assess functional stream integrity. Ecological Applications 12:498-510.  Giling DP, Grace MR, Mac Nally R, Thompson RM (2013) The influence of native replanting on stream ecosystem metabolism in a degraded landscape: can a little vegetation go a long way? Freshwater Biology 58:2601-2613.  Giorgio A, Ferreira G (2000) Phytobentos colonization in a lowland stream in Argentina. Journal of Freshwater Ecology 15:39-46.  Gomi T, Sidle RC, Noguchi S, Negishi JN, Nik AR, Sasaki S (2006a) Sediment and wood acumulations in humid tropical headwater streams: effects of logging and riparian buffers. Forest Ecology and Management 224:166-175. 147   Gomi T, Moore RD, Dhakal AS (2006b) Headwater stream temperature response to clear-cut harvesting with different riparian treatments, coastal British Columbia, Canada. Water Resources Research 42:W08437, DOI:10.1029/2005WR004162.  Gonçalves Jr. JF, Graça MAS, Callisto M (2007) Litter decomposition in a cerrado savannah stream is retarded by leaf toughness, low dissolved nutrients and a low density of shredders. Freshwater Biology 52:1440-1451.  Gonçalves Jr. JF, Couceiro SRM, Rezende RS, Martins RT, Ottoni-Boldrini BMP, Campos CM, Silva JO, Hamada N (2017) Factors controlling leaf litter breakdown in Amazonian streams. Hydrobiologia 792:195-207.  González E, Masip A, Tabacchi E (2016) Poplar plantations along regulated rivers may resemble riparian forests after abandonment: a comparison of passive restoration approaches. Restoration Ecology 24:538-547.  González E, Masip A, Tabacchi E, Poulin M (2017) Strategies to restore floodplain vegetation after abandonment of human activities. Restoration Ecology 25:82-91.  Gordon ND, McMahon TA, Finlayson BL, Gippel CJ, Nathan RJ (2004) Stream hydrology: an introduction for ecologists. 2nd edition. John Wiley & Sons, Chichester.  Grace MR, Imberger SJ (2006) Stream metabolism: performing and interpreting measurements. Water Studies Centre Monash University, Murray Darling Basin Comission and New South Wales Department of Environment and Climate Change. http://www.sci.monash.edu.au/wsc/tech-manual-v3.pdf  Griscom HP, Griscom BW, Ashton MS (2009) Forest regeneration from pasture in the dry tropics of Panama: effects of cattle, exotic grass, and forested riparia. Restoration Ecology 17:117–126.  Grizzel JD, Wolff N (1998) Occurrence of windthrow in forest buffer strips and its effects on small streams in Northwest Washington. Northwest Science 72:214-223.  Guariguata MR, Ostertag R (2001) Neotropical secondary forest succession: changes in structural and functional characteristics. Forest Ecology and Management 148:185-206.  Gulis V, Suberkropp K (2003) Leaf litter decomposition and microbial activity in nutrient-enriched and unaltered reaches of a headwater stream. Freshwater Biology 48:123-134.  148  Harper KA, Macdonald SE, Burton PJ, Chen J, Brosofske KD, Saunders SC, Euskirchen ES, Roberts D, Jaiteh MS, Esseen P (2005) Edge influence on forest structure and composition in fragmented landscapes. Conservation Biology 19:768-782.  Hedman CW, Van Lear DH, Swank WT (1996) In-stream large woody debris loading and riparian forest seral stage associations in the southern Appalachian Mountains. Canadian Journal of Forest Research 26:1218-1227.  Helfman GS, Collette BB, Facey DE (1997) The diversity of fishes. Blackwell Science, Massachussets.   Henderson PA, Walker L (1986) On the leaf litter community of the Amazonian blackwater stream Tarumázinho. Journal of Tropical Ecology 2:1-17.  Herbst GN (1980) Effects of burial on food value and consumption of leaf detritus by aquatic invertebrates in a lowland forest stream. Oikos 35:411-424.  Hodgson D, McDonald JL, Hosken DJ (2015) What do you mean, ‘resilient’? Trends in Ecology and Evolution 30:503-506.  Holling CS (1973) Resilience and stability of ecological systems. Annual Review of Ecology and Systematics 4:1-23.  Hoover TM, Pinto X, Richardson JS (2011) Riparian canopy type, management history, and successional stage control fluxes of plant litter to streams. Canadian Journal of Forest Research 41:1394-1404.  Houghton RA, Skole DL, Nobre CA, Hackler JL, Lawrence KT, Chomentowski WH (2000) Annual fluxes of carbon from deforestation and regrowth in the Brazilian Amazon. Nature 403:301-304.  Hunt L, Marrochi N, Bonetto C, Liess M, Buss DF, Silva CV, Chiu MC, Resh VH (2017) Do riparian buffers protect stream  invertebrate communities in South American Atlantic Forest agricultural areas ? Environmental Management 60:1155-1170.  IMAZON – Instituto do Homem e Meio Ambiente da Amazônia. http://imazon.org.br/?lang=en Acessed: January 2018.  IPAM – Instituto de Pesquisa Ambiental da Amazônia. http://ipam.org.br/glossario/arco-do-desmatamento/ 149  Acessed: January 2018.  Jackrel SL, Wootton JT (2014) Local adaptation of stream communities to intraspecific variation in a terrestrial ecosystem subsidy. Ecology 95:37-43.  Jackson KJ, Wohl E (2015) Instream wood loads in montane forest streams of the Colorado Front Range, USA. Geomorphology 234:161-170.  Johnson LB, Richards C, Host GE, Arthur JW (1997) Landscape influences on water chemistry in Midwestern stream ecosystems. Freshwater Biology 37:193-208.  Jones SE, Lennon JT (2015) A test of the subsidy-stability hypothesis: the effects of terrestrial carbon in aquatic ecosystems. Ecology 96:1550-1560.  Jones IL, Peres CA, Benchimol M, Bunnefeld L, Dent DH (2017) Woody lianas increase in dominance and maintain compositional integrity across an Amazonian dam-induced fragmented landscape. PLoS ONE 12:e0185527. https://doi.org/10.1371/journal.pone.0185527  Kassi N’Dja JK, Guillaume D (2008) Successional patterns of plant species and community diversity in a semi-deciduous tropical forest under shifting cultivation. Journal of Vegetation Science 19: 809-820.  Kaylor MJ, Warren DR (2017) Linking riparian shade and the legacies of forest management to fish and vertebrate biomass in forested streams. Ecosphere 8:e01845.  Kaylor MJ, Warren DR, Kiffney PM (2017) Long-term effects of riparian forest harvest on light in Pacific Northwest (USA) streams. Freshwater Science 36:1-13.  Kiffney PM, Richardson JS, Bull JP (2003) Responses of periphyton and insects to experimental manipulation of riparian buffer width along forest streams. Journal of Applied Ecology 40:1060-1076.  Kiffney PM, Richardson JS (2010) Organic matter inputs into headwater streams of southwestern British Columbia as a function of riparian reserves and time since harvesting. Forest Ecology and Management 260:1931-1942.  Kominoski JS, Marczak LB, Richardson JS (2011) Riparian forest composition affects stream litter decomposition despite similar microbial and invertebrate communities. Ecology 92:151-159.  150  Kominoski JS, Shah JJF, Canhoto C, Fischer DG, Giling DP, González E, Griffiths NA, Larrañaga A, LeRoy CJ, Mineau MM, McElarney YR, Shirley SM, Swan CM, Tiegs SD (2013) Forecasting functional implications of global changes in riparian plant communities. Frontiers in Ecology and the Environment 11:423-432.  Kominoski JS, Rosemond AD, Benstead JP, Gulis V, Maerz JC, Manning DWP (2015) Low-to-moderate nitrogen and phosphorus concentrations accelerate microbially driven litter breakdown rates. Ecological Applications 25:856-865.  Kreutzweiser DP, Good KP, Capell SS, Holmes SB (2008) Leaf-litter decomposition and macroinvertebrate communities in boreal forest streams linked to upland logging disturbance. Journal of the North American Benthological Society 27:1-15.  Kreutzweiser DP, Capell SS, Holmes SB (2009a) Stream temperature responses to partial-harvest logging in riparian buffers of boreal mixedwood forest watersheds. Canadian Journal of Forest Research 39:497-506.  Kreutzweiser D, Capell S, Good K, Holmes S (2009b) Sediment deposition in streams adjacent to clearcuts and partially harvested riparian buffers in boreal forest catchments. Forest Ecology and Management 258:1578-1585.  Kreutzweiser DP, Muto E, Holmes S, Gunn J (2010) Effects of upland clearcutting and riparian parital harvesting on leaf pack breakdown and aquatic invertebrates in boreal forest streams. Freshwater Biology 55:2238-2252.  Kuglerová L, Agren A, Janson R, Laudon H (2014) Towards optimizing riparian buffer zones: ecological and biogeochemical implications for forest management. Forest Ecology and Management 334:74-84.  Lamberti GA, Gregory SV (1996) Transport and retention of CPOM. In: Hauer R, Lamberti GA (eds). Methods in Stream Ecology. Academic Press, San Diego.  Laurance WF, Ferreira LV, Merona JMR, Laurance SG, Hutchings RW, Lovejoy TE (1998) Effects of forest fragmentation on recruitment patterns in Amazonian tree communities. Conservation Biology 12:460-464.  Laurance WF, Delamônica P, Laurance SG, Vasconcelos HL, Lovejoy TE (2000) Rainforest fragmentation kills big trees. Nature 404:836.  151  Laurance WF, Nascimento HEM, Laurance SG, Andrade AC, Fearnside PM, Ribeiro JEL, Capretz RL (2006a) Rain forest fragmentation and the proliferation of successional trees. Ecology 87: 469-482.  Laurance WF (2006b) Fragments and fire: alarming synergisms among forest disturbance, local climate change, and burning in the Amazon. In: Laurance WF, Peres CA (eds). Emerging Threats to Tropical Forests. The University of Chicago Press, Chicago.  Laurance WF, Sayer J, Cassman KG (2014) Agricultural expansion and its impacts on tropical nature. Trends in Ecology and Evolution 29:107-116.  Lawrence D (2004) Erosion of tree diversity during 200 years of shifting cultivation in Bornean rain forest. Ecological Applications 14:1855-1869.  Leal CG, Pompeu PS, Gardner TA, Leitão RP, Hughes RM, Kaufmann PR, Zuanon J, Paula FR, Ferraz SFB, Thomson JR, Nally RM, Ferreira J, Barlow J (2016) Multi-scale assessment of human-induced changes to Amazonian instream habitats. Landscape Ecology 31:1725-1745.  Leal CG, Barlow J, Gardner TA, Hughes RM, Leitão RP, MacNally R, Kaufmann PR, Ferraz SFB, Zuanon J, Paula FR, Ferreira J, Thomson JR, Lennox GD, Dary EP, Ropke CP, Pompeu PS (2018) Is environmental legislation conserving tropical stream faunas? A large-scale assessment of local, riparian and catchment-scale influences on Amazonian fish. Journal of Applied Ecology 55:1312-1326.  Lecerf A, Richardson JS (2010) Litter decomposition can detect effects of high and moderate levels of forest disturbance on stream condition. Forest Ecology and Management 259:2433-2443.  Lee P, Smyth C, Boutin S (2004) Quantitative review of riparian buffer width guidelines from Canada and the United States. Journal of Environmental Management 70:165-180.  Leitão RP, Zuanon J, Mouillot D, Leal CG, Hughes RM, Kaufmann PR, Villéger S, Pompeu PS, Kasper D, Paula FR, Ferraz SFB, Gardner TA (2017) Disentangling the pathways of land use impacts on the functional structure of fish assemblages in Amazon streams. Ecography 40:1-13.  Lewis WM, Hamilton SK, Saunders JF (1995) Rivers of northern South America. In: Cushing CE, Cummins KW, Minshall GW (eds) Ecosystems of the world: river and stream ecosystems. Elsevier Academic Press, London.  152  Liébault F, Gomez B, Page M, Marden M, Peacock D, Richard D, Trotter CM (2005) Land-use change, sediment production and channel response in upland regions. River Research and Applications 21:739-756.  Ligeiro R, Moretti MS, Gonçalves Jr. JF, Callisto M (2010) What is more important for invertebrate colonization in a stream with low-quality litter inputs: exposure time or leaf species? Hydrobiologia 654:125-136.  Likens GE, Bormann FH (1974) Linkages between terrestrial and aquatic ecosystems. BioScience 24:447-456.  Lovatt C, Kominoski JS, Sakamaki T, Macleod B, Richardson JS (2014) Leaf-litter leachate and light interactively enhance accrual of stream biofilms. Fundamental and Applied Liminology 184:297-306.  Lowe-McConnell RH (1975) Fish communities in tropical freshwaters: their distribution, ecology and evolution. Longman Inc., New York.  Lujan NK, German DP, Winemiller KO (2011) Do wood-grazing fishes partition their niche?: morphological and isotopic evidence for trophic segregation in Neotropical Loricariidae. Functional Ecology 25: 1327-1338.  Mallik AU, Kreutzweiser DP, Spalvieri CM (2014) Forest regeneration in gaps seven years after partial harvesting in riparian buffers of boreal mixedwood streams. Forest Ecology and Management 312:117-128.  Martin PA, Newton AC, Pfeifer M, Khoo M, Bullock JM (2015) Impacts of tropical selective logging on carbon storage and tree species richness: A meta-analysis. Forest Ecology and Management 356:224-233.  Martinelli LA, Filoso S (2008) Expansion of sugarcane ethanol production in Brazil: environmental and social challenges. Ecological Applications 18:885-898.  Martins RT, Rezende RS, Gonçalves Jr. JF, Lopes A, Piedade MTF, Cavalcante HL, Hamada N (2018) Effect of increasing temperature and CO2 on quality of litter, shredders, and microorganisms in Amazonian aquatic systems. Plos One 12:e0188791.  Marzolf ER, Mulholland PJ, Steinman AD (1994) Improvements to the diurnal upstream-downstream dissolved oxygen change technique for determining whole-stream metabolism in small streams. Canadian Journal of Fisheries and Aquatic Sciences 51:1591-1599.  153  Maser C, Sedell JR (1994) From the forest to the sea, the ecology of wood in streams, rivers, estuaries, and oceans. St Lucie Press, Florida.  Mckie BG, Malmqvist B (2009) Assessing ecosystem functioning in streams affected by forest management: increased leaf decomposition occurs without changes to the composition of benthic assemblages. Freshwater Biology 54:2086-2100.  McNamara S, Erskine PD, Lamb D, Chantalangsy L, Boyle S (2012) Primary tree species diversity in secondary fallow forests of Laos. Forest Ecology and Management 281:93-99.  McTammany ME, Benfield EF, Webster JR (2007) Recovery of stream ecosystem metabolism from historical agriculture. Journal of the North American Benthological Society 26: 532-545.  Montgomery DR, Buffington JM, Smith RD, Schmidt KM, Pess G (1995) Pool spacing in forest channels. Water Resources Research 31:1097-1105.  Moore RD, Spittlehouse DL, Story A (2005) Riparian microclimate and stream temperature response to forest harvesting: a review. Journal of the American Water Resources Association 41:813-834.  Moraes JM, Schüller AE, Dunne T, Figueiredo RO, Victoria RL (2006) Water storage and runoff processes in Plinthic soils under forest and pasture in Eastern Amazonia. Hydrological Processes 20:2509-2526.  Mosich TD, Bunn SE, Davies PM (2001) The relative importance of shading and nutrients on algal production in subtropical streams. Freshwater Biology 46:1269-1278.  Moyle PB, Cech JJ (2004) Fishes: an introduction to ichthyology. Prentice Hall, Upper Saddle River.  Mulholland PJ, Fellows CS, Tank JL, Grimm NB, Webster JR, Hamilton SK, Marti E, Ashkenas L, Bowden WB, Doods WK, McDowell WH, Paul MJ, Peterson BJ (2001) Inter-biome comparison of factors controlling stream metabolism. Freshwater Biology 46:1503-1517.  Murcia C (1995) Edge effects in fragmented forests: implications for conservation. Trends in Ecology and Evolution 10:58-62.  154  Nagy RC, Porder S, Neill C, Brando P, Quintino RM, Nascimento SA (2015) Structure and composition of altered riparian forests in an agricultural Amazonian landscape. Ecological Applications 25:1725-1738.  Naiman RJ, Décamps H (1997) The ecology of interfaces: riparian zones. Annual Review of Ecology and Systematics 28:621-658.  Naiman RJ, Décamps H, McClain ME (2005) Riparia: ecology, conservation, and management of streamside communities. Elsevier Academic Press, London.  Nepstad, DC, Uhl C, Serrão EAS (1991) Recuperation of a degraded Amazonian landscape: Forest recovery and agricultural restoration. Ambio 20:248–255.  Niyogi DK, Simon KS, Townsend CR (2003) Breakdown of tussock grass in streams along a gradient of agricultural development in New Zealand. Freshwater Biology 48:1698-1708.  Nunes SS, Barlow J, Gardner TA, Siqueira JV, Sales MR, Souza Jr. CM (2015) A 22 year assessment of deforestation and restoration in riparian forests in the eastern Brazilian Amazon. Environmental Conservation 42:193-203.  O’Connor BL, Harvey JW, McPhillips LE (2012) Thresholds of flow-induced be disturbances and their effects on stream metabolism in an agricultural river. Water Resources Research 48:W08504, doi:10.1029/2011WR011488.  Odum HT (1956) Primary production in flowing waters. Limnology and Oceanography 1:102-117.  Okuda T, Suzuki M, Adachi N, Quah ES, Hussein NA, Manokaran N (2003) Effective of selective logging on canopy and stand structure and tree species composition in a lowland dipterocarp forest in peninsular Malaysia. Forest Ecology and Management 175:297-320.  Owens M, Edwards RW, Gibbs JW (1964) Some reareation studies in streams. International Journal of Air and Water Pollution 8:469-486.  Palik B, Martin M, Zenner E, Blinn C, Kolka R (2012) Overstory and regeneration dynamics in riparian management zones of northern Minnesota forested watersheds. Forest Ecology and Management 271:1-9.  Paul MJ, Meyer JL, Couch CA (2006) Leaf breakdown in streams differing in catchment land use. Freshwater Biology 51:1684-1695.  155  Paula FR, Ferraz SFB, Gerhard P, Vettorazzi CA, Ferreira A (2011) Large woody debris input and its influence on channel structure in agricultural lands of Southeast Brazil. Environmental Management 48:750-763.  Paula FR, Gerhard P, Wenger SJ, Ferreira A, Vettorazzi CA, Ferraz SFB (2013) Influence of forest cover on in-stream large wood in an agricultural landscape of southeastern Brazil: a multi-scale analysis. Landscape Ecology 28:13-27.  Paula FR, Gerhard P, Ferraz SFB, Wenger SJ (2018) Multi-scale assessment of forest cover in an agricultural landscape of Southeastern Brazil: implications for management and conservation of stream habitat and water quality. Ecological Indicators 85:1181-1191.  Piggott JJ, Niyogi DK, Townsend CR, Matthaei CD (2015) Multiple stressors and stream ecosystem functioning: climate warming and agricultural stressors interact to affect processing of organic matter. Journal of Applied Ecology 52:1126-1134.  Piotto D, Montagnini F, Thomas W, Ashton M, Oliver C (2009) Forest recover after swidden cultivation across a 40-year chronosequence in the Atlantic forest of southern Bahia, Brazil. Plant Ecology 205:261-272.  Power ME (2003) Life cycles, limiting factors, and behavioral ecology of four Loricariid catfishes in a Panamanian stream. In: Arratia G, Kapoor BG, Chardon M, Diogo R (eds) Catfishes. Science Publishers, Enfield.  Putz FE (1990) Liana stem diameter growth and mortality rates on Barro Colorado island, Panama. Biotropica 22:103-105.  Quinn GP, Keough MJ (2002) Experimental Design and Data Analysis for Biologists. Cambridge University Press, Cambridge.  R Development Core Team (2016) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna.  Ramey TL, Richardson JS (2017) Terrestrial invertebrates in the riparian zone: mecanisms underlying their unique diversity. BioScience 67:808-819.  Ramírez A, Pringle CM, Wantzen KM (2008) Tropical stream conservation. In: Dudgeon D (ed). Tropical Stream Ecology. Academic Press, San Diego.  156  Reich PB, Peterson DW, Wedin DA, Wrage K (2001) Fire and vegetation effects on productivity and nitrogen cycling across a forest-grassland continuum. Ecology 82:1703-1719.  Ribeiro MC, Metzger JP, Martensen AC, Ponzoni FJ, Hirota MM (2009) The Brazilian Atlantic Forest: how much is left, and how is the remaining forest distributed? Implications for conservation. Biological Conservation 142:1141-1153.  Richardson JS, Danehy RJ (2007) A synthesis of the ecology of headwater streams and their riparian zones in temperate forests. Forest Science 53:131-147.  Richardson JS, Naiman RJ, Bisson PA (2012) How did fixed-width buffers become standard practice for protecting freshwaters and their riparian areas from forest harvest practices? Freshwater Science 31:232-238.  Richardson JS, Sato T (2015) Resource flows across freshwater-terrestrial boundaries and influence on processes linking adjacent ecosystems. Ecohydrology 8:406-415.  Richardson VA, Peres CA (2016) Temporal decay in timber species composition and value in an Amazonian logging concessions. Plos One 11:e0159035. doi:10.1371/journal.pone.0159035.  Ring E, Johansson J, Sandstrom C, Bjarnadótir B, Finér L, Libiete Z, Lode E, Stupak I, Saetersdal M (2017) Mapping policies for surface water protection zones on forest land in the Nordic-Baltic region: large differences in prescriptiveness and zone width. Ambio DOI 10.1007/s13280-017-0924-8.  Rodrigues RR (1999) A vegetação de Piracicaba e municípios de entorno. Circular Técnica do IPEF 189:1-18.  Rosenfeld JS, Huato L (2003) Relationship between large woody debris characteristics and pool formation in small coastal British Columbia streams. North American Journal of Fish Management 23:928-938.  Rueda-Delgado G, Wantzen KM, Tolosa MB (2006) Leaf-litter decomposition in an Amazonian floodplain stream: effects of seasonal hydrological changes. Journal of the North American Benthological Society 25:233-249.  Ruiz-Villanueva V, Piégay H, Gurnell AM, Marston RA, Stoffel M (2016) Recent advances quantifying the large wood dynamics in river basins: new methods and remaining challenges. Review of Geophysics 54: 611-652. 157   Sanpera-Calbet I, Chauvet E, Richardson JS (2012) Fine sediment on leaves: shredder removal of sediment does not enhance fungal colonization. Aquatic Sciences 74:527-538.  Schiller DV, Martí E, Riera JL, Ribot M, Marks JC, Sabater F (2008) Influence of land use on stream ecosystem function in a Mediterranean catchment. Freshwater Biology 53:2600-2612.  Schnitzer SA, Dalling JW, Carson WP (2000) The impact of lianas on tree regeneration in tropical forest canopy gaps: evidence for an alternative pathway of gap-phase regeneration. Journal of Ecology 88:655-666.  Schreuder HT, Gregoire TG, Wood GB (1993) Sampling methods for multiresource forest inventory. Wiley, New York.  Schwartz NB, Uriarte M, DeFries R, Bedka KM, Fernandes K, Gutiérrez-Vélez V, Pinedo-Vasquez MA (2017) Fragmentation increases wind disturbance impacts on forest structure and carbon stocks in a western Amazonian landscape. Ecological Applications 27:1901-1915.  Sedell JR, Reeves GH, Hauer FR, Stanford JA, Hawkins CP (1990) Role of refugia in recovery from disturbances: modern fragmented and disconnected river systems. Environmental Management 14:711-724.  Shoo LP, Catterall CP (2013) Stimulating natural regeneration of tropical forest on degraded land: approaches, outcomes, and information gaps. Restoration Ecology 21:670-677.  Sibley PK, Kreutzweiser DP, Naylor BJ, Richardson JS, Gordon AM (2012) Emulation of natural disturbance (END) for riparian forest management: synthesis and recommendations. Freshwater Science 31:258-264.  Sizer NG, Tanner EVJ, Ferraz IDK (2000) Edge effects on litterfall mass and nutrient concentrations in forest fragments in central Amazonia. Journal of Tropical Ecology 16:853-863.  Soares-Filho B, Rajão R, Macedo M, Carneiro A, Costa W, Coe M, Rodrigues H, Alencar A, (2014) Cracking Brazil’s Forest Code. Science 344:363-364.  Sovu MT, Savadogo P, Odén PC, Xayvongsa L (2009) Recovery of secondary forests on swidden cultivation fallows in Laos. Forest Ecology and Management 258:2666-2675. 158   Souza DR, Souza AL, Leite HG, Yared JAG (2006) Análise estrutural em floresta ombrófila densa de terra firme não explorada, Amazônia Oriental. Revista Árvore 30:75-87.  Stone MK, Wallace JB (1998) Long-term recovery of a mountain stream from clear-cut logging: the effects of forest succession on benthic invertebrate community structure. Freshwater Biology 39:151-169.  Stovall JP, Keeton WS, Kraft CE (2009) Late-successional riparian forest structure results in heterogeneous periphyton distribution in low-order streams. Canadian Journal of Fisheries and Aquatic Sciences 39:2343-2354.  Sutherland AB, Meyer JL, Gardner EP (2002) Effects of land on sediment regime and fish assemblage structure in four southern Appalachian streams. Freshwater Biology 47:1791-1805.  Tanaka MO, Fernandes JF, Suga CM, Hanai FY, Souza ALT (2016) Abrupt change of a stream ecosystem function along a sugarcane-forest transition: integrating riparian and in-stream characteristics. Agriculture, Ecosystems and Environment 207:171-177.  Taniwaki RH, Cassiano CC, Filoso S, Ferraz SFB, Camargo PB, Martinelli LA (2017) Impacts of converting low-intensity pastureland to high-intensity bioenergy cropland on the water quality of tropical streams in Brazil. Science of the Total Environment 584-585:339-347.  Taniwaki RH, Forte YA, Silva GO, Brancalion PHS, Cogueto CV, Filoso S, Ferraz SFB (2018) The native vegetation protection law of Brazil and the challenge for first-order stream conservation. Perspectives in Ecology and Conservation 16:49-53.  Tank JL, Rosi-Marshall EJ, Griffiths NA, Entrekin SA, Stephen ML (2010) A review of allochthonous organic matter dynamics and metabolism in streams. Journal of the North American Benthological Society 29:118-146.  The Nature Conservancy (2018) ICMS Ecológico. Available in: http://www.icmsecologico.org.br/site/index.php?option=com_content&view=article&id=48&Itemid=53 Accessed at 2/1/2018  Thogmartin WE, Gallagher M, Young N, Rohweder JJ, Knutson MG (2009) Factors associated with succession of abandoned agricultural lands along the Lower Missouri River, U.S.A. Restoration Ecology 17:290-296. 159   Turner MG, Gardner RH, O’Neill RV (2001) Landscape ecology in theory and practice: pattern and process. Springer, New York.  Uehlinger U, Naegeli MW (1998) Ecosystem metabolism, disturbance, and stability in a prealpine gravel bed river. Journal of the North American Benthological Society 17:165-178.  Uhl C (1987) Factors controlling succession following slash-and-burn agriculture in Amazonia. Journal of Ecology 75:377-407.  Uhl C, Buschbacher R, Serrão EAS (1988) Abandoned pastures in eastern Amazonia. I. Patterns of plant succession. Journal of Ecology 76:663-681.  Uhl C, Veríssimo A, Mattos MM, Brandino Z, Vieira ICG (1991) Social, economic, and ecological consequences of selective logging in an Amazon Frontier: the case of Tailândia. Forest Ecology and Management 46:243-273.  UN-REDD Programme (2016) Reducing Emissions from Deforestation and Forest Degradation in Developing Countries. Available in: http://www.un-redd.org/ Accessed at 2/1/2018.  Valente-Neto F, Koroiva R, Fonseca-Gessner AA, Roque FO (2015) The effect of riparian deforestation on macroinvertebrates associated with submerged woody debris. Aquatic Sciences 49:115-125.  Vasconcelos HL, Luizão FJ (2004) Litter production and litter nutrient concentrations in a fragmented Amazonian landscape. Ecological Applications 14:884-892.  Verissimo A, Barreto P, Mattos M, Tarifa R, Uhl C (1992) Logging impacts and prospects for sustainable forest management in an old Amazonian frontier: the case of Paragominas. Forest Ecology and Management 55:169-199.  Victor MAM, Cavalli AC, Guillaumon JR, Filho RS (2005) Cem anos de devastação: Revisitada 30 anos depois. Ministério do Meio Ambiente, Brasília.  Villela DM, Nascimento MT, Aragão LEOC, Gama DM (2006) Effect of selective logging on forest structure and nutrient cycling in a seasonally dry Brazilian Atlantic forest. Journal of Biogeography 33:506-516.  160  Wantzen KM, Yule CM, Mathooko JM, Pringle CM (2008) Organic matter processing in tropical streams. In: Dudgeon D (ed). Tropical Stream Ecology. Academic Press, San Diego.  Warren DR, Bernhardt ES, Hall Jr RO, Likens GE (2007) Forest age, wood, and nutriente dynamics in headwater streams of the Hubbard Brook Experimental Forest, NH. Earth Surface Processes and Landforms 32:1154-1163.  Warren DR, Kraft CE (2008) Dynamics of large wood in an eastern U.S. mountain stream. Forest Ecology and Management 256:808-814.  Warren DR, Kraft CE, Keeton WS, Nunery JS, Likens GE (2009) Dynamics of wood recruitment in streams of the northeastern US. Forest Ecology and Management 258:804-813.  Warren DR, Keeton WS, Bechtold HA, Rosi-Marshall EJ (2013) Comparing streambed light availability and canopy cover in streams with old-growth versus early-mature riparian forests in western Oregon. Aquatic Sciences 75:547-558.  Warren DR, Keeton WS, Kiffney PM, Kaylor MJ, Bechtold HA, Magee J (2016) Changing forests – changing streams: riparian forest stand development and ecosystem function in temperate headwaters. Ecosphere 7:e01435.10.1002/ecs2.1435  Webster JR, Benfield EF, Ehrman TP, Schaeffer MA, Tank JL, Hutchens JJ, D’Angelo DJ (1999) What happens to allochthonous material that falls into streams? A synthesis of new and published information from Coweeta. Freshwater Biology 41:687-705.  Wenger SJ (1999) A review of the scientific literature on riparian buffer width, extent and vegetation. University of Georgia, Athens.  Whitaker RH (1953) A consideration of climax theory: the climax as a population and pattern. Ecological Monographs 23:41-78.  White PS (1979) Pattern, process, and natural disturbance in vegetation. The Botanical Review 45:229-299.  White PS, Pickett STA (1985) Natural disturbance and patch dynamics: An introduction. In: Pickett STA, White PS (eds). The Ecology of Natural Disturbance and Patch Dynamics. Academic Press, New York.   161  Wing MG, Skaugset A (2002) Relationships of channel characteristics, land ownership, and land use patterns to large woody debris in western Oregon streams. Canadian Journal of Fisheries and Aquatic Sciences 59:796–807.  Wohl E, Lininger KB, Fox M, Baillie BR, Erskine W (2017) Instream large wood loads across bioclimatic regions. Forest Ecology and Management 404:370-380.  Wright JP, Flecker AS (2004) Deforesting the riverscape: the effects of wood on fish diversity in a Venezuelan piedmont stream. Biological Conservation 120:439-447.  Yeung ACY, Lecerf A, Richardson JS (2017) Assessing the long-term ecological effects of riparian management practices on headwater streams in a coastal temperate rainforest. Forest Ecology and Management 384:100-109.  Young MK, Mace EA, Ziegler ET, Sutherland EK (2006) Characterizing and contrasting instream and riparian coarse wood in western Montane basins. Forest Ecology and Management 226:26-40.  YSI (2012) 6-Series multiparameter water quality sondes: user manual. Yellow Spring Instruments, Yellow Springs.  Zarin DJ, Davidson EA, Brondizio E, Vieira ICG, Sá T, Feldpausch T, Schuur EAG, Mesquita R, Moran E, Delamonica P, Ducey MJ, Hurtt GC, Salimon C, Denich M (2005) Legacy of fire slows carbon accumulation in Amazonian forest regrowth. Frontiers in Ecology and the Environment 3:365-369.  Zenner EK, Olszewski SL, Palik BJ, Kastendick DN, Peck JE, Blinn CR (2012) Riparian vegetation response to gradients in residual basal area with harvesting treatment and distance to stream. Forest Ecology and Management 283:66-76.   162  Appendices  Appendix A.1 Candidate models tested and results of model selection for LW variables. Bold lines indicate models with deltaAIC≤2.   Candidate models for LW density and volume  Cand.models<-list() Cand.models[[1]] <- glm(LW~RFM+RFR+CWI+BA+SD+DIS,data=dados,family=gaussian) Cand.models[[2]] <- glm(LW~RFM+CWI,data=dados,family=gaussian) Cand.models[[3]] <- glm(LW~RFM+BA,data=dados,family=gaussian) Cand.models[[4]] <- glm(LW~RFM+SD,data=dados,family=gaussian) Cand.models[[5]] <- glm(LW~RFM+DIS,data=dados,family=gaussian) Cand.models[[6]] <- glm(LW~RFR+CWI,data=dados,family=gaussian) Cand.models[[7]] <- glm(LW~RFR+DIS,data=dados,family=gaussian) Cand.models[[8]] <- glm(LW~RFM,data=dados,family=gaussian) Cand.models[[9]] <- glm(LW~RFR,data=dados,family=gaussian) Cand.models[[10]] <- glm(LW~CWI,data=dados,family=gaussian) Cand.models[[11]] <- glm(LW~BA,data=dados,family=gaussian) Cand.models[[12]] <- glm(LW~SD,data=dados,family=gaussian) Cand.models[[13]] <- glm(LW~DIS,data=dados,family=gaussian)  Modnames<-paste("mod", 1:length(Cand.models), sep="") aictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE)  Results for LW density per channel length  Model selection based on AICc:         K   AICc Delta_AICc AICcWt Cum.Wt    LL mod10  3 -27.12       0.00   0.87   0.87 17.16 mod6   5 -23.27       3.85   0.13   1.00 18.30 mod2   8 -14.83      12.29   0.00   1.00 20.22 mod8   7  -5.93      21.19   0.00   1.00 13.47 mod3   8  -3.63      23.49   0.00   1.00 14.61 mod5   8  -2.12      25.00   0.00   1.00 13.86 mod4   8  -1.54      25.58   0.00   1.00 13.57 mod1  12   4.44      31.56   0.00   1.00 23.96 mod13  3   5.39      32.50   0.00   1.00  0.91 mod11  3   6.19      33.31   0.00   1.00  0.51 mod12  3   6.20      33.32   0.00   1.00  0.50 mod9   4   8.84      35.96   0.00   1.00  0.63 mod7   5  10.82      37.94   0.00   1.00  1.26      163  Appendix A.1  Results for LW density per channel area  Model selection based on AICc:         K   AICc Delta_AICc AICcWt Cum.Wt    LL mod10  3 -76.86       0.00   0.30   0.30 42.03 mod13  3 -76.64       0.22   0.27   0.57 41.92 mod11  3 -75.78       1.08   0.18   0.75 41.49 mod12  3 -75.28       1.58   0.14   0.88 41.24 mod9   4 -73.62       3.24   0.06   0.94 41.86 mod6   5 -72.41       4.45   0.03   0.98 42.87 mod7   5 -71.36       5.50   0.02   1.00 42.35 mod8   7 -66.96       9.90   0.00   1.00 43.98 mod5   8 -64.82      12.04   0.00   1.00 45.21 mod3   8 -64.49      12.36   0.00   1.00 45.05 mod4   8 -62.68      14.18   0.00   1.00 44.14 mod2   8 -62.39      14.47   0.00   1.00 44.00 mod1  12 -42.40      34.46   0.00   1.00 47.38  Results for LW volume per channel length  Model selection based on AICc:         K  AICc Delta_AICc AICcWt Cum.Wt    LL mod10  3 -3.28       0.00   0.79   0.79  5.24 mod6   5 -0.55       2.72   0.20   0.99  6.94 mod2   8  6.96      10.24   0.00   0.99  9.32 mod8   7  8.46      11.73   0.00   1.00  6.27 mod12  3 10.25      13.53   0.00   1.00 -1.53 mod13  3 10.89      14.16   0.00   1.00 -1.84 mod11  3 12.35      15.62   0.00   1.00 -2.57 mod5   8 12.64      15.91   0.00   1.00  6.48 mod3   8 12.78      16.06   0.00   1.00  6.41 mod4   8 12.88      16.16   0.00   1.00  6.36 mod9   4 13.63      16.91   0.00   1.00 -1.76 mod7   5 15.98      19.25   0.00   1.00 -1.32 mod1  12 26.13      29.41   0.00   1.00 13.12             164  Appendix A.1  Results for LW volume per channel area  Model selection based on AICc:         K    AICc Delta_AICc AICcWt Cum.Wt     LL mod12  3 -398.33       0.00   0.47   0.47 202.76 mod10  3 -395.69       2.64   0.13   0.59 201.45 mod11  3 -395.60       2.73   0.12   0.71 201.40 mod9   4 -395.56       2.77   0.12   0.83 202.83 mod13  3 -394.73       3.60   0.08   0.91 200.96 mod6   5 -393.64       4.69   0.04   0.95 203.49 mod7   5 -392.94       5.39   0.03   0.99 203.14 mod8   7 -390.46       7.87   0.01   0.99 205.73 mod5   8 -387.43      10.89   0.00   1.00 206.52 mod3   8 -386.84      11.49   0.00   1.00 206.22 mod4   8 -386.43      11.90   0.00   1.00 206.01 mod2   8 -386.00      12.33   0.00   1.00 205.80 mod1  12 -366.17      32.16   0.00   1.00 209.27  Candidate models for LW pool density  Cand.models<-list() Cand.models[[1]] <- glm(LWPOOL~RFM+RFR+CWI+BA+SD+DIS+LWDE+LWVO,data=dados,family=gaussian) Cand.models[[2]] <- glm(LWPOOL~RFM+CWI,data=dados,family=gaussian) Cand.models[[3]] <- glm(LWPOOL~RFM+BA,data=dados,family=gaussian) Cand.models[[4]] <- glm(LWPOOL~RFM+SD,data=dados,family=gaussian) Cand.models[[5]] <- glm(LWPOOL~RFM+DIS,data=dados,family=gaussian) Cand.models[[6]] <- glm(LWPOOL~RFM+LWDE,data=dados,family=gaussian) Cand.models[[7]] <- glm(LWPOOL~RFM+LWVO,data=dados,family=gaussian) Cand.models[[8]] <- glm(LWPOOL~RFR+CWI,data=dados,family=gaussian) Cand.models[[9]] <- glm(LWPOOL~RFR+DIS,data=dados,family=gaussian) Cand.models[[10]] <- glm(LWPOOL~RFR+LWDE,data=dados,family=gaussian) Cand.models[[11]] <- glm(LWPOOL~RFR+LWVO,data=dados,family=gaussian) Cand.models[[12]] <- glm(LWPOOL~RFM,data=dados,family=gaussian) Cand.models[[13]] <- glm(LWPOOL~RFR,data=dados,family=gaussian) Cand.models[[14]] <- glm(LWPOOL~CWI,data=dados,family=gaussian) Cand.models[[15]] <- glm(LWPOOL~BA,data=dados,family=gaussian) Cand.models[[16]] <- glm(LWPOOL~SD,data=dados,family=gaussian) Cand.models[[17]] <- glm(LWPOOL~DIS,data=dados,family=gaussian) Cand.models[[18]] <- glm(LWPOOL~LWDE,data=dados,family=gaussian) Cand.models[[19]] <- glm(LWPOOL~LWVO,data=dados,family=gaussian)  Modnames<-paste("mod", 1:length(Cand.models), sep="") aictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE)    165  Appendix A.1  Results of LW pool density per channel length  Model selection based on AICc:         K    AICc Delta_AICc AICcWt Cum.Wt    LL mod17  3 -108.08       0.00   0.27   0.27 57.64 mod14  3 -107.88       0.20   0.24   0.52 57.54 mod19  3 -106.36       1.72   0.11   0.63 56.78 mod15  3 -106.12       1.96   0.10   0.73 56.66 mod16  3 -106.06       2.02   0.10   0.83 56.63 mod18  3 -106.02       2.06   0.10   0.93 56.61 mod13  4 -103.47       4.61   0.03   0.95 56.79 mod8   5 -102.18       5.90   0.01   0.97 57.76 mod9   5 -102.06       6.02   0.01   0.98 57.70 mod11  5 -100.45       7.64   0.01   0.99 56.89 mod10  5 -100.24       7.84   0.01   0.99 56.79 mod12  7  -98.64       9.44   0.00   1.00 59.82 mod6   8  -97.93      10.15   0.00   1.00 61.77 mod5   8  -97.81      10.27   0.00   1.00 61.71 mod3   8  -95.75      12.33   0.00   1.00 60.67 mod7   8  -95.62      12.46   0.00   1.00 60.61 mod2   8  -94.61      13.47   0.00   1.00 60.10 mod4   8  -94.10      13.98   0.00   1.00 59.85 mod1  14  -62.14      45.94   0.00   1.00 68.40  Results of LW pool density per channel area  Model selection based on AICc:         K    AICc Delta_AICc AICcWt Cum.Wt    LL mod14  3 -148.81       0.00   0.91   0.91 78.00 mod8   5 -143.61       5.19   0.07   0.98 78.47 mod17  3 -139.17       9.64   0.01   0.99 73.18 mod5   8 -136.70      12.11   0.00   0.99 81.15 mod2   8 -136.42      12.39   0.00   0.99 81.01 mod6   8 -135.98      12.82   0.00   1.00 80.79 mod18  3 -135.58      13.23   0.00   1.00 71.39 mod16  3 -134.61      14.20   0.00   1.00 70.90 mod19  3 -134.21      14.60   0.00   1.00 70.70 mod15  3 -134.09      14.71   0.00   1.00 70.65 mod9   5 -133.18      15.62   0.00   1.00 73.26 mod13  4 -131.78      17.03   0.00   1.00 70.94 mod12  7 -131.32      17.49   0.00   1.00 76.16 mod10  5 -129.82      18.98   0.00   1.00 71.58 mod11  5 -128.68      20.13   0.00   1.00 71.00 mod7   8 -128.29      20.51   0.00   1.00 76.95 mod3   8 -127.58      21.23   0.00   1.00 76.59 mod4   8 -127.15      21.66   0.00   1.00 76.37 mod1  14 -105.01      43.79   0.00   1.00 89.84 166  Appendix A.2 Candidate models tested and results of model selection for respiration and primary production. Bold lines indicate models with deltaAIC≤2.   Respiration – 2 station - complete set of predictors  Cand.models<-list() Cand.models[[1]] <-glm(CR24~RFM+RFR+PC1RF+SD+LIT+CON+DIS,data=dados,family=gaussian) Cand.models[[2]] <-glm(CR24~RFM+PC1RF,data=dados,family=gaussian) Cand.models[[3]] <-glm(CR24~RFM+SD,data=dados,family=gaussian) Cand.models[[4]] <-glm(CR24~RFM+LIT,data=dados,family=gaussian) Cand.models[[5]] <-glm(CR24~RFM+CON,data=dados,family=gaussian) Cand.models[[6]] <-glm(CR24~RFM+DIS,data=dados,family=gaussian) Cand.models[[7]] <-glm(CR24~RFR+LIT,data=dados,family=gaussian) Cand.models[[8]] <-glm(CR24~RFR+CON,data=dados,family=gaussian) Cand.models[[9]] <-glm(CR24~RFR+DIS,data=dados,family=gaussian) Cand.models[[10]] <- glm(CR24~RFM,data=dados,family=gaussian) Cand.models[[11]] <- glm(CR24~RFR,data=dados,family=gaussian) Cand.models[[12]] <- glm(CR24~PC1RF,data=dados,family=gaussian) Cand.models[[13]] <- glm(CR24~SD,data=dados,family=gaussian) Cand.models[[14]] <- glm(CR24~LIT,data=dados,family=gaussian) Cand.models[[15]] <- glm(CR24~CON,data=dados,family=gaussian) Cand.models[[16]] <- glm(CR24~DIS,data=dados,family=gaussian)  Modnames<-paste("mod", 1:length(Cand.models), sep="") aictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE)  Model selection based on AICc:        df   AICc Delta_AICc AICcWt Cum.Wt     LL mod15  3  75.35       0.00   0.33   0.33 -34.08 mod16  3  76.25       0.89   0.21   0.54 -34.52 mod8   5  76.96       1.61   0.15   0.69 -31.81 mod12  3  77.11       1.76   0.14   0.83 -34.95 mod14  3  78.76       3.41   0.06   0.89 -35.78 mod13  3  78.78       3.43   0.06   0.95 -35.79 mod9   5  80.12       4.77   0.03   0.98 -33.39 mod11  4  81.30       5.95   0.02   0.99 -35.60 mod7   5  83.33       7.97   0.01   1.00 -35.00 mod5   8  88.12      12.77   0.00   1.00 -31.26 mod6   8  90.34      14.99   0.00   1.00 -32.37 mod10  7  90.37      15.02   0.00   1.00 -34.69 mod2   8  91.42      16.07   0.00   1.00 -32.91 mod4   8  94.09      18.74   0.00   1.00 -34.25 mod3   8  94.42      19.07   0.00   1.00 -34.41 mod1  13 110.48      35.13   0.00   1.00 -24.04      167  Appendix A.2  Primary Production – 2 station - complete set of predictors  Cand.models<-list() Cand.models[[1]] <-glm(GPP~RFM+RFR+PC1RF+SD+DIS,data=dados,family=gaussian) Cand.models[[2]] <-glm(GPP~RFM+PC1RF,data=dados,family=gaussian) Cand.models[[3]] <-glm(GPP~RFM+SD,data=dados,family=gaussian) Cand.models[[4]] <-glm(GPP~RFM+DIS,data=dados,family=gaussian) Cand.models[[5]] <-glm(GPP~RFR+DIS,data=dados,family=gaussian) Cand.models[[6]] <-glm(GPP~RFM,data=dados,family=gaussian) Cand.models[[7]] <-glm(GPP~RFR,data=dados,family=gaussian) Cand.models[[8]] <-glm(GPP~PC1RF,data=dados,family=gaussian) Cand.models[[9]] <-glm(GPP~SD,data=dados,family=gaussian) Cand.models[[10]] <-glm(GPP~DIS,data=dados,family=gaussian)  Modnames<-paste("mod", 1:length(Cand.models), sep="") aictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE)  Model selection based on AICc:        df  AICc Delta_AICc AICcWt Cum.Wt    LL mod5   5  9.13       0.00   0.36   0.36  2.10 mod7   4  9.73       0.60   0.27   0.63  0.19 mod1  11 11.27       2.14   0.12   0.75 16.37 mod8   3 11.44       2.31   0.11   0.86 -2.12 mod9   3 12.62       3.49   0.06   0.92 -2.71 mod10  3 13.41       4.28   0.04   0.97 -3.10 mod6   7 15.30       6.17   0.02   0.98  2.85 mod4   8 15.85       6.72   0.01   1.00  4.88 mod3   8 19.35      10.23   0.00   1.00  3.12 mod2   8 19.44      10.31   0.00   1.00  3.08  Excluding uninformative parameters  Cand.models<-list() Cand.models[[1]] <-glm(GPP~RFM+RFR+PC1RF+SD+DIS,data=dados,family=gaussian) Cand.models[[2]] <-glm(GPP~RFM+PC1RF,data=dados,family=gaussian) Cand.models[[3]] <-glm(GPP~RFM+SD,data=dados,family=gaussian) Cand.models[[4]] <-glm(GPP~RFM+DIS,data=dados,family=gaussian) Cand.models[[5]] <-glm(GPP~RFM,data=dados,family=gaussian) Cand.models[[6]] <-glm(GPP~RFR,data=dados,family=gaussian) Cand.models[[7]] <-glm(GPP~PC1RF,data=dados,family=gaussian) Cand.models[[8]] <-glm(GPP~SD,data=dados,family=gaussian) Cand.models[[9]] <-glm(GPP~DIS,data=dados,family=gaussian)  Modnames<-paste("mod", 1:length(Cand.models), sep="") aictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE)    168  Appendix A.2  Model selection based on AICc:       df  AICc Delta_AICc AICcWt Cum.Wt    LL mod6  4  9.73       0.00   0.42   0.42  0.19 mod1 11 11.27       1.54   0.19   0.61 16.37 mod7  3 11.44       1.72   0.18   0.78 -2.12 mod8  3 12.62       2.89   0.10   0.88 -2.71 mod9  3 13.41       3.68   0.07   0.95 -3.10 mod5  7 15.30       5.57   0.03   0.97  2.85 mod4  8 15.85       6.12   0.02   0.99  4.88 mod3  8 19.35       9.63   0.00   1.00  3.12 mod2  8 19.44       9.71   0.00   1.00  3.08  Excluding uninformative parameters – secound round  Cand.models<-list() Cand.models[[1]] <-glm(GPP~RFM+PC1RF,data=dados,family=gaussian) Cand.models[[2]] <-glm(GPP~RFM+SD,data=dados,family=gaussian) Cand.models[[3]] <-glm(GPP~RFM+DIS,data=dados,family=gaussian) Cand.models[[4]] <-glm(GPP~RFM,data=dados,family=gaussian) Cand.models[[5]] <-glm(GPP~RFR,data=dados,family=gaussian) Cand.models[[6]] <-glm(GPP~PC1RF,data=dados,family=gaussian) Cand.models[[7]] <-glm(GPP~SD,data=dados,family=gaussian) Cand.models[[8]] <-glm(GPP~DIS,data=dados,family=gaussian)  Modnames<-paste("mod", 1:length(Cand.models), sep="") aictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE)  Model selection based on AICc:       df  AICc Delta_AICc AICcWt Cum.Wt    LL mod5 4  9.73       0.00   0.51   0.51  0.19 mod6 3 11.44       1.72   0.22   0.73 -2.12 mod7 3 12.62       2.89   0.12   0.85 -2.71 mod8 3 13.41       3.68   0.08   0.94 -3.10 mod4 7 15.30       5.57   0.03   0.97  2.85 mod3 8 15.85       6.12   0.02   0.99  4.88 mod2 8 19.35       9.63   0.00   1.00  3.12 mod1 8 19.44       9.71   0.00   1.00  3.08      169  Appendix A.3 Results of correlation analysis between decomposition rates for the species and forest and stream environmental predictors. Values in bold indicate r ≥ 0.30.    KHe KSi KMa KLi PC1RF SD BA HEIG CANO LIGHT SED NIT TEMP CON DO DIS CWI CR24 KHe 1.00 0.24 0.23 0.19 0.04 -0.37 0.05 0.13 0.05 -0.18 0.25 -0.09 0.19 0.00 0.08 0.19 0.23 -0.05 KSi 0.24 1.00 0.56 0.25 0.25 -0.18 0.24 0.39 0.18 -0.32 -0.04 -0.15 -0.23 0.11 0.05 -0.22 -0.12 0.22 KMa 0.23 0.56 1.00 0.55 0.47 0.10 0.36 0.52 0.43 -0.52 -0.22 0.01 -0.33 -0.17 0.41 0.08 0.01 -0.14 KLi 0.19 0.25 0.55 1.00 0.34 0.08 0.23 0.38 0.28 -0.42 0.10 0.05 -0.18 -0.17 0.18 -0.03 0.30 0.03 PC1RF 0.04 0.25 0.47 0.34 1.00 0.66 0.87 0.85 0.91 -0.93 -0.22 -0.06 -0.50 -0.14 -0.12 -0.21 0.10 0.40 SD -0.37 -0.18 0.10 0.08 0.66 1.00 0.47 0.24 0.70 -0.54 -0.29 0.28 -0.44 -0.03 -0.25 -0.33 -0.07 0.31 BA 0.05 0.24 0.36 0.23 0.87 0.47 1.00 0.81 0.65 -0.75 0.00 -0.19 -0.28 -0.21 -0.03 0.07 0.16 0.31 HEIG 0.13 0.39 0.52 0.38 0.85 0.24 0.81 1.00 0.70 -0.80 -0.16 -0.26 -0.36 -0.17 -0.02 -0.07 0.13 0.35 CANO 0.05 0.18 0.43 0.28 0.91 0.70 0.65 0.70 1.00 -0.84 -0.34 0.14 -0.60 -0.04 -0.23 -0.38 0.14 0.42 LIGHT -0.18 -0.32 -0.52 -0.42 -0.93 -0.54 -0.75 -0.80 -0.84 1.00 0.15 0.15 0.44 0.14 0.00 0.21 -0.03 -0.30 SED 0.25 -0.04 -0.22 0.10 -0.22 -0.29 0.00 -0.16 -0.34 0.15 1.00 -0.26 0.54 0.05 0.09 0.17 0.10 0.09 NIT -0.09 -0.15 0.01 0.05 -0.06 0.28 -0.19 -0.26 0.14 0.15 -0.26 1.00 -0.10 0.15 0.16 0.04 0.31 0.04 TEMP 0.19 -0.23 -0.33 -0.18 -0.50 -0.44 -0.28 -0.36 -0.60 0.44 0.54 -0.10 1.00 0.22 0.28 0.51 0.10 -0.18 CON 0.00 0.11 -0.17 -0.17 -0.14 -0.03 -0.21 -0.17 -0.04 0.14 0.05 0.15 0.22 1.00 -0.27 -0.09 0.19 0.47 DO 0.08 0.05 0.41 0.18 -0.12 -0.25 -0.03 -0.02 -0.23 0.00 0.09 0.16 0.28 -0.27 1.00 0.65 0.25 -0.48 DIS 0.19 -0.22 0.08 -0.03 -0.21 -0.33 0.07 -0.07 -0.38 0.21 0.17 0.04 0.51 -0.09 0.65 1.00 0.43 -0.43 CWI 0.23 -0.12 0.01 0.30 0.10 -0.07 0.16 0.13 0.14 -0.03 0.10 0.31 0.10 0.19 0.25 0.43 1.00 0.11 CR24 -0.05 0.22 -0.14 0.03 0.40 0.31 0.31 0.35 0.42 -0.30 0.09 0.04 -0.18 0.47 -0.48 -0.43 0.11 1.00   170  Appendix A.4 Candidate set of models tested and results of model selection for decomposition rate (all species). Bold lines indicate models with deltaAIC≤2.   Cand.models[[1]]<-lme(K~RFM+RFR+SP+PC1RF+SD+SED+NIT+TEMP+CON+DO+DIS+CWI+CR24,random=~1|ID,method="REML",data=dados) Cand.models[[2]]<-lme(K~RFM,random=~1|ID,method="REML",data=dados) Cand.models[[3]]<-lme(K~RFR,random=~1|ID,method="REML",data=dados) Cand.models[[4]]<-lme(K~SP,random=~1|ID,method="REML",data=dados) Cand.models[[5]]<-lme(K~PC1RF,random=~1|ID,method="REML",data=dados) Cand.models[[6]]<-lme(K~SD,random=~1|ID,method="REML",data=dados) Cand.models[[7]]<-lme(K~SED,random=~1|ID,method="REML",data=dados) Cand.models[[8]]<-lme(K~NIT,random=~1|ID,method="REML",data=dados) Cand.models[[9]]<-lme(K~TEMP,random=~1|ID,method="REML",data=dados) Cand.models[[10]]<-lme(K~CON,random=~1|ID,method="REML",data=dados) Cand.models[[11]]<-lme(K~DIS,random=~1|ID,method="REML",data=dados) Cand.models[[12]]<-lme(K~CWI,random=~1|ID,method="REML",data=dados) Cand.models[[13]]<-lme(K~CR24,random=~1|ID,method="REML",data=dados) Cand.models[[14]]<-lme(K~PC1RF+RFM,random=~1|ID,method="REML",data=dados) Cand.models[[15]]<-lme(K~SD+RFM,random=~1|ID,method="REML",data=dados) Cand.models[[16]]<-lme(K~SED+RFM,random=~1|ID,method="REML",data=dados) Cand.models[[17]]<-lme(K~NIT+RFM,random=~1|ID,method="REML",data=dados) Cand.models[[18]]<-lme(K~TEMP+RFM,random=~1|ID,method="REML",data=dados) Cand.models[[19]]<-lme(K~CON+RFM,random=~1|ID,method="REML",data=dados) Cand.models[[20]]<-lme(K~DIS+RFM,random=~1|ID,method="REML",data=dados) Cand.models[[21]]<-lme(K~CWI+RFM,random=~1|ID,method="REML",data=dados) Cand.models[[22]]<-lme(K~CR24+RFM,random=~1|ID,method="REML",data=dados) Cand.models[[23]]<-lme(K~SED+RFR,random=~1|ID,method="REML",data=dados) Cand.models[[24]]<-lme(K~NIT+RFR,random=~1|ID,method="REML",data=dados) Cand.models[[25]]<-lme(K~TEMP+RFR,random=~1|ID,method="REML",data=dados) Cand.models[[26]]<-lme(K~CON+RFR,random=~1|ID,method="REML",data=dados) Cand.models[[27]]<-lme(K~DIS+RFR,random=~1|ID,method="REML",data=dados) Cand.models[[28]]<-lme(K~CWI+RFR,random=~1|ID,method="REML",data=dados) Cand.models[[29]]<-lme(K~CR24+RFR,random=~1|ID,method="REML",data=dados) Cand.models[[30]]<-lme(K~PC1RF+SP,random=~1|ID,method="REML",data=dados) Cand.models[[31]]<-lme(K~SD+SP,random=~1|ID,method="REML",data=dados) Cand.models[[32]]<-lme(K~SED+SP,random=~1|ID,method="REML",data=dados) Cand.models[[33]]<-lme(K~NIT+SP,random=~1|ID,method="REML",data=dados) Cand.models[[34]]<-lme(K~TEMP+SP,random=~1|ID,method="REML",data=dados) Cand.models[[35]]<-lme(K~CON+SP,random=~1|ID,method="REML",data=dados) Cand.models[[36]]<-lme(K~DIS+SP,random=~1|ID,method="REML",data=dados) Cand.models[[37]]<-lme(K~CWI+SP,random=~1|ID,method="REML",data=dados) Cand.models[[38]]<-lme(K~CR24+SP,random=~1|ID,method="REML",data=dados) Cand.models[[39]]<-lme(K~RFM+SP,random=~1|ID,method="REML",data=dados) Cand.models[[40]]<-lme(K~RFR+SP,random=~1|ID,method="REML",data=dados) Cand.models[[41]]<-lme(K~PC1RF+RFM+SP,random=~1|ID,method="REML",data=dados) Cand.models[[42]]<-lme(K~SD+RFM+SP,random=~1|ID,method="REML",data=dados)    171  Appendix A.4  Cand.models[[43]]<-lme(K~SED+RFM+SP,random=~1|ID,method="REML",data=dados) Cand.models[[44]]<-lme(K~NIT+RFM+SP,random=~1|ID,method="REML",data=dados) Cand.models[[45]]<-lme(K~TEMP+RFM+SP,random=~1|ID,method="REML",data=dados) Cand.models[[46]]<-lme(K~CON+RFM+SP,random=~1|ID,method="REML",data=dados) Cand.models[[47]]<-lme(K~DIS+RFM+SP,random=~1|ID,method="REML",data=dados) Cand.models[[48]]<-lme(K~CWI+RFM+SP,random=~1|ID,method="REML",data=dados) Cand.models[[49]]<-lme(K~CR24+RFM+SP,random=~1|ID,method="REML",data=dados) Cand.models[[50]]<-lme(K~PC1RF+RFR+SP,random=~1|ID,method="REML",data=dados) Cand.models[[51]]<-lme(K~SD+RFR+SP,random=~1|ID,method="REML",data=dados) Cand.models[[52]]<-lme(K~SED+RFR+SP,random=~1|ID,method="REML",data=dados) Cand.models[[53]]<-lme(K~NIT+RFR+SP,random=~1|ID,method="REML",data=dados) Cand.models[[54]]<-lme(K~TEMP+RFR+SP,random=~1|ID,method="REML",data=dados) Cand.models[[55]]<-lme(K~CON+RFR+SP,random=~1|ID,method="REML",data=dados) Cand.models[[56]]<-lme(K~DIS+RFR+SP,random=~1|ID,method="REML",data=dados) Cand.models[[57]]<-lme(K~CWI+RFR+SP,random=~1|ID,method="REML",data=dados) Cand.models[[58]]<-lme(K~CR24+RFR+SP,random=~1|ID,method="REML",data=dados) Cand.models[[59]]<-lme(K~SP+RFR+PC1RF+SD+SED+NIT+TEMP+CON+DIS+CWI+CR24,random=~1|ID,method="REML",data=dados) Cand.models[[60]]<-lme(K~DO,random=~1|ID,method="REML",data=dados) Cand.models[[61]]<-lme(K~DO+RFM,random=~1|ID,method="REML",data=dados) Cand.models[[62]]<-lme(K~DO+RFR,random=~1|ID,method="REML",data=dados) Cand.models[[63]]<-lme(K~DO+SP,random=~1|ID,method="REML",data=dados) Cand.models[[64]]<-lme(K~DO+RFM+SP,random=~1|ID,method="REML",data=dados) Cand.models[[65]]<-lme(K~DO+RFR+SP,random=~1|ID,method="REML",data=dados)  Modnames<-paste("mod", 1:length(Cand.models), sep="") aictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE)                      172  Appendix A.4  Model selection based on AICc:        df    AICc Delta_AICc AICcWt Cum.Wt Res.LL mod4   6 -931.96       0.00      1   0.99 472.45 mod36  7 -920.73      11.22   0.00   1.00 468.00 mod33  7 -918.10      13.86   0.00   1.00 466.69 mod30  7 -917.34      14.62   0.00   1.00 466.31 mod63  7 -916.27      15.69   0.00   1.00 465.77 mod32  7 -914.83      17.13   0.00   1.00 465.05 mod37  7 -914.72      17.24   0.00   1.00 465.00 mod34  7 -914.65      17.31   0.00   1.00 464.96 mod38  7 -914.06      17.90   0.00   1.00 464.67 mod35  7 -912.78      19.17   0.00   1.00 464.03 mod40  8 -911.30      20.66   0.00   1.00 464.48 mod31  7 -901.43      30.53   0.00   1.00 458.35 mod56  9 -899.55      32.41   0.00   1.00 459.82 mod53  9 -897.16      34.80   0.00   1.00 458.63 mod50  9 -896.83      35.13   0.00   1.00 458.46 mod65  9 -894.55      37.41   0.00   1.00 457.32 mod52  9 -893.62      38.34   0.00   1.00 456.85 mod58  9 -893.58      38.38   0.00   1.00 456.84 mod54  9 -893.40      38.56   0.00   1.00 456.74 mod57  9 -893.28      38.68   0.00   1.00 456.69 mod55  9 -891.71      40.25   0.00   1.00 455.90 mod51  9 -882.83      49.13   0.00   1.00 451.46 mod39 11 -859.67      72.29   0.00   1.00 442.40 mod47 12 -848.28      83.68   0.00   1.00 438.02 mod44 12 -845.66      86.30   0.00   1.00 436.71 mod41 12 -844.96      87.00   0.00   1.00 436.36 mod64 12 -843.87      88.09   0.00   1.00 435.81 mod43 12 -843.43      88.53   0.00   1.00 435.60 mod48 12 -843.32      88.64   0.00   1.00 435.54 mod45 12 -842.27      89.69   0.00   1.00 435.02 mod49 12 -841.85      90.10   0.00   1.00 434.81 mod46 12 -840.45      91.51   0.00   1.00 434.11 mod42 12 -834.94      97.02   0.00   1.00 431.35 mod11  4 -792.55     139.41   0.00   1.00 400.50 mod8   4 -789.86     142.10   0.00   1.00 399.15 mod5   4 -787.64     144.32   0.00   1.00 398.04 mod60  4 -787.36     144.60   0.00   1.00 397.90 mod7   4 -786.61     145.35   0.00   1.00 397.53 mod9   4 -786.38     145.58   0.00   1.00 397.41 mod12  4 -786.23     145.72   0.00   1.00 397.34 mod13  4 -785.94     146.02   0.00   1.00 397.19 mod10  4 -784.62     147.34   0.00   1.00 396.53 mod3   5 -776.19     155.77   0.00   1.00 393.43 mod6   4 -772.50     159.46   0.00   1.00 390.47 mod27  6 -766.21     165.75   0.00   1.00 389.58 mod24  6 -763.57     168.39   0.00   1.00 388.26 mod62  6 -760.78     171.18   0.00   1.00 386.86 mod23  6 -760.21     171.75   0.00   1.00 386.58 mod25  6 -760.03     171.92   0.00   1.00 386.49   173  Appendix A.4        df    AICc Delta_AICc AICcWt Cum.Wt Res.LL mod29  6 -759.77     172.19   0.00   1.00 386.36 mod28  6 -759.74     172.22   0.00   1.00 386.34 mod26  6 -758.27     173.69   0.00   1.00 385.61 mod59 17 -754.30     177.66   0.00   1.00 398.07 mod2   8 -733.00     198.96   0.00   1.00 375.33 mod20  9 -723.06     208.90   0.00   1.00 371.57 mod17  9 -720.35     211.61   0.00   1.00 370.22 mod14  9 -719.26     212.70   0.00   1.00 369.68 mod61  9 -717.78     214.18   0.00   1.00 368.94 mod21  9 -717.62     214.33   0.00   1.00 368.86 mod16  9 -717.50     214.46   0.00   1.00 368.79 mod18  9 -716.92     215.04   0.00   1.00 368.51 mod22  9 -716.49     215.47   0.00   1.00 368.29 mod19  9 -715.13     216.83   0.00   1.00 367.61 mod15  9 -704.57     227.39   0.00   1.00 362.33 mod1  21 -696.09     235.87   0.00   1.00 375.29              174  Appendix A.5 Candidate set of models tested and results of model selection for decomposition rates of He, Si, Ma, and Li species. Bold lines indicate models with deltaAIC≤2.  Cand.models<-list() Cand.models[[1]]<-glm(K~RFM+PC1RF+SD+SED+NIT+TEMP+CON+DO+DIS+CWI+CR24,data=dadosME,family=gaussian) Cand.models[[2]]<-glm(K~RFM,data=dadosME,family=gaussian) Cand.models[[3]]<-glm(K~RFR,data=dadosME,family=gaussian) Cand.models[[4]]<-glm(K~PC1RF,data=dadosME,family=gaussian) Cand.models[[5]]<-glm(K~SD,data=dadosME,family=gaussian) Cand.models[[6]]<-glm(K~SED,data=dadosME,family=gaussian) Cand.models[[7]]<-glm(K~NIT,data=dadosME,family=gaussian) Cand.models[[8]]<-glm(K~TEMP,data=dadosME,family=gaussian) Cand.models[[9]]<-glm(K~CON,data=dadosME,family=gaussian) Cand.models[[10]]<-glm(K~DIS,data=dadosME,family=gaussian) Cand.models[[11]]<-glm(K~CWI,data=dadosME,family=gaussian) Cand.models[[12]]<-glm(K~CR24,data=dadosME,family=gaussian) Cand.models[[13]]<-glm(K~PC1RF+RFM,data=dadosME,family=gaussian) Cand.models[[14]]<-glm(K~SD+RFM,data=dadosME,family=gaussian) Cand.models[[15]]<-glm(K~SED+RFM,data=dadosME,family=gaussian) Cand.models[[16]]<-glm(K~NIT+RFM,data=dadosME,family=gaussian) Cand.models[[17]]<-glm(K~TEMP+RFM,data=dadosME,family=gaussian) Cand.models[[18]]<-glm(K~CON+RFM,data=dadosME,family=gaussian) Cand.models[[19]]<-glm(K~DIS+RFM,data=dadosME,family=gaussian) Cand.models[[20]]<-glm(K~CWI+RFM,data=dadosME,family=gaussian) Cand.models[[21]]<-glm(K~CR24+RFM,data=dadosME,family=gaussian) Cand.models[[22]]<-glm(K~SED+RFR,data=dadosME,family=gaussian) Cand.models[[23]]<-glm(K~NIT+RFR,data=dadosME,family=gaussian) Cand.models[[24]]<-glm(K~TEMP+RFR,data=dadosME,family=gaussian) Cand.models[[25]]<-glm(K~CON+RFR,data=dadosME,family=gaussian) Cand.models[[26]]<-glm(K~DIS+RFR,data=dadosME,family=gaussian) Cand.models[[27]]<-glm(K~CWI+RFR, data=dadosME,family=gaussian) Cand.models[[28]]<-glm(K~CR24+RFR,data=dadosME,family=gaussian) Cand.models[[29]]<-glm(K~DO,data=dadosME,family=gaussian) Cand.models[[30]]<-glm(K~DO+RFM,data=dadosME,family=gaussian) Cand.models[[31]]<-glm(K~DO+RFR,data=dadosME,family=gaussian)  Modnames<-paste("mod", 1:length(Cand.models), sep="") aictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE)           175  Appendix A.5  Species He  Model selection based on AICc:        df    AICc Delta_AICc AICcWt Cum.Wt     LL mod5   3 -224.82       0.00   0.19   0.19 116.01 mod3   4 -223.60       1.21   0.10   0.30 116.85 mod6   3 -222.84       1.97   0.07   0.37 115.02 mod11  3 -222.62       2.20   0.06   0.43 114.91 mod8   3 -222.25       2.56   0.05   0.48 114.73 mod10  3 -222.22       2.60   0.05   0.54 114.71 mod22  5 -221.88       2.94   0.04   0.58 117.61 mod24  5 -221.87       2.95   0.04   0.62 117.60 mod14  8 -221.83       2.99   0.04   0.67 123.71 mod7   3 -221.53       3.29   0.04   0.70 114.37 mod29  3 -221.49       3.33   0.04   0.74 114.35 mod27  5 -221.41       3.41   0.03   0.78 117.37 mod12  3 -221.41       3.41   0.03   0.81 114.30 mod4   3 -221.37       3.45   0.03   0.84 114.29 mod9   3 -221.34       3.48   0.03   0.88 114.27 mod26  5 -221.04       3.77   0.03   0.91 117.19 mod28  5 -220.84       3.98   0.03   0.93 117.09 mod25  5 -220.59       4.23   0.02   0.96 116.96 mod23  5 -220.43       4.39   0.02   0.98 116.88 mod31  5 -220.38       4.44   0.02   1.00 116.86 mod2   7 -212.67      12.15   0.00   1.00 116.84 mod13  8 -210.68      14.14   0.00   1.00 118.14 mod17  8 -209.57      15.25   0.00   1.00 117.58 mod15  8 -209.56      15.25   0.00   1.00 117.58 mod20  8 -209.23      15.59   0.00   1.00 117.41 mod19  8 -208.98      15.84   0.00   1.00 117.29 mod21  8 -208.80      16.01   0.00   1.00 117.20 mod30  8 -208.73      16.08   0.00   1.00 117.17 mod18  8 -208.35      16.47   0.00   1.00 116.97 mod16  8 -208.23      16.59   0.00   1.00 116.92 mod1  17 -119.23     105.58   0.00   1.00 127.62                176  Appendix A.5  Species Si  Model selection based on AICc:        df    AICc Delta_AICc AICcWt Cum.Wt     LL mod3   4 -255.01       0.00   0.19   0.19 132.56 mod26  5 -254.73       0.27   0.16   0.35 134.03 mod24  5 -253.60       1.40   0.09   0.45 133.47 mod25  5 -253.41       1.60   0.08   0.53 133.37 mod27  5 -252.83       2.17   0.06   0.60 133.08 mod28  5 -252.65       2.36   0.06   0.65 132.99 mod22  5 -251.99       3.02   0.04   0.70 132.66 mod31  5 -251.86       3.14   0.04   0.74 132.60 mod23  5 -251.78       3.23   0.04   0.77 132.56 mod4   3 -251.46       3.54   0.03   0.80 129.33 mod8   3 -251.26       3.75   0.03   0.83 129.23 mod12  3 -251.17       3.83   0.03   0.86 129.19 mod10  3 -251.12       3.88   0.03   0.89 129.16 mod5   3 -250.80       4.21   0.02   0.91 129.00 mod7   3 -250.56       4.45   0.02   0.93 128.88 mod11  3 -250.33       4.67   0.02   0.95 128.77 mod9   3 -250.28       4.73   0.02   0.97 128.74 mod29  3 -250.03       4.98   0.02   0.98 128.62 mod6   3 -250.02       4.98   0.02   1.00 128.61 mod2   7 -240.89      14.11   0.00   1.00 130.95 mod14  8 -239.28      15.73   0.00   1.00 132.44 mod19  8 -238.86      16.14   0.00   1.00 132.23 mod21  8 -238.04      16.96   0.00   1.00 131.82 mod18  8 -237.88      17.12   0.00   1.00 131.74 mod20  8 -236.91      18.10   0.00   1.00 131.25 mod15  8 -236.80      18.21   0.00   1.00 131.20 mod17  8 -236.35      18.65   0.00   1.00 130.98 mod13  8 -236.31      18.70   0.00   1.00 130.95 mod16  8 -236.30      18.70   0.00   1.00 130.95 mod30  8 -236.30      18.70   0.00   1.00 130.95 mod1  17 -147.42     107.59   0.00   1.00 141.71                177  Appendix A.5  Species Ma  Model selection based on AICc:        df    AICc Delta_AICc AICcWt Cum.Wt     LL mod31  5 -257.15       0.00   0.20   0.20 135.24 mod4   3 -256.97       0.19   0.18   0.38 132.08 mod3   4 -255.75       1.40   0.10   0.48 132.93 mod28  5 -255.56       1.59   0.09   0.57 134.45 mod29  3 -255.37       1.78   0.08   0.65 131.28 mod24  5 -254.68       2.47   0.06   0.71 134.01 mod22  5 -254.18       2.97   0.05   0.76 133.76 mod8   3 -253.87       3.28   0.04   0.80 130.54 mod23  5 -253.31       3.85   0.03   0.83 133.32 mod26  5 -252.85       4.30   0.02   0.85 133.09 mod25  5 -252.81       4.35   0.02   0.87 133.07 mod27  5 -252.58       4.58   0.02   0.89 132.95 mod6   3 -252.25       4.91   0.02   0.91 129.72 mod9   3 -251.76       5.39   0.01   0.92 129.48 mod12  3 -251.53       5.62   0.01   0.94 129.37 mod5   3 -251.31       5.84   0.01   0.95 129.25 mod10  3 -251.23       5.92   0.01   0.96 129.22 mod7   3 -251.07       6.08   0.01   0.97 129.14 mod11  3 -251.07       6.09   0.01   0.98 129.13 mod30  8 -250.79       6.37   0.01   0.98 138.19 mod2   7 -250.10       7.05   0.01   0.99 135.55 mod16  8 -248.47       8.69   0.00   0.99 137.03 mod20  8 -248.12       9.03   0.00   0.99 136.86 mod21  8 -246.99      10.16   0.00   1.00 136.30 mod14  8 -245.93      11.22   0.00   1.00 135.77 mod19  8 -245.87      11.28   0.00   1.00 135.74 mod18  8 -245.60      11.55   0.00   1.00 135.60 mod17  8 -245.59      11.57   0.00   1.00 135.59 mod13  8 -245.56      11.59   0.00   1.00 135.58 mod15  8 -245.51      11.65   0.00   1.00 135.55 mod1  17 -157.90      99.25   0.00   1.00 146.95                178  Appendix A.5  Species Li  Model selection based on AICc:        df    AICc Delta_AICc AICcWt Cum.Wt     LL mod4   3 -277.88       0.00   0.17   0.17 142.54 mod11  3 -277.19       0.69   0.12   0.30 142.19 mod3   4 -276.57       1.31   0.09   0.39 143.34 mod29  3 -275.81       2.07   0.06   0.45 141.50 mod8   3 -275.76       2.11   0.06   0.51 141.48 mod9   3 -275.70       2.18   0.06   0.57 141.45 mod27  5 -275.47       2.41   0.05   0.62 144.40 mod6   3 -275.24       2.64   0.05   0.67 141.22 mod5   3 -275.15       2.73   0.04   0.72 141.17 mod7   3 -275.06       2.82   0.04   0.76 141.13 mod12  3 -275.01       2.87   0.04   0.80 141.11 mod10  3 -275.01       2.87   0.04   0.84 141.10 mod31  5 -274.12       3.76   0.03   0.87 143.73 mod23  5 -273.95       3.93   0.02   0.89 143.64 mod25  5 -273.80       4.07   0.02   0.92 143.57 mod22  5 -273.74       4.14   0.02   0.94 143.53 mod28  5 -273.66       4.22   0.02   0.96 143.50 mod24  5 -273.53       4.34   0.02   0.98 143.43 mod26  5 -273.35       4.53   0.02   1.00 143.34 mod2   7 -268.30       9.58   0.00   1.00 144.65 mod30  8 -265.09      12.79   0.00   1.00 145.34 mod16  8 -264.85      13.03   0.00   1.00 145.22 mod15  8 -264.73      13.15   0.00   1.00 145.17 mod18  8 -264.57      13.31   0.00   1.00 145.08 mod21  8 -264.35      13.53   0.00   1.00 144.98 mod20  8 -264.30      13.58   0.00   1.00 144.95 mod14  8 -263.97      13.91   0.00   1.00 144.78 mod13  8 -263.75      14.13   0.00   1.00 144.68 mod17  8 -263.72      14.16   0.00   1.00 144.66 mod19  8 -263.71      14.17   0.00   1.00 144.66 mod1  17 -165.35     112.52   0.00   1.00 150.68             179  Figure A.1 Allocation of riparian plots (squares) and stream transects (black dots) along the sampled stream reach.        0 m150 m10 m10 mD1 D2 D3 D4 D5WidthCAChannel- depth (D)- width- canopy (CA)• height• openness• light intensity Riparian forest- Stems ≥ 3.5 cm dbh• density • basal area 180  Figure A.2 Bar graphs showing means (± 1 S.D.) of dissolved oxygen. YR - Young regeneration. YRF - Young regeneration surrounded by forest. OR - Old regeneration. REF - Reference. YRPB – Buffer of young regeneration surrounded by pasture. MFPB – Buffer of mature forest surrounded by pasture.               A B C D E FManagement groupsDissolved oxygen (mg/L) 02468YR OR REF YRPBYRF MFPB181  Figure A.3 Relationships between k – He and suspended sediments (A), k-Si and temperature (B), k-Si and conductivity (C), and k-Si and discharge (D) in 24 streams. Codes in A: YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○). Codes and colors in B, C, and D: SI – Secondary initial (Δ, red line). SL – Secondary late (+, black line). MF – Mature forest (○, green line).        A BC0.0 0.5 1.0 1.5 2.0 2.5 3.00.0060.0080.0100.0120.014Suspended sediments (mg/L)k (g/day)25 26 27 28 290.0070.0080.0090.010Temperature (oC)K(g/day)14 16 18 200.0070.0080.0090.010Conductivity (uS/cm)K(g/day)0.05 0.10 0.150.0070.0080.0090.010Discharge (m3/s)K(g/day)Dk(Ln (% AFDM/day)) 0.0060.0080.0100.0120.0143.2.2.1.100Suspended se i ents (mg/L)0.0070.0080.0090.0102Temp rature (oC)20111Conductivit  (µS/cm)0.0070.0080.0090.0100.0070.0080.0090.0100.0.. 5D charge (m3/s)182  Figure A.4 Relationships between k-Ma and dissolved oxygen (A), k-Ma and respiration (B), k-Ma and dissolved oxygen (C), and k-Li and channel (D) width in 24 streams. Codes and colors in A and B: SI – Secondary initial (Δ, red line). SL – Secondary late (+, black line). MF – Mature forest (○, green line).Codes in B: YR - Young regeneration (x). YRF - Young regeneration surrounded by forest (◊). OR - Old regeneration (+). REF – Reference (Δ). YRPB – Buffer of young regeneration surrounded by pasture ( ). MFPB – Buffer of mature forest surrounded by pasture (○).   4.5 5.0 5.5 6.0 6.5 7.00.0030.0040.0050.0060.007Dissolved oxygen (mg/L)K(g/day)1 2 3 40.0030.0040.0050.0060.007Respiration (gO2/m2/day)K(g/day)4.5 5.0 5.5 6.0 6.5 7.00.0030.0040.0050.0060.007Dissolved oxygen (mg/Lk (g/day)2 3 4 5 60.00050.00100.00150.00200.00250.00300.0035Channel width (m)k (g/day)A BC D0.0030.0070.0040.0050.006Respiratio  (gO2/m2/day)4Dissolved oxygen (mg/L)6.5.54 7.6.0.0030.0070.0040.0050.0060.0030.0070.0040.0050.006Dissolved ox en (mg/L)6.5.5.. 7.06.Channel width (m)0.00050.00250.00100.00150.00200.00300.0035k(Ln (% AFDM/day))

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0367004/manifest

Comment

Related Items