Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Biomass, productivity and allocation patterns in tropical old-growth and logged-over forests in Ghana Addo-Danso, Shalom Daniel 2017

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

Item Metadata

Download

Media
24-ubc_2017_november_addodanso_shalom.pdf [ 3.77MB ]
Metadata
JSON: 24-1.0355268.json
JSON-LD: 24-1.0355268-ld.json
RDF/XML (Pretty): 24-1.0355268-rdf.xml
RDF/JSON: 24-1.0355268-rdf.json
Turtle: 24-1.0355268-turtle.txt
N-Triples: 24-1.0355268-rdf-ntriples.txt
Original Record: 24-1.0355268-source.json
Full Text
24-1.0355268-fulltext.txt
Citation
24-1.0355268.ris

Full Text

 BIOMASS, PRODUCTIVITY AND ALLOCATION PATTERNS IN TROPICAL OLD-GROWTH AND LOGGED-OVER FORESTS IN GHANA  by Shalom Daniel Addo-Danso  B.Sc., Kwame Nkrumah University of Science and Technology, 2005 M.Sc., Albert-Ludwigs University, 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)   August 2017  © Shalom Daniel Addo-Danso, 2017    ii Abstract Understanding how tropical forest structure and function change during the decades after logging is a key research challenge. This thesis reports functional traits, forest structure, biomass, net primary productivity (NPP) and allocation, as well as their controlling factors in an old-growth forest and a 54-year-old logged-over forest in Ghana. By analyzing root traits, I found fine-root biomass, root length, surface area, and root tissue density were higher in the logged-over forest, whereas the old-growth forest had higher specific root length and specific root area. I also found divergent exploitation strategies between the two forests; plants in the old-growth forest produced thinner roots, which increase resource uptake efficiency, while plants in the logged-over forest had thicker roots, which are associated with greater resource conservation. Through correlation analysis, I found that fine-root mass correlated positively to relative humidity, while absorbed photosynthetically active radiation and fine-root biomass were also positively correlated. Fine-root mass and soil K were also positively correlated, and fine-root necromass correlated positively with soil P. I then explored the relationships between leaf traits, taxonomic (e.g., species richness) or structural (e.g., tree diameter) variables and aboveground biomass (AGB) or coarse wood productivity (CWP) in the two forests. Leaf K related positively to tree biomass in the logged-over forest. Leaf N and P were significantly and positively related to tree productivity in the old-growth forest and logged-over forest. AGB and CWP were mostly explained by the structural variables. The shape and magnitude of the relationships between tree species richness and AGB or CWP differed between the two forests. In addition, I found that leaf area index, mean tree diameter and height were similar between the two forests, but stand density and basal area were higher in the logged-over forest than in the old-growth forest. Total biomass and annual NPP were comparable in both forests, but there was a shift in NPP allocation between wood and fine roots. I conclude that the forest structure, biomass and productivity of the logged-over forest have largely recovered, but the legacy of logging still persists, which is reflected in differences in functional traits and allocation patterns.     iii Lay summary In many tropical countries trees are selectively harvested from the forests for timber and to manufacture wood products. Usually, other trees and the soil are affected during harvesting. Yet, how tropical forests change during the decades after harvesting trees is not clear. I examined plant roots, and found that the way plants obtain water and nutrients from the soil differ in an old-growth forest and a 54-year-old harvested forest. I also found that tree sizes and tree height were similar in the two forests, but the harvested forest had more trees in a given area than the old-growth forest. Total biomass and productivity were similar in both forests. I conclude that the biomass and productivity in the harvested forest are again at pre-harvesting levels 54 years after the trees were cut, but some of the harvesting effects persist, which is seen in the changes in the traits and forest structure.                    iv Preface Research chapters of this thesis were written as a series of manuscripts for publishing in peer-reviewed journals. This thesis is partly based on a proposal by Yadvinder Malhi, Stephen Adu-Bredu and Lee White, which identified the broad research areas and research sites (http://gem.tropicalforests.ox.ac.uk/). I was responsible for developing research questions and selecting the sites for this work. I proposed and designed the experiments on functional traits, root methods, woody debris, and soil properties (Chapter 3-6). My contribution to all the research chapters involves experimental design, data collection, data analyses and preparation of draft manuscripts including tables and figures. Several research assistants were involved in the data collection and sample processing in the laboratory (see acknowledgments). Dr. Cindy E. Prescott and Dr. Robert D. Guy, my co-supervisors, and members of the supervisory committee (Dr. Peter Marshall, Dr. David I. Forrester and Dr. Stephen Adu-Bredu) provided help with the development of the research proposal and the editing of my thesis.   A version of Chapter 2 has been published: Addo-Danso SD, Prescott CE, Smith AR. 2016. Methods for estimating root biomass and production in forest and woodland ecosystem carbon studies: A review. Forest Ecology and Management 359: 332-351.   Shalom D. Addo-Danso conceived the idea, conducted the literature survey, performed the statistical analyses and wrote the manuscript.   Cindy E. Prescott helped with concept formation and provided manuscript edits.  Andrew Smith provided additional data on coarse root biomass and provided manuscript edits  A version of Chapter 3 is accepted pending revision: Addo-Danso SD, Prescott CE, Guy RD, Duah-Gyamfi A, Moore S, Owusu-Afriyie K, Marshall P, Forrester DI, Adu-Bredu S, Malhi Y. Root exploitation strategies differed in tropical old-growth and logged-over forests in Ghana. Biotropica.  v  Shalom D. Addo-Danso conceived the idea, designed the experiment, collected data, performed the statistical analyses and wrote the manuscript.   Cindy E. Prescott helped with concept formation and provided manuscript edits.  Stephen Adu-Bredu helped with experimental design and provided manuscript edits.  Robert D. Guy, Duah-Gyamfi A, Moore S, Owusu-Afriyie K, Marshall P, Forrester DI, and Malhi Y. helped with manuscript edits.                      vi Table of contents  Abstract……………………………………………………………………………………………………ii Lay summary……………………………………………………………………………………………..iii Preface………………………………………………………………………...…………………………...iv Table of contents………………………………………………………………………………………….vi List of tables……………………………………………………………………...………………………xiv List of figures……………………………………………………………………...……………………xviii List of symbols and abbreviations……………………………………………….…..………………..xxiii Acknowledgements………………………………………………………………...……………….....xxvii  Dedication…………………………………………………………….…..…………….…………........xxix Chapter 1: Introduction…………………………………………………………………………...……...1      1.1 Tropical forests and their relevance………………………………………………………...……….1         1.1.1 Threats to tropical forests………………………………………………………………...…...…2         1.1.2 Logging....……………………………………...…………………………...…………………...2             1.1.2.1 Impacts of logging……………………………………………………………………..…....3             1.1.2.2 Emerging ideas on logged-over forests……….………………………………………..……5      1.2 Biomass and net primary productivity…………………………………………………………...….6          1.2.1 Factors that affect biomass and net primary productivity………………………………...……..7            1.2.1.1 Biotic factors………………………………………………………………………...………8            1.2.1.2 Abiotic factors………………………………………………………………….…..………10            1.2.1.3 Other factors (methodology)………………………………………………………...……..12        1.3 Context: Ghanaian tropical forests……………………………………………………..…….…….13          1.3.1 Logging in Ghana………………………………………………………………….……..…….16      1.4 Thesis overview, conceptual framework and research questions………………………………….17          1.4.1 Forest structure and functional species composition…………………………………………..17  vii         1.4.2 Resource availability……………………………………….….………………………………18         1.4.3 Plant functional traits..................................................................................................................19         1.4.4 Chapter 2: Methods for estimating root biomass and productivity: A review and           global analysis…………………………………………………………………...…………………..20          1.4.5 Chapter 3: Root exploitation strategies differ in tropical old-growth forest and logged-over                  forest in Ghana……………………………………………………………………...……………….20         1.4.6 Chapter 4: Patterns and controls on root dynamics in tropical forests in Ghana,          West Africa………………………………………………………………………………..…………21         1.4.4 Chapter 5: Aboveground wood biomass and productivity in tropical old-growth and logged-           over forests: the importance of taxonomic variables, stand structural variables and traits……..…..21          1.4.5 Chapter 6: Biomass and productivity are similar, but allocation patterns differ between          tropical  old-growth and logged-over forests in Ghana………………………………………...……22         1.4.6 Chapter 7: Synthesis and general conclusions…………………………………………...…….22  Chapter 2: Methods for estimating root biomass and productivity: A review and global       analysis…………………………………………………………………………………………...……23      2.1 Synopsis………………………………………………………………………………………...….23       2.2 Introduction…………………………………………………………………………………...…....24       2.3 Methods………………………………………………………………………………….................26         2.3.1 Literature search and data compilation…………………………………………………...……26          2.3.2 Statistical Analyses………………………………………………………………………...…..29       2.4 Results and Discussion…………………….…………………………………………………..…...29          2.4.1 Literature review…………………………………………………………………………...…..29            2.4.1.1 Coarse-root biomass and productivity……...………………………………………..….…29                2.4.1.1.1 Direct methods..……….…………………………………………………………...…..29                  2.4.1.1.1.1 Root excavation…………………..………………………………………………...30                  2.4.1.1.1.2 Soil-pit/Soil-pit ingrowth…………...………………………………………......….31   viii                  2.4.1.1.1.3 Wall or soil trench profile....……………...……………………………………......34                   2.4.1.1.1.4 Soil-core…………………………………………………………………...……….35                2.4.1.1.2 Indirect methods………………………………………………………………...……...35                  2.4.1.1.2.1 Size-mass allometric equations……………………………………….……..……..36                   2.4.1.1.2.2 Root-shoot or Belowground-aboveground ratio………………………..………….37                  2.4.1.1.2.3 Ground-Penetrating Radar………………………………………………………....39                  2.4.1.1.2.4 Root biomass increment or difference……………………………………….…….41                  2.4.1.1.2.5 Fraction or percentage of wood productivity…………………………….…..…….42                   2.4.1.1.2.6 Root radial increment…………………………………………………………...….42             2.4.1.2 Fine-root biomass and productivity………………………………………………….….…43               2.4.1.2.1 Direct methods…………....……………………………………………….…………...44                   2.4.1.2.1.1 Soil-core…………………………………………………………………………....44                  2.4.1.2.1.2 Monolith……………...………………………………………………………...…..46                  2.4.1.2.1.3 Ingrowth-core…………………………………………………………………...….48                   2.4.1.2.1.4 (Mini) rhizotrons………………………………………………………………...…48                  2.4.1.2.1.5 Sequential-coring………………………………………………………………..…50                2.4.1.2.2 Indirect methods………….…………………………………………………………….51                   2.4.1.2.2.1 The Pipe model and others……………………………………………………...….51                    2.4.1.2.2.2 N budget…………………………………………………………………………....52          2.4.2 Comparison of methods…………………………………………………………………..……54             2.4.2.1 Coarse-root biomass estimates from soil-pit and soil-core methods…………………..…..54            2.4.2.2 Fine-root biomass estimates from soil-pit and soil-core methods……………………..…..55            2.4.2.3 Fine-root productivity estimates from ingrowth-core, (mini) rhizotrons and             sequential-coring methods and biome estimates……………………………………...……………56  2.5 Conclusions…………………………………………………………………………………...……….59    ix Chapter 3: Root exploitation strategies differ in tropical old-growth forest and logged-over  forest in Ghana………………………………………………………………………………...…………60      3.1 Synopsis…………………………………………………………………………………...……….60        3.2 Introduction……………………………………………………………………….………..………61      3.3 Materials and Methods………………………………………………..……………...…………….63         3.3.1 Study area………………………………………………………………………...…………….63         3.3.2 Study forests………………………………………………………………………...………….64          3.3.3 Root sampling and processing………………………………………………..………………..66            3.3.3.1 Root biomass……………………………………………………………………...………..66            3.3.3.2 Root morphology………………………………………………………..…………………67          3.3.4 Statistical analysis………………………………………………………………………..…….67         3.4 Results…………………………...………………………………………………………...……..68            3.4.1 Root biomass and distribution…...………………………………………….………..………68             3.4.2 Root morphological traits and distribution……...……………………………..…………….69             3.4.3 Root exploitation strategies………...…………………………………………...…………....72       3.5 Discussion……………………………………………………………………………..…………...72            3.5.1 Root biomass and distribution…………………………………………………...…………...72             3.5.2 Root morphological traits and distribution……………………………………..……………74             3.5.3 Root exploitation strategies differed in the old-growth forest and logged-over forest…...….76      3.6 Conclusions………………………………………………………………………………………...77  Chapter 4: Patterns and controls on root dynamics in tropical forests in Ghana, West Africa…....78      4.1 Synopsis……………………………………………………………………………………………78      4.2 Introduction………………………………………………………………………………………...79       4.3 Materials and methods……………………………………………………………………………..81          4.3.1 Study area………………………………………………………………………………………81          4.3.2 Root sampling….……………………………………………………………………………....81  x            4.3.2.1 Ingrowth-core method………………………………………………………………...……81             4.3.2.2 Sequential-coring method………………………………………………………..………...82          4.3.3 Root processing, characterization and weighing…………………………………...…………..83          4.3.4 Environmental and soil chemistry variables……………………………………..…………….84             4.3.4.1 Environmental variables………………………………………………………..………….84            4.3.4.2 Soil chemistry…………………………………………………………………..………….85          4.3.5 Calculations and statistical analysis….…...………………………………………..…………..85             4.3.5.1 Predicted root dry mass…....…………………………………………………..…………...85            4.3.5.2 Root productivity………………………………………………………………..…………86               4.3.5.2.1 Ingrowth-core…………………………………………………………...…………...…86               4.3.5.2.1 Sequential-coring……………………....……………………………..………….…….87                4.3.5.3 Root turnover rate………………………………………………………..………………88          4.3.6 Statistical analysis………………………………………………………..…………………….89       4.4 Results………………………………………………………………………..…………………….90          4.4.1 Patterns of environmental and soil chemistry variables…………………...…………………...90          4.4.2 Root mass distribution and relationship with environmental and soil chemistry          variables……………………………………………………………………………..……………….91         4.4.3 Root productivity and turnover rates from ingrowth-coring and sequential-coring methods…97         4.4.4 Root productivity and turnover rates in old-growth forest and logged-over forest…………....97      4.5 Discussion……………………………………………………………………………………..…...99         4.5.1 Root mass distribution and relationship with environmental and soil chemistry          variables……………………………………………………………………………………..……….99         4.5.2 Root productivity and turnover rates from ingrowth-coring and sequential-         coring methods……………………………………………………………………………………...101         4.5.3 Root productivity and turnover rates in old-growth forest and logged-over forest………..…103      4.6 Conclusions…………………………………………………………………………………….....105  xi Chapter 5: Aboveground wood biomass and productivity in old-growth and  logged-over forests: the importance of taxonomic variables, stand structural variables  and traits………………………………………………………………………………………………...106       5.1 Synopsis…………………………………………………………………………………………..106       5.2 Introduction……………………………………………………………………………………….107       5.3 Materials and methods…………………………………………………………….………..….…109         5.3.1 Study area………………………………………………………………………………..……109         5.3.2 Study forests…………………………………………………………………………..………110          5.3.3 Plot inventory…………………………………………………………………………..……..110            5.3.3.1 Data cleaning and gap-filling……………………………………………………..………111             5.3.3.2 Site-specific diameter-height equation…………………………………………..………..112          5.3.4 Aboveground biomass and coarse wood productivity estimation……………………………113          5.3.5 Taxonomic and stand structural variables…...…………………………………………..…....114         5.3.6 Leaf sampling and trait determination………………………………………………………..115           5.3.7 Statistical analyses………………………………….………………………………………...118       5.4 Results………………………………………………………………………………….…………121          5.4.1 Forests characteristics……………………………………………………………….………..121          5.4.2 Relationships between taxonomic, structural variables and aboveground biomass and          productivity………………………………………………………………………………………....122         5.4.3 Relationships between leaf traits and tree species biomass and productivity……………..….128          5.4.4 Relative importance of leaf traits, taxonomic and structural variables…………………….…131          5.4.5 Examining support for niche complementarity and the selection effect....……………….….132       5.5 Discussion…………………………………………………………………………………...........133          5.5.1 Forests characteristics…………………………………………………………………..…….133          5.5.2 Taxonomic variables relate to aboveground biomass and coarse wood productivity…...........134          5.5.3 Structural variables are better predictors of aboveground biomass and coarse wood   xii         productivity…………………………………………………………………………………........…136          5.5.4 Leaf chemical traits relate with tree biomass and productivity………………………........…137       5.6 Conclusions………………………………………………………………………………..…...…138 Chapter 6: Biomass and productivity are similar, but allocation patterns differ in old-growth forest and logged-over forest in Ghana……………………………………………………………………….139      6.1 Synopsis……………………………………………………………………..……………………139      6.2 Introduction……………………………………………………………………………………….140       6.3 Materials and method………………………………………………………...…………………...142         6.3.1 Study area……………………………………………………………………………………..142          6.3.2 Study forests…………………………………………………………………………………..142          6.3.3 Aboveground biomass……………………………….………………………………………..143            6.3.3.1 Leaf biomass……………………………………………………………………………...143            6.3.3.2 Stem biomass……………………………………………………………………………..144            6.3.3.3 Fine and coarse woody debris…………………………………………………………….146           6.3.4 Belowground biomass…………………………………………………………………….…..149             6.3.4.1 Fine- and small-root biomass……………………………………………………………..149            6.3.4.2 Coarse-root biomass………………………………………………………………………150         6.3.5 Aboveground net primary productivity……………………………………………………….150              6.3.5.1 Canopy productivity………………………………………………………………………150             6.3.5.2 Stem productivity…………………………………………………………………………151             6.3.5.3 Branch turnover productivity……………………………………………………………..152          6.3.6 Belowground net primary productivity……………………………………………………….152             6.3.6.1 Fine- and small-root productivity………………………………………………………...152            6.3.6.2 Coarse-root productivity………………………………………………………………….153          6.3.7 Allocation of net primary productivity……………………………………………………….153         6.3.8 Calculation of total biomass and net primary productivity….………………………………..154   xiii         6.3.9 Error propagation and statistical analyses…………………………………………………….155       6.4 Results…………………………………………………………………………………………….155          6.4.1 Forest structure………………………………………………………………………………..155          6.4.2 Aboveground and belowground biomass……………………………………………………..156         6.4.3 Aboveground and belowground net primary productivity……………………………………159         6.4.4 Allocation of net primary productivity……………………………………………………….160      6.5 Discussion…………………………………………………………………………………….…..163          6.5.1 Aboveground and belowground biomass……………………………………………………..163          6.5.2 Aboveground and belowground net primary productivity……………………………………165           6.5.3 Shifts in net primary productivity allocation to wood and fine roots………………………...168      6.6 Conclusions……………………………………………………………………………………….169 Chapter 7: Synthesis and general conclusions………………………………………………………..171      7.1 Limitations of research………………….………...……………………………………………...176      7.2 Future directions………………………………………………………………………………….176 References………………………………….……………………………………………………………178 Appendices…………………………………..…………………………………………………………..235             xiv List of tables  Table 2.1 Comparison of methods for estimating coarse-root biomass and productivity using selected criteria……………………………………………………………………………………………..............45 Table 2.2 Comparison of methods for estimating fine-root biomass and productivity using selected criteria……………………………………………………………………………………………………..53 Table 2.3 Fine-root productivity estimates (Mg ha-1 year-1) (mean ± SE) from ingrowth-core, (mini) rhizotrons and sequential-coring methods for different biomes…………………………………………..57 Table 3.1 Site, soil and stand structural characteristics of the two study forests in the Bobiri Forest Reserve in Ghana………………………………………………………………………………………….65 Table 3.2 Fine-root (diameter < 2 mm) biomass estimates (Mg ha-1, N = 12) in 0-30 cm soil depth in the old-growth forest and the 54-year-old logged-over forest…………………………………..…………….68 Table 3.3 Significant effects of, and interactions between forest type (old-growth forest and 54-year-old logged-over forest) and soil depth (0-10 cm, 10-20 cm, 20-30 cm) on fine-root (diameter < 2 mm) morphological traits……………………………………………………………………………………….70 Table 3.4 Fine-root (diameter < 2 mm) morphological traits (mean ± SE, N = 12) at 0-30 cm soil depth in the old-growth forest and the logged-over forest..……….……………………………….……………….72 Table 4.1 Decision matrix approach modified from Fairley and Alexander (1985)….…………………..88  Table 4.2 Correlation coefficients (r) between environmental, soil chemistry and fine-root variables (biomass, necromass and mass). Data for the old-growth forest and the 54-year-old logged-over forest are combined…………………………………………………………………………………………………..96  Table 4.3 Fine and small-root productivity (Mg ha-1 yr-1) and turnover rate (yr-1) via ingrowth-core and sequential-coring calculation approaches in the old-growth forest and the 54-year-old logged-over forest……………………………………………………………………………………………………….98  Table 5.1 Site-specific allometric equations relating tree diameter (D in cm) and tree total height (H in m) showing linear and second-order polynomial models. The Akaike Information Criteria (AIC), the Root Mean Squared Error (RMSE), adjusted R2 and coefficients are given for each model. The best model  xv (lowest AIC and RMSE, and highest adjusted R2) is shown in bold. Data for the old-growth forest and 54-year-old logged-over forest are combined……………………………………………………………….113 Table 5.2 Allometric equations used to estimate aboveground biomass (AGB) of trees (≥ 10 DBH cm) in  the old-growth forest and the 54-year-old logged-over forest…………………………………………...114 Table 5.3 Leaf traits measured, with abbreviations, units and what they indicate………………………117 Table 5.4 List of tree species sampled with codes, family, functional group or guild used for trait determination. Diameter, height and wood density are also shown…………………………………...…118  Table 5.5. Summary of the taxonomic and structural variables estimated in the study. Estimates are based on trees ≥ 10 cm DBH. Data for the old-growth forest and 54-year-old logged-over forest are combined…………………………………………………………………………………………………121  Table 5.6 Aboveground biomass (AGB) and Coarse wood productivity (CWP) estimates obtained from three allometric equations used in the study (Henry et al. (2010) and Chave et al. (2005)). Estimates are based on trees ≥ 10 cm DBH. Data for the old-growth forest and 54-year-old logged-over forest are combined. Different letters represent significant differences (p < 0.05) between the equations………...122 Table 5.7 Linear and second-order polynomial models relating taxonomic variables and aboveground biomass (AGB) or coarse wood productivity (CWP) in the old-growth forest and the 54-year-old logged-over forest. The Akaike Information Criteria (AIC), adjusted R2 and coefficients are given for each model. The best models have lowest AIC and highest adjusted R2 in bold. AGB values were estimated with the Henry equation (excluding height and wood density) and CWP included height and wood density……123  Table 5.8 Multiple linear predictor models of aboveground biomass (AGB) and coarse wood productivity (CWP) in the old-growth forest and the 54-year-old logged-over forest. All models are significant at p < 0.05. AGB values were estimated with the Henry equation (excluding height and wood density) and CWP included height and wood density………………………………………………………………………..128 Table 5.9 Leaf traits (mean ± SE) of tree species sampled in the old-growth forest and the 54-year-old logged-over forest………………………………………………………………………………………..130   xvi Table 5.10 Leaf traits (mean ± SE) for shade-bearers, pioneers and non-pioneer light demanders (NPLDs) in the old-growth forest and the 54-year-old logged-over forest………………………………………...131 Table 6.1 Allometric equations used to estimate tree, palm and liana biomass for the old-growth forest and 54-year-old logged-over forest………………………………………………………………………146  Table 6.2 Description of decay classes for coarse woody debris (CWD) (modified from Harmon et al. 1995)……………………………………………………………………………………………………..148  Table 6.3 Forest structural variables (mean ± SE) for old-growth forest and 54-year-old logged-over forest. Diameter, height, wood density, volume, basal area and stand density values are based on stems (≥ 10 cm DBH)…………………………………..………………………………………………………….156 Table 6.4 Biomass of each component (mean ± SE) of the total biomass (Mg ha-1) of the old-growth forest and the 54-year-old logged-over forest. Significant differences are denoted p < 0.05……………158 Table 6.5 Forest structural components and stem biomass (mean ± SE) for shade-bearers, pioneers and non-pioneer light demanders (NPLDs) in the old-growth forest and the 54-year-old logged-over forest. Estimates are based on trees (≥ 10 cm DBH)……………………………………………………………159 Table 6.6 Net primary productivity of each component (mean ± SE) of the total NPP (Mg ha-1 yr-1) of the old-growth forest and the 54-year-old logged-over forest……………………………………………….161 Table 6.7 Stem productivity (mean ± SE) for shade-bearers, pioneers and non-pioneer light demanders (NPLDs) in the old-growth forest and the 54-year-old logged-over forest. Estimates are based on trees (≥ 10 cm DBH)……………………………………………………………………………………………...162 Table A.1 Data used in the analysis of fine-and coarse-root biomass, including references……………235  Table A.2 Data used in the analysis of fine-root productivity, including references……………………236 Table A.3 Fine-root morphological traits compared in old-growth (unlogged) forests and logged/secondary regrowth forests in the tropics………………...……………………………………...239 Table A.4 Variance Inflation Factors (VIF) of taxonomic and structural variables. Variables selected for the regression model are in bold…………………………………………………………………………241   xvii Table A.5 Matrix showing correlation (Pearson product-moment) between the taxonomic and structural variables in the old-growth forest and the 54-year-old logged-over forest………………………………242 Table A.6 Linear and second-order polynomial models relating taxonomic variables and aboveground biomass (AGB) or coarse wood productivity (CWP). The Akaike Information Criteria (AIC), adjusted R2 and coefficients are given for each model. The best model (lowest AIC and highest adjusted R2) is shown in bold. Data for the old-growth forest and 54-year-old logged-over forest are combined……………...243 Table A.7 Leaf traits (mean ± SE) in the old-growth forest and the logged-over forest. Data for all tree species in the old-growth forest and 54-year-old logged-over forest are combined. Different letters represent significant differences (p < 0.05) between the forests…………………………………………244 Table A.8 Ten tree species with highest contribution to stem (≥ 10 cm DBH) biomass and productivity in old-growth forest…………………………………………………………………………………………245 Table A.9 Ten tree species with highest contribution to stem (≥ 10 cm DBH) biomass and productivity in 54-year-old logged-over forest…………………………………………………………………………..246 Table A.10 Ten tree species with highest contribution to stem (2-9.9 cm DBH) biomass and productivity in old-growth forest……………………………………………………………………………………...247 Table A.11 Ten tree species with highest contribution to stem (2-9.9 cm DBH) biomass and productivity in 54-year-old logged-over forest………………………………………………………………………..248 Table A.12 Diameter, length, density and volume (mean ± SE) of coarse woody debris for old-growth forest and 54-year-old logged-over forest……………………………………………………………….249 Table A.13 Diameter, length, density and volume (mean ± SE) by class of branch turnover (including woody input) for old-growth forest and 54-year-old logged-over forest………………………………..250 Table A.14 Tree species (≥ 10 cm DBH) distribution per ha into functional groups in the old-growth forest. Species in bold were restricted to the old-growth forest…………………………………………251 Table A.15 Tree species (≥ 10 cm DBH) distribution per ha into functional groups in the 54-year-old logged-over forest. Species in bold were restricted to the 54-year-old logged-over forest….…………..254   xviii List of figures  Figure 1.1 Map of Ghana showing forest zones and forest types………………………………….……..15 Figure 1.2 Designated forest uses in Ghana………………………………………………………………17 Figure 2.1 Distribution of the sites used in the analysis of fine- and coarse root biomass and productivity. With the exception of Africa and Australia, each point represents more than one site…………………...28 Figure 2.2 Coarse-root (> 2 mm) biomass estimates (Mg ha-1) (mean ± SE) from soil-pit and soil-core methods (A), and relationship between coarse-root biomass estimates (Mg ha-1) (mean ± SE) from soil-pit and soil-core methods (B, N = 11). Broken line is 1:1 relationship between the methods. Data were derived from the same sites, and relationship values are log-transformed (base 10)…….……………….54   Figure 2.3 Fine-root (≤ 2 mm) biomass estimates (Mg ha-1) (N = 9) from soil-pit and soil-core methods (A), and relationship between fine-root biomass estimates (Mg ha-1) from soil-pit and soil-core (B, N = 9). Broken line is 1:1 relationship between the methods. Data were derived from the same sites.….……….56 Figure 2.4 Fine-root productivity estimates (Mg ha-1 year-1) (mean ± SE) from ingrowth-core (N = 81), (mini) rhizotrons (N = 26) and sequential-coring (N = 67) methods (A), and fine-root productivity estimates (Mg ha-1 year-1) (mean ± SE) of tropical (N = 52), temperate (N = 44) and boreal (N = 79) forests. Data represent all observations used in the study. Different letters represent significant differences (p < 0.05)…………………………………………………………………………………………………..57 Figure 2.5 Relationship between fine-root productivity estimates (Mg ha-1 year-1) from sequential-coring and ingrowth-core (A, N = 66), ingrowth-core and (mini) rhizotrons (B, N = 25), and sequential-coring and (mini) rhizotrons (C, N = 11) compared at the same sites. Values of fine-root production are log- transformed (base 10). Broken line is 1:1 relationship between the methods…………………………….58 Figure 3.1 Fine root biomass estimates (Mg ha-1) to 30 cm soil depth in old-growth forest (black bars) and 54-year-old logged-over forest (grey bars). Different letters represent significant differences (p < 0.05, N = 36) among soil depth……………………………………………………………………………69  Figure 3.2 Vertical distribution of fine-root morphological traits (A-G) to 30 cm soil depth. Data are mean ± S.E for old-growth forest (black bars) and 54-year-old logged-over forest (grey bars). Different  xix letters represent significant differences (p < 0.05) among soil depth. SRA, specific root area; SRL, specific root length; RTD, root tissue density…………………………………………………………….71 Figure 4.1 Ingrowth core made of nylon mesh inserted into the soil (left A), and extracted core containing roots (right B)……………………………………………………………………………………………..83 Figure 4.2 Monthly environmental conditions recorded in the old-growth forest (black circles) and the 54-year-old logged-over forest (grey circles) during the study period: (A) Total rainfall; (B) Mean soil moisture content; (C) Mean Absorbed photosynthetically active radiation; (D) Mean air temperature; (E) Mean soil temperature, and (F) Mean relative humidity. Rainfall data were collected automatically from a weather station, ca. 18 km from the study forests………………………………………………………...93 Figure 4.3 Temporal changes in soil chemistry variables at 0-30 cm soil depth at different sampling times. Data are means ± SE for old-growth forest (black bars) and 54-year old logged-over forest (grey bars)……………………………………………………………………………………………………….94 Figure 4.4 Fine (A, B) and small (C, D) root biomass and necromass (mean ± SE) to 30 cm soil depth in old-growth forest and logged-over forest. Given are values of 25 soil cores per forest on each sampling date………………………………………………………………………………………………………...95 Figure 4.5 Root productivity (Mg ha-1 yr-1) and turnover rates (yr-1) to 30 cm soil depth for fine roots (A, C) and small roots (B, D) in old-growth (black bars) forest and 54-year old logged-over (grey bars) forest. Values are mean ± SE calculated from the approaches used to estimate root productivity and turnover rates. NS, Not significant (p > 0.05)……………………………………………………………..99 Figure 5.1 A tree with buttress roots (left photo, A), measuring tree diameter above buttress (right photo, B)…………………………………………………………………………………………………………111 Figure 5.2 A climber on a tree to harvest branches (left photo, A), and scanning leaves to determine traits (right photo, B)…………………………………………………………………………………………..117 Figure 5.3 Relationships between taxonomic variables (tree species richness, effective number of species and tree species diversity) and aboveground biomass (AGB). Relationships are shown for the old-growth forest (top panel, A-C) and the 54-year-old logged-over forest (down panel, D-F). Regression lines are  xx included for significant relationships. AGB values were estimated with the Henry equation (excluding height and wood density………………………………………………………………………………....124  Figure 5.4 Relationships between taxonomic variables (tree species richness, effective number of species and tree species diversity) and coarse wood productivity (CWP). Relationships are shown for the old-growth forest (top panel, A-C) and the 54-year-old forest (down panel, D-F). Regression lines are included for significant relationships. CWP values were estimated with the Henry equation (including height and wood density)……………………………………………………………….………………..125  Figure 5.5 Relationships between structural variables (tree diameter, basal area and tree density) and aboveground biomass (AGB). Relationships are shown for the old-growth forest (top panel, A-C) and the 54-year-old logged-over forest (down panel, D-F). Regression lines are included for significant relationships. AGB values were estimated with the Henry equation (excluding height and wood density)…………………………………………………………………………………………………...126 Figure 5.6. Relationships between structural variables (tree diameter, basal area and tree density) and coarse wood productivity (CWP). Relationships are shown for the old-growth forest (top panel, A-C) and the 54-year-old logged-over forest (down panel, D-F). Regression lines are included for significant relationships. CWP values were estimated with the Henry equation (including height and wood density)…………………………………………………………………………………………………...127  Figure 5.7. Bivariate relationships between leaf traits and tree biomass and productivity for old-growth forest (A) and 54-year-old logged-over forest (B-C)…………………….………………………………129 Figure 5.8. The relative importance of taxonomic and structural variables in explaining aboveground biomass (AGB) and coarse wood productivity (CWP) in the old-growth forest (black bars) and the 54-year-old logged-over forest (grey bars). Each bar represents the Pearson product-moment correlation coefficient between a variable and AGB or CWP. AGB values were estimated with the Henry equation (excluding height and wood density) and CWP included height and wood density……………………..132  Figure 6.1 Liana (left, A), measuring diameter of coarse wood debris (right B).……….………………149  xxi Figure 6.2 Litterfall trap (left photo, A), litter trapped in trees were not collected in the litter traps (right photo, B)………………………………………………………………………………………………….151 Figure 6.3 Monthly distribution of litterfall (A), and annual NPP allocation into canopy, wood and fine root components (B) in the old-growth forest and the 54-year-old logged-over forest………………….162 Figure A.1 Curves showing the fitting of observed and predicted fine (A-B) and small (C-D) root mass collected from the ingrowth-core and sequential-core methods. Prediction for the observed root mass from 50-100 min. were based on roots collected between 0-40 min. Closed circles are observed root mass and open circles are predicted values. General logarithmic equation of the curve is shown in Chapter 4…...256  Figure A.2 Root mass distribution for ingrowth-core and sequential-coring methods over five sampling times………………………………………………………………………………..…………………….257 Figure A.3 Linear and polynomial relationships between diameter (cm) and height (m) (A), and (B) correlation between tree heights predicted with site-specific equation and the Feldpausch equation for West Africa………………………………………………………………………………………………258 Figure A.4 Relationships between tree species richness and aboveground biomass (AGB) and coarse wood productivity (CWP) using Henry equations and Chave equation…………..……………………..259 Figure A.5 Relationships between standard deviation of wood density (functional diversity) and aboveground biomass (AGB) and coarse wood productivity (CWP) for the old-growth forest (A-B) and the 54-year-old logged-over forest (C-D)………………………………………………………………..260 Figure A.6 Relationships between community-weighted mean of wood density (functional dominance) and aboveground biomass (AGB) and coarse wood productivity (CWP) for the old-growth forest (A-B) and the 54-year-old logged-over forest (C-D)…………………………………………………………...261 Figure A.7 Tree diameter (cm) and height (m) distribution (≥ 10 cm DBH) in the old-growth forest and the 54-year-old logged-over forest………………………………………………………………………262 Figure A.8 Monthly distribution of litter components in the old-growth forest (black circles) and 54-year-old logged-over forest (grey circles) plots: (A) Leaf fall; (B) Twig; (C) Fruit; (D) Flower; (E) Seed; and (F) Unidentified………………………………………………………………………………………….263   xxii Figure A.9 Soil organic matter (MgC ha-1) to 1-m soil depth in old-growth forest (black circles) and 54-year-old logged-over forest (grey circles)…………………………………………………………...…..264                          xxiii List of symbols and abbreviations or other a                              Constant defining shape of curve b          Intercept  t          Time Δ           Change in a factor between consecutive dates AB           Alstonia boonei AZ                           Albizia zygia AGB AIC ANOVA Aboveground biomass Akaike Information Criterion Analysis of variance  ANPP APAR B Bh Bl Bm BA Aboveground Net Primary Productivity  Absorbed Photosynthetic Active Radiation Biomass Maximum root biomass Lowest root biomass Mean root biomass Basal area BGB BNPP C CE CM CV CZ Belowground biomass Belowground Net Primary Productivity  Carbon Coefficients Celtis mildbraedii Coefficient of variation  Celtis zenkeri  xxiv CRB CRP Coarse root biomass Coarse root productivity CWD CWM CWP D DBH EM ES FAO FRB FRP Coarse Woody Debris  Community-weighted mean  Coarse wood productivity Tree diameter Diameter at Breast Height  Electromagnetic wave  Effective number of species Food and Agriculture Organization Fine root biomass Fine root productivity FSD Forest Services Division GPR Gt C H LAI LDMC Leaf K Leaf N Leaf P N NP NS NPLD Ground-Penetrating Radar Gigatons carbon  Tree height Leaf Area Index  Leaf dry matter content Leaf potassium Leaf nitrogen concentration Leaf phosphorus concentration Necromass Nesogordonia papaverifera  Not significant  Non-pioneer light demander  xxv NPP P PM PMa PVC Net Primary Productivity Productivity  Pterygota macrocarpa Petersianthus macrocarpus Polyvinyl chloride  REDD RH RPa RMSE RSR RTD SD SE SO SR SOC Soil Ca Soil Mg Soil K Soil N Soil Na SLA SRA SRL T Reduced Emissions from Deforestation and Forest Degradation Relative humidity  Annual root productivity  Root mean squared error Root-shoot ratio Root tissue density Tree species diversity Standard error Sterculia oblonga Tree species richness Soil organic carbon Soil calcium  Soil magnesium  Soil potassium Soil nitrogen Soil sodium Specific Leaf Area Specific Root Area Specific Root Length  Annual root turnover rate  xxvi TD TS TSu Tree density Triplochiton scleroxylon Terminalia superba UNFCCC VIF WD United Nations Framework Convention on Climate Change Variance Inflation Factor Wood density                      xxvii Acknowledgements My first and foremost gratitude goes to God Almighty for giving me life, good health and strength. I am grateful to Dr. Cindy E. Prescott, who is not only an excellent supervisor, but a mentor too. Words cannot express how I appreciate your dedication, guidance and support during this study. I would also want to thank my co-supervisor Dr. Robert D. Guy for his support. I thank my supervisory committee, Dr. Peter Marshall, Dr. David I. Forrester and Dr. Stephen Adu-Bredu for their insights into different aspects of this research. I would like to thank Dr. David I. Forrester for taking interest in this research, and the valuable discussions that shaped my ideas. My appreciation also goes to Dr. Yadvinder Malhi and Dr. Stephen Adu-Bredu who gave me the opportunity to work on a project, which served as the basis for this research.   This doctoral research was funded in part by the Natural Environment Research Council (NERC) of UK, UBC Four Year Fellowship, Peter Rennie Memorial Award, Mary and David Macaree Fellowship (Faculty of Forestry, UBC) and the International Foundation for Science (IFS). Writing was supported by a NSERC Discovery grant to Dr. Cindy E. Prescott. This research would not have been possible without the financial support from these organizations.   I am grateful to the many field and laboratory assistants, especially Emmanuel Amponsah Manu, Jonathan Dabo, Afriyie Agyekum Kwabena, Adu Opoku-Ameyaw, Forzia Ibrahim, Michael Adu Sasu, Adu-Gyamfi Asamoah, Rita Oppong, Gideon Yawson, Joyce Mensah, Jacqueline Twintoh, Getrude Gyamfi, Lydia Bonsu, William Hagan Brown, Martin Larthbridge, Felix King Mensah, Kwaku Sarpong, Ntim Kofi, Bismark Aboagye and James Boadu. I also thank my colleagues at the Belowground Ecosystem Group and the Tree Physiology Laboratory for their encouragement. I appreciate Dr. Victor K. Agyeman, the Director-General of the Council for Scientific and Industrial Research (CSIR), Ghana for his mentorship and counsel. I also thank Dr. Daniel A. Ofori, the Director of CSIR-Forestry Research Institute of Ghana (FORIG) for helping to ignite my passion for research. I also want thank Dr. Joseph R.  xxviii Cobbinah, Dr. Ernest G. Foli, Dr. Paul P. Bosu, Mrs. Theresa Peprah, Dr. Lucy Amissah, Dr. Gloria D. Djagbletey and Dr. Mary Apetorgbor for their counsel and support in diverse ways.   My appreciation also goes to my pastors, David Quansah and Emmanuel Ayedzi, and the members of Mid-Country Chapel and Liberty House of Worship for being there for me always. Special thanks go to my parents, who supported me morally, spiritually and financially throughout my years of education. I am grateful to my siblings for their encouragement and support. I would like to express my outmost gratitude to my dear wife for her love, patience and understanding when I hardly stayed home.                    xxix Dedication  To: My parents (your labour is rewarded) My dear wife, Kesewaa and our girls, Ewura Esi, Maame Adobea and Nana Ama David Addo-Danso (your memory lives on) My-in-laws (your memory lives on)  1 Chapter 1: Introduction  1.1 Tropical forests and their relevance  Tropical forests cover about 10% of the Earth’s land surface, an area of ca. 1.2 billion ha, of which 49% is in the Neotropics, 34% is in tropical Africa and 17% is in tropical Asia and Oceania (FAO, 2001). In recent times, there is increased interest in tropical forests because of their critical roles in local economies and global environmental policies. Tropical forests (including those in Africa) provide valuable goods and services to support the socio-economic development of countries and local communities (Blay et al. 2008; Vira et al. 2015). Globally, about 1.2 to 1.5 billion people directly depend on tropical forests for timber, food, medicines and other products (Vira et al. 2015). Furthermore, tropical forests provide environmental services such as regulation of the water cycle, maintenance of soil fertility, flood regulation and recreation (Canadell and Raupach, 2008). These forests also provide sustenance and the cultural basis for indigenous forest-dwelling groups, including the Yanomami of Amazonia, the Baka of central Africa and the Dayak of Borneo, Malaysia (Lewis, 2006; Chazdon, 2014).   At the global scale, tropical forests are major contributors to biodiversity and terrestrial ecosystem productivity. They represent the largest terrestrial reservoir of biological diversity from gene to habitat levels, containing at least 50% of the Earth’s biodiversity (Heywood, 1995). Tropical forests are highly productive, accounting for 60-70% of the total biomass and productivity of global forest ecosystems (Pan et al. 2013; Ma et al. 2015). Tropical forests are also disproportionately important in the world carbon (C) budget, representing an estimated 55% of the global C pool in forests (Pan et al. 2011; Saatchi et al. 2011). Saatchi et al. (2011) estimated a total tropical forest C stock of 247 Gigatons carbon (Gt C), with 193 Gt C stored in the aboveground components and 54 Gt C stored belowground in roots. Tropical forests also regulate the transfer of this stored carbon into the atmosphere as carbon dioxide (CO2) (van der Werf et al. 2009; Pearson et al. 2017). Due to these important roles, small changes within the tropical forest biome could have major impacts at local and global levels.    2 1.1.1 Threats to tropical forests Forest degradation and deforestation are major threats to tropical forest ecosystems. Globally, deforestation (i.e. conversion of forests to another land use) led to a net forest loss of ca. 129 million ha from 1990 to 2015, representing an annual net loss rate of 0.13% (FAO, 2016). An estimated nearly 100-300 million ha (30-40%) of tropical forests are already degraded (Blaser et al. 2011), and only 24% of tropical forest has not been degraded or converted to other land uses (Mercer, 2015). Human activities, including agricultural expansion, logging, woodfuel collection and hunting are among the main drivers of deforestation and degradation in tropical forests (Laurance, 2000; Hosonuma et al. 2012; Rudel, 2013; Malhi et al. 2014; Mercer, 2015; Cusack et al. 2016).   The rapid loss and degradation of tropical forests have deleterious consequences for forest functioning and the global climate. Deforestation is the second largest contributor of CO2 emissions to the atmosphere after fossil fuel burning (van der Werf et al. 2009). Across 74 tropical and subtropical countries, Pearson et al. (2017) estimated an annual emission of 2.1 Gt C between 2005 and 2010, of which 53% was derived from logging, 30% from woodfuel harvest and 17% from fire. A study by Vitousek et al. (1986) on human appropriation of the products of photosynthesis estimated that deforestation and forest degradation had reduced annual productivity by 28%. Furthermore, the conversion of degraded forests to other land uses drives fragmentation of the remaining forests, which leads to additional C emissions, species loss, and changes in species composition (Lewis et al. 2015; Barlow et al. 2016; Brinck et al. 2017).   1.1.2 Logging  Logging, mostly through selective felling of large high-valued trees above a minimum size, is an integral component of forest management in the tropics (Blaser et al. 2011; Chazdon, 2014). Studies in tropical Asia, Africa and the Americas show that the rate of logging has expanded rapidly (Kotey et al. 1998; Asner et al. 2005; 2009; Laurance et al. 2006). Globally, about 3.9 million km2 (ca. 20%) of tropical  3 forest was allocated to selective logging between 2000 and 2005 (Asner et al. 2009). In Brazilian Amazonia, logged areas in the top five timber-producing states were estimated to range from 12,075 to 19,823 km2 between 1999 and 2002 (Asner et al. 2005). Laurance et al. (2006) reported that nearly half of forest resources in Gabon are currently under timber leases, and this figure could increase to over 75% of the remaining forest within a few years.   Logging intensity and practices vary widely across different tropical regions and countries (Hawthorne et al. 2012; Zimmerman and Kormos, 2012; Cho et al. 2013; Chazdon, 2014). Logging intensity (number of trees or tree volume removed) is usually higher in Southeast Asia than in Africa and the Americas (Chazdon, 2014; Malhi et al. 2014). For instance, 8-15 trees per ha are typically removed during each logging operation in Sabah, Malaysia (Chazdon, 2014), whereas typically 1-8 trees are removed per ha in many parts of Africa and South America (Ibrahima et al. 2010; Asamoah et al. 2011; Zimmerman and Kormos, 2012; Lewis et al. 2015). Different logging practices have been adopted ranging from conventional practices involving no planning to reduced-impact logging where harvesting operations are planned based on sound scientific and engineering principles (Blaser et al. 2011; Medjibe and Putz, 2012). Ultimately, the intensity of logging and the operational practices determine the impacts of logging on the residual stand structure and soil (Picard et al. 2012; Chazdon, 2014), and also affect the rate and duration of recovery of forest structure and functioning after logging (Edwards et al. 2010; Hawthorne et al. 2012; Rutishauser et al. 2015; Piponiot et al. 2016).   1.1.2.1 Impacts of logging  Logging is a pervasive form of disturbance, which negatively affects forest structure and functioning. In fact, logging is considered as the precursor to deforestation (Laurance, 2000; Chazdon, 2014), and accounts for more than 50% of tropical forest degradation (Hosonuma et al. 2012; Mercer, 2015). Logging opens up the forest through the construction of roads and loading yards, which provides access to people who engage in other activities, including shifting cultivation that lead to fragmentation and  4 eventual conversion to another type of land use (Laurance, 2000; Laurance et al. 2006; Kleinschroth and Healey, 2017). In addition, logging operations in most countries are wasteful (Laurance, 2000), and this makes forests more susceptible to fire by accumulating debris, which dry and become combustible.   Previous studies show that logging causes changes to the residual forest structure and shifts in species composition of flora and fauna (Ewers et al. 2015; Osazuwa-Peters et al. 2015a; Martin et al. 2015; Vaglio Laurin et al. 2016a). Typically, 40-50% of the canopy cover is removed, and about 40-70% of the remaining trees are damaged during logging operations (Chazdon, 2014). Tree height and stem density were reduced 10 years after logging in tropical forests in Uganda and French Guiana, respectively (Osazuwa-Peters et al. 2015a; Rutishauser et al. 2016). In a global meta-analysis of 19 published studies, Clark and Covey (2012) concluded that logging resulted in a significant reduction in tree species richness. An experimental study in Borneo (Ewers et al. 2015) showed that logging decreased the abundance of termites, beetles and earthworms, but increased the abundance of small mammals, amphibians and insectivorous birds.   Both above- and belowground biomass are lower in logged forests relative to old-growth (unlogged) forests (Hertel et al. 2007; Berenguer et al. 2014; Gatti et al. 2015). Berenguer et al. (2014) reported higher aboveground biomass (AGB) in old-growth forests than in logged forests in the eastern Amazon. In a global study that compared fine-root dynamics in primary tropical forests and logged forests, Hertel et al. (2007) reported significantly higher fine-root biomass in old-growth forests than logged forests. Furthermore, fine-root specific root area (SRA) was higher in logged forests than in old-growth forests in Indonesia, which was attributed to the lower tissue densities of light-demanding species in the logged forests (Leuschner et al. 2009). Aboveground productivity was significantly higher in unlogged forests than in logged forests in Borneo and Nepal (Saner et al. 2012; Gautam and Mandal, 2016). It is noteworthy that almost all the aforementioned studies (except Osazuwa-Peters et al. 2015a: 45-years after logging) were carried out within the first two decades after logging.   5 1.1.2.2 Emerging ideas on logged-over forests Logged forests are usually considered to be degraded, but new ideas are emerging on the relevance of these forests to sustain and recover critical ecosystem services and functions, including biodiversity, biomass and C storage as well as productivity (Gibson et al. 2011; Putz et al. 2012; Edwards et al. 2010; 2014; Lewis et al. 2015; Putz and Romero, 2015; Chaudhary et al. 2016). Logged forests have been reported to retain a large proportion of the biodiversity of old-growth forests or unlogged forests in several tropical regions (Berry et al. 2010; Gibson et al. 2011; Bicknell et al. 2014; 2015; Ding et al. 2017; Roopsind et al. 2017). In a study that assessed birds, bats and large mammal assemblages over a 5-year period in central Guyana, Bicknell et al. (2015) reported that structure and composition were similar before and after logging, and between logged and unlogged forests. Floral species richness was higher in a logged forest than in an old-growth forest in dipterocarp forests of Sabah, Borneo, whilst faunal species richness was lower in the logged forest (Berry et al. 2010). However, in most cases the difference in faunal species richness between habitats was no greater than 10% (Berry et al. 2010). On Hainan Island in south China, density and diversity of woody species was significantly higher in logged forests 35 to 40 years after logging than in old-growth forests (Ding et al. 2017).   Logged forests can recover above- and belowground biomass and C dynamics after logging. Sixteen years after logging, forests subjected to reduced-impact logging in Brazil had recovered 100% of their original aboveground biomass (West et al. 2014). In a global study that reviewed 109 studies on the conservation values of selectively logged forests, Putz et al. (2012) concluded that selectively logged forests retained about 76% of the carbon in the aboveground biomass. Fine-root biomass had recovered 7 years after logging in a rainforest in Cameroon (Ibrahima et al. 2010). Some studies have shown that selectively logged forests can be as productive as old-growth forests. Litterfall production was similar in old-growth forests and logged forests in Borneo, Malaysia (Saner et al. 2012). In southwestern Rwanda, annual total productivity was comparable in mixed montane forests and forests that had been logged and disturbed by other anthropogenic activities (Nyirambangutse et al. 2017). Logged forests are also important for  6 sustainable timber production (Sasaki et al. 2012), and for meeting the socio-economic needs of forest-fringe communities (Appiah et al. 2009).  Throughout the tropics, old-growth forests are rapidly disappearing, whilst areas under timber production are increasing (Asner et al. 2009; FAO, 2016). Selectively logged forests have become one of the major land uses in the tropics (Asner et al. 2009; Chaudhary et al. 2016). Logged forests will become increasingly important to the socio-economic development of countries, as well as efforts towards biodiversity conservation and climate-change mitigation such as the Reduced Emissions from Deforestation and Forest Degradation (REDD+) mechanism (Sasaki et al. 2012; Edwards et al. 2014). There is already concerted effort to monitor and generate data from networks of logged forests (e.g., Tropical managed Forests Observatory and FORMNET-B database) to enhance understanding of the functioning of these forests, and their relevance in providing ecosystem services and goods (Cho et al. 2013; Sist et al. 2015). Logged tropical forests are both extensive and also diverse (Gibson et al. 2011; Edwards et al. 2014), and many more studies are needed from different tropical regions to clarify the functions and services provided by these forests over time and space. This is particularly so for Africa, where logging intensity is generally low (Ibrahim et al. 2010; Asamoah et al. 2011), and the recovery of forests from disturbances is faster because of extensive history of disturbance in the region (Cole et al. 2014).   1.2 Biomass and net primary productivity  Key ecosystem properties, such as biomass and net primary productivity (NPP) affect the functioning of tropical forests (Barnes et al. 2016; Liang et al. 2016). Moreover, the allocation of NPP (i.e. fraction of NPP used by a plant part) affects plant growth and uptake of soil resources in forest ecosystems (Litton et al. 2007; Malhi et al. 2011). Biomass is the amount of live organic material present (per unit area) at any point in time. Often, biomass is inferred from the standing stock of single plant parts (e.g., stem or roots) or above- and belowground components of forests (Vogt et al. 1998; Lewis et al. 2013; Gautam and  7 Mandal, 2016). The biomass of a forest represents the long-term balance between growth and mortality of different plant parts and vegetation components in the ecosystem (Pan et al. 2013). Net primary productivity is the rate of production of new biomass (per unit area), after accounting for autotrophic respiration (Clark et al. 2001a). NPP also includes losses from plants through processes such as volatilization, herbivory and exudation, as well as export to mycorrhizal fungi and their symbionts (Clark et al. 2001a; Knapp et al. 2014). Forest NPP cannot be measured directly (Clark et al. 2001a), and therefore in the field it is determined as the change in biomass of live components within a specific time interval, and losses during the same interval (Clark et al. 2001a). Most studies of tropical forest NPP have quantified only aboveground components particularly canopy and stem (Clark et al. 2001b; Malhi et al. 2009), because quantifying the belowground components is challenging (Vogt et al. 1998). In this thesis greater attention is devoted to roots and belowground processes (Chapters 2, 3 and 4) to provide a better understanding of their roles in the functioning of tropical forests.   1.2.1 Factors that affect biomass and net primary productivity Biomass and NPP vary spatially and temporally within the same forest ecosystems (Day et al. 2013; Jucker et al. 2016), as well as between forest types (Lewis et al. 2013; Chisholm et al. 2013). There is much interest in the factors that control biomass and NPP in tropical forest ecosystems (e.g., Cleveland et al. 2011; Lewis et al. 2013; Fyllas et al. 2017). The factors that control biomass production in most cases also influence forest NPP (Pan et al. 2013; Knapp et al. 2014), although biomass and productivity are not always related (Keeling and Phillips, 2007; Johnson et al. 2016). Different factor (s) affect either biomass or NPP or both depending on other local conditions operating at specific sites (Lasky et al. 2014; Jucker et al. 2016). Biomass and NPP are related to biotic factors such as biodiversity, traits and forest structure, and abiotic factors such as climate and soil conditions. These factors may act independently or simultaneously to influence forest biomass and NPP (Cleveland et al. 2011; Lewis et al. 2013; Sande, 2016). The biotic and abiotic factors have both direct and indirect effects on biomass production (Poorter et al. 2015; Sande, 2016); for example abiotic factors can indirectly influence biomass or NPP through  8 their influence on biotic factors such as forest structure and biodiversity (Baraloto et al. 2011; Lewis et al. 2013; Durán et al. 2015; Fyllas et al. 2017).   1.2.1.1 Biotic factors  Biotic factors such as biodiversity, plant functional traits and forest structure affect the biomass accumulation and NPP of forest ecosystems (Cleveland et al. 2011; Durán et al. 2015; Poorter et al. 2015; Fyllas et al. 2017). Plant species vary within and among tropical forests (Phillips et al. 1994; Sullivan et al. 2017). This variability in species is commonly measured as richness (i.e. the number of species per area) or diversity (i.e. species richness and abundance). Other measures, including the effective number of species are also used (e.g., Jost, 2006; Poorter et al. 2015). Theoretical and experimental studies that manipulate tree species numbers suggest that ecosystems that have high species richness or diversity should contain more biomass and/or have higher productivity (Cardinale et al. 2007; Scherer-Lorenzen et al. 2007; Hector et al. 2011; Forrester and Bauhus, 2016). This is supported by global observational studies across the major forest biomes that reported positive relationships between species richness and forest productivity (Gillman et al. 2015; Liang et al. 2016). However, evidence from tropical forests is less consistent (Chisholm et al. 2013; Day et al. 2013), and is usually scale- and context dependent (Chisholm et al. 2013; Poorter et al. 2015; Sullivan et al. 2017). Typically, at large spatial scales (> 1 ha) or at the continental level, a weak relationship or no relationship has been observed between biodiversity or taxonomic variables and components of biomass or NPP (Chisholm et al. 2013; Poorter et al. 2015; Sullivan et al. 2017). After analyzing 25 plots across both tropical and temperate forests, Chisholm et al. (2013) reported that at large spatial scales (e.g., 1 ha) relationships between species richness and coarse wood productivity (CWP) were weak and often negative. Likewise, Sullivan et al. (2017) found that relationships between species diversity and carbon stocks within continents were weak for Asia and absent for Africa and Amazonia. Usually at such large scales the influences of climate and environmental heterogeneity on biomass and NPP dominate (Cleveland et al. 2011; Chu et al. 2016; Sullivan et al. 2017). In contrast, at local or fine spatial scales (< 1 ha) species richness or diversity are positively related  9 to forest biomass and NPP. Positive relationships between species richness and aboveground biomass or wood productivity have consistently been reported in tropical Africa (Day et al. 2013; Jucker et al. 2016; Vaglio Laurin et al. 2016a) and elsewhere (Balvanera and Aquirre, 2006; Jean-Ruiz and Potvin, 2011; Poorter et al. 2015; Sullivan et al. 2017). However, species richness was not related to biomass or productivity in a tropical forest in Guyana (Sande, 2016). Mechanisms such as niche complementarity and selection effect have been proposed to explain the influence of biodiversity or taxonomic variables on biomass and productivity (Fridley, 2001). Niche complementarity occurs when the interspecific interactions between plant species enhance the availability, capture and efficient use of resources by neighbours (Loreau et al. 2001; Bauhus and Forrester, 2016) or when different species in a species-rich plot or community show different niche partitioning, which enables a more complete use of available resources (Loreau et al. 2001). Selection effect occurs when a species-rich or highly diverse plot contains a dominant species that drives biomass or productivity (Fridley, 2001).   Plant functional traits correlate with key ecosystem functions and processes (Walker et al. 2014; Guerrero-Ramírez et al. 2016). Traits are characteristics that reflect the life history of individuals and enhance their performance in a particular ecosystem (Violle et al. 2007; Reich, 2014). Plant functional traits relate to tropical forests biomass and productivity (Cleveland et al. 2011; Finegan et al. 2015; Sande, 2016; Fyllas et al. 2017). In a pan-tropical study, which analyzed 113 moist and wet forests, Cleveland et al. (2011) reported a positive relationship between leaf P concentration and aboveground productivity in lowland forests. Stem productivity related positively with community-weighted mean specific leaf area, but negatively with leaf P in a wet tropical forest in central Guyana (Sande, 2016). Wood traits such as wood density also affect tree growth and forest aboveground biomass (AGB) (Lida et al. 2012; Slik et al. 2013; Finegan et al. 2015; Visser et al. 2016). A strong relationship is detected between plant functional traits and biomass or NPP because of their direct control on critical processes such as competition, photosynthesis, respiration and decomposition (Makita et al. 2012; Walker et al. 2014; Guerrero-Ramírez et al. 2016; Kunstler et al. 2016). For instance, fine-root respiration rates  10 declined with increasing specific root length and decreasing root tissue density in 13 tropical tree species in Peninsular Malaysia (Makita et al. 2012).   The structure of tropical forests is dynamic and complex (Baraloto et al. 2011; Pan et al. 2013). Stand structural variables, including leaf area, tree diameter, height, stand density and basal area affect the spatial variation in biomass and productivity (Hertel et al. 2009a; Lewis et al. 2013; Poorter et al. 2015; Jucker et al. 2016). Strong relationships between stand variables and biomass or productivity have been reported in many tropical forests. Stand basal area and tree diameter were the strongest predictors of spatial variation of AGB across forests at 74 sites in Peru and French Guiana (Baraloto et al. 2011). Similarly, tree diameter and basal area were related to CWP in a moist forest in Sierra Leone (Jucker et al. 2016). Indeed, stand density interactions with biotic factors such as species richness can indirectly affect forest biomass and productivity (Chisholm et al. 2013). Forest structure also influences root biomass and productivity in forest ecosystems. Leaf area, tree diameter and basal area related strongly and positively to fine-root biomass and fine-root productivity in moist forests in Indonesia (Harteveld et al. 2007; Hertel et al. 2009a).   1.2.1.2 Abiotic factors Climate and edaphic factors influence biomass and productivity in tropical forests (Hertel and Leuschner, 2011; Becknell et al. 2012; Quesada et al. 2012; Lewis et al. 2013; Taylor et al. 2017). Climate variables, including rainfall (or precipitation), temperature and light (solar radiation), have direct and indirect influence on forest biomass and NPP (Schuur, 2003; Cleveland et al. 2011; Dong et al. 2012; Michaletz et al. 2014; Chu et al. 2016; Fyllas et al. 2017). Aboveground biomass was positively related with rainfall during the driest months of the year, but negatively with temperature in forest plots across 12 African countries (Lewis et al. 2013). Mean annual temperature was the strongest predictor of aboveground productivity across all tropical forests (Cleveland et al. 2011), and globally, rainfall and temperature explained 44% of the variation in aboveground productivity across 150 tropical forest sites (Taylor et al.  11 2017). Stand-level tree growth rate was positively correlated with variation in solar radiation and negatively related to night-time temperatures in lowland forests in Panama, Malaysia and Thailand (Dong et al. 2012). Strong relationships have been reported between rainfall and temperature and fine-root biomass (e.g., Green et al. 2005; Jiménez et al. 2009) or fine-root productivity (e.g., Violita et al. 2016). Climate also affects biomass and NPP indirectly via its effects on forest structure and other environmental factors (Baraloto et al. 2011; Taylor et al. 2017). For instance, rainfall and temperature affect the distribution of forest tree species (Fauset et al. 2012; Amissah et al. 2014), which also determines the distribution of traits such as wood density that directly affect biomass production (Slik et al. 2013).   The positive effects of climate variables on biomass or NPP do not exist for all tropical sites, even within large geographic regions (Malhi et al. 2004; Baraloto et al. 2011; Cleveland et al. 2011; Hertel and Leuschner, 2011; Violita et al. 2016). For example, Baraloto et al. (2011) observed a weak direct relationship between mean annual precipitation (MAP) and AGB in the Amazon basin. No relationship was found between rainfall and CWP in 104 Neotropical forest plots (Malhi et al. 2004). Similarly, Hertel and Leuschner (2011) found no correlation between fine-root mass and temperature or precipitation in Paleo- and Neotropical forests. Fine-root productivity was not correlated with solar radiation in a rainforest in Jambi Province, Indonesia (Violita et al. 2016). Ultimately, the magnitude and direction of the effect of climate on tropical forests depends on the relationship with other factors, which also drive biomass and NPP (Dong et al. 2012; Quesada et al. 2012; Taylor et al. 2017).  Topography, soil texture and nutrient availability affect the spatial distribution of tropical forest biomass and productivity (Malhi et al. 2004; Paoli et al. 2008; Banin et al. 2014; Ledo et al. 2016; Quinto-Mosquera and Moreno, 2017). In particular, soil P is considered to have a major control on tropical forest biomass and NPP (Gower, 1987; Aragão et al. 2009; Cleveland et al. 2011; Quesada et al. 2012; Jucker et al. 2016). Nevertheless, the extent and direction of the effects of edaphic factors, especially soil nutrient availability varies among forest types (Paoli et al. 2008; Cleveland et al. 2011). It is typically expected  12 that forests growing on nutrient-rich soils will have higher biomass or productivity, but this is not always the case. For instance, the AGB estimates of forests on some nutrient-poor soils are similar or even higher than estimates from forests on fertile soils (e.g., Gourtlet-Fleury et al. 2011). The relationship between soil nutrients and forest biomass or productivity is complex. Positive (Quesada et al. 2012; Yuan and Chen, 2012a), negative (Lewis et al. 2013; Powers and Peréz-Aviles, 2013), and no relationship (Dewalt and Chave, 2004; Hertel and Leuschner, 2011) between soil nutrients and above- and belowground biomass or productivity has all been reported. These conflicting results reflect the wide variation in soil age and parent material in tropical forests (Porder and Hilley, 2010; Jiménez et al. 2014), and the influence of site-specific conditions that also affect both soil development and biomass production (Paoli et al. 2008; Quinto-Mosquera and Moreno, 2017).  1.2.1.3 Other factors (methodology)   The methodology used to quantify forest components affect biomass and productivity estimates (Nadelhoffer and Raich, 1992; Vogt et al. 1998; Cleveland et al. 2011; Šímová and Storch, 2017). Both direct and indirect methods are used to estimate above- and belowground biomass and NPP (Clark et al. 2001a; Vogt et al. 1998; Pan et al. 2013; Cleveland et al. 2015). For instance, aboveground biomass and productivity can be estimated using field-inventory, remote sensing-based (satellite-derived) and biogeochemical model-based methods (Pan et al. 2013; Cleveland et al. 2015), and all these methods have strengths and limitations (Clark et al. 2001a; Clark and Kellner, 2012; Cleveland et al. 2015). Root biomass and productivity have also been estimated with various methods including monolith, soil-core/sequential-coring, soil-pit, ingrowth-core and (mini) rhizotron methods (Vogt et al. 1998; Bledsoe et al. 1999; Levillain et al. 2011). There is no consensus in the literature on which method (s) is the most suitable (Nadelhoffer and Raich, 1992; Levillain et al. 2011; Girardin et al. 2013), and often the methods produce different estimates of biomass and productivity (e.g., Moser et al. 2010; Johnson et al. 2016).    13 A field-based approach involving the indirect use of allometric equations is the most widely used method to estimate forest aboveground biomass and productivity (e.g., Clark et al. 2001b; Lewis et al. 2013; Anderson-Teixeira et al. 2016). With this method, plot-level measurements are scaled to stand-level estimates using allometric equations, which incorporate tree diameter, height and wood density (Chave et al. 2005; Henry et al. 2010). Depending on the variables used (i.e. diameter and/or height or wood density) different estimates may be produced, even at the same sites (Henry et al. 2010; Ngomanda et al. 2014). Furthermore, site-specific and generalized allometric equations that require the same variables (e.g., diameter) also differ in the accuracy and precision of biomass estimates they produce (Ngomanda et al. 2014).  Root productivity is commonly estimated using ingrowth-coring, sequential-coring and (mini) rhizotron methods (Vogt et al. 1998; Finér et al. 2011a). These methods produce contradictory estimates of fine-root productivity for the same tropical forest sites (e.g., Metcalfe et al. 2007; Moser et al. 2010). It is therefore recommended to use multiple methods at the same sites to obtain a reasonable range of biomass and NPP estimates in tropical forests (Vogt et al. 1998; Cleveland et al. 2015).   1.3 Context: Ghanaian tropical forests  The forests in Ghana fall within the Guinean tropical forests region, which extends from Sierra Leone to Cameroon. Forests cover ca. 39% (9.2 million ha) of the 23.9 million ha total land area of Ghana (Affum-Baffoe, 2014), and lie within the high forest and savannah zones of the country. The tropical high forests have been categorized into seven subtypes (Figure 1.1) based on the distribution of rainfall and soil conditions (Hall and Swaine, 1981). The forest types are the Wet evergreen, Upland evergreen, Moist evergreen, Moist semi-deciduous Southeast and Northwest, Dry semi-deciduous Inner and fires zones, Southern marginal and outlier forests (Hall and Swaine, 1981). These forest types vary in structure and plant composition and diversity (Hackman, 2014; Vaglio Laurin et al. 2016a). There is a gradient of increasing rainfall and biodiversity from the Wet evergreen forests in the southwest to the Dry semi- 14 deciduous forests in the north and east of the country (Hall and Swaine, 1981; Vaglio Laurin et al. 2016b). Generally, the forestlands have been designated as on-reserve and off-reserve forest areas. On-reserve forests are those that were demarcated in the past and are managed by government for various economic and environmental purposes (Kotey et al. 1998). Off-reserve forests are those outside forest reserves, and consist of patches of old-growth forests, secondary forests, riparian forest strips, and other land uses, including farmlands, plantations and fallow lands (Kotey et al. 1998; Hansen et al. 2009).   Forests play critical roles in the socioeconomic development of Ghana, and are integral part of the livelihoods of most communities (Blay et al. 2008; Appiah et al. 2009; FIP, 2012). Timber and timber products are the fourth foreign-exchange earner, after gold, cocoa and oil exports (FIP, 2012). The forest sector directly employs 120,000 people, mostly in the log processing industry (FIP, 2012). The forests are also essential sources of timber and non-timber products for most local communities (Blay et al. 2008; Appiah et al. 2009). The forest sector is estimated to support the livelihood of more than 21 million people, particularly in rural communities (Blay et al. 2008). In addition Ghana’s forests are recognized as a critical part of the West African biodiversity hotspot (Myers et al. 2000), and contain 23 endemic floral and faunal species (Hackman, 2014).   Ghana’s forests are declining at a fast rate due to deforestation and degradation (Hansen et al. 2009; Affum-Baffoe, 2015). On average 45,931 ha of forests have been degraded annually since 1990 (Affum-Baffoe, 2015). Furthermore, forests were converted to other land uses at a rate of 0.3% (ca. 28, 000 ha per year) between 1990 and 2015 (FAO, 2016). The causes of forest degradation and deforestation are many, but the direct ones are agricultural expansion, logging, woodfuel and charcoal collection, infrastructure expansion, and mining (Appiah et al. 2009; Hansen et al. 2009; FIP, 2012; FAO, 2016).   15   Figure 1.1 Map of Ghana showing forest zones and forest types. Source: FIP (2012)      16 1.3.1 Logging in Ghana Timber production is central to the management of forest resources in Ghana (Kotey et al. 1998). About 24% of forestlands in the country are managed for timber production (Figure 1.2). Timber harvesting takes place in both reserve and off-reserve forestlands, and indeed of the 266 forest reserves, 216 have been selectively logged (Hawthorne and Abu-Juam, 1995; Bird et al. 2006). Logging is done by selectively harvesting 2-3 individuals (per ha) of high-valued species that reach diameter at breast height (DBH) > 50 cm (Asamoah et al. 2011; Adum et al. 2013). Logging is based on a polycyclic regime where a permit holder or concessionaire can re-enter a previously logged forest after 40 years (Hawthorne and Abu-Juam, 1995). Studies have reported widespread negative impacts of logging on the regeneration of economic timber species (Hawthorne et al. 2012; Duah-Gyamfi et al. 2014a), soil physical and chemical properties (Asamoah et al. 2011; Vaglio Laurin et al. 2016b), forest structure and aboveground biomass (Hawthorne et al. 2012; Djagbletey, 2014). Nevertheless, forest structure and functioning have been found to recover to pre-logging levels some years after logging (Swaine and Agyeman, 2008; Asase et al. 2012; 2014; Adum et al. 2013; Ofori-Boateng et al. 2013; Duah-Gyamfi et al. 2014b). In terms of abundance and composition of amphibian assemblages, logged and unlogged forests were indistinguishable 20 years after logging in the Suhuma, Krokosua Hills and Sui River Forest Reserves in western Ghana (Adum et al. 2013). Asase et al. (2012) reported significantly higher aboveground biomass in logged forests 29 years after logging than in old-growth forests in the Bia Conservation area, southwestern Ghana. Furthermore, soil physical and chemical properties were similar in old-growth and logged forests also in this area (Asase et al. 2014). This indicates that logged forests have the potential to sustain critical ecosystem services and functions, if they are managed well, and not subjected to further disturbances. More data are needed on the changes in both above- and belowground processes and functioning during post-logging recovery.    17  Figure 1.2 Designated forest uses in Ghana. Values are from Affum-Baffoe (2015).   1.4 Thesis overview, conceptual framework and research questions  More empirical studies are needed to clarify how forest structure, functional traits, biomass and net primary productivity and C allocation change during the decades after logging. This thesis aimed to quantify biomass and components of NPP, as well as their potential drivers in an old-growth forest and a nearby 54-year-old logged-over forest. The two forests are ca. 3 km apart, and influenced by similar parent material, but have distinct stand structural characteristics. To place the study in a context, I present a general framework for understanding the consequences of logging on certain factors, including forest structural attributes and functional species composition, resource availability and functional traits and how these may change during post-logging recovery relative to old-growth (unlogged) forests.   1.4.1 Forest structure and functional species composition Logging causes changes to forest structural attributes such as tree diameter and height, as well as stand basal area and tree density (Osazuwa-Peters et al. 2015a; Rutishauser et al. 2016; Vaglio Laurin et al. 2016a). Usually, logging reduces the density of large trees and increase the mortality of the remaining trees (Schulze and Zweede, 2006; Osazuwa-Peters et al. 2015b). The gaps created in the canopy also reduce the leaf area of forests (Hardwick et al. 2015), which have implications for forest productivity (Reich, 2012). Studies that have examined forest structural attributes have reported reduction in tree heights and stand basal area in logged forests relative to old-growth (unlogged) forests (Osazuwa-Peters ProductionProtectionBiodiversity conservationRecreation and socialNo designation 18 et al. 2015a; Rutishauser et al. 2016). Even, in the long-term (e.g., > 45 years) the effects of logging on the forest structure can still persist, with consequences for above- and belowground biomass in forest ecosystems (Hertel et al. 2007; Gatti et al. 2015; Osazuwa-Peters et al. 2015b; Yamada et al. 2016). In the Pasoh Forest Reserve in Malaysia, logging affected the size structure of dipterocarp populations through a reduction in the recruitment of seedlings 50 years after logging (Yamada et al. 2016). Although the density and basal area of small-stemmed saplings can recover few years after logging (Kariuki et al. 2006; Rutten et al. 2015), the turnover of large trees during post-logging recovery has greater influence on the structure and functioning of the forests (Schulze and Zweede, 2006).   Logging can also result in changes to the functional composition of species in forest ecosystems (Kariuki et al. 2006), but the extent of the change varies due to differences in logging intensity and local site conditions (Bonnell et al. 2011; Vaglio Laurin et al. 2016). Generally, it is expected that forests at different stages of post-logging recovery until the old-growth stage would be dominated by shade-intolerant pioneer species (Chadzon, 2014), and this has been supported by many empirical studies in different tropical regions (Kariuki et al. 2006; Hawthorne et al. 2012; Katovai et al. 2016). Thirty years after logging, pioneer tree species continued to dominate the species composition of a network of logged forests in southern Ghana (Hawthorne et al. 2012). In the Solomon Islands, a long-lived pioneer Campnosperma brevipetiola was the dominant species in the forests 50 years after logging (Katovai et al. 2016). Species differ substantially in their effects on ecosystem processes (Chapin, 2003); therefore any species functional group that dominates forest ecosystems will also have a greater influence on critical ecosystem functions and properties (Clark and Covey, 2012; Putz et al. 2012).   1.4.2 Resource availability  Logging creates gaps that change the environmental conditions and the availability of light, water and nutrients (Feldpausch et al. 2010; Chadzon, 2014; Hardwick et al. 2015). The changes in these conditions can facilitate the regeneration of species (Agyeman et al. 2016), but also affect processes like  19 mineralization, photosynthesis and respiration, which influence productivity and carbon storage (Grant, 2014). The changes in resource availability, particularly light can influence the C allocation patterns in forest ecosystems (Litton et al. 2007; Figueira et al. 2008). When the forest canopy is opened there is increased in the recruitment and growth of pioneers, which triggers increased competition for light capture, resulting in greater proportion of C to wood growth to maximizes the exposure of leaf area to favorable light conditions (Figueira et al. 2008). Moreover, the increase in the availability of belowground resources also results in reduced C allocation to fine roots relative to aboveground components (Litton et al. 2007).   1.4.3 Plant functional traits  The differences in species functional traits affect the survival and growth of plants in forest ecosystems (Reich, 2014). Pioneers and shade-tolerant species show contrasting leaf, wood and fine-root traits that relate closely to their resource uptake and regeneration strategies and growth pattern (Bauhus and Messier, 1998; Xiang et al. 2013; Reich, 2014). Typically, pioneer species show greater leaf acquisitive traits such as higher leaf N, P and SLA, but lower leaf toughness and dry matter content which are linked to greater photosynthetic and respiration rates compared to shade-tolerant species (Carreño-Rocabado et al. 2012; Chazdon, 2014). Wood traits that correlate with high tree growth rate, including increased hydraulic conductivity and lower wood density, are greater in pioneers than in shade-tolerant species (Chazdon, 2014). This indicates that in logged forests dominated by fast-growing pioneers there should a shift towards ‘fast’ acquisitive traits, whereas old-growth forests that contain high abundance of shade-tolerant species should have higher values of ‘slow’ conservative traits (Chazdon, 2014;). For instance, Leuschner et al. (2009) reported higher fine-root SRA in a selectively logged forest than an old-growth forest in Indonesia, which they attributed to the lower tissue densities of the fast-growing light-demanding trees species in the logged forests. Since functional traits correlate with many forest processes and functions (Harguindeguy et al. 2013; Walker et al. 2014; Knustler et al. 2016), the contrasting trait  20 dominance in pioneers and shade-tolerant species have implications for the growth and the functioning of forest ecosystems (Badgett et al. 2014; Finegan et al. 2015).   Based on the above framework, it was hypothesized that the logging legacy will persist, and will be reflected in different functional traits and forest structure, and reduced biomass and net primary productivity in the logged-over forest relative to the old-growth forest. The specific research questions addressed in each chapter are briefly provided below:   1.4.4 Chapter 2: Methods for estimating root biomass and productivity: A review and global analysis  The wide variation in biomass and productivity estimates may in part be an artefact of the methodology used to sample forest components. Currently, there is no consensus on which methods are most suitable to accurately estimate root biomass and productivity. In this chapter I conducted a literature survey on root biomass and productivity methods, and also compiled a global dataset to address the following questions: (1) What are the existing methods to assess root biomass and productivity (including new and lesser known ones)? (2) How do they compare in terms of cost, labour requirements, time efficiency, and accuracy? (3) How do fine- and coarse-root biomass and productivity estimates compare among different methods measured at the same sites?   1.4.5 Chapter 3: Root exploitation strategies differ in tropical old-growth forest and logged-over forests in Ghana Studies of post-logging recovery have rarely linked root biomass and root functional traits to the exploitation strategies of plant roots. However, knowledge of changes in exploitation strategies, and the associated trade-offs, is key to understanding biomass accumulation and productivity during post-logging recovery in tropical forests. In this chapter, the soil-core method and WinRhizo image analysis software were combined to quantify root biomass and morphological traits in an old-growth forest and a 54-year- 21 old logged-over forest to address these questions: (1) Do root biomass estimates and root morphological traits differ between an old-growth forest and a logged-over forest? (2) Do plant roots in an old-growth forest and a logged-over forest differ in their apparent strategies for exploiting soil resources?   1.4.6 Chapter 4: Patterns and controls on root dynamics in tropical forests in Ghana, West Africa Root dynamics vary spatially and temporally in tropical forests, but information on the controlling factors is scarce. In this chapter I quantified biomass, necromass, productivity and turnover rates of fine and small roots in an old-growth forest and a 54-year-old logged-over forest using ingrowth-core and sequential-coring methods. I also measured environmental (rainfall, air and soil temperatures, soil water content, relative humidity and absorbed photosynthetically active radiation) and soil chemistry variables (N, available P, K, Na, Mg, Ca, pH, base saturation), and correlated these with estimates of root biomass, necromass and mass (biomass plus necromass). The research examined these questions: (1) Are there seasonal variations in fine and small root biomass and necromass? If so, how are these values related to environmental and soil chemistry variables? (2) Are there differences in estimates of root productivity and turnover rates obtained by the ingrowth-core and sequential-coring methods? (3) Do rates of root productivity and turnover differ between the old-growth forest and logged-over forest?   1.4.7 Chapter 5: Aboveground wood biomass and productivity in tropical old-growth and logged-over forests: the importance of taxonomic variables, stand structural variables and traits Understanding the spatial patterns and drivers of aboveground stem biomass (AGB) and coarse wood productivity (CWP) in tropical forests is critical for ecosystem modeling and forest management. Leaf traits can be correlated with key forest processes and functions. In this chapter, I used regression analysis to examine the relationships between the taxonomic (tree species richness, effective number of species, tree species diversity) or structural variables (tree diameter, basal area, tree density) and AGB or CWP in an old-growth forest and a 54-year-old forest in Ghana. I also analyzed the bivariate relationships between leaf functional traits (leaf N, leaf P, leaf K, specific leaf area and leaf dry matter content) and tree species  22 biomass and productivity in 18 individuals ranging from pioneers to shade-tolerant species. I addressed the following questions: (1) Do taxonomic and structural variables relate to aboveground biomass and coarse wood productivity? If so, are the relationships consistent for the old-growth forest and logged-over forest? (2) What is the relative importance of the taxonomic and structural variables in explaining AGB and CWP? (3) Do leaf traits contribute to the interspecific variation in tree biomass and productivity in the old-growth forest and logged-over forest?   1.4.8 Chapter 6: Biomass and productivity are similar, but allocation patterns differ between tropical old-growth and logged-over forests in Ghana In the short-term, logging negatively affects forest biomass and productivity by reducing the density of large trees and increasing stem mortality. Nevertheless, forest structure and soil fertility recover during post-logging recovery, thus forest biomass and productivity could change over time. In this study I quantified components of biomass and net primary productivity in an old-growth forest and a 54-year-old forest to address the following questions: (1) Do above- and belowground biomass differ between the old-growth forest and logged-over forest?  (2) Are there any differences in the estimates of NPP, and its components in the two forests? (3) Is the allocation of NPP between canopy, wood and fine roots similar in the old-growth forest and logged-over forest?   1.4.9 Chapter 7: Synthesis and general conclusions  Here I present the main findings of the thesis, and the implications of the research for the management of logged-over forests. I also point out some limitations in the study, and possible future directions in relation to the key findings of the thesis.      23 Chapter 2: Methods for estimating root biomass and productivity: A review and global analysis  2.1 Synopsis  Currently, there is no consensus on which methods are most suitable to accurately study root biomass and productivity in forest and woodland ecosystems. In this synthesis, I compared existing methods for root biomass and productivity estimation based on relevant criteria that include cost, labor requirements, time efficiency, and accuracy and, also compared fine- and coarse-root biomass and productivity estimates from different methods measured at the same sites. Root excavation and soil-pit methods are commonly used to estimate coarse-root biomass, despite the high cost and labor required. Ground-Penetrating Radar is a very promising indirect approach to estimate coarse-root biomass, but may not be suitable for ecosystems with dense understory and soils with high organic matter and ion contents. The use of empirical models is accepted as a fast and cost-effective indirect approach to predict fine- and coarse-root biomass and productivity. Fine-root productivity is usually estimated with the (mini) rhizotron, sequential-coring and ingrowth-core methods. Coarse-root biomass estimates were not significantly different between soil-pit and soil-core methods. There was a significant positive relationship (R2 = 0.91, p < 0.0001) between fine-root biomass estimates obtained from soil-pit and soil-core methods. Mean fine-root productivity was 5.53 ± 1.00, 2.56 ± 0.41 and 1.46 ± 0.13 Mg ha-1 year-1 for tropical, temperate and boreal forests, respectively. Fine-root productivity estimates were lower in the ingrowth-core (2.04 ± 0.23 Mg ha-1 year-1) compared to the sequential-coring (3.70 ± 0.93 Mg ha-1 year-1) and (mini) rhizotrons (3.81 ± 0.46 Mg ha-1 year-1) methods. FRP estimates obtained from sequential cores and ingrowth cores were positively related (R2 = 0.29, p < 0.0001, N = 66). Based on the reviewed literature, multiple methods are recommended for yielding realistic estimates of fine- and coarse-root productivity.      1A version of this chapter has been published: Addo-Danso SD, Prescott CE, Smith AR. 2016. Methods for estimating root biomass and production in forest and woodland ecosystem carbon studies: A review. Forest Ecology and Management 359: 332-351.    24 2.2 Introduction  Fine and coarse roots are major contributors to the total biomass pools of forest and woodland ecosystems, and play critical roles in the cycling and allocation of carbon (C) and nutrients (Clark et al. 2001a; Brunner and Godbold, 2007; Malhi et al. 2011; Smyth et al. 2013; Raich et al. 2014). A significant fraction of C assimilated by plants through photosynthesis is transferred to roots and their symbionts (Litton et al. 2007; McCormack et al. 2015); this may even exceed the amount allocated to aboveground components (e.g., Moser et al. 2011). The carbon transferred belowground is estimated to account for 22-63% of the total gross primary productivity of forests (Litton et al. 2007). Despite these critical roles, roots have been understudied, and are poorly represented in many process-based ecosystem models, limiting the models ability to predict ecosystem responses to environmental changes and management practices (Smithwick et al. 2014; Warren et al. 2015). The uncertainty about root dynamics also hampers efforts to accurately estimate pool size for C accounting and climate mitigation measures such as the Reduced Emissions from Deforestation and Forest Degradation (REDD+) (Smyth et al. 2013; Yuen et al., 2013). This knowledge gap is partly attributable to methodological challenges in sampling roots to estimate biomass productivity and turnover (Vogt et al. 1998; Bledsoe et al. 1999; Makkonen and Helmisaari, 1999; Finer et al. 2011a,b).   Estimation of fine and coarse root biomass and productivity can be grouped into direct and indirect methods. Fine-root biomass and productivity have been estimated with direct methods that include soil-core/sequential-coring (Makkonen and Helmisaari, 1999; Lauenroth, 2000), monolith (Bledsoe et al. 1999; Makita et al. 2011), soil-pit (Millikin and Bledsoe, 1999; Park et al. 2007), ingrowth-core (Persson, 1979; Vogt et al. 1998) and (mini) rhizotrons (Taylor et al. 1990; Madji, 1996), and indirectly through the use of empirical models (Shinozaki et al. 1964a; Kurz et al. 1996). For coarse roots, direct methods include root excavation (Bledsoe et al. 1999; Niiyama et al. 2010), soil-pit/soil-pit ingrowth (Lawson et al. 1970; Kangas, 1992), wall or trench profiles and soil-core (van Noordwijk et al. 2000; Achat et al. 2008), while the indirect methods include, but are not limited to, size-mass allometric equations  25 (Whittaker et al. 1974; Kenzo et al. 2009; Brassard et al. 2011a), root-shoot or belowground-aboveground ratio (Keith et al. 2000; Levigne and Krasowski, 2007; Malhi et al. 2009), Ground-Penetrating Radar (GPR) (Butnor et al. 2001; Samuelson et al. 2015), and root biomass increment or difference (Steele et al. 1997; Kajimoto et al. 1999) as well as root radial increment (Zach et al. 2010; Moser et al. 2011).   There is no consensus in the literature on how best to estimate root biomass, productivity and turnover (Vogt et al. 1998; Milchunas, 2012; Finer et al. 2011a; Yuan and Chen, 2012b; Brunner et al. 2013). In a global study that compared fine root productivity (FRP) estimates across 186 stands, Finer et al. (2011a) reported similar FRP estimates between direct and indirect methods. In contrast, another global study reported significantly higher FRP estimates from indirect than direct methods (Yuan and Chen, 2012b). This uncertainty means that the choice of a method may be determined by considerations such as cost, labor availability, site constraints and individual preferences rather than accuracy and precision (Vogt et al. 1998; Levillain et al. 2011; Makita et al. 2011), with implications for modelling ecosystem C budget and allocation patterns. This lack of consensus therefore calls for critical evaluation of the assumptions, strengths and inherent limitations of the various methods to help investigators decide which method is best for their purposes.    It is often recommended to use multiple methods to quantify root dynamics (Vogt et al. 1998; Hendricks et al. 2006; Yuan and Chen, 2012b), but few studies compare methods at the same sites and at the same sampling time (e.g., Makkonen and Helmisaari, 1999; Hertel and Leuschner, 2002; Ostonen et al. 2005; Hendricks et al. 2006; Girardin et al. 2013; Sun et al. 2015). This study builds on earlier reviews (e.g., Vogt et al. 1998), but with greater emphasis on coarse roots due to the present recognition of their important roles in ecosystem productivity and C budgets (Clark et al. 2001a; Smyth et al. 2013; Doughty et al. 2014). In this review I address the following questions: (1) What are the existing methods to assess root biomass and productivity (including new and lesser-known ones)? (2) How do they compare in terms  26 of cost, labour requirements, time efficiency and accuracy? (3) How do fine- and coarse-root biomass and productivity estimates compare among different methods measured at the same sites?   2.3 Methods 2.3.1 Literature search and data compilation  Data was compiled through a literature search from journal platforms (Web of Science, Scirus, JSTOR and Google Scholar) and library sources using keywords and the terms ‘fine root’, ‘coarse root’, ‘root biomass and production or productivity’ and ‘belowground biomass allocation’. All data are from studies conducted in forest and woodland ecosystems (as they contain more than 60% of terrestrial C (Dixon et al. 1994), which should improve the clarity of the relationship between root biomass and productivity estimates provided by different methods.  Stands of all ages were used, including managed (irrigated, thinned and fertilized) and unmanaged stands. With respect to root sampling, additional criteria were: i) the study must have included the diameter used to define fine and coarse roots; ii) roots were sampled by using the soil-pit and soil-core methods to quantify biomass; iii) fine-roots were sampled using at least two of the direct methods (ingrowth-core, (mini) rhizotrons and sequential-coring) to estimate productivity; iv) sampling for fine-root productivity (FRP) should have lasted at least one vegetation season or 12 months; and v) data were collected from a single site and within the same period. Criteria used to identify fine and coarse roots are not uniform, and are usually defined based on arbitrary diameter classes (e.g., Nadelhoffer and Raich, 1992; Levigne and Krasowski, 2007; Finér et al. 2011a). From the database, fine roots were defined as ≤ 0.5 mm, ≤ 1 mm, ≤ 2 mm and ≤ 5 mm in diameter.   Coarse roots also ranged from > 2 mm to > 50 mm in diameter. However the majority of the studies defined fine and coarse roots as ≤ 2 mm and > 2 mm in diameter (see Appendix and references therein). These definitions were used to broadly classify fine and coarse roots in the first part of this review. These classifications have also been used in other reviews (e.g., Yuan and Chen, 2012b; Zhang and Wang, 2015). In the analysis of root biomass and productivity, fine and coarse roots were not standardized to  27 specific diameter classes (Finér et al. 2011a), but were considered to be as defined in the original studies (Nadelhoffer and Raich, 1992).   For the first objective, the database was critically assessed to extract information on existing root biomass and productivity methods (including new and less known ones), their operational principles and strengths and limitations. From the information gathered a matrix was developed to compare methods based on criteria such as ease of field application, cost-effectiveness, labor requirements, time efficiency, accuracy and impact on the ecosystem (Tables 2.1 and 2.2). For this review, time efficiency is considered to be the person-hours required to complete field (set-up and sampling) and laboratory processing (Levillain et al. 2011), and accuracy is the capacity for a method to provide accurate estimates. For the second objective, root biomass data were compiled from studies that compared more than one method at the same site. For fine and coarse roots biomass data were compared for the soil-pit and soil-core methods. Nine observations were obtained for fine root biomass (FRB), while eleven observations of coarse root biomass (CRB) were made from seven studies (Table A.1). The Voronoi trench was considered as a ‘soil pit’ since its field application is similar to the soil-pit method (Levillain et al. 2011). Moreover the study by Levillain et al. (2011) did not compare CRB (> 10 cm in diameter) between soil pit and soil cores, but was done for small roots (2-10 cm). Therefore the values produced by the two methods for the small roots were used in the analysis.    For fine root productivity (FRP), 66 observations were from studies comparing ingrowth-core and sequential-coring, while 25 observations were from comparative studies of ingrowth-core and (mini) rhizotrons. There were 11 observations from studies that compared sequential-coring and (mini) rhizotron methods (Table A.2). A total of 32 studies were considered for fine root productivity. Data for (mini) rhizotrons were compiled from studies that used both the minirhizotron and rhizotron to sample roots (e.g., Fahey and Hughes, 1994; Steele et al. 1997; Metcalfe et al. 2007a; Girardin et al. 2013). The  28 majority of data were from sites in North America, South American and Europe. There were few sites in Central America and Asia- with only one site in Africa and Australia, respectively (Figure 2.1).   Root biomass or productivity estimate from each site was considered as a single observation, even if methods were compared at different sites (Nadelhoffer and Raich, 1992; Finér et al. 2011a). In studies where data for forest floor and mineral soil were presented separately, they were combined to produce a single value. Data from multiple years were averaged. Several approaches have been developed to convert root-growth data from (mini) rhizotrons into biomass (Taylor et al. 1990; Metcalfe et al. 2007a; Milchunas, 2012), and to convert sequential-coring and ingrowth-core data into fine root productivity estimates (Nadelhoffer and Raich, 1992; Vogt et al. 1998; Hendricks et al. 2006; Brunner et al. 2013). In situations where data were presented for different approaches within the same method, the average value was used (e.g., Hertel and Leuschner, 2002; Hendricks et al. 2006; Metcalfe et al. 2007a). In a few cases where a range was given, the average was calculated between the lower and upper values (e.g., Noguchi et al. 2007).     Figure 2.1 Distribution of countries with sites used in the analysis of fine- and coarse root biomass and productivity. With the exception of Africa and Australia, each point represents more than one site.   29 2.3.2 Statistical Analyses  The data extracted were subsequently used to calculate average fine and coarse root biomass and FRP estimates for the different methods. The difference between the average estimates among fine and coarse root biomass and FRP methods was tested by student t-test and one-way analysis of variance (ANOVA). Differences in FRP estimates among the methods were analyzed separately for each biome using one-way ANOVA. Furthermore, FRP estimates were also compared for tropical, temperate and boreal forests using one-way ANOVA, followed by post-hoc Tukey’s HSD test. Regression analysis was used to evaluate relationships between FRB estimates from soil-pit and soil-core methods, and FRP estimates from sequential-coring and ingrowth-core, ingrowth-core and (mini) rhizotrons, as well as sequential-coring and (mini) rhizotrons methods. Root biomass and productivity data were converted to Mg ha-1 and Mg ha-1 year-1, and log-transformed prior to analysis when necessary. All analyses were performed with the GraphPad Prism 7 (GraphPad Software, Inc., California, USA) software package, with the significance level of p < 0.05.   2.4 Results and Discussion  2.4.1 Literature review 2.4.1.1 Coarse-root biomass and productivity  Due to their large size and spatial heterogeneity, coarse roots (mostly > 2 mm in diameter) can be difficult and expensive to quantify in the field. Studies have relied on both direct and indirect methods to estimate coarse root biomass and productivity.   2.4.1.1.1 Direct methods Coarse roots can be estimated using direct methods such as root excavation, soil-pit/soil-pit ingrowth, wall or trench profiles and soil coring. These methods generally produce reliable root-biomass estimates, but have strengths and limitations (Table 2.1). Therefore investigators should consider factors such as  30 field adaptability, cost effectiveness, labor requirements, and accuracy as well as their impacts on ecosystems (Table 2.1) before selecting a method.   2.4.1.1.1.1 Root excavation The root-excavation method has been used to estimate coarse root biomass (CRB) of individual trees and stands in tropical (e.g., Misra et al. 1998; Niiyama et al. 2010; Lima et al. 2012), temperate and boreal forest and woodland ecosystems (e.g., Bledsoe et al. 1999; Millikin and Bledsoe, 1999; Resh et al. 2003; Brassard et al. 2011a). The excavation method can generally be distinguished into two types, total root and root-ball excavation. With total root excavation, the entire root system of an individual tree within a designated radius is excavated manually (Bledsoe et al. 1999; Berhongaray et al. 2015) or with the help of machinery such as a tractor, backhoe or power shovel (Brassard et al. 2011a; Ryan et al. 2011; Fortier et al. 2015). The root-biomass estimates of individual trees can be related to aboveground biometric parameters such as diameter at breast height (DBH) to develop allometric equations, which can then be used to estimate belowground biomass of an entire stand (e.g., Haynes and Gower, 1995; Resh et al. 2003; Berhongaray et al. 2015). In root-ball excavation, a soil monolith is centred on the stump of a target tree (Bledsoe et al. 1999), and the roots (including lateral roots) are recovered following soil excavation either manually or mechanically, without removing the stump. Roots are then washed, dried and weighed. Although root-ball excavation is considered an expeditious and cost-effective method to estimate CRB (Miller et al. 2006), field application has mostly been limited to young plantations (e.g., Misra et al. 1998; Resh et al. 2003; Miller et al. 2006).   The main advantage of the excavation method is that it allows for direct harvesting of coarse roots concentrated mostly around the base of the stem (Bledsoe et al. 1999; Macinnis-Ng et al. 2010), resulting in an improved estimate of root biomass (Millikin and Bledsoe, 1999; Resh et al. 2003). For instance, in a Quercus douglasii stand in California, Millikin and Bledsoe (1999) reported higher CRB estimates using the excavation method compared to soil-pit and soil-core methods, and so recommended the excavation  31 method for estimating CRB. Excavation is considered the only method that will accurately quantify CRB in individual trees and whole stands (e.g., Snowdon et al. 2002; Cole and Ewel, 2006; Ouimet et al. 2008; Niiyama et al. 2010; Ryan et al. 2011), and is particularly recommended for sites with high spatial heterogeneity of roots (Resh et al. 2003).   However, the excavation method can be laborious and expensive (Danjon and Reubens, 2008). According to Danjon et al. (2005), it could take a whole day for five people to manually uproot twenty-four 50-year-old Pinus pinaster roots in a sandy spodosol soil, which is relatively well suited for excavation. The excavation method may result in sampling error as roots become broken or lost during excavation (Millikin and Bledsoe, 1999; Niiyama et al. 2010). Even under ideal conditions complete recovery of entire root systems is difficult, especially for large trees, which usually have extensive and deep root systems (Bledsoe et al. 1999). Indeed in a tropical primary forest in Malaysia, Niiyama et al. (2010) estimated biomass of lost roots to be ca. 19 Mg ha-1, accounting for 23% of the total CRB. Procedures for determining the biomass of broken lateral roots from intact laterals or harvesting of broken root fragments following stump excavation have been developed to correct for lost and broken roots (Santantonio et al. 1977; Millikin and Bledsoe, 1999; Niiyama et al. 2010; Brassard et al. 2011a). Excavation is also destructive (Danjon and Reuben, 2008), as excavation and tree root removal can seriously disturb soil a long distance from the target tree (Bledsoe et al. 1999; Cole and Ewel, 2006), and so may not be feasible if minimizing site disturbance is a priority.   2.4.1.1.1.2 Soil-pit/Soil-pit ingrowth  Soil-pit determination of root biomass is another direct method frequently used to estimate coarse root biomass and productivity (e.g., Kimmins and Hawkes, 1978; Fahey et al. 1988; Eamus et al. 2002; Rau et al. 2009; Smith et al. 2013; Costa et al. 2014). Soil pits have also been employed to determine lateral-root distribution in ecosystems (Bledsoe et al. 1999). With this method a number of trenches are excavated manually or mechanically, and the soil is collected and the roots processed to estimate root biomass  32 (Levillain et al. 2011). These biomass estimates can subsequently be used to establish allometric relationships with aboveground biometric parameters such as DBH and basal area to estimate total stand root biomass (Park et al. 2007). Studies have used varying pit dimensions and layouts, which typically depend on the size and spatial distribution of coarse roots (e.g., Rau et al. 2009; Chidumayo, 2014; Costa et al. 2014); greater numbers of pits being necessary at sites of high belowground heterogeneity (Bledsoe et al., 1999). A critical aspect of this method is ensuring that soil pits are excavated deep enough to recover more than 90% of the roots (Bledsoe et al. 1999). The sampling depth can be determined in a pilot trial to establish the maximum rooting depth. For instance, Park et al. (2007) reported that a sampling depth ranging from 0.95 m to 1.7 m reflected the natural variation in maximum root depth in mixed deciduous stands in New Hampshire, USA. The soil-pit method can also be used to quantify coarse root productivity by removing roots from a volume of soil, and monitoring the regrowth of roots over a period of time (Jordan and Escalante, 1980; Kangas, 1992; Lauenroth, 2000). Root productivity is assumed to be the amount of root biomass accumulated during the interval (Kangas, 1992). The basic principle behind the soil-pit ingrowth method is similar to the ingrowth-core method (discussed under section 2.4.1.2.1.3), even though no mesh bag is installed.   The soil-pit method has been found to provide good estimates of CRB (e.g., Millikin and Bledsoe, 1999; Jackson et al. 2009; Levillain et al. 2011; Smith et al. 2013). In a Eucalyptus plantation in Congo, Levillain et al. (2011) found a higher CRB estimate using the soil-pit method than using soil monoliths. Smith et al. (2013) also found higher CRB estimates from the soil-pit method compared to the soil-core method in an elevated-CO2 Betula pendula stand in the UK. However, they also reported lower CRB from soil pits compared to the soil cores in an elevated-CO2 mixed stand of Alnus glutinosa, B. pendula and Fagus sylvatica (Smith et al. 2013). The soil-pit method allows roots to be collected throughout the soil profile (Park et al. 2007), and therefore deeper roots are included in the root-biomass estimates (Rau et al. 2009). A large area is often sampled in each pit (relative to coring methods), which may improve  33 accuracy of root biomass assessments in ecosystems with greater heterogeneity of coarse roots (Macinnis-Ng et al. 2010).   To reduce the labour required to sample coarse roots, CRB is sometimes estimated from subsamples of soil from soil-pits.  This assumes uniform distribution of roots within pits (Rau et al. 2009), which may not be a valid assumption (Park et al. 2007). Nevertheless, because of the general reliability of the soil-pit method, it is often employed to validate the accuracy of root-biomass estimates provided by other methods (e.g., Park et al. 2007; Smith et al. 2013). For example, Smith et al. (2013) used a CRB estimate determined from soil pits to confirm the validity of data obtained through soil coring, and found a strong correlation between the two methods. Soil pits are also considered the best method for sampling roots in sites with stony soils where stones can hinder the ability to extract cores from soil and increase root heterogeneity and sampling bias (Park et al. 2007; Rau et al. 2009). Whilst the soil-pit method is excellent for sites of high spatial variability of roots, it is also recommended for ecosystems with low root heterogeneity (Bledsoe et al. 1999).   Despite its wide use, the soil-pit method has been shown to under-estimate CRB (e.g., Bledsoe et al. 1999; Park et al. 2007; Chidumayo, 2014) probably due to the exclusion of roots close to tree stems (Park et al. 2007) where coarse-root density is high. Soil-pit sampling also mostly excludes taproots, root crowns and lignotubers, thereby under-estimating the total biomass of coarse roots (e.g., Fahey et al. 1988; Millikin and Bledsoe, 1999; Park et al. 2007). Due to the destructive nature of soil pits, it is not considered to be suitable for directly sampling root biomass of individual trees (Rau et al. 2009), or for the estimation of coarse root productivity by soil-pit ingrowth (Jordan and Escalante, 1980; Kangas, 1992).      34 2.4.1.1.1.3 Wall or soil trench profile  The wall or soil-trench profile is another direct method used to estimate CRB (Epron et al. 1999; van Noordwijk et al. 2000; Achat et al. 2008). Wall or trench profiling involves digging a trench with a spade or excavator in a stand to expose the coarse roots. The soil surface is shaved and flattened to improve root observation (Achat et al. 2008), and a grid frame is installed onto the trench wall, usually with nails, down to a specified depth (Grant et al. 2012). Roots that are exposed along the vertical plane are carefully counted and mapped (Maurice et al. 2010). Wall profiles can also be photographed, and the images assembled to map the soil profile (Schmid and Kazda, 2002). Roots can be traced or images digitised to determine root length and other root metrics (e.g., number of tips, forks etc.) using image analysis software (Schmid and Kazda, 2002; Lobet et al. 2013). Root length can then be converted to root biomass per unit area (Metcalfe et al. 2007a). Alternatively, root biomass may be estimated from root volume data (e.g., Epron et al. 1999; Curt et al. 2001). The wall profile assumes a positive relationship between the number of roots on the soil vertical plane, and the root length density inside the soil column (Bledsoe et al. 1999). This method is less time-consuming than other direct methods and is non-destructive, and therefore provides the opportunity for investigators to quantify root systems in situ over large time scales (Curt et al. 2001). Curt et al. (2001) noted that observation and counting of 74 root profiles required about 30 person-days, compared to the 100 person-days required to completely excavate one mature tree. Root recordings yet to be converted to biomass can also be combined with root mass data obtained from other methods to produce a better estimate of CRB (e.g., Epron et al. 1999; Achat et al. 2008). For example, Achat et al. (2008) used the intersection of roots obtained from wall profiles to improve root biomass estimates from soil-cores for a Pinus pinaster stand in south-western France.   The wall profile method is, however, rarely applied, because it can substantially under-estimate CRB since the relative values produced have to be corrected (Bledsoe et al. 1999; Levillain et al. 2011), and therefore its use for estimating CRB has been discouraged. Furthermore, when using the wall profile method, soil properties must be considered in order to visualize undisturbed root growth. It may also be  35 difficult to distinguish live and dead roots, and to identify roots of different species (van Noordwijk et al. 2000; Curt et al. 2001). Although not well suited for estimating CRB (Bledsoe et al. 1999), the wall profile method is highly recommended for studies of root morphology and density distribution (Schmid and Kazda, 2002; Maurice et al. 2010; Grant et al. 2012).  2.4.1.1.1.4 Soil-core  Soil cores are also used to estimate CRB, although they may under-estimate CRB (Millikin and Bledsoe, 1999; Resh et al. 2003; Jackson et al. 2009; Major et al. 2012), as a result of the small volume of soil sampled (Levillain et al. 2011), and the infrequent encountering of large roots during coring (Butnor et al. 2012; Taylor et al. 2013). For instance, in a Eucalyptus plantation in Tasmania, Resh et al. (2003) found that soil cores under-estimated CRB by 9% compared to estimates from total-tree excavation. Jackson et al. (2009) also found that increasing CRB in response to CO2 enrichment was only detected when they shifted from the use of soil cores to soil pits at the Duke FACE experiment site in Durham, USA. On the other hand, Rau et al. (2009) and Smith et al. (2013) reported no significant difference in CRB estimates collected using soil-pit and soil-core methods. Resh et al. (2003) argued that soil-coring is a pragmatic method, as the accuracy is not usually affected by increasing sampling intensity, and it also allows for examination of the spatial distribution of coarse roots both vertically and laterally. It is also simpler, as sampling processes and data calculations are considerably less complicated than either the soil-pit or excavation method (Rau et al. 2009). The coring device can also be easily transported, and therefore the soil-core method is efficient for sampling roots in remote sites where accessibility by heavy-duty equipment is limited (Rau et al. 2009).   2.4.1.1.2 Indirect methods Due to the cost and labour requirements of direct methods for estimating root biomass (Table 2.1), many investigators have employed indirect techniques to estimate CRB and productivity. These indirect techniques include size-mass allometric equations, root-shoot or belowground-aboveground ratio, GPR,  36 root biomass increment or difference, fraction or percentage of wood productivity, and root radial increment. The aforementioned methods are increasingly being applied in field studies because they provide quick and non-destructive means to quantify the contributions of fine and coarse roots to total forest and woodland biomass and carbon storage.   2.4.1.1.2.1 Size-mass allometric equations  The size-mass allometric technique is the most common indirect way to estimate CRB (Whittaker et al. 1974; Bledsoe et al. 1999; Nàvar, 2009; Niiyama et al. 2010; Waring and Powers, 2017). To develop allometric equations, CRB is usually related to aboveground parameters like DBH or height (Snowdon et al. 2002; Brassard et al. 2011a). The basis for these equations is the assumption that coarse roots grow at a rate which is similar to wood biomass, and therefore the two components are related in size and age (Niiyama et al. 2010). Despite some exceptions (Bolte et al. 2004; Hunziker et al. 2014), the allometric relationships between wood biomass and CRB in many species are generally consistent (Whittaker et al. 1974; Jenkins et al. 2003; Kenzo et al. 2009; Nàvar, 2009), and it has been suggested that CRB in different sites and geographic regions could be estimated with published allometric equations developed from DBH (Brassard et al. 2011a). Different site- and species-specific equations, as well as generalized equations have been developed, and applied to estimate CRB in different regions of the world (Whittaker et al. 1974; Niiyama et al. 2010; Chidumayo, 2014; Waring and Powers, 2017). For instance, the allometric equations developed by Whittaker et al. (1974) for broadleaved-deciduous forest stands in New Hampshire have been used extensively to predict CRB in northern hardwood tree species such as Acer saccharum and Fagus grandifolia in the USA (Jenkins et al. 2003; Vadeboncoeur et al. 2007). Allometric equations are useful because they offer a cost-effective and reliable means to estimate stand-level CRB (Bledsoe et al. 1999; Xiao et al. 2003; Vadeboncoeur et al. 2007). Vadeboncoeur et al. (2007) reported that allometric equations developed to estimate CRB in F. grandifolia, A. saccharum and Betula alleghanien agreed well with root mass estimated from soil pits. Coarse root allometric equations are also  37 useful for evaluating temporal changes in root biomass and carbon stocks in forest ecosystems (Mokany et al. 2006; Niiyama et al. 2010).   Despite general acceptance, allometric equations may under- or over-estimate root biomass (Park et al. 2007; Kenzo et al. 2009; Brassard et al. 2011a). Park et al. (2007) reported a 32% lower CRB estimate from allometric equations than from soil pits in young hardwood stands in New Hampshire. Brassard et al. (2011a) also recorded large variations in Abies balsamea CRB using their own developed models and other allometric models from the literature. According to Niiyama et al. (2010) the accuracy and precision of root-biomass estimates are influenced by both the quantity and quality of data used for developing the allometric models. Some allometric equations do not correct for roots that are lost or broken during sampling, and this introduces uncertainty when applied at different sites. Larger trees are also sometimes excluded during the collection of data used to develop allometric models affecting the accuracy of extrapolated biomass estimates (Xiao et al. 2003). Most coarse root allometric equations are not validated at sites against multiple methods of determining CRB, which would increase accuracy when applied over large spatial scales (Bolte et al. 2004; Vadeboncoeur et al. 2007).   2.4.1.1.2.2 Root-shoot or Belowground-aboveground ratio The root-shoot ratio (RSR) or belowground-aboveground (BG/AG) biomass ratio is a common approach for estimating CRB. The ratio expresses a general relationship between root biomass and shoot biomass (Sanquetta et al. 2011; Luo et al. 2012; Poorter and Sack, 2012; Waring and Powers, 2017), which reflects the net effects of C allocation in a particular ecosystem (Mokany et al. 2006). CRB is estimated from aboveground biometric parameters such as DBH and biomass (Levigne and Krasowski, 2007; Malhi et al. 2009), based on the assumption of a positive relationship between root biomass and wood biomass (Eamus et al. 2002, Sanquetta et al. 2011; Poorter and Sack, 2012), which may not hold for all ecosystems (e.g., Barton and Montagu, 2006). Generalized and group-specific RSR exist for different tree species, forest and woodland types and biomes (Cairns et al. 1997; IPCC, 2006; Smyth et al. 2013; Yuen  38 et al. 2013; Waring and Powers, 2017). A RSR range of 0.10 - 1.16 has been estimated for the world’s forests and woodlands (IPCC, 2006), whilst, in a global literature study, Cairns et al. (1997) reported mean RSRs of 0.24, 0.26 and 0.27 for tropical, temperate and boreal forest biomes. In a recent global study, Waring and Powers (2017) estimated a RSR of 0.54 for tropical forests.   The RSR is a practical and cost-effective approach to estimate CRB. The method is being widely applied, especially in C dynamics studies (e.g., Malhi et al. 2009; Yuen et al. 2013; Smyth et al. 2013; Doughty et al. 2014), as the significance of coarse roots to the total belowground biomass and productivity becomes more apparent (Clark et al. 2001a; Waring and Powers, 2017). RSR has also become particularly useful since there are few available coarse-root allometric equations that can be extrapolated to other vegetation types and biomes. The approach is also useful for the calibration and validation of global and regional dynamic C-cycling models (Mokany et al. 2006). The RSR approach can however produce biased CRB estimates (Levigne and Krasowski, 2007; Yuen et al. 2013; Fortier et al. 2015). For instance, Levigne and Krasowski (2007) recorded greater CRB estimates using allometric equations than from RSRs in A. balsamea stands in New Brunswick, Canada. They also concluded that there was more belowground biomass in A. balsamea stands than estimated by the generic RSR often applied to coniferous forests (Levigne and Krasowski, 2007). This limitation is mainly due to the high variability that often characterizes RSRs (Mokany et al. 2006; Poorter and Sack, 2012; Nàvar, 2015; Waring and Powers, 2017). RSRs may also be affected by species traits, site, stand development characteristics and the methods used to quantify root biomass (Keith et al. 2000; Barton and Montagu, 2006; Waring and Powers, 2017). According to Keith et al. (2000) most generalized RSRs do not account for lignotubers, or the influences of resprouting in some tree species. The use of generalized and mean RSRs may therefore lead to biased estimates of CRB and C dynamics (e.g., Yuen et al. 2013). There is a general consensus that the use of site-specific RSRs is more accurate (Mokany et al. 2006; Levigne and Krasowski, 2007; Fortier et al. 2015), and therefore efforts should be made to generate site-specific ratios for specific forest types in different regions of the world.   39 2.4.1.1.2.3 Ground-Penetrating Radar One of the emerging indirect techniques to estimate CRB is the use of GPR. GPR is a geophysical close-range remote-sensing technique, which uses electromagnetic waves (EM) to obtain root images (Fisher et al. 1992; Butnor et al. 2001, 2012; Guo et al. 2013a). The principle underlying GPR operation is the ability of low-frequency radio and microwaves to penetrate into the subsoil to capture root images (Lorenzo et al. 2010). A GPR system usually consists of three main components - a radar control unit, an antenna and a display unit (Guo et al. 2013a). During sampling the GPR antenna is moved along established transects arranged in a grid or any other pattern, and the EM energy from the antenna propagates into the soil as waves (Butnor et al. 2012). The transmitted EM signal reflected back from the roots is digitized, recorded and processed (Guo et al. 2013a). For an in-depth discussion on GPR components, operational principles, signal-processing techniques and GPR-theoretical background see Butnor et al. (2001, 2003, 2012), Lorenzo et al. (2010) and Guo et al. (2013a). GPR has been used to estimate CRB in individual trees and stands in Europe (e.g., Lorenzo et al. 2010), Asia (e.g., Cui et al., 2011; Hirano et al. 2012) and North America (e.g., Butnor et al. 2001, 2003; Day et al. 2013; Raz-Yaseef et al. 2013; Samuelson et al. 2015). GPR can also detect and quantify coarse-root spatial distribution (Hirano et al. 2012) and tree decay patterns (Butnor et al. 2009), and determine the physical properties of trees and soil depth (Lorenzo et al. 2010). With this approach CRB is estimated based on an established relationship between GPR processed radargrams or profiles and destructively sampled root biomass (Butnor et al. 2003; Day et al. 2013).   GPR has some advantages that make it promising for root studies. In soils with low organic matter and ion contents, GPR can produce CRB estimates comparable with other methods. For instance, Day et al. (2013) reported that GPR-based estimate of CRB was comparable to soil-pit estimates, and greater than those obtained from soil cores. The GPR system is portable, and can therefore be used to scan large areas over a short period of time (Butnor et al. 2012). GPR can account for large spatial variability of tree roots within a short time, and can be used to repeatedly monitor and characterize roots at the same site (Stover  40 et al. 2007; Day et al. 2013). This is particularly important since repeated quantification of root systems in forest and woodland ecosystems can be very challenging (Bledsoe et al. 1999). GPR can also be utilized with other methods to improve efficiency and accuracy of root-biomass estimates over large areas (Butnor et al. 2003).   Despite these advantages, successful application of the GPR system is usually site-specific (Butnor et al. 2012), and comparisons between biomass estimates can only be achieved under comparable root, soil and environmental conditions (Butnor et al. 2001; Guo et al. 2013b; 2015). GPR-based root-biomass studies have mainly been conducted in plantations and isolated trees under controlled conditions (e.g., Cui et al. 2011; Day et al. 2013; Samuelson et al. 2015). These simplified environments are used because the ability of the GPR to detect and map roots is directly related to the characteristics of a site (Butnor et al. 2001; Lorenzo et al. 2010). According to Lorenzo et al. (2010), it may be difficult to obtain accurate GPR data in a dense and highly diverse ecosystem; this precludes most tropical forest ecosystems that are usually characterised by diverse structural attributes and life forms. In a Populus deltoides stand in the Atlantic Coast Flatwoods, Butnor et al. (2001) reported that GPR was not useful as understory vegetation and other fallen debris hindered the free movement of the antenna. Samuelson et al. (2015) circumvented this problem by mowing and raking the understory of Pinus palustris stands to improve mobility of the GPR system, but this will not be practicable in many ecosystems. Successful application of GPR has been limited to well-drained sandy soils (e.g., Butnor et al. 2001; Hirano et al. 2012; Day et al. 2013). According to Guo et al. (2013a), sandy soils with low concentrations of organic matter and ions are the most conducive sites for the use of GPR. The water content of roots is crucial to the detection of tree roots by GPR (Hirano et al. 2009a; Guo et al. 2013b), as it closely relates to the dielectric properties of the surrounding soil (Guo et al. 2013a). Under controlled conditions roots of Cryptomenia japonica with less than 20% water content were not identified by GPR (Hirano et al. 2009a). Furthermore GPR may not be able to distinguish overlapping and closely aligned roots (Butnor et al. 2001; Hirano et al. 2009a; Guo et al. 2015); these are often interpreted as a single root, leading to under-estimation of CRB (Butnor et al.  41 2001). GPR is also not able to differentiate roots into size and depth classes; any attempt is confounded by the roots’ orientation, surface-reflectance geometry, and the closeness of untargeted roots (Butnor et al. 2003; Tanikawa et al. 2013; Guo et al. 2015). It is also difficult to distinguish between live and dead roots with GPR (Butnor et al. 2001). However, under ideal conditions, GPR with advanced antenna frequency may detect (with less precision) decayed roots close to the soil surface (Butnor et al. 2005; Zhu et al. 2014). The reliability of GPR estimates needed to be validated with data from direct root biomass estimates (e.g., Stover et al. 2007; Butnor et al. 2012; Zhu et al. 2014), and further methodological developments could enhance the promise of this technique for quantifying root biomass.   2.4.1.1.2.4 Root biomass increment or difference This approach quantifies coarse-root productivity (CRP) based on changes in CRB over a specific time period (Sala et al. 1988; Steele et al. 1997), and productivity is computed as either the sum of biomass increments (e.g., Kajimoto et al. 1999; Upadhaya et al. 2005) or the biomass difference (e.g., Haynes and Gower, 1995; Hermle et al. 2010). Studies that use biomass increments as an index for productivity usually depend on root biomass data from soil cores taken across seasons for at least one year (e.g., Upadhaya et al. 2005). The latter approach however quantifies CRP as root biomass difference obtained from allometric equations (Bond-Lamberty et al. 2004; Nàvar, 2015). Most long-term studies use the biomass difference index to estimate CRP (e.g., Haynes and Gower, 1995; Hermle et al. 2010). The use of biomass data to estimate root productivity is simple and fast and therefore commonly used in C budget studies (e.g., Upadhaya et al. 2005; Hermle et al. 2010).   The approach has, however, several limitations; for instance because it is time-dependent it may under-estimate CRP. The use of biomass increment for instance mostly ignores contribution of root mortality and turnover, and losses due to exudation of C-containing compounds (Kajimoto et al. 1999; Hermle et al. 2010). This means that values obtained through this approach are relative, and do not reflect actual roots produced (Sala et al. 1988). Studies that estimate root productivity as CRB difference from  42 allometric equations suffer from the same limitations as the indirect root-biomass methods. Moreover subtracting two biomass estimates with their associated error terms can lead to a higher error for the final value (Sala et al. 1988; Bond-Lamberty et al. 2004).    2.4.1.1.2.5 Fraction or percentage of wood productivity  One indirect approach to estimate CRP is as a fraction/percentage of the aboveground wood productivity (Law et al. 2001; Malhi et al. 2009; Girardin et al. 2010), based on the assumption of a positive relationship between coarse-root biomass productivity and wood productivity in many ecosystems (Mokany et al. 2006; Poorter and Sack, 2012). For example, coarse-root productivity was estimated for a mature Pinus ponderosa forest in central Oregon by assuming it to be 25% of wood productivity (Law et al. 2001). Malhi et al. (2009) multiplied the stem productivity estimate by 0.21 to estimate CRP in selected sites in Brazil. This approach is simple, cost-effective and fast, and is increasingly being applied in C cycling and allocation studies, particularly in the tropics (e.g., Girardin et al. 2010; 2016; Doughty et al. 2014). It may, however, under-estimate CRP since it usually captures only production of large structural roots (Malhi et al. 2009). The mass of small coarse roots that are not accounted for by this approach exceeds that of the large structural roots in many ecosystems (Clark et al. 2001a; Girardin et al. 2010).   2.4.1.1.2.6 Root radial increment   Recently CRP has been estimated from direct measurements of the radial increment of roots (Moser et al. 2010, 2011; Zach et al. 2010). With this in situ approach, diameter tapes or dendrometer bands are used to record the diameter increments of superficially growing roots for a given period (Zach et al. 2010; Moser et al. 2011). For instance, Moser et al. (2010) manually measured coarse roots with a diameter tape in a moist montane forest in southern Ecuador, and found that CRB increased with elevation. When measurements are recorded with dendrometer bands, the bands are normally mounted on the roots and monitored for some period before recordings are made. Based on the diameter data, the relative volume of  43 each root segment is determined (Moser et al. 2011). These root-volume estimates, together with root-biomass and stand-density data, are then used to estimate stand-level CRP (Moser et al. 2010). This approach provides a direct means to estimate CRP, which may be an important step towards circumventing the difficulties associated with field monitoring of coarse roots.   However, this method suffers from the same limitation as allometric equations (section 2.4.1.2.1.1), since it fails to capture small coarse roots (e.g., Moser et al. 2011). In the study of Moser et al. (2011), only roots with diameters between 20 and 300 mm were equipped with dendrometer tapes, leading to the conclusion that CRP was likely under-estimated due to the lack of data from roots < 20 mm in diameter. Another criticism of this method is that during sampling, only superficial roots are usually considered (Zach et al. 2010), but more coarse roots may be distributed in the deep layers of the mineral soil (Levillain et al. 2011). Nevertheless, this method has been employed in the tropics, where visible aboveground roots and adventitious roots may be common (Vance and Nadkarni, 1991). Root-productivity estimates may also be affected by sampling bias and lack of replication. Despite some limitations, long-term in situ monitoring of coarse roots with dendrometer bands holds promise for future C dynamics studies in the tropics.   2.4.1.2 Fine-root biomass and productivity Fine roots (≤ 2 mm in diameter) contribute less than coarse roots to total root biomass (Vogt et al. 1998), but play important roles in the productivity and biogeochemical cycles of ecosystems. Fine-root biomass and productivity have also been estimated by both direct and indirect methods (Table 2.2). Direct methods include soil-core/sequential-coring, monolith, soil-pit, ingrowth-core and (mini) rhizotrons, while indirect approaches include N-budget, pipe-model and other empirical models.      44 2.4.1.2.1 Direct methods  2.4.1.2.1.1 Soil-core Fine-root biomass (FRB) is commonly estimated by the soil-core method (Bledsoe et al. 1999; Makkonen and Helmisaari, 1999; Oliveira et al. 2000). Soil cores or augers are usually used to sample fine roots, depending on the availability of equipment and investigator preference. One of the most important considerations is the diameter of the corer or auger used for sampling roots (Oliveira et al. 2000; Snowdon et al. 2002). According to Oliveira et al. (2000) the diameter of a corer should provide a good balance between sampling accuracy and intensity. Commonly used core diameters range from 1.9-15 cm (Oliveira et al. 2000; Levillain et al. 2011), but Snowdon et al. (2002) recommended corers of diameter >10 cm to estimate FRB. Field sampling may be based on random, systematic or stratified-random designs; an important consideration in designing field-sampling procedures is to avoid clumping of sampling locations so as to capture the spatial heterogeneity of root distribution in ecosystems.  The depth to which root cores are taken is crucial for accurate and reliable determination of FRB. Ideally, fine roots should be sampled to the maximum root-depth limit for the site (Bledsoe et al. 1999; Oliveira et al. 2000), but this may not always be possible, particularly at sites with stony soils (Park et al. 2007). Sampling depth should therefore be informed by a preliminary assessment of the concentration of fine roots at each depth. The soil-core method has been extensively used to estimate FRB for decades (see Vogt et al. 1998; Yuan and Chen, 2012b), and is preferred by most investigators (e.g., Bledsoe et al. 1999; Park et al. 2007). The utility of the soil-core method is its simplicity and its ability to capture well the spatial and temporal heterogeneity in FRB distribution over large scales (e.g., Makkonen and Helmisaari, 1999). According to Makkonen and Helmisaari (2001), soil cores are preferred for the study of annual and seasonal variations in FRB. Root-biomass data from the soil-core method is also used to validate the efficiency and accuracy of other methods (e.g., Day et al. 2013).     45 Table 2.1 Comparison of methods for estimating coarse-root biomass and productivity using selected criteria.   Method Type Operational principle Requirements for optimum Ease of field Cost Labor Time  Accuracy  Impact on        accuracy  application effectiveness requirements efficiency   ecosystem Coarse-root biomass                   Root excavation Direct Relates root biomass to aboveground Correction for lost and broken Difficult-not feasi Costly Laborious Inefficient High Destructive   parameters e.g., DBH roots ble in many sites      Soil-pit Direct Assumes roots are uniformly Soil pits must be deep enough to recover Difficult Costly Laborious Inefficient High Destructive   distributed in pits  about 90% of roots       Wall or soil trench  Direct Assumes a positive relationship Correction for relative root biomass Simple  Economical Less labor Efficient Low Non-destructive profile  between counted roots and root length estimates produced         density        Soil-core Direct Equates root biomass to roots in a soil Large diameter corer/auger Simple  Costly Laborious Inefficient Low Destructive   column preferably (> 10 cm) must be used       Mass-size allometric Indirect Woody components both above-and Validation of models for specific sites Simple Economical Less labor Efficient Low Non-destructive equations  belowground are related in size and age        Root-shoot or below Indirect Assumes a significant relationship Generation of site-specific ratios Simple  Economical  Less labor Efficient Low Non-destructive ground-aboveground  between root biomass and aboveground       ratio  biomass        Ground-Penetrating Indirect Based on geophysical close range  Well drained sandy soil, with low organic Simple-but usually Costly Less labor Efficient Low Non-destructive Radar (GPR)  remote sensing technique matter and ion content site specific                Coarse-root productivity           Soil ingrowth pit  Direct Assumes production to be propor More pits required to capture root growth Difficult Costly Laborious Inefficient High Destructive   tional to  root biomass accumulated in pits over a period of time and distribution                  Root biomass Direct/Indirect Based on changes in root biomass  Site-specific models to estimate Simple Depending  Laborious or  Could be High Destructive or increment/difference  over a period biomass difference  on approach simple-based efficient or  not based on        on preferred not  preferred        approach    approach   Fraction/percentage of wood productivity Indirect Assumes a general positive relationship between coarse- Accurate measurement of wood Simple Economical No labor Efficient Low Non-destructive   root and woody biomass parameters e.g., diameter       Root radial  Direct Direct measurements of root radial Capture of small coarse-roots Simple Economical Less labor Efficient Low Non-destructive increment  growth           46  The soil-core method has limitations that affect its field application and the estimates of FRB, and may over- or under-estimate FRB estimates. Soil compaction, which is a common problem with the use of cores, can result in over-estimation of root biomass. For example, compaction resulted in 10% over-estimation of FRB in a naturally regenerated hardwood forest in New Hampshire (Park et al. 2007). However, using a core with a large inner diameter may reduce soil compaction and improve root-biomass estimates (Bledsoe et al. 1999), but the use of large diameter cores in rocky or stony sites may be difficult (Park et al. 2007). The difficulty in separating organic debris from fine roots may lead to over-estimation of root biomass (Millikin and Bledsoe, 1999; Snowdon et al. 2002). Extracting sample cores and root-processing (washing, sorting, weighing) takes considerable time, so in most studies roots are under-sampled, which affects the reliability of FRB data (Levillain et al. 2011; Berhongaray et al. 2013; Taylor et al. 2013). In a study of a Eucalyptus plantation in Congo, Levillain et al. (2011) sampled 312 auger cores from the soil surface (0-10 cm) to achieve a 10% precision. In a young short-rotation Populus plantation in Belgium, Berhongaray et al. (2013) found that when processing extracted cores that the time spent washing, sorting and weighing roots represented 84-93% of the total time needed to process root samples. The high number of core samples required to get a good precision, as well as the processing time makes soil cores expensive and impractical in highly replicated experiments (Bledsoe et al. 1999). Field sampling and laboratory techniques such as randomization, temporal prediction and diameter-class accumulation curves have been developed to improve field and laboratory efficiency and also reduce root-processing time (Metcalfe et al. 2007b; Berhongaray et al. 2013; Taylor et al. 2013).   2.4.1.2.1.2 Monolith Fine-root biomass has also been quantified with the monolith sampling method (e.g., Lawson et al. 1970; Kimmins and Hawkes, 1978; Oliveira et al. 2000; Gautam and Mandal, 2012). The method requires cutting of a monolith (block) of the soil directly from the soil surface, or at different depths, from the wall of a trench. Soil monoliths can also be excavated with frames constructed from wooden or metallic  47 materials (Oliveira et al. 2000; Ibrahima et al. 2010). Investigators should conduct preliminary assessments to ensure that the size and number of monoliths is sufficient to account for the heterogeneity of root distribution at the site before using them in a large study.   Monoliths have some advantages - they provide large soil sample volumes, which increase the accuracy of FRB estimates (Taylor et al. 2013), and reduce the number of replicates needed to secure good estimates of FRB (Levillain et al. 2011). In a Eucalyptus plantation in Congo, 127 monoliths of volume 25 cm × 25 cm × 10 cm were required to achieve 10% sampling precision (Levillain et al. 2011). For 30% precision, the number of soil blocks required fell to 14, which could be accomplished in one person-day (Levillain et al. 2011). Monoliths are also suited for sites where roots may be distributed into the deep layers of the soil (Levillain et al. 2011). The monolith method is, however, beset with limitations (Table 2.2). Apart from being time-consuming and laborious, it can produce erroneous estimates of FRB due to inappropriate washing and oven-drying procedures. Because monoliths collect large soil volumes, it may be practically difficult to completely remove mineral particles attached to lower-order roots. Larger losses may also occur during washing, particularly in cases where frames are used to sample roots (Oliveira et al. 2000). There is, therefore, the need to apply a correction factor to account for root loss and adhering mineral particles, in order to obtain a comparable biomass estimate (e.g., Kimmins and Hawkes, 1978). The monolith method also requires considerable time for sorting and processing of fine roots. Levillain et al. (2011) reported that sieving operations related to monolith root extraction took 31 person-days, accounting for about 55% of the time spent on field sampling and processing. Under unfavourable conditions, soil blocks may collapse during excavation, which will increase the time spent in the field (Makita et al. 2011). Finally, excavation of soil blocks from deep soil layers along a trench was impossible in a broad-leaved deciduous forest in Japan because the soil was hardened (Makita et al. 2011).    48 2.4.1.2.1.3 Ingrowth-core  The root-ingrowth core (Persson, 1979), including various modifications (e.g., Vogt et al. 1998; Hirano et al. 2009b; Lukac and Godbold, 2010; Li et al. 2013; Van Do et al. 2016) is one of the most commonly used methods to estimate fine-root productivity and turnover. The method estimates the amount of fine root that grows into a defined volume of root-free soil within a specified period of time.  Its field application is based on the assumption that disturbances of roots and the surrounding soil during core, or net installation, do not affect root production during the ingrowth period (Lauenroth, 2000). This assumption may not apply in all sites and forest types (e.g., Hendricks et al. 2006). Studies have generally found that the method provides conservative root-biomass estimates, and could under-estimate FRP compared to sequential-coring and (mini) rhizotrons (see Table A.2). This may be due to constraints of mesh size on root ingrowth and rapid turnover of roots, which is missed before sampling (Hertel and Leuschner, 2002; Ademak et al. 2011; Milchunas, 2012). The ingrowth-core method may, however, provide fine-root productivity estimates comparable to other methods if the time elapsed between cores installation and root sampling is long enough to allow for maximum root colonization (e.g., Makkonen and Helmisaari, 1999; Ostonen et al. 2005; Jourdan et al. 2008). This is a very important consideration, particularly for sites with strong seasonal fluctuations in root growth. Despite these potential limitations, the ingrowth-core is used extensively because it is inexpensive and field-application is simple and straightforward. It is considered to be most effective in ecosystems with fast root growth such as the tropics (Vogt et al. 1998).   2.4.1.2.1.4 (Mini) rhizotrons Rhizotrons or minirhizotrons allow for fast and continuous in situ measurement of roots and root segments with moderate disturbance to sites (Taylor et al. 1990; Hendrick and Pregitzer, 1996; Vogt et al. 1998; Bernier and Robitaille, 2004). The two techniques are considered together because they both employ transparent interfaces, and the field applications involve similar processes, which include installation, root image collection, data processing, and conversion of root-growth parameters into  49 biomass production (Taylor et al. 1990; Madji, 1996; Milchunas, 2012; Girardin et al. 2013). Minirhizotron involves inserting clear tubes (usually small in size) into the ground at an angle, and a device such as camera is used to observe and capture root-growth at the soil-tube interface (Vogt et al. 1998; Roberti et al. 2014). Rhizotrons also have clear windows or screens installed on or both sides of soil profiles dug to a certain depth (Taylor et al. 1990; Fahey and Hughes, 1994; Metcalfe et al. 2007a), and fitted with materials such as black plastic sheets to mimic natural soil conditions. Root-growth is captured on the outer surface of the window or screen by either scanning or taking photos or manually tracing visible roots onto transparent sheets (Vogt et al. 1998; Janos et al. 2008; Girardin et al. 2013). Novel approaches that incorporate geographical information systems and remote-sensing techniques have been proposed to improve data capture and processing time (Silva, 2014). The root images or drawings are usually analyzed with software programs (Hendricks et al. 2006; Lobet et al. 2013). Lobet et al. (2013) identified 19 software programs that can be used to analyze root systems of plants. Since (mini) rhizotrons cannot directly produce FRB estimates, different approaches have been developed to convert root-growth data into mass production (Vogt et al. 1998; Bernier and Robitaille, 2004; Hendricks et al. 2006; Metcalfe et al. 2007a). The conversion used may, however, influence the accuracy of FRP estimates (Metcalfe et al. 2007a). Several studies have reported significantly greater FRP estimates from (mini) rhizotrons compared to other methods (e.g., Fahey and Hughes, 1994; Steele et al. 1997; Finér et al. 2011a). However, Moser et al. (2010) recorded higher FRP estimates for ingrowth-core compared to minirhizotron in Ecuadorian tropical montane forests.   The main disadvantage of rhizotron methods is the substantial cost involved in acquiring equipment and processing the root data. Hendrick and Pregitzer (1996) estimated the combined cost of minirhizotron and data processing equipment to be more than $US 15,000. Current costs would be substantially higher, which would be a major hindrance to many investigators. Installation of the minirhizotron tubes or rhizotron screens may change water regimes and soil matrix resistance to root penetration, which would affect the production estimates (Taylor et al. 1990; Hendricks et al. 2006). Root productivity can also be  50 under-estimated due to inappropriate scaling and calculation of soil volume sampled by minirhizotron images (Taylor et al. 2014). Root growth and mortality may also be stimulated along the tubes or screens of (mini) rhizotrons (Hendrick and Pregitzer, 1996; Milchanus, 2012). Moreover a large number of (mini) rhizotron tubes or screens may be required to properly capture root biomass distribution in highly heterogeneous sites.   2.4.1.2.1.5 Sequential-coring  Fine-root productivity can also be estimated by sampling soil cores at specific-time intervals for at least one year, and calculating the difference in biomass and necromass between the two periods (Vogt et al. 1998; Lauenroth, 2000; Olesinski et al. 2012). Each approach requires collecting root biomass, necromass or decomposition data independently or in combination before it can be applied to calculate FRP (Vogt et al. 1998). Among the three approaches the most widely used is the Minimum-Maximum formula, and the least is the Compartment Flow (Vogt et al. 1998; Hertel and Leuschner, 2002; Finér et al. 2011a; Brunner et al. 2013), which requires additional root-decomposition data to calculate FRP (Vogt et al. 1998). However the Compartment Flow formula may produce higher FRP estimates relative to the others (e.g., Hertel and Leuschner, 2002; Berhongaray et al. 2013). There are suggestions that the Minimum-Maximum formula may not be suitable for tropical ecosystems because of low seasonality in fine-root growth (Moser et al. 2010). New approaches such as forward, backward and continuous inflow estimates have also been developed to calculate FRP from sequential-coring data (Osawa and Aizawa, 2012; Van Do et al. 2016). The sequential-coring method can also be used to estimate mycorrhizal biomass and productivity (Vogt et al. 1998). The main disadvantage of sequential coring is that it may over- or under-estimate FRP (Hertel and Leuschner, 2002; Milchunas, 2012), because it assumes that fine-root biomass production and mortality occur asynchronously (Vogt et al. 1998; Lauenroth, 2000), although in many ecosystems the opposite trend is observed (e.g., Hendricks et al. 2006). Sequential-coring is also expensive and laborious, requiring a large number of soil-core replicates to produce good estimates.    51 2.4.1.2.2 Indirect methods  2.4.1.2.2.1 The Pipe model and others   FRB of individual trees and stand can be estimated indirectly with models including the pipe model (Shinozaki et al. 1964a; Niiyama et al. 2010). This is based on the pipe-model theory of tree form, which views a tree as an assemblage of pipe systems from the bottom to the top of the tree (Shinozaki et al. 1964a; 1964b). The contiguous nature of the pipes in a tree ensures that the cross-sectional area of an organ remains constant, even during branching events. The assumption of self-preservation in roots allows for the development of a simple model to predict root biomass (Carlson and Harrington, 1987; Richardson and zu Dohna, 2003). For in-depth discussion on the assumptions, theoretical justifications and calculations of the pipe model refer to Shinozaki et al. (1964a, 1964b) and Niiyama et al. (2010). The pipe model has rarely been used to determine FRB (Carlson and Harrington, 1987; Niiyama et al. 2010), although it is has been presented as a fast, non-destructive alternative approach to estimate FRB (Carlson and Harrington, 1987).   FRB can also be estimated from other models based on easily measurable aboveground metrics such as basal diameter, DBH, height and crown foliage. These models function on the basis of a strong relationship between FRB and aboveground variables at both tree and stand-levels (Kurz et al. 1996; Chen et al. 2004; Sun et al. 2015), although this may not apply to all sites (e.g., Helmisaari et al. 2007). Several models have been developed and used to predict FRB (e.g., Ammer and Wagner, 2002, 2005; Zerihun et al. 2007; Jurasinski et al. 2012). Models are a cost-effective means of determining FRB, particularly at large spatial scales (Kurz et al. 1996), at which aboveground biometric data may be the only available and dependable data available for most ecosystems. Chen et al. (2004) reported a good correlation between measured and model-estimated FRB of selected tree species from boreal and temperate forests. Unlike destructive methods, the accuracy of models does not depend on the spatial variation in root distribution in a stand (Lee et al. 2004). The use of models to predict FRB has some limitations, leading to uncertainties in the estimates they produce (Zerihun et al. 2007; Jurasinski et al.  52 2012). Models are unable to reflect the high temporal and spatial heterogeneity in FRB distribution common in most ecosystems (e.g., Ammer and Wagner, 2005; Helmisaari et al. 2007). Finally, some of the assumptions used to parameterize models may not hold true for all tree species and ecosystems (Ammer and Wagner, 2002; Lee et al. 2004).   2.4.1.2.2.2 N budget  The N-budget approach measures FRP as the product of annual N allocation to fine roots and the C: N ratio in fine roots (Nadelhoffer and Raich, 1992; Hendricks et al. 2006). This is based on the assumption that FRP can be predicted from N allocation since root production is largely driven by soil N (Nadelhoffer and Raich, 1992). To use this approach, complete data on N input, pools and fluxes at the ecosystem level must be available (Vogt et al. 1998; Hendricks et al. 2006). Several assumptions have to be satisfied to use the approach, including N limitation of plant growth (Vogt et al. 1998), which is not the case in many ecosystems (e.g., Cleveland et al. 2011). This technique has rarely been used to estimate FRP (see Finér et al. 2011a; Yuan and Chen, 2012b), and there are suggestions that its application should be limited to ecosystems with conditions similar to the N-poor sites for which it was developed (Vogt et al. 1998).     53 Table 2.2 Comparison of methods for estimating fine-root biomass and productivity using selected criteria.   Method Type Operational Requirements for optimum Ease of field Cost Labor Time  Accuracy  Impact on     principle  accuracy  application effectiveness requirements efficiency   ecosystem Fine-root biomass          Soil-core  Direct Equates root biomass to roots in a Large diameter corer/auger Simple  Costly Laborious Inefficient High Destructive    soil column preferably (> 10 cm) must be used       Monolith Direct Assumes root biomass is propor Good balance between monolith size Simple  Costly Laborious Inefficient High Destructive   tional to roots in soil block or   and number         monolith        Soil-pit Direct Refer to Table 2.1 Refer to Table 2.1 Difficult Costly Laborious Inefficient High Destructive Pipe model Indirect Based on the pipe-model theory Validation of model  Simple  Economical Less labor Efficient Low Non-destructive   of tree form        Allometric models Indirect Refer to Table 2.1 Validation of models Simple  Economical Less labor Efficient Low Non-destructive    for specific sites       Fine-root productivity          Ingrowth-core Direct Assumes that disturbances to roots Time and magnitude of root growth Simple  Economical Less labor Efficient Low Destructive   and soil do not affect root ingrowth for sites should be known        (Mini) rhizotrons Direct Images are captured, and converted Shorten observational intervals and  Simple  Costly Less labor Efficient High Non-destructive   to biomass production position tubes or plastic frames well      Sequential-coring Direct Assumes root production and Intervals between sampling dates must Simple  Costly Laborious Inefficient Low Destructive   mortality occur asynchronously  be shortened        N budget  Indirect Assumes that fine-root productivity N inputs, pools and fluxes must be deter Difficult Economical Less labor Efficient Low Destructive     is largely driven by soil N mined accurately                     54 2.4.2 Comparison of methods  2.4.2.1 Coarse-root biomass estimates from soil-pit and soil-core methods In the dataset of eleven sites evaluated during this study (Table A.1) estimates of mean CRB from soil-pit and soil-core methods were not significantly different (t = 0.06, p = 0.96, N = 11) (Figure 2.2A), and were weakly related (R2 = 0.32, p = 0.07, N = 11) (Figure 2.2B). However, in a study of a 6-year-old Eucalyptus plantation, Levillain et al. (2011) found slightly higher CRB estimates using soil-pits compared to soil-cores and concluded that auger coring was not a suitable method for estimating CRB. In contrast, Smith et al. (2013) reported a significant positive correlation (R2 = 0.40, p < 0.001) between CRB estimates from 30 × 30 × 30 cm soil pits and 8-cm-diameter soil cores in a CO2-enrichment study in the UK. The different results observed in these aforementioned studies could be explained by differences in the tree planting densities. Smith et al. (2013) used a planting density of 15,000 stems ha-1 to expedite tree ecological interactions whereas a density of 800 stems ha-1 was used by Levillain et al. (2011). The high planting density could have reduced the spatial heterogeneity of coarse roots.     Figure 2.2 Coarse-root (> 2 mm) biomass estimates (Mg ha-1) (mean ± SE) from soil-pit and soil-core methods (A), and relationship between coarse-root biomass estimates (Mg ha-1) (mean ± SE) from soil-pit and soil-core methods (B, N = 11). Broken line is 1:1 relationship between the methods. Data were derived from the same sites, and relationship values are log-transformed (base 10).    Soil-pit Soil-core0.00.51.01.52.02.5Coarse-root biomass (Mg ha-1)NSA0.0 0.5 1.0 1.50.00.51.01.5Soil-pitSoil-corelog10y = 0.63 x log10 x + 0.19R2 = 0.32p = 0.07B 55 Spatial heterogeneity of coarse root distribution and sampling strategies are major confounding factors when estimating in situ CRB. Estimates produced by the aforementioned techniques (see section 2.4.1.1) can vary widely and can be altered by planting density, site conditions, tree allometry, tree developmental age, site specific environmental conditions, or species specific differences (Resh et al. 2003). Indeed, the most detailed spatial sampling designs rarely capture the fact that the majority of coarse roots are concentrated below the tree stem potentially resulting in a significant under-estimate of CRB. As suggested by Resh et al. (2003) the analysis supports the assertion that, where destructive harvesting permits, the lowest cost method to achieve a high accuracy CRB estimation is a combination of soil coring and root ball excavation. Interpretation of coarse root data must include a consideration of the sampling technique used to ensure that under-estimates of CRB are not biased by techniques that do not account for the high concentration of coarse roots below the tree bole.   2.4.2.2 Fine-root biomass estimates from soil-pit and soil-core methods The mean FRB estimate from the dataset in Table A.1 was higher for soil cores (5.17 ± 0.93 Mg ha-1) than the estimate from soil pits (4.31 ± 0.08 Mg ha-1), but the difference was not significant (t = 0.601, p = 0.556, N = 9) (Figure 2.3A). Many studies at single sites have reported higher FRB estimates from the soil-core compared to the soil-pit method (e.g., Millikin and Bledsoe, 1999; Park et al. 2007 and references herein). There was a significant positive correlation between FRB estimates obtained from soil-core and soil-pit methods (Figure 2.3B), which may indicate the utility of both methods to quantify FRB. Some investigators (Bledsoe et al. 1999; Park et al. 2007) recommend the use of the soil-core method to quantify FRB, due to its simplicity and other advantages, discussed in section 2.4.1.2.1.4.     56 Figure 2.3 Fine-root (≤ 2 mm) biomass estimates (Mg ha-1) (mean ± SE) (N = 9) from soil-pit and soil-core methods (A), and relationship between fine-root biomass estimates (Mg ha-1) from soil-pit and soil-core (B, N =9). Broken line is 1:1 relationship between the methods. Data were derived from the same sites.  2.4.2.3 Fine-root productivity estimates from ingrowth-core, (mini) rhizotrons and sequential-coring methods and biome estimates  The mean FRP estimate obtained from studies in Table A.2 was significantly lower in the ingrowth-core method (2.04 ± 0.21 Mg ha-1 year-1, N = 81) compared to estimates provided by sequential-coring (3.70 ± 0.82 Mg ha-1 year-1, N = 67) and (mini) rhizotrons (3.81 ± 0.46 Mg ha-1 year-1, N = 26) methods (Figure 2.4A). The same pattern was observed when the methods were compared separately for the major biomes (Table 2.3). Earlier reviews and single studies that compared the three methods also reported lower FRP estimates from ingrowth-core than from other methods (e.g., Vogt et al. 1998; Hendricks et al. 2006; Moser et al. 2010; Finér et al. 2011a). Finér et al. (2011a) reported higher FRP estimates for sequential-coring and (mini) rhizotrons methods than the ingrowth-core method, although there were no significant differences in FRP estimates among the three methods at the stand level (Finér et al. 2011a). Similarly, in a global study that assessed FRP and C allocation in forest ecosystems, Nadelhoffer and Raich (1992) reported that sequential cores yielded higher values of FRP than ingrowth cores.    Soil-pit Soil-core02468Fine-root biomass (Mg ha-1)NSA0 2 4 6 8 100246810Soil-pitSoil-corey = 0.82x + 1.65R2 = 0.91p < 0.0001B 57   Figure 2.4 Fine-root productivity estimates (Mg ha-1 year-1) (mean ± SE) from ingrowth-core (N = 81), (mini) rhizotrons (N = 26) and sequential-coring (N = 67) methods (A), and fine-root productivity estimates (Mg ha-1 year-1) (mean ± S.E) of tropical (N = 52), temperate (N = 44) and boreal (N = 79) forests. Data represent all observations used in the study. Different letters represent significant differences (p < 0.05).   Table 2.3 Fine-root productivity estimates (Mg ha-1 year-1) (mean ± SE) from ingrowth-core, (mini) rhizotrons and sequential-coring methods for different biomes.  Biome Ingrowth-core (Mini) rhizotrons Sequential-coring Tropical 3.92 ± 0.37a 5.30 ± 0.59a 8.29 ± 3.30a Temperate 1.87 ± 0.44a 2.94 ± 0.53a 3.19 ± 0.80a Boreal 0.97 ± 0.10b 1.94 ± 0.52a 1.90 ± 0.23a Different letters represent significant differences (p < 0.05) among the methods.   There was a significant positive relationship between FRP estimates obtained from sequential-coring and ingrowth-core (R2 = 0.29, p < 0.0001, N = 66), and ingrowth-core and (mini) rhizotrons (R2 = 0.44, p = 0.0003, N = 25) methods (Figure 2.5A and 2.5B). However, the sequential-coring and (mini) rhizotrons methods were moderately related (R2 = 0.37, p = 0.05, N = 11, Figure 2.5C). Yuan and Chen (2012c) reported similar patterns of FRP between sequential-coring and ingrowth-core methods in a boreal mixed forest in Ontario, Canada, but others have reported contrasting relationships between sequential-coring and other methods (e.g., Steele et al. 1997; Hendricks et al. 2006; Moser et al. 2010). The methods Ingrowth-core (Mini) rhizotrons Sequential-coring012345Fine-root productivity (Mg ha-1 yr-1)ab bATropical Temperate Boreal02468Fine-root productivity (Mg ha-1 yr-1)abbB 58 explained less than 50% of the variation in FRP, indicating that other factors, including species, site conditions, sampling depth, fine-root size classification, resource availability, stand and environmental conditions also influence FRP (Rytter, 1999; Makkonen and Helmisaari, 2001; Ostonen et al. 2005; Hendricks et al. 2006; Finér et al. 2011a; Yuan and Chen, 2010; 2012b; Smith et al. 2013).      Figure 2.5 Relationship between fine-root productivity estimates (Mg ha-1 year-1) from sequential-coring and ingrowth-core (A, N = 66), ingrowth-core and (mini) rhizotrons (B, N = 25), and sequential-coring and (mini) rhizotrons (C, N = 11) compared at the same sites. Values of fine-root production are log- transformed (base 10). Broken line is 1:1 relationship between the methods.  FRP estimates significantly differed among the biomes (p < 0.0001, F2,172 = 15.04). Mean FRP was significantly higher in tropical forests (5.53 ± 1.00 Mg ha-1 year-1, N = 52) than in temperate (2.56 ± 0.41 Mg ha-1 year-1, N = 44) and boreal (1.46 ± 0.13 Mg ha-1 year-1, N = 79) forests (Figure 2.4B). This corresponds well with the FRP estimates reported by Finer et al. (2011a) for the three biomes. Fine-root productivity is high in tropical forests compared to the other biomes because ecosystem productivity is highest in these forests (Keeling and Phillips, 2007; Vicca et al. 2012). Given the strong relationship between Gross Primary Productivity and belowground productivity (Litton et al. 2007), it is expected that tropical forests would have high root productivity. In addition, FRP is mostly high in sites where nutrient availability is low (Vogt et al. 1996; Litton et al. 2007; Fernández-Martínez et al. 2014). Compared with temperate and boreal forests, tropical forests have low concentrations of soil nutrients, particularly phosphorus and potassium (Fernández-Martínez et al. 2014; Cusack et al. 2016).  0.0 0.5 1.0 1.5 2.0 2.50.00.51.01.52.02.5Sequential-coringIngrowth-corelog10 y = 0.5065 x log10x + 0.3468R2 = 0.29p < 0.0001A0.0 0.5 1.0 1.5 2.00.00.51.01.52.0Ingrowth-core(Mini) rhizotronslog10y = 0.42 x log10 x + 0.83R2 = 0.44p = 0.0003B0.0 0.5 1.0 1.5 2.0 2.50.00.51.01.52.02.5Sequential-coring (Mini) rhizotronslog10y = 0.21 x log10x + 1.05R2 = 0.37p = 0.05C 59 2.5 Conclusions  There is currently no consensus as to how best to estimate root biomass and productivity. Many investigators prefer root-excavation and soil-pit methods to estimate CRB, despite the high cost and labor required. GPR is a promising indirect approach to quantify CRB, but may not suitable for ecosystems with a dense understory and soils with high organic matter and ion contents. Empirical models are widely used to predict fine- and coarse-root biomass and productivity in C studies. Comparative studies have shown a consistent relationship between FRB estimates from the soil-pit and soil-core methods. For FRP, the ingrowth-core method consistently provided lower estimates than other methods. Based on the reviewed literature and comparative analysis the following are proposed: (i) the soil-pit method should be employed to estimate coarse-root biomass because it can serve as a compromise between cost and efficiency; (ii) (mini) rhizotrons should be favored over the other methods to estimate fine-root productivity; (iii) where cost and site conditions (e.g., in stony or on steep slopes) preclude the use of (mini) rhizotrons, the sequential-coring and ingrowth-core methods are suitable; (iv) the ingrowth-core method should be used with caution in sites where root growth is influenced by strong seasonal fluctuations, and when used, the period between cores/nets installation and root sampling should be extended to allow for maximum root colonization; and (v) multiple methods should be employed to estimate fine-root productivity, and more comparative studies of different methods on the same sites are needed.    60 Chapter 3: Root exploitation strategies differ in tropical old-growth forest and logged-over forest in Ghana1 3.1 Synopsis   Studies on post-logging recovery have rarely linked root biomass and morphological traits to the exploitation strategies of plant roots. I quantified root biomass, morphological traits and root exploitation strategies in an old-growth forest and in a 54-year-old logged-over forest influenced by similar parent material and climate. Fine (diameter < 2 mm) roots were sampled using the soil-core method to determine biomass and root traits. Seven morphological traits were considered: three associated with resource exploitation potential or an ‘extensive’ strategy (length, surface area and volume); and four traits which reflect exploitation efficiency or an ‘intensive’ strategy (root tip number, specific root area, specific root length and root tissue density). I found that fine-root biomass, root length, surface area, volume, and root tissue density were higher in the logged-over forest, whereas the old-growth forest had higher root tips/m root length, fine root specific root length and specific root surface area than the logged-over forest. I also observed divergent resource exploitation strategies between the forests. Plants in the old-growth forest also produced thinner roots, which can increase resource uptake efficiency and competitiveness. In contrast, plants in the logged-over forest had thicker roots, which are associated with greater resource conservation. The different strategies exhibited in these forests could partly reflect differences in stand structure and functional composition.        1A version of this chapter is accepted pending revision: Addo-Danso SD, Prescott CE, Guy RD, Duah-Gyamfi A, Moore S, Owusu-Afriyie K, Marshall P, Forrester DI, Adu-Bredu S, Malhi Y. Root exploitation strategies differed in tropical old-growth and logged-over forests in Ghana. Biotropica.     61 3.2 Introduction Plant roots are key components of forest ecosystems constituting about 20-30% of the total biomass (Brunner and Godbold, 2006). Roots provide anchorage, and are critical for the exploitation of water and nutrients from soil to aboveground tissues (Eissenstat, 2000; Fitter, 2002). In addition roots play major roles in the cycling and allocation of carbon (C) and nutrients in forest ecosystems (Hobbie et al. 2010; Prescott, 2010; Malhi et al. 2011). Soil C derived from roots and root-associated microbial biomass is considered a more stable input than that from aboveground components due to roots interaction with soil aggregate and clay minerals (Prescott, 2010). Between 22 and 75% of total forest productivity is allocated to roots and their symbionts (Malhi et al. 2011; McCormack et al. 2015). This significant flux of C belowground may influence root architecture and branching patterns (Nielsen et al. 1994; Thaler and Pagès, 1998), with concomitant effects on uptake of soil resources, belowground productivity and biogeochemical cycling (Bardgett et al. 2014).   Plants employ a range of strategies to capture water and nutrient resources, and to respond to changes in soil resources, mediated partly by the morphological traits of roots (Ostonen et al. 2007a; Hodge, 2009; Paz et al. 2015). Root traits associated with root quantity usually reflect exploitation potential or an ‘extensive’ strategy (Berntson, 1994; Lõhmus et al. 2006; Xiang et al. 2013). This strategy is represented by the quantity and spatial distribution of length, surface area, volume or biomass in the soil. Root functional traits, such as specific root length (SRL-root length per unit dry mass), specific root area (SRA-root surface area per dry mass), and root tissue density (RTD-root dry mass per volume) describe exploitation efficiency or an ‘intensive’ strategy (Fitter et al. 1991; Bauhus and Messier, 1999; Löhmus et al. 2006; Xiang et al. 2013). Plants associated with exploitation efficiency explore a large volume of soil per unit root biomass in order to capture soil resources.   Tropical forest structure and plant functional composition change during post-logging succession (Chazdon, 2014; Gatti et al. 2015; Vaglio Laurin et al. 2016), which may also induce shifts in root  62 dynamics (Leuschner et al. 2009; Barbhuiya et al. 2012). Most studies that have quantified root dynamics during tropical forest succession have focused on fine-root biomass (e.g., Hertel et al. 2007; Ibrahima et al. 2010; Barbhuiya et al. 2012; Gautam and Mendel, 2012), and a few fine-root morphological traits (Hopkins et al. 1996; Leuschner et al. 2009; Hansson, 2014). These studies have mainly reported higher fine-root biomass, but lower values for fine-root traits (SRL and SRA) in old-growth (unlogged) forests than in logged forests (Hopkins et al. 1996; Leuschner et al. 2009). For example, in a global study that compared fine-root (diameter < 2 mm) biomass in old-growth forests and logged tropical moist forests, Hertel et al. (2007) reported significantly higher fine-root biomass estimates in old-growth forests than logged forests.  By contrast, Hopkins et al. (1996) and Leuschner et al. (2009) reported higher fine-root (diameter < 2 mm) SRL and SRA in logged forests than old-growth forests in Australia and Indonesia respectively, which they attributed to the lower tissue densities of the fast-growing, light-demanding tree species in the logged forests. To determine if changes in root biomass and morphological traits during post-logging recovery reflect a potential shift in soil exploitation strategies of plants it is necessary to not only compare root traits associated with resource efficiency, as these studies did, but to also examine traits that depict the potential for resource exploitation. Understanding the changes in root exploitation strategies, and the associated trade-offs during post-logging recovery is important for predicting long-term C dynamics in tropical forests (Fischer et al. 2016). Here I addressed the following questions: (1) Do root biomass estimates and root morphological traits differ between an old-growth forest and a logged-over forest? (2) Do plant roots in an old-growth forest and a logged-over forest differ in their apparent strategies for exploiting soil resources?   I combined the soil-core method (Vogt et al. 1998), and WinRHIZO image analysis software to quantify root biomass and morphological traits in an old-growth forest and a nearby 54-year-old logged-over forest, which are influenced by similar parent material and climatic conditions. I considered seven root morphological traits: three associated with resource exploitation potential (length, surface area and  63 volume); and four, which reflect exploitation efficiency (root tip number, specific root area, specific root length and root tissue density) (Berntson, 1994; Lõhmus et al. 2006; Xiang et al. 2013).   3.3 Materials and Methods 3.3.1 Study area The study was conducted in the Bobiri Forest Reserve, situated in the Ejisu-Juabeng District of the Ashanti Region, in southern Ghana. The Bobiri Forest Reserve (latitude 6°44’N and longitude 1°23’W) covers an area of ca. 5,504 ha, and lies in the moist semi-deciduous forest zone (Hall and Swaine, 1981). The Bobiri forest was demarcated in 1936, and reserved in a pristine, unexploited state in 1939 (Foggie, 1947). The reserve lies on gently undulating terrain with a dominant slope of 6-7 percent; altitude between 183 m and 280 m a.s.l. The general slope is from northwest to southeast, and six streams flow in this direction to drain the reserve (Foggie, 1947). The 10-year mean monthly minimum and maximum temperatures range between 22.2 °C and 31.2 °C (Forestry Research Institute of Ghana, unpublished data). The mean temperature of the coolest month (August) ranges from 21.8 °C to 27.4 °C and for the hottest month (February) from 22.4 °C to 33.8 °C. The 10-year annual rainfall ranges from 1210 to 1800 mm, with a dry season lasting from December to mid-March. During the dry season, mean rainfall is less than 100 mm. June and September are the wettest months of the year in the area, with mean precipitation of ca. 246 and 209 mm per month, respectively (Forestry Research Institute of Ghana, unpublished data). The relative humidity averages 55% in the afternoon and 85% at night.   The Bobiri forest has deeply weathered, well-drained soils with a clay-enriched, illuvial subsoil developed on granite (Foggie, 1947). The soil has been described as Forest Ochrosol type (Hall and Swaine, 1981), which is the same as Acrisol for the FAO soil classification or Ultisol for the US soil classification system (Adjei-Gyapong and Asiamah, 2000; IUSS Working Group WRB, 2015). Soil texture varies from sandy loam to clay loam, passing into a grey leached sandy or silty soil in the periodically flooded river valleys, flats and swamps (Foggie, 1947). The structure of the forest is typical  64 of a tropical moist forest type, with the upper canopy layer consisting of a mixture of deciduous and evergreen species in approximately equal proportions (Hall and Swaine, 1981). Canopy height ranges from 30-50 m, with emergent trees up to 60 m tall.   3.3.2 Study forests  The Bobiri Forest Reserve has been compartmentalized and divided into four main blocks based on the designated use: research, butterfly sanctuary, protection old-growth forest (strict-nature forest) and production forest. For this study, plots were set up in the forests of the research block and the protection strict-nature block. The research forest (hereafter referred to as the ‘logged-over’ forest) covers an area of 64 ha. In 1959 African Woods Limited selectively harvested 172 trees mostly mahogany species (Forestry Department, 1958; Djagbletey, 2014), but no data exist on sizes of individual trees removed during the logging operation. No logging has been allowed after the first entry, but the evidence of old tracks is still noticeable in the forest. Prior to logging, part of the forest was placed under the Tropical Shelterwood Silvicultural System in 1955, where lianas and over-matured trees as well as defective trees were poisoned with sodium arsenite to allow desirable species to regenerate and grow (Forestry Department, 1958). The strict-nature forest is ca. 23 ha, and was described as an old-growth forest because of certain structural and compositional attributes, including: (1) multi-layered canopy structure, (2) different gap sizes, (3) the understory is composed of both shade tolerant and shade-intolerant species, (4) large coarse woody debris in all decay stages on the forest floor, and (5) the presence of ferns (Wirth et al. 2009; Chazdon, 2014). No commercial logging has been allowed in the protection (old-growth) forest, but the forest may have been influenced by past and present disturbances or climate extremes (Forestry Department, 1958; Shanahan et al. 2009; Fauset et al. 2012), which are common in most tropical mature forests (Chazdon, 2014). The two study forests are separated by about 3 km, but are similar in all factors related to energy budget, moisture and soil fertility. Detailed descriptions of the two study forests are provided in Table 3.1.    65 Table 3.1 Site, soil and stand structural characteristics of the two study forests in the Bobiri Forest Reserve in Ghana. Parameter Old-growth forest Logged-over forest Coordinates 6.7°N 1.3’W 6.9°N 1.3’W Elevation (a.s.l.)          268          276 Aspect          NW         NW Soil type Sandy loam Sandy loam Sand (%)   59.70 ± 1.85   58.30 ± 2.85 Silt (%)   17.60 ± 3.76   18.10 ± 1.60 Clay (%)   22.70 ± 2.70   23.60 ± 4.36 Bulk density (g cm-3)     0.66 ± 0.09     0.42 ± 0.06 Organic matter (%)     6.94 ± 0.94     6.79 ± 1.02 Organic C (%)     4.02 ± 0.55     3.94 ± 0.59 Total N (%)     0.40 ± 0.06     0.39 ± 0.07 Available P (Bray) (mg kg-1)   18.55 ± 1.76   15.26 ± 1.31 Available K (mg kg-1) 120.52 ± 8.96 109.09 ± 8.70 C:N ratio   10.05 ± 0.85   10.10 ± 0.49 K:P ratio     6.50 ± 1.75     7.15 ± 0.37 pH (H2O)     5.60 ± 0.06     5.90 ± 0.03 Base saturation (%)*   89.95 ± 1.74   98.93 ± 0.2 Mean DBH (> 10 cm)   21.50 ± 0.61   19.40 ± 0.43 Mean height (> 10 m)   17.80 ± 1.75   16.50 ± 0.19 Stand density (per ha)          475          755 Basal area (m2 ha-1)*    23.81± 0.26   30.58 ± 0.88 APAR (%)   92.18 ± 3.07   94.90 ± 0.26 Soil parameters are based on samples from 0-30 cm soil depth. Values are the mean and standard error (SE) of samples. *Parameters that significantly differ between study forests (p < 0.05). Initials refer to parameters as follows: a.s.l., Above sea level; N, nitrogen; P, phosphorus; K, potassium; APAR, absorbed photosynthetically active radiation; DBH, diameter at breast height.      66 3.3.3 Root sampling and processing  3.3.3.1 Root biomass In April 2015 (i.e. early rainfall season) roots were sampled using the soil-core method (Vogt et al. 1998; Addo-Danso et al. 2016). Samples were taken with a hand-driven soil auger (diameter 5.5-cm) to 30-cm depth in twelve plots (measuring 10 m x 10 m) randomly established across the study forests. Root samples were taken at three randomly located points within each plot. Roots were sampled to 30-cm depth because studies in the tropics, including Ghana, have found that roots are highly concentrated in this layer (e.g., Lawson et al. 1970; Metcalfe et al. 2008). Although some tropical trees may have deep roots (e.g., Huttel, 1975; Freycon et al. 2015), I assumed that the selected sampling depth would have a large fraction of the total root mass. During soil sampling, when an obstacle such as a large structural root or stone obstructed the auger, it was relocated within an area of ca. 25 cm2 until a suitable core was extracted.   Soil samples (N = 216) were kept separate for three depth intervals (0-10, 10-20, 20-30 cm), and transferred into plastic bags for laboratory processing at the Forestry Research Institute of Ghana, ca 18 km from the study site. Samples were refrigerated at 4 °C for 7 days before processing. Samples were soaked in plastic bowls, and washed thoroughly using a 0.25-mm sieve to remove soil particles and debris. Collected roots were separated into fine roots (diameter < 2 mm), and into live (biomass) and dead (necromass) based on visual inspection of morphological features such as colour, tensile strength, and cortex and periderm characteristics (Vogt et al. 1998; Leuschner et al. 2009). Live roots are usually light in colour (may not be white), not easily broken, turgid, and the cortex and periderm not easily separated. Dead roots are generally dark or brown, brittle and shrivelled with a non-turgid cortex, although this can be ambiguous when roots are not completely decomposed (Vogt et al. 1998). No attempt was made to separate roots into individual species or into trees and understory vegetation. Live roots were oven-dried at 60 °C to constant mass, and weighed to determine root biomass. Root biomass was expressed in g m-2, and later scaled to per ha as Mg ha-1.  67 3.3.3.2 Root morphology  Seven morphological traits were considered: three associated with resource exploitation potential (length, surface area and volume); and four that reflect exploitation efficiency (root tip number, specific root area, specific root length and root tissue density) (Fitter et al. 1991; Lõhmus et al. 2006; Børja et al. 2008; Xiang et al. 2013). Live roots were spread in a water-filled, transparent plastic tray and scanned using a high-resolution Epson Perfection v700 Photo/V750 Pro scanner, USA. Digital images were analyzed using WinRHIZO Basic 2013a software (Régent Instruments Inc., Québec, Canada) to estimate root length, surface area, volume, and number of root tips. After scanning, root fresh mass was measured and then the roots were oven-dried at 70 °C to determine their dry mass. Root length, surface area and volume data were expressed on a m2 ground surface area basis.   Root traits, including root tip number, specific root length (SRL), specific root area (SRA) and root tissue density (RTD) were calculated using the root dry mass and the estimated basic root traits. Root tip number (hereafter root tips/m root length) was calculated as the number of root tips divided by root length. Specific root length (m g-1) was calculated as root length divided by root dry mass, and specific root area (cm2 g-1) was obtained by dividing the root area by root dry mass (Ostonen et al. 2007b; Leuschner et al. 2009). Root tissue density (g cm-3) was calculated as root dry mass divided by the root volume (Eissenstat et al. 2015).   3.3.4 Statistical analysis I tested root biomass and morphological traits for normality by using a Shapiro-Wilk test. The majority of the estimates was not normally distributed, and so were log transformed. A student t-test was used to test differences in root biomass and morphological traits between the old-growth forest and the logged-over forest. I compared the differences in root biomass and morphological traits among soil depths using one-way Analysis of Variance (ANOVA). I used a post-hoc Tukey’s HSD test for multiple pairwise comparisons. Root biomass and morphological traits were also subjected to two-way analyses of variance  68 (ANOVA) with forest type (old-growth and logged-over forests) and soil depth (0-10 cm, 10-20 cm and 20-30 cm) as the main effects. All analyses were performed with Stata version 12.0 (Stata Corp LP, USA) and GraphPad Prism 7 (GraphPad Software, Inc., California, USA) software packages, with a significance level of p < 0.05.   3.4 Results 3.4.1 Root biomass and distribution  Fine-root biomass in the 0-30 cm soil depth was significantly higher in the logged-over forest (3.04 ± 1.10 Mg ha-1) than in the old-growth forest (1.42 ± 0.29 Mg ha-1) (p < 0.0001, Table 3.2). Fine-root biomass decreased with increasing soil depth in both forests (Figure 3.1). Fine-root biomass was highly concentrated in the upper 10 cm of soil, comprising ca. 53% and 43% of the total fine-root biomass in the old-growth forest and the logged-over forest, respectively. Root biomass was not affected by the interaction between forest type and soil depth (results not shown).    Table 3.2 Fine-root (diameter < 2 mm) biomass estimates (Mg ha-1, N = 12) in 0-30 cm soil depth in the old-growth forest and the 54-year-old logged-over forest.  Parameter Old-growth forest Logged-over forest Mean 1.42 ± 0.29a 3.04 ± 1.10b Total 51.25 109.5 CV (%) 35.84 62.69 Different lower-case letters indicate significant difference between forests (p < 0.05, N = 108). CV, coefficient of variation.    69         Figure 3.1 Fine root biomass estimates (Mg ha-1) to 30 cm soil depth in old-growth forest (black bars) and 54-year-old logged-over forest (grey bars). Different letters represent significant differences (p < 0.05, N = 36) among soil depth.   3.4.2 Root morphological traits and distribution  In both forests, fine-root traits representing resource exploitation potential (length, surface area and volume) were significantly higher in the upper most (0-10 cm) soil layer, and declined with depth (Figure 3.2A-C). In the logged-over forest, mean fine-root surface area declined with depth from 5.57 ± 0.46 m2 m-2 in the upper most soil layer to 2.55 ± 0.26 m2 m-2 in the 20-30 cm layer. In the old-growth forest, mean fine-root volume also declined from 420.8 ± 45.88 cm3 m-2 in the 0-10 cm layer to 231.7 ± 42.86 cm3 m-2 in the 20-30 cm soil layer. Fine-root length exhibited similar distribution patterns with soil depth in both forests, declining from the 0-10 cm to the 10-20 cm layer, and decreasing from the 10-20 cm layer to the 20-30 cm layer (Figure 3.2A). In contrast, fine-root traits associated with resource exploitation efficiency, except for RTD, increased with soil depth in both forests (Figure 3.2D-G). For example, mean fine-root SRL in the old-growth forest increased from 37.2 ± 5.85 m g-1 in the 0-10 cm soil layer to 51.75 ± 5.28 m g-1 in the 20-30 cm soil layer. Mean fine-root SRA in the logged-over forest was significantly higher in the 20-30 cm soil layer (Figure 3.2E). All fine-root traits, except root tip/m root length, SRL and  Fine-root biomass (Mg ha -1)0.0 0.5 1.0 1.5 2.0 2.520-30 10-20 0-10 Soil depth (cm)abbOld-growthLogged-overxyy 70 SRA differed between the forest types (Table 3.3). In addition to the forest type effect, soil depth significantly affected all fine-root traits except root tip/m root length. The interactions between forest type and soil depth affected only fine-root length, surface area and volume.   Table 3.3 Significant effects of, and interactions between forest type (old-growth forest and 54-year-old logged-over forest) and soil depth (0-10 cm, 10-20 cm, 20-30 cm) on fine-root morphological traits.  Morphological traits Forest type  Soil Depth  Forest type x Soil Depth F P-value   F P-value   F P-value Length (m m-2) 71.67 < 0.0001  31.58 < 0.0001 6.73 0.0150 Surface area (m2 m-2) 55.82 < 0.0001  36.77 < 0.0001 7.02 0.0011 Volume (cm3 m-2) 34.17 < 0.0001  37.18 < 0.0001 6.73 0.0015 Root tip/m root length 0.33 0.5662  14.06 < 0.0001 0.21 0.8070 SRA (cm2 g-1) 3.43 0.0653  6.01 0.0029  0.89 0.4102 SRL (m g-1) 1.19 0.2774  8.18 0.0004  0.73 0.4855 RTD (g cm-3) 7.00 0.0088  3.65 0.0277  0.22 0.8033 Data were analyzed by two-way ANOVA. Significant effects are denoted as p < 0.05, significantly different, and NS, Not significant (p > 0.05). SRA, specific root area; SRL, specific root length; RTD, root tissue density.                   71  Figure 3.2 Vertical distribution of fine-root morphological traits (A-G) to 30 cm soil depth. Data are mean ± SE for old-growth forest (black bars) and 54-year-old logged-over forest (grey bars). Different letters represent significant differences (p < 0.05) among soil depth. SRA, specific root area; SRL, specific root length; RTD, root tissue density.  72 3.4.3 Root exploitation strategies  Root morphological traits representing resource exploitation potential, including root length, root surface area and root volume were significantly higher in the logged-over forest than in the old-growth forest (Table 3.4). For instance, root length and surface area was 49% and 46% higher in the logged-over forest than in the old-growth forest. Except for RTD, fine-root traits that reflect resource exploitation efficiency were higher in the old-growth forest (Table 3.4). For example, total fine-root SRA was 15% higher in the old-growth forest than in the logged-over forest. The coefficient of variation values for traits related to fine-root exploitation potential was higher in the old-growth forest than in the logged-over forest (data not shown).   Table 3.4 Fine-root (diameter < 2 mm) morphological traits (mean ± SE, N = 12) at 0-30 cm soil depth in the old-growth forest and in the 54-year-old logged-over forest. Morphological traits        Old-growth forest         Logged-over forest P-value Length (m m-2) 1066.00 ± 102.80  2102.00 ± 172.60 < 0.0001 Surface area (m2 m-2) 5.88   ± 0.82  10.89 ± 0.87 0.0002 Volume (cm3 m-2)   897.60 ± 123.00  1507.00 ± 118.40 0.0008 Root tips/m root length 858.30 ± 47.43  489.50 ± 30.95 < 0.0001 SRA (cm2 g-1) 769.70 ± 54.01  652.20 ± 51.95 0.05 SRL (m g-1) 44.75 ± 3.27  40.27 ± 3.43 0.4008 RTD (g cm-3)   1.12 ± 0.09    1.46 ± 0.13 0.0338 SRA, specific root area; SRL, specific root length; RTD, root tissue density. Significant differences are denoted as p < 0.05.   3.5 Discussion 3.5.1 Root biomass and distribution  The high fine-root biomass in the upper soil surface is a common observation in all forest biomes (Finér et al. 2011b), and it is thought to be important for nutrient conservation in tropical forests (Cavelier, 1992; Zangaro et al. 2014). The mean fine-root biomass estimates (1.42-3.04 Mg ha-1) were lower than  73 estimates from other tropical moist forests in Ghana (2.80-10.20 Mg ha-1, Greenland and Kowal, 1960; Lawson et al. 1970; Jenik et al. 1971) and other parts of Africa (1.70-24.20 Mg ha-1, Huttel, 1975; Ibrahima et al. 2010; Ifo et al. 2015), but within the range reported for single sites in Asia (0.50-7.80 Mg ha-1, Sundarapandian and Swamy, 1996; Gautam and Mandal, 2012) and the Neotropics (0.08-39.50 Mg ha-1, Cavelier, 1992; Espeleta and Clark, 2007; Hertel and Leuschner, 2011; Zangaro et al. 2012; 2014). Large variation in root biomass in tropical moist forests is recognized in the literature (e.g., Cavelier, 1992), reflecting the influences of several biotic and abiotic factors on root biomass distribution (Vogt et al. 1996; Finér et al. 2011b). Other factors that may account for the large variation in root biomass estimates reported for tropical forests include sampling method, sampling depth, sampling period and the root diameter threshold used (Finèr et al. 2011b; Addo-Danso et al. 2016).   The higher mean fine-root biomass in the logged-over forest than in the old-growth forest could be due to the higher stand density and basal area of the logged-over forest (Table 3.1). This result agrees with some studies, which reported greater fine-root biomass in forest stands with higher density and tree basal area (Chen et al. 2004; Harteveld et al. 2007; Finér et al. 2011b; Lin et al. 2015). Stand density was found to be a major contributor to the variation in fine-root (diameter < 2 mm) biomass during post-logging recovery in subtropical forests in China (Lin et al. 2015). At the local, regional and global scales, fine-root biomass of trees and stands increased with stand basal area (Chen et al. 2004; Harteveld et al. 2007; Finér et al. 2011b; Lehtonen et al. 2016). Indeed, stand density and basal area are good predictors of fine-root biomass in some forest ecosystems (Addo-Danso et al. 2016; Lehtonen et al. 2016).   Contrary to the findings of this study, comparative studies in the tropics have usually reported higher fine-root biomass in old-growth (unlogged) forests than in logged forests (Hopkins et al. 1996; Harteveld et al. 2007; Hertel et al. 2007; Leuschner et al. 2009; Ibrahima et al. 2010; Barbhuiya et al. 2012; Gautam and Mandal, 2012). In a global study that compared fine-root (diameter < 2 mm) biomass in old-growth forests and logged tropical moist forests, Hertel et al. (2007) reported significantly higher fine-root  74 biomass estimates for old-growth forests than logged forests. Gautam and Mandal (2012) also reported higher fine-root (diameter < 2 mm) biomass in unlogged forest than in a selectively logged forest in Sunsari district, eastern Nepal. However, in broad-leaved evergreen subtropical forests in southeast China, Lin et al. (2015) reported a higher fine-root biomass estimate for a 50-year-old logged forest than in an old-growth forest. Moreover, fine-root (diameter < 1 mm) biomass was significantly higher in a highly disturbed forest than in an undisturbed wet evergreen forest in the Dibrugarh District of eastern Assam, India (Barbhuiya et al. 2012). The conflicting results from the aforementioned studies may be due to differences in logging intensity, soil characteristics, stand density and functional composition, and the time elapsed between logging events and studies (Espeleta and Clark, 2007; Hertel et al. 2007; Chazdon, 2014). For instance, increased logging intensity may lead to a reduction in fine root biomass (Hertel et al. 2007). With the exception of two studies (Hopkins et al. 1996; Ibrahima et al. 2010), all the others were conducted in Asia. Logging intensity is usually high in Asia (ca. 8-15 trees removed per ha) (Chadzon, 2014), which may account for the patterns observed in those studies. In fact, a study in a rainforest in Cameroon, reported that fine-root biomass recovered 7-years after logging, which was attributed to the low number of trees removed (1-2 trees per ha) during the logging operations (Ibrahima et al. 2010). Moreover, most of the studies mentioned above were carried out within 10 years after logging. However, previous studies show that fine-root biomass may increase with stand age as the forest canopy cover, stem density and basal area recover from logging (Hertel et al. 2007; Leuschner et al. 2009; Lin et al. 2015). Soil factors are known to affect fine-root biomass estimates in tropical forests (e.g., Espeleta and Clark, 2007; Harteveld et al. 2007).   3.5.2 Root morphological traits and distribution  The differences in fine-root traits among soil depths may be related to the changes in soil characteristics and resource availability at different soil layers (Borken et al. 2007; Ostonen et al. 2007a; Freschet et al. 2017). Soil organic matter and nutrient concentration decreased, while bulk density increased with soil depth in both forests (Addo-Danso SD, unpublished data). The high root length and surface area in the  75 upper 0-10 cm would allow plants to capture the nutrients that are concentrated in that layer. In contrast, the higher SRA and SRL may be an adaptation or acclimation or both to the higher soil bulk density in the deeper layers (Lõhmus et al. 1989; Freschet et al. 2017). Root traits, including root length, surface area, volume and RTD were also affected by forest type, and the interactions between forest type and soil depth suggesting that those traits may be sensitive to the spatial variation in above and belowground properties within the study forests. Root morphological traits are known to change spatially in response to variations in edaphic factors across single tropical forest landscapes (Metcalfe et al. 2008).   The logged-over forest had significantly higher values for fine-root traits related to exploitation potential (length, surface area and volume), while fine-root traits representing exploitation efficiency (root tip/m root length, SRA and SRL) were higher in the old-growth forest. The differences in root trait values in the old-growth forest and logged-over forest may be partly due to differences in stand structure and the dominant functional species composition or a combination of the two. The old-growth forest had higher abundance of shade-tolerant species (37%) than the logged-over forest (27%) (Addo-Danso SD, unpublished data). Shade-tolerant (usually late-successional) species are slow growing and so may have lower nutrient requirements, and therefore transfer less C to traits such as root length and surface area (Holdaway et al. 2011; Xiang et al. 2013). By contrast, the high root quantities in the logged-over forest would allow rapid exploitation of soil to sustain the high growth rate of shade-intolerant (early-successional) tree species, which have high demand for soil resources (Paz et al. 2015). Stem density and basal area are also major stand factors that influence root morphology (Børja et al. 2008; Hansson, 2014).   Often studies that have compared fine-root traits in other tropical regions (Table A.3) have revealed contrasting dominance of root traits in forests at different recovery stages. In moist forests in central Sulawesi, Indonesia, Leuschner et al. (2009) reported a higher fine-root SRA in the upper 20 cm of soil in a selectively logged forest than in an old-growth forest. Hopkins et al. (1996) also reported higher fine-root SRL in a selectively logged forest than an old-growth forest on a metamorphic soil in North  76 Queensland, Australia. These findings contradict the higher root SRL and SRA values I observed in the logged-over forest. Also, the lower fine-root RTD value obtained for the old-growth forest, contradicts results from Hansson (2014) and Zangaro et al. (2014) who reported higher fine-root (diameter < 2 mm) RTD in old-growth forests than in secondary forests in Albertine rift, Rwanda and Paraná state, Brazil. These mixed results show the dynamic responses of root morphological traits to different site conditions during post-disturbance succession.   3.5.3 Root exploitation strategies differed in the old-growth forest and logged-over forest I found divergent resource exploitation strategies between the study forests. Plants in the old-growth forest have root traits associated with resource exploitation efficiency or an ‘intensive’ strategy, while plants in the logged-over forest have root traits that maximize their capacity to exploit soil resources or an ‘extensive’ strategy (Lõhmus et al. 2006; Xiang et al. 2013). The results indicated that plants in the old-growth forest produced thinner roots and invest root biomass more efficiently, which increases the volume of soil explored per unit biomass to capture water and nutrients (Eissenstat, 1991; Bauhus and Messier, 1999). Plants with thinner roots, higher SRL and SRA, but lower RTD are likely to allocate more C to roots, which increases root proliferation, allowing for the capture of more soil resources (Lõhmus et al. 1989; Eissenstat et al. 2015). In contrast, plants in the logged-over forest may maximize soil resources by developing thicker long-lived absorptive roots for greater resource conservation (Reich, 2014).   The contrasting exploitation strategies could alter the processes that affect nutrient cycling and soil C storage in these forest ecosystems (Bardgett et al. 2014; Prieto et al. 2016). The old-growth forest had thinner roots (low root biomass, high SRL, high SRA) with low tissue density, which could stimulate rapid decomposition (Hobbie et al. 2010), and potentially lead to higher CO2 losses (Bardgett et al. 2014; Makita et al. 2016). On the other hand, fast decomposition may accelerate the chemical and microbial transformation of root litter into humus that may persist in the soil (Prescott, 2010). The logged-over  77 forest had thick roots (high root biomass, low SRL, low SRA) with high tissue density, which may produce long-lived tissues that decompose slowly (McCormack et al. 2012; Bardgett et al. 2014), but have limited root growth and transport capacity (Eissenstat et al. 2015). The observed root exploitation strategies in the two forests should provide future opportunities to link root morphological traits in tropical forests to processes such as root production, turnover, and decomposition.  3.6 Conclusions  I conclude that root morphological traits may be important indicators to infer the dominant soil resource exploitation strategies at different stages after logging. It appears that the logged-over forest and the old-growth forest depend on divergent strategies to capture limiting soil resources, which may reflect differences in stand density, basal area, and functional species composition. The root exploitation strategies of plants with different growth characteristics promote their survival and co-existence in forest ecosystems. I would like to examine the linkage of root morphology and resource exploitation, and how they are related to forest structure and functional composition during post-logging recovery in future research.    78 Chapter 4: Patterns and controls on root dynamics in tropical forests in Ghana, West Africa 4.1 Synopsis Roots play major roles in the carbon (C) and nutrient dynamics of forest ecosystems, but there is a paucity of information on the factors controlling root dynamics in tropical forests. Over a 1.7 year period, I quantified biomass, necromass, productivity and turnover rates of fine (diameter < 2 mm) and small (diameter 2-5 mm) roots in an old-growth forest and a 54-year-old logged-over forest using ingrowth-core and sequential-coring methods. I also measured environmental variables (rainfall, air, and soil temperatures, soil water content, relative humidity, and absorbed photosynthetically active radiation) and soil chemistry variables (N, available P, K, Na, Mg, Ca, base saturation and pH), and correlated these with estimates of root biomass, necromass and mass (biomass plus necromass). Fine-root biomass and necromass varied temporally in both forests. Fine-root mass correlated positively to RH (r = 0.67, p < 0.05). Absorbed PAR and fine-root biomass were also positively correlated (r = 0.61, p < 0.01). Both fine-root biomass and mass were strongly, positively correlated with soil Na (r = 0.8, p < 0.05); while fine-root necromass correlated positively with soil available P (r = 0.71, p < 0.1). Fine-root mass and exchangeable K were also positively correlated (r = 0.84, p < 0.05). The mean annual fine-root productivity estimate was significantly higher in the ingrowth-core method (20.86 ± 0.50 Mg ha-1 yr-1) than the sequential-coring method (4.25 ± 0.50 Mg ha-1 yr-1), but fine-root turnover rates did not differ between the methods. Small-root productivity estimates were marginally higher when calculated using the sequential cores. Based on estimates from the two methods, mean annual fine- and small-root productivity and turnover rate were similar in the old-growth forest and the logged-over forest. The results show that multiple factors influenced fine root dynamics, and that combining different methods at the same sites can provide reliable root data that can be used to clarify the changes in root dynamics during post-logging recovery in tropical forests.        79 4.2 Introduction  Roots constitute a large fraction of the total forest biomass and productivity (John, 1973; Noormets et al. 2015), and also play major roles in the carbon (C) and nutrient dynamics of forest ecosystems (Rasse et al. 2005; Robinson, 2007; Prescott, 2010; Anderson-Teixeira et al. 2016). Understanding the roles of roots has been hampered by methodological challenges and paucity of information on the factors that affect root dynamics in forest ecosystems (Vogt et al. 1996; 1998; Hertel et al. 2009a; Yuan and Chen, 2010; Finér et al. 2011a, b). Various methods are used to quantify root dynamics (Vogt et al. 1998), but there is no consensus on the most suitable for sampling roots to estimate root mass, productivity and turnover rates (Vogt et al. 1998; Milchunas, 2009). The ingrowth-core and sequential-coring methods are the most widely used to estimate root mass productivity and turnover rates in ecological studies (Finér et al. 2011a; Brunner et al. 2013; Addo-Danso et al. 2016). The ingrowth-core method estimates root productivity as the amount of roots that grow into a defined root-free soil volume over a period of time (Persson, 1979). The method assumes that disturbance to roots and soil during core installation does not alter root in-growth (Lauenroth, 2000); this assumption may be violated in some forest ecosystems (e.g., Hendricks et al. 2006). The sequential-coring method estimates root productivity by sampling a series of soil cores at specific time intervals for at least one year (McClaugherty et al. 1982), after which approaches such as maximum-minimum, positive increment, and decision matrix are used to calculate changes in live (biomass) and dead (necromass) root estimates between time intervals (McClaugherty et al. 1982; Vogt et al. 1998). Both methods have drawbacks (Vogt et al. 1998; Milchunas, 2009), and often produce contradictory results. At global and regional levels, estimates of fine-root productivity and turnover rates based on ingrowth-core and sequential-coring methods tend to be similar (Vogt et al. 1998; Finér et al. 2011a; Brunner et al. 2013; Addo-Danso et al. 2016). However, large differences between the two methods have been reported at local scales, even at the same sites (Hendricks et al. 2006; Moser et al. 2010; Sun et al. 2015). It has been recommended to compare multiple methods at the same sites to evaluate the shortcomings inherent to the various methods (Hendricks et al. 2006).    80 Root mass estimates vary spatially and temporally in many forest ecosystems (Tieney et al. 2003; Gei and Powers, 2015). In tropical forests, which experience seasonal dry periods, root mass distribution is usually related to water, temperature, and soil nutrient concentrations (Cavalier, 1992; Green et al. 2005; Hertel and Leuschner, 2011; Gei and Powers, 2015). However, attempts to relate changes in water, temperature or soil nutrient concentrations to root mass have had mixed success. Strong (Green et al. 2005; Jiménez et al. 2009), weak (Leuschner et al. 2006; Gei and Powers, 2015), and no relationship (Espeleta and Clark, 2007; Hertel and Leuschner, 2011) between fine-root mass and water, temperature or soil chemistry have all been reported. More studies are needed from sites in different tropical regions to clarify the relationship between root mass and environmental factors.   Root productivity and turnover rates change during post-logging recovery in tropical forests (Hertel et al. 2009a). The few studies to date have mainly reported higher root productivity and turnover rates in old-growth (unlogged) forests than in logged forests (Sundarapandian and Swamy, 1996; Harteveld et al. 2007; Hertel et al. 2009a; Barbhuiya et al. 2012). The lower estimates of root productivity and turnover in logged forests have been attributed to the reduction in canopy cover, tree basal area and stem density, as well as altered soil conditions. These studies were conducted within a few years of logging, but to better understand how logging affects root dynamics, requires assessment of root productivity and turnover rates after several decades, when forest structure and soil fertility have recovered (Herbohn and Congdon, 1998; Asase et al. 2014).   Here I addressed the following questions: (1) Are there seasonal variations in fine (diameter < 2 mm) and small (diameter 2-5 mm) root biomass and necromass? If so, how are these values related to environmental and soil chemistry variables? (2) Are there differences in the estimates of root productivity and turnover rates obtained by the ingrowth-core and sequential-coring methods? (3) Do rates of root productivity and turnover differ between the old-growth forest and a logged-over forest? I applied both ingrowth-core and sequential-coring methods to quantify root biomass, necromass, productivity and  81 turnover rates in two tropical forests in Ghana, an old-growth forest and a 54-year-old logged-over forest influenced by similar parent material, but with distinct stand structural characteristics (basal area and stand density). I also correlated periodic fine-root biomass and necromass measurements from sequential cores to environmental (rainfall, air and soil temperatures, soil water content, relative humidity (RH), absorbed photosynthetically active radiation (APAR)) and soil chemistry (N, available P, K, Na, Mg, Ca, base saturation and pH) variables.  4.3 Materials and methods  4.3.1 Study area  The study was conducted in the Bobiri Forest Reserve, situated in the Ejisu-Juabeng District of the Ashanti Region, in southern Ghana. The details of the study area and stand and soil characteristics of the two study forests are provided in Table 3.1.   4.3.2 Root sampling 4.3.2.1 Ingrowth-core method  In August 2013 a hand-driven soil auger (diameter 12 cm) was used to sample soil to 30-cm depth from twelve random locations at least 30 m apart in each forest. Roots were sampled to 30-cm depth because earlier studies in the tropics, including Ghana, revealed that roots are highly concentrated in this layer (John, 1973; Metcalfe et al. 2008). Therefore, I assumed that the selected sampling depth included a large portion of the total root mass in these forests. Collected samples were separated into organic and mineral layers to avoid mixing the two layers, which may differ in nutrient concentrations (Girardin et al. 2013). The soil was hand-sifted to manually remove the roots following the ‘temporal prediction technique’, which uses a maximum-likelihood approach to estimate root biomass after correcting for under-estimation of very fine (lower-order) roots (Metcalfe et al. 2007). The technique has been used to process roots in Africa (Hansson, 2014; Nyirambangutse et al. 2017) and elsewhere (Girardin et al. 2013).    82 Roots were carefully picked from the soil samples over a period of 40 minutes, split into 10-minute time intervals. In August 2013, twelve ingrowth bags made from cylindrical nylon mesh (Figure 4.1A, 12-cm diameter, 40-cm long) were installed with root-free soil into the same holes and depth of the original samples (Jourdan et al. 2008; Marthews et al. 2012). During installation care was taken to ensure that root-free soil bulk density was similar to the surrounding undisturbed soil (Vogt et al. 1998). This was repeated in each study forest. Roots were manually retrieved from the ingrowth bags at approximately 3-4-month intervals according to the ‘temporal prediction technique’ described above (Figure 4.1B). The root-free soil was replaced in each ingrowth bag, again ensuring that bulk density was similar to the surrounding soil. Collected samples were transferred into plastic bags for laboratory processing at the Forestry Research Institute of Ghana. Samples were either processed immediately or kept refrigerated (4 °C) for 2-4 days before processing. Roots were sampled at intervals of 3-4-months to capture seasonal trends in root growth and death (Metcalfe et al. 2008). Overall, 144 cores (12 cores x 2 study forests x 6 sampling times) were collected during the study period from November 2013 to April 2015.   4.3.2.2 Sequential-coring method  Soil was sampled with a hand-driven soil auger (diameter 5.5 cm) to 30-cm depth in 25 plots (20 m x 20 m) in each of the study forests. Roots were sampled to 30-cm depth at randomly located points within each plot throughout the study period. When an obstacle such as a large structural root or stone obstructed the auger, it was relocated within an area of ca. 25 cm2 until a suitable core was extracted. Soil samples were kept separate for the three depths (0-10, 10-20, 20-30 cm), and transferred into plastic bags for laboratory processing. Collected samples were either processed immediately or kept refrigerated (4 °C) for a maximum of 14 days before processing. Roots were sampled at approximately 3-4 month intervals to coincide with ingrowth-core collection; thus sampling lasted from November 2013 to April 2015, and a total of 900 cores (75 cores x 2 study forests x 6 sampling times) were collected during the study period.     83   Figure 4.1 Ingrowth core made of nylon mesh inserted into the soil (left A), and extracted core containing roots (right B). Photos: S.D. Addo-Danso   4.3.3 Root processing, characterization and weighing  Roots collected from the ingrowth-cores were rinsed with water through a 0.25-mm sieve to remove adhering soil particles and debris. Roots were separated into fine roots (diameter < 2 mm) and small roots (diameter 2-5 mm), and into live (biomass) and dead (necromass) based on visual inspection of morphological features such as colour, tensile strength, and cortex and periderm characteristics (Vogt et al. 1998; Hertel and Leuschner, 2002). Live roots are usually light in colour, not easily broken, turgid, and the cortex and periderm are not easily separated; dead roots are generally dark or brown, brittle and shrivelled with a non-turgid cortex. No attempt was made to separate roots into individual species or into trees and understory vegetation. Roots were not distinguished into root branch orders (Pregitzer et al. 2002) because dead roots, which do not lend themselves to such hierarchical designations, were included in this study (Hertel et al. 2013). Live and dead roots were oven-dried at 60 °C to constant mass, and weighed to determine root biomass and necromass, respectively. Root biomass and necromass were expressed in g m-2, and later scaled up to Mg ha-1. For the sequential cores, samples from different depths were composited before processing. To ensure comparable uniformity in root processing, the collected  84 roots were also hand-sifted and picked using the ‘temporal prediction technique’, and characterized following the same procedure as above.    4.3.4 Environmental and soil chemistry variables  4.3.4.1 Environmental variables Rainfall, air, and soil temperatures, soil water content, relative humidity (RH), and photosynthetically active radiation (PAR) were recorded during the study period. With the exception of rainfall, which was collected automatically at 20-min intervals from a Mini-met weather station (Skye Instruments Ltd, UK, ca. 18 km from the study forests), all variables were measured bi-weekly or monthly with handheld instruments in the plots where the sequential cores were collected. Soil temperature was measured at soil depth of 20 cm using a Checktemp@1 digital thermometer with a probe (± 0.03 °C accuracy, Model HI98510, HANNA Instruments, USA). Soil water content to soil depth 30 cm was measured using a digital soil moisture meter (±5 % accuracy, Model DSMM500, General Instruments, NY, USA). Air temperature and relative humidity were also recorded with a Hygro-Thermometer and a data logger (±1 °C and ±3 % accuracy, Model SDL500, EXTECH Instruments Corporation, USA). Photosynthetically active radiation was measured with a quantum sensor and meter (±5 % accuracy, Model MQ-200, Apogee Instruments, USA). PAR measurements were carried out simultaneously with two sensors under the forest canopy and on forest roads ca. 200-400 m from the plots to represent below- and above-canopy PAR, respectively. PAR measurements on the forest road were used as reference. All the PAR recordings under the canopy were systematically taken at 25 points ca. 1 m above the ground surface. PAR was measured near or immediately after solar noon when light was unobstructed by cloud cover (Cournac et al. 2002). PAR values were used to calculate percentage PAR interception (APAR) by dividing PAR values under the canopy by the corresponding PAR values above the canopy.      85 4.3.4.2 Soil chemistry  Soil was sampled 1-2 months prior to root sampling to 30-cm depth in five randomly located positions in each forest. Collected samples were kept separate for the three depths (0-10, 10-20 and 20-30 cm), and transferred into labeled plastic bags. At the laboratory the samples were air-dried, crushed and sieved after removing gravel, roots, and other debris. The soil samples were subsequently sent the Soil Research Institute of Ghana (SRI), about 30 km from the study site for chemical analysis 1-2 weeks after each sampling campaign. At SRI, the soil samples were ground and analyzed for concentrations of total nitrogen (N), available phosphorus (P), and exchangeable potassium (K), sodium (Na), magnesium (Mg), calcium (Ca), pH and base saturation. Soil total N was determined by the Kjeldahl macro-digestion technique (Bremner, 1960); available P via the Bray 1-P extraction method (Bray and Kurtz, 1945). Exchangeable cations were determined by the ammonium acetate method, buffered at pH 7 (Thomas, 1982). Potassium and sodium were determined using a flame photometer, and magnesium and calcium using the EDTA titration method. Percent base saturation was calculated as the sum of exchangeable cations (K, Na, Mg and Ca). Soil pH was measured in a 1:1 soil-liquid solution of deionized water using a pH meter.   4.3.5 Calculations and statistical analysis  4.3.5.1 Predicted root dry mass The dry mass of roots collected at 10-min intervals over a total of 40 min of sampling was used to predict root mass if all roots in each core had been collected. The cumulative dry mass extracted at 10-min intervals was plotted against time (i.e. 10, 20, 30, 40 min) using a logarithmic curve to predict root mass to 100 min, which I assumed to be the time needed for complete retrieval of all roots (Figure A.1). Previous studies have used 120 min as the maximum extraction time (e.g., Metcalfe et al. 2007; 2008; Nyirambangutse et al. 2017), but a preliminary assessment at the study forests showed no systematic change in root dry mass after 100 min. The logarithmic curve is described by the following equation:    86  Rm (g) = a log (t) + b       eqn. (1)  Where Rm (g) is the root mass extracted at time t in grams, a is the constant defining the shape of the curve and b is the intercept. Other curve formulas (e.g., power, exponential, second-order polynomial) are available, but the logarithmic curve has been found to be most suitable for predicting root mass (Metcalfe et al. 2007).   4.3.5.2 Root productivity  I calculated root productivity for the ingrowth-core and sequential-coring methods using root data from five sampling dates. The sequential-core root data collected in January 2015 were destroyed by fire, which damaged the laboratory where the samples were stored. The ingrowth-core data collected in the same month was also excluded. Before estimating root productivity for the two methods, I calculated mean root biomass and necromass per diameter class (fine- and small roots) for both study forests on each sampling date (Jourdan et al. 2008). The biomass and necromass values were converted to Mg ha-1 before further calculation and analysis. Root productivity is expressed in Mg ha-1 yr-1.   4.3.5.2.1 Ingrowth-core Root productivity from the ingrowth-core data was calculated using three conventional approaches: (1) the ‘maximum-minimum’ approach (McClaugherty et al. 1982), (ii) the ‘positive increment’ approach (Persson, 1978), and (iii) the sum of biomass and necromass (Neill, 1992). In principle the maximum-minimum approach is used to estimate root productivity from sequential coring data (Vogt et al. 1998; Finér et al. 2011a; references herein), but some studies have applied it to calculate productivity from ingrowth-core data (Ostonen et al. 2005; Berhongaray et al. 2013; Brunner et al. 2013). The maximum-minimum approach assumes a single annual pulse in root productivity, based on asynchronous occurrence between live and dead roots (Edwards and Harris, 1977; McClaugherty et al. 1982). With this approach, annual root productivity (RPa) was calculated (eqn. (2)) as the difference between the highest biomass  87 (Bh) value and the lowest biomass (Bl) value measured during a full year (McClaugherty et al. 1982). Only significant differences between the highest and lowest root biomass values were considered (Hertel and Leuschner, 2002).    RPa = Bh  Bl        eqn. (2)  The positive increment approach calculates annual root productivity (RPa) by summing the positive increments of biomass (B)+ and necromass (N)+ between sampling intervals (Persson, 1978). All observed differences in biomass and necromass were used to calculate root productivity as per eqn. (3) (Hendricks et al. 2006; Jourdan et al. 2008).    RPa = ∑ (B + N)+        eqn. (3)  Annual root productivity was also calculated as the sum of root biomass and necromass sampled at 3-4 month intervals for one year (Neill, 1992):    RPa = ∑ B + N        eqn. (4)  In this study, ingrowth-core root sampling exceeded one year (19 months-between time of ingrowth core installation to last root retrieval), and therefore the root productivity values calculated from the different approaches were scaled to one year.    4.3.5.2.1 Sequential-coring  Three approaches (maximum-minimum, positive increment, decision matrix) were used to calculate annual root productivity from sequential-core data. The maximum-minimum and positive increment approaches were discussed above. The decision-matrix approach uses the direction and relative  88 magnitude of changes in biomass and necromass during each sampling period to estimate productivity (McClaugherty et al. 1982; Fairley and Alexander, 1985). With this approach, annual root productivity (RPa) is estimated (eqn. (5)) by summing all calculated productivity estimates between consecutive sampling dates in a year (Fairley and Alexander, 1985). The conditions with which to select a suitable productivity formula are provided in Table 4.1. In this study all changes in biomass and necromass (whether significant or not) were included in the calculation of root productivity (Hertel and Leuschner, 2002; Brunner et al. 2013). Productivity estimates from the approaches were scaled to a yearly basis since sequential cores were taken beyond one year (17 months).     RPa = ∑ P        eqn. (5)   Table 4.1 Decision matrix approach modified from Fairley and Alexander (1985).       Biomass increase   Biomass decrease            │∆B│ < │∆N│   │∆B│ > │∆N│ Necromass increase P = ∆B + ∆N    P = ∆B + ∆N    P = 0 Necromass decrease P = ∆B     P = 0    P = 0 B, Biomass; N, Necromass; P, Productivity; ∆ change in B, N between consecutive sampling dates. Annual productivity estimate is calculated by summing the estimates from all sampling intervals within a year.    4.3.5.3 Root turnover rate  Turnover rate (yr-1) was calculated as annual productivity estimate (Mg ha-1 yr-1) obtained from the maximum-minimum approach divided by root biomass (Mg ha-1). Previous studies have used maximum, minimum and mean root biomass to calculate turnover rate (Aber et al. 1985; Hendrick and Pregitzer, 1993; Gill and Jackson, 2000). I determined turnover rate by dividing annual productivity by maximum (eqn. (6)) and mean (eqn. (7)) root biomass values, which are the most common approaches used in the literature (Finér et al. 2011a; Brunner et al. 2013):    89    T = RPa         eqn. (6)          Bh     T = RPa         eqn. (7)        Bm            where T is annual root turnover rate, RPa is the annual root productivity, Bh is maximum root biomass, and Bm is the mean root biomass, which was calculated as follows:    Bm = ∑ B        eqn. (8)              N  where N is the number of sampling times (N = 5).   4.3.6 Statistical analysis  All data were tested for normality with Shapiro-Wilk and the Kolmogorov-Smirnov tests, and for homogeneity of variances by Bartlett’s test. Data not meeting the assumption of normality were log-transformed, if possible. Temporal variations in root biomass and necromass over the study period were compared using one-way analysis of variance (ANOVA), followed when appropriate by comparison with post-hoc Tukey’s HSD test. If log transformation was not possible, means were compared using non-parametric single-factor analysis (Kruskal-Wallis) test. Differences in mean estimates of root productivity and turnover rates between the study forests (old-growth and logged-over) were evaluated with the Mann-Whitney U test using values obtained from the different approaches (refer to methods) as variables.    90 Correlations between environmental variables (rainfall, air and soil temperatures, soil water content, RH and APAR), soil chemistry (N, available P, K, Na, Mg, Ca, pH and base saturation) and fine-root biomass, necromass and mass (biomass plus necromass) data from sequential cores were examined using Pearson product-moment and Spearman rank order correlations. Pearson correlations were used for all soil chemistry variables, while the environmental factors were evaluated with Spearman rank-order correlations. I used this correlation approach because of the limited sample size (Powers and Peréz-Avilles, 2013). Sequential cores allow detection of seasonal changes in root dynamics (Hertel and Leuschner, 2002), whereas data from ingrowth cores can influence the direction and magnitude of relationships between root dynamics and abiotic factors (Côté et al. 1998). For this analysis the root data were pooled, and the total rainfall, and mean air temperature, soil temperature, soil moisture, RH and APAR of the 3 months prior to sampling were used (Sánchez-Gallén and Alvarez-Sánchez, 1996; Espeleta and Clark, 2007; Jiménez et al. 2009). This is reasonable given that data from sequential cores usually reflect root growth and mortality events that have occurred prior to sampling (Gaul et al. 2008). Soil chemistry data collected during each sampling interval were also used for the correlation analyses. Differences at p < 0.05 were considered highly significant, while differences at p < 0.10 were considered as marginally significant. All analyses were performed with GraphPad Prism 7 (GraphPad Software, Inc., California, USA).    4.4 Results  4.4.1 Patterns of environmental and soil chemistry variables  Environmental variables varied seasonally and temporally during the study period (Figure 4.2A-F). Total monthly rainfall was highest in September 2013 (237.0 mm), and lowest in January 2015 (0.0 mm) (Figure 4.2A). Mean monthly soil moisture content followed similar seasonal patterns in both forests. Soil moisture was high during the rainy seasons (September-November and May-July), but dropped in the dry season (December-January). There was, however, no significant correlation between rainfall and soil moisture content (r = 0.26, p > 0.05, N = 19). Mean monthly soil temperature lagged behind air  91 temperature, warming more during the dry season, and dropped in the rainy period (Figure 4.2D and 4.2E). Air and soil temperatures were moderately, positively correlated (r = 0.48, p = 0.0032, N = 19). Mean monthly relative humidity showed an opposite trend to air temperature. Relative humidity peaked when air temperature dropped (Figure 4.2F), but they were not correlated (results not shown). Monthly APAR also varied seasonally, and by forest type. Overall, the mean APAR over the study period was significantly higher in the logged-over forest than in the old-growth forest (p = 0.0063).   Some soil chemistry variables also differed temporally during the study period (Figure 4.3A-H). Soil available P differed up to 5-fold among the sampling periods in both forests. Available P was highest during the dry season (November-February) of 2014, and was also lowest in the middle of the dry season in 2015, probably in response to the high rainfall recorded during that period (Figure 4.3B). Soil P was consistently high in the old-growth forest, but none of the other variables showed a consistent difference between the two forests (Figure 4.3A-H).  4.4.2 Root mass distribution and relationship with environmental and soil chemistry variables I observed similar temporal patterns in fine and small root biomass and necromass in the old-growth forest and the logged-over forest (Figure 4.4). Fine-root biomass estimates were consistently higher than the corresponding necromass estimates, but there was no marked pattern for small root biomass and necromass (Figure 4.4C-D). In both forests, fine-root biomass differed significantly among the sampling periods (Old-growth, p = 0.0035, F4,55 = 4.44; Logged-over, p = 0.0022, F4,55 = 4.77). Fine-root biomass peaked in November, which coincided with the end of the minor rainfall season, and the start of the dry season. However, the lowest fine-root biomass estimates occurred during different sampling months, June and September for the old-growth forest and the logged-over forest, respectively (Figure 4.4A and 4.4B). Differences between peak and lowest fine-root biomass were significant in both the old-growth forest (p = 0.0076) and logged-over forest (p = 0.0004). For fine-root necromass, the peak was observed in June,  92 corresponding with the middle of the major rainfall season. For small roots, the lowest and the highest biomass and necromass estimates were observed in November and June, respectively.   The correlation analysis showed differing relationships between fine-root biomass, necromass and mass with environmental and soil chemistry variables (Table 4.2). With the exception of APAR (r = 0.61, p = 0.06, N = 10), no significant relationship was observed between fine-root biomass and the environmental variables (rainfall, air temperature, soil temperature, soil water content, RH). None of the environmental variables correlated with fine-root necromass. Fine-root mass was strongly, positively correlated with RH (r = 0.76, p = 0.01, N = 10). There was a very strong positive correlation between soil Na and fine-root biomass (r = 0.83, p = 0.0151, N = 8) and fine-root mass (r = 0.80, p = 0.0234, N = 8). Furthermore, fine-root mass and exchangeable K were also strongly, positively correlated (r = 0.84, p = 0.0086, N = 8). No other soil chemistry variables correlated with fine-root biomass and necromass. However, there was a marginally significant positive relationship between available soil P and fine-root necromass (r = 0.75, p = 0.05, N = 8, Table 4.2).    93  Figure 4.2 Monthly environmental conditions recorded in the old-growth forest (black circles) and the 54-year-old logged-over forest (grey circles) during the study period: (A) Total rainfall; (B) Mean soil moisture content; (C) Mean absorbed photosynthetically active radiation; (D) Mean air temperature; (E) Mean soil temperature, and (F) Mean relative humidity. Rainfall data were collected automatically from a weather station, ca. 18 km from the study forests.    050100150200250Rainfall (mm)A0510152025Soil moisture content ( vol. %)BA S O ND J FMAM J J A S ON D J FMA5060708090100APAR (%)C  2013       2014                               2015242628303234Air temperature ( °C) D182022242628Soil temperature (°C) EA S O N D J F M A M J J A S O N D J F M A5060708090100Relative humidity (%)F  2013       2014                               2015 94    Figure 4.3 Temporal changes in soil chemistry variables at 0-30 cm soil depth at different sampling times. Data are means ± SE for old-growth forest (black bars) and 54-year old logged-over forest (grey bars).  0.00.20.40.60.8 Total N (%)AOld-growth forestLogged-over forest07142128 Available P (mg kg-1)B050100150200 Exchangeable K (mg kg-1) CAug-Oct 13 Nov-Feb 14 Mar-May 14 Jan-Mar 15020406080Na (mg kg-1)D0200400600800Mg (mg kg-1)E01000200030004000 Ca (mg kg-1)F0306090120 Base saturation (%)GGAug-Oct 13 Nov-Feb 14 Mar-May 14 Jan-Mar 1502468 pH (H2O)H 95      Figure 4.4 Fine (A, B) and small (C, D) root biomass and necromass (mean ± SE) to 30 cm soil depth in old-growth forest and logged-over forest. Given are values of 25 soil cores per forest on each sampling date.           020406080Fine-root mass (Mg ha-1)Old-growth forestA BiomassNecromassNov 13 Mar 14 June 14 Sep 14 April 1502468Sampling dateSmall-root mass (Mg ha-1)CLogged-over forestBNov 13 Mar 14 June 14 Sep 14 April 15Sampling dateD 96 Table 4.2 Correlation coefficients (r) between environmental, soil chemistry and fine-root variables (biomass, necromass and mass). Data for the old-growth forest and the 54-year-old logged-over forest are combined.   Independent variable  Fine-root biomass   Fine-root necromass   Fine-root mass r p-value   r p-value   r p-value Environmental variable (N=10)        Rainfall (mm) 0.20 0.6022  -0.17 0.6409  0.51 0.1314 Air temperature (° C)  -0.52 0.1334  0.16 0.6558  -0.48 0.1663 Soil temperature (° C)  -0.20 0.5837  0.31 0.3866  -0.35 0.3283 Soil moisture content (vol. %) -0.17 0.6309  0.29 0.4208  -0.06 0.8754 Relative humidity (%) 0.49 0.1552  0.31 0.3571  0.76 0.0100** Absorbed PAR (%) 0.56 0.0667*  -0.10 0.7896  0.47 0.1786          Soil chemistry variable (N=8)         Total N 0.34 0.4144  -0.02 0.9601  0.32 0.4415 Available P (mg kg-1) 0.28 0.4999  0.71 0.0479*  0.56 0.1482 Exchangeable K (mg kg-1) 0.61 0.1069  0.25 0.5479  0.84 0.0086** Na (mg kg-1) 0.83 0.0151**  -0.06 0.9048  0.80 0.0234** Mg (mg kg-1) 0.04 0.9169  0.14 0.7421  0.10 0.8151 Ca (mg kg-1)  -0.28 0.5022  0.39 0.3384  -0.11 0.7880 pH (H2O) 0.15 0.7280  0.07 0.8639  0.17 0.6832 Base saturation (%) 0.33 0.4276  -0.45 0.2608  0.14 0.7483 Data were combined for the two forests, and the analysis was based on mean values of variables for each sampling period. Data for rainfall was based on total rainfall recorded prior to root sampling. Soil chemistry and root dynamics are for the 0-30 cm soil depth. Statistically significant values are bolded. P-values were ** < 0.05, highly significant, * < 0.10, marginally significant. APAR, absorbed photosynthetically active radiation was determined dividing PAR values under the canopy by the corresponding PAR values above the canopy expressed in percentage.      97 4.4.3 Root productivity and turnover rates from ingrowth-coring and sequential-coring methods Estimates of fine-root productivity from the ingrowth cores were 2-6-fold greater than estimates from the sequential cores (Table 4.3). The mean annual fine-root productivity estimate was significantly higher in the ingrowth-core method (20.86 ± 0.50 Mg ha-1 yr-1) compared to estimate produced by sequential-coring method (4.25 ± 0.50 Mg ha-1 yr-1) (p = 0.0002). However, the ingrowth cores gave lower small-root productivity estimates than the sequential cores (Table 4.3). The ingrowth-core method gave higher fine-root turnover rates than the sequential-coring method (Table 4.3), but the differences were not significant (p = 0.1143). The mean small-root turnover rate calculated from ingrowth cores was 58% higher than that obtained from sequential cores. In both methods, the sum of positive increment approach gave higher root productivity values than the other approaches. Turnover-rate estimates using mean biomass were higher than those from maximum biomass (Table 4.3).   4.4.4 Root productivity and turnover rates in old-growth forest and logged-over forest Fine-root productivity was similar in the old-growth forest and the logged-over forest (Figure 4.5). Mean annual fine-root productivity was 13.58 ± 4.38 Mg ha-1 yr-1 for the old-growth forest and 11.53 ± 4.07 Mg ha-1 yr-1 for the logged-over forest (Figure 4.5A). Mean small-root productivity was not significantly different in the two forests, although the logged-over forest had a slightly higher root productivity estimate (Figure 4.5B). There was also no significant difference in the rates of root turnover between the two forests. Fine-root turnover rate was similar between the old-growth forest (0.36 yr-1) and the logged-over forest (0.37 yr-1). Small roots ranged from 0.60 to 0.98 yr-1 for the old-growth forest and logged-over forest (Figure 4.5C and 4.5D).        98 Table 4.3 Fine and small-root productivity (Mg ha-1 yr-1) and turnover rate (yr-1) via ingrowth-core and sequential-coring calculation approaches in the old-growth forest and the 54-year-old logged-over forest.   Method Fine roots (< 2 mm)   Small roots (2-5 mm) Old-growth Logged-over   Old-growth Logged-over Productivity (Mg ha-1 year-1)        Ingrowth-core   Maximum-minimum 12.90 11.28  0.03 0.13   Sum of biomass and necromass 25.18 22.55  0.24 0.43   Sum of positive increment 28.33 24.90  0.35 0.53        Sequential-coring   Maximum-minimum 4.58 2.33  0.65 0.65   Decision matrix 4.29 4.03  0.91 1.12   Sum of positive increment 6.17 4.08  0.98 1.20  Turnover rate (year-1)        Ingrowth-core   Maximum biomass 0.21 0.22  0.77 0.34   Mean biomass 0.90 0.96  2.22 0.97        Sequential-coring   Maximum biomass 0.11 0.12  0.19 0.22   Mean biomass 0.24 0.19  0.74 0.85 Root data were measured to 30 cm soil depth.          99  Figure 4.5 Root productivity (Mg ha-1 yr-1) and turnover rates (yr-1) to 30 cm soil depth for fine roots (A, C) and small roots (B, D) in old-growth (black bars) forest and 54-year old logged-over (grey bars) forest. Values are mean ± SE calculated from the approaches used to estimate root productivity and turnover rates. NS, Not significant (p > 0.05).  4.5 Discussion 4.5.1 Root mass distribution and relationship with environmental and soil chemistry variables The temporal variations in fine and small-root biomass and necromass distribution in both forests are consistent with reports from Africa (Zewdie et al. 2008; Asaye and Zewdie, 2013; Assefa et al. 2017) and elsewhere (Tierney et al. 2003; Harteveld et al. 2007; Yuan and Chen, 2010). The temporal differences in root dynamics are likely related to changes in soil factors (Na, K and P concentrations), RH and APAR during the study.  Fine-root biomass and mass were strongly correlated with soil Na and K. Sodium is not regarded as essential to the growth and development of some plants (Maathuis, 2009), but it may be 05101520Productivity (Mg ha-1 yr-1)Fine rootsANSOld-growth forest Logged-over forest0.00.40.81.21.6Turnover rate  ( yr-1) CNS0.00.30.60.91.2Small rootsBNSOld-growth forest Logged-over forest0.00.40.81.21.6DNS 100 beneficial to plants, particularly when K is limited (Battie-Laclau et al. 2013). Sodium and K affect litter decomposition in forest ecosystems (Kaspari et al. 2014). Since fine-root mass represents a balance between root productivity and decomposition (Gill and Jackson, 2000), it is possible that Na and K affected root productivity or decomposition. Manipulative experiments are required to test the effect of Na or K on root productivity and decomposition in these forests. Fine-root necromass and available P were also correlated (Table 4.2), in agreement with both observational and fertilization studies (Ostertag, 2001; Powers and Peréz-Aviles, 2013). There are suggestions that roots tend to die more or turnover faster when soil nutrient availability is high (Ostertag, 2001; Yuan and Chen, 2010).   The positive correlation between relative humidity and root mass is consistent with results from experimental studies of temperate forests (Ostonen et al. 2012; Rosenvald et al. 2014). Ostonen et al. (2012) and Rosenvald et al. (2014) reported higher fine-root biomass in young Betula pendula and monoclonal hybrid Populus tremula x P. tremuloides plots exposed to elevated RH than those under near-ambient RH in Estonia, which they explained as a strategy for the trees to increase nutrient uptake. Increased RH can impede nutrient uptake via decreases in the transpiration rates of trees (Sollins et al. 2016), and therefore plants exposed to high RH tend to allocate more biomass to roots (Rosenvald et al. 2014). Fine-root biomass also correlated positively with APAR, in line with field studies on grasses in Europe (Fitter et al. 1998; Edwards et al. 2004). Studies that have assessed solar radiation and root dynamics in forest ecosystems are scarce, but no relationship was found between solar radiation and fine-root (diameter ≤ 2 mm) productivity in a rainforest on podzol soil in Jambi Province, Indonesia (Violita et al. 2016). Solar radiation does not directly penetrate into the soil, and therefore its effect on root dynamics may be indirect via changes in C allocation in plants (Rinnan et al. 2005). The relationships between root dynamics and the environmental and soil chemistry variables were based on correlations, and not experiments, so should be interpreted with caution.     101 4.5.2 Root productivity and turnover rates from ingrowth-coring and sequential-coring methods The higher fine-root productivity estimates obtained from the ingrowth-core method may have resulted from excessive root proliferation due to (i) disturbance during core installation and severing of roots, (ii) reduced competition in the cores, and (iii) increased nutrient availability (Nadelhoffer and Raich, 1992; Vogt et al. 1998; Hendricks et al. 2006; Milchunas, 2009). The data showed that the fine-root mass collected during the first sampling (three months after core installation) largely influenced the estimates of annual fine-root productivity. In both forests, fine-root mass from the ingrowth cores declined sharply after the first sampling, and was consistently lower than that produced by the soil cores (results not shown). This pattern suggests that the assumption that disturbance to roots and soil during core installation does not affect root productivity may not be valid in these forest ecosystems. Previous studies reported similar effects of disturbance on ingrowth roots (Hendricks et al. 2006; Jourdan et al. 2008). For instance, Hendricks et al. (2006) found that fine-root productivity values obtained from ingrowth cores were affected by disturbances created during short-term core installations in Pinus palustris-Aristida beyrichiana forests in Georgia, USA. Nevertheless, disturbance during core installation does not always lead to root proliferation (Neill, 1992; Hertel and Leuschner, 2002). Thus the impact of core installation and soil disturbance on root in-growth is not uniform, and may depend on the severity of disturbance and the characteristics of soil and root systems of sites. The sequential-coring method is also associated with some errors. Past studies show that the sequential cores may under- or over-estimate root productivity (Singh et al. 1984; Vogt et al. 1998). The sequential-core method under-estimates root productivity when root growth and decay occur simultaneously (Lauenroth, 2000; Hendricks et al. 2006), and also if seasonal root mass maxima and minima are not detected (Hertel and Leuschner, 2002; Milchunas, 2009).   The estimates of fine-root productivity from the ingrowth-core method are high compared with the range reported for tropical forests in a global comparison of root productivity estimates from different methods at the same sites (Addo-Danso et al. 2016). With mean values between 3.48 and 5.01 Mg ha-1 yr-1, the fine-root productivity estimates from the sequential-coring method in this study fall in the lower range of  102 values reported for tropical, temperate and boreal forests (Nadelhoffer and Raich, 1992; Vogt et al. 1998; Finér et al. 2011a; Addo-Danso et al. 2016). The higher fine-root productivity estimates obtained from ingrowth cores compared to sequential cores are in agreement with some studies (e.g., Persson, 1983; Hendricks et al. 2006; Mei et al. 2010; Moser et al. 2010), but in contrast to others (e.g., Jiménez et al. 2009; Makkonen and Helmisaari, 1999; Hertel and Leuschner, 2002; Addo-Danso et al. 2016). For example, in a study along an altitudinal transect in southern Ecuador, Moser et al. (2010) found that ingrowth cores yielded higher values of fine-root productivity than sequential cores. However, Jiménez et al. (2009) reported that fine-root productivity estimates from ingrowth cores and sequential cores were similar in a tropical terra firme forest in Caatinga, Columbia.   I observed comparable root turnover rates from the two methods. This may be due to the same calculation (maximum-minimum) approach used to estimate root turnover rates. It is widely recognized that root turnover rates are greatly influenced by the calculation approach used, even when the same method is applied (Gill and Jackson, 2000; Brunner et al. 2013; McCormack et al. 2014). The comparable fine and small-root turnover rates between the two methods are consistent with reports elsewhere (Hertel et al. 2009a; Finér et al. 2011a; Brunner et al. 2013). In a global study, which analyzed 186 stands from all biome types, Finér et al. (2011a) reported similar fine-root turnover rates from ingrowth cores and sequential cores. Likewise, Brunner et al. (2013) did not find significant differences in fine-root turnover rates obtained from the ingrowth-core and the sequential-coring methods, in a compilation of data from Fagus sylvatica, Picea abies, and Pinus sylvestris stands across Europe. In contrast, others have reported large differences in root turnover rate estimates produced by the two methods (Mei et al. 2010; Sun et al. 2015).   The discrepancies in root productivity and turnover rates in the aforementioned studies could result from the calculation approach used to determine root productivity and turnover rates (Finér et al. 2011a; Brunner et al. 2013). I used different approaches to calculate root productivity and turnover rates from the  103 ingrowth-core and sequential-coring data (Table 4.3). On average, the maximum-minimum approach yielded the lowest root-productivity values relative to the others, which confirms the findings from many other studies (Nadelhoffer and Raich, 1992; Hertel and Leuschner, 2002; Hendricks et al. 2006; Jourdan et al. 2008). The turnover rate was calculated by dividing root productivity by either mean root biomass or maximum root biomass. In line with other studies (Berhongaray et al. 2013; Brunner et al. 2013), using the mean root biomass gave higher turnover rates than using the maximum root biomass. Accordingly, Brunner et al. (2013) suggested that the mean root biomass should be used to estimate root turnover rates because it is more representative of the annual root biomass in the soil.   4.5.3 Root productivity and turnover rates in old-growth forest and logged-over forest The similar root productivity in the old-growth forest and the logged-over forest corroborates other studies from Africa (Hansson, 2014; Ibáñez, 2015; Nyirambangutse et al. 2017) reporting comparable root productivity in old-growth forests and logged forests. For example, Hansson (2014) did not find any significant difference in fine-root productivity between old-growth forests and sites affected by logging and other disturbances in the Albertine rift in southwestern Rwanda. In contrast, studies in Asia have usually reported higher fine-root productivity in old-growth forests than in logged forests (Sundarapandian and Swamy, 1996; Harteveld et al. 2007; Hertel et al. 2009a; Barbhuiya et al. 2012). The inconsistent results from the studies from Africa and Asia may be attributable to differences in logging intensity and rates of recovery after disturbance. Indeed, logging intensity is usually higher in Asia (ca. 8-15 trees removed per ha) than in Africa (ca. 1-5 trees removed per ha) during logging operations (Chazdon, 2014). Moreover, tropical African forests have been found to recover faster from past disturbances than forests in Asia (Cole et al. 2014).   The similar root-productivity estimates for the old-growth forest and the logged-over forest may be related to similarities in floristic composition of the two forests (Addo-Danso SD, unpublished data). While I did not directly assess the influence of species composition on root productivity, previous studies  104 have reported that floristic composition of forests affects estimates of root productivity (Yuan and Chen, 2010; Finér et al. 2011a; Kubisch et al. 2016). Although root productivity was not statistically different between the two forests in my study, the old-growth forest had marginally higher fine-root productivity estimates than the logged-over forest. Could this relate in any way to the different dominant root morphological traits in these forests? In Chapter 3, I found fine-root specific root length (SRL) to be higher, and root tissue density (RTD) to be lower in the old-growth forest than in the logged-over forest. Past studies demonstrate that stands with higher SRL, and lower RTD, are likely to have higher root growth (Jacob et al. 2014; Eissenstat et al. 2015; Kubisch et al. 2016). Future studies should test whether root morphological traits can be useful predictors of root biomass productivity in these forests.   Overall, the mean annual fine-root productivity estimates for the old-growth forest and the logged-over forest in the 30-cm soil depth (13.60-11.50 Mg ha-1 yr-1) fall within the range of values reported in comparative studies from Africa (3.90-19.50 Mg ha-1 yr-1, Hansson, 2014; Ibáñez, 2015; Nyirambangutse et al. 2017), but are higher than those from Asia (1.4-6.6 Mg ha-1 yr-1, Sundarapandian and Swamy, 1996; Harteveld et al. 2007; Hertel et al. 2009a; Leuschner et al. 2009; Barbhuiya et al. 2012). The annual small-root productivity estimates (0.36-0.37 Mg ha-1 yr-1) are lower than the mean value reported for boreal forests (0.63 Mg ha-1 yr-1, Yuan and Chen, 2010), and for wet evergreen forests in northeast India (0.56-1.66 Mg ha-1 yr-1, Barbhuiya et al. 2012).  Root turnover rates were similar in the old-growth forest and the logged-over forest. The result corroborates earlier studies, which reported comparable root turnover rates in old-growth forests and logged forests (Hertel et al. 2009a; Barbhuiya et al. 2012). This observation could be attributed to the similar soil properties of the two forests (Table 3.1).   The mean annual fine-root turnover rates are lower than global averages reported for tropical forests (0.8-1.44 yr-1; Gill and Jackson, 2000; Finér et al. 2011a). The fine-root turnover rates in the old-growth forest and the logged-over forest are also lower than the values in tropical wet forests in Iboubikro, Congo (0.71-1.32 yr-1, Ifo et al. 2015). Comparable studies of small-root turnover rates are scarce; however, the  105 small-root turnover rates obtained in this study (0.60-0.98 yr-1) are higher than values for unlogged and logged forests in northeast India (0.20-0.33 yr-1, Barbhuiya et al. 2012), and the mean value reported by Chen and Yuan (2010) for boreal forests (0.51 yr-1).   4.6 Conclusions Root biomass, necromass and mass correlated with some environmental (RH and APAR) and soil chemistry variables (Na, P and K), suggesting changes in these factors potentially affect root dynamics in these forests. Ingrowth-cores produced higher estimates of fine-root productivity and turnover rates than sequential cores. This indicates that the method used to sample roots affect root productivity estimates, and therefore the same method should be applied in studies that seek to evaluate root productivity at different sites. Calculating root productivity with the sum of positive increment and root turnover rate with mean biomass produced the highest estimates, supporting earlier findings of Brunner et al. (2013). Root production and turnover rates did not differ between the old-growth forest and the logged-over forest suggesting root structures and other belowground processes have recovered 54 years after logging. The results demonstrate that combining different methods at the same sites can provide reliable root data that can be used to clarify the changes in root dynamics during post-logging recovery in tropical forests.            106 Chapter 5: Aboveground wood biomass and productivity in old-growth and logged-over forests: the importance of taxonomic variables, stand structural variables and traits   5.1 Synopsis  Understanding the spatial patterns and drivers of aboveground stem biomass (AGB) and coarse wood productivity (CWP) in tropical forest is critical for ecosystem modeling and forest management. Leaf traits can be correlated with key forest processes and functions. I used regression analysis to examine the relationships between the taxonomic variables (tree species richness, effective number of species, tree species diversity) or structural variables (tree diameter, basal area, tree density) and AGB or CWP in an old-growth forest and a 54-year-old forest in Ghana. I also analyzed the bivariate relationships between leaf traits and tree species biomass and productivity in 18 individuals ranging from pioneers to shade-tolerant species. Five leaf traits were considered: three chemical traits (leaf N, leaf P and leaf K) and two morphological traits (specific leaf area and leaf dry matter content). Plot-level AGB and CWP were mostly explained by the structural variables. Tree diameter and tree density together emerged as the strongest predictors of AGB (old-growth forest: adjusted R2 = 0.95; logged-over forest: adjusted R2 = 0.97). In addition, tree species richness was positively related to AGB and CWP, but this was via the indirect influence of tree density. The taxonomic variables were most strongly related to AGB and CWP in the old-growth forest. In the logged-over forest, the relationship between tree species richness and AGB was linear and positive, but it was hump-shaped in the old-growth forest, indicating the variable nature of the relationships in these forests. Leaf K related positively to tree biomass (R2 = 0.66) in the logged-over forest. Leaf N and P were significantly and positively related to tree productivity in the old-growth forest and logged-over forest. Overall, the study showed that structural factors mediate the influence of tree species richness on AGB and CWP. Moreover, few leaf chemical traits explained the interspecific variation in tree-species biomass and productivity.       107 5.2 Introduction  Tropical forests cover about 10% of the Earth’s land surface, an area of ca. 1.2 billion ha (FAO, 2001), but represent the largest terrestrial reservoir of biological diversity (Mayaux et al. 2005). Tropical forests account for 70-72% of total biomass and productivity of global forest ecosystems (Pan et al. 2013; Ma et al. 2015). Stem biomass (hereafter aboveground biomass, AGB) and coarse wood productivity (CWP) are key components of the total forest productivity and carbon cycling (Malhi et al. 2004; Girardin et al. 2016). Considerable work has been undertaken to determine the spatial patterns and drivers of AGB and CWP in tropical forests at regional, continental, and global scales (e.g., Malhi et al. 2004; Lewis et al. 2013; Pan et al. 2013; Banin et al. 2014; Gillman et al. 2015; Sullivan et al. 2017).  Large variations also exist in the distribution of AGB and CWP at the site-level (e.g., Chisholm et al. 2013), due to differences in forest structure, soil conditions, species functional composition and disturbance regimes (Djagbletey, 2014; Jucker et al. 2016a; Ledo et al. 2016; Sande, 2016). There is recent interest in the relationship between taxonomic variables, including species richness and diversity, and AGB and CWP in tropical forests (e.g., Hector et al. 2011; Ruiz-Jaen and Potvin, 2011; Day et al. 2013; Chisholm et al. 2013; Gillman et al. 2015; Poorter et al. 2015; Sullivan et al. 2017). Most studies show that at fine spatial scales (<1 ha), species richness and diversity enhance biomass production (Chisholm et al. 2013; Sullivan et al. 2017) via niche complementarity (i.e., interspecific interactions between plant species enhance the availability, capture and efficient use of resources by neighbours) and selection effect (i.e., a species-rich community is more likely to contain a dominant species that drives productivity) (Fridley, 2001; Loreau et al. 2001). Spatial patterns in AGB and CWP are also influenced by stand structural attributes such as tree diameter, basal area and tree density (Slik et al. 2010; Paquette and Messier, 2011; Lewis et al. 2013; Vilà et al. 2013; Durán et al. 2015; Poorter et al. 2015; Forrester and Bauhus, 2016). Globally, basal area and tree diameter are strong predictors of AGB and CWP in tropical (Durán et al. 2015; Poorter et al. 2015), temperate and boreal forests (Paquette and Messier, 2011; Vilà et al. 2013). Tree density and basal area are known to influence canopy and fine-root architecture and morphology (Børja et al. 2008; Forrester and Bauhus, 2016), which affect light absorption and water uptake.   108 Tropical forest structure and species functional composition change during post-logging recovery (Chazdon, 2014), thus the relative importance of taxonomic and structural variables may also differ over time (Cai et al. 2016a; Sande, 2016). To date, only a few studies have examined how AGB or CWP change with species richness during secondary forest succession (Balvanera and Aguirre, 2006; Lasky et al. 2014; Sande, 2016). These studies have generally reported a strong positive relationship between species richness and biomass during the early stages of succession (Balvanera and Aguirre, 2006), but weak or no relationship later in succession (Balvanera and Aguirre, 2006; Lasky et al. 2014). More empirical studies are needed from other tropical sites to determine how the relationships between taxonomic or structural variables and AGB or CWP change during post-logging succession.   Plant functional traits drive key forest ecosystem processes and functions (Reich, 2014). Traits are characteristics that reflect the life history of individuals and enhance their performance in a particular ecosystem (Violle et al. 2007; Reich, 2014). Leaf traits such as N or P concentrations (leaf N and P) and specific leaf area (SLA; leaf area per unit dry mass) are correlated with critical processes such as photosynthesis, respiration and competition that influence tree growth (Walker et al. 2014; Kunstler et al. 2016). Trait-based studies in tropical forests at the community-level have reported strong relationships between leaf traits and stand growth and productivity (Finegan et al. 2015; Sande, 2016). However, at the tree species level, positive (Kunstler et al. 2016), weak (Wright et al. 2010; Visser et al. 2016) and no (Poorter et al. 2008; Hérault et al. 2011) relationships have been reported between leaf traits and tree growth. Additional studies are required to clarify relationships between leaf traits and tree growth in tropical forests.   In this study I examined the role of taxonomic and structural variables on AGB and CWP in an old-growth forest and a 54-year-old logged-over forest. These two forests provided the opportunity to assess how these relationships change during post-logging succession. Furthermore, I assessed how leaf traits influence the interspecific variation in tree biomass and productivity. I addressed the following questions:  109 (1) Do taxonomic and structural variables relate to aboveground biomass (AGB) and coarse wood productivity (CWP)? If so, are the relationships consistent for the old-growth forest and the logged-over forest? (2) What is the relative importance of the taxonomic and structural variables in explaining AGB and CWP? (3) Do leaf traits contribute to the interspecific variation in tree biomass and productivity in the old-growth forest and the logged-over forest?   I used scatter-plot and regression analyses to examine the relationships between taxonomic variables (tree species richness, effective number of species, tree species diversity) or structural variables (tree diameter, basal area and tree density) and AGB or CWP. Relationships were analyzed at a fine spatial scale (0.04 ha plots), where the effects of climate and environmental heterogeneity are inherently controlled. I also determined the bivariate relationships between leaf traits and tree biomass and productivity in 18 trees ranging from pioneers to shade-tolerant species in the old-growth forest and the 54-year-old logged-over forest. I considered five traits: three chemical traits (leaf N, leaf P and leaf K) and two morphological traits (specific leaf area and leaf dry matter content).   5.3 Materials and methods 5.3.1 Study area The study was conducted in the Bobiri Forest Reserve, situated in the Ejisu-Juabeng District of the Ashanti Region, in southern Ghana. The Bobiri Forest Reserve covers an area of ca 5,504 ha, and lies in the moist semi-deciduous forest zone (Hall and Swaine, 1981). The mean temperature of the coolest month (August) ranges from 21.8 °C to 27.4 °C and for the hottest month (February) from 22.4 °C to 33.8 °C. The annual rainfall ranges from 1210 to 1800 mm, with a dry season lasting from December to mid-March. During the dry season, mean rainfall is less than 100 mm.   The Bobiri forest has deeply weathered, well-drained soils with a clay-enriched, illuvial subsoil developed on granite (Foggie, 1947). Soil texture varies from sandy loam to clay loam (Foggie, 1947).  110 The reserve consists of a mixture of deciduous and evergreen species in approximately equal proportions (Hall and Swaine, 1981). Further details on the study area and the two study forests are provided in Table 3.1.  5.3.2 Study forests  Plots were established in the research and protection old-growth forest (strict-nature forest), which are described in detail in Chapter 3.   5.3.3 Plot inventory Twenty-five contiguous plots (20 m x 20 m) were established in each forest to inventory all trees ≥ 10 cm diameter at breast height (DBH). During the first census in 2013, each tree was mapped, marked with paint, and the diameter measured 1.3 m from the ground using a diameter tape. For trees with irregular boles or buttresses (Figure 5.1A), the diameter was measured ca. 50 cm above the buttress using a ladder (Figure 5.1B, Clark et al. 2001a; Marthews et al. 2012). Total tree heights were also measured on 50 trees in each forest using a LaserAce hypsometer (Measurement Devices Ltd. UK). Trees selected for height measurements included a range of functional groups or guilds (pioneers, non-pioneer light demanders and shade-bearers, Swaine and Whitmore, 1988) and DBH classes (10-20 cm, 20-50 cm and > 50 cm). Pioneers do not establish in the forest understory, Non-pioneer light-demanders (NPLDs) can survive in the understory during the seedling stage, but require light for further growth, and the shade-bearers are shade tolerant so survive and grow in the forest understory. Information on genus, family and species names was also collected following Hawthorne and Abu-Juam (1995) and Hawthorne and Jongkind (2006). The plots were re-censused in 2015 to incorporate new recruits and assess mortality.       111      Figure 5.1 A tree with buttress roots (left photo, A), measuring tree diameter above buttress (right photo, B). Photos: S.D. Addo-Danso  5.3.3.1 Data cleaning and gap-filling  Errors associated with data recording and entry, extreme outliers, and missing diameter values can introduce bias in biomass estimates (Muller-Landau et al. 2014; Talbot et al. 2014; Sheil et al. 2017). Therefore, I checked the data thoroughly for clerical errors (such as shifts in decimal points), and also identified extreme values that were biologically unrealistic within the census interval. Simple clerical errors were corrected, but trees that had diameter growth ≥ 4 cm year-1 or whose diameter shrank by ≥ 0.5 cm were considered outliers and excluded from further analysis (Talbot et al. 2014; Jucker et al. 2016), unless they were species that are known to grow very fast. As tree stems can shrink due to low water potential in the xylem during short-term dry periods (Shiel, 1995), small negative values were retained (Talbot et al. 2014). Trees that were not captured in the 2015 census, and were not found to have died were accounted for following the suggestions of Talbot et al. (2014). I used the mean diameter growth of all individuals in the same plot, belonging to the same tree size class (e.g., 10 cm ≤ diameter ≤ 20 cm, 20 cm ≤ diameter ≤ 30 cm; 30 cm ≤ diameter ≤ 40 cm, 40 cm ≤ diameter ≤ 50 cm, and diameter ≥ 50 cm) to estimate the diameters of missing trees. This procedure has been used in other studies to estimate the  112 diameters of missing trees in aboveground biomass research in the tropics (e.g., Jucker et al. 2016; Sande, 2016). Overall, less than 0.5% of the data were corrected for these errors.   5.3.3.2 Site-specific diameter-height equation  I used the height and diameter measurements from the 100 sample trees to generate an equation to estimate the heights of all trees in both forests. Two models were fitted to the data: linear and second-order polynomial (Figure A.3). The second-order polynomial fitted well to the diameter-height data (Table 5.1) because it had the lowest Akaike Information Criterion (AIC) and Root Mean Squared Error (RMSE), and the highest adjusted R2 (Fayolle et al. 2016). Although there is no consensus on the shape of diameter-height equations for tropical trees (Feldpausch et al. 2011; Fayolle et al. 2016), non-linear models including polynomials have been found to satisfactorily fit diameter-height data (Fayolle et al. 2016).  The height equation was:   H = a + b x D + c x D2         (1)  In addition, tree heights were predicted using an allometric equation developed by Feldpausch et al. (2011) for West Africa. I compared the heights predicted from the site-specific equation to those from the Feldpausch equation to assess whether they were closely related. The Feldpausch equation was:   log H = 0.8946 + 0.6365 x log D       (2)   where H is height, D is DBH, and a, b, c are coefficients. The site-specific equation and that of Feldpausch were strongly correlated (r2 = 0.98, p < 0.0001, N= 193, Figure A.3).      113 Table 5.1 Site-specific allometric equations relating tree diameter (D in cm) and tree total height (H in m) showing linear and second-order polynomial models. The Akaike Information Criteria (AIC), the Root Mean Squared Error (RMSE), adjusted R2 and coefficients are given for each model. The best model (lowest AIC and RMSE, and highest adjusted R2) is shown in bold. Data for the old-growth forest and 54-year-old logged-over forest are combined.  Model R2 adjusted AIC RMSE a b c Linear, H = a + b x D 0.73 680.7 5.76 7.636 0.4331 - Second-order polynomial, H = a+ b x D + c x D2 0.75 668.8 5.55 4.026 0.70005 -0.002657   5.3.4 Aboveground biomass and coarse wood productivity estimation  I calculated the aboveground biomass (AGB, kg) of individual trees in each plot using three allometric equations (Table 5.2). The allometric equations included locally derived equations (with and without tree height and wood density) (Henry et al. 2010; hereafter Henry equation), and the pan-tropical equation for moist forests (Chave et al. 2005; hereafter Chave equation). These equations were used to compare their AGB estimates, and to assess how they influence AGB-taxonomic relationships. Two of the equations employed diameter, total height, and wood density, so wood density values were obtained from the World Agroforestry Centre (ICRAF) and the global tropical forest wood-density database (Chave et al. 2009). For species with no wood density estimates, I used genus or family-level values (Poorter et al. 2015). For individuals that could not be identified, a plot-level mean wood density value was used. Plot-level AGB (Mg ha-1) was calculated by summing the biomass values of all trees in each plot.    Coarse wood productivity (CWP, kg yr-1) was estimated as the change in biomass of surviving trees between 2013 and 2015, plus biomass increment of new recruits in 2015 (Clark et al. 2001a). To estimate the biomass of recruits between 2013 and 2015, I subtracted the biomass of each individual with DBH = 10 cm in 2013 from its biomass in 2015. This assumes that the recruits had a DBH of 10 cm at the start of the census (Clark et al. 2001a; Talbot et al. 2014). Estimating the biomass increments of recruits with this  114 approach underestimates their contribution to total CWP (Sande, 2016). Plot-level CWP (Mg ha-1 yr-1) was estimated by summing the biomass increments of surviving trees and that of the recruits, and dividing by the census interval (in years), and is expressed on per-ha basis (Clark et al. 2001a). In this study, CWP excludes the contribution of tree mortality (Clark et al. 2001a), although mortality is a good predictor of forest productivity (Sande, 2016).    Table 5.2 Allometric equations used to estimate aboveground biomass (AGB) of trees (≥ 10 DBH cm) in  the old-growth forest and the 54-year-old logged-over forest. Source Location Equation Henry et al. (2010) Ghana Y = 0.30 x D exp (2.13) Henry et al. (2010) Ghana Y = 0.00347 + 0.002 ρD2H Chave et al. (2005) Pan-tropical Y = 0.0509 x ρD2H Y, biomass (in kg); H, height (in m); D, diameter at breast (DBH, in cm); ρ, wood density (in g cm-3)   5.3.5 Taxonomic and stand structural variables I calculated three taxonomic variables for each plot: species richness, Shannon’s index and effective number of species. Species richness is the number of tree species per plot (400 m2) and is calculated as:     𝑆𝑅 = ∑ 𝑝i𝑠𝑖=1           (3)  Shannon’s index quantifies the diversity of each plot, as a measure of species richness and abundance (Jost, 2006). Shannon’s index (H) was calculated as:  𝐻 = −∑ 𝑝i𝑠𝑖=1 ln(𝑝𝑖)          (4)  where SR is the species richness (hereafter tree species richness), H is the Shannon’s index (hereafter tree species diversity, SD) and pi is the proportion of tree species ith. Effective number of species (ES) is the number of equally common tree species in a plot required to give a value of a particular index (Jost,  115 2006). It is calculated as: exp (H). Previous studies have used either one or two of these taxonomic variables (Poorter et al. 2015; Jucker et al. 2016), but all were considered because each variable may relate differently to AGB and CWP (Day et al. 2013; Poorter et al. 2015).   Three structural variables: mean diameter, basal area, and tree density were calculated for each plot. Previous studies showed that these variables are closely related to AGB and CWP (Lewis et al. 2013; Poorter et al. 2015; Cai et al. 2016a). I calculated mean diameter (hereafter tree diameter) as the mean of all the tree diameters in each plot. Basal area (m2 ha-1) denotes the total cross-sectional area of all trees in a plot (0.04 ha), and was calculated as:  BA = 0.0000786 x ∑𝐷2/𝐴       (5)  where BA is basal area, D is DBH and A is plot area in ha. Tree density (m2) was calculated as the number of stems per plot.    5.3.6 Leaf sampling and trait determination   I considered five leaf traits: three chemical traits (leaf N, P and K concentrations); and two morphological traits (specific leaf area and leaf dry matter content). These leaf traits are known to be good predictors of tree growth (Kunstler et al. 2016; Sande, 2016). Leaf traits and the functional indicators are described in Table 5.3. Traits were measured on nine dominant species, which together composed > 50 % of the total basal area in each forest. The individuals included a range of functional groups or guilds and DBH classes (Table 5.4). Furthermore, the trees were selected on the basis that climbing was possible (Figure 5.2A) and that branches could be retrieved from all canopy positions, as it was not possible to climb some trees due to large buttresses or liana loadings. Live fresh leaves were collected from three branches detached from three canopy positions: (1) shaded lower portion, (2) upper half, where some amount of sunlight was received, and (3) outer canopy, where the leaves were exposed to direct sunlight (Pérez-Harguindeguy et  116 al. 2013). For each individual, mature fully expanded leaves with no herbivore damage were collected, stored in tightly sealed plastic bags, and sent to the laboratory for analysis. The leaves were collected in February-March (dry season) and July (wet season) of 2014, and the same trees were sampled during each visit.   Nine leaf samples (three replicates from each canopy position) from each tree were scanned using a flatbed scanner (CanoScan LiDE 110, Canon, Hanoi, Vietnam) attached to a laptop (Figure 2B). Only complete leaves were scanned, with rachis and the petiole attached. But for large-leaved species, including Musanga cecropioides and Pterygota macrocarpa, the leaves were cut into smaller pieces and scanned separately. Digital images were analyzed using IMAGEJ, v. 1.38 software (National Institute of Health, USA) to estimate leaf area. After scanning, leaf fresh mass was measured and then the leaves were oven-dried at 70° C to determine their dry mass. Leaf traits, including specific leaf area (SLA), and leaf dry matter content (LDMC) were calculated using the leaf area, fresh and dry mass values following standardized protocols (Pérez-Harguindeguy et al. 2013). Specific leaf area (cm2 g-1) was calculated as the one-sided leaf area divided by leaf dry mass. Leaf dry matter content (g g-1) was calculated as leaf dry weight divided by leaf fresh mass. The dried leaves were then ground, using a mortar and pestle and sent the Soil Research Institute of Ghana, to determine concentrations of N, P, and K. Leaf N was determined by the Kjeldahl technique (Bremner, 1960), leaf P by wet digestion technique followed by colorimetric determination of ammonium phosphomolybdate (Motsara and Roy, 2008), and leaf K using a flame photometer method (Motsara and Roy, 2008).         117 Table 5.3 Leaf traits measured, with abbreviations, units and what they indicate.  Trait Abbreviation Unit Indicator of: Leaf nitrogen concentration Leaf N % Light use efficiency; photosynthetic capacity Leaf phosphorus concentration Leaf P % Photosynthetic capacity Leaf potassium concentration Leaf K % Leaf conductance; photosynthetic capacity Specific leaf area SLA cm2 g-1 Photosynthetic capacity; soil fertility Leaf dry matter content LDMC g g-1 Leaf toughness; soil fertility        Figure 5.2 A climber on a tree to harvest branches (left photo, A), and scanning leaves to determine traits (right photo, B).           118 Table 5.4 List of tree species sampled with codes, family, functional group or guild used for trait determination. Diameter, height and wood density are also shown.  Tree species Code Family Functional group DBH (cm) Height (m) WD (g cm-3) Old-growth forest             Celtis mildbraedii CM Cannabaceae Shade-bearer 54.90 34.50 0.59 Nesogordonia papaverifera NP Malvaceae Shade-bearer 36.40 26.0 0.65 Terminalia superba TSu Combretaceae Pioneer 42.70 29.10 0.46 Musanga cecropioides MC Urticeae Pioneer 71.0 40.40 0.24 Triplochiton scleroxylon TS Malvaceae Pioneer 15.50 14.20 0.33 Sterculia oblonga SO Malvaceae NPLD 28.30 21.70 0.51 Albizia zygia AZ Fabaceae NPLD 50.90 32.80 0.49 Pterygota macrocarpa PM Malvaceae NPLD 15.70 14.40 0.49 Celtis zenkeri CZ Cannabaceae NPLD 23.80 19.20 0.61        Logged-over forest       Celtis mildbraedii CM Cannabaceae Shade-bearer 41.50 28.70 0.59 Nesogordonia papaverifera NP Malvaceae Shade-bearer 90.0 45.50 0.65 Terminalia superba TSu Combretaceae Pioneer 58.10 35.80 0.46 Triplochiton scleroxylon TS Malvaceae Pioneer 74.30 41.40 0.33 Alstonia boonei AB Apocynaceae Pioneer 32.0 23.70 0.32 Petersianthus macrocarpus PMa Lecythidaceae Pioneer 48.30 31.70 0.68 Sterculia oblonga SO Malvaceae NPLD 47.30 31.20 0.51 Pterygota macrocarpa PM Malvaceae NPLD 18.70 16.20 0.49 Sterculia rhinopetala SR Malvaceae NPLD 49.40 32.20 0.67 DBH, diameter at breast height; WD, wood density; NPLD, Non-pioneer light-demander   5.3.7 Statistical analyses  Differences in taxonomic variables (tree species richness, effective number of species and tree species diversity) and leaf traits (N, P, K, SLA and LDMC) between the old-growth forest and the logged-over forest were examined using a student t-test. Analysis of Variance (ANOVA) and post-hoc Tukey’s HSD tests were used to compare AGB and CWP estimates of the allometric models, as well as leaf traits among the three functional groups. All uncertainty in estimates of taxonomic variables, structural variables, leaf traits, AGB and CWP are given as standard error of the mean. I examined the relationships  119 between taxonomic and structural variables and AGB or CWP using scatterplots and regression analysis at the fine spatial scale (0.04 ha). Data for the two forests were analyzed separately and together. Relationships for AGB and CWP were analyzed separately because they are not always related in forest ecosystems (Keeling and Phillips, 2007; Knapp et al. 2014). For the taxonomic variables (tree species richness, effective number of species and tree species diversity), I tested the relationships with both linear and second-order polynomial models (Gillman et al. 2015; Sullivan et al. 2017). The model with the lowest AIC and highest adjusted R2 was selected as the best, but both models were considered to be equally supported if the difference in AIC was less than two units (Cai et al. 2016a).   Structural variables such as tree density can enhance or obscure any relationship between taxonomic variables and AGB or CWP (Chisholm et al. 2013; Forrester and Bauhus, 2016), so I repeated the analysis using multiple linear regression with AGB and CWP as the response variables and the taxonomic and structural variables as predictors. To avoid collinearity, I calculated the Variance Inflation Factor (VIF) for each variable (Table A4), after regressing each predictor on the remaining predictors (Dormann et al. 2013). Tree diameter, basal area, and tree density had VIF < 10 and were therefore selected for the model. Basal area had the lowest VIF, but it was excluded because it was correlated with tree density in the old-growth forest, and tree diameter in the logged-over forest (Table A5). Although tree species richness correlated with tree density in the old-growth forest (Table A5), it is the most commonly used taxonomic variable in AGB-biodiversity studies (e.g., Poorter et al. 2015; Forrester and Bauhus, 2016) and so was considered in the analysis. This allowed me to assess species richness effect on AGB or CWP after controlling for tree density and tree diameter. Prior to analysis, all trees that were not identified to species were removed, and the AGB and CWP estimates were log-transformed to meet the assumptions of normality, and homoscedasticity, and to reduce the effects of outliers. All regression analyses were performed using lm function of the lavaan package in R.3.1.3. (R packages v 3.1.3; The R Foundation for Statistical Computing Platform; R Development Core Team, 2015). In addition, I evaluated the relative importance of each variable by comparing the correlation coefficient between each variable and AGB or  120 CWP (Poorter et al. 2015). A higher correlation coefficient indicated greater importance in determining plot-level AGB or CWP.   Niche complementarity and a selection effect have been proposed to explain the positive relationship between taxonomic variables and AGB or CWP (Fridley, 2001; Loreau et al. 2001). A common approach used to assess the role of niche complementarity in tropical forests is to relate AGB or CWP to the functional diversity of tree species in a plot (Cavanaugh et al. 2014; Sullivan et al. 2017). It is assumed that in functionally diverse plots, tree species will partition resources or use resources more efficiently (Loreau et al. 2001). Functional diversity measures the variability of traits of the tree species in a plot (Cavanaugh et al. 2014). In this study, functional diversity was quantified as the standard deviation of wood density (WDSD) (Cavanaugh et al. 2014; Sullivan et al. 2017). Wood density was used because it is a good indicator of a tree species life history strategy (Lida et al. 2012), and relates to tree growth potential (Visser et al. 2016). More diverse plots should contain tree species with high variation in wood density, and greater potential for niche complementarity. Selection effect was also evaluated by relating AGB or CWP to the functional dominance of trees in a plot (Cavanaugh et al. 2014; Sullivan et al. 2017). Functional dominance refers to the abundance of a trait compared to other traits in a tree community (Cavanuagh et al. 2014). A few species tend to drive AGB in the study area (Djagbletey, 2014), so I assumed that the more diverse plots were more likely to contain a species with a dominant trait that influences biomass. Wood density is directly related to AGB (Lida et al. 2012), and usually tree species with high wood densities have higher AGB than those with low wood densities (Slik et al. 2013), so functional dominance was determined as the community-weighted mean (CWM) of wood density (Cavanaugh et al. 2014; Sullivan et al. 2017). Community-weighted mean was calculated as the mean wood-density value for each tree species weighted by the relative abundance of that species in a plot (Cavanaugh et al. 2014). I also analyzed the bivariate relationships between the leaf traits and tree biomass or tree productivity. Unless otherwise stated, all analyses were performed with the GraphPad Prism 7 (GraphPad Software, Inc., California) software package, with a significance level of p < 0.05.   121 5.4 Results  5.4.1 Forests characteristics  Tree species richness and effective number of species varied 3-4-fold across the 50 plots (ranging from 7-26 per 0.04 ha, Table 5.5). Tree species diversity ranged from 1.97 to 3.16, and did not differ significantly between the old-growth forest and the logged-over forest (t = 0.23, p = 0.82). Tree diameter varied 1.7-fold (15.69-28.14 cm), tree density varied 4-fold (10-41 per 400 m2) and basal area varied 9.6-fold (7.19-68.92 m2 ha-1). Aboveground wood biomass varied widely among the plots, ranging from 12.6 to 680.6 Mg ha-1, with a mean of 124.50 ± 9.67 Mg ha-1. Coarse wood productivity also ranged from 0.19 to 22.54 Mg ha-1 yr-1, with a mean of 4.22 ± 0.39 Mg ha-1 yr-1. On average, AGB and CWP estimates obtained from the Chave model were significantly higher than those obtained from the other models (Table 5.6). The Henry model, which excludes tree height and wood density, gave the lowest AGB and CWP estimates.   Table 5.5. Summary of the taxonomic and structural variables estimated in the study. Estimates are based on trees ≥ 10 cm DBH. Data for the old-growth forest and 54-year-old logged-over forest are combined.  Variable Minimum Maximum Mean ± S.E Taxonomic variables    Tree species richness 8 26 14.76 ± 0.56 Effective number of species 7 23 12.42 ± 0.50 Tree species diversity 1.97 3.16   2.49 ± 0.04     Structural variables    Tree diameter (cm) 15.69 28.14 21.01 ± 0.43 Basal area (m2 ha-1) 7.19 68.92 27.93 ± 1.78 Tree density (m2) 10 41 23.98 ± 1.18 DBH, diameter at breast height       122 Table 5.6 Aboveground biomass (AGB) and Coarse wood productivity (CWP) estimates obtained from three allometric equations used in the study (Henry et al. (2010) and Chave et al. (2005)). Estimates are based on trees ≥ 10 cm DBH. Data for the old-growth forest and 54-year-old logged-over forest are combined. Different letters represent significant differences (p < 0.05) between the equations. Variable Allometric equations Henry Henry (H and WD included)  Chave AGB (Mg ha-1)     Minimum 12.60 13.50  34.50 Maximum 65.03 267.4  680.6 Mean ± SE 37.53 ± 1.72a 94.76 ± 7.31b  241.20 ± 18.61c      CWP (Mg ha-1 yr-1)     Minimum 0.19 0.47  1.20 Maximum 1.31 8.86  22.54 Mean ± SE 0.60 ± 0.04a 3.41 ± 0.31b  8.68 ± 0.79c DBH, diameter at breast height    5.4.2 Relationships between taxonomic, structural variables and aboveground biomass and productivity  Aboveground biomass related positively with all the taxonomic variables (tree species richness, effective number of species and tree species diversity) (Table A.6). Among the taxonomic variables, tree species richness explained the highest portion of the variation in AGB (57%) and CWP (18%) among the plots. Within each forest, however, there were distinct relationships between AGB or CWP and the taxonomic variables (Figure 5.3 and 5.4). In the old-growth forest, AGB was strongly and significantly related to all the taxonomic variables, while the relationships with CWP were weak (Figure 5.3A-C and 5.4A-C). The relationship between tree species richness and AGB was best described by a quadratic (second-order polynomial) model, with a higher adjusted R2 and lower AIC (Table 5.7; Figure 5.3A). In the logged-over forest, only tree species richness related positively with AGB (R2 = 0.32, p = 0.0033, Figure 5.3D). The multiple regression analysis indicated that tree species richness had no effect on AGB and CWP (after  123 accounting for the effects of tree density and tree diameter; Table 5.8). Moreover, the different allometric equations used to estimate biomass produced distinct relationships (Figure A.4). For example, the Henry equation (without height and wood density) showed a quadratic relationship between tree species richness and AGB (R2 = 0.59, p < 0.0001), whilst the Chave equation resulted in a very weak hump-shaped relationship (R2 = 0.19, p = 0.02).   Table 5.7 Linear and second-order polynomial models relating taxonomic variables and aboveground biomass (AGB) or coarse wood productivity (CWP) in the old-growth forest and the 54-year-old logged-over forest. The Akaike Information Criteria (AIC), adjusted R2 and coefficients are given for each model. The best models have lowest AIC and highest adjusted R2 in bold. AGB values were estimated with the Henry equation (excluding height and wood density) and CWP included height and wood density.  Variable Model R2  adjusted AIC a b c Old-growth forest AGB       Tree species richness Linear, AGB = a + b x SR 0.62 97.58 2.192 -0.1626 -  Quadratic, AGB = a x SR2 + b x SR + c  0.68 95.15 -43.39 8.556 -0.2185 Effective number of species Linear, AGB = a + b x ES 0.40 109.10 6.533 1.879 -  Quadratic, AGB = a x ES2 + b x ES + c  0.46 108.00 -39.29 9.16 -0.2679 Tree species diversity Linear, AGB = a + b x SD 0.45 106.90 -36.57 26.67 -  Quadratic, AGB = a x SD2 + b x SD + c  0.49 106.60 -282.2 222.3 -38.54        Old-growth forest CWP       Tree species richness Linear, CWP = a + b x SR 0.20 41.16 -0.8244 0.2868 -  Quadratic, CWP = a x SR2 + b x SR + c 0.16 43.99 -1.683 0.4181 -0.004639 Effective number of species Linear, CWP = a + b x ES 0.23 40.21 -1.03 0.3291 -  Quadratic, CWP = a x ES2 + b x ES + c 0.20 43.05 -1.737 0.4423 -0.004187 Tree species diversity Linear, CWP = a + b x SD 0.23 40.20 -7.307 4.183 -  Quadratic, CWP = a x SD2 + b x SD + c 0.21 42.74 11.69 -11.12 3.044        Logged-over forest AGB Linear, AGB = a + b x SR 0.29 104.30 21.64 1.436 -   Quadratic, AGB = a x SR2 + b x SR + c 0.26 109.70 7.033 3.167 -0.04922 SR, Tree species richness; ES, Effective number of species; SD, Tree species diversity     124  Figure 5.3 Relationships between taxonomic variables (tree species richness, effective number of species and tree species diversity) and aboveground biomass (AGB). Relationships are shown for the old-growth forest (top panel, A-C) and the 54-year-old logged-over forest (down panel, D-F). Regression lines are included for significant relationships. AGB values were estimated with the Henry equation (excluding height and wood density).   Tree density, tree diameter, and basal area related positively and linearly with AGB, with tree density explaining 83% of the variation among the plots (results not shown). The strong relationship between AGB and tree density and basal area was maintained when the analysis was separated for the old-growth forest and the logged-over forest (Figure 5.5A-F). In the old-growth forest, tree diameter showed a positive association with AGB (R2 = 0.33, p = 0.0025), but the relationship was very weak in the logged-over forest (R2 = 0.15, p = 0.06; Figure 5.5D-F). Coarse wood productivity and tree diameter and basal area were also positively related in the forests (Figure 5.6). In the multiple regression analysis, the models that best predicted AGB were those that included both tree density and tree diameter (Table 5.8).    0 7 14 21 28015304560Tree species richnessAGB (Mg ha-1) Old-growth forestA R2 = 0.71, p < 0.00010 8 16 24 32020406080Tree species richnessAGB (Mg ha-1) Logged-over forestD R2  = 0.32, p = 0.00330 5 10 15 20015304560Effective number of speciesAGB (Mg ha-1) B R2 = 0.42, p = 0.00040 7 14 21 28020406080Effective number of speciesAGB (Mg ha-1) E R2 = 0.13, NS1.5 2.0 2.5 3.0 3.5015304560Tree species diversityAGB (Mg ha-1) C R2 = 0.47, p = 0.00011.5 2.0 2.5 3.0 3.5020406080Tree species diversityAGB (Mg ha-1) F R2  = 0.12, NS 125  Figure 5.4 Relationships between taxonomic variables (tree species richness, effective number of species and tree species diversity) and coarse wood productivity (CWP). Relationships are shown for the old-growth forest (top panel, A-C) and the 54-year-old forest (down panel, D-F). Regression lines are included for significant relationships. CWP values were estimated with the Henry equation (including height and wood density).             0 7 14 21 280246810Tree species richnessCWP (Mg ha-1 yr-1) Old-growth forestAR2 = 0.23, p = 0.010 8 16 24 320246810Tree species richnessCWP (Mg ha-1 yr-1) Logged-over forestD R2  = 0.10, NS0 5 10 15 200246810Effective number of speciesCWP (Mg ha-1 yr-1) BR2 = 0.26, p = 0.00890 7 14 21 280246810Effective number of speciesCWP (Mg ha-1 yr-1) E R2 = 0.12, NS1.5 2.0 2.5 3.0 3.50246810Tree species diversityCWP (Mg ha-1 yr-1) CR2 = 0.26, p = 0.00881.5 2.0 2.5 3.0 3.50246810Tree species diversityCWP (Mg ha-1 yr-1) FR2  = 0.11, NS 126    Figure 5.5 Relationships between structural variables (tree diameter, basal area and tree density) and aboveground biomass (AGB). Relationships are shown for the old-growth forest (top panel, A-C) and the 54-year-old logged-over forest (down panel, D-F). Regression lines are included for significant relationships. AGB values were estimated with the Henry equation (excluding height and wood density).           10 15 20 25 30015304560Tree diameter (cm)AGB (Mg ha-1) Old-growth forestR2 = 0.33, p = 0.0025A10 15 20 25 30020406080Tree diameter (cm)AGB (Mg ha-1) Logged-over forestR2  = 0.15, NSD0 15 30 45 60015304560Basal area (m2 ha-1)AGB (Mg ha-1) R2 = 0.78, p < 0.0001B0 20 40 60 80020406080Basal area (m2 ha-1)AGB (Mg ha-1) R2  = 0.45, p = 0.0002E0 8 16 24 32015304560Tree density (m2)AGB (Mg ha-1) R2 = 0.85, p < 0.0001C0 12 24 36 48020406080Tree density (m2)AGB (Mg ha-1) R2  = 0.60, p < 0.0001F 127  Figure 5.6. Relationships between structural variables (tree diameter, basal area and tree density) and coarse wood productivity (CWP). Relationships are shown for the old-growth forest (top panel, A-C) and the 54-year-old logged-over forest (down panel, D-F). Regression lines are included for significant relationships. CWP values were estimated with the Henry equation (including height and wood density).             10 15 20 25 30036912Tree diameter (cm)CWP (Mg ha-1 yr-1) Old-growth forestR2 = 0.24, p = 0.01A10 15 20 25 30036912Tree diameter (cm)CWP (Mg ha-1 yr-1) Logged-over forestR2  = 0.23, p = 0.01D0 15 30 45 60036912Basal area (m2 ha-1)CWP (Mg ha-1 yr-1) R2 = 0.31, p = 0.0037B0 20 40 60 80036912Basal area (m2 ha-1)CWP (Mg ha-1 yr-1) R2  = 0.52, p < 0.0001E0 8 16 24 32036912Tree density (m2)CWP (Mg ha-1 yr-1) R2 = 0.07, NSC0 12 24 36 48036912Tree density (m2)CWP (Mg ha-1 yr-1) R2  = 0.12, NSF 128 Table 5.8 Multiple linear predictor models of aboveground biomass (AGB) and coarse wood productivity (CWP) in the old-growth forest and the 54-year-old logged-over forest. All models are significant at p < 0.05 in bold. AGB values were estimated with the Henry equation (excluding height and wood density) and CWP included height and wood density.  Response variable Model Predictor variable Old-growth forest   Logged-over forest CE S.E t-value P-value   CE S.E t-value P-value Ln AGB SR + D + TD Intercept 0.55 0.05 11.09 <0.0001  0.82 0.03 25.59 <0.0001   SR -0.01 0.003 -1.72 0.0995  -0.0007 0.001 -0.65 0.526   D 0.02 0.002 9.58 <0.0001  0.02 0.001 16.60 <0.0001   TD 0.03 0.002 10.97 <0.0001  0.01 0.0008 19.33 <0.0001  D + TD Intercept 0.57 0.05 10.87 <0.0001  0.82 0.03 25.93 <0.0001   D 0.02 0.002 9.11 <0.0001  0.02 0.01 16.84 <0.0001   TD 0.02 0.001 16.94 <0.0001  0.01 0.001 25.91 <0.0001  SR + TD Intercept 0.97 0.06 16.72 <0.0001  1.28 0.06 20.39 <0.0001   SR 0.004 0.008 0.55 0.58  0.001 0.004 0.24 0.814876   TD 0.002 0.005 4.25 <0.0001  0.01 0.003 4.34 <0.0001             Ln CWP SR + D + TD Intercept -0.15 0.41 -0.35 0.7283  -0.39 0.51 -0.76 0.4538   SR 0.03 0.03 1.12 0.2766  0.004 0.02 0.247 0.8075   D 0.05 0.02 2.65 0.0149  0.06 0.02 3.00 0.0069   TD -0.003 0.02 -0.16 0.8774  0.02 0.01 1.96 0.0689  D + TD Intercept -0.2 0.42 -0.47 0.6404  -0.38 0.5 -0.77 0.4482   D 0.06 0.02 3.12 0.0049  0.06 0.02 3.10 0.005   TD 0.02 0.01 1.45 0.1609  0.003 0.01 2.93 0.0077  SR + TD Intercept 0.8 0.24 3.34 0.00294  0.9 0.31 2.88 0.00879   SR 0.06 0.03 1.79 0.0872  0.01 0.02 0.45 0.65934   TD -0.01 0.02 0.02 0.6194  0.02 0.01 1.11 0.27726 CE, Coefficients; SE, Standard error; SR, Tree species richness; D, tree diameter; TD, tree density   5.4.3 Relationships between leaf traits and tree species biomass and productivity  Leaf chemical traits were similar among the tree species in both the old-growth forest and the logged-over forest. However, leaf morphological traits, such as SLA varied widely among the tree species, ranging from 73.17 to 249.38 cm2 g-1 (Table 5.9). Celtis zenkeri in the old-growth forest had the highest leaf N concentration and SLA, while leaf P was greater in Terminalia superba in the logged-over forest. Celtis  129 mildbraedii in the logged-over forest had the highest LDMC, but the lowest SLA (Table 5.9). When the tree species were grouped into functional groups or guilds, there were no consistent patterns in the leaf traits among the shade-bearers, pioneers and non-pioneer light-demanders (Table 5.10). Nonetheless, the leaves of the non-pioneer light-demanders had higher N concentrations, while the shade-bearers had higher LDMC in both the old-growth forest and the logged-over forest. On average, leaf N and SLA were significantly higher in tree species in the old-growth forest than in the logged-over forest. (Table A.7).  Leaf K was the only trait that related strongly and positively to tree biomass (p = 0.0082), accounting for 66% of the variance in the tree species in the logged-over forest (Figure 5.7B).  In the old-growth forest, no relationship existed between the traits and tree biomass. Leaf P and leaf N were significantly and positively related to tree species productivity in the old-growth forest (R2 = 0.46, p = 0.04, Figure 5.7A) and logged-over forest (R2 = 0.48, p = 0.04, Figure 5.7C). No significant relationship was found between the tree biomass or productivity and the leaf morphological traits.    Figure 5.7. Bivariate relationships between leaf traits and tree biomass and productivity for old-growth forest (A) and 54-year-old logged-over forest (B-C). Initials are: AB, Alstonia boonei; AZ, Albizia zygia; CM, Celtis mildbraedii; CZ, Celtis zenkeri; MC, Musanga cecropioides; NP, Nesogordonia papaverifera; PM, Pterygota macrocarpa; PMa, Petersianthus macrocarpus; SO, Sterculia oblonga; SR, Sterculia rhinopetala; TS, Triplochiton scleroxylon; TSu, Terminalia superba.   0.12 0.14 0.16 0.18 0.20010203040Leaf P (%)Tree productivity (kg yr-1) Old-growth forestR2 = 0.46, p = 0.04CZCMTSuMCAZPMTSNPSOA0.0 0.2 0.4 0.6 0.8080160240320Leaf K (%)Tree biomass (kg) Logged-over forestR2  = 0.66, p = 0.0082NPTSSRPMSOPMaCMABTSuB2.0 2.5 3.0 3.5 4.0-4004080120Leaf N (%)Tree productivity (kg yr1) Logged-over forestR2  = 0.48, p = 0.04TS CMNPSOTSu PMABPMaCSR 130 Table 5.9 Leaf traits (mean ± SE) of tree species sampled in the old-growth forest and the 54-year-old logged-over forest.  Tree species Leaf N (%) Leaf P (%) Leaf K (%) SLA (cm2 g-1) LDMC (g g-1) Old-growth forest           Celtis mildbraedii 3.19 ± 0.10 0.18 ± 0.02 0.55 ± 0.15 218.13 ± 113.37 0.55 ± 0.0 Nesogordonia papaverifera 3.95 ± 0.59 0.16 ± 0.01 0.43 ± 0.01 208.07 ± 74.90  0.52 ± 0.09 Terminalia superba 3.54 ± 0.04 0.16 ± 0.01 0.30 ± 0.03 106.47 ± 27.50 0.46 ± 0.15 Triplochiton scleroxylon 3.05 ± 1.20 0.16 ± 0.01 0.62 ± 0.03 193.94 ± 28.73 0.48 ± 0.10 Sterculia oblonga 3.81 ± 0.11 0.14 ± 0.02 0.34 ± 0.02 115.22 ± 6.38 0.67 ± 0.17 Pterygota macrocarpa 3.69 ± 0.36 0.15 ± 0.03 0.44 ± 0.11 204.76 ± 28.58 0.35 ± 0.03 Musanga cecropioides 2.47 ± 0.0 0.18 ± 0.0 0.45 ± 0.15 100.16 ± 10.04 0.40 ± 0.02 Albizia zygia 3.87 ± 0.86 0.13 ± 0.0 0.27 ± 0.0 244.15 ± 56.64 0.51 ± 0.21 Celtis zenkeri 4.08 ± 1.13 0.19 ± 0.01 0.55 ± 0.06 249.38 ± 1.16 0.49 ± 0.10       Logged-over forest      Celtis mildbraedii 3.40 ± 0.18 0.09 ± 0.04 0.28 ± 0.04  73.17 ± 0.96 0.61 ± 0.08 Nesogordonia papaverifera 2.71 ± 0.79 0.14 ± 0.0 0.54 ± 0.06   88.74 ± 10.44 0.54 ± 0.01 Terminalia superba 3.02 ± 0.62 0.27 ± 0.14 0.44 ± 0.16   97.27 ± 8.36 0.43 ± 0.05 Triplochiton scleroxylon 3.26 ± 0.04 0.16 ± 0.02 0.42 ± 0.11 122.78 ± 26.81 0.53 ± 0.12 Sterculia oblonga 3.05 ± 0.31 0.14 ± 0.02 0.25 ± 0.09 114.33 ± 30.82 0.59 ± 0.04 Pterygota macrocarpa 3.36 ± 0.07 0.18 ± 0.03 0.30 ± 0.19 137.83 ± 36.81 0.47 ± 0.11 Alstonia boonei 2.92 ± 0.11 0.12 ± 0.0 0.22 ± 0.05 130.94 ± 6.81 0.35 ± 0.01 Petersianthus macrocarpus 3.02 ± 0.14 0.11 ± 0.04 0.38 ± 0.04   97.01 ± 6.37 0.56 ± 0.06 Sterculia rhinopetala 2.95 ± 0.0 0.16 ± 0.02 0.52 ± 0.11 128.61 ± 12.66 0.52 ± 0.03 N, nitrogen; P, Phosphorus; K, potassium; SLA, specific leaf area; LDMC, leaf dry matter content          131 Table 5.10 Leaf traits (mean ± SE) for shade-bearers, pioneers and non-pioneer light demanders (NPLDs) in the old-growth forest and the 54-year-old logged-over forest. Functional group Leaf N (%) Leaf P (%) Leaf K (%) SLA (cm2 g-1) LDMC (g g-1) Old-growth forest      Shade-bearers 3.57 ± 0.33 0.17 ± 0.07 0.49 ± 0.07 213.20 ± 55.54 0.53 ± 0.04 Pioneers 3.26 ± 0.37 0.17 ± 0.02 0.45 ± 0.07 133.50 ± 21.88 0.45 ± 0.05 NPLDs 3.86 ± 0.28 0.16 ± 0.01 0.45 ± 0.06 203.40 ± 30.03 0.50 ± 0.07       Logged-over forest      Shade-bearers 3.04 ± 0.25 0.11 ± 0.01 0.40 ± 0.05  86.31 ± 5.44 0.57 ± 0.03 Pioneers 3.06 ± 0.17 0.18 ± 0.04 0.36 ± 0.07 117.00 ± 9.84 0.44 ± 0.05 NPLDs 3.12 ± 0.11 0.16 ± 0.01 0.36 ± 0.08 126.90 ± 13.53 0.53 ± 0.04 N, nitrogen; P, Phosphorus; K, potassium; SLA, specific leaf area; LDMC, leaf dry matter content  5.4.4 Relative importance of leaf traits, taxonomic and structural variables  The plot-level AGB and CWP were mostly explained by tree density and basal area (Figure 5.5 and 5.6). The results of the multiple regression analyses confirmed the dominant effects of the structural variables on AGB and CWP (Table 5.8). Indeed, the effect of tree density and tree diameter on AGB remained unchanged even after accounting for the effect of tree species richness (Table 5.8). The taxonomic variables were also important in explaining the variations in AGB and CWP, particularly in the old-growth forest (Figure 5.8). At the tree species level, leaf K better explained the interspecific variation in tree biomass, whilst leaf N and leaf P were the most important traits explaining the interspecific differences in tree species productivity (Figure 5.7A and 5.7C; Figure 5.8).         132  Figure 5.8. The relative importance of taxonomic and structural variables in explaining aboveground biomass (AGB) and coarse wood productivity (CWP) in the old-growth forest (black bars) and the 54-year-old logged-over forest (grey bars). Each bar represents the Pearson product-moment correlation coefficient between a variable and AGB or CWP. AGB values were estimated with the Henry equation (excluding height and wood density) and CWP included height and wood density.   5.4.5 Examining support for niche complementarity and the selection effect  There was no relationship between the standard deviation of wood density (WDSD) and AGB or CWP (Figure A.5), indicating a lack of support for niche complementarity in these forests. The magnitude and direction of the relationship did not change when the analysis was repeated for all 50 plots (results not shown). Community-weighted mean (CWM) of wood density and AGB and CWP were also not related (Figure A.6).    0.0 0.2 0.4 0.6 0.8 1.0Tree densityBasal areaTree diameterTree species diversityEffective number of speciesTree species richnessAGB (Mg ha-1)Old-growthLogged-overA-1.0 -0.5 0.0 0.5 1.0LDMCSLALeaf KLeaf PLeaf NCorrelation with wood biomassB0.0 0.2 0.4 0.6 0.8 1.0CWP (Mg ha-1 yr-1)C-1.0 -0.5 0.0 0.5 1.0Correlation with wood productivityD 133 5.5 Discussion  5.5.1 Forests characteristics  The high spatial variation in tree species richness, tree species diversity, basal area and tree diameter is consistent with findings from other tropical studies (Chisholm et al. 2013; Sullivan et al. 2017) that assessed forest dynamics at fine spatial scales (<1 ha). Plot-level AGB and CWP also differed by an order of magnitude (Table 5.6), which suggests that a variety of factors are involved in shaping the distribution of stem biomass and productivity in these forests. Differences in species richness or diversity, forest structure, edaphic factors and disturbance intensity are known to affect the spatial distribution of forest AGB and CWP at the same sites (Day et al. 2013; Jucker et al. 2016; Ledo et al. 2016; Quinto-Mosquera and Moreno, 2017). The values for tree species richness and tree species diversity in these forests are within the range reported for other tropical forests (Day et al. 2013; Chisholm et al. 2013; Poorter et al. 2015; Vaglio Laurin et al. 2016a). The comparable tree species diversity in the old-growth forest and the logged-over forest corroborate the findings of studies from Uganda (Owing et al. 2016) and Malaysia (Hector et al. 2011) that also reported similar tree species diversity in old-growth (unlogged) forests and selectively logged forests. In contrast, Asase et al. (2014) and Ding et al. (2017) reported higher tree species diversity in logged forests than in unlogged forests in Ghana and China, respectively. The conflicting results in the aforementioned studies may be due to differences in logging intensity and practices, edaphic factors, climate and forest characteristics (Paoli et al. 2008; Chazdon, 2014; Day et al. 2013; Amissah et al. 2014).   The mean tree diameter and basal area values are within the range of values reported for moist tropical forests globally (Durán et al. 2015), and for single sites in South America (Poorter et al. 2015) and Ghana (Fauset et al. 2012). The mean AGB estimates (124.50 Mg ha-1) fall within the range of values reported for other forests in Ghana (114.20-575.40 Mg ha-1, Asase et al. 2012; Fauset et al. 2012; Lewis et al. 2013), but are lower than the mean values reported for tropical forests in central Africa, Malaysia, and Peru (262.30-445.00 Mg ha-1, Lewis et al. 2013; Rowland et al. 2014). The mean CWP estimate (4.22 Mg  134 ha-1 yr-1) is higher than the value reported for broadleaved evergreen forests in Europe (0.49 Mg ha-1yr-1, Vilá et al. 2013) but lower than global averages reported for pan-tropical forests (Chisholm et al. 2013; Sullivan et al. 2017). Estimates of forest components presented here are conservative because I eliminated all trees ≥10 cm DBH that were not identified to the species level prior to the analysis.  5.5.2 Taxonomic variables relate to aboveground biomass and coarse wood productivity  The positive relationship between tree species richness and AGB is consistent with observational (Jean-Ruiz and Potvin, 2011; Chisholm et al. 2013; Poorter et al. 2015; Jucker et al. 2016; Vaglio Laurin et al. 2016a; Forrester and Bauhus, 2016; Sullivan et al. 2017) and experimental (Hector et al. 2011) studies of tropical forest ecosystems at fine spatial scales (< 1 ha). A strong relationship between tree species richness and AGB is expected at fine spatial scales (< 1 ha) because the influence of climate and environmental conditions are inherently controlled (Loreau et al. 2001; Sullivan et al. 2017).   In this study, neither niche complementarity nor selection effect contributed to the positive relationship between taxonomic variables and AGB or CWP (Figure A.5 and A.6), suggesting that other mechanisms may be more important in explaining variation in biomass and productivity. Other studies also found no support for niche complementarity and selection effect in tropical forest ecosystems (e.g., Lohbeck et al. 2015; Prado-Junior et al. 2016). Indeed, some authors suggest that the effect of species richness on biomass or productivity at fine spatial scales (<1 ha) could be attributed to the indirect influence of stand structural attributes like tree density or basal area rather than niche complementarity and selection effect per se (Chisholm et al. 2013; Vilà et al. 2013; Yuan et al. 2016; Bohn and Huth, 2017). Stand structural attributes such as tree density can increase or decrease the strength of complementary interactions, and increase competition for limiting resources in forest ecosystems (Potter and Woodall, 2014; Forrester and Bauhus, 2016). Therefore, any impact of species richness or diversity on forest biomass or productivity may be indirect, via stand structure or other factors that affect both the composition and functioning of forests (Chisholm et al. 2013; Forrester and Bauhus, 2016; Muscerella et al. 2016). After analyzing 55,  135 265 inventory plots in five European countries, Vilà et al. (2013) reported an increase in wood productivity with increasing tree species richness, which they attributed to the positive association between stand basal area and tree species richness. After accounting for stem density, species diversity had no effect on AGB in subtropical old-growth forests on limestone and volcanic soils in Puerto Rico (Musceralla et al. 2016). This may have been the case in the forests I studied since the relationship between tree species richness and AGB disappeared after adjusting for tree density and tree diameter (Table 5.8). In contrast, other observational studies in tropical (Sullivan et al. 2017), temperate and boreal (Paquette and Messier, 2011; Forrester and Bauhus, 2016) forests reported that niche complementarity and selection effect resulted in strong relationships between species richness and AGB. Ultimately, the relative importance of the two mechanisms in influencing relationships between taxonomy and AGB or CWP depends on the forest type and successional stage (Paquette and Messier, 2011; Yuan et al. 2016), as well as other interacting factors operating at the local scale (Forrester and Bauhus, 2016).    The distinct relationships observed in the old-growth forest and the logged-over forest is consistent with findings from other studies, which showed that the shape and the direction of the relationship between tree species richness and AGB changed during post-disturbance forest recovery (Guo, 2003; Lasky et al. 2014; Yuan et al. 2016). The positive linear relationship between tree species richness and AGB in the logged-over forest is expected because during the early to intermediate stages of succession, a combination of germination, colonization and the establishment of pioneers and shade-tolerant species leads to an increase in both biomass and species richness (Guo, 2003; Chazdon, 2014; Lasky et al. 2014). However, during the later stages of succession, biomass accumulation eventually results in the competitive exclusion of short-lived pioneer cohorts in the canopy (Grime, 1973; Chazdon, 2014), and a hump-shaped or negative relationship develops (Guo, 2003; Lasky et al. 2014). The importance of the taxonomic variables was greater in the old-growth forest than in the logged-over forest (Figure 5.3 and 5.4), consistent with findings from studies that reported stronger biodiversity effects on biomass and  136 productivity in old-growth forests than in disturbed or secondary forests (Cai et al. 2016a; Sande, 2016; Yuan et al. 2016).   5.5.3 Structural variables are better predictors of aboveground biomass and coarse wood productivity  The dominant role of the structural variables (tree diameter, tree density and basal area) in explaining AGB and CWP corresponds well with the ‘vegetation quantity hypothesis’ (Lohbeck et al. 2015), which postulates that forest structural components are key drivers of ecosystem processes. The result also supports previous observational (Paquette and Messier, 2011; Lewis et al. 2013; Slik et al. 2013; Vilá et al. 2013; Durán et al. 2015; Zhang et al. 2017) and theoretical studies (e.g., Bohn and Huth, 2017) that showed a strong relationship between stand variables and AGB or CWP. Plot-level basal area was the strongest predictor (R2 = 0.79) of AGB across 145 forest plots in the tropics (Durán et al. 2015). In subtropical (Zhang et al. 2017), temperate (Vilá et al. 2013) and boreal (Paquette and Messier, 2011) forests, basal area, tree density and tree diameter positively influenced the distribution of AGB and CWP under different climates and soil conditions. The positive relationship between stand structures and AGB does not exist in all forest ecosystems (Slik et al. 2010; Lewis et al. 2013). For instance, Slik et al. (2010) found no relationship between stem density and AGB in 83 old-growth forests in Borneo, Malaysia.   The positive relationship between the tree diameter, basal area, tree density and AGB could be associated with the linkage between stand structures and factors essential for growth, such as light and water (Binkley et al. 2013; Forrester and Bauhus, 2016). Large trees, which dominate the forest canopy, tend to have high leaf area and leaf mass per area (Ryan et al. 2006), and also intercept more light and use it more efficiently for growth and biomass production (Binkley et al. 2013). Therefore plots that contain large trees are likely to accumulate more biomass (Sheil et al. 2017). Also, at higher stand densities, stand leaf area may increase, and the trees can capture more light for growth (Forrester and Bauhus, 2016). On  137 the other hand, higher tree density can result in intense competition for resources, which can affect forest productivity (Stephenson et al. 2011).   5.5.4 Leaf chemical traits relate with tree biomass and productivity  Leaf chemical traits (leaf N, P, K) were strongly related to tree biomass and productivity because they are related to processes such as the photosynthetic capacity and the respiration rates of trees (Pérez-Harguindeguy et al. 2013; Reich, 2014; Walker et al. 2014). Globally, the maximum rate of carboxylation of Rubisco and the maximum rate of electron transport increased with leaf N and P contents of forest trees (Walker et al. 2014). The pattern observed for leaf N and P and tree productivity is consistent with results from trait-based studies at the community level that reported strong relationships between leaf N or leaf P and stand growth (Mercado et al. 2011; Reich, 2012; Finegan et al. 2015; Sande, 2016). However, other studies reported a weak (Wright et al. 2010; Visser et al. 2016) or no (Poorter et al. 2008; Hérault et al. 2011) relationship between tree productivity and leaf N or leaf P at the species-level. Leaf N did not relate to tree growth in 5,524 individuals of 50 tree species in a wet tropical forest in French Guiana (Hérault et al. 2011). Comparative studies on leaf K concentration and tree biomass are scarce, but Tripler et al. (2006) reported that K fertilization led to increase in tree growth in many forest ecosystems. Potassium is considered an important modulator of photosynthetic rate and structural properties of forests (Lloyd et al. 2015). Indeed, K enhances CO2 assimilation in trees by regulating stomatal conductance and assimilation efficiency (Matthius, 2009; Erel et al. 2014).   The lack of relationship between leaf morphological traits and tree biomass or productivity agrees with a recent study, which reported a weak relationship between leaf mass per area (i.e. inverse of SLA) and the growth of 136 tree species in a moist tropical forest on Barro Colorado Island in Panama (Visser et al. 2016). When the tree species were grouped into guilds, there was no consistent pattern in the leaf traits of pioneers and shade-tolerant species. This contradicts the ‘leaf economic spectrum hypothesis’ (Reich, 2014), according to which, there should be an ‘acquisitive-conservation’ trait trade-off between pioneers  138 and shade-bearers. Usually, pioneer tree species are expected to have acquisitive leaf traits such as high SLA that ensure fast growth, whereas shade-tolerant species should have traits that allow for conservation of resources (e.g., high LDMC) (Chazdon, 2014; Reich, 2014).   5.6 Conclusions Stand structural variables were good predictors of AGB and CWP in these forests. Tree species richness related to AGB and CWP, but the effect was via the indirect influence of tree density. Both taxonomic and structural variables related better to AGB than with CWP, which reinforces the need to separate AGB and CWP when assessing the relationships among species richness, biomass and productivity. Relationships between the taxonomic variables and AGB were stronger in the old-growth forest than in the logged-over forest. I did not find evidence of niche complementarity and selection effect in either forest, suggesting that other mechanisms are important in driving the relationships. The shape and magnitude of the relationship between tree species richness and AGB depended on the allometric equation used to estimate tree biomass. Therefore, care should be taken when selecting allometric equations. Despite a wide interspecific variation in SLA, in both the old-growth forest and the logged-over forest, there was no evidence of a leaf ‘acquisitive-conservative’ trait trade-off between the pioneers and the shade-tolerant species, which is inconsistent with the ‘leaf economic spectrum hypothesis’. Overall, multiple factors, including tree diameter, basal area and tree density all influenced aboveground biomass in these forests.          139 Chapter 6: Biomass and productivity are similar, but allocation patterns differ in old-growth forest and logged-over forest in Ghana 6.1 Synopsis There is paucity of information on how biomass, net primary productivity, and the allocation of photosynthetic products to different plant components change in tropical forests after several decades of logging. In this study I quantified the components of above- and belowground biomass and NPP in an old-growth forest and a 54-year-old logged-over forest between August 2013 and June 2015. I also determined the allocation of NPP between canopy, wood and fine roots in both forests. Mean leaf area index, diameter, and height were similar in the two forests. Stand density and basal area (stem ≥ 10 cm diameter at breast height) were 35% and 26% higher in the logged-over forest than in the old-growth forest. Wood density (trees ≥ 10 cm DBH) was 0.57 ± 0.01 g cm-3 and 0.53 ± 0.01 g cm-3 in the old-growth forest and logged-over forest, respectively. Total biomass was similar in both forests (old-growth forest: 176.89 ± 20.38 Mg ha-1; logged-over forest: 180.02 ± 15.37 Mg ha-1). Aboveground biomass was 157.49 ± 20.28 Mg ha-1 and 155.18 ± 15.16 Mg ha-1, corresponding to more than 86% of the total biomass in the old-growth forest and logged-over forest, respectively. Total NPP was 8% higher in the old-growth forest (15.18 ± 1.88 Mg-1 ha-1 yr-1) than in the logged-over forest (13.89 ± 1.89 Mg-1 ha-1 yr-1). Stem (51-59%) and fine root (84-88%) productivity contributed the highest proportion of the aboveground and belowground NPP, respectively. In both forests, the fraction of NPP allocated to canopy was similar. However, greater NPP was allocated to fine roots in the old-growth forest, whereas in the logged-over forest greater NPP was allocated to wood. A shift in the allocation of NPP between wood and fine roots is consistent with recent theory, which suggests that a tradeoff between wood and fine roots dominates allocation patterns in forest ecosystems. The study highlights the need to consider successional stage, and the trade-offs that exist between different plant parts when modelling allocation in forests.      140 6.2 Introduction  Tropical forests cover a small land area (ca. 10%) of the Earth’s land surface (FAO, 2001) but account for more than 70% of the total global forest biomass and productivity (Pan et al. 2013; Ma et al. 2015). Tropical forests contribute significantly to the global carbon (C) cycle, representing 55% of the C pool in forests (Pan et al. 2011). Biomass accumulation and net primary productivity (NPP) are key descriptors of forest functioning (Malhi et al. 2009; Pan et al. 2013; Liang et al. 2016). Biomass represents the amount of live organic material present (per unit area) at any point in time. Net primary productivity is the rate of production of new biomass (per unit area), after accounting for autotrophic respiration. In addition, NPP also includes losses due to volatilization, flux to mycorrhizal symbionts and root exudation (Clark et al. 2001a). Understanding the spatial and temporal variation in NPP, and the driving factors has been the focus of large-scale research in tropical forests (e.g., Clark et al. 2001b; Malhi et al. 2009; Anderson-Teixeira et al. 2016). Also, the allocation of NPP (i.e. fraction of NPP used by a plant part) to canopy, wood and fine roots affects plant growth, C turnover and the uptake of resources in forest ecosystems (Litton et al. 2007; Malhi et al. 2011; Vicca et al. 2012). The fraction of NPP allocated to canopy influences leaf area, which directly affects the total productivity of forests (Reich, 2012). The fraction of NPP to fine roots and their symbionts influences the input of C to soil organic matter and the uptake of soil resources. Hence, any shifts in NPP allocation have the tendency to affect many processes that directly affect plant growth. Efforts devoted to quantifying total biomass and NPP, and understanding NPP allocation in tropical forests have mainly focused on the Amazon (e.g., Clark et al. 2001a; Aragão et al. 2009; Malhi et al. 2009; 2011; 2017; Jiménez et al. 2014; Girardin et al. 2010; 2016), and sites in Asia (Hertel et al. 2009b; Saner et al. 2012; Katayama et al. 2013; Kho et al. 2013; Kotowska et al. 2015; Gautam and Mandal, 2016), but African tropical forests are understudied (John, 1973; Nyirambangutse et al. 2017). For instance, a global study that analyzed NPP allocation in tropical forests did not include any data from Africa (Malhi et al. 2011). Given that African tropical forests differ in structure, floristic composition and richness from those in South America and Asia (Phillips et al. 1994; Banin et al. 2012;  141 Lewis et al. 2013; Sullivan et al. 2017), more empirical data are needed on these forests to accurately predict terrestrial ecosystem productivity and NPP allocation in forest ecosystems (Franklin et al. 2012).   Logging, mostly through selective harvesting of large, high-valued trees is central to forest management in the tropics (Kotey et al. 1998; Blaser et al. 2011; Chazdon, 2014). Most studies show that logging affects forest biomass and NPP through changes in the residual stand structure and resource availability (Hertel et al. 2007; Figueira et al. 2008; Saner et al. 2012; Berenguer et al. 2014; Gautam and Mandel, 2016). Notably, logging increases the recruitment of shade-intolerant species, and enhances the growth of saplings (Swaine and Agyeman, 2008; Chazdon, 2014), but also reduces the density of large trees and increases stem mortality (Figueira et al. 2008). The few studies that have compared above- and belowground biomass and NPP after logging have mainly reported higher total biomass and NPP in old-growth (unlogged) forests than in logged forests. Berenguer et al. (2014) reported higher aboveground biomass in old-growth forests than in logged forests in eastern Amazon. Fine-root biomass was significantly higher in moist tropical unlogged forests than logged forests globally (Hertel et al. 2007). Twenty-two years after logging, total biomass was 55-66% lower in logged forests than unlogged forests in Borneo, Malaysia (Saner et al. 2012). Similarly, total biomass and NPP estimates in old-growth forests were higher than those of logged forests in Sunsari district, eastern Nepal (Gautam and Mandal, 2016).  In addition, the allocation of NPP to wood productivity increased 3 years after logging in a wet forest in the Amazon (Figueira et al. 2008). In the aforementioned studies, the authors suggested that the removal of large trees, and high mortality of residual trees, contributed to the low biomass and NPP in the logged forests. Moreover, the amount of light reaching the forest floor also increased (Figueira et al. 2008). Nevertheless, forest structure, plant composition and soil fertility recover during post-logging recovery (Asase et al. 2012; 2014; Gatti et al. 2015; Alamgir et al. 2016). Therefore, it is imperative to determine the changes in biomass, NPP, and the allocation of NPP to different plant parts in tropical forests after several decades of logging, since productivity and C allocation change with stand development (Ryan et al. 1997; Litton et al. 2007; Anderson-Teixeira et al. 2016).   142 In this study, I quantified the components of biomass and NPP in an old-growth forest and a 54-year-old logged-over forest to address the following questions: (1) Do above- and belowground biomass differ between the old-growth forest and logged-over forest? (2) Are there differences in the estimates of NPP, and its components in the two forests? (3) Is the allocation of NPP between canopy, wood and fine roots similar in the old-growth forest and logged-over forest?  I hypothesized that total biomass and annual NPP would be higher in the old-growth forest than the logged-over forest. This hypothesis is consistent with the observation that logging reduces the density of large trees (Figueira et al. 2008) and increases stem turnover through time (Osazuwa-Peters et al. 2015a). I also predicted that a higher fraction of NPP would be allocated to wood productivity in the logged-over forest, whereas higher NPP would be allocated to fine roots in the old-growth forest. Usually, forests recovering from disturbance tend to allocate new biomass to aboveground components in response to increased competition to capture light (Bloom et al. 1985; Aragão et al. 2009; Anderson-Teixeira et al. 2016).    6.3 Materials and method 6.3.1 Study area  The study was conducted in the Bobiri Forest Reserve, situated in the Ejisu-Juabeng District of the Ashanti Region, in southern Ghana (find details in Chapter 3).   6.3.2 Study forests  For this study, plots were set up in the forests of the research block and the protection block. The research forest (hereafter referred to as the ‘logged-over’ forest) had been selectively logged 54 years earlier (Forestry Department, 1958; Djagbletey, 2014). No commercial logging has been allowed in the protection (old-growth) forest, which was set aside for ecological purposes (Forestry Department, 1958). The two study forests are similar in all factors related to energy budget, moisture and soil fertility (Chapter 3).    143 6.3.3 Aboveground biomass 6.3.3.1 Leaf biomass Leaf biomass was estimated from Leaf Area Index (LAI) and Specific Leaf Area (SLA) data (Hertel et al. 2009b). Leaf area index was determined indirectly using the Beer-Lambert law (eqn. 1):     LAI = -In (PARb/PARa)/k        (1)  where LAI is leaf area index (m2 m-2), PARb and PARa are photosynthetically active radiation (umol m-2 s-1) below and above the canopy, respectively, and k is the light extinction coefficient. I used the mean k value of 0.59, which is the value reported for broadleaf forests in the global metanalysis of Wang et al. (2014). Photosynthetically active radiation was measured with a quantum sensor and meter (±5 % accuracy, Model MQ-200, Apogee Instruments, USA). PAR measurements were carried out simultaneously with two sensors under the forest canopy and on forest roads ca. 200-400 m from the plots to represent below- and above-canopy PAR, respectively. All the PAR recordings under the canopy were systematically taken at 25 points ca. 1 m above the ground surface. PAR was measured biweekly or monthly from August 2013 to May 2015 near or immediately after solar noon when light was unobstructed by cloud cover. This indirect approach yields comparable LAI estimates with those from direct approaches (Asner et al. 2003). In fact, the mean LAI for the old-growth forest (5.87 ± 0.30 m2 m-2) and the logged-over forest (6.56 ± 0.19 m2 m-2) is close to the mean values of 3.9-6.0, which is the typical LAI range for tropical forests (Asner et al. 2003; Malhi et al. 2011; Clark et al. 2017). To determine SLA fresh leaves were collected and scanned to obtain the leaf area using IMAGEJ, v. 1.38 software (National Institute of Health, USA). Futher details on the equipment and the scanning procedure are provided in Chapter 5. The leaves were oven-dried at 70 °C to determine their dry mass. Specific leaf area (m2 g-1) was calculated as the one-sided leaf area divided by leaf dry mass. Leaf biomass was then determined using eqn. (2) following Hertel et al. (2009b):     144  Lb = LAI x 1/SLA         (2)  where Lb is leaf biomass (in g m-2), LAI is leaf area index (in m2 m-2), and SLA is specific leaf area (in m2 g-1).  The leaf biomass values were converted to Mg ha-1.   6.3.3.2 Stem biomass A 1-ha plot (100 m x 100 m) was established in each forest following the Global Ecosystem Monitoring (GEM) protocols manual (Marthews et al. 2012; gem.tropicalforests.ox.ac.uk) to inventory all vascular plants. The plots were subdivided into twenty-five 20 m x 20 m subplots. Plots of this size may be associated with sampling error (Clark et al. 2001a; Chave et al. 2004), but several NPP-related studies have used a plot size of 1 ha due to the laborious and expensive nature of such studies (e.g., Doughty et al. 2014; Girardin et al. 2016). In October 2013, all trees, shrubs, palms, and lianas ≥ 10 cm DBH were tagged, mapped, marked with paint, and the diameter measured at 1.3 m from the ground using a diameter tape. For trees with buttresses, the diameter was measured ca. 50 cm above the buttress using a ladder. Total tree heights were also measured on 50 individuals in each forest using a LaserAce hypsometer (Measurement Devices Ltd. UK). Trees selected for height measurements included a range of functional groups or guilds (pioneers, non-pioneer light demanders and shade-bearers, Swaine and Whitmore, 1988; Hawthorne, 1995) and DBH classes (10-20 cm, 20-50 cm and > 50 cm). Pioneers are shade intolerant species, and do not establish in the forest understory; non-pioneer light-demanders (NPLDs) can survive in the understory during the seedling stage, but require light for further growth; and the shade-bearers are shade tolerant so survive and grow in the forest understory.   Lianas were measured following Gerwing et al. (2006). Liana stems generally grow horizontally and are mostly non-circular in shape (Figure 6.1A), and hence measuring the diameter at DBH fails to better represent the structural characteristics. Liana diameter was recorded at three points: i) on the stem 130 cm from the main root position; ii) 130 cm vertically from the ground, and iii) 250 cm along the stem from  145 the rooting point. Liana stems with irregularities (e.g., nodes) were measured 5 cm below the anomaly (Gerwing et al. 2006). Lianas included all climbers that germinate in the forest floor, with their main rooting point inside a subplot. The structure of the palms made it impossible to measure the diameter, so heights were estimated visually. Palm height highly correlates with the AGB (Ansari et al. 2013). To account for biomass of small stemmed plants (2-9.9 cm DBH), five 10 m x 10 m nested subplots were established in the main plot. To account for the spatial variation in the plots, four of the subplots were placed at the corners and one in the middle of the main plot. All woody stems were tagged, mapped, and their diameters were measured with a tape to a resolution of 1 mm. Information on species names was collected following Hawthorne and Abu-Juam (1995) and Hawthorne and Jongkind (2006).  I calculated the biomass (kg) of individuals using different allometric equations to reflect the variations in the various life forms (Table 6.1). Biomass of trees (≥ 10 cm DBH) was estimated by using a locally derived allometric equation based on diameter, tree height and wood density (Henry et al. 2010). Wood density values were obtained from the World Agroforestry Centre (ICRAF, 2008) and the global tropical forest wood-density database (Chave et al. 2009). For those species with no wood density, I used the genus or family level values (Poorter et al. 2015). For individuals that could not be identified, plot-level wood density value was used. I used the height and diameter measurements from the 100 trees to generate an equation (3) to predict the heights of all trees in both forests. The height equation was:   H = a + b x D + c x D2       (3)   where H is height, D is DBH, and a = 4.026, b = 0.70005, c = -0.002657 are coefficients.  Biomass of trees (2-9.9 cm DBH) was also estimated using a locally derived allometric equation based on only DBH (Henry et al. 2010). The strength of this set of allometric equations is that it included tree species from all guilds, including pioneer, non-pioneer light demanders and shade-bearers, which  146 characterize the floristic composition and diversity range of tropical forest trees (Chazdon, 2014; Agyeman et al. 2016). The equations have already been applied to estimate tree biomass in moist forests in Africa, including Ghana (e.g., Asase et al. 2012; Ngomanda et al. 2014; Asase and Tetteh, 2016). Palm biomass was estimated with an allometric equation (Table 6.1) specific for palm following Brown (1997). For lianas, I estimated the biomass using locally derived equations for primary and secondary moist forests (Addo-Fordjour and Rahmad, 2013). Plot-level biomass (Mg ha-1) was obtained by adding the biomass of all the trees, palms and lianas in each subplot, and expressed on a per ha basis.   Table 6.1 Allometric equations used to estimate tree, palm and liana biomass for the old-growth forest and 54-year-old logged-over forest.  Life form Equation Source Tree (≥ 10 cm DBH) Y = 0.00347 + 0.002 ρD2H Henry et al. (2010) Tree (2-9.9 cm DBH) Y = 0.30 x D exp (2.13) Henry et al. (2010) Palm Y = 10.0 + 6.4 x H Brown (1997) Liana (primary forest) Y = 1.077 + 0.850 ln D Addo-Fordjour and Rahmad (2013) Liana (secondary forest) Y = 0.236 + 1.128 ln D Addo-Fordjour and Rahmad (2013) Y, biomass (kg); H, height (m); D, diameter at breast (DBH, cm); ρ, wood density (g cm-3)   6.3.3.3 Fine and coarse woody debris   Fine woody debris (diameter < 2 cm) was sampled within quadrats (0.5 m x 0.5 m) from 25 randomly selected locations in each forest (Marthews et al. 2012). Samples were sent to the laboratory, and separated into components of leaves, fruits, seeds, and twigs and unidentified (litter that was not grouped into leaves, twigs or reproductive organs). The twigs were oven-dried at 70 °C to constant mass, and weighed to determine necromass. The necromass (g m-2) was expressed in per ha.   Coarse woody debris (CWD) was quantified using the line-intersect method (Van Wagner, 1968) following the sampling procedures described by Addo-Danso (2012). In this study CWD refers to all  147 fallen dead debris (diameter ≥ 2 cm), excluding standing dead trees and stumps. CWD pieces were only considered if more than 50% of their thickness was above the soil surface. Standing dead trees were quantified separately during the second inventory for stem biomass. Four 100-m long transects were established randomly in each forest. Sample points were systemically positioned at 20 m intervals along each transect, and the point edges were marked with plastic strings secured to the ground. CWD pieces with a minimum diameter ≥ 2 cm at the point of intersection with any of the transect lines were tallied, and length and diameter recorded (Figure 6.1B). The lengths of CWD pieces were measured along the centre of axis regardless of shape (Marshall and Davis, 2002), and restricted to portions that were within the mininmum diameter threshold. For forked pieces, the fork segment with diameter  ≥ 2 cm was tallied and measured with a caliper. For non-circular/elliptical CWD pieces, diameters were recorded in two perpendicular directions, and their geometric mean calculated. Some large CWD have void spaces, so the volume of the piece was estimated after accounting for the void spaces (eqn. 5) following suggestions of Clark et al. (2002). During sampling, care was taken to discard the pieces that were collected on the lines because the same lines were used to quantify branch turnover production.   CWD pieces were separated into three diameter size classes, 2-5 cm (Class 1), 5-9.9 cm (Class II) and ≥ 10 cm (Class III) (Palace et al. 2012). Furthermore, CWD pieces were categorised into one of five decay classes (Table 6.2) modified from Harmon et al. (1995), based on the presence of certain physical features and structural integrity. The CWD pieces were sent to the laboratory to determine the density for each decay or size class. To determine the density of CWD pieces in the 1-3 decay class, I cut rectangular plugs from wood sections that were removed with a machete or saw in the field. The volume of each plug was estimated from the height, width and depth (Palace et al. 2012). For plugs that were not rectangular, and pieces in decay class 4-5, the volume was determined using the water displacement method (Larjavaara and Muller-Landau, 2011). To avoid bias from structurally less stable pieces (especially decay 5), pieces were first wrapped in a cling film to prevent water from penetrating the wood and saturating the pieces. The plugs and the other pieces were oven-dried at 80 °C to constant mass, and  148 weighed to determine their dry mass. The density and volume data were then used to quantify necromass (Palace et al. 2012). The volume (m3 ha-1) of the CWD pieces was estimated following van Wagner (1968):   V = (π2 ∑ D2)/8L       (4)  Vv = πr2 (h/3)         (5)   Where V is the volume of piece per ha (m3 ha-1), Vv is volume of cone (m3), D is piece diameter (in cm), L is the total length of each transect (m), r is the radius of cone (cm), and h is the height of cone (m).    Table 6.2 Description of decay classes for coarse woody debris (CWD) (modified from Harmon et al. 1995)  Decay class  Description 1 Firm wood with extensive and firm bark; leaves and twigs still attached 2 Solid wood, but starting to decay; bark intact, but starting to fall off; leaves and twigs absent 3 Non-solid wood, but firm when pressure applied; bark, leaves and twigs absent 4 Soft rotten wood; outer portions case-hardened and inner decomposing; absence of bark, leaves and twigs 5 Soft rotten wood; elliptical in shape; no bark, leaves and twigs      149   Figure 6.1 Liana (left, A), measuring diameter of coarse woody debris (right B). Photos: S.D. Addo-Danso   6.3.4 Belowground biomass  6.3.4.1 Fine- and small-root biomass Fine (diameter < 2 mm) and small (diameter 2-5 mm) root biomass was determined using the soil-core method (Vogt et al. 1998). Samples were taken with a hand-driven soil auger (diameter 5.5-cm) to 30-cm depth in twelve plots (measuring 10 m x 10 m) randomly established across the study forests. Root samples were taken at three randomly located points within each plot. During soil sampling, when an obstacle such as a large structural root or stone obstructed the auger, it was relocated within an area of ca. 25 cm2 until a suitable core was extracted. Soil samples were kept separate for three depth intervals (0-10, 10-20, 20-30 cm), and transferred into plastic bags for laboratory processing at the Forestry Research Institute of Ghana, ca. 18 km from the study site. Samples were refrigerated at 4 °C for 7 days before processing. Samples were soaked in plastic bowls, and washed thoroughly using a 0.25-mm sieve to remove soil particles and debris. Collected roots were separated into fine roots (diameter < 2 mm) and small roots (diameter 2-5 mm), and into live and dead based on visual inspection of morphological features such as colour, tensile strength, and cortex and periderm characteristics (Vogt et al. 1998). Live  150 and dead roots were oven-dried at 70 °C to determine their dry mass. Estimates are presented for only root biomass (live roots), and are also combined for all three soil depths.    6.3.4.2 Coarse-root biomass Coarse roots (diameter > 5 mm) are extremely difficult to sample in the field (Addo-Danso et al. 2016). Coarse-root biomass (CRB) was therefore estimated as a product of the stem biomass and a root-shoot ratio of 0.21 (Malhi et al. 2009). This root-shoot ratio was estimated by Malhi et al. (2009) based on values reported for the tropics in the global root distribution analyses of Jackson et al. (1996) and Cairns et al. (1997). This approach is commoly used in tropical C studies (e.g., Jiménez et al. 2014; Nyirambangutse et al. 2017), but it may over- or under-estimate CRB because root-shoot ratios are driven by different factors, including forest types, site characteristics and climate (Addo-Danso et al. 2016; Waring and Powers, 2017).  6.3.5 Aboveground net primary productivity   6.3.5.1 Canopy productivity  The annual canopy productivity was estimated using litterfall production (Clark et al. 2001a).  The use of litterfall as a proxy for canopy productivity is based on the assumption that under a near steady-state condition the annual litterfall produced by a stand is approximately equivalent to the canopy production (Clark et al. 2001a; Malhi et al. 2011). However, this is not always the case in forests where litterfall spatial pattern is aggregated, rather than uniform (Malhi et al. 2011). Twenty-five litter traps measuring 0.5 m x 0.5 m were installed in the middle of the 20 m x 20 m subplots in each forest. The traps were made from polyvinyl chloride (PVC) tube frames and fine nylon mesh netting (1-mm pore size) (Figure 6.2A, Marthews et al. 2012). The traps were placed ca. 1 m above the ground surface to avoid possible disturbances from mammals (Chave et al. 2010). Litter was collected every 15 days to minimize decomposition in the traps (Clark et al. 2001a). The litter samples were sent to the laboratory and, separated into leaves, twigs (diameter < 2 mm, including bark), fruits, flowers, seeds and unidentified.  151 The components were oven-dried at 60 °C to constant mass, and weighed to determine the dry mass. The sum of all the components was used to estimate canopy productivity. The canopy productivity was underestimated because the values were not corrected for missing components and losses (Clark et al. 2001a). Losses due to herbivory and in situ decomposition were not considered in the calculation. Furthermore, fruits and leaves produced by understory plants were not captured. Litter trapped in the canopy (Figure 6.2B), and leaves from plants such as palm were not captured. Hence, the annual canopy productivity estimates are considered conservative.       Figure 6.2 Litterfall trap (left photo, A), litter trapped in trees were not collected in the litter traps (right photo, B). Photos: S.D. Addo-Danso   6.3.5.2 Stem productivity  In June 2015, the plots were re-censused to incorporate new recruits and also to assess mortality. Stem productivity was determined as the change in biomass of surviving trees, palms and lianas between the second and first censuses, plus the increment of new recruits and mortality (Clark et al. 2001a). The data was thoroughly checked for errors, anomalous diameter changes and missing diameter values to avoid bias in the estimates (Muller-Landau et al. 2014; Shiel et al. 2017). Details of the data cleaning and gap- 152 filling procedures were based on suggestions of Talbot et al. (2014) and Sheil et al. (2017), and are provided in Chapter 5.   Change in biomass of surviving individuals was estimated as the biomass difference between 2013 and 2015. To estimate the biomass of recruits between 2013 and 2015, I subtracted the biomass of each individual with DBH of 10 cm or 2 cm (for small stemmed plants) in 2013 from its biomass in 2015. This assumes that the recruits had a DBH of 10 cm or 2 cm at the start of the census (Clark et al. 2001a; Talbot et al. 2014). For mortality, dead individuals in the second census were assigned values of zero (Muller-Landau et al. 2014; Talbot et al. 2014). This ensured that the approaches used to estimate biomass of recruits and that of dead individuals were comparable (Talbot et al. 2014). Plot-level stem NPP (Mg ha-1 yr-1) was estimated as the sum of change in biomass of surviving individuals and recruits minus that of mortality divided by the census interval (in years).   6.3.5.3 Branch turnover productivity  Fallen branches, and other woody debris (diameter ≥ 2 cm) were assessed on the same transect lines used to quantify CWD necromass. Surveys were repeated on the four 100-m transect lines at three-month intervals in each forest. Data collected on transects periodically were used to determine branch turover NPP (Mg ha-1 yr-1). The density and volume of same size-decay classes were estimated following the same procedure described under CWD. The field sampling was carried out from October 2013-May 2015.  6.3.6 Belowground net primary productivity  6.3.6.1 Fine- and small-root productivity  The ingrowth-core method was used to quantify fine- and small-root productivity (Vogt et al. 1998). A soil auger was used to extract soil samples to soil depth of 30-cm from twelve random locations in each forest. The soil was hand-sifted to manually remove the roots following the ‘temporal prediction technique’, which uses a maximum-likelihood approach to estimate root biomass after correcting for  153 under-estimation of very fine roots (Metcalfe et al. 2007). Details of the ‘temporal prediction technique’ are provided in Chapter 4. After root removal, ingrowth cores made of cylindrical mesh bags (12-cm diameter, 40-cm long) were installed. During installation care was taken to ensure that root-free soil bulk density was similar to the surrounding undisturbed soil (Vogt et al. 1998). Roots were manually retrieved from the ingrowth bags at approximately 3-4-month intervals according to the ‘temporal prediction technique’. Collected samples were transferred into plastic bags for laboratory processing at the Forestry Research Institute of Ghana. Samples were either processed immediately or kept refrigerated (4 °C) for 2-4 days before processing. Overall, roots were collected from November 2013 to April 2015. The annual root productivity (Mg ha-1 yr-1) was estimated as the sum of live and dead roots produced (Neill, 1992), scaled to one year.   6.3.6.2 Coarse-root productivity  I estimated coarse-root productivity (CRP) by multiplying stem productivity by 0.21 (Malhi et al. 2009). The BNPP estimate excluded the contributions from mycorrhizae and root exudate, which can be substantial (Clark et al. 2001a).   6.3.7 Allocation of net primary productivity  In this study allocation refers to the fraction of total NPP used by a plant part (Malhi et al. 2011; Jiménez et al. 2014). Allocation as defined here follows after Malhi et al. (2011) and Jiménez et al. (2014), which is different from that of Litton et al. (2007), who described allocation as a fraction of total gross primary production. Total NPP was separated into canopy (all litterfall components), wood (stem, branch, and coarse root) and fine root (fine root and small root together). Coarse-root NPP was added to wood NPP because it was estimated as a fraction of stem NPP. Also, I combined the fine- and small-root NPP into a single term (fine-root NPP). Given that some authors defined fine roots to include roots ≤ 5 mm in diameter (Nadelhoffer and Raich, 1992; Finer et al. 2011), this is a reasonable term to use.    154 6.3.8 Calculation of total biomass and net primary productivity  I calculated aboveground biomass (AGB), belowground biomass (BGB) and total biomass using the following equations:   AGB = Leaf biomass + Stem biomass + Fine woody debris + Coarse woody debris (6)    BGB = Fine-root biomass + Small-root biomass + Coarse-root biomass                 (7)  Total Biomass = AGB + BGB          (8)  Aboveground productivity (ANPP), belowground productivity (BNPP) and total net primary productivity (NPP) were also calculated using the following equations:   ANPP = Canopy NPP + Stem NPP + Branch turnover NPP     (9)  BNPP = Fine-root NPP + Small-root NPP + Coarse-root NPP    (10)  Total NPP = ANPP + BNPP        (11)  The ANPP did not include losses to biogenic volatile organic compounds and herbivory. Losses due to root herbivory and leaching, as well as exudation and mycorrhizae biomass production were not accounted for in BNPP. These missing terms are critical components of the forest productivity (Clark et al. 2001a), and therefore the NPP estimates should be considered conservative. However, a previous global study showed that excluding these small NPP terms did not substantially alter NPP and allocation patterns in tropical forests (Malhi et al. 2011).    155 6.3.9 Error propagation and statistical analyses  Data on biomass and biomass change are associated with uncertainties, including systematic and random errors. An attempt was made to ensure that measurements were made without large biases. The major uncertainty is likely related to that associated with random error. All uncertainty in forest structural variables, biomass and NPP estimates are given as standard error of the mean (Saner et al. 2012; Girardin et al. 2016). Errors related to a combination of components (e.g., total NPP) were propagated using standard rules of quadrature (Aragão et al. 2009; Jiménez et al. 2014). This assumes that uncertainties were independent and normally distributed. The data were tested for normality using a Shapiro-Wilk test. A student t-test was used to test differences in forest structural variables, including diameter, height, volume as well as biomass and NPP components between the two forests. Wood density was not normally distributed, and could not be log-transformed satisfactorily, so the difference between the forests was tested with the non-parametric Mann-Whitney U test. Differences in forest structure, stem biomass and stem NPP among the functional groups or guilds were tested using one-way Analysis of Variance (ANOVA). I used post-hoc Tukey’s HSD test for multiple pairwise comparisons. Biomass and NPP estimates are reported as Mg ha-1 and Mg ha-1 yr-1. All analyses were performed with the GraphPad Prism 7 (GraphPad Software, Inc., California) software package, with a significance level of p < 0.05.   6.4 Results  6.4.1 Forest structure  Leaf area index, diameter, height and volume estimates were similar between the old-growth forest and the logged-over forest (p > 0.05, Table 6.3, Figure A.7). For instance, mean diameter was 20.17 ± 0.59 cm for the old-growth forest and 19.25 ± 0.51 cm for the logged-over forest. Stand density (≥ 10 cm DBH) was higher in the logged-over forest (783 individuals ha-1) than in the old-growth forest (503 individuals ha-1). Furthermore, basal area (≥ 10 cm DBH) was significantly greater in the logged-over forest (31.21 ± 2.38 m2 ha-1) than in the old-growth forest (23.07 ± 2.38 m2 ha-1) (p = 0.0371). The trees (≥ 10 cm DBH) in the old-growth forest had significantly higher wood density than those of the logged- 156 over forest (p = 0.0012). The mean diameter of stems (2-9.9 cm DBH) was 3.80 ± 0.16 cm for the old-growth forest and 4.27 ± 0.20 cm for the logged-over forest.   Table 6.3 Forest structural variables (mean ± SE) for old-growth forest and 54-year-old logged-over forest. Diameter, height, wood density, volume, basal area and stand density values are based on stems (≥ 10 cm DBH).  Variable Old-growth forest Logged-over forest P-value Leaf Area Index (m2 m-2)           5.87 ± 0.30           6.56 ± 0.19 0.0697 Diameter (cm)         20.17 ± 0.59         19.25 ± 0.51 0.2692 Height (m)         16.63 ± 0.31         16.22 ± 0.24 0.4267 Wood density (g cm-3)           0.57 ± 0.01           0.53 ± 0.01 0.0012 Volume (m3 ha-1)           9.50 ± 1.20         12.47 ± 1.60 0.1937 Basal area (m2 ha-1)         23.07 ± 2.18         31.21 ± 2.38 0.0371 Stand density (stems ha-1) 783 503     - DBH, diameter at breast height. Differences at p < 0.05 were significant.    6.4.2 Aboveground and belowground biomass Mean AGB was comparable between the old-growth forest (157.49 ± 20.28 Mg ha-1) and the logged-over forest (155.18 ± 15.16 Mg ha-1) (p = 0.9384, Table 6.4). The mean leaf biomass was 7.70 ± 2.43 Mg ha-1 for the old-growth forest and 4.72 ± 1.46 Mg ha-1 for the logged-over forest. Tree stem (≥ 10 cm DBH) was the highest contributor to total AGB in both forests, comprising 54% and 67% of the old-growth forest and logged-over forest, respectively. The contributions of species guilds to total stem biomass differed between the two forests (Table 6.5). In the old-growth forest, the shade-bearers (shade tolerant species) had significantly greater biomass compared to the pioneers (shade intolerant species) and the non-pioneer light-demanders. On the other hand, the NPLDs contributed more in the logged-over forest than the other guilds. The 10 tree species that contributed most biomass in the ≥ 10 cm size class accounted for 70% and 75% of the stem biomass in the old-growth forest and logged-over forest, respectively. In contrast, the 10 tree species accounting for most biomass in the < 10 cm size class,  157 contributed less than 18% of the total aboveground biomass. In both forests, the Malvaceae family contributed the largest proportion of the stem biomass (≥ 10 cm DBH), while the Papilionaceae family contributed more to the biomass of small-stemmed trees (2-9.9 cm DBH). Celtis mildbraedii, a shade tolerant species, accounted for the greatest proportion of stem biomass in the old-growth forest, while T. scleroxylon, a pioneer, recorded the highest biomass in the logged-over forest (Table A.8 and A.9).  The total contribution of lianas and palm to AGB was very low, < 1% in both forests. Coarse woody debris was 60% higher in the old-growth forest (43.61 ± 16.84 Mg ha-1) than in the logged-over forest (17.34 ± 8.39 Mg ha-1), and was the second largest contributor (28%) to AGB in the old-growth forest. Fine woody debris was 0.19 ± 0.01 Mg ha-1 for the old-growth forest and 0.25 ± 0.04 Mg ha-1 for the logged-over forest.   Belowground biomass was higher in the logged-over forest than in the old-growth forest, but the difference was not significant (p = 0.8480, Table 6.4). In both forests, CRB accounted for 85% and 94% of the total BGB in the old-growth forest and the logged-over forest, respectively. Mean fine-root biomass was significantly higher in the logged-over forest (3.04 ± 1.10 Mg ha-1) than in the old-growth forest (1.42 ± 0.29 Mg ha-1) (p < 0.0001). Overall, the total biomass (AGB plus BGB) did not differ between the old-growth forest (176.89 ± 20.38 Mg ha-1) and the logged-over forest (180.02 ± 15.37 Mg ha-1) (p = 0.9636, Table 6.5). Aboveground biomass contributed more to the total biomass, accounting for more than 86% of the total biomass in both forests. The largest contributor to the total biomass was tree stem (≥ 10 cm DBH), accounting for 48% and 58% of the biomass in the old-growth forest and the logged-over forest. CWD (24%) and small-stemmed trees (15%) were the second most important biomass sinks in the old-growth forest and logged-over forest, respectively.        158 Table 6.4 Biomass of each component (mean ± SE) of the total biomass (Mg ha-1) of the old-growth forest and the 54-year-old logged-over forest. Significant differences are denoted p < 0.05. Biomass component      Old-growth forest    Logged-over forest P-value Aboveground biomass (Mg ha-1)    Leaf        7.70 ± 2.43        4.72 ± 1.46 0.4 Stem (< 2-9.9. cm DBH)    Tree      18.59 ± 1.93      27.11 ± 4.46 0.1172 Liana         0.47 ± 1.05         1.05 ± 0.26 0.1508 Stem (≥ 10 cm DBH)    Tree      85.27 ± 9.78    103.50 ± 10.83 0.3548 Liana         0.08 ± 0.001         0.08 ± 0.02 0.131 Palm         1.58 ± 0.14         1.13 ± 0.08 0.20 Fine woody debris (< 2 cm)         0.19 ± 0.01         0.25 ± 0.04 0.3624 Coarse woody debris (≥ 2 cm)       43.61 ± 16.84       17.34 ± 8.39 0.212 AGB     157.49 ± 20.28     155.18 ± 15.16 0.9384     Belowground biomass (Mg ha-1)    Fine root        1.42 ± 0.29        3.04 ± 1.10 <0.0001 Small root         0.09 ± 0.01         0.07 ± 0.01 0.60 Coarse root       17.89 ± 2.05       21.73 ± 2.27 0.3448 BGB       19.40 ± 2.07       24.84 ± 2.52 0.8480 Total biomass     176.89 ± 20.38     180.02 ± 15.37 0.9636 AGB, Aboveground biomass; BGB, Belowground biomass; DBH, diameter at breast height         159 Table 6.5 Forest structural components and stem biomass (mean ± SE) for shade-bearers, pioneers and non-pioneer light demanders (NPLDs) in the old-growth forest and the 54-year-old logged-over forest. Estimates are based on trees (≥ 10 cm DBH). Variable Shade-bearers Pioneers NPLDs Old-growth forest     Diameter (cm) 22.41 ± 1.24a 23.13 ± 1.59a 18.98 ± 0.88a Height (m) 17.90 ± 0.64a 18.14 ± 0.77a 16.12 ± 0.47a Wood density (g cm-3)   0.64 ± 0.01a   0.50 ± 0.02b   0.55 ± 0.01b Basal area (m2 ha-1)   9.62 ± 1.34a 11.99 ± 5.06a   5.60 ± 0.84a Stem biomass (Mg ha-1) 40.26 ± 7.00a 25.98 ± 6.94ab 16.61 ± 4.50b     Logged-over forest    Diameter (cm) 20.02 ± 0.89A 19.28 ± 0.94A 19.31 ± 0.76A Height (m) 16.64 ± 0.47A 16.42 ± 0.53A 16.17 ± 0.36A Wood density (g cm-3)   0.53 ± 0.01A   0.53 ± 0.01A   0.54 ± 0.01A Basal area (m2 ha-1) 10.08 ± 1.16A   4.44 ± 0.59B 16.71 ± 2.19AC Stem biomass (Mg ha-1) 36.00 ± 5.80A 11.90 ± 1.84B 55.11 ± 9.73AC Different letters represent significant differences (p < 0.05) among the guilds or functional groups   6.4.3 Aboveground and belowground net primary productivity   Mean annual ANPP was similar between the old-growth forest (7.63 ± 1.24 Mg ha-1 yr-1) and the logged-over forest (7.40 ± 0.59 Mg ha-1 yr-1) (Table 6.6). Mean canopy productivity was the same for both forests (2.14 Mg ha-1 yr-1, Table 6.6). Leaf productivity was the largest contributor (74%) to the annual canopy productivity, followed by twigs. There was a strong seasonality in litterfall production, particularly leaf production; attaining maximum towards the end of the dry season (February-March) (Figure 6.3A and Figure A.7). A less obvious peak could also be observed during the rainfall season (June-July) in the old-growth forest. Stem productivity accounted for ca. 37% and 51% of the aboveground NPP in the old-growth forest and the logged-over forest, respectively. Mean annual stem productivity (≥ 10 cm DBH) was 2.84 ± 0.50 Mg ha-1 yr-1 and 3.76 ± 0.40 Mg ha-1 yr-1 for the old-growth forest and the logged-over forest, respectively. On average, annual stem NPP was comparable among the shade-bearers, pioneers  160 and the non-pioneer light demanders (old-growth forest, p = 0.4207, F2,71 =0.88; logged-over forest, p = 0.7904, F2,70 = 0.24, Table 6.7). However, few differences were apparent among the guilds in the old-growth forest. C. mildbraedii, T. scleroxylon and Nesogordonia papaverifera were the most productive species, accounting for more than 40% of the total stem productivity in both forests (Table A.8 and A.9). Mean annual productivity of small-stemmed individuals (2-9.9 cm DBH) was 1.03 ± 0.16 Mg ha-1 yr-1 for the old-growth forest and 0.59 ± 0.13 Mg ha-1 yr-1 for the logged-over forest. Canopy productivity contributed equally (28%) to the ANPP in both forests. The annual branch turnover productivity in the old-growth forest was ca. 2-fold greater than that of the logged-over forest. Branch turnover accounted for 21% of the total ANPP in the old-growth forest.   Mean annual BNPP was comparable between the old-growth forest (7.55 ± 1.41 Mg ha-1 yr-1) and the logged-over forest (6.49 ± 1.80 Mg ha-1 yr-1). Fine-root productivity comprised 88% and 84% of BNPP in the old-growth forest and the logged-over forest, respectively. The total annual NPP did not differ significantly between the two forests (p = 0.9497), but was relatively higher in the old-growth forest (15.18 ± 1.88 Mg ha-1 yr-1) than the logged-over forest (13.89 ± 1.89 Mg ha-1 yr-1) (Table 6.6). In the old-growth forest, both annual ANPP and BNPP contributed equally (ca. 50%) to the total NPP, while ANPP accounted for 53% of the total NPP in the logged-over forest.   6.4.4 Allocation of net primary productivity  The allocation of NPP to canopy was similar between the old-growth forest (14%) and the logged-over forest (15%) (Figure 6.3B). There was a shift in allocation between wood and fine root in the old-growth forest and logged-over forest. In the old-growth forest, 40% of NPP was allocated to wood productivity, and 46% to fine-root productivity. The fraction of NPP allocated to wood and fine-root productivity in the logged-over forest was 44% and 41%, respectively.     161 Table 6.6 Net primary productivity of each component (mean ± SE) of the total NPP (Mg ha-1 yr-1) of the old-growth forest and the 54-year-old logged-over forest. Significant differences are denoted p < 0.05. Productivity component       Old-growth forest      Logged-over forest P-value Aboveground NPP (Mg ha-1 yr-1)   Canopy    Leaf        1.59 ± 0.30        1.62 ± 0.33 0.7439 Twig (< 2 cm)         0.40 ± 0.01         0.33 ± 0.08 0.3396 Fruit         0.04 ± 0.01         0.04 ± 0.01 0.246 Flower         0.05 ± 0.01         0.04 ± 0.01 0.3357 Seed         0.03 ± 0.02         0.08 ± 0.03 0.2978 Unidentified         0.03 ± 0.02         0.03 ± 0.03 0.5869 Stem (< 10 cm DBH)    Tree        1.03 ± 0.16        0.59 ± 0.13 0.06 Liana         0.01 ± 0.01         0.04 ± 0.02 0.1429 Stem (≥ 10 cm DBH)    Tree        2.84 ± 0.50        3.76 ± 0.40 0.1564 Branch turnover (≥ 2 cm)         1.61 ± 0.58         0.87 ± 0.23 0.2669 ANPP         7.63 ± 1.24         7.40 ± 0.59 0.9594     Belowground NPP (Mg ha-1 yr-1)   Fine root         6.64 ± 1.40        5.42 ± 1.79 0.6103 Small root         0.31 ± 0.14         0.28 ± 0.15 0.8902 Coarse root         0.60 ± 0.11         0.79 ± 0.08 0.1572 BNPP         7.55 ± 1.41         6.49 ± 1.80 0.8997 TNPP       15.18 ± 1.88       13.89 ± 1.89 0.9497 ANPP, Aboveground net primary productivity; BNPP, Belowground net primary productivity; DBH, diameter at breast height; Total NPP, Total net primary productivity.          162 Table 6.7 Stem productivity (mean ± SE) for shade-bearers, pioneers and non-pioneer light demanders (NPLDs) in the old-growth forest and the 54-year-old logged-over forest. Estimates are based on trees (≥ 10 cm DBH). Forest type   Shade-bearers   Pioneers   NPLDs Old-growth forest    1.16 ± 0.30a   0.69 ± 0.28a   0.66 ± 0.31a Logged-over forest    1.38 ± 0.24A   1.34 ± 0.38A   1.13 ± 0.21A DBH, diameter at breast height. Different letters represent significant differences (p < 0.05) among the guilds or functional groups.      Figure 6.3 Monthly distribution of litterfall (A), and annual NPP allocation into canopy, wood and fine root components (B) in the old-growth forest and the 54-year-old logged-over forest. S O N D J F M A M J J A02468Litterfall (Mg ha-1 month-1)2013 2014Logged-over forestOld-growth forestAOld-growth forest Logged-over forest020406080100NPP allocation (%)Fine rootWoodCanopyB 163 6.5 Discussion  6.5.1 Aboveground and belowground biomass  The mean total biomass of the old-growth forest and the logged-over forest are within the range of values reported for tropical moist forests in Africa (Djomo et al. 2010; Gautam and Pietsch, 2012; Ekoungoulou et al. 2015) and elsewhere (Sierra et al. 2007; Saner et al. 2012; Ngo et al. 2013; Berenguer et al. 2014; Gautam and Mandal, 2016). The mean AGB estimates (155.18-157.49 Mg ha-1) are comparable with values reported for tropical forests in central Africa (163-191 Mg ha-1, Doetterl et al. 2015) but lower than the mean values reported for lowland tropical forests in Asia (205.0-334.98 Mg ha-1, Ngo et al. 2013; Kenzo et al. 2015). The high AGB compared to BGB in the two forests corroborates the findings of many studies in the major forest biomes (e.g., Helmisaari et al. 2002; Djomo et al. 2010; Cai et al. 2016b). The CRB estimates (17.89-21.73 Mg ha-1) fall within the range of values reported for other lowland mature and degraded tropical forests (11.68-164.50 Mg ha-1, Hertel et al. 2009b; Ngo et al. 2013; Anderson-Teixeira et al. 2016; Gautam and Mandal, 2016). The large variation in CRB reported in the literature could be partly attributed to methodological differences, sampling depth and the root diameter threshold used. Currently, there is no consensus on the preferred method to sample coarse roots for biomass estimation (Addo-Danso et al. 2016), and the coarse root category is defined based on arbitrary diameter thresholds (e.g., > 2 mm or > 5 mm to 50 mm) by different authors (e.g., Moser et al. 2010; Anderson-Teixeira et al. 2016). Coarse roots are critical to the forest C budget, therefore it is important to standardize the methods used to quantify these roots to accurately estimate C pools and fluxes.   The comparable total biomass estimates of the old-growth forest and the logged-over forest is partly attributable to the (i) forest structure, and (ii) edaphic conditions. Forest structural attributes, including wood density, tree diameter and height are closely related to both above- and belowground biomass in tropical forests (Carlson et al. 2016; Miyamoto et al. 2016). Wood density was significantly higher in the old-growth forest, but diameter and height did not differ between the two forests (Table 6.3). It appears the differences in wood density alone did not affect the biomass values (Table 6.5), and this could be due  164 to the fact that wood density does not always correlate with biomass in all forest ecosystems (e.g., Slik et al. 2010). Indeed, I did not find any relationship between community-weighted mean of wood density and AGB in the studied forests (Figure A.6, Chapter 5). It is likely that the diameter and height distribution had more influence on the biomass values. Logging typically affect the forest structure, by removing large trees and reducing tree heights (Osazuwa-Peters et al. 2015a; Rutishauser et al. 2016). However, it appears that after 54 years the vertical structure of the logged-over forest has recovered, as reflected in the biomass estimates. Gatti et al. (2015) reported that the distribution of tree size and height of selectively logged forests in Ghana, Sierra Leone, Cameroon and Gabon recovered to old-growth forest levels after 30-50 years. Indeed, Alamgir et al. (2016) reported similar AGB between intact forests and those degraded by logging in Australia, which they attributed to the advanced stage of regeneration shown in the similar tree size distribution in the forest types. Previous studies at the global (Andrade et al. 2017), regional (Rutishauser et al. 2015) and local (Asase et al. 2012) scales show that tropical forests that are subject to low intensity selective logging (targeting 3-7 trees ha-1) are likely to recover their structure and biomass after a few decades. Nevertheless, it is possible that the other structural variables, including stand density and basal area, which were significantly higher in the logged-over forest could have accounted for the greater root biomass and tree stem (≥ 2 cm DBH) biomass in that forest. Stand density and basal area are strong predictors of root and stem biomass of forest ecosystems in all biomes (Finér et al. 2011b; Paquette and Messier, 2011; Durán et al. 2015). Edaphic conditions also affect the distribution of biomass, including tree biomass (Kenzo et al. 2015), root biomass (Powers and Peréz-Aviles, 2013), and coarse woody necromass (Martins et al. 2014). With the exception of base saturation, all the edaphic conditions, including elevation, as well as the soil physical and chemical properties was comparable between the old-growth forest and the logged-over forest (Table 3.1). The lack of clear differences in the edaphic conditions between the two forests could have resulted in the similar total biomass estimates.   The comparable total biomass of the two forests is inconsistent with findings from other tropical studies elsewhere (Sierra et al. 2007; Saner et al. 2012; Berenguer et al. 2014; Gautam and Mandal, 2016).  165 Previous studies have usually reported higher total biomass in old-growth forests than in logged forests. In the eastern Amazon, Berenguer et al. (2014) reported significantly higher total biomass estimates for old-growth forests than logged forests. Aboveground biomass and BGB were 1.5-fold greater in an unlogged forest than in a 20-year-old logged forest in a dipterocarp forest in Malaysian Borneo (Saner et al. 2012). However, Asase et al. (2012) reported significantly higher AGB in 29-year-old logged forest than in unlogged forest in southwestern Ghana. There are plausible explanations for the contradictory findings from the aforementioned studies: (i) logging intensity varies; (ii) the time elapse between logging events and the studies differ; (iii) forest composition and structure varies; (iv) edaphic conditions differ; and  (v) a combination of these factors.   6.5.2 Aboveground and belowground net primary productivity   The mean annual total NPP estimates for the two forests (13.89-15.18 Mg ha-1 yr-1) fall within the range of values reported for global tropical forests (Clark et al. 2001b; Anderson-Teixeira et al. 2016; Vogt et al. 2016). Vogt et al. (2016) reported annual total NPP between 4 and 43.30 Mg ha-1 yr-1 for 96 tropical forest sites. The annual total NPP for mature tropical forests ranged from 8-36.17 Mg ha-1 yr-1 across 845 plots in 178 sites in Africa, Asia, Oceania and the Americas (Anderson-Teixeira et al. 2016). The annual total NPP values are lower than the mean value (22 Mg ha-1 yr-1) estimated by John (1973) for a 40-year-old secondary moist forest in southern, Ghana. The comparable annual total NPP of the old-growth forest and the logged-over forests is in agreement with the findings from some studies (Noormets et al. 2015; Nyirambangutse et al. 2017). Equally, other studies have reported significantly higher total NPP in unlogged forests than logged forests (Saner et al. 2012; Cai et al. 2016b; Gautam and Mandal, 2016). The annual total NPP was comparable for the two forests because the subcomponents offset each other. For instance, canopy productivity was the same, but stem productivity (≥ 10 cm DBH) was higher in the logged-over forest, while the old-growth forest had higher branch turnover and fine-root productivity (Table 6.6).    166 I used litterfall production as a proxy for canopy productivity (Saner et al. 2012), and the results obtained for the two forests are in line with findings from other studies (Burghouts et al. 1992; Chave et al. 2010; Saner et al. 2012; Paudel et al. 2015; Anderson-Teixeira et al. 2016). In a regional study, which analyzed litterfall patterns in 186 sites in tropical South America, Chave et al. (2010) reported comparable annual litterfall production between old-growth forests and secondary forests. Litterfall production was similar between unlogged forests and logged forests in Borneo, Malaysia (Saner et al. 2012). The mean annual litterfall production (2.14 Mg ha-1 yr-1) is lower than the 9.90-10.54 Mg ha-1 yr-1 reported for 40-year-old secondary forests in Ghana (John, 1973). The mean litterfall production values fall within the range of values (2-11 Mg ha-1 yr-1) reported by Zhang et al. (2014) for global forests based on 400 litterfall datasets. The litterfall production estimates of the two forests are underestimated because components like litter trapped in the canopy and large leaves and palms were not included. Litterfall peaked during the dry season (Figure 6.3A and Figure A.8); a similar seasonal pattern has been observed in Ghanaian forests (e.g., John, 1973) and other tropical forests elsewhere (e.g., Chave et al. 2010; Zhang et al. 2014; Girardin et al. 2016). The high litterfall during the dry season is associated with reduced rainfall and high solar radiation (Wright and Schaik, 1994; Chave et al. 2010; Zhang et al. 2014). Water shortage during the dry season may lead to high leaf abscission to reduce evapotranspiration (Ballestrini et al. 2011). High solar radiation can also lead to increased leaf fall (Ballestrini et al. 2011), which in turn may trigger the flushing of new leaves and flower production (Wright and Schaik, 1994).  The stem productivity estimates are higher than the mean value (1.79 Mg ha-1 yr-1) reported by Fauset et al. (2012) for 19 old-growth forest plots across a rainfall gradient in Ghana. The stem productivity estimates are lower than the global mean for tropical forests (Vicca et al. 2012; Anderson-Teixeira et al. 2016), and for single sites in Africa (Nyirambangutse et al. 2017) and South America (e.g., Chave et al. 2008; Aragão et al. 2009; Jiménez et al. 2014; Girardin et al. 2016; Quinto-Mosquera and Moreno, 2017). The variability in stem NPP could be attributed to the local differences in both biotic and abiotic factors, including edaphic conditions and species composition (Cleveland et al. 2011; Jucker et al. 2016; Quinto- 167 Mosquera and Moreno, 2017). Nevertheless, one cannot rule out the influence of allometric equations used to estimate biomass changes in such studies. Allometric equations directly affect stem productivity values (Chave et al. 2008; Anderson-Teixeira et al. 2016).  In the present study, I used a locally derived allometric equation to estimate tree biomass (Henry et al. 2010). However, most of the other studies (e.g., Girardin et al. 2016; Quinto-Mosquera and Moreno, 2017 and references therein) estimated tree biomass with the widely used Chave equation for moist forests (Chave et al. 2005), which is known to overestimate tree biomass in African forests (Ngomanda et al. 2014). In Chapter 5, I found that the mean stem productivity estimate obtained from the Chave equation was 61% higher than that from the Henry equation. The lower stem productivity estimate in the old-growth forest could be associated with age-related decline in stem productivity in forest ecosystems, which is well established in the literature (Ryan et al. 1997; Tang et al. 2014; Anderson-Teixeira et al. 2016). In a recent pantropical study, Anderson-Teixeira et al. (2016) reported that stem productivity declined with stand age in diverse tropical forests. A number of mechanisms, including high mortality, reduced leaf area, change in C allocation, and hydraulic limitation has been proposed to explain the decline in stem productivity as forest stands age (Ryan et al. 1997; Drake et al. 2010; Tang et al. 2014).   Compared with other studies the branch turnover (or CWD input) of 0.87-1.61 Mg ha-1 yr-1 was low (Clark et al. 2002; Palace et al. 2008; 2012; Doughty et al. 2014; Silver et al. 2016). Values exceeding 9 Mg ha-1 yr-1 have been reported for similar tropical lowland moist forests (Palace et al. 2012; Silver et al. 2016). Few studies have compared woody debris input, including branch turnover in unlogged forests and logged forests. However, the high branch turnover in the old-growth forest is consistent with the findings from some studies (Pauletto, 2006; Gautam and Mandal, 2016). Branch turnover input was higher in an unlogged forest than in a selectively logged forest in Sunsari district, eastern Nepal (Gautam and Mandal, 2016). Pauletto (2006) also reported higher woody debris input in an unlogged forest than in 11-12 year-old logged forest in Mato Grosso, Brazil. In contrast, Palace et al. (2008) reported higher woody debris input in a logged forest than in unlogged forest in the Tapajos National Forest, Para state, Brazil. Studies  168 in temperate forests have also given mixed results on woody debris input in old-growth forests and logged or secondary forests (e.g., Cai et al. 2016b; Holdaway et al. 2016). The higher branch turnover productivity of the old-growth forest is due to high mortality, which generally increases with stand age (Harmon, 2009). Mortality is the main process that creates woody debris input in forest ecosystems (Harmon, 2009). Using a simple equation from Palace et al. (2008) I calculated the mortality rate (%) by dividing branch turnover productivity by aboveground biomass (Table 6.4), and had a higher mortality rate for the old-growth forest (1.0% yr-1) compared to the logged-over forest (0.6% yr-1).   The substantial contribution of fine-root productivity to the total NPP is already recognized in the literature (e.g., Clark et al. 2001b; Jiménez et al. 2014; Noormets et al. 2015). The mean fine-root productivity estimates are higher than global averages reported by Anderson-Teixeira et al. (2016) for mature (>100 year) tropical forests (5.50 Mg ha-1 yr-1) and by Noormets et al. (2015) for managed and unmanaged forests (3.40-3.46 Mg ha-1 yr-1). The total BNPP of the two forests (Table 6.6) is higher than the value reported for a lowland undisturbed forest in Colombia (Jiménez et al. 2014), and for montane tropical forests in Rwanda (Nyirambangutse et al. 2017).   6.5.3 Shifts in net primary productivity allocation to wood and fine roots  The NPP allocation results agree with both empirical (Litton et al. 2007; Malhi et al. 2011; Jiménez et al. 2014; Girardin et al. 2016) and modeling (Dybzinski et al. 2011; Wolf et al. 2011) studies that suggest that shifts in wood and fine-root NPP dominate the allocation pattern in fore