We have limited understanding of how tropical canopy foliage varies along environmental gradients, and how this may in turn affect forest processes and functions. Here, we analyse the relationships between canopy leaf area index (LAI) and above ground herbaceous biomass (AGBH) along environmental gradients in a moist forest and miombo woodland in Tanzania. We recorded canopy structure and herbaceous biomass in 100 permanent vegetation plots (20 m × 40 m), stratified by elevation. We quantified tree species richness, evenness, Shannon diversity and predominant height as measures of structural variability, and disturbance (tree stumps), soil nutrients and elevation as indicators of environmental variability. Moist forest and miombo woodland differed substantially with respect to nearly all variables tested. Both structural and environmental variables were found to affect LAI and AGBH, the latter being additionally dependent on LAI in moist forest but not in miombo, where other factors are limiting. Combining structural and environmental predictors yielded the most powerful models. In moist forest, they explained 76% and 25% of deviance in LAI and AGBH, respectively. In miombo woodland, they explained 82% and 45% of deviance in LAI and AGBH. In moist forest, LAI increased non-linearly with predominant height and linearly with tree richness, and decreased with soil nitrogen except under high disturbance. Miombo woodland LAI increased linearly with stem density, soil phosphorous and nitrogen, and decreased linearly with tree species evenness. AGBH in moist forest decreased with LAI at lower elevations whilst increasing slightly at higher elevations. AGBH in miombo woodland increased linearly with soil nitrogen and soil pH. Overall, moist forest plots had denser canopies and lower AGBH compared with miombo plots. Further field studies are encouraged, to disentangle the direct influence of LAI on AGBH from complex interrelationships between stand structure, environmental gradients and disturbance in African forests and woodlands.
Citation: Shirima DD, Pfeifer M, Platts PJ, Totland Ø, Moe SR (2015) Interactions between Canopy Structure and Herbaceous Biomass along Environmental Gradients in Moist Forest and Dry Miombo Woodland of Tanzania. PLoS ONE 10(11): e0142784. https://doi.org/10.1371/journal.pone.0142784
Editor: Eric Gordon Lamb, University of Saskatchewan, CANADA
Received: February 4, 2015; Accepted: October 27, 2015; Published: November 11, 2015
Copyright: © 2015 Shirima et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: All data on Leaf Area Index are available at (http://dx.doi.org/10.6084/m9.figshare.1497894 or http://www.york.ac.uk/environment/research/kite/resources/#tab-2),the rest of the data are from Climate Change Impact, Adaptation and Mitigation programme in Tanzania (CCIAM), available upon request from project PI (firstname.lastname@example.org) and requires co-authorship with core project researchers or proper acknowledgement when used.
Funding: The Norwegian State Education Loan Fund (http://www.lanekassen.no/) and Norwegian Government through the Climate Change Impacts, Adaptation and Mitigation programme at Sokoine University of Agriculture in Tanzania (http://www.drpgs.suanet.ac.tz/) funded DDS. MP is funded by European Research Council under the Project number 281986 and PJP was funded by the Ministry for Foreign Affairs of Finland through the CHIESA project (http://chiesa.icipe.org/). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Tree species vary in their capacity to use abiotic resources, promoting coexistence among life forms at different growth stages [1,2]. Morphological differences among tree crowns, for example, enhances a forest community’s capacity to exploit light resources and fix carbon, regulating stand-scale biomass production . Canopy leaf area is the main regulator of radiation absorption and can block up to 95% of visible light from reaching the forest floor . Light extinction through dense, multi-layered vegetation creates a strong vertical energy gradient, shaping microclimate and soil properties within the forest . In turn, microclimate, light availability, soil moisture and soil fertility interact to regulate plant growth in sub-canopy layers .
Few studies have explored interrelationships between tree diversity, canopy structure and biomass in the Afro-tropics [7,8], and no study has quantified the relationship between canopy structure and herbaceous biomass, or how this varies along gradients of soil nutrients and anthropogenic disturbance. Forests and woodlands in Eastern Africa are under considerable pressure from increasing human populations [9–11]. Land use and climate change interact to modify natural variability in canopy leaf area, which declines with degradation pressure and increases with water availability . The loss of canopy trees due to selective logging and high intensity fires results in forests with simpler vertical structure and reduced functional capacity . This may affect ecosystem productivity, including woody and herbaceous aboveground biomass.
The majority of studies looking at variations in carbon stocks and biodiversity in Africa focus on moist forests [14–17], while miombo woodlands remain understudied . Moist forests tend to have higher biodiversity and carbon value per unit area, but they cover much smaller areas compared with miombo in both Southern and Eastern Africa (0.064 vs. 2.7 million km2 [9,19]). In Tanzania, miombo woodlands comprise around 90% of the total forested area , providing essential resources to rural communities, particularly wood-based energy (firewood and charcoal) and other non-timber products [21,22]. Miombo tree assemblages are predominantly deciduous with open canopies, on soils that have low nutrient content, are well drained, highly leached, acidic and low in organic matter . Moist forests, on the other hand, are predominantly evergreen, with denser canopies on more nutrient-rich soils . Understanding drivers of structure and function in both these vegetation types, especially their role in the carbon cycle, is a priority for research .
Here, we analyse the relationships between canopy leaf area index (LAI) and above ground herbaceous biomass (AGBH) in a moist forest and miombo woodland in Tanzania. Our objectives are to test the hypotheses that forest structural attributes (tree richness and size distribution) can be used to predict canopy leaf area (LAI), and that AGBH is negatively related to LAI due to light extinction through the canopy. We explore the extent to which our findings vary along environmental and disturbance gradients, and between moist forest and miombo woodland systems.
Materials and Methods
We conducted our study in Hanang and Dirma forest reserves in Tanzania (Fig 1). Permission to conduct the fieldwork was granted by the Manyara region and Hanang district administrative secretaries. Hanang forest reserve (forest extent: 58.71 km2) is a central government catchment reserve, spanning an elevation range of 1860–3418 m (Latitude: -4.44°, Longitude: 35.40°). The reserve receives a mean annual rainfall of 895 mm depending on elevations, ranging from 878 mm at lower elevations to over 1000 mm at higher elevations. Mean annual temperatures range from 17°C at the lowest elevations to 13°C at the highest . Grasses and thickets dominate the highest elevations and steepest slopes. Moist forest with canopy species such as Albizia gummifera and Cassipourea malosana dominate at mid to high elevations, and are interspaced irregularly by moorland patches. The forest grows on volcanic soils ranging from sandy to humus rich loams in the upland moorlands and upper montane areas .
Dirma village forest reserve (miombo extent: 63.5 km2, Fig 1) spans an elevation range of 1500–1700 m (Latitude: -4.70°, Longitude: 35.44°). The reserve receives a mean annual rainfall of 796 mm (range: 787–804 mm) and mean annual temperature of 19.5°C (range: 19–20°C) . Miombo woodland assemblages characterized by Brachystegia spiciformis and Julbernadia globiflora dominate vegetation cover in the reserve. The woodlands grow on poor soils that are low in nutrients and vegetation cover is highly influenced by frequent fires and anthropogenic disturbances, similar to miombo woodlands elsewhere in Africa [9,28].
Vegetation and soil surveys
We surveyed 100 vegetation plots of 20 m × 40 m (60 plots in moist forest and 40 plots in miombo woodland) in March 2012. The plots were placed systematically along an elevation gradient (from 1583 m and 1900 m minimum elevations in miombo woodland and moist forest respectively), and separated by a minimum distance of 400 m. In each plot, we recorded all tree individuals with diameter at breast height (dbh) ≥ 5 cm, and identified each of these trees to species-level. Where species identification by a botanist was not possible in the field, voucher specimens were collected and identified at Arusha National Herbarium in Tanzania. We recorded geographical location and elevation using a handheld GPS (Map76cx). We measured tree height in the field whenever conditions allowed, using a Suunto-hypsometer. The remaining tree heights were estimated from our field data using biome-specific height-dbh regression equations. We recorded the number of tree stumps in each plot as an indicator of anthropogenic disturbance. Selective logging is widespread in forest and woodland reserves in sub-Saharan Africa, due to acute poverty, increased human populations and weak forest governance . Apart from logging, forest and woodlands in Africa also experience other disturbances such as herbivory and frequent fires.
To estimate AGBH, litter biomass and tree seedling density, each plot was sub-divided into eight (10 m × 10 m) subplots. Aboveground herbaceous plant materials (clipped at ground level) and litter materials were collected from five (1 m × 1 m) quadrats, and tree seedlings (dbh < 5cm) were counted in 2 m × 2 m quadrats, placed at random within four alternate subplots (S1 Fig). We recorded the total fresh weight of herbaceous and litter samples in the field. We collected a subset of each of these samples, which were then oven dried in the laboratory to a constant weight at 70°C for 48 h to obtain dry mass estimates. Results were applied to the total fresh weights to obtain the total herbaceous and litter dry mass per plot, referred to as aboveground herbaceous biomass (AGBH) and litter biomass.
Soil samples were collected at three depths (0–15 cm, 15–30 cm and 30–60 cm) at five points per plot, i.e. from each of the four corners and from the centre of the main plot. We aggregated samples for each depth into composites for subsequent analyses. The resulting 300 soil samples were air-dried and sieved through a 2 mm wire mesh and analysed for soil pH (at 1:2.5 soil: H2O), organic carbon (Walkley-Black method), available phosphorous (Bray II), total nitrogen (Kjeldahl method), potassium and sodium (ammonium acetate 1.0 M pH7.0 extraction) at Seliani Agricultural Research Institute, Arusha, Tanzania.
LAI data acquisition and processing
All photographs were taken during the wet season in March 2012. Leaf area index was estimated following standard protocols [12,31]. We took 13 hemispherical photographs in each of four subplots (10 m × 10 m) using a Nikon D3100 camera equipped with a hemispherical fish-eye lens (S1 Fig). The camera was mounted on a tripod at 1 m above ground, looking vertically upward from beneath the canopy. The levelled hemispherical photographs were acquired normal to a local horizontal datum, orienting the optical axis of the lens to local zenith. We measured under overcast conditions whenever possible to minimize anisotropy of the sky radiance .
CAN-EYE analysis software estimates LAI in digital images based on gap fraction for specific viewing directions. CAN-EYE estimates LAI as plant area index, as is the case with other indirect measurements. Thus, our LAI estimates include materials such as stems, trunks, branches, twigs and plant reproductive parts . However, it is not possible to know if leaves are present behind the stems, branches or trunks. Therefore, masking some parts of the plants to keep only the visible leaves is not correct and could lead to large underestimation of the actual LAI value, depending on the way leaves are grouped in other parts of the plant. Furthermore, during the growing season in both deciduous woodland and evergreen forest, the total vegetation surface is mainly composed of leaf area, and by a lesser part of twigs, branches and stem surface . Also according to  branches and boles contributed to total LAI by less than 5% in three relatively dense stands of conifers.
Hemispherical images were pre-processed by first extracting blue-channel pixel brightness values and then applying a threshold algorithm for separating sky from vegetation . Resultant binary images were analyzed using the free canopy analysis software CAN-EYE V6.3.8 [31,37]. For each site, we derived LAI corrected for foliage element clumping , limiting the field of view of the lens to values between 0° and 60° to avoid mixed pixels. Values of LAI from the four subplots were averaged per plot for subsequent analyses.
Forest and woodland structural attributes
Tree richness was estimated as the total number of tree species per plot. Stem and seedling density were estimated as the numbers of tree stems and seedlings per ha. We used Pielous’s index (J) to estimate tree species evenness  and the Shannon diversity index (H′) to estimate diversity . We estimated the quadratic mean diameter (QMD) for all trees as , where is the arithmetic mean diameter and S2 is the variance of tree diameters in a plot. QMD has a strong correlation to stand volume and basal area, and is a preferred measure of stand structure over the arithmetic mean diameter . We estimated predominant height (PDH) of the forest and woodland stands as the average height of the 100 tallest trees per hectare . Quantified variables were categorised into stand structural and environmental variables for subsequent modelling (Table 1).
Modelling vegetation structure and links to environmental drivers
We used generalized linear models (GLM) with Gaussian distribution error and identity link function [43,44] to explore the relationships between stand structural and environmental variables versus LAI. We also explored relationships between stand structural variables (including LAI), and environmental gradients versus AGBH. We developed these models for moist forest and miombo woodland, separately.
In a first step, we fitted two subsets of models focussing on: structural variables as predictors of either LAI or AGBH; and environmental variables as predictors of either LAI or AGBH. Each of these models included disturbance as additional predictor and first term interactions between all predictors. We then combined structural and environmental predictors, disturbance and first term interactions between predictors, into one single model for each response in each vegetation type.
Exploratory analysis using smoother functions  indicated nonlinear relationships between LAI, tree richness and predominant height. We therefore fitted relationships including quadratic terms for these predictors (see also S1 Table for details on final global models). We used Pearson correlation (r) and variance inflation factor (VIF) to assess collinearity among structural and environmental predictor variables . In cases of high collinearity between two predictor variables (|r| > 0.5 and VIF > 3.0), we retained the predictor showing a stronger univariate relationship with the response variable .
We used stepwise model selection based on the Akaike Information Criterion to identify optimal models from the global models [43,48]. The relative contributions of predictor variables were determined by the percentage reduction in explained deviance (D2) . We used likelihood ratio tests to compare subset models with the global model . We validated residual spread and estimated the predictive error using leave-one-out cross-validation , implemented using the“cv.glm” function in R , in conjunction with the mean squared error of prediction . Moreover, we used paired Mann-Whitney-Wilcox tests between observed and predicted LAI or AGBH to assess the significance of mean squared error of prediction as a measure of model bias .
Structural and environmental attributes
We identified 97 tree species from 46 families in moist forest, and 62 species from 29 families in miombo woodland. These two vegetation types differed significantly in their Shannon diversity and tree richness (Table 1). They also differed significantly in LAI, AGBH, soil pH, soil phosphorus, soil nitrogen, soil potassium and soil sodium (Table 1).
Structural and environmental influences on LAI
Structure and environmental variability (combined) explained 76% of the deviance in moist forest LAI and 82% in miombo woodland LAI, outperforming all other models (Table 2). The parameter estimates and the likelihood ratio tests for the combined models were significantly different from zero (P≤0.05), suggesting that these models explained variation in LAI better than the global (unreduced) models for both vegetation types (Table 2 and S2 Table; Forest: AIC = 58.6, LRT3.6, P = 0.001, miombo: AIC = 16.8, LRT17, P = 0.014). Further, the AIC weights show that the combined models consistently outperform structure-only or environment-only models for LAI, in both moist forest and miombo woodland (Table 2).
LAI exhibited a substantial non-linear increase with predominant height and a strong linear increase with tree species richness in moist forest (S2 Table, Fig 2). Additionally, the LAI of moist forest increased weakly with disturbance, but decreased with soil nitrogen (see S2 Table, S2 Fig). Under high disturbance, LAI of moist forest increased with soil nitrogen and decreased with soil pH (S2 Table, Fig 2).
(A) LAI shows a non-linear relationship with predominant height, (B) linear relationship with tree species richness, (C) soil nitrogen, (D) soil pH under high disturbance levels, (E) stem density and (F) tree species evenness when all other variables are set to their mean values (S2 Table, combined models). Solid lines plot fitted partial regressions from generalized linear models, with standard errors of the mean in dotted lines.
In miombo woodland, LAI decreased linearly with tree species evenness and increased strongly with stem density (S2 Table, Fig 2). The LAI of miombo woodland increased with soil phosphorous and nitrogen (S2 Table). Moreover, tree species richness interacted with soil phosphorous and pH in that LAI decreased with richness in soils rich in phosphorous and at high pH while LAI increased with richness where soil phosphorous concentrations and pH were low (S2 Table, Fig 3). Plant species richness was negatively related to LAI when soil potassium was low but at high soil potassium, the relationship between richness and LAI was positive (S2 Table, Fig 3). There was an overall positive relationship between predominant height and LAI, and the relationship was stronger at relatively high soil potassium (S2 Table, Fig 3).
(A) LAI shows linear relationships with tree species richness at high soil phosphorous, (B) with tree species richness at high soil potassium, (C) with tree species richness at high soil pH, and (D) with predominant height at low and high soil potassium, when all other variables are set to their mean values (S2 Table, combined model). The solid lines are the fitted partial regression lines from generalized linear models of the relationships between LAI and labeled variables (Low and High levels of P = phosphorous, K = potassium and pH, respectively), with standard errors of the mean in dotted lines.
Structural and environmental influences on AGBH
Structural and environmental variability (combined) explained around 25% and 45% of deviance in AGBH of moist forest and miombo woodland, respectively. The parameter estimates and the likelihood ratio tests for these two models were significantly different from zero (P ≤ 0.05), suggesting that they explained deviance in AGBH better than the global model in miombo woodland (Table 3 and S3 Table; miombo: AIC = 46.2, LRT69.38, P = 0.001), while in moist forest only the parameter estimates were significant (P ≤ 0.05). The AIC weights shows that the combined models are the most optimal models for predicting AGBH in miombo woodland and moist forest (Table 3).
AGBH in moist forest decreased linearly with tree species richness, but followed an inverted hump-shape curve in response to LAI (S3 Table, Fig 4). AGBH in moist forest decreased with LAI at lower elevations whilst increasing with LAI at higher elevations (S3 Table, Fig 4). AGBH in miombo woodland increased linearly with soil nitrogen and soil pH (S2 Fig), but decreased linearly with LAI at high levels of disturbance (S3 Table, Fig 4).
(A) AGBH show a non-linear relationship with leaf area index, (B) linear relationship with tree species richness, (C) a non-linear relationship with LAI at high elevation in moist forest, and (D) linear relationship with LAI in miombo woodland at high levels of disturbance, when all other variables are set to their mean values (S3 Table, combined models). The solid lines are the fitted partial regression lines from generalized linear models of the relationships between AGBH and labeled variables (Low-Elv, High-Elv, and Low-Stump and High-Stump are low and high levels of elevation and disturbance gradients, respectively), with standard errors of the mean in dotted lines.
Structural and environmental influences on LAI
Foliage density and tree sizes affect the ability of woody vegetation types to intercept light and atmospheric nutrients . Thus, differences in canopy leaf area are likely to feed through to plant growth and biomass production and the vegetation type’s carbon sequestration potential. We found substantial differences in structural and environmental variables between the moist forest and miombo woodland, which in turn correlated with differences in LAI and herbaceous biomass.
Structural and environmental variables (combined) explained over 70% of the deviance in LAI and performed better (AIC weights of over 90%) than either structural or environmental variables alone, in both vegetation types, which suggests that forests and woodlands respond to the environment based on morphological and physiological adaptations . We have shown that LAI increases with tree richness in moist forest and additionally with tree stem density in miombo woodland, similar to findings in tropical lowland and montane forests in Ecuador . Our data indicated that tree richness and stem density had low correlation (about 50%), suggesting that the observed patterns of LAI vs richness and LAI vs stem density in miombo woodland were not artifacts of one-another. This finding lends support to the hypothesis of resource use complementary and higher productivity in more diverse forests . Whereas in less diverse and less productive miombo woodland, LAI decreased with tree species evenness, suggesting that a few dominant species with relatively low multi-layer canopy which characterize the canopy leaf area.
We show that while LAI increases with tree height, this relationship is not linear and LAI saturates or decreases for forest stands featuring larger trees, a pattern we find for moist forest only. Similar findings have been reported for Acadia forest in the US , where the gradual decline in LAI with tree height was attributed to severe branch abrasion and loss of new foliage as trees grow taller . Yet, our finding contradicts a global meta-analysis, which reports positive, non-saturating relationships between remotely-sensed LAI and field-measured tree height across broadleaf forests and savanna . The difference in findings may partly arise from the different spatial scales and resolutions used in Yuan et al.’s  study, and uncertainty introduced by different instruments used to estimate LAI at different sites in this meta-analysis. However, it should also be noted that LAI estimated from hemispherical images saturates in high-biomass biomes due to methodological constraints such as inability to differentiate leaves from other parts of the plant such as trunks, branches or twigs .
We show that anthropogenic disturbance and/or its interaction with soil nutrients influenced variability in LAI of moist forest. In miombo woodland, anthropogenic disturbance had less influence on LAI compared to soil nutrients or interactions between soil nutrients and structural variables. Miombo woodlands are relatively highly disturbed (selective logging, frequent fires grazing and shifting cultivation), have more open canopies and are more limited by soil nutrients , compared to moist forest. Therefore, disturbance inferred from stump frequency will likely correspond to a more pronounced effect on LAI in moist forest than in miombo woodland. Further studies are encouraged to disentangle anthropogenic effects from those of soil and other structural attributes on the moist forest and miombo woodland canopy characteristics.
Structural and environmental influences on AGBH
As the herbaceous layer is likely to affect soil, surface fluxes, and ground-dwelling organisms, understanding how canopy structure interacts with this layer can aid in understanding the responses of sub-canopy biodiversity and ecosystem processes to climate and land use [31,53]. Previous studies suggested that trees can benefit herbaceous vegetation through amelioration of harsh environmental conditions and increased nutrient availability [6,56]. However, trees have also been reported to suppress herbaceous biomass by altering light availability and soil fertility on the forest floors .
In miombo woodland, structural and environmental variability explained 45% of the deviance in AGBH, and performed better (AIC weight of over 90%) than either structural or environmental gradients alone. In moist forest, structural and environmental variability performed better than structural variables (AIC weights, 57% vs 43%) but by smaller margin than in miombo woodland. AGBH in the moist forest decreased with increasing LAI, which increased with tree species richness. This suggests that tree richness may suppress AGBH by increasing canopy density thereby reducing light availability at the forest floor limiting herbaceous plant growth. Patterns are complex, though, as indicated by the U-shaped association between AGBH and LAI. The slight increase in AGBH with LAI at high elevations suggests a reduced impact of light limitation due to declines in tree growth at higher altitudes . This is indicative of a positive effect of environmental stress gradients (adverse climate, shallow soils and low radiant energy) on plant growth at high elevations, which promotes plant coexistence  due to low competition intensity [60–62].
In miombo woodland, AGBH was higher under denser vegetation canopies, suggesting an amelioration of harsher environments characterized by either drought or frequent fires. Miombo features open canopies, and sub-canopy plant growth is likely to be soil nutrient and water-limited rather than restricted by light availability. Unsurprisingly, disturbance, which further prevents canopy closure, interacts with soil nutrients and LAI to regulate biomass in the herbaceous layer. Trees’ multiple effects on the herbaceous layer partly depend on interactions between tree canopy (i.e. shedding), soil fertility and moisture availability . Soil nutrients from decaying tree stumps or ring-backed roots and deposits from biomass burning have been associated with increased herbaceous biomass in miombo woodland [23,64]. Meanwhile, human activities may affect AGBH depending on disturbance type (i.e. grazing and fire). For example, herbaceous biomass decreased linearly with LAI at high disturbance levels and increased linearly with soil nitrogen and pH, indicating high local variation in herbaceous plant growth in response to stand structure characteristics and soil nutrients in miombo woodlands.
Tree diversity, sizes and environmental variability affect canopy leaf area and herbaceous biomass in both moist forest and miombo woodland. Our findings provide support for the hypothesis of niche complementarity, with higher tree diversity enabling a better use of canopy space optimizing light capture through forest canopies. Since canopy leaf area is a good indicator of a vegetation’s photosynthetic capacity, it is likely that high species richness in moist forests facilitates higher productivity and biomass production. Denser moist forest canopies, on the other hand, should have negative impact on herbaceous biomass than less dense canopy miombo woodlands . Our data provide support for this hypothesis, but suggest complex interrelationships between environmental and structural parameters interacting to drive variability in herbaceous biomass. Anthropogenic disturbance modifies canopy structure and herbaceous biomass, and also affects abiotic parameters including soil nutrients . Monitoring of structural components (i.e. tree species diversity, sizes, forms and LAI), anthropogenic disturbances and their interactions with environmental factors is important for effective management of human-modified moist forest and miombo woodland ecosystems in Tanzania, and elsewhere in Africa.
S1 Fig. A sketch diagram of plot design used in the moist forest and miombo woodland of Hanang district in Tanzania.
S2 Fig. Relationships between LAI or AGBH, and structural, environmental variables, and their combination in moist forest and miombo woodland.
(Forest 1) LAI show linear relationships with soil nitrogen in moist forest, (Miombo 2) with soil nitrogen, (Miombo 3) soil phosphorous, and (Miombo 4) soil potassium in Miombo woodland. (Forest 5) AGBH show linear relationships with elevation in moist forest, (Miombo 6) with soil nitrogen, (Miombo 7) soil pH and (Miombo 8) with disturbance in miombo woodland when all other variables are set to their mean values. The solid lines are the fitted partial regression lines with 95% shaded confidence band.
S1 Table. Global models used in predicting LAI and AGBH in moist forest and miombo woodland of Hanang district in Tanzania.
Models used either structural, environmental variables, or a combination of both. Note: Only predicator variables with variance inflation factor (VIF) ≤ 3 and Pearson correlation coefficient (r) ≤ 50% were included in the model.
S2 Table. Results of generalized linear models predicting LAI as a function of structural variables, environmental variables, and their combination in moist forest and miombo woodland of Hanang district in Tanzania.
S3 Table. Results of generalized linear models predicting aboveground herbaceous biomass (AGBH) as function of structural variables and environmental variables, and their combinations in moist forest and miombo woodland of Hanang district in Tanzania.
We thank Hamidu Seki, Rashi Khasim, John Herbet, George Bulenga, Osca Bakombezi, Godgift Swai, Godbless Lema and Hussein Shelukindo who assisted during fieldwork. Mr. Mhoro and Canicius Kayombo helped in plant identification. We thank P.K.T Munishi for his guidance and support during the design of this study.
Conceived and designed the experiments: DDS ØT SRM. Performed the experiments: DDS. Analyzed the data: DDS. Contributed reagents/materials/analysis tools: DDS MP PJP ØT SRM. Wrote the paper: DDS MP PJP ØT SRM.
- 1. Kohyama T (1993) Size structured tree populations in gap-dynamic forest the forest architecture hypothesis for the stable coexistence of species. J Ecol 81: 131–143.
- 2. Ishii H, Azuma W, Nabeshima E (2013) The need for a canopy perspective to understand the importance of phenotypic plasticity for promoting species coexistence and light-use complementarity in forest ecosystems. Ecol Res 28: 191–198.
- 3. Reich PB (2012) Key canopy traits drive forest productivity. Proc Biol Sci 279: 2128–2134. pmid:22279168
- 4. Turner IM (2001) The ecology of trees in the tropical rain forest: Cambridge University Press. 290 p.
- 5. Hardwick SR, Toumi R, Pfeifer M, Turner EC, Nilus R, et al. (2015) The relationship between leaf area index and microclimate in tropical forest and oil palm plantation: Forest disturbance drives changes in microclimate. Agric For Meteorol 201: 187–195.
- 6. Sagar R, Pandey A, Singh J (2012) Composition, species diversity, and biomass of the herbaceous community in dry tropical forest of northern India in relation to soil moisture and light intensity. The Environmentalist 32: 485–493.
- 7. Gatti RC, Castaldi S, Lindsell JA, Coomes DA, Marchetti M, et al. (2015) The impact of selective logging and clearcutting on forest structure, tree diversity and above-ground biomass of African tropical forests. Ecol Res 30: 119–132.
- 8. Veenendaal E, Torello-Raventos M, Feldpausch T, Domingues T, Gerard F, et al. (2015) Structural, physiognomic and above-ground biomass variation in savanna–forest transition zones on three continents–how different are co-occurring savanna and forest formations? Biogeosciences 12: 2927–2951.
- 9. Campbell B, Frost P, Byron N (1996) Miombo woodlands and their use: Overview and key issues. In: Campbell B, editor. The miombo in tansition: Woodlands and welfare in Africa. Bogor, Indonesia: Center for International Forestry Research (CIFOR). pp. 1–5.
- 10. Murphy PG, Lugo AE (1986) Ecology of tropical dry forest. Annu Rev Ecol Syst 17: 67–88.
- 11. Shirima D, Munishi PKT, Lewis SL, Burgess ND, Marshall AR, et al. (2011) Carbon storage, structure and composition of miombo woodlands in Tanzania’s Eastern Arc Mountains. Afr J Ecol 49: 332–342.
- 12. Pfeifer M, Lefebvre V, Gonsamo A, Pellikka PK, Marchant R, et al. (2014) Validating and linking the GIMMS leaf area index (LAI3g) with environmental controls in tropical Africa. Remote Sens 6: 1973–1990.
- 13. Mograbi PJ, Erasmus BFN, Witkowski ETF, Asner GP, Wessels KJ, et al. (2015) Biomass Increases Go under Cover: Woody Vegetation Dynamics in South African Rangelands. PLoS ONE 10: e0127093. pmid:25969985
- 14. Willcock S, Phillips O, Platts P, Balmford A, Burgess N, et al. (2014) Quantifying and understanding carbon storage and sequestration within the Eastern Arc Mountains of Tanzania, a tropical biodiversity hotspot. Carbon Balance and Manag 9: 2.
- 15. Marshall A, Willcock S, Lovett J, Balmford A, Burgess N, et al. (2012) Measuring and modelling above-ground carbon storage and tree allometry along an elevation gradient. Biol Conserv 154: 20–33.
- 16. Munishi P, Shear T (2004) Carbon storage in afromontane rain forests of the Eastern Arc Mountains of Tanzania: their net contribution to atmospheric carbon. J Trop For Sci 16: 78–93.
- 17. Fischer R, Ensslin A, Rutten G, Fischer M, Costa DS, et al. (2015) Simulating Carbon Stocks and Fluxes of an African Tropical Montane Forest with an Individual-Based Forest Model. PLoS ONE 10: e0123300. pmid:25915854
- 18. Williams M, Ryan CM, Rees RM, Sambane E, Fernando J, et al. (2008) Carbon sequestration and biodiversity of re-growing miombo woodlands in Mozambique. For Ecol Manage 254: 145–155.
- 19. FAO (2010) Global forest resources assessment Rome, Italy: FAO. 340 p.
- 20. Tanzania URo (1998) National Forest Policy. Dar es Salaam: Ministry of Natural Resources and Tourism. 59 p.
- 21. Dewees PA, Campbell BM, Katerere Y, Sitoe A, Cunningham AB, et al. (2010) Managing the Miombo woodlands of southern Africa: policies, incentives and options for the rural poor. Journal of natural resources policy research 2: 57–73.
- 22. Campbell B, Angelsen A, Cunningham A, Katerere Y, Sitoe A, et al. (2007) Miombo woodlands–opportunities and barriers to sustainable forest management. CIFOR, Bogor, Indonesia. Available: http://www.cifor.cgiar.org/miombo/docs/Campbell_BarriersandOpportunities.pdf (4th November 2008).
- 23. Frost PGH (1996) The ecology of miombo woodlands. In: Campbell B, editor. The miombo in transition: Woodlands and welfare in Africa Bogor Indonesia: CFIOR. pp. 11–57.
- 24. Thomas PA, Packham JR (2007) Ecology of woodlands and forests, description, dynamics and diversity. Cambridge, UK: Cambridge University Press. 528 p.
- 25. Gibbs HK, Brown S, Niles JO, Foley JA (2007) Monitoring and estimating tropical forest carbon stocks: making REDD a reality. Environ Res Lett 2: 045023.
- 26. Platts PJ O P, Marchant R (2015) AFRICLIM: high-resolution climate projections for ecological applications in Africa. Afr J Ecol 53: 103–108.
- 27. Lovett JC, Pocs I (1993) Assessment of the conditions of the catchment forest reserves, A botanical appraisal. Dar es Salaam, Tanzania.: Forest Division ⁄ NORAD. 300 p.
- 28. Scholes R, Kendall J, Justice C (1996) The quantity of biomass burned in southern Africa. J Geophys Res 101: 23667–23676.
- 29. Baas SP, Keßler PJA, Slik JWF, ter Steege H, Raes N (2011) Human and environmental influence on plant diversity and composition in Ben En National Park, Vietnam. J Trop For Sci 23: 328–337.
- 30. Kissinger G, Herold M, De Sy V (2012) Drivers of deforestation and forest degradation: a synthesis report for REDD+ policymakers.
- 31. Pfeifer M, Gonsamo A, Disney M, Pellikka P, Marchant R (2012) Leaf area index for biomes of the Eastern Arc Mountains: Landsat and SPOT observations along precipitation and altitude gradients. Remote Sens Environ 118: 103–115.
- 32. Jonckheere I, Fleck S, Nackaerts K, Muys B, Coppin P, et al. (2004) Review of methods for in situ leaf area index determination: Part I. Theories, sensors and hemispherical photography. Agric For Meteorol 121: 19–35.
- 33. Bréda NJ (2003) Ground‐based measurements of leaf area index: a review of methods, instruments and current controversies. J Exp Bot 54: 2403–2417. pmid:14565947
- 34. Gonsamo Gosa A (2009) Remote sensing of leaf area index: enhanced retrieval from close-range and remotely sensed optical observations. Helsingiensis Instituti Geographici Universitatis.
- 35. Fournier R, Landry R, August N, Fedosejevs G, Gauthier R (1996) Modelling light obstruction in three conifer forests using hemispherical photography and fine tree architecture. Agric For Meteorol 82: 47–72.
- 36. Jonckheere I, Muys B, Coppin P (2005) Allometry and evaluation of in situ optical LAI determination in Scots pine: a case study in Belgium. Tree Physiol 25: 723–732. pmid:15805092
- 37. Weiss M, Baret F (2014) CAN-EYE V6.313 User Manual. In: (INRA) FNIoAR, editor. Available: http://www6pacainrafr/can-eye/Documentation-Publications/Documentation.
- 38. Weiss M, Baret F, Myneni R, Pragnère A, Knyazikhin Y (2000) Investigation of a model inversion technique to estimate canopy biophysical variables from spectral and directional reflectance data. Agronomie 20: 3–22.
- 39. Pielou EC (1969) An introduction to mathematical ecology. Biometrische Zeitschrift 13: 219–220.
- 40. Shannon CEW (1949) The mathematical Theory of Communication. Urbana,: Univ. of Illinois Press. 117 p.
- 41. Van Laar A, Akça A (2007) Forest mensuration: Springer. 389 p.
- 42. West PW, West PW (2009) Tree and forest measurement: Springer. 191 p.
- 43. Zuur A, Ieno EN, Walker N, Saveliev AA, Smith GM (2009) Mixed effects models and extensions in ecology with R; Gail M, Krickeberg K, Samet JM, Tsiatis A, Wong W, editors. New York: Springer. 574 p.
- 44. Crawley MJ (2009) Mixed-Effects Model. The R Book. West Sussex. England: John Wiley & Sons Ltd. 627–660 p.
- 45. Crawley MJ (2012) The R book: John Wiley & Sons.
- 46. Dormann CF, Elith J, Bacher S, Buchmann C, Carl G, et al. (2013) Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography 36: 27–46.
- 47. Zuur AF, Ieno EN, Elphick CS (2010) A protocol for data exploration to avoid common statistical problems. Methods Ecol Evol 1: 3–14.
- 48. Murtaugh PA (2009) Performance of several variable-selection methods applied to real ecological data. Ecol Lett 12: 1061–1068. pmid:19702634
- 49. James G, Witten D, Hastie T, Tibshirani R (2013) An introduction to statistical learning: Springer. 441 p.
- 50. Canty A, Ripley B (2014) Package ‘boot’. Available at CRAN. Available: http://cran.rproject.org/web/packages/boot/index.html.
- 51. Quinn GP, Keough MJ (2002) Experimental design and data analysis for biologists: Cambridge University Press. 557 p.
- 52. Gilliam FS, Roberts MR (2003) Interactions between the herbaceous layer and overstory canopy of eastern forests: A mechanism for linkage. In: Gilliam FSR, R M., editor. The herbaceous layer in forests of eastern North America. Oxford, UK: Oxford University Press. pp. 198–223.
- 53. Unger M, Homeier J, Leuschner C (2013) Relationships among leaf area index, below-canopy light availability and tree diversity along a transect from tropical lowland to montane forests in NE Ecuador. Trop Ecol 54: 33–45.
- 54. DeRose RJ, Seymour RS (2010) Patterns of leaf area index during stand development in even-aged balsam fir-red spruce stands. Can J For Res 40: 629–637.
- 55. Yuan Y, Wang X, Yin F, Zhan J (2013) Examination of the quantitative relationship between vegetation canopy height and LAI. Adv Meteor 2013: 6.
- 56. Scholes RJ, Archer SR (1997) Tree-grass interactions in savannas. Annu Rev Ecol Syst 28: 517–544.
- 57. Gilliam FS (2007) The ecological significance of the herbaceous layer in temperate forest ecosystems. Bioscience 57: 845–858.
- 58. Coomes DA, Allen RB (2007) Effects of size, competition and altitude on tree growth. J Ecol 95: 1084–1097.
- 59. Callaway RM (2007) Interaction between competition and facilitation: Springer.
- 60. Bertness MD, Callaway R (1994) Positive interactions in communities. Trends Ecol Evol 9: 191–193. pmid:21236818
- 61. Malkinson D, Tielbörger K (2010) What does the stress‐gradient hypothesis predict? Resolving the discrepancies. Oikos 119: 1546–1552.
- 62. Price MF, Byers AC, Friend DA, Kohler T, Price LW (2013) Mountain geography: physical and human dimensions: Univ of California Press.
- 63. Ludwig F, Kroon H, Prins HH, Berendse F (2001) Effects of nutrients and shade on tree‐grass interactions in an East African savanna. J Veg Sci 12: 579–588.
- 64. Runyan CW, D'Odorico P, Lawrence D (2012) Effect of repeated deforestation on vegetation dynamics for phosphorus-limited tropical forests. J Geophys Res Biogeosci 117: G01008.
- 65. Ribeiro NS, Syampungani S, Matakala NM, Nangoma D, Ribeiro-Barros AI (2015) Miombo Woodlands Research Towards the Sustainable Use of Ecosystem Services in Southern Africa.