INTRODUCTION
Dominant plant species may regulate surrounding environment to influence other plant species diversity and composition (Angelini et al. Reference ANGELINI, ALTIERI, SILLIMAN and BERTNESS2011, Peh et al. Reference PEH, LEWIS and LIOYD2011). According to Grime (Reference GRIME1998), ecosystem properties, such as biomass production and diversity, are determined by the traits of the dominant species. Dominant plant species are termed foundation species if they determine the structure and composition of communities at local and regional scales (Caro Reference CARO2010, Dayton Reference DAYTON and Parker1972, Ellison et al. Reference ELLISON, BANK, CLINTON, COLBURN, ELLIOTT, FORD, FOSTER, KLOEPPEL, KNOEPP, LOVETT, MOHAN, ORWIG, RODENHOUSE, SOBCZAK, STINSON, STONE, SWAN, THOMPSON, HOLLE and WEBSTER2005). However, increasing abundance of the dominant plant species may have contrasting effects on co-occurring species (Dickson & Gross Reference DICKSON and GROSS2013). For example, a Gilbertiodendron dewevrei-dominated forest at Ituri reserve in the Democratic Republic of Congo had a comparable tree species richness (dbh ≥ 10 cm) with adjacent mixed forest (Djuikouo et al. Reference DJUIKOUO, PEH, NGUEMBOU, DOUCET, LEWIS and SONKÉ2014, Makana et al. Reference MAKANA, TERESE, HIBBS, CONDIT, Losos and Leigh2004), while tree species richness (dbh ≥ 10 cm) was lower in G. dewevrei-dominated forest in Dja Faunal reserve of Cameroon compared with adjacent mixed forests (Peh et al. Reference PEH, SONKÉ, SÉNÉ, DJUIKOUO, NGUEMBOU, TAEDOUMG, BEGNE and LEWIS2014). Removal of dominant plant species may have a significant impact on the remaining species (Dayton Reference DAYTON and Parker1972), because dominant species can create and maintain habitats that support other taxa of a community (Martin & Goebel Reference MARTIN and GOEBEL2013, Smee Reference SMEE2012).
Miombo woodlands, dominated by the genera Brachystegia and Julbernardia, are the most extensive (range: 2.7–3.2 million km2) deciduous woodland type in south-central and East Africa (Campbell et al. Reference CAMPBELL, FROST, BYRON and Campbell1996). However, plant species structure and composition in miombo woodlands has recently changed rapidly due to anthropogenic activities, such agricultural expansions, and local-climatic variability in the region (Frost Reference FROST and Campbell1996, Spinage Reference SPINAGE2012). These changes may cause decline in species richness or abundance and consequently influence species recruitment patterns and succession (Backéus et al. Reference BACKÉUS, PETTERSSON, STRÖMQUIST and RUFFO2006). For example, intensive removal of species of Brachystegia and Julbernardia, which are associated with ectomycorrhizas, have deep roots and produce slowly decomposing litter (Frost Reference FROST and Campbell1996), may affect other species recruitment and subsequent succession. Moreover, dominant woody species in miombo woodland often have high basal area and above-ground biomass, which are important in carbon cycling and other regulatory functions of the woodland (Munishi et al. Reference MUNISHI, MRINGI, SHIRIMA and LINDA2010, Ryan & Williams Reference RYAN and WILLIAMS2010). Yet there is limited information on how these dominant species interact with non-dominant woody species and affect community properties.
In this study we explored the relationships between the abundance of dominant miombo tree species richness, evenness, diversity and vertical structural heterogeneity of non-dominant tree species in wet and dry miombo woodlands. Although resprouting from surviving stems and root stocks is the main form of regeneration in miombo woodlands (Chidumayo Reference CHIDUMAYO2013), the dominant tree species from the genera Brachystegia and Julbernardia are known to have low recovery rates after major disturbances because of their low dispersal ability and short-lived seeds (Frost Reference FROST and Campbell1996). A previous study suggests that a change in the abundance of dominant plant species may cause changes in the growth patterns of non-dominants and their resource acquisition strategy (Tilman Reference TILMAN1985). We hypothesize (1) that there will be a negative relationship between the relative abundance of dominant species (dbh ≥ 5 cm) and the species richness, diversity, evenness and vertical structural heterogeneity of non-dominant trees, because dominant miombo tree species can suppress other tree species after escaping the ‘fire trap’ (at 3–6 m height; Frost Reference FROST and Campbell1996), (2) anthropogenic disturbances will reduce the negative effects of species dominance on Shannon diversity, evenness and vertical structure heterogeneity because frequent disturbance tends to promote plant species diversity in tropical forests (Connell Reference CONNELL1978).
MATERIALS AND METHODS
Study area
Miombo woodlands occupies about 90% of forested land from the north-west to the central, and along the eastern coast to regions further south in Tanzania (White Reference WHITE1983). They occupy a wide range of altitude (10–2000 m asl) and climate (mean annual rainfall: 500–1400 mm, mean annual temperature: 15°C–30°C; Frost Reference FROST and Campbell1996). Similar ecosystems occur in North-Central and West Africa (Sudanian or Guinea savanna woodlands), but unlike miombo woodlands they lack the dominance of the genera Brachystegia and Julbernardia. Instead they are dominated by Isoberlinia among others, mainly from Caesalpiniaceae (Ernst Reference ERNST1988, Frost Reference FROST and Campbell1996).
Miombo woodlands occur on nutrient-limited soils and at various macro- and micro-climates, and experience high disturbance that influences their vegetation structure and compositions (Campbell et al. Reference CAMPBELL, FROST, BYRON and Campbell1996). They are categorized as wet miombo woodlands in areas with above 1000 mm or dry in areas with less than 1000 mm mean annual rainfall (Frost Reference FROST and Campbell1996, Munishi et al. Reference MUNISHI, TEMU and SOKA2011, White Reference WHITE1983). Tree canopy cover varies from closed to open, with closed canopy in wet and open canopy in dry miombo woodlands (Frost Reference FROST and Campbell1996). The maximum height of mature tree canopies ranges between 18–27 m (Frost Reference FROST and Campbell1996, Malimbwi et al. Reference MALIMBWI, SOLBERG and LUOGA1994). We used AFRICLIM, which is a high-resolution climate projections dataset for Africa (Platts et al. Reference PLATTS, OMENY and MARCHANT2014) to categorize miombo woodlands into wet and dry miombo woodlands (Table 1).
A list of main variables estimated (mean ± SE) from the surveyed wet and dry miombo woodlands in Tanzania. A comparison of the main variables using Mann–Whitney–Wilcox test (U-test) between plots from dry and wet miombo woodlands.
We surveyed miombo woodlands located in Chunya, Hanang, Iringa Rural, Kilolo, Kilombero, Mufindi, Mbeya Rural and Mbozi districts (Figure 1). The districts were selected to represent a wide range of climatic conditions in miombo woodlands, and within each district, miombo woodlands were selected to capture a wide range of topographic gradients (Table 1). We surveyed randomly positioned plots along altitudinal gradients in each district between May 2011 and March 2012, and a total of 48 and 98 plots were measured in wet and dry miombo woodlands, respectively.
Miombo woodland study locations in Tanzania.
Data collection
We used rectangular plots of 20 × 40 m for the vegetation survey in wet and dry miombo woodlands (Shirima et al. Reference SHIRIMA, TOTLAND, MUNISHI and MOE2014). Rectangular plots were preferred over circular because they are widely used in vegetation surveys and suitable for capturing variations in heterogeneous environments (Goslee Reference GOSLEE2006, Scott Reference SCOTT1998, Stohlgren et al. Reference STOHLGREN, FALKNER and SCHELL1995). Plots were laid systematically along altitudinal gradients, at 400 m inter-plot distance to avoid within-site spatial autocorrelation. Inter-plot distances of 100 m to 1 km have previous been used for vegetation surveys in miombo woodlands (Banda et al. Reference BANDA, SCHWARTZ and CARO2006, Munishi et al. Reference MUNISHI, TEMU and SOKA2011). We used a hand-held Garmin Map76cx GPS to record the geographic location and altitude of each plot.
We measured tree stem diameter at breast height (dbh), tree height, and recorded species identity in each of the 146 plots (total 11.68 ha). Multi-stemmed individuals branching below 1.3 m were treated as separate individual stems. Tree heights were measured using a calibrated wooden rod and a Suunto hypsometer. We counted the number of stumps after tree felling in each plot and estimated the distance (km) from the nearest access road as indicators of disturbance from human activities. We identified tree species in the field where possible; otherwise, voucher specimens were collected and later identified at the Tanzania National Herbarium in Arusha.
Statistical analysis
We estimated the relative abundance of each species from individual species basal area divided by the total basal area of all species. We used an abundance distribution curve to identify the two most abundant species in wet and dry miombo woodlands, and derived two species groups (dominants and non-dominants) according to their relative abundance (Grime Reference GRIME1998).
Tree species were ranked by their relative abundance in ascending order and cumulative abundances for each species, where 100% frequency means that the species is present in all plots and 100% cumulative abundance corresponds to the most abundant species (Mariotte et al. Reference MARIOTTE, BUTTLER, KOHLER, GILGEN and SPIEGELBERGER2013). In each woodland type, two tree species were grouped arbitrarily as dominant (combined frequency greater than 90% and highest cumulative abundance), and the remaining tree species as non-dominants (Grime Reference GRIME1998, Mariotte et al. Reference MARIOTTE, BUTTLER, KOHLER, GILGEN and SPIEGELBERGER2013). Tree species richness were estimated as the total number of tree species, tree species diversity using Shannon's diversity index (Shannon Reference SHANNON1948), and evenness using Pielou's index (Pielou Reference PIELOU1969), in the non-dominant group in each plot. Since species richness is highly sensitive to sample size (Chao et al. Reference CHAO, GOTELLI, HSIEH, SANDER, MA, COLWELL and ELLISON2014), we calculated species rarefactions (using the Mao Tau rarefaction) to compare the two woodland types and estimated species richness of the non-dominants using Chao 2 estimator in EstimateS 8.2.0 (Colwell et al. Reference COLWELL, CHAO, GOTELLI, LIN, MAO, CHAZDON and LONGINO2012).
We estimated the vertical structural heterogeneity of the non-dominant tree species, using the species profile index (Hsp: Lei et al. Reference LEI, WANG and PENG2009, Pretzsch Reference PRETZSCH1996). This index is derived from the Shannon diversity index (H), and is based on grouping tree species into different height classes in a stand. These classes were defined relative to the height of the tallest tree in a stand (Class 1: within 81–100% of the tallest tree, Class 2: 50–80% of the tallest tree, Class 3: <50% of the tallest tree; Pretzsch Reference PRETZSCH1998). Individual tree heights were allocated to their appropriate classes, and Hsp is the proportion of each individual species occurring in the three classes, relative to the total number of trees species in the plot, as follows:
\begin{equation*}
H_{sp} = - \sum\nolimits_{i = 1}^s {\cdot \sum\nolimits_{j = 1}^B {\left\{{\begin{array}{*{20}c}
{p_{ij} \times \ln p_{ij} \,{\rm if}\,p_i > 0} \\
\cdot \\
{{\rm otherwise}\,0} \\
\end{array}} \right.}}
\end{equation*}
where Hsp = species profile index, S = tree species richness, B = number of height classes (3), pij = proportion of species i in class j.
The species profile index varies with the number of tree species and classes. To compare plot values therefore, we calculated a relative measure of the species profile index (RHsp) in each plot:
\begin{equation*}
RH_{sp} = \frac{{H_{sp}}}{{H_{spMax}}}{\rm where}\,H_{spMax} = \ln (S \times B)
\end{equation*}
where Hsp = species profile index and HspMax = maximum species profile index, respectively.
We used generalized least square regressions to fit separate models of tree species richness, Shannon diversity, evenness and the relative species profile index as response variables against the relative abundance of the dominant tree species, disturbance (distance from nearest access road and number of stumps) and interactions between disturbance and relative abundance of the dominant tree species as predictor variables. Exploratory analysis indicated non-linear relationships between tree richness, Shannon diversity, evenness and disturbance (distance from nearest access road) and the relative abundance of dominant tree species were therefore fitted using quadratic terms. Generalized least square models were preferred over multiple linear regressions to account for high heterogeneity among predictors in the dataset caused by large variation among different areas sampled (Zuur et al. Reference ZUUR, IENO, WALKER, SAVELIEV and SMITH2009). Each model was fitted by including one nominal weight (miombo woodland type) as a variance-covariate structure using restricted maximum likelihood (RML), because RML estimates stable variance components (Zuur et al. Reference ZUUR, IENO, WALKER, SAVELIEV and SMITH2009). We validated the final models and assessed their goodness-of-fit by observing the residual patterns (Zuur et al. Reference ZUUR, IENO and ELPHICK2010). All statistical analyses were done with the R software, version 3.1.0.
RESULTS
A total of 217 woody plant species (dbh ≥ 5 cm) from 48 families and 122 genera were recorded in 146 plots, amounting to a sampled area of 11.68 ha (Table 1, Appendix 1). The richness and the Shannon diversity of the non-dominant tree species were significantly higher in wet than in dry miombo woodlands (Table 1, 2). However, species rarefaction curves showed a similar pattern in species richness between wet and dry miombo woodlands, with slightly higher estimated tree richness in wet than in dry miombo woodland (Chao2 estimator, Figure 2a, b). Moreover, stem density and basal area of the non-dominant tree species were significantly higher in wet than in dry miombo woodlands (Table 2). The two most abundant species in wet miombo woodland were Brachystegia spiciformis Benth. and Brachystegia microphylla Harms, while Brachystegia spiciformis and Julbernardia globiflora (Benth.) Troupin were the most abundant species in dry miombo woodland, all from Caesalpiniaceae (Appendix 1, Figure 3a, b). Dominant tree species represented 37% and 45% of all tree stems in wet and dry miombo woodland, respectively (Table 2). In general, there was a relatively high dominance of small trees of both dominant and non-dominant tree species in the woodlands. Moreover, there were few large individual trees with dbh >50 cm of the dominant tree species and none of non-dominant tree species (Figure 4).
Structural attributes of non-dominants and dominant tree species of wet and dry miombo woodlands from eight districts (Figure 1) in Tanzania. A comparison of estimates, tree species structural characteristics using Mann–Whitney–Wilcox test (U-test (W)) between plots from dry and wet miombo woodlands.
Tree species rarefaction curves (Mao Tau function), indicating sampling efforts in wet (a) and dry (b) miombo woodlands sampled plots in Tanzania. The rarefaction curves in solid lines and 95% confidence intervals in dashed line, obs = number of observed species and Chao2 = the estimated species richness from 48 plots in wet and 98 plots in dry miombo woodlands.
Cumulative abundance as a function of frequency, showing the two most abundant tree species based on their relative basal area for the sampled plots in wet (a) and dry (b) miombo woodlands of Tanzania.
The distribution of tree stems (dbh ≥ 5 cm) in different diameter size classes in miombo woodlands of Tanzania.
Tree species richness was negative and linearly related to the relative abundance of the dominant tree species (P = 0.03, Table 3, Figure 5a), and had a hump-shape relationship with disturbance (distance to nearest access roads; P = 0.001, Table 3, Figure 5b). Tree species Shannon diversity had a negative non-linear relationship with relative abundance of the dominant tree species (P = 0.001, Table 3, Figure 5c). However, a significant interaction between relative abundance and disturbance shows that disturbance to some extent modified this relationship (P = 0.005, Table 3, Figure 5d): at high disturbance the relationship became significantly less negative compared with at low and medium disturbance. Tree species evenness had a non-linear negative relationship with the relative abundance (Table 3, Figure 6a). However, a significant interaction between relative abundance and disturbance shows that disturbance to some extent modified this relationship (P = 0.001, Table 3, Figure 6b): as was the case with diversity, at high disturbance the relationship became significantly less negative compared with at low and medium disturbance (Table 3, Figure 6b). The relative species profile index had a negative linear relationship with the relative abundance of the dominant tree species (P = 0.001, Table 3, Figure 6c). There was a significant interaction between the relative abundance of the dominant tree species and disturbance (P = 0.034, Table 3, Figure 6d): at high disturbance, there was no relationship between relative species profile index and disturbance whereas there were significant negative relationships at low and medium disturbances.
The relationships between tree species richness, Shannon diversity, evenness, and relative species profile index of the non-dominants and relative abundance of the dominant tree species in miombo woodlands of Tanzania. Generalized least squares models, showing significant variables (α ≤ 0.05) only.
The relationships between non-dominant tree species richness and relative abundance of dominants (a), tree species richness and disturbance (distance from road, (b)), Shannon diversity index and relative abundance of dominants (c), and relative abundance of dominants and the three disturbance levels (d), when all other variables are set to their medians in miombo woodlands of Tanzania. Plots show partial regression lines from generalized least square regression models of the relationships between tree species richness, Shannon diversity and the labelled variables (L-Stumps, M-Stumps and H-Stumps are Low, Medium and High number of stumps, respectively and represent disturbance levels).
The relationships between non-dominant tree species evenness and tree species relative abundance (a), tree species evenness and relative abundance of the dominants, and the three disturbance levels (b), when all other variables are set to their medians, tree species profile index and tree species relative abundance (c), and relative species profile index and relative abundance of the dominants, and the three disturbance levels (d), when all other variables are set to their medians in miombo woodlands of Tanzania. The plots show partial regression lines from generalized least square regression models of the relationships between tree species evenness, relative species profile index and the labelled variables (L-Stumps, M-Stumps and H-Stumps are Low, Medium and High number of stumps, respectively and represent disturbance levels).
DISCUSSION
We found negative relationships between tree species richness, Shannon diversity and evenness, and the relative abundance of dominant tree species in both wet and dry miombo woodlands. In habitats with intermediate resource levels, competition among dominant plant species tends to outweigh their facilitation effects on other plant species (Angelini et al. Reference ANGELINI, ALTIERI, SILLIMAN and BERTNESS2011, Bertness & Callaway Reference BERTNESS and CALLAWAY1994, Huston Reference HUSTON1979). Also high rates of biomass production by the dominant tree species can constrain space and nutrient availability to other plant species (Grime Reference GRIME1998). Previous studies have shown that re-sprouting from stems and root suckers are the main forms of tree species regeneration in miombo woodlands (Backéus et al. Reference BACKÉUS, PETTERSSON, STRÖMQUIST and RUFFO2006, Chidumayo Reference CHIDUMAYO2013). Our results indicate that the mean stem basal area of dominant tree species was slightly lower than that of non-dominant tree species in the two woodland types. Both non-dominants and dominants had a high number of stems in the low diameter size classes, which may indicate a good regeneration but also intensive competition between dominants and non-dominants in miombo (Backéus et al. Reference BACKÉUS, PETTERSSON, STRÖMQUIST and RUFFO2006). In addition, there were more large stems of dominant than non-dominant species, perhaps due to selective harvesting. The increase in canopy size and biomass of the large-stemmed dominants may suppress non-dominant species (Munishi et al. Reference MUNISHI, MRINGI, SHIRIMA and LINDA2010). Dominant miombo tree species can exploit limited soil nutrients more effectively than non-dominants because they have extensive ectomycorrhizal root systems (Frost Reference FROST and Campbell1996), which enhances their biomass production (Bâ et al. Reference BÂ, DUPONNOIS, MOYERSOEN and DIÉDHIOU2012, Diédhiou et al. Reference DIÉDHIOU, GUÈYE, DIABATÉ, PRIN, DUPONNOIS, DREYFUS and BÂ2005, Frost Reference FROST and Campbell1996). Nevertheless, our results suggest a good recovery, particularly after selective harvesting, which is the main anthropogenic disturbance factor in miombo woodlands (Backéus et al. Reference BACKÉUS, PETTERSSON, STRÖMQUIST and RUFFO2006, Chidumayo Reference CHIDUMAYO2013).
We found a hump-shaped pattern between Shannon diversity and the relative abundance of the dominant tree species, and the interactions between relative abundance of the dominant tree species and disturbance (number of stumps). This may imply that the influence of disturbance on biotic interactions is determined by disturbance intensity (Connell Reference CONNELL1978). However, the dominant tree species can assimilate nutrients, such as extractable phosphorus and water, throughout the soil profile and store considerable quantities of carbohydrates over long periods, thereby buffering the system against losses through fire, herbivory and year-to-year fluctuations in climate (Bâ et al. Reference BÂ, DUPONNOIS, MOYERSOEN and DIÉDHIOU2012, Chidumayo & Gumbo Reference CHIDUMAYO and GUMBO2010, Munyanziza Reference MUNYANZIZA1994). Although, it is well established that plant species diversity in miombo woodlands is shaped by historical disturbances (Dewees et al. Reference DEWEES, CAMPBELL, KATERERE, SITOE, CUNNINGHAM, ANGELSEN and WUNDER2011, Frost Reference FROST and Campbell1996, Runyan et al. Reference RUNYAN, D’ODORICO and LAWRENCE2012), we did not have adequate estimates of disturbances, especially those that are more linked to dominant tree species. Thus, further studies are required to disentangle the underlying mechanism for the observed hump-shaped pattern.
We found a non-linear relationship between tree species richness and disturbance (distance to access road), which suggests that vegetation in plots near the road are recovering faster after disturbance compared with plots that are far from an access road. A previous study has documented that there is intensive harvesting of trees along roads, targeting tree species suitable for charcoal and timber production (Ahrends et al. Reference AHRENDS, BURGESS, MILLEDGE, BULLING, FISHER, SMART, CLARKE, MHORO and LEWIS2010, Schwartz & Caro Reference SCHWARTZ and CARO2003). We found a negative non-linear relationship between Shannon diversity, evenness and relative abundance of the dominant tree species at low disturbance (low number of stumps), suggesting that disturbance can also reduce tree species diversity (Connell Reference CONNELL1978). Apart from selective harvesting, other forms of disturbance such as frequent fires have an impact on plant diversity in miombo woodlands (Frost Reference FROST and Campbell1996). For example, previous results from fire experiments in miombo woodlands of Zambia have shown that disturbances from fire play a crucial role in maintaining species diversity and composition in the woodland ecosystem (Trapnell Reference TRAPNELL1959). Moreover, regular fire occurrences promote rapid pulsing of nutrient release from otherwise slowly decomposing litter and herbaceous biomass (Chamshama & Vyamana Reference CHAMSHAMA, VYAMANA, Bongers and Tennigkeit2010). Miombo woodlands in Tanzania, like in other parts of Africa, have experienced climatic and anthropogenic disturbances for decades (Campbell et al. Reference CAMPBELL, FROST, BYRON and Campbell1996), which has varying impacts on the species diversity in the woodland ecosystem (Frost Reference FROST and Campbell1996, Spinage Reference SPINAGE2012). Furthermore, our results show that observed tree species richness differ significantly between wet and dry miombo woodland, but the estimated richness (Chao 2) and rarefaction pattern suggested that the wet and dry miombo woodlands may have little difference in tree richness if sampled adequately. The actual observed tree species richness and diversity was from a wide range of families and genera, similar to previous studies (Banda et al. Reference BANDA, SCHWARTZ and CARO2006, Munishi et al. Reference MUNISHI, TEMU and SOKA2011).
We observed a significant negative relationship between the relative species profile index and the relative abundance of the dominant miombo tree species. This suggests that dominant miombo tree species are supressing the non-dominant tree species and hence dominate the higher canopy stratum (Pretzsch Reference PRETZSCH1998). Moreover, the relative species profile index decreased with increasing relative abundance of dominant tree species at low disturbance (low number of stumps), which further suggest that dominant tree species are supressing the non-dominant tree species. The vertical structure of miombo woodlands is characterized by a uniform canopy of the dominant tree species within single sites, with large areas ranging from a discontinuous shrub layer (Frost Reference FROST and Campbell1996) to a homogeneous overstorey canopy. Strong interspecific competition for space between the most dominant tree species and other tree species at different growth stages may result in niche partitioning among tree species (Peterson et al. Reference PETERSON, RICE and SEXTON2013), which could promote vertical size differentiation among trees if exposing the understorey species to more space and light resources.
We found a negative association between tree species richness, Shannon diversity, evenness and relative profile index of the non-dominant and relative abundance of the dominant tree species. It is possible that dominant miombo tree species out-compete other tree species due to their extensive root systems with ectomycorrhizal associations (Bâ et al. Reference BÂ, DUPONNOIS, MOYERSOEN and DIÉDHIOU2012, Frost Reference FROST and Campbell1996), which enhance their ability to access limited nutrients. This competition effect may be enhanced because these dominant tree species may not be the main targeted in selective logging because of their relatively low preference in charcoal and timber uses (Ahrends et al. Reference AHRENDS, BURGESS, MILLEDGE, BULLING, FISHER, SMART, CLARKE, MHORO and LEWIS2010, Schwartz & Caro Reference SCHWARTZ and CARO2003). Moreover, dominant miombo tree species are known to have a high recovery rate after mild disturbance or after escaping the ‘fire trap’, because of their ability to coppice from surviving stems or root suckers (Frost Reference FROST and Campbell1996). It will likely require further efforts to understand how dominant miombo tree species influence trees species diversity under contrasting local physiographic and anthropogenic disturbance factors.
ACKNOWLEDGEMENTS
We thank the Norwegian Government, through its Climate Change Impacts, Adaptation and Mitigation programme, and the Norwegian State Education Loan Fund for funding this study. Hamidu Seki, Rashi Khasim, John Herbet, George Bulenga, Osca Bakombezi, Godgift Swai and Godbless Lema assisted with field work. Botanists Yahaya Abeid, Moses Mwangoka and Canicius Kayombo helped with plant identification. We thank Peter Frost, two anonymous reviewers for their constructive comments.
A list of species encountered in plots (n = 146) surveyed in miombo woodlands of Tanzania.
