A slippery slope: logging alters mass–abundance scaling in ecological communities


Correspondence author. E-mail: umesh.srinivasan@gmail.com


  1. Natural ecosystems face ever-increasing anthropogenic threats from activities such as logging. It is therefore important to: (i) understand anthropogenic impacts on key community characteristics with implications for community structure and function and (ii) identify metrics with mechanistic underpinnings, providing a more functional understanding of anthropogenic impacts on biodiversity. Mass–abundance scaling is the most fundamental property of ecological communities, with implications on energy and resource partitioning and on how species and population traits translate to community structure and function.

  2. In habitat patches representing a continuum of logging intensity, I examined the impact of logging on mass–abundance scaling in understorey bird species using mist netting and bird ringing. I used regression quantiles to estimate the slope and intercept of upper bound of the polygonal local-scale mass–abundance relationship.

  3. I show that this slope becomes more negative (and the intercept higher) as logging intensity increases. Logging might therefore significantly alter resource and energy partitioning among species in natural communities.

  4. The mechanism underlying this pattern is likely to be an interaction between resource depletion, body size and density compensation. Larger species face more severe resource shortages from depleted resource availability and are expected to decline in abundance following logging, leading to more negative mass–abundance scaling. Density compensation by smaller species following large species decline can ‘push up’ the intercept of the mass–abundance relationship. This might result in both steeper slopes and higher intercepts in logged habitats.

  5. Synthesis and applications. This study shows for the first time that anthropogenic habitat change can alter fundamental community properties such as mass–abundance scaling and that this property of ecological communities was better at detecting logging impacts than other standard community measures. Identifying key metrics that provide a functional understanding of anthropogenic impacts on biodiversity is a crucial need, for both the assessment of these impacts and the continued monitoring of habitat change. This study highlights the need to supplement commonly used community descriptions with more mechanistic measures of human impacts on biodiversity.


Perhaps the most consistent and fundamental property of ecological communities is the negative relationship between body size (or body mass) and population size (or abundance) (Damuth 1981). This ubiquitous relationship forms a vital bridge in understanding the links between species and population characteristics on the one hand and emergent community structure and dynamics on the other (White et al. 2007; Emmerson 2011), including phenomena such as resource partitioning (Pagel, Harvey & Godfray 1991; Griffiths 1992), food web structure (Jennings & Mackinson 2003) and trophic ordering (Cohen, Jonsson & Carpenter 2003; Stouffer, Rezende & Amaral 2011). The disruption of mass–abundance scaling in any community, therefore, should also indicate changes in the way that resources and energy in the community are partitioned and cause downstream impacts on community structure and function.

A gamut of anthropogenic activities increasingly imperils community structure and function. The most severe impacts come from activities that cause habitat disruption, such as logging and fragmentation (Korfanta, Newmark & Kaufman 2012; Laurance et al. 2012). Selective logging, which is the predominant logging regime in the tropics (Edwards & Laurance 2013), is well known to impact the structure and function of plant and animal communities (e.g. Velho et al. 2012), although to a lesser extent than other land-use changes such as agriculture or plantation (Gibson et al. 2011; Edwards & Laurance 2013). Although the impacts of various logging practices on ecological communities have received much attention, there is little information on how logging affects fundamental community properties such as mass–abundance scaling. Most studies are still limited to examining changes in ‘standard’ community metrics (such as species richness, evenness or abundance) in response to logging (e.g. Edwards et al. 2012). There is therefore a severe paucity of information on the more mechanistic aspects of community response to habitat disruption, especially from tropical systems, where most of the world's biodiversity is concentrated (Gibson et al. 2011).

The precise nature of the mass–abundance relationship is scale dependent (review in White et al. 2007). At global or continental scales, the relationship between the log of abundance and the log of body mass is linear and negative (Damuth 1981; Carbone & Gittleman 2002), whereas at more local scales, the pattern is polygonal or triangular, with an upper bound on abundance with increasing body mass (Blackburn et al. 1993; Currie 1993; Blackburn & Gaston 1997; inset in Fig. 1). This upper bound is likely to be energetically determined (e.g. Blackburn, Lawton & Perry 1992) and represents the maximum possible abundance of a species, given its body size.

Figure 1.

A graphical representation of patterns in increasingly negative mass–abundance scaling from various community-level demographic processes. The solid line represents the upper bound of the local-scale mass–abundance relationship in an intact community. Mass–abundance scaling could become more negative from (a) solely a reduction in the abundances of large species (stippled line), (b) solely an increase in the abundances of small species (dashed line) or (c) a combination of both (a) and (b) (dotted line). Inset: The local-scale mass–abundance relationship. An example log–log plot of abundance (standardized by effort) on the y-axis and body mass on the x-axis, demonstrating the triangular relationship between mass and abundance at local scales. The solid line indicates the upper bound of the mass–abundance relationship at the local scale.

Changes in community structure in response to logging should result in altered mass–abundance scaling (specifically, changes in the slope and intercept of the upper bound; see inset in Fig. 1), with downstream impacts on community function. I tested whether logging, known to adversely impact wild populations and ecological communities (e.g. Thiollay 1992; Flaspohler, Temple & Rosenfield 2001; Velho et al. 2012), affected the mass–abundance relationship in understorey insectivorous birds. Logging and habitat degradation depress resource availability for insectivores (Burke & Nol 1998; Zanette, Doyle & Tremont 2000; Sodhi et al. 2009), and variation in resource availability is strongly expected to impact the mass–abundance relationship (Carbone & Gittleman 2002; Carbone et al. 2007).

Large species are more severely impacted by logging or fragmentation and tend to become less abundant in modified habitats because of, for instance, higher absolute energy requirements, large range sizes and lower population densities (e.g. Crooks 2002; Sodhi, Liow & Bazzaz 2004). I expected that larger species would decline preferentially more than small species in areas that were more intensively logged, ‘pulling down’ the right end of the mass–abundance polygon (Fig. 1). I therefore predicted a steep negative upper bound to the mass–abundance relationship for communities in logged areas when compared with that for intact communities in unlogged forest. In other words, I expected a negative relationship between the slope of the upper bound of the mass–abundance relationship and logging intensity.

A more negative slope of the upper bound of the mass–abundance relationship might arise from one of the three processes: (i) a reduction in the abundances of large species; (ii) an increase in the abundances of small species or; (iii) a combination of (i) and (ii). These three cases would lead to differences in the manner in which the slopes and intercepts of the mass–abundance relationship might be altered with logging (Fig. 1). I therefore also examined the change in the intercept of mass–abundance scaling in response to logging, to make inferences about which of the three processes outlined above might result in altered mass–abundance scaling.

Materials and methods

Study Area

I carried out fieldwork in the Eaglenest Wildlife Sanctuary, Arunachal Pradesh, India, in the Eastern Himalaya global biodiversity hotspot (Fig. 2). I sampled montane wet temperate forest (Champion & Seth 1968) at elevations centred around 2000 m a.s.l, because the species richness of understorey insectivorous bird species peaks at this elevation (data from Rasmussen & Anderton 2005) and, more generally, because bird species richness in this area is among the highest in the world (Orme et al. 2005). A large number of bird species can therefore be sampled to more reliably estimate local-scale mass–abundance relationships. Eight polygonal plots were established in this area, ranging in size from 2·5 to 4·2 hectares (Fig. 2). These plots represented a gradient in the intensity of past logging, from intact forest to intensively logged forest (logging occurred until 2002, after which it was halted). Patches were selected for sampling based on the interviews with past logging managers, who provided semi-quantitative information on the quantum of logging in the study area.

Figure 2.

Map of the study area. The black dot on the outline map of India indicates the position of Eaglenest Wildlife Sanctuary, the study area. Sampling plots are outlined in white, and each white dot represents the location of a mist net.

Sampling Methods

I sampled vegetation on each of these plots using the point-centred quarter method (Mitchell 2007) to estimate tree density (tree defined as an individual with diameter at breast height ≥10 cm). I sampled vegetation in approximately 25 (range 24–28) point-centred quarters in each of the larger sampling plots and used tree density as the variable representative of, and inversely proportional to, the intensity of past logging.

I sampled birds using mist netting and bird ringing in each of the eight plots from 23 March to 2 May 2011. This period is the early breeding season for understorey bird species in the study area (Rasmussen & Anderton 2005; pers. obs.). Mist netting (12 m, four-shelf nets; between 24 and 28 nets placed systematically per plot and run simultaneously) was carried out for three consecutive days per plot, from 05·00 to 12·00. Each individual captured was ringed, weighed and released, and the shelf in which each bird was captured was also noted. The lowest shelf of the mist net was numbered one, the next highest two and so on. For each species, the ‘mean shelf’ in which individuals were captured was calculated. Only species captured, on average, lower than shelf three, were included in the analysis. This was done to limit the analysis to understorey species with largely equal capture probabilities to enable cross-species comparisons of abundances and exclude canopy species that might be captured ‘by chance’ in mist nets. Although mist netting has clear advantages as a sampling method for forest birds (Karr 1981), it is possible that capture rates might not always reflect actual abundances (Thiollay 1994). However, results from mist netting have been validated through comparisons with other techniques (e.g. Barlow & Peres 2004), and classifying species captured below a certain height in mist nets provides an objective criterion for inclusion in analyses. One concern might be that large species are less likely to be captured by mist nets. However, the species considered in these analyses are small- to medium-sized understorey insectivores (5–165 g) that are capturable by mist nets used in this study (although we did also capture much larger birds such as pheasants). There are no large understorey insectivores (>250–300 g) in the study area. I also excluded carnivorous and frugivorous species, limiting the analysis to species that are predominantly insectivorous (based on diet data from Rasmussen & Anderton 2005).

Analytical Methods

For each plot, I estimated the upper bound of the local-scale log(mass)–log(abundance) relationship (henceforth, mass–abundance scaling, or the mass–abundance relationship) with quantile regressions using the package ‘quantreg’ (Koenker 2012) in the statistical software R (R Development Core Team 2012). Quantile regression is recommended to derive quantitative information from the boundaries of polygonal relationships (Scharf, Juanes & Sutherland 1998), especially since quantile regression models are robust to small sample sizes and outliers in the dependent variable (Scharf, Juanes & Sutherland 1998). Further, this modelling technique enables the detection of limiting factors on species responses (Vaz et al. 2008). Quantile regression, therefore, is suited to model the upper bound of the local-scale mass–abundance relationship, especially because the upper bound is likely to be energetically limiting on species (Blackburn & Gaston 1997; McGill 2008). Further, because of the polygonal nature of the data, results from ordinary least squares regression modelling are likely to be misleading, as fundamental assumptions such as homoscedasticity are likely to be violated.

The mean body mass of each species was obtained from data collected during the study. The number of individuals captured (excluding recaptures) was standardized by effort (number of net hours per plot) for each species and used as a proxy for abundance. The number of individuals is not the same as population size. However, given the filtering of species based on capture height and diet, and the subsequent analysis used, this metric is likely to be a suitable representation of population size for the following reasons: only understorey insectivorous species were considered, assuming that all these species would have similar capture probabilities using mist nets. In other words, across species, the number of individuals would consistently and equally be a fixed proportion of true population size. Given this assumption, the slope describing the relationship between mass and number of individuals would be identical to the slope of true population size on mass. Note that using raw number of captures versus an effort-standardized measure of the number of captures affects only the intercept of mass–abundance scaling, but not the slope.

The mean and standard errors of the quantile regression slope and intercept were estimated for each plot. While assessing potential patterns between the mean slope and intercept of mass–abundance scaling and the intensity of past logging, I sought to take into account the uncertainty associated with each of the slope and intercept estimates. I therefore used weighted regressions of the slopes and intercepts on tree density, with each of the slope estimates weighted by the inverse of the square of its standard error. The contribution of each point to the regression fit is therefore inversely related to the variance of the estimate.

It is possible that a significant relationship between mass–abundance scaling and logging intensity might emerge simply because of the selection of a particular quantile to estimate the slope of the upper bound of the mass–abundance relationship. To test the robustness of the relationship between the slopes and intercepts estimated from regression quantiles and logging intensity, I examined the patterns emerging from the selection of various quantiles to estimate the upper bound of the mass–abundance relationship (Table 1).

Table 1. Summary statistics from weighted regressions modelling the upper bound of the mass–abundance relationship (through regression quantiles) on tree density. Results from the higher regression quantiles show significant (or marginally significant) impacts of tree density on the slope of the mass–abundance relationships. The relationship for the intercept is somewhat weaker
F 1,6 R 2 P F 1,6 R 2 P

Apart from examining how mass–abundance scaling responds to logging, I also investigated the relationship between logging intensity and standard community metrics such as species richness (using the Chao 1 estimator), species diversity (using the Shannon–Wiener index) and community similarity [using non-metric multidimensional scaling (NMDS), with abundance data and the Bray–Curtis dissimilarity index]. In addition, I also examined rarefaction species richness curves for all the sampling plots to indicate completeness of sampling the bird community. All community-level analyses were carried out in the package ‘vegan’ (Oksanen et al. 2011) in the software R (R Development Core Team 2012).

Given the relatively small spatial scale of the study area, and the proximity of plots to each other (Fig. 2), spatial autocorrelation might result in the emergence of spurious relationships emerging from the spatial arrangement of plots alone. I tested for the presence of spatial autocorrelation in the slopes and intercepts of the mass–abundance relationships across plots using Moran's I. Moran's I is a measure of spatial autocorrelation that ranges from −1 (indicating complete dispersion) to + 1 (perfect correlation). A value of zero indicates randomness in spatial pattern. To confirm the independence of plots, I also checked the proportion of individuals that were captured and banded on one plot but recaptured on another plot.

A further concern is that logging intensity might be related to plot size or topology, again leading to patterns driven by plot characteristics alone, rather than by any true relationship between tree density and mass–abundance scaling. I therefore also tested for any relationship between tree density and (i) plot size, and (ii) plot perimeter–area ratios. A final concern is related to the variability in plot sizes–larger plots are expected to encompass more bird territories, and there is a possibility that the abundances of species with larger territories (likely to be larger species) are artificially inflated in larger plots. For a set of nine species (that represent a body mass gradient and were captured in almost all plots), I therefore checked for a relationship between plot size and the number of individuals captured.


The minimum tree density (on the most intensively logged plot) was 76·3 trees per hectare, whereas the highest tree density was 192·3 trees per hectare (in intact forest with no history of logging). Tree density in intact forest, therefore, was two-and-a-half times that in the most intensively logged plot. In all, 1100 individual birds (range across plots: 68–223) of 57 understorey species (range across plots: 24–38) were used in the analysis. (82 individuals belonging to 25 species were excluded based on capture height and diet.)

Individual-based rarefaction curves indicate that sampling of the understorey insectivorous bird community was reaching an asymptote in almost all the sampled plots (Fig. 3a). Species richness (as estimated by the Chao 1 estimator) and species diversity (as estimated by the Shannon–Wiener index) did not show any predictable relationship with logging intensity (Fig. 3c,d). Further, community similarity, as investigated by non-metric multidimensional scaling (NMDS) also did not reveal any significant pattern with logging intensity (Fig. 3b). However, given that relatively undisturbed plots appear to vary widely in community composition, and more logged plots tend to cluster closer together (Fig. 3b), it is possible that logged plots might be more homogeneous in terms of community composition than intact plots.

Figure 3.

Rarefaction curves for the bird community in different sampling plots (a) Line widths are proportional to tree density on the plots. Non-metric multidimensional scaling (NMDS on two axes) of community similarity in the eight sampling plots (b; stress = 0·062). Size of the circle is directly proportional to tree density. Species richness (c) and diversity (d) patterns in response to tree density. The dashed line in and the open circles in represent intact forest.

Despite the proximity of the sampling plots to each other (Fig. 2), I was unable to detect any spatial autocorrelation in the slopes and intercepts of the mass–abundance relationship using Moran's I (e.g. using the 0·75 regression quantile to estimate slopes and intercepts, Moran's I for slopes = −0·02, = 0·23; Moran's I for intercepts = −0·28, = 0·18). Also, only 10 of 1100 individuals (<1%) were banded on one plot, but recaptured in another, indicating that the sampling plots were largely independent. This proportion is similar to those used by other studies to infer independence of sampling units (e.g. Barlow & Peres 2004). Further, there was no relationship between plot area and tree density (Pearson's = 0·13, = 0·76) or plot perimeter–area ratio and tree density (Pearson's = −0·22, = 0·59). For a set of nine species (ranging in body mass from 6·1 to 71·3 g; mean = 20·3 g), there was no relationship between plot size and the number of individuals captured, indicating that biases in sampling related to body size are unlikely.

As expected, the estimated slopes of the mass–abundance relationships were positively correlated with tree density across plots (Fig. 4a, Table 1; results from the weighted regression). Therefore, the slope of the upper bound of the mass–abundance relationship was related negatively to logging intensity, indicating that mass–abundance scaling was modified by logging. Also, the intercepts of the mass–abundance relationship appeared to be negatively correlated with tree density for certain quantiles (Fig. 4b), but not for others (Table 1).

Figure 4.

The relationship between the mean slope (a) and intercept (b) of the upper bound of the mass–abundance relationship and logging intensity (as measured by tree density). As logging intensity increases (and tree density declines), this slope becomes progressively more negative and the intercept higher. The slope and intercept estimates in the figure are derived from the 0·75-regression quantile of the triangular mass–abundance relationship. The solid lines indicate the best fit from a weighted regression model (points weighted by the inverse of their variances). The curves indicate the 95% confidence interval of the slope of the fitted model.


As reported from other sites and for other taxa (Blackburn & Gaston 1997), local-scale mass–abundance relationships were polygonal (triangular) for insectivorous bird species in the Eastern Himalaya (inset in Fig. 1). Importantly, the slopes and intercepts of the upper bound of the mass–abundance relationship appeared to be related to the degree of logging that the habitat patch had experienced. Specifically, the slope became more steeply negative as logging intensity increased (Fig. 4a), whereas the intercepts became larger with greater logging intensity (Fig. 4b).

The slope of the upper bound of the local-scale mass–abundance relationship is an important property of ecological communities. Energetics very likely determines this upper limit to the polygonal relationship (Blackburn et al. 1993; McGill 2008). The possibility that the most abundant species in a community (near this upper bound) are energy limited is supported by empirical data–the values of the slope of these upper bounds in a large number of data sets are largely consistent with the predictions of energetic models (Blackburn et al. 1993). A change in this slope would therefore indicate fundamental changes in the way that energy and resources are partitioned between members of the community.

I found that the slope of the upper bound of the mass–abundance relationship in insectivorous bird species appears to vary predictably with logging pressure (Fig. 4), becoming more negative with increased logging intensity. This is the first time that an anthropogenic impact on habitat structure has been shown to affect mass–abundance scaling, one of the fundamental properties of any ecological community. In contrast, field studies and mesocosm experiments from freshwater, marine and intertidal systems indicate that the mass–abundance relationship is robust to anthropogenic disruption, food web disturbances and trophic cascades (Duran & Castilla 1989 in Marquet et al. 2005; Marquet, Navarette & Castilla 1990; Jonsson, Cohen & Carpenter 2005; O'Gorman & Emmerson 2011; but see Rossberg et al. 2008).

In taxa such as birds, species turnover (especially at small geographical scales) and changes in body size might not be able to compensate for changes in abundance to maintain mass–abundance scaling. Given this, certain questions remain to be answered: What mechanism can account for the change in the mass–abundance relationship in response to logging? What implications would this have for community structure and functioning, energy allocation in communities in modified landscapes and for biodiversity conservation in general?

These results, and the initial predictions of this study, are at variance with prior predictions that examine the impact of interspecific competition on mass–abundance scaling (notably, as examined by the mean slope of the relationship, rather than the upper quantile). Cotgreave & Stockley (1994) predict that in habitats with higher interspecific competition, mass–abundance scaling should be less negative (or even positive) because competition reduces species abundances, leading to a weaker relationship between abundance and mass, on average. However, I found an increasingly negative slope in more intensively logged habitats, which might reflect higher competition for resources in logged forest. Under this suggestion, competition could interact with body mass and resource shortages to change the slope of the upper bound of the mass–abundance polygon, but in the opposite direction to that expected by Cotgreave & Stockley (1994). Insect abundance and biomass do decline with logging and other anthropogenic threats (Sodhi et al. 2009; Berry et al. 2010), and insectivorous species are very likely to face resource crunches in degraded areas (see Burke & Nol 1998; Zanette, Doyle & Tremont 2000; Sodhi et al. 2009 for how fragmentation results in food shortages for insectivorous birds). As resources decline, larger species face greater resource shortages because of their higher absolute energy requirements (Sodhi et al. 2008), leading to higher interspecific competition between larger species. Larger species would therefore depress abundances of other large species to a greater extent than smaller species, making the slope of the upper bound more negative. This mechanism predicts an opposite pattern from that predicted by Cotgreave & Stockley (1994), and increasingly negative slopes in more intensively logged habitats from this study might reflect higher competition for resources in logged forest.

A related mechanism that can produce the patterns observed from this study (independent of competition) is simply resource depletion. As pointed out by Lawton (1990), species at the upper bound of the mass–abundance triangle are very likely to be resource limited–these species occur at abundances that should be very close (if not identical to) their carrying capacities. As productivity or resource availability falls, so should carrying capacity. However, the carrying capacity of all species at the upper bound of the mass–abundance relationship would not fall equally, but more so for large, rare species (Hutchinson 1959) occurring at the right tail of the mass–abundance triangle. Therefore, this corner of the triangle would fall lower on the y-axis with resource reduction.

Energy partitioning at the upper bound of the mass–abundance polygon also appears to change with logging. If the slope of the upper bound were −0·75, it would indicate that the ‘dominant’ species (that lie along this line) partition energy equally among themselves (Damuth 1981). However, the value of this slope in intact forest is much lower (roughly −0·25), indicating a significant energetic dominance by larger species (Griffiths 1992) in undisturbed communities. However, this dominance of large species in access to energy seems to erode predictably with logging – more and more steep slopes (and rising intercepts) indicate that smaller species appear to benefit in terms of access to resources from logging, whereas larger species tend to suffer.

Of the several ways in which mass–abundance scaling might be altered, there are two processes that might yield similar patterns in mass–abundance scaling, but have different consequences for ecosystem functioning: (i) simply a decline in the abundance of large species or; (ii) a decline in the abundance of a large species balanced to some extent by the addition of a weight-equivalent but less abundant species. For example, if the compensating species foraged in a different manner than the original species, this can actually change community functioning. In the case of this study, however, given that all the species in the analyses are understorey insectivorous birds, it is unlikely that community function would be altered in either of the two scenarios outlined above.

As shown in Fig. 1, different mechanisms (increase in abundance of small species, decline of large species or a combination of both) could result in steeper mass–abundance scaling with logging. These data indicate that both depressed abundances of large species and increased abundances of small species drive progressively more negative mass–abundance scaling due to logging, possibly from density compensation (MacArthur, Diamond & Karr 1972; O'Gorman & Emmerson 2011). Density compensation implies that as resources decline and large species reduce in abundance, smaller species utilize resources that would ordinarily be monopolized by large species and rise in abundance (line c in Fig. 1). This would lead to mass–abundance scaling having both steeper slopes and higher intercepts in response to logging, as reported here (Fig. 4). However, it must be noted that density compensation can only occur after large species decline, and therefore, the underlying mechanism behind altered mass–abundance scaling is very likely to be the reduced abundances of large species. If steeper mass–abundance scaling were to be solely because of the increased abundance of smaller species (with no impact on larger species), the patterns in Fig. 5 would not occur–rather, all the lines in Fig. 5 would theoretically converge to a single point (also see line b in Fig. 1). This further points to large species' declines (followed by density compensation) as the most feasible driver of altered mass–abundance scaling in logged habitats. However, caution is advised in interpreting these findings in the light of density compensation because: (i) the effort-standardized measure of abundance might not be suitable to compare mass–abundance intercepts; (ii) relatively weaker evidence from these data for a consistent relationship between the intercept and logging intensity for different regression quantiles (Table 1) and; (iii) relatively weak empirical evidence for density compensation in non-insular faunas (e.g. Peres & Dolman 2000; Larsen, Lopera & Forsyth 2008).

Figure 5.

Regression quantiles from the upper bound of the local-scale mass–abundance relationship from habitat patches with varying logging intensities. The thickness of the line is inversely proportional to logging intensity. As can be seen, thicker lines (from relatively more intact forest) have low intercepts and shallower slopes, whereas the thinner lines (from relatively more logged forest) have higher intercepts and steeper slopes. This is indicative of progressively intensifying large species declines and density compensation by smaller species as logging pressure increases.

One of the important implications of this work is highlighting the inadequacy of conventional community metrics when used in isolation in detecting the impacts of anthropogenic habitat modification. Measures such as species richness and species composition have been used extensively in making judgments about the value and conservation importance of modified landscapes (e.g. Edwards et al. 2012). These measures often do not reveal the ‘hidden’ impacts of anthropogenic activities on ecological communities (e.g. Baralato et al. 2012). Conventional metrics such as species richness and species diversity do have value in making certain inferences–for example, in evidencing the adequacy of sampling and to examine the impacts of more severe land-use changes such as agriculture or habitat fragmentation. Further, they are also simple to understand and communicate to policy makers and to a wider audience. However, they do suffer from shortcomings and need to be supplemented with more robust and informative metrics when making inferences about the value of modified landscapes for biodiversity (see Baralato et al. 2012).

Even from this data set, species richness, diversity and community similarity showed no patterns with logging intensity (Fig. 3). Using the same data set, however, logging appears to have detrimental impacts on the structure and functioning of ecological communities (Fig. 4), reinforcing the observation that there is no replacement for primary forest in conserving the full complement of biodiversity and ecosystem function, especially in the tropics (Gibson et al. 2011).

This study deals only with changing patterns in a bird community in response to logging at a single site. Further, this study was carried out at a relatively small spatial scale. Assessments based on multiple taxonomic groups (that could vary in their responses), and at larger scales, would be necessary to examine the universality of these patterns. However, I expect that given the energetic and resource-use changes in response to habitat modification, these patterns should be repeatable across sites and other taxa. Given the massive quantum of logging in the tropics (20% of all tropical forests were logged between 2000 and 2005; Edwards & Laurance 2013) and the presence of long-term monitoring programmes, there is ample opportunity to investigate patterns in mass–abundance scaling for a range of taxa in a large selection of sites.

These results reinforce the vulnerability of large species to anthropogenic habitat modification and integrate previous research on body size and extinction risk (review in Sodhi, Liow & Bazzaz 2004) with metabolic and energetic theory. This also provides a novel approach in examining the impacts of habitat degradation on ecological communities, relevant to examining the responses of a range of taxa that face an ever-increasing constellation of disruptive influences such as hunting, habitat loss and fragmentation.


I thank G. Rana, S. Rai, B. Tamang and D. Subba for assistance in the field, and I. Glow and N. Tsering for logistic help. S. Quader and J. Krishnaswamy helped with key conceptual and statistical issues, and the two anonymous reviewers and the editors substantially improved the conceptual content and style of this manuscript. I thank S. Balachandran and A. Rahmani of the Bombay Natural History Society for bird rings. The Arunachal Pradesh Forest Department provided permission to carry out this study, and US thanks T. Tapi, K.S. Jayachandran, M. Tasser and P. Ringu for their support. Thanks to N. Velho for her support through all stages of this study.