Fragmented tropical forests can maintain tree diversity and function, but not under high deforestation

Fragmented tropical forests can maintain tree diversity and function, but not under high deforestation | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL This is a preprint and has not been peer reviewed. Data may be preliminary. 6 October 2025 V2 Latest version Share on Fragmented tropical forests can maintain tree diversity and function, but not under high deforestation Authors : Bruno X. Pinho 0000-0002-6588-3575 [email protected] , Lourens Poorter , Dimitri Justeau-Allaire , Fabian Fischer , Jerome Chave , and Isabelle Maréchaux 0000-0002-5401-0197 Authors Info & Affiliations https://doi.org/10.22541/au.175647864.44539581/v2 466 views 233 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Habitat loss threatens biodiversity, yet the independent effects of habitat fragmentation remain debated. Resolving this debate is critical for managing biodiversity in tropical forests undergoing rapid land-use change, but it requires a sampling design difficult to implement empirically. We developed a virtual framework by extending a spatially explicit, trait- and individual-based model to simulate tropical forest landscapes and disentangle the impacts of forest loss, fragmentation, and patch size on tree diversity and function. Deforestation reduced species richness, evenness, and biomass at patch and landscape scales. Smaller patch size and fragmentation further exacerbated these declines—except for neutral to positive fragmentation effects on landscape-scale species richness—but only in landscapes with less than 30% forest cover. Trait-mediated species turnover underpinned these changes, leading to increased dominance by fast-growing, high-dispersal species. These results suggest that conservation efforts should prevent further fragmentation and prioritize the restoration of large patches in highly deforested regions. Introduction Land-use change is the leading global driver of biodiversity loss and the disruption of ecosystem services (Lewis et al. 2015; IPBES 2018). Converting natural habitats to agricultural, urban, or other human uses not only reduces total habitat area but also fragments what remains (Taubert et al. 2018). While habitat loss is a recognized cause of biodiversity decline, the independent role of fragmentation remains debated (Riva & Fahrig 2023; Gonçalves-Souza et al. 2025). Disentangling the effects of habitat amount and fragmentation across scales is critical for effective conservation (Arroyo-Rodríguez et al. 2020). Yet, these factors are often tightly correlated and their effects context-dependent in real landscapes, making their individual impacts difficult to isolate empirically. To address this, we develop an individual-based mechanistic model to simulate tropical forest dynamics in human-modified landscapes and clarify how habitat amount and fragmentation affect tree diversity and function at patch and landscape scales. Chase et al. (2020) showed that small patches support fewer species than expected by species–area relationships, due to “ecosystem decay” processes such as edge effects and dispersal limitation. Yet, Riva & Fahrig (2023), using the same fragSAD database (Chase et al. 2019), found that patch-level decay does not necessarily translate into reduced landscape-scale biodiversity under fragmentation, since several small patches can collectively contain more species through increased compositional dissimilarity compared to a few large patches of the same total area. Meta-analyses also suggest that, when habitat amount is controlled, fragmentation effects on richness are often weak and positive (Fahrig 2017, 2020), challenging the primacy of large patches (Diamond 1975) and highlighting the value of small patches (Riva & Fahrig 2022). Conversely, recent syntheses report consistent biodiversity declines in fragmented landscapes, with turnover failing to compensate for local losses (Gonçalves-Souza et al. 2025). These divergent results highlight the need for deeper mechanistic insights (Fletcher et al. 2023), but progress is limited by constraints of empirical approaches. First, properly testing fragmentation effects requires multiple landscape replicates spanning independent habitat amount and fragmentation gradients (Fahrig 2017), yet empirical studies rarely meet this. Riva & Fahrig (2023) analyzed overlapping subsets of patches within the same area, risking conflation of fragmentation effects with metacommunity dynamics—such as rescue effects (Hanski 1998), where species in smaller patches persist through immigration from larger surrounding patches excluded from the sample. Second, empirical work seldom examines habitat amount-fragmentation interactions, despite theoretical and some empirical evidence that fragmentation can exacerbate biodiversity loss especially at low habitat amount due to increased extinction risk and dispersal limitation (Andrén 1994; Fahrig 1997; Villard & Metzger 2014; Zhang et al. 2024). Third, focusing narrowly on diversity metrics like species richness or evenness can obscure community shifts; weak patch size effects on richness may mask high turnover with disturbance-sensitive species replaced by disturbance-adapted taxa (Chase et al. 2020), potentially driving functional impoverishment (Pinho et al. 2025). Overall, given the vast spatial and temporal scales of this debate, empirical studies remain necessarily constrained by their sampling schemes (Estes et al. 2018), which often yield limited, correlated gradients of habitat amount and fragmentation and exhibit unbalanced geographic and taxonomic coverage. For example, just three of 117 studies in the FragSAD database analyzed tropical trees (Chase et al. 2019), despite tropical forests containing over two thirds of global biodiversity and facing severe land-use pressures (Giam 2017). We address these knowledge gaps by simulating tree community structure and ecosystem functioning in human-modified tropical forest landscapes. Spatially explicit modeling complements empirical studies (Fisher et al. 2018; Maréchaux et al. 2021) and offers a mechanistic understanding of biodiversity change. However, current landscape-scale individual-based tropical forest models (e.g., Köhler et al. 2003; Chetcuti et al. 2020) are limited: they use few functional groups, assume uniform dispersal and recruitment, and impose seed production and dispersal externally rather than deriving them from demographic processes. To overcome these limitations, we extend the spatially explicit, individual-based, trait-based model TROLL (Fig. 1a, Maréchaux & Chave 2017), which uniquely simulates diversity, structure, and functioning of tropical forest tree communities—a rare capability in vegetation modeling (Maréchaux et al. 2021). To disentangle habitat amount and fragmentation effects across scales, we simulate 5 × 5 km tropical forest landscapes spanning independent gradients of these variables (Fig. 1b). This virtual experiment tests key findings from Chase et al. (2020) at the patch scale and Riva & Fahrig (2023) at the landscape scale, while extending previous work by explicitly evaluating interactions among drivers across scales and trait-mediated community assembly. Habitat amount is expressed as total forest cover, and fragmentation as the number of patches per simulated landscape. Patch size (the area of an individual patch) predicts patch-level biodiversity and biomass, whereas total forest cover and fragmentation serve as predictors at both patch and landscape scales. We hypothesize that habitat loss primarily drives biodiversity and biomass decline, with fragmentation exerting additional scale- and context-dependent effects: (1) smaller patch size reduces species richness, evenness, and biomass within patches; (2) fragmentation promotes species turnover among patches, potentially increasing landscape-scale diversity despite local decay; (3) the effects of patch size and fragmentation are strongest under low habitat amount; and (4) fragmentation and smaller patches shift community composition toward species with traits associated with rapid growth and high dispersal ability. Finally, we evaluate whether species turnover offsets patch-scale ecosystem decay (Chase et al. 2020) or enhances diversity at the landscape scale (Riva & Fahrig 2023). Model overview We used TROLL, an individual-based, spatially explicit, trait-based model of tropical forest dynamics that builds on advances in plant physiology (Maréchaux & Chave 2017). The source code is written in C++, with a new interface in R ( rcontroll package; Schmitt et al. 2023) . TROLL simulates the recruitment, growth, dispersal, and mortality of individual trees through carbon assimilation and allocation processes in a light-limiting environment that is explicitly computed within a voxel space of 1 m 3 . Physiological and demographic processes are mediated by morphological and physiological traits that can vary around species-specific means. A comprehensive description of TROLL’s structure, processes, and parameterization is provided in Appendix A. TROLL has been applied to a range of ecological questions, including biodiversity and ecosystem functioning relationships in undisturbed tropical forests (Maréchaux & Chave 2017), tropical forest resilience to disturbances (Schmitt et al. 2020), inferences of tree allometry from remote sensing (Fischer et al. 2019), and responses to windthrow (Rau et al. 2022). However, despite its potential to address cross-scale forest dynamics, TROLL’s applications have been so far restricted to the plot scale, with simulations of local tree communities under constant external seed rain and no explicit landscape effects on plant demography or dispersal. To extend TROLL’s applications to fragmented tropical landscapes, we implemented several novel, empirically grounded mechanisms, as follows. Model developments Trait-based tree fecundity and establishment In previous versions of the TROLL model, it was assumed that the equalizing trade-off between seed production and seedling survival, driven by variation in seed size (Moles 2018), would result in a “zero-sum game”, so that the number of seedling opportunities produced per timestep and per mature tree was assumed to be constant across species. Therefore, seeds of different species in the seed bank had the same probability to establish, proved that they had enough light to sustain a positive carbon balance. We here relax this assumption by explicitly simulating tree species fecundity and establishment rates, based on the following trait-based models (Visser et al. 2016): \[\log\left(f\right)=-4.234-1.223\times\log_{10}\left(m_{s}\right)+0.108\times\text{wd}_{t}-0.0005\times\text{lma}_{t}\] \[-0.0564\times\log_{10}(\text{maxdbh}_{t})\] (Eq. 1) \[\text{logit}\left(E_{r}\right)=4.33+1.428\times\log_{10}\left(m_{s}\right)+0.098\times\text{wd}_{s}-0.000013\times\text{lma}_{s}\] \[-2.421\times\log_{10}(\text{maxdbh}_{s})\] (Eq. 2) where f is the tree fecundity, defined as the number of seeds produced per year per adult tree basal area, and E r is the establishment rate, defined as the number of new recruits (i.e., seedlings) per seeds arriving per year per m 2 (Visser et al. 2016). The other terms represent functional traits: m s is the species’ average seed dry mass (g), while wd t , lma t and maxdbh t (respectively wd s , lma s and maxdbh s ) are the tree-level (respectively species’ average) wood density (g cm -3 ), leaf mass per area (g m -2 ) and maximum diameter at breast height (mm). Only 60% of the produced seeds are actually dispersed to account for pre-dispersal seed predation (Jackson et al. 2022). To calculate a species’ establishment probability in each site (1-m 2 grid-cell), we multiplied its relative local seed abundance by its Er . Although some of the traits included in the models above are not expected to be directly related to tree fecundity or initial establishment (e.g. wood density), they were selected in the best models from Visser et al. (2016) and may distinguish plant life-history strategies (Grime & Pierce et al. 2012). This is, to our knowledge, the most comprehensive study of functional traits’ effects on vital rates across the life cycle of tropical forest trees, based on multiple censuses of the 50-ha forest dynamics plot on the Barro Colorado Island (BCI), Panama. By incorporating these demographic models from Visser et al. (2016) into TROLL, we assume that neotropical forest regions exhibit a similar range of variation in tree functional traits—a reasonable assumption (Pinho et al. 2021)—and that trait effects on demographic rates are consistent across regions. This latter assumption in particular warrants further exploration as additional data become available from other regions. Trait-based dispersal We estimated tree maximum dispersal distance ( maxDD , in m) using a global model (Tamme et al. 2014) based on tree height ( height , in m) and species’ mean seed mass \(\left(m_{s}\right)\ \)and dispersal syndrome (S s ), as follows: \(\log_{10}\left(\text{maxDD}\right)=2.95+\ S_{s}-0.32\times\log_{10}\left(m_{s}\right)+0.60\times\log_{10}\left(h\text{eig}ht\right)\) (Eq. 3) where S s =0, -0.72, -1.03 and -1.43 for animal-dispersed species, ant-dispersed species, wind-dispersed and ballistic-dispersed species, respectively, and m s is the species average seed dry mass in mg. maxDD was then assumed to be the 95 th percentile of a Rayleigh distribution of scale parameter \(\sigma\) (which, by definition of the cumulative distribution function of the Rayleigh distribution, leads to\(\sigma=0.41\times maxDD).\ \)We used this distribution as the tree’s dispersal kernel, thus allowing for eventual long-distance dispersal events. Edge-driven mortality Finally, the widely observed increasing tree mortality rates with edge proximity was additionally implemented using the model from Laurance et al. (1998): \(D_{r}=\ \left\{\begin{matrix}1\ \text{if}\ E_{d}\geq 100\ m\\ b_{0}\times\exp\left(-0.19+1.94\times E_{d}^{-0.66}\right)\text{if}\ E_{d}<100\ m\\ \end{matrix}\right.\ \) (Eq. 4) Where D r is the multiplicative factor increasing death rate due to edge effects relative to forest interior, E d is the distance to the edge (m), and b 0 (= 1.09) is the inverse of the edge mortality rate at the maximum distance of edge effects (i.e., 9% increase in mortality), here defined as 100 m. We defined this maximum edge effect distance because it is in this forest portion where desiccated conditions and fast decay in tree species composition are mostly observed, although other abiotic and biotic edge effects usually penetrate much further (Laurance et al. 2018). Trait and climate forcing data We compiled trait data for 164 tree species from French Guiana forests (Baraloto & Forget 2007; Baraloto et al. 2010), including leaf mass per area, leaf nitrogen and phosphorus concentration, wood density, maximum stem diameter, seed mass, and dispersal syndrome. These species largely cover the spectrum of global relationships in tree functional traits (Maynard et al. 2022). They include a prevalence of species with high-density leaves and woods (Supporting Information, Fig. S1), i.e. conservative resource-use, that are typical for the flora of French Guiana and a large part of the Amazon (ter Steege et al. 2025), and that may be particularly sensitive to landscape modification (Pinho et al. 2025). Climate forcing was based on daily data from the GUYAFLUX eddy flux tower (Aguilos et al. 2018). Design and simulation of landscape scenarios We used the neutral landscape generator flsgen (Justeau-Allaire et al. 2022) to create fine-grained virtual tropical forest landscapes spanning wide, independent gradients of forest loss and fragmentation. We generated 307 landscapes covering six discrete forest cover levels (10, 20, 30, 40, 50, 60%), 11 fragmentation levels defined by the number of patches (1, 4, 8, 16, 32, 48, 64, 80, 96, 112, 128), and five replicates per forest cover–fragmentation combination (except single-patch scenarios), plus one control landscape with no forest loss. Replicates maximized patch size distribution variability within combinations. Patch sizes ranged from 1 to 1,500 hectares across landscapes, though low forest cover and high fragmentation yielded narrower ranges (see Fig. 2 solid and dashed lines), with all patches constrained to a square shape (Fig. S2). A minimum 10-meter gap between patches ensured clear spatial separation, approximating a two-lane road width. Each landscape also incorporated 20 regenerating patches of 1 hectare each, representing natural regeneration in abandoned tropical fields (Crouzeilles et al. 2017). We first simulated a 5 × 5 km old-growth forest by running TROLL for 600 years from bare ground, applying uniform external seed rain for the first 300 years (following Maréchaux & Chave 2017). After this spin-up, external seed rain ceased, and forest dynamics relied solely on seed production and dispersal explicitly simulated within the virtual landscape. This old-growth state served as the baseline for all disturbed landscape simulations. Disturbance was inmposed by extracting patch coordinates from flsgen outputs, killing trees and restricting recruitment outside remnant patches . We assumed a uniformly ‘hostile’ matrix without remnant trees or recruitment (except within 20 regenerating patches), leaving matrix heterogeneity (Fletcher et al. 2025) for future work. Disturbed landscapes were simulated for an additional 600 years, comparable to the age of long-fragmented tropical forests like the Brazilian Atlantic Forest (Joly et al. 2014). Our simulated landscape size aligns with prior studies on landscape effects on tree diversity (e.g., Watling et al. 2020; Arasa-Gisbert et al. 2022; Pinho et al. 2025). To account for broader-scale metacommunity processes beyond this extent, we implemented toroidal boundaries in all simulations. This allowed seeds dispersing beyond one edge to re-enter from the opposite edge, thereby simulating a continuous, borderless environment without artificial edge effects. Model output analysis To assess forest cover and fragmentation effects at patch and landscape scales, we followed Chase et al. (2020) and Riva & Fahrig (2023). At the patch level, we performed spatially constrained rarefaction by sampling each forest patch 100 times using 1-ha plots (100 × 100 m), sampled with replacement, averaging response variables across samples. At the landscape level, full inventories ensured sampling effort scaled proportionally with patch size, avoiding bias related to patch size–tree density relationships. We included individuals ≥10 cm DBH to align with most empirical studies (e.g., Chase et al. 2019). We measured fragmentation by the number of patches, which we consider a more intuitive metric, and obtained consistent results (not shown) with mean patch size, the metric used by Riva & Fahrig (2023), which is highly correlated under fixed forest cover. Response variables included species richness and evenness, quantified as Hill numbers of order q = 0 (species richness) and q = 2 (Simpson diversity, i.e. the effective number of dominant species; Chao et al. 2014), as well as above-ground biomass (AGB) density (Mg ha -1 ). We focused on AGB instead of stem density, as biomass loss can be masked when using stem counts alone, especially near forest edges where few large trees are replaced by many smaller ones (Santos et al. 2008; Laurance et al. 2018). This was supported by a negative correlation between stem density and AGB in our models, indicating AGB better reflects ecosystem decay in our study system. We also measured community functional composition using trait means based on individual-level traits. All variables were assessed at patch and landscape scales. Patch-scale responses were modeled using linear mixed-effects models with individual patch size and landscape-scale forest cover and fragmentation (number of patches) as predictors, including two- and three-way interactions. Simulated landscapes were specified as random effects with random intercepts to account for nesting of patches within landscapes. At the landscape scale, where predictors and responses share the same spatial scale, multiple linear regression was used with forest cover, fragmentation, and their interaction as predictors, without random effects. Non-linear effects were addressed by examining interactions between other drivers and forest cover. To test if compositional dissimilarity among patches mediates biodiversity responses, we decomposed incidence-based Jaccard dissimilarity into turnover and nestedness components (Baselga 2010). Turnover quantifies species replacement; nestedness indicates smaller patches’ species forming subsets of larger ones. We calculated patch-size ratios (larger patch area divided by smaller) for all patch pairs within landscapes, then related these ratios to turnover and nestedness using linear models per landscape (Fig. S3). These models yielded slopes describing the effects of patch size (i.e. patch-size ratio) on turnover or nestedness, which we related to slopes from species diversity ~ patch size models. This examined whether higher turnover between small and large patches offsets biodiversity loss via disturbance-adapted species gains (Chase et al. 2020). At the landscape scale, we tested if increased turnover in fragmented landscapes increases species richness (Riva & Fahrig 2023) by relating fragmentation effects on turnover and nestedness to fragmentation effects on species diversity. This framework links compositional changes to biodiversity outcomes across scales. At the species level, we quantified changes in species’ relative abundance from simulation start to end at patch and landscape scales, relating these to functional traits via multiple regressions. Patch-level analysis assessed the effect of patch size on per-species change in relative abundance within landscapes, then related per-species slopes (averaged across landscapes) to traits. Landscape-level analyses related per-species abundance changes directly to traits across landscapes. All analyses were conducted in R 4.3.3 (R Core Team 2024). Results Ecosystem decay from patch to landscape scales At the patch level, forest cover was the dominant driver of biodiversity and biomass, with individual patch size also exerting substantial influence on per-hectare species richness, evenness, and biomass (Fig. 2, see also Table S1 and Fig. S4). Effects varied along the forest cover gradient, with greatest gains at low cover. For example, increasing forest cover from 10% to 20% corresponded to a 20% rise in per-hectare species richness and a 45% increase in evenness, while at higher cover (50% to 60%), gains were smaller (~3% and 5%, respectively). Above-ground biomass increased moderately, rising nearly 4% per 10% cover increase at lowest forest cover and less than 1% at highest cover. One standard deviation increases in individual patch size (≈3.6-fold increase) led to 4–5% gains in per-hectare richness and 8–20% gains in evenness, with the strongest effects observed in landscapes characterized by low forest cover, especially when fragmentation was also low. Fragmentation had smaller effects on richness (~3% gain per one standard deviation increase of about 33 patches), but did not significantly affect evenness or biomass at the patch level. At the landscape level, forest cover remained the primary driver of post-disturbance biodiversity and biomass trajectories (Fig. S5), increasing species richness, evenness, and biomass density (Fig. 3, Table S2). Species richness increased modestly with forest cover, rising by approximately 10% as cover increased from 10% to 60%, while fragmentation had no significant effect on richness across forest cover levels. Evenness and biomass responded positively to increasing forest cover, with evenness increasing by 2–6% at intermediate to high forest cover and biomass rising steadily but modestly (up to around 5%) across the gradient. Fragmentation effects were strongly context dependent: it sharply reduced both evenness (by 20% to 46%) and biomass (up to 5%) at low forest cover (30%). Patch size and fragmentation effects on species compositional dissimilarity Slopes (effect sizes) from patch-level models relating patch-size ratios to pairwise species turnover (x-axis) and patch size to species richness (y-axis) were negatively correlated across landscapes (Fig. 4a). Landscapes where larger patch-size ratios were associated with greater compositional turnover tended to show weaker positive effects of patch size on richness, consistent with turnover-driven compositional changes dampening richness gains. Conversely, landscapes where patch-size ratios were more strongly related to nestedness—particularly under low forest cover and low fragmentation—showed stronger positive patch size effects on richness (Fig. 4b). At the landscape level, average pairwise dissimilarity increased with forest loss and fragmentation, driven primarily by species turnover, which became the dominant component under extreme disturbance (Fig. S6). At low forest cover, fragmentation increased turnover but decreased nestedness (Fig. S6), paralleling its contrasting effects on landscape diversity: a weak positive effect on richness linked to increased turnover and a strong negative effect on evenness associated with decreased nestedness (Fig. 5). Winner-loser species and trait replacements Forest loss, fragmentation, and patch size drove trait-mediated shifts in species composition at patch and landscape scales (Tables S3–S4). Changes in community-weighted mean traits (Fig. 6) and shifts in species abundances (Fig. S7) indicated that decreasing patch size and increasing fragmentation favored—at the patch and landscape scales, respectively—species with lower maximum height, wood density, LMA, seed mass, and leaf nutrient concentrations, especially under low forest cover. These trait shifts were generally more pronounced in landscapes with low forest cover, except for seed mass that increased with patch size more strongly in high- than low-forest-cover landscapes (Fig. 6). Discussion Using a novel trait-based modelling framework, we simulated tropical forest dynamics in human-modified landscapes to assess the independent effects of habitat loss, fragmentation, and patch size on biodiversity and biomass. Forest loss was the primary driver of declines at both patch and landscape scales: species richness, evenness, and biomass all decreased with increasing deforestation. Effects of patch size and fragmentation intensified as forest cover declined—positive patch-size effects peaked in highly deforested landscapes, while negative fragmentation effects on landscape-scale evenness and biomass were likewise greatest, despite neutral richness responses. Highly deforested landscapes became dominated by a few fast-growing, highly dispersive species, especially in small patches and fragmented landscapes. These results parallel observed tropical forest degradation and offer new mechanistic insights into the scale- and context-dependent impacts of habitat loss and fragmentation on biodiversity. Patch-scale ecosystem decay in highly deforested landscapes: trait-mediated impacts of edge effects and isolation At the patch level, biodiversity declines observed within standardized samples support the notion of ecosystem decay described by Chase et al. (2020). This decay—manifested as reduced species richness, evenness, and biomass in smaller patches—is stronger than expected from passive sampling alone, likely due to intensified edge effects, dispersal limitation, and demographic stochasticity. Importantly, our analysis explicitly tested how patch size effects interact with landscape-scale forest cover and fragmentation, extending prior work (Chase et al. 2020). According to the Habitat Amount Hypothesis (HAH; Fahrig 2013), local biodiversity is primarily driven by the total habitat amount in the surrounding landscape rather than patch size or isolation (Watling et al. 2020). Consistent with this, forest cover was the main driver of patch-level biodiversity, with fragmentation effects on species richness, when significant, relatively weak and positive (Table S1). Reducing individual patch size exacerbated biodiversity loss mainly in landscapes with low forest cover (< 30%) (Fig. 2). This indicates small patches retain biodiversity when embedded in landscapes with high forest cover, likely due to immigration from larger species pools and rescue effects (Hanski 1998), consistent with HAH and supported by empirical studies (Watling et al. 2020). Conversely, in heavily deforested landscapes, small patches become edge-affected and isolated, dominated by a few disturbance-adapted, highly dispersive species (Santos et al. 2008; Tabarelli et al. 2012). Fragmentation interacted with forest cover and patch size to influence patch-level biodiversity (Table S1). Patch size effects were strongest in landscapes with low forest cover and fragmentation (Fig. 2), where few large patches occur amid many smaller ones. Large patches retain interior, edge-free habitat supporting disturbance-sensitive species absent from edge-dominated small patches (Santos et al. 2008), resulting in strong richness differences. These patterns suggest patch-scale biodiversity loss is primarily driven by pervasive edge effects, with large patches shaping biodiversity under low habitat cover, while metacommunity dynamics mitigate patch-level decay with increased connectivity under high forest cover. Changes in functional composition further elucidate these patterns. In highly deforested landscapes, smaller patches were dominated by species with traits suited to high light and disturbance (Fig. 6a)—soft leaves with high nutrient concentrations, low wood density, short stature, and fast maturity (Grime & Pierce 2012). These shifts reflect trait-mediated filtering from increased mortality and canopy openness near edges (Laurance et al. 1998), causing biomass declines in small, edge-affected patches (Fig. 2; de Paula et al. 2011). Notably, seed mass was less sensitive to patch size in more deforested landscapes (Fig. 6a), as all patches became dominated by small-seeded species—a likely result of large-seed dispersal limitation with reduced connectivity. Fragmentation increases turnover and maintains richness under high forest cover but leads to impoverishment under low forest cover At the landscape level, fragmentation did not affect species richness (Fig. 3a), as increased turnover among patches compensated for local losses (Figs. 5 and S6). This pattern aligns with various taxa studies where local (α) diversity declines are offset by increased β-diversity, potentially maintaining or enhancing γ-diversity (Fahrig 2020; Riva & Fahrig 2023). However, a recent synthesis found turnover often insufficient to rescue biodiversity in fragmented landscapes (Gonçalves-Souza et al. 2025). Biodiversity persistence depends on dispersal capacity and environmental heterogeneity—scale-dependent factors varying across studies. Our simulations covered 25-km² landscapes and may overestimate species richness by underrepresenting dispersal limitations prominent at larger scales (Gonçalves-Souza et al. 2025). Larger-scale studies may reveal steeper richness declines due to increased patch isolation and reduced connectivity for a given habitat amount, but also capture broader environmental gradients, which may enhance turnover that partially offset losses (Riva & Fahrig 2023). Importantly, despite stable landscape-scale species richness with fragmentation, our simulations revealed substantial taxonomic and functional reorganization, especially in highly deforested landscapes (Fig. 3). Fragmentation caused pronounced declines in species evenness, consistent with extinction-debt—where sensitive species persist temporarily at low abundance before extirpation (Haddad et al. 2015). Meanwhile, generalist pioneer trees proliferate and dominate altered landscapes (Fig. 6b; Tabarelli et al. 2012; Pinho et al. 2025). These opportunistic species are characterized by traits favoring rapid growth (soft tissues) and effective dispersal (small seeds). As ecosystem processes hinge on traits of dominant species (Grime 1998), this pioneer dominance may undermine forest functioning and resilience under low forest cover. Notably, the capacity of remnant forests to absorb and store carbon may decline (Fig. 3c), as these “winning” species grow rapidly but have higher mortality rates (Wright et al. 2010). Benefits and limitations of the modeling framework We used a spatially explicit, individual-based, trait-driven model combined with a landscape generator spanning independent gradients of forest loss and fragmentation. As discussed above, this overcame limitations of previous studies. However, our framework also relies on some simplifying assumptions, reflecting current data availability and process understanding. We assumed fecundity and dispersal depend entirely on tree size and few key functional traits, based on established empirical relationships (Tamme et al. 2014; Visser et al. 2016). While some noise and missing covariates are possible, our simulated trait-based patterns align with recent empirical findings (Arasa-Gisbert et al. 2022; Pinho et al. 2025). Our model of edge effects on tree mortality used distance from forest edges (Laurance et al. 1998), rather than explicit trait-based microclimatic or edaphic mechanisms (Camargo & Kapos 1995; Nunes et al. 2022). We assumed a homogeneous hostile matrix surrounding patches, although matrix heterogeneity may affect biodiversity persistence in fragmented landscapes (Fletcher et al. 2025). Future empirical work could help refine these assumptions and guide specific data collection to improve mechanistic understanding of these processes. The species pool (164 species) examined in this study encompasses much of the global range of tree functional strategies (Maynard et al. 2022), but regional biogeography and environment influence species and trait persistence under disturbance (Betts et al. 2019; Banks-Leite et al. 2022). Our conservative pool represents largely undisturbed Amazonian forests (ter Steege et al. 2025) and likely reflects a worst-case scenario for fragmentation effects. In contrast, regions with a history of biogeographic instability or repeated disturbance cycles may show attenuated impacts due to prior species loss or adaptation (Betts et al. 2019). Replicating our study with different species pools and climates can reveal how biogeographic and climatic context mediate responses to forest loss and fragmentation. Implications for fragmentation research and tropical forest management This study provides mechanistic insights into the fragmentation-biodiversity debate, showing context-dependent, non-linear fragmentation effects. We provide strong evidence that fragmentation exacerbates forest loss impacts in tropical forest landscapes by intensifying edge effects and dispersal limitation. These findings support predictions from metacommunity (Zhang et al. 2024) and stochastic spatial models (Rybicki et al. 2020) and the fragmentation threshold hypothesis (Andrén 1994), indicating effects shift from neutral or positive to strongly negative below critical habitat thresholds (Villard & Metzger 2014). Our results also highlight the importance of integrating functional with taxonomic metrics to fully capture landscape modification effects. Species diversity alone can hide significant floristic and functional shifts driven by trait-mediated processes, which reduce forest resilience to future disturbances and climate extremes (Schmitt et al. 2020; Sakschewski et al. 2016). Two key conservation messages emerge. First, preserving and restoring native forest cover is crucial: biodiversity and biomass decline sharply below 30% cover, supporting established thresholds (Arroyo-Rodriguez et al. 2020, 2021; Banks-Leite et al. 2021). Second, fragmentation is most harmful where deforestation is greatest. The few reports of positive fragmentation effects on tropical forest trees—such as those synthesized by Fahrig (2020) and Riva & Fahrig (2023)—often come from regions that have undergone long-term biogeographic instability, such as Mexico’s forests (e.g., Hernandez-Ruedas et al. 2014), which experienced repeated Pleistocene expansion–retraction cycles at the northern edge of Neotropical forests (Graham 1999). This history may have favored disturbance-adapted species (Betts et al. 2019) that now show muted responses to recent landscape change (Hernandez-Ruedas et al. 2014). Recent deforestation there may also reflect unrealized extinction debt (Haddad et al. 2015). By contrast, geologically and ecologically stable regions that experienced extensive historical human-driven deforestation—like the Brazilian Atlantic Forest—show clear evidence of taxonomic and functional impoverishment with landscape modification, consistent with our findings (Santos et al. 2008; Tabarelli et al. 2012; Pinho et al. 2025). Overall, although fragmentation may appear benign—or even beneficial—in intact landscapes or certain taxa, it consistently drives major losses of tree diversity and function under low forest cover. In these highly modified landscapes, conservation must therefore focus on halting further fragmentation and restoring large patches to safeguard tropical forest biodiversity and ecosystem functioning. Acknowledgements This study was supported by the French National Research Agency (ANR) under the ”Investissements d’avenir” program (ANR-16-IDEX-0006 and ANR-10-LABX-25-01). LP was supported by ERC Advanced Grant PANTROP (grant nr. 834775). We thank Marco Visser for valuable insights into trait-based modeling of tree fecundity and seedling establishment. We also appreciate the IT support provided by Frédéric Theveny, Thomas Arsouze and Philippe Verley for IT support. Our gratitude extends to Jannah Oliveira for creating the illustration of our methodological framework (Fig. 1). Figure 1. Methodological framework for simulating tropical forest dynamics in human-modified landscapes. (a) Scheme illustrating key processes in the extended TROLL model. Novel developments (blue arrows/yellow circles) include seed production (f, eq. 1), establishment rate (Er, eq. 2), and maximum dispersal distance (maxDD, eq. 3), all modeled using functional traits—seed mass (SM), wood density (WD), leaf mass per area (LMA), and maximum diameter (maxDBH). Edge-driven mortality (Dr) is modeled as a function of edge proximity. Existing processes, such as explicit voxel-based light competition, carbon assimilation, and allocation to leaf production and wood growth, follow Maréchaux & Chave (2017). Illustrative tree shapes are shown for clarity, although the model simulates cylindrical trunk/crown geometry (see Methods and Appendix A). (b) Landscape simulation approach. We first simulate a 5 × 5 km old-growth forest from bare ground, which serves as the baseline for all subsequent simulations of disturbed landscapes (n = 307) spanning independent gradients of forest cover and fragmentation, with forest cover-fragmentation combinations replicated across different patch size distributions. Figure 2. Drivers of patch-level biodiversity and biomass. Model predictions showing three-way interactions between landscape-scale forest cover (rows, ‘FC’) and fragmentation (number of fragments; line colors), and individual patch size (x-axis), as drivers of species richness (left column), species evenness (middle column), and aboveground biomass (AGB, right column), with values standardized as the mean of 100 randomly placed 1-ha (100×100 m) samples per patch across simulated forest patches. Landscapes were simulated across six levels of forest cover, but only results for highest forest cover (60%, top row), intermediate forest cover (30%, middle row), and lowest forest cover (10%, bottom row) are shown. Shaded areas around model fit lines represent 95% confidence intervals. Figure 3. Drivers of landscape-level biodiversity and biomass. Relationships between habitat fragmentation (number of patches; x-axis) and landscape-scale species richness and evenness (measured as effective number of species and dominant species, i.e. Hill numbers of order q = 0 and q = 2, respectively) and above-ground biomass in simulated landscapes across different forest cover levels (point and line colors). Each point represents a simulated landscape, excluding the control landscape (n = 306). Shaded areas around model fit lines represent 95% confidence intervals. Figure 4. Relationships between model estimates (i.e., effect sizes) predicting patch size effects on species richness (y-axis; see Figs. 2 and S3) and effects of patch-size ratios (i.e. pairwise relative differences in patch size) on Jaccard’s compositional dissimilarity (x-axis), partitioned into turnover (a) and nestedness (b) components (see Figs. S5 and S6). Each point represents a simulated landscape with a given forest cover (colors) and fragmentation level (point size), excluding those with a single forest patch and the control scenario (n = 301). Pearson’s correlation coefficients (with 95% confidence intervals) are –0.32 (–0.42, –0.21) for turnover (a) and 0.24 (0.13, 0.35) for nestedness (b). Figure 5. Associations between fragmentation effects (model slopes) on landscape-scale species diversity (y-axis)—species richness (upper panels) and species evenness (lower panels)—and fragmentation effects on average compositional dissimilarity between forest patches (x-axis), partitioned into turnover (left panels) and nestedness (right panels) components. Each point and error bar represents the slope estimate and standard deviation from models fitted to groups of 50 simulated landscapes (10 fragmentation levels × 5 replicates) with the same forest cover level (colors). Figure 6. Community-weighted functional trait mean responses to (a) individual patch size at the patch level and (b) fragmentation (number of patches) at the landscape level, across simulated landscapes with varying forest cover (colors). In (a), each line shows model predictions of trait means, representing trends across 2,940 individual forest patches pooled from 50 simulated landscapes (10 fragmentation levels × 5 replicates) at a specific forest cover level. In (b), each point summarizes data from one simulated landscape at a given forest cover and fragmentation level. WD: wood density; LMA: leaf mass per area. References Aguilos, M., Hérault, B., Burban, B., Wagner, F. & Bonal, D. (2018). What drives long-term variations in carbon flux and balance in a tropical rainforest in French Guiana? Agricultural and Forest Meteorology, 253–254, 114–123. Andrén, H. (1994). Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos, 71, 355–366. Arasa-Gisbert, R., Arroyo-Rodríguez, V., Meave, J. A., Martínez-Ramos, M. & Lohbeck, M. (2022). Forest loss and treeless matrices cause the functional impoverishment of sapling communities in old-growth forest patches across tropical regions. Journal of Applied Ecology, 59, 1897–1910. Arroyo-Rodríguez, V., Fahrig, L., Tabarelli, M., Watling, J. I., Tischendorf, L., Benchimol, M., et al. (2020). Designing optimal human-modified landscapes for forest biodiversity conservation. Ecology Letters, 23, 1404–1420. Arroyo-Rodríguez, V. et al. (2021). Preserving 40% forest cover is a valuable and well-supported conservation guideline: reply to Banks-Leite et al. Ecology Letters, 24, 1114–1116. Banks-Leite, C., Betts, M. G., Ewers, R. M., Orme, C. D. L. & Pigot, A. L. (2022). The macroecology of landscape ecology. Trends in Ecology & Evolution, 37(6), 480–487. Banks-Leite, C., Larrosa, C., Carrasco, L. R., Tambosi, L. R. & Milner-Gulland, E. J. (2021). The suggestion that landscapes should contain 40% of forest cover lacks evidence and is problematic. Ecology Letters, 24, 1112–1113. Baraloto, C. & Forget, P.-M. (2007). Seed size, seedling morphology, and response to deep shade and damage in neotropical rain forest trees. American Journal of Botany, 94(6), 901–911. Baraloto, C., Paine, C. E. T., Poorter, L., Beauchene, J., Bonal, D., Domenach, A.-M., et al. (2010). Decoupled leaf and stem economics in rain forest trees: Decoupled leaf and stem economics spectra. Ecology Letters, 13(11), 1338–1347. Baselga, A. (2010). Partitioning the turnover and nestedness components of beta diversity. Global Ecology and Biogeography, 19, 134–143. Betts, M. G., Wolf, C., Pfeifer, M., Banks-Leite, C., Arroyo-Rodríguez, V., et al. (2019). Extinction filters mediate the global effects of habitat fragmentation on animals. Science, 366, 1236–1239. Camargo, J. L. C. & Kapos, V. (1995). Complex edge effects on soil moisture and microclimate in central Amazonian forest. Journal of Tropical Ecology, 11, 205–221. Cazalis, V. (2022). Species richness response to human pressure hides important assemblage transformations. Proceedings of the National Academy of Sciences, 119, e2107361119. Chao, A., Chiu, C.-H. & Jost, L. (2014). Unifying Species Diversity, Phylogenetic Diversity, Functional Diversity, and Related Similarity and Differentiation Measures Through Hill Numbers. Annual Review of Ecology, Evolution, and Systematics, 45, 297–324. Chase, J. M., Liebergesell, M., Sagouis, A., May, F., Blowes, S. A., Berg, Å., et al. (2019). Frag SAD: A database of diversity and species abundance distributions from habitat fragments. Ecology, 100(12), e0 2861. Chase, J. M., Blowes, S. A., Knight, T. M., Gerstner, K. & May, F. (2020). Ecosystem decay exacerbates biodiversity loss with habitat loss. Nature, 584, 238–243. Chetcuti, J., Kunin, W. E. & Bullock, J. M. (2020). Habitat fragmentation increases overall richness, but not of habitat-dependent species. Frontiers in Ecology and Evolution, 8. Crouzeilles, R, Ferreira, M. S., Chazdon, R. L., Lindenmayer, D. B., Sansevero, J. B. B., Monteiro, L., et al. (2017). Ecological restoration success is higher for natural regeneration than for active restoration in tropical forests. Science Advances, 3, e1701345. de Paula, M. D., Costa, C. P. A. & Tabarelli, M. (2011). Carbon storage in a fragmented landscape of Atlantic Forest: the role played by edge-affected habitats and emergent trees. Tropical Conservation Science, 4, 349–358. Diamond, J. M. (1975). The island dilemma: Lessons of modern biogeographic studies for the design of natural reserves. Biological Conservation, 7, 129–146. Estes, L., Elsen, P. R., Treuer, T., Ahmed, L., Caylor, K., Chang, J., et al. (2018). The spatial and temporal domains of modern ecology. Nature Ecology and Evolution 2, 819–826. Fahrig, L. (2013). Rethinking patch size and isolation effects: the habitat amount hypothesis. Journal of Biogeography, 40, 1649–1663. Fahrig, L. (2017). Ecological responses to habitat fragmentation per se. Annual Review of Ecology, Evolution, and Systematics, 48(1), 1–23. Fahrig, L. (2020). Why do several small patches hold more species than few large patches? Global Ecology and Biogeography, 29(4), 615–628. Fisher, R. A., Koven, C. D., Anderegg, W. R. L., Christoffersen, B. O., Dietze, M. C., Farrior, C. E., et al. (2018). Vegetation demographics in Earth System Models: A review of progress and priorities. Global Change Biology, 24, 35–54. Fletcher, R. J., Smith, T. A. H., Kortessis, N., Bruna, E. M. & Holt, R. D. (2023). Landscape experiments unlock relationships among habitat loss, fragmentation, and patch‐size effects. Ecology, 104(5), e4037. Giam, X. (2017). Global biodiversity loss from tropical deforestation. PNAS 114, 5775–5777. Graham, A. (1999). The Tertiary history of the northern temperate element in the northern Latin American biota. American Journal of Botany, 86, 32-38. Grime, J. P. & Pierce, S. (2012). The evolutionary strategies that shape ecosystems (1st ed.). Wiley. Haddad, N. M., Brudvig, L. A., Clobert, J., Davies, K. F., Gonzalez, A., Holt, R. D., et al. (2015). Habitat fragmentation and its lasting impact on Earth’s ecosystems. Science Advances, 1, e1500052. Hanski, I. (1998). Metapopulation dynamics. Nature, 396, 41–49. Hernández-Ruedas, M. A., Arroyo-Rodríguez, V., Meave, J. A., Martínez-Ramos, M., Ibarra-Manríquez, G., Martínez, E., et al. (2014). Conserving tropical tree diversity and forest structure: The value of small rainforest patches in moderately-managed landscapes. PLOS ONE, 9, e98931. IPBES (2018). The IPBES assessment report on land degradation and restoration. Montanarella, L., Scholes, R. & Brainich, A. (eds.). Secretariat of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services, Bonn, Germany. Jackson, E. E., Wright, S. J., Calderón, O., Bullock, J. M., Oliver, T. & Gripenberg, S. (2022). Pre‐dispersal seed predation could help explain premature fruit drop in a tropical forest. Journal of Ecology, 110(4), 751–761. Joly, C. A., Metzger, J. P. & Tabarelli, M. (2014). Experiences from the Brazilian Atlantic Forest: Ecological findings and conservation initiatives. New Phytologist, 204(3), 459–473. Justeau-Allaire, D., Blanchard, G., Ibanez, T., Lorca, X., Vieilledent, G. & Birnbaum, P. (2022). Fragmented landscape generator (flsgen): A neutral landscape generator with control of landscape structure and fragmentation indices. Methods in Ecology and Evolution, 13, 1412–1420. Köhler, P., Ditzer, T., Ong, R. C. & Huth, A. (2003). Simulating the long-term response of tropical wet forests to fragmentation. Ecosystems, 6, 114–128. Laurance, W. F., Camargo, J. L. C., Fearnside, P. M., Lovejoy, T. E., Williamson, G. B., Mesquita, R. C. G., Meyer, C. F. J., Bobrowiec, P. E. D. & Laurance, S. G. W. (2018). An Amazonian rainforest and its fragments as a laboratory of global change. Biological Reviews, 93, 223–247. Laurance, W. F., Ferreira, L. V., Rankin-de Merona, J. M. & Laurance, S. G. (1998). Rain forest fragmentation and the dynamics of Amazonian tree communities. Ecology, 79, 2032–2040. Lewis, S. L., Edwards, D. P. & Galbraith, D. (2015). Increasing human dominance of tropical forests. Science, 349, 827–832. Maréchaux, I. & Chave, J. (2017). An individual-based forest model to jointly simulate carbon and tree diversity in Amazonia: Description and applications. Ecological Monographs, 87, 632–664. Maréchaux, I., Loubet, P., Courbaud, B., Chave, J., Davi, H., Fournier, R. A., et al. (2021). Tackling unresolved questions in forest ecology: The past and future role of simulation models. Ecology and Evolution, 11, 3746–3770. Maynard, D. S., Bialic-Murphy, L., Zohner, C. M., Crowther, T. W., Ferlian, O., Aubrey, D. P., et al. (2022). Global relationships in tree functional traits. Nature Communications, 13, 3185. Moles, A. T. (2018). Being John Harper: Using evolutionary ideas to improve understanding of global patterns in plant traits. Journal of Ecology 106, 1–18. Nunes, M. H., Berenguer, E., Lennox, G. D., Ferreira, J., Gardner, T. A., Barlow, J., et al. (2023). Edge effects on tree architecture exacerbate biomass loss of fragmented Amazonian forests. Nature Communications, 14, 8129. Pinho, B. X., Santos, B. A., Tabarelli, M., Lopes, A. V., Lopes, S. F., Melo, F. P. L., et al. (2021). Functional biogeography of Neotropical moist forests: Trait–climate relationships and assembly patterns of tree communities. Global Ecology and Biogeography, 30, 1430–1446. Pinho, B. X., Santos, B. A., Tabarelli, M., Lopes, A. V., Lopes, S. F., Melo, F. P. L., et al. (2025). Winner–loser plant trait replacements in human-modified tropical forests. Nature Ecology & Evolution, 1–15. doi:10.1038/s41559-024-02592-5. R Core Team (2023). R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available at: https://www.R-project.org/ Rau, E.-P., Fischer, F. J., Maréchaux, I., Chave, J. & Salzet, G. (2022). Wind speed controls forest structure in a subtropical forest exposed to cyclones: A case study using an individual-based model. Frontiers in Forests and Global Change, 5. Riva, F. & Fahrig, L. (2022). The disproportionately high value of small patches for biodiversity conservation. Conservation Letters, 15(3), e12881. Riva, F. & Fahrig, L. (2023). Landscape‐scale habitat fragmentation is positively related to biodiversity, despite patch‐scale ecosystem decay. Ecology Letters, 26(2), 268–277. Rybicki, J., Abrego, N. & Ovaskainen, O. (2020). Habitat fragmentation and species diversity in competitive communities. Ecology Letters, 23, 506–517. Sakschewski, B., von Bloh, W., Boit, A., Poorter, L., Peña-Claros, M., Heinke, J., Joshi, J. & Thonicke, K. (2016). Resilience of Amazon forests emerges from plant trait diversity. Nature Climate Change, 6(11), 1032–1036. Santos, B. A., Peres, C. A., Oliveira, M. A., Grillo, A., Alves-Costa, C. P. & Tabarelli, M. (2008). Drastic erosion in functional attributes of tree assemblages in Atlantic forest fragments of northeastern Brazil. Biological Conservation, 141(1), 249–260. Schmitt, S., Salzet, G., Fischer, F. J., Maréchaux, I. & Chave, J. (2023). rcontroll: An R interface for the individual‐based forest dynamics simulator troll. Methods in Ecology and Evolution, 14, e14215. Schmitt, S., Maréchaux, I., Chave, J., Fischer, F. J., Salzet, G., & Hérault, B. (2020). Functional diversity improves tropical forest resilience: Insights from a long‐term virtual experiment. Journal of Ecology, 108, 831–843. Shugart, H. H., Asner, G. P., Fischer, R., Huth, A., Knapp, N., Le Toan, T. & Shuman, J. K. (2015). Computer and remote‐sensing infrastructure to enhance large‐scale testing of individual‐based forest models. Frontiers in Ecology and the Environment, 13(9), 503–511. Tabarelli, M., Peres, C. A. & Melo, F. P. L. (2012). The ‘few winners and many losers’ paradigm revisited: Emerging prospects for tropical forest biodiversity. Biological Conservation, 155, 136–140. Tamme, R., Götzenberger, L., Zobel, M., Bullock, J. M., Hooftman, D. A. P., Kaasik, A. & Pärtel, M. (2014). Predicting species’ maximum dispersal distances from simple plant traits. Ecology, 95, 505–513. Taubert, F. et al. (2018). Global patterns of tropical forest fragmentation. Nature 554, 519–522. ter Steege, H. et al. (2025). Functional composition of the Amazonian tree flora and forests. Communications Biology 8, 1–17. Villard, M.-A. & Metzger, J. P. (2014). Beyond the fragmentation debate: A conceptual model to predict when habitat configuration really matters. Journal of Applied Ecology, 51, 309–318. Visser, M. D., Bruijning, M., Wright, S. J., Muller-Landau, H. C., Jongejans, E., Comita, L. S. & de Kroon, H. (2016). Functional traits as predictors of vital rates across the life cycle of tropical trees. Functional Ecology, 30, 168–180. Watling, J. I., Arroyo-Rodríguez, V., Pfeifer, M., Baeten, L., Banks-Leite, C., Cisneros, L. M., et al. (2020). Support for the habitat amount hypothesis from a global synthesis of species density studies. Ecology Letters, 23, 674–681. Zhang, H., Chase, J. M. & Liao, J. (2024). Habitat amount modulates biodiversity responses to fragmentation. Nature Ecology & Evolution, 8, 1437–1447. Information & Authors Information Version history V1 Version 1 29 August 2025 V2 Version 2 06 October 2025 Copyright This work is licensed under a Non Exclusive No Reuse License. Keywords ecosystem decay forest loss functional traits habitat fragmentation individual-based model land-use change spatially explicit trait-based modeling troll Authors Affiliations Bruno X. Pinho 0000-0002-6588-3575 [email protected] Institute of Plant Sciences, University of Bern, Bern, Switzerland View all articles by this author Lourens Poorter Forest Ecology and Forest Management Group, Wageningen University, Wageningen, Netherlands View all articles by this author Dimitri Justeau-Allaire AMAP, Univ Montpellier, INRAE, CIRAD, CNRS, IRD, Montpellier, France View all articles by this author Fabian Fischer School of Biological Sciences, University of Bristol, Bristol, UK View all articles by this author Jerome Chave Centre de Recherche sur la Biodiversité et l’Environnement, CNRS, INPT, IRD, Université de Toulouse, Toulouse, France View all articles by this author Isabelle Maréchaux 0000-0002-5401-0197 AMAP, Univ Montpellier, INRAE, CIRAD, CNRS, IRD, Montpellier, France View all articles by this author Funding Information Agence Nationale de la Recherche ANR-16-IDEX-0006 and ANR-10-LABX-25-01 Isabelle Maréchaux Metrics & Citations Metrics Article Usage 466 views 233 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Bruno X. Pinho, Lourens Poorter, Dimitri Justeau-Allaire, et al. Fragmented tropical forests can maintain tree diversity and function, but not under high deforestation. Authorea . 06 October 2025. DOI: https://doi.org/10.22541/au.175647864.44539581/v2 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.175647864.44539581/v2","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9feff6bbaa8e300f',t:'MTc3OTMyODU3Nw=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();