Forest quality mitigates extinction risk in humid tropical vertebrates


 Reducing deforestation underpins efforts to conserve global biodiversity. However, this focus on retaining forest cover overlooks the multitude of anthropogenic pressures that can degrade forest quality in ways that may imperil biodiversity. Here we use the latest remotely-sensed measures of forest structural condition and associated human pressures across the global humid tropics to provide the first estimates of the importance of forest quality, relative to forest cover, in mitigating extinction risk for rainforest vertebrates worldwide. We found tropical rainforests of intact structural condition and minimal human pressures played an outsized role in reducing the odds of species being threatened or having a declining population. Further, the effects of forest quality in mitigating extinction risk were stronger when small amounts of high quality forest remained within species geographic ranges, as opposed to when large extents were forested but of low quality. Our research underscores a critical need to focus global environmental policy and conservation strategies toward the targeted protection of the last remaining undisturbed forest landscapes, in concert with strategies aimed at preserving, restoring and reconnecting remnant forest fragments across the hyperdiverse humid tropics.

the 2030 Agenda for Sustainable Development 3 and the United Nations Framework Convention 24 on Climate Change 4 typically mandate the maintenance and restoration of forest cover, without 25 prescribing clear targets to achieve the preservation of native forest quality. 26 Tropical rainforests, the most biodiverse terrestrial ecosystems on Earth 15 , are currently 27 undergoing an accelerated rate of conversion and degradation [16][17][18] . Less than half of the global 28 tropical rainforest estate remains in its native state characterized by tall, closed-canopy stands 29 free from deleterious human activities 6,7 . The steady degradation and loss of these 'best of the 30 last' remaining rainforests foreshadows a disproportionately high rate of imminent extinctions, 31 given their hyperdiversity 7,15,19 . Yet, there is a lack of direct evidence on whether native tropical 32 rainforests of intact structural condition and minimal human pressures are associated with 33 reduced species extinction risk. Such evidence on the potential for undisturbed native rainforests 34 to mitigate species extinction risk is critical for supporting the inclusion of forest quality targets Condition Index (SCI), a globally consistent measure of forest structure, enables identification of 42 taller, older, more structurally complex, closed-canopy rainforests (hereafter "structurally intact 43 forests") 6 . Structurally intact forests may deteriorate in quality with anthropogenic pressures such 44 as settlements, roads, fire, selective logging and hunting, and the adverse impacts of such 45 pressures on biodiversity may surpass those of deforestation alone 5 . To capture such pressures, 46 the Forest Structural Integrity Index (FSII) 6 combines the SCI with the Human Footprint (HFP) 21 47 to distinguish rainforests of intact structural condition and minimal human modification 48 (hereafter "high integrity forests"). 49 Here, we present the first assessment of the global importance of native high integrity forests 50 in mitigating species extinction risk, compared with structurally intact forests and forest cover 51 alone (i.e., without consideration of either structural condition or integrity). We use two IUCN 52 Red List of Threatened Species 22 measures of extinction risk: (1) threatened status and (2) 53 declining population for 16,396 mammal, bird, reptile, and amphibian species whose geographic 54 ranges overlap the tropical rainforest biome 23 . We classified species as either rainforest endemic 55 or non-endemic on the basis of extent of range overlap with the tropical rainforest biome and 56 association with rainforest habitats 15 , expecting the potential effects of forest quality in 57 mitigating extinction risk would be stronger for endemic or rainforest dependent species than for 58 non-endemics. Within species ranges, we used the SCI and FSII datasets to calculate the area 59 (km 2 ) of structurally intact and high integrity forests, relative to the area of structurally degraded 60 and low integrity forests. We also pooled all SCI values representing forest to calculate the total 61 area of forest cover within species ranges, relative to non-forest area. We then used a generalized 62 linear modeling framework that accounts for the phylogenetic non-independence of species 24 to 63 test whether greater area of high integrity forests within species ranges is linked to a reduced 64 odds of species: (i) being threatened, and (ii) having a declining population, relative to greater 6 (e.g., endemic and non-endemic mammals, reptiles and amphibians being threatened, Fig. 1) than 93 others (e.g., endemic birds having a declining population). Inconsistent with our expectations, 94 the strength of the effects of high integrity forests in mitigating extinction risk was largely 95 similar for endemic and non-endemic vertebrates (95% CIs overlapped each other, Fig. 1). 96 We found robust support for high integrity forest fragments playing a more important role in 97 mitigating species extinction risk than larger forested extents of low integrity. Endemic as well 98 as non-endemic vertebrates had a lower probability of being threatened and having a declining 99 population when small amounts of high integrity forest remained within species humid tropical 100 ranges, as opposed to when large extents were forested but of low integrity (Fig. 2). Evidence for 101 this finding is the strong positive statistical interaction (95% CIs did not overlap zero and FDR-102 adjusted p < 0.05) between forest cover and integrity in 10 out of 16 models testing for such 103 interactions on both response variables for each taxonomic group. A further four interactions 104 tended to be positive albeit statistically non-significant (95% CIs overlapped zero and FDR-105 adjusted p > 0.05; Supplementary Table 3). Across all endemic and non-endemic vertebrate 106 groups, we also found species had a lower probability of being threatened and having declining 107 populations when small amounts of high integrity forest remained within species humid tropical 108 ranges, compared with when larger extents remained structurally intact but of low integrity (Fig. 109 3). Support for this finding lies in the strong positive statistical interaction between forest 110 condition and integrity in nine out of 16 models testing for such interactions on both response 111 variables for each taxonomic group. A further six interactions tended to be positive albeit 112 statistically non-significant (Supplementary Table 4). These patterns were consistent for the 113 interactions between forest cover and condition on both response variables (Extended Data Fig.   114 1, Supplementary Table 5). However, inconsistent with these widespread trends, endemic birds, 7 reptiles and amphibians had a lower probability of having declining populations only when large 116 extents of high quality forest remained within species ranges (Figs. 2-3, Extended Data Fig. 1), 117 as supported by the negative interactions between forest cover, integrity and condition on 118 declining population probability in six out of 48 models (Supplementary Tables 3-5). 119 Reducing deforestation is a central pillar of global biodiversity conservation efforts 1-4 . Yet, 120 this attention on maintaining forest cover alone ignores the many human pressures that can 121 degrade the quality of forest cover in ways severely detrimental to biodiversity. Leveraging the 122 latest advances in remote sensing, we provide the first estimates of the importance of forest 123 quality, relative to forest cover, in mitigating extinction risk for humid tropical vertebrates 124 worldwide. Our analyses reveal the last remaining high integrity forests play a significant role in 125 reducing species extinction risk for all vertebrate groups, and serve as critical habitats not only 126 for species that occur exclusively in these ecosystems, but also for species that use them as 127 refugia or on a seasonal basis (e.g., wintering migratory birds 15 ). Forest cover is known to have a 128 positive effect on biodiversity, relative to human land-uses such as agriculture and 129 development 19 . However, when compared with high integrity forests for the first time in this 130 study, forest cover was linked to a higher likelihood of species extinction risk, reflecting how 131 structural degradation and human pressures within forest cover alone can adversely affect 132 biodiversity. Furthermore, structurally intact forests tended to be associated with higher odds of 133 species extinction risk than high integrity forests, suggesting structural intactness alone can be 134 insufficient to prevent biodiversity loss without also limiting human pressures within intact 135 forests.

136
Large, well connected forest landscapes are known to be essential for biodiversity 137 conservation, especially in an era of climate change 13,14,19 . However, our research shows the 138 effects of forest quality in mitigating extinction risk for humid tropical vertebrates were 139 amplified when small amounts of high integrity forest remained within species ranges, compared 140 with when large extents were forested or remained structurally intact but were of low integrity.

141
Our findings add to the growing evidence that high integrity forest fragments can play a vital 142 supporting role in limiting biodiversity loss by providing refugia or habitat for numerous species, 143 and are thus worthy of inclusion in conservation planning 7,27,28 . Nevertheless, small forest 144 fragments face a higher likelihood of loss than larger forested tracts because of the severe land-145 use pressures around them and improved access for resource extraction 29 . Furthermore, 146 sensitivity to isolation in forest fragments may likely explain the higher probability of declining 147 populations even in high integrity forest fragments for endemic birds, reptiles and amphibians 30 , 148 potentially signaling the presence of an extinction debt for these vertebrate groups in fragmented 149 landscapes 19 . Therefore, proactively prioritising the protection of high integrity forest fragments 150 from loss, while simultaneously setting targets for restoring degraded forest fragments and re-151 establishing landscape connectivity is of paramount importance for limiting biodiversity loss 27 .

152
The positive role of high integrity forests in mitigating species extinction risk remained  Tables 8-10). 163 These findings suggest high integrity forests may be particularly important for species that are 164 threatened because of restricted range area and forest fragmentation within these small ranges.

165
Human influence on the terrestrial biosphere is not limited to tropical rainforests but extends 166 over much of Earth's land surface 21 . Therefore, future research needs to take advantage of global 167 forest integrity 31 and ecosystem intactness 32 datasets to comprehensively quantify the importance 168 of high integrity forest as well as non-forest ecosystems for biodiversity in all of the world's 169 terrestrial biomes. Large-scale environmental perturbations such as climate change can interact 170 with the many human pressures impacting forest systems and their biodiversity 33 . However, 171 native high integrity forests are known to be more resilient to climate stressors than degraded 172 forests. Large, contiguous, high integrity forest landscapes also provide connectivity across 173 wide-ranging environmental gradients that can facilitate adaptive responses of species such as 174 disperal to track shifting climate 13 . Consequently, the importance of the last remaining high 175 integrity tropical rainforests for biodiversity conservation is likely to increase over time, given 176 forests that are already degraded will likely experience intensifying pressures exacerbated by 177 climate change 34 .

178
Tropical rainforests harbor the overwhelming majority of the world's terrestrial 179 biodiversity 15 . However, these hyperdiverse ecosystems are also under overwhelming human 180 pressures worldwide [16][17][18] , such that the accelerating trends in their loss and degradation predict a 181 highly diminished and fragmented rainforest estate over the next few decades, depauperate in 182 much of the biodiversity extant today and with limited ecosystem services for humanity 35 . We 183 provide robust evidence of the global significance of native high integrity forests in mitigating 184 species extinction risk, emphasizing the necessity to ensure conservation strategies aim to 185 preserve and restore forest quality, as opposed to maintaining forest cover alone. A unique 186 opportunity to advance biodiversity conservation is at hand, given 86% of the last remaining 187 high integrity tropical rainforests remain unprotected 7 . Focusing conservation efforts on these 188 imperiled ecosystems through environmental policies and management actions geared at their 189 preservation will advance biodiversity conservation outcomes, particularly but not exclusively 190 for threatened and restricted range species 7 .

191
Our findings demonstrate a clear and urgent need for the targeted preservation of the last 192 remaining high integrity forest landscapes in tandem with strategies aimed at protecting, 193 restoring and reconnecting remnant forest fragments across the global humid tropics 7,27,28,36,37 . 194 On the basis of the evidence presented here, we argue the single most important policy action 195 nations can take to prevent catastrophic biodiversity loss in tropical rainforests is to commit to a 196 global target of "no net loss in area and integrity" of these endangered ecosystems 38 . Such 197 aggressive forest quality retention targets are urgently needed to 'bend the curve' on species loss 198 in the Anthropocene 9,39 , and ensure the CBD's post-2020 Global Biodiversity Framework stands 199 a realistic chance "to put biodiversity on a path to recovery for the benefit of the planet and its 200 people" by 2030 20 .

202
No statistical methods were used to predetermine sample size. The experiments were not 203 randomized and investigators were not blinded to allocation during experiments and outcome 204 assessment.

205
Geographic range maps 206 We conducted our analyses across the tropical and subtropical moist broadleaf forest biome, 207 which encompasses the present-day distribution of tropical rainforests around the Equator and 208 between the Tropics of Cancer and Capricorn 23 . Despite covering a mere 14% of Earth's 209 terrestrial area 23 , these forests are home to over half of the world's vertebrate species 15 , such that 210 the continued loss and degradation of these imperilled ecosystems is likely to result in a 211 disproportionately high number of extinctions. We followed the protocols in Pillay et al. (2021, 212 In press) 15 to obtain the latest established geographic range maps for all species of mammals 22 , 213 birds 40 , reptiles 41 and amphibians 22,42 . The original datasets contained range maps for 5,566 214 mammals, 11,125 birds, 10,064 reptiles and 6,684 amphibians, and include ranges for species 215 that are extinct as well as polygons based on uncertain data. 216 We filtered all geographic range map datasets with three successive IUCN Red List of 217 Threatened Species spatial attributes to remove extinct species and records based on uncertain 218 data. First, we retained only species known to be "Extant", while discarding polygons 219 representing parts of a species range where it was reported to be "Possibly extant", "Possibly 220 extinct", "Extinct" and "Presence uncertain". Second, we filtered this list of extant species to 221 retain only those that are "Native" and "Reintroduced", while discarding polygons representing 222 parts of a species range where it was reported to be "Introduced", "Vagrant", "Origin uncertain" 12 retain only "Resident" and "Non-breeding" parts of the range for mammals (the only ones 225 remaining for mammals after the first two filters above). For birds, we retained "Resident", 226 "Breeding", "Non-breeding" and "Passage" parts of the range, while discarding "Seasonal 227 occurrence uncertain". For amphibians, we retained "Resident" parts of the range, which was the 228 only one remaining after the first two filters above. The final list of amphibians from the IUCN

229
Red List after this third filter included 6,607 species. However, this list of amphibians from the 230 IUCN do not comprise all known species. Therefore, we included range maps for 659 additional our analyses showed that 10 species from this list are now regarded as extinct. Therefore, we 235 discarded these 10 species. After performing these filters, our list of species for subsequent 236 analyses included 5,529 mammals, 10,935 birds, 10,054 reptiles and 7,264 amphibians, for a 237 total of 33,782 species of extant terrestrial vertebrates worldwide. 238 We projected all geographic range maps to the World Mollweide projection prior to analyses, 239 and used Python code implemented with the ArcPy module in ArcGIS Pro 2.5.0 to perform a 240 union of the range map of each species with the map of the tropical rainforest biome. This 241 procedure allowed us to distinguish parts of the global range of species that overlap the tropical 242 rainforest biome, should there be such overlap for a given species. Thereafter, we used species-243 level attributes from the IUCN Red List of Threatened Species to obtain data on the major 244 habitats in which each species occurs to limit some forms of commission or false positive errors 245 that may occur with range maps. Specifically, these errors include species whose ranges may 246 overlap with the tropical rainforest biome but do not actually use the forests within that biome 15 .

247
For species having range overlap with the tropical rainforest biome, we retained only species 248 reported to occur in the tropical rainforest habitat types listed in the IUCN Habitats Classification 249 Scheme 43 . We merged this list of species reported to occur in tropical rainforest habitats with the 250 list of species whose ranges overlap the tropical rainforest biome to retain 3,327 mammals, 7,704 251 birds, 3,828 reptiles and 5,298 amphibians, for a total of 20,157 species with both range overlap 252 and habitat association with tropical rainforests 15 . We note that we discarded additional species 253 from this dataset on the basis of matching species names with those in the respective 254 phylogenetic trees (for the final list of species in this study, see Statistical analyses).

255
Definiton of tropical rainforest endemic species 256 We defined endemism to tropical rainforests on the basis of the criteria established by Pillay et 257 al. (2021, In press) 15 . We considered a species to be endemic if (1) 80-100% of its global range 258 overlapped with the tropical rainforest biome, and (2) it was near-exclusively reported from the 259 tropical rainforest habitat types listed in the IUCN Habitats Classification Scheme 43 . We did not 260 exclude wetlands, rocky and cave habitats from this second criterion, making the reasonable 261 assumption that for species with > 80% range overlap with the tropical rainforest biome and 262 nearly exclusively associated with rainforest habitats, these three other habitat types are likely to 263 be within tropical rainforests (e.g. bats that roost in caves within rainforest habitats).

264
Tropical rainforest structural condition and integrity indices and forest cover 265 We used two indices of tropical rainforest quality in our analyses -the Structural Condition  Hansen et al. 2020 7 , we pooled and categorized the area of SCI values ranging from 2 to 5 (> 293 25% canopy cover and > 5 m canopy height) as low SCI or structurally degraded forest, and the 294 area of SCI values ranging from 14 to 18 (> 75% canopy cover and > 15 m canopy height) as 295 high SCI or structurally intact forest. We note some secondary and selectively logged forests 296 have structural attributes similar to this high SCI class. When validating the SCI dataset, it was 297 observed ~20% of older secondary forests were within the high SCI class 6 . Older secondary 298 forests may not have all the structural intactness characteristics associated with forests that have 299 never undergone anthropogenic degradation. However, current remote sensing capabilities do not 300 allow discriminating these older secondary forests from unlogged native forests. Overall, the 301 high SCI forests in our data are largely representative of structurally intact native forests typical 302 of the humid tropics 7 . We followed a similar procedure to pool and categorize the area of FSII 303 values ranging from 1 to 5 as low FSII or low integrity forest and the area of FSII values ranging 304 from 14 to 18 as high FSII or high integrity forest. These high integrity forests represent 305 rainforests of not only intact structural condition but also low human pressures, specifically HFP 306 values ≤ 4 7 .

307
Predictor variables 308 We calculated the relative difference between the area under high and low SCI forest within the 309 humid tropical range of a species as: . Similarly, we calculated the 310 relative difference between the area under high and low FSII forest within the humid tropical 311 range of a species as: These calculated values range between -1 and + 312 1 and represent the relative percentage difference between the area under high and low SCI and 313 FSII forests within the humid tropical range of a species. Therefore, a value of -1 signifies 100% of the humid tropical range of a species is encompassed by low SCI or low FSII forest, whereas a 315 value of + 1 means 100% of the humid tropical range of a species is covered in high SCI or high 316 FSII forest. 317 We also calculated the relative difference between the area of forest and non-forest cover 318 within the humid tropical range of a species as: . We used the lowest SCI 319 value of 1 to identify stands < 5 m tall, disturbed since 2012 or with canopy cover < 25%, which 320 are considered highly disturbed, and categorized the area under this pixel value as non-forest 7 . 321 We pooled and categorized the remaining SCI values from 2 to 18 as forest. Similar to the SCI 322 and FSII relative difference values, these calculated values of forest cover also range between -1 323 (signifying 100% of the humid tropical range of a species consists of non-forest) and + 1 324 (signifying 100% of the humid tropical range of a species is forested). We thereby brought all 325 forest cover, condition and integrity data (the predictor variables in this study) to a consistent 326 scale for further analyses.

328
The response variables in this study are binary -threatened/non-threatened and declining 329 population/not declining in population. To achieve this binary classification, we defined species Increasing and Stable categories as not declining in population, while discarding species in the 335 We used a generalized linear modeling framework, specifically logistic regression, for 337 statistical inference. Our primary units of analyses -species -cannot be considered as 338 independent because of the variable degree of evolutionary relatedness between the species in 339 each taxonomic group. To account for the potential effect of evolutionary dependence, we first 340 obtained phylogenetic trees for mammals 47 , birds 48 , reptiles 49 and amphibians 50 , and matched the 341 species lists from the previous steps to discard species not in the respective phylogenetic trees.

342
Our list of species after this step comprised 3,217 mammals, 6,674 birds, 3,735 reptiles and  Table 1). For each taxonomic group, we 347 partitioned species into rainforest endemic and non-endemic groups. Next, we randomly sampled 348 100 trees out of 10,000 available full phylogenetic trees for each taxonomic group, as  We parameterized identical models for endemic and non-endemic species in each taxonomic 354 group to test whether greater area of high integrity forests within species ranges is linked to a 355 reduced odds of species: (i) being threatened, and (ii) having a declining population, relative to 356 greater area of structurally intact forests and forest cover alone. Prior to analyses, we 357 standardized each predictor variable (forest cover, condition and integrity) to have a mean of 0 358 and a standard deviation of 1 (z-transformation). We tested for the effects of the three predictor variables on the respective response variable (threatened status or declining population) by 360 parameterizing them as additive effects in multiple phylogenetic logistic regression models, and 361 used the standardized partial coefficient of each predictor variable as a measure of its effect on 362 the response variable 19,25 . In this form of multiple logistic regression, the exponentiated 363 standardized partial coefficient of a given predictor variable represents the odds of a 1-unit 364 increase in that variable on the response, controlling for the effects of the other predictor Firth's correction implemented in the phyloglm function via the parameter "logistic_MPLE" 24,51 . 381 of additive models and two predictor variables for interaction models, which risks inflating type 383 1 error rate. Therefore, we used a FDR procedure (graphically sharpened method 53 ) which 384 corrects for multiple comparisons in comparative extinction risk modeling. We calculated FDR-   Code availability 396 Custom Python and R code used for geospatial and statistical analyses will be uploaded to 397 GitHub/Zenodo upon acceptance.    Supplementary information is available for this paper and is appended below for peer-review.
525 Figure 1. Effects of forest cover, structural condition and integrity on the threatened status and declining population trend of tropical rainforest mammals, birds, reptiles and amphibians. Point estimates represent median standardized odds of species being threatened (circles) or having a declining population (squares) generated by exponentiating standardized coefficients (log odds) of 100 phylogenetic logistic regressions to obtain standardized odds ratios, and thereafter converting to percentage odds to aid interpretation. Each regression was performed with one phylogenetic tree randomly drawn from 10,000 available trees for each taxonomic group, and separate models were parameterized for rainforest endemic and non-endemic species for each response variable. Error bars represent median 95% confidence intervals generated with 2,000 parametric bootstraps in each regression. See Supplementary Tables 1-2 for sample sizes and model results, respectively.

Lower extinction risk
Higher extinction risk Figure 2. Predicted probabilities of mammals, birds, reptiles and amphibians being threatened or having declining populations as a function of forest cover area and the area of forests of varying integrity within species humid tropical ranges. Median predicted probabilities were generated from 100 phylogenetic logistic regressions. See Supplementary Tables 1 and 3 for sample  Extended Data Figure 2. Effects of forest cover, structural condition and integrity on the threatened status and declining population trend of tropical rainforest mammals, birds, reptiles and amphibians after excluding 2,751 and 2,155 species in IUCN criterion B for threatened status and declining population response variables, respectively. Point estimates represent median standardized odds of species being threatened (circles) or having a declining population (squares) generated by exponentiating standardized coefficients (log odds) of 100 phylogenetic logistic regressions to obtain standardized odds ratios, and thereafter converting to percentage odds to aid interpretation. Each regression was performed with one phylogenetic tree randomly drawn from 10,000 available trees for each taxonomic group, and separate models were parameterized for rainforest endemic and non-endemic species for each response variable. Error bars represent median 95% confidence intervals generated with 2,000 parametric bootstraps in each regression. See Supplementary Tables 6-7 for sample sizes and model results, respectively.

Lower extinction risk
Higher extinction risk Extended Data Figure 3. Predicted probabilities of mammals, birds, reptiles and amphibians being threatened or having declining populations as a function of forest cover area and the area of forests of varying integrity within species humid tropical ranges. This analysis was performed after excluding 2,751 and 2,155 species in IUCN criterion B for threatened status and declining population response variables, respectively. Median predicted probabilities were generated from 100 phylogenetic logistic regressions. See Supplementary Tables 6 and 8 for samples sizes and model results, respectively. Extended Data Figure 5. Predicted probabilities of mammals, birds, reptiles and amphibians being threatened and having declining populations as a function of forest cover area and the area of forests of varying condition within species humid tropical ranges. This analysis was performed after excluding 2,751 and 2,155 species in IUCN criterion B for threatened status and declining population response variables, respectively. Median predicted probabilities were generated from 100 phylogenetic logistic regressions. See Supplementary Tables 6 and 10 for sample sizes and model results, respectively. Supplementary Table 2. Results of multiple phylogenetic logistic regression models contrasting the effects of forest cover, structural condition and integrity on the threatened status and declining population trend of humid tropical (a) mammals, (b) birds, (c) reptiles and (d) amphibians worldwide. Estimates represent median standardized coefficients (log odds) from 100 phylogenetic logistic regressions. The 95% confidence intervals of estimated coefficients were generated with 2,000 parametric bootstraps in each regression. Adj. p-value represents the False Discovery Rate (FDR)-adjusted p-value. We fit identical models separately for endemic and non-endemic species within each taxonomic group. Standardized coefficients generated by the models were exponentiated to obtain standardized odds ratios, and further converted to percentage odds using the formula [e (b) -1] ´ 100 (Fig. 1).  Supplementary Table 3. Results of phylogenetic logistic regression models testing for interactions between forest cover and integrity on the threatened status and declining population trend of humid tropical (a) mammals, (b) birds, (c) reptiles and (d) amphibians worldwide. Estimates represent median standardized beta coefficients (log odds) from 100 phylogenetic logistic regressions. The 95% confidence intervals of estimated coefficients were generated with 2,000 parametric bootstraps in each regression. Adj. p-value represents the False Discovery Rate (FDR)-adjusted p-value. We fit identical models separately for endemic and non-endemic species within each taxonomic group. See Fig. 2 for predicted probabilities generated from these results.

Supplementary Table 3 (continued).
Notes -A positive coefficient for the interaction term would suggest that the effect of forest integrity on the response variable is stronger when small amounts of high integrity (i.e., structurally intact and low pressure) forest remain within species humid tropical ranges, as opposed to when large extents are forested but of low integrity. In contrast, a negative coefficient for the interaction term would indicate that the effect of forest integrity on the response variable is stronger when large extents of high integrity forest remain within species ranges, as opposed to small fragments, irrespective of integrity.  Supplementary Table 4. Results of phylogenetic logistic regression models testing for interactions between rainforest structural condition and integrity on the threatened status and declining population trend of humid tropical (a) mammals, (b) birds, (c) reptiles and (d) amphibians worldwide. Estimates represent median standardized coefficients (log odds) from 100 phylogenetic logistic regressions. 95% confidence intervals of estimated coefficients were generated with 2,000 parametric bootstraps in each regression. Adj. p-value represents the False Discovery Rate (FDR)-adjusted p-value. We fit identical models separately for endemic and non-endemic species within each taxonomic group. See Fig. 3 for predicted probabilities generated from these results.

Supplementary Table 4 (continued).
Notes -A positive coefficient for the interaction term would suggest that the effect of forest integrity on the response variable is stronger when small amounts of high integrity (i.e., structurally intact and low pressure) forest remain within species humid tropical ranges, as opposed to when large extents are structurally intact but of low integrity (i.e. high human pressure). In contrast, a negative coefficient for the interaction term would indicate that the effect of forest condition on the response variable is stronger when large extents of high integrity forest remain within species humid tropical ranges, as opposed to small fragments, irrespective of integrity.  Supplementary Table 5. Results of phylogenetic logistic regression models testing for interactions between forest cover and rainforest structural condition on the threatened status and declining population trend of humid tropical (a) mammals, (b) birds, (c) reptiles and (d) amphibians worldwide. Estimates represent median standardized coefficients (log odds) from 100 phylogenetic logistic regressions. 95% confidence intervals of estimated coefficients were generated with 2,000 parametric bootstraps in each regression. Adj. p-value represents the False Discovery Rate (FDR)-adjusted p-value. We fit identical models separately for endemic and non-endemic species within each taxonomic group. See Extended Data Fig. 1 for predicted probabilities generated from these results. Notes -A positive coefficient for the interaction term would suggest that the effect of forest structural condition on the response variable is stronger when small amounts of structurally intact forest remain within species humid tropical ranges, as opposed to when large extents are forested but structurally degraded. In contrast, a negative coefficient for the interaction term would indicate that the effect of forest structural condition on the response variable is stronger when large extents of structurally intact forest remain within species humid tropical ranges, as opposed to small fragments, irrespective of condition.  Supplementary Table 7. Results of multiple phylogenetic logistic regression models contrasting the effects of forest cover, structural condition and structural integrity on the threatened status and declining population trend of humid tropical (a) mammals, (b) birds, (c) reptiles and (d) amphibians after excluding 2,751 and 2,155 species in IUCN criterion B for threatened status and declining population, respectively. Estimates represent median standardized coefficients (log odds) from 100 phylogenetic logistic regressions. The 95% confidence intervals of estimated coefficients were generated with 2,000 parametric bootstraps in each regression. Adj. p-value represents the False Discovery Rate (FDR)-adjusted p-value. We fit identical models separately for endemic and non-endemic species within each taxonomic group. Standardized coefficients generated by the models were exponentiated to obtain standardized odds ratios, and further converted to percentage odds using the formula [e (b) -1] ´ 100 (Extended Data Fig. 2). Notes -A positive coefficient for the interaction term would suggest that the effect of forest integrity on the response variable is stronger when small amounts of high integrity (i.e., structurally intact and low pressure) forest remain within species humid tropical ranges, as opposed to when large extents are forested but of low integrity. In contrast, a negative coefficient for the interaction term would indicate that the effect of forest integrity on the response variable is stronger when large extents of high integrity forest remain within species ranges, as opposed to small fragments, irrespective of integrity. Notes -A positive coefficient for the interaction term would suggest that the effect of forest integrity on the response variable is stronger when small amounts of high integrity (i.e., structurally intact and low pressure) forest remain within species humid tropical ranges, as opposed to when large extents are structurally intact but of low integrity (i.e. high human pressure). In contrast, a negative coefficient for the interaction term would indicate that the effect of forest condition on the response variable is stronger when large extents of high integrity forest remain within species humid tropical ranges, as opposed to small fragments, irrespective of integrity. Notes -A positive coefficient for the interaction term would suggest that the effect of forest structural condition on the response variable is stronger when small amounts of structurally intact forest remain within species humid tropical ranges, as opposed to when large extents are forested but structurally degraded. In contrast, a negative coefficient for the interaction term would indicate that the effect of forest structural condition on the response variable is stronger when large extents of structurally intact forest remain within species humid tropical ranges, as opposed to small fragments, irrespective of condition.  Supplementary Table 11. Phylogenetic signal parameter α measuring the strength of the phylogenetic correlation. When α = 1, evolution is approximately by Brownian motion on a given phylogeny and α > 1 indicates low phylogenetic correlations among species. α parameter estimates are provided for all additive and interaction models for all species.  Table 12. Phylogenetic signal parameter α measuring the strength of the phylogenetic correlation. When α = 1, evolution is approximately by Brownian motion on a given phylogeny and α > 1 indicates low phylogenetic correlations among species. α parameter estimates are provided for all additive and interaction models after excluding 2,751 and 2,155 species in IUCN criterion B for Threatened status and Declining population, respectively.