Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Letter
  • Published:

Global forest loss disproportionately erodes biodiversity in intact landscapes


Global biodiversity loss is a critical environmental crisis, yet the lack of spatial data on biodiversity threats has hindered conservation strategies1. Theory predicts that abrupt biodiversity declines are most likely to occur when habitat availability is reduced to very low levels in the landscape (10–30%)2,3,4. Alternatively, recent evidence indicates that biodiversity is best conserved by minimizing human intrusion into intact and relatively unfragmented landscapes5. Here we use recently available forest loss data6 to test deforestation effects on International Union for Conservation of Nature Red List categories of extinction risk for 19,432 vertebrate species worldwide. As expected, deforestation substantially increased the odds of a species being listed as threatened, undergoing recent upgrading to a higher threat category and exhibiting declining populations. More importantly, we show that these risks were disproportionately high in relatively intact landscapes; even minimal deforestation has had severe consequences for vertebrate biodiversity. We found little support for the alternative hypothesis that forest loss is most detrimental in already fragmented landscapes. Spatial analysis revealed high-risk hot spots in Borneo, the central Amazon and the Congo Basin. In these regions, our model predicts that 121–219 species will become threatened under current rates of forest loss over the next 30 years. Given that only 17.9% of these high-risk areas are formally protected and only 8.9% have strict protection, new large-scale conservation efforts to protect intact forests7,8 are necessary to slow deforestation rates and to avert a new wave of global extinctions.

This is a preview of subscription content, access via your institution

Access options

Buy this article

Prices may be subject to local taxes which are calculated during checkout

Figure 1: Spatial distribution of the six variables used to predict species’ IUCN Red List response variables.
Figure 2: Effects of four predictors on the status of 19,432 vertebrate species worldwide.
Figure 3: Predicted probabilities of species status as a function of recent forest loss and total forest cover within a species range.
Figure 4: Projected increases in the number of threatened species under three scenarios of future forest loss.

Similar content being viewed by others


  1. Joppa, L. N. et al. Filling in biodiversity threat gaps. Science 352, 416–418 (2016)

    Article  ADS  CAS  PubMed  Google Scholar 

  2. Newbold, T. et al. Has land use pushed terrestrial biodiversity beyond the planetary boundary? A global assessment. Science 353, 288–291 (2016)

    Article  ADS  CAS  PubMed  Google Scholar 

  3. Andrén, H. Effects of habitat fragmentation of birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos 71, 355–366 (1994)

    Article  Google Scholar 

  4. Betts, M. G., Forbes, G. J. & Diamond, A. W. Thresholds in songbird occurrence in relation to landscape structure. Conserv. Biol. 21, 1046–1058 (2007)

    Article  PubMed  Google Scholar 

  5. Barlow, J. et al. Anthropogenic disturbance in tropical forests can double biodiversity loss from deforestation. Nature 535, 144–147 (2016)

    Article  ADS  CAS  PubMed  Google Scholar 

  6. Hansen, M. C. et al. High-resolution global maps of 21st-century forest cover change. Science 342, 850–853 (2013)

    Article  ADS  CAS  PubMed  Google Scholar 

  7. Peres, C. A. Why we need megareserves in Amazonia. Conserv. Biol. 19, 728–733 (2005)

    Article  Google Scholar 

  8. Gibson, L. et al. Primary forests are irreplaceable for sustaining tropical biodiversity. Nature 478, 378–381 (2011)

    Article  ADS  CAS  PubMed  Google Scholar 

  9. Brooks, T. M. et al. Global biodiversity conservation priorities. Science 313, 58–61 (2006)

    Article  ADS  CAS  PubMed  Google Scholar 

  10. Kareiva, P., Watts, S., McDonald, R. & Boucher, T. Domesticated nature: shaping landscapes and ecosystems for human welfare. Science 316, 1866–1869 (2007)

    Article  CAS  PubMed  Google Scholar 

  11. Mendenhall, C. D., Karp, D. S., Meyer, C. F., Hadly, E. A. & Daily, G. C. Predicting biodiversity change and averting collapse in agricultural landscapes. Nature 509, 213–217 (2014)

    Article  ADS  CAS  PubMed  Google Scholar 

  12. Fahrig, L. When does fragmentation of breeding habitat affect population survival? Ecol. Modell. 105, 273–292 (1998)

    Article  Google Scholar 

  13. Phalan, B., Onial, M., Balmford, A. & Green, R. E. Reconciling food production and biodiversity conservation: land sharing and land sparing compared. Science 333, 1289–1291 (2011)

    Article  ADS  CAS  PubMed  Google Scholar 

  14. Wilson, E. O. Half-Earth: Our Planet’s Fight For Life. (W. W. Norton & Company, 2016)

  15. International Union for Conservation of Nature World Congress. Motion 48: Protection of primary forests, including intact forest landscapes. (2016)

  16. Tracewski, Ł. et al. Toward quantification of the impact of 21st-century deforestation on the extinction risk of terrestrial vertebrates. Conserv. Biol. 30, 1070–1079 (2016)

    Article  PubMed  Google Scholar 

  17. International Union for Conservation of Nature. IUCN red list of threatened species. Version 2016.3 (2017)

  18. BirdLife International and NatureServe. Bird Species Distribution Maps of the World Version 5.0 (BirdLife International, 2015)

  19. Kim, D.-H. et al. Global, Landsat-based forest-cover change from 1990 to 2000. Remote Sens. Environ. 155, 178–193 (2014)

    Article  ADS  Google Scholar 

  20. Tilman, D., May, R. M., Lehman, C. L. & Nowak, M. A. Habitat destruction and the extinction debt. Nature 371, 65–66 (1994)

    Article  ADS  Google Scholar 

  21. Wearn, O. R., Reuman, D. C. & Ewers, R. M. Extinction debt and windows of conservation opportunity in the Brazilian Amazon. Science 337, 228–232 (2012)

    Article  ADS  CAS  PubMed  Google Scholar 

  22. Sanderson, E. W. et al. The human footprint and the last of the wild. Bioscience 52, 891–904 (2002)

    Article  Google Scholar 

  23. Hanski, I. Metapopulation dynamics. Nature 396, 41–49 (1998)

    Article  ADS  CAS  Google Scholar 

  24. Brooks, T. M. et al. Habitat loss and extinction in the hotspots of biodiversity. Conserv. Biol. 16, 909–923 (2002)

    Article  Google Scholar 

  25. Balmford, A. Extinction filters and current resilience: the significance of past selection pressures for conservation biology. Trends Ecol. Evol. 11, 193–196 (1996)

    Article  CAS  PubMed  Google Scholar 

  26. Ripple, W. J. et al. Bushmeat hunting and extinction risk to the world’s mammals. R. Soc. Open Sci. 3, 160498 (2016)

    Article  ADS  PubMed  PubMed Central  Google Scholar 

  27. Benítez-López, A. et al. The impact of hunting on tropical mammal and bird populations. Science 356, 180–183 (2017)

    Article  ADS  CAS  PubMed  Google Scholar 

  28. Bellard, C ., Genovesi, P . & Jeschke, J. M. Global patterns in threats to vertebrates by biological invasions. Proc. R. Soc. Lond. B 283, 20152454 (2016)

    Article  Google Scholar 

  29. Clavel, J., Julliar, R. & Devictor, V. Worldwide decline of specialist species: toward a global functional homogenization. Front. Ecol. Environ. 9, 222–228 (2011)

    Article  Google Scholar 

  30. Betts, M. G. et al. A species-centered approach for uncovering generalities in organism responses to habitat loss and fragmentation. Ecography 37, 517–527 (2014)

    Article  Google Scholar 

  31. Potapov, P. et al. The last frontiers of wilderness: tracking loss of intact forest landscapes from 2000 to 2013. Sci. Adv. 3, e1600821 (2017)

    Article  ADS  PubMed  PubMed Central  Google Scholar 

  32. Potapov, P. et al. Mapping the world’s intact forest landscapes by remote sensing. Ecol. Soc. 13, 51 (2008)

    Article  Google Scholar 

  33. Olson, D. M. et al. Terrestrial ecoregions of the world: a new map of life on Earth: a new global map of terrestrial ecoregions provides an innovative tool for conserving biodiversity. Bioscience 51, 933–938 (2001)

    Article  Google Scholar 

  34. IUCN and UNEP-WCMC. The World Database on Protected Areas (WDPA) (2015)

  35. Butchart, S. H. et al. Improvements to the red list index. PLoS ONE 2, e140 (2007)

    Article  ADS  PubMed  PubMed Central  Google Scholar 

  36. Butchart, S. H. et al. Measuring global trends in the status of biodiversity: red list indices for birds. PLoS Biol. 2, e383 (2004)

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  37. Hoffmann, M. et al. The changing fates of the world’s mammals. Phil. Trans. R. Soc. Lond. B 366, 2598–2610 (2011)

    Article  Google Scholar 

  38. BirdLife International. IUCN Red List for birds. (2017)

  39. Böhm, M. et al. The conservation status of the world’s reptiles. Biol. Conserv. 157, 372–385 (2013)

    Article  Google Scholar 

  40. Wildlife Conservation Society and Center for International Earth Science Information Network, Columbia University. Last of the Wild, v2: Global Human Footprint Dataset (Geographic). (2005)

  41. Venter, O. et al. Sixteen years of change in the global terrestrial human footprint and implications for biodiversity conservation. Nat. Commun. 7, 12558 (2016)

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  42. Venter, O. et al. Data from: Global terrestrial Human Footprint maps for 1993 and 2009. (2016)

  43. ESRI. ArcGIS Desktop: Release 10.1 (Environmental Systems Research Institute, 2012)

  44. R Core Team. R: A Language and Environment for Statistical Computing. (2013)

  45. Seligman, M. Rborist: extensible, parallelizable implementation of the random forest algorithm. (2015)

  46. Hurlbert, A. H. & Jetz, W. Species richness, hotspots, and the scale dependence of range maps in ecology and conservation. Proc. Natl Acad. Sci. USA 104, 13384–13389 (2007)

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  47. Augustin, N. H., Mugglestone, M. A. & Buckland, S. T. An autologistic model for the spatial distribution of wildlife. J. Appl. Ecol. 33, 339–347 (1996)

    Article  Google Scholar 

  48. Dormann, C. F. Assessing the validity of autologistic regression. Ecol. Modell. 207, 234–242 (2007)

    Article  Google Scholar 

  49. Bivand, R. et al. spdep: spatial dependence: weighting schemes, statistics and models. (2017)

  50. Pike, N. Using false discovery rates for multiple comparisons in ecology and evolution. Methods Ecol. Evol. 2, 278–282 (2011)

    Article  Google Scholar 

  51. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. B 57, 289–300 (1995)

    MathSciNet  MATH  Google Scholar 

  52. Swets, J. A. Signal Detection Theory and ROC Analysis in Psychology and Diagnostics: Collected Papers (Psychology Press, 2014)

  53. United Nations. United Nations Statistics Division. Standard Country and Area Codes Classifications (M49). (2013)

  54. Bahn, V. & McGill, B. J. Testing the predictive performance of distribution models. Oikos 122, 321–331 (2013)

    Article  Google Scholar 

  55. Gilleland, M. E. verification: weather forecast verification utilities. (2015)

  56. Dray, S., Legendre, P. & Peres-Neto, P. R. Spatial modelling: a comprehensive framework for principal coordinate analysis of neighbour matrices (PCNM). Ecol. Modell. 196, 483–493 (2006)

    Article  Google Scholar 

  57. Lee, D. CARBayes: an R package for Bayesian spatial modelling with conditional autoregressive priors. J. Stat. Softw. 55, 13 (2013)

    Google Scholar 

  58. McMillen, D. McSpatial: nonparametric spatial data analysis. (2013)

  59. Klier, T. & McMillen, D. P. Clustering of auto supplier plants in the United States. J. Bus. Econ. Stat. 26, 460–471 (2012)

    Article  Google Scholar 

  60. Cardillo, M. et al. Multiple causes of high extinction risk in large mammal species. Science 309, 1239–1241 (2005)

    Article  ADS  CAS  PubMed  Google Scholar 

  61. Bates, D. M. lme4: Mixed-effects modeling with R. (2010)

  62. Ives, A. R. & Garland, T. Jr. Phylogenetic logistic regression for binary dependent variables. Syst. Biol. 59, 9–26 (2010)

    Article  PubMed  Google Scholar 

  63. Petersen, R. et al. Mapping Tree Plantations with Multispectral Imagery: Preliminary Results for Seven Tropical Countries. (World Resources Institute, 2016)

  64. MacDicken, K. G. Global Forest Resources Assessment 2015: what, why and how? For. Ecol. Manage. 352, 3–8 (2015)

    Article  Google Scholar 

  65. Giam, X . et al. Reservoirs of richness: least disturbed tropical forests are centres of undescribed species diversity. Proc. R. Soc. Lond. B 279, 67–76 (2012)

    Article  Google Scholar 

  66. Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G. & Jarvis, A. Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol. 25, 1965–1978 (2005)

    Article  Google Scholar 

Download references


Funding from the National Science Foundation (NSF-DEB-1457837) and the College of Forestry IWFL Professorship in Forest Biodiversity Research to M.G.B. supported this research. We are grateful for comments from A. Hadley, U. Kormann, J. Bowman, C. Epps and C. Mendenhall on earlier versions of this manuscript.

Author information

Authors and Affiliations



M.G.B., C.W., S.H.M.B., W.J.R. and T.L. conceived the study, C.W., M.G.B. and T.L. analysed the data, and M.G.B. and C.W. wrote the first draft of the paper with subsequent editorial input from C.W., B.P., S.H.M.B., K.A.M. and A.D.

Corresponding authors

Correspondence to Matthew G. Betts or Christopher Wolf.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Additional information

Reviewer Information Nature thanks J. Barlow, L. Gibson and the other anonymous reviewer(s) for their contribution to the peer review of this work.

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data figures and tables

Extended Data Figure 1 Receiver operating characteristic (ROC) curves for the models predicting status of forest exclusive species.

Class was included as a fixed effect (as in our main results) for the ‘All species’ group. The other results (by class) are based on models fit to each class separately. The left column is based on results where the model was fit to the entire dataset. The right column shows ROC curves for predictions using a fourfold cross-validation scheme where the probability of species being threatened was predicted for each of four regions with the model fit using data from all other regions. P values are based on the Mann–Whitney U statistic and test whether the population AUC is greater than 0.5 (that is, better than random predictions). Results are presented both with (bottom row) and without (top row) the spatial autocovariate.

Extended Data Figure 2 Model results for models fit by class (mammals, amphibians, birds) and for all classes together (All).

Each row shows standardized coefficient estimates and 95% confidence intervals (as error bars) for each single model. All covariates are shown in this figure.

Extended Data Figure 3 Sensitivity analysis results.

The plotted variable is the estimated standardized coefficient for the forest loss × cover term with 95% confidence interval (as error bars). Each column corresponds to a different sensitivity analysis (other covariates are not shown). ai, In general, we found that our primary results were robust to the inclusion of absolute latitude as a predictor variable (a), the restriction of the dataset to tropical species only (b), the exclusion of species listed as threatened based on small geographic range (c), using a 75% pixel-scale threshold for the forest loss and forest cover variables (d), standardizing forest loss and gain by forest cover (that is, dividing forest loss and gain by forest cover so that these variables can be interpreted as approximate percentages of species’ forested range) (e), accounting for potential phylogenetic dependence using generalized linear mixed models with random intercepts by taxonomic order (and by class for the ‘all species’ model) (f), using high-resolution species’ range maps and covariate maps (approximately 5 km), clipping species ranges based on altitudinal limits, and setting forest loss and cover to zero in regions of known tree plantations (g), including forest loss and gain from 1990–2000 by adding 1990–2000 and 2000–2014 forest change variables (h), and the inclusion of year of initial species description as a main effect and in a three-way interaction term with forest loss × cover (i).

Extended Data Figure 4 Estimated standardized coefficients for each model term (with 95% confidence intervals as error bars) when a quadratic forest loss × cover2 interaction (forest loss × cover2) is included in the model.

This allows for the effect of loss to vary quadratically with cover. A significant and positive forest loss × cover2 interaction term would suggest that the (negative) effects of forest loss are greatest in areas with both high and low proportions of forest cover. However, this term was non-significant for most taxa and response variables, indicating that the linear model for the interaction is more parsimonious.

Extended Data Figure 5 The effect of forest loss (for 2% additional loss) in relation to total forest cover using quadratic models.

These models allow the effect of forest loss to vary nonlinearly as a function of forest cover, allowing us to test the hypothesis that forest loss is detrimental to species at both high and low levels of forest cover. However, the quadratic model reveals very similar results to the linear model. The exception is when ‘declining trend’ is used as the response; species’ populations were more likely to be in decline when forest amount is very low (the habitat threshold hypothesis), and upon initial intrusion into intact forests (the initial intrusion hypothesis). For statistical significance of the quadratic models, see confidence intervals in Extended Data Fig. 4, far right panel. For context, the histograms (grey bars) show the (normalized to maximum 100%) distributions of forest cover across species. For example, if one bar in a panel is twice as high as another, then twice as many species have average forest cover of this percentage in their ranges. The black lines show the cumulative percentages of species with at most x per cent forest cover. For example, approximately half of forest-optional species have 50% forest cover or less.

Extended Data Figure 6 Results of multiple spatial models (estimates and 95% confidence intervals as error bars) for forest exclusive species when status (that is, whether or not a species is threatened) is used as the response.

Coefficients across multiple models that account for spatial autocorrelation were very similar. ‘Method’ indicates the procedure (if any) used to account for spatial autocorrelation: non-spatial ordinary GLM (non_spatial), autologistic model with spatial autocovariate (AL_b), autologistic model using 50 nearest neighbours in the spatial weights matrix (AL_b_50), Moran eigenvector filtering (filtering), spatial autoregressive model (SAR_approx), or Bayesian condition autoregressive model (CAR_Bayes). Details on each method are given in the sensitivity analyses section of the Methods.

Extended Data Figure 7 Relationship between forest loss 1990–2000 (from ref. 34) and 2000–2014 (from ref. 7).

Overall, rates of forest loss are temporally autocorrelated; species ranges with high forest loss in the 1990s also show high forest loss in 2000s. However, this relationship is strongly affected by data availability; approximately 12.1% of forest loss data are missing across the globe and as we expected, the more data missing from a species range, the weaker the relationship between 1990s and 2000s rates of forest loss. The plots show correlations (in red; top right of each panel) between forest loss across the two time periods for various levels of missing data. Each point corresponds to a single species and the x and y axis values indicate average values of each variable across its range. Panel titles show the proportion of missing 1990–2000 forest loss data in species ranges. For example, the top left panel contains results for species with between 0% and 4% of their ranges missing 1990 forest data (owing to clouds, lack of satellite coverage, and so on). The correlation between 1990–2000 and 2000–2014 forest loss is highest for species with the least missing data.

Extended Data Figure 8 Country-level forest net loss (that is, change in percentage forest cover) for the 1990–2000 and 2000–2015 periods according to the Food and Agriculture Organization’s (FAO) Global Forest Resources Assessment.

Based on these data, the correlation between 1990–2000 and 2000–2015 forest loss is 0.705. Weighting by country area increases the correlation to 0.805. The relatively high correlation suggests that the spatially explicit recent (2000–2014) forest loss data that we used is closely related to less recent (1990–2000) forest loss.

Extended Data Figure 9 Sensitivity of our results to alternative categories of threat.

In the main text we considered a species to be ‘threatened’ if it fell into the IUCN Red List category Vulnerable, Endangered or Critically Endangered. We conducted further analysis considering as threatened only species that are Endangered and Critically Endangered, and again for only species that are Critically Endangered. Dots show estimated standardized coefficients for each model term (with 95% confidence intervals as error bars) for all main effects and the forest loss × cover interaction term. Our overall conclusions were consistent across these different definitions of threat.

Extended Data Figure 10 Maps showing the methods used to quantify historical forest loss.

First, we used random forests (a machine-learning method) to estimate potential forest cover globally (within forest biomes33). ac, This model was fit using current forest cover within intact forest landscapes36 and bioclimatic and other predictor variables66 (a; see Methods). We then subtracted current forest cover (b; Hansen et al.6) from this map to obtain estimated historical forest loss (c). The map of land is taken from

Supplementary information

Supplementary Information

This file contains Supplementary Tables 1-5.

PowerPoint slides

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Betts, M., Wolf, C., Ripple, W. et al. Global forest loss disproportionately erodes biodiversity in intact landscapes. Nature 547, 441–444 (2017).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:

This article is cited by


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Quick links

Nature Briefing Anthropocene

Sign up for the Nature Briefing: Anthropocene newsletter — what matters in anthropocene research, free to your inbox weekly.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing: Anthropocene