Probabilistic projections of human modification of the terrestrial biosphere from 2020 to 2040
Abstract
Conservation actions are more effective when they anticipate future threats rather than respond only after loss occurs. Most spatial forecasts focus only on land-cover change and fail to capture the broader set of pressures that degrade biodiversity. To close this gap, we develop global probabilistic forecasts of cumulative human modification (HM), a continuous measure that combines 16 anthropogenic threats to biodiversity classified under the IUCN threat taxonomy. Our approach uses a convolutional long short-term memory network trained on the historical HM time series from 1990 to 2020 to generate forecasts at 1 km resolution and 5-year intervals from 2020 to 2040. The model ingests sequences of historical HM maps along with socioeconomic, biophysical, and protected-area covariates to generate forecasts of the full distribution of potential change. Model outputs outperformed a persistence baseline, which assumes that HM remains unchanged from the last observation. Conditional on the continuation of historical relationships and dynamics, the forecasts indicate that between 2020 and 2040, more than 4.5 million km² of land classified as natural (HM < 0.1) in 2020 could be lost, an area larger than India. A further 2.9 million km² of moderately modified land (HM < 0.4) could become highly modified. Upper-bound (97.5th percentile) forecasts also identify large areas of currently natural land as potential hotspots of future human modification, suggesting the potential expansion of human pressure beyond the current conversion frontiers by 2040. By providing projections of future human pressure as continuous measures with explicit estimates of uncertainty, rather than broad land-use categories, these forecasts support more informed conservation decisions. They can help identify where protection or stewardship may prevent future loss, where restoration may be more feasible, and where precautionary action is most warranted. They can also be incorporated into spatial planning, allowing future pressure and uncertainty to be considered alongside representativeness, connectivity, biodiversity, and ecosystem services in support of the Kunming-Montreal Global Biodiversity Framework.
Related
- Probabilistic inference of homonymous and heteronymous recurrent inhibition in human muscles from large-scale motor neuron recordings
- Mapping Mesolimbic Dopamine Projections to Subcortical Circuits Underlying Adaptive Behavior: A Human 7T Diffusion Tractography Atlas
- Three-dimensional printing of human scala tympani models: surface modification to reduce cochlear implant insertion forces with a physiological lubricant
- Interacting effects of human presence and landscape modification on birds and mammals
- Genetic modification of primary airway epithelial cell cultures defines MX1 as a barrier for zoonotic influenza A viruses in the human respiratory epithelium
BibTeX
@misc{moncrieff2026probabilistic,
title = {{Probabilistic projections of human modification of the terrestrial biosphere from 2020 to 2040}},
author = {Glenn Moncrieff and Maria Voigt and James Oakleaf and David Theobald and Katie Walker and James Fitzsimons and Joseph Kiesecker and Christina Kennedy},
howpublished = {EcoEvoRxiv},
year = {2026},
month = sep,
doi = {10.32942/x2s11g},
url = {https://ecoevorxiv.org/repository/view/15284/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Probabilistic projections of human modification of the terrestrial biosphere from 2020 to 2040
AU - Glenn Moncrieff
AU - Maria Voigt
AU - James Oakleaf
AU - David Theobald
AU - Katie Walker
AU - James Fitzsimons
AU - Joseph Kiesecker
AU - Christina Kennedy
PY - 2026
DA - 2026/09/30
PB - EcoEvoRxiv
DO - 10.32942/x2s11g
UR - https://ecoevorxiv.org/repository/view/15284/
AB - Conservation actions are more effective when they anticipate future threats rather than respond only after loss occurs. Most spatial forecasts focus only on land-cover change and fail to capture the broader set of pressures that degrade biodiversity. To close this gap, we develop global probabilistic forecasts of cumulative human modification (HM), a continuous measure that combines 16 anthropogenic threats to biodiversity classified under the IUCN threat taxonomy. Our approach uses a convolutional long short-term memory network trained on the historical HM time series from 1990 to 2020 to generate forecasts at 1 km resolution and 5-year intervals from 2020 to 2040. The model ingests sequences of historical HM maps along with socioeconomic, biophysical, and protected-area covariates to generate forecasts of the full distribution of potential change. Model outputs outperformed a persistence baseline, which assumes that HM remains unchanged from the last observation. Conditional on the continuation of historical relationships and dynamics, the forecasts indicate that between 2020 and 2040, more than 4.5 million km² of land classified as natural (HM < 0.1) in 2020 could be lost, an area larger than India. A further 2.9 million km² of moderately modified land (HM < 0.4) could become highly modified. Upper-bound (97.5th percentile) forecasts also identify large areas of currently natural land as potential hotspots of future human modification, suggesting the potential expansion of human pressure beyond the current conversion frontiers by 2040. By providing projections of future human pressure as continuous measures with explicit estimates of uncertainty, rather than broad land-use categories, these forecasts support more informed conservation decisions. They can help identify where protection or stewardship may prevent future loss, where restoration may be more feasible, and where precautionary action is most warranted. They can also be incorporated into spatial planning, allowing future pressure and uncertainty to be considered alongside representativeness, connectivity, biodiversity, and ecosystem services in support of the Kunming-Montreal Global Biodiversity Framework.
ER -