A generative deep learning approach for global species distribution prediction DOI Creative Commons
Yujing Yan, Bin Shao, Charles C. Davis

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 16, 2024

Abstract Anthropogenic pressures on biodiversity necessitate efficient and highly scalable methods to predict global species distributions. Current distribution models (SDMs) face limitations with large-scale datasets, complex interspecies interactions, data quality. Here, we introduce EcoVAE, a framework of autoencoder-based generative trained separately nearly 124 million georeferenced occurrences from taxa including plants, butterflies mammals, their distributions at both genus levels. EcoVAE achieves high precision speed, captures underlying patterns through unsupervised learning, reveals interactions via in silico perturbation analyses. Additionally, it evaluates sampling efforts interpolates without relying environmental variables, offering new applications for exploration monitoring.

Language: Английский

Tracking hidden dimensions of plant biogeography from herbaria DOI Creative Commons
Barnabas H. Daru

New Phytologist, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 14, 2025

Plants are diverse, but investigating their ecology and evolution in nature across geographic temporal scales to predict how species will respond global change is challenging. With breadth, herbarium data provide physical evidence of the existence a place time. The remarkable size collections along with growing digitization efforts around world possibility extracting functional traits from preserved plant specimens makes them invaluable resources for advancing our understanding changing distributions over time, biogeography, conserving communities. Here, I synthesize core aspects biogeography that can be gleaned herbaria distributions, attributes (functional biogeography), conservation globe. advocate collaborative, multisite, multispecies research harness full potential these while addressing inherent challenges using macroecological investigations. Ultimately, present untapped opportunities enable predictions species' responses inform effective planning.

Language: Английский

Citations

1
Predicting undetected native vascular plant diversity at a global scale DOI Creative Commons
Barnabas H. Daru

Proceedings of the National Academy of Sciences, Journal Year: 2024, Volume and Issue: 121(34)

Published: Aug. 12, 2024

Vascular plants are diverse and a major component of terrestrial ecosystems, yet their geographic distributions remain incomplete. Here, I present global database vascular plant by integrating species distribution models calibrated to species’ dispersal ability natural habitats predict native range maps for 201,681 into unsurveyed areas. Using these maps, uncover unique patterns diversity, endemism, phylogenetic diversity revealing hotspots in underdocumented biodiversity-rich regions. These hotspots, based on detailed species-level show pronounced latitudinal gradient, strongly supporting the theory increasing toward equator. trained random forest extrapolate under unbiased sampling identify overlaps with modeled estimations but unveiled cryptic that were not captured estimations. Only 29% 36% extrapolated inside protected areas, leaving more than 60% outside vulnerable. However, unprotected harbor attributes make them good candidates conservation prioritization.

Language: Английский

Citations

4
Climate change alters the future of natural floristic regions of deep evolutionary origins DOI Creative Commons
Samuel Minev-Benzecry, Barnabas H. Daru

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: Nov. 2, 2024

Abstract Biogeographic regions reflect the organization of biotas over long evolutionary timescales but face alterations from recent anthropogenic climate change. Here, we model species distributions for 189,269 vascular plant world under present and future climates use this data to generate biogeographic based on phylogenetic dissimilarity. Our analysis reveals declines in beta diversity years 2040 2100, leading a homogenization regions. While some boundaries will persist, change alter separating realms. Such boundary be determined by altitude variation, heterogeneity temperature seasonality, past velocity. findings suggest that human activities may now surpass geological forces shaped floristic millions years, calling mitigation impacts meet international biodiversity targets.

Language: Английский

Citations

1
A generative deep learning approach for global species distribution prediction DOI Creative Commons
Yujing Yan, Bin Shao, Charles C. Davis

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 16, 2024

Abstract Anthropogenic pressures on biodiversity necessitate efficient and highly scalable methods to predict global species distributions. Current distribution models (SDMs) face limitations with large-scale datasets, complex interspecies interactions, data quality. Here, we introduce EcoVAE, a framework of autoencoder-based generative trained separately nearly 124 million georeferenced occurrences from taxa including plants, butterflies mammals, their distributions at both genus levels. EcoVAE achieves high precision speed, captures underlying patterns through unsupervised learning, reveals interactions via in silico perturbation analyses. Additionally, it evaluates sampling efforts interpolates without relying environmental variables, offering new applications for exploration monitoring.

Language: Английский

Citations

0