Predicting species occurrence patterns from partial observations

arXiv:2403.18028v1 Announce Type: new
Abstract: To address the interlinked biodiversity and climate crises, we need an understanding of where species occur and how these patterns are changing. However, observational data on most species remains very limited, and the amount of data available varies greatly between taxonomic groups. We introduce the problem of predicting species occurrence patterns given (a) satellite imagery, and (b) known information on the occurrence of other species. To evaluate algorithms on this task, we introduce SatButterfly, a …

abstract arxiv biodiversity climate cs.ai cs.cv cs.lg data however patterns q-bio.pe species type understanding