gbif-biodiversity

To model species distributions from GBIF occurrence data, download records via the pygbif API, apply spatial thinning to reduce sampling bias, and fit presence-background SDMs using elapid to generate suitability GeoTIFF rasters.

Spatial thinning reduces sampling bias in GBIF occurrence records by retaining at most one record per specified latitude/longitude grid cell using GeoPandas, preventing clustered occurrence data from skewing MaxEnt-style species distribution model predictions.

Yes, you can use scikit-learn to build BRT (Boosted Regression Trees) models from spatially thinned GBIF records, complementing MaxEnt-style SDMs fitted with elapid, and stack the resulting suitability rasters for species richness mapping.

Reduce sampling bias in GBIF records before fitting MaxEnt models by applying grid-based spatial thinning using GeoPandas, which filters coordinate-clustered occurrence data to retain only one record per defined latitude/longitude grid cell.

Yes, this approach fits species distribution models using elapid and rasterio to output suitability GeoTIFF rasters, which can be thresholded and stacked across multiple species to generate species richness maps for ecological analysis.

Species distribution modeling with GBIF data requires geospatial raster environmental predictors to fit presence-background models, using rasterio to process the raster layers and generate predicted suitability outputs for ecological range mapping.