Station-informed targets
ERA5-Land fields are combined with station residuals to build transparent high-resolution targets without using CHELSA as training data.
The paper in three steps
Open methods, not a black box.
The XCLIM paper describes a reproducible pipeline designed to retain year-to-year climate variability instead of publishing only climatological normals.
Deep-learning downscaling
An enhanced EDSR model learns monthly 1/120° temperature and precipitation fields from coarse climate inputs and terrain.
Year-specific correction
A post-inference 3D-IDW station correction anchors each year to observations before BIO01–BIO19 are computed.
Read the backbone
XCLIM: an open, station-informed deep-learning pipeline for 1/120° monthly climate time series (1950–2025).
Open the preprint →
The live atlas
Plain GeoTIFFs, ready for analysis.
Every layer has a stable URL, explicit variable name and HTTP Range support. Use curl, open the URL directly in terra, or generate a batch script.
Global archive
Global 9 km BIO01–BIO19 layers spanning 1950–2025 for broad-scale ecological modelling and historical comparison.
Browse coverage →CONUS 1 km layers
Independent GeoTIFFs for 1986–2025, served directly for selective access and reproducible station-scale work.
View the API →R-facing toolkit
The companion xbioclim package computes BIOCLIM variables from monthly rasters with a native, raster-aware core.
Meet the toolkit →The toolkit
One public layer, two development paths.
R packagestable
xbioclim
Compute the 19 standard bioclimatic variables from monthly climate data with a high-performance native core and raster-aware workflows.
R packagedevelopment
xbioclim-devel
The working line for new algorithms, performance experiments and API changes before they move into the stable package.
Follow development →Start with one URL
Discovery to analysis in one line.
Select a year and variable from the API, then let your existing geospatial tools do the rest.
library(terra)
bio01 <- rast("https://data.xbioclim.org/api/download/1987/bio01_1987.tif?dataset=1km")
plot(bio01)