Universal kriging with hydrogeologic covariates

Universal kriging can use an explicit trend and named raster or point/grid covariates. Geometry must match by default; requested alignment is recorded. The fit checks missing coverage, constants, duplicates, collinearity, and rank. Covariates should be scientifically defensible and available throughout the prediction domain. Statistical association is not causality.

uk <- ps_interpolate(points, methods="UK", template=template,
                     trend=Z~land_surface+valley_distance,
                     covariates=list(land_surface=dem, valley_distance=valley))