The scipy.interpolate.RectBivariateSpline interpolator is a special-case of regular grid interpolation (corresponding to scipy.interpn(method="splinef2d"), while our 2D point interpolation/resampling can use any other method of interpn ("linear" which is the default, "cubic", etc).
We should keep using the interpolator directly for computational efficiency (instead of multiple calls of Raster.interp_points()), but generalize to any resampling method, with a consistent default ("linear" as elsewhere).
The
scipy.interpolate.RectBivariateSpline
interpolator is a special-case of regular grid interpolation (corresponding toscipy.interpn(method="splinef2d")
, while our 2D point interpolation/resampling can use any other method ofinterpn
("linear"
which is the default,"cubic"
, etc). We should keep using the interpolator directly for computational efficiency (instead of multiple calls ofRaster.interp_points()
), but generalize to any resampling method, with a consistent default ("linear" as elsewhere).Opened https://github.com/GlacioHack/geoutils/issues/559 for this, and we should also add a
config
option for this!