This is misleading: none of the given examples (median, max, sd) can actually be used with apply_neighborhood: the former processes take an array as input and produce a scalar, while apply_neighborhood expects processes that receive a raster data cube and produce a new raster data cube.
I think the only pre-defined process that makes practical sense to pass to apply_neighborhood is run_udf at the moment
I propose to just remove the following from the cookbook
Pre-defined processes (median, max, sd etc.) can be applied to spatial, temporal or even spatio-temporal neighbourhoods with apply_neighborhood. Due to lack of implementation by back-ends, this is not covered at this point.
Note that the VITO backend does implement apply_neighborhood, so that second sentence also makes little sense to keep as-is.
Raised by @MartinSchobben at https://github.com/Open-EO/openeo-processes-dask/pull/215#issuecomment-1883670997
(from: https://github.com/Open-EO/openeo.org/blame/10b68cee63b1dac17d8a4902a02f325563be6743/documentation/1.0/cookbook/README.md#L1198)
This is misleading: none of the given examples (median, max, sd) can actually be used with
apply_neighborhood: the former processes take an array as input and produce a scalar, whileapply_neighborhoodexpects processes that receive a raster data cube and produce a new raster data cube.I think the only pre-defined process that makes practical sense to pass to
apply_neighborhoodisrun_udfat the momentI propose to just remove the following from the cookbook
Note that the VITO backend does implement
apply_neighborhood, so that second sentence also makes little sense to keep as-is.