What are some best practices for calculating local geostatistics across the boundaries of different projections?
I am thinking of using the State Plane Coordinate System to represent a national-level dataset. However I will be conducting analysis at the local metropolitan area level. I'd like to hear how people have handled analysis of zones that cover multiple projections (like the Kings County state plane in the Seattle region).
EDIT: the following was previously a comment:
We are applying convolutions on rasterized land use data to learn significant land use features. It seems like area will be the most important feature to preserve, but distance is a close second. My hope was that by dividing in the country into 120+ projections, that SPCS would essentially make such distortions insignificant. But that of course introduces inconsistencies when convolving over the boundary between two different projections