Possible ways to speed up spatio-temporal GP modeling

That sounds amazing! Although I haven’t calculated, I think the number of missing data points for a given month is > ~90% in the full space-time grid. That’s because the satellite returns to the same location only every few days, and even if it comes back, it passes over a section (i.e., swath) of the land. Then there is a quality issue, so we toss out lots of data points.

If there is a simple toy example for this full grid + indexing, that would be extremely helpful.

Thank you so much!