Hi @marco.guzzetti , thanks for trying to troubleshoot this and for updating us.
This makes sense. I think the critical thing here is that the extent of the LULC extends beyond the Area of Interest by at least the distance of the “max blending distance”. When this is not the case, the nodata values from the edges of the LULC seem to creep deep into the AOI during the convolution steps. (This is evident in nodata appearing in intermediate/T_air.tif). I think we may be able to improve the model’s behavior there, but for now you found a good workaround by using a larger LULC.
The User’s Guide suggests this column should be included here, but I wonder if that is simply a mistake in the documentation. I think it makes more sense for energy_sav to appear in the buildings result. @jesseG do you agree this is just a mistake in the UG?
And I’ll let Jesse also comment on your other energy_savings question since I think he’s thought about that more than I have.