Conservation of mass with conservative regridding

@andersy005 @rabernat @dcherian

With the newer version of the Python xesmf library, I have a plausible result. But my question is:

I am downscaling lightning frequency (number of flashes per number of seconds in a month). When I add up the total lightning freq over an entire high resolution domain shouldn’t the total lightning freq in the downscaled (coarse) conservatively regridded domain by the same? I thought mass should be conserved.
There is difference in the hi-res vs the regridded
Hi-res grid: 2.276
regrid: 0.000834

Perhaps I am doing something wrong orshould I regrid the actual hi-res GLM data files that are flash counts for every minute before I calculate lightning freq and the monthly average.