I was performing regression based downscaling on gridded climate data (observed, to be served as predictand in the model). There are around 300 grid point in my study area. Do i have to perform downscaling on each grid point?
If you are supposing to downscale your data spatially from coarse resolution (e.g., 1deg) to high resolution (e.g., 1km) then it better is to take the average of all grids and set your regression equation for simplicity. After predicting your target variables using a regression equation, you can apply regression coefficients to your raster image at each grid using the raster calculator in ArcGIS and predict your target variable at high resolution.