I am trying to calculate the mean of a raster using zone data from another raster. The zone data was created using the Whitebox clump tool where each group of pixels have a unique ID number. There are ~800,000 unique clumps and the rasters are fairly big (27700, 31511) so I am avoiding converting the clumps to vector format because this causes memory issues. I would like the output of this analysis to be another raster where the clump IDs in the original raster are replaced with the mean value of that clump or a table containing the clump IDs and mean value of that clump. This process is the same as ArcMaps Zonal Statistics but I would like to use python and open source packages. I tried this using the code below that I wrote but it is way too slow for the size of data I'm working with.

clumps = raster of clumps 
IDs = np.unique(clumps)
values = raster of values
means = clumps
for id in IDs:
    mask = clump
    mask = np.where(mask == id, 1, 0)
    mean_value = mask*values
    mean_value[mean == 0] = np.nan
    mean_value = np.nanmean(mean_value )
    mean_value = mean_value.astype(np.int64)
    means[means == id] = mean_value


I found a solution by tiling the input rasters and running the above code for each tile. I used the clump output from Whitebox Tools as my zone data.


You could use WhiteboxTools' ZonalStatistics tool to do this. As a Python script, it would look like the following:


There should be no problem running this tool with 800,000 features from the clump output.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.