I have a relatively big raster file on which I want to conduct geographically weighted regression. Size of raster ~ 2 GB and number of rows is around 100,000

One way might be to convert from raster to polygon and then use the GWR in ArcGIS. However, the raster to polygon conversion is taking too long.

Is there any function in R or python which can conduct GWR on big rasters? Alternatively, any approach on how to sample a huge raster so that I can run GWR on the smaller sample?

  • do you mean geographically weighted regression ? please specify. – radouxju Apr 15 '14 at 15:40
  • yes, i mean geographically weighted regression (also expanded in question). thanks! – user1186 Apr 15 '14 at 16:01
  • 1
    if your raster is too big, I would first try resampling at a lower resolution (with gdal, for instance, and double pixel size will bring you down to approx 500 Mb) – radouxju Apr 17 '14 at 6:09

Apparently, there is a GWR function in R. I have not applied it myself, so I have no idea if it can be applied to rasters, but I did find this document for the command in R:


and this GWR python library:



There are some multiple regressions (including GWR) functions in SAGA GIS that offer many options of rasters as predictors, dependent variables and outputs. I hope it can help you.

  • Can you elaborate more? Which functions? How to use them? – Andre Silva Apr 19 '16 at 13:54
  • Hi! If you enter in the tool libraries in SAGA GIS and go to "Spatial and Geostatistics' and then into "Regressions", you will see many functions of GWR and Multiple Regressions, with options to specify rasters, and shapefile (points or polygons) as predictors, dependent variables and results. After you imported your data into SAGA GIS, it just works like the toolbox of ArcGis. You also have the option of using SAGA GIS toolbox within QGIS, too. – Vitor Vieira Vasconcelos Apr 21 '16 at 1:15

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.