I have two set of rasters, with data that should correlate...one are measured data, one are modelled data. The correlation is weak, however, when looked at with a normal regression analysis. I suspect that different parts of the base terrain need different modelling input, or that in some parts there is a shift between the correlating cells.

So now I try to visualize where these residuals are highest between my two rasters, and where the real outliers are and whether they are in similar terrain. Is there any tool that would allow me to directly compare the rasters in GIS?

I run ArcGIS 9.3 with Spatial Analyst and Statistics, and also have QGIS and GRASS on my Linux-based machine. What would be easiest? I am not trying to do a full-on analysis, I am just trying to see where the odd data are and if there is a systematic spatial pattern behind them.

Cheers, Peter

  • This maybe of interest: gis.stackexchange.com/questions/20127/… – dmci Mar 20 '13 at 13:13
  • Hi! Thanks! I saw this topic, but I hoped to get around R...but I will look more into it, seems worthwhile – Peter Mar 20 '13 at 13:24
  • My first instinct is also R, though you can use the Geographically Weighted Regression tool in ArcGIS if you can aggregate your raster values using a polygon feature class. GWR won't accept raster inputs, if I remember correctly. – Christopher Aug 7 '16 at 12:41

Maybe a simple raster difference (raster1 minus raster2) will give you a starting point?

I would use ArcGIS for that, but simply because i know how to do it in ArcGIS and don't know how in QGIS.

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.