I have a question about multiple raster layer regression using R or Arcgis. I would like to predict biomass using some Vegetation Index - VI (NDVI, EVI, SAVI, OSAVI...). and generate R² and RMSE maps for my study area.

At first moment, I cut the VI images and stacked them with biomass, because I have for some areas biomass field plot. So, I create a stack with biomass field and VI's. Then I extracted the value (Biomass and VI's for each pixel where I have biomass value) and generated a model to predict biomass (I have used R and regsubset function to do it). Doing this a got, for example (Biomass = B0 + B1*NDVI + B2*SAVI) I could apply this model in my VI's images to extrapolate, but I will have only biomass maps.

I think to get R², RMSE maps, I must do a different process, someone could help me?

  • 1
    Do you want a local or global model? In either case you would make estimates to each pixel but how the model is approached is quite different, with very different assumptions. It wold be helpful if you provide code illustrating what you have already tried and specify R and ArcGIS in your tags so we know what software you are working with. – Jeffrey Evans Mar 6 '18 at 16:29
  • I think you need a spatial regression technique. Start with spatial error and spatial lag in R. Or you could do a point pattern analysis. – Mox Mar 6 '18 at 17:53
  • Thanks Jeffrey for your help. I want a local model to estimate to each pixel. In my R code I just did a linear regression using regsubset and applied lm() then I used the model to predict the biomass in my whole area. I just mentioned ArcGIS because it would be a option... but i prefer R – Synet Mar 7 '18 at 13:34
  • I will do it Mox, if you have some example, please, share – Synet Mar 7 '18 at 14:16

Your Answer

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

Browse other questions tagged or ask your own question.