I have two tiff files in R (one modelled and one observed). Both tiffs show the spatial maps of gross primary productivity from 2000-2010. The study area is HinduKush Himalaya (http://rds.icimod.org/Home/DataDetail?metadataId=3924). I would like to find the agreement between these two tiff files statistically.

I go ahead with plotting a scatterplot in R. The y axis is the modelled gross primary producitivty and x is modis gross primary productivty. Looking at the scatterplot, most of the data is present on the left side. I have two concerns in this scatterplot. Firstly there is a lack of linearity and the amount of overplotting. What can I apply to my code to overcome the problem?



#resample to bring both rasters to same resolution. GPPMODIS1 is 0.01 and rastergpp1 is 0.5


#plot scatterplot

This is the scatterplot between 2 rasters


This is the map of residuals map of residuals

This was achieved by

  • 2
    $\begingroup$ Hello, and welcome to CrossValidated! First question, second question. I deduce that this is your third question on the same end of analyzing tiff files. You should instead show progress by editing the broad original question rather than putting separate questions. On this scatterplot: you mention your data is in a raster. What is each of the dots in this scatterplot? Is your data spatial? $\endgroup$ – Nuclear03020704 Aug 12 at 11:40
  • 1
    $\begingroup$ Please specify what particular aspect of "odd looking" you have in mind. Is it the spacial pattern, is it the lack of linearity, is it the amount of overplotting that turns a large region basically black, is it the jitter that is applied to discrete data, is it the intercept of the regression line that cannot be seen of just a preference for ggplot2? Please make it more clear, what defines a good answer for your question. $\endgroup$ – Bernhard Aug 12 at 13:13
  • $\begingroup$ I apologise for the unclear question. I edited the question. $\endgroup$ – Hallie Sheikh Aug 12 at 13:36
  • $\begingroup$ If I understand correctly, each (x,y)-pair corresponds to a certain spatial location (on a grid?). Mybe you could show us a map of the residuals from that regression? $\endgroup$ – kjetil b halvorsen Aug 12 at 15:01
  • $\begingroup$ @kjetil According to the description in the question as well as the code, this is a scatterplot of collocated values on two grids. Each dot corresponds to one pair of overlapping grid cells. One of the grids had to be resampled to create these overlaps--and therein is part of the problem--so it's important to understand this workflow. $\endgroup$ – whuber Aug 12 at 15:54

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.