# How to interpret semivariogram parameters from two different raster images?

I know the definitions of the components of a semivariogram. However, I would want to know how they could be interpreted when applied to actual scenarios. For instance, I have two EVI (enhanced vegetation index) rasters computed from two different images (years 2000 and 2010). I calculated the semivariogram for each hoping I could compare them both, in terms of nugget effects, range, and sill values. First, can i do this? Second, how would i interpret a high range value for one and a low range for the other in terms of EVI? What about the rest of the properties: nugget effects and sill?

Example:

Image 1: nugget = 0.04,sill = 0.02, range = 6000, type = spherical, lag size = 750, no. of lags = 10, anisotropy = false

Image 2: nugget = 0.02,sill = 0.006, range = 1500, type = spherical, lag size = 250, no. of lags = 10, anisotropy = false

How would I "translate the variogram parameters into statements about the underlying EVI spatial patterns"?

• Please clarify what you mean by "interpret." In most applications this would mean translating the variogram parameters into statements about the underlying spatial patterns, which requires subject matter expertise. What kind of "comparison" do you have in mind? How did you calculate the variogram, given there are a huge array of options and choices to make in the variogram-estimation procedure? – whuber Aug 14 '18 at 12:29
• @whuber I picked 50 random points in both images, then calculated the variogram based on the points (EVI values). In short, I have nugget, sill, and range values from each image. And you are right, how would I "translate the variogram parameters into statements about the underlying EVI spatial patterns"? – GISnew Aug 14 '18 at 13:37
• But how exactly did you estimate the variograms? What lags did you choose, what were the lag populations, what was the neighborhood search method, are the variograms isotropic or directional, did you use the usual estimator or some other estimator (such as a robust one), what models did you fit to estimate the nugget, range, and sill, and how did you fit them? All this information is needed just to understand what you might be looking at: and you need to supply it with the question, not in the comments. – whuber Aug 14 '18 at 16:10