geospatialgstatcovariogram

minimum sample size for estimating a reliable variogram


I have thirty three SO2 readings from sensing equipments installed at 33 locations spread over a large region (one observation per city). The sample size is fixed and can not be increased because of the number of equipments installed. Now (geo)-statistically it's a quite a small sample as many authors state that the sample size should be between 100-150 to get reliable variogram. Can somebody guide me as to how to deal with such a small data? Further, it actually will be a spatio-temporal analysis, but at each time instant again there are 33 observations. What I actually want to ask is that what is the impact of increasing the area size (resolution), whereas the number of overall sample points still remains the same, on variogram estimation?

I want to post an image but don't have enough reputation.

In the above both cases the relative distance between different pairs still remain the same. That's dist(z_1, z_2)/dist(z_2, z_3) remain the same on both scales.


Solution

  • Variogram estimation does not get any better if you increase area size or resolution, but do not increase the amount of sample data from which you compute the sample variogram.