I have to perform a mapping of a DVB-T field (the TV signal), per every location I consider the median in time of the measurements, there are some issues, e.g. the variance seems to be proportional to the intensity of the field because of the measuring instrumentation.
My first idea was to use kriging.
Following R's gstat package vignette I have plotted the directional variograms and one of the variables that are measured shows a clear anisotropy.
How do I deal with it? Is there a favourite framework?
Looking at literature a change in variable is suggested but since my data are not in a grid I do not know how to deal with it.
I have looked at this post Problems estimating anisotropy parameters for a spatial model
How do you know from the cloud variogram if there are enough points to consider aisotropy?
In a second dataset which is less numerous a low number of samples for a range of distance, that is lower than the range corresponds to a drop in the variogram. I am not sure if this is a phisiological drop due to the field or if it is due to the low-sampling.
Is there any way to find out?
For doing this kind of analysis I am using gstat in R. Is geoR a better framework? Or is there a better package in absolute?
There are two measurement campaigns, one with 110 locations, the other with 35, in a territory which is approximately 4 km * 4 km, but I have also simulated the field, so from the simulated field I can extract as many samples as I want.
Any pointer to relavant literature is very welcome.
At the moment I am reading [Webster, Oliver] Geostatistics for Environmental Scientists.