Dear all Gstat users, I use Gstat and I've some questions for you. - I decided to use Gstat for declustering my data set with the polygonal method (Goovaerts, Geostatistics for Natural Resources Evaluation, 1997, pp.79-81; Isaaks and Srivastava, An Introduction to Applied Geostatistics, 1989, pp.238-239). Well, though this method could be improved according to Dubois (How representative are samples in a sampling network?, Journal of Geographic Information and Decision Analysis, vol.4, no.1, pp.1-10, on-line), I decided to apply it just as weights to apply to my data depending on polygonal area. I added an index column to my data set, so that every sample could be seen as unique. Using Gstat I executed the following command: #col 1 contains unique identifier -index- I put 'average' because maybe I've values coming from different times on the same sample point or laying in the same cell (see following discussion). >data(clus): 'zinc.eas', x=2, y=3, v=1, average, max=1; >mask: 'mask.map'; >predictions(clus): 'distarea'; Using PCRaster I executed the following pcrcalc script: >#! --clone mask.map --unitcell --nondiagonal ># transform distarea to ordinal/nominal class: is 'ordinal' the same in this case? >distcls=nominal(distarea*1000); # *1000 is to create different classes for <> decimal values (if the case) until 3rd significant decimal digit >area=areaarea(clump(distcls)); >idealarea=maptotal(area)/maparea(area); #idealarea stands for declustered mean area >report declarea=area/idealarea; #calculate weights for samples Finally using Gstat again I created a new file containing my data with a column weight added: >data(a): dummy; >data(): 'zinc.eas', x=2, y=3, v=1, average; # prediction locations >method: map; >masks: 'declarea'; >set output = 'decl_data.eas'; Now I could multiply all my column data by weights with the software I want and use them. Well my questions for you are: - do you think my method works well or there's something I did wrong? If I've missing values in a column I should repeat the procedure for that column, given that weights change because I've sample points that are missing (with their own area) in my study area. - Does anybody know how Gstat treats data which have been sampled on different points, but that lie on the same map cell (for example xa=181072 ya=333611, xb=181062 yb=333601 and cell size 20m (->cellarea 400m2)? Does it average them? - another problem in converting *.eas file->PCRaster map, is that coordinates are treated as the center of the cell; when I do the back conversion I don't obtain original coordinates, but the center of the cell. So, if I want to be nearest to my original coordinates I've to choose a very little cell size. Anyway this is not always useful due to huge map created if points lie far away. Besides, kriging in geostatistics uses relative distances for weights: if they are changed, weights change. Is there a solution to this problem? Thanks in advance for help. --- Giovanni De Ferrari
