"Beyer, Christof" wrote: > > according to the gstat-manual gaussian simulation on a regular grid using a > mask map for simulation locations is done in gstat following a recursively > refining random path on that grid. > My question is in which sequence locations defined by the "data > ():...;"-command will be simulated by gstat. Are they simulated just > line-by-line (which to my understanding of conditional simulation would not > make very much sense) or in a random sequence. > In a completely random sequence, without the refining (as no information about a pattern is present). You could turn this randomization off by setting
set rp = 0; # no random path in the command file. > Is there a danger of > introducing bias to the simulated values if gstat simulates locations which > are in fact regularly gridded but defined by the "data ():...;"-command? It may not be as bad as you think. The bias is a consequence of using a local neighbourhood to approximate a global neighbourhood. If you use a regular sequence with a large neighbourhood I'd expect almost perfect simulations. I haven't experimented with it though, so who am I to tell you how large the neighbourhood should be. > > I'd be very happy if someone could help me with that! > No problem. -- Edzer
