Dear geostats aficionados, Should you have a rainy week-end ahead, here are a few questions that may keep you (shortly) away from the TV.
We all know that because variogram models have to be positive definite, the number of options to fit complex spatial structures is relatively limited What is the most common practice when the analysed phenomenon clearly presents nested structures that can't modelled properly with the allowed functions? (e.g. imagine several exponential structures shifted by a certain distance and with increasing sills, or see e.g the case study in fig 9 in J. Soil Science, 1986, 37 617-639, McBratney and Webster) Forcing the fitting of such nested structures by combining approximatively several models seems to be the general rule but it is frustrating to see that so much information regarding the spatial structure is getting lost in such a way. Is kriging based on the first model to calculate residuals which will in turn be interpolated using a second variogram model a reasonable option? How would one deal with the nugget effect of the second model (which in theory should have been already extracted with the first variogram model)? Many thanks for any feedback and thoughts. Have a nice w-e Gregoire __________________________________________ Gregoire Dubois (Ph.D.) European Commission (EC) Joint Research Centre Directorate (DG JRC) Institute for Environment and Sustainability (IES) TP 441, Via Fermi 1 21020 Ispra (VA) ITALY Tel. +39 (0)332 78 6360 Fax. +39 (0)332 78 5466 Email: [EMAIL PROTECTED] "The views expressed are purely those of the writer and may not in any circumstances be regarded as stating an official position of the European Commission."