Thanks for your help, I used to interpret cross validation on a graph where actual values are plotted against estimated values. With your informations about gstat cross validation output file and more theory about cross-validation (I got a book), I would be alright. I don't need more documentation. Thanks Guillaume.