REGIONALIZATION OF PRECIPITATION DATA IN THE HARZ MOUNTAINS

Universal kriging with simple altitude-dependent trend functions prove d to be a suitable technique for regionalization of average annual pre cipitation in the Harz mountains. The inherent usage of the present sp atial autocorrelation of residuals improves the simple estimation by o rdinary least squares regression. This improvement turns out to be rem arkable on the lee-side. The superiority of universal to ordinary krig ing, measured by means of crossvalidations is evident on the luff as w ell as on the lee-side. The iteration procedure for the estimation of semivariograms exhibited quick convergence after only few steps.