The basic idea underlying the proposed approach to spatial-temporal modeling is to combine, at the different levels, the potential of various techniques and to exploit their performance. A three level procedure is used. Firstly low level polynomial interpolation removes a general trend; then finite element method follows to match local variations. The last step is performed assuming the topography to be a realization of a continuos, normal, stationary, ergodic and isotropic stochastic process. Thus a filtering technique, based on covariance estimation and collocation, is applied. Finally the three levels can be computed together by using an integrated approach.
展开▼