Analysis of anisotropic variogram models for prediction of the Curonian lagoon data

Abstract

The anisotropy in particular environmental phenomena is detected when behavior of a physical process differs in different directions. In this paper geometric and zonal anisotropies are considered. Various methods of geostatistical analysis, also isotropic and geometrical anisotropic variogram models are compared and applied for the Curonian lagoon depth data. The results demonstrate that after robust estimation, i.e. elimination of outliers and after elimination of geometric anisotropy the precision of prediction and adequacy of models are much better. All computations have been performed by means of gstat, base and spatial packages of R system. Prediction results are compared with the results of research where outliers and geometric anisotropy were not eliminated.

Key words:

• variogram
• robustness
• trend
• geometric anisotropy
• zonal anisotropy
• universal kriging