On Robustness for Spatio-Temporal Data

The spatio-temporal variogram is an important factor in prediction through kriging, especially fields such as environmental sustainability or climate change, where data analysis based on this concept. However, the traditional estimator, which commonly employed for these purposes, extremely sensitive to outliers. We approach problem two ways paper. First, new robust estimators are introduced, defined M-estimators of original transformation. Second, we compare classical estimate against a one, identifying outliers way. To accomplish this, use multivariate scale-contaminated normal model produce reliable approximations sample distribution estimators. In addition, define and study class paper, including real-world applications, order determine whether there any significant differences between temporal lags and, if so, can reduce number considered analysis.