Geostatistical analysis using K-splines in the geoadditive model

Abstract

In geostatistics, both kriging and smoothing splines are commonly used to predict a quantity of interest. The geoadditive model proposed by Kammann and Wand (2003) represents a fusion of kriging and penalized spline additive models. The fact that the underlying spatial covariance structure is poorly estimated using geoadditive models is a drawback. We describe K-splines, an extension of geoadditive models such that estimation of the underlying spatial process parameters and predictions of the spatial map are performed with the same accuracy and precision as in kriging.