Visual Comparison of Moving-Window Kriging Models

Kriging is a spatial prediction method that is core to the geo-statistical paradigm. Commonly applied in the environmental sciences, it enables a prediction at an unsampled location coupled with a measure of confidence in its accuracy. Many variations of kriging exist, some of them complex, especially those that allow many parameters to vary spatially. Calibrating such a kriging model and interpreting its results can therefore be quite daunting. We suggest that visualization and visual exploration can help with this task. In particular, we focus on the moving-window kriging model, evaluating three newly developed robust variants of this model against a basic counterpart. We use star icon maps and plots to visually explore model results to evaluate model parameterization, specification, and performance.