Spatially adaptive covariance tapering

Covariance tapering is a popular approach for reducing the computational cost of spatial prediction and parameter estimation for Gaussian process models. However, previous work has shown that tapering can have poor performance for example when the data is sampled at spatially irregular locations. In this work we introduce a computationally convenient non-stationary taper method in order to improve the performance of tapering in the case of spatially irregular observation locations or when non-stationary covariance models are used. Numerical experiments are used to show that the performance of both kriging prediction and parameter estimation can be improved by allowing for spatially varying taper ranges.