|
Abstract:
|
With the development of new remote sensing technology, large or even massive spatial datasets covering the globe become available. Statistical analysis of such data is challenging. We propose a semiparametric approach to modeling and inference for massive spatial datasets. In particular, a Gaussian process with additive components is considered, with its covariance structure coming from two components: one part is flexible without assuming a specific parametric covariance function but is able to achieve dimension reduction; the second part is parametric and simultaneously induces sparsity. The inference algorithm for parameter estimation and spatial prediction is devised. The resulting spatial prediction method that we call sparsity-induced kriging (SIK), is applied to simulated data and a massive satellite dataset. The results demonstrate the computational and inferential benefits of SIK over competing methods and show that SIK is more flexible and more robust against model misspecification.
|
