2025 Volume 33 Issue 3 Pages si-91-si-97
This study reviews recent R packages for spatial statistical modeling, focusing on scalable alternatives to Gaussian processes (GPs), which are limited by high computational cost. We categorize key approximation strategies into low-rank basis methods, covariance tapering, and sparse precision matrix approaches such as the SPDE method. Among them, the sdmTMB package, which supports generalized linear mixed models, stands out for its computational efficiency, flexibility, and ease of use. We demonstrate its practical utility through a case study on fish distribution data, highlighting its ability to model spatiotemporal variation and terrain-constrained processes.
