Statistical package for computing precision covariance matrices via modified Cholesky decomposition

Abstract

We introduce a statistical package designed to compute precision covariance matrices using modified Cholesky decomposition, tailored for Atmospheric General Circulation Models (AGCMs). By incorporating predefined localization radius structures, the package achieves sparse precision matrix estimation, enhancing computational efficiency and scalability for large-scale atmospheric models. Leveraging ensemble modeling, the method ensures robustness and accuracy, with direct applications in data assimilation tasks. The package also features seamless integration with prominent climate datasets, such as ERA5 and MPAS, enabling streamlined workflows for climate modeling and prediction.