This is the official code repository for the paper "A Novel ML-NWP Coupled Optimization Approach on a Reconstructed WRF Model".
Based on the Python programming language, we reconstructed the Fortran-version mesoscale numerical weather prediction model WRF in a GPU-accelerated tensor computing environment, enabling its seamless integration into various artificial intelligence (AI) frameworks. The currently reconstructed pyWRF model incorporates the complete thermodynamical framework of WRF and one microphysics parameterization scheme (WSM6), which can basically reproduce the simulation results of the original Fortran-based WRF model.
Due to the extensive codebase and complex structure, the repository is scheduled for release around June 2026.
Citation: Chu, H., Y. Cao, R. Wang, et al., 2026: A novel ML-NWP coupled optimization approach on a reconstructed WRF model. J. Meteor. Res., 40 (x), XXX－XXX, doi: 10.1007/s13351-026-5154-1.