AINPP PB LATAM

This dataset provides a curated and harmonized collection of gridded precipitation fields derivedfrom the GSMaP Near Real-Time (NRT) and GSMaP Moving Vector with Kalman filter (MVK, V8)products, specifically designed to support reproducible machine learning experiments in precipitationnowcasting. The dataset is structured to enable supervised learning, where GSMaP-NRT fields are usedas input data and the higher-quality GSMaP-MVK fields serve as forecast targets, allowing systematicevaluation of models under realistic data latency constraints. The dataset covers the AINPP Latin Americadomain (latitudes \u221255\u00b0 to 33\u00b0, longitudes \u2212120\u00b0 to \u221223\u00b0) on a regular 0.1\u00b0 grid, yielding spatial fieldsof 880\u00d7970 pixels per time step. The temporal extent spans January 1, 2018 to December 31, 2024, withan hourly temporal resolution. To prevent temporal data leakage and ensure fair model development,the data are explicitly partitioned into training (2018\u20132022), validation (2023), and test (2024) subsets.A standardized preprocessing pipeline is applied to both products, consisting of a log1p transformationfollowed by Z-score normalization. The normalization parameters are computed exclusively from theGSMaP-MVK training period and consistently applied to both NRT and MVK across all temporalsplits. The dataset is stored in a compressed Zarr (v2) format with chunking optimized for deep learningworkflows, and the normalization parameters are distributed to allow exact inverse transformation backto physical precipitation units. This dataset provides a robust and scalable foundation for precipitationnowcasting benchmarks and reproducible experimentation.