mlcast-dataset-IT-DPC-SRI: Data Processing Pipeline for the IT-DPC-SRI Radar Archive

This repository contains the source code used to generate the IT-DPC-SRI dataset, a 16-year (2010–2025) Analysis-Ready Cloud-Optimized (ARCO) archive of Italian radar precipitation estimates. Overview The software implements the end-to-end data processing workflow required to harmonize heterogeneous historical radar products from the Italian Civil Protection Department (DPC) into a unified, cloud-native Zarr datacube. Key Processing Steps Data Ingestion: Parses raw radar Surface Rainfall Intensity (SRI) products from multiple historical formats, including legacy OPERA BUFR, ODIM HDF5, and operational GeoTIFFs. Spatial Harmonization: Reprojects varying historical spatial domains and map projections onto a standardized 1200×1400 km Transverse Mercator target grid at 1 km resolution. Temporal Alignment: Aligns historical observations with varying temporal resolutions (15-minute, 10-minute, and 5-minute intervals) onto a consistent 5-minute time grid without artificial interpolation. ARCO Generation: Compresses (via Blosc/Zstd) and chunks the harmonized arrays into a CF-compliant Zarr store optimized for efficient spatial slicing and cloud-native access. Purpose and Reproducibility This software is released to ensure full computational reproducibility and transparency of the data harmonization pipeline described in the associated Scientific Data data descriptor (in preparation). Source Repository: github.com/mlcast-community/mlcast-dataset-IT-DPC-SRI