Retrieval accuracy validation data for a physics-driven multi-task deep learning framework using SMAP and WindSat observations

This dataset contains processed supporting data for retrieval-accuracy validation of a physics-driven multi-task deep learning framework for simultaneous retrieval of ocean-atmosphere parameters from SMAP and WindSat observations. The archive includes density-scatter validation points, annual spatial bias and RMSE fields, zonal mean profiles, seasonal retrieval distributions, daily error time series, public input-data source links, and a Python script for reproducing the validation plots.