ASTRA: AMSR2-to-SMAP Image-to-Image TRAnslation for global daily 36 km soil moisture (2012-07-03 to 2023-12-31)

ASTRA is a global daily 36-km soil moisture product generated using a deep learning framework that transforms AMSR2-based products into SMAP-consistent soil moisture estimates. ASTRA employs a Swin-Unet architecture trained on SMAP SM retrieved with the single-channel algorithm at vertical polarization, using AMSR2 Land Parameter Retrieval Model SM, vegetation optical depth, and skin temperature as inputs. The dataset is distributed as annual NetCDF files on the SMAP EASE-Grid 2.0 36 km grid for 2012-07-03 to 2023-12-31. Two additional static layers, OMEGA_median.nc and h_median.nc, provide the median values of ω and h retrieved from ASTRA-based τ-ω radiative transfer model inversion. Soil moisture is reported in volumetric units (m³ m⁻³).