Improving soil freeze–thaw retrieval from spaceborne L-Band measurements based on diurnal amplitude variation

Soil freeze–thaw (FT) cycles impact soil functions and atmosphere–land interaction, but accurate measurements are very limited. Since surface dielectric properties and microwave emissions are sensitive to the FT state, brightness temperature (TB) measurements at L-band allow retrieval of the FT state. We have demonstrated the potential of a soil FT retrieval algorithm from Soil Moisture Active Passive (SMAP) TB measurements. This retrieval algorithm is formulated regarding Diurnal Amplitude Variation (DAV), which is defined as the difference in TB observations of ascending and descending orbits. The DAV-FT algorithm uses globally fixed parameters. However, parameters should vary regionally considering factors like land cover type, terrain, and climate regions. We introduce Overall Classification Accuracy (OA) to characterize the extraction of DAV annual variation under different parameters. Then, the parameter optimization process, akin to maximum likelihood estimation, selects a combination of parameters to extract the annual variation of the DAV optimally. The DAV-FT algorithm uses optimized parameters, and the results show that compared to using fixed parameters, (a) the area with OA > 0.7 increases from 54.43% to 89.36%; (b) consistency with ERA5-Land and SMAP data has improved in southwestern North America, the Qinghai–Tibet Plateau, and southwestern Eurasia, with regions showing over 0.7 consistency reaching 81.28% for ERA5-Land and 79.54% for SMAP-FT; and (c) in situ stations with higher accuracy outnumber those with lower accuracy (48.11% versus 22.97% for fixed parameters, 35.14% versus 33.51% for SMAP FT). Furthermore, the algorithm achieves the highest median (0.92) and median accuracy (0.88), compared to fixed parameters and SMAP.