A methodology to improve daytime passive microwave soil moisture retrievals
Abstract. Passive microwave radiometer observations are crucial for global soil moisture (SM) monitoring, yet daytime retrievals are affected from significant uncertainties. The radiometric observations used for SM retrieval are the product of both surface emissivity and effective temperature (Tef f ). However, the precise estimation of Tef f is heavily affected by the strong vertical temperature gradients between the soil surface and the effective emitting layer. Accurate estimation of the emitting layer’s Tef f during SM retrieval using radiative transfer schemes like the Land Parameter Retrieval Model is crucial. In LPRM Tef f is estimated by a globally generic approach using the vertically polarized Ka-band (T BKaV ), which measures only the topmost layer of the surface, however, SM estimation is done by lower frequencies like the C-band. Radiation at C-band is emitted from deeper soil layers as compared to what T BKaV senses, causing a difference in the sensed temperatures, which becomes more apparent through the day, because strong solar illuminance heats up soil layers at different rates. The discrepancy between Tef f and surface temperature is a function of land cover type, which becomes particularly prominent in bare soils due to the significantly deeper penetration of lower frequencies compared to the Ka-band. To tackle this issue, we propose a novel approach: Globally, where the microwave polarization difference index (MPDI) is invariant from night-to-day, SM and vegetation optical depth (VOD) are assumed constant at this sub-daily timescale. Nighttime SM and VOD are then used as inputs to invert the radiative transfer equation and simulate Tef f . Linear regression between simulated Tef f and T BKaV over time-invariant locations then yields slope and intercept maps. These globally variable regression slopes and intercepts are then used with T BKaV to improve Tef f estimates which, as a consequence yield improved SM retrievals. The optimized daytime retrievals show globally reduced bias both on bare and vegetated land covers, when compared to nighttime retrievals and to ERA5-Land data. The method provides a robust bias reduction technique that can be used over multiple satellites for the C- and X-band frequencies.