Error characterization of L-band passive microwave radiometry in soil moisture and VOD retrievals

Soil Moisture (SM) and vegetation are the two critical drivers of land-atmosphere interactions. Low-frequency passive microwave remote sensing has enabled SM and vegetation dynamics estimation at global scales by analyzing the interactions of their emissions using a radiative transfer model (RTM). Vegetation is characterized in an RTM as vegetation optical depth (VOD or ) and single scattering albedo (). In the presence of dense vegetation, microwave emissions are subjected to multiple scattering that must be accurately modeled by an RTM. Research is underway to improve vegetation representation in an RTM to achieve accurate SM and VOD retrievals. Understanding the relative strengths of various RTMs coupled with the specific impacts of RTM model parameterization on SM and VOD retrievals remains a grey area of research. In this work, we attempt to characterize errors propagated from RTMs and their parameterization while retrieving SM and VOD from L-band brightness temperature data from Soil Moisture Active Passive (SMAP) mission at 10 sites situated across various biomes. Three RTMs are considered, including Mo et al.s - model, Schwank et al.s 2-Stream (2S) model, and Feldman et al.s 1st order model. The surface roughness (h) and parameters are considered to study the effects of model parameterization. SM and VOD are retrieved concurrently using a multi-temporal RTM inversion scheme. A large bias is observed among the retrievals due to the variation of h, and RTM. The 2S model SM retrievals performed slightly better than other two models. The performance of VOD retrievals are assessed by comparing with other recently developed L-band VOD products and optical vegetation indices. A statistical approach is used to characterize the propagation of bias and uncertainty contributions from h, and RTM on SM and VOD retrievals. The source of error in the retrievals is found to be predominantly from followed by RTMs. The uncertainty due to increases under dense vegetation conditions. In vegetated regions, VOD retrievals are more sensitive to model parameters and choice of RTM than SM. This study can contribute to assessing errors due to h, and RTM at different frequencies, which are useful for developing collaborative retrieval algorithms for the upcoming Copernicus Imaging Microwave Radiometer (CIMR) mission.