Dynamic vegetation parameter retrieval algorithm for SMAP L-band radiometer observations

Abstract

Vegetation Optical Depth (VOD), obtained from passive microwave sensors, quantifies Vegetation Water Content (VWC) and complements conventional vegetation indices. Recent studies on Soil Moisture (SM) and VOD retrieval algorithms identified that VOD is more susceptible to errors due to the Radiative Transfer Model (RTM) and its parameterization than SM. The present work aims to address this limitation. We initially characterized the error propagation from ω and h parameters in VOD through synthetic experiments. These experiments also indicate notable propagation of errors from assuming a temporally constant ω in VOD retrievals, which could be resolved using a time-varying ω parameter.

To improve the VOD characterization, we proposed a Dynamic Vegetation Parameter retrieval Algorithm (DVPA) to retrieve VOD and ω simultaneously, along with a temporally constant h parameter applied to L-band SMAP brightness temperatures. DPVA is based on the Two-Stream emission model (2S-EM) RTM. Retrievals are obtained using a novel multi-temporal inversion coupled with a regularization scheme. SMAP Level-3 SM is supplied as one of the critical inputs. DVPA, as a proof-of-concept, is applied to ten reference sites with varying vegetation conditions. The retrieved VOD and ω from DVPA are compared with optical vegetation indices and SMAP baseline VOD product (Regularized Dual Channel Algorithm-RDCA). DVPA VOD estimates outperform SMAP RDCA VOD in terms of correlation (R) and lagged correlation with vegetation indices. Regularization ensured optimum filtering of retrieval noise from the VOD retrievals. Retrieval of dynamic ω helped to resolve errors in VOD, resulting in improved correspondence with vegetation growth patterns compared to SMAP baseline VOD retrievals. Given its generic structure, DPVA is scalable and applies to other passive microwave sensors.