Spatio-temporal dynamics of L-band zeroth-order vegetation scattering albedo from SMAP observations in tropical forests

Abstract

The effective scattering albedo (ω) is a key parameter in the zero-order radiative transfer equation (known as the τ-ω model) for passive microwave retrieval of soil moisture (SM) and vegetation optical depth (VOD), quantifying the scattering energy loss as microwave radiation passes through the vegetation canopy. The scattering effects of vegetation are influenced by time-dependent factors such as plant geometry, vegetation water content, and canopy structure, suggesting that ω may vary over time. However, in the current τ-ω model-based retrieval algorithms used by orbiting L-band sensors, namely the Soil Moisture and Ocean Salinity (SMOS) and the Soil Moisture Active Passive (SMAP), ω is generally assumed to be time-invariant and assigned a fixed value according to land cover types. In this study, we aim to analyze and understand the spatio-temporal dynamics of ω, its relationship with vegetation water stress and the driving factors behind microwave scattering characteristics over tropical forests. By assuming a vegetation transmittance of zero for rigorously selected dense forest areas in the tropics, we calculated ω from the SMAP L-band radiometer observations during 2018–2023. Regarding the spatial distribution of ω, we observed distinct spatial dynamics within the same land cover type. The lowest ω values were typically found in the northeastern Amazon. Additionally, ω exhibited clear temporal dynamics, displaying a unimodal pattern in the Amazon and a bimodal pattern in the Congo. Clear polarization dependence of ω was observed, with values consistently higher at Horizontal (H-) polarization compared to Vertical (V-) polarization. Despite this, the seasonal patterns of ω are similar at both H- and V-polarizations. The seasonal variation of ω was found to be asynchronous with soil water availability indicated by root zone soil moisture (RZSM) across different regions. The shortest time lags (0–30 days) between ω and RZSM were observed in the densely vegetated northeastern Amazon, while the longest occurred in the northeastern Congo. A machine-learning based interpretation of the spatial variability of ω and time lag indicates that the values of ω are strongly and inversely related to canopy height, while the time lag is mainly associated with precipitation and soil water content. Our results deepen the understanding of the spatio-temporal dynamics of ω and could contribute to the improvement of SM and VOD retrieval algorithms, thereby enhancing the utility of these variables as indicators for monitoring vegetation carbon dynamics and phenology in dense tropical forests.