ESTIMATION OF CRITICAL PARAMETERS IN THE SMOS LAND RETRIEVAL ALGORITHM

S. Mike^1,2, Y. Zhou¹, A. Mialon³, M. Holmberg⁴, R. Crapolicchio⁵, Y. Kerr³

¹Federal Research Institute for Forrest, Snow and Landscape Research WSL, ²GAMMA Remote Sensing Research and Consulting AG, ³CESBIO, Université de Toulouse, ⁴Finnish Meteorological Institute, Earth Observation Research, ⁵ESA-ESRIN

The SMOS land retrieval algorithm simultaneously derives soil moisture (SM) and vegetation optical depth (VOD) from L-band Brightness Temperatures (BT) measured at multiple observation angles and different polarizations. Retrievals are obtained by minimizing the mismatch between observed SMOS BTs and simulated forward operator BTs. In operational retrievals, simulated brightness temperatures (BTs) rely on the L-band Microwave Emission of the Biosphere (L-MEB) model, which incorporates the Tau-Omega (TO) Microwave Emission Model (MEM). Here, however, we employ an upgraded version of L-MEB that uses the Two-Stream (2S) MEM to simulate BTs over vegetated surfaces. The 2S-MEM has a wider range of applicability and includes the TO-MEM when the single-scattering albedo is set to ω = 0.

Independent of whether TO-MEM or 2S-MEM is used, the forward operator (the MEM) requires additional parameters beyond the retrieved SM and VOD. These include effective soil and vegetation temperatures, the H-Q-N roughness model parameters (h, n), and the single-scattering albedo (ω), which characterizes vegetation scattering relative to extinction. This study investigates different approaches to estimate (h, n, ω), and compares their plausibility with the operational values currently used in SMOS retrievals. In addition to SMOS BTs, the methods incorporate SM simulated by the European Centre for Medium-Range Weather Forecasts (ECMWF). The finally estimated parameters (h, n, ω) are differentiated and analyzed according to their values within various WWF ecoregions and IGBP land cover types. This strategy of spatial parameter allocation leverages the complementarity between the WWF classification, which is ecological and biodiversity-oriented, and the IGBP classification, which is satellite-based and land-cover-oriented for climate modeling.

Temporal variability in the estimated parameters is also analyzed. Pronounced seasonal cycles are identified and quantified through Fourier analysis of the time series h(t), n(t), and ω(t) of estimated critical retrieval parameters. Results suggest that accounting for seasonally varying h and w could enhance SMOS retrievals of (SM, VOD). However, this raises the question of what constitutes “improvement” in the context of SM retrieval; whether retrievals should be assessed primarily by their agreement with sparse in-situ measurements, or by other quality indicators that are more directly related to the water fluxes between the soil and the lower atmosphere.

To evaluate potential benefits of the estimated critical retrieval parameters, we examine whether VOD retrieved with the newly estimated parameters (h, n, ω) provides improved information compared to VOD retrieved with standard parameter settings. This assessment includes comparisons with Above Ground Biomass (AGB) and an analysis of the temperature dependence of boreal forest VOD.