Uncertainty quantification of GEOS-5 L-band radiative transfer model parameters using Bayesian inference and SMOS observations

• Uncertainty in large-scale radiative transfer model (RTM) parameters is sampled using MCMC simulation.
• Differences between GEOS-5 and SMOS brightness temperatures (Tb) are minimized.
• Maximum a posteriori density (MAP) parameter values found with MCMC and PSO do agree.
• Relative posterior uncertainty is less than 25% of MAP value for RTM parameters.
• Estimated total Tb simulation and observation errors are consistent with actual misfit.

Uncertainties in L-band (1.4 GHz) microwave radiative transfer modeling (RTM) affect the simulation of brightness temperatures (Tb) over land and the inversion of satellite-observed Tb into soil moisture retrievals. In particular, accurate estimates of the microwave soil roughness, vegetation optical depth and scattering albedo for large-scale applications are difficult to obtain from field studies and often lack an estimate of uncertainty. Here, a Markov Chain Monte Carlo (MCMC) simulation method is used to determine satellite-scale estimates of RTM parameters and their posterior uncertainty by minimizing the misfit between long-term averages and standard deviations of simulated and observed Tb at multiple incidence angles, at horizontal and vertical polarizations, and for morning and evening overpasses. Tb simulations are generated with the land model component of the Goddard Earth Observing System (version 5) and confronted with Tb observations from the Soil Moisture Ocean Salinity satellite mission. The maximum a posteriori density (MAP) parameter values reduce the root-mean-square differences between observed and simulated long-term Tb averages and standard deviations to 3.4 K and 2.3 K, respectively. The relative uncertainty of the posterior RTM parameter estimates is typically less than 25% of the MAP parameter value, whereas it exceeds 100% for literature-based prior parameter estimates. It is also shown that the parameter values estimated through Particle Swarm Optimization are in close agreement with those obtained from MCMC simulation. The MCMC results for the RTM parameter values and the uncertainties presented herein are directly relevant to the need for accurate Tb modeling in global land data assimilation systems.