Parameters Optimization Of An L-band Radiation Transfer Model,the CMEM,using The SMAP Brightness Temperature Observations

Soil moisture is an important parameter in the Earth’s water circulation system.Accurate estimation of the initial state of soil moisture is also important for numerical weather prediction and climate prediction.The data assimilation is an important method to obtain high-precision space-time continuous soil moisture.In direct assimilation,an important prerequisite for assimilation of brightness temperature observations into land surface models is that there is no bias between the brightness temperature observations and the brightness temperature simulated by a radiation transfer model.It is often very complicated and difficult to simulate the unbiased brightness temperature through a radiation transmission model on the global scale.One of the reasons is that the parameters as estimated by a radiation transmission model in a local experiment are not always suitable for simulating brightness temperature observations derived from the spaceborne radiometers.The uncertainty of the parameters of a radiation transfer model also affects the simulation of the top of the atmosphere brightness temperature and the inversion of soil moisture from the satellite brightness temperature observations.In response to these aformentioned problems,this thesis used an L-band radiation transmission model,the CMEM,to conduct an in-depth study of selecting parameterization scheme of the radiation transmission model(CMEM),optimizing the CMEM parameters,and quantizing the CMEM parameter uncertainty:Firstly,a CMEM parameterization scheme that simulates optimal brightness temperature on the global scale is selected.The correlation coefficient,normalized center rms error,mean deviation,and standard deviation ratio were used as the statistical metrics.Using these metrics and the SMAP brightness temperature data,the brightness temperature simulations obtained from different CMEM parameterization schemes were compared and analyzed.The CMEM parameterization scheme with the most optimal simulation of the SMAP brightness temperature observations on the global scale was determined by analyzing the sensitivity of each CMEM module.The results indicated that the roughness module of the CMEM is more sensitive to the simulated SMAP data than the vegetation optical thickness module and the soil dielectric module since the SMAP brightness temperature data simulated by the CMEM significantly varied for different roughness module parameters.The ability of the CMEM to simulate SMAP brightness temperature data has significant seasonal variation as the simulation performance from April to October is better than that from November to March at the horizontal polarization.Compared with the horizontally polarized brightness temperature,the vertically polarized brightness temperature data simulated by the CMEM has a higher correlation coefficient,a lower normalized center rms difference,an average deviation closer to zero,and a standard deviation ratio closer to one.The optimal CMEM parameterization scheme for simulating SMAP brightness temperature on the global scale is found: the roughness module adopts Wsimple parameterization,the soil dielectric module adopts Wang parameterization,and the vegetation optical thickness module adopts Jackson parameterization.Secondly,a CMEM parameter optimization study was carried out.The SCE-UA algorithm was used to optimize the CMEM parameters by reducing the long-term bias and the long-term standard deviation difference between brightness temperature simulated by the CMEM and brightness temperature obtained from the SMAP satellite at different polarizations and orbits.Then,the parameter optimization results was analyzed based on four aspects: brightness temperature data simulated by the CMEM before and after the parameter optimization,the optimized parameter values,the sensitivity to soil moisture for the CMEM before and after parameter optimization,and the optimization performance of the SCE algorithm.The results demonstrated that the long-termaveraged bias and long-term standard deviation difference between CMEM simulation and SMAP brightness temperature were reduced after the parameter optimization.Residuals were still found for the bias and standard deviation difference after the parameter optimization.However,these residuals are free from the parameters and reflected in the model structure simplification,the inaccuracy of the forcing data,and the errors in the brightness temperature observations.The results also revealed that the optimized parameters have more extensive meanings.The optimized effective parameter values not only reduced the parameter errors but also compensated for the residuals caused by the model structure simplification in the CMEM,the forcing data inaccuracy,and the errors of the SMAP brightness temperature observations.Research also found that the optimized parameters increased the sensitivity to soil moisture for the CMEM model.In the end,the uncertainty of quantitative CMEM parameters was studied.The Bayesian inference and MCMC simulation method were used to estimate the CMEM parameters by reducing the difference between the long-term mean and the standard deviation of the brightness temperature simulations and the SMAP brightness temperature observations in terms of dual orbits,the ascending and decending orbits,and dual polarizations,the horizontal and vertical polarizations.Consequently,the parameters uncertainty was estimated from the parameter posterior distribution.Two cases were used for the parameter estimation.The case 1 only estimated the parameters and the case 2 not only estimated parameters but also estimated the standard deviation of long-term bias residuals and the standard deviation of the standard deviation difference residuals.Results indicated that the vegetation structure parameters have the greatest uncertainty,and the uncertainty reached 62% of maximum value for the posterior probability parameter.The uncertainty of the roughness parameters reached 25% of maximum value for the posterior probability parameter.Vegetation single-scattering albedo had the lowest uncertainty,and its uncertainty was only 5% of maximum value for the posterior probability parameter.Morover,the estimated parameters were similar with respect to the numerical value and spatial distribution for the case 1 and case 2.The estimated parameter uncertainty in the case 2,however,was larger than that in the case 1.The residual standard deviation of brightness temperature simulated by the parameter with the maximum posterior probability was smaller than that as estimated by the DREAM algorithm.