Code 614.3 LIS Science Seminar

Dr. Gabriëlle J. M. De Lannoy, Post Doctoral Research Fellow, Center for Research on Environment and Water, Calverton, MD
Abstract BIAS FILTER: A fundamental assumption of the Kalman filter is that both the observations and model predictions are unbiased. Land surface models are usually biased in at least a subset of the simulated variables even after calibration. On-line forecast bias estimation may therefore be needed for data assimilation. In situ soil moisture observations in a small agricultural field (OPE3) were merged with Community Land Model (CLM2.0) simulations using different algorithms for state and bias estimation with and without bias correction feedback. The different bias correction schemes were tested to study the impact of the state correction on depending model fluxes. The best variant for state and bias estimation depends on the nature of the model bias: an improper bias correction scheme could distort the water balance. The lack of knowledge of the bias `dynamics’ in time and space and the approximation of the bias uncertainty structure limit successful bias estimation and correction to directly observed state variables. However, all assimilation schemes including bias correction algorithms yield far improved state analysis results compared to standard state filter analyses. ADAPTIVE FILTER: Data assimilation aims at providing an optimal estimate of the overall system state, not only for an observed state variable or location. However, large scale land surface models are typically column-based and purely random ensemble perturbation of states will lead to block-diagonal a priori error covariances. This facilitates the filtering calculations, but compromises the potential of data assimi lation to influence (unobserved) neighboring state variables. Here, a combination of an ensemble Kalman filter and an adaptive covariance correction method is explored to optimize the variances and retrieve the off-block-diagonal correlations in the a priori error covariance matrix. In a first time period, all available soil moisture profile observations in a small agricultural field are assimilated into the CLM2.0 land surface model to find the adaptive second order a priori error information. After that period, only observations from single individual soil profiles are assimilated with inclusion of this adaptive information. It is shown that assimilation of a single profile can partially rectify the incorrectly simulated soil moisture spatial mean and variability. The largest reduction in the root spatial mean square error in the soil moisture field varies between 7 and 22%, depending on the soil depth, when assimilating a single complete profile every 2 days during 3 months with a single time-invariant covariance correction.