Skip to main content
ARS Home » Research » Publications at this Location » Publication #280825
Title: Strategies for minimizing the impact of systematic errors on land data assimilation

Author
item Crow, Wade
item Yilmaz, Mustafa
item HAN, E - Science Systems, Inc

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 6/2/2012
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: Data assimilation concerns itself primarily with the impact of random stochastic errors on state estimation. However, the developers of land data assimilation systems are commonly faced with systematic errors arising from both the parameterization of a land surface model and the need to pre-process observations to remove systematic differences commonly found in modeled versus observed land surface states. This poster will provide and illustrate two examples of these challenges. First, the analytical basis underlying the pre-assimilation rescaling of soil moisture observations will be developed and compared to empirical re-scaling approaches commonly applied (in a pre-processing step) during the assimilation of remotely-sensed surface soil moisture retrievals into land surface models. Conditions under which these empirical re-scaling procedures do and do not conform to this analytical basis will be clarified and the impact of such non-conformity on assimilation results quantified. Second, the impact of uncertainty in two key parameters: surface soil moisture storage capacity and the effective water-flux vertical diffusivity will be examined to clarify the impact of errors in these parameters on the ability of surface soil moisture retrievals to constrain deeper soil moisture states using sequential data assimilation. These results will lead to specific conclusions regarding strategies for minimizing the impact of miss-specifying either parameter on land data assimilation results. Combined, results from both cases will be used to define new land data assimilation approaches which are most robust to the potential degrading impact of systematic errors in land surface models and observations.