Land Surface Data Assimilation: Improved Results with Multivariate Assimilation Including Parameter Estimation.

Abstract:

Sequential data assimilation in combination with land surface models has traditionally focused on assimilation of soil moisture contents to improve the initial conditions of atmospheric circulation or rainfall-runoff models. Limitations of this approach are related to the quality and resolution of remote sensing data and the representation of subsurface flow in the land surface model, amongst others. This presentation focuses on ways to improve predictions with land surface models using sequential data assimilation. One approach is multivariate data assimilation, where different data types are jointly assimilated. We will show two examples; one is the joint assimilation of land surface temperature and brightness temperature, the second example is the joint assimilation of neutron counts and land surface temperature. A second approach to improve predictions is estimation of land surface parameters in a consistent stochastic framework. This will be illustrated with a synthetic example. A third approach is by using a more sophisticated land surface model. A sequential data assimilation framework implemented together with the land surface model ParFlow-CLM in a high performance computing environment will be presented, and its performance illustrated for a real-world example.