Using data assimilation to improve climate models: Model Error and parameter estimation

PULIDO M.

Petropolis, Rio de Janeiro, Brazil

Workshop; CLIVAR-VAMOS Workshop on modeling and predicting climate in the Americas; 2012

WCRP

Data assimilation can be used to determine sources of climate model errors. We apply 4D variational assimilation principles to estimate the missing subgrid momentum forcing of a simple general circulation model. The technique is evaluated using twin experiments. Then, some estimations of the missing forcing from analysis are presented. In order to improve models, data assimilation techniques is also used to estimate the optimal parameters in a simple general circulation model. I will present two successful examples: the estimation of gravity wave parameters using a genetic algorithm and the estimation of convective parameters using ensemble Kalman filtering in a simple general circulation model. In the last part of the talk I will show some of our current efforts to use ensemble Kalman filtering to estimate model error, in particular a missing momentum forcing in a simple chaotic model, the Lorenz 96 model.