Abstract |
Numerical weather prediction models include a set of parameterizations that represents the effect of small scale or complex phenomena that cannot be modeled explicitly. Associated to this parameterizations there is a relatively large number of parameters whose value is a-priori unknown. Finding an optimal value for these parameters (i.e. the value that produces a representation of the atmospheric circulation which is close to the observed one) is a numerically intensive and difficult task. On the other hand, data assimilation consists of a group of methodologies that combines numerical weather predictions with observations to provide an estimation of the state of system (i.e. the atmosphere, the ocean, etc). Data assimilation has been also extended in order to use the available observations to find the optimal value of the uncertain parameters in a model. In this talk a description of how parameters are estimated using ensemble based data assimilation techniques will be provided. Examples of parameter estimation will be presented using a simple general circulation model and an state-of-the-art numerical weather prediction model and the impact of the estimated parameters upon the short range forecast and climate simulations will be discussed.
|