Estimation of parameters in climate models using data assimilation

PULIDO M

Montevideo

Workshop; FAPESP week; 2016

AUGM FAPESP

Data assimilation is the statistical combination of observations from different instruments and information from model integrations to obtain a state of the system as close as possible to the state of the atmosphere. The most important applications of data assimilation techniques are to use the analysis, i.e. state estimation, as initial conditions for weather forecasts and as the state of the atmosphere for climate studies. A new area that is emerging is the application of data assimilation concepts to improve numerical models. I will talk on the work of our research group during last years which has been focussed on the use of techniques based on variational assimilation and ensemble Kalman filter to estimate parameters of physical schemes of atmospheric numerical models.  Results of the estimation of parameters using data assimilation techniques will be presented for a non-orographic gravity wave scheme (Pulido et al 2012), a convection scheme (Ruiz et al 2013), and a subgrid-scale orography scheme (Tandeo et al 2015). There two major challenges in parameter estimation: model error and the strong nonlinearities associated to the paremeters. The Expectation-Maximization algorithm combined to the ensemble Kalman filter appears to partially overcome these difficulties but more research is needed, also particle filters look a promising technique for parameter estimation in nonlinear regimes.  Recently a combination of the ensemble Kalman filter with an offline regression technique was succesfully used to uncover the functional form of subgrid-scale schemes (Pulido, et al 2016). These early results show data assimilation has a great potential for model development which is still largely unexplored.Pulido M., S. Polavarapu, T. Shepherd and J. Thuburn, 2012. Q. J.  Roy. Meteorol. Soc.,  138, 298--309. DOI: 10.1002/qj.932.Pulido M., G. Scheffler, J. Ruiz, M. Lucini and P. Tandeo, 2016: In press in  Q. J.  Roy. Meteorol. Soc. DOI: 10.1002/qj.2879.Ruiz J., M. Pulido and T. Miyoshi, 2013.  J. Meteorol. Soc. Japan.  91, 79--99.Tandeo P., M. Pulido and F. Lott, 2015. Q. J.  Roy. Meteorol. Soc., 141, 383--395.