MLEF Applications to Parameter Estimation

1. MLEF Applications to Parameter Estimation Dusanka Zupanski CIRA/Colorado State University Fort Collins, Colorado CG/AR Research Progress Meeting August 16, 2005 Dusanka Zupanski, CIRA/CSU Zupanski@CIRA.colostate.edu

2. OUTLINE • Major sources of forecast uncertainty • Initial Conditions and Parameter estimation • Conclusions and future work Dusanka Zupanski, CIRA/CSU Zupanski@CIRA.colostate.edu

3. Major sources of forecast uncertainty • Initial conditions • Model errors (e.g., errors in dynamical equations, errors due to unresolved scales) • Parametric errors (errors in empirical parameters) • Forcing errors • Boundary conditions (e.g., lateral boundaries) • These uncertainties are not independent!  • All sources of uncertainty should be taken into account simultaneously within a unified mathematical approach.  • Verified analysis and forecast uncertainty

4. Initial Conditions and Parameter estimation RAMS (non-hydrostatic) model Korteweg-de Vries-Burgers model(KdVB) Nstate =101+1Nens =102Nobs =101 Nstate =54000+1Nens =50Nobs =7200 The method is applicable to simple and complex models. Dusanka Zupanski, CIRA/CSU Zupanski@CIRA.colostate.edu

5. CONCLUSIONS AND FUTURE WORK • Unified methodologies for estimating all major sources of analysis and forecast uncertainties are desirable. • Ensemble-based methods, such as the MLEF, can address multiple sources of uncertainties within the same mathematical (probabilistic) framework. • Preliminary results indicate that estimation of initial conditions and empirical parameters works well within the MLEF approach. • Further experiments with time varying empirical parameters and more complex model errors will be performed in the future. • Employ different dynamical models (atmospheric, carbon transport, hydrological, etc.). Dusanka Zupanski, CIRA/CSU Zupanski@CIRA.colostate.edu