Testing variational estimation of process parameters and initial conditions of an earth system model

Abstract

We present a variational assimilation system around a coarse resolution Earth System Model (ESM) and apply it for estimating initial conditions and parameters of the model. The system is based on derivative information that is efficiently provided by the ESM's adjoint, which has been generated through automatic differentiation of the model's source code. In our variational approach, the length of the feasible assimilation window is limited by the size of the domain in control space over which the approximation by the derivative is valid. This validity domain is reduced by non-smooth process representations. We show that in this respect the ocean component is less critical than the atmospheric component. We demonstrate how the feasible assimilation window can be extended to several weeks by modifying the implementation of specific process representations and by switching off processes such as precipitation.

• data assimilation
• climate modelling
• coupled ocean–atmosphere model
• earth system model
• automatic differentiation
• adjoint model

7. Acknowledgements

This work was supported by the European Community within the 6th Framework Programme for Research and Technological Development under contract no. 212643 (THOR) to Hamburg University and supported in part also through the DFG funded CLISAP excellence initiative. The presented research is building on work on an assimilation system developed around only PlaSim that was funded by NERC through its QUEST programme. M. Scholze contributed through that phase and acknowledges support through a CLISAP fellowship. P. Herrmann was supported through a Max-Planck Fellow Grant to D. Stammer. The authors wish to thank E. Kirk for his advice throughout this study and valuable comments on the manuscript.

Notes

¹Available via http://www.cen.uni-hamburg.de/en/research/cen-models/cesam.html.