|Institution:||University of Bremen|
|Room:||GEO building, room 5420|
|Phone:||+49 421 218 - 65448|
|Other webpage(s):||Charlotte's MARUM web page|
State estimates by inverse modeling using the MIT general circulation model
My PhD project is part of the PalMod project funded by the Federal Ministry of Education and Research (BMBF). The main goal of PalMod is to model the complete last glacial cycle, i.e. the last 120.000 years from the last interglacial to the Anthropocene. A good understanding of the past states of the Earth's climate and the underlying processes are crucial to make predictions for the future.
Modeling the climate is still a challenging task as there are a lot of uncertainties, for example regarding the initial conditions, boundary conditions or the parameterization of sub-grid-scale processes. This is especially true for the modeling of a paleo climatic states. Another way of learning about past climate states is by looking at proxy data. But paleo proxies usually have a high spatial sparsity and high uncertainties in age and in the proxy value. Furthermore, fields which are extrapolated from proxy data with a statistical method are not consistent with the physics of the climate system. Combining a model and proxy data via data assimilation/inverse state estimation produces estimates which are consistent with the underlying model physics and also with the proxy data and therefore give a good chance to attain a better estimate with less uncertainties. It is important to develop different data assimilation methods for different climate models to test their abilities and limitations.
In my project I work with a global configuration of a coupled ocean-sea-ice model: the MIT general circulation model (MITgcm). The MITgcm gives the possibility to use the so called adjoint method to perform data assimilation. This technique needs the adjoint of the model code which can be produced by so called automatic differentiation. The MITgcm was developed in such a way that this is possible for its model code but automatic differentiation cannot be applied to a conventional ``forward'' model such as the Community Earth System Model (CESM). Therefore, the main goals of my project are firstly to use the adjoint method for state estimates and secondly to implement a data assimilation technique into the MITgcm which can also be used in the CESM.
The estimates shall be performed by assimilating water-isotopes data from the pre-industrial state and specific times in the last glacial cycle, e.g. from the last glacial maximum (LGM). The results of the newly implemented data assimilation technique can be compared to results obtained from utilizing the adjoint method. Subsequently, the same data assimilation method will be implemented into the CESM by a colleague and the results from my project can be used as a benchmark.
|Prof. Dr. Michael Schulz||University of Bremen|
|Dr. André Paul||University of Bremen|
|Dr. Ute Merkel||University of Bremen|
|Dr. Martin Losch||Alfred Wegener Institute, Helmholtz Center for Polar and Marine Research (AWI), Bremerhaven|