Understanding past climate states is crucial for predicting future climate change. Validating climate models by applying them to climate states that were very different from today's, is important to
deepen our trust in models and hence, in their predictions for the future climate. The Last Glacial Maximum (19-23 ka BP, LGM) was a climatic state substantially different from today and the main
patterns of the ocean circulation during this time remain uncertain. Results from proxy data and models show significant differences. Combining climate models with observational data via inverse
modelling, i.e. data assimilation, is a powerful means to obtain more reliable estimates of the climate's state. Data assimilation is frequently used in the field of weather forecasting, but it is still
not well-established in the community of paleo-climatology.
In my PhD project I employ the Massachusetts Institute of Technology general circulation model (MITgcm) and work on the development and application of data assimilation methods to estimate the
LGM ocean state. A water-isotopes module is used which simulates water-isotopes in the whole water column such that d18O data reconstructed from benthic and planktonic formainifera can be assimilated.
The project consists of two parts. In the first part I utilize the so called adjoint method for variational data assimilation. The adjoint method is comparatively powerful and has been applied successfully in the past, but it requires so called “automatic differentiation” (AD) of the model code. The MITgcm was developed for this purpose but AD cannot be applied to many other models. Therefore, the second part of my project consists of the development and application of a new method which does not require AD. The new method will be applied to estimate the LGM ocean state and the results can be compared to results obtained from the adjoint method.