Admissions > PhD by research > Research Projects > Parameter Estimation in Earth System Science: development of a robust method for non-linear model calibration

Parameter Estimation in Earth System Science: development of a robust method for non-linear model calibration

Supervisors: Dr. Marko Scholze and Dr. Tilo Ziehn

New NERC funded project

Earth system models and their components have rapidly increased in complexity in the last decade with a constant addition of processes increasing the non-linearity of these models. Processes implemented in these models are often parametrised and in most cases only a large uncertainty range can be given for individual parameters. Therefore, the uncertainty associated with them might substantially contribute to the overall model output uncertainty.

Quantitative parameter estimation through observational data assimilation techniques is an emerging area in Earth System science, which can help to reduce the uncertainties associated with the model parameters and model predictions. The challenge is that current estimation methodologies have various insufficiencies especially in the case of highly non-linear models. The most advanced methodology available in this field, four-dimensional variational assimilation (4D-Var), accumulates observed information into the model and estimates parameters via an optimisation procedure. However, case studies using a terrestrial biosphere model have shown, that the identified minimum in the physical parameter domain is not always a global minimum. The highly non-linear behaviour of the model makes it likely that the minimum is only local.

The student's work will build on the existing Carbon Cycle Data Assimilation System (CCDAS, http://www.ccdas.org and [1]). CCDAS is a modelling tool that uses atmospheric CO2 concentration observations and remotely sensed vegetation activity to constrain process parameters in a terrestrial ecosystem model. CCDAS is currently the only tool of its kind that is able to quantify terrestrial carbon fluxes complete with error bars fully consistent with the major global-scale observations.

The student will develop a novel approach combining the variational method with ensembles methods (e.g. Monte Carlo technique) to test the robustness of the optimisation. In order to derive a generic approach this will be done using a hierarchy, in terms of complexity, of models (CCDAS itself can be run at various levels of complexity). This approach will also potentially identify model weaknesses and point to areas for future model improvements.

The student engaged in this project will have the benefit of working as part of the National Centre for Earth Observation (NCEO, http://www.nceo.ac.uk). The NCEO aims to exploit data from Earth observation satellites to monitor and predict climate and environmental changes. The project is of wide interest to Earth observation and climate change institutions (such as research and academic institutions and government agencies and NGOs) because of its application on the global carbon cycle, but is also important basic science. .

Candidates should have a good honours degree in a subject with strong mathematical content. The closing date for applications is Friday 14th August.

References

  1. Rayner, P. et al., 2005, Two decades of terrestrial carbon fluxes from a Carbon Cycle Data Assimilation System (CCDAS), Global Biogeochemical Cycles 19, doi:10.1029/2004GB002,254.
Last updated: 4/2/09