Please use this identifier to cite or link to this item: http://hdl.handle.net/1959.14/139082
0 Visitors0 Hits0 Downloads
Assimilating reflectance data into a ecosystem model to improve estimates of terrestrial carbon flux
International Symposium on Physical Measurements and Spectral Signatures in Remote Sensing (10th : 2007) (12 - 14 March 2007 : Davos, Switzerland)
Schaepman, Michael E.; Liang, Shunlin; Groot, Nikée E. and Kneubühler, Mathias. Proceedings of the 10th International Symposium on Physical Measurements and Spectral Signatures in Remote Sensing (ISPMSRS'07), p.1-6
Ecosystem models are valuable tools for understanding the growth of vegetation, its response to climatic change and its role in the cycling of greenhouse gasses. Data Assimilation (DA) of synoptic coverage Earth Observation (EO) data into ecosystem models provides a statistically optimal mechanism for constraining the model state vector trajectory both spatially and temporally. EO "products" such as leaf area index (LAI) are attractive candidates for assimilation, but it is difficult to assign accurate uncertainty estimates to such products (a critical requirement of DA) and, more importantly, they are derived on the basis of assumptions that may be contradictory to those in the ecosystem model. An attractive alternative, therefore, is to assimilate reflectance data; the uncertainty in which is more easily understood. The assumptions made in generating the reflectance data are independent of assumptions in the ecosystem model and may consequently be treated as additional sources of uncertainty. To achieve this it is necessary to build a canopy reflectance model into the assimilation scheme. This paper describes the coupling of a canopy reflectance model to a simple ecosystem model. Reflectance data are assimilated over a boreal forest and improvements in predicted carbon fluxes are shown with comparison to field data. Previous work has highlighted problems of lost samples due to snow cover, resulting in poorly constrained flux estimates during winter months. This issue is addressed by incorporating a snow reflectance model. Results utilising the EnKF as a parameter estimator are also discussed.