ESA Data Assimilation Projects: An Earth Observation Land Data Assimilation System for Multiple Wavelength Domains
Quaife, Tristan1; Lines, Emily2; Davenport, Ian1; Styles, Jon3; Lewis, Philip2; Gurney, Robert1
1University of Reading, UNITED KINGDOM; 2UCL, UNITED KINGDOM; 3Assimila, UNITED KINGDOM
Land surface models are important components of many atmospheric circulation models used to predict weather and climate. There has recently been much focus on using Data Assimilation techniques to integrate Earth Observation data with land models to improve estimates of both state and parameters. Typically, however, the level of physical detail used to represent the interaction of electromagnetic radiation with the surface in such models is not sufficient to enable prediction of intrinsic satellite observations (reflectance, brightness temperature and so on) and consequently these are not assimilated directly into the models. A seemingly attractive alternative is to assimilate high-level products derived from satellite observations (such as leaf area index or photosynthetic flux) but these are often only superficially related to the corresponding variables in land surface models due to conflicting assumptions between the two. Furthermore Data Assimilation requires well founded estimates of the uncertainties in observations and the errors in such products are often poorly characterised.
Atmospheric Data Assimilation communities have made significant progress over the last two decades by assimilating radiance measurements from satellites, rather than making retrievals of the variable of interest and assimilating these. There is some evidence to suggest that a similar approach may be beneficial to the land surface Data Assimilation problem. To achieve this it is necessary to use a forward operator to predict the necessary satellite observations directly from the land surface model. However, it is also likely necessary to replace existing radiative transfer schemes inside the land surface models themselves so that they are physically consistent with the forward operator. In addition it is desirable to have a scheme that is able to handle observations from any part of the electromagnetic spectrum that data is likely to come from (optical, thermal or microwave).
This paper describes a project funded by the European Space Agency to develop a data assimilation scheme for the land surface and forward operators to translate between models and the intrinsic observations acquired by satellite missions. A forward operator using a hybrid Geometric Optic Radiative Transfer approach to predict forest canopy reflectance at optical wavelengths has been adapted so that it is capable of partitioning shortwave radiation into reflected and absorbed (vegetation and soil) components required by land surface models. Furthermore this forward operator is capable of describing the dependency of satellite observed land surface temperature as a function of viewing angle and is thus capable of assimilating measurements from the thermal infra-red region. The operator has been coupled to a simple land surface scheme. The rationale behind the design of the underlying model is to represent the physics of the water and energy balance in as parsimonious manner as possible, using a force-restore approach, but describing the physics of electromagnetic radiation scattering at the surface sufficiently well that it is possible to assimilate the intrinsic observations made by remote sensing instruments. Results are shown from initial Data Assimilation experiments based at flux tower sites.