Error Characterisation of the ESA Essential Climate Variables
European Space Agency, UNITED KINGDOM
The ESA Climate Change Initiative aims to provide long-term satellite-based products for climate. Error characterisation of remote sensing measurements is a complicated issue which is especially important for climate applications where long-term, consistent records of the Earth system are required. Although most methods used to derive satellite products assume that errors in the measurements and prior data are random and zero in the mean, errors may in fact be systematic rather than random and arise from a variety of sources which may be difficult to characterise, such as uncertainty in instrument calibration, forward model parameter errors or approximations in the representation of the physics of the radiative transfer calculations. These errors can be estimated a posteriori through error propagation analyses or validation against other measurements, but this is not a simple task, and such errors can easily obscure the useful climate signal. Here we outline how the common assumption in satellite retrievals that the measurement error is random and unbiased can be relaxed and suggest how the optimality of satellite retrievals could potentially be improved and the retrievals made bias-aware using statistical information contained in the measurements and prior data, potentially reducing systematic errors. Drawing on techniques developed in the field of data assimilation for the optimal combination of satellite data and model predictions in the presence of systematic errors in the measurement or forecast, we suggest how a fully characterised post-retrieval adjustment could be applied to the retrieved products.