For many historic Earth Observation (EO) datasets that are key to climate science and meteorological re-analysis, uncertainty information is absent, generic, unrealistic and/or partial. This is true both for Fundamental Climate Data Records (FCDRs – i.e., observed radiances) and [Thematic] Climate Data Records (CDRs — containing geophysical products). This is clearly unsatisfactory: long-term records of meteorological and climate variables from space-based observations need to provide trustworthy information about variability and change over decades, so that they can be used for rigorous science, decision-making and climate services (including re-analysis).
Improved realism and rigour are needed in the generation of FCDRs and CDRs in relation to stability and uncertainty information, and this is the aim of the H2020 project "Fidelity and Uncertainty in Climate data records from Earth Observation" (FIDUCEO).
Defensible uncertainty estimates need to be realistic (demonstrably not underestimating uncertainty) and traceable (obtained by assessing and propagating all known effects that introduce uncertainty). FIDUCEO will develop realistic, traceable uncertainty information for four FCDRs (AVHRR, HIRS, MW humidity sounders, Meteosat VIS). New versions of these FCDRs with harmonised calibration across the full sensor series will be developed in a common format that supports calculation of observation error covariances or alternative ensemble members. Example geophysical CDR products will be derived from each FCDR, with uncertainty estimates obtained by uncertainty propagation. Beyond the products directly generated, the project aims to demonstrate and promote better handling of uncertainty in EO datasets and applications by providing cookbooks and tools.
The seminar will give an overview of the project, its rationale and its approach. How in practice metrological principles apply to EO will be illustrated with reference to one FCDR. Aims include stimulating discussion on how improved uncertainty information can be relevant within data assimilation, and highlighting the opportunities to be involved in trail blazing applications of FIDUCEO data.