On the usefulness of additional cloud and tracking information for the assimilation of AMVs

On the usefulness of additional cloud and tracking information for the assimilation of AMVs
Date Published
EUMETSAT/ECMWF Fellowship Programme Research Report
Document Number
Francis Warrick
Abstract Currently the vast majority of available AMVs are not assimilated into NWP models due to various quality control steps. The aim of this study is to examine whether additional information on the properties of the tracked clouds, or estimates from within the AMV derivation on the reliability of the tracking and height assignment steps, can be used to extend the use of AMVs in global NWP and improve the impact they have on NWP forecast accuracy.
Three sources of additional information available from the AMV derivation were considered with a view to enhance the assimilation of AMVs in global NWP. The additional information characterises the reliability of the pixel-level cloud-top pressure estimate used for height assignment in EUMETSAT AMVs, and the cloud type or phase of the scene, and variability of the tracked motion, in NOAA AMVs. Departure statistics are evaluated in order to see whether the additional information is able to discriminate between different departure regimes which would allow a refinement of either the quality control or observation-error modelling.
Firstly, a test dataset was supplied by EUMETSAT which applied filtering to pixels with a less reliable cloud-top pressure estimate from EUMETSAT’s Optimal Cloud Analysis (OCA) product. This was found to be effective at reducing background departures of EUMETSAT’s Meteosat-11 AMVs compared to the unfiltered dataset. Preliminary assimilation trials with the one-month dataset available showed some promising benefits for short-range forecasts in the tropics, though a longer trial would be required to corroborate this finding.
Secondly, an evaluation of NOAA AMVs showed that for some areas AMVs with a different cloud type or phase had different background departures to each other. However, generally it was found that the type of cloud is already implied by the AMV location and pressure and therefore mostly already accounted for in the current quality control choices or observation error modelling.
Thirdly, information on the consistency of tracking clusters in GOES AMVs was also looked at, it was found that in some areas the more consistent the cluster vectors the lower the background departures were, however in other areas the reverse was true. It is hence not straightforward to use this quantity to refine the assimilation of AMVs.
The study hence recommends that EUMETSAT generates a longer dataset that would allow a deeper investigation of the applied filtering and may ultimately lead to a useful refinement of the EUMETSAT AMV processing. A further recommendation is that additional cloud and tracking information should be included by EUMETSAT and other AMV producers within the AMV product to allow NWP centres to try using this information to further improve AMVs’ contribution to forecast performance.
URL https://www.ecmwf.int/en/elibrary/81564-usefulness-additional-cloud-and-tracking-information-assimilation-amvs
DOI 10.21957/1c5cfdf425