|Title||Revision of the HIRS cloud detection at ECMWF|
|Series/Collection||EUMETSAT/ECMWF Fellowship Programme Research Reports|
|Authors||Krzeminski, B, Bormann, N, Kelly, GA, McNally, A, Bauer, P|
|Event Series/Collection||EUMETSAT/ECMWF Fellowship Programme|
|Place of publication||Shinfield Park, Reading|
HIRS radiances are assimilated routinely at ECMWF. Assimilation of clear-sky radiances requires cloudcontaminated observations to be identified and rejected. The old operational cloud detection procedure was found to leave residual cloud contamination in the assimilated scenes. In this document, we describe the recent revision of the HIRS cloud detection procedure. The HIRS cloud detection scheme aims at finding clear-sky channels within each field of view. It exploits the differences (fg-departures) between observed radiances and their simulated first guess clear-sky counterparts and interchannel gradients. Selected HIRS scenes were analysed to understand the cause of undetected cloud contamination. It was found that noise in the fg-departures often masked the presence of a cloud or was misinterpreted as a cloud signature. Smoothing was introduced to the fg-departures to reduce the noise. Thresholds used in the cloud detection have also been revised. The revised cloud detection is shown to detect cloud contamination more reliably, more consistent with the results from the IASI cloud detection. Departure histograms after cloud screening are more symmetric and comparisons with MSG imagery suggest that the revised algorithm detects clouds more reliably. Forecast experiments did not show statistically significant impact of the revision on the forecast accuracy. The revised cloud detection scheme was applied operationally from cycle 35r2.