Comparison of model-derived and radar-observed freezing-level heights: Implications for vertical reflectivity profile-correction schemes

In the current operational Met Office scheme for deriving rainfall rates from radar, the height of the enhanced radar return associated with melting snow, the bright band, is determined using the height of the 0 degreesC isotherm obtained from the Met Office Unified Mesoscale Model (UM) forecast. In this paper the potential errors of using model forecast heights as input to the bright-band correction scheme are investigated. The UM and European Centre for Medium-Range Weather Forecasts (ECMWF) model forecasts of wet-bulb 0 degreesC isotherm heights (WBZ) are compared with the height of the 'step' increase in reflectivity in the vertical profiles recorded by the vertically pointing 94 GHz Galileo cloud radar at Chilbolton. High-frequency radars do not measure an enhanced 'bright band' at the melting layer, rather a sudden increase of reflectivity as the ice particles become coated in water. This sudden step can be used to locate accurately the height of the WBZ. Results show that the UM predicts the WBZ height with a root-mean-square error of 147 m and a bias of 15 m. This is within the worst-case tolerance of 200 m of the operational bright-band correction scheme. Factors that may influence the accuracy are the presence of deep isothermal layers and timing errors concerning the passage of fronts, but these are found not to be serious. An investigation into the deterioration of the UM 36-hour forecast shows that half the error is introduced at the initialization time, highlighting the fact that further improvements can be achieved through a better definition of the initial state of the atmosphere. The ECMWF forecast is for a longer lead time, but comparison for the same lead-time errors as the UM shows that the performance of the two models is comparable. An alternative approach is to derive the bright-band height from volumetric radar scans at different elevations. This study suggests that, at least in the UK, operational model predictions of the freezing-level height are within the specified 200 m error, but that the use of volumetric scans, even under idealized conditions, cannot achieve this accuracy.