The variation of reflectivity with height is understood to be one of the most significant sources of error affecting the quantitative accuracy of surface precipitation estimates derived from radar data, particularly at long range. The radar samples at increasing altitude with range and reflectivity measurements aloft require correction to account for the VPR between the measurement height and the surface.
The operational VPR correction method currently employed as part of the Met Office radar processing chain uses an iterative approach to fit an idealised reflectivity profile shape to data measured by the lowest usable elevation scan available at each pixel only. The profile shape is constructed separately for each radar pixel using parameterisations of the bright-band and low-level orographic growth, model-derived values of the freezing level and satellite-derived cloud-top heights. Different profile shapes are used if the sampled precipitation type is diagnosed as being rain, graupel, snow or warm rain. Considerable bias errors may result at long range when the true observed reflectivity profile above the bright band deviates significantly from the assumed VPR.
A modified scheme is described which aims to reduce the bias errors at long range by deducing the VPR shape above the bright band from reflectivity measurements at all available radar scan elevations. All radars in the UK network scan at least 5 different elevations to long range at angles between 0° and 4°. The profile shape above the bright band is prescribed by first identifying distinct rainfall segments within a radar scan and using the available radar data in each segment to compute an average reflectivity profile between the diagnosed freezing level and cloud top heights. The derived gradient of reflectivity with height for each rainfall segment is then applied to construct the idealised VPR profile at each pixel in that region. The new method maintains the physically-based pixel-by-pixel approach currently used to reflect the spatially inhomogeneous reflectivity profiles observed across the radar domain. In contrast, applying a locally-defined statistical method to derive the profile shape aloft avoids the requirement of assuming a climatological profile shape above the bright band. The linear regression parameters computed for each region can be incorporated into a radar data quality product.
Results demonstrating the performance of the new multiple scan elevation method using operational radar data will be presented using case studies in different rainfall conditions and longer-term verification. The impact of the choice of tuning parameters used to derive the VPR on the quantitative accuracy of surface rainfall estimates will be discussed.