Abstract/Summary

A major source of errors in radar-derived quantitative precipitation estimates (QPEs) is the vertical reflectivity profile (VPR). A feature of particular importance is the radar “bright band”: a reflectivity enhancement due to the melting of large snowflakes that occurs in the majority of high latitude rainfall, and which if misrepresented can cause order of magnitude errors in surface QPEs. Recent upgrades of several national weather radar networks to dual polarisation provide opportunities to refine the identification of bright band in operational radar measurements, and to improve subsequent determination and corrections for VPR. This thesis applies information from the linear depolarisation ratio (LDR) to improve classification of VPRs at the pixel scale. Using a unique dataset of high resolution vertical profiles, values of LDR in the melting layer are shown to provide a seven-fold increase in probability of detection of non-bright band reflectivity profiles over the current UK operational criterion. In the context of the Met Office local VPR correction scheme, an LDR-based classification step alone is shown to produce improvements in bias and RMSE of more than 1 mm h−1 for high rain rates in non-bright band conditions. The high resolution vertical profile dataset is then further used to improve the parameterisation of VPR shapes for correction at the local scale. Using a simulation method adapted from previous literature, three possible non-bright band VPR shapes are defined and their performance objectively compared to a control (no correction) profile. The most skilful profile is applied to a high intensity non-bright band case study in combination with LDR-based VPR classification, yielding further improvements to QPE bias and RMSE. Improvements to the stratiform profile are also investigated. The conclusions of this thesis indirectly support the use of local over global VPR corrections to maximise the accuracy of radar QPEs at the sub-kilometre scale.