An Error-Based Selection Criterion to Mitigate the Effects of C-Band Attenuation in the UK Radar Composite

A new radar compositing criterion was implemented on the UK operational system in November 2012. In this work we describe the background and derivation of this criterion, and demonstrate its effects in intense precipitation.

Operational radar networks with significant overlap provide a selection of horizontally coincident data for rain rate estimation. Increasing demand for accuracy in radar products, from both commercial and scientific users, make effective exploitation of this overlap essential. To achieve this, radar compositing algorithms must be able accurately and consistently to discriminate between data of different quality.

The algorithms currently used within the radar community are broadly divisible into two groups: those that weight or select on the basis of range, and those which select for maximum measured reflectivity. Maximum-reflectivity algorithms maximise probability of detection at the expense of bright band contamination and increased false alarm ratio. Range- or height-based criteria minimise uncertainties due to non-homogeneous vertical profile of reflectivity (VPR) – which can be considerable in stratiform precipitation – but are vulnerable to underestimation and missed detections. This is particularly an issue where low elevation scans are subject to attenuation.

Beam attenuation can cause significant problems for millimetre and sub-millimetre wavelength radars. Attenuation at C-band leads to underestimation of moderate to intense precipitation, and in extreme cases can result in total beam extinction, leaving gaps in the composite even where some radars have detected rain. X-band measurements can be affected even by light precipitation. With increasing use of radar composites for NWP, and of X-band radars for network gap-filling and urban flood monitoring, it is important that compositing algorithms can account for and mitigate this issue.

This work proposes a method of combining information from multiple sources to identify “best data” for radar composites. A calculated rain rate can be expressed as a function of measured reflectivity through the corrections and transformations applied by the processing system. By propagating measurement uncertainty through these functional transformations, an analytical estimate of rain rate error can be obtained. This estimate is used to define an error-based quality index for composite data selection.

The method is tested within the UK centralised processing system (RADARNET), which uses a single-scan compositing algorithm with no vertical averaging. Until 2012 the algorithm selected, at each point, the surface rain rate calculated from the reflectivity measurement with the lowest available beam height. The new criterion, replacing height, is calculated using three transformations: an attenuation correction, VPR adjustment and reflectivity-to-rain-rate conversion. This yields a quality index that varies with measurement position, signal strength and estimated attenuation.

The compositing skill of the quality index is assessed using radar and gauge data from June 2012: a month of exceptional rainfall across the UK, featuring several high-intensity events and widespread flooding. Composites generated by the UK RADARNET, using both height- and quality-based compositing methods, are compared with each other and with gauge accumulations. The quality index method achieves visible improvements in strongly attenuated cases, where regions of extinction and underestimation are infilled with data from adjacent radars. Quantitatively, the quality index is shown to improve gauge-radar agreement in moderate to intense precipitation.