Dual-polarisation radar rainfall estimation

Operational dual-polarisation radars are being installed across much of Europe, presenting the opportunity for better rainfall rate (R) estimates than available with conventional radars which provide only reflectivity (Z). The dual-polarisation provides additional information which, when used in combination with Z, have the potential to provide more accurate estimates of rainfall rates and better data quality. Operational radars have noisy polarisation parameters, meaning the benefit is lost when used at individual pixels. I developed a technique to estimate drop concentration from the distribution of data from a number of radar pixels over small areas, constraining the Z-R relationship. I have assisted with the setup of this algorithm on the Meteo France operational dual-polarisation radars.

Radar Bright-Band Correction

A further source of error in radar derived rainfall products is caused by melting snowflakes, creating the 'bright band'; I have investigated this showing the potential for improved correction for this effect using dual polarisation radar.

Statistics of Rainfall

The nature of rainfall is of importance to radar estimates, the scales over which it changes have implications on how accurately rainfall can be estimated. By studying the statistics of rainfall using drop counting raingauge and disdrometer, the natural changes in rainfall we would measure with radar can be examined, for instance the effect of changing drop size distributions or the implications of scan strategy on rainfall accumulation measurements (this work was also continued in the undergraduate dissertaion of Igobe Malidza under my supervision). As part of the FFIR - FRANC project, I worked on:

Radar Attenuation

In heavy flood producing rainfall, attenuation of the radar beam becomes a major problem, resulting in serious underestimations of rainfall accumulations. I have investigated potential corrections to account for attenuation to allow significant improvement on traditional correction methods; this included the use of dual-polarisation and observed changes in the background noise. This technique has the potential to differentiate radome attenuation from storms for correction algorithms.
Emission for outreach

Radar Hail Detection

Heavy rainfall events sometimes also include hail, hail in itself is a hazard, but in the UK is unlikely to be large enough to cause significant damage/danger. Hail however has a very different signiture in the radar, when the rainfall rate is infered from hail, significant errors are introduced. This project will use the new radar capabilities to detect hail and therefore flag data with potentially large errors as such, This is a different approach to attempting to correct the hail to an equivilent rainfall rate, which is likely to still retain large errors, since hail affected rain is small scale and transient.

Beam Blocking Correction

Radar beams need to scan at as low an angle to the ground as possible to minimise the distance rain must fall from radar beam to reach the ground. This means that often the radar beam is partially cut off by close in objects so as to only partially illuminate the volume that would be assumed without the blocking. This partial illumination means that lower reflectivities are measured and hence rainfall rates estimated. The traditional correction is from theodolite (or more recently high resolution digital terrain map - though this doesn't allow for trees and buildings) to give a consistant value in each direction. Experience however suggests this doesn't well capture the full extent of blocking, and that blocking is not consistant. This project will investigate the potential to use the radar data itself to infer the extent of blocking and hence allow a better correction, and investigate the variability of blocking.

Radar-Gauge Representitivity

Radar to raingauge comparisons still show large errors (factor of 1.6 are about the best achieved for multiple event studies with very carefully controlled radars), more than can be explained by the commonly cited reasons why radar-gauge comparisons are not good (Z-R relationships in particular). In this project I will further investigate the possibility of using the radar data itself to estimate the likely representivity to raingauge, and hence the expected statistical discrepency.