Extended Abstract

B3.1A Multiscale Observation Error Estimation Scheme for QPE UsingLocal High Resolution RadarP E SHUCKSMITH 1 , G L AUSTIN 1 , L SUTHERLAND-STACEY 11 University of Auckland, Auckland, New Zealand, pshucksmith@orcon.net.nz1. IntroductionIt is well known that there are myriad sources of error associated with the retrieval ofsurface rainfall rate from radar reflectivity measurements made aloft (for example, seeVillarini and Krajewski, 2010). Some of these sources of error, such as the uncertaintyin the observed rainfall’s drop size distribution and the resulting effect on the Z-Rrelationship, have been researched extensively over the past several decades. Fewerstudies have focussed on the fundamental sampling error associated with themeasurement of precipitation at spatial resolutions and at sampling intervals that arecoarse compared to precipitation systems’ decorrelation length and time.From the studies that have been undertaken it is evident that this source of error is verysignificant and could dominate over other sources of error in many cases. For instance,Fabry et al. (1994) studied one hour of high resolution data, downgrading it to a range ofspatial and temporal resolutions. For this case, they found that a 39% error wasintroduced into a 5 minute rainfall accumulation solely from downgrading a radar’sspatial and temporal resolution to 2 km and 5 minutes respectively. Similar results wereobserved by Piccolo and Chirico (2005) who looked at error due to temporal sampling atthree different accumulation spatial scales and by Jordan et al. (2000) who found thatamong other results, the error due to decreasing sampling resolution to 5 km was of theorder of 100% of the mean rainfall rate.Lack of attention to this error source could perhaps be because there is little which canbe done to improve the spatial sampling of existing fixed radar networks withoutincreasing the density of the network or changing the fundamental parameters of theradars. Decreasing the sampling interval of the same radars may be impractical foroperational reasons, particularly if a slow Doppler scan is incorporated in the scan cycle.As well as affecting a radar’s ability to measure fine scale structure in precipitationsystems, the spatial resolution of a radar, if it is coarser than the observed rain field, canalso lead to an overestimation of precipitation area and intensity. These latter effectsare due to non-uniform beam filling and incomplete beam filling problems. The temporalresolution, dependent upon the radar’s sampling rate, determines a radar’s ability toobserve a rain field’s evolution as well as to better track its movement. Fabry et al.(1994) found this to be particularly important regarding 5 minute rainfall accumulations.The ability to estimate the magnitude of this error using only the low resolution data,along with supplementary data such as that from rain gauges, would allow, aftercombination with estimates of other error sources, the creation of an ensemble of-132-