Analysis, Evaluation And Utilization Of3D Mosaic Reflectivity From Ground-based Radars And Hourly Precipitation Data From Automatic Weather Stations At Yangtze-Huai River Basin During The Mei-Yu Season Of2007

The Mei-Yu season of year2007persisted for38days from19June to26July.Rainstorms during this period triggered the worst flood events in the Huai River valley since1954. During this period, two types of non-conventional data (3D mosaic reflectivity databased on measurements from ground-based weather radars and hourly rainfall data fromAutomatic Weather Stations) are collected, analyzed and evaluated, forming the data basis forfuture investigation of mesoscale features of the Meiyu precipitation processes. The presentstudy consists of the following three parts.1. Generation and evaluation of3D mosaic reflectivity data. By applying the Dopplerweather radar3-D digital mosaic system developed by the State Key Laboratory of SevereWeather at Chinese Academy of Meteorological Sciences, we produced3D mosaic reflectivitydata based on measurements from29weather radars at the Yangtze-Huai River Basin(28-36°N,110-122°E). The mosaic reflectivity data are compared to TRMM PrecipitationRadar observations at the same times and locations. To make reasonable comparison, the twodata sets are interpolated to the same horizontal and vertical resolutions. It is found that:(1) The two radar reflectivity data sets agree with each other at the height range of2.5-4km. During the2007Meiyu period, the fractions of rainy profiles (maximum reflectivity≥18dBZ) are very close between the two datasets:6.85%vs.7.28%(2.5km),7.55%vs.7.60%(3km),7.98%vs.7.66%(3.5km), and8.27%vs.7.62%(4km). Moreover, the probabilitydistribution functions of radar reflectivity from the two datasets at each level within this heighrange are also similar with the same peak at20-25dBZ. At each instantaneous time, horizontaldistributions of radar reflectivity from the two datasets present essentially the same features interms of size of precipitation area, pattern and location of precipitation system, as well aslocation of local intense precipitation.(2) At heights of and above5km, more significant disagreements are found between thetwo datasets. The ground-based dataset overestimates radar reflectivity and size of rainy arearelative to the TRMM data. The TRMM data presents maximum radar reflectivity at about5km (i.e., signal of brightness band near the0oC level in stratiform precipitation region) and a sharp decrease of radar reflectivity with height from5km to7km. In contrast, theground-bsed mosaic reflectivity dataset does not show obvious brightness band around5kmand the reflectivity decreases with height from5km to7km at a slower rate than the TRMMdata.2. Quality-controll of the AWS hourly rainfall records, and production andevaluation of a gridded hourly rainfall dataset. We collected hourly rainfall records from7127automatic weather stations (AWSs) including555national stations and6572regionalstations at the Yangtze-Huai River Basin. The average distance between the AWSs is about7km. First, these hourly rainfall records are strictly quality-controlled through four procedures,i.e., the comparison to mosaic radar reflectivity at3km, the internal and spatial consistencycheck, and the extreme check. Then, the quality-controlled hourly rainfall records areinterpolated onto a0.07°×0.07°grid using an improved Cressman interpolation method. Inorder to retain local intense rainfall centers, weighting of the stations is proportional to therainfall intensity. Moreover, the gridded hourly rainfall product is evaluated, not onlyqualitatively by comparison with the mosaic radar reflectivity at3km but also quantitativelyby application of two cross validation methods. Finally, to further check the hourlyprecipitation records from the regional AWSs, another gridded rainfall data is produced withonly rainfall records at those regional AWSs and compared to rainfall records at the nationalAWSs. About87%of the differences between the two data sources are less than2.5mm hr-1.This lends further support to the high quality of the gridded hourly rainfall product.3. Evaluation of CMORPH using the gridded hourly rainfall data based on AWSobservations. The gridded hourly rainfall product is utilized to evaluate CMORPH, a satelliteprecipitation product at high resolutions (30min and0.07°). It is found that:(1) CMORPHexhibits good performance in depicting the spatial pattern of hourly precipitation. Theaveraged pattern-correlation coefficient is0.44. However, CMORPH tends to underestimateoccurrences of intense local rainfall: the occurrence frequency of heavy rainfall (>10mm hr-1)is1.10%(CMORPH) versus1.31%while that of>20mm hr-1is0.19%(CMORPH) versus0.37%.(2) CMORPH largely captures the diurnal variations of region-averaged rainfallamount as revealed by the surface observations, with peaks occurring in the morning (about05hr Beijing time) and minima appearing in the early afternoon (13-14hr). However, CMORPHunderestimates the region-averaged rainfall amount during08-13hr by about19%andoverestimates it during21-01hr by22%.