Laserfiche WebLink
<br />I <br />I <br />I <br /> <br />'I <br /> <br />AUGUST 1984 <br /> <br />DONEAUD, IONESCU-NISCOV AND MILLER,JR. <br /> <br />1609 <br /> <br />ft <br />t <br /> <br />scans (-10 min); 2) 1, 2 and 4 h periods; and 3) <br />growing and decaying periods. For all time intervals, <br />the sampling frequency was the radar scanning fre- <br />quency (-10 min). Only clusters with data from four <br />or more scans and with both identifiable growing and <br />decaying periods were selected. Of the 587 1981 <br />Bowman clusters previously identified, 284 clusters <br />met the above criteria. Merging (or splitting) clusters <br />were classified separately. <br />The growing and decaying periods of storms were <br />determined in three ways: using maximum reflectivity <br />(ZMX), maximum echo height (MEH), and maxi- <br />mum areal extent (MAREA). On average, a typical <br />cluster reached its maximum growth after about 56% <br />of its lifetime. On a statistical basis, the maximum <br />area was reached first, followed by the maximum <br />reflectivity and then the maximum echo height. The <br />results regarding ATI-RERV relationships were about <br />the same for the growing or the decaying period as <br />determined by anyone of the three criteria (Table <br />2). The log-standard errors of estimate seem to be a <br />little larger during the decaying period. The average <br />rain rate is a little greater during the growing period <br />of the typical cluster compared to the decaying period. <br />The differences in means, according to a t-test, are <br />not significant (a > 0.05), except in the case where <br />ZMX was used to divide the cell duration. In that <br />case (Table 2), the average rain rate of the growing <br />period is -10% greater than that of the decaying <br />period. This result differs from what Griffith et al. <br />(1976) derived in south Florida from digitized Miami <br />WSR-57 observations. Using 5 min time increments, <br />the authors found rain rates of 15.60 and 7.92 mm <br />h-1 for growing and decaying periods, respectively, <br />giving an average of 11.76 mm h - I. The differences <br />between the average rain rates in North Dakota and <br />Florida for the entire cell duration may reflect a <br />climate difference between Florida and North Dakota. <br /> <br />Perhaps more significant are the differences in rain <br />rates during the growing and the decaying periods at <br />the two locales. In Florida, R for the decaying period <br />is hair of that for the growing period, while in North <br />Dakota R for the decaying period is only - 10% less <br />than that of the growing period. <br />The mode of the distribution (the 4-5 mm h-I <br />category) for both the growing and the decaying <br />periods of the North Dakota clusters represents ap- <br />proximately 50% of the observations (Fig. 7). The <br />A TI reflectivity threshold (the MART) in both situ- <br />ations is 25 dBz. If small clusters were discarded, th:c <br />average rain rate (R = 4.8 mm h-I) could be <br />considered independent of the A TI in more than two- <br />thirds of the cases. <br />A scan-by-scan calculation (-10 min intervals) of <br />the average rain rates was also performed. The rainfall <br />rates showed a similar overall frequency distribution <br />(unimodal with a right skewness). The rain volume <br />versus single-scan, area-time product correlation <br />dropp(:d to 0.94, and the (logarithmic) standard error <br />of estimate increased to 0.20 (Table 3). Table 3 is a <br />composite table displaying RER V versus A TI corre- <br />lations for the NDCMP and Quadra data sets. The <br />Quadra data were recorded in the tropical Atlantic <br />Ocean. The rain amounts were determined by mul- <br />tiplying the radar-determined rain area by a rain rate <br />averaged over a long period of time (Lovejoy and <br />Austin, 1979). In North Dakota, the average lifetime <br />of the clusters was somewhat longer than -1.0 h <br />(Fig. 2). The data were computed by averaging values <br />at 10 min time increments (Table 3). For the NDCMP <br />data, as the averaging time interval approaches storm <br />lifetimes, the RER V j A TI correlations become higher, <br />from 0.94 to 0.98. This strengthens the accuracy of <br />the rain volume estimates obtained using the recently- <br />developed RERVjATI technique (Doneaud et al., <br />1981; Ionescu-Niscov, 1982), and the concept that <br /> <br />TABLE 2. RER V versus A TI 25 and rain rate regression analyses for the growing and the decaying periods of a cluster <br />(Bowman, North Dakota; 1981 Summer experiment, 284 clusters). <br /> <br /> Split cluster criteria <br /> Maximum area (25 dBz) ZMX MEH <br /> Growing Decaying Growing Decaying Growing Decaying <br /> period period period period period period <br /> - <br />Correlation coefficient 0.98 0.98 0.98 0.98 0.98 0.98 <br />Slope 1.06 1.02 1.05 1.03 1.06 1.02 <br />Intercept 3.56 . 4.06 3.88 3.92 3.69 ' 4.12 <br />Logarithmic standard error <br />of estimate 0.15 0.18 0.16 0.17 0.16 0.17 <br />Average rain rate (mm h-') 5.02 4.92 5.23 4.81 5.11 4.93 <br />Standard deviation of rain <br />fates (mm h-') 2.10 2.72 2.32 2.20 2.43 2.24 <br /> <br />t. <br /> <br />