Laserfiche WebLink
<br />Radar engineers from OSF performed a special calibration of the Denver radar on January 31, 1996. They <br />discovered the previous calibration was 0.3 dB low and recommended that prior level II data be adjusted <br />by that amount, which was done in our analyses. Consequently, Denver radar data from the 1995-96 <br />winter are believed to be quite accurate. A calibration of the Cleveland radar .was not accomplished until <br />May 1997, when a significant calibration error was discovered. Recordings of the prior winter's" delta . <br />system calibration" were used by an OSF engineer to estimate corrections for each storm period. These <br />corrections ranged from +0.6 to + 1.2 dB and have been used in analysis of Cleveland's level II data. <br />These corrections are believed to have an RMS uncertainty (not worst case) of about 0.7 dB (William <br />Urell, personal communication). Unfortunately, some uncertainty exists in the Cleveland reflectivity <br />measurements, but that uncertainty appears to be much less than necessary to account for the difference <br />in avalues calculated for Cleveland and Denver. Consequently, these differences are believed to be real. <br /> <br />The reason for the a value differences is not obvious but is presumably related to the different <br />microphysical processes between lake effect (Cleveland) and upslope (Denver) snowstorms. As <br />discussed in section 6, Denver snow densities tend to be greater than Cleveland's, which could be related <br />to the differences seen on figure 2. For example, aggregates of dendritic crystals are frequent inland from <br />Lake Erie, especially as the winter progresses and air-water temperature differences decrease (Jiusto and <br />Weickmann, 1973). Dendritic aggregates are known to produce low density snow but also produce <br />relatively high reflectivities for a given snowfall rate (Ohtake and Henmi, 1970). <br /> <br />Figure 3 shows all 698 hourly observations for Cleveland gages 1 and 2 plotted against radar estimates <br />using the recommended Ze-Srelation. The solid 1:1 line and dashed least squares regression equation are <br />plotted for reference on figure 3 and all similar figures to follow. The figure 3 R value is 0.66, so only <br />44 percent of the variance in hourly observations is explained by the relationship. A stronger association <br />between gage observations and radar estimates would clearly be desirable. Substantially better agreement <br />is found over entire storm periods as discussed in section 8. <br /> <br />0.15 <br /> <br />........ <br />c <br />'-"" <br /> <br />w <br />30.10 <br />V> <br /> <br />>0- <br />J... <br />:J <br />o <br />::r: <br /> <br />~0.05 <br /> <br />~ I <br /> <br /> <br />Ze = 330 52.0 <br />CLE gages 1,2 <br />N=698, R=0.66 <br /> <br />0.00 <br />0.00 0.05 0.10 0.15 <br />Radar-est. Hourly SWE (in) <br /> <br />Figure 3. - Plot of all 698 hourly observations for Cleveland gages 1 and 2 against radar estimates using the noted <br />Zs-S relation. The solid 1:1 line represents hypothetical exact agreement and the dashed line is the linear least <br />squares regression equation. An R-value of 0.66 was calculated for this data set. <br /> <br />15 <br />