Laserfiche WebLink
<br />I is important to realize that estimation of the "best" relationship between radar reflectivity factor and <br />$ciPitatiOn, either rain or snow, is not an exact process. Furthermore, the relation is not fixed but varies <br />ong and even within storms. Radar can only be used to estimate precipitation. However, radar <br />s owfall estimates can be reasonably accurate, especially over time periods of several hours, as will be <br />sown. <br /> <br />4. SELECTION OF RANGE BINS TO COMPARE WITH INDIVIDUAL GAGES <br /> <br />4 1 Comparisons of Array-Average and Single Bin Snowfall Estimates <br /> <br />Ai potential problem of comparing point snowfall observations from gages or snow boards with array- <br />ayeraged radar estimates is discussed in section 9.1 of Super and Holroyd (1996). They used arrays of <br />at least 9 range bins to calculate hourly snowfall estimates. Arrays were always 3 km in range by at least <br />3 I degrees in azimuth, resulting in an area of 9 km2 at 60 km range, increasing to 18 km2 by 120 km range. <br />jdditional pairs of radials were added for decreasing ranges less than about 50 km, so the array area over <br />eth gage was never much less than 9 km2. <br /> <br />T~is spatial averaging could compound the already present problem of mismatching of the volumes <br />represented by radar and gage observations. When significant spatial gradients exist, the greatest snowfall <br />rJtes would be expected to affect small areas, which occasionally could affect an individual gage. But <br />ateraging over 9 km2 or more would be expected to reduce ("smooth out") unusually heavy rates. <br />Cofl nversely, averaging over large areas would sometimes be expected to result in greater radar-estimated <br />a' cumulations than observed by a gage located near a minimum in the snowfall pattern. <br /> <br />S per and Holroyd (1996) presented a comparison of snowfall estimates between individual range bins <br />iq the 3-km by 5-degree array over Cleveland's gage 1, nearest the radar, and the estimated average from <br />tlie entire array. Correlation coefficients were large between pairs of individual bin estimates, and radar <br />st.OWfall estimates were similar in magnitude. These findings suggested little difference between use of <br />~ lSingle range bin near the gage and the array average. . <br /> <br />S1milar comparisons have since been done for all three gages nearest the Cleveland WSR-88D and for <br />I <br />tlle three gages nearest the Denver radar. The same 2-month Cleveland data set and 3-month Denver data <br />set used in the calculations presented by Super and Holroyd (1996) were employed in the comparisons <br />rJported here. Gage numbers are the same as in their tables 3 and 5, which also list specific locations. <br /> <br />Juation (1) was used in estimation of hourly SWE accumulations. The <<and pvalues used were 318 <br />Jd 1.5 for Cleveland and 155 and 1.6 for Denver as recommended by Super and Holroyd (1996). These <br />7lues have since been changed based on analysis of larger data sets, but those changes should not affect <br />:r following discussion. <br /> <br />Tflble 2 shows the linear R values (correlation coefficients) between hourly SWE from gage <br />measurements and radar estimates, both for the array averages and for single range bins. R values are also <br />I <br />gten between snowfall accumulations estimated for single bins and array averages. <br /> <br />Table 2 shows large R values between single range bin and array average snowfall estimates. Values <br />r~ge between 0.95 and 0.99 with a single exception. If the unusually small R value of 0.90 for Denver's <br />gage 2 is ignored, the next smallest value for that gage is 0.96. Therefore, in general, any single range <br />bin explains between 90 and 98 percent of the variance (R2) in these array estimates of snowfall. <br /> <br />8 <br />