WMOD00282 - Laserfiche WebLink

Home Browse Search

Although use of the WSR-88D with a 0.50 tilted narrow beam has reduced the range effect reported by Wilson (1975), further improvement is still needed, especially beyond the 60-km range. Extrapolation of radar measurements from beam altitudes to ground level using the vertical profile of Ze appears to be a viable approach far significantly improving radar- estimated precipitation at moderate-to-far range (Joss and Waldvogel, 1990; Joss and Lee, 1995). A vertical profile extrapolation scheme will be developed for a future version of the snow algorithm if resources permit. The R values of tables 8 and 9 are similar even though the optimization scheme was separately applied to each gage in table 8, and a single Ze-S relation was used for table 9. This finding is comparable to the results of Smith et a1. (1975) who noted that their similar optimization scheme, "does not noticeably improve the correlation between the radar and gage estimates of the hourly point rainfall amounts." Although the basic scatter of the data appears to remain the same no matter what Ze-S relation is applied (within reason), the optimization scheme provides an objective and effective means for estimating the "best fit" a and ~ values to quantify radar-estimated snowfall. The table 9 R values are similar for gages No.1 to 5 and the same for gage No.1 at the 36- km range and gage No.4 at the 115-km range. The standard error of estimate values are also very similar to those of table 8. These results suggest that the basic radar "signal" is not significantly diminished to at least a 115-km range. This finding provides hope that a method can be developed to improve WSR-88D estimates of snowfall at ranges to and beyond 100 km. However, a note of caution is in order. The data set used herein combined all storm types for the 2-month period, including at least two major synoptic storms which produced the heavier snowfalls. Results may not be as encouraging when lake effect storms are considered separately in the future because such storms tend to. have shallow cloud systems and light to moderate snowfall accumulations at the Cleveland gage sites. The optimization scheme was applied to upwind arrays of range bins (sec. 4.2) used with gage No.1 observations. The resulting a value was 354, and the ~ value was 1.25. However, these results likely were contaminated by ground clutter in the operator-designated box with zero suppression imposed around gage No.1, the only Cleveland gage with nearby clutter. Whenever the advection scheme "moved" the upwind array a few kilometers west of the gage, a region of known clutter likely increased the Z. values in some range bins. Software to exclude this cluttered area has not yet been developed. However, a partial test of the advection scheme was applied as follows. Equation (8) was applied to. the upwind arrays, and correlation coefficients were calculated for the resulting pairs of radar- estimated snowfall and gage-observed snowfall. Only gage No.2 had a slight improvement in R value (0.82) over table 9. The lesser value for gage No.1 (0.68) might be explained by clutter contamination. Gage No. 3's R value was almost unchanged (0.65), but the gage No.4 value of 0.56 and gage No.5 value of 0.10 are well below the R values of table 9. These preliminary results do not indicate any improvement at near and moderate ranges with application of the advectian scheme, and indicate a'definite degrading of predictability at far ranges. This result suggests that the current advection scheme, using the V AD wind profile and a fixed 1.0-m S.l fall speed for all crystals, offers no advantage over simply assuming that snowflakes fall vertically and using Ze measurements directly over the gages. One common feature of figures 3 through 7 is the lack of a linear relationship between radar- estimated and gage-observed snowfall. Of course, the predominance of data points would be expected to lie above the 1:1 line for gages No.3 through 5 because of the range effect. 30 1 1