WMOD00562 - Laserfiche WebLink

Home Browse Search

nly 7 percent of the gage observations of figure 3 were greater than 0.050 inch h-1, and the largest value as 0.150 inch h-1. (Of the 1794 hours with detectable snowfall along the line of all five gages located est-northeast of Cleveland, only 2 hours exceeded 0.150 inch.) So estimates of the Ze-S relation for s owfall accumulations greater than 0.050 inch h-1 are based on limited sample sizes. Consequently, the r commended Ze-S relation should be used with some caution for heavier snowfalls. onsiderable overplotting of data pairs at the lighter hourly accumulations exists on figure 3 because of e highly skewed distribution of the hourly snowfall accumulations. Median hourly accumulations are nly about 0.01 inch for both gages 1 and 2. As discussed by Super and Holroyd (1996), similar median alues were found for all Cleveland and Denver gages. he average gage-observed and radar-estimated hourly accumulations for the data of figure 3 are the s e, 0.019 inch, as required by the optimization technique. But figure 3 shows that the optimized I' lation results in frequent radar overestimation at light hourly accumulations and frequent u derestimation at heavy accumulations. This feature is common in similar plots for Denver gages, and h s been noted with Ze-R relations as well where R is rainfall (see Smith et al., 1984). At least a partial e planation for this "mismatch" is likely related to the several orders of magnitude differences in pling volume between the radar range bins and the gages (or snow board) used in table 5. lthough 1 hour temporal averaging has been used in this analysis, significant spatial variations s metimes likely existed over the volume of a range bin. Assume that during a given hour, a gage h ppened to sample near a spatial maximum, as should occasionally occur. That gage point measurement ould then be markedly greater than the radar's spatial average sampled in the overlying range bin. Such a point would appear on figure 3 as a radar underestimation associated with a high gage measurement. a similar manner, assume that the gage happened to sample near a spatial minimum, markedly less than t e radar's overhead spatial average. This point would appear on figure 3 as a radar overestimate sociated with a light gage measurement. Because of the inherently very large differences in sampling v lumes between the two measurement methods, one cannot be certain of the "true" snowfall. A very d nse gage network over an area larger than a range bin would be required to measure "true" snowfall o the same spatial scale as the radar. But it is usually very difficult to find representative gage sites well- p otected from the wind as discussed by Super and Holroyd (1996). So finding many adequate sites near t one another is impractical in almost all situations. In summary, one cannot be certain whether the a parent "mismatches" of figure 3 are simply caused by the differences in sampling volume of gages and I' dar, or by other factors such as an improper choice of CTF or even limitations in equation (1). 5 2 SAA Development for Albany large volunteer network of snow "spotters" were organized and maintained by Albany NWS WFO P rsonnel. The spotters made hourly measurements of SWE and SD in the area around Albany, NY, d ring the 1995-96 winter as discussed in section 3.2 of Super and Holroyd (1996). Only that portion o the network expected to receive dry snow (no bright band) was activated for each storm. These data ere carefully reviewed and provided by John Quinlan of the Albany NWS WFO for 12 events with u eful surface data, but one had missing radar data. So useable observations are available from the 11 s ow storms listed in table 6. 16