Laserfiche WebLink
<br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br />I <br /> <br />gage measurements by over a factor of two (Super and Holroyd 1997a). Similar or <br />greater errors have been reported with rainfall. Consequently, the errors caused by <br />using Level III data to estimate SWE are not regarded as serious. Of course, the <br />method of selecting the "most representative" dBZ value within each Level III <br />interval was tested and improved using the five storms of table 2. Future <br />comparisons should be accomplished with independent storms to determine <br />whether the degree of agreement is generally as good as suggested by table 2. <br /> <br />The most serious Level III overestimates beyond 150 km are believed related to the <br />shift in Level II frequency distributions with range as previously discussed. <br />Further adjustments could be made to bring the Level II and III estimates into even <br />better agreement. Testing with other storms data sets would be required to <br />demonstrate whether such "fine tuning" was justified. <br /> <br />4. SUMMARY AND CONCLUSIONS <br /> <br />This applied research addressed two tasks related to estimating snowfall with <br />NEXRAD radar measurements and Reclamation's Snow Accumulation Algorithm <br />(SAA). The first task was to begin development of a range correction scheme based <br />on the measured vertical profile of reflectivity, converted into vertical profiles of <br />radar-estimated SWE (VPS). The VPS of Minnesota winter snowfall, based on a <br />sample of nine storms, has been shown to increase downward in the lowest 3 km <br />above the radar. That is, maximum SWE estimates were always at the lowest <br />estimated sampling level about 380 m above the radar. <br /> <br />Linear least squares regression equations were fit to averaged SWE values for the <br />five lowest radar beam tilt centers at about 35 kIn range. These equations, with <br />one exception, explained 93 to 100 percent of the variance indicating that the VPS <br />is linear. Although the slope of the equations varied among storms, use of the <br />median case worked reasonably well in correcting range underestimation. <br />Estimates at 150 km range would have been within a factor of two of the assumed <br />"true" value in all 9 storms and within about 20 percent of "true" for 6 of the 9 <br />storms. These results are believed to justify continued pursuit of a range correction <br />scheme for the SAA based on the radar observed vertical profile of reflectivity. <br /> <br />The second task was to determine the feasibility of using "degraded" Level III <br />reflectivity data instead of high resolution (0.5 dBZ) Level II data with the SAA. <br />Although having less resolution (4.0 to 5.0 dBZ), the Level III product is readily <br />available in near real time from NIDS vendors. Moreover, Level III reflectivities <br />have been, and are being, archived for several radars in the GCIP LSA-NC. The <br />ability to use Level III reflectivities with the SAA would allow for radar estimation <br />of SWE over most of the LSA-NC for a two winter period. This capability would <br />greatly enhance estimation of spatial SWE accumulations in view of the limited <br />number of suitable precipitation gages in protected locations in the LSA-NC. <br /> <br />22 <br />