Laserfiche WebLink
<br />8. SUMMARY AND DISCUSSION <br /> <br />This extension of the previous work for the OSF produced a different style of field testing of the SAA <br />under the GCIP program. The algorithm was modified to accept Level ill data from NIDS providers in <br />near real time for a series of five radars across the Dakotas and Minnesota. Products were provided via <br />the Internet in the 4-km HRAP grid so as to be useful for forecast groups. Accumulations of Sand SD <br />were produced for a variety of time intervals up to 24-hours, ending at 12 UTC each day. The products <br />of the five radars were combined in a mosaic to show regional accumulations. <br /> <br />Working with the NIDS data was generally successful. The mosaic process indicated that one or two <br />radars appeared to be calibrated differently from the others, as shown by Sand SD discontinuities across <br />lines equidistant between the radars. <br /> <br />Virga was a persistent problem. An experimental procedure eliminated most virga without sacrificing <br />the reliability of the algorithm in widespread, intense storms. That algorithm still needs further testing <br />and adjustment before becoming part of the operational version of the SAA. <br /> <br />The SAA failed to match surface observations during a snowstorm in arctic air. An analysis indicated <br />that the storm was shallow and had temperatures in the dendritic growth band for snow crystals. The <br />radar beam generally was above the clouds, missing the rapid crystal growth close to the ground. <br />Furthermore, dendritic crystals have the least density as snow on the ground. A change in a few <br />adaptable parameters could have remedied the problem, but such was not possible in the routine <br />production of products from the NIDS data stream. <br /> <br />Though desired in the specifications for tasks, it was not possible to derive local parameters of alpha and <br />beta for radars in Alaska, Washington, and Illinois. There was insufficient quality data for those sites. <br />Analyses of the California (Sierra Nevada) data indicated that the radar beam was far above the snow <br />growth zones, which resulted in small alpha values. <br /> <br />A separate program was written to combine many days of SAA files to produce composite accumulations <br />for three partitions of area coverage: scattered, moderate, and widespread. The output gave guidance for <br />adjusting the hybrid scan file for inadequate or excessive suppression of clutter. The same products <br />using widespread storm data could be useful in determining adjustments in the occultation correction file. <br /> <br />As part of the virga investigations, experimental coding was produced to generate images of the vertical <br />profile of reflectivities. The images gave insights into the changing vertical structure of the storms. Parts <br />of the code could be used for producing a better algorithm that is sensitive to vertical gradients. There is <br />potential for better performance with virga and bright band events and for a better range correction <br />scheme. <br /> <br />In general, this extension of effort has shown that the original SAA tends to be robust in an operational <br />mode. Therefore, no major modifications to the operational versions of the SAA were made. There are <br />lingering blemishes to work on, such as virga and bright band effects, but for now, forecasters can be <br />alerted to their effects by the natures of the patterns (rings and intense gradients) in the SAA output. <br /> <br />18 <br />