WMOD00282 - Laserfiche WebLink

Home Browse Search

These measurements illustrate that most hours with snowfall in both geographical locations had relatively low accumulations. Half the hours had accumulations of 0.01 inch or less. Yet relatively low accumulations are important as indicated by half the snowfall totals usually occurring at hourly accumulations less than or equal to 0.03 to 0.04 inch. These similarities are especially interesting because most hours of snowfall in the Cleveland area are from lake effect storms and most in the Denver area are from upslope storms. Cleveland gage No.1 had the fewest hours with detectable snowfall in that area, partly because it was not installed in time for the first early-November storm and had one later period of missing data. But had no hours been missed, gage No. 1 still would have received fewer hours of snowfall than gages located farther from the radar which are more influenced by lake effect storms. When it is realized the gage No. 4 had 30 h of missing data because it also was not operational for the first storm, the snowfall frequency did not vary much for gages No.2 to 5. The Cleveland line of gages was located along relatively flat terrain as shown by their limited range of elevations in table 5. ~ ~ ~j Denver gage No. rleceived fewer hours of snowfall t~ gage No. %at almost the same range. This result likely occurred because gage No. .i'is-i~cated 127 m higher in elevation on a minor ridge which may produce some local orographic uplift. Gage No.3 had a higher snowfall frequency than higher gage No.2, probably because gage No.3 is located farther west, near the foothills of the Rocky Mountains. Snowfall was infrequent at gage No.4, in a broad mountain valley, likely because of "rain shadow" effects. Even gage No.5, located high in the mountains, received only 136 hours of snowfall during the dry 3-month period. 17 6. Calculation of Radar-Estimated Snowfall Accumulations Once arrays of range bins were obtained for each volume scan, the next step was to provide any needed adjustment to the Level II Z. data. Adjustments were provided by William Urell, OSF radar engineer, after he made careful calibration checks and studied the hourly records of overall WSR-88D performance from Albany, Cleveland, and Denver over the entire winter. For example, an offset of +0.3 dB was applied to all Denver data before January 31, 1996, the date of a major calibration, and no correction was needed thereafter. Higher adjustments were decided upon for the period of Cleveland data reported herein, ranging between +0.7 and +1.1 dB depending on the storm period. Mr. Urell estimated the RMS (root mean squared) error of the radar estimates at about 0.7 dB. Corrected Z. values from the lowest available tilt (0.50) were always used to estimate snowfall accumulations using particular values of a and J3 for equation (1). It is important to always convert recorded values in dBZ to snowfall rate before any averaging is done. Although often done, averaging values of Z. in mm6 mm-3 or in dBZ over time or space can cause significant bias. For example, assume a = 300, J3 = 1.4, and 3 adjoining range bins are averaged which have Ze values of 20, 25, and 30 dBZ. Use of the average of 25 dBZ would lead to an estimated snowfall of 0.041 inch h-l. However, first converting each range bin's Z. to snowfall rate leads to an average of 0.051 inch hot, a 24-percent higher value in this example. The approach used throughout this study always involved converting each range bin's Ze value to snowfall rate before averaging all range bins in an array. In like manner, precipitation rates, not values of Ze' were averaged over time. In the results presented here, average array snowfall rate for each volume scan was weighted equally with every other volume scan that had a start time (near the time of 0.50 scanning) within the