WMOD00282 - Laserfiche WebLink

Home Browse Search

Algorithm does not average two radials that happen to have the same integerized azimuth angle. Instead, it keeps track of the exact azimuth angles in an array and scans the array each time a particular azimuth is needed. The nearest azimuth to the nominated azimuth is selected. That technique produces no gaps in coverage as can happen with simple integerized azimuth angles. The range has been limited to 115 kIn for the 2.50 tilt and to 10 kIn for the 3.50 tilt. The ranges needed for those tilts for the three sites (Denver, Cleveland, Albany) are within those values according to the expanded hybrid sector files. That range limitation saves memory space in the computer environment. It does not affect any results. The reflectivity storage arrays are in named common rather than in a direct access file. The advection scheme needs rapid access to the stored data in almost a random access fashion. Reserving space in named common makes the search go faster than disk access. The Snow Algorithm differs from the PPS algorithm by saving the biased reflectivity numbers (byte or b2) rather than floating point dBZ's or precipitation rates. Converting all of the reflectivities to precipitation rates before the rest of the manipulations takes time, and converting them aU back takes additional time. Furthermore, the RATE_TABLE has the same precipitation rate for a variety of reflectivities. If nothing is ever done to a range bin's reflectivity (averaging, occultation boost), the conversion from biased reflectivity to precipitation rate and back to biased reflectivity almost always results in a different number for weak echoes. The PPS conversion thereby substitutes adversely quantized numbers for the real ones. The subroutine OCCULT is called to add 1, 2, 3, or 4 dBZ to the radial beyond particular ranges indicated by the occultation file. This addition amounts to adding twice that number (or 2,4,6, or 8) to the biased reflectivity. The PPS routine has the same adjustment but is coded differently. Both versions have the same 0.20 azimuth resolution. Then NOSPIKES is called to "weed out" isolated bins above a minimum threshold for precipitation detection and outlier bins above a maximum threshold. The former are replaced by zeros (in biased dBZ units) and the latter by either zeros if the outlier is not isolated from other outliers or averages of the neighbors if the outlier is isolated. The Snow Algorithm uses similar logic to that in the PPS routines but has a smaller array for. corrected values. NOSPII\ES replaces PPS routines A3133S, A31330, A3133G, and A3133N. Once the biased reflectivity data are thereby cleaned, the three snowfall totals files are opened. The names are automatically calculated from the date and time. The I-hour and 3-hour files are always closed after a Level II file has been processed, but the storm total file remains open. The I-hour and 3-hour files may be reopened if necessary, or a new file may be created. Separate files record snow accumulation for vertically-falling and advected snow, ' respectively. Only the advected snow routine is required by the MOU for this version of the Snow Algorithm. The assumption of vertically-falling snow is included here because of its simplicity and speed of calculation. Moreover, as discussed in section 9, initial results indicate the present advection scheme does not improve snowfall estimates over the assumption of vertically-falling snow. The PRECIP subroutine is then called. This version does not make any adjustment for the time delay between reflectivity observations and the arrival of the snow at the surface. It integrates the snowfall under two assumptions. The fastest calculations are for assuming no wind speed (vertically falling snow). The routine then calls ADVECT to determine the part of the sky from which advected snow is arriving at each ground location. It is assumed that the snow has a vertical fall speed of 1.0 m s-\ typical of aggregated snow. Future revisions 43