Laserfiche WebLink
<br />2. PROBE Recalibration <br /> <br />Although a rigorous quality control program was inexistence during the <br />PROBE data collection, there was still a possibility that biases couldappeiar .. <br />(Holman and McInerney, 1983). This was particularly true for pressure and <br />humidity. The basic problem in this application was to remove low frequency <br />noise, a difficult concept. The approach was to remove the meteorological <br />signal to find drift. This technique was somewhat arbitrary.and was <br />constrained by the type of data, scales and data processing capabilities. <br />Local effects should not be considered as contribution to the biases. This <br />would be difficult to defend and would possibly mask the phenomena the analysis <br />is trying to detect; and the objective analysis scheme is designed to filter <br />such effects (depending upon the radius of influence and number of passes). <br /> <br />In the final scheme, smoothed fields were found by taking 48-hr average,s of <br />various parameters during periods when benign weather provided fairly <br />steady-state conditions. Since the largest dimension of the PROBE network was <br />approximately 250 km, this was the largest wavelength of phenomena that could <br />be adequately described. The Barnes technique has applied to the 48-hr st8Ltion <br />averages (trends in the case of pressure) using a radius of 150 km which <br />effectively eliminated wavelengths shorter than 150 km leaving a smoothed <br />field. Individual station offsets from this analysis were objectively analyzed <br />using two passes and a 30 km radius of influence. Large offset errors could be <br />easily discerned in this second analysis. Examination of time vs parameter <br />plots (GAPF) confirmed problems. <br /> <br />Figure 11.9 shows an example smoothed temperature field. Stations shol,ing <br />a significant departure from the field were easily identified and questionable <br />temperature, relative humidity and pressure data could be eliminated in this <br />fashion. Wind direction data had correction factors inserted into the <br />recalibration file if the biases were of the same sign and approximate <br />magnitude throughout the season. Wind speed was assumed to be correct if <br />nothing was seen on the original GAPF check. The steps in recalibrating the <br />data are summarized below. <br /> <br />1. The station parameters were turned off for periods which were susp,ect, <br />based on preliminary examination of GAPFplots and day logs. This was done <br />for the periods of 48-hr averages, days of interest and for 24-hrs prior to <br />a case study. <br /> <br />2. For the closest 48-hr period of near steady-state weather conditions, <br />each station 48-hr average parameters or in the case of pressure, 48-hr <br />pressure trend were found. Speed was excluded from this analysis. An <br />objective analysis of these averages (trends) was performed using a radius <br />of influence of 150 km and two passes of the Barnes technique. <br /> <br />3. Using bilinear interpolation of the four nearest grid points, the 48-hr <br />average (trend offset) was determined. <br /> <br />4. An objective analysis of those offsets was performed using a radius of <br />30 km with two Bands' passes to pinpoint stations having unacceptably large <br />offsets. <br /> <br />30 <br />