WMOD00477 - Laserfiche WebLink

Home Browse Search

206 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 4 humidity (personal communication, Potts, 1985). Richner and Phillips (1981) concluded that the repro- ducibility of the VIZ sonde is quite satisfactory. De- viations in multiple sonde accents were found to lie within the accuracies specified by the respective sensor manufacturer. 3. Field evaluation analyses a. Rawinsonde and radiometer comparison During SCPP, rawinsondes were launched from Kingvale during the storm periods at 3-h intervals for one shift per day (up to 12 h, four launches). Launches also were made on clear days to compare the vapor observations under optimal conditions. Eighty-nine rawinsonde soundings were analyzed for the 1984-85 season. The comparison between the rawinsonde and the radiometer water vapor measurements is shown in Fig. 1. The radiometer data used in plotting Fig. 1 were averages over the first 20 min following the rawinsonde launch. Three sets of statistics were calculated from the da- taset. Separate statistics were calculated for pairs of ob- servations measured in the presence of liquid water, pairs without liquid water and all pairs. The scatter diagram and accompanying statistics shown in Fig. 1 were determined from the 89 rawinsondejradiometer vapor observation pairs. Pairs evaluated in the absence of liquid water provided the greatest correlation coef- ficient at 0.96, with a root-mean-square (rms) difference of 0.05 cm. Based on the means, slopes and y-inter- . cepts, the radiometer generally measured less vapor 120 l.00 E 0.80 .!d 0:: 0 0... <( > 0:: 0.60 w. f- W L 0 5 OAD <( 0:: .' 020 0.08.0~' 0.40 0.20 0.60 RAWINSONDE VAPOR (em) than the rawinsonde. The bias was near 0.03 cm. Pairs with liquid water showed slightly lower correlation and slightly larger rms differences. The combined data (all pairs) yield a correlation coefficient of 0.94 with arms difference of 0.07 cm. These differences were between 13 and 15% of the mean. Hogg et aI., (1983), also com- pared radiometer and rawinsonde data. They found a 0.17 cm rms difference for a 6-month dataset taken at Denver, Colorado, and a 0.26 cm rms difference for a 3-month field test at Sterling, Virginia. These differ- ences were over twice that found with the Kingvale dataset. Since the radiometer used at both Denver and Sterling was nearly identical to the one used in King- vale, causes for the differences must be attributed to differences in rawinsonde sensors or data reduction techniques. b. Radiometer-radiometer comparisons During the 1984-85 CaSE program, the USBR and NOAA radiometers were collocated for a period of 3 weeks for direct comparisons. The radiometers were placed adjacent to each other at the base of the Park Range in northwest Colorado, near the town of Steam- boat Springs. Due to physical restraints within the local parking area, the radiometer trailers had to be aligned perpendicular to each other. The distance between the reflectors was 20 m. Two modes of operation, fixed vertical and azimu- thal scanning, were utilized for direct comparison. Ad- ditional comparisons between the two radiometers and a rawinsonde launched from a local site were made on rOt + + cl~kit. ~jt~h~qu~ ~6:955 Average ogoute d~rence - .053 RMS . .066an i'.IJmb"'Sy'~fb&";;S 0 . 27 1i:f.pe . 0 988&, ~~~epl:; ~:~~55 ......1.... POirS in ~qlJd water ,. 0 mm CorrebleD, coefffi<~nt . 0.93 Average aR'f;~r,; ~~nce B .070 Numbesy~~ +" 62 11.kJpe. 1.0175 ~~cepl: . ~~t'7 . ..Y... ~ope . 1.008(,,, tercept " -~1' J ~~ G : ~.."546 0.80 1.00 120 FIG. 1. Scatter plot showing rawinsonde precipitable water vapor (ordinate) vs. radiometer retrieved water vapor measured at Kingvale, California.