WMOD00275 - Laserfiche WebLink

Home Browse Search

JOURNAL OF HYDROMETEOROLOGY I ITABLE 1. Collocated snow course and SNOTEL SWE on 30 March. I Ye~r I 1999 1998 1997 I 1996 I 1995 I 199t 1993 1992 I 1991 I 1990 I Av~ I . erable scatter when snow course values were low. ThIS thr~shold did not consistently apply to water years 1998 I and 1999. I The SWE point data were interpolated to a 1-km grid ovJr the Gunnison, San Juan, and Salt-Verde basins, I and SWE volumes were summarized for different eleva- ioJ bands. For the Gunnison, San Juan, and Salt-Verde I . 1500 I 1000 1500 I 0 ~ 1500 .1 b. 1998 ~ 1000 en ...:l 500 ~ E-< o ~ 0 1500 I 1000 708 SNOTEL stations were not operated continuously and snow course measurelllents are periodic. The SWE was determined for each 1-km grid block by considering all stations within a 200-km radius, determined as an op- timal distance by Fassnacht et al. (2003). A residual was obtained at each grid block where a recording station was located by subtracting the observed SWE from the computed SWE. This residual was regressed to a datum of 5000 m using a constant lapse rate. From the com- mon datum, the lapsed residuals were gridded using the inverse-distance-weighting-squared technique. The gridded residual surface was then regressed to the basin surface and subtracted from the hypsometrically de- rived SWE surface. The choice of datum was a matter of scaling, and through the specific location was unim- portant (Fassnacht et al. 2003). To compute root-me an- square error (rmse) and related error statistics, the sta- tion value was removed from the calculation for the coincident grid block, and a residual was computed be- tween the station SWE and the regressed SWE. The following four data combinations for the inter- polation were used: (i) SNOTEL only, (ii) snow course only, (iii) collocated SNOTEL plus snow course for measurement days only (mmt), and (iv) collocated SNOTEL plus snow course for all days, with linear in- terpolation of snow course values between measure- ment days (linear fill). An additional data combination was tested (collocated SNOTEL plus snow course for a 2-day linear interpolation around the measurement day, i.e., a partial fill); however, the results were no different than linear fill for the 2 days around the mea- surement point and mmt on the day of measurement. The number of reporting snow course sites (from linear interpolation) exceeded the number of SNOTEL ob- servations during peak snow accumulation (Fig. 3). 4. Results Following Serreze et al. (1999), we began by exam- ining spatial correlations for all collocated SNOTEL and snow course sites (Table 1). The average Pearson's correlation coefficient for SWE at collocated sites ranged from 0.79 in 1993 to 0.96 in 1996. While the correlation was lowest in 1993, the difference between the means (-5 mm) was small; however, using means masks extreme site-specific differences that occur when SWE values are medium to high. To illustrate the varia- tions in the correlation, scatterplots of all SWE data from SNOTEL and snow course sites for 1993, 1998, and 1999 are presented in Fig. 4. Consistent with Table 1, SWE differences are most pronounced in 1993. Large differences occurred when SWE at the snow course sites is greater than 500 mm, but there was still consid- 500 o o ',. ? VOLUME 7 Correlation Avg SWE (mm) Snow course SNOTEL 0.94 0.94 0.92 0.96 0.87 0.92 0.79 0.89 0.87 0.94 0.90 135 217 269 205 247 166 345 199 230 143 213 153 242 283 212 273 181 340 205 240 147 223 8.1993 . . . C. 1999 500 1000 snowcourse SWE, mm 1500 FIG. 4. Measured SWE for SNOTEL vs collocated snow course site~ along the 1:1 line for (a) above-average (1993), (b) average (1998), and (c) below-average (1999) snow years.