WMOD00493 - Laserfiche WebLink

Home Browse Search

;~?C' .. 1988-89 simulations. Giorgi et a1.(1992a) defined bias as the sum of the differences between observed and model predicted daily average precipitation in each subregion, divided by the total number of station days for that subregion and season. 1bis model bias is a direct measure of the model error simulating a given variable over a region. Their study was based on 390 daily observation stations in an area of -4.4 million km2 in the West. The station density was about 11,000 km2 per gauge with few high elevation gauges. We focus on the assessment of local area precipitation in complex terrain using a higher density of -930 km2 per gauge with 11 gauges above -2700 m ms!. Modeled precipitation was compared with surface precipitation data from NWS (National Weather Service) offices and cooperative stations, and the Soil Conservation Service's SNOTEL (snow telemetry) remote high elevation automatic stations. Figure 2a shows a time-series comparison between observed high elevation cumulative precipitation and MM4 predicted cumulative precipitation in the Gunnison region during the 1982-83 water year (October 1 to September 30). Predictions by MM4 were averaged over 8 grid points in the vicinity of 31 precipitation gauge sites. Model daily average values and gauge averages were accumulated to form model and observed daily cumulative means. This daily cumulative mean gives a measure of the total precipitation observed from the beginning of a water year to a given date. Precipitation observations were stratified by three elevation zones: < 2100 m, 2100 to 2700 m, and > 2700 m ms! and similarly accumulated into three daily cumulative means. Figure 2a shows the time series of daily cumulative precipitation observed in three elevation zones from October 1,1982, to May 15, 1983. The striking difference between the high elevation SNOTEL data and the low and moderately high observation sites illustrates the problem of model verification over complex terrain. In the model validation by Giorgi et al. (1992a), few precipitation observations were available at high elevations such as the SNOTEL sites; consequently ,their analyses may have overestimated the model's positive bias (overprediction of precipitation) in these regions. Figure 2a shows that MM4 predicted the high elevation cumulative precipitation' remarkably well through winter and spring in 1982-83; however, it significantly overpredicted warm season convective precipitation. Similar convective problems occurred in the simulations of the 1988-89 dry period, but cumulative totals were significantly less in the Gunnison. MM4 model simulations started on January 1, 1988, and continued through April 1989. When comparing 1982-83 with 1988 analyses in figure 2b, note that results are from different periods when simulations were available, and that the ordinate was scaled to include the excessive summer precipitation in 1982-83. Winter cumulative precipitation was reasonably well modeled; however, from mid-May to September, convective precipitation was poorly estimated. This crossover point in late spring marks the MM4 change from mesosynoptic prediction of orographic precipitation to orographically triggered convective precipitation. It clearly shows MM4's limited ability to simulate convective precipitation. Figure 1 also shows an example of summer simulations of the strong summer monsoon flow predicted by MM4 on July 20, 1983, and the resulting 24-h precipitation field predicted over the West. The maximum of . 5.25 cm over Colorado illustrates the overprediction problem in mountainous regions. Precipitation measurements in the Gunnison Basin indicated an area average of 0.2 cm with a maximum of 1.0 cm on July 20, 1983, hence, the need for Oark model simulations to' improve convective parameterizations for both MM4 and Rhea models. a OCTOBER.DECEMBER1~~3 2500 12000 c ~ t 1500 ~ ~1000 ~ il :; ~5OO o - MM4 Model -H- Gauges> 2700 m msl -'- Gauges 210010 2700 m msl -t- Gauges < 21oo'm msl ~ao N 0 J F M A M J i A S Month b JANUARY - DECEMBER 1988 800 ~700 e .5.600 1500 ~~ Go 1300 jj . ~200 a 100 oJ F M A M MonthJ A SON 0 - MM4 Model -H- Gauges > 2700 m msl - Gauges 2100 10 2700 m msl -t- .Gauges < 2100 m msl Figure 2. Time-series of cumulative precipitation predicted by MM4 and corresponding observations in the Gunnison Basin during: (a) the El Nino 1982-83 and (b) dry 1988-89 periods. Three elevation zones stratify the observed data clearly demonstrating the effect of elevation on precipitation during both periods. MM4 predictions closely matched observed cumulative precipitation above -2700 m msl in 1983 from October to mid-May; however, the model grossly overestimated precipitation during the convective summer months. In 1988, MM4 predicted light precipitation .that matched the low elevation observations from January to March, and October to December. Its estimate increased sharply and exceeded all observations from mid-April to September, again indicating a problem in predicting convective precipitation.