WMOD00527

• Title: Quantitative Precipitation Forecasting for Improving Reservoir Operations
• Date: 4/1/1995
• Weather Modification - Doc Type: Report

2.6 Statistical Results

2.6.1 Precipitation

When comparing the orographic precipitation model estimates to the average observed
(recorded) precipitation, the linear correlation coefficient, r, between model computed and
observed l2-h total precipitation (for 371 periods) was 0.73. Summation to 24-h periods
increased r to 0.80. This increase in correlation is believed to primarily be caused by reduced
errors associated with the timing of representativeness of Oakland rawinsonde data; the
longer period reduces timing errors over the longer measurement interval. The overall
average ratio of l2-h observed precipitation accumulation to 12-h model precipitation
estimate was 0.87. The difference could be caused by overprediction by the model,
undercatch of actual precipitation because of gauge limitations, or a combination of both.

J'

-

The sample size of37112-h periods was large enough to allow study of the results by 700-mb
wind direction class (directions were rounded to the nearest 10-degree azimuth). Systematic
wind direction dependent differences between the 22-site average observed and model
precipitation values were found (table 5). The model overpredicted the average precipitation
with southwest flow, but underpredicted precipitation with northwest flow. The correlation
(r) was highest for winds most nearly perpendicular to the Sierra Nevada Range. These
systematic differences can be used to good advantage for operational forecasting by adjusting
model estimates based upon the observed/model ratio by 700-mb wind direction class.

Table 5. - Model and observed 12-h precipitation averages (in.) for a group of 22 precipitation gauge
sites in the American River Basin.

Direction Averages (in.) Ratio Corr. Sample
Class (deg) Obs. Model OIM r size
330-350 0.19 0.00 info 0.00 17
300-320 0.28 0.13 2.06 0.23 41
270-290 0.38 0.34 1:12 0.69 81
240-260 1.02 1.08 0.94 0.78 120
210-230 0.77 1.06 0.73 0.67 76
170-200 0.54 1.08 0.50 0.30 36
Overall 0.66 0.76 0.87 0.73 371

j
;

2.6.2 Simulated Inflow to Folsom Reservoir

Linear regression analysis was performed of predicted versus observed peak inflow to Folsom
Reservoir (sample size = 63). Inflow predictions used orographic precipitation model
estimates as input for the HED71 rainfall-runoff simulation hydrologic model. These
calculations yielded a linear correlation coefficient of 0.87, a regression line slope of 0.87, an
intercept of -5900 ft3/s, and a standard error of estimate of 20,700 ft3/s (fig. 5). This line slope
is the same as the overall ratio of observed to model average precipitation noted in the
precipitation analysis section above. Standard error of estimated inflow peaks is
encouragingly small, especially considering that the peaks range from 6000 ft3/s to as high

14