5—MODELING APPROACH AND ASSUMPTIONS
WDC023080001.ZIP/TAF 5-6
5.2.6 Salt Creek Hydrologic Validation Summary
To validate the results of the hydrology calibration, HSPF was run for Salt Creek from
January 1996 through September 1999. Table 5-3 includes statistical summaries of the
calibration and validation results.
TABLE 5-3
Summary of Hydrologic Calibration and Validation - Annual Flow and Correlation Coefficients
Rolling Meadows Elmhurst Western Springs
Calibration Period (1991-1995)
Mean Observed Annual Flow (in.) 16.3 21.8 21.4
Mean Simulated Annual Flow (in.) 15.6 20.9 21.2
Difference (percent) -4.3 -4.1 -0.9
R-Squared Daily 0.61 0.64 0.76
R-Squared Monthly 0.72 0.81 0.86
Validation Period (1996-Sept. 1999)
Mean Observed Annual Flow (in.) 16.3 23.5 23.9
Mean Simulated Annual Flow (in.) 17.1 22.4 22.7
Difference (percent) 4.9 -4.7 -5
R-Squared Daily 0.42 0.56 0.59
R-Squared Monthly 0.45 0.51 0.55
For a hydrology calibration, the percent difference between simulated and observed flows
often is used as a measure of the calibration’s accuracy. A difference of less than 10 percent
is considered a very good calibration, difference of 10 to 15 percent is considered good, and
a difference between 15 and 25 percent is considered fair (Donigian, 2000).
Table 5-3 shows differences between simulated and observed flows of less than 5 percent for
the calibration, indicating a very good calibration. For the validation period, the differences
are in the range of 5 percent, also indicating a very good calibration.
R-squared, or the coefficient of determination, sometimes is used as a statistical measure of the
quality of a calibration. When analyzing daily values, an R-squared value of 0.8 to 0.9 is
considered very good, 0.7 to 0.8 is considered good, and 0.6 to 0.7 is considered fair. When
analyzing monthly values, an R-squared value of 0.85 or higher is considered very good, 0.75
to 0.85 is considered good, and 0.65 to 0.75 is considered fair (A. Donigian, personal
communication, 2001).
For the hydrology calibration, the daily R-squared values indicate a range from fair to good,
while the monthly values indicate a range from fair to very good. For the validation, the daily
R-squared values indicate a range from poor to fair, while the monthly values indicate a range
from poor to fair. The poor values tend to be more toward the upper portions of the watershed,
which are more influenced by the heavy point-source discharges during low-flow periods.