7073

• UCREFRP Catalog Number: 7073
• Author: Brewer, M. and J. Renne.
• Title: Application of the Instream Temperature Model to the Green River Basin
• USFW Year: 1989.
• USFW - Doc Type: Draft.
• Copyright Material: NO

Instream Temperature Model " "DRAFT" "
Page 7
March 27, 1989
S = . 6745 S-r . x
where
Ti ith predicted temperature from heat transport model
Tj jth predicted temperature from regression model
N = number of predicted temperatures in sample set
i j
ST.x modified standard difference of estimate
Calibration of the Temperature Model was accomplished by adjusting
the internally-defined model coefficient for air temperature until
the probable difference was minimized while increasing the
coefficient of determination. Adjusting the air temperature
coefficient accounts for transposition of air temperature from the
Grand Junction weather station to each node in the network. An
assumption is made that the validation gage regression models and
observed water temperatures are accurate.
The mean and probable differences for the Green River validation
node near Jensen on an annual basis were .49°C and 1.39°C, respec-
tively. Regressed and predicted water temperatures for this node are
plotted in Figure 2. The model under-predicted water temperatures at
this node for six of the fourteen time periods (Table 5).
Water temperatures were under-predicted for three time periods at
the Green River node near-Green River, Utah. The mean difference at
this node was .83°C and the probable error was 1.23°C. Regressed
and predicted water temperatures at this node are plotted in figure
3.
The average mean error for all validation nodes was .66°C and the
associated probable error was 1.32°C (Table 5). The 50% confidence
interval of .66 ± 1.23°C is within the Biel fluctuation of water
temperatures throughout the stream network.
CONCLUSIONS
The USGS gages within the study area had useful instream temperature
data.- However, most of the gages had missing data and several gages
had two or more years in a row with no water temperature data. In
addition, none of the gage data were developed from accurate 24-
hour averages; it can only be assumed that instream temperatures
acquired from the USGS represent temperatures somewhere between the
minimum and maximum for the day recorded.
The Regression model was used to fill in missing temperature data