The three-dimensional numerical model UnTRIM was used to model hydrodynamics and heat transport in Upper Klamath Lake, Oregon, between mid-June and mid-September in 2005 and between mid-May and mid-October in 2006. Data from as many as six meteorological stations were used to generate a spatially interpolated wind field to use as a forcing function. Solar radiation, air temperature, and relative humidity data all were available at one or more sites. In general, because the available data for all inflows and outflows did not adequately close the water budget as calculated from lake elevation and stage-capacity information, a residual inflow or outflow was used to assure closure of the water budget.
Data used for calibration in 2005 included lake elevation at 3 water-level gages around the lake, water currents at 5 Acoustic Doppler Current Profiler (ADCP) sites, and temperature at 16 water-quality monitoring locations. The calibrated model accurately simulated the fluctuations of the surface of the lake caused by daily wind patterns. The use of a spatially variable surface wind interpolated from two sites on the lake and four sites on the shoreline generally resulted in more accurate simulation of the currents than the use of a spatially invariant surface wind as observed at only one site on the lake. The simulation of currents was most accurate at the deepest site (ADCP1, where the velocities were highest) using a spatially variable surface wind; the mean error (ME) and root mean square error (RMSE) for the depth-averaged speed over a 37-day simulation from July 26 to August 31, 2005, were 0.50 centimeter per second (cm/s) and 3.08 cm/s, respectively. Simulated currents at the remaining sites were less accurate and, in general, underestimated the measured currents. The maximum errors in simulated currents were at a site near the southern end of the trench at the mouth of Howard Bay (ADCP7), where the ME and RMSE in the depth-averaged speed were 3.02 and 4.38 cm/s, respectively. The range in ME of the temperature simulations over the same period was ?0.94 to 0.73 degrees Celsius (?C), and the RMSE ranged from 0.43 to 1.12?C. The model adequately simulated periods of stratification in the deep trench when complete mixing did not occur for several days at a time.
The model was validated using boundary conditions and forcing functions from 2006 without changing any calibration parameters. A spatially variable wind was used. Data for the model validation periods in 2006 included lake elevation at 4 gages around the lake, currents collected at 2 ADCP sites, and temperature collected at 21 water-quality monitoring locations. Errors generally were larger than in 2005. ME and RMSE in the simulated velocity at ADCP1 were 2.30 cm/s and 3.88 cm/s, respectively, for the same 37-day simulation over which errors were computed for 2005. The ME in temperature over the same period ranged from ?0.56 to 1.5?C and the RMSE ranged from 0.41 to 1.86?C.
Numerical experiments with conservative tracers were used to demonstrate the prevailing clockwise circulation patterns in the lake, and to show the influence of water from the deep trench located along the western shoreline of the lake on fish habitat in the northern part of the lake. Because water exiting the trench is split into two pathways, the numerical experiments indicate that bottom water from the trench has a stronger influence on water quality in the northern part of the lake, and surface water from the trench has a stronger influence on the southern part of the lake. This may be part of the explanation for why episodes of low dissolved oxygen tend to be more severe in the northern than in the southern part of the lake.