Construction, calibration, and validation of the RBM10 water temperature model for the Trinity River, northern California

We constructed a one-dimensional daily averaged water-temperature model to simulate Trinity River temperatures for 1980–2013. The purpose of this model is to assess effects of water-management actions on water temperature and to provide water temperature inputs for a salmon population dynamics model. Simulated meteorological data, observed streamflow data, and observed water temperatures were used as model inputs to simulate a continuous 34-year time series of historical daily mean water temperature at eight locations along 112.2 river miles from Lewiston Dam near Weaverville, California, downstream to the Klamath River confluence. To demonstrate the utility of the model to inform management actions, we simulated three management alternatives to assess the effects of bypass flow augmentation in a drought year, 1994, and compared those results to the simulated historical baseline, referred to as the “No Action” alternative scenario. Augmentation flows from the Lewiston Dam bypass consist of temperature-controlled releases capable of cooling downstream water temperatures in hot times of the year, which can reduce the probability of disease outbreaks in fish populations. Outputs from the Trinity River water-temperature model were then used as inputs to an existing water-temperature model of the Klamath River to evaluate the effect of augmentation flow releases on water temperatures in the lower Klamath River. 

We structured the Trinity River water-temperature model in River Basin Model-10 (RBM10), which uses a simple equilibrium flow model, assuming discharge in each river segment on each day is transmitted downstream instantaneously. The model uses a heat-budget formulation to quantify heat flux at the air-water interface. Inputs for the heat budget are calculated from daily mean meteorological data, including net shortwave solar radiation, net longwave atmospheric radiation, air temperature, wind speed, vapor pressure, and a psychrometric constant needed to calculate the Bowen ratio. The modeling domain was divided into eight reaches ranging in length from 8.8 to 20.6 miles, which were calibrated and validated separately with observed water temperature data collected irregularly from 1980 to 2013. Root mean square errors of observed and simulated water temperatures for the eight reaches ranged from 0.25 to 1.12 degrees Celsius (°C). Mean absolute errors ranged from 0.18 to 0.89 °C. For model validation, a k-fold cross-validation technique was used. Validation root mean square error and mean absolute error for the eight reaches ranged from 0.24 to 1.11 °C and from 0.18 to 0.89 °C, respectively.

Augmentation scenarios were based on historical hydrological and meteorological data, combined with prescribed flow and temperature releases from Lewiston Dam provided by the Bureau of Reclamation. Water releases were scheduled to achieve targeted flows of 2,500, 2,800, and 3,200 cubic feet per second in the lower Klamath River from mid-August through late September, coinciding with the upstream migration of adult fall-run Chinook salmon (Oncorhynchus tshawytscha). Water temperatures simulated at river mile 5.7 on the Klamath River showed a 5 °C decrease from the No Action historical baseline, which was near or greater than 23 °C when augmentation began in mid-August. Thereafter, an approximate 1 °C difference among augmentation scenarios emerged, with the decrease in water temperature commensurate to the level of augmentation. All augmentation scenarios simulated water temperatures equal to or less than 21 °C from mid-August through late September. Water temperatures equal to or greater than 23 °C are of particular interest because of a thermal threshold known to inhibit upstream migration of salmon. When temperatures exceed this approximate 23 °C threshold, Chinook salmon are known to congregate in high densities in thermal refugias and show extended residence times, which can potentially trigger epizootic outbreaks such as of Ichthyophthirius multifiliis (“Ich”) and Flavobacterium columnare (“Columnaris”) that were the causative factors of the Klamath River fish kill in 2002. A model with the ability to simulate water temperatures in response to management actions at the basin scale is a valuable asset for water managers who must make decisions about how best to use limited water resources, which directly affect the state of fisheries in the Klamath Basin.