Review of an EPA Thermal Model of the Snake and Columbia Rivers

Water temperatures in the lower Columbia River are known to exceed temperatures of 20 degrees C as frequently as 15 percent of the days of the year. Whether this is a natural condition or one exacerbated by the construction of dams is not well understood. As temperatures exceed this level, they pose potential harm to threatened and endangered salmonids.

The Environmental Protection Agency developed a model that suggested that removal of four lower Snake River dams would improve the thermal environment in the lower Columbia River. GEI evaluated the data and the model used by EPA and found significant problems with both. Because of the lack of detailed thermal data associated with corresponding meteorological data, EPA resorted to mixing and matching data sets from various time periods from the 1970’s to the 1990’s and various locations across the Columbia basin into a single unified data set.

As a result, a percentage of important modeled variables such as relative humidity violated assumptions of heat transfer equations producing theoretically impossible results. This created bias upwards of the predicted temperatures of the “with reservoir” condition. Other problems included errors in estimating the surface and volumes and differing residence times of the reservoirs versus the natural rivers and the variation in thermal (stratification) behavior.

The raw data showed that thermal violations were predominantly occurring in reaches upstream of those modeled by the EPA. Thus incoming water was already in violation and that dam removal at best only reduced violation days by 4-5 percent. Finally, the concept of using a point-value in excess of a specific temperature as a measure of violation or compliance did nothing to assess the difference between a 0.1 C violation over 20 C from a 5.0 C violation, yet the biological differences could be potentially very large or imperceptible.

Review of more detailed models over shorter reaches of the river suggested that the reservoirs were having no greater effect on temperature than could be statistically determined with more rigorous data sets. In other words, these models concluded that it was statistically impossible to determine whether reservoirs were having a warming-effect, as the modeled effects were smaller than the error variances in the data.