Projection of Climate Change Effects on Stream Temperatures in the U.S.

Abstract

The nonlinear regression model developed by Mohseni et al. [1998] and modified by Erickson et al. [l998a], was applied to weekly stream temperatures from 584 stream gaging stations in the US. On average, there were 12 years of record per gaging station in the US Geological Survey database. The revised model gave slightly smaller Nash~ Sutcliffe Coefficients (0.91 ± 0.08 on average) than the original model [Mohseni et al., 1998] (0.93 ± 0.06 on average). Highest stream temperatures simulated with the modified model were, however, more realistic which is essential to any projection of stream temperature effects on ecosystems under 2xC02 climate conditions. The stream temperature model has four model parameters; their dependence upon geographical factors was investigated. No relationship between latitude and the upper bound of weekly stream temperatures was evident, except in the Midwest. The lower bound of weekly stream temperatures was about 0 °c for most of the continent, except in the South and Southeast. A weak relationship was found between model parameters mean seasonal air temperatures. It is, therefore, less likely that model parameters would change under a warmer climate condition as has been projected by general circulation models under a 2xC02 climate scenario.