Potential Effects of Climate Change on Ecologically Relevant Streamflow Regimes
We assessed the climate-driven changes in ecologically relevant flow regimes expected to occur by the year 2100 in streams across the conterminous United States. We used long-term daily flow measurements from 601 gauged streams whose watersheds were in relatively natural condition to characterize spatial variation in 16 flow regime variables selected for their ecological importance. Principal component analysis of these 16 variables produced five uncorrelated factors that described patterns of spatial covariation in flow regimes. These five factors were associated with low flow, magnitude, flashiness, timing, and constancy characteristics of the daily flow regime. We applied hierarchical clustering to the five flow factors to classify the 601 streams into three coarses and eight more finely resolved flow regime classes. We then developed a random forest model that used watershed and climate attributes to predict the probabilities that streams belonged to each of the eight finely resolved flow regime classes. The model had a prediction accuracy (per cent correct classification) of 75%. We used the random forest model with downscaled climate (precipitation and temperature) projections to predict site-specific changes in flow regime classes expected by 2100. Thirty-three per cent of the 601 sites were predicted to change to a different flow regime class by 2100. Snow-fed streams in the western USA were predicted to be less likely to change regimes, whereas both small, perennial, rain-fed streams and intermittent streams in the central and eastern USA were predicted to be most likely to change regime.