Generalized Least Squares Regression Applied to Emerging Trend Detection in Streamflow and 21st Century Streamflow Projections in the Northern Rocky Mountains under Greenhouse Forcing Scenarios
Jeannine-Marie St. Jacques1, Yang Zhao2, Suzan L. Lapp1, Elaine M. Barrow1, David J. Sauchyn1
1Prairie Adaptation Research Collaborative (P.A.R.C.), University of Regina, SK, Canada; 2Department of Mathematics and Statistics, University of Regina, SK, Canada

Recent research on the detection of emerging climate change trends in the northern Rocky Mountains, North America, has concluded that the region is running out of surface water supplies due to global warming. Reaching such a conclusion from a statistically simplistic analysis of instrumental streamflow records is problematic, given the severe residual autocorrelation of hydrological data, the short length and discontinuity of most streamflow gauge records, and human impacts. The 20th century hydroclimatology of the northern Rocky Mountains is heavily influenced by recurring large-scale climate patterns: the Pacific Decadal Oscillation (PDO), the El Nino-Southern Oscillation (ENSO), and the North Atlantic Oscillation/Arctic Oscillation (NAO/AO). Generalized least squares (GLS) regression addresses residual autocorrelation and allows reliable significance testing of any predictor coefficients, and hence, is highly suitable for hydrological modeling. We modeled northern Rocky Mountain river discharges using GLS regression equations with a linear trend and these climate indices as predictors. The GLS equations captured a major portion of streamflow variability. We demonstrated that streamflows are declining at most gauges due to hydroclimatic changes (probably from global warming) and severe human impacts. Next, using archived global climate model runs from the Coupled Model Intercomparison Project Phase 3 (CMIP3), we projected the PDO, ENSO and NAO for the 21st century for the B1, A1B and A2 SRES emission scenarios. These projected climate indices were then used as inputs into the GLS regression models, giving projected northern Rocky Mountain river discharges for the 21st century. These projections showed further generally declining trends in surface water availability for 2010-2099.

Keywords: Emerging trend detection in hydrological data; Generalized least squares regression; Northern Rocky Mountains; Streamflow projections under greenhouse forcing scenarios

Biography: The presenter holds a Ph.D in Biology, from Queen's University, Kingston, Canada; and a M.A. in Mathematics, from Berkeley, California, USA. She is currently a Research Associate at the Prairie Adaptation Research Collaborative, University of Regina, Saskatchewan, Canada. Her research interests are in the climatology and hydrology of the Canadian western interior and the verification and improvement of statistical methods used in paleoclimate reconstructions.