CHANGES IN SNOWMELT RUNOFF TIMING IN WESTERN NORTH AMERICA UNDER A ÔBUSINESS AS USUALÕ CLIMATE CHANGE SCENARIO
IRIS T. STEWART*
Scripps Institution of Oceanography, La Jolla, California
DANIEL R. CAYAN
Scripps Institution of Oceanography and
US Geological Survey, La Jolla, California
MICHAEL D. DETTINGER
US Geological Survey, Scripps Institution of Oceanography,
La Jolla, California
Revised & resubmitted to journal Climate Change, Feb 2003
Abstract
Spring snowmelt is the most important contribution to the annual flow of many rivers in western North America. In the present climate, spring and summer snowmelt discharge is relatively dependable and not prone to producing floods, so it is a critical source of water supply. If climate changes, however, this seasonal pattern may be disrupted. A shift in the initial spring pulses of snowmelt runoff towards earlier in the water year already has been observed in historic runoff records from the 1948-2000 period in many rivers. The present study uses Parallel Climate Model (PCM) simulations of the 1995-2099 climate under a business-as-usual greenhouse-gas emissions scenario to project likely responses of snowmelt runoff timing in snowmelt-dominated rivers across western North America. Twenty-first Century streamflow timings, measured here by the temporal center of mass of streamflow (CT) each year, were projected with simple and multiple linear regression analyses relating CT variations to temperature (TI) and precipitation (PI) indices. The PCM simulations project a warming trend in TI through the year 2099, and, in response, streamflow timing comes increasingly early. The projected CT changes are consistent in magnitude and direction with those observed over western North America for the past five decades. The projected timing trend most strongly affects the Pacific Northwest, Sierra Nevada, and Rocky Mountain regions, where rivers eventually run 30-40 days earlier at many gages. By contrast, the modest PI changes projected by PCM do not cause significant changes in CT. The responses of CT to the simultaneous changes in TI and PI are dominated by the changes in TI. Regression model-based changes in CT generally agree well with those produced by physically-based hydrologic simulations of selected rivers in the Pacific Northwest and in the Sierra Nevada of California.
Change in annual surface air temperature [¼C] (red dots) and total precipitation [mm] (green and orange dots) for one 2070-2099 business-as-usual climate simulation minus one 1951-1989 historical simulation. Contours show temperature and precipitation changes for three different future-climate simulations (with the same model) vs. the same historical simulation.
OBSERVED trends in the temporal center of mass of yearly
hydrographs (CT). The color of the symbols corresponds to a given magnitude of
the trend, which is given here in terms of the corresponding overall timing
shift [days] for the 1948-2000 historical period. Larger circles indicate
statistically significant trends at 90% confidence level, smaller circles
correspond to trends that are not significant at the 90$ confidence level.
20-year averages of projected changes in CT [days] as determined by regression with a temperatures at the historical average time of snowmelt runoff, and compared to the average CT of the 1951-1980 climatology. Projected CT is averaged over the (a) 2000-2019, (b) 2040-2059, and (c) 2080-2099 periods.