Consistent long-term observations of snow amount (snow depth or snow water equivalent) are very few, and generally are insufficient to identify the complex interaction of winter precipitation, temperature, and local topography that determine long term variations in snow accumulation. However, through the use of long-term gridded climatalogical fields, it is possible to simulate patterns of winter snow accumulation for a period approaching 85 years. These simulated data can serve as a surrogate for observations. We use the Variable Infiltration Capacity (VIC) macroscale hydrology model, implemented over the Columbia River basin at a spatial resolution of 1/8 degree, with further discretization of topographic effects through the use of subgrid elevation bands. We examine variations in simulated snow accumulation and ablation for the period 1918-2001 across two transects in the mountainous Pacific Northwest. Where snow course observations are available (typically on or about April 1), we compare trends in simulated and observed snowpack.

Others have found that a long-term warming trend in much of the west has led to an earlier spring freshet in many unregulated river basins. These trends are a complex mixture of natural climate variability (temperature and precipitation) and regional warming trends at longer time scales associated with climate change. In large snow melt dominated river basins such as the Columbia River basin in the Pacific Northwest, the first indications of global warming are not likely to be discernible in streamflow observations, but rather should be observed as a loss of winter snowpack in the relatively small areas of the basin where winter temperatures are close to freezing much of the time. We hypothesize that statistically significant trends in snowpack in these sensitive areas may already be discernible given the moderate warming that has been observed over the 20th century. These changes may provide early warning of climate change impacts and opportunities for long-term climate monitoring. We evaluate, by comparing long-term streamflow records as observed and as simulated by VIC, whether such trends exist in the Columbia River basin, and runoff generating source areas that are most responsible for observed trends.