Skip to main content

Skip to navigation

Departments & Programs


Climate uncertainty and snow variability in the 21st Century

Max, min, and ensemble seasonal snow accumulation (snw) trends | a., Seasonal average snow accumulation (snw) trends, 2000-2060. We plot the  time series from the runs with the maximum  and minimum NH area-weighted linear snw trend (max, open red circles; min, blue open triangles) and the 40-member ensemble (grey plus signs). Fitted to each set of data is locally weighted (nonlinear) regression to illustrate the spread in the ensemble. The shading represents the 95% confidence around the median nonlinear regression estimate (red for the maximum trend, blue for the minimum, grey for the ensemble trend), constructed from 1000 bootstraps of the data with replacement. The spread across all three shaded regression estimates represents the hemispheric spread in the seasonal trend. b., spatial fields of the snw trend differences (max minus min) among the runs plotted in a. where the trends are expressed as a percentage of the 1970-1999 seasonal mean. c., histogram for the 40 seasonal (NDJFM) 50-year hemispheric snw trends. d., Ensemble composites. For each grid point, we select and plot the maximum and minimum 50-year snw trend among the 40 members (first two panels); these are composites of all 40 members, not trend fields from a single run. The third map shows their difference. The final map is the ensemble average of all 40 members’ 50-year NDJFM snw trends. Stippling in the max and min snw trend plots indicate the statistical significance of the linear trend at each grid point run. Black dots are for significance at the 5% level, white dots are for significance at the 1% level.  Stippling in the ensemble plot indicate the ensemble signal to noise in tas trends, with white plus signs indicating an ensemble signal greater than two times the ensemble noise. Black plus signs indicate a signal greater than one times the noise.

Mountain ice and snowpack supplies water for up to 40 percent of the world's irrigation, plays a critical role in providing hydropower and a host of ecosystem functions in forests, riparian, and downstream communities. Climate models project that global warming from increased greenhouse gas concentrations will induce important changes in the quantity and timing of runoff from these mountain systems. Given the strategic importance of seasonal snow accumulation and the timing of its runoff for downstream systems we ask a single model ensemble (CCSM3.0 21st C. large ensemble) three questions: (1) Over what timescales and dimensions does the distribution of snow accumulation and melt runoff change in the Northern Hemisphere? The CCSM3.0 runs show a salient decline in both snow accumulation and cover in the boreal winter, concurrent with increasing trends in both temperature and precipitation with March snow accumulations declining upwards of 40-50% below the 1970-1999 average in continental interiors over a 50-year period. Snow cover shows similar declines, with trends increasing in early season runoff and decreasing in late season runoff and snowmelt when downstream communities require it. (2) Where are such trends driven by temperature or precipitation variability?  Trends in snow accumulation are largely dominated by rapid warming, even in continental interiors where snow accumulation is historically driven by precipitation variability. (3) What are the implications of the large-scale internal variability in accounting for any of these equally likely trends? There are important implications for natural variability's accounting for a large fraction of uncertainty in climate model projections. We illustrate that a single model initialized with only minor atmospheric differences can have statistically significant temperature trends at the scale of on spatial scales as large as the Northern Hemisphere and on timescales as long as 50 years.