Improving predictions of summer climate change in the United States

Across vast, agriculturally intensive regions of the United States, the spread in predictions of summer temperature and soil moisture under global warming is curiously elevated in current climate models. Some models show modest warming of 2-3C degrees and little drying or slight moistening by the 22nd century, while at the other extreme are simulations with warming as large as 7-8C degrees and 20-40% reductions in soil moisture. We show this region of large spread arises from differences in simulations of snow albedo feedback. During winter and early spring, models with strong snow albedo feedback exhibit large reductions in snowpack and hence water storage. This water deficit persists in summer soil moisture, with reduced evapotranspiration yielding warmer temperatures. Comparison of simulated feedback strength to observations of the feedback from the current climate's seasonal cycle suggests the inter-model differences are excessive. At the same time, the multi-model mean feedback strength agrees reasonably well with the observed value. We estimate that if the next generation of models were brought into line with observations of snow albedo feedback, the unusually wide divergence in simulations of summer warming and drying over the US would shrink by roughly one third to one half.