One of the most challenging questions about climate change is how Earth’s warming atmosphere will affect water availability across the globe. Climate models present a range of possible scenarios—some more extreme than others—which can make it difficult for cities, states, and countries to plan ahead. Now, however, in a new study, Padrón et al. suggest a way to reduce uncertainty using precipitation patterns from the past.
A rule of thumb for global warming’s impact on Earth’s water availability that was sometimes proposed in the past was that dry regions will get drier and wet regions will get wetter, also known as the DDWW hypothesis. But mounting evidence suggests the reality is more complicated. In 2014, for example, one of the study’s authors and colleagues found that the DDWW hypothesis checks out less than half the time when applied to historical precipitation records.
Climate models can generate different results because of the inherent variability of climate or because of flaws in the models. To reduce the models’ contribution to this uncertainty, the authors gave 36 climate models the task of accurately reproducing historical precipitation records from 1976 to 2005. The best-performing historical models, they hypothesized, would converge in their predictions of water availability in the future, defined as precipitation (rain and snow) minus evapotranspiration (water consumed by plants and evaporated by the Sun).
The hypothesis proved correct. When the team used only historically accurate models to simulate water availability from 2006 to 2100, they produced more consistent forecasts. All the models were run under a business-as-usual scenario in which nothing is done to reduce greenhouse gas emissions.
The historically accurate models lacked many of the extreme changes that the full ensemble of climate models predicted, the team found. Although the possibility of extreme drying in Europe, western North America, and South Africa remained, for example, these models predicted that it would be 5 times less likely to occur compared to the prediction by the full 36-model ensemble. Overall, the constrained group of models showed that previous projections of very extreme future changes in water availability were less likely to occur on more than 70% of Earth’s land surface.
The historically skillful models held some other surprises, including a higher confidence of drying in the Amazon, where the 36-model ensemble predictions are inconclusive. Atmospheric circulation patterns are complex, making it extremely difficult to predict precisely how changes in the water cycle will play out as the atmosphere warms. But the new study suggests that models that agree with historical observations are the most likely to produce reliable predictions.