| 英文摘要: | As atmospheric carbon dioxide concentrations rise, some regions are expected to warm more than others. Now research suggests that whether warming will intensify or slow down over time also depends on location. When we need to wrap our head around a very complex problem, it is helpful to simplify and make approximations. Many of the methods we currently employ to understand climate change, arguably one of the most complex problems around, use approximations of linearity and aggregation of regional effects to global averages. In reality, all natural systems are nonlinear, and none of us live in a global average world. Deviations from these assumptions are particularly important when we are concerned with climate change adaptation strategies. However, integrated assessment models1 — the tools developed to inform adaptation decisions — are often based on linear approximations of climate change. As they report in Nature Climate Change, Peter Good and colleagues2 investigate sources of regional nonlinearities in climate model projections of future warming. Integrated assessment models are important decision tools for policy makers. They represent the complex relationships between the earth system and social and economic realms3. Because they include so many different processes, their representation of the earth system is necessarily very simple, often consisting of only a few equations. Many of these models assume linearity in the response of climate to an external forcing. In a linear system, doubling a perturbation doubles the response. In the context of global warming, the perturbation might be an increase in carbon dioxide concentrations. The resulting increase in surface temperatures is the system response. In a linear climate, the temperature response to a doubling of carbon dioxide levels would be exactly the same as the temperature response to a subsequent doubling. Making this approximation proves powerful when we are interested in the general behaviour of the climate system. However, when making decisions about adaptation and strategies, projections based on linear global assumptions are of limited use, and we need to take a closer look at how well they hold up in different locations and for different climate change scenarios. This is what Good et al.2 have done using a framework, developed in previous work4, that allows them to separate the climate's response to an external forcing (such as a doubling or quadrupling of atmospheric carbon dioxide) into its linear and nonlinear components. Nonlinearities in climate have previously been studied both in observational warming trends5 and in future model projections6. What distinguishes the work of Good et al.2 from previous studies is their focus on regional patterns of nonlinearity. The metric they use to quantify nonlinearity is a spatially varying 'doubling difference' — the difference between the temperature change caused by the first and that caused by the second doubling of carbon dioxide. Positive doubling differences imply that the second doubling of carbon dioxide leads to a stronger warming than the first (Fig. 1). Good et al.2 make use of doubling difference patterns to identify physical mechanisms driving nonlinear regional warming, including the Atlantic Meridional Overturning Circulation, an ice surface reflectivity feedback in the high latitudes and evapotranspiration over land. The authors focus their analysis on the Met Office Hadley Centre climate model, but corroborate their results using four other global climate models. Adding the four models to their analysis shows that nonlinearities increase the spread in projected warming among the models. As the forcing increases, so does the uncertainty associated with the climate's response.
|
