A new climate feedback framework (REFRAME)

全球气候变化预测仍然存在很大的不确定性，减少这些不确定性是制定政策的迫切要求。这些不确定性的核心是气候反馈--这些过程可以放大或抑制全球变暖，以应对外部气候压力，例如温室气体造成的压力。评估这些反馈的一个主要挑战是它们的不稳定性：现在的反馈与未来的不同，虽然我们只能观察前者，但我们需要后者来限制未来的气候预测。我认为，气候反馈的明显不稳定性是由于当前反馈框架中缺少物理因素，该框架假设反馈过程完全依赖于全球平均地表温度。相反，最近的研究表明，反馈也通过低云量的变化和太阳反射对表面温度的模式做出反应。因此，ReFrame将引入一个新的反馈框架，从数量上解释这种缺失的物理现象。然后它将利用这个新的框架来(1)量化气候变化和反馈之间的耦合，观察表明这在气候模型中是不现实的；以及(2)限制未来的气候变化，重点是强烈放大反馈和高全球变暖水平的风险。REFRAME将利用我之前的工作阐明地表温度模式和反馈之间的物理联系，从而最佳地利用我的专业知识。这一努力是及时和新颖的：我最近展示了统计学习在限制云反馈方面的力量，创造了新的机会，从这个项目将利用的短暂观测记录中限制气候变化。通过解决气候动力学理论中的一个根本问题，该项目有可能在与政策相关的规模上减少气候预测和碳预算的不确定性，从而影响未来10-20年的气候政策。

Substantial uncertainties remain in global climate change projections, and reducing these is an urgent requirement for policymaking. At the heart of these uncertainties are climate feedbacks - processes that can amplify or dampen global warming in response to an external climate forcing, e.g. due to greenhouse gases. A major challenge in assessing these feedbacks is their inconstancy: present-day feedbacks differ from those in the future, and while we can only observe the former, we need the latter to constrain future climate projections.I propose that the apparent inconstancy of climate feedbacks is due to missing physics in the current feedback framework, which assumes feedback processes to solely depend on global-mean surface temperature. Instead, recent work has shown that the feedbacks also respond to patterns of surface temperature, through changes in low cloud amount and hence solar reflection. REFRAME will thus introduce a new feedback framework that quantitatively accounts for this missing physics. It will then exploit this new framework to (1) quantify the coupling between climate variability and the feedbacks, which observations suggest is unrealistic in climate models; and (2) constrain future climate change, focusing on the risk of strongly amplifying feedbacks and high global warming levels.REFRAME will leverage my prior work elucidating the physical linkages between surface temperature pat- terns and the feedbacks, thus making optimal use of my expertise. The effort is timely and novel: I have recently demonstrated the power of statistical learning to constrain cloud feedback, creating new opportunities to constrain climate change from the short observational record which this project will exploit. By addressing a fundamental problem in the theory of climate dynamics, the project has the potential to reduce uncertainties in climate projections and carbon budgets on policy-relevant scales, thereby impacting climate policy over the next 10-20 years.