Constraining the Emergent Constraints

Abstract

An accurate estimate of equilibrium climate sensitivity (ECS) is pivotal to humankind’s responses, including both the mitigation and adaptation, to future global climate change (not necessarily that of a distant future). However, the uncertainty in estimates of ECS remains large, as shown in the past assessments by the Intergovernmental Panel on Climate Change (IPCC) (see IPCC, 2013), though the level of understanding on the physics and dynamics of Earth’s climate system has improved considerably during the past four decades since the appearance of the Charney report (Charney et al., 1979). To narrow the gap in ECS estimates, a new approach, called the emergent-constraint method, has been developed during the past two decades. In this approach, a particular climate variable [referred to as the “predictor” in Brient (2020)], which is observable and hence available in the present climate conditions, for instance the changes in albedo or low-cloud fraction per degree of surface temperature variation, is first singled out as a variable that has a clear and definite relationship with the ECS [referred to as the “predictand” by Brient (2020)], i.e., the relationship is consistent across multi-model ensembles. Then, the ECS (predictand) can be constrained based on the observed probability distribution of that particular climate variable (predictor). By “emergent” it is meant that, while the ECS is basically a theoretical and unobservable value, it may emerge from the observable shorter-term variations of the past and present climate. It is unsurprising that, due to the complexity of the climate system and the inter-linkage of physical processes therein, various emergent constraints have “emerged” during the past two decades. Caldwell et al. (2018) systematically evaluated the robustness/weakness and the correlation of the existing 19 emergent constraints in the literature. While confirming shortwave cloud feedback as the main contributor to the correlations among the emergent constraints, Caldwell et al. (2018) cast more doubt than confidence on about 10 of the total 19 emergent constraints. Hall et al. (2019) further suggested a possible use of the emergent constraints in constraining climate extremes, teleconnections, and tipping points of the climate system. In this issue of Advances in Atmospheric Sciences, Brient (2020) provides a thorough survey on the concept of emergent constraints while emphasizing some fundamental issues associated with the concept—namely, the importance of physical understanding, observational uncertainties, and statistical inference in the uncertainty of emergent constraints. Furthermore, the emergent constraints on the changes in the earth system, in a wider sense than the ECS, including the hydrological cycle, carbon cycle, and regional patterns of climate change are also briefly reviewed, though understandably these constraints are even less robust given the lack of available observational data and more uncertain representation in models. Based on 11 available emergent constraints providing the best estimates of the ECS, Brient (2020) tentatively presents a combined “a posteriori” distribution of ECS, which is similar to the “a priori” distribution, but skewed toward a higher ECS [Fig. 4 in Brient (2020)]. However, the emergent-constraint-based posterior distribution does not narrow the spread in the original ECS distribution, suggesting the need for further constraining the emergent constraints. Given the accumulation of massive data about the climate system in the age of big data, the utilization of available data in constraining the ECS cannot be more natural. However, some fundamental issues should be addressed carefully before emergent constraints can really reduce the uncertainty in estimates of climate sensitivity. Indeed, several theoretical assumptions have been made implicitly when applying emergent constraints to constrain the