Rashed Mahmood, Markus G. Donat, Pablo Ortega, Francisco J. Doblas-Reyes & Yohan Ruprich-Robert
Providing accurate and reliable information about future climate is a key part of informing adaptation policy, and reducing uncertainty by constraining future projections can contribute to better decision making. This paper details a new method of reducing uncertainty from internal climate variability by constraining climate projections for the next 20 years using decadal predictions, aligning the projections with natural climate variability phases. The team were able to improve the accuracy of their constrained projections when compared against the unconstrained ensemble, as well as projecting stronger summer warming in the Sahara, and western and southern Asia out to 2035. These findings help provide more accurate information about future climate changes, contributing to effective climate policy.
Climate change is an ongoing risk to many aspects of everyday life, including in the future. Policy and investment decisions on climate change adaptation require accurate, reliable and actionable climate information for the decades to come. Decadal climate predictions utilise initialisation (supplying the simulations with data about the current state of the climate system at the start of the prediction) to constrain the natural, internal climate variability in the predictions and reduce their uncertainty. This, however, makes them computationally expensive, often limiting them to only up to 10-year predictions. Longer climate projections, which typically provide climate change estimates until the end of this century, cannot utilise initialisation in this way, leaving them with larger uncertainty. Recent studies have explored ways of using decadal predictions to constrain projections instead, and this study presents a new approach to implement this at the global scale. Improving projections in this way can help provide more useful climate information to inform adaptation policy and reduce the risks posed by future climate change.
The team have developed a new method to provide near-term climate information with increased accuracy by selecting members from a large ensemble of climate projections that closely resemble a decadal prediction, using only those ensemble members that best align with natural climate variability. This approach enables skilful 20-year predictions with greater accuracy than the original unconstrained ensemble and can also provide different climate change estimates compared to the unconstrained ensemble, in particular by reducing the uncertainty in near-term projections. The team looked at several case studies of putting this technique into practice and they found that their constraint method reduces the probability of small summer temperature increases in the Sahara, and western and southern Asia, thereby suggesting a stronger summer warming in these regions until 2035. This is consistent with a warmer North Atlantic in the near-term projections until 2035 and illustrates how this new constraint method can leverage on global variability patterns and their interconnections. This new method is being applied in this study to single-model ensembles but has the capacity to be scaled up to large multi-model ensembles. This is subject to ongoing research and could form a valuable part of future climate projection techniques.
Methods and data
This paper uses large ensembles of climate projections and decadal predictions, both consisting of 40 members and performed using the Community Earth System Model. The projections ensemble ran from 1920 to 2100, with future emission scenarios (from 2006 onwards) using a high-emission future scenario (RCP8.5). The decadal prediction ensemble was initialised in each year from 1954 to 2015, with CMIP5 forcings used up to 2005 and RCP8.5 for 2006 onwards. Observational data used to evaluate the simulations came from the HadISST1.1 (sea surface temperature) and HadCRUT4.6 (global surface air temperature) datasets. Sea surface temperature anomalies were used to choose the 10 projection ensemble members with the highest pattern correlations with the initialised predictions, with the first 20 years of the projections after initialisation being the focus of this study.
This study describes a new methodology to improve the accuracy of near-term multidecadal climate projections, further than 10 years in the future. Improving these projections allows the earlier, more confident adoption of climate adaptation policies aimed at reducing the severity of climate change impacts.
Targeted adaptation to near-term climate change requires accurate, reliable, and actionable climate information for the next few decades. Climate projections simulate the response to radiative forcing, but are subject to substantial uncertainties due to internal variability. Decadal climate predictions aim to reduce this uncertainty by initializing the simulations using observations, but are typically limited to the next 10 years. Here, we use decadal predictions to constrain climate projections beyond the next decade and demonstrate that accounting for climate variability improves regional projections of 20-year average temperatures. Applying this constraint to climate projections of the near future until 2035, summer temperatures over land regions in Asia and Africa tend to show stronger changes within the warming range simulated by the larger, unconstrained, ensemble—consistent with a warm phase in North Atlantic variability. This improved regional climate information can enable tailored adaptation to climate changes in the coming decades.