Abstract
How the pattern of the Earth’s surface warming will change under global warming represents a fundamental question for our understanding of the climate system with implications for regional projections. Despite the importance of this problem there have been few analyses of nonlinear local temperature change as a function of global warming. Individual climate models project nonlinearities, but drivers of nonlinear local change are poorly understood. Here, I present a framework for the identification and quantification of local nonlinearities using a time-slice analysis of a multi-model ensemble. Accelerated local warming is more likely over land than ocean per unit global warming. By examining changes across the model ensemble, I show that models that exhibit summertime drying over mid-latitude land regions, such as in central Europe, tend to also project locally accelerated warming relative to global warming, and vice versa. A case study illustrating some uses of this framework for nonlinearity identification and analysis is presented for north-eastern Australia. In this region, model nonlinear warming in summertime is strongly connected to changes in precipitation, incoming shortwave radiation, and evaporative fraction. In north-eastern Australia, model nonlinearity is also connected to projections for El Niño. Uncertainty in nonlinear local warming patterns contributes to the spread in regional climate projections, so attempts to constrain projections are explored. This study provides a framework for the identification of local temperature nonlinearities as a function of global warming and analysis of associated drivers under prescribed global warming levels.
Highlights
IntroductionAs the world warms there are observed patterns of local warming, such as arctic amplification and increased warming over land than ocean (Braganza et al 2003, Drost and Karoly 2012)
Previous studies have analysed the validity of assuming the pattern of warming will not change (through methods often referred to as ‘pattern scaling’ first proposed by Santer et al (1990)) and found that for single- or multi-model ensemble average changes the assumption of linearity largely holds (Frieler et al 2012, Tebaldi and Arblaster 2014, Seneviratne et al 2016, King et al 2018, Osborn et al 2018, Tebaldi and Knutti 2018)
The efficacy of pattern scaling has been discussed in reports from the Intergovernmental Panel on Climate Change (e.g. IPCC fifth assessment report; Collins et al 2013) with the view that in general, under transient global warming, the use of pattern scaling for large-scale projections holds
Summary
As the world warms there are observed patterns of local warming, such as arctic amplification and increased warming over land than ocean (Braganza et al 2003, Drost and Karoly 2012). The pattern of warming may change in a warmer world, with some locations experiencing an accelerated or decelerated warming per unit change in global warming as illustrated using idealised examples in figure 1(a). Nonlinear local change could occur due to factors such as changes in non-greenhouse gas forcings (Good et al 2015, King et al 2018), movement in storm tracks (Arblaster and Meehl 2006, Knutti et al 2015) or shifting sea-ice boundaries. Previous studies have analysed the validity of assuming the pattern of warming will not change (through methods often referred to as ‘pattern scaling’ first proposed by Santer et al (1990)) and found that for single- or multi-model ensemble average changes the assumption of linearity largely holds (Frieler et al 2012, Tebaldi and Arblaster 2014, Seneviratne et al 2016, King et al 2018, Osborn et al 2018, Tebaldi and Knutti 2018). The factors determining why different models may produce varying degrees of nonlinear change have not been explored previously
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call