A novel method to test non-exclusive hypotheses applied to Arctic ice projections from dependent models

Abstract

A major conundrum in climate science is how to account for dependence between climate models. This complicates interpretation of probabilistic projections derived from such models. Here we show that this problem can be addressed using a novel method to test multiple non-exclusive hypotheses, and to make predictions under such hypotheses. We apply the method to probabilistically estimate the level of global warming needed for a September ice-free Arctic, using an ensemble of historical and representative concentration pathway 8.5 emissions scenario climate model runs. We show that not accounting for model dependence can lead to biased projections. Incorporating more constraints on models may minimize the impact of neglecting model non-exclusivity. Most likely, September Arctic sea ice will effectively disappear at between approximately 2 and 2.5 K of global warming. Yet, limiting the warming to 1.5 K under the Paris agreement may not be sufficient to prevent the ice-free Arctic.

Refining projected multidecadal hydroclimate uncertainty in East-Central Europe using CMIP5 and single-model large ensemble simulations

“…Action taken in this direction introduced the application of diverse model ranking methodologies, ranging from studies using correlation, root-mean-square error, and variance ratio (Boer and Lambert 2001;Gleckler et al 2008) to the application of prediction indices (Murphy et al 2004) or to those taking a Bayesian approach (Min and Hense 2006). In addition, the concern of interdependency of CMIP models (Sanderson et al 2015) has been re-evaluated recently (Olson et al 2019). Regarding the target area of model evaluation Garfinkel et al (2020) studied the sources of CMIP5 intermodel spread in precipitation changes globally, however, ample analyses are targeted at more regional areas, e.g., the North-Atlantic (Perez et al 2014), parts of Europe (Coppola et al 2010;Pieczka et al 2017), Africa (Brands et al 2013;Dyer et al 2019;Yapo et al 2020), South-America (Lovino et al 2018), or Asia (Ahmed et al 2019).…”

A Markov chain method for weighting climate model ensembles

“…An unresolved issue in using weights for models is that models have interdependence, due to the sharing of computer codes, parameterizations, etc. (Olson et al, 2019). Abramowitz et al (2019) points out that model dependence can play a crucial role when assembling the models into an ensemble.…”

“…Although Markov chains have been used frequently in the literature for the prediction of future time series (e.g., Bai and Wang, 2011;Pesch et al, 2015), to the best of our knowledge, this is the first time the method has been applied to building weighted climate model ensemble means. In this paper, we use the "memoryless" property of Markov chains at each time step to capture the dynamic change in models' fit through the time series.…”

Bayesian Inference of Spatio-Temporal Changes of Arctic Sea Ice

“…Related research on calibrating spatial binary outputs from computer models of Antarctic ice sheets can be found in Chang et al (2016a,b). More recently, Olson et al (2019) proposed a new method for assessing the dependence of Arctic sea ice on climate variables (in the form of output from climate models), for the purpose of forecasting the Arctic sea ice. In contrast to these papers, the article by Director et al (2017Director et al ( , 2019 takes a decidedly statistical approach, using a spatio-temporal statistical model to forecast selected contours of sea-ice concentration.…”