The rapid loss of sea ice in the Arctic is one of the most striking manifestations of climate change. As sea ice melts, more open water is exposed to solar radiation, absorbing heat and generating a sea-ice–albedo feedback that reinforces Arctic warming. Recent studies stress the significance of this feedback mechanism and suggest that ice-free summer conditions in the Arctic Ocean may occur faster than previously expected, even under low-emissions pathways. Here we use an integrated assessment model to explore the implications of a potentially rapid sea-ice-loss process. We consider a scenario leading to a full month free of sea ice in September 2050, followed by three potential trajectories afterward: partial recovery, stabilization, and continued loss of sea ice. We analyze how these scenarios affect the efforts to keep global temperature increase below 2°C. Our results show that sea-ice melting in the Arctic requires more stringent mitigation efforts globally. We find that global CO2 emissions would need to reach zero levels 5–15 years earlier and that the carbon budget would need to be reduced by 20%–51% to offset this additional source of warming. The extra mitigation effort would imply an 18%–59% higher mitigation cost to society. Our results also show that to achieve the 1.5°C target in the presence of ice-free summers negative emissions would be needed. This study highlights the need for a better understanding of how the rapid changes observed in the Arctic may impact our society.
The rapid decline of Arctic sea-ice extent in the past few decades is one of the most evident indicators of global warming. One direct consequence of this phenomenon is the sea-ice–albedo feedback (SIAF), which amplifies Arctic temperature changes. A better understanding of the processes leading to this accelerated sea-ice loss has been recognized as one of the “grand challenges” of climate science [Kattsov et al., 2010], as Arctic changes are going to have profound climatic [Liu et al., 2012], ecological [Post et al., 2013], economic [Gautier et al., 2009; Smith and Stephenson, 2013], and societal [Laidler et al., 2008; Newton et al., 2016] implications not only for the northern regions, but for the entire globe [Lenton et al., 2008; Duarte et al., 2012].
Since satellite data records began in 1978, Arctic sea-ice extent has been showing persistent and significant reductions across all months. Generally, this decrease has been most pronounced for the month of September, at the end of the melting season. In September 2012, when the last record minimum was registered, Arctic sea-ice extent was 3.3 × 106 km2, equivalent to a 50% reduction compared to the sea-ice cover during the early 1980s. On average, Arctic sea-ice extent declined at a rate of 7.5% per decade in the period 1979–2001, and at 22.2% per decade within the period 2001–2015 [ARC, 2015]. The retreat in the extent of sea ice is part of an ongoing, more abrupt decline in ice thickness, volume, and age [ARC, 2015].
An ice-free summer (i.e., a sea-ice extension of less than 1 million square kilometers) in the Arctic Ocean is now projected to occur earlier than previously expected [Stroeve et al., 2007, 2012]. According to the IPCC AR5, it could occur around 2040–2060 [IPCC, 2013] under a high-emissions scenario (viz. RCP8.5). Other authors [Wang and Overland, 2012; Overland and Wang, 2013], who combine information from climate models with observational data, suggest that an Arctic sea-ice-free summer by the 2030s is more likely. If the estimated volume of lost sea ice is used as proxy for future sea-ice extent, some authors [Maslowski et al., 2012] project ice-free summers in the Arctic Ocean by 2020 or even earlier. Finally, given the vulnerability of sea ice, some authors [Serreze et al., 2007] anticipate that an abrupt episode due to natural climate variability could also trigger a faster transition of sea-ice loss.
Due to the inertia of the climate system, sea-ice loss during the first half of the century is not expected to be strongly dependent on the emissions scenario [Overland et al., 2014]. Therefore, it turns out that an Arctic sea-ice-free summer is also possible under low-emissions pathways [Mahlstein and Knutti, 2012]. For instance, within the Coupled Model Intercomparison Project 5 (CMIP5), this is the case for 15% of the simulations ran through 2100 under a 2°C scenario [Notz, 2015]. The results of these models show that the trends of sea-ice loss until 2050 are indeed very similar for high- and low-emissions scenarios [Hezel et al., 2014].
A direct consequence of sea-ice-free conditions is the occurrence of the SIAF, which refers to the process when sea ice (with high albedo/reflectivity) melts and more open water (with low albedo) is exposed to solar radiation, therefore absorbing more energy and generating a self-reinforcing warming mechanism. Although this is not the only feedback mechanism associated to sea-ice loss, it has been identified to have a central role in recent Arctic temperature amplification [Screen and Simmonds, 2010]. Various studies have estimated the change in albedo due to sea-ice melting [Riihelä et al., 2013] and some have estimated the impact on regional or global radiative forcing. Flanner et al.  estimated the global annually averaged radiative forcing caused by the sea-ice loss in the Arctic between 1979 and 2008 to be around 0.11 W m−2. Hudson  calculated that a complete removal of Arctic sea ice would result in an annual forcing of about 0.7 W m−2, and that “a more realistic sea-ice-free summer scenario,” with no sea ice for 1 month in September, would result in an annual forcing of about 0.29 W m−2.
Here we study the implications of mitigating climate change to below 2°C levels in the presence of the SIAF. This feedback is currently not incorporated into integrated assessment models [Stern, 2016; Knutti and Hegerl, 2008] and, therefore, its direct implications have not been addressed yet. We study the consequences of a sea-ice-free month in September, as described by Hudson , assuming that it may occur by 2050. Given the current trends, this scenario seems plausible, although it may be considered a rapid transition by some climate models. Finally, based on the current debate about a potential recovery of Arctic sea ice in a low-carbon scenario, we consider three different sea-ice processes afterward: partial recovery, stabilization, and continued sea-ice loss.
2 Model and Scenarios
2.1 Integrated Assessment Model
We use the Dynamic Integrated Climate–Economy model (DICE version 2013R), which is an integrated assessment model that has been used in the analysis of the implications of different climate policies and pathways [Nordhaus, 1992; Moore and Diaz, 2015; González-Eguino and Neumann, 2016]. In our study, an optimal control rate for fossil fuel and industrial CO2 emissions is sought that maximizes the net present value of cumulative economic welfare from 2010 to 2100, subject to a constraint on global temperature change. In this approach, economic welfare corresponds to net welfare, that is, the damages from climate change and the mitigation costs have already been deducted. More information on the DICE model can be found in the model documentation [Nordhaus and Sztorc, 2013].
2.2 SIAF Scenarios
Given the high uncertainty related to projections of Arctic sea-ice melting, we characterize the scenarios of extra radiative forcing in a stylized way taking into account the possible projections of Arctic sea-ice evolution discussed in the literature. The proposed scenarios explore the implications of a transition toward an entire month free of sea ice in the Arctic Ocean in September 2050, with three different developments thereafter: stabilization, partial recovery, and continued loss (no recovery) of sea ice. We estimate the additional efforts required to maintain global mean temperature change to below 2°C (called 2°C Scenario). We compare these three SIAF scenarios with a baseline mitigation scenario that does not account for SIAF (Figure 1).
The scenarios are summarized as follows:
2°C: This is the baseline mitigation scenario. It assumes no extra global radiative forcing from Arctic sea-ice loss. This scenario introduces the constraint that global mean temperature change should be below 2°C in 2100, but allowing for overshooting. To be consistent with RCP scenarios from the literature [van Vuuren et al., 2011; IIASA, 2016] we consider that the exogenous radiative forcing from non-CO2 factors increases from 0.25 to 0.4 W m−2 and that land-use emissions are reduced progressively to zero by 2100, which are the values associated to the RCP2.6 Scenario.
2°C_SIAF_Stabilization: This scenario assumes a linear increase in the extra radiative forcing due to SIAF from 0.11 W m−2 in 2010 to 0.29 W m−2 in 2050. The value of 0.29 W m−2 is associated with an Arctic sea-ice-free September as estimated by Hudson . From 2050 onward, the Arctic sea-ice cover is assumed to remain stable at 0.29 W m−2 until 2100. This stabilization scenario is consistent with current model predictions of sea-ice loss under a low carbon scenario and by 2100 [Hezel et al., 2014].
2°C_SIAF_Recovery: This scenario matches 2°C_SIAF_Stabilization up to 2050 followed by a partial recovery of the Arctic sea-ice cover thereafter. We consider here that recovery takes place at half the speed of loss occurred during the period 1980–2010, leading to an extra radiative forcing of 0.19 W m−2 in 2100 (Figure 1). This recovery process is consistent with some recent studies, which suggest that Arctic sea-ice loss could be reversible and recover quite quickly in a cooling climate [Notz, 2009; Serreze, 2011; Tietsche et al., 2011].
2°C_SIAF_No-Recovery: This scenario also matches 2°C_SIAF_Stabilization up to 2050. However, in this case sea-ice loss continues during the second half of the century with the extra radiative forcing reaching 0.51 W m−2 by 2100. This scenario is consistent with some studies that contemplate the existence of a threshold or “tipping point” in the process of Arctic sea-ice loss [Holland et al., 2006; Wadhams, 2012].
3 Results and Discussion
For all four scenarios the optimal CO2 emissions mitigation effort to reach the 2°C target by 2100 is computed (Figure 2a). We assume that emissions reduction starts in 2020. The results indicate that fossil fuel and industrial CO2 emissions need to reach zero (corresponding to an emissions control rate of 1) by 2065 in the scenario without SIAF (2°C scenario). However, in the presence of SIAF the mitigation effort would need to be more stringent and emissions would have to be zero between 5 and 15 years earlier, depending on the scenario.
We analyze the implications of SIAF in terms of carbon budget available for limiting temperature increases to 2°C (Figure 2b). According to the DICE model, the CO2 budget from 2015 onward (in a 2°C scenario) is 1122 gigatons of CO2 (GtCO2), which is consistent with the range estimated by the IPCC-AR5 [IPCC, 2014b] and the most recent literature [Rogelj et al., 2016]. In the presence of SIAF, the carbon budget ends up being significantly lower, lying between 660 and 862 GtCO2, depending on the scenario. The carbon budget is one of the clearest indicators that the small window of opportunity for achieving the 2°C goal is reduced considerably in the presence of Arctic sea-ice loss.
CO2 concentration in the atmosphere is one of the control variables proposed to evaluate the status of the planetary boundary of climate change [Steffen et al., 2015]. In the presence of SIAF the CO2 concentration in the atmosphere would have to peak earlier and be lower during the entire century (Figure 2c). The peak in CO2 concentrations would decrease from 476 parts per million (ppm) to 447–459 ppm and be below current values (400 ppm) by the end of the century.
Given the strong current connection between emissions and economic activities, decarbonization efforts inevitably affect the global economy. A primary source of information on multidecadal costs of mitigation stems from integrated models such as the one used in this study [IPCC, 2014a]. A relevant economic indicator is the CO2 price, which quantifies the marginal costs of mitigation, that is, the cost per additional unit of CO2 reduction. The extra warming from SIAF requires more stringent mitigation efforts and, to achieve these, a higher carbon price is needed as cheaper mitigation options are progressively exhausted. In the presence of SIAF the price of CO2 must be higher, both now and in the future (Figure 3a). In 2020 the global carbon price in the mitigation scenario without SIAF is estimated at US$34 per ton of CO2 ($ tCO2−1). However, if we consider the SIAF effect, the price of CO2 increases by 38%–89%. The DICE model explicitly includes a backstop technology, which refers to a future technology that can eventually replace all fossil fuels at an initially “high” cost (350 $ tCO2−1 in 2010) declining over time due to technological progress. This explains the decreasing CO2 price in all scenarios during the second half of the century (Figure 3a). The global CO2 price reported here assumes full participation of all countries and well-functioning markets in all sectors. Therefore, it provides a benchmark for the lowest cost under these idealized implementation conditions.
In addition to the CO2 price we estimate the evolution of the mitigation cost in the period 2010–2100. The presence of SIAF would increase the present value of mitigation cost from U.S. $8 trillion to a range between U.S. $9.7 trillion and U.S. $11.5 trillion. Although these extra costs are significant, and can vary depending on the discount rates used [Nordhaus, 2007; Stern, 2008], they are relatively low when compared to the projected global economic growth during the century, with the estimated economic benefits that the Arctic provides to our economy in terms of regulation of the climate system [Euskirchen et al., 2013] or with our own estimations of the future extra damages from SIAF in a no climate policy scenario (see Appendix S1, Supporting Information). The mitigation cost as a percentage of the economic output (Figure 3b) is below 1% in the short-run and up to 3.7% at the maximum point (mid-century) of the worst-case scenario.
4 Sensitivity Analysis
The implications of the analysis are dependent on various factors. Here we concentrate on the uncertainties in relation to the date of sea-ice-free conditions and the choice of the climate target in 2100.
A key issue in studies of Arctic sea-ice loss is the date at which the sea-ice-free condition may occur. Predictions from different models and projections vary greatly, so that there is much uncertainty in relation to this timing. Table 1 shows the results of a sensitivity analysis of the date for a full month free of sea ice in the Arctic between 2040 and 2060. The extra radiative forcing depicted in Figure 1 is modified, assuming a linear trend that would reach a radiative forcing of 0.29 Wm−2 in 2040 or 2060. After 2050 the same rates from recovery/stabilization/no-recovery scenarios of Section 2 are applied. The results show that the earlier the date of the first sea-ice-free month occurs, the lower becomes the carbon budget. In the worst case situation of one month free of Arctic sea ice by 2040 without recovery, the carbon budget would have to be reduced from 1122 GtCO2 to 549 GtCO2, implying a 51% reduction. Similarly, in that situation the mitigation cost would increase from US $8 trillion to 12.7 trillion.
Table 1. Sensitivity analysis of timing for a sea-ice-free Arctic. The table shows the results of alternative timings of a full month free of Arctic sea ice between 2040 and 2060. A linear trend is set in order to achieve a radiative forcing of 0.29 W m−2 by that date, as in Figure 1. In all scenarios, after 2050 a recovery/stabilization/no-recovery trend is applied to the value achieved in 2050, as the first column shows. The table presents results of carbon budget for 2015 onward (GtCO2) and present value of mitigation cost between 2010 and 2100 (trillions of 2005 U.S.$)
Extra-Radiative Forcing 2050/2100 (W m−2)
Carbon Budget from 2015 (GtCO2) and Reduction with Respect to 2°C (%)
Mitigation Cost 2010–2100 (Trillions of 2005 U.S.$) and Reduction with Respect to 2°C (%)
Another important issue is the climate target selected. We assess the sensitivity of the mitigation effort toward climate targets ranging from 1.5°C to 2.5°C in 2100 (Figure 4). The results show that the carbon budget in a scenario without SIAF is 2112 GtCO2 for a 2.5°C target and 282 GtCO2 for the 1.5°C target. According to the DICE model, the 1.5°C target is not attainable for any of our SIAF scenarios. This implies that the way to achieve that 1.5°C target in the presence of SIAF would be through “negative emissions,” that is, extraction of CO2 from the atmosphere, an option that it is not captured in the DICE model. Mitigation costs increase with the level of climate target stringency. The result for mitigation costs show that in a scenario without SIAF the cost would be U.S. $4.1 trillion for a 2.5°C target and U.S. $11.2 trillion for a 1.75°C target. In the presence of SIAF the cost for reaching the 1.75°C target would increase to a value in the range between U.S. $14 trillion and U.S. $17.4 trillion.
Finally, we do not include in the sensitivity analysis of this section other sources of uncertainty discussed by Hudson  (e.g., changes in cloud cover or albedo) or the remaining uncertainties in the current levels of radiative forcing from SIAF. For example, Pistone et al.  doubled the current estimations for SIAF of Flanner et al.  and Hudson  used here. Moreover, our study only captures the SIAF forcing effect quantified by Hudson , whereas other authors, such as Caldeira and Cvijanovic , estimate that if all other feedbacks were also considered, a full year free of Arctic sea ice could produce a net radiative global forcing as large as 3 W m−2.
Arctic sea ice is a key indicator of global climate change because of its sensitivity to warming and its role in amplifying climate change through the SIAF. However, this feedback has not been incorporated into integrated assessment models and therefore its direct implications have not been addressed. Moreover, recent trends in sea-ice extent and volume in the Arctic show greater and faster losses than generally obtained from physical models. Although there is much uncertainty about the timing of the first sea-ice-free summer in the Arctic, it is imperative to study this phenomenon now, even if in stylized form [Lenton and Ciscar, 2012], in order to better understand its implications and guide current decision making. This is especially important if abrupt losses of Arctic sea ice—such as those of 2007 and 2012—become a recurrent phenomenon in the coming years.
Our study reveals the significant consequences of rapid Arctic sea-ice loss for keeping climate change to low levels. The sooner the sea-ice-free condition occurs, the more difficult it will be to control climate change, especially if sea-ice recovery does not occur. Emissions reductions should increase significantly compared to current mitigation scenarios that do not include Arctic sea-ice loss. Existing energy infrastructures would have to be replaced quicker and policy instruments that could make such improvements feasible would need to be adopted earlier. Therefore, the already difficult task of achieving the targets of the Paris Agreement may become even more challenging. Our results show that the only way in our scenarios to achieve the 1.5°C target in the presence of SIAF would be through negative emissions, which imply more risks and uncertainties for the future [Rogelj et al., 2015; Hansen et al., 2016]. The implications of considering Arctic sea-ice-free conditions for the transformation of the global energy system are severe.
Our current analysis, based on stylized scenarios of extra-radiative forcing, should be complemented in the future by considering more precise scenarios of sea-ice loss derived directly from physical models. In addition to the SIAF addressed in this study, other related feedbacks occurring in the Arctic should also be considered. A better understanding of the socioeconomic implications of the rapid environmental changes occurring in the Arctic also requires a more intensive collaboration between the integrated assessment community and climate scientists.
The authors thank Paul Hezel, Anthony Patt and two anonymous reviewers for valuable comments. This study received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement number 642260 (TRANSrisk project). M.G.-E. and I.A. acknowledge financial support from the Ministry of Economy and Competitiveness of Spain (ECO2015-68023) and the Basque Government (IT-799-13). M.B.N. (RYC-2013-13628) and S.H.. (RYC-2012-12167) acknowledge financial support from the Ramón y Cajal Research Fellowship of the Ministry of Economy and Competitiveness of Spain. The authors declare that there are no conflicts of interest. All data and model outputs of this article are available upon request to M.G.-E. (email@example.com).
María de Maeztu Excellence Unit 2023-2027 Ref. CEX2021-001201-M, funded by MCIN/AEI /10.13039/501100011033