The value of early methane mitigation in preserving Arctic summer sea ice

We develop four main emissions scenarios from 2020 to 2200: current policy, net zero CO₂ emissions, strong methane action, and then combined CO₂ and methane action. The current policy case is constructed from multiple data sources (text S1 and figure S1 available online at stacks.iop.org/ERL/17/044001/mmedia), and can be considered as middle-of-the-road among various existing future scenarios with minimal climate action and results in around 4 °C of warming by end of century (figure S1(a)). For CO₂ mitigation, we focus on net zero CO₂ pathways in which no further CO₂ are being added to the atmosphere through human activities beyond what can be removed by human intervention. Net zero CO₂ is chosen given recent convergence around the need for mid-century net zero CO₂ emissions to stabilize the climate and achieve the Paris Agreement temperature goals (IPCC 2018), and as many countries and businesses around the world have pledged to reach 'net zero' around mid-century (Levin et al 2020). For methane mitigation, we apply a 55% reduction from the current policy baseline (figure S2), which is considered achievable within 10 years with existing technologies (Ocko et al 2021). Note that we do not consider other constraints such as behavioral changes and socio-economic factors. We focus on CO₂ and methane mitigation given that other greenhouse gases and black carbon emissions play a limited role over the investigated timeframe (figure S3). In addition, the benefit of black carbon mitigation may be somewhat offset by concurrent mitigation of co-emitted cooling pollutants (Stohl et al 2015, Harmsen et al 2020). However, this assumption is not intended to downplay the role of black carbon emissions in contributing to climate change, especially over the short term.

For each of the three mitigation scenarios, we develop several pathways that encompass different mitigation magnitudes and implementation timelines, including nine CO₂ mitigation pathways, 18 methane mitigation pathways, and some combinations of the two (figure S2 and text S1). Global mean temperature responses to all emissions pathways are simulated by the reduced-complexity climate model Model for the Assessment of Greenhouse Gas Induced Climate Change (MAGICC) version 6 (Meinshausen et al 2011a). See text S2 for more details. All global mean temperatures presented in this paper are relative to the pre-industrial (1850–1900) level.

To determine the probability of a seasonally ice-free Arctic for each emissions pathway, we use a distribution of temperature thresholds that represent the range of global mean temperatures above which Arctic sea ice extent could fall below 1 M km² every summer—which is considered 'nearly sea ice-free' (Wang and Overland 2009). This temperature distribution has a mean of 2.45 °C and a 68% confidence interval of 2.2 °C–2.7 °C relative to the pre-industrial level (figure S4) following Iseri et al (2018). The mean value of the distribution is derived from a subset of Coupled Model Intercomparison Project (CMIP) 5 models that represents present-day sea ice conditions reasonably well and projects ice-free summer conditions to occur during 2041–2060 under the representative concentration pathway (RCP) 8.5 scenario (Massonnet et al 2012). This subset of models project ice-free summer to occur above 2.44 °C global mean temperature increase on average with a range of 1.9 °C–3.1 °C. A Gaussian distribution is then applied to represent the uncertainties of summer sea ice temperature thresholds. It is important to note that this distribution is a simplified assumption of the probability density of certain temperature threshold which remains difficult to constrain (Stroeve and Notz 2015, Notz and SIMIP Community 2020, Senftleben et al 2020).

We use the joint kernel density estimation (KDE; figure S4) method to estimate the likelihood of a seasonally ice-free Arctic over time for each emissions pathway. Joint KDE, also known as joint probability density function, is commonly used to characterize the relationship between non-parametric distributions by indicating the likelihood of two events occurring at the same time. In this case, the two events are 'global mean temperature reaches X degrees' and 'summer sea ice temperature threshold is Y degrees'. Once the probability of any X and Y occurring at the same time is determined (shown in figure S4(c)), one could integrate the density function in the domain where X is larger than Y (area below the diagonal line in figure S4(c)) to determine the total probability of global mean temperature exceeding summer sea ice temperature threshold. For each resulting temperature trajectory in our analysis, we use joint KDE to calculate the probability of exceeding the temperature threshold of summer sea ice every decade through 2100 (figure S4(d)). The level of likelihood is defined according to the IPCC uncertainty guidance—unlikely (0%–33%), as likely as not (33%–66%), likely (66%–100%), very likely (90%–100%) and virtually certain (99%–100%) (Mastrandrea et al 2010).

The uncertainties of the risk estimates for summer sea ice mainly come from three sources: emission scenarios (including mitigation magnitude and timeline), MAGICC simulations of global temperature outcomes, and the response of summer sea ice to certain levels of temperature increase.

The uncertainty of future emission scenarios is analyzed by applying various pathways to each mitigation scenario (section 2.1 and text S1). We do not attempt to represent all the numerous possible emission scenarios the world could pursue, rather a reasonable range of scenarios that are aligned with climate policy discussions and/or technological feasibility.

The uncertainty of global temperature simulations comes from both the physical understanding of the climate system (often reflected in climate sensitivity) and the approach of utilizing a reduced-complexity climate model. MAGICC is one of many reduced-complexity climate models that are widely used to rapidly reproduce the global aggregated characteristics of complex GCMs (Meinshausen et al 2011a), and its performance has been evaluated by previous studies (Meinshausen et al 2011b, Ocko et al 2018; see text S2 for more details). To address the uncertainty of climate sensitivity, we conduct a 190 member ensemble simulation with parameterizations that mimic more complex GCMs and carbon cycle models. The range of temperature outcomes for any given year would reflect inter-model differences in their climate response to anthropogenic emissions, which is then incorporated into the joint probability calculations. However, we note that MAGICC does not simulate unforced variability (i.e. internal variability) of GCMs which could further broaden the range of temperature outcomes in any given year and affect the risk estimate for summer sea ice. It also does not include any spatial variations in radiative forcing or temperature response, thus the spatial variations of summer sea ice are not discussed in this study.

The uncertainty of summer sea ice response to certain levels of temperature increase is represented by varying the distribution of summer sea ice temperature threshold. Current estimations of when the Arctic will become nearly ice-free in summer contain large uncertainties (e.g. Armour et al 2011, Massonnet et al 2012, Li et al 2013, Guarino et al 2020, Notz and SIMIP Community 2020, Senftleben et al 2020). GCMs exhibit a wide range of nearly ice-free dates (and corresponding global mean temperatures) likely due to differences in initial ice volume and thickness, seasonal cycle of sea ice properties, and strength of various feedbacks in the Arctic region. For instance, the latest CMIP6 models estimate that the first nearly ice-free September will occur above ∼2.2 °C increase in global mean temperature on average with a range of 0.9 °C–3.2 °C (Notz and SIMIP Community 2020). Considering the large temperature range, we conduct two sets of sensitivity tests with higher and lower summer sea ice temperature thresholds, respectively, to present a range of outcomes due to uncertainties of summer sea ice response to temperature. The lower temperature threshold has a mean of 2.2 °C and 1.95 °C–2.45 °C as the 68% confidence interval; the higher temperature threshold has a mean of 2.7 °C and 2.45 °C–2.95 °C as the 68% confidence interval.

Our analysis shows that methane mitigation can play a major role in reducing the likelihood of a seasonally ice-free Arctic. Neither CO₂ nor methane mitigation alone is enough to preserve summer sea ice through this century, let alone beyond, but combined they can lower the likelihood of losing summer sea ice to below 30% by the end of the century and even through 2200, a level which is characterized as unlikely. This can be done without relying on a large amount of negative emissions.

If global CO₂ emissions were to peak today and reach net zero in 2050—the goal of many climate policies focused on not exceeding 1.5 °C level of long-term warming—we find that the likelihood of an ice-free Arctic summer would be 50%, as likely as not, by 2100 (figure 1; blue lines) and 69%, likely, by 2200 (figure S5). If sustained net negative CO₂ emissions are achieved after reaching net zero, the likelihood of an ice-free Arctic summer will further decrease depending on the scale of negative emissions (figure S5). For example, achieving −10Gt (−20Gt) of CO₂ emissions per year from 2060 onwards would lead to a 39% (26%) likelihood of an ice-free Arctic summer by 2200. However, achieving these amounts of negative CO₂ emissions depends on the large-scale deployment of technologies that are not yet scalable. This means that while preserving summer sea ice is possible with strong CO₂ policies alone, it is far from certain in the absence of reductions in emissions of other major greenhouse gases (most prominently methane). While strong CO₂ policies are likely to have the effect of reducing emissions of other pollutants, this is not guaranteed; for example, carbon capture and storage solutions for fossil fuel systems will not reduce methane emissions from fossil fuel supply chains; several major methane sources are not related to energy use (e.g. agriculture and waste); and the energy transition from coal to natural gas and potentially biogas (i.e. methane from fermentation of organic matter) complicates matters further.

The continued risk of losing summer sea ice despite dramatic cuts in CO₂ is primarily due to the major role that methane plays in current and future warming (Naik et al 2021). Methane emissions are projected to nearly double by the end of the century compared to today's level in the absence of mitigation measures, with the majority of the emissions emanating from the livestock, oil and gas, and landfill sectors as populations grow (Höglund-Isaksson et al 2020, Ocko et al 2021; figure 2(a)). These emissions may contribute to nearly 1 °C increase in global mean temperature by 2100 (Ocko et al 2021), which would increase the risk of losing summer sea ice. Ocko et al show that half of the anticipated methane emissions can be reduced with existing technologies, which can lead to a 30% decrease in near-term warming rate and avoid more than half a degree Centigrade of warming by 2100 (Ocko et al 2021). While methane mitigation alone cannot prevent a seasonally ice-free Arctic (figure 1; orange lines), a combination of strong methane and CO₂ policies (all technically feasible methane mitigation measures implemented by 2030 and net zero CO₂ by 2050) can reduce the likelihood of an ice-free Arctic summer by 2100 to as low as 19%, and thus unlikely (figure 1; green lines). Through 2200, the likelihood would remain below 30%; thus unlikely to be ice free (figure S5). Note that the extent to which methane mitigation can help preserve summer sea ice is dependent on the magnitude of the mitigation (figure S6), and therefore the potential benefit could be reduced if not all feasible measures were implemented or it could grow as technologies improve.

Zoom In Zoom Out Reset image size

The benefit of combined methane and CO₂ policies on summer sea ice (80% lower likelihood compared to current climate policy) is much greater than the simple sum of benefits from two single gas policies (57% lower likelihood: 50% from net zero CO₂ and 7% from strong methane policy), even though the global mean temperature benefits are linearly additive between the two. This is because the probability of losing summer sea ice is more sensitive to global mean temperature change when it is close to the temperature threshold of summer sea ice. Combined mitigation lowers the global mean temperature to below 2 °C on average, which has little overlap with the summer sea ice temperature threshold, while CO₂ or methane only mitigation can only lower global temperature to around 2.5 °C and 3.5 °C, respectively. This amplifying effect of combined mitigation actions remains robust across varying summer sea ice temperature thresholds (figure 1(b); vertical bars). Therefore, multiple mitigation actions tend to amplify the benefit of each other in lowering the probability of losing summer sea ice. This insight highlights the need to consider climate impacts beyond temperature when evaluating the benefit of emissions mitigation polices, because the overall climate benefit is not necessarily proportional to global mean temperature.

The benefits of methane mitigation in lowering the risk of losing summer sea ice are also dependent on when mitigation measures are deployed (figure 2). Each decade of delay in implementing a strong methane policy would reduce the benefits of such policy on reducing the risk of losing summer sea ice.

To examine how the timeline of methane mitigation affects the risk of losing summer sea ice, we analyze two sets of emissions scenarios with different timelines for starting and achieving a 55% reduction in anthropogenic methane emissions below the reference scenario: 'slow mitigation' and 'delayed mitigation' (figure 2). The methane mitigations are added on top of achieving net zero CO₂ emissions by 2050. The slow mitigation scenarios begin implementing methane mitigation measures in 2020 and reach 55% reduction relative to the reference scenario in 10–80 years (figure 2(a)). The delayed mitigation scenarios start implementation between 2020 and 2080 and reach 55% reduction within a decade (figure 2(c)). These scenarios are illustrative of a wide diversity of plausible pathways to achieving a 55% cut in anthropogenic methane emissions, with varying start dates and implementation rates.

The slow mitigation scenarios show similar probability of losing summer sea ice throughout the century (figure 2(b)), while the delayed mitigation scenarios show higher probability of losing summer sea ice each decade the implementation is delayed (figure 2(d)). This indicates that the benefit of methane mitigation is more sensitive to the implementation start year than the implementation rate. This is likely due to the shifting of the center of the joint probability distribution with temperature change (figure S7). Early methane mitigation, regardless of implementation rate, reduces methane emissions immediately and rapidly slows the rate of warming, which ultimately results in a lower end-of-century temperature compared to delayed methane mitigation. Therefore, the chances of global mean temperature exceeding the summer sea ice temperature threshold stays small throughout the century. This is indicated as the peak density of the joint probability pattern stays further away from the integration domain (under the diagonal line), so that the integrated probability is small and less sensitive to changes in global mean temperature due to different implementation rates (figures S7(b) and (c)). On the other hand, delayed action results in a higher end-of-century temperature that shifts the peak density of the joint probability pattern closer to the integration domain, thus the integrated probability is larger and more sensitive to the changes in global mean temperature due to different implementation start dates (figure S7(d)). The greater importance of early implementation date than fast implementation rate is also evident with other levels of concomitant CO₂ mitigation (figure S8).

A key result is that if mitigation begins in 2020, the rate of implementing methane mitigation has a small influence on the likelihood of a seasonally ice-free Arctic throughout the century. In other words, immediate mitigation of methane emissions allows us to achieve a 55% methane emissions reduction over several decades and still retain most of the benefit in reducing the probability of losing summer sea ice. However, slower implementation of early mitigation would lead to a faster rise in global mean temperature (Ocko et al 2021), which could affect the temporal evolution of summer sea ice conditions, as sea ice extent is directly linked to global mean temperature (Notz and Stroeve 2018). It could also worsen other climate impacts in the near-term such as ocean warming and extreme weather events (IPCC 2019, Fischer et al 2021).

Further, there are benefits of early methane action irrespective of the ultimate CO₂ mitigation timeline (figure 3). Figure 3 shows the probabilities of having a seasonally ice-free Arctic by 2100 under net zero CO₂ pathways with various implementation timelines (blue markers) and in combination with fast deployment of strong methane mitigation (green markers). Two groups of CO₂ emissions scenarios are included: one group peaks today and reaches net zero between 2050 and 2080; the other group peaks in 10–30 years from now and reaches net zero three decades after peaking (figure 3(b)). The end-of-century probability of losing summer sea ice under these CO₂ emissions scenarios ranges from 49% to 83% and increases with slowed or delayed CO₂ mitigation (figure 3(a)). The risk of losing summer sea ice is linearly correlated with the cumulative CO₂ emissions prior to net zero being achieved (figure S9).

Zoom In Zoom Out Reset image size

When strong methane mitigation (figure 3(c); 55% below baseline by 2030) is deployed in addition to these net zero CO₂ emissions scenarios, the end-of-century probabilities of losing summer sea ice are consistently reduced by 24%–30% (figure 3(a)). We also test the sensitivity of the added benefit of strong methane policy to two other summer sea ice temperature thresholds (described in section 2.3). The results show that strong methane policy brings substantial reductions to the probability of losing summer sea ice across various timelines of net zero CO₂ emissions and different choices of temperature thresholds (vertical bars in figure 3(a)). It also shows the value of early CO₂ action given that less CO₂ is emitted in pathways where net zero is achieved in earlier decades.