Volume 9, Issue 7 e2021EF002152
Research Article
Open Access

Possibility of Stabilizing the Greenland Ice Sheet

Xiuquan Wang

Corresponding Author

Xiuquan Wang

Canadian Center for Climate Change and Adaptation, University of Prince Edward Island, Charlottetown, PE, Canada

School of Climate Change and Adaptation, University of Prince Edward Island, Charlottetown, PE, Canada

Correspondence to:

X. Wang,

[email protected]

Contribution: Conceptualization, Methodology, Software, Validation, Formal analysis, ​Investigation, Resources, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration, Funding acquisition

Search for more papers by this author
Adam Fenech

Adam Fenech

Canadian Center for Climate Change and Adaptation, University of Prince Edward Island, Charlottetown, PE, Canada

School of Climate Change and Adaptation, University of Prince Edward Island, Charlottetown, PE, Canada

Contribution: Resources

Search for more papers by this author
Aitazaz A. Farooque

Aitazaz A. Farooque

Canadian Center for Climate Change and Adaptation, University of Prince Edward Island, Charlottetown, PE, Canada

School of Climate Change and Adaptation, University of Prince Edward Island, Charlottetown, PE, Canada

Faculty of Sustainable Design Engineering, University of Prince Edward Island, Charlottetown, PE, Canada

Contribution: Resources

Search for more papers by this author
First published: 08 July 2021

Abstract

Recent acceleration in the retreat of the Greenland ice sheet under a warming climate has caused unprecedented challenges and threats to coastal communities due to the rising sea level and increasing storm surges. This raises a critical question from a climate mitigation perspective: Would there still be a chance to stabilize the Greenland ice sheet if the carbon reduction goals of the Paris Agreement could be met? Here, we show that there is indeed a possibility for stabilizing the Greenland ice sheet with the low-emission scenario of RCP2.6. In particular, RCP2.6 would potentially limit the warming in Greenland below 1°C within next 30 years and constrain its loss of ice sheet coverage below 10%. After 2050, the annual mean temperature in Greenland is likely to be stabilized and no further loss is expected to its ice sheet. However, the effective window for this chance will be closing after 2020. If no effective carbon reduction policies are being taken now, we are very likely to enter a continuous warming pathway and lose the chance of stabilizing the Greenland ice sheet.

Key Points

  • Future changes in the Greenland ice sheet coverage under various emissions scenarios are quantified using a regional climate model

  • The RCP2.6 emission scenario has the potential to stabilize the Greenland ice sheet after 2050

  • The effective window for this possibility will be closing after 2020, if no effective carbon reduction policies are being taken

Plain Language Summary

Quantifying the effects of global carbon reduction policies on shrinking the Greenland ice sheet is extremely important from a climate mitigation perspective, yet it remains poorly understood. In particular, it is unknown whether there is still a possibility for stabilizing the Greenland ice sheet given that some effective carbon reduction policies can be implemented following the 2015 Paris Climate Agreement. If the chance still exists, then the next question is when it is likely to happen. Here, we address these two questions for the first time by quantifying the effects of different carbon emission scenarios on stabilizing or minimizing the warming in Greenland and the loss to its ice sheet coverage. We show that the low-emission scenario of RCP2.6 is the only one with a possibility of stabilizing the Greenland ice sheet after 2050, by constraining the loss of ice sheet coverage below 10%. Our findings are of great significance not only for providing some hope for climate activists to slow down the loss of the Greenland ice sheet and protect our coastal communities from the rising sea level, but also for urging all nations to take immediate actions to reduce carbon emissions.

1 Introduction

The rising sea level together with increased storm surges has become one of the most challenging issues for coastal communities in the context of global warming. The retreating ice sheet in Greenland under a warming climate is one of the main contributors to sea level rise and has attracted tremendous research interests in recent years (Enderlin et al., 2014; Morlighem et al., 2014; Mouginot et al., 2019; Nghiem et al., 2012; Robinson et al., 2012; Shepherd et al., 2019; Simonsen et al., 2019; Sundal et al., 2011; Van Tricht et al., 2016). Particularly, modeling the surface mass balance of the Greenland ice sheet through regional climate models has been the main focus of previous studies (Alexander et al., 2015; Aschwanden et al., 2019; Fettweis et al., 2013; Lenaerts et al., 2019; Mouginot et al., 2019; Noël et al., 2018; Rae et al., 2012; Shepherd et al., 2019), with the purpose of projecting the ice sheet evolution over time. These studies suggest that there is still a large amount of uncertainties for simulating the temporal evolution of the Greenland ice sheet and significant improvements to model performance are still required. Although accurate estimation of the mass loss of the Greenland ice sheet and its contribution to sea level rise is essential for understanding and increasing the resilience of coastal communities to climate change (Thorne et al., 2018; Vitousek et al., 2017), it is also critically important to explore the effects of global carbon reduction policies on the shrinking Greenland ice sheet from a climate mitigation perspective. In particular, it is unknown whether there is still a possibility for stabilizing the Greenland ice sheet given that some effective carbon reduction policies can be implemented following the 2015 Paris Climate Agreement. If the chance still exists, then the next question is when it is likely to happen.

In order to answer these two questions, here we attempt to quantify the effects of different carbon emission scenarios on stabilizing or minimizing the warming in Greenland and the loss to its ice sheet coverage. Three representative concentration pathways (RCPs) are considered to reflect the impacts of low (RCP2.6), medium (RCP4.5), and high (RCP8.5) emissions of greenhouse gases caused by human activities. Our results suggest that the low-emission scenario of RCP2.6 does have the potential to stabilize the warming climate in Greenland after 2050s and prevent further loss to its ice sheet coverage. By contrast, the local climate in Greenland is likely to warm up continuously throughout this century under both RCP4.5 and RCP8.5. Due to the continuous warming, the spatial extent of ice sheet is expected to decrease by 15% (RCP4.5) and 25% (RCP8.5) by the end of this century relative to the reference period of 1970–2000. Our results highlight the urgency of taking immediate carbon reduction actions to minimize the warming in Greenland and stabilize its ice sheet. Otherwise, we are very likely to enter a continuous warming pathway associated with fast retreating the Greenland ice sheet and increasing sea level.

The paper is organized as follows: Section 2 describes the methodology and data used in this study; Section 6 presents the results for historical climate validation and future climate projection for Greenland, as well as future changes in the spatial coverage of Greenland ice sheet; and Section 10 provides in-depth discussions on the key results and states the main conclusions.

2 Methods and Data

2.1 Global Climate Model

In this study, the Hadley Centre Global Environment Model version 2 with Earth Systems components (denoted as HadGEM2-ES) is used to provide large-scale boundary conditions for high-resolution regional climate simulations over Greenland. The HadGEM2-ES provides a coupled atmosphere-ocean configuration with a vertical extension in the atmosphere to include a well-resolved stratosphere (Collins et al., 2008). In particular, its standard atmospheric component has 38 levels extending to 40 km height, with a horizontal resolution of 1.25° of latitude by 1.875° of longitude, which produces a global grid of 192 × 145 grid cells. This is equivalent to a surface resolution of about 208 × 139 km at the equator, reducing to 120 × 139 km at 55° of latitude (Collins et al., 2008). Its vertically extended version has 60 levels extending to 85 km height and can thus be used for investigating stratospheric processes and their influence on global climate. The oceanic component of HadGEM2-ES utilizes a latitude-longitude grid with a longitudinal resolution of 1° and a latitudinal resolution of 1° between the poles and 30° North/South, from which it increases smoothly to one third of a degree at the equator (C Jones et al., 2011). It provides 360 × 216 horizontal grid points and 40 unevenly spaced vertical levels (a resolution of 10 m near the surface) for ocean coverage (Collins et al., 2008). The earth-system configuration of HadGEM2-ES includes dynamic vegetation, ocean biology, and atmospheric chemistry to better reflect the internal interactions and feedback within the Earth system under a changing climate (C Jones et al., 2011). For example, the TRIFFID dynamic global vegetation model is incorporated into the HadGEM2-ES to allow time-varying land-use distributions in order to simulate the internal changes of land surface and its evolution in response to climate change (Cox, 2001; C Jones et al., 2011).

2.2 Regional Climate Model

The PRECIS (Providing REgional Climates for Impacts Studies) regional climate modeling system is used here to generate high-resolution climate projections over Greenland, driven by the boundary conditions from HadGEM2-ES under three emissions scenarios, including RCP2.6, RCP4.5, and RCP8.5. These three RCP emissions scenarios are commonly used to investigate the potential impacts of low (RCP2.6), medium (RCP4.5), and high (RCP8.5) emissions of greenhouse gases caused by human activities (Jubb et al., 2013; Van Vuuren et al., 2011). The PRECIS is a comprehensive physical model with consideration of both the atmosphere and land surface components of the regional climate system (R Jones et al., 2004). This enables it to simulate the physical processes within the climate system, including dynamical flows, atmospheric cycles, clouds, precipitation, radiative processes, and land surface dynamics (Wilson et al., 2005). The PRECIS provides two horizontal resolutions (i.e., ∼50 and ∼25 km) with 19 vertical levels using a hybrid coordinate system. It can be applied to any area of the globe to generate detailed climate scenarios and thus has been widely used in many recent studies for regional climate projections (Geetha et al., 2019; González-Zeas et al., 2019; Guo et al., 20172019; Wang et al., 2016; Zhou et al., 2018; Zhu et al., 2017). In this study, the PRECIS model is used to carry out one historical run from 1960 to 2005 and three future runs from 2006 to 2100 under three RCPs. All PRECIS runs are performed at a spatial resolution of 25 km.

2.3 Historical Climate Data

In order to validate the performance of the PRECIS historical simulations, we collected high-resolution gridded data sets from two public sources. One is the gridded Climatic Research Unit (CRU) Time-Series (TS) data version 4.03 (denoted as CRU, available at: http://dx.doi.org/10.5285/10d3e3640f004c578403419aac167d82). The CRU data provide month-by-month variations in climate over the period of 1901–2018, covering the global land surface (excluding Antarctica) with high-resolution grids (0.5 arc degree, ∼50 km). The CRU TS4.03 data variables are produced using angular-distance weighting (ADW) interpolation and include various climate variables, such as cloud cover, diurnal temperature range, frost day frequency, potential evapotranspiration, precipitation, daily mean temperature, monthly average daily maximum and minimum temperature, and vapor pressure for the period of 1901–2018 (Harris et al., 2014). Another data set used in this study is the gridded climate data from the WorldClim version 2.0 (denoted as WC, website: http://worldclim.org/version2). The WC data provide average monthly climate data for minimum, mean, and maximum temperature as well as precipitation for the period of 1970–2000. The WC data are generated by applying the ANUSPLIN-based spatial interpolation to the major climate databases from weather stations and the SRTM elevation database (Fick & Hijmans, 2017; Hijmans et al., 2005). The WC data provide climate data for different spatial resolutions from 30 arc seconds (∼1 km) to 10 arc minutes (∼18 km). In this study, the WC data with a resolution of 18 km are selected to minimize its spatial mismatch with the PRECIS modeling outputs. To deal with the spatial mismatch between these gridded data sets and the PRECIS outputs, both CRU and WC data sets are resampled to the 25 km grid spacing of PRECIS runs over Greenland. Note that the comparisons are conducted only for the common period (i.e., 1970–2000) covered by all data sets. This period is also used as the reference period for calculating future climatic changes projected by the PRECIS model.

3 Results

3.1 Reproduction of Historical Climate Over Greenland

In order to validate the performance of the PRECIS model in simulating historical climate over Greenland, here we compare the PRECIS outputs for the period of 1970–2000 to two observational gridded data sets (i.e., CRU and WC). As for climate variables, we focus on mean temperature (Tmean), maximum temperature (Tmax), minimum temperature (Tmin), and total precipitation (Precip), which are commonly used in many studies to evaluate the performance of climate models (Cannon, 2015; Fenech et al., 2002; Herger et al., 2018; McSweeney et al., 2015; Mendlik & Gobiet, 2016; Pierce et al., 2009). We first evaluate the capability of the PRECIS model in reproducing the spatial patterns of annual averages for these four variables. Following that, the seasonality of historical climatology simulated by PRECIS is validated.

Figure 1 shows the comparisons of Tmean, Tmax, and Tmin among PRECIS, CRU, and WC for the reference period, while their differences in these three temperature variables are presented in Figure S1 (see supporting information). First of all, it should be noted that there are significant discrepancies between the two observational gridded data sets (i.e., CRU and WC). In detail, the CRU data set consistently presents higher values in Tmax than the WC data set does across Greenland, with a median difference of 5.3°C. By contrast, the Tmin values of the CRU data set in the northeast are extremely lower than the WC data set although slightly higher values in Tmin are still observed in the remaining regions, resulting in a median difference of −2°C. This suggests that the diurnal temperature range presented in CRU is generally wider than that in WC, which further lead to a slight difference in Tmean by 1.3°C between CRU and WC. The major reason for these discrepancies is likely to be the lack of weather stations with long-term climate records, which makes it extremely challenging to provide sufficient coverage of the spatial variations in local climate over Greenland (Box, 2002; Cappelen, 2020; Schomacker et al., 2017; Steffen & Box, 2001; Steffen et al., 1996). In other words, a slight change or update in point-based weather data might be exaggerated considerably in the process of generating gridded-based data sets. Besides, the spatial mismatch in the gridded data sets (i.e., CRU at 50 km and WC at 18 km) could be another reason for the large discrepancies. Given that the focus here is on the evaluation of the PRECIS model, we will mainly discuss the comparisons of PRECIS simulations to these two gridded data sets (rather than investigating the exact reasons causing the inconsistency between them). In general, the model's performance is deemed to be reasonable, as long as it can generate similar results from at least one observational data set.

Details are in the caption following the image

Comparisons of Tmean, Tmax, and Tmin between model simulations from PRECIS and observational gridded data sets from Climatic Research Unit (CRU) and WorldClim (WC), 1970–2000. The spatial maps for the three temperature variables are shown in (a–i) and their statistical comparisons are shown in the boxplots presented in the right column (j–l).

As shown in Figures 1 and S1, the spatial patterns of Tmean, Tmax, and Tmin presented in the WC data set can be well reproduced by the PRECIS model, while apparent discrepancies between PRECIS and CRU are observed. Since the WC data set has a higher spatial resolution than the CRU does, it is reasonable to give higher credibility to spatial patterns generated by the WC data set. Thus, the similarity between PRECIS and WC indicates the reasonable performance of PRECIS in simulating the spatial patterns of three temperature variables across Greenland. In addition, PRECIS can reasonably reproduce the spatially averaged Tmean, Tmax, and Tmin presented by the WC data set. For example, the median values of Tmean, Tmax, and Tmin of WC data set (i.e., −19.5°C, −16.5°C, and −22.4°C) can be well simulated by PRECIS (i.e., −19.7°C, −16.6°C, and −23.2°C).

The comparisons of annual total precipitation between PRECIS and two observational data sets from CRU and WC are shown in Figure 2. Both CRU and WC data sets show that the high-precipitation areas (with the annual total precipitation over 1,000 mm) in Greenland are largely located in the southeastern regions, while the annual total precipitation in northwestern areas is usually less than 600 mm. In general, this spatial pattern of annual total precipitation can be captured by PRECIS. However, apparent spatial variations are present in PRECIS simulations in comparison to the smoothed patterns manifested in both CRU and WC. This reflects the added-value of high-resolution regional climate models with regard to capturing the spatial variability of precipitation, which has been widely recognized in previous studies (Guo et al., 2020; Loikith et al., 2018; Lucas-Picher et al., 2017; Rummukainen, 2016; Singh et al., 2017; Xu et al., 2018). While the spatial pattern of annual total precipitation in Greenland can be generally captured by PRECIS, it should be noted that PRECIS tends to overestimate the amount of precipitation in most of the coastlines, especially the high-precipitation areas in the southeastern coastlines. By contrast, most of the inland precipitations are underestimated by PRECIS, resulting in an overall underestimation of annual total precipitation over Greenland (i.e., 46% to CRU and 37% to WC).

Details are in the caption following the image

Comparisons of annual total precipitation simulated by PRECIS and two observational data sets from Climatic Research Unit (CRU) and WorldClim (WC) for the period of 1970–2000.

In order to evaluate the performance of PRECIS in simulating the seasonality of Tmean, Tmax, Tmin, and Precip over Greenland, here we further compare the monthly averages of these variables between PRECIS and two observational data sets from CRU and WC for the reference period (shown in Figure S2). Although the seasonality to Tmean shown in two observational data sets is in good alignment with each other, it should be noted that the diurnal temperature ranges for all months in CRU are consistently larger than those in WC. In addition, some misalignments in monthly precipitation are also observed between CRU and WC. This further confirms the large uncertainty in observational data sets for Greenland. The comparisons show that the PRECIS model performs reasonably well in capturing the seasonality of Tmean, Tmax, and Tmin over Greenland. However, the seasonality in precipitation simulated by PRECIS seems to be smoother than that of two observational data sets. In particular, the PRECIS model tends to overestimate the rapid decreases in precipitation when temperature warms up in spring or cools down in fall. As the extent of sea ice can change dramatically during these two seasons, the poor performance of PRECIS may suggest that the modeled precipitation is less sensitive to the influence of sea ice in the context of Arctic (Kopec et al., 2016). In other words, this may imply that the PRECIS is less reliable in capturing the influence of sea ice on the Arctic atmospheric simulations (Rinke et al., 2006).

3.2 Projections of Future Climate Over Greenland

In order to evaluate the potential warming in Greenland, here we consider three emission scenarios including RCP2.6, RCP4.5, and RCP8.5. This allows us to quantify the likely impacts of mitigation scenarios through carbon reduction on the local climate of Greenland. The future climate projections for Greenland based on PRECIS are divided into three 30-year periods (i.e., 2020s, 2050s, and 2080s) to help explore the short-, medium-, and long-term effects of different mitigation scenarios. The spatial patterns of future temperature and precipitation under different RCPs are shown in Figures S3–S6. It can be seen that the high-emission scenario (i.e., RCP8.5) would result in the most significant warming from 2020s to 2080s. In particular, the median of annual mean temperature in Greenland under RCP8.5 would rise from −17.3°C in 2020s, to −14.9°C in 2050s, and then to −12.1°C by the end of this century, suggesting an average warming by 4.8°C. It appears that the medium-emission scenario (i.e., RCP4.5) can relatively slow down the warming trend but still projects an average warming of 3°C in annual mean temperature from −17.5°C in 2020s to −14.5°C in 2080s. By contrast, the low-emission scenario (i.e., RCP2.6) is likely to constrain the average warming in Greenland below 1°C from −17.3°C in 2020s to −16.2°C in 2080s. In particular, the mean annual temperature is likely to be stabilized to −16°C with some minor variability less than 0.5°C after 2050s. Similar patterns are also observed for Tmax and Tmin under three RCPs. In detail, Tmax is expected to increase from −14.4°C in 2020s to −9.5°C in 2080s under RCP8.5 (with an average increase of 4.9°C), while Tmin is expected to increase from −20.6°C in 2020s to −15.2°C in 2080s under RCP8.5 (with an average increase of 5.4°C). In comparison, the expected warmings in Tmax and Tmin from 2020s to 2080s under the medium-emission scenario (i.e., RCP4.5) are 3°C and 3.1°C, respectively. The low-emission scenario (i.e., RCP2.6) would potentially limit the average increase in both Tmax and Tmin to 1°C from 2020s to 2080s. The average Tmax in Greenland is likely to be stabilized to roughly −13.5°C after 2050s under RCP2.6, while its average Tmin is expected to be stabilized to −20°C.

The annual total precipitation over Greenland is likely to increase under all RCPs although the magnitude of increasing precipitation can vary significantly depending on the warming trends under various emission scenarios. Particularly, the most significant increase in annual total precipitation would be expected under the high-emission scenario (i.e., RCP8.5) and the least increase is projected for the low-emission scenario (i.e., RCP2.6). This apparent correlation between warming temperature and increasing precipitation does reflect the impacts of global warning on regional hydrological cycles. In other words, a warming climate can intensify the hydrological cycle in multiple ways such as increasing cloudiness, latent heat fluxes, and precipitation (Neelin et al., 2017; Wang et al., 2014). In detail, the median value of annual total precipitation over Greenland under RCP8.5 is expected to increase from 289 mm in 2020s to 428 mm in 2080s, suggesting an increase by 48%. In comparison, the projected increase in annual total precipitation under RCP4.5 can drop to 27% from 276 mm in 2020s to 350 mm in 2080s. The increase can be further reduced to 7.2% from 291 mm in 2020s to 312 mm in 2080s under the low-emission scenario (i.e., RCP2.6). This suggests that the local climate in Greenland is highly sensitive to the increasing greenhouse-gas concentration in the atmosphere. Many previous studies suggest that some minor changes to the local climate of Greenland can lead to significant damages to its ice sheet, which further contribute to the rising sea levels (Aschwanden et al., 2019; Cuffey & Marshall, 2000; Dowdeswell, 2006; Harper et al., 2012). Therefore, it is extremely important to reduce carbon emissions according to the pathways specified by RCP2.6 such that the changes to the local climate of Greenland can be minimized.

In order to evaluate the potential changes to the seasonality of temperature and precipitation in Greenland, here we further compare future projections of monthly Tmean, Tmax, Tmin, and Precip to the observations in the reference period of 1970–2000. The comparisons of seasonal patterns of these four variables are shown in Figure S7 and the changes for future periods under various RCPs relative to the reference period are presented in Figures S8–S11. In general, there would be no noticeable changes in terms of seasonal patterns of temperature and precipitation, suggesting that no seasonal shifts would be expected in Greenland. For example, the highest temperature typically occurs in July, while the lowest temperature is usually expected in February. Although such a seasonal pattern in Greenland is unlikely to change in the context of global warming, both monthly temperature and precipitation are projected to increase throughout this century. Similar to the annual averages shown in Figures S3–S6, the magnitude of increases in monthly averages of temperature and precipitation is also correlated to emission scenarios. In other words, the high-emission scenario (i.e., RCP8.5) would result in the most significant increases in both temperature and precipitation, while the least increases are expected under the low-emission scenario (i.e., RCP2.6). In addition, it is worthwhile to mention that the mean temperature in summer (i.e., June, July, and August) under RCP8.5 would be completely above freezing by the end of this century. This would speed up the melting of ice sheet over Greenland and lead to significant increases to the mean sea level. The projected increase in precipitation can also be potentially balanced by the increase in surface melt due to the continuously warming temperature.

While inevitable warming is expected for all months under all RCPs, it should be noted that cold months (e.g., December, January, and February) are likely to experience larger magnitude of increases in temperature than warm months (e.g., June, July, and August). For example, the increase in February's mean temperature in 2080s under RCP8.5 can be as high as 10°C but the increase in July under the same scenario is only about 5°C, suggesting a weakening seasonal pattern of temperature in Greenland. Besides, the magnitude of warming for Tmin is usually higher than that of Tmax, indicating that certain decreases in diurnal temperature would be expected. In comparison, the pattern of monthly precipitation changes is less clear than that of temperature although all months are likely to experience more precipitation under a warming climate. Considering that the PRECIS model does not fully capture the seasonality of monthly precipitation for the reference period (see Figure S2), we are less confident to draw any further conclusions about the potential changes in seasonal patterns of precipitation.

Here, we should note that, similar to other regional climate models used in previous studies (Alexander et al., 2015; Fettweis et al., 2013; Lenaerts et al., 2019; Noël et al., 2018; Rae et al., 2012), the PRECIS model does provide a full set of surface variables (e.g., runoff, evapotranspiration, sublimation, and snow mass) which could be used to understand the future evolution of the Greenland surface mass balance. However, we feel less confident to further evaluate these surface variables due to its poor performance in simulating precipitation, considering that: (a) precipitation dominates the mass input for the Greenland ice sheet mass balance, and (b) precipitation is an atmospheric variable which is less likely to be affected by various surface conditions of the Greenland ice sheet (either unknown or not well understood) than any surface variables. Instead, we decide to use the temperature projections only to investigate if it is possible to stabilize the Greenland ice sheet given that the PRECIS does perform reasonably well in simulating near-surface air temperature over Greenland.

3.3 Future Changes in the Greenland Ice Sheet Coverage

Since the ice sheet over Greenland has a great potential to raise the sea level in the context of global warming, here we further explore how the spatial extent of ice sheet over Greenland would change in the context of global warming. Due to the poor performance of climate models in simulating precipitation (including the PRECIS model used in this study), we only use future temperature projections to address this question. In particular, we use the concept of ice cap climate to approximately determine if an area will be covered by ice sheet or not. The ice cap climate is defined as a climate with no mean monthly temperature above 0°C (Hess, 2011). This climate is usually found around the North and South Pole as well as on the top of high mountains. Since the temperature never exceeds the melting point of ice, all snow or ice will accumulate and form a large ice sheet over time. As demonstrated in our validation analysis, the PRECIS model performs very well in reproducing the mean temperature over Greenland. Therefore, here we use the high-resolution projections of mean temperature from PRECIS to help quantify the potential changes in the spatial coverage of ice cap climate over Greenland. The results can thus be used to further estimate the spatial extent of ice sheet in Greenland.

As shown in Figure 3, the coverage of ice cap climate over Greenland for the reference period of 1970–2000 simulated by PRECIS is first compared to the results from two observational data sets (i.e., CRU and WC). It should be noted that there are some apparent discrepancies in terms of ice cap climate coverage generated by these two observational data sets. Particularly, the CRU data set shows that 69.7% of Greenland is covered by ice cap climate, while the coverage estimated by WC is 66.9%. In comparison, the simulated coverage of ice cap climate from PRECIS is 65.5%, suggesting a very slight underestimation of WC data set (by 1.4%). There are some notable spatial mismatches between PRECIS and two observational data sets in terms of the location of ice cap climate, especially in the far north. However, considering the lack of long-term observational data in the far north of Greenland and the uncertainties with existing observational data sets, it is reasonable to accept the performance of PRECIS.

Details are in the caption following the image

Comparison of ice cap climate between PRECIS and two observational data sets from Climatic Research Unit (CRU) and WorldClim (WC) for the period of 1970–2000.

Figure 4 shows the projected coverage of ice cap climate over Greenland for three future periods under RCP2.6, RCP4.5, and RCP8.5. It can be seen that the high-emission scenario would result in significant shrinkage in ice cap climate coverage throughout this century. In particular, the coverage of ice cap climate under RCP8.5 would be shrinking continuously from 65.5% in the reference period to 40.5% by the end of this century, indicating a significant loss of ice sheet coverage by 25%. Similarly, the coverage of ice cap climate under RCP4.5 would also be shrinking continuously and the end-of-century coverage is like to be 50.3%, suggesting a 15.2% loss in ice sheet coverage. By contrast, the ice cap climate coverage under the low-emission scenario would be stabilized to around 56% after 2050s, although a continuous decrease is still projected from now to 2050s. That means the RCP2.6 scenario has the potential of limiting the loss of ice sheet coverage in Greenland below 10% before the middle of this century and no further loss would be expected afterward. This further highlights the importance of pursuing the low-emission scenario to minimizing the sea level rise caused by melting ice sheet.

Details are in the caption following the image

Projected ice cap climate coverage over Greenland. Note that the maps here show the steady-state coverage of ice cap climate by assuming that the climate will hold constant for a long period. In other words, the regions beyond the ice cap climate coverage (filled with white color) are not necessarily ice free at the end of a certain time period (i.e., 2020s, 2050s, or 2080s). The ice sheet in these regions might just start to melt but will eventually disappear as the air temperature no longer meets the conditions for ice cap climate.

4 Conclusions and Discussions

In this study, we have investigated the possibility of stabilizing the Greenland ice sheet in the context of global warming with consideration of three RCP emission scenarios. The PRECIS regional climate model was used to generate high-resolution climate projections over Greenland driven by the boundary conditions from the HadGEM2-ES global climate model. Due to its poor performance in precipitation, we only used the temperature projections from PRECIS to evaluate the changes in the ice cap climate coverage over Greenland. This allows us to estimate the changes in the spatial extent of the Greenland ice sheet under different emission scenarios in order to assess its possibility of stabilization. It should be noted that many previous studies have evaluated the evolution of the Greenland ice sheet over time by modeling its surface mass balance with various regional climate models (Alexander et al., 2015; Aschwanden et al., 2019; Fettweis et al., 2013; Lenaerts et al., 2019; Mouginot et al., 2019; Noël et al., 2018; Rae et al., 2012; Shepherd et al., 2019). However, these studies suggest that there are still a considerable number of uncertainties toward the modeling and estimation of the Greenland ice sheet surface mass balance, making the results less credible than the estimation from our approach based on temperature projections.

The results show that the ice cap climate coverage of Greenland would be shrinking continuously throughout this century under both RCP8.5 and RCP4.5, suggesting that the spatial extent of ice sheet would decrease by 15% (RCP4.5) and 25% (RCP8.5) by the end of this century. In comparison, the low-emission scenario (RCP2.6) does have the potential of limiting the loss of ice sheet coverage in Greenland below 10% before the middle of this century and no further loss would be expected afterward. The results from this research are extremely important to understand the effects of different carbon emission scenarios on stabilizing or minimizing the warming in Greenland and thus the loss in its ice sheet coverage, which is further linked to the rising sea level. In particular, our results suggest that both the high- and medium-emission scenarios would result in continuous warming up in Greenland and thus significant loss to its ice sheet. However, the low-emission scenario does show a great potential in minimizing the warming in the local climate and the loss of ice sheet before 2050s. Most importantly, no significant changes would be expected after 2050s if the low-emission scenario could be met. Therefore, it is extremely important to reduce carbon emissions by referring to the low-emission scenario defined by RCP2.6 in order to minimize the sea level rise caused by melting ice sheet in Greenland. It also should be noted that the effective window for this chance is very short. In particular, as shown in Figure S12, the CO2 equivalent concentrations of RCP2.6 and RCP4.5 will start to diverge in 2020. Our results for RCP4.5 and RCP8.5 suggest that the local climate of Greenland is likely to warm up continuously throughout this century. That means, if no effective carbon reduction policies are being taken now in accordance with the low-emission scenario of RCP2.6, we are very likely to enter a continuous warming pathway and lose the chance of stabilizing the Greenland ice sheet.

It should be noted that there are a few caveats about the results from this study. First, we know that the mass of the Greenland ice sheet can vary year by year. It may gain mass this year because of the increase in snowfall and its total mass might drop significantly next year due to the considerable melting of ice and snow. In this sense, the Greenland ice sheet will probably never be able to stabilize and it might be meaningless to discuss the possibility of stabilization. However, our definition about stabilization here is established from a long-term climate change perspective. Specifically, we consider the ice cap climate coverage for three different 30-year periods to assess the stabilization possibility of the Greenland ice sheet. The ice cap climate of each period is used to represent its steady-state ice sheet coverage, which means that you will probably never see it happening at a specific year during the 30-year period because it simply represents the expected outcome for this period. Once the steady-state ice sheet coverage for each 30-year period is determined, we can compare the results between two periods. If no significant changes are found from one period to another, the ice sheet is deemed as stabilized for these two periods.

Second, we only consider one global climate model (i.e., HadGEM2-ES) which has been proven to be one of the most reliable models for the Greenland ice sheet modeling among all the CMIP5 models (Barthel et al., 2020). This is because our intention here is to test the possibility of stabilizing the Greenland ice sheet rather than quantifying the related uncertainties. Since the outputs from the most reliable model are usually in good agreement with the ensemble mean of CMIP5, it is reasonable to believe that our results from HadGEM2-ES can represent the majority of CMIP5 models. We understand that using a single model can hardly capture the uncertainty range generated by multiple models, but our goal here is not to quantify the multi-model uncertainties through an ensemble modeling approach. In fact, since a possibility means that there is still a chance, there is no need to test all models if the focus is on the stabilization possibility. In other words, as long as one model shows a chance of stabilization, it can be concluded that there is a possibility although the probability of the chance is depending on the likelihood of the model itself. By using one of the most reliable models for Greenland ice sheet modeling, here we aim to maximize the likelihood of the selected model and ensure the credibility of our results.

Third, we only use downscaled temperature projections from the PRECIS regional climate model to calculate the spatial coverage of ice cap climate, which is further used to estimate the extent of ice sheet. Although PRECIS does provide direct diagnostics for surface water balance to help estimate the mass loss of ice sheet, we do not use them here for the possibility testing. The reason is because the well-recognized poor performance of both global and regional climate models (including the models used in this study) in simulating precipitation, which is a critical variable for estimating the surface mass balance of ice sheet. Since the models selected in this study demonstrate a reasonable performance for simulating temperature over Greenland, our approach based on ice cap climate is expected to have a higher credibility.

Last but not least, our research is not intended for providing accurate estimation of the mass loss of the Greenland ice sheet over time. This is because the ice cap climate coverage is incapable of estimating the depth change of melting ice sheet, especially for those regions where air temperature continues to warm up and no longer meets the condition for ice cap climate. In other words, it is reasonable to assume that the ice sheet covered by ice cap climate will remain there permanently, but it is hard to tell the exact timeline about when the ice sheet beyond the ice cap climate coverage will start to melt and eventually disappear.

Acknowledgments

This research was supported by the National Key Research and Development Plan (2016YFA0601502), Natural Science and Engineering Research Council of Canada, Atlantic Canada Opportunities Agency, and Atlantic Computational Excellence Network (ACENET).

    Data Availability Statement

    The gridded CRU data set used in this study is available at: http://dx.doi.org/10.5285/10d3e3640f004c578403419aac167d82. The gridded climate data from the WorldClim version 2.0 is available at http://worldclim.org/version2.