Economic impacts of melting of the Antarctic Ice Sheet

The Antarctic Ice Sheet (AIS) is losing mass under global warming and holds enough freshwater to raise sea level by approximately 58 meters, making it a potentially dominant source of future sea level rise (SLR). In economic modeling, AIS processes are often treated as a missing risk or stylized tipping element, leaving a gap between physical ice-sheet dynamics and economic impact assessments that commonly rely on abstract temperature-damage functions. Prior studies have explored the West Antarctic Ice Sheet (WAIS) using simplified tipping frameworks or stylized SLR scenarios, while broader SLR impact studies implicitly capture some AIS contribution without isolating it. This paper addresses this gap by building a reduced-form, process-informed AIS melting emulator and coupling it to a spatially resolved coastal impact and adaptation model to quantify country-level costs under different adaptation assumptions. It further embeds this coupled system within a modular Integrated Assessment Model (META) to evaluate the incremental contribution of AIS melting to the social cost of carbon. The study’s purpose is to integrate advances in glaciology and coastal impact modeling to generate realistic SLR projections from AIS mass loss, estimate heterogeneous global coastal impacts and adaptation responses, and quantify implications for climate policy via the social cost of carbon.

The paper situates its contribution within several strands: (1) AIS as a missing or stylized risk in economic models of climate change and tipping points, where previous economic work often used abstract temperature-damage functions disconnected from ice-sheet processes; (2) two prior AIS-focused economic studies: Diaz and Keller (2016) integrated a stochastic tipping formulation for WAIS into DICE, and Nicholls et al. (2008) used stylized SLR scenarios with the FUND IAM; (3) broader literature on global SLR economic impacts (e.g., Nicholls & Tol; Hinkel et al.; Anthoff et al.; Diaz’s CIAM) that implicitly include AIS contributions and emphasize the role of adaptation; (4) recent advances in glaciology enabling process-based reduced-form emulation of AIS dynamic responses, such as LARMIP-2, with recognition of additional processes like hydrofracturing and Marine Ice Cliff Instability (MICI) that may amplify tail risks; (5) IPCC AR6 synthesis providing benchmark SLR projections and highlighting deep uncertainties, particularly in high-end outcomes linked to AIS dynamics. The study builds on this literature by covering the whole AIS (not just WAIS), using process-based reduced-form dynamics calibrated to multi-model ensembles, explicitly modeling coastal adaptation at high spatial resolution, and integrating the results into an IAM framework to assess the social cost of carbon.

The study develops a coupled physical-economy framework and embeds it within the META IAM. Key components: 1) AIS melting emulator: - Surface Mass Balance (SMB) modeled in reduced form by scaling global mean surface temperature (from RCP scenarios via the FAIR/META climate modules) to Antarctic continental temperature (per Garbe et al.), relating continental warming to accumulation changes (Frieler et al., approx. +5% ±1% accumulation per degree), mapping accumulation to AIS mass balance with lagged interaction with dynamics, and applying an adjustment for SMB turning negative around +6.5 K above preindustrial. - Dynamic contributions modeled using Levermann et al. (LARMIP-2) reduced-form linear response functions emulating basal ice shelf melt-driven SLR contributions from 16 ice-sheet models across five Antarctic basins (East Antarctica, Ross, Amundsen, Weddell, Antarctic Peninsula). Global mean surface temperature is mapped to subsurface ocean temperature (CMIP5) then to basal melt (observations), then to SLR via response functions. Sensitivity analyses include (a) DeConto et al. projections incorporating hydrofracturing and MICI (RCP8.5), and (b) ABUMIP extreme ice shelf melt experiments as a worst-case. 2) Global-to-local SLR downscaling: - Statistical downscaling maps global mean SLR to local SLR for 12,148 DIVA coastal segments using flexible cubic functions fitted to probabilistic local SLR projections (Kopp et al.). Fit quality under RCP4.5: median absolute error 0.003 m over 2010–2100 (mean 0.005 m; 97.5% within 0.018 m). 3) Coastal impact and adaptation modeling (CIAM): - Segment-level planner minimizes discounted sum of adaptation costs (protection via dikes/seawalls; proactive managed retreat/relocation) and residual damages (permanent inundation land loss, wetland ecosystem service losses, storm surge flood damages) over rolling ~40-year horizons. Two bounding scenarios are run: no adaptation (no protection, no proactive retreat; only reactive responses) and least-cost/optimal adaptation. Segment costs are aggregated to national and global totals. 4) Integration with META IAM: - META provides modules for other SLR sources (thermal expansion, glaciers/small ice caps via IPCC AR5 ranges; GIS via Nordhaus replication), probabilistic climate-carbon cycle (FAIR), and socioeconomics (RCP/SSP). AIS module is integrated to assess incremental impacts of AIS melting. - For SCC calculation, CIAM’s segment-level outcomes are simplified to national linear SLR damage functions with slope parameters drawn from symmetric triangular distributions bounded by the no-adaptation and least-cost estimates, capturing uncertainty in adaptation degree. The SCC is computed as the welfare-based present value of damages from a marginal tonne of CO2 emitted in 2020 (2020 USD), under standard META settings (e.g., pure rate of time preference 1%, elasticity of marginal utility 1.5, mixed levels/growth damages). 5) Uncertainty and simulation: - Monte Carlo sampling propagates uncertainties in AIS SMB and dynamics, other SLR sources (when applicable), and META parameters. For SLR and costs, 50,000 draws are used (convergence verified). CIAM is run for the 5th, 50th, 95th percentile SLR paths to compute adaptation choices and costs; META SCC simulations use 10,000 draws. Discounting for NPV cost summaries uses a 4% consumption discount rate (sensitivity to pure time preference shown for SCC). Data and code availability are provided (CIAM, META repositories; Zenodo dataset).

- AIS contribution to global mean SLR by 2100 (median [mean], 90% CI): RCP2.6: 0.15 m [0.18 m], 0.03–0.45 m; RCP4.5: 0.16 m [0.21 m], 0.02–0.55 m; RCP8.5: 0.20 m [0.26 m], 0.01–0.71 m. SMB is a small negative contributor; dynamic processes dominate, with largest regional contributions from Weddell, East Antarctica, and Ross. - Total SLR from all sources by 2100 (median): RCP2.6: 0.38 m; RCP4.5: 0.47 m; RCP8.5: 0.62 m (broadly consistent with IPCC AR6, somewhat lower for non-AIS sources in META). - Global annual costs of incremental SLR from AIS melting (2020 USD): No adaptation: RCP4.5: $180bn/yr (2050) rising to $1.04trn/yr (2100, ~0.1% global GDP); RCP2.6: $167bn/yr (2050), $911bn/yr (2100); RCP8.5: $201bn/yr (2050), $817bn/yr (2100). Cost profiles exhibit non-monotonicity over short horizons due to sequencing of impacts. - Optimal/least-cost adaptation reduces costs by roughly an order of magnitude: RCP4.5: $23bn/yr (2050), $86bn/yr (2100); RCP2.6: $23bn/yr (2050), $66bn/yr (2100); RCP8.5: $24bn/yr (2050), $126bn/yr (2100). Protection costs are generally smaller than retreat/residual damages but reach ~$29bn/yr by 2100 under RCP8.5. - Cost components under no adaptation: permanent inundation, relocation/retreat, and storm surge flooding contribute similar magnitudes overall; storm-related costs grow fastest and dominate toward century end. - Heterogeneity across countries: Mid-century (2040–2060 average) costs as % of GDP on RCP4.5 are highly unequal and concentrated in Small Island Developing States. Examples: No adaptation—Maldives 51.39%, Marshall Islands 42.01%, Micronesia (FSM) 10.11%, Kiribati 5.88%, French Polynesia 3.97%, Tonga 2.85%, Tuvalu 2.83%, Antigua and Barbuda 2.19%, Netherlands 2.04%, Nauru 1.45%. Least-cost adaptation—Maldives 0.63%, Marshall Islands 0.46%, Tonga 0.39%, Bahamas 0.25%, Kiribati 0.24%, Tuvalu 0.17%, Micronesia (FSM) 0.22%, Netherlands Antilles 0.15%, French Polynesia 0.13%. By 2100, SIDS remain prominent; under least-cost adaptation, Australia also faces high costs relative to GDP due to coastal concentration of assets and slower GDP growth. - Tail risk and sensitivity: Using DeConto et al. (RCP8.5 with hydrofracturing and MICI), NPV costs to 2100 at a 4% discount rate are comparable to main estimates. ABUMIP worst-case experiments yield much higher NPV costs; under no adaptation median NPV is $18.8 trillion by 2100 (~22% of 2020 world GDP), with extremely wide uncertainty. - Social cost of carbon (SCC) impacts: AIS melting increases SCC on average by about 7% under lower-to-mid emissions—RCP2.6–SSP1: +7.1% (base SCC ~ $34/tCO2); RCP4.5–SSP2: +7.0% (base ~ $52/tCO2). Under RCP8.5–SSP5, average SCC increases by 53.3% (base ~ $33/tCO2), with a strongly right-skewed distribution and cases of SCC more than doubling. Lower pure time preference (0.1%) magnifies SCC impacts and tail risk; higher (2%) reduces them. - Uncertainty: Cost uncertainty driven by AIS SLR uncertainty (conditional on RCP) is large and grows over time; 90% confidence intervals often exceed median estimates, especially under RCP8.5 by 2100.

The study directly links physically informed AIS melting processes to localized coastal impacts and global welfare, addressing the gap between ice-sheet dynamics and economic assessments. Findings show that while AIS-induced SLR poses a growing global economic challenge, the aggregate burden can be reduced dramatically through economically efficient coastal adaptation strategies that optimize protection and proactive retreat. However, benefits and burdens are distributed very unevenly: Small Island Developing States face disproportionately high costs relative to GDP, raising equity and international support considerations for adaptation and loss and damage. Significant deep uncertainty arises from AIS dynamics themselves, particularly the upper tail of potential outcomes associated with processes like hydrofracturing and MICI, which can drive very large costs and strongly elevate the social cost of carbon on high-emissions pathways. Integrating AIS dynamics within an IAM reveals that AIS melting materially increases the SCC—even under moderate scenarios—and can substantially raise it under high emissions, reinforcing the case for stronger mitigation and robust, proactive adaptation. These findings underscore the importance of targeted adaptation planning, risk management under uncertainty, and the social value of reducing scientific uncertainty about AIS behavior.

The paper contributes a process-informed, reduced-form AIS melting model coupled with high-resolution coastal impact and adaptation modeling (CIAM), embedded within the META IAM to quantify the incremental economic and policy implications of AIS melting. It shows that optimal adaptation can reduce AIS-related coastal costs by roughly an order of magnitude, that costs are highly heterogeneous and concentrated among SIDS and some coastal nations, and that AIS melting adds several dollars to the social cost of each tonne of CO2, with a pronounced tail risk under high-emissions scenarios. Policy implications include the need for well-resourced, proactive coastal adaptation; international support mechanisms for highly exposed, low-capacity countries; and strengthened mitigation consistent with the heightened SCC. Future research should focus on narrowing AIS dynamic uncertainties (including hydrofracturing and MICI), incorporating geophysical interactions with other tipping elements and the GIS, improving local SLR and coastal risk data, capturing interdependencies across coastal segments, relaxing perfect foresight assumptions in adaptation planning, and exploring AIS hysteresis within economic assessments.

- The META implementation does not model geophysical interactions between AIS and other tipping elements (e.g., GIS, albedo feedbacks). - Uncertainty characterization focuses on AIS contributions to SLR (and total global mean SLR within META) and omits important local SLR uncertainties. - CIAM-related uncertainties (engineering costs, exposure data, behavioral responses) are not fully propagated; DIVA flood exposure is sensitive to extreme sea-level datasets. - CIAM omits some damage channels (e.g., saltwater intrusion) and treats coastal segments independently, neglecting interdependencies. - Adaptation planners are assumed to have perfect foresight about SLR percentiles, underestimating costs of decision-making under uncertainty and hedging. - Confidence intervals likely understate true uncertainty ranges; ABUMIP is used only as a worst-case sensitivity. - IAM-level assumptions (discounting, damage persistence) influence SCC results; sensitivity to pure time preference is noted. - AIS hysteresis is not explored in depth due to mostly monotonic warming in RCP scenarios, despite emulator capability.