The global costs of extreme weather that are attributable to climate change

The study investigates the economic costs of extreme weather events that are attributable to anthropogenic climate change by combining extreme event attribution (EEA) results with socio-economic cost data. Extreme weather imposes direct damages to assets and indirect losses in economic activity; given the difficulty of consistently quantifying indirect losses, the paper focuses on direct damages and mortality. The authors compile Fraction of Attributable Risk (FAR) estimates from frequentist EEA studies and match them to disaster impact data (primarily EM-DAT) to quantify the portion of event costs attributable to climate change. Due to data limitations in temporal and spatial coverage, the goal is to demonstrate the utility of an attribution-based aggregation method rather than to provide definitive global totals. The approach provides complementary evidence to Integrated Assessment Models (IAMs) and suggests many IAMs may understate current climate damages, particularly those from extremes.

The paper builds on the development of EEA (e.g., Allen; Stott et al.) and prior applications linking event attribution to economic losses (e.g., Frame et al. on Hurricane Harvey and flood/drought damages). It contrasts its event-aggregation approach with IAMs such as DICE and PAGE, which typically use polynomial damage functions based on deviations in mean temperature and often omit or ad hoc represent extreme events. Using DICE 2016R parameters, IAM-based estimates for 2000–2019 total US$4.04 trillion in climate damages, compared with US$2.86 trillion from the attribution-based aggregation; however, these metrics are not directly comparable (stock vs. flow, and scope of impacts). The paper also compares results with the FUND model, which provides sectoral and regional damage functions for tropical and extra-tropical storms; FUND’s estimate of current tropical cyclone damages (~0.08% of GDP) exceeds the attribution-based storms estimate (~0.06% of GDP), and FUND’s estimated mortality rate also exceeds the attribution-based estimate, highlighting data and methodological differences. The literature also notes selection biases and methodological challenges in EEA (e.g., framing choices, conditioning, and regional coverage), and underscores the need to better incorporate extremes in macroeconomic modeling.

- Core concept: Use Fraction of Attributable Risk (FAR) from EEA to apportion observed event costs to anthropogenic climate change. - FAR definition: FAR = 1 − P0/P1, where P0 is the probability of the event in a counterfactual world without anthropogenic GHG and P1 is the probability under the observed climate with anthropogenic GHG. - Event cost attribution: For each event i with observed cost (damages and/or mortality valuation), climate change-attributed cost CCcost_i = FAR_i × cost_i. - Data sources and assembly: - EEA studies: 118 attribution studies providing 185 matched events (2000–2019) with FARs; preference hierarchy for multiple studies uses journal quality (SJR rank), then closest spatio-temporal match to economic data; WWA rapid studies assigned the dataset-average SJR proxy. - Economic impacts: EM-DAT disaster database for direct economic damages and mortality; events include heatwaves, floods, droughts, wildfires, storms, and cold events. Indirect losses are excluded due to measurement challenges. - Mortality valuation: Uniform Value of Statistical Life (VSL) set at US$7.08 million (average of US DoT 2020 and UK Treasury values), applied globally for equity and comparability with IAMs. - Sample characteristics: - 185 events spanning 52 countries; 154 with increased risk due to climate change, 24 with decreased risk, 7 unchanged. - Event-type distribution: heatwaves (31%), floods (37%), droughts (16%), wildfires (4%), storms (5%), cold events (6%). - Geographic coverage uneven: Africa (10%), Asia (28%), Americas (24%), Europe (20%), Oceania (18%). - Extrapolation to global totals (2000–2019): - Global average method: Compute mean FAR per event type from studies, then multiply by total EM-DAT costs (damages and mortality) for that event type globally over 2000–2019. - Regional average method: Compute mean FAR per event type per continent, multiply by EM-DAT costs for matching region–type cells; where no regional FAR exists, substitute global event-type FAR. - The global average method is preferred for headline results due to fewer sensitivities to sparse regional cells. - Sensitivity: Also computed using median FARs by event type. Uncertainty characterized via one standard deviation around mean FARs by type to give annual ranges. - Key formulas: - FAR = 1 − P0/P1 - CCcost_i = FAR_i × cost_i

- Event-level attribution (185 matched events): - Net mortality attributable to climate change: 60,951 deaths (75,139 deaths in events made more likely vs. 14,187 fewer deaths in events made less likely). - Attributed statistical value of life (VSL) cost: US$431.8 billion. - Attributed direct economic damages: US$260.8 billion (53% of total recorded damages for these events). - Hazard shares of attributed damages: storms >64%, heatwaves 16%, floods 10%, droughts 10%, wildfires 2%, cold events −2% (net decrease). - Global extrapolation (2000–2019, global average FAR method): - Total climate change-attributed costs: US$2.86 trillion over 20 years, averaging US$143 billion per year. - Split: ~US$90 billion per year from statistical loss of life; ~US$53 billion per year from direct economic damages. - Using median FARs by event type yields a higher estimate: US$167 billion per year. - Interannual variability: Lowest year 2001 (~US$23.9 billion); highest year 2008 (~US$620 billion), with peaks driven largely by mortality-heavy events (e.g., 2003 European heatwave; 2008 Cyclone Nargis; 2010 Russia heatwave and Somalia drought). - Uncertainty range (±1 SD on event-type FARs): ~US$58–US$228 billion per year; storms dominate absolute differences; floods show largest standard deviation. - Damage-only peaks: 2005 and 2017 driven by US hurricane seasons (Katrina/Rita/Wilma: US$123 billion attributed; Harvey/Irma/Maria: US$139 billion attributed). - Income group distribution: - High-income countries account for ~47% of climate-attributed economic costs, influenced by high asset exposure (notably US storms) and better data availability. - Relative burden as share of GDP: Low-income countries near ~1% of GDP per annum on average versus ~0.2% for high-income, largely due to higher climate-attributed mortality rates in lower-income settings. - IAM comparisons: - DICE (2016R parameters) estimates US$4.04 trillion climate damages (2000–2019) vs. attribution-based US$2.86 trillion; metrics are not directly comparable (flow vs. stock; scope differences). - FUND suggests tropical cyclone damages ~0.08% of GDP vs. attribution-based storms ~0.06% of GDP; FUND mortality ~0.00015% of population vs. attribution-based ~0.00009% per annum. - Additional metrics: - Average global FARs by type: heatwaves 77%; floods 19% (wide range including decreases in some regions); droughts 44%; wildfires 60%; storms 60%; cold events −79% (decreasing risk). - Estimated climate change-attributed number of people affected (global average extrapolation): ~1.4 billion (not monetized in totals).

The study demonstrates that a substantial portion of the observed human and economic costs from recent extreme weather is attributable to anthropogenic climate change, quantifying an average of US$143 billion per year over 2000–2019. This directly addresses the research aim of providing an event-based, empirical aggregation complementing IAMs. The findings indicate that extremes—particularly storms and heat—are already imposing large climate-attributed burdens, with loss of life dominating total costs and a disproportionately high relative impact on low-income countries. The comparison with IAMs underscores that models centered on mean temperature changes likely under-represent current damages from the tails of the climate distribution (extremes). Policy implications include the importance of mitigation to reduce FARs of damaging extremes and the immediate potential for adaptation (e.g., heatwave action plans, early-warning systems, and protective infrastructure) to reduce attributed costs, as illustrated by the reduced mortality in France following the 2003 heatwave. The methodology can inform loss and damage assessments and provide evidentiary support in climate litigation by bridging causality (FAR) with monetized impacts.

By aggregating event attribution (FAR) with observed disaster impacts, the paper provides a global estimate of climate change-attributed costs of extreme weather: US$2.86 trillion over 2000–2019 (US$143 billion annually), with the majority from mortality. Results suggest that commonly used IAMs likely understate near-term damages from extremes due to their focus on average temperature changes and limited treatment of tail risks. The contribution is methodological and empirical: a scalable framework using EEA to monetize climate-attributed portions of disaster costs and to compare with macro-model outputs. Future research should expand EEA coverage across regions and event types, improve alignment between event definitions and impact boundaries, develop and calibrate intensity-based damage functions, incorporate indirect and cascading losses, better capture small but frequent events, and refine extrapolation with richer regional and hazard-specific FARs and higher-resolution economic loss data.

- Data coverage and bias: EEA studies are unevenly distributed geographically (underrepresentation in Africa, South America, and parts of Oceania) and by event type (overrepresentation of heatwaves; underrepresentation of storms, droughts). Regional extrapolations may rely on very few studies for some region–hazard cells. - Framing and methodological variability in EEA: Different definitions (frequency vs. intensity), conditioning choices (e.g., ENSO, SST), and counterfactual specifications can alter FARs; spatial and temporal boundaries of events may not align precisely with EM-DAT impact footprints. - Extrapolation assumptions: Use of average FARs (global or regional) assumes representativeness of studied events for all events of the same type and region; data gaps necessitate global substitutions, limiting regional specificity. - Economic impact measurement: EM-DAT primarily records direct damages and mortality; indirect losses (productivity, supply-chain disruptions, service outages) and many human impacts (morbidity, mental health, education loss) are excluded, leading to underestimation. EM-DAT omits many small, frequent events. - Mortality valuation: A uniform VSL (US$7.08 million) is applied globally for equity and comparability, but this ignores country-specific valuations and distributional welfare considerations. - Publication and selection bias: EEA studies tend to focus on major, high-impact events and on events that increased in likelihood; events that became less likely or did not occur are underrepresented. - Comparability with IAMs: Attribution-based estimates are stocks (event costs at points in time) and a subset of climate impacts, while IAM damages are flows relative to GDP and aim to cover a broader set of climate effects; comparisons are informative but not one-to-one. - Insurance and monetary data limitations: Insurance losses are sparse and skewed toward high-income countries; monetary damage data quality varies across regions and events.