Extreme Event Attribution
Extreme event attribution is the climate-science subdiscipline that quantifies how human-caused climate change has altered the probability and intensity of specific weather and climate events. It moves climate science from the abstract (“the planet has warmed”) to the specific (“the 2021 Pacific Northwest heat dome that killed 619 people in British Columbia was virtually impossible without anthropogenic warming”). Born in Myles Allen’s 2003 (Nature) commentary on the European heatwave of 2003, formalised by Stott-Stone-Allen 2004 (Nature) calculation that the EU 2003 event was 2× more likely under climate change, the field has matured into a rapid-response operational service with peer-reviewed methodologies, large-ensemble modelling infrastructure, and increasingly direct application in litigation, loss-and-damage finance, and policy. The World Weather Attribution (WWA) collaboration (founded 2014 by Geert Jan van Oldenborgh KNMI + Friederike Otto Oxford/Imperial + Heidi Cullen Climate Central + Maarten van Aalst Red Cross Climate Centre, now led by Otto, van Aalst, Robert Vautard, Sjoukje Philip, Sarah Kew, Mariam Zachariah) publishes attribution analyses of major events within days to weeks. The IPCC AR6 WG1 Chapter 11 (Seneviratne-Zhang-Adnan-Badi-Dereczynski-DiLuca-Ghosh-Iskandar-Kossin-Lewis-Otto-Pinto-Satoh-Vicente-Serrano-Wehner-Zhou 2021) issued the strongest formal attribution statements to date, with very-high-confidence attribution for heat extremes globally. This note compiles methodologies, headline attributed events, the operational rapid-attribution pipeline, the legal applications, and the loss-and-damage interface.
1. Methodology
1.1 Risk-based attribution
Classical framework (Allen 2003; Stott-Stone-Allen 2004; National Academies of Sciences, Engineering, and Medicine 2016 report). Probability ratio:
PR = P1 / P0
where P1 = probability of event ≥ observed magnitude in factual (current-climate) world, P0 = probability in counterfactual (pre-industrial or natural-only) world. Fraction of Attributable Risk:
FAR = 1 − P0 / P1
For the 2003 EU heatwave (mean summer T France +5 °C above 1961–90), Stott 2004 found FAR ≈ 0.75 → 4× more likely under climate change.
1.2 Storyline / conditional attribution
Shepherd 2016 (Curr Opin Env Sustain) + Shepherd-Boyd-Calel-Chapman-Dessai-Dima-West-Fowler-Goodess-Hall-Hallegatte-Hegerl-Knutti-Kretschmer-Lenton-Pinto-Raible-Shaw-Sillmann-Stainforth-Stott-Sutton-Woollings-Zappa 2018 Climatic Change: “Given the observed circulation pattern, how much did climate change modify the magnitude (e.g., heat intensity, precipitation intensity)?” Conditions on dynamic features (jet position, atmospheric blocking, ENSO state). Useful when:
- Dynamic changes uncertain or absent (e.g., circulation patterns of European blocking).
- Insurance / decision-relevant magnitude rather than frequency framing.
Trenberth-Fasullo-Shepherd 2015 (Nature Climate Change) “Attribution of climate extreme events” advocated for storyline + thermodynamic framing.
1.3 Single-model large ensembles
CESM Large Ensemble (CESM-LE, Kay-Deser-Phillips-Mai-Hannay-Strand-Arblaster-Bates-Danabasoglu-Edwards-Holland-Kushner-Lamarque-Lawrence-Lindsay-Middleton-Munoz-Neale-Oleson-Polvani-Vertenstein 2015 BAMS): 40 members of CESM1 driven by RCP 8.5 + identical forcings, differing only in initial-condition perturbation (round-off-level disturbance of atmosphere). Subsequent ensembles: CanESM5 (50 members), MPI-GE (Maher-Milinski-Suarez-Bonneau-Bittner-Brovkin-Jungclaus-Marotzke 2019), UKESM (15), CESM2-LE (Rodgers-Lee-Rosenbloom-Timmermann-Danabasoglu-Deser-Edwards-Kim-Simpson-Stein-Stuecker-Yamaguchi-Bódai-Chung-Huang-Kim-Lamarque-Lombardozzi-Wieder-Yeager 2021 100 members + CMIP6 forcings).
Each ensemble provides “ergodic” sampling of internal variability for a single model state — directly support PR estimation by counting events exceeding threshold across members.
1.4 Multi-model ensembles
CMIP6 DAMIP (Detection and Attribution Model Intercomparison Project, Gillett-Shiogama-Funke-Hegerl-Knutti-Matthes-Santer-Stone-Tebaldi 2016 GMD): coordinated runs of hist (full forcing) + hist-nat (natural only) + hist-GHG (GHG only) + hist-aer (aerosol only) — supports detection of human influence + apportionment by forcing type.
1.5 Fingerprinting
Hegerl-Zwiers 2011 (WIREs Climate Change) optimal detection: project observations onto spatial-temporal “fingerprint” patterns of expected response, weighting by inverse internal variability. Scaling factor β tests for presence + magnitude of forced signal. Used for global-mean and field-pattern detection; less commonly for single events.
1.6 Operational rapid pipeline (WWA methodology)
van Oldenborgh-vanderWiel-Vecchi-Yiou-Sjoukje-Kew-Philip-Vautard 2021 (Climatic Change) “Pathways and pitfalls in extreme event attribution” formalised WWA seven-step protocol:
- Trigger: select event by societal impact + climate metric well-defined.
- Define event class: e.g., 5-day max T over region.
- Observed return time: fit GEV or GP to historical data, get current return period.
- Trend in observations: fit nonstationary model with time covariate.
- Model evaluation: select models that pass climatology + variability checks.
- Model attribution: factual vs counterfactual probability + intensity change.
- Synthesis: combine observational + model estimates.
Time from event to publication: typically 10–40 days. Pre-registered methodology + open data (climexp.knmi.nl). Climate Central + Climameter (Coppola-Faranda, IPSL) provide similar rapid analysis.
2. Headline attributed events — heat
2.1 European 2003 heatwave
July–August 2003 Western Europe summer +2.3 °C above 1961–90 (France +3.1 °C); ~70 000 excess deaths Europe-wide (Robine-LeRoy-Cheung-vanOyen-Griffiths-Michel-Herrmann 2008). Stott-Stone-Allen 2004 first formal attribution: ≥2× more likely with anthropogenic forcing; FAR > 0.5 with 90 % confidence. 2003 events that were 1-in-1000-yr without warming became 1-in-100-yr with.
2.2 Russian 2010 heatwave
July 2010 Moscow +7.6 °C above climatology; ~55 000 excess deaths (Otto-Massey-vanOldenborgh-Jones-Allen 2012 GRL): blocking high “anchored” circulation; thermodynamic component ~80 % attributable to climate change, dynamic component dominant for that magnitude. Rahmstorf-Coumou 2011 PNAS found 80 % probability for new heat records would be human-caused.
2.3 Pacific Northwest “heat dome” June 2021
Lytton British Columbia reached 49.6 °C (121.3 °F) Jun 29 2021 — 4.6 °C above previous Canadian record; town burned down Jun 30. Philip-Kew-vanOldenborgh-Anslow-Seneviratne-Vautard-Coumou-Ebi-Arrighi-Singh-vanAalst-Pereira-Marghidan-Wehner-Yang-Li-Schumacher-Hauser-Bonnet-Luu-Lehner-Gillett-Tradowsky-Vecchi-Rodell-Stull-Howard-Otto 2021 (WWA + Earth Syst Dyn 2022): event return period (current climate) ~1-in-1000-yr; under counterfactual (no climate change) “virtually impossible” — 150× rarer (very wide uncertainty 5×–10 000×). 619 heat-related deaths BC alone; ~1 000 Pacific NW.
2.4 Europe 2022 summer
Hottest European summer on record at time, +1.7 °C above 1991–2020 mean. WWA (Vautard-vanOldenborgh-Bonnet-Li-Robin-Kew-Philip-Soubeyroux-Dubuisson-Viovy-Reichstein-Otto-Garcia-Herrera 2023): 5–6 °C added by climate change in heatwave intensity; ~61 000 excess deaths (Ballester-Quijal-Zsofie-Schumacher-Toschen-Mahmoud-Lavigne-Mahecha-Robinson-Achebak-Buenestado-Galmes-Pannet-Boulart-Achebak-Tamerius-Honaa-Cheriff-Ballester-Antonescu-Capraru-Antonescu-Ballester 2023 Nature Medicine).
2.5 India + Pakistan March-April 2022
Pre-monsoon temperatures +5 °C above climatology over NW India + Pakistan. Zachariah-Vautard-Schumacher-Vahlberg-Heinrich-Marghidan-Singh-Stephens-Arrighi-vanAalst-Otto 2022 (WWA): 30× more likely with climate change; intensified by 1 °C.
2.6 Sahel July 2023
Sahel temperatures 45 °C+; WWA: 1.5–3 °C added; ~10× more likely.
2.7 Mediterranean July 2023 “Cerberus” + “Charon”
Italy + Greece + Spain. WWA: 50× more likely.
3. Headline attributed events — extreme precipitation + storms
3.1 Hurricane Harvey Aug 2017
Houston metro: 1 500 mm rainfall in 4 days, ~$125 B damages (NOAA NCEI, second-costliest US natural disaster after Katrina). 89 deaths. Risser-Wehner 2017 (GRL) found return period for 7-day precipitation ~9 000-yr in current climate; ~3× more likely under climate change with intensity +15 %. vanOldenborgh-vanderWiel-Sebastian-Singh-Arrighi-Otto-Haustein-Li-Vecchi-Cullen 2017 (Env Res Lett): observations 15 % more intense + 3× more probable than 1900 climate. Wang-Wang-Wang-Vecchi-Reed-Wehner 2018 (J Climate) similar.
3.2 Pakistan August 2022 floods
500–900 mm rainfall over Sindh + Balochistan in monsoon season; 33 M people displaced, ~$30 B damages, ~1 700 deaths. Otto-Zachariah-Saeed-Siddiqi-Shahzad-Fahad-Hassan-Abrar-Sieck-Beermann-Heinrich-Aalst-Vautard-Vecchi-Vahlberg-Singh-Arrighi-Marghidan 2022 (WWA): heavy precipitation amplified by climate change with intensities ~75 % more likely (medium confidence due to large internal variability + circulation uncertainty), though 5-day rainfall over Sindh ~50 % more intense than 1.2 °C cooler climate.
3.3 Storm Daniel / Libya Derna floods Sep 2023
Cyclone Daniel devastated Libya (10 900+ deaths Derna alone after Wadi Derna dams Mansour + Belad failed Sep 11; Mediterranean storm with 414 mm 24-h rain). Climameter (Coppola-Faranda 2023): up to 50× more intense than analogue events. WWA (Zachariah-Kotroni-Lagouvardos-Vahlberg-Pinto-Otto 2023): 10× more probable + 50 % more intense rainfall over Libya; dams + governance failure compounded.
3.4 Spain November 2024 DANA floods (Valencia)
Storm “DANA Mediterránea” Oct 29-Nov 3 2024 produced 491 mm in 8 hours in Chiva (Valencia province). 224+ confirmed deaths, ~€10 B damages. Climameter analysis (Faranda 2024): cyclonic precipitation 12 % more intense than analogues. Climate-change role: warming Mediterranean Sea drove higher atmospheric moisture content.
3.5 Hurricane Helene + Milton 2024
Helene (Sep 24-29 2024) made landfall NW Florida Cat 4 then catastrophic inland flooding in Appalachians (Asheville NC 460 mm; 230+ deaths). Milton (Oct 9-10 2024) Cat 5 over Gulf with rapid intensification (Hurricane Hunter measured 935 hPa). WWA rapid attribution: rainfall amplified 10 % + intensity 13 % for Helene + Milton. Reed-Wehner-Wang-Tezarsky-Stansfield-Bacmeister-Stone 2025 attributing Helene rainfall.
3.6 Tropical cyclone literature
Kossin-Knapp-Olander-Velden 2020 PNAS: TC intensity distribution has shifted (more major TCs in same number of total TCs). Bhatia-Vecchi-Knutson-Murakami-Kossin-Dixon-Whitlock 2019 (Nature Comms): rapid intensification ~20 % more frequent globally since 1982. Knutson-Camargo-Chan-Emanuel-Ho-Kossin-Mohapatra-Satoh-Sugi-Walsh-Wu 2020 (BAMS) IPCC-consensus statement: future TC intensities ~5 % higher, rainfall 10–15 % higher, sea-level rise multiplies surge.
Reed-Wehner-Stansfield 2022 (Geophys Res Lett) “Forecasted attribution of the human influence on Hurricane Florence” — pre-event attribution of expected Hurricane Florence (2018) rainfall, validated post-event; new “pre-attribution” mode.
4. Headline attributed events — drought
4.1 Western US megadrought 2000–2024
Williams-Cook-Smerdon-Cook-Anchukaitis-Bolles-Cook-Coats-Marvel-Smith 2022 (Nature Climate Change) “Rapid intensification of the emerging southwestern North American megadrought in 2020–2021”: 2000–2021 driest 22-yr period in 1 200-yr tree-ring reconstruction; ~42 % of soil-moisture deficit attributable to anthropogenic warming via increased evaporative demand. Cook-Mankin-Marvel-Williams-Smerdon-Anchukaitis 2020 (Earth’s Future).
4.2 Cape Town Day Zero 2018
Otto-Wolski-Lehner-Tebaldi-vanOldenborgh-Hogesteeger-Singh-Holden-Fucik-vanGarderen-NewmanThakur 2018 (Env Res Lett): 3-yr drought 5–7× more likely with anthropogenic forcing.
4.3 Western European 2022 drought
Vicente-Serrano-Domínguez-Lorente-Reig-Beguería-Sanz-Lasanta-Schrier 2022 + WWA Schumacher-Hauser-vanOldenborgh-Sippel-Knutti-Seneviratne 2022: 5–20× more likely under climate change; +1 °C drove +6 % evaporative demand.
4.4 Horn of Africa drought 2020–23
Five consecutive failed rainy seasons (Oct-Dec 2020, Mar-May 2021, Oct-Dec 2021, Mar-May 2022, Oct-Dec 2022 — “Deyr”/“Gu”); 23 M people food-insecure. Kimutai-Pinto-Funk-Otto-Vahlberg-Marghidan-Walton-Stephens 2023 (WWA): climate-change-induced drought increased by 100× under counterfactual modelling, but attribution complicated by La Niña.
4.5 Amazon 2023 drought
Worst Amazon drought on record (Rio Negro Manaus reached 12.7 m, lowest since 1902). Clarke-Otto-Stuart-Smith-Harrington 2023 (WWA): climate change 30× more likely; El Niño role separated.
4.6 Argentina 2022–23 drought
Soybean crop devastated (~$20 B loss). WWA: 60× more likely.
5. Headline attributed events — wildfires
5.1 California fires 2020
August Complex (1 m acres burned, largest in California history), Creek Fire, North Complex. Williams-Abatzoglou-Gershunov-Guzman-Morales-Bishop-Balch-Lettenmaier 2019 (Earth’s Future): vapour-pressure-deficit driven; climate change has doubled cumulative forest area burned in W US 1984–2015. Goss-Swain-Abatzoglou-Sarhadi-Kolden-Williams-Diffenbaugh 2020 California 2020 attribution.
5.2 Australia Black Summer 2019–20
24 M ha burned; 34 lives + 3 billion animals killed; $130 B damages. WWA + van Oldenborgh-Krikken-Lewis-Leach-Lehner-Saunders-vanWeele-Haustein-Li-Wallom-Sparrow-Arrigo-Singh-Hauser-Allen-Otto 2021 (Nat Hazards Earth Syst Sci): “fire weather” risk 30 % higher than 1 °C cooler climate (lower bound).
5.3 Canada 2023 wildfires
Record 18.5 M ha burned (>2× previous record), $9 B+ damages, evacuated 232 000 people, smoke covered E US + Europe. Jain-Castellanos-Lonborg-Castellanos-Flannigan 2024 (Comm Earth Env): fire weather 50 % more likely + intense due to climate change. Boer-Resco-Bowman 2024 attribution.
5.4 Maui Lahaina fire Aug 2023
Aug 8 2023 dry hurricane Dora trade winds + Lahaina built environment + invasive grasses → 102 deaths, $5.5 B damages. Drought + fire-weather component attributed (multi-factor, Climate Central + Hawai’i State analyses).
5.5 Greece fires 2023
Rhodes + Evros + Parnitha; >190 000 ha burned; >20 deaths.
5.6 Chile 2024 Valparaíso fire
Feb 2024 - 130+ deaths.
6. Marine heatwaves
Hobday-Alexander-Perkins-Smale-Straub-Oliver-Benthuysen-Burrows-Donat-Feng-Holbrook-Moore-Scannell-SenGupta-Wernberg 2016 (Prog Oceanogr) categorisation: MHW = 5+ consecutive days of SST >90th percentile. Categories I (moderate) through IV (extreme) by multiples of climatology threshold.
6.1 Notable MHWs
- The Blob (NE Pacific 2014–16): +2.5 °C SST anomaly persistent over Gulf of Alaska; toxic algae bloom; salmon fishery losses (Bond-Cronin-Freeland-Mantua 2015 GRL).
- Bering Sea 2017–19: catastrophic for snow crab; >10 B individuals lost 2018–22 (Litzow 2024).
- N Atlantic 2023: SST anomalies reached +5σ Jun 2023 (NOAA), 1.3 °C above 1991–2020 mean; persistent through 2024 (Buontempo 2023 commentary); contributed to Atlantic hurricane season intensification.
- N Atlantic + global ocean 2024: warmest ocean temperatures on record continuing.
- Australian Great Barrier Reef MHWs: 1998, 2002, 2016, 2017, 2020, 2022, 2024 (5th mass coral bleaching) (Hughes-Anderson-Connolly-Heron-Kerry-Lough-Baird-Baum-Berumen-Bridge-Claar-Eakin-Gilmour-Graham-Harrison-Hobbs-Hoey-Hoogenboom-Lowe-McCulloch-Pandolfi-Pratchett-Schoepf-Torda-Wilson 2018 Science).
- 2024 fourth global coral bleaching event (NOAA Coral Reef Watch + ICRI declaration Apr 2024) — affecting 73 % of world coral reefs by Oct 2024.
6.2 AMOC slowdown link
Caesar-Rahmstorf-Robinson-Feulner-Saba 2018 (Nature) identified cold blob SE of Greenland as AMOC-slowdown fingerprint; weakening trend +cold N Atlantic + warm Gulf Stream. Compound link to Mediterranean MHWs disputed but plausible.
7. Compound and cascading events
Zscheischler-Westra-vanderHurk-Seneviratne-Ward-Pitman-AghaKouchak-Bresch-Leonard-Wahl-Zhang 2018 (Nature Climate Change) on compound events: combinations of multiple hazards whose interaction generates greater impact than sum. Examples:
- Heat + drought + fire (California 2020, Australia 2019-20).
- Heat + humidity (wet-bulb T) + crop failure (S Asia annual stress; Im-Pal-Eltahir 2017 Sci Adv on lethal wet-bulb thresholds Persian Gulf 2070+).
- Drought + crop failure → food crisis: Syria 2007-10 drought → rural displacement (Kelley-Mohtadi-Cane-Seager-Kushnir 2015 PNAS link to Arab Spring conflict; subsequently disputed by Selby 2017 Polit Geogr; emerging consensus that link existed but was small relative to other political drivers).
- Compound flooding (pluvial + fluvial + surge, see hydrology-and-water-cycle).
8. Methodological frontiers
8.1 ML for attribution
Faranda-Bourdin-Ginesta-Krouma-Mesnard-Yiou-Coppola-Messori 2022 (Weather Clim Dyn) Climameter approach: identify dynamical analogues + compare counterfactual; ML accelerates analogue search. Pasini 2023 review of neural-network attribution.
8.2 Operational rollout
- Climateprediction.net (climateprediction.net, Oxford): public computing distributed ensemble for attribution science; weather@home regional. WeatherRisk service in development.
- ECMWF S2S sub-seasonal-to-seasonal predictions provide operational ensemble for rapid event attribution within hours.
- UK Met Office ARCS (Attribution of Climate Change Service).
- Climate Central WeatherPower + Climate Shift Index (Jan 2022+) provides daily probability ratio of T anomalies attributable to climate change for thousands of cities.
8.3 Pre-attribution / forecast attribution
Reed-Wehner-Stansfield 2022 attributed forecast Florence rainfall before landfall; emerging “pre-attribution” paradigm for managing storm response.
9. Legal and policy use
9.1 Litigation citing attribution science
- Urgenda Foundation v Netherlands (2015, 2018, 2019 Hoge Raad): Dutch Supreme Court ordered state to reduce GHG emissions 25 % by 2020 vs 1990; cited human-rights duty informed by climate-science consensus.
- Friends of the Irish Environment v Ireland (Supreme Court 2020 IESC 49): struck down Ireland’s National Mitigation Plan as inadequate.
- Milieudefensie et al. v Royal Dutch Shell (Hague District Court May 26 2021): ordered Shell to reduce CO2 45 % by 2030; cited IPCC + EPA 2018 + scientific consensus on attributable harm. On appeal Nov 2024.
- Held v Montana (Held 2023, decided Aug 14 2023 Mont 1st Jud Dist; affirmed Mont Sup Ct Dec 2024): 16 youth plaintiffs won that Montana’s MEPA fossil-fuel-friendly limitation violated state constitutional environmental-rights clause; attribution science central to standing + causation findings.
- Juliana v United States (filed 2015): youth climate suit, dismissed at 9th Cir 2020 + 2024 voluntary dismissal under settlement.
- KlimaSeniorinnen v Switzerland (ECHR Grand Chamber, Apr 9 2024): European Court of Human Rights ruled Switzerland violated Articles 8 + 6 of the European Convention by inadequate emissions targets exposing elderly women to attributable heatwave mortality risk; first ECHR climate case decided on the merits. Cited attribution science explicitly.
- Duarte Agostinho v Portugal + 32 other states (ECHR 2024): dismissed for non-exhaustion + lack of jurisdiction but advanced framing.
- Carême v France (ECHR 2024): dismissed for victim-status reasons.
- Conservation Law Foundation v ExxonMobil (D Mass 2016–): “everett terminal” CWA action citing climate adaptation duty.
- US state AGs (NY People v ExxonMobil — NY Supreme Court ruling 2019 dismissed; Massachusetts AG v Exxon — Mass Sup Jud Ct 2022 allowed claims to proceed); Honolulu, San Francisco, NYC suits proceeding 2024.
- Smith v Fonterra (NZ 2024): NZ Supreme Court allowed climate-tort claim against largest dairy company to proceed Feb 2024.
9.2 Inter-American Court of Human Rights advisory
Chile + Colombia requested 2023; hearings Apr-May 2024; opinion pending 2025.
9.3 ICJ advisory opinion on climate change
UNGA Resolution 77/276 (Mar 29 2023, Vanuatu-led) requested ICJ advisory opinion on states’ obligations re GHG emissions. Hearings Dec 2024 Hague (96 states + 11 intl orgs argued). Opinion expected 2025.
10. Loss and Damage Fund
Established at COP27 Sharm el-Sheikh Nov 2022 after 30+ years AOSIS + LDC advocacy. Operationalised COP28 Dubai Dec 2023:
- World Bank as interim 4-yr host (controversial; some pushed for stand-alone treaty).
- Board (Co-Chairs Richard Sherman SA + Jean-Christophe Donnellier France 2024).
- Initial pledges
$700 M COP28: UAE $100 M, Germany $100 M, UK £40 M ($50 M), Italy €100 M, France €100 M, EU €25 M, US $17.5 M, Japan $10 M, Norway/Denmark/Ireland smaller. - First disbursements anticipated 2025; Implementation Mechanism Implementation Initial Disbursement Arrangement (IIDA) designed 2024.
- Funding needs estimated ~$400 B yr-1 by 2030 (Markandya-Cawood-Eyre-Kourdoulou-Boyd 2022).
Attribution science directly feeds L&D mechanism (NoT v Lp); justice arguments rely on quantified harm attribution to high-emitter states. Burke 2023 “How attribution science changes the conversation” Nature Climate Change.
11. IPCC AR6 attribution statements (Ch 11)
Seneviratne-Zhang-Adnan-Badi-Dereczynski-DiLuca-Ghosh-Iskandar-Kossin-Lewis-Otto-Pinto-Satoh-Vicente-Serrano-Wehner-Zhou 2021 Chapter 11 selected:
- “It is virtually certain that hot extremes (including heatwaves) have become more frequent and intense across most land regions since the 1950s” (>99 %).
- “Human-induced climate change is the main driver of these changes” (high confidence).
- “The frequency and intensity of heavy precipitation events have increased since the 1950s over most land area for which observational data are sufficient” (high confidence).
- “Agricultural and ecological droughts in some regions due to increased evapotranspiration” (medium-to-high confidence).
- “Proportion of tropical cyclones that are major (cat 3-5) has likely increased over the past four decades” (medium confidence at the global scale).
- “Compound events” framework adopted Chapter-wide.
12. Tipping element attribution
While individual events are attributable, tipping elements (Greenland ice sheet, AMOC, Amazon dieback, permafrost) need longer time + spatial framing:
- Greenland mass loss attribution: Bevis 2019 PNAS, Briner 2020 Nature; current rates unprecedented over millennia.
- AMOC: Caesar 2018, Boers 2021, Ditlevsen-Ditlevsen 2023.
- See glaciology-and-cryosphere for ice-sheet attribution details.
13. Limitations and uncertainties
13.1 Model dependence
Single-model results often don’t generalise. Hauser-Engelbrecht-Fischer 2022 (Comms Earth Env) found CMIP6 models with high ECS over-attribute extremes; multi-model frameworks essential.
13.2 Internal variability
For high-amplitude rare events, internal variability dominates; large ensembles + storyline approach essential.
13.3 Compound + dynamic vs thermodynamic
Risk-based attribution under-represents dynamical extremes (atmospheric blocking, jet meanders) whose change under warming is contested.
13.4 Definitional sensitivity
Event-class definition (spatial scale, season, duration) materially affects PR estimates. Cattiaux-Ribes 2018 (Bull Am Meteorol Soc) sensitivity analysis.
13.5 Observations
Reanalysis biases (ERA5, MERRA-2, JRA-55); station coverage gaps in Africa + S America + Pacific.
13.6 The “low-likelihood high-impact” framing
AR6 Ch 1 introduced LLHI to flag tipping risks even at low probability — relevant to attribution of unprecedented events.
14. Major attribution research groups
- World Weather Attribution (WWA): Imperial College London + Oxford + KNMI + IPSL + Princeton + Red Cross Climate Centre + ICCCAD Bangladesh + others. Rapid analyses + capacity building.
- Climate Central (Princeton NJ): Climate Shift Index for daily temperature; rapid analyses with WWA.
- Climameter (Faranda et al., CNRS-IPSL Paris): Analogue-based attribution using ECMWF reanalysis; rapid (24-h after event).
- NOAA (BAMS “Explaining Extremes” annual): peer-reviewed compendium of extreme-event attribution studies since 2012 (Peterson-Stott-Herring 2012 first ed).
- UK Met Office Hadley Centre: ARCS service in development; led by Peter Stott + Nikos Christidis.
- Lawrence Berkeley National Laboratory: Mike Wehner extreme precipitation + TC attribution.
- NCAR: Boulder, large-ensemble work; CESM-LE.
- Universities: Oxford, Imperial, Princeton, ETH Zürich, Reading, Wageningen, UVA.
15. BAMS “Explaining Extreme Events” series
Annual special issue since 2012 (Peterson-Stott-Herring 2012 first); ~30 events analysed per year. Recent editions:
- 2022 events (Aug 2023 release).
- 2023 events (Sep 2024 release, ed Jacob-Stott-Wehner-Hauser-Cattiaux).
Highlights human influence detected for majority of high-temperature events (>90 %); mixed for precipitation + drought; rare for cold extremes (generally show reduced probability under warming, consistent with thermodynamics).
16. Climate Shift Index and operational dailies
Climate Central Climate Shift Index (CSI, Jan 2022+):
- Daily probability ratio of T anomalies vs counterfactual climate at each location.
- CSI = 0 (no climate-change signal) to CSI = 5 (extraordinary attribution).
- Coverage: ~3 200 US cities + ~1 000 global cities.
- Web + API + Twitter feed.
Pacific Northwest June 2021: peak CSI = 7 (off-scale) at Lytton BC.
Sea-surface CSI added 2023 — daily ocean attribution.
17. Loss and damage attribution science
17.1 Carbon Majors
Heede 2014 Climatic Change “Tracing anthropogenic carbon dioxide and methane emissions to fossil fuel and cement producers, 1854–2010” — identified 90 producers responsible for ~63 % of cumulative emissions. Ekwurzel 2017 Climatic Change scaled to attribute climate consequences proportionally — temperature ~50 mK + sea level ~30 mm from major producers’ emissions. Critical for litigation + reparations.
17.2 Source attribution
Otto 2024 Annual Review attribution review; Burger 2023 mortality cost of carbon by emitter source.
17.3 Adaptation finance proxies
Attributable damages drive loss-and-damage allocation discussions. Mark Carney 2019 + GFANZ later focus on physical-risk disclosure (TCFD/ISSB).
18. Public communication
18.1 Climate Outreach (Oxford)
Communication training including for attribution science.
18.2 Media coverage
Increased coverage of attribution science; SciLine + Climate Matters (CCNet). Risks: oversimplification (“caused by climate change” rather than probabilistic framing).
18.3 Climate fingerprint visualisation
Realistic + counterfactual ensemble visualisations to communicate FAR/PR.
19. Future directions
19.1 Real-time attribution
Forecast-attribution: pre-event probabilistic statements when event imminent. Reed 2022 hurricane example.
19.2 Sub-seasonal-to-seasonal
S2S Project (WMO-WCRP) provides ensemble forecasts at 2–6 week lead; supports attribution + early-warning.
19.3 ML acceleration
Faranda 2022 Climameter; emerging neural emulators of climate-extreme statistics. Watson 2022 ML for stochastic weather generators.
19.4 Compound + cascading
Frontier in attribution: multi-hazard events (heat + drought + fire), and trans-system cascades (drought → harvest failure → price shock → conflict).
19.5 Tipping points
Detection + attribution of approach to tipping (AMOC weakening, Arctic sea-ice decline) — Boers 2021 statistical early-warning approaches.
19.6 Africa + Global South
Funded WWA capacity building (FCDO + Wellcome); local researchers in Pakistan, Bangladesh, Nigeria, Kenya, Senegal increasingly leading their own attribution.
19.7 Pre-bunking
Pre-event communication framing to counter mis/disinformation; Otto 2023 attribution pre-bunking commentary.
20. Cumulative attribution corpus
By end-2024:
- 600+ peer-reviewed attribution studies (Carbon Brief tracker).
- WWA has produced 70+ rapid attributions since 2015.
- BAMS annual compendium ~30 events/yr × 12 yr = 360+ events.
- Studies cover: heatwaves (dominant), extreme precipitation, droughts, wildfires, marine heatwaves, tropical cyclones (rainfall + intensification), cold extremes, snow drought, sea-level extremes.