Increasing large wildfires over the western United States linked to diminishing sea ice in the Arctic

Large wildfires across the western United States have increased in number and total burned area in recent decades, imposing major socioeconomic and health burdens. Disentangling human and climate drivers is challenging because ignition, suppression, land use, and forest management interact with natural variability. Prior work links anthropogenic warming to enhanced fuel aridity and shifting fire regimes, yet detection and attribution remain complicated. This study asks whether declines in Arctic sea ice causally contribute to worsening autumn fire weather and increased large wildfires in the western U.S. and assesses the importance of this linkage relative to other climate modes (e.g., ENSO) and global warming. The purpose is to identify and quantify an Arctic-driven teleconnection mechanism that modulates regional circulation, hydroclimate, and fuel aridity relevant to fire hazards on interannual to interdecadal time scales.

Previous studies show human activity can mask climate signals in regional fire activity through ignition patterns, suppression, and land management, while anthropogenic climate change has increased forest wildfire activity via enhanced fuel aridity and expanded upslope burning. Globally, climate influences fire through lightning in boreal/permafrost regions and by modulating fuel availability or aridity depending on regime. Arctic amplification has coincided with rapid summer–autumn sea-ice loss, and modeling and statistical analyses have suggested links between sea-ice decline and midlatitude temperature/precipitation extremes, including possible impacts on western U.S. fire weather and annual wildfire activity. However, comprehensive, quantitative evaluation of Arctic impacts on regional burning and their role in observed fire weather change has been lacking, and consensus on mechanisms has been limited due to short records and model signal-to-noise issues.

The study combines observation/reanalysis diagnostics, targeted climate model sensitivity experiments, and multi-model ensembles. Key components: (1) Observation/reanalysis analysis: Arctic sea-ice concentration (SIC) from HadISST (1981–2019) and ERA5 meteorology are used to compute a regional Fosberg Fire Weather Index (FFWI) over the western U.S. (124 W–97 W; 32.5 N–48 N). Correlations are computed between July–October SIC (Pacific sector of the Arctic: 120 E–135 W; 70 N–80 N) and September–December FFWI. Composite analyses compare years with minimum (SIC−) and maximum (SIC+) SIC using ±1 standard deviation thresholds. Detrended analyses remove long-term trends to isolate interannual variability. Additional reanalyses and fire weather indices (e.g., FWI from GEFF-ERA5 and GFWED; multiple precipitation datasets) test robustness. (2) CESM-RESFire sensitivity experiments: A 40-year control with climatological SST/SIC provides initial conditions. Two experiments perturb Arctic SIC and co-located SST (July–October) to multi-year means of observed minimum (SICexp−: 2007, 2008, 2012, 2016, 2017, 2019) and maximum (SICexp+: 1983, 1984, 1985, 1986, 1994, 1996) SIC years within the Pacific Arctic sector, keeping other forcings (GHGs, aerosols), extra-polar SSTs, lightning and population (for ignitions) constant. Fire feedbacks (emissions, land cover change) are disabled to isolate forcing response. Outputs include 500 hPa geopotential height (Z500), precipitation, surface radiation, temperature, humidity, and fire variables (burned area, fire occurrence, mean fire size). A fire-favorable circulation index (Z500i) is obtained by projecting Z500 anomalies onto an ERA5-derived pattern linked to high FFWI. Kernel density estimation summarizes joint distributions. Extreme burning years are defined by BA above the 95th percentile of SICexp+; bootstrap resampling (10,000 iterations) estimates changes in probability and intensity. (3) CMIP6 multi-model analysis: Atmospheric-only amip and amip-piForcing (AMIP SSTs/sea ice with preindustrial forcings) ensembles (up to 15 and 4 models, respectively) are analyzed to assess the relative roles of surface boundary conditions vs. atmospheric/land forcings. Differences between SIC− and SIC+ years are evaluated. (4) Signal-to-noise-maximizing pattern (S/NP) filtering is applied to joint fields (surface air temperature, precipitation, FFWI, Z500) across ERA5 and amip/amip-piForcing ensembles to separate forced responses associated with ENSO and Arctic sea-ice variability and compare their contributions. (5) Statistical tests: Two-sided t-tests assess significance; False Discovery Rate controls multiple testing for gridded fields. Dynamic/thermodynamic diagnostics in CESM-RESFire decompose temperature tendencies into dynamic and physical components to attribute circulation and thermal structure changes.

- Observational teleconnection: Strong negative correlation between July–October Arctic SIC and September–December western U.S. FFWI (r = −0.68; p < 0.01). After detrending, correlation remains significant (r = −0.50; p < 0.01), indicating robustness across interannual and interdecadal scales. Composite analyses show enhanced fire-favorable weather and increased fractional burned area during September–December following SIC decline. - CESM-RESFire response: Prescribed summer–autumn Arctic sea-ice loss produces a Z500 dipole with cyclonic anomalies over Alaska and anticyclonic anomalies over the western U.S., mirroring reanalysis. Associated changes include suppressed clouds/precipitation and increased solar radiation, shifting Z500i and FFWI positively (Z500i p = 0.01; FFWI p = 0.06). Regional burned area increases (p = 0.04) due to both higher fire occurrence (p = 0.04) and larger mean fire size (p < 0.01). The largest monthly burned area increase is ~12.5% in November. Probability of extreme burning years (>95th percentile BA of SICexp+) is about fourfold higher under SICexp−, with 14–15% higher intensity. - Dynamics/thermodynamics: Reanalysis and model show enhanced meridional temperature gradient near ~60 N and reduced near ~80 N, implying a poleward shift of the polar jet and storm tracks. Resulting hydroclimate anomalies include drier western/midwestern U.S. and wetter Pacific Northwest/Alaska–Canada, with hotter, drier surface conditions and reduced relative humidity over the western U.S., favoring fuel aridity. - CMIP6 ensembles: amip and amip-piForcing show similar SIC− vs. SIC+ fire weather responses, indicating dominance of SST/sea-ice boundary conditions over direct atmospheric/land forcing by GHGs, aerosols, and land use. Across 15 amip models, all capture correct sign of SIC–Z500i and Z500i–FFWI correlations; 12/15 capture negative SIC–FFWI correlation, 5 significantly so. S/NP filtering identifies ENSO (S/NP1) and Arctic-driven (S/NP3; detrended S/NP8) patterns. Arctic-driven changes in FFWI and precipitation are about half the total but preserve spatial contrasts and are comparable in magnitude to ENSO contributions; the two act constructively in worsening western U.S. fire weather during SIC− years. - Mechanism: Summer–autumn sea-ice loss increases surface solar absorption and turbulent heat fluxes, generating a cyclonic anomaly over the Arctic/Alaska, enhanced warm-air advection downstream, and an anticyclonic anomaly over the western U.S., reinforcing a poleward jet shift, reduced precipitation, higher temperature and vapor pressure deficit, and elevated FFWI that prolong and intensify autumn fire conditions. - Trend implication: The S/NP3 Arctic-driven index shows an increasing trend consistent with declining Arctic sea ice, suggesting an increasingly prominent role in future western U.S. fire weather.

The study provides mechanistic evidence that Arctic summer–autumn sea-ice decline preconditions hotter, drier autumn and early winter fire weather in the western U.S. via a hemispheric teleconnection. The induced circulation dipole and poleward jet shift reduce precipitation and humidity and raise temperatures and vapor pressure deficit over the region, increasing FFWI, fire occurrence, fire size, and total burned area. Model sensitivity experiments reproduce reanalysis patterns, and CMIP6 amip/amip-piForcing ensembles corroborate that oceanic boundary conditions (SST/sea ice) primarily drive the observed fire weather variability, rather than direct atmospheric or land-based anthropogenic forcings alone. S/NP filtering shows Arctic-driven effects are comparable to ENSO in magnitude and combine constructively, helping explain observed interannual to interdecadal variability. These findings directly address the research questions by identifying the physical mechanism and quantifying the Arctic teleconnection’s contribution relative to other modes, thereby refining attribution of recent increases in large western U.S. wildfires, particularly in autumn.

This work demonstrates a robust teleconnection linking diminishing Arctic sea ice in summer–autumn to worsening autumn fire weather and increased large wildfires over the western United States. Using observations, targeted CESM-RESFire sensitivity experiments, and CMIP6 ensembles, the study identifies a consistent circulation response featuring a dipole between Alaska and the western U.S., a poleward-shifted jet, and hotter, drier surface conditions that elevate FFWI and expand burned area. Arctic-driven fire weather changes over recent decades are of similar magnitude to ENSO influences and act constructively with them. Given ongoing Arctic sea-ice decline and indications that the Arctic-driven pattern is strengthening, this teleconnection is likely to increasingly modulate western U.S. fire hazards. Future research should improve representation of Arctic–midlatitude teleconnections in ESMs, incorporate ocean–atmosphere coupling and fire–climate feedbacks, extend observational records, and develop predictive frameworks for compound extremes to inform adaptation strategies in fire-prone regions.

- The CESM-RESFire sensitivity experiments fix extra-polar SSTs and repeat climatological atmospheric forcings (GHGs, aerosols), omitting broader global warming effects outside the perturbed Arctic and limiting full-system responses. - Fire–climate feedbacks (emissions, land cover change) are disabled; prior work suggests these are secondary but may modulate regional outcomes. - Ocean–atmosphere coupling is not included in the sensitivity setup; coupling can amplify or modify responses. - Observational/reanalysis records are relatively short, and internal variability plus model signal-to-noise limitations introduce uncertainty in attribution and pattern ranking. - CMIP6 model spread in Arctic–midlatitude teleconnection responses remains substantial, reflecting challenges in simulating complex dynamics. - Composite year selections differ slightly among datasets/experiments due to data availability windows, which may affect exact magnitudes though not qualitative conclusions.