pubmed.ncbi.nlm.nih.gov

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

  • ️Fri Jan 01 2021

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

Yufei Zou et al. Nat Commun. 2021.

Abstract

The compound nature of large wildfires in combination with complex physical and biophysical processes affecting variations in hydroclimate and fuel conditions makes it difficult to directly connect wildfire changes over fire-prone regions like the western United States (U.S.) with anthropogenic climate change. Here we show that increasing large wildfires during autumn over the western U.S. are fueled by more fire-favorable weather associated with declines in Arctic sea ice during preceding months on both interannual and interdecadal time scales. Our analysis (based on observations, climate model sensitivity experiments, and a multi-model ensemble of climate simulations) demonstrates and explains the Arctic-driven teleconnection through regional circulation changes with the poleward-shifted polar jet stream and enhanced fire-favorable surface weather conditions. The fire weather changes driven by declining Arctic sea ice during the past four decades are of similar magnitude to other leading modes of climate variability such as the El Niño-Southern Oscillation that also influence fire weather in the western U.S.

© 2021. Battelle Memorial Institute.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1
Fig. 1. Observation- and reanalysis-based teleconnection between Arctic sea ice and regional fire.

a Spatial distributions of the correlation (shading in the Arctic as denoted by the purple-green color bar) between seasonal average Arctic sea-ice concentrations (SIC) in summer and autumn (July to October) and a seasonal and regional average fire weather index (FFWI) over the western U.S. in the following autumn and early winter (September to December), and the difference of seasonal average FFWI (shading in North America as denoted by the blue-red color bar) between the years with minimum (SIC−: red up-pointing triangles in b) and maximum (SIC+: blue down-pointing triangles in b) Arctic SIC. The difference of seasonal (September to December) average geopotential height at 500 hPa between the SIC− and SIC+ years is also shown in a (contours with negative values in dashed lines; unit: m). Stipples in a mark regions that are significantly different from 0 at the 0.05 significance level of two-sided t-tests, and hatching in a denotes statistically significant regions based on the stricter FDR method (see Methods section) with local gridded p-value ≤pFDR*=0.0023 at the threshold of αFDR=0.10. b Time series of seasonal and regional average SIC (seasonal mean from July to October; normalized by its 1981–2010 climatological mean and standard deviation; note its scale on the left Y-axis is inverted to directly compare temporal variations of both time series), FFWI (seasonal mean from September to December), and their correlation. The region definitions for the Pacific sector of the Arctic and the western U.S. are outlined by the red and cyan boxes in a, respectively. The horizontal dashed lines denote the ±1 standard deviations of normalized SIC as thresholds for selecting the SIC± years. c The composite of monthly FFWI (solid lines with dots and error bars) and fractional burned area change of large wildfires (vertical bars) over the western U.S. Error bars in c denote ±1 standard deviations of monthly FFWI in each group. Dot sizes for monthly FFWI in c Denote the 0.05 (large),0.1 (medium), and non-significant (small) significance levels of two-sided t-tests for monthly FFWI differences between the years with minimum (FFWI_SIC−) and maximum (FFWI_SIC+) Arctic SIC, respectively.

Fig. 2
Fig. 2. CESM-RESFire simulated Arctic sea ice and regional fire teleconnection.

a Spatial distributions of the seasonal average (July to October) sea-ice concentration (SIC; unit: 100%) difference (color shading in the Arctic Ocean as denoted by the purple-green color bar) between the SICexp− and SICexp+ experiments, and the seasonal average (September to December) burned area difference (color shading in North America as denoted by the blue-red color bar) in response to the sea-ice perturbation. The difference of geopotential height at 500 hPa (Z500; unit: m) between SICexp− and SICexp+ is also shown (contours with negative values in dashed lines). The region definitions for the Pacific sector of the Arctic and the western U.S. are outlined by the red and cyan boxes, respectively. Stipples in a show regions that are significantly different from 0 at the 0.1 significance level of two-sided t-tests, and hatching in a denotes statistically significant regions based on the stricter FDR method (see Methods section) with local gridded p-value ≤pFDR*=0.017 at the threshold of αFDR=0.20. b Two-dimensional joint distributions of the seasonal mean fire-favorable circulation index (Z500i; standardized by first removing the 40-year mean and then normalizing by the standard deviation of Z500i from the SICexp+ experiment; unitless) and fire weather index (FFWI; also standardized by the 40-year mean and standard deviation of FFWI from the SICexp+ experiment; unitless) based on the kernel density estimation (KDE) for SICexp− (red shading) and SICexp+ (blue contours). The legends for color shading and contours are attached aside, and corresponding 1-d KDE distributions for each index in SICexp− (red) and SICexp+ (blue) are also shown along the x- and y-axis. c As in b, but for the comparison of FFWI and regional total burned area (BA; unitless) that are both standardized. d As in b, but for the comparison of FFWI and regional mean fire occurrence (NFIRE; unitless) that are both standardized. e As in b, but for the comparison of FFWI and regional mean fire size (FIREsize; unitless) that are both standardized. Note that the model-based fire variables are averaged over the coarse model grid cells that describe statistical properties of fire ensembles at each grid cell rather than individual properties of each single fire.

Fig. 3
Fig. 3. Physical processes underlying the Arctic sea ice and regional fire teleconnection.

a Zonally averaged (170 °W to 60 °W; as shown in g) temperature (T; color shading; unit: K) difference in autumn and early winter (September to December) between the years with minimum (SIC−) and maximum sea-ice concentration (SIC+) based on the original ERA5 reanalysis data. The time average of zonally averaged temperature in the SIC+ years is also shown (contours; unit: K). b As in a, but for the temperature difference between the years with minimum (SICnotrd−) and maximum sea-ice concentration (SICnotrd+) based on the detrended ERA5 reanalysis data. c As in a, but for the temperature difference between the experiments with minimum (SICexp−) and maximum sea-ice concentration (SICexp+) based on the CESM-RESFire simulations. d–f As in a–c, but for zonally averaged zonal wind (U; color shading; unit: m s−1) difference based on the original ERA5 reanalysis data, the detrended ERA5 reanalysis data, and the CESM-RESFire simulations, respectively. g–i As in a–c, but for wind circulation at 500 hPa (arrows; unit: m s−1) and total precipitation rate (PREC; color shading; unit: mm d−1) differences based on the original ERA5 reanalysis data, the detrended ERA5 reanalysis data, and the CESM-RESFire simulations, respectively. j–l As in a–c, but for surface relative humidity (cyan contours with negative values in dashed lines; unit: %) and surface air temperature (SAT; color shading; unit: K) differences based on the original ERA5 reanalysis data, the detrended ERA5 reanalysis data, and the CESM-RESFire simulations, respectively. Stipples in a–l show regions that are significantly different from 0 at the 0.1 significance level of a two-sided t-test.

Fig. 4
Fig. 4. Arctic sea ice and regional fire weather teleconnection in the CMIP6 amip experiment.

a Time series of normalized seasonal and regional average sea-ice concentration (SIC; unitless), a fire-favorable circulation index (Z500i; unitless), and a fire weather index (FFWI; unitless) based on the observational/reanalysis data and 15 amip model ensemble. The shading along lines of the time series denotes ±1 standard deviations of the model ensemble, and the vertical bar shading denotes years with minimum (pink; SIC−) and maximum (blue; SIC+) SIC for the composite differences in b and c. Note the SIC scale on the top left Y-axis is inverted for direct comparison with the other two variables. b The total difference in precipitation rates (PREC; color shading; unit: mm d−1) in autumn and early winter (September to December) between the SIC− and SIC+ years based on the 12 amip model ensemble. c As in b, but for the total difference in the fire weather index (FFWI; color shading; unitless). d The Arctic-driven (S/NP3) changes in precipitation rates (PREC; color shading; unit: mm d−1) in autumn and early winter (September to December) between the SIC− and SIC+ years based on the 12 amip model ensemble. e As in d, but for the Arctic-driven (S/NP3) changes in the fire weather index (FFWI; color shading; unitless) based on the 12 amip model ensemble. Stipples in b–e show regions that 2/3 amip models agree on the signs.

Fig. 5
Fig. 5. A schematic diagram for the teleconnection between Arctic sea-ice loss and increasing fire hazards over the western U.S.

The L and H denote the cyclonic low pressure and anticyclonic high pressure circulation anomalies, respectively, induced by preceding Arctic sea-ice loss as suggested by the CESM-RESFire model sensitivity results shown in Fig. 2a. (Background image by NASA/Goddard Space Flight Center Scientific Visualization Studio).

Similar articles

Cited by

References

    1. Radeloff VC, et al. Rapid growth of the US wildland-urban interface raises wildfire risk. Proc. Natl Acad. Sci. USA. 2018;115:3314–3319. doi: 10.1073/pnas.1718850115. - DOI - PMC - PubMed
    1. Dennison PE, Brewer SC, Arnold JD, Moritz MA. Large wildfire trends in the western United States, 1984-2011. Geophys. Res. Lett. 2014;41:2928–2933. doi: 10.1002/2014GL059576. - DOI
    1. Thomas, D., Butry, D., Gilbert, S., Webb, D., & Fung, J. The Costs and Losses of Wildfires: A Literature Survey (National Institute of Standards and Technology (NIST, 2017).
    1. Fann N, et al. The health impacts and economic value of wildland fire episodes in the US: 2008-2012. Sci. Total Environ. 2018;610:802–809. doi: 10.1016/j.scitotenv.2017.08.024. - DOI - PMC - PubMed
    1. Syphard AD, Keeley JE, Pfaff AH, Ferschweiler K. Human presence diminishes the importance of climate in driving fire activity across the United States. Proc. Natl Acad. Sci. USA. 2017;114:13750–13755. doi: 10.1073/pnas.1713885114. - DOI - PMC - PubMed