• Abatzoglou, J. T., and C. A. Kolden, 2013: Relationships between climate and macroscale area burned in the western United States. Int. J. Wildland Fire, 22, 10031020, https://doi.org/10.1071/WF13019.

    • Search Google Scholar
    • Export Citation
  • Colle, B. A., and C. F. Mass, 1998a: Windstorms along the western side of the Washington Cascade Mountains. Part I: A high-resolution observational and modeling study of the 12 February 1995 event. Mon. Wea. Rev., 126, 2852, https://doi.org/10.1175/1520-0493(1998)126<0028:WATWSO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Colle, B. A., and C. F. Mass, 1998b: Windstorms along the western side of the Washington Cascade Mountains. Part II: Characteristics of past events and three-dimensional idealized simulations. Mon. Wea. Rev., 126, 5371, https://doi.org/10.1175/1520-0493(1998)126<0053:WATWSO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Dennison, P. E., S. C. Brewer, J. D. Arnold, and M. A. Moritz, 2014: Large wild fire trends in the western United States, 1984–2011. Geophys. Res. Lett., 41, 29282933, https://doi.org/10.1002/2014GL059576.

    • Search Google Scholar
    • Export Citation
  • Deser, C., R. Knutti, S. Solomon, and A. S. Philips, 2012: Communication of the role of natural variability in future North American climate. Nat. Climate Change, 2, 775779, https://doi.org/10.1038/nclimate1562.

    • Search Google Scholar
    • Export Citation
  • Dudney, J., L. M. Hallett, L. Larios, E. C. Farrer, E. N. Spotswood, C. Stein, and K. N. Suding, 2017: Lagging behind: Have we overlooked previous-year rainfall effects in annual grasslands? J. Ecol., 105, 484495, https://doi.org/10.1111/1365-2745.12671.

    • Search Google Scholar
    • Export Citation
  • Durran, D. R., 1986: Another look at downslope windstorms. Part I: On the development of supercritical flow in an infinitely deep, continuously stratified fluid. J. Atmos. Sci., 43, 25272543, https://doi.org/10.1175/1520-0469(1986)043<2527:ALADWP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Durran, D. R., 1990: Mountain waves and downslope winds. Atmospheric Process over Complex Terrain, Meteor. Monogr., No. 45, Amer. Meteor. Soc., 59–81.

  • Durran, D. R., and J. B. Klemp, 1987: Another look at downslope winds. Part II: Nonlinear amplification beneath wave-overturning layers. J. Atmos. Sci., 44, 34023412, https://doi.org/10.1175/1520-0469(1987)044<3402:ALADWP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Griggs, T., K. K. R. Lai, H. Park, J. K. Patel, and J. White, 2017: Minutes to escape: How one California wildfire damaged so much so quickly. New York Times, www.nytimes.com/interactive/2017/10/12/us/california-wildfire-conditions-speed.html.

  • Hughes, M., A. Hall, and J. Kim, 2009: Anthropogenic reductions of Santa Ana winds. California Climate Change Center Rep. CEC-500-2009-015-D, 19 pp., www.energy.ca.gov/2009publications/CEC-500-2009-015/CEC-500-2009-015-D.pdf.

  • Keeley, J. E., 2017: Why were the wine country fires so destructive? The Conversation, U.S. Geological Survey, http://pubs.er.usgs.gov/publication/70192771.

  • Keeley, J. E., and C. Fotheringham, 2003: Impact of past, present, and future fire regimes on North American Mediterranean shrublands. Ecol. Stud., 160, 218262, https://doi.org/10.1007/0-387-21710-X_8.

    • Search Google Scholar
    • Export Citation
  • Keeley, J. E., and A. D. Syphard, 2016: Climate change and future fire regimes: Examples from California. Geosciences, 6, 37, https://doi.org/10.3390/geosciences6030037.

    • Search Google Scholar
    • Export Citation
  • Keeley, J. E., C. J. Fotheringham, and M. A. Moritz, 2004: Lessons from the 2003 wildfires in southern California. J. For., 102, 2631.

    • Search Google Scholar
    • Export Citation
  • Littell, J. S., D. McKenzie, D. L. Peterson, and A. L. Westerling, 2009: Climate and wildfire area burned in western U.S. ecoprovinces, 1916–2003. Ecol. Appl., 19, 10031021, https://doi.org/10.1890/07-1183.1.

    • Search Google Scholar
    • Export Citation
  • Monteverdi, J. P., 1973: The Santa Ana weather type and extreme fire hazard in the Oakland–Berkeley Hills. Weatherwise, 26 (3), 118121, https://doi.org/10.1080/00431672.1973.9931644.

    • Search Google Scholar
    • Export Citation
  • Monteverdi, J. P., and B. L. Wood, 1973: The December 1972 freeze and its effects on the eucalyptus forest of the Oakland-Berkeley Hills. Weatherwise, 26 (4), 160167, https://doi.org/10.1080/00431672.1973.9931653.

    • Search Google Scholar
    • Export Citation
  • Pagni, P., 1993: Causes of the 20 October 1991 Oakland Hills conflagration. Fire Saf. J., 21, 331339, https://doi.org/10.1016/0379-7112(93)90020-Q.

    • Search Google Scholar
    • Export Citation
  • Polade, S. D., A. Gershunov, D. R. Cayan, M. D. Dettinger, and D. W. Pierce, 2017: Precipitation in a warming world: Assessing projected hydro-climate of California and other Mediterranean climate regions. Nature Sci. Rep., 7, 10783, https://doi.org/10.1038/s41598-017-11285-y.

    • Search Google Scholar
    • Export Citation
  • Smith, C., B. Hatchett, and M. Kaplan, 2018: A surface observation based climatology of Diablo-like winds in California’s Wine Country and western Sierra Nevada. Fire, 1, 25, https://doi.org/10.3390/fire1020025.

    • Search Google Scholar
    • Export Citation
  • Sullivan, M., 1997: Firestorm!: The Story of the 1991 East Bay Fire in Berkeley. City of Berkeley, 180 pp.

  • Swain, D. L., B. Langenbrunner, J. D. Neelin, and A. Hall, 2018: Increasing precipitation volatility in twenty-first-century California. Nat. Climate Change, 8, 427433, https://doi.org/10.1038/s41558-018-0140-y.

    • Search Google Scholar
    • Export Citation
  • U.S. Fire Administration, 1991: The East Bay hills fire. USFA Tech. Rep. USFA-TR-060, 120 pp., www.usfa.fema.gov/downloads/pdf/publications/tr-060.pdf.

  • Westerling, A. L., T. J. Brown, A. Gershunov, D. R. Cayan, and M. D. Dettinger, 2003: Climate and wildfire in the western United States. Bull. Amer. Meteor. Soc., 84, 595604, https://doi.org/10.1175/BAMS-84-5-595.

    • Search Google Scholar
    • Export Citation
  • Westerling, A. L., D. R. Cayan, T. J. Brown, B. L. Hall, and L. G. Riddle, 2004: Climate, Santa Ana winds and autumn wildfires in southern California. Eos, Trans. Amer. Geophys. Union, 85, 289296, https://doi.org/10.1029/2004EO310001.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    The remnants of the Fountaingrove neighborhood northeast of Santa Rosa after the Wine Country wildfire. Picture was taken on 14 Oct 2017 and was made available through a Creative Commons license granted by Gary Cedar Photography.

  • View in gallery

    (a) Geographical and terrain features of the wildfire region on 8–9 Oct 2017, with the largest fires shown by red areas. (b) The terrain of a broader region, the location of the vertical cross section and sounding locations described in the model simulation section, and the positions of two surface observation sites discussed in the text.

  • View in gallery

    NASA MODIS visible imagery at approximately 2000 UTC (left) 8 and (right) 9 Oct 2017.

  • View in gallery

    Daily averaged AQI at locations over the North Bay region (northern zone) and for locations around San Francisco (Coast and Central Bay).

  • View in gallery

    Fire extents of wildfires between 1939 and 2016 (color shading), with the perimeters of the 2017 Wine Country fires shown by the red outlines. The border of Sonoma County is shown by a black line. Image provided by the Sonoma County Land Trust (https://sonomalandtrust.org/explore/fire-history.html).

  • View in gallery

    NOAA/NWS Climate Division data for the California North Coast Drainage Division for the (a) May–Sep temperature anomaly (oF) from the 1960–90 normal, (b) May–Sep precipitation (in.), and (d) May–Sep PDSI for 1948–2017. (c) Precipitation for Nov–Apr.

  • View in gallery

    (a) Temperature (oC), (b) relative humidity (%), (c) wind direction (o), and (d) wind gusts (m s–1) at the Santa Ana RAWS site from 15 Jul to 15 Oct 2017.

  • View in gallery

    (top). (a) The 10-h fuel moisture percentages at the RAWS site at Santa Rosa from 13 Apr to 10 Oct 2017. (b) Histogram of the frequency of 10-h fuel moisture for the 1–15 Oct time window for 1991–2017 (local standard time).

  • View in gallery

    (bottom). Maximum wind gusts (kt) during the 8–10 Oct 2017 Wine Country wind event. Winds are from METAR, MADIS, and USDA RAWS sites.

  • View in gallery

    Wind gusts (kt) at the (a) Santa Rosa and (b) Hawkeye RAWS sites north of the Bay Area. Locations noted in Figs. 2 and 16.

  • View in gallery

    Surface and 500-hPa NOAA/NWS analyses at 1200 UTC 8 Oct, 0000 UTC 9 Oct, and 1200 UTC 9 Oct 2017. The surface analysis was produced by the NOAA/NWS Weather Prediction Center (WPC) and the upper-air analysis is from NOAA/NWS NAM analysis.

  • View in gallery

    Surface weather observations, including wind (barbs, kt) and sea level pressure (contours, plotted every 2 hPa) over Northern California for 0000 and 1200 UTC 8 Oct and 0000 and 0600 UTC 9 Oct 2017.

  • View in gallery

    Domain used for the WRF-ARW simulations of this event. The outer domain has 36-km grid spacing, with the inner domains (d02, d03, and d04) using 12-, 4-, and 1.33-km grid spacing, respectively.

  • View in gallery

    Sea level pressure (solid lines, contour interval of 2 hPa), 10-m winds (barbs, kt), and 925-hPa temperature (color shading, oC) for 1200 UTC 8 Oct and 0000, 0600, and 1200 UTC 9 Oct 2017 from a WRF simulation initialized at 1200 UTC 8 Oct. A portion of the 12-km domain is shown.

  • View in gallery

    (left) Sea level pressure (solid lines, contour interval of 2 hPa), 10-m winds (barbs, kt), and 925-hPa temperatures (color shading, oC), and (middle) 850- and (right) 700-hPa heights (solid lines, contour interval of 6 gpm at 850 hPa and 15 gpm at 700 hPa), temperatures (color shading, oC), and winds (barbs, kt) at (a)–(c) 1800 UTC 8 Oct, (d)–(f) 0000 UTC 9 Oct, and (g)–(i) 0600 UTC 9 Oct from the 4-km domain of a WRF simulation initialized at 1200 UTC 8 Oct 2017.

  • View in gallery

    Maximum winds in the lowest 250 m (color shading, kt) and winds at roughly 250 m above the surface (barbs, kt) at (a) 1800 UTC 8 Oct, and (b) 0000, (c) 0600, and (d) 1200 UTC 9 Oct 2017. Also shown is (e) the terrain (m) of the corresponding domain and the location of the model vertical cross section (white line).

  • View in gallery

    Vertical cross sections (location shown in Fig. 16; heights in hPa) of simulated potential temperature (solid lines, K), sustained wind speed (color shading, kt), and wind vectors within the cross section at 1800 UTC 8 Oct, and 0000, 0600, and 1200 UTC 9 Oct 2017.

  • View in gallery

    Model vertical soundings at the (left) midstream and (right) upstream locations at (a),(b) 1800 UTC 8 Oct, and (c),(d) 0000, (e),(f) 0600, and (g),(h) 1200 UTC 9 Oct 2017. Temperature (red, oC), dewpoint (green, oC), and wind speed (barbs, kt) are shown.

  • View in gallery

    Backward model trajectories ending at the Santa Rosa RAWS site at 50 (black), 200 (blue), 500 (red), and 1,000 m (orange) AGL. The altitudes of each of the trajectories and the underlying terrain heights (green line) are shown in the insets. Terrain heights from the WRF Model 1.33-km domain are shown with color/gray shading.

  • View in gallery

    The 10-h surface (10 m) wind gust forecast (kt) of the HRRR system valid at 0700 UTC 9 Oct 2017. The model was initialized at 2100 UTC 8 Oct 2017. Wind gusts are derived by determining the highest wind speed in the lowest 1 km.

  • View in gallery

    Sea level pressure analysis (hPa) at 0600 UTC 9 Oct and 24-, 48-, and 72-h GFS model forecasts valid at the same time.

  • View in gallery

    Red-flag warning and maximum wind gust graphic released on 6 Oct 2017 by the San Francisco Bay office of the National Weather Service.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 37 37 37
PDF Downloads 27 27 27

The Northern California Wildfires of 8–9 October 2017: The Role of a Major Downslope Wind Event

View More View Less
  • 1 Department of Atmospheric Sciences, University of Washington, Seattle, Washington
© Get Permissions
Open access

Abstract

A series of large wildfires began over the terrain north of San Francisco, California, during the evening of 8 October 2017 and spread across nearly 250,000 acres, including areas near the towns of Santa Rosa and Napa. These “Wine Country” wildfires were the most destructive in California history, with 44 deaths; the loss of 9,000 buildings; damage to approximately 21,000 structures; $10 billion of insured losses; and substantially greater total economic loss.

This paper describes the synoptic and mesoscale conditions that were associated with the wildfires, with strong, easterly “Diablo” winds playing a central role in both initiating and supporting the fires. The climatological conditions preceding the fires are reviewed, including near-normal precipitation and above-normal temperatures during the summer, as well as much above-normal precipitation the previous winter, which led to abundant dry grass that provided fuel for the wind-driven fires.

High-resolution meteorological modeling realistically simulated the strong winds associated with this event. Importantly, operational mesoscale forecast models provided excellent forecasts of the high winds several days in advance. It appears that a vulnerable power system, urbanization of fire-prone areas, flammable invasive species, and poor communication of dangerous conditions contributed to this catastrophic event. The potential for mitigating or preventing such destructive wildfires using skillful weather prediction is examined, as well as the possible role of global warming.

© 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

CORRESPONDING AUTHOR: Clifford F. Mass, cmass@uw.edu

Abstract

A series of large wildfires began over the terrain north of San Francisco, California, during the evening of 8 October 2017 and spread across nearly 250,000 acres, including areas near the towns of Santa Rosa and Napa. These “Wine Country” wildfires were the most destructive in California history, with 44 deaths; the loss of 9,000 buildings; damage to approximately 21,000 structures; $10 billion of insured losses; and substantially greater total economic loss.

This paper describes the synoptic and mesoscale conditions that were associated with the wildfires, with strong, easterly “Diablo” winds playing a central role in both initiating and supporting the fires. The climatological conditions preceding the fires are reviewed, including near-normal precipitation and above-normal temperatures during the summer, as well as much above-normal precipitation the previous winter, which led to abundant dry grass that provided fuel for the wind-driven fires.

High-resolution meteorological modeling realistically simulated the strong winds associated with this event. Importantly, operational mesoscale forecast models provided excellent forecasts of the high winds several days in advance. It appears that a vulnerable power system, urbanization of fire-prone areas, flammable invasive species, and poor communication of dangerous conditions contributed to this catastrophic event. The potential for mitigating or preventing such destructive wildfires using skillful weather prediction is examined, as well as the possible role of global warming.

© 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

CORRESPONDING AUTHOR: Clifford F. Mass, cmass@uw.edu

The Northern California wildfires of October 2017 were associated with a strong, well-forecast, downslope wind event. Problems with a vulnerable power system, urbanization, and other nonmeteorological issues contributed to the most damaging wildfire in California history.

During the evening of Sunday, 8 October 2017, a series of catastrophic wildfires began over the terrain of north-central California and spread rapidly across approximately 250,000 acres, including the region near the towns of Santa Rosa and Napa. Known as the “Wine Country” wildfires due to the extensive vineyards and wineries of the region, the fires were the most destructive1 in California history: destroying roughly 9,000 structures; damaging 21,000 buildings (www.insurance.ca.gov/0400-news/0100-press-releases/2017/upload/nr135Statewideclaims.pdf); and resulting in over $10 billion (U.S. dollars) of insured losses (Fig. 1). Forty-four people lost their lives, several hundred were hospitalized, and millions of people were exposed to smoke from the fires. The Tubbs fire, the most destructive single fire in California history in terms of lost structures, descended the hills northeast of Santa Rosa, burning 36,807 acres and killing 22 individuals. Other major blazes included the Nun’s fire (54,382 acres), which spread east and north of the city of Sonoma, and the Atlas fire (51,624 acres), north of the city of Napa. The areas of the major fires are shown in Fig. 2, with the fire swaths generally extending from the upper slopes of the regional terrain toward lower elevations to the southwest.

Fig. 1.
Fig. 1.

The remnants of the Fountaingrove neighborhood northeast of Santa Rosa after the Wine Country wildfire. Picture was taken on 14 Oct 2017 and was made available through a Creative Commons license granted by Gary Cedar Photography.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

Fig. 2.
Fig. 2.

(a) Geographical and terrain features of the wildfire region on 8–9 Oct 2017, with the largest fires shown by red areas. (b) The terrain of a broader region, the location of the vertical cross section and sounding locations described in the model simulation section, and the positions of two surface observation sites discussed in the text.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

Terrain maps (Fig. 2) show that the region north of San Francisco is characterized by a collection of ridges generally oriented northwest to southeast, with the crests rising to approximately 800 m (2,500 ft) over the southern portion of the domain and roughly 1,100 m (3,600 ft) at the latitude of the most northern fires. To the east of the main wildfire area of 8–9 October 2017 lies the Central Valley of California and farther east there is the Sierra Nevada, on whose western slopes additional fires were initiated that day. A lower area, known as the Burney Gap, is found over Northern California between the northern terminus of the Sierra Nevada and the southern Cascade Mountains. As will be described later, the interaction of the large-scale flow field with terrain was a critical element of the October 2017 wildfire event.

Strong winds played an essential role in both the initiation and maintenance of the Wine Country wildfires, with gusts reaching 60–95 kt (where 1 kt ≈ 0.5144 m s–1) near and downstream of the crests of the regional terrain. Over central/Northern California, such strong, dry offshore winds are known as “Diablo” or north winds and are close cousins to the Santa Ana winds of Southern California; all of these winds are most frequent during the fall and early winter when high pressure builds inland over the Intermountain West, producing an offshore pressure gradient and dry low-level flow with an easterly component. Many Diablo wind events possess many of the characteristics of a terrain-induced downslope windstorm and are generally strongest at night (e.g., Smith et al. 2018). The term Diablo winds can be traced back to the early 1990s, when the name was popularized by the San Francisco National Weather Service office and local media. Although a few papers have described the general synoptic and local wind patterns associated with this feature (Monteverdi 1973; Monteverdi and Wood 1973: Pagni 1993), none has examined Diablo winds in detail, making use of modern modeling and observational approaches.

The strong winds of the October 2017 event not only aided fire growth and spread, but also initiated many of the fires. Recent comprehensive evaluations (http://calfire.ca.gov/communications/downloads/newsreleases/2018/2017_WildfireSiege_Cause.pdf) by CalFire, the responsible agency of the State of California, found that most of the investigated Wine Country fires were initiated by strong winds damaging the electrical system either due to falling trees/branches or failures of power poles or lines. The winds associated with the Wine Country wildfires were well forecast by a number of operational high-resolution numerical weather prediction systems, such as the National Oceanic and Atmospheric Administration/National Weather Service (NOAA/NWS) High Resolution Rapid Refresh (HRRR) forecast model

An important characteristic of the Wine Country fires was their explosive and nearly simultaneous initiation during the evening of Sunday, 8 October, coincident with the start of the strong winds. Driven by these winds, the fires rapidly increased in size, moving quickly south and southwest into populated areas over a period of 3–6 h. To illustrate, the Tubbs fire initiated around 2145 Pacific daylight time (PDT) on 8 October near the town of Calistoga and traveled over 19 km in the first 3 h (Griggs et al. 2017). The sudden initiation and rapid spread of these fires contributed substantially to the great loss of life and property attending this event.

Beyond the physical damage, economic loss, direct injuries, and deaths, the smoke from the Wine Country fires seriously degraded air quality for millions of people in the San Francisco Bay Area. National Aeronautics and Space Administration (NASA) Moderate Resolution Imaging Spectroradiometer (MODIS) visible imagery showed no fires north of San Francisco during the early afternoon of 8 October; one day later there were massive smoke plumes directed toward the west and southwest (Fig. 3). According to the Bay Area Air Quality Measurement District, the air quality was the worst on record for a number of locations around the San Francisco Bay Area (the record goes back to 1999) (www.npr.org/sections/health-shots/2017/10/12/557394636/this-weeks-air-quality-is-worst-on-record-for-san-francisco-bay-area). To illustrate the air quality impacts of the fires, Fig. 4 shows the air quality index (AQI)2 for the first half of October based on air quality measurements over the North Bay area (where the fires were located) and in the region near San Francisco. The first 8 days of the month had good or moderate air quality, with a rapid transition to unhealthy or very unhealthy conditions on 9 October that continued for several days. Although delayed a few days by initially easterly winds, dense smoke reached San Francisco a few days later as the winds turned more northerly. Very poor air quality continued for over a week in the Wine Country region north of San Francisco and for several days around the city, resulting in increased hospitalizations for asthma and other air-quality-related morbidity (https://newrepublic.com/article/145259/toxic-air-california-public-health-crisis).

Fig. 3.
Fig. 3.

NASA MODIS visible imagery at approximately 2000 UTC (left) 8 and (right) 9 Oct 2017.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

Fig. 4.
Fig. 4.

Daily averaged AQI at locations over the North Bay region (northern zone) and for locations around San Francisco (Coast and Central Bay).

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

The wildfires of 8–9 October 2017 were severe but not unprecedented for the region; wildfires driven by Diablo winds have occurred over central and Northern California many times in the past. Perhaps the most similar in impact was the 1991 Oakland Hills firestorm (a.k.a., the Tunnel or East Bay Hills fire), which occurred on the hillsides of northern Oakland and southeastern Berkeley, California, during the weekend of 19–20 October 1991 (U.S. Fire Administration 1991; Pagni 1993; Sullivan 1997). Driven by northeasterly winds reaching 57 kt, the fire killed 25; injured 150; destroyed nearly 2,843 homes and 437 apartments; and resulted in insured economic losses estimated at $1.5 billion in 1991 dollars. Strong easterly winds also contributed to the 1964 Hanley fire, which burned 83,000 acres northeast of Santa Rosa, encompassing a region nearly identical to the Tubbs fire of October 2017. Another catastrophic wind-driven wildfire occurred in 1923 around the campus of the University of California, Berkeley. Known as the Berkeley fire, it consumed 640 structures and resulted in major improvements in regional building codes and increased capacity for local water mains (www.berkeleypubliclibrary.org/sites/default/files/files/inline/bplhstrm_979.467_st76_the_story_of_the_berkeley_fire.pdf). Figure 5 shows the areas burned by wildfires since 1939, with the boundaries of the major 2017 Wine Country wildfires shown by red lines. Not only have there been numerous wildfires over the region during the past 78 years, but major previous fires have occurred within the boundaries of the 2017 burn areas.

Fig. 5.
Fig. 5.

Fire extents of wildfires between 1939 and 2016 (color shading), with the perimeters of the 2017 Wine Country fires shown by the red outlines. The border of Sonoma County is shown by a black line. Image provided by the Sonoma County Land Trust (https://sonomalandtrust.org/explore/fire-history.html).

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

Tree mortality due to drought, insect infestation, or other causes does not appear to be a significant contributor to the Wine Country fires (Keeley 2017). This finding is supported by the State of California–U.S. Department of Agriculture (USDA) Forest Service Tree Mortality Viewer website (http://egis.fire.ca.gov/TreeMortalityViewer/), which shows very little tree mortality in the region of the October 2017 fires.

This paper examines the meteorological context of the Wine Country wildfires of October 2017, using both observations and high-resolution numerical modeling. It is shown that a well-predicted synoptic evolution led to a localized downslope wind event, one that was skillfully forecast in the preceding days by high-resolution mesoscale models. The implications of increasing mesoscale forecast skill for the prediction and management of such wildfires, the role of the urbanization and fire suppression of the surrounding environment, the potential to reduce the frequency and intensity of such fires, and the impacts of anthropogenic global warming are also discussed.

ANTECEDENT METEOROLOGICAL CONDITIONS.

A major question regarding central California coastal wildfires, such as the October 2017 event, is the role of antecedent weather conditions. There is a considerable literature suggesting that wildfire frequency and area burned in the predominantly grassland/shrub environment of the hills north of San Francisco are relatively insensitive to the temperature and precipitation of the preceding summer, since a normal summer is sufficient for drying most fuels (e.g., Keeley and Fotheringham 2003; Abatzoglou and Kolden 2013; Keeley and Syphard 2016). In fact, most central California coastal locations have little precipitation from June through September during normal years (generally less than 0.5 in. month–1). Furthermore, many of the largest fires are associated with periods of dry, offshore flow that can rapidly desiccate even initially moist fine fuels (e.g., Keeley et al. 2004). Enhanced precipitation during the prior winter, as occurred for the October 2017 wildfires, appears to increase the danger of wildfires during the next warm season, since it encourages the growth of herbaceous fuels (e.g., grass), which dry during the subsequent summer and provide substantial fuel availability for late summer/autumn fires (Westerling et al. 2003, 2004; Littell et al. 2009; Dudney et al. 2017). The association of many central California wildfires with strong offshore flow results in the nearly simultaneous initiation of multiple fires, a dangerous situation that can overwhelm the resources of responders, as it did during the October 2017 Wine Country fires.

The strong winds of 8–9 October 2017 struck coastal central California during the period when surface conditions are typically dry, following the climatologically warm and dry period from May through September. Figure 6 presents the historical weather conditions for the U.S. Climate Division encompassing much of the fire area (California Division 1, North Coast Drainage) for 1948–2017. May–September 2017 mean temperatures were at record levels and approximately the same as in 2015: about 2.5°F (1.4°C) above the long-term average (Fig. 6a). The summer (May–September) precipitation was low (about 2 in.), but typical for that period (Fig. 6b). In contrast, the prior winter’s (November 2016–April 2017) precipitation for that climate division (57 in.) was approximately 20 in. above normal (Fig. 6c); as a result, the Palmer drought severity index (PDSI; Fig. 6d), which provides a measure of soil dryness, indicated only slightly drier than normal conditions for the warm season (May–September). Such a minimal difference from normal soil moisture conditions reflected a wetter than normal PDSI early in the summer, declining to drier than normal conditions by the time of the wildfire.

Fig. 6.
Fig. 6.

NOAA/NWS Climate Division data for the California North Coast Drainage Division for the (a) May–Sep temperature anomaly (oF) from the 1960–90 normal, (b) May–Sep precipitation (in.), and (d) May–Sep PDSI for 1948–2017. (c) Precipitation for Nov–Apr.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

The daily conditions at the USDA Remote Automated Weather Stations (RAWS) site above Santa Rosa3 (576 ft) for the 3-month period leading up to and including the fire (15 July–15 October) are shown in Fig. 7. The temperatures during the days leading to the event (Fig. 7a) were not exceptional (∼29°C) and were equaled or exceeded many times during the previous months. Relative humidity during the event, under strong northeasterly flow, was the lowest of the period (∼10%), but not appreciably lower than during other dry events that summer (Fig. 7b). During much of the summer there was a large diurnal variation in relative humidity (from roughly 100% in the morning to about 45% during the afternoon), which was replaced by a drier regime during the week before the wildfire. The wind direction at Santa Rosa varied greatly during the summer (Fig. 7c), and was highly correlated with relative humidity; specifically, higher relative humidity accompanied winds from the south and west, while the lowest humidities were associated with northeasterly, offshore flow. Maximum wind gusts were modest during most of the summer (Fig. 7d), typically reaching 7–11 m s–1, with the strong winds on 8–9 October (∼30 m s–1) being a clear outlier.

Fig. 7.
Fig. 7.

(a) Temperature (oC), (b) relative humidity (%), (c) wind direction (o), and (d) wind gusts (m s–1) at the Santa Ana RAWS site from 15 Jul to 15 Oct 2017.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

Figure 8 presents the 10-h fuel moisture at the Santa Rosa RAWS site from 13 April through 10 October 2017. Ten-hour fuel moisture is a measure of the moisture content of vegetation with a diameter of 1/4–1 in. Throughout the late spring and summer, the fuel moisture varied between roughly 22% and 6%, with the latter values associated with dry, offshore flow periods. The week prior to the incident, characterized by dry, warm offshore flow, had 10-h fuel moistures around 6%, similar to other dry periods during the summer. Finer fuels such as grasses would be expected to respond even more quickly than 10-h fuel to environmental conditions and, thus, would have been primed to burn as the easterly winds accelerated on 8 October. A histogram of historical (1991–2017) 10-h fuel moisture observed at the Santa Rosa RAWS site for the 1–15 October period indicates the values of 6% or less are not rare, encompassing approximately 14% of all hours for that time period.

Fig. 8
Fig. 8

(top). (a) The 10-h fuel moisture percentages at the RAWS site at Santa Rosa from 13 Apr to 10 Oct 2017. (b) Histogram of the frequency of 10-h fuel moisture for the 1–15 Oct time window for 1991–2017 (local standard time).

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

In summary, although the summer of 2017 was warmer than normal over central coastal California, heavy winter precipitation and typical summer rainfall left the surface moisture in a state that was dry, but not exceptional, for early October (Fig. 8b). With substantial precipitation the preceding winter, the growth of grasses was enhanced, making above normal amounts of fuels available for burning during the late summer and early fall (Dudney et al. 2017; www.sfgate.com/news/article/green-hills-Bay-Area-California-rain-drought-grass-10992664.php).

THE MAXIMUM WINDS OBSERVED DURING THE 8–9 OCTOBER WILDFIRE EVENT.

Wind speeds during the event were highly variable across the region, with gusts ranging from 10–20 kt over lower elevations to 60–95 kt near and downstream of the crests of the regional terrain. Figure 9, which shows winds from aviation routine weather report (METAR), Meteorological Assimilation Data Ingest System (MADIS), and USDA RAWS sites, illustrates this variability. Immediately above Santa Rosa, a center of wildfire damage, winds at the RAWS site reached 59 kt, while winds as strong as 94 kt4 were found on the upper slopes of a ridge northwest of that town. Strong, but lesser, winds were also observed near the ridgetops east of Berkeley and Oakland, California, and along the western slopes of the Sierra Nevada, which was another locus of wildfire initiation.

Fig. 9
Fig. 9

(bottom). Maximum wind gusts (kt) during the 8–10 Oct 2017 Wine Country wind event. Winds are from METAR, MADIS, and USDA RAWS sites.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

An important characteristic of the high winds during this event was their short (∼12 h) duration, with a rapid increase during the evening of Sunday, 8 October, and a rapid decline during the morning of 9 October. The winds at the Santa Rosa and Hawkeye RAWS stations (Fig. 10) illustrate this evolution, with high winds limited to roughly a 10–12-h period and maximum gusts reaching 59 and 69 kt, respectively. As described later, the magnitude and rapid increase of the winds were well predicted by high-resolution numerical models initialized during the preceding days.

Fig. 10.
Fig. 10.

Wind gusts (kt) at the (a) Santa Rosa and (b) Hawkeye RAWS sites north of the Bay Area. Locations noted in Figs. 2 and 16.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

The northeasterly winds on and downstream of the crests of regional terrain were highly unusual in their magnitude. For example, at the Hawkeye RAWS site, 45 km to the northwest of Santa Rosa, the maximum wind gusts during the 8–9 October 2017 event (79 mph, 69 kt) were not only the strongest northeasterly gusts on record (dating back to September 1993), but were the strongest winds from any direction. At the Santa Rosa RAWS site (dating back to 1991), the northeasterly winds (68 mph, 59 kt) were the second strongest on record from the northeast quadrant. In contrast, winds at lower elevations were moderate and not unusual.

THE SYNOPTIC AND MESOSCALE EVOLUTIONS ACCOMPANYING THE EVENT.

The synoptic-scale evolution of the 8–9 October wildfire/windstorm was typical of strong autumn offshore wind events over the central coastal California region, including the passage of an upper-level trough across the Pacific Northwest, the movement of high surface pressure to the north and east of the region, and the development of offshore, northeasterly flow in the lower troposphere (Fig. 11). Aloft (500 hPa), an upper-level trough crossed the northwest coast and moved southeastward across the western United States, followed by an upper-level ridge. Strong 500-hPa northerly flow developed over Northern California between 1200 UTC 8 October and 0000 UTC 9 October, immediately before the initiation of the surface wind event. These northerly winds declined rapidly after 1200 UTC 9 October, as the trough moved quickly to the east.

Fig. 11.
Fig. 11.

Surface and 500-hPa NOAA/NWS analyses at 1200 UTC 8 Oct, 0000 UTC 9 Oct, and 1200 UTC 9 Oct 2017. The surface analysis was produced by the NOAA/NWS Weather Prediction Center (WPC) and the upper-air analysis is from NOAA/NWS NAM analysis.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

At the surface, sea level pressure gradients were initially weak over Northern California (1200 UTC 8 October), with high pressure over the eastern Pacific extending into the Pacific Northwest and a front stretching across the northern portion of California. As high pressure and cool air extended east and southeast into the northern Rockies and Nevada, the front pushed southeastward and a trough developed along the California coast, with a large offshore pressure gradient developing over the northeastern portion of California. The coastal trough strengthened further by 1200 UTC 9 October, followed by rapid weakening.

Figure 12 shows surface observations over Northern California and Nevada and a mesoscale sea level isobaric analysis during the event. At 0000 UTC 8 October, winds were light over most of the region, with northerly flow beginning to invade from the north. The largest pressure gradients were over the northwest portion of the domain between Pacific high pressure and lower pressure over the Central Valley of California. Twelve hours later (1200 UTC), high pressure pushed into Oregon, producing an enhanced meridional pressure gradient over Northern California. As a result of these pressure changes, northerly winds accelerated to 20 kt over portions of the Central Valley and the coastal waters. By 0000 UTC 9 October, as pressure increased over the northern portion of the domain and high pressure extended southeastward over Nevada, pressure gradients greatly increased over northern and northeastern California, with strengthening northerly and northeasterly winds extending over the region north of San Francisco. Six hours later (0600 UTC 9 October), near the time of strongest winds, pressure gradients continued to intensify northeast of San Francisco, as pressure increased further over Nevada and eastern Oregon. Two windward ridge–lee trough pressure couplets were evident: a stronger one across the Sierra Nevada and a weaker one associated with the terrain between the Central Valley and the coast. With the enhanced pressure gradients, northerly and northeasterly winds strengthened north of San Francisco.

Fig. 12.
Fig. 12.

Surface weather observations, including wind (barbs, kt) and sea level pressure (contours, plotted every 2 hPa) over Northern California for 0000 and 1200 UTC 8 Oct and 0000 and 0600 UTC 9 Oct 2017.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

WRF MODEL SIMULATIONS.

To explore the structure and dynamics of this wind/wildfire event, an Advanced Research version of the Weather Research and Forecasting (WRF) Model (WRF-ARW) simulation was initialized at 1200 UTC 8 October and run through 0000 UTC 10 October 2017. Initialization and boundary conditions were obtained from the 1/4° resolution NOAA/NWS Global Forecast System (GFS) grids for forecast hours 0–36. A nested configuration of 36-, 12-, 4-, and 1.33-km grid spacing was applied, with the 1.33-km domain encompassing the terrain associated with the wind/fire event (Fig. 13). The model configuration was the same as the highly tested University of Washington real-time WRF configuration, using WRF-ARW V3.7.1, Yonsei University (YSU) planetary boundary layer (PBL), Thompson microphysics, RRTM for GCMs (RRTMG) radiation, and the Noah land surface model with multiple parameterization options (Noah-MP), but with one exception: the number of vertical levels was increased from 38 to 75. Testing other options (e.g., MYNN PBL and 125 vertical levels) produced only minor changes in the results.

Fig. 13.
Fig. 13.

Domain used for the WRF-ARW simulations of this event. The outer domain has 36-km grid spacing, with the inner domains (d02, d03, and d04) using 12-, 4-, and 1.33-km grid spacing, respectively.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

Figure 14 shows the simulated sea level pressure and 925-hPa temperatures from the 12-km nest of the WRF simulation initialized at 1200 UTC 8 October, roughly 12 h prior to the initiation of strong winds. At 1200 UTC 8 October, cool low-level air and associated high sea level pressure were moving into the Pacific Northwest, with a substantial pressure gradient over Northern California. During the next 12 h, the cool air and high pressure pushed inland and southward into northern Nevada, resulting in a strengthened pressure gradient over the northeastern section of California. By 0600 UTC 9 October, when the winds north of San Francisco were approaching their peak, cool air had reached southern Nevada, with an intense offshore pressure gradient building along the California–Nevada border. Interactions of strengthening northeasterly winds with the Sierra Nevada and the terrain north of San Francisco produced the observed wavelike perturbations in sea level pressure northeast of San Francisco and strong northerly winds north of the Bay Area. The pressure gradients weakened by 1200 UTC 9 October and attenuated further by 1800 UTC (not shown). These and other simulated fields document the close fidelity of the synoptic/mesoscale evolution of the modeled event to the observed evolution (e.g., Figs. 11 and 12).

Fig. 14.
Fig. 14.

Sea level pressure (solid lines, contour interval of 2 hPa), 10-m winds (barbs, kt), and 925-hPa temperature (color shading, oC) for 1200 UTC 8 Oct and 0000, 0600, and 1200 UTC 9 Oct 2017 from a WRF simulation initialized at 1200 UTC 8 Oct. A portion of the 12-km domain is shown.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

An important mesoscale feature of this event is a low-level pressure ridge and associated northerly flow that developed over the western side of the Central Valley of Northern California; as this pressure ridge extended southward, strong pressure gradients and associated gusty northerly and northeasterly winds moved toward and over the Wine Country region. This feature also influenced the vertical structure of the air approaching the terrain north and east of San Francisco. To explore the three-dimensional structural development of this feature, Fig. 15 shows simulated sea level pressure and height, temperature, and wind at the surface, 850 hPa, and 700 hPa over Northern California from the 4-km WRF domain during the period of wind intensification (1800 UTC 8 October–0600 UTC 9 October).

Fig. 15.
Fig. 15.

(left) Sea level pressure (solid lines, contour interval of 2 hPa), 10-m winds (barbs, kt), and 925-hPa temperatures (color shading, oC), and (middle) 850- and (right) 700-hPa heights (solid lines, contour interval of 6 gpm at 850 hPa and 15 gpm at 700 hPa), temperatures (color shading, oC), and winds (barbs, kt) at (a)–(c) 1800 UTC 8 Oct, (d)–(f) 0000 UTC 9 Oct, and (g)–(i) 0600 UTC 9 Oct from the 4-km domain of a WRF simulation initialized at 1200 UTC 8 Oct 2017.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

At the surface, the Central Valley sea level pressure ridge is evident at 1700 UTC 8 October, weakens due to surface heating at 0000 UTC [1700 Pacific daylight time (PDT)] 9 October, and then greatly strengthens and extends southward by 0600 UTC. There is some suggestion of cooler air moving through the Burney Gap at the north end of the Sierra Nevada (see terrain in Fig. 2) and then flowing southward into the Central Valley. At 850 hPa, the Central Valley pressure ridge develops during the same 12-h period, bounded by the troughs resulting from downslope (easterly) flow descending the Sierra Nevada and coastal mountains. Upslope flow and resulting adiabatic cooling on the coastal mountains contributed to the Central Valley ridging at 850 hPa. At 700 hPa, the large-scale flow turned from northerly to easterly during the 12-h period, resulting in downslope warming in the lee of the Sierra Nevada and coastal mountains. Cooler air is found over the Central Valley, with only a slight hint of the pressure ridge in the 700-hPa geopotential heights. Little evidence of the mesoscale pressure ridge or other terrain effects is noted at 500 hPa (not shown).

Turning to the 1.3-km domain, Fig. 16 shows low-level (250 m above the surface) wind barbs and the strongest winds in the lowest 250 m (an attempt to display a parameter reflecting the maximum expected surface gust) for the period from 1800 UTC 8 October through 1200 UTC 9 October. At 1800 UTC 8 October, winds were less than 20 kt over most of the populated regions around Santa Rosa, with the strongest winds (reaching 40 kt) limited to the immediate lee of the terrain crests. Six hours later (0000 UTC), winds accelerated modestly over the interior, but remained below 30 kt over most of the Wine Country region, with the exception of near the highest terrain. The situation changed radically by 0600 UTC, with areas immediately downstream of the terrain crests experiencing wind gusts of 65–75 kt, with gusts greater than 50 kt reaching the northwest section of Santa Rosa. By 1200 UTC the winds had weakened greatly, with strong winds limited to near the crests of the regional terrain.

Fig. 16.
Fig. 16.

Maximum winds in the lowest 250 m (color shading, kt) and winds at roughly 250 m above the surface (barbs, kt) at (a) 1800 UTC 8 Oct, and (b) 0000, (c) 0600, and (d) 1200 UTC 9 Oct 2017. Also shown is (e) the terrain (m) of the corresponding domain and the location of the model vertical cross section (white line).

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

A series of vertical cross sections directed southwest–northeast across the coastal mountains to the Central Valley and roughly parallel to the lower-tropospheric winds (locations shown in Figs. 2 and 16) illustrates the high-amplitude wave structure and temporal evolution aloft that accompanied the Wine Country wind event (Fig. 17). Earlier in the day (1800 UTC), there is evidence of modest mountain wave/downslope wind acceleration on the lee slopes of the terrain above Santa Rosa, with even stronger downslope acceleration associated with another crest to the northeast (Fig. 17a). A stable layer is noted upstream and above the terrain crest east of Santa Rosa. By 0000 UTC, surface heating resulted in a deeper mixed layer upstream of the terrain crest, with little wind acceleration compared to the initial time. During the next 6 h, increasing winds upstream of the crest and a lowering of the stable layer (as diurnal heating ended and cooler air moved into the Central Valley), resulted in substantial amplification of the mountain wave and associated downslope flow, with an elevated area of strong winds extending eastward over the lowlands. By 1200 UTC, the surface layer cooled and the wave amplitude declined substantially, resulting in attenuation of the downslope winds.

Fig. 17.
Fig. 17.

Vertical cross sections (location shown in Fig. 16; heights in hPa) of simulated potential temperature (solid lines, K), sustained wind speed (color shading, kt), and wind vectors within the cross section at 1800 UTC 8 Oct, and 0000, 0600, and 1200 UTC 9 Oct 2017.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

Additional insight into the control of the terrain-related wave amplification by the upstream flow is provided by vertical soundings (Fig. 18) immediately upstream of the Santa Rosa terrain crest (midstream) and over the Central Valley of California (upstream, locations shown in Figs. 2 and 16). At 1800 UTC, the winds at both locations were northerly in the lower troposphere, with only a small easterly component at midstream. By 0000 UTC 9 October, a deepening dry-adiabatic surface layer, driven by surface heating, extended to 870 hPa at midstream and 800 hPa upstream, with a small rotation of wind direction toward easterly at both locations compared to the previous time. The situation changes substantially by 0600 UTC, with the winds above approximately 950 hPa strengthening and turning sharply toward the northeast at both locations. At the same time, the sounding upstream becomes more stable above crest level (between 900 and 800 hPa), with a critical level (wind reversal) around 700 hPa. Mountain waves and associated downslope winds may undergo amplification in the presence of a stable layer near crest level or an elevated environmental critical level, an elevation at which the environmental wind speed with respect to a mountain barrier goes to zero (Durran 1990). Finally, at 1200 UTC, when the winds north of San Francisco had lessened, the upstream flow had weakened, with a considerable increase of stability between 900 and 800 hPa at the midstream sounding.

Fig. 18.
Fig. 18.

Model vertical soundings at the (left) midstream and (right) upstream locations at (a),(b) 1800 UTC 8 Oct, and (c),(d) 0000, (e),(f) 0600, and (g),(h) 1200 UTC 9 Oct 2017. Temperature (red, oC), dewpoint (green, oC), and wind speed (barbs, kt) are shown.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

There is an extensive body of literature that describes the conditions producing downslope windstorms associated with high-amplitude mountain waves (e.g., Durran 1986; Durran and Klemp 1987; Durran 1990; Colle and Mass 1998a,b). Many of these events occur when strong winds approach the crest and a stable layer or critical level positioned at or above crest level helps trap wave energy within the lower troposphere. As noted above, such conditions were observed during the initiation of the strong winds during the Wine Country wind event and were successfully reproduced by numerical models.

Finally, Fig. 19 presents air parcel trajectories ending at the Santa Rosa RAWS site at several elevations (50, 200, 500, and 1,000 m), based on the simulation in the 1.33-km WRF domain. The air reaching the Santa Rosa RAWS at the lowest level (50 m) passed through the Burney Gap, just north of the main Sierra Nevada barrier, at approximately 2,000 m and then descended rapidly to roughly 400 m over the Central Valley, followed by modest ascent and decent over the terrain northeast of Santa Rosa. In contrast, the trajectories ending at 200, 500, and 1,000 m originated in the middle troposphere (around 4,000 m), moved southward over Nevada, turned sharply toward the west, and then descended rapidly over the western slopes of the Sierra Nevada. These trajectories then ascended and descended over the coastal terrain northeast of Santa Rosa. With such a history of descent, the relative humidity of the air reaching the Wine Country region was quite dry, dropping below 20% in a number of locations.

Fig. 19.
Fig. 19.

Backward model trajectories ending at the Santa Rosa RAWS site at 50 (black), 200 (blue), 500 (red), and 1,000 m (orange) AGL. The altitudes of each of the trajectories and the underlying terrain heights (green line) are shown in the insets. Terrain heights from the WRF Model 1.33-km domain are shown with color/gray shading.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

FORECAST QUALITY OF REAL-TIME, MESOSCALE FORECASTING SYSTEMS.

An important feature of the 8–9 October 2017 windstorm/wildfire event was the realistic simulation of the strong winds both in the research simulation shown above and in operational predictions from high-resolution, operational forecasting systems during the preceding days. For example, the HRRR modeling system is initialized and run each hour to 18 h with 3-km grid spacing using the WRF-ARW model. Every run on 8 October predicted a strong wind event over the Santa Rosa/Wine Country region that evening. For example, the 10-h HRRR surface wind gust forecast valid at 0700 UTC 9 October predicted 10-m wind gusts as high as 50–55 kt over the terrain north and east of Santa Rosa (Fig. 20).

Fig. 20.
Fig. 20.

The 10-h surface (10 m) wind gust forecast (kt) of the HRRR system valid at 0700 UTC 9 Oct 2017. The model was initialized at 2100 UTC 8 Oct 2017. Wind gusts are derived by determining the highest wind speed in the lowest 1 km.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

Operational high-resolution WRF forecasts by the California and Nevada Smoke and Air Committee (CANSAC) group, which is associated with the Desert Research Institute (DRI) in Reno, Nevada, are made twice daily using 2-km grid spacing for a domain covering California and Nevada. Their predicted surface winds, made 30 h prior to the event, were from the north or northeast over the Wine Country area, with sustained winds of 35–45 kt over and near the ridges of the regional terrain (not shown); these predictions were quite similar to observations. CANSAC runs, as well as operational National Centers for Environmental Prediction (NCEP) models [e.g., GFS, North American Mesoscale Forecast System (NAM)] made several days before the incident, realistically predicted the upper-level trough passage, increasing pressure gradients, and strong winds during the evening in question. As shown in Fig. 21, the NOAA/NWS GFS forecasts valid at the height of the event (0600 UTC 9 October) were virtually the same for all model cycles initialized during the previous 72 h. Furthermore, the wind gust forecast, released Friday morning by the San Francisco Bay Area office of the National Weather Service based on WRF runs with 3-km grid spacing, predicted winds over 55 mph in the hills above Santa Rosa (Fig. 22). Based on this excellent forecast guidance, the National Weather Service provided highly accurate and threatening warnings during the days before the event. For example, during the morning of Friday, 6 October, the San Francisco NWS office released a red flag warning for the northeast Bay Area valid 1100 local time (LT) Sunday through 0500 LT Tuesday, predicting low humidity, gusts to 55 mph (48 kt), and rapid spreading of any fires (Fig. 22). The type of downslope windstorm associated with the Wine Country event, one associated with a critical level just above the crests of the regional terrain, may be particularly predictable since critical-level existence and elevation are controlled by the more predictable synoptic-scale flow (D. Durran 2018, personal communication).

Fig. 21.
Fig. 21.

Sea level pressure analysis (hPa) at 0600 UTC 9 Oct and 24-, 48-, and 72-h GFS model forecasts valid at the same time.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

Fig. 22.
Fig. 22.

Red-flag warning and maximum wind gust graphic released on 6 Oct 2017 by the San Francisco Bay office of the National Weather Service.

Citation: Bulletin of the American Meteorological Society 100, 2; 10.1175/BAMS-D-18-0037.1

The ability of operational, high-resolution prediction systems to provide realistic forecasts of the strong winds over and downwind of the terrain associated with such wildfire/downslope wind events offers substantial potential for saving life and property. Based on accurate high-wind forecasts, vulnerable citizens can be evacuated and firefighting resources enhanced and prepared. Preemptive power outages in vulnerable areas could reduce the potential for electrical fires, the apparent cause of the October 2017 event. For example, based on the high-wind forecasts of the previous days, power lines in the hills above Santa Rosa and Napa could have been deenergized a few hours before the winds hit the region. Since it appears that most of the fires of the October 2017 event were initiated by the strong winds (www.mercurynews.com/2017/10/10/pge-power-lines-linked-to-wine-country-fires/; http:/calfire.ca.gov/communications/downloads/newsreleases/2018/2017_WildfireSiege_Cause.pdf) interacting with the power infrastructure, many of the fires might have been prevented by preemptive power outages. Experimentation with such preemptive power outages guided by high-resolution model forecasts and observations is already taking place in San Diego, California (www.energymanagertoday.com/socal-wildfire-outages-0173692/), and has recently been proposed by the Pacific Gas and Electric Company (PG&E), the power utility for the North Bay area.

ANTHROPOGENIC CONTRIBUTIONS TO CENTRAL CALIFORNIA WILDFIRES.

The lack of correlation between late summer/autumn wildfires over central coastal California and summer precipitation or temperature weakens the potential connection of the current coastal California wildfires with anthropogenic global warming. Even without additional warming, coastal California summers are sufficiently dry and warm to desiccate fine and larger fuels, with dry offshore flow rapidly reducing relative humidity and fuel moisture in any case. Most of the largest coastal California wildfire events have been associated with strong offshore winds: Santa Ana and sundowner winds over Southern California, and Diablo winds over the central and northern coastal areas of the state. There is no reason to expect that strong offshore winds in autumn have increased or will increase due to anthropogenic global warming; in fact, the opposite may be the case, since model projections suggest that anthropogenic global warming will warm the interior more than the coastal zone, preferentially reducing pressure over the inland region, thus reducing the offshore pressure gradient (e.g., Hughes et al. 2009). As described earlier in the paper, one weather/climate variable that does correlate with fall wildfire frequency and area burned over coastal California, above-normal precipitation during the previous winter, does not appear to have been enhanced by the anthropogenic global warming of the past decades. As shown in Fig. 6, there has been no observed trend during the past half-century in extreme winter precipitation over coastal California. Furthermore, studies using climate model simulations driven by increasing greenhouse gas concentrations (e.g., Swain et al. 2018) do not suggest an increase in heavy precipitation events during the contemporary period. Furthermore, the potential for increasing winter precipitation over coastal California later in the present century is uncertain, with some climate model ensembles suggesting general drying over the region and others projecting increased frequency in heavy precipitation events (e.g., Deser et al. 2012; Polade et al. 2017; Swain et al. 2018). Finally, there is no long-term trend in the number of wildfires over coastal California (Dennison et al. 2014), a trend that might be consistent with a small role of anthropogenic climate change driven by increasing greenhouse gases, or the result of compensation between a number of competing factors. For example, active suppression of fires could be reducing the number of wildfires, while fires are enhanced by population growth increasing ignitions and flammable structures, the spread of flammable invasive species, and increasing temperatures driven by global warming and land-use changes. Further research is also needed to evaluate whether the period of dry fuels might extend later into the fall and winter as the planet warms during the upcoming century.

There are a number of factors that can contribute to an increasing wildfire threat in the Wine Country region, even if the climate remains unchanged. A large population increase has occurred across coastal California during the past 75 years, with much of that growth occurring at the urban–wildland interface. For example, in Sonoma County, which encompasses many of the Wine Country fires, the population has increased sixfold since 1940 (http://sonomacounty.ca.gov/CAO/Public-Reports/About-Sonoma-County/Population-Growth/). Such population growth not only increases the numbers of vulnerable population and structures, but leads to more human-initiated fires from the electrical distribution system and other causes. To illustrate this increased vulnerability, the Hanley fire of 1964 struck essentially the same area as the Santa Rosa Tubbs fire of 2017, but with no loss of life and destroying only about 100 structures (Keeley 2017).

Fire is no stranger to the region, with regular fires being a normal part of the regional ecology. Suppression of fire during the past century has enhanced vegetative density far beyond that occurring under a natural fire regime, contributing to larger and more catastrophic fires (www.firesafesonoma.org/main/sites/default/files/CWPP%20Final.pdf). Native chaparral can burn with great intensity, and highly flammable invasive species have been brought into the area by human activity, including Eucalyptus trees and cheatgrass (Bromus Tectorum; www.cal-ipc.org/resources/library/publications/ipcw/report21/), also known as “grassoline.”

As noted earlier, the expansion of the electrical distribution system, much of it vulnerable to strong winds, provides multiple points of wildfire initiation. This problem has been worsened by the local utility’s (PG&E) use of reclosers, devices that repeatedly attempt to reenergize power lines that experience a short circuit. Specifically, the continued attempt to reenergize power lines potentially leads to more fires (https://spectrum.ieee.org/energywise/energy/the-smarter-grid/utilities-probed-as-potential-cause-of-california-fires). Reprogramming reclosers not to reenergize circuits during wildfire season, something done by San Diego Gas and Electric, could potentially reduce the number of fires.

Several other problems also contribute to the increasing costs of California wildfires. For example, reduction in state and local support for firefighting resources and personnel, including the failure of the “mutual aid” approach to wildfire control, may have hurt the response for the Wine Country wildfires (www.sfchronicle.com/news/article/Unlike-in-Wine-Country-fires-mutual-aid-is-12411585.php). In addition, an ineffective warning communication system led to unnecessary loss of life and property (www.sfchronicle.com/bayarea/article/Alerts-reached-few-in-Sonoma-County-in-early-12477770.php).

SUMMARY AND CONCLUSIONS.

This paper describes the catastrophic fires that struck the hill country north of San Francisco during the evening and early morning hours of 8–9 October 2017. These Wine Country wildfires were some of the most costly in California history, with 44 deaths; damage to approximately 21,000 structures; the destruction of roughly 9,000 buildings; and over $10 billion of insured losses.

The central cause of this wildfire event was the strong winds that developed during the evening of 8 October, as high pressure and cooler temperatures pushed inland across the Pacific Northwest and then southward into Nevada, resulting in a strong offshore pressure gradient and accompanying high winds. These winds, which reached 60–95 kt over and to the lee of the crests of the regional terrain, were associated with a downslope windstorm event produced by a high-amplitude mountain wave. Strong downslope winds were the result of strong flow approaching the terrain, with both a stable layer and a critical level above terrain crests. Winds developed rapidly during the evening of 8 October and weakened abruptly about 12 h later. At some locations within the wildfire area, the winds of 8–9 October were the strongest on record. It appears that most of the Wine Country fires were initiated by a failing power distribution infrastructure, with repeated attempts to reenergize lines increasing the potential for fires.

The climatological conditions preceding the fires included near-normal precipitation and above-normal temperatures during the summer, and much above-normal precipitation the previous winter, which led to abundant dry grass that provided fuel for the wind-driven fires. Considering that a normal summer/early fall is sufficiently dry and warm to desiccate fine and larger fuels, and the lack of evidence that the frequency or strength of strong offshore winds will increase as Earth warms, it does not appear that California coastal wildfire events during the late summer and autumn have been enhanced by anthropogenic global warming. But even without a climate change component, wildfire vulnerability in the region has increased substantially during the past decades, due to rapidly growing population in the urban–wildland interface, the suppression of fire and attendant growth in fuel availability, and the spread of highly flammable, nonnative plant species.

High-resolution modeling realistically simulated the winds associated with this event. Importantly, operational mesoscale forecast models provided excellent forecasts of the strong offshore-directed winds several days in advance, allowing the National Weather Service to provide skillful warnings of the potential for large wildfires in the Wine Country area. Skillful weather prediction offers the potential for mitigating or preventing such catastrophic wildfires, such as by evacuating vulnerable populations or by preemptively deenergizing vulnerable electrical infrastructure.

ACKNOWLEDGMENTS

This research was supported by the National Science Foundation through Grant AGS-1349847. Mr. Conor McNicholas created the maximum wind gust map. Professor Brian Harvey of the UW School of Environment and Forest Science and Dr. Brian Potter of the USDA Forest Service Fire Science Group provided important insights on wildfire science. Ms. Beth Tully prepared several of the graphics used in this study.

REFERENCES

  • Abatzoglou, J. T., and C. A. Kolden, 2013: Relationships between climate and macroscale area burned in the western United States. Int. J. Wildland Fire, 22, 10031020, https://doi.org/10.1071/WF13019.

    • Search Google Scholar
    • Export Citation
  • Colle, B. A., and C. F. Mass, 1998a: Windstorms along the western side of the Washington Cascade Mountains. Part I: A high-resolution observational and modeling study of the 12 February 1995 event. Mon. Wea. Rev., 126, 2852, https://doi.org/10.1175/1520-0493(1998)126<0028:WATWSO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Colle, B. A., and C. F. Mass, 1998b: Windstorms along the western side of the Washington Cascade Mountains. Part II: Characteristics of past events and three-dimensional idealized simulations. Mon. Wea. Rev., 126, 5371, https://doi.org/10.1175/1520-0493(1998)126<0053:WATWSO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Dennison, P. E., S. C. Brewer, J. D. Arnold, and M. A. Moritz, 2014: Large wild fire trends in the western United States, 1984–2011. Geophys. Res. Lett., 41, 29282933, https://doi.org/10.1002/2014GL059576.

    • Search Google Scholar
    • Export Citation
  • Deser, C., R. Knutti, S. Solomon, and A. S. Philips, 2012: Communication of the role of natural variability in future North American climate. Nat. Climate Change, 2, 775779, https://doi.org/10.1038/nclimate1562.

    • Search Google Scholar
    • Export Citation
  • Dudney, J., L. M. Hallett, L. Larios, E. C. Farrer, E. N. Spotswood, C. Stein, and K. N. Suding, 2017: Lagging behind: Have we overlooked previous-year rainfall effects in annual grasslands? J. Ecol., 105, 484495, https://doi.org/10.1111/1365-2745.12671.

    • Search Google Scholar
    • Export Citation
  • Durran, D. R., 1986: Another look at downslope windstorms. Part I: On the development of supercritical flow in an infinitely deep, continuously stratified fluid. J. Atmos. Sci., 43, 25272543, https://doi.org/10.1175/1520-0469(1986)043<2527:ALADWP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Durran, D. R., 1990: Mountain waves and downslope winds. Atmospheric Process over Complex Terrain, Meteor. Monogr., No. 45, Amer. Meteor. Soc., 59–81.

  • Durran, D. R., and J. B. Klemp, 1987: Another look at downslope winds. Part II: Nonlinear amplification beneath wave-overturning layers. J. Atmos. Sci., 44, 34023412, https://doi.org/10.1175/1520-0469(1987)044<3402:ALADWP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Griggs, T., K. K. R. Lai, H. Park, J. K. Patel, and J. White, 2017: Minutes to escape: How one California wildfire damaged so much so quickly. New York Times, www.nytimes.com/interactive/2017/10/12/us/california-wildfire-conditions-speed.html.

  • Hughes, M., A. Hall, and J. Kim, 2009: Anthropogenic reductions of Santa Ana winds. California Climate Change Center Rep. CEC-500-2009-015-D, 19 pp., www.energy.ca.gov/2009publications/CEC-500-2009-015/CEC-500-2009-015-D.pdf.

  • Keeley, J. E., 2017: Why were the wine country fires so destructive? The Conversation, U.S. Geological Survey, http://pubs.er.usgs.gov/publication/70192771.

  • Keeley, J. E., and C. Fotheringham, 2003: Impact of past, present, and future fire regimes on North American Mediterranean shrublands. Ecol. Stud., 160, 218262, https://doi.org/10.1007/0-387-21710-X_8.

    • Search Google Scholar
    • Export Citation
  • Keeley, J. E., and A. D. Syphard, 2016: Climate change and future fire regimes: Examples from California. Geosciences, 6, 37, https://doi.org/10.3390/geosciences6030037.

    • Search Google Scholar
    • Export Citation
  • Keeley, J. E., C. J. Fotheringham, and M. A. Moritz, 2004: Lessons from the 2003 wildfires in southern California. J. For., 102, 2631.

    • Search Google Scholar
    • Export Citation
  • Littell, J. S., D. McKenzie, D. L. Peterson, and A. L. Westerling, 2009: Climate and wildfire area burned in western U.S. ecoprovinces, 1916–2003. Ecol. Appl., 19, 10031021, https://doi.org/10.1890/07-1183.1.

    • Search Google Scholar
    • Export Citation
  • Monteverdi, J. P., 1973: The Santa Ana weather type and extreme fire hazard in the Oakland–Berkeley Hills. Weatherwise, 26 (3), 118121, https://doi.org/10.1080/00431672.1973.9931644.

    • Search Google Scholar
    • Export Citation
  • Monteverdi, J. P., and B. L. Wood, 1973: The December 1972 freeze and its effects on the eucalyptus forest of the Oakland-Berkeley Hills. Weatherwise, 26 (4), 160167, https://doi.org/10.1080/00431672.1973.9931653.

    • Search Google Scholar
    • Export Citation
  • Pagni, P., 1993: Causes of the 20 October 1991 Oakland Hills conflagration. Fire Saf. J., 21, 331339, https://doi.org/10.1016/0379-7112(93)90020-Q.

    • Search Google Scholar
    • Export Citation
  • Polade, S. D., A. Gershunov, D. R. Cayan, M. D. Dettinger, and D. W. Pierce, 2017: Precipitation in a warming world: Assessing projected hydro-climate of California and other Mediterranean climate regions. Nature Sci. Rep., 7, 10783, https://doi.org/10.1038/s41598-017-11285-y.

    • Search Google Scholar
    • Export Citation
  • Smith, C., B. Hatchett, and M. Kaplan, 2018: A surface observation based climatology of Diablo-like winds in California’s Wine Country and western Sierra Nevada. Fire, 1, 25, https://doi.org/10.3390/fire1020025.

    • Search Google Scholar
    • Export Citation
  • Sullivan, M., 1997: Firestorm!: The Story of the 1991 East Bay Fire in Berkeley. City of Berkeley, 180 pp.

  • Swain, D. L., B. Langenbrunner, J. D. Neelin, and A. Hall, 2018: Increasing precipitation volatility in twenty-first-century California. Nat. Climate Change, 8, 427433, https://doi.org/10.1038/s41558-018-0140-y.

    • Search Google Scholar
    • Export Citation
  • U.S. Fire Administration, 1991: The East Bay hills fire. USFA Tech. Rep. USFA-TR-060, 120 pp., www.usfa.fema.gov/downloads/pdf/publications/tr-060.pdf.

  • Westerling, A. L., T. J. Brown, A. Gershunov, D. R. Cayan, and M. D. Dettinger, 2003: Climate and wildfire in the western United States. Bull. Amer. Meteor. Soc., 84, 595604, https://doi.org/10.1175/BAMS-84-5-595.

    • Search Google Scholar
    • Export Citation
  • Westerling, A. L., D. R. Cayan, T. J. Brown, B. L. Hall, and L. G. Riddle, 2004: Climate, Santa Ana winds and autumn wildfires in southern California. Eos, Trans. Amer. Geophys. Union, 85, 289296, https://doi.org/10.1029/2004EO310001.

    • Search Google Scholar
    • Export Citation
1

Most destructive in terms of buildings destroyed, second most destructive in terms of deaths.

2

The AQI is based on PM2.5 observations, with PM2.5 being the concentration of small particles of a size of 2.5 µm and smaller.

3

The Remote Atmospheric Weather System (RAWS) data were acquired from the DRI online archive (https://raws.dri.edu/).

4

Observed at the KCAGEYSE11 station available from http://weatherunderground.com.

Save