1. Introduction
On 8 August 2023, a wind-driven wildfire pushed across the city of Lahaina, located in West Maui, Hawaii, resulting in at least 100 deaths and an estimated economic loss of 4–6 billion dollars (Moody’s 2023). During the same period, other wind-driven wildfires spread across the Kula region of central Maui and the western Kohala area on the northwest corner of the island of Hawaii (Fig. 1). The Lahaina wildfire death toll was the greatest wildfire fatality event in the United States in over a century, only exceeded by the 1918 Coquet Fire (Minnesota, estimated 453 deaths) and the 1871 Peshtigo Fire (Wisconsin, estimated 1200–1500 deaths). Considering the serious consequences of the 2023 Maui wildfire, it is important to understand the meteorological conditions associated with the event and to evaluate its predictability. These are the goals of this paper.
Wildfire areas (red fill) on the Hawaiian Islands on 7–9 Aug 2023 based on satellite observations. Map produced by the NASA Fire Information for Resource Management System (FIRMS; https://firms.modaps.eosdis.nasa.gov/map/#d:today;@0.0, 0.0, 3.0z). The locations of the Lihue and Hilo radiosonde stations are shown with red stars.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
Figure 2 presents topographic, annual precipitation, and land surface maps for Maui. The eastern portion of Maui is dominated by the Haleakala volcano, whose peak reaches 3055 m (10 023 ft), while western Maui includes a single, dormant volcanic barrier with a crest of 1764 m (5788 ft). With persistent northeasterly trade winds during much of the year, strong upslope flow and enhanced precipitation occur on the northeast side of the huge Haleakala massif and over the higher terrain of the West Maui Mountains (Fig. 2b). In contrast, distinct rain shadows exist downstream of the barriers, such as over the northwest coast of Maui surrounding the town of Lahaina, where the average annual precipitation is 277 mm (10.9 in.). Dry conditions also occur west of Haleakala over the uplands of central Maui. The leeside dry areas are dominated by grassland and shrubland (Fig. 2c), which become highly flammable when they are cured during the warm/dry summers and early autumns of the region.
Maui terrain (m) and (a) locations, (b) annual precipitation, and (c) land cover. Images from Giambelluca et al. (2013) in (b) and Trauernicht et al. (2015) in (c).
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
The wildfire danger in Hawaii has increased greatly during recent decades as flammable nonnative grasses have invaded abandoned agricultural lands on the lee sides of all the islands (Pickett and Grossman 2014; Trauernicht et al. 2015). Currently, about 0.48% of the Hawaiian land area burns each year, which is greater than either the United States as a whole (0.30%) or the relatively fire-prone western United States (0.46%) (Trauernicht et al. 2015; Trauernicht and Lucas 2016). The overwhelming majority of Hawaiian wildfire ignitions are human caused, with the greatest area burned in August during the typically dry summer season (Trauernicht and Lucas 2016). The fuel loads of Hawaiian rangelands are often very high, sometimes exceeding values over western North America by as much as an order of magnitude (Zhu et al. 2021).
Maui has a history of wind-driven wildfires over the dry, leeward portions of the island. Reports back into the nineteenth century describe powerful, destructive winds near and east of Lahaina called the Kaua’ula Wind.1 Such winds occurred infrequently, would begin over the western foothills of the West Maui Mountains in the vicinity of the Kaua’ula Valley, and then extended westward with great ferocity and roaring sound, resulting in tree damage and toppled houses. Kaua’ula Winds often lasted several days, and since they showed no signs in the clouds, they were sometimes called Obake or “ghost” winds. Recently, on 24 August 2018, several wind-driven wildfires were reported on the western side of Maui as Hurricane Lane passed to the south (Nugent et al. 2020). While Hurricane Lane produced heavy precipitation on the eastern side of Maui, strong downslope winds over the western slopes of the West Maui Mountains contributed to an 896-ha fire near Lahaina, which destroyed 21 structures and 30 vehicles, and a 119-ha fire near Kaanapali to the north of Lahaina. The year 2019 was a particularly wildfire-prone year on Maui, with 10 117 ha burned. For example, on 22 October 2019, strong winds led to wildfires spreading across 406 ha between the Kapalua airport near the northwest Maui coast and the slopes of the West Maui Mountains. Although the most significant Maui wildfires occur during summer or early autumn, some have taken place under strong winds during the winter. For example, on 26 December 2020, strong downslope winds led to a wildfire that charred more than 304 ha of dry brush over leeward West Maui near Olowalu (roughly 10 km southeast of Lahaina), while strong trade winds on 22 January 2001 resulted in a fire covering 324 ha from Olowalu to Launiupoko, located south of Lahaina.
The goal of this paper is to describe the meteorological conditions preceding and accompanying the August 2023 Maui wildfire event, as well as to examine the predictability of these conditions. Major questions include the following:
-
What was the distribution of strong, damaging winds in both space and time?
-
What were the origins of the strong winds? What mesoscale effects were crucial?
-
How unusual were the large-scale conditions associated with the Maui wildfires? What aspects of the large-scale flow were important for this event? What role was played by Hurricane Dora, which passed to the south of Hawaii?
-
Were the precipitation and temperatures during the prior months unusual?
-
How predictable were the meteorological conditions associated with this wildfire?
2. The low-level winds associated with the event
Strong winds played a central role in the West Maui wildfire of 7–8 August 2023. Specifically, powerful downslope easterly winds descended into western Maui and rapidly pushed the wildfire westward. Strong winds provide oxygen to wildfires and advect embers and superheated gases forward of the flame front. The strong winds caused substantial damage to the local power infrastructure, with arcing power lines igniting grass and other vegetation. In addition, the strong, dry downslope winds, accompanied by low relative humidity (as low as 30%),2 promoted rapid drying of the extensive grass and range vegetation found upwind (east) of populated areas.
A map of the strongest observed winds during the 7–8 August event is shown in Fig. 3. Delineation of the distribution of strongest winds was made difficult by the lack of weather observations over West Maui and the loss of power over large portions of the island during the event. The strongest available winds from instrumentation were both 30 m s−1 (58 kt; 1 kt ≈ 0.51 m s−1): downstream of the West Maui Mountains (south of Lahaina) and in the lee of the Haleakala volcano. In addition, northerly winds accelerated at low levels in the gap between the West Maui Mountains and Haleakala, with wind speed at the southern portion of the gap reaching 28 m s−1 (54 kt) at Maalaea Bay. As noted later, both damage reports near Lahaina and model simulations strongly suggest that the winds were considerably stronger over portions of West Maui, with gusts to 31–41 m s−1 (60–80 kt).
Observed maximum gusts (kt) on 7–8 Aug 2023. The Pua Nui Way Tempest site is indicated by a red circle.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
A qualitative estimate of maximum winds can be derived from wind-related damage reports by the Hawaiian Emergency Management Agency (HEMA), the County of Maui, and the Hawaii Electric Company. Several dozen power poles were toppled, some of which were wooden poles that were snapped off midsection. Numerous trees were downed, and roof sections were blown off. A January 2024 after-action report of the Maui Police Department3 documented catastrophic wind damage before and after wildfire initiation. Shortly after midnight Hawaii standard time (HST; 8 August), fires and tree downing were noted near Olinda in central Maui. By 0500 HST, strong winds were noted east of Lahaina, with roofs blown off buildings, numerous fractured utility poles, and large trees downed over roadways. During the subsequent morning hours, there were many reports of severe damage, including downed trees, major roof damage, and failed wall structures. Such damage is consistent with winds of at least 31–41 m s−1 (60–80 kt) (FEMA 2020,4 NOAA5).
Determination of the exact timing of the strongest winds in the Lahaina area was made difficult by the lack of operational wind observations over western Maui and the extensive power outages that occurred on 8 August. Several nongovernmental weather sites observed large wind accelerations over western Maui late on the day of 7 August and into the early morning of 8 August before the power failed. For example, surface observations were available from the Tempest network site at Pua Nui Way (Fig. 4), located about 6.8 km southeast of downtown Lahaina (location of a white circle labeled Tempest in Fig. 2a and the red circle in Fig. 3). The wind speed at this location was relatively low until approximately 0000 UTC 8 August (1400 HST 7 August), after which it accelerated, reaching 30 m s−1(58 kt) around midnight HST. The power at that site failed at approximately 0230 HST 8 August, preventing observations during the period of maximum winds and fire spread on that day. The wind direction at this site was initially northerly, turned southerly, and then shifted to north-northeasterly during the period of greatest acceleration late on 7 August. Sea level pressure initially possessed typical diurnal/semidiurnal variability and evinced high-frequency variability during the last portion of the record when the wind was accelerating. Importantly, the wind acceleration period was associated with a pressure decline, with a decrease of 7 hPa during the final hours before power failure. As described later, such pressure falls were associated with the hydraulic jump connected with a high-amplitude mountain wave. Finally, the period of strong winds was associated with a shift to lower relative humidity, consistent with the downslope flow.
Wind speed (kt), wind direction (°), sea level pressure (hPa), and relative humidity (%) at the Tempest weather site at Pua Nui Way in south Lahaina. Time in UTC (date on top, hour on the bottom).
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
In addition to surface wind observations, the power quality sensors from Whisker Laboratories,6 found in thousands of homes throughout the United States, provided insights regarding when winds became strong enough to damage the power infrastructure. Around 2247 HST 7 August (0847 UTC 8 August), power sensors on central Maui, west of the Haleakala volcano, indicated numerous faults. At approximately 1215 HST 8 August, a brush fire was reported in the nearby upcountry Maui that burned over 1000 acres near the towns of Olinda and Kula. Subsequently, around 0500 HST 8 August (1500 UTC), extensive faults were noted from sensors near Lahaina, followed by a regional loss of power roughly 90 min later. These power faults implied that damaging winds started to push into central Maui before midnight HST and then extended into western Maui around 0500–0600 HST. This is consistent with a report (and video) of downed power lines that initiated a fire over eastern Lahaina around 0637 HST 8 August.
After the initial reports of downed power lines and a grass fire over eastern Lahaina near Lahainaluna Road around 0637 HST 8 August, an attempt was made to suppress the fire, which was reported to be controlled by approximately noon HST. Unfortunately, by 1530 HST, the wildfire over the grasslands east of Lahaina appeared to reignite and expanded rapidly westward under increasingly gusty conditions; by 1630 HST, the fire had begun to move into Lahaina.
The explosive fire growth is suggested by imagery from the National Weather Service radar on Molokai. The radar beam from the lowest (0.5°) elevation angle is shown for several times on 8 August in Fig. 5. The center point of the radar beam at this elevation angle was approximately 1200 m above Lahaina. At 0700 HST 8 August, the radar delineated the higher terrain of the West Maui Mountains. By 1600 HST, the radar had picked up returns from a smoke plume over western Maui near Lahaina, with the smoke extending over the channel to Lanai. At 1700 HST, the plume had intensified and extended farther to the west, with the smoke spreading farther west by 1800 HST. An important implication of this radar sequence is that there was easterly flow from Lahaina westward to Lanai between 1600 and 1800 HST at approximately 1100–1200 m MSL (the height of the radar beam over Lahaina). Tracking the forward (western) edge of the smoke suggests a westward speed of approximately 6 m s−1(11 kt) at that level.
Reflectivity from the lowest elevation angle (0.5°) from the National Weather Service radar on Molokai on 8 Aug 2023. The radar location is noted by a white circle.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
3. The synoptic configuration associated with the Maui wildfire
The low-level synoptic situation during the 8 August event was characterized by a broad area of high pressure to the north of Hawaii and a far smaller area of low pressure associated with Hurricane Dora to the south. Figure 6 shows sea level pressure, 10-m winds, 850-hPa winds, and 850-hPa temperatures at 1200 UTC 8 August 2023 based on the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) analysis. In addition, standardized pressure anomalies from climatology were computed using climatologies derived from the Climate Forecast System Reanalysis (CFSR) 6-hourly product obtained on a 0.5° grid for the period January 1979–December 2010 (Saha et al. 2010), with departures from climatology computed from the NCEP GFS analyses obtained on a 0.5° grid.
(a) Sea level isobars (hPa) and standardized pressure anomalies as well as 10-m winds, (b) 850-hPa heights and standardized temperature anomalies with 850-hPa winds, and (c) 925-hPa heights, wind speed standardized anomalies, and winds. All fields are valid at 1200 UTC 8 Aug 2023. A yellow star shows the location of Maui.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
Sea level pressure was unusually high north of Hawaii, exceeding three standard deviations above normal, while the low pressure anomaly associated with Dora was far south of Hawaii and much smaller in scale (Fig. 6a). Heights at 850 hPa, near the crest level of the West Maui Mountains, indicated a strong high to the north, a lobe of higher heights projecting toward Hawaii, and a very small low to the south with the hurricane (Fig. 6b). A highly unusual warm 850-hPa temperature anomaly (exceeding six standardized deviations from climatology) was centered northeast of Maui. With enhanced high pressure to the north, the low-level north–south pressure/height gradient was increased as were the lower-tropospheric easterly trade winds. The greatly strengthened easterly winds at low levels are illustrated in Fig. 6c, which shows an unusually strong easterly flow at 925 hPa, with the easterly wind anomalies near Maui reaching six standard deviations above normal.
The unusual nature of the high pressure/ridging to the north of Maui during the event can be evaluated by comparing heights and pressures at 30°N, 155°W, approximately 1000 km north of Maui, with the daily-mean historical heights using the NOAA/NCEP reanalysis dataset, which is available on the ESRL website.7 As shown in Fig. 7, the NOAA reanalysis dataset revealed that the daily 850-hPa heights at that location on 8 August 2023 (1652 m) were the highest observed in August for the entire record (1948–2023). 9 August 2023 was in second place. At 925 hPa (not shown), heights on 9 August were in the first place (924 m), with 8 August in the third place for August. Finally, sea level pressure at that point on 8 August was in the second place for August (1028.8 hPa, not shown). In summary, the sea level pressures and lower-tropospheric geopotential heights north of Hawaii were at or near record-breaking levels for August during the wildfire event. One should note that higher pressures and geopotential heights have been observed to the north of Hawaii, with correspondingly stronger easterly winds, during cooler/wetter periods of the year when wildfire danger is less.
Daily 850-hPa heights (m) during August only at a point 1000 km north of Maui. The red arrow indicates the height on 8 Aug.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
Satellite imagery reflected the large anticyclone to the north of Hawaii and the suppression of clouds by the associated subsidence. GOES-18 visible (channel 2) imagery at 1900 UTC 7 and 8 August 2023 are shown in Fig. 8. On 7 August, a large area of suppressed clouds was associated with the high pressure, with an arc of enhanced clouds on its western edge. A cloud mass associated with Hurricane Dora is noted to the south, with little evidence of hurricane-related distortion of the cloud field near the latitude of Maui. One day later, the suppressed cloud area had pushed past Hawaii, and Hurricane Dora, 1300 km to the south, had weakened and moved past Maui.
GOES-18 visible satellite imagery (channel 2) at (left) 1900 UTC 7 Aug and (right) 1900 UTC 8 Aug. The Hawaiian Islands and western North America are highlighted in yellow.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
There are only two radiosonde sites on the Hawaiian Islands: Lihue (PHLI) on Kauai and Hilo (PHTO) on the Big Island, with western Maui being substantially closer to Hilo (locations shown in Fig. 1). In addition, radiosonde drift under the easterly trade wind flow would bring Hilo radiosondes even closer to Maui during their ascent. Figure 9 shows the lower-tropospheric soundings at these locations at 0000 and 1200 UTC 8 August and 0000 UTC 9 August, which encompasses the period immediately before and during the Lahaina wildfire event. For reference, the crest of the West Maui Mountains is at roughly 1500 m (∼850 hPa).
Radiosonde data at (top) Lihue and (bottom) Hilo, Hawaii, for 0000 UTC 8 Aug–0000 UTC 9 Aug 2023. Temperatures are in degrees Celsius and winds are in knots.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
At Lihue, the base of a stable layer (an inversion) was found near 850 hPa at the initial time and descended by approximately 50 hPa over the subsequent 24 h. The Lihue soundings indicate that easterly winds increased to approximately 26 m s−1 (50 kt) near 850 hPa during the period shown. In contrast, the Hilo sounding started (0000 UTC 8 August) with a stable-layer base near 850 hPa (near the crest level of the West Maui Mountains) that subsequently ascended to roughly 750 hPa (∼2700 m), which would increase the potential for downslope winds on the western slopes of the West Maui Mountains. The easterly winds near 850 hPa at Hilo were approximately 8–10 m s−1 (15–20 kt) weaker than at Lihue.
The Hilo sounding indicated a critical level (reversed wind direction from easterly to westerly) at approximately 350 hPa (∼8000 m) at 0000 UTC 8 August, with the critical level absent in the later soundings. In contrast, at Lihue, a critical level was initially observed near 500 hPa (∼5900 m) and rose to approximately 400 hPa (∼7700 m) in the subsequent two soundings. The subtropical inversion during the event was generally 400–800 m lower than climatology; the inversion base during August is usually located around 2200 m (∼790 hPa) at Lihue and 2400 m (∼770 hPa) for Hilo (Cao et al. 2007). The inversion’s displacement from climatology during the August 2023 event is consistent with stronger-than-normal anticyclogenesis over and to the north of Maui and the warmer-than-normal 850-hPa temperature anomalies noted above.
Strong winds normal to a terrain barrier and a stable level above the crest level are favorable for the development of high-amplitude mountain waves and strong downslope winds to the lee of the barrier (e.g., Klemp and Lilly 1975; Durran 1990). As shown above, such conditions were evident in the flow approaching the West Maui Mountains on 7–8 August 2023. In addition, similarly favorable conditions occurred around the Kohala Range on the northwest side of the island of Hawaii. Consistent with such conditions, a high-resolution Weather Research and Forecasting (WRF) Model simulation produced a downslope windstorm on the lee side of the West Maui Mountains during the August wildfire event (described in the next section) and on the western side of the Kohala Range. As discussed by Yang et al. (2005), Carlis et al. (2010), and Li and Chen (2017), an inversion base at or below the crest level is unfavorable for leeside downslope winds, which may help explain why downslope winds were attenuated over Kauai and Oahu. As shown by the Lihue sounding, the stable layer was considerably lower over the northern islands, descending to or below the crest of their terrain, which was lower on those islands. Hsiao et al. (2021) noted the role of strong trades in enhancing warming and drying on the lee slopes of the Hawaiian terrain.
To further explore whether the easterly flow approaching the Hawaiian Islands on 8 August was unusual, the climatology of the 850-hPa wind speed at the two radiosonde sites was acquired from the Storm Prediction Center’s Sounding Climatology website and is shown in Fig. 10. The strongest observed wind at each radiosonde site during the event is shown by the red dot in the figure. For Lihue, the 850-hPa wind speed during the event (28 m s−1, 54 kt) was truly unusual, far stronger than any such wind during the summer months. In contrast, although the 850-hPa wind speed at Hilo (12 m s−1, 24 kt) was stronger than normal, it was still within the envelope of the strongest historical winds during the summer. At Hilo, the strongest easterly winds were at higher elevations, reaching 22 m s−1 (43 kt) at 796 hPa (2134 m).
Maximum daily 850-hPa winds at Lihue (PHLI) and Hilo (PHTO) radiosonde sites, with the maximum value during the 7–8 Aug 2023 event shown by a red dot. Data and plots were acquired from the NOAA/NWS Storm Prediction Center.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
4. High-resolution simulations
To understand the origin and predictability of the strong winds associated with the 8 August Maui wildfire, high-resolution WRF Model simulations were completed. WRF version 4.1.3 was applied using Thompson microphysics, the YSU boundary layer scheme, and the Grell–Freitas cumulus parameterization.8 The domain structure is shown in Fig. 11. The simulation shown below was initialized at 0000 UTC 8 August and run for 36 h. An outer 12-km grid was initialized using the 0000 UTC 8 August NOAA/NWS GFS analysis with boundary conditions from GFS forecasts, which are available on a quarter-degree grid. Within the 12-km grid, there are nests of 4 km, 4/3 km, and 444 m, with the latter domain covering Maui.
WRF grid structure used for the simulations described in this work.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
The potential for surface wind gusts was estimated from the strongest predicted winds within the lowest 250 m above the surface, which is shown for the highest-resolution (444-m) domain (Fig. 12). The 6-h forecast for 0600 UTC 8 August (2000 HST 7 August) indicated strong winds (reaching 60 kt, 31 m s−1) on the lee (western) slopes of the West Maui Mountains east of Lahaina, with moderately strong gusts (45–50 kt, 23–26 m s−1), extending over the coast south of Lahaina. Strong winds were also found over the upland area of central Maui downstream/west of Haleakala. Six hours later (1200 UTC, 0200 HST 8 August), winds had greatly increased along the western slopes of the West Maui Mountains as well as central Maui, the latter associated with the observed power outages and fire starts in that area near that time. By 0000 UTC 9 August (1400 HST 8 August), powerful gusts reaching 36 m s−1 (70 kt) were predicted to extend into Lahaina. Finally, at 0600 UTC 9 August (2000 HST 8 August), the forecast winds had weakened and pulled back from the coastal zone near Lahaina.
Estimated maximum surface wind gusts (color shading and wind barbs; 1 kt) from the 444-m WRF simulation initialized at 0000 UTC 8 Aug 2023 for 0600, 1200, and 1800 UTC 8 Aug and 0000 and 0600 UTC 9 Aug 2023. The location of Lahaina is indicated by a star.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
A close-up view of the forecast surface meteorological fields over West Maui at 0000 UTC 9 August (1400 HST 8 August), near the time that the wildfire was raging in Lahaina, is shown in Fig. 13. The estimated surface wind gusts at this time were as high as 70 kt (36 m s−1) around Lahaina (left panel), with the surface winds reversing direction offshore. This wind pattern is consistent with the predicted sea level pressure distribution (middle panel), including a classic mountain-wave pressure couplet, with high pressure on the windward side and a well-defined trough along the lower slopes of the West Maui Mountains. Winds accelerated down the strong pressure gradient on the lee slopes, weakening and reversing off the coast where the pressure gradient reversed. Finally, the surface relative humidity forecast shows near saturation over the high terrain of the West Maui Mountains, plunging to under 30% near Lahaina on the central West Maui coast.
(a) Near-surface wind gusts and wind direction (kt), (b) sea level pressure, 925-hPa temperature, and 10-m winds (color shading; °C; hPa), and (c) surface relative humidity (%) from a 24-h WRF forecast valid at 0000 UTC 9 Aug 2023 (444 m grid spacing). The location of the cross section in Fig. 14 is also shown.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
To examine the origin of the strong winds on West Maui, a series of west–east vertical cross sections from near Lahaina and across the West Maui Mountains using the 444-m WRF simulations initialized at 0000 UTC 8 August are shown in Fig. 14 (the position of the cross section is shown in Fig. 13). These cross sections show the total horizontal wind speed (color shading), winds within the plane of the cross section (arrows), and potential temperature (K).
Vertical cross sections from near Lahaina eastward across the West Maui Mountains at 0600, 1200, and 1800 UTC 8 Aug and 0000 and 0600 UTC 9 Aug 2023. Horizontal wind speed (kt; color shading), winds within the plane of the cross section, and the potential temperature (solid lines; K) are shown. The location of Lahaina is indicated by the gray vertical line. The cross-sectional location is shown in Fig. 13.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
At 0600 UTC 8 August (2000 HST 7 August), a mountain wave had developed over and to the west of the crest of the West Maui Mountains, with sustained easterly winds reaching 50 kt (26 m s−1) over the upper western slopes. By 1200 UTC (0200 HST 8 August), the wave amplitude had increased and the low-level winds on the upper slopes had strengthened and extended westward. A hydraulic jump is apparent on the midslope. These strong easterly winds and the hydraulic jump continued to descend the terrain through 1800 UTC (0800 HST 8 August) and pushed to lower elevations (and Lahaina) by 0000 UTC 9 August (1400 HST 8 August), followed by a weakening and retreat of the strong winds 6 h later. In these cross sections, a stable layer is noted just above the crest level of the West Maui Mountains.
Time–height cross sections of predicted meteorological conditions upstream of the West Maui Mountains are shown in Fig. 15 (the location indicated as upstream in Fig. 13a). The approximate crest-level pressure (∼850 hPa) is indicated by the black line. The approaching wind speed near the crest level increased from approximately 30 to 50 kt (15–26 m s−1) between 0000 and 1500 UTC 8 August (Fig. 15a). The relative humidity, potential temperature, and horizontal winds are shown in Fig. 15b, with major changes during the 12-h period before strong winds descended into Lahaina. Early in the period, dry air and a stable layer were found well below the terrain crest level, a situation unfavorable for a downslope wind response on the lee (western) side of the West Maui Mountains (Yang et al. 2005). During the subsequent 12 h, the stable layer ascended to near or just above the crest level; such changes, accompanied by stronger winds, created a favorable environment for the development of downslope windstorms on the western slopes of the regional terrain.
Time–height cross sections of (a) wind speed (kt) and (b) relative humidity (%). Potential temperature (K) and wind speed (kt) for a location upstream of the Maui Mountains (location shown in Fig. 13).
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
An important question is whether the above simulations are realistic. Figure 16 shows comparisons of 4/3-km model forecasts with observations at the two closest NOAA/NWS ASOS sites to Lahaina (Lanai Airport, 29.5 km to the southwest, and Kahului Airport, 25 km to the east). The evolution of the model predictions closely follows observations. At Lanai Airport, northeasterly winds increased rapidly after 1800 UTC 8 August, accompanied by a large decline in relative humidity and a rise in temperature. At Kahului, both observations and model prediction indicate a modest strengthening of northeasterly winds, declining relative humidity, and rising temperatures starting around 1800 UTC 8 August.
Hourly observations (blue) and model forecasts (red) of sustained winds (speed and direction), temperature, and relative humidity at Lanai Airport and Kahului Airport on 8–9 Aug 2023.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
Even with a scarcity of surface observations over West Maui, there is strong observational evidence for the existence of a high-amplitude mountain-wave response, with a strong downslope windstorm, an intense pressure trough, and a hydraulic jump structure. For example, a photo taken near sunset on 8 August from Wailea, about 32 km southeast of Lahaina, shows upslope clouds on the western slopes and crest of the West Maui Mountains and smoke rising in a hydraulic jump feature near the West Maui coast and just offshore (Fig. 17a). Pressure observations available near Lahaina before the power failures (such as in Fig. 4 for the Pua Nui Way Tempest site) document a sharp decline in pressure as wind speed increased. Such pressure/wind evolution is consistent with the model forecast of a pressure trough leading strong downslope winds down the western slopes of the West Maui Mountains (Fig. 13b). Another image taken a few hours earlier from Kaanapali, approximately 3.5 km to the north, documents strong easterly winds (with whitecaps over the water) extending offshore north of Lahaina (Fig. 17b). The shallow nature of the smoke over land and the upward jump in the smoke near the coast is also suggested by this image.
(a) The view looking northwest toward Lahaina from Wailea at 1856 HST 8 Aug 2023. Image taken by Rob Markwardt. Upslope clouds over and east of the crest of the West Maui Mountains are evident, as is a hydraulic jump feature in smoke to the west. (b) View looking south from Kaanapali toward Lahaina around 1500 HST. Image courtesy of Christopher and made available by HawaiiNewsNow.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
Another semiquantitative evaluation approach is to compare model winds with the westerly movement of the smoke plume produced by the wildfire. As described above (Fig. 5), the Molokai radar at around 1700 HST (0300 UTC 9 August) documented smoke moving westward at approximately 11 kt (6 m s−1) from Lahaina to Lanai at roughly 1100–1200 m MSL (∼890 hPa) over Lahaina and lower to the west. This westward speed is consistent with the simulated winds along a similar vertical cross section (Fig. 18).
An east–west vertical cross section including Lahaina and the eastern portion of Lanai from a WRF forecast valid 0300 UTC 9 Aug 2023. This cross section from a WRF simulation was initialized at 0000 UTC 8 Aug with a grid spacing of 4/3 km. The red line indicates the height of the lowest radar beam (0.5° elevation angle) of the Molokai radar.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
Another indirect way to verify the model wind field is to compare simulated surface winds to the reflectance of the water shown by high-resolution NASA Terra MODIS imagery around noon on 8 August (Fig. 19). Over water, high wind speeds (associated with surface water waves) correspond to dark colors, while lighter/whiter shades indicate weak winds and a flatter ocean surface. As typical during strong northeasterly trade winds, a region of weak winds (light colors) extends downstream (southwest) from the terrain of Maui. However, strong winds were evident southwest of the high terrain of western Maui from southern Lahaina to southward down the coast as well as south of the large north–south low-level gap between Haleakala and the West Maui Mountains. These strong winds were associated with a downslope flow extending westward from the West Maui Mountains and a gap flow on the eastern side of that terrain barrier. This satellite image can be compared to a 21-h WRF forecast (4/3 km domain) of sustained winds for 2100 UTC (1100 HST) 8 August 2023 (Fig. 19, right panel). The predicted wind pattern closely corresponds to the water reflectance, with strong model winds associated with darker regions, and weak simulated winds collocated with the whiter areas on the satellite imagery.
Visible imagery from the MODIS Terra satellite midday on 8 Aug 2023 and WRF 21-h forecast of sustained 10-m winds at 2100 UTC (1100 HST). The location of Lahaina is shown by a star.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
5. The predictability of the meteorological conditions associated with the Maui wildfire
The synoptic-scale and regional forecasts by global and regional numerical weather prediction (NWP) models skillfully predicted the meteorological conditions associated with the Maui wildfires, forecasting the unusually strong trade winds, the passage of Hurricane Dora to the south, and the extreme winds over western Maui and the Kohala region of northwest Hawaii. To illustrate the short-term model skill of surface winds on Maui, forecasts of maximum gust potential from the experimental NOAA/Global Systems Laboratory (GSL) HRRR-HI model (3-km grid spacing), valid near the time of the initial fire reports around Lahaina (1700 UTC 8 August, 0700 HST 8 August), are shown in Fig. 20 for various HRRR model initialization times (and thus projections) before the wildfire. The 29-h forecast (initialized 1200 UTC 8 August) predicted gusts exceeding 70 kt (36 m s−1) near Lahaina and over upcountry central Maui near Kula (also an area of strong winds and wildfire). The 17-h forecast (initialized 0000 UTC 8 August) was very similar, although with slightly weaker winds. The HRRR winds near Lahaina strengthened modestly for the 11- and 5-h forecasts (initialized 0600 and 1200 UTC 8 August). Thus, for the day before the Lahaina wildfire, the high-resolution Hawaii HRRR forecasts (run hourly but termed experimental) provided highly accurate predictions of very strong downslope winds over western Maui.
Maximum gust potential for the HRRR forecasts valid at 1700 UTC 8 Aug (0700 HST) based on initializations at 1200 UTC 7 Aug, and 0000, 0600, and 1200 UTC 8 Aug 2023. The location of Lahaina is shown by a star.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
Of great importance, model forecasts provided clear warning of the potential for strong winds over western Maui nearly a week in advance. To illustrate the extended predictability of the event, nested WRF simulations (36–12–4–4/3 km) were run for various initialization times (0000 UTC on 3, 5, 7, and 8 August 2023), with the domain initialized with and provided boundary conditions from the operational NOAA GFS forecasts (Fig. 21 shows the domain structure).
Nested grid structure for WRF forecasts.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
Figure 22 shows the forecast surface wind gusts (strongest winds in the lowest 250 m) at 1700 UTC 8 August for the Maui portion of the 4/3-km domain for the various initialization times. The wind distribution was very similar at the various projection times, indicating the great predictability of this event. All projections indicated strong winds over western Maui, with the longest projection time (137 h) associated with the strongest winds over the area in question. Strong winds were also predicted over the upcountry central Maui for all four forecasts.
Estimated gust strength (kt) at 1700 UTC 8 Aug 2023 for WRF forecasts initialized at 0000 UTC 3, 5, 7, and 8 Aug 2023. The location of Lahaina is indicated by a white star.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
The Honolulu National Weather Service put out both a red flag warning and a high wind warning on 7 August, a day before the Maui wildfire event, and these warnings were maintained through the event (Fig. 23).9 Although these messages provided warning of damaging winds and the potential for wildfire, they predicted similar conditions on the leeward sides of all the Hawaiian Islands. Thus, they did not identify the highly localized extreme conditions suggested by the HRRR model over western Maui, central Maui, and along the Kohala region of northwest Hawaii. As noted above, these HRRR model forecasts verified well in these regions, since strong winds and wildfires were observed over northwestern Maui, central Maui, and the northwest Kohala coastal region of the Big Island.
Red flag and high wind warnings issued by the Honolulu office of the National Weather Service at 0926 HST 8 Aug.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
6. The role of Hurricane Dora
During the August Maui wildfire event, Hurricane Dora passed roughly 1300 km south of the island of Hawaii as a category 4 storm (with sustained winds of 115 kt, 59 m s−1). As suggested by satellite imagery (e.g., Fig. 8) and official hurricane track information (Fig. 24), Hurricane Dora passed sufficiently south of Hawaii so that the islands were not directly influenced by the low-level circulation of the hurricane. This interpretation is supported by the sea level pressure, 925-hPa wind, and 850-hPa height/pressure standardized anomalies presented earlier in the paper.
Path and intensity of Hurricane Dora. Image courtesy of the National Weather Service.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
If the direct wind effects of Hurricane Dora on Maui were small because of the storm’s compact size and trajectory far to the south of the island, an important question is whether Hurricane Dora could have indirectly contributed to the strong winds near Lahaina. Specifically, did Dora influence the flow approaching the island through the modification of the synoptic-scale circulation to its north? To address this question, numerical experiments were undertaken in which Hurricane Dora was greatly weakened before its passage south of the Hawaiian Islands. Two 12-km WRF simulations were completed over an extended domain (Fig. 21), with both initialized at 0000 UTC 3 August 2023 when the developing storm was located over the eastern Pacific, just west of Central America. The first simulation (the control) used observed sea surface temperatures, with initialization and boundary conditions coming from the operational GFS global forecasts. The second simulation [attenuated storm (AS)] was identical except that the sea surface temperature in a 2° square area following the storm was cooled to 23.9°C (75°F) to weaken the disturbance over time.
The 138-h forecast of 850-hPa heights and winds valid at 1800 UTC 8 August 2023 for both the control and attenuated hurricane runs are shown in Fig. 25. The control run had the hurricane south of the Island of Hawaii in a location very close to the observed (e.g., compared to the location in Fig. 7). In contrast, cooling sea surface temperatures near the hurricane greatly attenuated its amplitude, leaving only a weak trough displaced to the south at 850 hPa. The 850-hPa heights and winds near Maui were slightly reduced for the weakened storm simulation.
The 850-hPa winds (wind barbs and shading) and heights (white lines) at 1800 UTC 8 Aug 2023 for (a) a run with observed sea surface temperatures and (b) one with cooled temperatures around Hurricane Dora. Only a portion of the full domain is shown (see Fig. 21 for the full domain grid structure).
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
To examine the implications of an attenuated Hurricane Dora on surface winds, a 4/3-km regional domain was nested in the two large-domain 12-km runs (see Fig. 21 for grid structures). Figure 26 shows the forecast gusts at the surface (strongest winds in the lowest 250 m) around Maui for 1200 and 1800 UTC 8 August and 0000 UTC 9 August 2023 for both simulations. Both the control and AS runs predicted strong winds over West Maui and central Maui to the lee of Haleakala. Importantly, there is little consistent difference between the leeside winds over western Maui between the two simulations. For 1200 and 1800 UTC 8 August, the AS simulation had stronger winds near Lahaina and central Maui than the run with the full-strength hurricane. In contrast, the forecast winds were about the same at 0000 UTC 9 August.
Simulated maximum surface wind gusts (kt) for both the control and weak hurricane simulations at 1200 and 1800 UTC 8 Aug and 0000 UTC 9 Aug 2023 for WRF runs initialized at 0000 UTC 3 Aug 2023. The forecasts are from a nested 4/3-km domain simulation. The location of Lahaina is indicated by a star.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
7. Antecedent climatological conditions
As noted in the introduction, the 2023 Maui wildfire was initiated in grass/rangeland vegetation, mainly on abandoned agricultural land. Such vegetation is generally characterized as 1–10-h dead fuels, meaning that it can be susceptible to fire after a few hours of dry conditions (e.g., low relative humidity/high vapor pressure deficit, and strong winds). Even after a wet period, such fuels can rapidly become receptive to fire, lessening their sensitivity to antecedent weather/climatic conditions. Recent research has suggested that wetter-than-normal conditions during the previous growing season (generally winter) make a greater contribution to summer wildfire potential in Hawaii than drying conditions during or immediately preceding the time of fire by increasing the density of fuel available to burn (Trauernicht 2019).
Figure 27 presents the accumulated precipitation at the Lahainaluna 361.1 climate site, located 3.2 km east of Lahaina from 1 October 2022 to 7 August 2023. This precipitation plot reflects the climatologically large seasonality of Hawaiian precipitation, with relatively wet winters followed by dry summers during which seasonal grasses cure and become highly flammable. The 2022/23 water year had a dry autumn, a wet winter and early spring that fostered grass growth, and then a dry summer. As described below, the accumulated precipitation during the prior water year was well above normal.
Accumulated precipitation (in.) from 1 Oct 2022 to 7 Aug 2023 at the Lahainaluna 361.1 climatological site near Lahaina, Maui. Data and plots were made using the NOAA xmACIS2 website.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
To evaluate whether the precipitation preceding the Maui wildfire was unusual, the precipitation amounts for two seasonal periods at the Lahainaluna 361.1 site are shown in Fig. 28 for 2004–23. For the 1 June–7 August period, the average summer precipitation is modest (approximately 0.45 in.), with occasionally much wetter summers (2013, 2020) of 1–2 in. (Fig. 28a). Summer 2023 was slightly wetter than normal at this site (0.75 in.). The water-year (1 October–7 August) precipitation at Lahainaluna 361.1 for 2022/23 (22.23 in.) was well above normal (9.15 in.) and just slightly less than the record year of 2004 (Fig. 28b). Such above-normal precipitation would encourage greater than normal grass growth before the 2023 Maui wildfire.
Warm-season (1 Jun–7 Aug) and water-year (1 Oct–7 Aug) precipitation at the Lahainaluna 361.1 climatological site on Maui.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
Although temperature is not available at Lahainaluna 361.1, it is available at the nearby Puukolii 457.1 climate site, located approximately 5.6 km north-northeast of Lahaina. The average summer temperature for 1 June–7 August 2023 at Puukolii 457.1 (76.3°F) was slightly below normal (77.3°F) (Fig. 29).
Daily temperature range and climatological daily extreme maximum (red shading), minima (blue shading), and average range (brown shading) at the Puukolii 457.1 climatological site.
Citation: Weather and Forecasting 39, 8; 10.1175/WAF-D-23-0210.1
In summary, the climatological conditions preceding the August 2023 Maui wildfire were characterized by near-normal summer conditions for temperature and precipitation but much greater than normal precipitation for the preceding water year.
8. Summary and conclusions
On 8–9 August 2023, a wind-driven wildfire moved across the city of Lahaina, located in West Maui, Hawaii, resulting in at least 100 deaths and an estimated economic loss of 4–6 billion dollars. During the same period, wind-driven wildfires spread across the Kula region of central Maui and the western Kohala area on the northwest corner of the island of Hawaii. This event represents the deadliest wildfire in the United States in over a century and was produced by a downslope windstorm over the western slopes of the West Maui Mountains, with gusts estimated to reach 60–80 kt (31–41 m s−1). It appears that the fires were ignited by wind-damaged electrical infrastructure, with the fires spreading rapidly in highly flammable invasive grasses and other nonnative vegetation.
The synoptic situation during the fire was characterized by a highly anomalous lower-tropospheric anticyclone to the north of Maui and the westward passage of Hurricane Dora approximately 1300 km to the south. Regionally, the wildfire synoptic/mesoscale environment was associated with the stronger than normal easterly flow near the crest level of the West Maui Mountains (at around 850 hPa) and a subtropical inversion layer that was initially below the mountain-crest level, but which ascended to near-crest level as leeside winds accelerated.
High-resolution WRF simulations, with grid spacing down to 444 m, were used to examine the mesoscale structure and evolution associated with the wildfire. Such model runs realistically simulated the observed structure, magnitude, and distribution of the strong winds, which were also accompanied by low relative humidity declining below 30%. Model vertical cross sections revealed that the strong winds were associated with a high-amplitude mountain wave and accompanying downslope windstorm, with a hydraulic jump near the coast.
Model verification for this event is made difficult by the lack of operational surface observing stations over northwest Maui and the extensive loss of power before the wildfire, which took amateur weather stations offline. This paper provided alternative evidence of model fidelity, including radar imagery of smoke, photographic documentation of a hydraulic jump feature, reflectance patterns off the nearby ocean, qualitative damage reports, and limited data from amateur weather stations and power-quality sensors.
The meteorology associated with the Maui wildfires was highly predictable by operational and research numerical models. The experimental NOAA/GSL HRRR-HI model predicted the strong winds at least 29 h ahead and WRF Model simulations, driven by the operational NOAA/NWS GFS modeling system, provided a skillful forecast of West Maui winds 137 h (5.5 days) before the event. The National Weather Service distributed both high wind and red flag warnings the day before the event but did not highlight extreme winds over western Maui.
A major question deals with the role of Hurricane Dora, which passed approximately 1300 km to the south of Maui. Although Dora’s low-level wind circulation was spatially limited and too far south to have directly produced strong easterly winds over Maui, an important question is whether the indirect effects of the hurricane on the synoptic-scale flow could have been important. To address this question, two extended, large-domain WRF runs were made, one with the observed sea surface temperatures and the other with a cooled SST patch following the storm. The disturbance rapidly weakened during the latter simulation, and the impacts of the weakened storm over Maui were evaluated. It was found that the simulation with a highly attenuated hurricane had similar low-level winds over northwest Maui, suggesting that Hurricane Dora had little impact on the 8–9 August event.
Finally, the antecedent climatological conditions were evaluated. It was found that the conditions preceding the August 2023 Maui wildfire were characterized by near-normal summer conditions for temperature and precipitation but much greater than normal precipitation for the preceding water year. Such excess precipitation would have encouraged the growth of flammable nonnative grasses and other vegetation.
Like several recent large wildfires in the western United States, this event was associated with a high-amplitude mountain wave resulting from strong winds and a stable or critical layer just above the crest level and was highly predictable with the current mesoscale modeling technology. For example, downslope windstorm-related wildfires include the 2018 Camp Fire (Northern California; Mass and Ovens 2021), the 2017 Wine Country Fires (Bay Area; Mass and Ovens 2019), the 2017 Thomas Fire (Southern California; Fovell and Gallagher 2018), and the 2021 Marshall Fire (Front Range of Colorado; Fovell et al. 2022; Benjamin et al. 2023; Juliano et al. 2023). Most were also associated with dry grasses and shrubs adjacent to urban areas, with many of the homes poorly constructed to resist wildfire. The predictability of the meteorological conditions and apparent dangers of the associated urban development suggest the potential for substantially reducing loss of life and damage in the future.
The climatological relative humidity during August at Lahaina is 70%.
The simulations were also run with the scale-aware version of the YSU scheme (the Shin–Hong scheme; Shin and Hong 2015). The results were nearly identical to those using the original YSU parameterization.
Red flag warnings over Hawaii require a Keetch–Byram drought index (KDBI) of greater than or equal to 600, a relative humidity of less than or equal to 45% for multiple hours, and surface (10-m) wind speeds greater than or equal to 17 kt for 2 h or more. A high wind warning is called when forecasts are for sustained winds of 35 kt or greater for an hour or more, and/or gusts of 50 kt or higher.
Acknowledgments.
This work was supported by the National Science Foundation through NSF Grants AGS-2042105 and AGS-2344105. We acknowledge the Tempest data made available by WeatherFlow, Inc. (Tony McGee and Marty Bell), the assistance of Kevin Kadama, John Bravender, and Jerome Saucier of the Honolulu office of the National Weather Service, and useful conversations with Professor Steve Businger of the University of Hawaii, Manoa. Robert Marshal of Whisker Labs provided important insights into the effects of wind on the power infrastructure. Eric Lau of the National Weather Service Pacific Region provided information about historical wind events. Two anonymous reviewers and Paul Schlatter, NWS Boulder, provided excellent comments and suggestions that substantially improved the manuscript.
Data availability statement.
All data used in this manuscript are freely available at the following sources or by request from the corresponding author, including the following: WRF Model output, including namelists, is available upon request. GFS grids for model initialization are available from the National Centers for Environmental Information (NCEI) at https://www.ncei.noaa.gov/products/weather-climate-models/global-forecast. Surface observations are available from Tempest.
REFERENCES
Benjamin, S. G., E. P. James, E. J. Szoke, P. T. Schlatter, and J. M. Brown, 2023: The 30 December 2021 Colorado front range windstorm and Marshall Fire: Evolution of surface and 3-D structure, NWP guidance, NWS forecasts and decision support. Wea. Forecasting, 38, 2551–2573, https://doi.org/10.1175/WAF-D-23-0086.1.
Cao, G., T. W. Giambelluca, D. E. Stevens, and T. A. Schroeder, 2007: Inversion variability in the Hawaiian trade wind regime. J. Climate, 20, 1145–1160, https://doi.org/10.1175/JCLI4033.1.
Carlis, D. L., Y.-L. Chen, and V. Morris, 2010: Numerical simulations of island-scale airflow and the Maui vortex during summer trade-wind conditions. Mon. Wea. Rev., 138, 2706–2736, https://doi.org/10.1175/2009MWR3236.1.
Durran, D. R., 1990: Mountain waves and downslope winds. Atmospheric Processes over Complex Terrain. Meteor. Monogr., Vol. 23, Amer. Meteor. Soc., 59–81, https://doi.org/10.1007/978-1-935704-25-6_4.
Fovell, R. G., and A. Gallagher, 2018: Winds and gusts during the Thomas Fire. Fire, 1, 47, https://doi.org/10.3390/fire1030047.
Fovell, R. G., M. J. Brewer, and R. J. Garmong, 2022: The December 2021 Marshall Fire: Predictability and gust forecasts from operational models. Atmosphere, 13, 765, https://doi.org/10.3390/atmos13050765.
Giambelluca, T. W., Q. Chen, A. G. Frazier, J. P. Price, Y. Chen, P. Chu, J. K. Eischeid, and D. M. Delparte, 2013: Online rainfall atlas of Hawai’i. Bull. Amer. Meteor. Soc., 94, 313–316, https://doi.org/10.1175/BAMS-D-11-00228.1.
Hsiao, F., Y.-L. Chen, H. Van Nguyen, D. E. Hitzl, and R. Ballard, 2021: Effects of trade wind strength on airflow and cloudiness over Oahu. Mon. Wea. Rev., 149, 3037–3062, https://doi.org/10.1175/MWR-D-20-0399.1.
Juliano, T. W., and Coauthors, 2023: Toward a better understanding of wildfire behavior in the wildland-urban interface: A case study of the 2021 Marshall Fire. Geophys. Res. Lett., 50, e2022GL101557, https://doi.org/10.1029/2022GL101557.
Klemp, J. B., and D. R. Lilly, 1975: The dynamics of wave-induced downslope winds. J. Atmos. Sci., 32, 320–339, https://doi.org/10.1175/1520-0469(1975)032<0320:TDOWID>2.0.CO;2.
Li, L., and Y.-L. Chen, 2017: Numerical simulations of two trapped mountain lee waves downstream of Oahu. J. Appl. Meteor. Climatol., 56, 1305–1324, https://doi.org/10.1175/JAMC-D-15-0341.1.
Mass, C. F., and D. Ovens, 2019: The Northern California wildfires of 8–9 October 2017: The role of a major downslope wind event. Bull. Amer. Meteor. Soc., 100, 235–256, https://doi.org/10.1175/BAMS-D-18-0037.1.
Mass, C. F., and D. Ovens, 2021: The synoptic and mesoscale evolution accompanying the 2018 camp fire of Northern California. Bull. Amer. Meteor. Soc., 102, E168–E192, https://doi.org/10.1175/BAMS-D-20-0124.1.
Moody’s, 2023: Moody’s RMS estimates USD 4 to 6 billion in economic losses from Hawaii wildfires. https://www.rms.com/newsroom/press-releases/press-detail/2023-08-22/moodys-rms-estimates-usd-4-to-6-billion-in-economic-losses-from-hawaii-wildfires.
Nugent, A. D., R. J. Longman, C. Trauernicht, M. P. Lucas, H. F. Diaz, and T. W. Giambelluca, 2020: Fire and rain: The legacy of Hurricane Lane in Hawai‘i. Bull. Amer. Meteor. Soc., 101, E954–E967, https://doi.org/10.1175/BAMS-D-19-0104.1.
Pickett, E., and I. Grossman, 2014: Western Maui community wildfire plan. Hawaii Wildfire Management Organization, 137 pp., https://dlnr.hawaii.gov/forestry/files/2024/01/Western-Maui-CWPP.pdf.
Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1058, https://doi.org/10.1175/2010BAMS3001.1.
Shin, H. H., and S.-Y. Hong, 2015: Representation of the subgrid-scale turbulent transport in convective boundary layers at gray-zone resolutions. Mon. Wea. Rev., 143, 250–271, https://doi.org/10.1175/MWR-D-14-00116.1.
Trauernicht, C., 2019: Vegetation–rainfall interactions reveal how climate variability and climate change alter spatial patterns of wildland fire probability on Big Island, Hawaii. Sci. Total Environ., 650, 459–469, https://doi.org/10.1016/j.scitotenv.2018.08.347.
Trauernicht, C., and M. P. Lucas, 2016: Wildfire ignition density maps for Hawaii. Forest and Natural Resource Management Series RM-21, University of Hawai‘i Cooperative Extension Service, 10 pp., www.ctahr.hawaii.edu/oc/freepubs/pdf/RM-21.pdf.
Trauernicht, C., E. Pickett, C. P. Giardina, C. M. Litton, S. Cordell, and A. Beavers, 2015: The contemporary scale and context of wildfire in Hawai‘i. Pac. Sci., 69, 427–444, https://doi.org/10.2984/69.4.1.
Yang, Y., Y.-L. Chen, and F. M. Fujioka, 2005: Numerical simulations of the island-induced circulations over the island of Hawaii during HaRP. Mon. Wea. Rev., 133, 3693–3713, https://doi.org/10.1175/MWR3053.1.
Zhu, T. R., C. M. Litton, C. P. Giardina, and C. Trauernicht, 2021: Moisture availability and ecological restoration limit fine fuels and modeled wildfire intensity following non-native ungulate removal in Hawaii. J. Appl. Ecol., 58, 2207–2219, https://doi.org/10.1111/1365-2664.13952.
