Mesoscale model simulations of the meteorological conditions during the
June 2, 2002 Double Trouble State Park wildfire.
Joseph J. Charney and Daniel Keyser
1. Introduction
As a thick morning mist cleared in the early afternoon on 2 June 2002, the New
Jersey Forest Fire Service (NJFFS) lookout at the Cedar Bridge Fire Tower in eastcentral New Jersey reported a smoke plume in the northeastern corner of the Double
Trouble State Park (DT), near the Jake’s Branch of the Toms River (Figs. 1a,b), at 1309
EDT (1709 UTC) (NJFFS 2003). Firefighters from two NJFFS units immediately
responded to the fire, arriving on the scene within six minutes of the initial report, at 1314
EDT (1714 UTC) (Table 1). Eleven minutes later, as 20 ms-1 winds were reported at the
scene of the fire, the firefighters were forced to abandon their initial attack and pull back
to establish defensible positions around the town of Beachwood, NJ, less than 3 km
downwind of the fire. By the sunset that evening, 361 wildland firefighters had been
called in to fight a fire that burned 1,300 acres (Fig. 2), forced the closure of the Garden
State Parkway, damaged or destroyed 36 homes and outbuildings, directly threatened
over 200 homes, forced the evacuation of 500 homes, and caused an estimated $400,000
in property damage (NJFFS 2003).
Prior to the initial attack, at 1415 UTC, weather observations at the Lakewood
Fire Tower indicated modest surface winds of 4 ms-1 , temperatures near 24°C, and a
relative humidity of 62%. Two hours after the initial attack, a spot weather forecast
prepared by the Mount Holly National Weather Service Forecast Office (KMHO)
predicted winds of 8 ms-1 with gusts to 14 ms-1, a temperature of 28°C , and a relative
humidity of 28% at 1500 UTC (Table. 2). The Lakewood Fire Tower reported winds in
excess of 20 ms-1 at 1425 UTC, coinciding with the time when the firefighters were
forced to abandon the initial attack on the fire. Clearly, a profound change occurred in
the atmosphere between the late morning and the early afternoon of 2 June, resulting in a
substantial drop in relative humidity accompanied by an increase in the surface wind
speeds. Since there are no strong local variations in the land surface characteristics (e.g.,
terrain, land use characteristics) in the immediate area of the fire that could entirely
account for the observed surface conditions, it is hypothesized that they developed in
response to a mesoscale weather phenomenon organized on a larger scale than the fire
itself.
The Eastern Area Modeling Consortium (EAMC), in East Lansing, MI, operates a
real-time mesoscale modeling system that predicts the weather conditions throughout the
north-central and northeastern United States on a daily basis (Heilman et al. 2003; Zhong
et al. 2005). At the time of the fire, the EAMC simulations indicate that dry, highmomentum air at the surface was not present throughout the entire northeastern United
States on this day. Instead, the simulations show that dry, windy conditions are most
evident over southeastern PA and southern NJ. Furthermore, since the horizontal
distribution of moisture is not well correlated with the coastline or other surface moisture
sources, the mechanism for the moisture and wind anomalies are hypothesized to derive
from larger than local scale atmospheric conditions aloft.
Kalpan et al. (2006) analyzed of the atmospheric conditions associated with the
DT wildfire, concentrating on the different scales of atmospheric phenomena that
contributed to the development of the surface weather conditions. This paper presents
observations and model results that depict the atmospheric conditions at the surface and
aloft at the time of the fire, drawing connections between atmospheric conditions aloft
and the evolution of the surface conditions associated with the observed fire behavior.
No attempt is made in this paper to simulate the evolution of the fire. Instead, we focus
on the utility of a mesoscale model for predicting and understanding the weather
conditions that contributed to the observed fire behavior during the DT wildfire. Section
2 describes the mesoscale atmospheric model employed by the EAMC to study this
event. Section 3 describes the observed fire behavior and meteorological conditions.
Section 4 presents the mesoscale model simulations and highlights the weather conditions
that contributed to the observed fire behavior. Section 5 consists of discussion and
conclusions.
2. Mesoscale model description
The EAMC mesoscale modeling system is built around the Penn State
University/National Center for Atmospheric Research Mesoscale Model version 5.3
(MM5). The MM5 is a nonhydrostatic, primitive equation, mesoscale model that uses a
terrain-following pressure coordinate system. The EAMC formulation of the model
employs physical subgrid-scale parameterizations to account for radiative transfer
(Mlawer et al. 1997), mixed-phase cloud microphysics (Reisner et al. 1998), cumulus
convection (Kain and Fritsch 1990), boundary layer turbulence (Janjic 1990), and landsurface exchange processes (Chen and Dudhia, 2001). The EAMC runs the MM5 twice
daily initialized at 0000 and 1200 UTC using the National Centers for Environmental
Prediction (NCEP) operational Eta model output. The model produces 48-h simulations
over 36 and 12-km domains, and 24-h simulations for two 4-km domains. The model
outputs are available every hour for display and analysis purposes and the hourly outputs
are also archived for forecast validation. The outermost 36 km domain covers the entire
continental United States, with a one-way nested 12 km domain covering the entire northcentral and northeastern United States. Additional two one-way nested 4 km domains are
generated from the 12 km simulation, one covering the Great Lakes Region and another
covering most of New England (see Fig. 1 of Zhong et al. 2005). The model employs 36
vertical levels, with 15 vertical levels located in the lowest 2000 m to better resolve
planetary boundary layer (PBL) structures that can impact fire behavior.
Although the real-time simulation of the weather associated with the DT fire
reproduced some of the observed surface phenomena, the overall simulation was
insufficient for a detailed analysis, particularly in its simulation of the PBL. For the
purposes of this case study, the model was initialized using the NCEP North American
Regional Reanalysis (Mesinger et al. 2006).
3. Observed Fire Behavior and Meteorological Conditions
Table 1 summarizes the variations in weather conditions, fire behavior, and
firefighting activities reported by the NJFFS on 2 June (NJFFS, 2003). The fire started
from a camp fire that had been left smoldering on the night of 1 June. The fire was small
enough to escape detection at fire observation towers until 1700 UTC, due to poor
visibility and high relative humidity in the morning (Steve Maurer, NJFFS, personal
communication). The Keech-Byram Drought Index (Keech and Byram 1968) at the time
of the fire was 105, which indicates a long-term, climatological drought at the
“moderate” level. Other fuel indices calculated at the time of the fire suggest that the fuel
conditions were at moderate risk for fire. However, NJFFS (2003) reports that frost
damage occurred in the area due to unusual freezing temperatures that occurred some
weeks before the fire. As a result, the drought and fuel indices for the area
underestimated the amount of dead and dry fuel that existed at the time and location of
the wildfire.
Within 25 min. after the initial observation and 11 min. after the initial attack, the
fire was found to be growing too rapidly to be contained. Wind speeds at the Lakewood
Fire Tower were reported to be in excess of 20 ms-1 at 1725 UTC. Radio logs from the
Lakewood Fire Tower indicated that the fire was advancing at >34 m per minute during
the initial attack. The fire crews were forced to retreat and light back fires in an effort to
deprive the fire of fuel and prevent it from sweeping through the town of Beachwood
less than 3 km to the east. A combination of a surface wind shift from westerly to
northwesterly and the actions of the firefighters prevented the fire from entering the town,
but continued efforts were required throughout the day to protect homes, outbuildings,
and other structures to the south of the town.
Observed surface winds and mixing ratios at 1200 UTC and 1800 UTC 2 June
and 0000 UTC 3 June are shown in Fig. 3. The observations indicate that wind speed
increased and mixing ratios decreased in New Jersey between 1200 and 1800 UTC (Figs.
3a,b), which roughly coincides with the time when the fire started to grow out of control.
The surface winds also changed from south-southwesterly to northwesterly between 1200
UTC 2 June and 0000 UTC June 3 (Figs. 3a,b), indicating the passage of a dry cold front
through the area of the fire. Drying is better defined than cooling in the Mid-Atlantic
region in conjunction with this frontal passage. KMHO provided a spot forecast for the
fire location, shown in Table 2. The forecast shows the winds growing stronger through
the afternoon and then decreasing in the early evening, with relative humidity varying
between 30% and 45% during the same time period.
The observed atmospheric conditions aloft at 1200 UTC 2 June and 0000 UTC 3
June are shown in Fig. 4. At 1200 UTC, a pronounced upper level trough over northern
New England is associated with strong northerly flow at 850 hPa (Fig. 4a) and a sharp
temperature gradient along the St. Lawrence River. At 300 hPa (Fig. 4b), a jet streak
north and west of the Great Lakes is clearly represented. By 0000 UTC, the trough has
advanced to the south and east, with the northerly flow and cold air at 850 hPa
progressing southward through NY, northern PA, and northern NJ (Fig. 4c). At 300 hPa,
the jet streak has advanced to the east and south over southern New England.
Geostationary Operational Environmental Satellites (GOES) water vapor imagery for
these times (Figs. 5a,b) clearly delineates a ribbon of dry air sagging southward over the
same location as the jet streak diagnosed in Figs. 4b,d, suggesting that the upper level
trough and jet streak are associated with the arrival of dry air aloft in southern New
England.
Atmospheric soundings at 1200 UTC 2 June and 0000 UTC 3 June from Upton,
NY are shown in Fig. 6. The soundings indicate that a deepening surface-based mixed
layer during the day coincides with pronounced drying at the surface during this 12-h
period. Wind profile observations from New Brunswick, NJ (Fig. 7) and the surface
meteograms from Atlantic City, NJ (Fig. 8a) and McGuire Air Force Base (Fig. 8b),
show that surface wind speeds and gusts also increased during this time period. The
meteograms also show a wind shift and gradual decrease in dew point temperature
between 1300 and 1800 UTC, which are consistent with the passage of a dry front.
The frontal passage diagnosed in the surface observations and the development of
a dry slot in the trough in the upper air analyses could be a critical pieces of evidence for
explaining the surface drying and gustiness that were observed at the time of the outbreak
and spread of the fire. It is hypothesized that coupling between a deepening mixed layer
and a mid-tropospheric dry layer occurs in the post-frontal environment and leads to the
downward mixing of dry air and momentum in the PBL. Since strong winds (Byram
1954; Fahnstock 1965; Brotak and Reifsnyder 1977, and Simard et al. 1987) and low
relative humidity (Lansing 1939; Davis 1969; Simard et al. 1987) are known to be
conducive to large fire development and rapid fire spread, this sequence of
meteorological events could help explain the observed extreme fire behavior in this case.
4. Mesoscale model simulations
While the observed weather conditions provide evidence that dry, windy, and
warm conditions developed in east-central New Jersey on the afternoon of 2 June 2002,
there is not sufficient temporal and spatial detail in the observational record to document
the conditions in the immediate vicinity of the wildfire, nor to draw a causal connection
between the weather conditions and the observed fire behavior. The MM5 model
described in section 2 was employed to generate spatial and temporal detail that
augments the observational record to provide a more complete understanding of the
sequence of events that occurred during the wildfire. All of the model results discussed
below were generated using the 4-km model domain covering the northeastern United
States and were initialized at 1200 UTC 1 June 2002.
Figure 9 shows the surface winds and relative humidity from the MM5
simulation, valid at 1800 UTC on 2 June. Note that the model predicts low relative
humidity in southern NJ and across the mid-Atlantic states at this time. Surface wind
speeds are generally around 10 ms-1 across the northeast, with stronger winds evident in
eastern and southern NJ. Figure 10 indicates that a pool of dry air aloft is in place at
1800UTC, over most of the mid-Atlantic region. This 700 hPa pool of dry air shows the
horizontal extent of the dry feature evident above 750 hPa in the 0000 UTC 3 June
observed sounding at OKX (Fig. 6b). A Skew-T diagram computed from the simulation
at the fire location (39.9˚N, 74.1˚W) valid at 1800 UTC (Fig. 11) shows that the
simulated pool of dry air at the fire location is much deeper and drier than observed at
OKX 6 h later.
Figures 12a,b show northwest-southeast vertical cross sections of wind speed and
relative humidity through the fire location valid at 1700 UTC and 1800 UTC 2 June,
respectively (see Fig. 10 for the cross section location). Note how there is evidence of a
coupling between the dry air at the surface and the deep pool of very dry air aloft. This
coupling indicates that as the mixed layer deepens through the morning and early
afternoon, the very dry air mixes with the relatively moist air near the ground, producing
a sudden and dramatic drying of the surface air. The wind vectors within the cross
section do not show a noticeable influence of this mixing process on the surface wind
speeds. However, a time-height cross section of relative humidity (Fig. 13a) and wind
speed (Fig. 13b) at the fire location shows that both drying and an increase in surface
wind speed accompanies the mixed layer growth in the early afternoon. A time series of
surface relative humidity (Fig. 14a) and surface wind speed (Fig. 14b) at the fire location
support this interpretation, clearly demonstrating the dramatic change in surface
conditions at the fire location between 1300 and 1800 UTC.
Fig. 15a shows simulated maps of the Haines Index (HI) and the Ventilation
Index (VI) for 1800 UTC 2 June. While the (HI) did “light up” over the mid-Atlantic
region at the time of interest, with values of 5 indicating a high potential for a blow-up
fire to occur in the absence of strong surface winds, the HI did not change dramatically
from the morning through the early afternoon (not shown). Thus, the HI failed to capture
the change in surface conditions that are hypothesized to have contributed to the extreme
fire behavior observed during the fire. Figure 15b shows the VI at the same time at the
HI. The VI is defined as the mixed-layer depth multiplied by the mixed-layer average
wind speed (e.g. Hardy et al. 2002). The VI showed locally high values over southern
New Jersey, but this index also failed to effectively capture or highlight the spatial or
temporal changes in meteorological conditions that accompanied the wildfire. Since the
HI and the VI are the two most commonly used operational fire weather indices, it is
clear that new indices and diagnostics that are sensitive to the meteorological conditions
documented in this case could have been useful when preparing to fight this fire.
5. Discussion and Conclusion
The Double Trouble State Park wildfire in east-central New Jersey presented
many challenges to firefighters, and destroyed some $400,000 worth of property. The
tools available to the firefighters and the NWS were helpful in effectively fighting the fire
and preventing any loss of life or catastrophic loss of property. Nevertheless, it is clear
that in this case there were some meteorological conditions occurring at the fire location
that were not anticipated by the available tools. For instance, it is apparently quite
important that fire-weather forecasters take into account the potential for dry air aloft to
be mixed down to the surface within a few hours of sunrise.
The MM5 simulations suggest that there is much to be learned about interactions
between air aloft and the surface weather conditions that are known to influence fire
behavior. The results presented here suggest that interactions between the PBL and deep
pools of dry air aloft can generate localized areas of very dry, high momentum air at the
surface. While the pool of mid-tropospheric dry air is detectable in observations in this
case, it is more straightforward to detect and quantify in mesoscale simulations. These
results are supported by the findings of Mills (2005), Mills (2003), [and others…I still
need to dig them out…--Jay]
There is a clear indication that new products that diagnose specific types of
weather conditions aloft simulated by atmospheric models could provide additional
predictive power and early warnings of atmospheric conditions conducive to rapid fire
growth. There is an opportunity to develop indices and diagnostics that raise the “threat
level” when a pool of dry air is in place, and then can indicate localized areas of extreme
danger where the PBL is predicted to grow such that the dry air aloft will be mixed down
to the surface.
Case studies built around mesoscale model simulations, like the one presented
here, have the potential to radically alter our understanding of fire-weather interactions.
Formulating new indices based on a physical understanding of how a fire modifies the
atmosphere and how the atmosphere can impact fire behavior is an active area of fire
weather research. By conducting similar case studies for significant wildfires both in the
eastern region and throughout the United States and the world, it is possible to better
understand these interactions and improve all facets of fire-atmosphere interaction
prediction, assessment, and understanding.
6. References
Brotak, E.A., and W. E. Reifsnyder, 1977: Predicting major wildland fire occurrence.
Fire Management Notes 38, 5-8.
Byram, G. M., 1954: Atmospheric conditions related to blowup fires. United States
Department of Agriculture Forest Service Station Paper 35, 31 pp.
Carlson, T. N., S. G. Benjamin, G. S. Forbes, and Y.-F. Li, 1983: Elevated mixed layers
in the severe storm environment - Conceptual model and case studies. Mon. Wea.
Rev., 111, 1453-1473.
Charney and Keyser, 2004: NWP conference paper on mixed-layer impact on fire
indices and diagnostics.
Chen, F., and J. Dudhia. 2001: Coupling an advanced land surface-hydrology model with
the Penn State/NCAR MM5 modeling system. Part I: model implementation and
sensitivity. Mon. Wea. Rev., 129, 569-585.
Davis, R. T., 1969: Atmospheric stability forecast and fire control. Fire Control Notes,
30, 3-14,15.
Fahnstock, G. R., 1965: Texas Forest Fires. United States Department of Agriculture
Forest Service Research General Technical Report, SO-16, 19 pp.
Heilman, et al. (2003): EAMC Overview paper from the recent FFM conference
Hardy, C.; Ottmar, R. D.; Peterson, J.; Core, J. 2001. Smoke management guide for
prescribed and wildland fire -- 2000 edition. PMS 420-2. NFES 1279. Boise, ID:
National Wildfire Coordination Group. 226 pp.
Karyampudi, V. M., M. L. Kaplan, S. E. Koch, and R. Zamora, 1995: The influence of
the Rocky Mountains on the 13-14 April 1986 severe weather outbreak. Part I:
Mesoscale lee cyclogenesis and its relationship to severe weather and dust storms.
Mon. Wea. Rev., 123, 1394-1422.
Keetch, J. J., and G. M. Byram, 1968: A drought indexfor forest fire control. USDA
Forest Service ResearchPaper SE-38, Southeastern Forest Experiment Sta-tion,
Asheville, NC, 33 pp
Lansing, L., 1939: Weather preceding forest fires in New Hampshire. Bul. Amer. Met.
Soc., 20, 10-26.
Mesinger, F., G. Dimego, E. Kalnay, K. Mitchell, P. C. Sharfran, W. Ebisuzaki, D. Jovic,
E. Rodgers,E. Berbery, M. B. Ek, Y. Fan, R. Grumbine, W.Higgins, H. Li, Y. Lin,
G. Manikin, D. Parrish,W. Shi, 2006: North American Regional Reanalysis. Bull.
Amer. Meteor. Soc. 87. 343-360
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997:
Radiative transfer for inhomogeneous atmospheres: RRTM, a validated
correlated-k model for the longwave. J. Geophys. Res., 102, 4353–4356.
New Jersey State Forest Fire Service (NJFFS), 2003: Division forest firewardens report
of large fire and problem fire analysis: Jakes Branch Wildfire B06-02-02 (2002),
20 pp.
Simard, A. J., J. E. Eenigenburg, and S. L. Hobrla, 1987: Predicting extreme fire
potential. Proceedings of the Ninth Conference on Fire and Forest Meteorology,
American Meteorological Society, San Diego, California, April 21-24, 1987, 148157.
Table 1: Table 1: Sequence of events during the Double Trouble Fire on June 2, 2002
(taken from NJFFS, 2003). All times are EDT (UTC - 4 hours). Blue entries indicate
weather-related observations.
Time (EDT)
Activity Reported
1015
wind: west at 8 mph, gusting to 20 mph; temp: 75; RH: 62%
1300
Estimated start time of fire
1309
Fire reported by Cedar Bridge Fire Tower to Section Forest Firewarden
1314
Direct Attack on fire by NJFFS firefighters begins
1325
40 mph winds reported at fire site; Direct Attack abandoned
1326
Backfiring operations begin north and east of the fire
1331
Request for Garden State Parkway to be closed
1335
Backfiring operations begin south of the fire
1347
Request for aerial support for fighting the fire
1351
Fire jumps Double Trouble Road and approaches the Garden State
Parkway
1400
Fire is officially declared to be a Major Fire
1401
Fire has crossed the Garden State Parkway
1408
First report of a house being burned
1418
Wind shift reported on the fire line
1423
Fire crews prepare for structure protection
1451
Wind shift to the north reported; former right flank of the fire becomes
the head fire
1457
Lakewood Tower reports winds from the north at 35 mph
1536
Big wind shift reported on the fire line
1537
Lakewood Tower reports winds shifting to the East, North East
1538
Right flank becomes head fire
1553
Fire has been diverted south of the line of homes located just east of the
Garden State Parkway
1559
Wind shift reported on fire line; Electric lines down roadway
1601
House on fire
1604
Evacuation order issued for homes in the area
1610
Wind shift reported on fire line
Time
Temp
RH
Wind Conditions
1500 EDT
82 ˚F
28%
WNW 16 GUST 28 MPH
1700 EDT
80 ˚F
27%
WNW 16 GUST 26 MPH
1900 EDT
77 ˚F
30%
NW 14 GUST 22 MPH
2100 EDT
71 ˚F
35%
NW 12 GUST 18 MPH
0000 EDT
63 ˚F
45%
NW 8 MPH
0300 EDT
57 ˚F
55%
NW 5 MPH
0600 EDT
55 ˚F
65%
NW 5 MPH
Table 2: Spot forecast issued by the Mount Holly National Weather Forecast Office. All times are EDT
(UTC - 4 hours).