SATELLITE 
INTERPRETATION 

DISCUSSION

NOAA/NESDIS
Regional and Mesoscale Meteorology Team
Cooperative Institute for Research in the Atmosphere (CIRA)
Colorado State University   Fort Collins, Colorado

 

 

October 2, 2002

Colorado wildfires from GOES-11; 08-09 June 2002 


Figure 1


Table 1

I.  Introduction.


As the summer of 2002 approached, much of the western United States was entering its third, and most severe, year of drought (Fig. 1).  By the end of the first week in June, a large fire, known as the Ponil complex had already burned nearly 70,000 acres in north central New Mexico.  In Colorado, there had been 10 large wildland fires by 7 June.  Each of these had burned areas ranging from 102 – 103 acres (Table 1).  Furthermore, the Denver National Weather Service (NWS) office had already issued eleven Red Flag Warnings (extreme fire danger) – four times the normal for that early in the year.  NWS fire weather meteorologists, U.S. Forest Service personnel, and state and county agencies were all warning that things would be getting a lot worse as the drought wore on.  They were right.  The year 2002 turned out to be the worst wildland fire season in all of Colorado’s recorded history.

II.  Special GOES Data Collection.

At the same time that the fire season was becoming progressively worse, a meteorological research field program known as IHOP (International H2O Project) was underway in central Oklahoma. In support of this program, NOAA/NESDIS had decided to temporarily activate its reserve geostationary weather satellite (GOES-11) which had been stored in orbit awaiting potential failure of either one of the operational weather instruments (GOES-8/GOES-10).  The activation period ran from 29 May through 20 June 2002.  Additionally, an unusual collection schedule during the three-week period allowed for data collection at a rate of one imaging scan every five minutes over the entire continental United States.  This scan rate is three times the normal.  For this reason, very high time resolution GOES data were being collected on June 8th and 9th, when five large forest fires began – two of which turned out to be the largest in Colorado recorded history.  This page describes these two days through various data sets, though the discussion is focused primarily on the 5-minute interval GOES-11 imagery.

III.  GOES 3.9 µm Data.

A brief sidebar is in order on the use of the GOES channel-2, shortwave infrared imagery ( a channel whose peak response is at 3.9 µm) in wildland fire applications. The 3.9 µm channel is different from the other five imaging channels in that it is characterized by an intense, sub-pixel response to heat.  (Note: During this period the size of a GOES-11 channel-2 pixel at central Colorado’s latitude and longitude was approximately 2.4 km east-west by 6.0 km north-south.)  Channel-2’s strong sensitivity to sub-pixel “hot-spots” makes it ideal for use in the early detection of wildland fires, especially in forested areas where fires burn hot.

The reason for the more effective sub-pixel response in the 3.9 µm channel is evident when one looks at Fig. 2 (left side) which compares Plank radiances at shorter, versus longer, infrared wavelengths.The plot illustrates the more rapid response of the shorter wavelength sensors to increasing heat.  Fig. 2 (right) compares the response in the 3.9 µm sensor directly with the temperatures as measured by the 10.7 µm channel, a “window” channel that measures the actual average temperature of the pixel being scanned.  In this example, the sub-pixel, hot-spot is assumed to average 500oK and the remainder of the area covered by the pixel is at 300oK.  Note that if only 5% of the pixel is at 500oK, the 3.9 µm measured brightness temperature of the pixel is 360oK, while the corresponding 10.7 µm average pixel brightness temperature is less than 320oK.  When there is no sub-pixel heat source, or when the entire pixel reaches the 500oK temperature, the two channels match one another within a degree or two Celcius.  The sudden appearance and persistence of a very hot pixel in the 3.9 µm imagery is often an early indication of a significant fire.  (Lack of persistence over two or three sequential images may be a sign of instrument noise, or of a short-lived, controlled burn.) 

It should be noted that the channel-2 pixel on the current GOES instruments saturates at preset values, and the saturation value depends on the satellite.  For the currently operational satellites, the values are as follows: GOES-8, 336 oK, GOES-10, 221.5 oK, GOES-11, 338 oK, and GOES-12, 336 oK.  At and above the saturation temperature, pixels values for the imagery are arbitrarily set to 273 oK, except on GOES-11, where the value is set to 163 oK.  For GOES-8, -10, and -12, the linear color table (wherein warmer temperatures are darker) becomes completely black at saturation, while for GOES-11 the table “wraps around,” and the pixel suddenly turns pure white.  Note: For a complete discussion of the attributes of GOES Channel-2 see: http://www.cira.colostate.edu/ramm/goes39/cover.htm


 


Figure 2


Figure 3


Loop 1

IV.  Scan Frequency and Data Delivery.

The first significant wildland fire to be reported on 8 June 2002 began about 20 mi. northwest of Grand Junction, CO in Long Canyon.  It was triggered when smoldering duff, which had been ignited by lightning on the late afternoon of June 3rd, flared-up.  This occurred when surface winds turned from southeast to southwest, and began gusting to over 30 kt.  At the same time, dewpoints dropped 5-10 degrees F.  The first call from the public reporting the fire came into the Grand Junction, CO, Bureau of Land Management Fire Protection District dispatch center at 18:12 UTC.  Of interest here is that the fire could actually be seen as a hot spot on the GOES-11, 3.9 µm imagery on the 18:03 UTC image (though it may have gone unnoticed at that time, since the single, darkened pixel was difficult to see).  Apart from the fire meteorologist’s skill, however, there are other factors that currently work against the utility of this potentially powerful tool – the two most important of which are scan frequency and delays in data delivery.

First, consider the current GOES-8 and GOES-10 data that are collected at a routine collection interval of 15-minutes. Utilizing GOES-11 imagery at 15-min intervals to simulate that mode, the first time the hot spot becomes truly obvious is at 18:15 UTC (Fig. 3c). Going back to 18:03 UTC (Fig. 3b), one can identify the very dark pixel in the same spot, but it is prudent to observe a series of at least three images showing a hot pixel in the same location before thinking, “possible wildland fire.” In this case – since there’s a hot spot on the 18:03, 18:15 and 18:34 UTC images – the fire weather meteorologist would want to notify the appropriate response agency shortly after viewing the third image. Unfortunately, under the current routine data (15-min) delivery schedule, satellite images are not loaded onto the forecaster’s display system (called AWIPS – Advanced Weather Information Processing System) until about 20-25 minutes after initial scan time. Thus, the fire weather meteorologist would not be able to obtain the requisite series of three images until about 18:55 UTC, some 43-min after the fire was reported by the public.

Now let’s look at the GOES-11, 5-min interval data (Loop 1).  Even if one didn’t notice the warm pixel(s) at 18:03 and 18:07 UTC, the saturated pixel at 18:15 UTC certainly gets your attention.  Looking back, one would find the overheated pixels on the 18:03, 18:07 and 18:15 UTC images.  This series of three provides the requisite confidence that one may be looking at an actual wildland fire, and the decision can be made as soon as the 18:15 image is in hand.  This step occurs more rapidly in this case, because satellites operating in special rapid-scan mode have their imagery delivered to AWIPS within 8-min of their initial scan time.  In this case then, the appropriate responding agency could have been notified within 11 - 12 minutes after the first public report of this fire.  In less populated regions – or at night – GOES imagery can often provide the first indication of a wildland fire in progress.

Future GOES systems are planned that allow for routine rapid interval scanning schedules (5-min scanning is planned for GOES-R) at twice the spatial resolution of today’s satellites.  Furthermore, future forecast data display systems include plans for more rapid delivery of imagery, so that lead times will be cut even further.  However, there are partial solutions available now.  Today’s rapid-scan schedule operates at an uneven imaging interval mode (Table 2) due to conflicting scheduling obligations, but the delivery time to the AWIPS is still around 8-min.   Rapid scan operations can be initiated by NWS forecasters for any legitimate forecasting need.  During fire season, it is strongly recommended that fire weather meteorologists call for rapid-scan imaging on red flag days.


 


Table 2


Figure 4


Loop 2

V.  A Late Detection and a False Alarm.

According to a Garfield County, Colorado (press release) a wildland fire, that came to be called the Coal Seam Fire, was first reported at about 19:00 UTC on 8 June 2002.  It eventually burned more than 12,000 acres, and destroyed 29 homes and 14 outbuildings in and around the town of Glenwood Springs, Colorado. It was started by an underground coal seam that has been burning since the mid-1970's.  The seam runs from West Glenwood to New Castle, Colorado.  Over the first 45-minutes of its life the fire grew appreciably in size and at 19:45 UTC, Glenwood Springs firefighters requested helicopter support.  By that time, the fire had burned a long narrow swath nearly a mile long.  However, the fuel-type involved at that time was primarily grass, with a small percentage of mixed brush, so the sub-pixel heat was insufficient to show up on GOES-11, channel-2 imagery (Fig. 4). 

It wasn’t until later in the afternoon that the Coal Seam fire became well defined (Loop 2).  Though darkened pixels can be seen (off and on) between 19:45 and 20:15, the fire doesn’t become obvious until 21:15 UTC.Thus, detection of the Coal Seam fire would have lagged the public report by at least an hour.  An informal test amongst a number of CIRA researchers finds that before a darkened pixel becomes visually evident, the temperature difference between the 3.9 µm and 10.7 µm pixel must be roughly 10oC – 15oC.  This represents about 2% of the pixel area for the 200K example illustrated in Fig. 2.  For a 2.4 x 5.9 km (~3,500 acre) pixel, this would still require nearly 70 acres to be burning vigorously.  It is hoped that the better resolution planned for future-GOES will allow detection of these cooler burning and/or smaller wildland fuels.  With the resolution planned for the GOES-R instruments, the channel-2 pixel size will be roughly 875 acres and the burn area would only have to be about 18 acres.

Another factor that could easily have delayed the Coal Seam fire detection was that a control burn was taking place in Eagle county.  That fire was hot enough to darken a pixel or two for several scans.  This hot spot could easily have distracted the fire weather meteorologist from less obvious indicators.  Worse, the control burn likely would have led to a false alarm, though it has become increasingly clear that emergency managers and responders would rather receive a number of false alarms rather than miss one significant fire.

VI.  The Hayman Fire.

A fire which was to burn 137,760 acres of Colorado forest land, and would eventually destroy 133 private homes, 1 business, plus 466 outbuildings began late on the afternoon of 8 June 2002.  It would become the largest wildland fire ever recorded in Colorado. The first report of the Hayman fire was received at 22:55 UTC.  It was reported by the person who had started it -- when it was still relatively small.  On satellite imagery, the fire didn’t really become obvious until around 23:45 UTC (Fig. 5).  The most likely reason for the delayed signal is that a field of cumulus cloudiness had developed over the area (Fig. 6).  This is not a very common occurrence on red flag days.  However, when it does happen, detection by satellite can be delayed appreciably.  Another instance when cloudiness can delay detection are those days on which lightning from a low precipitation thunderstorm starts the fire.  In those cases, detection by satellite will be delayed until thunderstorm anvils clear the area. 


 


Figure 5


Figure 6


Loop 3

VII. Expansion of fires on 09 June 2002.

The last fire to be described in this discussion will be the "Missionary Ridge" Fire (in extreme southwestern Colorado) which was first reported by the public at 20:32 UTC on 9 June 2002.  The first indication on the GOES-11 imagery came at 20:55 UTC, and the involved pixels heated rapidly after that (Loop 3).  The fire itself grew to 6,800 acres in one day.  By the time it was contained, it had burned a total of 70, 485 acres to become the second largest fire in Colorado recorded history.  Some extremely interesting fire behavior took place within the Missionary Ridge complex on 17 June 2002 when the fire spawned a damaging tornado, outside of the plume itself.  Material on this event can be found at this location.   Note also the large fire that appears in western Colorado, just south of the Long Canyon fire area. This was the so-called Dierich Fire that burned nearly 3,000 acres -- a small fire by 2002 standards.

Figure 7 shows a GOES-11 visible image of the four primary Colorado fire plumes on the late afternoon of 9 June.  A thin plume of smoke can also be seen crossing into southwest Colorado from large fires burning in Arizona at the time.  Notice that the Hayman Fire plume is extremely large.  Interestingly, the rising column of hot air was warm enough, and deep enough, to produce convective clouds above it (Fig. 8) even though the sounding in that day was extremely dry (Fig. 9).
 


 


Figure 7


Figure 8


Figure 9

The convection was sufficiently intense to produce near-precipitation-intensity echoes (Loop 4) on the Denver WSR-88D.  Loop 5 shows the plume in GOES-11 visible satellite imagery.  Note that the thick smoke extends all the way up into southwest Nebraska, with thinner smoke on its west side covering most of the northern Front Range.  Finally, Loop 6 shows a series of web camera images from 9 June 2002. The camera belongs to the Colorado Department of Public Health and Environment. This series of hourly images shows the smoke plume arriving over the Denver metro area on the late afternoon of the ninth.

 


Loop 4


Loop 5


Loop 6

VIII.  MODIS Imagery.  MODIS (Moderate Resolution Imaging Spectroradiometer) is an instrument aboard the Terra (EOS AM) and Aqua (EOS PM) satellites.  Terra's orbit around the Earth is timed so that it passes from north to south across the equator in the morning, while Aqua passes south to north over the equator in the afternoon.  Terra MODIS and Aqua MODIS are viewing the entire Earth's surface every 1 to 2 days at 250 km resolution.  The following are examples of MODIS images, in the visible channel, from 9 June 2002 over Colorado.  Three different image sizes are offered -- at resolutions of 1 km, 500m, and 250m -- and should be downloaded according to the speed of your access line. The visible channel on GOES-R will have horizontal resolution in the visible channel comparable to the 500m MODIS image.

 


1 km (372 KB)


500m (1.2 mb)


250m (3.4 mb)

Interesting MODIS images can be found at:  http://rapidfire.sci.gsfc.nasa.gov/gallery/?2002167-0616

To view fire hotspots featured on MODIS imagery see:  http://www.fs.fed.us/eng/rsac/fire_maps.html

General information on satellite fire detection can be found by accessing:  http://www.ssd.noaa.gov
 
 

Back to top

We welcome your comments and discussion at ramsdis@comet.ucar.edu

Previous RAMMT Satellite Interpretation Discussions

Previous RAMMT Satellite Interpretation Discussions in reverse chronological order
 

Information Contact: John Weaver

CIRA/RAMM WebMaster: Hiro Gosden

Authors: John Weaver and Dan Bikos

Data Aquisition/Analysis: Daniel Coleman (CIRA) 
and Dolores Kiessling (COMET)

Last Updated: October 2, 2002