The most interesting Satellite Observations of Forest Fires

Introduction: Fires in the Hell's Canyon area

Map showing the area covered by the satellite image
A map showing the area covered by the satellite on this pass shows the western U.S. is in deep night (the sun is just rising over eastern Hudson Bay). We can only collect data from the satellite when it is in view of our ground station. The bold diagonal green line shows the path of the satellite on the ground, that is, where it passed directy overhead. The map label states that this was a descending pass which means the satellite, named NOAA-14, was moving toward the south. As it moves along its orbit it scans the earth below on both sides of the ground track. Some of the scans are plotted, the total area covered is inside the green area.

Image of what the satellite saw
As the satellite was scanning southern Oregon and Idaho the signal was lost as it dropped below our horizon. A view of what the satellite saw shows that the signal loss was not completely sudden, it weakened causing noise, the little white speckles along the bottom of the image (also the top), and also had a couple dropouts causing the horizontal black bands at the bottom of the image. This view was made from channel 3, one of the 5 channels of the AVHRR sensor on this satellite. Note the shape distortion near the left and right edges of the image. The satellite scans a single strip outward from its ground track, the farthest points are distorted by the curvature of the earth. Take note of the state of Oregon on this image. A better view will be shown next.

Geographically corrected view
Computer software can be used to make an image of Oregon with the distortion removed. The same white speckles of image noise and black bands of signal dropout may now be seen more clearly. Again this is AVHRR channel 3 which is in the mid infrared. At this wavelength the earth appears to glow faintly. Variations in this glow show where it is warmer (brighter) or cooler (darker). Clouds are quite a bit cooler than the surface and appear black in this view since they block the glow of the warmer earth below them. The brightest spots in this image, above the image noise at the bottom, are hot spots. These are forest fires. The northern part of the eastern border of Oregon follows the Snake River as it flows through Hell's Canyon. Several small fires are visible in Hell's Canyon in this image.

More detailed view of Hell's Canyon area
The northern part of Oregon's eastern border follows the Snake River through Hell's Canyon. The easternmost curve in this border is deep within the canyon. This is a spectacular area, more details are available from a local newspaper web site, The Wallowa County Chieftain. The Hell's Canyon fires are shown more clearly on this view centered on Hell's Canyon. Fires of a similar size are visible in Idaho to the east and larger fires are seen to the southwest. Such fires show well on nighttime views such as this, but perhaps a bit surprisingly they also often show well on daytime views as well. The images list the exact date and time as well as the name of the satellite that aquired the data.

Shaded relief maps
The geographically registered channel 3 image may be used to determine the locations of the fires, especially using the latitude/longitude scales along the sides of the image. However, the fire locations are more clearly shown by transferring them to a shaded relief map. The channel 3 image hints at the topography, often the higher elevations are cooler and so appear darker. But variations in heat loss cause exceptions, this effect is seen in the small lake (Cascade Reservoir) outlined in dark green on the map at Lat 44.6, Long -116.1. On the channel 3 image the lake appears brighter than the surrounding valley since the water looses heat more slowly. So the shaded relief map shows elevations better but not as well as a color shaded relief map. Color is not used on the fire maps here since it makes the fires harder to see. Black and white is used for the map and magenta for the fires. Even so some of the smaller fires may be hard to see, especially in rough terrain. For that reason a variation called a finder map is also made. The finder map shows each fire point enlarged enough to be easily seen. These maps are called Hotspot maps since it is not guaranteed that all detected spots are real fires.

Day to day changes
Day to day changes are quite easily seen on the satellite images. A channel 3 image from the next night shows the fires to the southwest of Hell's Canyon are considerably more visible.

Color composite images
As stated above, fires may also be seen in the daytime. Daytime views are very useful when searching for smoke from a fire. Smoke is not easily seen at night, but often it is quite apparent in the daytime. A color daytime view of the state of Oregon shows a number of smoke plumes originating from areas of fire. This is a fairly complex view and needs some explanation. The view was aquired fairly late in the day as seen on the pass map which shows that for Oregon the sun was less than 20 degrees above the horizon (the sun was setting along the red line through Texas and the midwest). It also shows that Oregon was near the edge of the viewing area so is not seen as sharply as at other times, especially on the west side. This daytime view is much like the other daytime views shown on this site, it is a combination of 3 of the AVHRR channels. As mentioned above, the AVHRR sensor has 5 channels: 1=visible, 2=near infrared, 3=mid infrared, 4,5=thermal infrared. This combination of 3 channels is referred to as a color composite image. This image was made by coloring channel 1 green, channel 2 red, and channel 3 blue. Many combinations of 3 channels are possible, this was chosen to match somewhat the standard infrared Landsat views that are very familiar in remote sensing (and they tried to match color infrared film). Space does not permit a detailed discussion of just what all the colors mean (nor is that necessary here), but one item is of interest. We normally think of vegetation as being green. If our eyes were sensitive to a wider range of wavelengths we would probably think of the color of vegetation as near infrared since this wavelength is highly reflected back from plants. Since the near infrared, channel 2, was colored red here, red areas indicate vegetation. This image shows three main features: surface features in various colors (red, blue, purple, gray, some green), clouds in white and yellow, and smoke in bright yellow green. Channel 3 which is colored blue shows, in daytime views, both some reflected sunlight and some glow from the warm earth below. Channels 1 and 2 may be considered to be all reflected sunlight (channels 4 and 5 are almost entirely infrared emmission from the view below).

Discussion of the smoke plumes
So how do we know that the prominent yellow green areas are smoke from fires? Because they act like smoke is expected to act. Quite often they start from visible fires. This same area can be viewed in channel 3 only, and shows the fire hotspots. Comparing the channel 3 view and the color composite shows the yellow green plume at the southeast corner of Oregon originates from a cluster of fires. Another obvious smoke plume is seen from the small fire in Hell's Canyon. The fires southwest of Hell's Canyon, near Ukiah and Dale, show very prominent smoke plumes. Not every visible fire shows a clear smoke plume and not every smoke plume shows a visible fire. A possible fire without much smoke may be seen as a small bright spot just above the southern border of Oregon at about longitude -121.4. It certainly does not show a dense plume, although perhaps the smoke starts out blowing to the southeast and then curves around toward the northeast forming the faint plume at about longitude -120, just above the southern border of Oregon (however this plume might be from other fires farther south in California). A good example of smoke with no visible fire may be seen at about longitude -122.5, latitude 43.3. This area certainly looks much like the other smoke areas but channel 3 does not show bright spots here. There are a number of reasons why a real fire might not be seen on the channel 3 image. This area is near the edge of the scan, so not only is the resolution degraded (which causes a loss of sensitivity for seeing fires) but the view angle is more oblique. The satellite as viewed from this location on the ground would appear very low in the northeastern sky, so low that it could easily be blocked by hills and mountains in rough terrain. This effect definitely causes a problem in seeing the fires directly on channel 3. However, the smoke plumes have no such problem, they rise above the hills and are easily seen. Wind variations with time and altitude can cause some very interesting and complex shapes in the smoke plumes. The smoke from the fires near Dale and Ukiah clearly show this effect, near the ground the smoke blows south or southeast but higher up it swings around and heads northeast (with perhaps some of the highest parts blowing off to the east in long narrow streamers). Why is the smoke yellow green on these views? This is due to the way the smoke particles scatter the sunlight in the different wavelength channels. Looking at the channel 3 image shows that no smoke is visible, this wavelength of infrared cuts right through the smoke. The smoke does show up in channels 1 and 2 which are colored green and red, and green and red when mixed makes yellow (for light, not paint). If the smoke were also visible on the channel 3 image the color composite would show it as more white. That is why the clouds appear more white (except where they are cold which is dark on channel 3, then they are yellow). The smoke shows up a bit better in channel 1, the visible channel which is colored green, than in channel 2, the near infrared channel which is colored red. So there is more green in the mix of green and red, giving a yellow-green result.

Fire scars
Large fires leave a visible mark on the satellite images for quite a while after the fire itself is gone. The large fires southwest of Hell's Canyon, near Dale and Ukiah, are a good example. A daytime color composite image shows a morning view of the area. A number of fire scars are visible but those from the fires mentioned above are seen as blue/gray patches near the left edge of the image at about latitude 44.7 to 45.1. Compare the earlier nighttime view to see just where the fire scars would be expected. There is a minor north/south position shift between these two images. An attempt to correctly geographically register the images is made but sometimes small errors are still apparent. Other fire scars are visible on the color composite, the most prominent is the deep blue, almost purple, area in Hell's Canyon itself. The royal blue patches to the southeast of Hell's Canyon are water bodies. A later color composite image shows the fire scars even more clearly (a few clouds look like popcorn scattered over the image).

List of images
This section gives a convenient list of the images discussed above, plus a few extras.

1996 Aug 25 09:59 UT     NOAA-14 nighttime view     Fires

1996 Aug 26 09:47 UT     NOAA-14 nighttime view     Fires

1996 Aug 31 01:06 UT     NOAA-12 evening view     Smoke

1996 Sep 01 14:41 UT     NOAA-12 morning view     Fire scars

1996 Sep 08 20:34 UT     NOAA-14 daytime view     Fire scars