The images and descriptions in these pages are experimental observations only and must be interpreted with some understanding of the limitations of the available data and processing techniques. It is suggested that the early entries be read if you are new to this site. They contain more detailed descriptions of the images which will help you become more familiar with their nature and limitations. Also some features visible in the images are described in the early entries but these descriptions are dropped in later entries to avoid repetition. Some of these features may resemble fires or smoke, it is very helpful to be aware of what an area contains before interpreting what an image shows. One resource for topographic information is the Color Landform Atlas of the United States. The data is from the NOAA polar orbiter AVHRR sensor which scans a thin ribbon of the earth beneath at 6 times per second. From its altitude of a bit over 500 miles it views this strip of about 1800 miles long across its path as it orbits the earth at about 4 miles per second. These strips together form an image of the earth beneath, more sharply directly below at about a 1 km (0.6 miles) resolution, less sharp farther out on the scans. We currently collect data mainly from two such satellites, NOAA-12 and NOAA-14. These satellites continuously transmit the data they collect, anyone with the proper equipment can receive it. The satellites are in sun-synchronous orbits, that is, orbits that pass over an area about the same time each day. NOAA-14 is about mid-day and mid-night, NOAA-12 is more near the sun rise/set line, on the day or night side depending on the time of year. The AVHRR data from these satellites consists of 5 channels. One is in the visible region, one in the near infrared, and the rest further into the infrared. Large fires are often seen in channel 3, the mid infrared channel, near where the fire spectrum peaks (see the Physics section of Frank Monaldo's Primer on the Estimation of Sea Surface Temperature for details on the spectral coverage of the AVHRR sensor). Channel 3 alone sometimes gives an incomplete view of what is or is not a fire. A number of factors cause both false positives (detecting non-existent fires) and false negatives (missing real fires). At night channel 3 pretty much gives an idea of how warm or cold are the features it views. In summer lakes often are warmer at night than the land around them and appear bright in the channel 3 image, sometimes quite bright. Lakes also appear bright in channel 4 which is less sensitive to real fires so both channels are used together. Real fires may be missed for the following reasons: Too small. A significant fraction of the resolution element of the sensor must contain fire to be detected. This becomes more of a problem near the far edges of the scan lines. Clouds may block the view. Light from a fire can penetrate through a fair amount of smoke, but thick clouds will block the view. Below detection threshhold. Some areas suspected from viewing channel 3 itself may not be labeled as hot spots. The detection algorithm is tuned to avoid giving too many false detections so may miss some actual fires if they are somewhat weak in the image. Coverage is not available. Obviously if the satellite does not view a region it cannot detect any fires in it. Non-existent fires may be indicated for the following reasons: Poor signal adds noise to the image. When the signal is blocked or degraded, due to the satellite being low in the sky and going behind some obstruction, the images show a random speckle pattern. The methods used to indicate abnormally hot areas are fooled by this image noise. Often these false positives occur in straight bands across the image. Such aligned points should be ignored. The edges of clouds cause problems. Channels 3 and 4 are compared to see if there are abnormally hot areas. Spotty clouds are a problem for the current method used for this. The current method does not work well during the day. Daytime images give a lot of false positives and are currently avoided for detecting hot spots. They are useful for imaging smoke plumes from the fires. Because of the current limitations the generated maps are labeled as hot spot detection maps.