On March 22, 2013 at about 7:53 pm EDT people in the mid-Atlantic states and beyond witnessed a bright, multi-colored fireball caused by a boulder-sized bolide streak across the night sky from the northwest to the southeast. Meteors have been in the news lately, particularly on February 15, 2013 when one exploded near Chelyabinsk, Russia. As with the event over Russia, inevitably the question asked around CIMSS is “Did the geostationary satellites see the flash or the heat signature?”
The current generation of geostationary imagers were not designed to detect transient and/or sub-pixel events such as asteroids entering the atmosphere, lightning, and fires. However, in some cases they do. Fires are the most well known example, and have been an area of research at SSEC/CIMSS for almost 20 years. Active fires occupy only a fraction of a pixel from a geostationary satellite, making their detection a challenge but not impossible. The Wildfire Automated Biomass Burning Algorithm (WFABBA), developed at SSEC/CIMSS, automates the process in near real-time and provides fire detection and characterization from GOES, Meteosat Second Generation, MTSAT, and COMS. Lightning flashes are frequently below the satellites’ minimum detectable intensity and are of such short duration that they are easy to miss, though there has been some work identifying lightning in SEVIRI images (opens as a PDF). SEVIRI has also detected the light and heat signatures of an asteroid entering the atmosphere at least once.
In the case of the Russian meteor, the condensation trail left behind was seen by multiple platforms, but the satellite scans happened at the wrong time to capture the heat signature from the event. After the March 22 fireball, HansPeter Roesli contacted us regarding a hotspot he saw in a 3.9 micron GOES-13 image taken that night:
Scientists at CIMSS immediately went to work looking up information about the fireball and found estimated paths provided by the American Meteor Society derived from hundreds of sightings:
The WFABBA flagged the hotspot as a pair of fire pixels, and rather intense ones at that:
The orange pixel is hotter than the red one, with a brightness temperature of 317.7 K for the orange one and 305 K for the red one. The clouds, by comparison, are around 256-257 K. This led the WFABBA to estimate the fire radiative power of the orange pixel at nearly 700 MW, which is at the high end of detected fires. This rules out the source of the hotspot as a fire on the ground. A fire of that intensity would persist long enough to be seen more than once and given the population density of the area such a hot and/or large fire would have drawn attention. No fires that could account for the hotspot were reported in that portion of southern New York state.
Meteors often break-up at around 50 miles up, which means that the hotspot wasn’t actually over New York state, but rather somewhere to the south. Applying a parallax correction for altitudes of 80 km and 85 km (approximately 50 miles) put the source of the hotspot basically right on the fireball path estimated by the American Meteor Society:
Eureka! The spatial match between the path and the hotspot is great, and the hotter pixel is in the direction of motion.
Unfortunately, the spatial match means little if the times of the ground and satellite observations do not match. Observers on the ground saw the fireball at approximately 7:53 pm EDT (23:53 UTC). GOES-13 scanned the hotspot at precisely 7:33:43 pm EDT (23:33:43 UTC). While a couple seconds or even a minute could be explained away, the 20 minute difference is too large to ignore.
The hotspot is almost certainly real. The GOES-13 image is devoid of noise otherwise. The hotspot was present in the original data stream from GOES-13 to NOAA’s receiving station at Wallops Island, VA. Having two adjacent hot pixels, neither of which maxed-out (saturated) the sensor, while having no noise pixels otherwise is highly improbable, and in the years of examining this type of data for fires such pairs have virtually always proven to be fires or reflections off of something on the ground.
Since the hotspot is not a fire, and no meteors were seen at 7:33 pm that night, the likely cause of the hotspot is reflection. But off of what? It was not a reflection off of the ground – the sun had already set. Planes are too small and also would not have been sunlit at that hour. There are, however, approximately 13,000 manmade objects (and debris clouds) being tracked by NORAD at varying altitudes, many of which would have been illuminated by the sun. The first candidate to come to mind is the ISS, however it was over the Southern Hemisphere at the time and thus could not have been the source. The object would have to have been at sufficient altitude to reflect sunlight (it was after dark on the surface) and be of sufficient size to reflect sunlight but also not be too bright in the visible wavelengths. The location over the Earth would vary with altitude:
The size and altitude of the object are not known, but since it was picked up by two adjacent pixels there are constraints on its size relative to its altitude. Identifying which of the NORAD-tracked objects could have been in the line of sight between GOES-13 and 41.4 N and 74.5 W would be a substantial computational effort. On April 1 the wannabe-prankster author of this blog contemplated what the calculations would look like if the object was the USS Enterprise as portrayed in the recent “rebooted” 2009 Star Trek movie. The idea of an April Fool’s Day post was abandoned but the example is still somewhat helpful to illustrate the concept. Using an overhead view of the Enterprise produced by Tobias Richter and factoring in the east/west oversampling performed by GOES-13 (adjacent pixels overlap by a bit less than 50%), one could speculate that the Enterprise would fit within two pixels in a fashion similar to this:
The green and yellow boxes approximate the GOES-13 pixel footprint. The pixel locations were selected to roughly approximate the relative radiance difference between the two pixels – the green one has to have about 1.7 times more reflective surface than the yellow one to create the temperature difference observed by GOES-13. Admittedly this is very, very rough and completely ignores the oblique sun angle and complicated surface features of the Enterprise. It also ignores the point spread function of the pixel – not every part of the pixel contributes the same relative amount of energy to the detector, the parts near the center count more than the parts near the edges, and some of the area outside the pixel footprint counts as well. This is one of the features of satellites that makes detection of sub-pixel features like fires a challenge.
Running the numbers and using the size of the Enterprise given by its designer (725.35 m), the Enterprise would have been in orbit approximately 31,000 km above the Earth’s surface, just north of the equator, to produce a reflection like the hotspot that was observed. As a geostationary satellite, GOES-13 orbits at 35790 km right above the equator.
This example does not rule out a smaller source in a lower orbit. The physics of detection of sub-pixel features with a GOES Imager are such that a lower orbiting and/or smaller but highly reflective object could produce a reflection of sufficient intensity to be detected by GOES. And while it is highly improbable instrument noise cannot be ruled out completely – if you eliminate the impossible, whatever remains, however improbable, must be the truth. We can rule out the presence of a Federation starship, however, as Starfleet regulations forbid interference with historical events and with species incapable of superluminal travel, and because the Enterprise does not actually exist. (yet)
The next step in investigating this hotspot is to take the orbits provided by NORAD and calculate the positions of the objects at the time the pixel was detected, and see which ones could have been near the line of sight. With luck, one or more candidates will be identified. It is conceivable that these types of hotspots occur with some regularity. If they occur over water, no automated algorithm will draw attention to them and they will be treated as noise. Over land they are hard to distinguish from WFABBA fires simply because they look the same as fires do. Fire can be ruled out here due to the magnitude and location of the detected spot, but doing so on a regular basis would require examining the thousands to tens of thousands of fire pixels detected every day. An automated technique to find these reflections would likely involve calculating the orbital positions of candidate objects and filtering those for the ones that could be in the right place and time to be seen.
Particularly eagle-eyed readers may have noticed that in the first image the pixels look rectangular, but in later images they appear square and the image looks stretched in the east-west direction. This is due to the oversampling performed by GOES in the east-west direction, as illustrated in the Enterprise example.