Finding Hotspots
An autonomous sensor developed by NASA proves its worth in firefighting.
In the early morning hours of Oct. 26, 2006, an arsonist started a small fire in Riverside County, Calif., (Esperanza), and 18 hours later, 24,000 acres were involved. Later that same day the Governor declared a state of emergency, citing “conditions of extreme peril.” The California Office of Emergency Services (OES) asked NASA’s Ames Research Center, Moffett Filed, Calif., for help. The OES was aware that NASA had operated the ER-2, a U-2 reconnaissance aircraft specially modified for imaging Earth resources. Indeed, the ER-2 often flew the Daedalus Thematic Mapper Simulator, a sensor package capable of imaging in 12 spectral bands, matched to wavelength regions of the LANDSAT sensors. When the OES called on NASA-Ames, they didn’t know if Ames was still operating something like the Daedalus sensor, they simply hoped that some help would be available. They had no idea how fortunate they would be.
Unfortunate events, propitious timing
In years past, NASA officials had looked at the success of remotely-piloted aircraft (RPA) in military applications and realized that a sensor package that could be mounted on such a flexible platform would be valuable. NASA-Ames researchers had been tasked with developing instrumentation and demonstrating that scientifically valuable data could be gathered from an RPA. They began working on an upgraded version of the Thematic Mapper Simulator—a lighter, smaller, more autonomous sensor package, with lower electrical power requirements.
Vincent Ambrosia at NASA-Ames recognized that the engineering goal of developing an accurate RPA-mounted scientific sensor meshed with the pragmatic goal of providing rapid detection of wildfires. “With the possible exception of the Dept. of Homeland Security,” he said, “the Forest Service has the most immediate use for UAV-based sensor data for both tactical and strategic applications.” So began a four-year collaboration with the National Forest Service. A starting point was provided by the moderate resolution imaging spectrometer (MODIS) instrument currently flying on the Terra and Aqua Earth observing satellites. Data from two of MODIS’s 36 spectral bands are input into a fire detection algorithm that currently provides daily fire location information for the U.S.
The general requirements were defined: The new sensor system needed to be compatible
with an RPA, should be capable of general earth resources imaging, and must be
able to provide input for the fire detection algorithm used by the Forest Service
and other agencies. With the bar set, Ambrosia and his colleagues set to work.
This false color image of the
hills between Los Angeles and Palm Springs during last autumn’s
Esperanza fire was generated by NASA’s AMS-wildfire. Data from
two of the 12 spectral channels of the instrument combine to reconstruct
the temperature on the ground to within ±0.5°C.
|
NASA’s Altair remotely
piloted aircraft is a civilian version of General Atomics Aeronautical
Systems’ Predator B. The aircraft is controlled from a ground
station, meaning pilots need not be risked for “3-D” missions—those
dull, dirty, and dangerous. The instrument pod attached to the belly
contains NASA’s wildfire sensor package that was put to good
use during the recent Esperanza Fire in Southern California. (All
photos: NASA) |
|
Making it real
To be optimally effective, the new system needed to be capable of delivering readily-interpreted, accurate images of affected areas in near real-time, so that firefighting resources can be efficiently deployed.
That said, the optical instrument of the autonomous modular sensor (AMS) is contained within a cylindrical package mounted inside a pod attached to the belly of the aircraft. The sensor is a line scanner, with a rotating elliptical mirror providing a cross-track scan. Input light reflected off the input mirror is sent through a telescope, which collimates the beam at a reduced diameter of one inch. The light then enters a 12-band spectrometer assembly. Eight wavelength bands, ranging in width from 20 to 140 nm, cover the spectral range from 420 nm in the visible to 1.05 µm in the near infrared (IR). These bands are dispersed through a prism and collected by a custom array of silicon photodiodes.
Two additional near IR bands, from 1.55 to 1.75 µm and from 2.08 to 2.35 µm, are routed through filters and dichroic mirrors to two thermoelectrically-cooled InGaAs detectors. The final two bands cover, first, from 3.60 to 3.79 µm and, second, from 10.26 to 11.26 µm. These two bands, specially defined for wildfire applications, are selected by dichroic mirrors and a dual-band bandpass filter. The bandpass filter sits in front of a sandwiched detector with an InSb detector on top of an HgCdTe detector, cooled by a Stirling cycle cryogenic cooler. The “modular” in autonomous modular sensor refers to the ability to change out the 12-band wildfire spectrometer with one of two others, with different spectral bands optimized for ocean and atmospheric detection.
During operation, the output from the detectors is digitized into 716 16-bit cross-track pixels. The pixel angular resolution is selectable at 1.25, 2.5, or 5 milliradians (mr). For the wildfire work the 2.5 mr selection, which is about 100 ft/pixel for data acquired from 40,000 ft, provides the appropriate tradeoff between resolution and coverage.
The digitized data is combined with navigational and inertial sensor data to precisely determine the location and orientation of the sensor. In addition, the data is also processed with topographical information. The end product is an intensity map for each wavelength that is already geo-rectified for accurate representation.
The data from the final two bands, the mid-infrared at 3.7 µm and the far-IR around 11 µm, is also processed in the fire detection algorithm. Onboard blackbody sources calibrate each detector well enough so that the algorithm can report the temperature of the detected fires up to a temperature of 1,000° C with an accuracy of half a degree C. A second product of the AMS-wildfire sensor is the output from the algorithm, a vector map of hotspots.
The AMS, including heaters, coolers, the optical assembly, the digitizer and the processor, weighs only about 120 kg. With blackbody calibration sources and detector heaters on, the system can use up to 50 A at 28 VDC. But this small package, with some help from the processor on board the ALTAIR unmanned aircraft, provides not only the raw data, but a finished product formatted according to a geographical information systems standard, which makes it accessible with commonly available programs such as Google Earth. The system was built; now it needed to be proven.
Trial by fire
A flight demonstration program had been planned for late-summer of 2006. From August through October, the latter half of the official fire season, the ALTAIR was to fly four or five 24-hour flights over the western states. But the FAA, concerned over the safety and security of the national airspace, was reluctant to give approval for the series of remotely-piloted flights. Finally, on Oct. 20, approval was granted, but, with the end of the fire season only days away, the demonstration program was severely curtailed. All that was left was a single flight from NASA’s Dryden Flight Research Center, Edwards, Calif., up to the neighborhood of Yosemite National Park, viewing a couple of controlled burns along the way. The flight completed, engineers unmounted the AMS pod and prepared to store it for the next fire season.
Then the California OES, looking for help to handle the quick-moving Esperanza fire, asked the governor to request assistance from NASA. The ALTAIR was just around the corner at Dryden. After hurricane Katrina, when RPA were available to help but not given clearance by the FAA, the rules had been changed. General Atomics, which built and operates the ALTAIR, was eager to help, particularly after they were not permitted to assist with the San Diego fires of 2004, right in their backyard. Under the new FAA rules, with an emergency declared, the aircraft was cleared to fly.
Saturday at 3:47 p.m. the AMS-wildfire was airborne and headed for the fire site. FAA rules prohibit all crewed aircraft from flying over a wildfire at night. Firefighters at the Incident Command Post had to rely on limited ground observations and infrequent satellite data. The ALTAIR with the AMS-wildfire instrument flew a 16-hour mission, providing near real-time coverage of the fire throughout the night, data that was otherwise unavailable. A day later the fire was contained. The AMS had proven its worth.
— Richard Gaughan, is founder and Chief Engineer at Mountain
Optical Systems Technology. www.mountainoptical.com
|
