[from "The Near-Earth Asteroid Tracking (NEAT) Program: An Automated System for Telescope Control, Wide-Field Imaging, and Object Detection," by Steven H. Pravdo, David L. Rabinowitz, Eleanor F. Helin, Kenneth J. Lawrence, Raymond J. Bambery, Christopher C. Clark, Steven L. Groom, Stephen Levin, Jean Lorre, and Stuart B. Shaklan (JPL) and Paul Kervin, John A. Africano, Paul Sydney, and Vicki Soohoo (AFRL), Astronomical Journal, 117, 1666 (1999)]
3. OPERATING PROCEDURES
3.1 Observing modes and planning
The "survey" is the primary NEAT observing mode. It is designed to discover new objects. In preparation for a night of observing, the first task is to create an observing script that lists the position of the search field. For this purpose we use a sequencing program, run once at the start of each 6-night run. Figure 2 shows a hypothetical 6-night search pattern planned for 1998 September 14-19. The program takes into account the time to expose and read out each image, as well as the number of nights per run and their duration. It thereby determines a search pattern that will uniformly sample the areas of the sky close to the ecliptic and to opposition. In order to keep the telescope pointed near to the meridian, the program targets a given night's search along strips of sky, each perpendicular to the ecliptic, and separated in longitude from one another by 11.25°. With the length of a given strip chosen so that it can be searched in ~45 minutes (~20 fields for the current NEAT system), the search is completed by the time the next strip approaches the meridian. By shifting the longitude of the search strips each night by 2.25°, the program creates a search pattern that uniformly samples the ecliptic within 45° of opposition after 5 nights. Search areas covered on the first night are repeated on the sixth, thus yielding positions with 6-day separation for any objects moving slowly enough (< ~0.2 deg/day) to be detected on both nights.
There are additional constraints that shape the search pattern. We generally choose longer search strips within 15° of opposition in order to increase the coverage there. The latitude of the observatory, 20.7° N, and design of the telescope limit the available Declinations to > -38°. The time of year limits the hour angles between about 3 hours west at astronomical twilight and 3 hours east at astronomical dawn. Observations within 15° of the galactic plane are also avoided because confusion with stars thwarts asteroid detection. Finally, in cooperation with the Spacewatch search (Scotti, Gehrels, & Rabinowitz 1991), we avoid the relatively small areas of sky that they search each month. The search pattern in Figure 2 shows large gaps where galactic plane appears, and small holes closer to opposition revealing the typical areas searched by Spacewatch.
In addition to the survey search positions, a few positions are scripted each night to follow up objects discovered on previous nights or lunations. Weather and schedule permitting, NEAT follows up all candidate NEOs, comets, or other bodies with unusual orbits or properties. Criteria for deciding if an object is worthy of follow up are described below (Sec 3.3). It is also possible to insert new positions into the observing script while the night-time observations are in progress, thus permitting follow-up observations in near real-time. Several recently occurring gamma-ray burst fields have also been observed using this near real-time method.
For each target position, the observing script also may be used to specify the observation time to ~10 sec precision. This feature has been used to test the capabilities of the NEAT system for tracking artificial satellites. Because of the high rates of motion for these objects, the exposures must be obtained within 1 minute of the time they reach their scripted positions. The script may also be used to specify image binning (the summing of neighboring pixels in both the horizontal and vertical directions as an image is read out). Binning reduces the time to read the image in proportion to the number of pixels summed. During tests of satellite tracking, this option has also been used to increase the observation rate.
Once the observing script has been loaded, and just before the end of nautical twilight, on-site operators prepare the telescope for operation. They remove the mirror cover, open the dome, clear the previous night's data from the disks of the control and analysis computers, and start the control program. The program takes over, pointing the telescope and acquiring images as scripted. The program also acquires dark-current images at one-hour intervals, taken with the shutter closed but with the same exposure time used for the search images. The search program that runs on the analysis computer subtracts these dark-current images from each sky image as part of the analysis procedure (discussed below). If bad weather interrupts the observing, the operators can pause the control program until the weather clears. It will continue where it left off. After a pause or for any other reason, the program will skip a scripted exposure if there is not enough time to obtain an entire triplet of images before the target position has set below 10° elevation. The control program will proceed with the next position on the script until there are no more, or until the operators stop the program at the start of morning nautical twilight.
While the control computer executes the observing script, an auxiliary program ("the analysis manager") runs on the analysis computer, monitoring log files generated by the control program. As soon as the control program has acquired a complete triplet of images, the analysis manager adds their file names to a processing queue. Appendix A gives a more complete description of the operations system. For each triplet in the queue, the analysis manager launches an additional program (described in Appendix B) to search for asteroids and to record their magnitudes and astrometric positions. Up to four instances of this search program can be run in parallel, each using one of the four CPUs of the analysis computer, and each analyzing a different triplet. For each asteroid, the search program also records 9 small sub-arrays or "patches" of image data (about 25x25 pixels each), 3 from each image in the triplet. Of the three patches taken from a given image, one is centered on the measured position for the asteroid, while the other two are centered on the positions where the asteroid appears in the other two images of the triplet. The 9 patches are later examined by eye to validate the detection (see discussion below).
For each analyzed triplet, the search program typically records 50 Kbytes of data (patches plus positions and magnitudes). A typical 10 hour night may yield ~270 triplets, or 15 Mbytes of information. This compares with 26 GB of raw image data and corresponds to a data "compression" by the processing system of a factor of ~2000. The processed data are further compressed and transmitted via modem and commercial phone line to JPL as soon as night-time observations are completed. At JPL, team members use a screening program called PATCHVIEW to visually inspect the 9 patches associated with each asteroid, and also to check the consistency of the measured positions. This serves as a final check of validity of each detection, and to pick out especially interesting objects for follow-up. Such objects are immediately reported to the world-wide observing community via the Minor Planet Center (MPC).
Figure 3a shows an example of the PATCHVIEW display for a given asteroid. The 9 patches are displayed as a 3 x 3 matrix. Column 1 (left) shows the three patches from the first exposure. Columns 2 and 3 (middle and right) show the three patches from the 2nd and 3rd exposures, respectively. If the asteroid is a valid detection, it should appear centered only within the diagonal patches running from the upper left (row 1, column 1) to the lower right (row 3, column 3). These are the locations where the search program found the asteroid in exposures 1, 2, and 3, respectively. The asteroid should not appear centered in any other patch in the matrix. These "veto" patches show the same locations as the diagonal patches, but at the times when the asteroid had not yet moved there, or when the asteroid had already moved away. For example, columns 2 and 3 of row 1 are patches from images 2 and 3, respectively, showing where the asteroid had appeared in exposure 1. Similarly, columns 1 and 2 of row 3 are patches from exposures 1 and 2 showing where the asteroid would appear in exposure 3.
Visual examination of these 9 patches is an efficient method to quickly identify the most common source of false positives from the search program: faint stars at the limit of detection. An example is shown in Figure 3b An object appears centered in the diagonal patches, but also in the veto patches. This observation clearly shows a star, and not an asteroid. The software incorrectly finds an asteroid, here, because it has only marginally detected the star. Because of the influence of random noise, and variation in atmospheric conditions (seeing and extinction), a faint star can appear above the detection threshold in one exposure, but below the threshold in the other two. The search program occasionally detects these faint stars (and other image artifacts) in such a way that they appear to be observations of a moving object.
For each asteroid, the PATCHVIEW program also displays ancillary information, such as the time, magnitude, position, and rate of motion (ecliptic and equatorial coordinates). A plot of the ecliptic rate of motion is used to decide if the asteroid has an interesting rate of motion. If the motion is outside the boundaries for the motion expected of main-belt asteroids (empirically determined, see Rabinowitz 1991), it is scheduled for follow-up and reported to the MPC as an interesting object. PATCHVIEW also calculates the deviation of each asteroid's measured positions from linear motion. If the deviation is larger than would be expected from measurement error, a decision may be made to reject the object, or to make further confirmatory observations before reporting it.
Each object detected with NEAT is assigned a unique, unpronounceable name. First a number is constructed based on the elapsed time of an exposure starting with the beginning of 1995, incremented by the object number within the exposure. This number is then translated into a 6-alphanumeric character name. Base 36 (26 letters + 10 digits) is used. The largest number is therefore 366 = 2176782339. For 10 years of observations with one exposure every 10 seconds this allows 70 unique names for objects per exposure. When an object is confirmed by recovery on a subsequent day with NEAT or other observers, it is assigned an official preliminary designation by the MPC.
At the end of the evening, while the night's haul of data is downloaded and screened at JPL, an archive program (Klimesh 1998) is run remotely at the Maui site to compress all the raw image data collected during the night (lossless compression by a factor ~2) and store it to Digital Linear Tape (using a Quantum DLT7000 tape drive). These tapes are later shipped to JPL for incorporation into the SkyMorph archive (Pravdo et al. 1998). This is a separate research program cooperatively run by JPL and NASA Goddard Space Flight Center. The goal is to create a data base of images and object information (brightness, shape, and position versus time for asteroids, comets, stars, galaxies, etc.) derived from the NEAT data and accessible on the internet. To date, more than 25,000 NEAT images have been archived by the SkyMorph project. The archive can be accessed via the World Wide Web at address http://skys.gsfc.nasa.gov/skymorph/obs.html.