IAC-06-B6.1.2 SURVEY AND CHASE: A NEW METHOD OF OBSERVATIONS FOR THE MICHIGAN ORBITAL DEBRIS SURVEY TELESCOPE (MODEST) Kira J. Abercromby ESCG/Jacobs, Houston, TX, USA [email protected] Patrick Seitzer University of Michigan, Ann Arbor, MI, USA [email protected] Heather M. Rodriguez ESCG/Jacobs, Houston, TX, USA [email protected] Edwin S. Barker, Mark J. Matney NASA Johnson Space Center, Houston, TX, USA [email protected], [email protected] ABSTRACT The Michigan Orbital DEbris Survey Telescope (MODEST) is normally used to survey the geosynchronous orbit (GEO) environment to obtain a statistical assessment of the debris population. Due to the short time that the object is in the field-of-view (usually five minutes), it is common practice to assume a circular orbit when calculating the orbit from this limited observational arc. Some objects in the GEO regime are GEO-transfer orbit (GTO) objects which are observed at their apogee or objects with varying eccentricities such as those with high area-to-mass ratios. For these objects, an assumed circular orbit (ACO) prediction would not be accurate. After MODEST was modernized in March 2005 and brought under computer control, it became possible to use the telescope in modes other than tracking at the sidereal rate. Three data runs were conducted to test the orbit determination program, the ability to transfer data effectively between site locations, and to determine if objects could be re-acquired with this method. We report in this paper our initial efforts to determine full orbits based on follow-up observations after the initial detection in survey mode. Our long-term goal is to construct a system which can detect and follow-up an object in any orbit at GEO. This paper reports our first steps towards that goal. During the three data collections, only circular orbit objects were obtained. Although this allowed for testing of the process, further tests must be conducted in an effort to target GTO and high area-to-mass objects. INTRODUCTION However, in an effort to retain the survey capabilities of MODEST, a process was tested With the modernization of the Michigan Orbital that allowed the retention of some survey data DEbris Survey Telescope (MODEST) in March while doing follow-ups during parts of the 2005, observations in modes other than tracking evening. at the sidereal rate were possible. This allowed for efforts to commence determining the These first efforts are based on determining environment more accurately by obtaining orbits for geosynchronous orbit (GEO) objects accurate orbits on all types of objects. The best which are in circular or near-circular orbits (with method through which one would determine an eccentricities less than 0.05). There are three orbit of an object is to track it specifically. reasons for doing so: 1 • A GEO object that starts in a circular or low objects were circular uncorrelated targets eccentric orbit will remain in such an orbit, (UCTs) and the other three were correlated unless it has a high area-to-mass (A/M) and targets (CTs) with circular orbits. is subject to significant solar radiation forces. Finally, in January 2006, a “survey and chase” • Circular orbits are the simplest to predict run was conducted similarly to the October 2005 and follow-up. run where the survey was completed on night • There exists a large number of well- one and the follow-up observations were observed, bright objects in the public United conducted on subsequent nights. By acquiring a States Satellite Catalog at GEO which are on 24 hour arc of data, it was possible to use the near-circular orbits and can serve as truth non-constrained eccentricity orbit propagator tests. How well do our derived orbits agree (referred to here as the non-constrained-ecc orbit with the published ones? propagator) to calculate the orbit. Unfortunately, none of the objects recovered Our long-term goal is to construct a system were UCTs. which can detect and follow-up an object in any orbit at GEO. This paper reports our first steps Our results on the frequency of follow-up towards that goal. observations to determine an orbit from a short time arc survey observation agree in general The new method of observing entails using with those reported earlier by the ESA debris survey mode detections over a 5.3 minute time- group using the ESA Space Debris Telescope span and propagating the orbits to predict (SDT) on Tenerife1, 2. One significant difference specific right ascensions and declinations at is that our initial observation period can be up to future times. During the follow-up observations, twice as long as obtained with the ESA SDT . longer time arcs are obtained and thus more accurate orbits are calculated. To date, survey MODEST and chase has been used on three observing runs. Since early 2001, the University of Michigan Beginning in July 2005, MODEST surveyed for and NASA have been using MODEST to the first few hours of the night. Data was observe orbital debris in the GEO environment. reduced in real-time at the telescope and sent to MODEST is a 0.6/0.9-m Schmidt telescope the National Aeronautics and Space located at the Cerro Tololo Inter-American Administration (NASA) Orbital Debris Program Observatory (CTIO) in Chile. A brief Office at Johnson Space Center (JSC) in description of the system follows: for more Houston, Texas. Using assumed circular orbit details see References 3 and 4. (ACO) predictions, successful follow-up observations were conducted on those The telescope is equipped with a thinned 2048 x previously detected objects on the same night. 2048 pixel charge coupled device (CCD) with a With only an orbit propagator using a field-of-view (FOV) of 1.3 degrees square and constrained eccentricity (referred to here as the 2.318 arc-second pixels. In a 5 second exposure constrained-ecc orbit propagator) available, two through a broad R filter a limiting magnitude of R targets were recovered both with circular orbits. = 18th is reached with a signal to noise (S/N) of 10. Our next try was during the October 2005 run. In survey mode, the telescope tracks at the The entire first night was observed in survey sidereal rate a fixed right ascension (RA) and mode. Using an ACO prediction, MODEST declination (DEC) point close to the anti-solar was used in follow-up mode on the second and point and outside of Earth shadow. During the third night specifically targeting the fields where exposure the charge on the CCD is shifted the projected orbits placed the objects. On night backwards such that GEO objects appear as one, nine objects were detected, of which five point sources or short streaks, and stars appear were reacquired on the later nights. Two of the drift as fixed length streaks. In the 5.3 minutes 2 Fig. 1: An example of a survey detection sequence from MODEST. Each sub-image is 2.5 arc-minutes square. The GEO object is at the center of each sub-image, while the horizontal streaks are stars. it takes a station-keeping GEO object to across object increases if the first follow-up the FOV, up to 8 independent positions can be observations are conducted within one hour of measured. A minimum of 4 detections are the initial observation. The fields where the required in this 5.3 minute window for a real object is predicted to be are calculated for detection. All correlated objects are visually every minute up to either four or eight hours examined to guard against false detections. after the initial detection. Following the results of the second test run in October 2005, the user A circular orbit works well for most GEO objects, has an option to determine the orbit using a but for objects such as those in a GEO-transfer circular orbit assumption (fixing the orbit (GTO) or with high A/M ratios this process eccentricity) or to allow the eccentricity to be is not ideal. By using follow-up observations, determined. As mentioned previously, if the either later in the same night or the next night, a actual orbit is eccentric the time arc of five longer time arc can be acquired such that a more minutes may not be sufficient to determine the accurate orbit is obtained. Figure 1 shows a orbit accurately using ACO. sample detection sequence of a GEO object. The initial orbital parameters are calculated The system is designed to detect objects with using initial radius and velocity vectors in angular rates between +/- 2 arc-seconds/second equatorial coordinates (Earth-centered inertial) in hour angle (HA) and +/- 5 arc-second/seconds based on a method shown in detail in Reference in declination. 5. The program uses 100 Monte Carlo simulations to determine the best orbit, and ten Data is reduced in real-time at the telescope and possible orbits are sent to the prediction file. time stamped positions of survey objects are sent For each orbit a residual is calculated, which is to NASA JSC in Houston, Texas, where the a goodness of fit of the orbit to the detections. orbit and predictions for future location are The fields where the object is predicted to be calculated. are computed by using the orbit with the smallest residual. These predicted fields are With such a short-arc observation (5.3 minutes listed for every minute over the user specified out of a total orbital period of 1440 minutes), time period. Observations can be added to the determination of a full parameter orbital solution original set of detections to obtain a new is extremely uncertain, and hence the need for prediction field for later in the same night or follow-up (chase) observations to determine the another night. orbit accurately. The standard technique is to calculate a constrained-ecc (or circular) orbit While calculating the orbit, the program based on the first 5.3 minutes of data. determines whether or not the object is in shadow at the given time step. If it is in ORBIT PREDICTION shadow, those fields are marked with the words “In Shadow” and are not sent in the prediction The orbit prediction process determines the RA files to the telescope. and DEC of the object at a specific time beginning 30 minutes after initial detection or In these data sets presented in this paper, no anytime thereafter. However, depending on the correlation of the object to the catalogue is done orbit of the object, the likelihood of finding the while the observations are being conducted. To 3 validate the prediction process, correlation with objects were detected and follow-up the United States Satellite Catalogue was measurements were conducted two times within conducted after the observing run was that same night. After the observing was completed. completed for the night, a correlation with the catalogue found all three of the objects detected CHASE MODE were correlated objects and were in circular orbits. Predictions of where objects will be are sent from Houston to the telescope in Chile as The second night in this test run (DOY 192) quickly as possible. Turn-around time from found six objects for follow-up. Four of the six receipt of initial survey positions in Houston to objects were found on the first follow-up of the receipt of predicted future positions at the evening and three of those were found on a telescope was generally less than 1 hour. second follow-up attempt. The three found on both follow-ups were all CTs and were in Our initial experiments in follow-up (or chase circular orbits; however, the one that could not mode) used the same observing technique as be found on the second attempt was a UCT. survey mode: the telescope tracked a predicted field at sidereal rate and the CCD was drift The third night (DOY 193) of the July 2005 run, scanned backwards to remove the effect of the we found four objects and all four were tracking. This allowed us to use the same reacquired two times during that night. Three of observing and reduction software as survey the objects were circular orbit CTs while one of mode, at the cost of observing time. the objects was a circular orbit UCT. Given a list of an object’s predicted position and The July run proved that with a fast turn of less time, the telescope was set at the position than 30 minutes, GEO objects can be reacquired approximately 7 minutes ahead of the current with a predicted orbit. Since all of the objects starting time. Thus the object would drift across found were in circular orbits, the time from the the center of the field about 10 minutes after the last observation to the first follow-up sequence started. This technique is largely observations needs to be quicker to find insensitive to both timing and along-track errors, eccentric orbit objects. Due to the fact all of the making it an excellent choice for our first objects had absolute magnitudes of brighter than experiments. Future observations will have the 16.5, it is likely that the objects that were not telescope come up on the object’s predicted found had eccentric orbits although it could be a position, tracking at the predicted rates. brightness variation from the objects as well. Even with a variation of 1.5 in magnitude, the As per survey observations, a minimum of four object would still have been within the limiting measured positions were required on these chase magnitude of the system. In addition, a study on observations for an object to be considered a real the reacquisition of an eccentric orbit object detection. when using ACO is discussed in the future observation section. RESULTS October 2005 July 2005 Both the October 2005 and January 2006 runs The July 2005 trial was conducted within a night were conducted over multiple nights with the and was the first test to see if the turn around first night being used for normal survey mode. time of the observations was short enough to The October 2005 run covered DOYs 278 – 280. reacquire objects. The first two hours of each In the first night, nine objects were detected. night were used for normal survey mode and Each of objects was propagated to the next night follow-up objects were determined from those (DOY 279) using an ACO. This was done as a observations. On day of year (DOY) 189, three test of the propagator to see if 24 hours was too 4 long to recover objects. Thirteen objects were constrained orbits were tested and observed observed the second night, although only nine during this run. It was found though that the were observed the previous night. This posed a non-constrained-ecc orbit propagator solution problem. How can one tell which objects were proved to find a better orbit even when the object linked together? Without correlating the objects, was in a near circular orbit. The third night of each detection on the second night was the October run found multiple objects in the combined with the detections from the first predicted fields although only five of the objects night, and the orbits were calculated and sent to seemed to be good matches with low residuals. the telescope for observations. It was noted which set of observations was more likely the Post-observation correlation showed that three true linkage based on the residual from the orbit of the five objects were CTs and two objects fit to the observations; however, both sets of were UCTs. All five objects had circular orbits. fields were observed. DOY 278 Residual DOY 279 Residual Residual After the January 2006 data collection, it was objects circ objects circ ecc determined that the orbit with the lowest 0001 1.69 0001 2.449 1.516* residual was usually the non-constrained-ecc 0002 2.261 0002 836.4 err - div by 0 orbit propagator solution and the priority was 0003 1.279 0003 41.11 29.79 changed to reflect that. However, for this run, 0004 2.047 1.847* both predictions were sent to MODEST. The 0004 0.9743 0005 4.354 1.782* residuals are shown in Table 1. In the table, the 0006 26.48 25.99 first column shows the detections for DOY 278 0005 1.02 0007 1.531 1.180* (objects 0001 – 0009). To explain the table 0006 0.9334 0008 4.489 3.196* better, an example is the first object, 0001, 0009 285.5 274.8 which on DOY 0001 had a residual from a 0007 0.8841 0010 3.28 2.111* circular orbit of 1.7. The predicted orbit was 0008 1.393 0011 2.331 1.855* propagated to DOY 279. Observations from 0012 162.6 162.3* that field showed an object detected (0001). 0009 0.8802 0013 5.618 2.104* Combining those two observations yields a Table 1: Residuals for night 1 and night 2 of October residual on the constrained-ecc orbit of 2.5 and 2005 data collection. The objects with the on the non-constrained-ecc orbit of 1.5. Both asterisk in the eccentric column depict the orbits orbit types were sent to MODEST for further that are still circular. The object numbers do not inspection. Another example displays the correlate with each other but are the order in which they were taken within the night. results when two objects are found in the predicted field. On DOY 279, two objects were found at object 0003’s predicted location January 2006 (objects 0003 and 0004). Both objects were combined independently with object 0003 and The final testing run was conducted in January predicted field locations were determined and 2006. During this run, the priority scheduling sent to MODEST. was used for the first time. The JSC staff determined the priority of the follow-up It was quickly apparent that the naming observations based on likelihood of future convention of these prediction files would need detection (shadow, predicted locations, etc.), the to be examined. A system in place now keeps goodness of fit of the orbit, and the magnitude of track of the objects in order of observation so the object. In these data shown in this paper, no that when the run is finished the lineage of the correlation of the object to the catalogue is done detections would be preserved. while the observations are being conducted. From prior experiments, fainter objects are more Because of the longer time arc (~24 hours), an often UCTs than CTs so it is assumed in this non-constrained-ecc orbit could be calculated study that a fainter object is more likely to be a with accuracy. Both the non-constrained and UCT rather then a CT. 5 Object Year DOY object # Number of Obs ECC Inc MM RAAN A 2006 031 0001 1 0 7.3951 0.995396 67.839 A 2006 032 0102 2 1.29E-03 7.3652 0.99759 67.583 A 2006 032 0210 3 1.01E-03 7.3686 0.997588 67.6212 A 2006 032 1012 4 9.83E-04 7.369 0.997587 67.6269 A TLE Prediction 8.42E-04 7.35 0.9976 67.45 B 2006 031 0011 1 0 5.0908 1.007234 79.6896 B 2006 032 1103 2 7.71E-04 5.0697 1.007703 78.992 B 2006 032 0311 3 1.09E-03 5.0686 1.007676 78.976 B 2006 032 1113 4 9.86E-04 5.0695 1.007663 78.9807 B TLE Prediction 7.36E-04 5.05 1.0077 78.85 C 2006 031 0015 1 0 4.9342 1.001831 81.1507 C 2006 032 1521 2 3.25E-03 4.8742 0.99886 78.4943 C TLE Prediction 6.77E-04 4.86 0.9988 78.42 Table 2: Predicted Orbits for three objects observed in January 2006. As the number of observations of the object increases, the accuracy of the orbit increases. To combat the issue of multiple objects found in node (RAAN), respectively. In each of the the predicted field, each object is assumed to be figures, the final value is the predicted value the object in question until the residuals are from the two-line element set and is shown as examined. If the residual of the orbit, or an open symbol and is offset to ensure goodness of fit, was low it was determined that visibility. it was the proper object. If the residual was high, the results were still sent to MODEST but Eccentricity, seen in Figure 2, has the largest the priority was set to a much larger number increase in accuracy when more observations (priority of 1 is high priority and priority of 9 is were added and the observed orbital arc was low priority) to ensure that all viable extended. In the first observation set, the observations were taken first. On the normal eccentricity is basically set to zero, but as one survey mode of data collection (DOY 030), 10 adds more observations the eccentricity can be objects were detected. The following day (DOY defined and refined. The inclination, shown in 031) field locations based on an ACO were sent Figure 3, shows the least amount of change in to MODEST. Objects were found in six of the accuracy by adding observations, although there 10 fields. It is possible that by not finding an is convergence toward the predicted value. object where the circular orbit predicts it to be Even though these objects are in circular orbits, that those objects were not in circular orbits. adding observations lends an increase in This information can be used in future accuracy for MM for all three objects as seen in observations by widening the search area for the Figure 4. The final orbital parameter, RAAN object. (shown in Figure 5), has less of a change in value when adding more observations. Like Orbit Comparisons inclination, this means that the RAAN over a small time arc is likely accurate for dealing with A comparison of the orbits using the circular nearly circular objects and both values may be orbit assumption and the eccentric orbit unbiased by ACO for objects with moderate or propagator was conducted on three objects from high inclinations. the January 2006 observations. The results are shown in Table 2 and in Figures 2, 3, 4, and 5 for eccentricity (ecc), inclination (inc), mean motion (MM), and right ascension of ascending 6 3.50E-03 85 3.00E-03 Obj A TLE Obj A 80 2.50E-03 Obj B Obj A TOLbEj C Obj B grees75 TOLbEj B Obj A 2.00E-03 TLE Obj C de TLE Obj B CC N, Obj C E 1.50E-03 RAA70 TLE Obj C 1.00E-03 65 5.00E-04 60 0 1 2 3 4 0.00E+00 Number of Observations 0 1 2 3 4 Fig. 5: Number of Observations versus RAAN, Number of Observations degrees. Same description of symbols as Figure 2. Fig. 2: Number of Observations versus the Eccentricity. Each filled point represents an FUTURE OBSERVATIONS observation. TLE value (open symbol) is offset from the last observation to ensure visibility. Due to the long-term goal of constructing a system which can detect and follow-up an object 8 in any orbit at GEO, a system must be designed 7.5 to obtain observations on objects which are not 7 Obj A circular. A subset of CT eccentric orbit objects s egree6.5 TOLbEj B Obj A detected previously with MODEST during n, d 6 TLE Obj B survey mode were propagated for times of 30 nclinatio5.5 OTLbEj C Obj C mobisneurtveast ionth. r oTuhgihs tes8t ingh owuarss depsaigsnt edt htoe selea sift I 5 there was a systematic search pattern that could 4.5 be followed after the survey observations were obtained to maximize the possibility of follow- 4 0 1 2 3 4 on observations as well as the necessary hand- Number of Observations off time to ensure observation. When the code Fig. 3: Number of Observations versus Inclination, could compute an eccentric solution on the short degrees. Same description of symbols as arc from the survey, both solutions are provided, Figure 2. however, for some of the objects the eccentric solution could not converge on an orbit and thus 1.01 only the circular orbit prediction was calculated. 1.008 1.006 Obj A It was assumed that the telescope would be TLE Obj A pointed at the RA and DEC at the specific time on1.004 Obj B Moti1.002 TLE Obj B of the circular orbit prediction. Then, it was n Obj C determined how long the object would have a e TLE Obj C M 1 been in the FOV. Two objects are shown here 0.998 as representative examples of the subset tested. Object A had an eccentricity of 0.18 and object 0.996 B had an eccentricity of 0.48. Both objects were 0.994 observed with MODEST during normal survey 0 1 2 3 4 Number of Observations operations. Both figures show the circular orbit predicted (red line straight line), the eccentric Fig. 4: Number of Observations versus Mean orbit prediction (green line, or top curved line), Motion, degrees. Same description of symbols the “true” orbit based on the two-line element as Figure 2. (TLE) predictions (blue line, or the shortest 7 line), and the actual observations (black line at Figure 8 shows object B with the circular, the beginning of the UT). eccentric, and TLE predicted orbits, and actual observations. Object B has an eccentricity of Figure 6 shows the predicted RA and DEC for 0.48. If the telescope had been centered on a given time for the circular, eccentric, and this object at the location predicted by the TLE orbits for an object (object A listed as obs) circular orbit, we would have never seen the with eccentricity of 0.18. The object would object in the 30 minute turn around assigned have been detected with the circular orbit for this project as seen in Figure 9. In general, assumption up to 1.35 hours from the first half of the eccentric objects tested would have prediction and 2.5 hours after the first been recovered as long as the observations detections. This is shown in Figure 7. After took place within 30 minutes of the initial 2.5 hours after the initial detection, the TLE observation. This study showed that if survey prediction states the object would have gone observations are to continue as well as follow- into shadow, but prior to entering the shadow up observations, two telescopes will be the likelihood of detecting this object would necessary so that the hand-off between the have been good. systems is as fast as possible and definitely within 30 minutes of the initial observations. Fig. 6: Object A Universal Time versus RA and DEC. Fig. 8: Object B Universal Time versus RA and DEC. Fig. 7: Image of the FOV after 1.35 hours after the first prediction (2.5 hours after detection) showing Fig. 9: Image of the FOV after 30 minutes showing the “true” position of the object is still in the FOV the object would not be in the FOV if a circular if a circular orbit is assumed. orbit was assumed. 8 CONCLUSIONS REFERENCES The process of using follow-up (chase) 1. R. Musci, et al. Orbit Improvement for GEO observations to obtain a longer arc on a specific objects using follow-up observations, object was successful. By using a quick turn Advances in Space Research, 34, 912, 2004. around method of orbit determination and propagation, one can obtain better orbital 2. R. Musci, et al. Orbit improvement for GTO parameter accuracy. Real-time detection is objects using follow-up observations, possible with MODEST using fast reacquisition Advance in Space Research, 35, 1236, 2005. of an object. 3. P. Seitzer, et al. A Survey for Space Debris It was surmised through these trial runs that in in Geosynchronous Orbit in the Proceedings order to obtain eccentric orbit objects, the orbits of AMOS 2001 Technical Conference, for the follow-up observations would need to be Maui, HI, 2001. updated as new observations of the object arrived. To obtain eccentric objects, one hour 4. P Seitzer, et al. Results from the was too much time between the initial NASA/Michigan GEO Debris Survey, the observations and the first follow-up Proceedings of AMOS 2004 Technical measurement. Yet there exists a population of Conference Proceedings, Maui, HI, 2005. faint objects at GEO which are on eccentric orbits, as determined by the ESA SDT6, so 5. R.R. Bate, D.D. Muller, and J.E. White. determining eccentric orbits is not just an Fundamentals of Astrodynamics, Dover academic question. Publications, New York, NY, 1971. In addition, the bookkeeping of what object was 6. T. Schildknecht, et al. Optical Observations matched with what follow-up observation can be of Space Debris in High-Altitude Orbits, in very confusing when observing multiple fields the Proceedings of the Fourth European on multiple nights. This needs to be tracked Conference on Space Debris, Darmstadt, carefully so that the validity of the orbital Germany, 2005. linkages are preserved. The residuals increase when adding more observations and so depending on a specific value of the residual to determine the goodness of fit of an orbit is not appropriate. As more testing commences, the values will be tabulated and a priority will include the number of observations as well as the magnitude, location, and residual. In the future, more correlation studies within a night and night to night will be conducted using these methods and building on the knowledge obtained. Future work on this project would be to continue to use MODEST in survey mode and have a second telescope dedicated to the follow- up observations. In this fashion, the knowledge of the statistical environment is gathered as well as orbital elements and photometric properties on specific objects. 9