The Second International Asteroid Warning Network Timing Campaign: 2005 LW3

The Earth close approach of near-Earth asteroid 2005 LW3 on 2022 November 23 represented a good opportunity for a second observing campaign to test the timing accuracy of astrometric observation. With 82 participating stations, the International Asteroid Warning Network collected 1046 observations of 2005 LW3 around the time of the close approach. Compared to the previous timing campaign targeting 2019 XS, some individual observers were able to significantly improve the accuracy of their reported observation times. In particular, U.S. surveys achieved good timing performance. However, no broad, systematic improvement was achieved compared to the previous campaign, with an overall negative bias persisting among the different observers. The calibration of observing times and the mitigation of timing errors should be important future considerations for observers and orbit computers, respectively.


Introduction
The International Asteroid Warning Network (IAWN) 72 was established to create an international group of organizations involved in detecting, tracking, and characterizing near-Earth objects.The IAWN has conducted multiple observational campaigns to exercise the preparedness of the international community for the different components of the mitigation response of a potential impact hazard: discovery, recovery and tracking, physical characterization, and risk assessment (Reddy et al. 2019(Reddy et al. , 2022a(Reddy et al. , 2022b)).
In order to assess the quality of the observational data used in the orbit determination process, the 2021 campaign targeted asteroid 2019 XS during its close approach to Earth (Farnocchia et al. 2022).Because of the asteroid's high rate of motion in the sky, timing errors significantly contributed to the astrometric positional errors in the plane of sky.Timing errors can be due to several reasons, such as infrequent or unsuccessful synchronization, a delay between the command to open the shutter and the actual motion of the device, a finite travel time of the shutter over the focal plane, numerical rounding, or confusion between the exposure start time and mid-time.By breaking down the plane-of-sky positional errors in the along-track (i.e., aligned with the plane-of-sky motion) and cross-track (i.e., perpendicular to the plane-of-sky motion) directions, Farnocchia et al. (2022) assessed the accuracy of the time of the observations reported to the Minor Planet Center (MPC).The timing errors were typically smaller than 1 s.However, there was an overall negative bias; i.e., the reported observation times were earlier than they should have been, on average.
Thanks to the possibility of reporting astrometric uncertainties using the ADES format (Chesley et al. 2017),73 a second outcome of the 2019 XS campaign was that of validating the reliability of the reported uncertainties by comparing them to the cross-track errors, which are not affected by timing or trailing and therefore reflect purely positional errors.A fraction of the reported uncertainties appeared to be optimistic, especially when smaller than 0 2, thus suggesting that some sources of error are not fully captured in the uncertainty estimation process.
To address both timing accuracy and uncertainty quantification issues, Farnocchia et al. (2022) recommended calibrating observation times against GNSS satellites74 and presented a recipe to derive realistic, conservative uncertainties.
Near-Earth asteroid 2005 LW3 was discovered in 2005 June by the Siding Spring Survey (MPEC 2005-L19). 75The Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.asteroid, with an inferred diameter between 100 and 300 m based on its absolute magnitude H = 21.7, made a close approach to Earth on 2022 November 23 at a distance of 1.1 million km (about 3 lunar distances; see top left panel of Figure 1) from the geocenter.Figures 1 and 2 show the observing conditions for 2005 LW3 around the close approach.
The asteroid was bright enough (bottom right panel of Figure 1) to be easily observable, moving rapidly in the sky (albeit not quite as fast as 2019 XS), and more accessible from the Northern Hemisphere (Figure 2).We aimed to conduct a timing campaign in the second half of 2022, and 2005 LW3 was the only candidate with a plane-of-sky rate of motion greater than  2″ s −1 (bottom left panel of Figure 1) that could be observed after the encounter (high elongation and bright, as shown by the top right and bottom left panels of Figure 1, respectively), thus constraining the orbit enough to reliably analyze the timing errors.We therefore selected 2005 LW3 as a target for a second IAWN timing campaign. 76

Observations
Table 1 lists the 82 stations that participated in the campaign, and Figure 3 shows their geographical location.The broad geographical coverage is encouraging in terms of ensuring that observers are available in case of short observing windows that limit favorable observing conditions to only specific locations.Thirty-eight stations participated in both the 2019 XS campaign and the 2005 LW3 one, while 33 stations only participated in the 2019 XS campaign, and 44 stations only participated in the 2005 LW3 campaign.
We observed 2005 LW3 from 2022 November 21.9 UTC to 25.0 UTC, i.e., the days around the closest approach, when the object was moving faster in the sky.The earliest campaign observations were collected by SONEAR Observatory in Brazil, and a total of 1046 astrometric observations of the plane-of-sky position of 2005 LW3 were reported by campaign participants to the MPC.The observations can be found in the MPC's Daily Orbit Update Minor Planet Electronic Circulars (MPECs) 2022-W151, 77 2022-W161, 78 2022-W171, 79 2022-W190, 80 2022-W220, 81 The majority (664, i.e., 63%) of these observations were reported using the ADES format, thus including uncertainty information.The full ADES records can be accessed using the xml version of the abovementioned MPECs. 88 Table 2 shows the star catalogs used to reduce the astrometric observations from this campaign.Of the observations, 89% were reduced using one of the Gaia star catalogs or ATLAS-2, which uses Gaia sources, thus removing any concern for star catalog biases (e.g., Eggl et al. 2020) affecting the accuracy of the measured astrometric positions.Older, less accurate catalogs should be deprecated for the purpose of astrometric reduction.

Analysis of the Residuals
We analyzed the astrometry collected as part of the campaign against JPL solution 27 (Table 3).This orbital solution for 2005 LW3 is based on the optical data reported to the MPC through 2023 January 8. 89 To keep timing errors from biasing the orbital solution, around the November 22 close approach, we only selected single detections from J04, T12, D03, and V28.These four stations calibrated their timing to better than 0.1 s by observing GNSS satellites 90 prior to observing 2005 LW3.Finally, we included radar observations collected between 2022 November 22 and 27. 91 We projected the astrometric residuals in the along-and cross-track directions, i.e., in the directions parallel and perpendicular to the motion of 2005 LW3 in the sky.The exact procedure is presented in detail in Farnocchia et al. (2022) and is therefore not repeated here.The cross-track direction is not affected by timing errors and so is informative about the purely astrometric positional errors.For each tracklet,   4 shows the weighted rms, which is the rms of the residuals normalized by the observer-provided uncertainty.A weighted rms of 1 means that the claimed uncertainty exactly matches the noise in the data, a weighted rms < 1 means that the reported uncertainties are overestimated, while a weighted rms > 1 means that the noise in the data is not captured by the reported uncertainties.As already found for 2019 XS (Farnocchia et al. 2022), the reported uncertainties tend to become more and more optimistic (up to a factor of 10) as they become smaller, especially when <0 2. The implication is that some error sources may not be fully captured in the uncertainty quantification.
The timing error is the along-track residual divided by the rate of motion.The right panel of Figure 4 shows the timing error of each tracklet of observations collected as part of the campaign.As already noticed for the 2019 XS campaign (Farnocchia et al. 2022), most tracklets have timing errors within 1 s.For 28 stations, there is no measurable time bias, while 26 stations have biases within 0.5 s.Seven participating stations have timing errors exceeding 1 s, though one of them has large and biased cross-track errors, so positional errors could fully explain its along-track errors without invoking timing errors.Overall, there is still a negative time bias of around −0.2 s, which means that, on average, the reported times are 0.2 s earlier than they should be.
Despite the persistence of an overall time bias, it is worth noting that 14 stations managed to significantly improve compared to the 2019 XS campaign, while a couple stations performed worse than in the previous campaign.As an example, Leo Observatory (station code V17) did not report uncertainties and had a −0.6 s timing bias for the observations of 2019 XS during the 2021 close approach.For this campaign, V17 adopted the recommendations by Farnocchia et al. (2022) to quantify the uncertainties and calibrate the timing using GNSS satellites, which helped improve the observation accuracy.The V17 results can be seen in Figure 5.The right panel shows the cross-track residuals of V17, which are small and consistent with the uncertainties as reported by the observer using the ADES format.The right panel shows the timing errors, which are statistically consistent with zero.The first five astrometric positions were derived by stacking multiple images, while the last five were derived from single images.Due to the brightness of 2005 LW3, stacking was not necessary, but measuring positions using both approaches shows that no timing error is being introduced by the stacking procedure.When stacking, it is important to keep in mind that separate sets of exposures should be used for different stacks.
Surveys in the United States generally achieved good timing performance (see Figure 4).The Catalina Sky Survey (Christensen et al. 2018; station codes 703 and G96) has errors of the order of a few hundredths of a second, statistically consistent with zero.The same is true for Pan-STARRS (Wainscoat et al. 2022; station code F52), which represents a significant improvement over the −0.5 s bias observed during the 2019 XS campaign.The ATLAS survey (Tonry et al. 2018) has errors between −0.1 and −0.2 s for the stations in Hawaii (codes T05 and T08) and essentially no bias for the two more recent stations in South Africa (code M22) and Chile (code W68).

Discussion
Timing errors become a key consideration when observing objects close to Earth that move rapidly in the sky.The faster the object, the larger the astrometric position displacement error induced by a timing error.If not mitigated, timing errors affect the accuracy of orbits determined by fitting these observations.The effect becomes even more severe if the timing errors are systematic, not only for individual tracklets but even across different stations and nights.
Measures can be taken as part of the orbit determination process to mitigate timing errors.For example, Farnocchia et al. (2022) described how to augment an observation's covariance matrix to account for a time uncertainty.For this very reason, the ADES format includes fields that observers can use to communicate estimated random and systematic errors in time and other fields to be used by orbit computers to set time biases and weights.Therefore, debiasing and weighting schemes can be developed based on statistical analysis of the residuals analogous to the statistical schemes devised for the purely positional terms of the astrometric errors (Vereš et al. 2017;Eggl et al. 2020).However, it remains preferable for timing problems to be addressed at the source, with observers correcting for time biases directly, so that the information from astrometric observations can be fully leveraged.To support the quantification and resolution of these biases, we provided each participating observer with a detailed report of the results from this campaign.
Compared to the first timing campaign targeting 2019 XS (Farnocchia et al. 2022), some observers were able to report more accurate observation times.However, there was no clear, systematic improvement, and an overall negative time bias persists.This is especially true for follow-up observers.Thus, more work is needed to improve timing accuracy, e.g., by ensuring that observers take full advantage of tools like Project Pluto's GNSS page 92 to properly and regularly calibrate their timing.
A similar conclusion holds about the reliability of astrometric uncertainties.It is encouraging to see a larger fraction of the observing community adopting the ADES format and reporting their uncertainty estimates, which provide key information for orbit computers to properly weight the data.However, such uncertainties can prove optimistic, especially when <0 2. Astrometric weighting schemes will need to account for this issue, e.g., by incorporating some (possibly station-specific) minimum credible error.
Future campaigns will be announced on the IAWN website 93 and possibly with an editorial notice by the MPC.

Figure 1 .
Figure 1.Geocentric and heliocentric distance (top left), solar elongation and phase angle (top right), plane-of-sky rate of motion (bottom left), and V-band magnitude (bottom right) of 2005 LW3 as a function of time.The vertical lines correspond to the times of perihelion and closest approach to Earth.

Figure 2 .
Figure 2. R.A. and decl. of 2005 LW3 as a function of time.The thickness of the line is related to the brightness of the object.

Figure 3 .
Figure 3. Locations of the 82 ground-based observation sites that participated in the campaign.

Figure 4 .
Figure 4. Left: weighted rms of the cross-track residuals of each tracklet as a function of the astrometric position uncertainties.Right: estimated mean timing error for each tracklet of observations considered in our analysis.Crosses correspond to survey data, and dots correspond to follow-up data.Three outliers with timing errors between 2 and 4 s fall outside the plot.The two dashed lines correspond to the orbital uncertainty mapped into the along-track direction in the plane of sky and converted to time using the rate of motion.

Figure 5 .
Figure 5. Cross-track (left panel) and timing (right panel) errors (crosses and dots) and 1σ error bars for a tracklet of observations collected by V17 during the campaign.The two dashed lines correspond to the orbit solution uncertainty in the along-track direction scaled by the rate of motion.Crosses correspond to positions obtained by stacking images, while dots are for positions measured from single images.