Saturday, 10 March 2012

Finding needles in haystacks, minor planet style

Early in February the Minor Planet Center's (MPC) Bright Recovery Opportunities page was listing NEO 2006 AL8 as a bright, well placed but fast moving evening target, ideal for a recovery attempt.

It had been discovered on 08 Jan 2006 by Gordon Garradd and Rob McNaught from Siding Spring and then followed for 7 weeks. With a period of 6 years, this was its first return to the Sun since discovery and unfortunately the 3-sigma uncertainty in the predicted ephemeris position by 02 Feb 2012 covered a massive 54° of sky, 33° to the east of the nominal position and 21° to the west. However it was also predicted to be 16th mag. and moving at about 27"/min, so if it could be caught in the telescope field of view it ought to be a very noticeable moving object. The Moon was due to be full on Feb. 8th and so many fainter NEO targets were already getting blotted out by the strengthening moonlight, searching for 2006 AL8 seemed like a good way to use an otherwise not particularly useful sky.

First step in the search was to retrieve all the existing astrometry for the NEO via the MPC's MPC Database option, one of the standard options on the left of most of the MPC's web pages. This astrometry was then pasted into Bill Gray's Find_Orb orbit determination program to generate an orbit and have a look at the residuals from the original astrometry. One of the functions of the program allows the filtering out of any astrometry with residuals larger than a set maximum and this was done over a series of steps, starting with a fairly generous limit of 2 arc-seconds, then gradually reducing to a harsh 0.75 arc-seconds, causing a number of the positions to be ignored in the orbit determination. Care has to be taken here, the intention is to only remove positions with large random errors and pushing the limit too small can start to drop good positions and degrade the quality of the solution.

The resulting orbit was then used to generate an ephemeris for the current night and this was compared with the MPC's nominal prediction which was generated from an orbit which only excluded positions with residuals in excess of about 2 arc-seconds. The Find_Orb prediction placed 2006 AL8 to the east of the MPC prediction, but only a couple of degrees from the MPC place. A few other orbits with some different selections of positions were calculated and all placed the NEO to the east of the MPC position, with relatively little scatter.

At 18:28 UT on 02 Feb I started my search at the MPC nominal position and then used the hint from Find_Orb to move east along the line of variation, taking 21 x 4 second exposures of each overlapping field. As each set of images was being exposed they were blinked using Astrometrica, looking for the moving object and in the 6th field 2006 AL8 was glaringly obvious, moving west to east at mag +16.3R, 1.2° from nominal.

NEO 2006 AL8 recovery field on 2012 Feb 02
NEO 2006 AL8 can be seen to the lower right, moving right to left in this video made from 21 images, each frame a 4 second exposure.

It had only taken 25 minutes from taking the first image to recover it. Another set of images was obtained the next night and the two nights of astrometry were sent off to the MPC at 18:40 UT on 03 Feb. Within the hour the MPC published MPEC 2012-C17 announcing the recovery.

Large uncertainties like the 54° in this case need not be quite as daunting as they might initially seem. If the object is bright enough to be seen in individual images and careful handling of the available astrometry is done, in the majority of cases the correct side of an uncertainty area can be chosen, halving the potential search and often also helping to give a reasonable clue to the whereabouts within the remaining uncertainty area. The same process often helps with newly discovered NEOs on the NEO Confirmation page when uncertainties are large, though often this is made more difficult when there are just a handful of positions for a new object, making it impossible to decide which positions are more in error than others, compared with the recovery of an existing object where there may be dozens or even hundreds of positions available.