It all started very early in the morning on December 21st, sitting with Maia in my parents' living room, trying to get her back to sleep. We were watching Early Today, when they told the story of 2007 WD5, the asteroid that might hit Mars on January 30th. They said that the probability of impact had already been upgraded from 1-in-350 (0.3%) to 1-in-75 (1.3%), but that the orbit couldn't be refined further until more observations were made in early-to-mid January.
I had been planning to work about 4 hours that day anyway, so I decided to look into what was already known about this asteroid. I found that the orbit was based on 25 observations spanning 29 days (from November 20th to December 19th). From my past experience with refining orbits, I knew that there was barely enough time remaining before the potential impact to get an accurate orbit. What we know about an orbit depends strongly on the length of time covered by the observations, and as we extrapolate beyond this "observed arc" the uncertainty in our predictions grows and grows, especially if that arc is small. I knew that one way to refine the orbit without waiting until January would be to find "precovery" observations that had gone unrecognized in images taken prior to the discovery date (Nov 20). I also had access to and experience with the Sloan Digitial Sky Survey data... this was essentially the same work that I did for my PhD thesis. So I did a little exploring.
As luck would have it, the object had passed through the area of the sky repeatedly scanned by the SDSS-II Supernova Survey on November 8th & 9th, just as it was in the middle of its close approach to the Earth. It was moving along at a pretty good clip at the time (10 arcseconds/minute), so I knew that I would be looking for a fairly long streak in the images. That is, if it was there at all! Of the three runs imaged on those two days, only one was taken in the right place at the right time to catch the asteroid. Figuring this out was a fair amount of work, because the long thin stripes of the SDSS runs can take 8 hours or more to be imaged from beginning to end, and the asteroid would move more than a degree in that time. Plus, it had been a while since I had downloaded the Supernova Survey data to the University of Chicago computer where I run my code, so I had to figure it out by hand.
In the end, I narrowed the likely position to one field in one run of the SDSS data, so first I looked at that image. Here's what I found:
(Thanks to Steve Kent for putting together this "Publicity Shot," which is currently the SDSS Image of the Week.) The three colored streaks come from three separate images taken through different colored filters. I was sure I had the right object because the length of the streaks, their angle across the sky, and the rough brightness all matched the predictions. However, it was about 55 arcseconds away from the predicted position! This was how I knew that I would have a significant effect on the best-fit orbit.
I fed the calibrated images through Astrometrica to get positions for each of the streaks, and then corrected the observation times for their positions within the field. I submitted this info to the Minor Planet Center, and that was the end of my involvement! The MPC is the clearinghouse for data like this, and they calculate and publish orbits for more than 300,000 objects every day. I knew that the folks at JPL would get my data from the MPC, and calculate new odds for the Mars impact. I also knew that I'd had a significant effect on the impact probability because I had extended the observed arc by more than 40%, but I had no idea whether that probability would go up or down based on my input. I just had to wait and see.
I couldn't be patient, however, and emailed the director of the Near Earth Object Program Office at JPL. Don Yeomans had come to speak to my students at the Astro-Science Workshop back in 2004, so I had connections. I asked him to keep me in the loop regarding the calculation of the impact odds, but Don did one step better. He asked for my information, and put me into the press release! He also mentioned my name on NPR. This has spurred a lot of interest from various folks, including the UAA Advancement Office, the Anchorage Daily News, and even the Discovery Channel. So many thanks to Don! It's been a great Christmas present.
Now, I've forgotten to mention that this ties in nicely with an undergraduate research project that I'm developing for a class at UAA this coming semester. The ultimate goal is to package it for use in astronomy labs across the nation. The students examine astronomical images in order to determine the coordinates of an asteroid at a particular time, and then use the free Find_Orb software to calculate a refined orbit for the asteroid. They then use Find_Orb to add random errors to the observational data, and produce a number of asteroid "clone" orbits that represent other plausible orbital paths for the asteroid. (These are called "Monte Carlo" orbits.) Plugging all of these clones into a planetarium program like Starry Night, the students can then see that orbits can't be computed instantly when an asteroid or comet is discovered, as in Deep Impact. The predicted position of an asteroid always has some level of uncertainty, and with a large number of clones you can begin to visualize this region of uncertainty. This is essentially how JPL calculates their impact odds: they generate a large number of clones, then simulate the passage of time, and the fraction of these clones that hit Mars determines the probability of impact. I had actually given up on doing this part with my students, because the asteroid with the highest known probability of hitting the Earth (Apophis) has only a 1-in-45,000 chance of impact. This means that I would have needed some 50,000 clones, and this was infeasible with the tools I'm using. But even a 1-in-75 chance could be estimated with just a few hundred clones. Plus, the short amount of time between discovery and potential impact meant that a number of other complexities would be removed. So, I will definitely be using this object in my class project, even though the impact will have already happened (or not) by that time.
Anyway... before my contribution, here were the circumstances at Mars on the date of impact:
The horizontal-ish white line represents the orbit of Mars, and the blue line represents the best-fit orbit of 2007 WD5. The near-vertical string of white dots are JPL's clones of the asteroid. 1-in-75 of these clones hits Mars, even though the best-fit orbit is just a close approach to Mars. Here is an updated plot representing the effect of my observations (and ONLY my observations):
Now, nearly 1-in-25 of the clones hits Mars, but notice that the best-fit orbit has switched sides of Mars. The close approach would now happen on the opposite side of the planet! Also, notice the change in the scale of the image. The uncertainty region has shrunk significantly.
UPDATE: The JPL NEO Program Office has just updated the impact odds again, based on 4 observations taken on December 29th & 31st. Here are the new circumstances:
Notice that the scale has changed again, and that the uncertainty region has shrunk again as well. The impact probability is now down to 1-in-28 (3.6%), which is not very much different than before. The best-fit orbit has also moved a bit further away from the planet. If that best-fit orbit doesn't change much from here on, but the uncertainty region continues to shrink, the result will be a definite miss.
Here are a few final thoughts regarding 2007 WD5:
- It passed within 4.5 million miles of Earth on November 1st. This sounds small when you call it 0.048 astronomical units, but large if you quote it as 18.7 times the distance to the Moon. This was nothing to be concerned about, although the object bears watching in the future if it misses Mars.
- The best-fit orbit passes about 13,000 miles from the surface of Mars. If it were passing that close to the Earth, it would be only 3% of the distance to the Moon! That would be an amazing show.
- The impact on Mars would produce an explosion equivalent to 3 megatons of TNT. This is about 200 times the output of the atomic bomb dropped on Hiroshima, or 2-3 of the highest-yield weapons in the current US arsenal. Without the radiation effects, of course.
- The resulting crater would be about 1 kilometer (0.6 miles) wide. This is comparable to Barringer/Meteor Crater, which is 1.2 km wide. According to Wikipedia, that impact killed instantly everything within 2 miles, caused severe flash burns within 7 miles, leveled everything within 9-14 miles with a shock wave moving at 1,200 mph, and created hurricane-force winds up to 25 miles away. But enough dust wasn't thrown up to change the global climate. Of course, the thinner Martian atmosphere (surface pressure 1% that of Earth) could definitely change these effects.
- The impact energy would be less than that responsible for the Tunguska Blast of 1908, which was probably equivalent to 10-15 megatons of TNT.
- The time from discovery to potential impact is only a little over 2 months. If the same object were hurtling towards Earth and we had the same amount of lead-time, the only option would be evacuation of a 2,000 square mile area. That is, if astronomers had sufficient time to calculate the location of the impact! As I pointed out above, it takes time to calculate a fairly precise orbit, so the lead-time on the evacuation time could be as little as a week or two. And that's if it didn't hit in the ocean, then tsunamis would be a problem over a large coastal area. This is why I'm hoping for an impact on Mars, as a wake-up call that we need to have a plan on hand for when a space-rock is headed our way. We need a plan, we need to practice that plan, and we need an organization with the mandate to enact it when the time comes (especially if we were to try to divert it).
