LFHST - Buie Journal - 1996 February

Getting ready for the new Pluto images.

Now that the new Pluto observations are moving along toward execution on HST, I must now turn my attention to some preparatory work. Dr. Alan Stern and I have been working on our previous pictures of Pluto. We first started to work on these images last January, January 1995, that is. That month we had time to spend on trying to understand the pictures and how to get the best information from them.

Our work at that time lead us to understand that getting good maps of Pluto would not be easy. The size of Pluto makes our job very difficult. Unlike Jupiter, or even Neptune, Pluto is very small and is just barely resolved by HST. Any finite size telescope has a fundamental limit on how small an object it can make out. From the ground most modern telescope are limited at the same level by the atmosphere. In space, the resolution is completely determined by the size of the mirror in the telescope. When you work at the limit of a telescope you must understand everything about what it does in the process of focusing the light to make an image.

So, to get a map of Pluto we found that we would have to remove some blurring effects HST has on the image we see. This blurring comes from the point-spread function (PSF) and is present at some level in all optical systems, whether it's HST or your hand-held 35mm camera. When HST first launched, it's PSF was very, very bad due to a misshapen mirror. After HST was repaired, the PSF was improved but not removed. When we realized this work needed to be done we found ourselves out of time to work on this project. Alan and I both had other projects that were demanding our attentions. Alan had a rocket flight experiment to prepare and I had other ground-based observations to conduct. All of these activities distracted us until late in 1995 when we were able to return to our analyses.

All right, so what's the point here? The time-table for the Live from Hubble Space Telescope program set a very deadline for our work. After all, if we haven't figured out what is in the older data, what will we be able to do with new data? So, I've spent the better part of January and February trying to finish our analysis of the old Pluto images before we get the new data.

All of this work is leading up to a press release and a TV broadcast on our results. Thursday, March 7, Alan and I will be in Washington DC for a presentation of results which will be just one week prior to the unveiling of the new Pluto images as part of the Live From HST project. Because of the press release, I can't show you the results (yet) of our work but I can let you in on a few of the steps leading up to the release.

Our project consisted of images taken at four times such that Pluto had rotated by about 90 degrees between each time. The images clearly show regions of light and dark on the surface but need to have the effects of the PSF removed. I constructed a computer model, or description, of Pluto as it must look and then created steps in the computer that would apply all the effects caused by the PSF. From this I can then make a guess as to what Pluto looks like, blur the image, and then compare to the actual pictures. The Pluto image I want is a grid of pixels that I can relate to the brightness of it's surface. For this model I divided up the visible disk of Pluto into an 8x8 grid of pixels. Then I used my program to vary the brightness of each pixel until the collection of pixels most closely matched the image. The approach works well because we know how big Pluto is in the image. If we didn't know the size this trick wouldn't work.

Each image gives us a set of pixels that covers a part of the surface of Pluto. If we take all of these partial sets, we can add them up to get a complete map of the surface. I don't know if this sounds simple or not but to me this is pretty routine stuff. It certainly didn't sound like 2 months of work to me. But let's look at this a little closer. I was able to pretty quickly write a computer program to find the best set of pixels for each image. But, the program takes about 3-4 hours to run for each image on the fastest computer we have here at Lowell Observatory. We have 20 images in all so running all the images through the computer takes about 2 days before I get an answer. That's a pretty long time.

Have you ever written computer programs? When I was first learning to write programs, my teachers said that it was impossible to write a program that was correct on the first try. I thought at the time that this was silly. If you were really careful, certainly it must be possible to write a good program. Well, I've been doing this for nearly 20 years now and I have to say that those teachers were right. Somehow, a detail here or there will slip through and there will be something wrong with the program. Sometimes the problems are easy to spot, like all the pixels coming out completely black. Usually, the problems are very, very difficult to notice. How do know the answer is right? After all, I can't just pick up an atlas off my shelf and check to see what Pluto must look like. This is the first time this has ever been done and I have to use other means to see if the map I get out is correct.

You're probably already guessing that I had a few problems along the way and you'd be right. I found a bad file name (foc410.fit instead of foc410n.fit). I found missing statements that were needed to rotate the images properly. There were others too, and each time I found a mistake that meant another 2 days before I could check the answer. Two months can go by real fast this way. In the meantime, everyone is yelling for final maps and images to get ready for this upcoming press release. At the Space Telescope Science Institute (STScI), they had to redo the press images and video animation sequences several times because of errors I found. Do I have the right answer yet? If you were asking yourself that question already, good for you. I don't know. All I can do is check answers and compare images and do my best to eliminate mistakes. I'm pretty sure I've got a pretty good map pulled out from the data but in the process there are some things that might work even better than my current computer program. Unfortunately for this press release, it will take much too long to try my new ideas so we'll just have to settle for what I've been able to do so far. Next year, maybe I'll have an even better map.

This is a pretty good indication of what it means to do original scientific research. You take an idea, some data, a bag of computer or mathematical tricks, and try to make sense of it. Along the way you learn new tricks, get new data, or get a new idea and before you know it, last year's work has been made obsolete. As you proceed, you never stop checking and re-checking your results. When you finally get the right answer you'll know because the answer is the same no matter how you get there. These new observations we'll be getting on Pluto will become useful checks of the work I've done on the data from two years ago. If nothing looks the same we'll probably have to take another hard look at Pluto to see if we can understand what we're seeing. Anytime I can get new data on Pluto, it helps me to understand Pluto a little bit better and also helps me to believe in my previous results. It also seems that for all the work, you never quite finish (at least until a spacecraft flys by Pluto and then we'll really know what it looks like). As soon as the results have been made public on March 7, I will be posting information on the maps and images we have. Stay tuned, it's only going to get more interesting from here!

Back to my home page -o- Back to my Pluto page -o- Back to my LFHST page

Marc W. Buie, Southwest Research Institute