Leonid Analysis Diary

The story so far:

Since my high school days, I have been trying to determine once and for all whether there is any fine structure in meteor streams. I have made many efforts to make this happen, and my best shot came with the Leonid outburst in 1999. That provided plenty of data, but it was all on videotapes, and a proper analysis required that I get the data into numeric form -- that is, digitize the video and write software that would recognize each Leonid in the video and compile data on it. I tried to write such software just after the 1999 outburst, but the computers of those days were not up to the task. I therefore put off the effort until 2007, when I built a computer system adequate to digitize the video. Dr. Jenniskens generously assigned a summer intern to carry out the digitization process, and the result was nearly a terabyte of digitized imagery from four cameras covering three hours centered on the peak of activity. However, in 2007, I did not have the time to analyze the data, so it sat for over three years.

Then in November of 2010, I had set aside my main project on interactive storytelling and decided that it was time to complete my work on the Leonids. I set to work rewriting everything in Java. Things proceeded quite well at first; by early December I had fully analyzed nearly six hours of data and produced a huge catalogue of some six thousand Leonids. At that point, however, everything went to hell. I had been analyzing the video from the left side of the aircraft, which faced south. The spherical coordinates in that area were rather “flat” -- that is, they were far from either celestial pole. This permitted me to make some simplifications in the analysis that worked just fine. Moreover, the cameras were pointed at an area of the sky rich in bright stars, which made it easier for my algorithms to register exact locations in the sky.

But then I moved to the right side of the aircraft, which pointed north. There weren’t as many bright stars there, so my algorithms had to struggle to identify an adequate set of stars to register locations. Worse, the fields of view are close to the north celestial pole, so the extreme curvature of the spherical coordinates there exposed many dumb simplifications in my algorithms. Bit by bit, I nailed down the problems and corrected them. As of today, January 8th, 2011, I am still struggling with these problems.

January 8th
January 10th
January 11th
January 14th
January 15th
January 28th
January 30th
February 1st
February 3rd
February 6th
February 10th
February 13th
February 16th
February 25th
February 27th: a problem with heights
March 12th: first results!
March 17th: absolute luminances
March 19th: decelerations
March 20th: average flare counts per minute
March 22nd: Light Curves, Part 1
March 23rd: I screwed up!
March 24th: Light curves, corrected
March 25th: Analysis of ablation
March 26th: Concluding the analysis of light curves
March 27th: First crack at nonrandomness -- success!
April 2nd: Weird results (probably wrong)
April 20th: A revolting development
April 23rd: Correlation coefficients
April 25th: Suspicions confirmed!