Fine Structure Analysis

For decades, visual observers have claimed that shower meteors appear to come in clusters. It seems as if you’ll watch for ten minutes seeing nothing, and then half a dozen will burst upon you in a matter of seconds. The problem is, human perceptions of randomness aren’t objective enough to build a case. It is not difficult to carry out a rigorous statistical test that will settle the question. As early as 1968, Czechoslovakian astronomers using radio data carried out such a test and were able to establish that shower meteors are randomly distributed. A variety of subsequent studies have confirmed this result. However, all such studies have been hampered by the low observed rates of shower meteors. These low rates have imposed a serious statistical constraint on studies of temporal non-randomness: the time intervals studied have typically been between one minute and ten minutes. We have not had sufficient data to examine time intervals below one minute. A recent analysis of 1999 video Leonid data by Jenniskens and Gural found no reason to believe that Leonids are clustered at time scales of less than one second, but the data did suggest the possibility of nonrandomness on time scales of ten seconds. Again, the amount of data was insufficient for a full-scale examination.


The objective for this effort is to attack the problem from the other direction. Instead of asking whether shower meteors are distributed randomly in time, I assume that they are nonrandomly distributed and then ask, what could cause such a nonrandom distribution? The most likely answer is that a group of meteors might be released from the cometary head near perihelion, and that the members of such a group would slowly spread out over time. This spreading-out process would not be omnidirectional; orbital motion would tend to spread the members out along the orbital path, so that they form what I shall call a "streamer": a line lying along the orbital path populated with meteors.

When the earth intersects this line, meteors in the streamer will hit the earth’s atmosphere. If the earth were stationary relative to the Leonid orbit, then we would see a sequence of meteors at the same position in the sky. However, the earth is moving relative to the Leonids, which means that the streamer will move across the surface of the earth. This motion can be calculated using some rather messy spherical trig.

I developed an elaborate experiment to test this hypothesis, but the equipment failed and I lost everything. However, there is so much data available from a wide variety of observers that it might be possible to confirm or reject the hypothesis with this data. Consider, for example, the following fragments of data taken from actual reports to the meteorobs mailing list:

Time UT
Obs #1
Obs #2
Obs #3
Obs #4
10:27
14
17
10
19
10:28
21
12
14
17
10:29
21
16
7
12
10:30
17
8
14
16
10:31
12
18
12
22
10:32
12
22
8
22
10:33
21
15
15
10
10:34
13
10
17
5
10:35
14
7
13
17
10:36
16
14
9
15
10:37
24
13
11
18


Remember, this is not in any way fudged data, so it shows the messiness of real data. Note that Observer #1 recorded a slight increase in instantaneous rate at 10:28 - 10:29. With enough imagination, you can see this same slight increase at 10:29 for Observer #2, 10: 30 - 10:31 for Observer #3, and 10:31 - 10:32 for Observer #4. Again, Observer #1 sees a bit of a local slowdown at 10:31 - 10:32, which shows up at 10:32 for Observer #3 and 10:34 for Observer #4. You might see a few other interesting relationships in the data.

Of course, you must realize that this is just a fragment of data and by itself it proves nothing. I offer this table to illustrate what I’m looking for. If I can get lots of data from lots of observers all over North America, I can apply statistical tests that will clearly show whether these apparent relationships have any substance to them.

So here’s what I intend to do: first, I must collect a vast amount of data so that I can have a much, much larger version of this table. Moreover, I need to set it up for higher temporal resolution, not just these one-minute bins. Then I must calculate the expected correlations based on the motion of the Leonids relative to the Earth. Lastly, I must apply those correlations to the statistical tests to determine if there really is something going on here.