MOVIE - Actual Observations and Latest Results

Introduction

At the last two IMCs I presented the structure of our video system MOVIE and the way the analysis software for video meteor observations works ([1],[2]). I could also show first results from investigations based on the already processed 1993 Perseids. As the analysis program worked out to be quite stable and efficient I focused my work in the last few month on minor software improvements and mainly on the inspection of all video tapes we had recorded before. With new video observations in 1995 and the resulting availability of a bigger database I could check some investigation results from last year and carry out new research for other meteor showers.
In the first part this paper describes changes and additions I made to the software and sketches the impact those have on the accuracy of the meteor parameter. Then some of the new observations and results are introduced in detail.

Last year I showed that the analysis program works efficient and obtains all meteor parameters with a reasonable quality ([2]). Nevertheless there have been some limitations I mentioned as a task for future improvements. One of those was the measurement of the linear meteor position in the processed image. Whereas the position of stars can be determined precisely by calculating the geometric centre of their bloomed image, meteors are often elongated with the 'core' far away from the centre of the object. Thus, one has to point a crosshair manually at the meteor position in an enlarged image which results in less accurate values. This procedure might have a distinct effect on final meteor coordinates because I measured only the start and end position of each meteor.
To overcome this problem I have changed the software to consider meteor positions in all single video frames along the meteors trail. This extra work takes in average less than one minute per meteor but has several positive effects on the obtained meteor parameter.
First of all there are now at least three (normally five or more) position measurements per meteor available. The analysis program calculates an average meteor trail from all of them and obtains a corrected start and end position of much higher accuracy. Thus, the meteor position accuracy does not depend any longer from manual positioning of the crosshair but is limited by other effects. It turned out that the average deviation of single meteor positions from the mean trail is of the order of 5', so the meteor trail resulting from all positions is even more accurate.
Another effect of considering all video frames is the opportunity to get meteor light curves because the measurements are not anymore restricted to the video frames with the maximum brightness. All linear meteor positions are known, so the meteor brightness in all single frames can be determined. By doing this an considerable amount of extra data that does not fit in the PosDat specification occurs, so all those numbers have to be stored separately. I have chosen an easy solution by creating an extra ASCII file for every meteor beside its normal database entry. These files contain all data coming up during the analysis process. They have standard names and the same internal structure. Thus, they can be easily edited and parsed by other analysis programs. An example for an PosDat entry and the appropriate text file is given in Figure 1.

Record# REF_NO HOUR MIN SEC MAG VEL TYPE RABEG DECBEG RAEND DECEND ACC ID REM
  572     15    0    41  50 0.4  2   10  278.37 32.45 279.38 32.29  1 AAP 0.5s

                            meteor parameter
observer                         : MOVIE
observing site                   : 11131
centre of field of view          : Alpha = 255°   Delta =  40°
limiting magnitude               : 5.9 mag
date                             : 22/23.04.95
time                             : 00:41:50.12 -  00:41:50.60 UT
duration                         : 0.48 s
starting point                   : Alpha = 278.37°    Delta =  32.45°
end point                        : Alpha = 279.38°    Delta =  32.29°
maximum brightness               : 0.5 mag
mean velocity                    : 2°/s
number of reference stars        : 10
mean position error of stars     : 6.2'
mean position error of the meteor: 3.2'
supposed meteor shower           : LYR (Lyrids)
expected velocity                : 2°/s
deviation from radiant position  : 1
single positions
00:41:50.12 UT  Alpha = 278.37° (278.42 °)  Delta = 32.45° ( 32.36 °)  h = +2.5 mag
00:41:50.16 UT  Alpha = 278.46° (278.48 °)  Delta = 32.44° ( 32.37 °)  h = +2.3 mag
00:41:50.20 UT  Alpha = 278.54° (278.49 °)  Delta = 32.42° ( 32.40 °)  h = +1.7 mag
00:41:50.24 UT  Alpha = 278.63° (278.59 °)  Delta = 32.41° ( 32.41 °)  h = +1.4 mag
00:41:50.28 UT  Alpha = 278.71° (278.71 °)  Delta = 32.40° ( 32.40 °)  h = +1.2 mag
00:41:50.32 UT  Alpha = 278.79° (278.77 °)  Delta = 32.38° ( 32.42 °)  h = +1.0 mag
00:41:50.36 UT  Alpha = 278.88° (278.85 °)  Delta = 32.37° ( 32.40 °)  h = +0.5 mag
00:41:50.40 UT  Alpha = 278.96° (278.95 °)  Delta = 32.36° ( 32.37 °)  h = +0.6 mag
00:41:50.44 UT  Alpha = 279.05° (279.06 °)  Delta = 32.34° ( 32.35 °)  h = +0.5 mag
00:41:50.48 UT  Alpha = 279.13° (279.20 °)  Delta = 32.33° ( 32.34 °)  h = +0.6 mag
00:41:50.52 UT  Alpha = 279.21° (279.23 °)  Delta = 32.32° ( 32.28 °)  h = +0.8 mag
00:41:50.56 UT  Alpha = 279.30° (279.29 °)  Delta = 32.30° ( 32.26 °)  h = +1.3 mag
00:41:50.60 UT  Alpha = 279.38° (279.14 °)  Delta = 32.29° ( 32.22 °)  h = +1.8 mag

At the beginning all standard information of the observing session is listed, followed by the main meteor data that appear in the PosDat file. The next three lines contain information about the quality of the plate constant fit including the number of reference stars and the mean errors of reference stars and single meteor positions. Next the supposed meteor shower together with the expected meteor velocity at the given distance from the radiant and the deviation of the backtraced meteor trail from the standard radiant appears. Finally all single meteor position are listed. The first values for Alpha and Delta are calculated according to the mean meteor trail whereas the values in curved brackets give the originally measured positions.
The calculation of the meteor shower is another new feature of the analysis program. The first thing I realised after having implemented this routine was that there are meteors which should not be there! During the analysis of this years Quadrantid observation from January 3/4 I found a considerable number of Delta Cancrids (about half the sporadic rate). This is surprising because according to IMOs meteor shower list they 'officially' start to occur not before January 5. May be we should think about softening these thresholds. Normally the software would not recognise such shower meteors because it checks only for active showers in the meteor shower list.
The automatic shower determination makes later data analysis not only easier but it helped also to find out a weak point in the coordinate calculation procedure. During the processing of the '93 Perseids I found out that the calculated meteor positions are still quite sensitive to the chosen reference stars in the field of view. Sometimes only little modifications of these stars resulted in considerable changes of the radiant deviation from the backtraced meteor.
The explanation for this effect is as follows: Due to the recording properties of our video system we are limited to bright stars as reference objects. It is in contrast to the analysis of meteor photographs sometimes difficult to find enough reference stars that are well positioned all around the meteor trail. Furthermore bright stars tend to appear in close groups as illustrated in Figure 2.

Figure 2: unequal star distributions result in considerable meteor position errors

What happens if I take all bright stars in this figure as reference objects? Since almost all of them are concentrated at the right side the plate constants will describe the equatorial coordinates in that part of the image very accurate. However, the few important stars on the left side near the meteor have almost no effect in the whole coordinate transformation. In fact they have larger errors in the resulting 'right side optimised' fit and might eventually be removed from the reference star list because the programs automatically considers them to be useless.
Up to now there is nothing implemented to avoid that effect, the user has to decide manually which stars are used for the position calculation. In the future I hope to overcome the problem using weighted plate constants. This concept implies that each star gets a certain weight in the plate constant calculation to avoid negative influences from imperfect star distributions. Those weights might be assigned either by hand or automatically using the number of reference stars in each area or the distance from the meteor trail as the main parameter.

Analysis of the Video Tapes

In spring '95 I did the tremendous job of analysing all video tapes with hundreds of meteors we had recorded up to that time. I finished at the end of April with a database containing 587 records. Most of them are Perseids from the 1993 maximum but there is also a considerable amount of 1995 Quadrantids and Lyrids stored. Actually we do not have any more unanalysed tapes and all the meteor data are available for everybody in IMOs central meteor database.
The next things was to have a closer look at those meteor files and come up with new investigations as suggested last year.

One of the main aims of our meteor work is the accurate determination of meteor radiant positions and the search for possible sub-radiant structures. Last year I presented a first radiant plot for the Perseids in 1994 ([2]) which was still far away from being optimal. At that time we observed most meteors in the Summer constellations with only a few meteor recordings in other directions which gave the plot a strange longish shape. Having now all the Perseid data from the 1993 maximum available, too, the new plot in Figure 3 is much more accurate and shows smaller details. It was created using 228 Perseid recordings from both years.

Figure 3: radiant plot for 228 Perseids recorded in '93 and '94

The 'theoretic' radiant position given in IMOs meteor shower list is without doubt confirmed by this plot. There even seem to be some sub-radiant structures visible, but they are most probably artefacts from inaccurate meteor data. All 1994 meteors have been analysed with an older analysis program version, so position errors may be larger for those.
On January 3/4, 1995, we were able to observe the Quadrantid maximum under good circumstances near Hannover/Germany ([3],[4]). This time we had chosen a field of view near the radiant to record meteors from a wider area around it. The drive of our mounting did not work at those -8 degrees centigrade, so in fact the radiant slowly shifted into the field of view and we recorded short meteors in all directions. There was another problem with the lens heating which caused a thick ice layer at the optics and reduced the limiting magnitude drastically for a certain time. Nevertheless we finally found more than 100 meteors on the tapes and I could produce the radiant plot given in Figure 4.

Figure 4: radiant plot for 39 Quadrantids recorded in '95

Again the theoretic radiant given by the green circle with a diameter of 5° is confirmed by our video observations. The peak is even sharper than the one of the Perseids and there is no sub-radiant structure visible in this plot. The absence of such structures up to a resolution of about 1° is one of the main results of this analysis.
The last plot shows the chance to obtain accurate radiant positions from even less shower meteors. It proves the possibility to investigate minor meteor showers and shows once more the power of video observations. On April 22/23, 1995, we observed the maximum of the Lyrids and recorded about 50 meteors within 5 hours of observation. In that night we changed the field of view two times having its centre west of the radiant, at the radiant itself and east of the radiant. The resulting plot for 32 Lyrids is more accurate than all other plots we have produced so far. This is both due to excellent positioning of the video system and a good distribution of reference stars in the fields of view.

Figure 5: radiant plot for 32 Lyrids recorded in '95

This time the sharp peak of the radiant lays with Alpha = 271° and Delta = 33° almost directly at the expected position. Again there are no distinct sub-structures visible in this plot.

Meteor Images and Sequences

At the last IMC I showed how easy it is to produce nice meteor shower images from a given video image collection ([2]), here I can present a new picture of the 1995 Lyrids. The meteors look different in this picture because now we recorded the sky directly at the radiant as described above. Furthermore a larger portion of the sky reaching from Bootes at the right side until Cygnus in the left corner is displayed. To achieve such an wide angle view I merged three independent shower images into the final picture. Radiant effects like different trail lengths are quite obvious.

Figure 6: Lyrid shower image from recordings on April 22/23, 1995

In addition to another shower image showing the '95 Quadrantids I created an especially interesting animation displaying also the dynamic aspect of meteors appearing and disappearing near the radiant. Within a few seconds 18 Quadrantids near the radiant become visible in this sequence. You can notice different meteor velocities depending on their radiant distance which makes the radiant effect even more clear.
Unfortunately it is impossible to reproduce the animation in this paper, but the 120kB MPEG MOVIE is directly available from me and can be accessed in the World Wide Web under the URL http://www.informatik.rwth-aachen.de/I6/Colleagues/molau/meteore/quad95.mpg.

Meteor Cluster Effects

One of my favourite aims in meteor astronomy is the proof of the existence or non-existence of meteor clusters. In 1992 I analysed our visual Perseid observations ([5]) and found no clue for any type of clustering. Even though the meteor timings have already been very accurate and we missed almost no event due to our computer based observation method ([6]), the distribution of meteor distances behaved completely as expected for random spread particles with no clustering.
This year I repeated the calculation for our video observations which are still more objective because there is no 'plotting time' at all involved in this method. The data set of the '93 Perseids met the requirements closest because that time we recorded more than 300 meteors within one night of observation. Due to the fact that the exponential distribution holds only for random events with a constant probability, I had to take the variable hourly meteor rates into consideration. So I first calculated the shower activity profile and used that result to normalise the shower activity to a constant level as described in [7]. Finally I considered the same 20s intervals as in the '92 investigation ([5]) and finished with the graph given in Figure 7.

Figure 7: meteor cluster analysis with 20s intervals for 337 meteor pairs on August 11/12, 1993

Each column in this graph represents the probability that two successive meteors appear with the respective time difference (averaged in 20s intervals) related to the total of 337 meteor pairs. It is quite obvious that the theory for randomly distributed particles fits the video observations very well. In other words, there is no cluster effect at all in these data from the 1993 Perseid maximum.
I was still curious whether clustering takes place on smaller time scales. May be there are meteor clusters in the order of some seconds which are completely smoothed out by averaging in 20s intervals. Thus, I repeated the calculation for the shortest possible interval length of one second which is the time resolution of our video system. This time cumulative probabilities were applied to get enough meteor pairs for each slot and avoid statistic fluctuations resulting from small samples. The result is given in Figure 8.

Figure 8: meteor cluster analysis of the same data set with cumulative 1s intervals

Again there is obviously no difference between clusterless theory (exponential distribution) and video observation in the graph from the lower left to the upper right corner. However, once you look close enough at the most interesting first intervals you will surprisingly realise that here we always observed more meteor pairs than expected. To make things clearer I have superimposed another graph showing the relative differences between observation and theory. There is a clear surplus up to 12s distance between successive meteors which is just the expected outcome in presence of meteor clusters. Thus Figure 8 gives a first vague clue that there really is a kind of small scale clustering during the Perseid maximum. Further calculations showed that the observation can be most accurately described with a low cluster level of about 1.5 percent (i.e. from 100 meteors only one or two did not origin from mutual independent meteoroids).
Of course we have to handle this result with great care since it relies only on a very data sample. 57% surplus of meteor pairs with distances up to 1 second comes from 11 observed pairs instead of the expected 7. 30% surplus up to 3 seconds meteor distance results from 21 pairs instead of 16,1. Beside that further inaccuracies were introduced by the normalisation of the meteor activity. Thus, it is only a first indication for clustering which definitely has to be confirmed in the future.

As mentioned in the first part of the paper I have used a new algorithm to determine meteor positions with the side effect that all video frames are analysed. Consequently we find information about meteor development expressed by their changing brightness along the trail in the additional data file. With those data it is possible to calculate mean light curves for different showers and compare them with one another.
In a first attempt I have chosen 86 Perseids and 19 Lyrids with starting and end points within the field of view. There were at least 6 brightness measurements for each meteor available and the maximum brightness was usually high to avoid errors from less accurate faint meteors.
In the analysis each meteor was normalised to a standard duration of 100 frames, all brightness values between the measurements have been interpolated. The result is given in Figure 9.

Figure 9: mean light curves for 86 Perseids and 19 Lyrids

First of all we can see that the average Perseid was brighter than the average Lyrid, but that is not the main point of this investigation. More important is the fact that Perseids seem to have symmetric light curves centred at their point of maximum brightness whereas Lyrids have a steeper descending then ascending branch. It is difficult to get more information from this diagram because both graphs terminate already shortly after the maximum. At this point the moving meteor 'core' is not visible anymore and consequently I cannot measure further meteor positions.
It is without doubt interesting to do further research and find out whether asymmetric light curves are the rule or the exception of the rule.

Meteor Parallaxes

As in 1994 we have also this year not yet been successful in obtaining double station video observations with our Dutch friends from the NVWS. Even though we observed the Quadrantids together we could not record even one single meteor from both station. The main reason was different weather on places only 30 km apart from each other (they just started their observation when we had to finish due to clouds), but also problems with our frozen lens and simply misfortune are to blame for that. It seemed that Mr. Murphy was not far away from Hannover that night.
In return we recorded an amazing -4 mag Delta Capricornid fireball at 22:52 UT on January 5, 1995 , which moved spectacular slowly along the southern horizon and was visibly for 2 seconds. After submitting the observation to the Fireball Data Centre I received the information that the fireball was photographed from three stations of the European Fireball Patrol, too. Even though the orbit of the meteoroid has not been determined yet I can present a video image and a photography of our second double station meteor (see [8] for our first success) here.

Figure 10: a -4 mag DCA fireball from January 5, 1995, recorded with MOVIE near Hannover (left) and with an all-sky-camera at EN station #75 in Beuterode (right)

From our 1995 Lyrids observation I can present another interesting result by comparing the radiant plots from plottings of two visual observers and our video system MOVIE.
Even though both observers have experiences from almost 10 years of observation they did not made any plottings in the last few years. Hence, their 'first try' after a long break is of course not comparable with the results an active observer would achieve. Nevertheless it is amazing how obvious the differences in accuracy are. Figure 11 shows both radiant plots with the same pixel resolution, the first one obtained from 42 plottings of the visual observers, the second one from 23 video meteors.

Figure 11: radiant plot of the '95 Lyrids, obtained from 42 plotting of two visual observers (left) and 23 video meteors (right)

It is worth to mention that both analysis yield almost the same radiant, the position obtained from the much less accurate plottings was less than one degree north of the video meteor radiant.

Video Meteor Observation in the Internet

Beside the analysis work I recently dealt with another important subject. Some members of the IMO council had the idea of publishing observation hints, results and other material in the Internet to further increase the accessibility of meteor related information and the awareness of IMO. This is especially important in times of a rapid growing availability of computer networks and electronic information exchange. The basis for this project is the World Wide Web, a user friendly information system that came up at the beginning of the nineties and developed very fast from that time on. The access to the Web is not anymore a privilege for students and other on-line users of the Internet but it is more and more available for everybody over on-line services like CompuServe. WWW documents are written in a simple language called HTML, so creating own pages is no real problem . There are several interactive HTML browser for different hardware platforms available which are mouse controlled and really easy to use. They show you the requested documents consisting of formatted texts, inline images, links to other documents and all types of files which can be downloaded just with the click of your mouse.
It is planned to develop pages for all kinds of meteor observation and I was asked to prepare a document about video work. This page is available since May under its URL http://www.imo.net/video/. It contains a basic introduction into video observation, a complete literature reference list with links to all documents directly accessible in the Web, addresses of contact persons, links to other meteor astronomy related pages and several examples of video observation results. The material has been revised by several other video observers and gives everybody who wants to deal with the subject a good basis to start with.

Figure 12: the intro of IMOs video observer pages displayed with NetScape

The Internet is a fast living community, so documents and Websites change very quick. Many things will have been added or improved when this paper is published. In every case it is worth to have a look at this site because you will always find links to our latest investigations here.

Conclusions

Video observation has once more proved to be a powerful tool in meteor astronomy. With larger databases more accurate investigations have been possible and will be carried out in the future, too. In the moment we can concentrate on new video observations and investigations as well as software improvements because all our video tapes are analysed. The meteor data is available from IMO to invite their use by other researchers. Double station observations are still one of the major remaining tasks especially with a growing number of available video systems in the observer community. First meteoroid orbit calculations are expected soon.

Acknowledgements

I appreciate the ongoing materialistic and idealistic support of our work from the Archenhold-Observatory Berlin. Thanks to Thomas Rattei, who originally wrote the meteor shower detection routine, and Dieter Heinlein, who sent me information and photographs of the recorded fireball. Last but not least special thanks to Felix Bettonvil and Marc Neijts, who again joint our double station project in January, and Kathrin Düber, who took part in all visual and video observation and was always an important aid in the 'hot times' of the equip- ment set-up.

References