[meteorite-list] AMS Fireball Program Stats & Trajectory Solver

From: Mike Hankey <mike.hankey_at_meteoritecentral.com>
Date: Wed, 20 Mar 2013 11:30:37 -0400
Message-ID: <CAJak_qWXxS-w_VHyaPCTDzKJhuBQfCpmraoprisjY9mE8Ev_Hg_at_mail.gmail.com>

Hello Meteor List Readers,

I wanted to share a new paper we published on the AMS site earlier
this week with you.

http://amsmeteors.org/2013/03/ams-fireball-stats-analysis/

Its a statistical analysis / presentation of the fireball report data
logged by the AMS since 2005.

While the data shows a continuous increase in reports and events
logged with the AMS, it is not possible to make any conclusions about
fireball rates from the AMS reports because the environment is not
controlled. Increases in traffic, public interest, mobile devices, etc
all contribute to a more connected world and thus more people
reporting the things they see in the sky.

With that being said, it is still interesting to look at the data grow
over time. We have also implemented a classing system to differentiate
the size of events reported.

Also of interest to these lists is a new capability we are in the
process of developing that allows us to automatically plot
trajectories from fireball events within a small margin of error.

Right now our algorithm is pretty simple. For the start and end points
of each witness we compute the intersection of these points with all
other witnesses. This results in 100s or 1000s of points for both the
start and end. We then apply some error suppression to weed out bad
reports and then simply average the rest of the points to determine
the estimated 2d trajectory. Believe it or not this is actually
working very well! We are still tweaking these algorithms and data
suppression/balancing routines and the accuracy varies from event to
event, but right now we are at least able to determine the bearing
accurately most of the time and we are able to plot the trajectory
accurately with a margin of error that varies from event to event.

We have about 10 events that have trajectories plotted from videos
and/or radar returns/meteorite finds that we are using as a control
group. For some of these we can plot the trajectories from witness
reports to within < 2 KM of the video plot. Others are still off by up
to 40-50km.

Sutters Mill is a good example of an event we are having trouble with.
There were over 100 witnesses and a large % of them came from the San
Francisco Bay Area. Most of these users reported bearings slightly
south of reality and this is weighing the averages down. If we run the
algorithms with only witnesses who self rated their
astronomy/observing experience at >=2 we can plot the trajectory
within 2km of the trajectory published in the science article. The
challenge though, is knowing which algorithm or ratings set to use
when the actual solution is unknown. We still have a few ideas that we
are working on, but if anyone has suggestions about how to improve
this system please contact me.

A KMZ file containing the trajectory estimates from all of the
significant events over the USA since 2008 is posted with the paper.
All of these events had sonic booms associated with them and large
volumes of reports. I know some of these trajectories are off, so
please keep in mind this is literally version 1 of the trajectory
solver. We've already made a few improvements since this KMZ was
published and we still have work we can do to make it better.

On a related note, I went to the Congressional Hearing on Asteroid and
Meteor threats yesterday in Washington DC. It was very interesting,
the hearing is not yet available on CSPAN, but it should be soon. I
think everyone on these lists would enjoy watching it. They discussed
the DOD systems that detect fireball events, but did not get into many
details or talk about what they plan to do with this data. I hope the
DOD data analysis related to the most significant fireball events
logged with the AMS in the last 5 years will eventually be released.
This page http://neo.jpl.nasa.gov/fireballs/ needs some more content.

The best line of the hearing came from Congressman Ralph M. Hall from
TX who asked, "If an asteroid was going to hit Oklahoma, could we
shoot a laser at it and split it so that one half hit California and
the other half hit New York?"

I appreciate any private or public criticism or suggestions about the
paper and trajectory solver.

Thanks,

Mike Hankey
Freeland MD
http://amsmeteors.org
Received on Wed 20 Mar 2013 11:30:37 AM PDT