Mcfadden 2021

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 32

Uppsala University

Bachelor of Science in Physics

Degree Project 15 ECTS

Optimization of Meteor Triangulations


Using Timed Observations
Supervisor: Eric Stempels
Subject reader: Andreas Korn

Author: Vendla Niklasson

June 3, 2021
UPPSALA UNIVERSITET 2 (32)

Table of Contents
Abstract .............................................................................................................................................. 3
1 Introduction ..................................................................................................................................... 4
2 Background ...................................................................................................................................... 5
2.1 Meteor observation ................................................................................................................................. 5
2.2 Swedish Allsky Meteor Network .............................................................................................................. 6
2.3 Sweden’s approach .................................................................................................................................. 6
3 Problem statement ........................................................................................................................... 7
4 Method ............................................................................................................................................ 7
4.1 Mathematical description of existing methods......................................................................................... 7
4.1.1 Intersecting planes method ........................................................................................................................................7
4.1.2 Lines of sight method .................................................................................................................................................8
4.1.3 Multiparameter fit ......................................................................................................................................................9
4.1.4 Monte Carlo..............................................................................................................................................................10
4.2 Optimizing the trajectory ....................................................................................................................... 10
4.2.1 Chi-Square test .........................................................................................................................................................10
4.2.1 Powell’s method .......................................................................................................................................................11
4.3 Coordinate transformations ................................................................................................................... 12
4.3.1 Topocentric coordinates to geocentric cartesian coordinates .................................................................................12
4.3.2 Transformation between geodedic- and geocentric cartesian coordinates.............................................................13
4.3.3 Calculating the radiant for the meteor .....................................................................................................................14
4.4 Application of methods on Swedish data ............................................................................................... 14
4.4.1 Perseid meteor shower ............................................................................................................................................14
4.4.2 Intersecting plane method for first estimation of the trajectory .............................................................................15
4.4.3 Calculating the trajectory by using the lines of sight method ..................................................................................15
4.4.4 Adding time variable to the optimization.................................................................................................................16
4.4.5 Adding time offsets between the stations ...............................................................................................................17
5 Results ........................................................................................................................................... 17
5.1 Looking at a single meteor event............................................................................................................ 17
5.2 All results ............................................................................................................................................... 26
5.2.1 Intersecting planes method ......................................................................................................................................26
5.2.2. Lines of sight method ..............................................................................................................................................27
5.2.2. Adding the time variable .........................................................................................................................................28
6 Evaluation and conclusions............................................................................................................. 31
Recommendations ....................................................................................................................................... 31
8 Appendix ................................................................................................ Error! Bookmark not defined.
UPPSALA UNIVERSITET 3 (32)

Abstract
Meteors are light events appearing in the sky, looking like strips of light. A meteor occurs when a space
object, a meteoroid, enters the earth’s atmosphere at high velocity and starts to glow. Meteor
triangulation is an important part in astronomical research, by mapping streams of material in our solar
system the knowledge about the solar system is extended. By making measurements with meteor
cameras and calculate the meteors trajectories with triangulation methods, the meteors origin in space
can be derived. Previous research has shown that by including time measurements, and not only use
triangulation techniques with spatial measurements, the trajectory for a meteor can be calculated more
accurately. This project seeks to explore if including the time information from the Swedish meteor data
would create more precise solutions for the meteor’s paths through the atmosphere. This was done by
investigating how the time variable can be implemented and running the Swedish 2020 Perseid data
through the algorithm that implements the time variable. The accuracy of the methods is determined by
how exact the Perseids radiant position is calculated and also by looking at the computed velocities for
the meteor events. The results from this project show that the time variable should be considered in the
Swedish meteor triangulation algorithm. With inclusion of the time variable, the radiant position of the
Perseids can be more closely determined and the velocities for the meteors are in the expected range for
Perseid meteors. But it turns out that the timestamps between the meteor cameras are not synchronized,
which means that time offsets between the stations are required.

Sammandrag
Meteorer är ljusfenomen som kan ses på natthimlen, de ser ut som ljusstreck och kallas även stjärnfall.
En meteor inträffar när ett objekt från rymden, en meteoroid, kommer in i jordatmosfären med hög
hastighet och börjar glöda. Meteortriangulering är en viktig del i astronomisk forskning, genom att
kartlägga strömmar av material i solsystemet kan kunskapen om vårt solsystem breddas. Meteorers
ursprung kan beräknas genom att göra mätningar med meteorkameror, för att sedan med
trianguleringsmetoder beräkna dess banor genom atmosfären.

Tidigare forskning har visat att man kan få mer noggranna bestämningar för meteorspår genom att
inkludera tidmätningar i beräkningarna, i Sverige används just nu endast trianguleringsmetoder som
bygger på spatiella mätningar. Målet med det här projektet är att undersöka om inkludering av
tidmätningar från den svenska meteordatan kan ge mer exakta lösningar för meteorspår. Detta
undersöktes genom att kolla på hur tidsvariabeln kan integreras i den svenska trianguleringsalgoritmen
och sedan testköra svensk perseiddata från år 2020. Precisionen i metoderna avgörs genom att kolla på
hur exakt perseidernas radiantposition kan bestämmas och genom att kolla på meteorernas beräknade
hastigheter i atmosfären. Resultaten från det här projektet visar att tidsvariabeln borde tas i beaktning i
den svenska trianguleringsalgoritmen. Genom att inkludera tidsvariabeln kan radiantpositionerna
bestämmas mer exakt och hastigheterna är inom det förväntade spannet för perseidmeteorer. Men det
visade sig att tidmätningarna mellan de olika kamerorna är osynkroniserade, så tidsoffsets måste
inkluderas i algoritmen.
UPPSALA UNIVERSITET 4 (32)

1 Introduction
Meteoroids are space objects that enter the earth’s atmosphere with high velocity. The shockwave in
front of an incoming meteoroid strongly heats the surrounding air, which causes a bright glow. The trail
of this glow can then be observed as a meteor, often visible with the naked eye. Meteors are often seen in
meteor showers; these meteor showers originate from streams of material in the solar system. When
passing Earth, some of this material might enter the Earth’s atmosphere, creating strips of light in the
sky. By mapping these streams of material and calculating where they originate from, our knowledge
about the solar system can be extended. Therefore, observing meteors is a valuable part of astronomical
research. Meteors is also interesting for the general public, the spectacular light from a meteor can catch
the interest of most people, and sometimes give rise to concern. It is hard to estimate distances to objects
that move with cosmic velocities, which can trick an observer to believe that a meteor is a light event
happening nearby. Some objects make it to the earth’s surface without burning up, for these cases there is
high interest in finding the meteorite and calculate where the meteorite hit the ground. Without proper
measurements and calculations, it is hard to estimate how the meteor has been moving.

An effective way to make meteor observations is to use meteor cameras that record the night sky. During
a cloudless night a meteor camera can observe multiple meteors. Using the results from these
observations one can calculate the meteors trajectories in the atmosphere. Every meteor in a meteor
shower appear to come from the same origin in the sky, this celestial point is called the radiant. To check
the accuracy of the calculated trajectory, the radiant of all meteors from the same shower are compared.
If the calculations are exact the radiant for each meteor should be the same.

There are multiple approaches to calculate the trajectory of the meteor, the common factor is that all of
them are based on vector geometry. By extracting information of the meteor’s azimuth- and altitude
angles, and the longitude- and latitude coordinates for the place of observation, the path of the meteor
through the atmosphere can be estimated. But the meteor stations happen to measure more than just
angles and coordinates, there is also an exact time stamp for every observation. This parameter, the time,
is currently not used in the Swedish algorithm. As in all fields of physics, there is an ambition to take as
many variables as possible into consideration, and especially when this data is already available, it
should be considered.

Previous research has shown that including the time variable can produce a more precise solution (Gural
2012). The idea of this project is to include the time variable and see if it is suitable for the measurements
made in Sweden. Some potential complications of including the time parameter are clock setting errors
and nonsynchronization concerning the meteor cameras.
UPPSALA UNIVERSITET 5 (32)

2 Background

2.1 Meteor observation

Figure 1 Image describing the observation of a meteor

To be able to map the streams of material in the solar system, we want to know where in the solar system
meteors originate from. Meteoroids are anonymous and dark objects when they are outside the earth’s
atmosphere, but once entering the atmosphere they can be observed, and the angle of incidence can be
derived. Observing the night sky is like looking at a flat sheet, it is impossible for a single observer to
know how far away an object is. When observing a star for example, it is impossible to know how far
away the star is, but the direction for which the observer has to look can be determined. This direction is
called the line of sight and is described by the azimuth and altitude angles for the observation, see Figure
1.

Figure 2 Image illustrating meteors that appear to


emerge from the same point, the radiant
Meteors from the same shower travel with the same direction, parallel to each other. This causes an
optical illusion; when meteors from the same shower are observed, it looks like they originate from the
same point in the sky. Driving a car through a snow fall can offer the same experience, when looking at
the parallel travelling snowflakes through the windshield, it looks like they emerge from the same point.
The point that the meteors from a single shower appear to originate from is called the radiant. By
UPPSALA UNIVERSITET 6 (32)

deriving the radiant for a shower, their velocity direction can be calculated which is valuable information
when calculating where the meteors come from. Using the statement that meteors from the same shower
should have the same radiant positions, different approaches to trajectory estimation can be reviewed. If
the radiant positions differ a lot in the same shower, the trajectory solution is not very precise. And the
opposite, if the radiant positions are close to each other, it indicates that the solution for the meteors
trajectories is accurate.

2.2 Swedish Allsky Meteor Network


Sweden has 12 meteor observation stations. Together these form the Swedish Allsky Meteor Network,
the purpose of the network is to track meteors that can be seen from Sweden. Each station is equipped
with a meteor camera that records the night sky to observe meteors. The cameras react to sudden changes
of light, when a change of light is seen, the cameras take pictures of the sky and catch the trail of the
meteor. The cameras take 30 pictures per second. Every pixel in the picture frame is associated with
azimuth and altitude coordinates, calibrated with the stars as reference. Data is extracted from every
picture, providing the azimuth and altitude angle for every meteor snapshot along with a precise time
stamp.

Figure 3 Picture of a meteor event taken with a meteor camera (Swedish Allsky Meteor Network 2021),
with permission from Eric Stempels.

2.3 Sweden’s approach


There are different triangulation approaches to calculate the path of the meteor through the Earth’s
atmosphere. The method used in Uppsala is the intersecting plane method followed by the lines of sight
method. Intersecting plane method often gives a sturdy first estimation of the track, by defining
geometrical planes for which both the meteor and the observing point must pass through. If at least two
geometrical planes can be estimated, the intersection between them gives an estimation of the trajectory.
To improve the solution, the lines of sight method is applied. The lines of sight method is based on a
method for linear regression, referred to as the least squares approach. The method of least squares is to
minimize the sum of the squares of the residuals. In the lines of sight method, the residuals are the
distance between the lines of sight and the proposed trajectory.

There are several more methods to consider and the algorithm for calculating the trajectory can always be
made more precise. Each meteor camera provides timing information for every observation and this data
might be useful in seeking a more complete solution. One method called the multiparameter fit includes
the time parameter and minimizes the trajectory to the observed points, based on the time information at
each observation. The algorithm in Sweden currently does not handle the time variable. It is of interest to
UPPSALA UNIVERSITET 7 (32)

see if the inclusion of timed observations in the algorithm could give us a more precise solution. Previous
research shows that the multiparameter fit method gives a more accurate solution (Gural 2012).

3 Problem statement
What is the best application for the Swedish network; an algorithm based on geometrical information or
an algorithm including timed observations?
How can the method of multiparameter fit, described in the article by Gural, be implemented on Swedish
meteor data?

4 Method

4.1 Mathematical description of existing methods

4.1.1 Intersecting planes method


The intersecting planes method is a common approach for meteor trajectory estimation. Measurements
from at least two stations is needed. The meteor cameras measure the azimuth and altitude of the
observed meteor as it moves across the sky. From each station a series of lines of sight is calculated.
Then, an estimation of the best fitting geometric plane passing through the station and its lines of sight is
made. This is made by taking the cross product between the lines of sight at each station and taking the
mean value of them. Now there are two different geometrical planes that the meteor has been passing
through. By taking the cross product of the two planes normal vectors, the directional vector of the
meteor’s path is computed.

𝑣̅! × 𝑣̅" = 𝑛&! (1)

The normal vector for each plane, 𝑛! is derived by taking the cross product of the lines of sight, 𝑣! and
𝑣" at the same station, equation (1). Repeating the same procedure for the lines of sight of the second
stations provides another normal vector, 𝑛" .

𝑛&! × 𝑛&" = 𝑣̅#$%&'($)* (2)

By taking the cross product of the normal vector for the corresponding planes, intersecting the planes, the
directional vector of the meteor, 𝑣̅#$%&'($)* is obtained.

𝑛&!
UPPSALA UNIVERSITET 8 (32)

𝑛&"

𝑣̅"
𝑣̅!
𝑣̅#$%&'($)*

Figure 4 Visual description of the intersecting plane method. Two stations, with the minimum of two
lines of sight per station.
Most often there are more than two stations collecting data of the same meteor. The same calculation is
made between all the different pairs of stations. To obtain a good quality solution, the angle between the
lines of sight between the stations should be as large as possible. If two stations record the meteor from a
similar angle, the angle between the sets of lines of sight is smaller, closer to the margin of error, which
makes the calculated trajectory less reliable. It can be corrected by taking a weighted average of the
calculated directional vectors, where the angle between the sets of lines of sight is considered as weights.
This provides a decent solution for the meteor’s direction (Gural 2012).

4.1.2 Lines of sight method


Each station makes a set of observations for the meteor’s positions, during its way through the
atmosphere. These observations can be formulated as multiple lines of sight, graphically shown as a fan
of rays originating from the coordinates of the station. For a perfect solution, all of the lines of sight,
from all stations, should intersect with the same 3D-line, which is the meteor’s trajectory. Technically,
the solution for the trajectory is the 3D-line with a minimal distance to all lines of sight, this is called the
lines of sight method, LoS. To obtain the trajectory with a minimal distance to all lines of sight, a linear
regression approach is needed. Often, the well-known least squares approach is used, which is the one
the Swedish algorithm use. The method of least squares minimizes the sum of the squared residuals.
Residuals, in this case, are the distance between the estimated trajectory and the lines of sight. (Vida;
Brown; Campbell-Brown; Wiegert and Gural. 2020)
UPPSALA UNIVERSITET 9 (32)

Figure 5 Visual description of the LoS method, three stations with multiple lines of sight

4.1.3 Multiparameter fit


The meteor cameras collect several parameters, the idea of the multiparameter fit is to solve for multiple
parameters simultaneously. Unlike the intersecting plane method where the solution is only based on
geometrical properties, this method uses more of the available information. One parameter is the time
parameter, information about the time at each observation can be valuable for meteor trajectory
estimation. The time parameter is included by using a 3D-propagation model and minimize relative to the
timing information at each point.
Most meteors only last for a few seconds before they burn up, and they enter the atmosphere at high
velocity. Because of this, most meteors follow a linear path. The heavy and bright meteors, that make it
deeper into the earth’s atmosphere, cannot be assumed to follow a linear path, as they noticeable bend
due to earth’s gravity. The software for multiparameter fit made by Gural assumes that the meteor’s
propagation follows one of three choices (Gural 2012).
𝑋(𝑡) = 𝑣+ 𝑡 (3)
𝑋(𝑡) = 𝑣+ 𝑡 − 𝑎𝑡 " (4)
!
𝑋(𝑡) = 𝑣+ 𝑡 − 𝑎𝑒 +( (5)
Most meteors follow a linear path, equation (3), meteors that penetrate deeper into the Earth’s
atmosphere follow a quadradic- or exponential path, equation (4) and (5). But as most meteors are small,
the estimation that a meteor follows a linear path can be made (Gural 2012).
To be able to minimize the solution relative to the time, the solution requires an initial guess for the
meteor’s position and velocity. One way to calculate an initial guess for the trajectory is to use the
intersecting plane method followed by the LoS method, which gives a firm first guess for the meteor’s
trajectory. A first guess for the meteor’s velocity can be made by taking the travelled distance between
the observation of maximum - and minimum height above sea level and divide this distance with the time
it took for the meteor to travel between the two observation points. The propagation model predicts
where the meteor should be observed along the track, the goal is to find a trajectory that satisfies the
measurements. The trajectory solver minimizes a cost function with respect to the actual measurements.
UPPSALA UNIVERSITET 10 (32)

In the case of unsynchronized meteor cameras, where the time stamps between the different stations is
not comparable, due to clock setting errors or other reasons, time offsets between the stations are needed.
The timing offsets are then integrated in the cost function as free variables, where the trajectory solver
finds the best fitting time offsets between the cameras.

4.1.4 Monte Carlo


The Monte Carlo method builds on previously described techniques, the intersecting plane method and
the lines of sight method, what is new in this method is that noise is added to the data and the trajectory
is recalculated, as a way to handle the uncertainty in the measurements. Intersecting plane method and
LoS are first applied to obtain a first estimate of the trajectory vector. Then taking the angular residuals
between the measurements and the estimated trajectory, these residuals are used as an estimate of the
uncertainty. By taking the standard deviation of these uncertainties, Gaussian noise is added to the data
and the trajectory is recalculated. Redoing this procedure, provides a number of trajectories that fits
within the measurement uncertainty. The best solution is chosen for the trajectory which has the most
consistent trajectory, the trajectory that responds the least to the Gaussian noise. This is done by
comparing the observed dynamic between different stations. This method is often used in trajectory
estimation, but it is not currently integrated in the Swedish algorithm and will not be used during this
project. (Vida; Brown; Campbell-Brown; Wiegert and Gural. 2020)

4.2 Optimizing the trajectory

4.2.1 Chi-Square test


In the LoS- and multiparameter fit method, the directional vector is minimized relative to the distance-
and time residuals. In the first method, LoS, the vector is minimized in relation to the distance residuals,
which is done using the method of least squares.
.
(𝑥$ /𝑥&,- )"
𝑥'" =5 (6)
𝑘
$/0

A Chi-Square test is formulated so the quality of the proposed solution can be evaluated, equation 6.
Where 𝑥'" is the chi-square value, 𝑥$ is the residual (the length between the line of sight and the
trajectory), and 𝑥&,- is the expected residual due to measuring inaccuracy.

Combining this, with a chi-square test for the time residuals and solve for both of them simultaneously
will give a trajectory solution that includes the time variable. As discussed above, most meteor’s follow a
linear path. Assuming the meteor follow a linear path, its position propagation model can be formulated
as:

𝑋(𝑡) = 𝑥0 + 𝑣̅ ∙ 𝑡 (7)

Where 𝑋(𝑡), is the distance in km from the first observation of the meteor, 𝑣̅ is its velocity and 𝑡 is the
time in seconds from the first observation and 𝑥0 is the distance for the first observation. For a first
estimate, 𝑥0 should be zero, but when calculating the best fitting slope 𝑥0 might not be zero due to
measurement errors. By making an estimate for what the variables 𝑥0 and 𝑣̅ , the position for the meteor
can be predicted. Each observation has a time stamp, this is now used to calculate the time residuals.
Calculating the observations distance from the start coordinate gives a distance to solve for 𝑡 in equation
UPPSALA UNIVERSITET 11 (32)

(7). The calculated distance depends on the estimated trajectory vector. The difference between the
calculated time and the observed time are the time residuals. A chi-square test is formulated for the time
residuals:
.
(𝑡$ /𝑡&,- )"
𝑡'" =5 (8)
𝑘
$/0

Where 𝑡'" is the chi-square value for the time residuals, 𝑡$ is the time residual and 𝑡&,- is the expected
time residual, according to the accuracy of the measuring equipment. When these two chi-square tests are
formulated, they can be combined into a combined cost function to solve the trajectory, for both these
parameters.
.
(𝑥$ /𝑥&,- )" + (𝑡$ /𝑡&,- )"
𝑟' = 5 (9)
𝑘
$/0

Where 𝑟' is the combined value of 𝑡'" and 𝑥'" .

The residuals in equation (9) are normalized, divided with the expected residual. But it is hard to know
exactly how large the expected residual is, the measuring equipment for the angle determination has to be
recalibrated regularly to make sure it operates as expected. Measuring time however is easier, meaning
the time measurements should be more reliable than the angle measurements. To make sure the
optimization takes both the spatial- and time measurements into consideration, the chi square value for
them separately has to be checked so that they are in balance.
.
(𝑏 − 1) ∙ (𝑥$ /𝑥&,- )" + 𝑏 ∙ (𝑡$ /𝑡&,- )"
𝑟' = 5 (10)
𝑘
$/0

Integrating the factor 𝑏, equation (10), the balance between the spatial- and time residuals can easily be
corrected if the optimization tend to lean more on one of the two residuals.

4.2.1 Powell’s method


The optimization is done by using Powell’s conjugate direction method. Powell’s method is used to find
the minimum of a function of several variables, without having to calculate derivatives, it only uses
function values. The method goes through different stages to find the minimum, and each stage has a
sequence of 𝑛 + 1 one-dimensional searches. 𝑛 is the number of variables of the function. In other
words, the function as two loops, one outer loop and one inner loop. The only purpose of the outer loop
is to repeat the inner loop until no significant improvements are made. The inner loop does 𝑛 number of
searches along every linearly independent direction. The last search is along the direction connecting the
end point obtained by the 𝑛 previous searches with the starting point of the stage. Then, the first of the 𝑛
search directions is replaced by the last search direction. This is repeated until the minimum of the
function is obtained, in this case the minimum of the function is found when the minimum of the cost
function, equation (10), is obtained (Fang, Sun and Gu, 2015). Powell’s method is integrated in the
algorithm from the python package scipy, the name of the function is fmin powell.
UPPSALA UNIVERSITET 12 (32)

4.3 Coordinate transformations

4.3.1 Topocentric coordinates to geocentric Cartesian coordinates


For meteor trajectory estimation, the geocentric Cartesian coordinates are preferred, geocentric
coordinates originate in the center of the earth. The data from the meteor cameras are given in
topocentric coordinates, which are coordinates in the reference of earth’s surface. This means that a
coordinate transformation is required to transform the coordinate system from the Earth’s surface to the
center of Earth. Also, a transformation between altitude-, azimuth coordinates to Cartesian coordinates is
needed, as the calculations is preferably done in Cartesian coordinates.

Figure 6 Geocentric (red) vs topocentric (blue) coordinate system

Beginning with the transformation to cartesian coordinates, the lines of sight vectors need to be
transformed from topocentric-geographical coordinates to geocentric Cartesian coordinates. Starting with
a vector pointing in the negative x-direction:

−1
𝑣̅ = A 0 B
0

And then rotating it to the right altitude, 𝛼, which is a rotation around the y-axis. The rotation has to be
counterclockwise around the y-axis, so the rotation angle is negative. 𝑅 2 is the matrix for this rotation:

1 0 0
𝑅 2 = A0 cos(− 𝛼) sin(𝛼) B
0 sin(− 𝛼) cos(− 𝛼 )

Then this same vector is rotated according to the azimuth, 𝛾, which translates to rotation around the z-
axis:
UPPSALA UNIVERSITET 13 (32)

cos(𝛾) − sin(𝛾) 0
𝑅 3 = A sin(𝛾) cos(𝛾) 0B
0 0 1

After these two steps, the vector is described in Cartesian coordinates in the topocentric coordinate
system. To get the vector in geocentric coordinates the topocentric (blue) coordinate system will be
rotated for its location in longitude and latitude.

First the topocentric (blue) coordinate system in figure 6 is rotated for its latitude 𝜑. 90-𝜑 degrees around
the y-axis.

cos(90 − 𝜑) 0 sin (90 − 𝜑)


𝑅4 = A 0 1 0 B
−sin (90 − 𝜑) 0 cos (90 − 𝜑)

Then the coordinate system is rotated for its longitude, 𝜆. The goal is to line up the blue coordinate
system with the red, geocentric, coordinate system. This is done by rotating −𝜆 degrees around the z-
axis.

cos(−𝜆) sin( 𝜆) 0
𝑅5 = Asin(− 𝜆) cos(−𝜆) 0B
0 0 1

Doing these four rotations, the coordinates for the lines of sight vectors in cartesian geocentrical
coordinates are obtained:

𝑣̅6&) = 𝑅 2 𝑅 4 𝑅 3 𝑅5 𝑣̅

4.3.2 Transformation between geodedic- and geocentric Cartesian coordinates

The reference points for all of the lines of sight are the coordinates for the stations, which are obtained in
longitude, latitude and height above sea level. Conversion between Cartesian coordinates and geodetic
coordinates for an ellipsoid can be written as:

(𝑁 + ℎ)𝑐𝑜𝑠𝜑𝑐𝑜𝑠𝜆
𝑥
M𝑦P = Q (𝑁 + ℎ)𝑐𝑜𝑠𝜑𝑠𝑖𝑛𝜆 X
𝑧 (𝑁(1 − 𝑒 " ) + ℎ)𝑠𝑖𝑛𝜑

Where 𝜑 is the latitude, 𝜆 is the longitude and h is the height above the ellipsoid.
UPPSALA UNIVERSITET 14 (32)

𝑁 is the radius of curvature of the prime vertical

𝑎
𝑁=
Y1 − 𝑒 " 𝑠𝑖𝑛" 𝜑

𝑎 is the semi major axis of the ellipsoid. 𝑒 is the first eccentricity, corrected for the flatness of the Earth,
!
𝑓. The flattening factor is a unitless number, 𝑓 = "78.":;""<:=< (Maus, 2010).

𝑒 = Y2𝑓 − 𝑓 "

The eccentricity, 𝑒, is gained from equation (). Following the conversion above provides the Cartesian
coordinates for the stations, which are the reference points for the lines of sight. (Reit, Bo-Gunnar. 2009).

4.3.3 Calculating the radiant for the meteor


To evaluate different trajectory estimation approaches, the radiant for the meteor events has to be
calculated. Meteors from the same shower are originating from the same point in the sky, the radiant.
Looking at multiple meteors from the same shower and comparing their radiant positions, which should
be close or equal to each other, gives a measurement of the accuracy of the method. Calculating the
radiant is quite complex, all observations are made on the rotating Earth, and we want to know from
which direction in space the meteors came from. This means that there are several aspects to take into
consideration, the Earth’s rotation around its own axis and the rotation around the sun.

To be able to compare the coordinates for the radiant, they need to be described in the equatorial
coordinate system. In this project, we use a geocentric equatorial coordinate system, originating from the
centre of Earth with the vernal equinox as the x-direction and the north pole as the z-direction, the y-axis
is defined in reference to the x- and z-axis so it matches a right-handed coordinate system. The vernal
equinox is the intersecting point between the celestial equator and the ecliptic where the sun is going
from south to north declination. The coordinates are described in two angles, right ascension and the
declination. The right ascension (RA) is the angle from the vernal equinox along the celestial equator, 0°
RA means that the object is located in the direction of the vernal equinox. The declination is the angle
from the celestial equator, positive in the north direction, negative in the south direction. 0° Dec means
that the object is located in the plane of the celestial equatorial.

The conversion to the geocentric equatorial coordinate system is made by using a software, the Python
package ephem. The software calculates the Earth’s position at the time the observation was made and
transforms the altitude and azimuth angles of incidence of the meteor (where the meteor enters the
atmosphere), to equatorial coordinates, RA and Dec.

4.4 Application of methods on Swedish data

4.4.1 Perseid meteor shower


There is a lot of data available from the Swedish Allsky Meter Network, in this project the Perseids from
2020 will be the study object. The Perseids is a meteor shower that can be seen from the northern
hemisphere in August. Its radiant is around 48° RA and 58° Dec, shifting a little bit throughout August,
between 37°-63° RA and 56°-58° Dec. The Perseids velocities is usually around 60 km/s (Jürgen 2016).
UPPSALA UNIVERSITET 15 (32)

The data is chosen by filtering out the events that lies 10° away from the expected radiant position for a
Perseid meteor.

4.4.2 Intersecting plane method for first estimation of the trajectory


The Swedish Allsky Meteor Network provide data files for every observed meteor. One of the files
contain two observations from every station that observed the meteor, these two observations is the first
and the last pixel where the meteor was seen. This file is read and used for making an initial guess for the
velocity vector of the meteor. The input data contains the geographic coordinates for the station and the
name of the station. There are two observations for each station, so there is azimuth and altitude values
for the first- and the second observation.

Data from this first file is converted to two lines of sight per station by doing coordinate transformation
from the topocentric altitude and azimuth coordinates, to geocentric Cartesian coordinates. Cartesian
geocentric position coordinates are obtained by transforming the geographic coordinates for each station.
Now the initial guess for the velocity vector can be calculated by using the intersecting plane method.

Taking the cross product for each pair of two vectors from the same station, gives a geometrical plane,
which passes both the meteor trajectory and the station. Then again taking the cross product, but this time
between every pair of normal vectors between the planes, these product vectors are initial guesses for the
velocity vectors. One problem about using cross-product to determine the velocity vector is that it could
point in the wrong direction, depending on which order the normal vectors are multiplied. This is easily
corrected by checking that all the vectors are going in the right direction. The intersection between every
line of sight and the plane intersecting the line provides Cartesian point, finding the highest point and the
lowest point gives an estimate for the direction of the meteor. To check that the directional vectors are in
the right direction, the scalar product between the directional vector and the vector between the highest
and the lowest point should be larger than zero.

A good guess for the initial vector is taking a weighted mean value of every coordinate for these vectors,
providing a mean value-vector as the initial guess for the meteor’s directional vector. Weighting the
mean value for the angle between the stations lines of sight is done naturally, the length of the directional
vectors depends on the angle between the planes of intersection. If the angle between them is small, the
product of their normal vectors will also be small. Therefore, directional vectors that is obtained from
planes with small angles between them will contribute less in the mean value.

4.4.3 Calculating the trajectory by using the lines of sight method


When the initial guess is obtained, it can be used as a reference for improving the solution for several
lines of sight. This method is described as the lines of sight method, where you use several lines of sight
and find a trajectory that has the minimal distance to all lines of sight.

This is done by using another type of file that the Swedish Allsky Meteor Network provides, these files
contain all the observations for each station for one meteor, which means they contain a lot more
information than the previous file. Every line in this file tells at what altitude and azimuth the meteor was
observed, the name for the station and a date and time stamp. Lines of sights are created in the same way,
and the names for the stations are associated with their coordinates given in the previous file. The
directional vector can now be optimized for the minimal distance between the lines of sight vectors and
the estimated meteor track.

First, a Chi-Square test is formulated so the quality of the proposed solution can be evaluated:
UPPSALA UNIVERSITET 16 (32)

.
(𝑥$ /𝑥&,- )"
𝑥'" =5 (6)
𝑘
$/0

Where 𝑥'" is the chi-square value, 𝑥$ is the residual (the length between the line of sight and the
trajectory), and 𝑥&,- is the expected residual due to measuring inaccuracy.

4.4.4 Adding time variable to the optimization


Previous steps provide a good initial vector for the trajectory solution. The goal of this project is to
optimize the solution further by including the time variable. To do this, a new cost function for both the
distance residuals and the time residuals has to be formulated. A chi-square test for the distance residuals
is already formulated from previous phases. Combining this with a chi-square test for the time residuals
and solve for both of them simultaneously will give a trajectory solution that includes the time variable.

As discussed above, most meteor’s follow a linear path. Assuming the meteor follow a linear path, its
position propagation model can be formulated as:

𝑋(𝑡) = 𝑥0 + 𝑣̅ ∙ 𝑡 (7)

Where 𝑋(𝑡), is the distance in km from where the first observation of the meteor, 𝑣̅ is its velocity and 𝑡 is
the time in seconds from the first observation and 𝑥0 is the distance for the first observation. For a first
estimate, 𝑥0 should be zero, but when calculating the best fitting slope 𝑥0 might not be zero due to
measurement errors.

By making an estimate for what the variables 𝑥0 and 𝑣̅ , the position for the meteor can be predicted. Each
observation has a time stamp, this is now used to calculate the time residuals. Calculating the
observations distance from the start coordinate gives a distance to solve for 𝑡 in equation (7). The
calculated distance depends on the estimated trajectory vector. The difference between the calculated
time and the observed time are the time residuals. A chi-square test is formulated for the time residuals:
.
(𝑡$ /𝑡&,- )"
𝑡'" =5 (8)
𝑘
$/0

Where 𝑡'" is the chi-square value for the time residuals, 𝑡$ is the time residual and 𝑡&,- is the expected
time residual, according to the accuracy of the measuring equipment.

When these two chi-square tests are formulated, they can be combined into a combined cost function to
solve the trajectory, for both these parameters.
.
(𝑥$ /𝑥&,- )" + (𝑡$ /𝑡&,- )"
𝑟' = 5 (9)
𝑘
$/0
UPPSALA UNIVERSITET 17 (32)

Where 𝑟' is the combined value of 𝑡'" and 𝑥'" .

Then Powell’s method is applied once again to minimize the solution, but for the combined cost function
(9). By using Powell’s method, the best fitting trajectory vector and variables for the position model will
be obtained.

4.4.5 Adding time offsets between the stations


In the article A new method of meteor trajectory determination applied to multiple unsynchronized video
cameras, Gural assumes that the cameras are unsynchronized and takes that into consideration by
applying time offsets between the stations (Gural 2012). The offsets are applied as free variables and
solved for the best fitting trajectory. Suspecting that the timestamps from the Swedish data are a bit
unsynchronized, time offsets are added. For an event with 𝑛 number of stations, 𝑛 − 1 time offsets are
integrated, meaning that we no longer use the absolute timestamp, only the relative time within each
station. One station is randomly selected as the reference point, and the other stations gets an offset as a
free variable. The trajectory is solved in the same way with Powell’s method, but with 𝑛 − 1 more
variables.

5 Results

5.1 Looking at a single meteor event


To evaluate all the steps in this project, a single event will first be reviewed. Following a meteor seen
2020-08-12 and evaluating all the results for each step in the algorithm.

Figure 7 Picture describing from which stations the meteor was seen (Swedish Allsky Meteor
Network 2021), with permission from Eric Stempels.
5.1.1 Intersecting plane method
For the intersecting plane method, a smaller file is used, containing two lines of sight per station. The
meteor 2020-08-12 was recorded by five different stations.
meteor_2020-08-12T23-48-06-240/meteor_2020-08-12T23-48-06-240-dat.txt

Station Longitude Latitude Azimuth 1 Azimuth 2 Altitude 1 Altitude 2 Time [s]


UPP 17.6477° 59.8376° 300.6° 296.4° 29.5° 22.5° 1597276086.000
UPPSALA UNIVERSITET 18 (32)

UAA 17.1300° 59.8000° 307.0° 301.2° 33.2° 25.3° 1597276085.940


HBRG 14.6447° 60.2411° 21.2° 2.5° 66.1° 67.3° 1597276086.000
VARF 16.4500° 59.6300° 318.3° 314.4° 33.4° 29.3° 1597276086.210
HAR 10.75175° 60.21191° 76.19° 77.82° 22.37° 18.6° 1597276084.700
Table 1: Data containing two lines of sight per station, meteor 2020-08-12

Table 1 contains the data for the meteor, degrees for the angles and seconds for the time stamp. To make
an estimation for the track, the intersecting plane method is applied. The first steps are to convert these
10 lines of sight to Cartesian coordinates and convert the stations geographical coordinates to Cartesian
coordinates.

𝑥[ 𝑦[ 𝑧̂
0.09785772 -0.75501701 0.64836193
0.09565243 -0.83796736 0.5372721
0.0441273 -0.68569037 0.72655454
0.04638991 -0.79491967 0.60493858
0.08475085 0.1735753 0.98116712
0.11494992 0.04743621 0.99223804
-0.09159285 -0.60611316 0.79008707
-0.09114512 -0.67657407 0.73071205
-0.16998424 0.88178218 0.43996084
-0.18765036 0.90735573 0.37615544
Table 2: Lines of sight, cartesian coordinates

Table 2 shows the 10 lines of sight in Cartesian coordinates, obtained by coordinate transformation. Each
pair of lines from the same station span a geometrical plane for which the meteor and the station lies in.

𝑥[ 𝑦[ 𝑧̂
0.99515843 0.06825291 -0.07071949
0.99887253 0.04302614 -0.02006035
0.96755838 0.22087927 -0.12265044
0.99579634 -0.05524015 0.07306288
-0.95185452 -0.26249325 0.15833592
Table 3: Normal vectors for each geometrical plane

Taking the cross product between every pair of vectors, the normal vectors describing the planes are
obtained, table 3. The directional vectors for the meteor track can be estimated by taking the cross
products between every pair of normal vectors in Table 3.

𝑥[ 𝑦[ 𝑧̂
-0.00167361 0.05067653 0.02535813
0.00724922 0.05363139 0.15377119
-0.0010802 0.14313135 0.12293869
0.0077565 0.09025466 0.19625553
-0.00084626 0.10310259 0.17899993
-0.00203548 0.09295653 0.09802314
-0.00154688 0.13906286 0.22124267
UPPSALA UNIVERSITET 19 (32)

-0.00936285 0.19282746 0.27339883


-0.00277821 0.03645387 0.04373261
-0.01043201 0.22721556 0.3139704
Table 4: Estimated directional vectors for the meteor’s trajectory

Table 4 shows all the estimated directional vectors from taking the cross products between each pair of
normal vectors. By taking an average of all these vectors we get the directional vector from the
intersecting plane method:

]]]]]⃗
𝑣>? = [ 0.00744507, −0.57002979, −0.82159029]

This vector 𝑣
]]]]]⃗
>? corresponds to the radiant 49.2° RA and 55.2° Dec.

(b) The computed meteor track using (a) Distance residuals, the shortest distance
the IP method between each observation and the computed
track

Figure 8 The computed directional track and the distance residuals, the distance between the actual
measurements and the computed track.

Figure 9 The travelled distance from the start coordinate (which is determined by the coordinate
with the maximum height) vs the time
UPPSALA UNIVERSITET 20 (32)

In Figure 9 it is clear that the time information does not match the calculated trajectory. The distance
residuals in Figure 8 (b) are also far off the computed track, at least for some observations.

5.1.2. Lines of sight method


When applying the LoS method, a larger file describing the event is used. This file contains more
measured observations, which creates more lines of sight. Each line in this file contains information
about the name of the station, the altitude- and azimuth angles and a time stamp.

Altitude [°] Azimuth [°] Station Date Timestamp


66.13 21.15 HBRG 2020-08-13 01:48:06.00
66.44 18.75 HBRG 2020-08-13 01:48:06.08
66.59 17.51 HBRG 2020-08-13 01:48:06.11
66.78 16.14 HBRG 2020-08-13 01:48:06.19
66.84 14.59 HBRG 2020-08-13 01:48:06.22
67.01 11.37 HBRG 2020-08-13 01:48:06.30
67.11 9.86 HBRG 2020-08-13 01:48:06.32
67.15 6.14 HBRG 2020-08-13 01:48:06.40
67.12 4.48 HBRG 2020-08-13 01:48:06.43
67.07 2.84 HBRG 2020-08-13 01:48:06.51
67.28 2.51 HBRG 2020-08-13 01:48:06.54
33.21 307.03 UAA 2020-08-13 01:48:05.94
31.81 305.87 UAA 2020-08-13 01:48:06.04
30.28 304.64 UAA 2020-08-13 01:48:06.14
28.98 303.74 UAA 2020-08-13 01:48:06.24
27.25 302.58 UAA 2020-08-13 01:48:06.34
25.44 301.25 UAA 2020-08-13 01:48:06.44
25.30 301.20 UAA 2020-08-13 01:48:06.54
29.49 300.61 UPP 2020-08-13 01:48:06.00
28.95 300.32 UPP 2020-08-13 01:48:06.03
28.26 300.10 UPP 2020-08-13 01:48:06.08
27.59 299.65 UPP 2020-08-13 01:48:06.12
27.13 299.21 UPP 2020-08-13 01:48:06.16
26.48 298.80 UPP 2020-08-13 01:48:06.20
25.88 298.41 UPP 2020-08-13 01:48:06.24
25.27 298.13 UPP 2020-08-13 01:48:06.28
24.87 298.00 UPP 2020-08-13 01:48:06.32
24.33 297.51 UPP 2020-08-13 01:48:06.36
23.37 296.94 UPP 2020-08-13 01:48:06.40
22.42 296.45 UPP 2020-08-13 01:48:06.43
22.45 296.40 UPP 2020-08-13 01:48:06.48

Table 5 Data from three stations for multiple lines of sight for the meteor 2020-08-12, containing
one line of sight per row. The Norwegian station HAR is not providing information for this file.
VARF, Västerås, is not in this file either.

𝑥[ 𝑦[ 𝑧̂
0.08524745 0.17318577 0.98119293
UPPSALA UNIVERSITET 21 (32)

0.08980557 0.15626401 0.98362418


0.09223012 0.14765393 0.98472937
0.09552088 0.13824037 0.9857816
0.09678914 0.12769154 0.9870799
0.10100676 0.10597861 0.98922503
0.10369834 0.09594093 0.9899707
0.10775325 0.07108607 0.991633
0.10919967 0.05992494 0.99221189
0.11056329 0.04884375 0.99266815
0.1145969 0.04742808 0.99227926
0.04385594 -0.68541815 0.72682777
0.04494762 -0.70674797 0.70603613
0.04622747 -0.72922815 0.68270735
0.04586121 -0.74707839 0.66315204
0.04536206 -0.76991506 0.63653208
0.04695848 -0.79337987 0.60691291
0.04638991 -0.79491967 0.60493858
0.09764018 -0.75508266 0.64831826
0.0967981 -0.76186227 0.64046546
0.09377048 -0.76983408 0.63131813
0.09402112 -0.77836466 0.62073222
0.09599774 -0.78462484 0.6124935
0.09600321 -0.7925681 0.6021787
0.09621588 -0.79982588 0.5924703
0.09488301 -0.80668849 0.58331029
0.09328721 -0.81095004 0.57763096
0.09549867 -0.81768983 0.56768244
0.09536188 -0.82842301 0.5519252
0.09427998 -0.83852865 0.53663861
0.0952347 -0.83841383 0.53664943
Table 6 Lines of sight, cartesian coordinates

The data from table 5 is converted to lines of sight in Cartesian coordinates. Using Powell’s method, the
trajectory is optimized by minimizing the distance residuals, the distance between the trajectory and each
line of sight. The result from the IP method is used as a first guess for the directional vector.

𝑣⃗@)A = [ 4.57054010 ∙ 10BC , −0.547243752, −0.836973158]

The vector 𝑣⃗@)A is the resulting vector from the LoS method and corresponds to the radiant 48.7° RA and
56.8° Dec.
UPPSALA UNIVERSITET 22 (32)

(a)Adding
4.1.3. The computed meteor
the time track using the
variable (b) Distance residuals, the shortest distance between
LoS method each observation and the computed track

Figure 10 The computed directional track and the distance residuals. The distance residuals are the
distances between the actual measurements and the computed track.

Figure 11 The travelled distance from the start coordinate (which is determined by the coordinate with
the maximum height) vs the time

Looking at Figure 10 (b) and Figure 11, it is obvious that this trajectory fit the actual measurements
better than the IP-solution in 4.1.1. The residuals are a lot smaller in Figure 10 (b) and the time
measurements fit the trajectory better.

5.1.3. Adding the time variable


Adding the timestamps from Table 5 to the solution and solving for the time and distance simultaneously
gives another solution to the directional vector and the radiant position.

𝑣⃗($D& = [ 0.30094051, -0.5145356, -0.82183076]


UPPSALA UNIVERSITET 23 (32)

The vector 𝑣⃗@)A is the resulting vector from the LoS method and corresponds to the radiant 48.2° RA and
58.2° Dec.

(a) The computed meteor track when (b) Distance residuals, the shortest distance
adding the time variable between each observation and the computed
track
Figure 12 The computed directional track and the distance residuals. The distance residuals are the
distances between the actual measurements and the computed track.

(b) The travelled distance vs time for the (a) The time residuals, the difference
meteor, the slope is the fitted between the actual time measurements
propagation model, and the blue and the time according to the
points are the observations. propagation model

Figure 13 (a) Showing the fitted slope, travelled distance as a function of time and (b) showing the time
residuals.

The velocity for the meteor, the slope in Figure 13 (a) is 66.5 km/s, which is high for a Perseid meteor.
UPPSALA UNIVERSITET 24 (32)

5.1.4. Adding the timing offsets between the stations


As seen in Figure 11 and Figure 13 (b), it almost looks like the time stamps follow different tracks. This
could be because of clock setting errors between the cameras. To compensate for this possible effect,
timing offsets between the stations are added, providing another result for the directional vector.

𝑣⃗)EEF&( = [ 0.33503881, -0.69450296, -0.95234333]

The vector 𝑣⃗)EEF&( corresponds to the radiant position 49.4° RA and 54.2° Dec.

(a) The computed meteor track (b) Distance residuals, the shortest distance
when adding the time variable between each observation and the computed
track
Figure 14 The computed directional track and the distance residuals. The distance residuals are the
distances between the actual measurements and the computed track.

UAA, a station in Uppsala is randomly chosen as the fixed time. The other two stations in Table 5, UPP
and HBRG get an offset as a free variable and the optimization finds the best fitting offsets for them in
reference to UAA.
UPPSALA UNIVERSITET 25 (32)

(a) The travelled distance vs time for the (b) The time residuals, the difference
meteor, the slope is the fitted between the actual time measurements
propagation model, and the blue and the time according to the
points are the observations. propagation model

Figure 15 (a) Showing the fitted slope as travelled distance as a function of time and (b) showing the
time residuals.

The velocity for the meteor, the slope in Figure 15 (a) is 57.6 km/s, which is in the range of expected
velocity for a Perseid meteor. UAA was chosen as the fixed time frame and the timestamps from UPP
was moved -0.021 seconds and the timestamps from HBRG was moved -0.211 seconds.
UPPSALA UNIVERSITET 26 (32)

5.2 All results

5.2.1 Intersecting planes method


The first step is to identify the Perseid meteor shower, this is done by plotting all 52 meteor events that
were observed by the Swedish Allsky Network during August.

(a) All meteor events during August (b) Choosing the Perseid meteor events
Figure 9 (a) Graph showing radiant positions for all meteor events in August 2020 using the IP method.
The value d is the median value for the deviation between the radiant positions and the median value
radiant (the orange point). (b) The meteor events in a radius of 12° from the median value radiant are
chosen.

Figure 9 Histograms showing the spread in Declination and Right Ascension for the set of Perseid meteor
events, using IP method.

Once the 52 meteor events are plotted, the Perseid meteor shower can be identified, see Figure 8 (a). The
mean value radiant is calculated and marked in the plot. 40 meteors with a radiant position deviating less
than 12° from the median value radiant is chosen as the set of Perseid meteors, see Figure 8 (b). The
standard deviation for the radiant positions using the intersecting plane method is 4.86°.
UPPSALA UNIVERSITET 27 (32)

5.2.2. Lines of sight method

Figure 10 Graph showing the radiant positions for the set of perseid meteor events. Eight of the meteor
events change its radiant positions when LoS is applied. The set shown in this figure excludes these eight
events.

Adding the LoS method to the chosen set of 40 Perseid meteor events and using more lines of sight gives
a different result. Eight of the meteors from the set chosen in 5.2.1 differ a lot when LoS is applied. This
can happen because of various reasons, one of them is that the stations that recorded the meteor did it
from similar angles, this can create big differences when calculating the trajectory with all lines of sight.
Figure 10 shows the radiant positions for the set of Perseid meteor events, excluding the eight events that
differ more than 10° from the IP solution. The standard deviation for the radiant positions using the LoS
method is 4.80°, meaning the radiant positions are closer when the LoS method is applied.

Figure 11 Histograms showing the spread in Declination and Right Ascension for the set of Perseid
meteor events, using LoS method.
UPPSALA UNIVERSITET 28 (32)

5.2.2. Adding the time variable

Figure 12 Graph showing the radiant positions for the set of Perseid meteor events when the time
variable is included. The set shown in this figure excludes the same eight events that differ a lot from the
IP solution.

The mean value for the radiant positions is 4.47°, meaning the radiant positions are closer when the time
variable method is included in the algorithm.

Figure 13 Histograms showing the spread in Declination and Right Ascension for the set of Perseid
meteor events, when the time variable is included in the algorithm.

When solving for the time parameter the velocity for the meteor is also obtained. Perseid meteors should
have velocities around 60 km/s, velocities above 60 km/s are unlikely.
UPPSALA UNIVERSITET 29 (32)

Figure 14 Histogram of the velocities for the meteors

The velocity for the meteors is shown in figure 12, except for six events that have velocities above 90
km/s. Knowing this is very unlikely, these might be the result of nonsynchronization between the
cameras.

5.2.2. Adding time offsets for the stations


To compensate for possible nonsynchronization between the stations, the same set is solved with the time
variable with offsets between the stations.

Figure 15 Graph showing the radiant positions for the set of Perseid meteor events when the time
variable and timing offsets is included. The set shown in this figure excludes the same eight events that
differ a lot from the IP solution.
UPPSALA UNIVERSITET 30 (32)

The mean value for the radiant positions is 4.57°, meaning the radiant positions are slightly further away
from each other when solving for the timing offsets.

Figure 16 Histograms showing the spread in Declination and Right Ascension for the set of Perseid
meteor events, when the time variable is included in the algorithm.

Figure 17 Histogram of the velocities of the meteors

Plotting the velocities when solved with time offsets, only two meteors get velocities that is above 90
km/s. There is also a more distinct peak at 57-60 km/s, which is the expected velocity for a Perseid
meteor.
UPPSALA UNIVERSITET 31 (32)

6 Evaluation and conclusions


Currently, the Swedish algorithm only uses the first file presented in the method, the file containing two
lines of sight per event. When more lines of sight are used, the radiant positions are closer to each other,
which means the solution is more accurate. This can be seen in the comparison of the IP-solutions and
the LoS-solutions, for the IP-solutions only two lines of sight per station was used, but in the LoS method
all available lines of sight were used. The chi square test gives solutions with smaller residuals when the
LoS method is used, even for events with only two stations the LoS method gives a better result.

By also solving the trajectory for the time, more of the data is used. The radiant positions are also slightly
closer when adding the time variable to the solution. But relying on the absolute timestamp is not the
way to go with the Swedish data. It can be seen in various plots that the time does not seem to follow a
monotonous line, for example Figure 13 (b). The velocities are also higher than the expected velocity for
a Perseid meteor, or unphysical for a meteor at all, see Figure 12. This indicates that there is a
nonsynchronization between the meteor cameras. When the trajectories are solved with the time offsets,
fewer events have unphysical velocities and there is a more distinct peak at the expected velocity
distribution for Perseid meteors, see Figure 17. The fact that the velocities are more reasonable when
solving with time offsets says that there are offsets between the cameras. What the offsets depend on is
still unknown, but it could possibly be delays in the reading of the camera or writing to disc. There are
still uncertainties in the spatial measurements, because there still are some unphysical meteor tracks,
even with time offsets. When a meteor has a velocity direction towards a station, the measurements is not
as reliable, which seems to be the case for most of the unphysical meteor events.

6.1 Recommendations
As the trajectory solutions are more accurate with the time adjustment, a recommendation would be to
implement such an algorithm in the Swedish routine. It is also clear that using more lines of sight gives a
better solution, an algorithm for the LoS method with all lines of sight exists, but it is not used for all
meteor events. All recorded meteors and their solutions are published on the Swedish Allsky Network
website, the solutions that are published are solved with the smaller file which does not contain all lines
of sight. A recommendation would be to implement all of the lines of sight for all meteor events. Time
offsets will need to be considered when implementing an algorithm that solves for the time, the absolute
time stamps cannot be trusted at the moment. There are benefits with absolute timestamps, for example
events that only get recorded from two stations with only two lines of sight will create a mathematically
“perfect” solution but might not be close to the actual track. The lack of measurements in these events
makes it hard to know the uncertainty of the solution, absolute timestamps can be used to review the
uncertainty for these events. To investigate the reason for the clock setting errors, meteors with variable
light intensity can be examined, a meteor with variable light intensity look like a small flash at some
point on the track. Looking at when this flash occurs in the different cameras, the absolute time can be
defined from this. This might show that there are static offsets at every station, if these are corrected the
trajectories can be solved using the absolute time stamp.
UPPSALA UNIVERSITET 32 (32)

7 References
Gural, Peter S. 2012. A new method of meteor trajectory determination applied to multiple
unsynchronized video cameras. Meteoritics & Planetary Science 47(9): 1405-1418.
https://onlinelibrary.wiley.com/doi/epdf/10.1111/j.1945-5100.2012.01402.x (Collected 2021- 02-25)
Maus, S., 2010. An ellipsoidal harmonic representation of Earth's lithospheric magnetic field to degree
and order 720. Geochemistry, Geophysics, Geosystems, 11(6).

Vida, Denis; Brown, Peter G; Campbell-Brown, Margaret; Wiegert, Paul and Gural, Peter S. 2020.
Estimating trajectories of meteors: an observational Monte Carlo approach -I. Theory. Monthly Notices
of the Royal Astronomical Society 491(3): 3996-4011.
https://academic.oup.com/mnras/article/492/4/5313/5731420?searchresult=1 (Collected 2021-02-25)
Reit, Bo-Gunnar. 2009. On geodetic transformations. Rapportserie: Geodesi och Geografiska
informationssystem. Gävle: Lantmäteriet.
https://www.lantmateriet.se/contentassets/4a728c7e9f0145569edd5eb81fececa7/rapport_reit_eng.pdf
(Collected 2021-05-05)
Fang, C., Sun, H. and Gu, J., 2015. Powell’s method-based nonlinear least-squares data fitting for the
Mittag-Leffler relaxation function. Mathematics and Mechanics of Solids, 22(5), pp.1058-1067.
https://journals-sagepub-com.ezproxy.its.uu.se/doi/pdf/10.1177/1081286515616284 (Collected 2021-05-
20)
Rendtel, Jürgen. 2016. 2017 Meteor Shower Calendar. International Meteor Organization.
https://www.imo.net/files/meteor-shower/cal2017.pdf (Collected 2021-05-20)

You might also like