Académique Documents
Professionnel Documents
Culture Documents
ABSTRACT
This document describes the process of determining the orbital elements of asteroid 1998
YP11. The observations were carried out by “Cretaceous Crustaceans” team (Juhee Bae, Marie
Fulmer, Konrad Komorowski) at Etscorn Observatory (New Mexico Tech, 34° 4' 21" N; 106° 54'
52" W, IAU code: 719) between 6/17/08 and 7/12/08 as a part of the Summer Science Program
2008 curriculum.
Orbital elements calculated using Gaussian orbit determination method on data from Fig. 1:
-1-
INTRODUCTION
The goal of the project is to determine the orbital elements of a near-Earth asteroid 1998
YP11. It will be achieved by performing observations of the asteroids at 4 different nights. Various
Python programming language scripts (the core one performing Gaussian orbit determination) will
be then employed to determine a set of 6 standard orbital elements of the asteroid.
Materials
All the data was collected using CCD imaging with a Takahashi 6” refractor. The properties
of the CCD chip are summarized in Fig. 3.
Observations were normally carried out with a time span of 8 ± 4 days. Each of the analyzed
images was taken with a 2×2 binning in order to reduce noise and simplify analysis, which reduced
the images resolution to 1092 px × 736 px. Usual exposure time was 120 sec, with the exception of
the last observation when a shorter exposure of 90 sec was needed to avoid recording asteroid’s trail –
an effect of its faster motion against the background of stars.
Method
Data reduction and observations preparation were performed using a collection of 4 scripts
written in the Python programming language. These included:
-2-
Positions of the asteroid on consecutive observation nights were predicted using a Python
interpolation script (linear interpolation between the previous and next ephemeris data). Ephemeris
data was supplied by the Summer Science Program 2008 at Socorro faculty.
Accurate positions of the asteroid on the obtained CCD images were then determined using
a plate reduction script written in the Python programming language. The script’s algorithm was
prepared by Dr. Kevin Krisciunas. The script itself included modules reading data from an external
CSV1 file, weighting the objects’ positions based on the brightness of adjacent pixels and accounting
for the proper motion of the stars.
Gaussian method was the core of the orbital determination script, also written in Python. Its
computer implementation had been prepared by Dr. Agnes Kim. Monte Carlo method was adopted
for error analysis – the probability distribution function used for randomizing input variables was a
standard Gaussian curve; and the same p.d.f. was assumed for the results interpretation. Gaussian
method was refined by topocentric to geocentric coordinates correction and account for the travel
time of light.
The orbital elements were then inserted into the ephemeris generator script. The ephemeris
was then generated for the time of the other observation, to make sure that the calculated elements
are correct.
1
Comma Separated Value
-3-
DATA AND ANALYSIS
Data collection
The positions of the asteroid (see Appendix 5) were determined accurately from CDD
images taken at four different nights (see Appendices 1 – 4) using the plate reduction script:
Dec (°)
20.00
#4 15.00
10.00
5.00
0.00
19.400 19.375 19.350 19.325 19.300 19.275 19.250 19.225
RA (h)
Fig. 4 – declination vs. right ascension graph for 1998 YP11 observations
Classical orbital elements are given with mean anomaly, M, at the time of perihelion.
However, the crude mean anomaly received from the Gaussian method corresponds to the time of
the middle observation of the object and does not have to be equal to 0° (perihelion). The OD
founds the time (t2) of previous passage though the perihelion (M2 = 0) using the following relation:
ߤ
ܯଵ ൌ ܯଶ ට ሺ ݐെ ݐଶ ሻ
ܽଷ ଵ
ܯଵ
ݐଶ ൌ ݐଵ െ
ߤ
ට
ܽଷ
Data analysis
Four different orbit determinations were performed – each for a different set of observations.
The results of each of them are summarized in Appendices 6 – 9.
-4-
In each of the 4 cases Monte Carlo analysis was performed. It turned out that in some
iterations of the Monte Carlo analysis the script would encounter errors – non-converging series,
significantly different from the one of others or values of cosine/sine greater than one. In such cases
the particular partial results were ignored. Unavoidably, such omissions did contribute to the
skewness of the general result.2
To counteract this unwanted result, the script mirrored each data (here: orbital elements)
against the normal value after obtaining the primary Monte Carlo set of results. Obtained in such a
way new set of partial results was then added to the old one. As the skewness of the new set was
opposite to the skewness of the old one, their sum added to zero. Such a set of results was called
normalized.
It was then noticed that for calculations performed in a shorter time span or with irregular
dates rate of such error omissions is high. What directly corresponds with the skewness:
1
0.9
0.8 a
0.7
0.6 e
0.5 i
0.4
0.3 Ω
0.2 ω
0.1
0 M
#1 : #2 : #3 #1 : #2 : #4 #1 : #3 : #4 #2 : #3 : #4
(5.8%) (38.6%) (0.2%) (10.5%)
Fig. 5 – skew as a fraction of uncertainty (absolute value) for primary partial orbital data for 4
observations sets; in brackets – percentage of failed iterations
Sets for which the results were significantly distorted (as #1:#2:#4 and #2:#3:#4) are not
scientifically valuable. Therefore, they should not be included in the final result.
The uncertainty itself is also a very important factor. Standard deviation from the mean of
the partial data was assumed to be also the result’s uncertainty:
2
Non-skewed result was considered to be the one obtained with zero error analysis, i.e. after performing the
OD for input variables as they were – with no randomization.
-5-
S.d.m. of partial orbital data for 4 observations sets
5
4.5
4 a [AU]
3.5
e
3
2.5 i [°]
2 Ω [°]
1.5
ω [°]
1
0.5 M [°]
0
#1 : #2 : #3 #1 : #2 : #4 #1 : #3 : #4 #2 : #3 : #4
Fig. 6 – Standard deviation from the mean of partial orbital data for 4 observations sets
Skewness check excluded sets #1:#2:#4 and #2:#3:#4 (to a lesser extend). Uncertainty check
excluded sets #1:#2:4 and #2:#3:#4. The only relatively precise and unskewed orbital elements
calculation had been performed based on observations set #1:#3:#4.
This was to be expected, as observations #1, #3 and #4 were performed against the longest
time span. Also, they were the most regular ones. These two factors minimized statistical errors
resulting from distribution of observations. Consequently, the orbital elements will be calculated
basing on observations #1, #3 and #4. Observation #2 will be used for sanity check of the elements.
-6-
Data consistency check
The orbital elements obtained were then inserted into the ephemeris generator script, to
determine the position of the asteroid at the time of observation #2. The following result was
obtained:
Notice the very high ephemeris uncertainty as compared to the actual discrepancy between
the ephemeris and observation. It is because uncertainty in the ephemeris generator is calculated
regardless of the source of orbital elements data. It is only a coincidence that the ephemeris point was
very close in time to the data points used to calculate orbital elements. This radical closeness in time
minimized the error.
However, if the ephemeris point was a few months apart from the time of measurements
taken at the Etscorn observatory, then the actual discrepancy between the prediction and observation
could have been as much as 1h 43m in RA and 10° 31’ in Dec.
-7-
CONCLUSION
The orbital elements are certainly correct within the specified error bonds according to every
possible internal measure of verification (i.e. consistency of orbital elements received from different
combinations of observations, ephemeris check for the other observation).
The most reliable external method of checking the calculated orbital elements is Jet
Propulsion Laboratory’s Horizons ephemeris computation system.3 Query for 1998 YP11’s orbital
elements on 2008-04-27 07:54 (calculated perihelion passage revealed):
It can be seen that the difference between the JPL Horizons system values (assumed to be
very precise) and the ones calculated in this paper is always smaller than the estimated uncertainty. It
means that measurements, calculations were accurate and error estimation was honest. This means
that the results are scientifically significant and worth of submission to the Minor Planet Center.
Note on uncertainty of M: in my orbital elements the uncertainty was calculated for the
time of perihelion passage, i.e. when M = 0° (without any uncertainty). However, there is no
possibility of querying the JPL Horizons system by the value of M to return a specific date, so this
uncertainty could not be directly tested. However in the primary calculation of orbital elements based
on observations #1, #3 and #4 the uncertainty in M was 2.226832°. Assuming a similar value for
2008-04-27 07:54, this orbital element is also fully confirmed by JPL Horizons system.
Source of errors
Lack of precision of the observational techniques was an evident source of error. It was
revealed in the discrepancies between calculated orbital elements and the ones taken from JPL
Horizons system. However, it was accurately predicted in the Monte Carlo error analysis and
3
http://ssd.jpl.nasa.gov/?horizons
-8-
confirmed in its magnitude by the Horizons system. The solution for a smaller error would be to use
1x1 image binning, do shorter exposures and more advanced pixels brightness weighting to obtain
more accurate positions of the asteroid and reference stars on the CCD images. Also, the fact that RA
lines converge at the celestial poles, what distorts the interpolated system of coordinates, should be
taken into account to achieve higher measurements precision.
Another source of error is the very short time of observations – less than 6 weeks in total. It
has been shown by skewness/s.d.m. partial data analysis how profoundly unevenly spaced/short-
termed observations affect the uncertainty in the result. The only solution to that problem would be
to observe the asteroid in a greater time span.
Orbital elements have been determined using only one set of three observations, as the rest of
possible combinations have been found to distort the results/lower their accuracy in an unacceptable
way. More observations would make it possible to perform combination of multiple results to achieve
higher accuracy.
-9-
APPENDICES
- 10 -
# 2UCAC RA J2000 Dec J2000 μRA μDec X Y
6 40732020 19 18 5.580 +25 12 21.43 0.7 -9.0 153.99 22.98
7 40732125 19 18 47.330 +25 10 31.25 -3.3 -9.8 85.00 70.00
8 40732083 19 18 42.479 +25 10 02.94 -5.9 3.1 113.00 82.96
9 40732099 19 18 44.011 +25 09 58.51 1.1 1.8 104.00 84.99
10 40732167 19 18 51.887 +25 09 06.37 2.4 -1.2 57.98 107.00
11 40732194 19 18 54.721 +25 09 33.12 -0.5 -6.4 41.00 95.00
12 40732197 19 18 55.097 +25 10 38.32 4.6 6.9 38.98 66.99
As 1998 YP11 - - - - 77.00 63.00
Appendix 3 – contents of the CCD image from observation #3 on 2008-06-18 at 05:44:18.359 ±60 s
(proper motion is measured in mas/yr)
Normalized Normalized
Element Initial value Uncertainty Skew
value uncertainty
a [AU] 1.655172 0.167479 -0.001292 1.656465 0.167475
e 0.361543 0.058366 -0.006200 0.367742 0.058691
i [°] 14.065205 2.306594 -0.238857 14.304062 2.318807
Ω [°] 145.649424 1.977235 +0.272037 145.377387 1.995760
ω [°] 73.191064 3.395554 -0.441414 73.632477 3.423949
- 11 -
Normalized Normalized
Element Initial value Uncertainty Skew
value uncertainty
M(t) [°] 26.469491 4.676070 0.679451 25.790040 4.724935
Perihelion 2008-04-22 13d 18h 5m -5d 1h 32m 2008-04-21 16d 3h 8m
passage time 12:30:11.081461 34.585347s 47.068368s 03:07:24.730182 46.839596s
Appendix 6 – orbital elements calculated from observations #1, #2, #3, 5 000 iterations,
94.2% success rate, middle observation on 2008-06-22 at 05:48:58.000234 ± 60 s
Normalized Normalized
Element Initial value Uncertainty Skew
value uncertainty
a [AU] 1.563310 0.076771 -0.071829 1.635139 0.105133
e 0.333427 0.030312 -0.026868 0.360295 0.040505
i [°] 12.980434 1.221062 -1.087472 14.067906 1.635096
Ω [°] 146.443520 1.140067 0.983712 145.459808 1.505785
ω [°] 71.796584 1.886706 -1.626096 73.422680 2.490722
M(t) [°] 28.515636 2.534486 2.178086 26.337550 3.341766
d h m
Perihelion 2008-04-27 7d 6h 35m -1 16 42 2008-04-27 9d 1h 44m
passage time 01:34:22.388478 7.322609s 29.897610s 16:32:57.769458 04.944411s
Appendix 7 – orbital elements calculated from observations #1, #2, #4, 5 000 iterations,
61.4% success rate, middle observation on 2008-06-22 at 05:48:58.000234 ± 60 s
Normalized Normalized
Element Initial value Uncertainty Skew
value uncertainty
a [AU] 1.685026 0.078131 -0.001129 1.686156 0.078135
e 0.377115 0.025047 -0.001488 0.378604 0.025090
i [°] 14.506337 1.049504 -0.060833 14.567170 1.051213
Ω [°] 145.476670 0.983357 +0.082997 145.393673 0.986805
ω [°] 73.533011 1.599789 -0.117595 73.650606 1.604025
M(t) [°] 29.315447 2.220552 0.168586 29.146860 2.226832
d h m h m
Perihelion 2008-04-28 4 14 14 22 58 2008-04-27 5d 19h 28m
s
passage time 11:36:59.627291 00.851720 40.543812 07:54:00.851276 35.232159s
s
Appendix 8 – orbital elements calculated from observations #1, #3, #4, 5 000 iterations,
99.8% success rate, middle observation on 2008-07-01 at 06:35:44.546 ± 60 s
Normalized Normalized
Element Initial value Uncertainty Skew
value uncertainty
a [AU] 1.585063 0.128419 -0.015778 1.600841 0.129377
e 0.340312 0.048136 -0.009486 0.349798 0.049059
i [°] 13.031847 2.000277 -0.382916 13.414763 2.036493
Ω [°] 146.916947 2.092853 +0.492963 146.423985 2.150019
ω [°] 71.282602 3.286904 -0.730204 72.012806 3.366866
M(t) [°] 32.676157 4.549375 +1.010619 31.665538 4.660039
- 12 -
Normalized Normalized
Element Initial value Uncertainty Skew
value uncertainty
d h m d h m
Perihelion 2008-04-26 14 2 09 -2 20 46 2008-04-26 14d 5h 31m
s
passage time 01:37:09.664462 11.877144 36.182565 18:08:46.086318 12.435960s
s
Appendix 9 – orbital elements calculated from observations #2, #3, #4, 5 000 iterations,
89.5% success rate, middle observation on 2008-07-01 at 06:35:44.546 ± 60 s
- 13 -