Question |
PEP |
JPL (see note below) |
Hannover |
Paris |
A. What coordinate origins are available in your code? |
Geocentric, heliocentric, and solar-system barycentric. Chosen
according to accuracy requirements. |
Solar system barycentric coordinates are used in the integrator and
for basic computation in fits. Integrator can also do Sun centered
coordinates. Station positions are geocentric and reflector positions
are Moon centered. |
Geocenter, solar system barycenter, selenocenter |
Solar System barycentric coordinates |
B. What coordinate system do you use for LLR analysis? |
General relativistic isotropic. |
Solar system barycentric coordinates or their Earth-Moon differences
are used for fits after space and time transformations. |
Mean ecliptical system J2000, origin in geocenter. |
Solar system barycentric coordinates or their Earth-Moon
differences |
C. What formulation do you use to transform times (e.g., the LLR
time measurements in UTC)? |
Proper time (UTC) is converted to coordinate time (CT, a constant
offset from TDB) via either an analytic approximation or an integrated
time ephemeris, according to the accuracy requirements. |
Station UTC → TDB. Moyer dot product formulation used for time
transformation. |
UTC > TAI > TT > TDB (Hirayama), geocentric proper time
TCG |
Transformations from UTC to UT1 or TT according to IERS Conventions
2003. Between TT and TDB, use of the transformation integrated together
with equations of motion. |
D. How do you handle mass contributions from kinetic energy of
bodies? |
Implicit in the general relativistic n-body equations of
motion. |
EIH N-body equations of motion. |
EIH N-body equations of motion. |
EIH N-body equations of motion. |
E. What relativistic corrections do you apply in tandem with
coordinate transformations (e.g., Lorentz contraction of Earth)? |
Lorentz contraction of Earth and Moon. |
Vector Lorentz transformation, constant (L_C for Earth) and
potential on scale (including annual variation) for both Earth and
Moon. |
Lorentz and Einstein contraction for Earth and Moon. |
None in the dynamical part, Lorentz and Einstein contraction for
Earth and Moon in the LLR reduction of data. |
F. Do you account for relativistic contributions from
Earth rotation affecting position of stations in solar system? Can
be thought of as skew in Lorentz transformation. |
Other than the obvious direct effect of diurnal rotation, no. |
Yes, for the big effect. Was something more subtle intended? |
No. |
No. |
G. Is eccentricity of earth's orbit taken into account in the
rescaling of proper distances of earth positions due to (1 + γ
Usun/c²) factor? |
No. |
Yes, the Sun's distance is used in the potential so annual and
smaller effects are present. |
Yes. |
Yes. |
H. What is used for orientation of Earth for gravity field
perturbation? Does orbit integration include figure-figure effects or
other small interesting effects? Is there an anomalous eccentricity rate
in the integration or solution? |
Earth orientation given by precession and IAU 2000 nutation series.
Figure-figure interaction not included in orbit. No anomalous eccentricity
rate. |
Orientation has constant angles, precession and obliquity
expressions, and 18.6 yr nutation. No figure-figure effect on orbit.
Anomalous eccentricity rate is in solution, but not in integration. |
Precession, Nutation, obliquity, GMST and the difference to the
Earths principal axis system. Figure-figure effects only in the rotation
of the Moon. No eccentricity rate taken into acount. |
Orientation is integrated together with other equation of motion.
Figure-Figure effect only in the rotational motion of the Moon. No
eccentricity rate taken into account. |
Note: For JPL integrator documentation, see Standish and Williams,
Chapter 8 for the Explanatory Supplement. The DE421 memos contain
information on the solution used. Our 2001 paper discusses dissipation
in the Moon.
Question |
PEP |
JPL |
Hannover |
Paris |
A. How do you handle post-Newtonian dynamics in your equations
of motion? |
N-body equations of motion as shown in Will (1981) for isotropic
coordinates. |
EIH equations of motion for Moon and planets. |
N-body Einstein-Infeld-Hoffmann equations of motion. |
EIH equations of motion for Moon, planets and 5 asteroids |
B. How many asteroids do you treat as individuals, and how many
diffuse distributions do you include to describe the asteroid
population? |
98 asteroids plus one ring. 8 of the 98 asteroid masses are
estimated separately while the other 90 are modeled by five density
classes with individual radii assigned from the best available
information. |
About 300 asteroids, no ring. See Folkner et al DE421 memo. |
7 asteroids as point masses. |
About 300 asteroids, one ring. Its orientation is integrated
together with equations of motion. |
C. Do you include any natural satellites of planets? |
Only the Moon. |
No planetary satellites other than Moon. |
Only the Moon. |
Only the Moon |
D. Do you handle radiation pressure on Earth/Moon? How? |
Yes, assuming uniform albedo on each body, neglecting thermal
inertia. |
No radiation pressure in integrator. |
Not in integrator. In the modelled range by adding 3.7mm*cos(synodic
angle), as shown in Vokrouhlicky(1997) |
No. |
Question |
PEP |
JPL |
Hannover |
Paris |
A. In the context of planet/system rotation, do you switch to a
proper time description at the body that varies according to
position? |
Yes for Mars only (the rotation being modeled as an essentially
uniform angular velocity in proper time). The rotation of the Earth is
determined empirically, rather than parametrically, and is thus tied to
the time "flow" of the observations. Similarly, the integrated rotation
of the Moon is tied to the coordinate time system of the
integration. |
No proper time for Moon rotation. Earth stations use standard
UTC→TDB. Moon retroreflectors and rotation use TDB. Other bodies'
data analysis and rotations are in planetary software, not LLR
software. |
Earth rotation empirically, neglected for the Moon. |
No. |
B. From where do you get Earth orientation information? |
IERS |
UT1/PM from a JPL file. Precession, obliquity and nutations come
from theories, but we solve for constant orientation angles, rates and 4
nutation periods (18.6 yr, 9.3 yr, 1 yr, 1/2 year) × 4
coefficients each. |
IERS C04 series |
Integrated together with equations of motion in the dynamical part.
Fixed to IERS 2003 + C04 series in the reduction to
observations. |
C. What is your Earth gravity model (degree and order, which if
any terms are fit parameters in your model)? |
Through fourth degree, zeroth order. No coefficients solved for.
Values from SAO (1973) standard Earth model. |
Zonals degrees 2-4 for Earth. We are adding J2 and its secular rate
as possible fit parameters, but that is not operational yet. |
Complete second degree, 3rd and 4th zonals. No coefficients solved for.
Values from EGM 2008 |
Through fourth degree, zeroth order. J2 and J3 are solved for, J4 is
fixed to EGM96 value. |
D. What is your lunar gravity model (degree and order; which
terms are fit vs. used from external sources)? What bodies apply
torque to the Moon through the gravity field? |
Through third degree, third order. Solving for J2, J3, C22, C31,
C32, C33, S31, S32, S33. Others taken to be zero. |
For Moon, J2, and degree,order 3,0-3 and 4,0-4. J2 and libration
beta and gamma are used to compute C22. J2, beta, gamma, and all
third-degree coefficients can be fit. LP150Q is source for some gravity
coefficients, others are fit; fixed vs fit can vary. See DE421 memo for
that set. Direct torques are from Earth, Sun, Venus, Mars, and
Jupiter. |
Complete up to fourth order from LP165P. Solving for C20, C22, C30,
C31, C32, C33, S32. Torques from Earth and Sun. |
Through fourth degree, fourth order. Solving for J2, C22, J3, C31,
C33, S31, S32. Other fixed to LP150Q values. |
E. How do you handle Earth tidal dissipation?
| Earth orientation is handled purely empirically. The effect of
Earth tidal drag on the lunar orbit is modeled by a solved-for lag angle
on the tidal response. |
The Standish and Williams paper gives our integrator Earth tide
expression from Moon and Sun with 3 degree-2 Love numbers and 3 time
delays (semidiurnal, diurnal, and zonal). We lately changed the diurnal
and semidiurnal time delays to allow for linear frequency dependence
adding 2 more parameters. The 5 time-delay parameters can be fit, but
only 2 are typically fit. |
Solved for an lag angle. |
Variations of its inertia matrix and degree-2 potential
coefficients are computed from delayed positions of tide generating
bodies. 3 Love numbers and 3 time delays are taken into
account. |
F. What dissipation model do you have for describing lunar
orientation? many fit parameters? |
Three different models. One uses constant time-delay response (two
parameters); one has constant-Q response (two parameters, exclusive of
previous model); one has a fluid core loosely coupled to the mantle
(five parameters). |
In the integrator, we have a degree-2 Love number and time delay for
tides on Moon, plus a fluid-core/solid-mantle boundary (CMB) dissipation
term. During fits we can solve for or constrain up to 5 out-of-phase
periodic libration terms to account for tidal dissipation vs frequency;
currently we fit 3, constrain 1, and do not solve for 1. See our 2001
paper. |
Two models. constant time delay (two parameters) and a fluid core
model (two parameters), not solved for core parameters at the
moment |
Similar to Earth's one, but with only one Love number (k2) and one
time delay (τ) for second degree harmonics, both fitted. |
G. Do you handle geocenter motion, as seen by SLR? If so, what
model/input? |
No. |
No geocenter motion, but we solve for 3-D vector rates for stations,
which should absorb any secular geocenter rates. |
No. |
No. |
H. What is tidal model for physical librations? Which bodies
cause tides? What fluid core effects are present? Does integrator have
figure-figure terms? Are there relativistic rotation or other
effects? |
See 3F for tidal model and fluid core. Only Earth is considered for
lunar tides. Figure-figure interaction is included in libration
integration. No relativistic rotation effects. |
In the libration integration, solid-body tides on the Moon are from
Earth, but not from Sun, which is <1% of Earth. We also have spin
distortion, though that has very small variation. Libration model has
fluid core moment, CMB dissipation and CMB oblateness. We have
Earth-Moon degree-2 figure-figure effects, but no relativistic rotation
effects. |
see 3F, time delayed variation of the tensor of inertia. Tides from
Earth and Sun. Fluid core moment, CMB dissipation and Earth-moon degree2
figure-figure effects. |
Tidal model: see 3F. Tides are generated by Earth and Sun. No fluid
core. Geodesical torque is taken into account. |
Question |
PEP |
JPL |
Hannover |
Paris |
A. What is your Earth tidal model? What degree/order? How many
Love numbers? Which ones are fit vs. used from external sources? |
Degree-independent response to perturbing potential characterized by
two Love numbers and a time lag. The three parameters are fit. |
Degree-2 tidal displacements on Earth are from Moon and Sun with 2
displacement Love numbers. There is also a K1 amplitude correction and a
pole tide contribution. Displacement Love numbers are not fit. |
IERS 2003 tidal corrections, 4 love numbers (2nd and 3rd degree)
which are not fitted. IERS 1992 tidal corrections due to polar
motion | IERS Conventions 2003 (V. Dehant's subroutine) |
B. Do you handle continental drift, post-glacial rebound,
and other local motions? How many parameters per station? Are
these fit or used from external sources? |
Yes, Three solved-for velocity components per station. |
We can solve for 3 coordinate rates per station. We currently fit
rates for McDonald, OCA and APO, while Haleakala and Matera motions are
fixed to plate motion model. |
Yes, three solved-for velocity components per station. APOLLO
modeled by NUVEL1A model |
Tectonic plate motion are taken into account. Values are fixed to
ITRF's ones, the possibility to fit them is not used. |
C. Do you model the fiducial point of the telescope? |
If necessary. |
The McD 2.7 m data has a constant bias applied to bring ranges to
the intersection of axes. I believe the MLRS applies a range correction
to refer to a marker. Other stations we assume intersection of axes when
not told otherwise. We d o not apply any eccentricities. |
If necessary. |
No. |
D. Do you handle thermal expansion of the telescope? |
If necessary. |
No telescope thermal expansion is modeled, but we would be
interested if the stations have estimates or suggestions. Thermal
expansion of telescope length should already be accounted for if there
is internal calibration, but mount motion is open for improvement. |
No. |
No. |
E. Do you handle atmospheric loading? |
No. |
No. |
Yes. |
Yes. |
F. Do you handle ocean loading? |
No. |
No. |
Yes. |
Yes. |
G. Do you handle hydrologic loading? |
No. |
No. |
No. |
No. |
Question |
PEP |
JPL |
Hannover |
Paris |
A. Do you obtain partial derivatives via integration of the
variational equations, or use numerical difference methods. |
Integration. |
We have three programs for dynamical partials: 1 integrates
variational equations and 2 use finite differences, where one
differences separate integrations and the other integrates two cases
simultaneously. |
Both methods. |
Numerical differences only. |
B. If both are used, when is each method used, generally? |
Numerical differences are used only for checking the accuracy of the
integrated partials. |
Long-time existing dynamical partials tend to be from variational
equations, while newer partials are going into a finite difference
program. We have had noise build up for some partials during the
variational equations integration and
have switched some partials. |
Numerical derivatives are used for relativistic parameters. |
N/A |
C. For dynamical and geometrical partials, are the up and down
legs calculated separately or is one leg calculated and doubled? |
Calculated Separately |
I think we do partials for both up and down legs in most cases, but
we would have to check each partial in the code. Some partials, e.g.
biases, do not depend on the up or down leg. |
One leg calculated and doubled. |
Calculated separately |
Question |
PEP |
JPL |
Hannover |
Paris |
A. What is your integration method or methods? |
15th-order Adams-Moulton. The starting procedure uses the Nordsieck
method. |
We use the Krough variable step size variable order integrator
running quadruple precision. |
Adams-Bashfort multistep with variable step size |
12th-order Adams PECE. The starting procedure uses ODEX. |
B. Do you simultaneously integrate the solar system and
Moon? |
No. |
Do simultaneous integration of Moon, planets, Pluto and physical
librations. |
Yes. |
Yes. |
C. If not, how are these products merged? |
Three iterations of the cycle, first taking the Moon as given while
integrating the rest of the solar system and then taking the solar
system as given while integrating the Moon. The initial Moon is the
Brown mean lunar theory. |
N/A |
N/A |
N/A |
D. If not, what are the epochs for the two integrations
corresponding to initial conditions? |
The same: May 1968. |
N/A |
N/A |
N/A |
E. What is your integration step size (or sizes)? |
1/8 day for the Moon, 1/2 day for the solar-system n-body
integration, various sizes for other single-body integrations. |
Variable step size. |
Variable, output every 0.3 days. |
Fixed, about 0.055... |
F. Do you simultaneously integrate the Moon orbit and
rotation? |
Yes. |
Do simultaneous integration of Moon, planets, Pluto and physical
librations. |
Yes. |
Yes. |
G. If not, how are these products merged? |
N/A |
N/A |
N/A |
N/A |
H. Do you determine the location of the Sun by integrating the
equations of motion,or impose center of mass to dictate the Sun's
position? |
Impose fixed center of mass. |
We use a relativistic center of mass expression to get the Sun. |
By integrating the equations of motion. |
Integrating its equations of motion. |
Question |
PEP |
JPL |
Hannover |
Paris |
A. What relativistic corrections do you apply to light
propagation? |
Shapiro delay due to the Sun and, in the case of cislunar or lunar
targets, due to the Earth, but not due to the Moon. |
Shapiro time delay is calculated for gravity of Sun and Earth. I
thought we
did Moon too, but do not see it in our code. |
Shapiro delay due to Sun and Earth. |
Shapiro delay due to the Sun and to the Earth, but not due to the
Moon. |
B. Do you apply the Shapiro light-time correction iteratively or
in a single step? Are up and down legs computed separately, or is
one leg calculated and doubled? |
Single step. |
Shapiro delay and atmospheric delay are iterated for up leg, but not
for do wn leg (which is known much better at first iteration). |
Single Step. |
Iteratively. Up and downleg computed separately. |
C. What is your atmospheric propagation model, and what input
parameters do you require from the observation? |
Mendes and Pavlis (2004). Pressure, temperature, relative humidity
-- obtained from site measurements when available and assumed to be
nominal otherwise. |
Marini and Murray model is used for atmospheric delay. We use
temperature, pressure, humidity and laser wavelength. Elevation angle
is calculated internal ly. |
Mendes and Pavlis (2004). Pressure, temperature, relative
humidity. |
Marini and Muray (1973), pressure, temperature, relative humidity,
laser wavelength |
D. How do you handle range bias? Do you have multiple bias
parameters for each station (how many)? |
Solve-for constant bias parameters. As many parameters per station
as needed (as determined by history of changes in measurement
system). |
We have overall (full time span) range biases for stations MCD (2.7
m), OCA, HAL and APO plus shorter time biases where we suspect them from
residuals or station information. Haleakala had several sets of biases
depending on changes in different rings of lenses. We know the dates for
those changes. Currently, our solutions do not solve for an overall bias
for the MLRS using it as the zero standard. Our present solutions use 40
bias parameters, 39 biases and one slope. With zero overall MLRS bias,
the other four overall biases are within ± 0.2 nsec. |
Solve-for constant (full time span) and time dependent bias parameters.
All together 43 bias parameters and one slope. |
Possibility to solve for 40 biases (constant and drift for each).
But only 2 constants are used up to now. |
E. In order to extract LLR science, what additional
observational inputs does your model process (this is separate from the
use of importing analysis results, as in other questions)? |
Delay or delay-rate observations of other solar-system bodies, such
as radar to Mercury and Venus, orbiter and lander tracking data on Mars,
and flyby normal points for Jupiter and Saturn. |
Station specific inputs, in addition to range, range uncertainty and
time, are pressure, temperature, humidity and laser wavelength. We can
scale the normal point uncertainty with a factor and also add RSS noise.
This adjusted uncertainty is used for weighting during the least-squares
fit. The UT/PM table has uncertainties that limit adjustment of a
stochastic UT/PM correction subject to a correlation time. |
Meteorology. |
Some parameters involved in the dynamical part are first fitted to
planetary
observations. Then a fit to LLR data is done, checking that it does not
deteriorate planetary residuals. |
F. What weighting scheme do you typically employ for your data
sets? Do you scale errors, convolve them with a constant, etc.? |
Weighted least-squares using reported errors, except that errors may be
scaled to balance residuals. |
We use weighted least-squares fits. We can fit LLR data alone or LLR
+ planetary, combining files from the planetary programs. Usually it is
the former since the Earth's orbit is well known now, but for a public
ephemeris like DE421 it is the latter. Linear constraints between
parameters can be imposed. There are no a priori uncertainties. Matrix
inversion uses a square root computation. We get solution parameter
corrections and uncertainties. We also get the correlation matrix, and
normal point residuals from linear corrections. After the solution we
can use the residuals to get the weighted rms residual overall or by
year with a choice of station and retroreflector. Amplitude spectra of
residuals can be co mputed with a periodogram. |
We use an iterative, weighted least squares adjustment of the LLR data.
The weighting is calculated with respect to the Normal Point uncertainty
for every single observation. We can scale this uncertainty, e.g.,
depending on station or observation time. We introduce no a priori
uncertainties. We get normal point residuals, correlation matrix,
corrections to the solution parameters and their uncertainties. |
For LLR, weighting of a set of data is computed according the
standard deviation of the residuals. Things are different for planetary
observation (more information on request). |
Question |
PEP |
JPL |
Hannover |
Paris |
A. Do you estimate the parameters of the lunar model
simultaneously with the parameters of the solar-system model? (This
requires incorporating the solar-system data in the LLR analysis, of
course.) |
Yes. |
|
N/A |
No. First fit is done to planetary observations, then to LLR's
ones. |
B. What options do you have for working with and solving the
normal equations (e.g: a priori parameter values, with an uncertainty
for single parameters or a covariance for a related group; rank
deficient solution; elimination of "uninteresting parameters" without
distorting the estimate or covariance)? |
Parameters can be constrained with individual uncertainties or with
full covariance of related parameters. Rank-deficient matrices can be
inverted by eigenvalue elimination. Uninteresting parameters can be
projected away in a preliminary stage, leaving a reduced set of normal
equations that yields correct estimates and covariances for the interesting
parameters. |
|
See 8F |
LLR fit: no options, just solving normal equations. For planetary
fit, more options are available (more information on request). |
C. What tools do you have for examining post-fit residuals
(e.g., Fourier transform or other spectral analysis)? |
FFT, weighted Fourier transform, histograms. |
|
Nothing special, MatLab standard tools. |
None is used up to now. |
D. What parameters can your software estimate? This is likely
to be a very long list, with batches of parameters requiring explanatory
notes. |
Briefly: masses, asteroid class densities, initial conditions,
station positions and velocities, coordinates of fixed targets on Moon
or planets, Earth precession and nutation coefficients, piecewise linear
adjustments to polar motion and UT1, gravitational harmonic coefficients
for all bodies (limited to J2 for the Sun), Moon moment-of-inertia
ratios, Moon core and tidal dissipation parameters mentioned above,
observation biases, radii and topography grids (and spins) for other
planets, interplanetary plasma density, metric parameters beta and
gamma, EP violation, de Sitter precession rate, G-dot, ad hoc
coefficients of post-Newtonian terms in the integrated motions and in
light propagation delay and in transformation between proper time and
coordinate time. |
Info provided; will display separately. |
All together: about 220 can be solved, but not all are estimated in
every analysis. I will prepare a list. |
For LLR: 188 can be solved! But only 59 of them are fitted to LLR:
initial conditions, stations and reflector's positions, potential
coefficients, tides parameters, biaises, ... Other parameters can be
fitted to planetary observations: asteroid masses, Sun's J2, initial
conditions ... more information on request. |