Terrestrial

Reference System and Frame

Datum transformations
E. Calais
Purdue University - EAS Department
Civil 3273 ecalais@purdue.edu

Need for aTerrestrial Reference

GPS geodesy estimate position and
displacement (velocity) of points on Earth
Positions and velocities:
Are not direct observations, but estimated
quantities
Are not absolute quantities

Need for a Terrestrial Reference in which

(or relative to which) positions and velocities
can be expressed.

Need for a Terrestrial Reference

In practice, GPS data processing can be
seen as a 2-step process:
1. From phase and pseudorange observables
to (relative) site positions (and their
associated variance-covariance = the whole
thing is sometimes called quasiobservations)
2. From quasi-observations to positions and
velocities in a well-defined reference frame

Need for a Terrestial Reference:

Datum Defect

Geodetic data are not

sufficient by themselves
to calculate
coordinates!
Ex. of triangulation data
(angle measurements):
origin, orientation, and
scale need to be fixed
Ex. of distance
measurements: origin
and orientation need to
be fixed, scale is given
by the data

Datum Defect
3 points (in 3D):

2 independent distance measurements

2 independent angle measurements
2 independent height difference measurements
But 9 unknowns

4 points: 12 unknowns, 9 data

Datum defect = rank deficiency of the matrix that relates
the observations to the unknowns
Solution: define a frame!
Fix or constrain a number of coordinates
3D network: rank deficiency is 7 (or 6) => fix 3 (or 2) points

Fixing a Frame, but

Too many constraints?
Ex.: fixing 2 points in a geodetic solutions means that one
fixes:
Scale
Orientation
Origin twice

A priori constraints used on site positions have an

impact on the final uncertainties: overconstraining
typically results in artificially small uncertainties
The Frame has to be well defined (stable) through
time: time-dependent model

System vs. Frame

Terrestrial Reference System (TRS):
Mathematical definition of the reference in which
positions and velocities will be expressed.
Therefore invariable but inaccessible to users in
practice.

Terrestrial Reference Frame (TRF):

Physical materialization of the reference system
by way of geodetic sites.
Therefore accessible but perfectible.

The ideal TRS

Tri-dimensional orthogonal (X,Y,Z).
Base vectors have same length.
Geocentric: origin close to the Earths center
of mass (including oceans and atmosphere)
Equatorial: Z-axis is direction of the Earths
rotation axis
Rotating with the Earth.

3D similarity
Under these conditions, the transformation of
cartesian coordinates of any point between 2 TRSs
(1) and (2) is given by a 3D similarity:

X (2) = T1,2 + "1,2 R1,2 X (1)

Where:
X(1) and X(2) are the position vectors in TRS(1) and TRS(2)
T1,2 = translation vector

!1,2 = scale factor

R1,2 = rotation matrix

3D similarity
X

(2)

= T1,2 + "1,2 R1,2 X

(1)

Also called a Helmert, or 7-parameter,

transformation.
If translation (3 parameters), scale (1
and rotation (3 parameters) are
! parameter)
known, then one can convert between TRSs
If there are common points between 2 TRSs,
one can solve for T, , R: minimum of 3
points.

Linearization
3D similarity between TRS1,X1 and
TRS2, X2 can also be written as:

X 2 = X1 + T + DX1 + RX1
With D = scale factor and:

" T1 %
$ '
T = $T2 '
$ '
#T3 &

# 0 "R3 R2 &
%
(
R = % R3
0 "R1 (
%
(
0 '
$"R2 R1

Time dependence
X1, X2, T, D, R are generally functions of time
(plate motions, Earths deformation).
Differentiation w.r.t. time gives:
X 2 = X 1 + T + D X1 + DX 1 + R X1 + RX 1

D and R ~ 10-5 and Xdot ~ 10 cm/yr

! DXdot and RXdot negligible, ~ 0.1 mm/100
years, therefore:
X 2 = X 1 + T + D X1 + R X1

Estimation

The above equations can be written as:

X 2 = X1 + T + DX1 + RX1 " X 2 = X1 + A#

X 2 = X 1 + T + D X1 + R X1 " X 2 = X 1 + A#

with:

! " = [T ,T ,T ,D,R ,R ,R ]
1 2 3
1
2
3

!
!

The least-squares solutions are:

!

" = [T1, T2 , T3 , D , R1, R 2 , R 3 ]

#. . . . .
.
.&
%
(
z "y (
%1 0 0 x 0
A = %0 1 0 y "z 0
x(
%
(
%%0 0 1 z y "x 0 ((
.
.'
$. . . . .

" = (AT !
Px A)#1 AT Px (X 2 # X1 )
" = (AT Pv A)#1 AT Pv ( X 2 # X 1 )

Where Px and Pv are the weight matrix for station positions and
velocities, respectively
!

Summary
If we know the position and velocities of a number of geodetic sites
that define a reference frame, we can:
1. Use observations at these sites in addition to observations at sites of
unknown position and velocity = reference, or fiducial, sites
2. Equate the position and velocity at these reference sites to these
known values
3. Solve for 7 or 14 transformation parameters
4. Apply the transformation to the position and velocity of the other sites.
(3 and 4 can actually be done in one step)

Choice of reference sites:

Global distribution
Position and velocity precise and accurate
Error on their position/velocity and correlations well known

So, now we need to define a reference frame

The international Terrestrial

Reference System: ITRS
Definition adopted by the IUGG and IAG
Tri-dimensional orthogonal (X,Y,Z), equatorial (Zaxis coincides with Earths rotation axis)
Non-rotating (actually, rotates with the Earth)
Geocentric: origin = Earths center of mass, including
oceans and atmosphere.
Units = meter and second S.I.
Orientation given by BIH at 1984.0.
Time evolution of the orientation ensured by
imposing a no-net-rotation condition for horizontal
motions.

The no-net-rotation (NNR) condition

Objective:
Representing velocities without referring to a particular plate.
Solve a datum defect problem: ex. of 2 plates 1 relative velocity to solve for 2
absolute velocities (what about 3 plates?)

Mean: describe velocities in a frame that minimizes plate velocities (actually

their angular momentum), over the whole Earths lithosphere:

# X " Vdm = 0
L

Can be discretized as:

$Q "
p

=0

(= condition equation)

p #P

Where Qp = inertia tensor of plate p, p = angular velocity of plate p, P = total

number of plates
Qp : depends !
on plate geometry, but thickness variations are ignored.

The no-net-rotation (NNR) condition

The no-net-rotation
condition states that the
total angular momentum of
all tectonic plates should
be zero.
See figure for the simple
(and theoretical) case of 2
plates on a circle

VA/A=0
VA/NNR

P la

te 1

P la

te 2

VB/A
VB/NNR

Earth s radius R

MA = R X VB/NNR
MB = R X VA/NNR
M = 0 VB/NNR = VA/NNR

The no-net-rotation condition

The NNR condition has no impact on
relative plate velocities.
It is an additional condition used to
define a reference for plate motions
that is not attached to any particular
plate

The international Terrestrial

Reference Frame: ITRF
ITRF is defined by the positions (at a given epoch)
and velocities of a set of geodetic sites (+ associated
covariance information) = dynamic datum
These positions and velocities are estimated by
combining independent geodetic solutions and
techniques.
Combination:
Randomizes systematic errors associated with each
individual solutions
Provides a way of detecting blunders in individual solutions
Accuracy is equally important as precision

History of the ITRF

1984: VLBI, SLR, LLR, Transit
1988: TRF activity becomes part of the IERS => first
ITRF = ITRF88
Since then: ITRF89, ITRF90, ITRF92, ITRF93,
ITRF94, ITRF96, ITRF97
Current: ITRF2000
ITRF improves as:
Number of sites with long time series increases
New techniques appear
Estimation procedures are improved

ITRF2000

Solutions used: 3 VLBI, 1 LLR, 7 SLR, 6 GPS, 2 DORIS

Up to 20 years of data
GPS sites defining the ITRF are all IGS sites
Wrms on velocities in the combination: 1 mm/yr VLBI, 1-3 mm/yr SLR and GPS

ITRF and the user

ITRF distributed as:
A table (incomplete covariance information)

A Solution Independent Exchange format file =

SINEX file (complete covariance information)

ITRF and the User

Coordinates precise to better than 1 cm and velocities to better

than 1 mm/yr for sites with long (10 years and more) observation
spans (figure from Altamimi et al., 2002).

ITRF and the user

Velocities in ITRF2000 at some IGS sites

ITRF and the user

Why use it?
Because it provides the most accurate global
reference frame available
Because precise GPS orbits (from IGS) use it
If everyone uses it, then, by default, all positions
and velocities will be expressed in the same (and
globally consistent) geodetic frame easy to
compare

ITRF and the user

Decide on a number of GPS sites to include in your
data processing
Estimate daily station positions by inverting phase
and pseudorange data = daily solutions
Constrain the solutions to ITRF by:
1. First possibility: Fixing, or tightly constraining, the position
(and velocity) of reference sites to their ITRF values:
2. Second possibility:
1. Estimating a 14 parameter transformation between a loosely
constrained solution and the ITRF, using common sites
2. Applying the transformation to the loosely constrained solution

ITRF and the User

Since the ITRF is global, the best procedure is
actually to:
1. Include a minimum of reference station in daily GPS
solutions = regional solutions
2. Combine (loosely constrained) daily regional solutions with
global solutions
3. Estimate 14 parameter transformation using a global set of
reference sites, well distributed
4. Apply the transformation.

Also, constraining station positions and/or velocities

will bias the solution if their coordinates and/or the
associated variance are incorrect.

ITRF and the User

Does ITRF solve the datum defect
problem?
Yes, but it is a space-based technique,
therefore:
If orbits are fixed => reference frame
already tightly imposed
If EOP are fixed => network orientation
imposed

ITRF and the User

How to ensure that the implementation of the
reference frame minimally affects the final solution
(estimates and their variance)?
Minimal constraints: for a solution to be in a given
frame, the transformation parameters between that
solution and the frame must be null.
Application of minimal constraints:
Remove constraints from solution if they were implemented
=> loose solution
Decide on a number of sits that will be used to estimate the
transformation parameters (= to implement the reference
frame)
Add condition equation that reflect the null transformation

Minimal Constraints

Link with Inertial System

Conventional Inertial System:

Orthogonal system, center = Earth center of mass, defined at standard epoch
J2000 (January 1st, 2000, 12:00 UT)
Z = position of the Earths angular momentum axis at standard epoch J2000
X = points to the vernal equinox
It is materialized by precise equatorial coordinates of extragalactic radio
sources observed in Very Long Baseline Interferometry (VLBI) = Inertial
Reference Frame.
First realization of the International Celestial Reference Frame (ICRF) in 1995.

Systems are defined and frames are realized in the framework of an

international service: the International Earth Rotation Service, IERS
(http://www.iers.org/)
Mission of the IERS: To provide to the worldwide scientific and technical
community reference values for Earth orientation parameters and reference
realizations of internationally accepted celestial and terrestrial reference
systems

Link with Inertial System

The Earth's orientation is defined as the rotation
between:
A rotating geocentric set of axes linked to the Earth (the
terrestrial system materialized by the coordinates of observing
stations) = Terrestrial System
A non-rotating geocentric set of axes linked to inertial
space (the celestial system materialized by coordinates of
stars, quasars, objects of the solar system) = Inertial System

The transformation from Inertial to Terrestrial can be

made using a standard matrix rotation:

xi = P
inertial

precession

N
nutation

spin

Polar motion

xt
terrestrial

Link with Inertial Frame

Rotation parameters are given in astronomical tables
provided by the IERS: for the most precise
positioning, they are used as a priori values and
sometimes adjusted in the data inversion.
In practice, the IERS provides five Earth
Orientation Parameters (EOP) :
Celestial pole offsets (dPsi, dEps): corrections to a given
precession and nutation model
Universal time (UT1) = UT corrected for polar motion,
provided as UT1-TAI
2 coordinates of the pole (x,-y) of the CEP with respect to
Earths geographic pole axis = polar motion

Questions?
How to transform ITRF velocities to
plate-fixed velocities?
Are velocities at reference sites really
constant?
Most legal reference frames are static