• Introduction to GAMIT/GLOBK

1
Introduction to GAMIT/GLOBK
Release 10.3
T. A. Herring, R. W. King, S. C. McClusky
Departement of Earth, Atmospheric, and Planetary Sciences
Massachussetts Institute of Technology
Contents
1. GPS Measurements and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1 Phase and pseudorange observations . . . . . . . . . . . . . . . . . . . 3
1.2 Modeling the motions of satellites and stations . . . . . . . . . . . . 6
1.3 Parameter estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2. Automatic Processing with GAMIT and GLOBK . . . . . . . . . . . . . 10
2.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Summary of program flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Editing the control files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Using sh_gamit and sh_glred . . . . . . . . . . . . . . . . . . . . . . . . 25
3. Evaluating Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4. Estimating Velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
16 July 2008
 2
Preface
GAMIT/GLOBK is a comprehensive GPS analysis package developed at MIT, the
Harvard-Smithsonian Center for Astrophysics (CfA), and the Scripps Institution of
Oceanography (SIO) for estimating station coordinates and velocities, stochastic or
functional representations of post-seismic deformation, atmospheric delays, satellite
orbits, and Earth orientation parameters. Although the software is currently maintained
by the three authors of this document at MIT, many people have made substantial
contributions. The orbital integration and modules used in computing the theoretical
phase observable have their origins in the Planetary Ephemeris Program (PEP) written by
Michael Ash, Irwin Shapiro, and Bill Smith at Lincoln Laboratory in the 1960’s., with
later contributions by Bob Reasenberg and John Chandler at MIT. The codes for
processing GPS observations were developed at MIT in the 1980’s by Chuck
Counselman, Sergei Gourevitch, Yehuda Bock, Rick Abbot, and King. GAMIT attained
its current form through the efforts of Bock, Danan Dong, Kurt Feigl, Peng Fang,
Herring, King, McClusky, Peter Morgan, Mark Murray, Berkhard Schraffin, Seiichi
Shimada, and Paul Tregoning. GLOBK was developed by Herring at CfA for
combination of VLBI data and modified at MIT to incorporate GPS data. Details of
these contributions may be found in the references listed at the end of this manual.
Funding for the early development of GAMIT was provided by the Air Force Geophysics
Laboratory, and for GLOBK by NASA. Current support for development and support of
the scientific community comes primarily from the National Science Foundation.
To control processing the software uses C-shell scripts (stored in /com and mostly named
to begin with sh_ ) which invoke the Fortran or C programs compiled in the the /libraries,
/gamit, and /kf directories. The software is designed to run under any UNIX operating
system supporting X-Windows; we have implemented thus far versions for Solaris, HP-
UX, IBM/RISC, DEC, LINUX, and Mac OS-X. The maximum number of stations and
atmospheric parameters allowed is determined by dimensions set at compile time and can
be tailored to fit the requirements and capabilities of the analyst's computational
environment. Installation instructions are given in the README file distributed with the
software.
The first chapter of this manual provides some theoretical background for readers not
familiar with high-precision GPS analysis. Chapter 2 describes setup of tables and
commands for automatic processing to obtain time series of daily position estimates, and
Chapter 3 provides a guide to evaluating your results. In Chapter 4 we discuss various
approaches to estimating station velocities from observations spanning several years.
More detailed documentation is available in the longer (and older) GAMIT Reference
Manual and GLOBK Reference Manual. There are also tutorials available on-line at
http://www-gpsg.mit.edu. The most up-to-date information about the commands is
available through help files, invoked by typing the name of the shell-script or program
without arguments.
16 July 2008
 3
1. GPS Measurements and Analysis
1.1 Phase and pseudorange observations
High-precision geodetic measurements with GPS are performed using the carrier beat
phase, the output from a single phase-tracking channel of a GPS receiver. This
observable is the difference between the phase of the carrier wave implicit in the signal
received from the satellite, and the phase of a local oscillator within the receiver. The
phase can be measured with sufficient precision that the instrumental resolution is a
millimeter or less in equivalent path length. For the highest relative-positioning
accuracies, observations must be obtained simultaneously at each epoch from several
stations (at least two), for several satellites (at least two), and at both the L1 (1575.42
MHz) and L2 (1227.6 MHz) GPS frequencies. The dominant source of error in a phase
measurement or series of measurements between a single satellite and ground station is
the unpredictable behavior of the time and frequency standards ("clocks") serving as
reference for the transmitter and receiver. Even though the GPS satellites carry atomic
frequency standards, the instability of these standards would still limit positioning to the
several meter level were it not for the possibility of eliminating their effect through signal
differencing.
A second type of GPS measurement is the pseudo-range, obtained using the 300-m-
wavelength CA ("coarse acquisition") code or 30-m-wavelength P ("protected") code
transmitted by the satellites. Pseudo-ranges provide the primary GPS observation for
navigation but are not precise enough to be used alone in geodetic surveys. They are
useful, however, for synchronizing receiver clocks, resolving ambiguities, and repairing
cycle slips in phase observations.
For a single satellite, differencing the phases (or pseudo-ranges) of signals received
simultaneously at each of two ground stations eliminates the effect of bias or instabilities
in the satellite clock. This measurement is commonly called the between-stations-
difference, or single-difference observable. If the stations are closely spaced,
differencing between stations also reduces the effects of tropospheric and ionospheric
refraction on the propagation of the radio signals If the ground stations have hydrogen-
maser oscillators (with stabilities approaching 1 part in 1015 over several hours), then
single differences can, in principle, be useful, as they are for VLBI. In practice, however,
it is seldom cost effective to use hydrogen masers and single difference observations in
GPS surveys. Rather, we form a double difference by differencing the between-station
differences also between satellites to cancel completely the effects of variations in the
station clocks. In this case the observations are just as accurate with low-cost crystal
oscillators as with an atomic frequency standard.
Since the phase biases of the satellite and receiver oscillators at the initial epoch are
eliminated in doubly differenced observations, the doubly differenced range (in phase
units) is the measured phase plus an integer number of cycles. (One cycle has a
wavelength of 19 cm at L1 and 24 cm at L2 for code-correlating receivers; half these
16 July 2008
 4
values for squaring-type receivers used prior to the mid-1990s.) If the measurement
errors, arising from errors in the models for the orbits and propagation medium as well as
receiver noise, are small compared to a cycle, there is the possibility of determining the
integer values of the biases, thereby obtaining from the initially ambiguous doubly
differenced phase an unambiguous measure of doubly differenced range. Resolution of
the phase ambiguities improves the uncertainties in relative position measures by about a
factor of 1.5 for 24-hr sessions, 3 for 8-hr sessions and more than 5 for short sessions.
(see, e.g., Blewitt [1989], Dong and Bock [1989] ).
GAMIT incorporates difference-operator algorithms that map the carrier beat phases into
singly and doubly differenced phases. These algorithms extract the maximum relative
positioning information from the phase data regardless of the number of data outages,
and take into account the correlations that are introduced in the differencing process.
(See Bock et al. [1986] and Schaffrin and Bock [1988] for a detailed discussion of these
algorithms.) An alternative, (nearly) mathematically equivalent approach to processing
GPS phase data is to use formally the (one-way) carrier beat phases but estimate the
phase offset due to the station and satellite clocks at each epoch. This approach is used
by the autcln program in GAMIT to compute one-way phase residuals for editing and
display.
In order to provide the maximum sensitivity to geometric parameters, the carrier phase
must be tracked continuously throughout an observing session. If there is an interruption
of the signal, causing a loss of lock in the receiver, the phase will exhibit a discontinuity
of an integer number of cycles (“cycle-slip”). This discontinuity may be only a few
cycles due to a low signal-to-noise ratio, or it may be thousands of cycles, as can occur
when the satellite is obstructed at the receiver site. Initial processing of phase data is
often performed using time differences of doubly differenced phase ("triple differences",
or "Doppler" observations) in order to obtain a preliminary estimate of station or orbital
parameters in the presence of cycle slips. The GAMIT software uses triple differences in
editing but not in parameter estimation. Rather, it allows estimation of extra bias
parameters whenever the automatic editor has flagged an epoch as a cycle slip that cannot
be repaired. Various algorithms to detect and repair cycle slips are described by Blewitt
[1990], and also in Chapter 4 of the GAMIT Reference Manual.
Although phase variations of the satellite and receiver oscillators effectively cancel in
doubly differenced observations, errors in the time of the observations, as recorded by the
receiver clocks, do not. However, the pseudo-range measurements, together with
reasonable a priori knowledge of the station coordinates and satellite position, can be
used to determine the offset of the station clock to within a microsecond, adequate to
keep errors in the doubly differenced phase observations below 1 mm.
A major source of error in single-frequency GPS measurement is the variable delay
introduced by the ionosphere. For day-time observations near solar maximum this effect
can exceed several parts per million of the baseline length. Fortunately, the ionospheric
delay is dispersive and can usually be reduced to a millimeter or less by forming a
16 July 2008
 5
particular linear combination (LC, sometimes called L3) of the L1 and L2 phase
measurements:
φ LC = 2.546 φ L1 − 1.984 φ L2 (1)
(See, e.g., Bender and Larden [1985], Bock et al. [1986], or Dong and Bock [1989])
Forming LC, however, magnifies the effect of other error sources. For baselines less than
a few kilometers the ionospheric errors largely cancel, and it is preferable to treat L1 and
L2 as two independent observables rather than form the linear combination. The station
separation at which the ionspheric errors exceed the phase noise depends on many factors
(receiver, antenna, multipath environment, latitude, time of day, sunspot activity) and
must be determined empirically by analyzing the data with both observable types.
In examining phase data for cycle slips, it is often useful to plot several combinations of
the L1 and L2 residuals. Single-cycle slips in L1 or L2 will appear as jumps of 2.546 or
1.984 cycles, respectively, in LC. Single-cycle slips in both L1 and L2 (a more common
occurrence) appear as jumps of 0.562 cycles in LC, which, though smaller, may be more
evident than the jumps in L1 and L2 because the ionosphere has been eliminated. If the
L2 phase is tracked using codeless techniques, the carrier signal recorded by the receiver
is at twice the L2 frequency, leading to half-cycle jumps when it is combined with full-
wavelength data. Hence, a jump of a "single" L2 cycle will appear as 0.992 in LC and
simultaneous jumps in (undoubled) L1 and (doubled) L2 will appear as 1.554 cycles in
LC. Another useful combination is the difference between L2 and L1 with both
expressed in distance units
φ LG = φ L2 − 0.779 φ L1 (2)
called "LG" because the L2 phase is scaled by the "gear" ratio (f2/f1 = 60/77 =
1227.6/1575.42). In the LG phase all geometrical and other non-dispersive delays (e.g.,
the troposphere) cancel, so that we have a direct measure of the ionospheric variations.
One-cycle slips in L1 and L2 are of course difficult to detect in the LG phase in the
presence of much ionospheric noise since they are equivalent to only 0.221 LG cycles.
If precise (P-code) pseudorange is available for both GPS frequencies, then a "wide-lane"
(WL) combination of L1, L2, P1, and P2 can be formed which is free of both ionospheric
and geometric effects and is simply the difference in the integer ambiguities for L1 and
L2:
WL = n2 - n1 = φ L2 − φ L1 + (P1 + P2 ) (f1 - f2)/(f1 + f2) (3)
The WL observable can be used to fix cycle slips in one-way data [Blewitt, 1990] but
should be combined with LG and doubly differenced LC to rule out slips of an equal
number of cycles at L1 and L2.
These various combinations of phase and pseudorange observations are used not only in
data editing, but also in resolving the phase ambiguities. When the LC observable is
16 July 2008
 6
used, we determine the L1 and L2 ambiguities by first resolving n2 - n1 (“widelane”)
and then n1 (“narrow lane”). If precise and unbiased pseudoranges are available, the
widelane ambiguities can be resolved for baselines up to thousands of kilometers under
any ionosphere conditions. For measurements prior to 1995, and possibly prior to 2000,
inter-channel receiver biases can corrupt the pseudoranges and it is necessary to use the
phase observations alone with a constraint on the ionopshere to resolve the widelane
ambiguities (see, e.g., Blewitt, 1989; Dong and Bock, 1989; Feigl et al., 1993]. GAMIT
gives you the option of selecting the method to be used, either pseudoranges (LC_AUTCLN)
or ionospheric constraints (LC_HELP). When using the pseudorange approach with
different receiver types, it is important to use the satellite-dependent differential code
biases (DCBs) computed from tracking data by the Center for Orbit Determination in
Europe (CODE) [http://www.aiub.unibe.ch/ionosphere.html]. If the wide-lane ambiguity for a
given doubly differenced combination has been resolved, resolving the narrow-lane
ambiguity for that combination depends on the level of noise from the receiver,
multipathing, and the troposphere, and the accuracy of the models employed for the
position and motion of the stations and satellites. It is generally more difficult to resolve
these ambiguities for the longest baselines.
1.2 Modeling the motions of the satellites and stations
A first requirement of any GPS geodetic experiment is an accurate model of the satellites'
motion. The (3-dimensional) accuracy of the estimated baseline, as a fraction of its
length, is roughly equal to the fractional accuracy of the orbital ephemerides used in the
analysis. The accuracy of the Broadcast Ephemerides computed regularly by the
Department of Defense using pseudorange measurements from < 10 tracking is typically
1-5 parts in 107 (2-10 m), well within the design specifications for the GPS system but
not accurate enough for the study of crustal deformation. By using phase measurements
from a global network of over 100 stations, however, the International GNSS Service
(IGS) [Beutler et al., 1994a], is able to determine the satellites' motion with an accuracy
of 1 part in 109 (2 cm; 5-20 cm in earlier years). For GPS surveys prior to 1994, the
global tracking network was much smaller but can still be used to achieve accurate
results for regional surveys. If we estimated orbital parameters and include in our
analysis observations from widely separated stations whose coordinates are well known,
the fractional accuracy of the baselines formed by these stations is transferred through the
orbits to the baselines of a regional network. For example, a 10 mm uncertainty in the
relative position of sites 2500 km apart introduces an (approximate) uncertainty of only 1
mm in the components of a 250 km baseline. This scheme can be used successfully even
with regional fiducial stations, transferring, for example, the relative accuracy of 250-500
km baselines to a network less than 100 km in extent, a helpful approach with surveys
conducted prior to the availability of precise global orbits.
The motion of a satellite can be described, in general, by a set of six initial conditions
(Cartesian position and velocity, or osculating Keplerian elements, for example) and a
model for the forces acting on the satellite over the span of its trajectory. To model
accurately the motion, we require knowledge of the acceleration induced by gravitational
attraction of the sun, moon, and higher order terms in the Earth's gravity field, and some
16 July 2008
 7
means to account for the action of non-gravitational forces due, for example, to solar
radiation pressure and gas emission by the spacecraft's batteries and attitude-control
system. For GPS satellites non-gravitational forces are the most difficult to model and
have been the source of considerable research over the past 20 years (see Colombo
[1986] Lichten and Bertiger [1989], Beutler et al. [1994b] for more discussion).
In principle, a trajectory can be generated either by analytical expressions or by
numerical integration of the equations of motion; in practice, numerical integration is
almost always used, for both accuracy and convenience. The position of the satellite as a
function of time is then read from a table (ephemeris) generated by the numerical
integration. In GAMIT the integration is performed by program arc using equations
given by Ash [1972].
Besides the orbital motion of a satellite, we must take into account meter-level offsets
between its center of mass and the phase-center of the transmitting antenna, including
temporary excursions of several decimeters lasting up to a half-hour during the
maneuvers the satellites execute to keep their solar panels facing the Sun when the orbital
plane is nearly aligned with the Earth-Sun direction. For the satellites in each orbital
plane, this alignment occurs for several weeks twice a year, the so-called "eclipse
season". Yoaz Bar-Sever and colleagues at JPL have spent considerable effort
developing models of the satellites' orientation, even to point of making the behavior
more predictable by getting DoD to apply a small bias about the yaw axis—a change that
was implemented gradually between June, 1994, and November, 1995. See Bar-Sever
[1996] for a complete discussion.
The position of the ground station in the Earth-centered inertial system defined by the
satellites’ orbits is affected by a number of geophysical phenomena. These include the
Earth’s rotation, precession and nutation of the spin axis in inertial space, motion of the
spin axis with respect to the crust (“wobble”), luni-solar solid-body tides, and loading of
the crust by ocean tides and the atmosphere. All of these effects are incorporated into
the model of the phase and pseudorrange observations computed in program model.
In modeling the phase and pseudorange observations, we must also take into account
changes in the apparent distance due to variations in the phase centers of the transmitting
and receiving antennas. With matched ground antennas in a regional network, these
effects nearly cancel, but for longer baselines and/or different antenna types they can
amount to several centimeters in estimated heights. Until recently, it was not possible to
model phase center differences and variations for the GPS satellites, but recent work
within the IGS (http://www.aiub.unibe.ch/download/igsws2004/Antenna_Effects/THPM2_Mader.pdf) has
produced models for all but the earliest GPS satellites. These models, together with
“absolute” models for the ground antennas will replace “relative” models for ground
antennas sometime in 2006.
Also part of the phase and pseudorange model is the propagation delay caused by the
neutral part of the Earth’s atmosphere. This effect is represented by a time-dependent
16 July 2008
 8
“zenith delay”, a “mapping function” that extends the delay to other elevation angles, and
a simple function for north-south and east-west gradients. The zenith delay and gradient
parameters are usually estimated in the analysis. The details of these models are
described in Chapter 7 of the GAMIT Reference Manual.
1.3 Parameter estimation
GAMIT (program solve) incorporates a weighted least squares algorithm to estimate the
relative positions of a set of stations, orbital and Earth-rotation parameters, zenith delays,
and phase ambiguities by fitting to doubly differenced phase observations. Since the
functional (mathematical) model relating the observations and parameters is non-linear,
GAMIT produces two solutions, the first to obtain coordinates within a few decimeters,
and the second to obtain the final estimates (See the discussion in Section 2.2 of the
GAMIT Reference Manual.)
In current practice, the GAMIT solution is not usually used directly to obtain the final
estimates of station positions from a survey. Rather, we use GAMIT to produce
estimates and an associated covariance matrix ("quasi-observations") of station positions
and (optionally) orbital and Earth-rotation parameters which are then input to GLOBK or
other similar programs to combine the data with those from other networks and times to
estimate positions and velocities [Feigl et al., 1993; Dong et al., 1998]. GLOBK uses a
Kalman filter (equivalent to sequential least squares if there are no stochastic parameters
in the solution) which operates on covariance matrices rather than normal equations and
hence requires that you specify a non-infinite a priori constraint for each parameter
estimated (see, e.g., Herring et al. [1990]). In order not to bias the combination, GAMIT
generates the solution used by GLOBK with loose constraints on the parameters. Since
phase ambiguities must be resolved (if possible) in the phase processing, however,
GAMIT also generates several intermediate solutions with user-defined constraints
before loosening the constraints for its final solution. These steps are described in detail
in Section 3.4 of the GAMIT Reference Manual.
In parameter estimation based on least-squares, the conventional measure of goodness-of-
fit is the χ2 (chi-square) statistic, defined for uncorrelated data as the sum of the squares
of each observation residual (post-fit observed minus computed observation, “o-c”)
divided by its assigned uncertainty. In a GPS analysis parameter correlations arise (even
if not represented in the original data weights) so the computation of χ2 in GAMIT or
GLOBK involves a complex matrix operation (see Dong et al., [1998]), but the idea is
the same. The value of χ2 is usually normalized by dividing by the “degrees of freedom”
(df), the number of observations minus the number of parameters estimated, so that the
ideal value for properly weighted, independent random observations is 1.0. Conversely,
with whatever a priori weight is assigned to the observations, multiplying the estimated
parameter uncertainties by the square root of χ2/df yields the “formal errors” (“formal
standard deviations”) of the parameter. With white noise, these formal uncertainties will
be realistic and will be inversely proportional to the square root of the number of
observations (i.e., will depend strongly on the sampling interval).
16 July 2008
 9
The uncertainties in a GPS analysis, however, cannot be treated with white noise
statistics because errors with temporal correlations dominate both the phase observations
and estimates of station coordinates (the quasi-observations input to GLOBK). In the
phase residuals (from cview or the sky plots produced by sh_gamit), the visible noise—
from multipathing and tropospheric fluctuations—is typically correlated over spans of
15-30 minutes. This implies that only samples taken at these intervals are independent,
and, to a first approximation, we would get realistic uncertainties by multiplying the
formal errors by the square root of the ratio of this interval to the sampling interval
used—e.g., for the 2 minute sampling commonly used in solve, we would increase the
uncertainties by a factor of 3-5. There also errors with longer correlation times that do
not show up in the residuals but are absorbed into the parameter adjustments. Assessing
the magnitude of these errors requires us use noise visible in the residuals (phase or
coordinates) to infer the character of the noise at lower frequencies. See Mao et al.
[1999]; Dixon et al. [2000], Williams [2003], and Herring [2003] for two approaches to
this problem. Once we have adopted a particular weighting of the data, it is often
possible to use external knowledge of the expected behavior of the coordinates or
velocities to validate the uncertainties; see McClusky et al. [2000], Davis et al. [2003],
and McCaffrey et al. [2006].
In GAMIT/GLOBK there are several ways you can control the uncertainties you obtain
for coordinates and velocities, and it is important for you to keep clearly in mind how
each of these operates. The uncertainties generated by solve and passed to GLOBK in
the h-file are determined by the a priori error assigned to the phase observations and by
the sampling interval—solve does not rescale by the square root of χ2/df (“postfit nrms”
in the q-file). In the initial (“preliminary”) solution, we normally assign an uncertainty of
10 mm to each one-way L1 phase. By Equation (1), the assigned uncertainty in an LC
phase becomes 32 mm. The mean rms of one-way LC residuals is typically ~6-9 mm, so
the nrms computed by solve is 0.2-0.3. In the second (“final”) solution, we normally
reweight the observations using a constant and elevation-dependent term computed in
data editing by program autcln from the actual (one-way LC) phase residuals. In order to
keep the overall weighting approximately the same as with the 10 mm constant error, the
values computed by autcln (table in file autcln.post.sum) are multiplied by an arbitrary
factor of 1.7 (in script sh_sigelv) before being input to solve (via the N-file). We chose to
use inflated values of the a priori phase error and not rescale by the nrms in order to
generate coordinate uncertainties that (in the presence of correlated noise) are
approximately realistic with 2-minute sampling. An equally valid approach would be to
rescale by the nrms (i.e. make χ2/df =1.0) and compensate later for the unrealistically low
coordinate uncertainties. With whatever weighting you use in solve, you can increase the
coordinate uncertainties used by GLOBK by rescaling all covariances on the h-file or by
adding white noise or random-walk noise to the variances of individual stations.
To obtain meaningful estimates of crustal motion, it is necessary to define a reference
frame by imposing constraints on the solution. These come in two common flavors:
With finite constraints, we assign realistic a priori uncertainties to the coordinates and
velocities of one or more sites. This is the only option available in GAMIT and it is also
16 July 2008
 10
available in GLOBK. A minimum (non-redundant) constraint with this approach would
involve restricting translation by fixing or tightly constraining the three coordinates of
one site, and restricting rotation by constraining Earth orientation. Constraining
additional sites provides redundancy, but can distort the network and remove ability to
detect errors in those sites except through an increase in χ2. The second type of
constraint, available through the glorg program of GLOBK, is generalized. With this
approach, we choose as large a set as possible of sites with good a priori values and
minimize their adjustments while estimating an overall translation and rotation of the
network. Since all of the frame-defining sites are free to move, outliers can be readily
detected and removed. Moreover, with generalized constraints there can be no internal
distortion of the network: all realizations of the reference frame will differ only by a
translation and rotation. See Dong et al., [1998] for a mathematical description of each
of these approaches.
2. Automatic Processing with GAMIT and GLOBK
GAMIT is composed of distinct programs which perform the functions of preparing the
data for processing (makexp and makex), generating reference orbits for the satellites
(arc), computing residual observations (O-C's) and partial derivatives from a geometrical
model (model), detecting outliers or breaks in the data (autcln), and performing a least
squares analysis (solve). Although the modules can be run individually, they are tied
together through the data flow, particularly file-naming conventions, in such a way that
most processing is best done with shell scripts and a sequence of batch files set up a
driver module (fixdrv) for modeling, editing, and estimation. Though the data editing is
almost always performed automatically, the solution residuals can be displayed or plotted
so that problematic data can be identified (cview).
Likewise, GLOBK operates through distinct programs, which can be invoked with a
single command or run separately. The primary functions are to combine quasi-
observations--either GAMIT/GLOBK “h-files” or the internationally accepted SINEX
format--from multiple networks and/or epochs (glred or globk), and to impose on this
solution a reference frame appropriate to the scientific objective (glorg). Note that globk
and glred are the same program, just called in different modes: glred to read data from
one day at a time for generating time series, globk for stacking multiple epochs to obtain
a mean position and/or velocity.
The full sequence of steps to take you from phase data to time series is accomplished
with two shell scripts: sh_gamit looks for raw or RINEX data over a range of days and
invokes the GAMIT programs to produce constrained and loose estimates of coordinates
together with sky plots of phase data as a record of the processing; sh_glred uses the
GAMIT results to produce time series of day-to-day repeatability or a combined h-file
that may be further combined with those from other epochs to estimate station velocities.
The only preparation required is assembling the meta-data from station logs; setting up
16 July 2008
 11
the control files, most of which are common to all analyses of a particular era; and
assembling the non-IGS phase data in one or more directories on your system.
2.1 Setup
The first step in running the scripts is to create an experiment directory and to link or
copy into it the standard control and data tables. All of the required templates and tables
reside in ~gg/tables, where ~gg is a required alias pointing to the highest level of the
GAMIT/GLOBK installation. Executing the script sh_setup will invoke sh_links.tables to
link into the experiment ./tables directory all of the standard data tables (see Section 4.1)
and will copy into the experiment ./tables directory the seven control and data files listed
below:
process.defaults: Edit this file to specify your computation environment, sources for
internal and external data and orbit files, start time and sampling interval, and
instructions for archiving the results.
: Edit this file to specify which local and IGS stations are to be used and
sites.defaults
how station meta-data are to be handled.
: This file should contain the receiver and antenna type and height of
station.info
instrument (HI) values as a function of time for all occupations of the stations you will
use. A template containing currenent values for IGS and many other continuous stations
is maintained by the Scripps Orbit and Permanent Array Center (SOPAC;
http://sopac.ucsd.edu/processing/gamit/) and copied daily to source/updates/tables in the
GAMIT/GLOBK distribution directory on chandler.mit.edu. If the RINEX files for your
local network contain IGS-standard receiver and antenna codes and the correct height
information refers to the ARP, then sh_gamit will append these to the SOPAC file
automatically. If not, then you must manually enter the GAMIT codes and height values
into the file. You can give station.info priority over the RINEX headers using the xstnfo
option in sites.defaults.
coordinate files : sh_gamit maintains in the experiment ./tables directory two files of a
priori coordinates. The file ending in .apr (set as aprf in process.defaults, itrf00.apr in the
template) contains the Cartesian coordinates (position and velocity) of stations you wish
to have unchanged throughout the processing. The L-file ( lfile. ), which can contain
either Cartesian position and velocity (default, same as .apr file) or spherical position
(GAMIT old-style, use -oldfmt_lfile option in sh_setup), is updated after each day is
processed if the adjustments exceed a specified value (0.3 m by default). If you have
good coordinates for stations not in the apr file, you should append these to the apr file if
you want them to be unchanged. For any station that does not have coordinates in lfile.,
sh_gamit will attempt to calculate coordinates via a pseudorange solution, or (if use_rxc Y in
process.defaults) use the coordinates in the RINEX header. When GAMIT runs, these
coordinates will be updated from the phase solution so that for successive days on which
the same station is observed, the accurate coordinates will be used.
16 July 2008
 12
and sittbl. : Edit these files to set the appropriate options for your analysis. Make
sestbl.
sure that any station for which you specify tight constraints in sittbl. has accurate
coordinates in the apr file.
autcln.cmd: This file will usually not require editing unless you encounter unusual data
during the processing.
When sh_gamit is executed, RINEX observation files are stored in the experiment /rinex
directory. Although this directory can be created automatically and populated by files
translated from raw data or located anywhere on your system ( rnxfnd in process.defaults), it
is usually convenient to create /rinex in advance and copy into it all of the RINEX files
from your experiment. RINEX files for continuous sites to be added to
To obtain a time series from multiple days of GAMIT processing, you need to set up two
control files and and possibly an a priori file for GLOBK.
Once you have edited appropriately the template files, you can start the processing from
within the experiment directory by giving sh_gamit simply the 4-character code for the
experiment and a range of days to process:
sh_gamit -expt emed -d 1999 235 236 237 238 >&! sh_gamit.log
The time span can also be specified using -s to indicate a range of
consecutive days, or -r to indicate that you want to process a single day
before the current date. You may also override some of the parameters specified in
process.defaults:
-orbit Type of orbit (IGSP IGSR IGSF SIOP SIOR SIOF)
-eops EOP series to be used (usno [default], bull_b)
-sessinfo Sampling interval, #epochs, start time (HH MM) (e.g. 30 2880 0 0 )
-copt List of files to compress in the day directory (default x, k, autcln.out, d)
-dopt List of files to delete from the day directory (default C-files only)
-pres Plot phase residuals as postscript skyplots (default no)
-nogifs Do not create gif files of sky plots (default is to create from postscript)
-netext Add net--specific suffix (character) to day directory names (e.g. 035r)
-noftp Don’t try to ftp any data from outside your system (default is to ftp)
-mailto E-mail address for summary
Most of the time the parameters may be omitted in favor of the values you have specified
in process.defaults for the whole experiment. The overrides are useful, however, if you
wish to test the effect of processing a day with a different orbit, EOP table, or session
length, in which case you can create second directory for the same day by appending a
character to its name (e.g., -netext t ). Finally, you may launch sh_gamit from anywhere on
the system by specifying the full path name of the processing directory with -dir .
When the script runs, it will write to the screen a record of each step, which you may
choose to redirect to a file (e.g., >&! sh_gamit.log). Though the current version of this log is
16 July 2008
 13
cryptic in parts, you should be able to use it together with GAMIT.fatal file, and the source
code for sh_gamit to identify the point and reason for failure should that occur. We
welcome user feedback to improve the displayed information.
When processing of each day is completed, sh_gamit will send a mail message to you
giving the number of stations used, the rms of the one-way phase residuals for the two
best and two worst stations from the AUTCLN postfit summary file, the nrms val