Geoscientific Researchand the

Global Positioning System

Recent Developments and Future Prospects

A Report of the

University Navstar Consortium

Summer, 1994

ii

iii

Contents

1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11.1 State of the Art. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2

1.1.1 Complex Deformation in Plate Boundary Zones. . . . . . . . . . . . . . . . . . . . . . . . . . . . .21.1.2 Capturing Earthquakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21.1.3 Continuously Operating GPS Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41.1.4 Volcano Monitoring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41.1.5 Post-Glacial Rebound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .51.1.6 Global Climate Change. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .51.1.7 Ocean Circulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .61.1.8 Atmospheric Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .61.1.9 Space-Based GPS Meteorology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .71.1.10 Probing the Ionosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7

1.2 The Future . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .71.2.1 Permanently Operating Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .71.2.2 Multipurpose National Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .81.2.3 Global Change Research. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .81.2.4 Space-Based Meteorology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9

2 Contributors to This Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9

3 UNAVCO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9

4 Scientific Opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .114.1 Deformation of the Earth’s Lithosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11

4.1.1 Plate Boundary Processes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12 4.1.1.1 Mantle Dynamics and Tibet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .14 4.1.1.2 The Pacific-North American Transform Boundary . . . . . . . . . . . . . . . . . . . .14 4.1.1.3 Complexity of Continental Convergence in the Eastern Mediterranean . . . .14 4.1.1.4 Southeast Asia and Indonesia Tectonics. . . . . . . . . . . . . . . . . . . . . . . . . . . . .17 4.1.1.5 The Interior Western U.S. and the Transition to the Stable Plate Interior. . .18

4.1.2 Volcanic Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20 4.1.2.1 Kilauea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22 4.1.2.2 The Yellowstone Caldera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23

4.2 The Earthquake Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .244.2.1 Loma Prieta. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26

4.3 Deep Earth and Whole Earth Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .274.3.1 Earth Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27

4.4 Interaction of Mantle Convection and Tectonics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .294.4.1 Vertical Motions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .294.4.2 Horizontal Motions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .304.4.3 Post-Glacial Rebound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .31

4.5 Ice Dynamics and Sea Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .334.5.1 Ice Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .33

iv

4.5.1.1 Mass Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .33 4.5.1.2 Ice Sheet Dynamics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .34

4.5.2 Sea Level: The Link Between Ice Dynamics and Tectonics. . . . . . . . . . . . . . . . . . . .344.6 Precise Mapping and Navigation in Oceanography. . . . . . . . . . . . . . . . . . . . . . . . . . . .35

4.6.1 Mapping of Global Sea Height . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .364.6.2 Acoustic Doppler Current Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .374.6.3 Ocean Warming. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .374.6.4 Seafloor Mapping. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .394.6.5 Other Shipboard Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .39

4.7 Atmospheric Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .404.7.1 Ground-Based GPS Meteorology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .414.7.2 Future Opportunities in Ground-Based GPS Meteorology. . . . . . . . . . . . . . . . . . . .424.7.3 Space-Based Meteorology with GPS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .444.7.4 Probing the Ionosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .45

4.8 Instrument Arrays and Data Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .45

5 Accomplishments and Ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .465.1 Subsurface Geodesy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .465.2 Measurement of Local Geoid Slope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .475.3 Precise Airborne Surface Altimetry Based on GPS Positioning. . . . . . . . . . . . . . . . . . .475.4 Airborne Geophysical Observations of Long Valley Caldera. . . . . . . . . . . . . . . . . . . . .485.5 GPS Measurement of Relative Motion of the Cocos and Caribbean Plates

and Strain Accumulation Across the Middle America Trench . . . . . . . . . . . . . . . . . . . . .495.6 Constraints on Deformation of the Resurgent Dome, Long Valley Caldera,

California from Space Geodesy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .495.7 GPS Monitoring of Crustal Strain in Southwest British Columbia with the

Western Canada Deformation Array.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .505.8 GPS Used to Monitor Ground Movements at Augustine Volcano, Alaska . . . . . . . . . . .505.9 Use of Airborne Kinematic GPS and Laser Altimetry for Determining Volume

Changes of Mountain Glaciers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .515.10 Use of GPS Technology in Glaciology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .525.11 Solar Hygrometer for GPS Field Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .535.12 Non-Equilibrium Forces on GPS Satellites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .545.13 Sub-Daily Earth Rotation During Epoch’92 from GPS . . . . . . . . . . . . . . . . . . . . . . . .545.14 Atmospheric Excitation of Rapid Polar Motions Measured by GPS. . . . . . . . . . . . . . .555.15 Large Scale Combination of GPS and VLBI Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . .555.16 Antarctic Search for Meteorites (ANSMET) Project. . . . . . . . . . . . . . . . . . . . . . . . . . .565.17 Landslide Hazard and GPS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .575.18 Crustal Deformation at Convergent Plate Boundaries. . . . . . . . . . . . . . . . . . . . . . . . .585.19 The GPS Flight Experiment on Topex/Poseidon. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .595.20 Use of GPS for Tethered Satellite Position Determination. . . . . . . . . . . . . . . . . . . . . .605.21 Status of GPS/Acoustic Measurements of Seafloor Strain Accumulation

Across the Cascadia Subduction Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .615.22 GPS in Volcanology: Monitoring of Icelandic Volcanoes. . . . . . . . . . . . . . . . . . . . . . .615.23 GPS Geodesy and Seismology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .625.24 Microplate Versus Continuum Descriptions of Active Tectonic Deformation . . . . . . .63

v

5.25 GPS Detects Vertical Surface Displacements Caused by AtmosphericPressure Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

5.26 Ionospheric Monitoring Using GPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655.27 Atmospheric Noise in Airborne Gravimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655.28 Improved GPS Vertical Surveying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 665.29 Sensing the Atmosphere with an Orbiting GPS Receiver . . . . . . . . . . . . . . . . . . . . . . . 675.30 GPS Sensing of Atmospheric Water Vapor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 675.31 GPS in Antarctica . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 685.32 Fast Technique for Computing Precise Aircraft Acceleration from GPS Phase. . . . . . 69

6 Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 706.1 GPS Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

6.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 706.1.2 Static Receiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726.1.3 GPS Ocean Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

6.2 Ancillary Data Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746.3 Infrastructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 756.4 Error Budget . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

6.4.1 Data Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 756.4.2 Multipath . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 766.4.3 Receiver Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 776.4.4 Atmosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 786.4.5 Second Order Ionospheric Delay. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 796.4.6 Satellite Motions and Satellite Antenna Characteristics . . . . . . . . . . . . . . . . . . . . . . 796.4.7 Impact of SA and AS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

vi

1

1 Overview

The Global Positioning System (GPS), a nowfully operational 26-satellite constellation,represents one of those rare technologicaldevelopments which suddenly opens up wholenew opportunities for scientific advance, withbroad influence on commerce and society atlarge. It was put in place by the U.S.Department of Defense primarily to provideaccurate geographic position location andnavigation for a wide variety of militarypurposes, through distance ranging to thesatellites. In designing the system, DoDadopted an enlightened dual-use policy,providing for civilian use of the system forpositioning at a limited level of accuracy (about100 meters), but one adequate for manyapplications.

The tremendous boom in civilian use of GPS inrecent years for a multitude of purposes—shipand private boat navigation, vehicle fleetlocation, emergency vehicle location, aircraftand spacecraft guidance, and commercialsurveying, to name a few—reveals the dual-usepolicy to have been remarkably foresightful. Alarge and rapidly growing commercial industryhas come into existence because of GPS, bothwithin the U.S. and abroad. Yet the designers ofthe system could hardly have foreseen thescientific bonanza of GPS applications thatcame about through the discovery that bycarefully measuring the phase of the radiosignals from the GPS satellites, as well as theranges to them, it is possible in principle todetermine positions with an accuracy to fiveorders of magnitude better than the system wasdesigned for. That is, using radio transmissionsreceived from satellites 20,000 kilometers up, itis now possible to determine relative positionsat or above the Earth’s surface to better than 1centimeter.

That this discovery originated in the spacegeodetic community was a naturalconsequence of the intensive effort over manyyears, under the sponsorship of NASA’sCrustal Dynamics Project, to develop thetechnologies for applying Very Long BaselineInterferometry (VLBI) and Satellite LaserRanging (SLR) to the direct measurement ofthe relative velocities of the Earth’s greattectonic plates on a global scale. Thus, theoriginal geoscientific applications of GPS hadthe same focus, but it was soon realized that thepotential applications extend far beyond platetectonics.

Realization of the scientific potential of GPShas produced a tremendous sense of excitementacross the geosciences—Earth science,atmospheric science, oceanography, polarscience, and geospace physics—and a strongnew sense of interdisciplinary community hasdeveloped. Much of the effort of the pastdecade has been focused on eliminatingsources of error that limit accuracy, and withsuch an intensity that the accuracy goals whichseemed only a distant possibility just a fewyears ago have already been exceeded. We arejust now in a period in which investigations inall areas of the geosciences are beginning tobear fruit, and are proving successful beyondexpectations. Scientific output is on an upwardcurve which parallels the rapid improvement inmeasurement accuracy and a decline inhardware costs (Figure 1).

The decade to come promises a proliferation ofscientific applications of GPS which willparallel and be closely connected with aproliferation in commercial and civilapplications. The intent of this report is to givean overview of developments in geoscientificapplications of GPS of recent years and anindication of things to come. What is alreadyclear is that there will be a continuum of diversescientific applications of GPS fromfundamental investigations into how the Earth

2

works to highly applied investigations withimmediate focus on diverse matters of publicsafety, such as earthquake hazard, movement oflandslides, climate change, sea level rise,coastal erosion, volcanic eruptions, landsubsidence, and the integrity of dams and otherman-made structures. But the distinctionbetween basic and applied science in the GPSrealm is not meaningful, and across the boardGPS will play an important part in organizednational and international programs such as theEarthquake Hazard Reduction Program and theInternational Decade of Natural HazardReduction. In this arena, the scientific

Figure 1. Trends in receiver cost, horizontalaccuracy, and vertical precision for GPS sci-entific applications (Ware).

0

10

20

30

40

50

mill

imet

ers

1985 1986 1987 1988 1989 1990 1991 1992 1993

Year

GPS Surveying

Vertical Accuracy or Precision

10-3

10-210-2

10-1

100

101

GPS Surveying

Horizontal Accuracy

part

s pe

r mill

ion

0

40

80

120

160

$K/U

nit

Geodetic GPS Receiver

UNAVCO $K/Unit

purc

hase

or u

pgra

de

1985 1987 1989 1991 1993

UNAVCO GPS ReceiverPurchase or Upgrade

$K/r

ecei

ver

mill

imet

ers

GPS SurveyingHorizontal Accuracy

GPS SurveyingVertical Precision

160

120

80

40

102

101

100

50

30

10

0

20

40

0

mill

imet

ers

community and the civil and commercialsectors already form an interdependent andmutually supporting network.

Below are given some highlights of recentdevelopments in the application of GPS togeoscientific research, followed by a look to thefuture, recognizing that the whole field isdeveloping with breathtaking speed, and thatmuch of what is today could not have beenforeseen only a few years ago.

1.1 State of the Art

1.1.1 Complex Deformation in PlateBoundary Zones

In contrast with predecessor technologies, GPSreceivers are highly mobile and relativelycheap. This means in principle that largenumbers of them can be deployed anywhere onEarth that benchmarks can be emplaced. Thus,the deformation patterns of active regions ofthe Earth’s surface can be observed to the detailrequired to adequately describe them. Thendynamic models can be built around theseobservations, through which we can infer thereal physical processes within the Earth andwhat causes them. This will be especiallyimportant in the boundary zones where tectonicplates interact, for example that between thePacific and North American plates in California(Figure 2), which, far from being knife-sharp,are broad deforming regions with complexnetworks of active faults and interveningblocks which interact and evolve with time.

1.1.2 Capturing Earthquakes

The University Navstar Consortium(UNAVCO) now provides equipment and tech-nical and logistical support for scores of GPSfield projects worldwide (Figure 3), and mostof these involve investigations of the tectonics

3

of plate boundary zones. For a large number ofthese regions, repeated GPS surveys of geo-detic networks spanning active fault zones havemade it possible to “capture” earthquakes—that is, determine the deformation field over awide area surrounding the locus of fault slip-page. Rapid deployment of GPS receiversfollowing the several large earthquakes ofrecent years in California have measured thepost-seismic deformations, which decay overlong time periods. Such measurements comple-ment information from seismographsconcerning the depth, orientation, and amountof fault slippage.

In tandem, geodesy and seismology are tellingus much about the mechanics of earthquake-generating fault motions and the larger tectonicframework which produces the stresses whichcause them. The magnitude 6.6 Northridgeearthquake of 1994, which devastated largesections of Los Angeles, was determined, by

GMT Aug 4 19:40 ../aug4/gk930702_none.all.mainpcf

122˚W 121˚W 120˚W 119˚W 118˚W 117˚W 116˚W 115˚W 114˚W

32˚N

33˚N

34˚N

35˚N

36˚N

37˚N

100 km

BRSH

FTOR

BLAN

BLHL

VNDN

LOSP

MADC

MUNS

CENT

TWIN

FIBR

BLUF

PVER

OVRO

JPL1

PEAR

NIGU

MOJA

DEAD

BLKB

MONP

PIN1

SOLJ

10 mm/yr

122˚W 121˚W 120˚W 119˚W 118˚W 117˚W 116˚W 115˚W 114˚W

32˚N

33˚N

34˚N

35˚N

36˚N

37˚N

95% confidence, scaled σ

Fig. 2. Observed Vel. wrt Pacific Plate

Figure 2. Observed velocity vectors and 95%confidence ellipses for the combined GPS andVLBI data in southern California (Feigl et al.,1993). Velocities are referenced to the PacificPlate.

combined analysis of GPS and seismologydata, to be the result of a previously unknown“blind” thrust fault at depth in the Los AngelesBasin (Figure 4). But the regional pattern ofdeformation of which that fault was a part, notdirectly related to the San Andreas system, hadalready started to emerge from repeated GPS

Figure 3. Locations of UNAVCO-supportedGPS surveying projects.

UNAVCO Campaigns 1986 - 1994

GMT Jan 25 18:23 Hudnut and Murray/USGS

118˚ 40' 118˚ 20'34˚ 00'

34˚ 20'

PACO

CALA

PVEP

GLEN

MULH

NORT

PICO

RESE

SAFR

2131

JPL1

50 mm10 km

Northridge GPS Displacements

Hudnut and Murray, USGS

Figure 4. Northridge GPS displacements andmodeling of slip zone at depth from surfacedeformation (Hudnut and Murray, January 25,1994).

34o 20'

34o 30'

118o 40' 118o 30'

4

surveys of the Los Angeles and Ventura Basinsjust in recent years (Figure 5).

1.1.3 Continuously Operating GPSNetworks

Within plate boundary zones, the relativemovement of the adjoining plates causes stressto accumulate, which is relieved by slip onfaults by elastic strain. For an extended periodthereafter, anelastic deformation continues at adecaying rate, as stresses again build, until sliprecurs; this is the earthquake cycle. As is onlytoo well known, such fault slip usually isassociated with earthquakes. Analysis of thetime and space variation in deformation duringand after fault movements using GPS, as wellas traditional geodetic tools, tells us muchabout the rheology of the crust and underlyingmantle and the tectonic processes at work, andmay shed light on the frequency and amount offuture fault slip.

Figure 5. Relative velocities of sites in the Ven-tura Basin region, California (Donnellan etal., 1993). The error ellipses represent 95%confidence.

The 15-station Permanent GPS Geodetic Array(PGGA) (Figure 6), a recently establishednetwork of continuously recording GPSreceivers in southern California, operated bythe Scripps Institution of Oceanography,captured the Northridge earthquake, as it hasother recent earthquakes, such as themagnitude 7.4 Landers earthquake of 1992(Figure 7). The virtues of having fixedinstruments which continuously monitornetworks of baselines in order to observepatterns of movement of the Earth after, during,and (perhaps in the future) before anearthquake is evident. A similar network isoperational in Japan, and these networks areforerunners of much more extensive networksin the future.

1.1.4 Volcano Monitoring

The success of detailed earthquake monitoringin predicting the recent eruptions of Mt.Pinatubo in the Philippines and Mt. Redoubt inAlaska saved thousands of lives and points outthe use of geophysical tools to forecast eminenteruptions. As in earthquake research, the GPSreceiver has become complementary to theseismograph for scientific investigations ofactive volcanos, while at the same time

Figure 6. Permanent GPS geodetic array(PGGA), southern California (Bock).

238˚

238˚

240˚

240˚

242˚

242˚

244˚

244˚

246˚

246˚

32˚ 32˚

34˚ 34˚

36˚ 36˚

Blythe

Chatsworth

Harvest Platform JPL

Pinyon Flat

Scripps

Vandenberg

Parkfield

Lake Mathews

Palos Verdes

Goldstone

Yucaipa

China Lake

Bommer Canyon

Ensenada

Monument Peak

Permanent GPS Geodetic Array (PGGA) - Southern California

5

providing an important new tool for volcanohazard assessment and eruption warning.

In recent years, GPS networks have beenestablished in numerous areas of activevolcanism—for example Hawaii, Alaska,Yellowstone, Iceland, New Zealand, andItaly—and as noted in this report are alreadyproviding scientific insight into complexvolcanic processes. GPS can give three-dimensional views of the movement of volcanosurfaces in response to seismic or aseismicactivity; this has provided a new means forinferring the volume and movement of magmaat depth.

1.1.5 Post-Glacial Rebound

Large parts of the Earth’s surface which weredepressed under the weight of huge continental

242˚

242˚

243˚

243˚

244˚

244˚

33˚ 33˚

34˚ 34˚

35˚ 35˚

10

30

30

30

100

100

300300

1000

PIN1

DS10

JPL1

SIO2

PIN1

DS10

JPL1

Theoretical

Bock et al.

Blewitt et al.

20 mm

Figure 7. Observed versus modeled displace-ments from the PGGA tracking network for theLanders earthquake (Bock et al., 1993). Thecontours show displacement magnitude in mil-limeters for the elastic dislocation model.

ice sheets during the last glaciation are stillslowly rebounding at rates measurable by GPS.From rates of post-glacial rebound, inferencescan be made about the rheological properties ofthe Earth’s mantle. Such information is ofcrucial importance to modeling the slowconvective circulation of the mantle, whichultimately drives plate tectonics.

Previous studies were limited to estimatingvertical rates from inferred ancient shorelines;GPS, on the other hand, can directly measureboth vertical and horizontal motions anywherea benchmark can be emplaced. The horizontalrate is much more sensitive to the verticalvariation of mantle viscoelastic properties andto the ice loading history than the vertical rate.GPS measurements should make it possible todistinguish between competing models ofmantle rheology within a decade.

The problem is complicated for glaciated areasnear plate boundaries, where multiple deforma-tions may superimpose; in coastal areas, botheffects need to be distinguished from realeustatic sea level rise. This requires well-distributed, high-accuracy GPS measurementscombined with geologic observations andmodels. The Pacific Northwest is a good exam-ple of where this is important: inferences maybe possible about whether the region is in dan-ger of a great earthquake, provided the tectonicsignal due to subduction of the Juan de Fucaplate beneath North America can be distin-guished from that due to post-glacial reboundand real changes in sea level.

1.1.6 Global Climate Change

A plausible result of anthropogenic warming ofthe atmosphere through addition of greenhousegasses would be melting of the Earth’s icesheets and mountain glaciers. This would leadto a rise in sea level which could have veryserious consequences for low-lying coastalpopulation centers in terms of inundation,

6

coastal erosion, and salt encroachment inaquifers. On the other hand, it is also plausiblethat atmospheric warming could lead toincreases in precipitation which could cause atleast some ice masses to grow.

Although some mountain glaciers do seem tobe shrinking, at present it is not at all clearwhether the polar ice caps are shrinking orgrowing. The problem is lack of data. There issome consensus, based on tide gauge data, thatsea level has risen on the average of about 2mm/year over the last several decades, butmeasurement of sea level change isexceedingly difficult due to the superpositionof innumerable effects. GPS is being applied tothe climatic temperature question in a numberof new ways which are likely to produce somerapid advances.

The topographic profile of mountain glaciersare being measured remotely from aircraftpositioned by GPS with a view towarddetecting temporal changes. The same is beingapplied to the Antarctic ice sheets and rapidlymoving ice streams, along with directmeasurements of the ice surface using GPS.These observations will complementdeveloping techniques for measuring changesin ice topography from spacecraft.

Tide gauge sites have begun to be tied to GPSnetworks; ultimately, tying a global network oftide gauges to a common GPS-based referenceframe will greatly facilitate the determinationof absolute sea level changes.

Warming of the oceans will accompanywarming of the atmosphere. An experiment hasbeen carried out to measure the time oftransmission of acoustic waves through thesurface layers of the ocean along very longoceanic paths. Accurate distancemeasurements were achieved with GPS.Repeat measurements will look for changes in

average transmission velocity due to warmingof the ocean.

1.1.7 Ocean Circulation

GPS is providing unprecedented orbit accuracyfor the TOPEX/Poseidon spacecraft, which ismeasuring ocean topography in conjunctionwith the World Ocean Circulation Experiment(WOCE). Ocean topography is the basis forinferring large-scale circulation of ocean cur-rents. Improved orbits lead to improveddetermination of ocean circulation on all scalesfor both time variable and time averagemotions, which in turn leads to improvement inclimate models. Small geographicallycorrelated orbit errors make possible the deter-mination of absolute geostrophic circulation(net flow through the full water column), whichcannot be determined globally otherwise. GPSis providing a reference frame for acousticDoppler current profilers, which measure thecurrent profiles in the upper few hundredmeters of the ocean from ships by providingaccurate ship speed and heading. By a quite dif-ferent approach, deep ocean current velocitiesare being measured by use of GPS receivers ona surface ship and buoy, while tracking viasonar a probe drifting in a deep current.

1.1.8 Atmospheric Sensing

Water vapor plays a crucial role in atmosphericprocesses acting over a wide range of temporaland spacial scales. Knowledge of thevariability of water vapor is central tounderstanding weather phenomena and, overthe long term, climate. Water vapor variationscause changes in velocity, and thus refraction,of the GPS signals passing through theatmosphere. It was thought early on that errorsin GPS positioning due to water vapor couldwell limit the ultimate utility of GPS forscientific uses. But quite the opposite occurred:methods were developed, not only to correct

7

for path delays, but to turn the problem aroundand use the delays as a tool to remotely sensevariations in atmospheric water vapor.

GPS-based meteorology has thus become anextremely valuable new tool for theatmospheric sciences. The method relies on theuse of networks of continuously operating GPSreceivers simultaneously observing multipleGPS satellites. Recently developed water vaporradiometers which are highly accurate andportable can be used as a companion tool to fixbiases in small-aperture arrays. The method hasbeen shown to be highly successful in testsinvolving placement of GPS receivers at theNOAA Wind Profiler sites in the U.S.midcontinent.

1.1.9 Space-Based GPS Meteorology

This approach takes the well-known radiooccultation methodologies for studyingplanetary atmospheres and applies it to theEarth. Observations of signal delays alongsuccessive paths to rising or setting GPSsatellites, as observed by a small satellite inlow-Earth orbit (LEO), can be inverted to avertical profile of refractivity through theatmosphere. Refractivity is a function oftemperature and water vapor.

Thus, in principle, if the temperaturedistribution is known or can be modeled, thevertical distribution of water vapor can beinferred, or vice versa. The method provides ahigh degree of vertical resolution, but lowresolution horizontally. Because ground-basedGPS meteorology involves high resolutionlaterally, but low resolution vertically, themethods are highly complementary. A similarstatement can be made about space-based GPSmeteorology in comparison with the imagingradiometers used in the familiar NOAAweather satellites. A demonstration experimentis under way to collect occultation data from aGPS receiver accompanying a commercial

LEO satellite. The single receiver should beable to record more than 500 occultations perday with nearly global coverage.

1.1.10 Probing the Ionosphere

This ionized part of the Earth’s atmosphereaffects the GPS signals dispersively—that is,the delay experienced by a signal is a functionof its frequency. Dual frequency receivers cancorrect for this effect, and in the processdetermine the degree of ionization of theionosphere along the path of the radio signals.Precise dual frequency measurements byground and space-based receivers will probethe ionosphere along many paths.Measurement of total electron content alongthese paths will be assembled into high-resolution 3-D electron distribution maps bycomputerized tomography to study irregularand transient mesoscale structures, such asequatorial bubbles and the mid-latitude trough.Study of the ionosphere with GPS signals willlead to a better understanding ofcommunication problems associated withionospheric disturbances.

1.2 The Future

Based on the experience of the last severalyears, it is quite likely that many newapplications of GPS in all areas of thegeosciences will emerge over the next severalyears which we cannot now foresee. On theother hand, we can identify certain thingswhich, if adequately nurtured, have areasonable expectation of reaching fruitionwithin the next 5–10 years.

1.2.1 Permanently OperatingNetworks

Permanently operating GPS networks arelikely to proliferate in the future in earthquake-

8

prone regions with the continuing decline inreceiver costs, expansion of data storage andnetworking capability, and increasing numbersof people trained in GPS technology, bothwithin the U.S. and abroad. Such networks willcomplement seismic networks alreadyestablished in many such regions. Because ofthe same factors, regional GPS networks inmany areas of the world will be outfitted withpermanent receivers, which will havesignificant logistical advantages over thecurrent mode involving massive shipmentsfrom the U.S. or Europe and the necessity ofcoordinating large, complex field campaigns.Such networks will provide frameworks forlocal scientific or civil surveys. Along the samelines, there is likely to be a strong movement toestablish continuously monitoring GPSnetworks on active volcanoes worldwide.

1.2.2 Multipurpose National Network

Several government agencies intend toestablish GPS networks within the U.S. for awide variety of applications. A significantopportunity exists for coordinating theseefforts toward the establishment of a dense,high-quality national GPS network whichcould serve many purposes—scientific, civil,and commercial—while providing significantcost leveraging for all parties concerned. Ofcentral importance is the deployment of high-quality GPS receivers which would satisfy allrequirements.

Such a network would simultaneously serve theFAA’s program for instrumenting airports;NOAA’s programs in meteorology, sea level,and geodetic networks; the DOE ARMprogram in water vapor climatology; the CoastGuard’s program to provide differential GPS(DGPS) service in coastal areas; and NSF andNASA programs in active tectonics andsatellite tracking.

At the same time, the DGPS capability wouldbe available to a wider community of users,such as transportation companies, emergencyservices, commercial surveyors, and schools. Itcould be used by the space physics communityto study ionospheric effects which have a majorimpact on commercial and governmentcommunications systems. Such a system wouldbe a case of technology transfer of genuinelyhigh value on a large scale.

1.2.3 Global Change Research

Continuing reduction in errors in orbits ofsatellites measuring ocean topography throughGPS tracking will lead to much betterdetermination of ocean circulation. This couldlead to the computation of flux divergences ofheat to and from the atmosphere from the oceanat a level of accuracy required to infer directlya greenhouse warming.

A network of continuously operating GPSstations at coastal tide gauges, tied to a globalreference frame, will allow much betterdetermination of absolute sea level trends,while providing input to models of isostaticreadjustments and shifts in the Earth’s center ofmass due to redistributions of water betweenthe ice caps and the oceans. A network of GPSstations will be established around theAntarctic to monitor elastic crustal strains dueto ice addition or subtraction from its ice caps.

Extensive GPS networks established on icecaps and glaciers will be used to monitor iceflow, in an effort to determine long-termpatterns of accumulation or wastage.

An oceanographic experiment to look forsecular trends in long range acoustic traveltimes between fixed sites positioned by GPSwill be complemented by similar experimentsinvolving drifting buoys, also positioned byGPS. Such experiments should be able to detecta greenhouse warming signal within a decade.

9

There is much potential for using DGPS andkinematic surveying to map coastal zones indetail sufficient to study erosion rates, stormdamage, and ecosystem modification.

1.2.4 Space-Based Meteorology

Following demonstration phases for using GPSoccultation measurements from LEO satellitesto determine vertical refractivity profiles in theatmosphere, a significant opportunity exists forexpanding the effort significantly.

If provision could be made for outfitting thehundreds of LEO communications satellitesplanned for the next decade with suitable GPSreceivers, occultation measurements could bemade with a density adequate for providingvaluable input to daily meteorological forecastmodels on the mesoscale globally. Further, overthe longer term such measurements willprovide important input to climate models.

Simulations suggest that changes in thetemperature profile of the atmosphere due togreenhouse warming should be detectable in amatter of years (Figure 8). The potentialbenefits of such a program, resulting from apartnership of the academic, government, andcommercial sectors, clearly are great.

2 Contributors to This Report

UNAVCO convened a two-day workshop inSeptember, 1993, at which the participantscrafted the framework for this report. Otherindividuals subsequently contributed substan-tively to the report. The work of all thecontributors, who are named below, is verymuch appreciated.

John Beavan, Lamont-Doherty EarthObservatory

Robin Bell, Lamont-Doherty EarthObservatory

Michael Bevis, University of Hawaii

Roger Bilham, University of Colorado

Yehuda Bock, Scripps Institution ofOceanography

George Born, University of Colorado

George Davis, University of Arizona

Roy Dokka, Louisiana State University

Michael Exner, University Corporation forAtmospheric Research

Thomas Herring, Massachusetts Institute ofTechnology

Michael Jackson, University of Colorado

Myr on McCallum, UNAVCO

Mar cia McNutt, Massachusetts Institute ofTechnology

William Melbourne , Jet PropulsionLaboratory

Charles Meertens, UNAVCO and Universityof Utah

Richard O’Connell, Harvard University

John Orcutt, Scripps Institution ofOceanography

Robert Reilinger, Massachusetts Institute ofTechnology

Christian Rocken, UNAVCO

Leigh Royden, Massachusetts Institute ofTechnology

John Rundle, University of Colorado

Robert Schutz, University of Texas

David Simpson, Incorporated ResearchInstitutions for Seismology

Robert Smith, University of Utah

Wayne Thatcher, U.S. Geological Survey

Randolph Ware, UNAVCO

3 UNAVCO

The University Navstar Consortium(UNAVCO) provides information and

10

Figure 8. Phase delay versus altitude in the tropics for a model atmosphere, for current and dou-bled CO2 (Yuan et al., 1993).

scientific infrastructure to investigators makinguse of the GPS satellites for Earth science andrelated research. University investigatorsassisted by UNAVCO are conducting researchthat is important to the nation and its citizens.Topics include earthquake, volcano, sea level,polar ice dynamic, weather forecasting, globalclimate, and natural resource explorationresearch.

UNAVCO’s charter, defined by its SteeringCommittee, follows:

UNAVCO is a national program, governed byuniversities and funded by NSF to provide GPSequipment and technical support to university-

based investigators in the geosciences, and topursue a research program that complementsand enhances the programs of universities. Theaim of UNAVCO is to extend the capabilities ofthe university community, nationally and inter-nationally, to better understand the behavior ofthe Earth and the global environment; and tofoster the transfer of knowledge and technologyfor the betterment of life on Earth.

In pursuit of this charter, UNAVCO plans tocontinue its primary role in supporting researchconducted by NSF-supported Earth scienceinvestigators. During the past 10 years,remarkable progress in the understanding ofimportant Earth science problems has been

11

achieved by this community, with assistancefrom the UNAVCO facility. The facilityprovides investigators with a standardized GPSequipment pool, assistance in planning andexecution of high-accuracy GPS surveyingprojects, technical support, data analysis, anddata archiving.

During the past decade, regional GPS geodesycampaigns of growing size were conducted toestablish epoch and re-survey measurements.Large, coordinated projects demanded highlevels of equipment and facility staff resources.New procedures, including multi-modaloccupation strategies, and rapid surveying, arenow under development. We anticipate thatthese procedures will lead to large increases inscientific productivity, and significant changesin the facility as it works with the community todevelop and support these new procedures.

In addition, the UNAVCO community willcontinue to encourage and support the use ofGPS technology in other geoscience areas. Thefacility plans to provide the GPS-relatedinformation and scientific infrastructure neededby geoscience researchers as they addressnational problems, and to identify and exploitsynergies between the various geosciencedisciplines.

4 Scientific Opportunities

From the beginning, the space geodetic com-munity realized that in principle it is possible toachieve accuracies in GPS positioning severalorders of magnitude better than that specifiedby the system design for civilian use, providedthat significant sources of error could be man-aged. As a result of intensive research by manygroups, accuracies have increased rapidly anddramatically. Applications to complexities ofdeformation in tectonic plate boundary zones,active volcanic regions, whole-Earth dynamics,

and other fields of research have mushroomed.Atmospheric scientists have learned to use theGPS signals to probe the atmosphere throughwhich the signals pass, opening vast potentialfor understanding and forecasting weather, cli-matic trends, and ionospheric fluctuations. GPScan accurately track or navigate any sort ofinstrumented platform: ship or buoy, airplane,snowmobile, or spacecraft. Thus, GPS hasopened up a huge variety of scientific applica-tions encompassing the whole of thegeosciences. This section provides a perspec-tive on these applications, today and in thefuture.

4.1 Deformation of theEarth’s Lithosphere

Plate tectonics has provided a unifying theoryfor understanding the kinematics of global-scale plate motions for large regions of theEarth that behave in a rigid framework. Acentral achievement of the Crustal DynamicsProject was the direct measurement of rates ofplate motion on a global scale using thetechniques of Very Long BaselineInterferometry (VLBI) and Satellite LaserRanging (SLR). Now, GPS has brought about anew era. The very rapid improvements inposition accuracy have made it possible tomeasure baselines on the large scale as well asVLBI and SLR. More importantly, the highdegree of portability and relatively low cost ofGPS receivers means that a far greater numberand density of measurements are possible, sothat active geologic processes can be observeddown to the local scale.

Thus, we are now in a position to directlyobserve how plates deform internally withinnon-rigid tectonic regimes and across regionsof rapid and complex tectonism. Thisinformation is crucial to understanding thedynamics of Earth processes, including the

12

distribution of displacements and finite strain,parameters leading to understanding thestresses responsible for Earth processes.Inferences about lateral and verticalpartitioning of displacements across plateboundaries, as well as intraplate regions ofactive seismicity and Quaternary tectonism, areneeded to determine rates of tectonicdeformation and earthquake occurrence.Understanding the vertical distribution of strainand how it is partitioned between surfacedeformation and that inferred fromobservations such as heat flow and earthquakedata is critical to understanding the role ofcoupling between the mantle and the lowercrust and to how rheology controls tectonicprocesses.

Real-time and periodic monitoring ofearthquake zones and regions of activevolcanism offers new information on the long-and short-term rates of deformationaccompanying these processes. Thisinformation may be the key metric in accurateprediction of earthquakes and volcanoes, abroad societal goal of the Earth sciencecommunity. These problems can for the firsttime be addressed by systematic GPSmeasurements, coupled with realistic three-dimensional models of these active processes,and integrated with other geological andgeodetic information.

4.1.1 Plate Boundary Processes

Local and regional departures from rigid platebehavior occur along interplate boundaries andwithin domains of intraplate deformation.Where plate boundary deformation involvescontinental lithosphere, it is commonlydistributed over complex networks of faultsand folds up to several thousand kilometerswide, with zones of intense deformation alongthe margins of less deformed crustal fragmentsof dimensions from ten to hundreds ofkilometers. This spatial complexity is

demonstrably tied to temporal variations in thepattern of motions. In contrast to global platemotions, the complex history of deformationwithin regions of plate boundary interactionand intraplate deformation is known in mostareas only at the crudest level of detail.

Actively deforming subregions are complexlaboratories that evolve over time and involve amultitude of interrelated physical and chemicalprocesses. Thus, tectonic analysis of activelydeforming subregions requires an integratedapproach involving the use and appreciation ofdiverse data sets. It requires a comprehensiveunderstanding of the nature and distribution ofrock units and structures within the crust andlithosphere, an appreciation of the relationshipof the subregional tectonic framework to theregional and plate tectonic framework, andfinally an understanding of the geologicalhistory.

Because it offers the unparalleled means tomeasure the subregional contemporary strainand rates of deformation as it occurs, GPS canplay a major role in the study of contemporarydeformation. These include: (1) establishingthe recent and present-day displacement fieldby tracking the changes of position of groundpoints through time, (2) establishing rates ofcontemporary vertical and lateral deformation,(3) interpreting how the displacements arepartitioned with respect to slip of surface faults,rigid-body rotation, permanent strains, andvolumetric strains, (4) interpreting themechanisms by which the regional deformationis accomplished, (5) tracking how thedeformation accrues through time, (6) defininghow surface deformation relates to the overallcrustal structure, and (7) assessing regions ofseismicversus aseismic (dormant) regions ofexpected earthquakes. Thus, the significantopportunity is to define comprehensively thedisplacement field(s) through time, includingvertical changes, tilts, translations, rotations,and wholesale strains.

13

Ultimately the “paths” of displacements,viewed collectively, will permit investigators todistinguish among models of crustal/lithospheric deformation. Understanding plateboundary deformation is particularly importantbecause these zones host most of the world’scatastrophic earthquakes and volcanoes(Alpine/Himalayan and Pacific Rim).Moreover, the boundaries of plates, if properlyunderstood, are the only locations that canprovide us with information on crust-mantleinteraction and also provide some of the bestinsights into lithospheric rheology.

Problems in plate boundary deformation needto be addressed simultaneously at severaldifferent levels. First, one needs to map theactive strain field within the global plateboundary system, by use of measurements onbaselines of perhaps 100 km. Second, oneneeds to understand how strain is partitioned, inspace and time, along zones of high strain thatseparate less deformed crustal blocks withinplate boundaries, and how these high strainzones are expressed geologically. The latterneed not, and probably should not, be carriedout on a global scale, but may be undertaken inselected regions that are logistically accessibleand geologically well expressed. Someexamples of outstanding problems in each ofthese areas might be as follows:

1. Where are the regions undergoing rapid rates oflithospheric strain adjacent to and within the rela-tively rigid plates? What is the width of these zonesof distributed deformation, and how is deformationpartitioned within them?

2. How is slip partitioned between multiple faultswithin a plate boundary? How do fault zonesdevelop and evolve through time? How are faultzones localized within the crust (for example by pre-existing crustal structures)? What conditions lead tothe abandonment of existing fault systems and thedevelopment of new ones?

3. How is the deformation observed in the upper layersof the Earth expressed at deeper levels in flow of thelower crust and upper mantle? How are vertical axisrotations accommodated at depth?

4. How do strain measurements from geodetic obser-vations compare with long-term geologic estimatesof deformation? How much plate boundary slipoccurs aseismically? How much deformation is notaccommodated by slip on faults? How do geodeti-cally determined rates compare to rates ofdeformation deduced from summation of momentsof contemporary seismicity?

The contribution of GPS to each of these sets ofquestions is obvious. However, GPS canprovide only a snapshot of 2-D motions at theEarth’s surface. Our ultimate goal is tounderstand how the current deformation at theEarth’s surface is the expression of 3-Ddynamic systems that operate over time scalesof perhaps tens of millions of years.

Thus the greatest long-term challenge instudies of regional plate boundary deformationis the successful merging of precise geodeticdata with other geophysical data and carefulgeologic mapping to provide the most completedescription of how these complex deformationsystems reflect motions occurring at depthwithin the lower crust and mantle, and howthey have evolved through time.

Ultimately, we must seek to understand thedynamic processes that drive surfacedeformation within zones of plate boundarydeformation. Thus GPS will provide a morecomplete foundation to address even morefundamental questions such as:

1. What are the driving mechanisms for deformationwithin a plate boundary, such as boundary condi-tions imposed by the relative velocities of adjacentplates, potential energy from excess topography andsubsurface loads, contributions from local dynamicsystems within a plate boundary (e.g. localized sub-duction boundaries, back-arc systems), andinfluence of processes in the asthenosphere?

2. How are motions in the upper crust coupled tomotions in the uppermost mantle? Over what time-scale and length-scale are these motions coupled?How does this reflect the rheology of thelithosphere?

14

Examples of tectonic processes are discussedbelow.

4.1.1.1 Mantle Dynamics and Tibet

At present, techniques for measuring flowwithin the mantle of a regional scale are notavailable, but recent work that combinescrustal kinematics, gravity data, velocitystructure, and theoretical models for fluidconvection show promise for the future. Theimplications for our understanding of mantledynamics are enormous, as can be illustrated byconsidering some basic questions about thedeformation history of Tibet (Figure 9).

A number of studies suggest that deformationwithin Tibet involves movement of crustalfragments, with dimensions of tens to hundredsof kilometers, in directions oblique or evenorthogonal to the overall direction ofconvergence between India and Asia. Datafrom northeastern Tibet show that, on a lengthscale of tens of kilometers, major deformationhas switched from one system of faults toanother and from one type of deformation toanother over time intervals of less than amillion years. If mantle flow beneath Tibet canbe linked spatially to the active movements ofsizable crustal fragments, then the spatialvariability of mantle flow must correspond tothe motions of individual crustal fragments(which can be measured using space-positioning techniques), whereas the temporalvariability of mantle flow must occur over thesame time scale as changes in the direction andrate of movements within the crust (which canbe dated using traditional geologicaltechniques and remote sensing data).Alternatively, if mantle flow can be shown to berelated to crustal motion only at the scale of theentire plate boundary, then one might suspectthat the temporal variability of flow in themantle corresponds to that of the entire plateboundary zone.

4.1.1.2 The Pacific-North AmericanTransform Boundary

The plate tectonics paradigm has providedEarth scientists with the conceptual frameworkwith which to understand the kinematic historyof and strain rates along plate boundaries suchas the Pacific-North American transform.Instead of a single break, the transform isseveral hundred kilometers wide and composedof hundreds of fault-bounded slivers that havetranslated, rotated, and deformed over time.Plate circuit reconstructions provide aquantitative upper limit on integrated, long-term shear across this broad transform zone,but provide no details on where and how shearis partitioned across this complex boundary.For example, in southern California theboundary consists of the San Andreas faultsystem and the newly recognized easternCalifornia shear zone (ECSZ) (Figure 10).

Comparison of the integrated net slip along theECSZ with motion values across the entiretransform boundary indicate that between 9%and 23% of the total relative plate motion hasoccurred and continues to occur along theECSZ since its inception in late Miocene time.In this setting, GPS offers the unparalleledopportunity to track the lateral motions of themyriad of fault-bounded blocks that comprisethe plate boundary. This powerful tool can alsobe used to examine the build-up of strain priorto and following major earthquakes, thusproviding critical data relating to the formationof earthquakes.

4.1.1.3 Complexity of ContinentalConvergence in the Eastern Mediterranean

The Mediterranean region is one of the bestdocumented examples of the spatialcomplexity found in zones of plate boundarydeformation (Figure 11). At the largest scalethis region comprises a zone of continentalconvergence and collision between the slowly

15

Figure 9. Schematic map of Cenozoic tectonics in eastern Asia (from Tapponnier et al., 1982).Heavy lines are major faults, thin lines are less important faults. Open barbs indicate subduction,solid barbs indicate intra-continental thrust faults.

16

moving European and African continents. Atevery smaller scale, tectonic systems withinthis broad region exhibit convergent, divergent

Figure 10. The Pacific-North American trans-form boundary in the western U.S., highlight-ing the location of the eastern California shearzone (modified from Dokka and Travis, 1990).

and strike-slip motions in a large number ofdifferent directions.

It is not uncommon for very different types ofcrustal motion and deformation to existadjacent to one another, such as occurs in theAegean-Hellenic subduction system in theEastern Mediterranean where an area of activeback-arc extension beneath the Aegean Seaexists adjacent to a zone of shortening andconvergence along the Hellenic trench.Geological constraints on the timing andmagnitude of extension and thrusting show thatthere is a close connection between thrustingand extension in space and time, but thedetailed kinematic and dynamic relationshipsbetween thrusting, extension, and subductionremains poorly understood phenomena.

Figure 11. Late Cenozoic examples of thrust belts associated with coeval back arc extensionalsystems in the Mediterranean region (from Royden, 1988).

17

In the complex zone of interaction between theEurasian, Arabian, and African plates, there iscontinent-continent collision in eastern Turkeyand subduction of the leading edge of theAfrican plate beneath Eurasia along theHellenic and possibly the Cyprus arcs. GPSobservations indicate that western, central, andeast-central Turkey is de-coupled from theEurasian plate and is moving as a more or lesscoherent unit about an Euler pole located northof the Sinai peninsula (Figure 12). Other space-based measurements of crustal motion inGreece and along the Hellenic arc suggest thatthis coherent motion includes southern Greeceand the south-central Aegean Sea.

4.1.1.4 Southeast Asia and IndonesiaTectonics

In southeast Asia, four major tectonic plates(Australia, Eurasia, Pacific, and Philippine Sea)interact with each other, mostly alongsubduction zones (Figure 13). Their present-day kinematics is derived from inversion ofearthquake slip vectors along their boundariesas well as transform fault azimuths and oceanfloor magnetic anomalies. Within the kinematicframe defined by these four major plates, abroad zone of active crustal deformation coverscentral and southeast Asia. It extends from theBaikal rift in the north to the Indonesiansubduction arc in the south where most of thestress induced by plate motions is released.

The present-day deformation pattern insoutheast Asia combines the effects of the Java,Sumatra, Mariana, and Philippinessubductions, as well as the effects of the India-Eurasia (Himalaya-Tibet), Australia-Asia(Banda arc), and Australia-Pacific (NewGuinea) collisions. It has been imaged bygeological and seismological studies but nodirect measurements of the velocity field overthe whole area has been proposed so far. GPS isthe only existing tool that offers the opportunityto directly measure a regional velocity field for

southeast Asia. This result is a critical basis fortesting current geophysical models at differentspatial and temporal scales, from platekinematics to intracontinental deformation(Central Asia), plate boundary zonedeformation (Indonesian arc), and co-seismicdeformation.

In particular, the Indonesian archipelago hasmany active tectonic processes that result inphenomena such as mountain building,continental accretion, destruction of continents,emplacement of ophiolites, development ofsubduction zones, arc polarity reversal, andbasin closure. Sumatra has been cited as aclassic example of oblique plate convergence.

Eastern Indonesia is a region of broaddeformation where the continental margins ofAustralia and Southeast Asia and the oceanicplates of the Pacific and Philippine Sea arecoming together at rates of 80 to 110 mm/yr.Plate tectonic concepts do not explain theevolution of Indonesia because deformation isnot confined to the edges of well-defined, rigidplates. Eastern Indonesia is at the stage of

Figure 12. Velocities for GPS and SLR sites ina Europe-fixed reference frame, and their 95%confidence ellipses (Oral, 1994). Note theabrupt drop in the magnitude of velocities forsites north of the North Anatolian Fault andthe change in orientation in northeastern Tur-key

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

ANATOLIAN PLATE

EURASIAN PLATE

ARABIAN PLATE

AFRICAN PLATE

Marmara Sea

Black Sea

Mediterranean Sea

Aegean Sea

North Anatolian Fault

East Anatolian F

Dea

d S

ea F

BitlisZagros

Thrust and Fold Belt Hellenic Arc

Greater Caucasus

Cyprean Arc

Lesser Caucasus

Florence Rise

Lake Van

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

500 km

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

500 km

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20 mm/yr20 mm/yr

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

20˚ 24˚ 28˚ 32˚ 36˚ 40˚ 44˚ 48˚

30˚

32˚

34˚

36˚

38˚

40˚

42˚

44˚

18

Figure 13. Plate boundaries, seismicity, and relative velocities determined by GPS for the Indone-sian arc and adjacent regions (Calais, 1994).

95˚

95˚

100˚

100˚

105˚

105˚

110˚

110˚

115˚

115˚

120˚

120˚

125˚

125˚

130˚

130˚

135˚

135˚

140˚

140˚

-15˚ -15˚

-10˚ -10˚

-5˚ -5˚

0˚ 0˚

5˚ 5˚

10˚ 10˚

SENT

AUKE

WAMETIMI

BIAKSORO

KALA

KUPA

MAU2

UJUN

DAKY

BAKO

XMAS

COCO

PADA

BINAD957

COCO

XMAS

AUKE

WAMETIMI

SENT

SOROBIAK

DAKY

KUPA

KALAMAU2

UJUN

BAKO

PADA

BINAD957

GPS

NUVEL-1

100 mm/yr

INDONESIAN ARC 89-90-91-92-93

1 sigma ellipses - vectors refer to the ITRF92

assembly of crustal elements, both continentaland oceanic, that will eventually form acomplex mountain belt. The strain field withinsuch deforming regions is a first-orderobservation that can constrain ideas about thebehavior of the crust and upper mantle inmountain building.

Of particular interest is whether the brittle,shallow portion of the lithosphere, which isbroken into fault-bounded blocks and producesmost of the directly observable effects oftectonics, offers significant resistance to platemotions, or if it merely responds to horizontalshear stresses from a strong, ductile uppermantle. If the latter case holds, then the large-scale deformation of the lithosphere can be

represented by a continuum, and the surfacefaulting has little to do with plate dynamicsexcept in an average sense. An importantgeological consequence of this view is that thejuxtaposition of rock types in mountain beltsmay be as much a consequence of small localmechanical and geometric irregularities as it isa consequence of plate motions.

4.1.1.5 The Interior Western U.S. and theTransition to the Stable Plate Interior

Much interest has been focused on themechanism and mode of Late Cenozoiclithospheric extension of the western U.S.Cordillera encompassing the Basin and Rangeprovince, the world’s largest region of

19

intraplate continental extension (Figure 14).Tectonic process of the Basin-Range aregenerally thought to be locally driven,including crust-mantle coupled volcanism;“soft” shear coupling between the stableinterior, the San Andreas transform fault, andthe Cascadia subduction zone; localizedtopographic load-induced stresses; andcoupling of quasi-plastic flow in the lower crustwith surface deformation. Active tectonism ofthe Basin-Range is directly manifest byextensive earthquakes and large Quaternaryfault scarps which occur primarily along itsmargins. Historically eight large earthquakes,dominated by normal- to oblique-slip, reflectongoing tectonism that is measurable withGPS.

Measurements of contemporary tectonism inthe Basin-Range are limited to a few VLBI/SLR ~1000 km baselines which reveal ~1.0cm/yr of east-west extension. These rates agreeremarkably well with displacement ratesdeduced from summations of seismic momentsof historic earthquakes of ~1 cm/yr as well asrates deduced from models of intraplatedeformation tied to a fixed mantle framework,and suggest that the brittle mode ofdeformation accompanying earthquakesaccounts for ongoing tectonism. However, littleis known of how deformation is distributedacross this wide region of extension or of themechanisms responsible for this unusual styleof intraplate deformation.

On a regional scale, VLBI measurements froma single baseline across the Basin-Range transi-tion to the Stable Plate interior compared to thestable U.S. interior give a clue to the lateral dis-tribution of intraplate strain release across thetransition between the Basin-Range and stableinterior of ~5 mm/yr east-west extension. Thisrate is remarkably similar to the first GPS mea-surements against historic trilateration/triangulation data, as well as an annual resur-vey of a GPS network across the Wasatch fault,

Utah, which revealed 5 to 9 mm/yr northeastextension. This rate notably exceeds theinferred long term ~1 mm/yr rate deduced fromHolocene slip determined from paleoseismicmeasurements. One interpretation of these datais that the now dormant Wasatch fault is storingstrain energy at high rate preceding a futureearthquake, which has important societal rami-fications. However, without further analyses,additional GPS observations, and comparisonwith older geodetic data these high rates cannot be verified.

These observations point out an importantproblem that can be addressed by GPS: How do

Figure 14. Horizontal strain and displacementfrom 1993 to 1992 GPS surveys on the centralWasatch fault (Smith, 1994). Error ellipse cor-responds to one standard error.

20

rates of contemporary deformation comparewith Holocene geologically deduced rates, anddo deviations of these contemporary rates indi-cate short-term processes such as rapid strainaccumulation preceding large earthquakes?This question has major implications for theoccurrence of earthquakes on known or hiddenfaults. Low-angle listric and basal detachmentfaults were first observed in the Basin andRange based upon geologic mapping and seis-mic reflection data. These observations suggestan important paradox in tectonics, namely theoccurrence low-angle normal faults at rela-tively shallow crustal depths, despite the lackof unequivocal evidence for low-angle faultingfrom earthquake data. This an important prob-lem which can be addressed using an integratedapproach of fault studies, earthquake monitor-ing, and high-precision GPS measurements.

The widespread seismicity, well-preservedQuaternary faults, and active extension of theBasin and Range thus provide excellentopportunities to evaluate rates of deformationusing GPS and such processes questions as“characteristic” earthquakes, space-timeclustering of earthquakes and stress couplingbetween the lower and upper crust, andtopographically driven deformation. It is aparticularly well-suited region for studyingactive continental extensional mechanismswith GPS.

4.1.2 Volcanic Processes

Geologic and historical records show that over500 of the world’s volcanoes can be consideredto be “active” or capable of producing aneruption. The vast majority of these volcanoesare associated with subduction-zone plateboundaries. Although constituting only about15% of the annual global magma production,subduction-zone volcanoes can be explosiveand have dramatic effects on humanpopulations. In contrast, most magma isgenerated at plate spreading centers at remote

mid-ocean ridges. The remainder of volcanicactivity is located along mantle hotspot tracks.

Current geologic and geophysical volcanoresearch investigations are examining theorigins of magma, the physical and chemicalproperties of magmas, hydrothermal fluids andgasses, the structure of volcanoes and volcanoplumbing, the mechanism of magma transport,and the relationships of volcanoes to tectonicboundaries and earthquakes. GPSmeasurements of crustal deformation are at theforefront of new techniques for studying thesevolcanic processes. A better understanding ofthese processes is contributing to improvedvolcano hazard assessments and volcanoeruption predictions.

Surface deformation measurements offer onlythe means of constraining the volume andgeometry of subsurface crustal magma sources.Traditionally done with trilateration andleveling, precision GPS surveys of groundmotion now offer unprecedented simultaneousthree-dimensional views of volcanodeformation on scales of hundreds of meters totens of kilometers, and can operate in remoteregions not requiring line of sight toobservation points. When tied to regional GPSsurvey networks, a direct measure of therelationship of the volcanic and tectonicframework can also be obtained. Examples ofthis application are studies of rifting andvolcanism in Iceland, investigations ofdeformation of the Yellowstone caldera and itsrelationship to a large continental hotspot, andextensive USGS efforts to monitor such activefeatures as Kilauea volcano, Hawaii, and somethreatening Alaskan and Cascades volcanoes.

In general, volcano deformation can beseparated into two types: that due to roughlyequi-dimensional magma chambers and thatdue to rifting or dike intrusion. From the studyof some basaltic shield volcanoes, notably inHawaii and Iceland, we know that as magma

21

rises from the mantle it collects in shallowmagma chambers. As the magma chamber fills,the Earth’s surface uplifts and stretches. Theseinflationary periods are interrupted byeruptions or intrusions into flanking rifts,which cause withdrawal of magma from thechamber and attendant subsidence andcontraction.

The earliest models of the inflation-deflationcycle used point centers of dilatation (the socalled Mogi source). The surface displace-ments produced by a Mogi source are nearlyindistinguishable from that produced by aspherical magma chamber, and similar to thatgenerated by other equi-dimensional models. Itturns out that given only vertical displacementdata it is nearly impossible to determine theshape of the magma chamber. Horizontal andvertical data together provide much tighter con-straints on source models. One of the benefitsof GPS over traditional methods (leveling andtrilateration) is that GPS determines all threecomponents of station displacement. Much ofthe same arguments hold true of rifting. Riftingcan occur episodically, accompanied by exten-sive ground cracking and swarms of shallowearthquakes that migrate downrift at velocitiesof a few tens of centimeters per second. Exam-ples of volcanic processes are discussed in thefollowing sections.

Interpretations of the crustal deformation dataobtained in volcanic studies, such as the studiespresented in the following sections, are mostmeaningful when combined with othergeophysical and geologic data. For example,seismicity patterns can be used to locate andidentify potential faults related to magmapathways, whereas GPS measures bothaseismic and seismic deformation associatedwith magmatic activity. Combined with precisegravity measurements, GPS and verticalmotion observations can also be used todistinguish between changes in elevation due totectonic strain and changes due to magma flux.

Hazards from volcanic events range from localphreatic (hot water and steam) blasts and lavaflows to large catastrophic ash flows and airfalls. Volcanic eruptions also producesecondary flooding and landslides. The long-term effects of volcanic eruptions are wellknown to effect globe-scale weather changesresulting from ash particulates deposited in theatmosphere. The primary goal of volcanic riskresearch, however, is to specify the timing andlocations of potential volcanic eruptions. Thelocation and long-term history of volcanoes aregenerally known from geologic and historicalrecords. However, it is difficult to determinethe short-term timing, style, and volume of aneruption at times scales of a few minutes tohours.

The success of detailed earthquake monitoringin predicting the recent eruptions of Mt.Pinatubo, Philippines, and Mt. Redoubt,Alaska, saved thousands of lives and points outthe use of geophysical tools to forecast eminenteruptions. This class of research will benefit byparallel efforts in deformation monitoring forinflation and deflation of the surface which areassociated with magma or hydrothermal fluidmovements, to constrain the geometry andvolume of magma chambers, plumbing ofconduits, dikes and sills and associatedfaulting, and to assess incipient earthquaketriggering.

Repeated GPS surveys taken over months andyears give insight into the long-term deforma-tion signal associated with volcanic processes.High precision static GPS surveys can be usedwhere deformation rates may only be mm/year.In many cases however, the sequence of erup-tive activity can take only hours to days andrates accompanying deformation are cm/hour.Deformation can be continuously monitoredwith GPS receivers deployed at remote sitesand with data transmitted to a central recordingsite where processing can be done in near real-time kinematic differential modes.

22

Orbit information broadcast directly from theGPS satellites are sufficient for analyzing shortbaselines (~10 km), but where receivers areseparated by large distances (>10 km), moreprecise satellite orbits are required for real-timeprocessing. The U.S. Geological Survey iscurrently operating a small continuous trackingGPS network on St. Augustine volcano inAlaska and will install a network in Hawaii.Monitoring is more effective where augmentedwith repeated static surveys which can coveradditional sites over a larger area and whichcan be used to determine spatial coherence ofthe deformation signal over longer timeintervals.

New techniques are being developed to makeremote sensing measurements of deformationusing aircraft. For example, in an experiment atthe Long Valley caldera in California, repeatedaerial laser altimeter measurements are beingmade which will be used to produce an imageof ground motion. Kinematic GPS is used forprecise, centimeter-level positioning of theaircraft. Radar interferometry can detect mapsurface displacements at the centimeter level ofhorizontal accuracy, but requires GPS tocontrol the coordinates of satellite images.Similarly, radar interferometry from space islikely to have much potential in observingdeformations associated with volcanic activity,but likewise will require GPS calibration on theground. These remote sensing methods notonly have the potential to cover large andpossibly remote areas, but also do not requireobservers to be physically located on apotentially hazardous volcano.

4.1.2.1 Kilauea

Geodetic and seismic data have been used tomodel the geometry and depth of rift zoneintrusions. Kilauea is a case in point. OnKilauea, these models indicate that the dikesare steeply dipping bladelike structures a fewmeters thick, typically at depths of between 2

and 4 km. Less is known about the deeper partof the rift system and the long-termdeformation accompanying intrusion withinplate-boundary rift zones.

Deformation and seismicity data demonstratethat magmatic activity and slip on Kilauea’sfault systems, including the subhorizontalfaults that generated the magnitude 7.2 1975Kalapana earthquake, are coupled. Yet thesubsurface geometry of these faults and thedeep structure of Kilauea’s rift zones are poorlyunderstood. Figure 15 shows the averagestation velocities for the three-year period withrespect to a point on the northwestern side ofthe island of Hawaii. Stations on the centralsouth flank of Kilauea moved SSE, away fromthe rift zones, at rates of up to 10 cm/yr. Incontrast, stations north of the rift zone, only 6or 7 km from the rift, are not moving withinmeasurement errors.

Figure 15. Displacements near the Kilauea riftzone determined by GPS compared with modeldisplacements (Owen et al., 1994).

155˚ 30'W

155˚ 30'W

155˚ 20'W

155˚ 20'W

155˚ 10'W

155˚ 10'W

155˚ 00'W

155˚ 00'W

154˚ 50'W

154˚ 50'W

19˚ 00'N 19˚ 00'N

19˚ 10'N 19˚ 10'N

19˚ 20'N 19˚ 20'N

19˚ 30'N 19˚ 30'N

19˚ 40'N 19˚ 40'N

.

20 km 100 mm/yr

Observed

Predicted

23

Velocities also decrease dramatically to zerotoward the more distal ends of the east andsouthwest rift zones. There seems to be a zonethat is actively displacing seaward. Lateral gra-dients in velocity (strain-rates) are very large.In fact, these strain-rates are at least an order ofmagnitude greater than those across the SanAndreas fault! The rapidly displacing sectionof the south flank coincides roughly with thesource region of the 1975 earthquake, and theseismically active portion of the south flank.

The steep gradients in the surface velocity fieldshow intriguing relationships to structuralelements of the volcano as expressed in thetopography and bathymetry. The rapidlydisplacing zone corresponds quite well to theextent of the Hilina fault system. Thebathymetry off the south flank contains featuresindicative of gravitationally driven slumping.A series of topographic basins offshore of therapidly displacing south flank block must begeologically active, since the high sedimentflux off Kilauea would quickly fill anydepression.

The western extent of the rapidly displacingzone may correspond to a prominent linearfeature in the bathymetry, once referred to asthe Papau Seamount, now interpreted by someto be the boundary of the Hilina mega-landslidesystem. However, there is no evidence forlateral “tear” faults at either the eastern orwestern edges of the rapidly deforming zones.

4.1.2.2 The Yellowstone Caldera

A GPS survey in 1987 established a densenetwork of 60 stations planned to measurecrustal deformation of the active Yellowstonecaldera and adjacent Hebgen Lake and Tetonnormal fault zones. Comparisons of data froma 1993 survey with GPS surveys in 1989 and1991 show significant crustal deformationwhich is characterized by subsidence andcontraction over the 60 km-long Yellowstone

caldera superimposed upon a regional tectonicstrain field. The principle components of theregional strain rate tensor, estimated using amethod of simultaneous reduction afterexcluding data for stations outside the caldera,are 1.1±0.2 x 10-7/yr of northwest contractionand 0.6±0.2 x 10-7/yr of northeast extension.After removing the regional strain componentsfrom the observed GPS field, the residuals ofthe inner-caldera sites revealed up to 20±7 mmof horizontal displacement, directed radiallytoward the center of the caldera, and up to70±20 mm of caldera-wide subsidence. Thiscorresponds to an average subsidence rate of 17mm/yr.

Forward models made with a 3-D boundaryelement technique can explain the calderadeformation with a shallow NE-elongateshaped, sill-like body beneath the caldera atdepths of 3 km to 6 km (Figure 16) which lostvolume at a rate of 0.02 km3/yr, similar to ratesindependently estimated for magmatic fluidrelease. This deflationary model was confirmedwith a 3-D volumetric source model derived bysimultaneous inversion of the horizontal andthe vertical GPS displacement data. Theinverse model indicates that the largestdeflation is quite shallow, between the surfaceand 3 km depth, in the depth range of extensivehydrothermal fluids and possible magmaticfluids in Yellowstone.

It is evident from the regional GPSobservations that the extent of the currentYellowstone GPS network is insufficient tomeasure crustal deformation associated withthe long-wavelength signature of theYellowstone hotspot. For example, topographicand geoid data suggest that the influence of thehotspot extends up to 500 km from the caldera.To examine these large-scale effects of thehotspot signature, the new Yellowstone GPSsurveys are designed to reobserve much of theexisting network and to make ties to the high-

24

accuracy GPS networks of surrounding stateswhich have approximately 100 km spacing.

4.2 The Earthquake Cycle

It has recently been estimated that within 50years more thana third of the world’s popula-tion will live in seismically and volcanicallyactive zones. As a result, the international com-munity is attempting to devise means to reduceand mitigate the occurrence of these devastat-ing natural hazards. The International Councilof Scientific Unions, together with UNESCO,the World Bank and a number of U.S. agenciesincluding—NSF, the DOE, the USGS andNOAA—have endorsed the 1990s as the Inter-national Decade of Natural Disaster Reduction(IDNDR). These organizations are planning a

ï�°ï|̤ï} �ï|�Ü

-120 -1

00-8

0

-80

-60

-60

-40

-40

-20

0

contour interval 20 mm 20 mm 20 km

111o

45o

110o

44o

Contour Interval 20 mm Horizontal displacements with 95% confidence ellipses Scale: 20 mm 20 km

Figure 16. Three-dimensional deformation ofthe Yellowstone caldera showing subsidence,contraction, and regional strain (Meertens andSmith, 1994).

variety of scientific programs in an attempt tounderstand how such natural disasters might bepredicted, or at least mitigated.

Because both earthquakes and volcanoes are sostrongly associated with significant crustaldeformation, it is expected that GPSmeasurements will play a central role in manyof these scientific projects. Virtually all of theseprojects will include GPS field observation,and in addition will use numerical models tointerpret and understand the measurements;thus, modeling occupies a central focus in theseintegrated studies of earthquakes.

It is generally recognized that observations leadto the definition of distinct intervals during themore-or-less repetitive earthquake cycle,referred to as “coseismic,” “postseismic,” and“interseismic.” The coseismic period isidentified as the short (approximately minutes)interval associated with the earthquake itself,when the majority of the slip occurs and elasticand inertial effects dominate.

The postseismic interval is a somewhat looselydefined period comprising days to perhaps adecade in which further non-inertial slip(aseismic slip or creep) takes place, togetherwith the surface deformation related to inelasticstress relaxation within the upper part of theasthenosphere. Both of these effects yieldsurface deformation that decays approximatelyexponentially. The interseismic intervalcomprises most of the long period betweenearthquakes, during which plate motion reloadsthe plate boundary, and which can berepresented to a reasonable approximation bysimple rigid plate motion. There are as yet noindisputable geodetic indications of a“preseismic” process interval, in whichchanges of an as-yet unobserved type mightoccur prior to the earthquake.

Resolution requirements for the variousintervals depend to a certain extent on the

25

nature of what space-time frame features of theevent are of interest. For example, if details ofthe coseismic slip distribution are needed, thenaccuracy of better than 100 mm over timeperiods of days and spacial scales 1–100 km orlarger are required. On the other hand,delineation of the physics of postseismicdeformation requires accuracy of 10–100 mmover time scales of hours and spacial scales ofagain 1–100 km. The interseismic phase isobviously characterized by the same time-space scales that characterize plate tectonicmotion. The only factor known aboutrequirements for preseismic deformation is thatthe space-time accuracy and resolutioncurrently available is not sufficient to resolveindisputable precursors.

In order to understand the physical meaning ofthe processes operating during each interval,models must be used to interpret the data. Thereare two broad categories of earthquake models:“kinematic” and “dynamic.” Althoughrecognizing the complexities of crustaldeformation, most models have attacked theproblem of strain accumulation and release ona single prescribed fault.

Models that prescribe the slip on the fault andthen calculate the resulting surface deformationby numerically evaluating the appropriateGreen’s function are called kinematic models.Using these Green’s functions, calculationshave been carried out in which a series ofearthquake slips are specified on fault planes intime and space. Together with relative pathvelocities, which are known from rigid platemotion models, the surface deformation canthen be calculated at any time or location on thesurface on the model Earth. Self-consistentkinematic models of earthquake cycle can bedeveloped in which a history of earthquakes isprescribed. In these models, both the slipdistribution along the fault as well as the timeslip are given, and the resulting surfacedeformation is calculated. Models of this kind

have been constructed for both purely elasticand layered viscoelastic-gravitational media.

The other kind of models are those in which theslip itself during a series of earthquakes is cal-culated as a consequence of a given set ofdynamical governing equations, boundary con-ditions, and frictional properties on the fault.Once the earthquake history on the model faulthas been computed, together with the slip ineach event, the entire history of time-dependentdeformation of the free surface can be calcu-lated. Using this approach, earthquakeresearchers have begun to focus on the dynam-ics of interacting fault systems, and a series offascinating new possibilities have emerged thatdefine many new directions and lines ofresearch.

For example, these models, which are funda-mentally nonlinear in character due to thenonlinear frictional forces acting on the faultsurface, have been shown to produce a fullrange of rich, complex dynamical behaviorincluding limit cycles as well as chaotic behav-ior. Thus, at the same time that our ability toaccurately observe crustal deformation hasblossomed with the advent of GPS technology,dramatic improvements have developed ourability to realistically simulate the physics of anearthquake source process, and as a conse-quence predict the surface deformation that canbe measured by space geodetic observations. Itis therefore reasonable to expect that it mayshortly be possible to critically test models forcompeting ideas of the physics of earthquakenucleation using space geodetic data.

A substantial and growing database ofcentimeter-level accuracy geodetic observa-tions of deformation along active fault zonesnow exists. Important regions covered includethose parts of California through which the SanAndreas fault passes; the tectonically activeOwens and Chalfant valleys and Long Valleycaldera; the Wasatch range; Yellowstone; the

26

Delani, Totschunda, and Fairweather faults; aswell as several islands sensitive to deformationalong the Aleutian trench in Alaska; many partsof Japan; the Alpine fault in New Zealand; theAnatolian fault in Turkey; and the Pozzuloi-Campi Flegrei caldera in Italy, to name just asubset. A critical issue is how to make best useof these data. Many of the previous efforts havebeen devoted to verification of the predictionsof regional and global kinematic plate motionmodels. A most challenging problem is that ofusing precise geodetic data to search for pre-cursors to major earthquakes on these faults.

There are only three possible alternatives thatcan explain the lack of observations ofprecursors to date. The first is that there exist noprecursors to observe at the precision that wehave available to us with our currentmeasurement techniques. If this is the case, onewould like to prove it theoretically so that nomore money is spent on fruitless searches. Thesecond is that the space-time coverage of theeffected region is inadequate. If this is the case,one would like to determine, using theory, whatconstitutes adequate coverage. The thirdalternative is that the precursory signals arepresently there in the data, but we do not knowhow to recognize them. This alternative impliesthat theoretical models for the earthquakenucleation processes must be constructed toprovide predictions of the correspondinggeodetic signature. Observational techniquesmust then be devised to optimize chances fordetection of these signals.

4.2.1 Loma Prieta

An opportunity to model the phases of theearthquake cycle was provided by the magni-tude 7.1 Loma Prieta earthquake of October 17,1989. This event was the largest earthquake inCalifornia in 40 years, and the largest earth-quake in the San Francisco Bay Area since1906. Following the earthquake, a GPS net-work in the Santa Cruz–Watsonville area that

had been surveyed by the California Depart-ment of Transportation (Caltrans) in the yearprior to the earthquake was remeasured. Whilethe Caltrans measurements were made withsingle frequency receivers, it was found thatday-to-day repeatability of the baseline vectorswas 0.5–1.0 cm in the horizontal componentsand 3.0–3.5 cm in the vertical.

The GPS baseline vectors were combined withother data—Electronic Distance Measurements(EDM), Very Long Baseline Interferometry(VLBI), and spirit leveling—to obtain a com-plete picture of the three-dimensionalcoseismic deformation field. An iterative quasi-Newton method was used to find the disloca-tion geometry that best fits all of the availablegeodetic observations. When weighted prop-erly, all of the data can be accounted for by asingle dislocation surface that is consistent withthe aftershock distribution. The strike and dipof the best-fitting dislocation is consistent withmoment tensor solutions.

GPS has been the principal tool used in study-ing postseismic deformation following theLoma Prieta earthquake. Strain diffusion awayfrom the epicentral region is predicted by theo-retical models of viscoelastic relaxation belowthe seismogenic crust, and has been suggestedas a mechanism to explain sequential rupturingalong a fault system. Transient postseismicdeformation has generally been explainedeither by viscous relaxation of a ductile(asthenospheric) layer underlying an elastic(lithospheric) plate, or by downward propaga-tion of aseismic slip along an extension of thefault zone into the lower crust. The horizontaldeformation field in the epicentral region priorto the 1989 earthquake was already well-established on the basis of 20 years of geodolitemeasurements and several years of GPSmeasurements.

Following the 1989 earthquake two arrayscrossing the San Andreas in the epicentral

27

region and northwest of the rupture zone wereestablished and surveyed multiple times.Because of the extensive pre-earthquake moni-toring, it is possible to identify features in thepresent-day deformation field that are attribut-able to the earthquake. The postseismicvelocity field in the epicentral region of theLoma Prieta earthquake (Figure 17) differs sig-nificantly from displacement rates measured inthe two decades preceding the event. The post-earthquake displacement rates along the BlackMountain profile, which crosses the SanAndreas fault 44 km northwest of the LomaPrieta epicenter, do not differ significantly fromthose determined from 20 years of trilaterationmeasurements. However, station velocitiesalong the Loma Prieta profile, which passesthrough the epicentral region, significantlyexceed pre-earthquake rates within 20 km ofthe Loma Prieta rupture.

There is also significant San Andreas normalshortening centered near Loma Prieta. Contraryto expectation, the post-Loma Prieta GPS datacannot be fit with slip on a single fault plane, orby distributed down-dip deformation followingthe earthquake. The data can be fit, however,with slip on two fault planes: strike-slip on afault coincident with the rupture plane, andreverse-slip on a fault within the Foothillsthrust belt more or less coincident with the Ber-rocal fault. Slip within the Foothills thrust beltappears to have been triggered by the stressingduring the Loma Prieta mainshock.

4.3 Deep Earth and WholeEarth Applications

4.3.1 Earth Rotation

Geodetic techniques are able to sense theresponse of the mantle to internally andexternally applied torque. With currentlyavailable systems and a data set which now

spans over a decade, deviations of the Earth’smotion from that predicted by geophysicaltheories with amplitude of 0.3 mm can now bemeasured, for signals that are coherent with themajor tidal forcing signals. Results of thisaccuracy have been used to infer the flatteningof the core-mantle boundary, to bound theanelastic properties of the mantle in the diurnal

20' 10' 122 50' 40' 30' 20'

30'

20'

10'

37 3 cm/yr

20 km

Model velocities

Mt. Oso

Bürgmann et al., 1993 - Figure

Post-seismicvelocities

reduced by pre-seismic rates

Figure 17. Excess postseismic station veloci-ties (secular displacement rates subtractedfrom postseismic rates) assuming the velocityof Mt. Oso was unaffected by the earthquake(Bürgmann et al., 1994). Error ellipses (3sigma) for the sites near Loma Prieta are notshown for clarity. They are the same order ofmagnitude as the error ellipses at other sites.Also shown are the velocities (filled arrows)computed from a dislocation model involvingslip on two faults; strike slip on the Loma Pri-eta fault and thrusting on a reverse fault in theFoothills thrust system. The two shaded rect-angles show the projections of the faults usedto compute the modeled station displacements.The first fault lies in the Loma Prieta after-shock zone, from 6 to 17 km depth, dipping 70°SW, with a strike-slip displacement rate of 0.10m/yr. The second fault slips 0.12 m/yr on areverse fault dipping 47° SW, striking subpar-allel to the San Andreas, from 1.4 to 5.8 kmdepth.

28

frequency band, to detect the influence of thesolid inner core on the rotational dynamics ofthe mantle, and to place upper bounds onconductivity at the base of the mantle and thestrengths of the magnetic field near the core-mantle boundary. The effects of the oceanshave also been detected, and the influence ofthe Earth’s atmosphere are also now becomingevident.

The applications of geodetic systems, includingGPS, cover all time scales from the diurnal andsemidiurnal bands of the major tidal forcesbeing applied to the Earth to decade durationvariations whose study is limited only by theduration of the geodetic data themselves. In theformer of these time scales, the forcingfunctions are well known and the geodeticsystems can measure the response of the Earthto these forces. In the latter of these time scales,the forces themselves are not well understoodand geodetic systems provide informationabout convolution of the forcing function withthe response of the Earth. For very long-termmeasurements, GPS must rely on VLBImeasurements to provide a link between long-term variations in the Earth’s rotation and aninertial reference frame.

For deep Earth studies, geodetic measurementscan provide accurate information about thedynamical interactions of the fluid core andmantle. Much of this information arises from aresonance in the rotation of the Earth due to thedifferential rotation of the core and mantle. Thetorques applied to the mantle by the core aresensitive measures of the flattening of the core-mantle boundary and deformation of the mantlearising from the pressure at the core-mantleboundary as the elliptical core rotatesdifferentially with respect to the mantle. Sincethe frequencies of these torques are in thenearly diurnal band, the measurements of thenutations of the Earth and the associatedsurface deformations provide unique measures

of the dynamical mantle properties in thisfrequency regime.

For example, competing models for the fre-quency dependence of the shear-strain Q of themantle predict anelastic effects on the rotationwhose amplitudes at the surface differ by a fewmillimeters. While these values are small, theyare detectable with current data sets. Similarly,magnetic coupling between the core and mantlecould also produce few millimeter deviationswhich are currently detectable.

The major differences between a geophysicaltheory for the motion of the Earth’s body axisin space and those observed by VLBI areshown in Figure 18. The large annual signature(amplitude 2 mas) is interpreted as being due toa deviation of the core-mantle boundary fromits hydrostatic shape by about 4.5%. The longperiod variations reflect deviations in the 18.6year nutation whose accurate determination iscurrently limited by the duration of high qualityVLBI data.

Not so evident in the figure is a semi-annualsignal (amplitude 0.5 mas) which arises partlyfrom the excess flattening of the core mantleboundary (0.3 mas), partly from ocean tides(0.6 mas), partly from the anelasticity of themantle (0.3 mas), and partly from the deviationof the ellipticity of the whole Earth from thatcomputed assuming hydrostatic equilibriumand geophysical models of the density distribu-tion of the Earth. Although non-hydrostaticmodels of the Earth derived from seismictomography have been available for a numberof years, a rotational dynamic model for theEarth has yet to be constructed with such data.When such a consistent model is determinedthe anomalies due to the ellipticity of the wholeEarth and fluid core should be eliminated.

Because the forcing functions are known forthe nutations, both the in-phase and out-of-phase components of the response of the Earth

29

can be determined, with the latter interpreted asarising from dissipative processes in the Earth.For example, 0.4 mas of the annual signal is outof phase and thought to arise from magneticcoupling between the core and mantle, whileother out of phase components are consistentwith anelasticity of the mantle.

The accuracy of these nutation results, alongwith other rotational components such as diur-nal and semidiurnal variations in the rotationrate, is now so high that deficiencies in theaccepted tidal displacement models also needto be examined, particularly in the light of theeffects of mantle inhomogeneities and anelas-ticity. Although these effects are expected to beless than 1 mm of surface displacement, they

Figure 18. Differences between positions of anaxis attached to the mean mantle of the Earth(body axis) in inertial space and that predictedby the IAU 1980 nutation series and observedwith very long baseline interferometry (VLBI).The positions are expressed as two angles: (a)nutation in longitude and (b) nutation in obliq-uity. One milliarcsecond (mas) is equivalent to30-mm displacement at the surface of theEarth (courtesy Herring, 1993).

are within the realm of detectability with thecurrent systems and data sets.

Global geodetic networks also allow themotion of the rotation axis to be accuratelydetermined. Much of the free nutation, orChandler wobble, has been shown to correlatewith variations in atmospheric massdistributions. However, it is likely that a part ofthe wobble is excited by large earthquakes andrelatively rapid tectonic motions (that havebeen termed “silent earthquakes”). At thepresent, large earthquakes account for only afraction of the relative motions of plates atconvergent zones, and for even less at strikeslip plate boundaries.

The detection of aseismic motions on plateboundaries would add greatly to ourunderstanding of plate motions, and wouldcontribute to understanding the development oflarge scale strain fields associated with largeearthquakes on plate boundaries. Continuedglobal monitoring of Earth rotation shouldallow the detection of changes in the rotationaxis associated with such motions.

The establishment of global GPS networks willpermit more precise determination of therotation of the whole Earth and its relation tothe transfers of mass and momentum betweenvarious parts. More importantly, it will allowthe determination of the spatial structure of thewhole Earth deformation field, and allow thenature of the loading and momentum transfer tobe better determined.

4.4 Interaction of MantleConvection and Tectonics

4.4.1 Vertical Motions

The Earth’s surface adjusts vertically inresponse to temporal changes in the forces

30

applied at the base of the lithosphere, such asmight be caused by changes in temperature,mineral phase, or pattern of convection. Formost processes, the predicted vertical motionsare too slow to be detected by GPSinstrumentation at present, or even with likelyimprovements in the next 10 years, with twopossible exceptions. The first is the possibilitythat broad continental plateaus, such as Tibetand the Altiplano, might experience rapid upliftas the result of delamination of the dense lowerlithosphere that becomes convectively unstableon account of over-thickening of thelithosphere at continental convergence zones.

For example, it has been proposed that 8million years ago the Tibet Plateau abruptlyrose by 1 to 3 km within the span of a millionyears on account of delamination of its cold,thickened lithospheric root. The rise of thisabrupt and massive continental feature had amajor influence on climate in southern Asia bytriggering the onset of the monsoons. At ratesof a few millimeters per year, vertical uplift inresponse to convective destabilization of thelower lithosphere could be measured and usedto understand how the dynamics of the lowerlithosphere affects continental tectonics.

The second example is the uplift of the Earth’ssurface as the lithosphere rides over mantle hotspots. From analysis of depth anomalies in theocean basins, it is clear that the uplift is of theorder of a kilometer or two, and it reaches itsfull height within a few million years.Investigation of the rate and pattern of upliftprovides important information on the thermaland dynamic properties of hot spots and howthey interact with the lower lithosphericboundary layer. Nothing is known about thisprocess for hot spots underlying continentallithosphere, such as Yellowstone (Figure 19)and the many African hot spots, because welack a vertical reference for measuring upliftassociated with the passage of the lithosphereover these hot spots. GPS provides the only

hope of recording this process of hot spot-continental lithosphere interaction, butdetecting this small signal will requireimprovements in the vertical accuracy.

4.4.2 Horizontal Motions

Although it is generally acknowledged thatlithospheric plate motions are driven by deeperseated convective flows, the exact nature of thecoupling is not known. Recent comparisons ofplate motions derived from geologic data overa few million years with geodeticmeasurements over a few years indicate thatthe plate motions are the same over both timescales. Yet we know that the motion at plateboundaries is not continuous, but isaccomplished in large part by episodicearthquakes.

There is also evidence that the seismic (andsubseismic) motions on plate boundaries mayfollow larger scale spatial and temporalpatterns; there may be an alternation ofearthquakes on convergent and transform plateboundaries on a time scale of decades.Deciphering such patterns and processes isclearly important for understanding thedevelopment of stress and strain along plateboundaries, and relating the earthquake cycleto plate motions.

Geodetic networks on the scale of hundreds ofkilometers should be able to measure thedisplacements associated with major slip atplate boundaries, and detect its evolution withtime as it diffuses away from the plateboundary into the interior of a plate, where themotion becomes more uniform. Such data willdirectly bear on the nature of the couplingbetween the plates and the underlying mantle,and will allow the determination of the uppermantle rheology that controls the coupling.

Such measurements will also address anothercentral problem. To first order, lithospheric

31

Figure 19. The Yellowstone hot spot swell (Waschbusch and McNutt, 1993). Based on theobserved elevation of the topography and the predicted rate of motion of the North Americanplate relative to the hot spot framework, the rate of uplift of the eastern side of the swell is approx-imately 0.1 to 1 mm/year, but it is difficult to extract the long-wavelength signal of the swell fromtopographic noise created by older Cenozoic tectonics in the western U.S. Direct observations ofthe rate of uplift would improve our understanding of the hot spot phenomenon.

400 800 1200 1600 2000 2400 2800 3200 3600meters (contour interval 400m)

42

44

46

48

42

44

46

48

244 246 248 250 252 254

244 246 248 250 252 254

42

44

46

48

42

44

46

48

244 246 248 250 252 254

244 246 248 250 252 254

Yellowstone Topography

plates behave as rigid bodies; this is a centralpremise of plate tectonics. Nevertheless, platesdo deform, and measuring the deformation iscentral to determining the limits of simple platetectonic descriptions of surface motions, and tounderstanding processes that may ultimatelylead to the formation of new plate boundaries.This is apparent in the case of continentalportions of plates, which are not rigid and inwhich incipient rifts form. The extent ofdeformation in oceanic plates is not known atpresent, and measurements of whether any

large scale deformation is occurring is ofconsiderable interest.

4.4.3 Post-Glacial Rebound

To date, the principal constraints on therebound of the Earth’s surface in response tothe glacial unloading have come from geologicobservations of the vertical position withrespect to present-day sea level of formershorelines with ages less than 18,000 years.

32

This approach is limited because the reboundsignal can only be measured where the landsurface intersects the sea and only the verticalcomponent of the surface readjustment can beestimated.

Furthermore, the vertical datum, present-daysea level, has been continually changingposition over the past 18,000 years on accountof the draining of the glaciers into the oceanand changes in the geoid due to massredistribution accompanying melting andrebound. GPS measurements promise toprovide a much richer data set to solve for theEarth’s response to unloading in several ways:

1. GPS measures horizontal, as well as verticalmotions. Horizontal motions are much more sensi-tive to radial variations in viscosity than are verticalmotions. And they are also sensitive to variations inmodels of the ice loading history. Differencesamong predictions from currently viable modelsexceed 2 mm/yr. Within 5-10 years, GPS observa-tions will resolve much of the current debate aboutmantle viscosity structure.

2. GPS observations can provide data anywhere areceiver can be located, not just at a coastline. Awider geographical distribution of displacementdata will make it possible to separate out the “noise”of local deformation due to other causes. In order toinvestigate the regional differences in reboundcaused by lateral variations in mantle viscosity, wecan study the desiccation of inland seas like the Araland Caspian Seas using local GPS surveys to mea-sure the rebound and periodic aero-gravity surveysto measure the rate of desiccation. These aero-gravity surveys will require precise navigation byGPS in order to separate gravitational from inertialaccelerations of the moving platform.

3. Understanding the contribution of post-glacialrebound to contemporary deformations is critical foreffectively monitoring present-day sea level rise andfor properly interpreting possible tectonic deforma-tions. For example, recent models of present-daypost-glacial rebound predict onshore uplift and off-shore subsidence at rates of a few mm/yr along thePacific Northwest coast of the U.S. and Canada.This is similar to rates reported from repeated level-ing and tide gauge measurements and interpreted toreflect on-going subduction of the Juan de Fucaplate beneath North America. On the basis of con-temporary deformation, paleoseismic studies, and

comparison to similar plate boundary settings, it hasbeen suggested that this region could be vulnerableto extremely large (> magnitude 9) subductionearthquakes. GPS can contribute to our understand-ing of ongoing processes by providing 3-D estimatesof contemporary deformation which should helpseparate contributions from tectonic and post-gla-cial mechanisms. This example illustrates theimportance of improving the vertical resolution ofGPS, integrating geologic and geodetic observa-tions, and developing theoretical models tounderstand causative processes.

4. The rebound of the Earth’s surface from glacialunloading at the longest wavelength can be inferredfrom temporal variations in the degree 2 gravityfield as detected by monitoring variations in the pre-cession of satellite orbits. The precision of suchestimates can be improved with better tracking ofsatellites and by instrumenting satellites with GPSreceivers.

Like the withdrawal of Quaternary glaciers, thedrying-up of Quaternary lakes, and theassociated lithospheric unloading, provides ameans for estimating upper mantle viscosity.The Basin-Range intraplate regime represents alarge 800-km wide area of active extensionassociated with dominant normal faulting.However, a more important signature of theextensional process is the relatively hightopography of this region—that is, a 1200-meter central topographic high surrounded bylarge depressions on the province boundariesmarked by the presence of large, but nowdesiccated, Quaternary lakes: Lake Lahantonon the west and Lake Bonneville on the east.The excess 250 m high of the central uplift hasbeen attributed to mantle flow upwelling in thecenter of the Basin-Range, actively extendingthe lithosphere in a general east-west direction,opening at least 100% over its 20-million yearhistory. Province-wide measurement by GPS ofthe topographic high compared to thesurrounding lows can provide importantconstraints on processes such as mantle flow,viscosity and ultimately on how the Basin-Range works.

33

Furthermore, the upper mantle beneath LakeBonneville reflects a recognized region ofcrustal rebound of more than 300 m in the past10,000 years, attributed to a low-viscosityupper mantle of 1019 mPa. This value is wellbelow the viscosity of the surroundingcontinental mantle of 1021 mPa, and ishypothesized to be reflective of a mobilemantle. GPS measurements of these largeQuaternary lake shorelines can provide anadditional dimension and greatly add to thedetail of contemporary deformation of thisregion, which is postulated to materially effectthe capability of nearby faults, such as theWasatch fault.

4.5 Ice Dynamics and SeaLevel

The world’s oceans and the major ice sheetsform a linked system which drive much of theworld’s climate change. The oceans aredominated by circulation systems whichtransport heat and material from the equators tothe poles and back. This heat transportsubsequently drives much of the local climaticsystems. Ice sheets, as slow moving andisolated as they appear to be, have tremendouspotential for modifying the ocean circulationsystem. Recent evidence indicates thatcollapsing ice sheets not only cause sea levelrises but also rapid changes in climate. Thusdeciphering the dynamics of the oceans anddynamics of major ice sheets both are criticaltargets in the ongoing effort to understand therecord of past climate changes and predictfuture climate change.

4.5.1 Ice Dynamics

Over the last 20,000 years, the marine icesheets of the Northern Hemisphere andAntarctica have been the largest contributors tosea level change. Evidence is building that

rapid sea level rise results from an extremenon-linear response of these ice sheets to long-term climate change. Two approaches can betaken to understanding the link to sea level riseand the major ice sheets. One is to gauge thepresent mass balance of the remaining icesheets of East and West Antarctica andGreenland, while the second is to understandthe processes which control the rapid collapseof major ice sheets. Both these problemsrequire the precise positioning available fromGPS.

4.5.1.1 Mass Balance

The present mass balance estimates of themajor ice sheets is based on limited poorlypositioned data. These results to date areunsatisfactory. Changes in the surface providesinsight into the total flux of material into andout of the ice sheet. Tracking these changes insurface elevation to better than 1 cm may benecessary to identify the changes.

The ideal approach would be to map the surfaceremotely or from aircraft. A number of spacemissions have attempted this—i.e., Geosat,Seasat, and ERS-1—but the results continue tobe as ambiguous both due to the difficulty inrecovering accurate results from steeplysloping regions, ambiguities in the orbits, andthe lack of full coverage of the East and WestAntarctic ice sheets. Future missions aretargeted at this problem.

Aircraft are also capable of recovering accurateice surface elevation over a wide region.Surface-based work provides the highestaccuracy measurements, but continues to beplagued by uncertainties in location of thestations, as well as the difficulty of installingground locations. It has been proposed that themountain glaciers in temperate climates may bea significant contributor to short-term climatechange. Monitoring the surface of the mountainglaciers to better than 1 cm may provide

34

important insights into their contribution to thepresent rise in sea level.

A more direct determination of the massbalance of ice sheets can be made by takingadvantage of the well-understood elasticresponse of the solid Earth to changes insurface loads. That is, to use the Earth as aspring balance to weigh the changes in ice sheetmasses. For example, a continent-wide (~ 2100km radius) uniform change in thickness of theAntarctic ice sheet sufficient to change sealevel by 1 cm would lead to a vertical motion ofthe rock at the edge of the ice sheet of 4 mm anda horizontal displacement across the continentof 2 mm. For a given change in sea level (icemass), the displacements scale inversely withthe width of the ice sheet. Thus a uniformchange in the thickness of the West Antarcticice sheet (~ 700 km radius) sufficient to cause a1 cm change in sea level would triple thesedisplacements, giving elastic displacementscomparable to the accompanying change in sealevel.

A change in mass balance on the scale of an icestream (~ 70 km) would lead to decimeter levelelastic displacements for cm level changes insea level. Since changes in sea level of metersare envisioned to accompany global change,the resulting deformation of the elastic Earthshould be diagnostic of the change in ice sheetmass.

4.5.1.2 Ice Sheet Dynamics

To understand the entire ocean-ice sheet systemit is important to examine how ice sheetscollapse. Studies of both the climate record andthe remaining ice sheets indicates that thecollapse of an ice sheet may be a non-linearresponse to climate change. This non-linearresponse may be attributed to the bimodaldynamics now recognized for marine icesheets.

One mode is represented by interior ice whichis fixed to the bed and moves primarily byinternal deformation with surface velocities ofa few tens of meters per year. The “classic”parabolic ice sheet profile is indicative of thisinterior ice reservoir.

The second mode is characterized by icegliding over its bed which is channeled into icestreams up to 100 kilometers in width andseveral hundred kilometers in length, withspeeds of hundreds to thousands of meters peryear. These ice streams, which have very lowsurface profiles, are the conduits that drain theinterior ice reservoir.

Most of the potential sea level increase islocked within the interior ice reservoir. Amigrating ice stream system may reduce thesize of an interior ice reservoir at rates muchfaster than predicted by steady-state ice sheetmodels where the size of the reservoir changesonly in response to changes in precipitation.

The end result of this migration is a rapid sealevel rise. This migration of the interior-ice/ice-stream boundary, frequently called the onset ofice streaming, may be governed only by long-term climate change and the sub-glacialconditions which control ice streaming. Inother words, rapid sea level rise caused bymarine ice sheets may be largely independentof short-term changes in temperature andprecipitation.

4.5.2 Sea Level: The Link Between IceDynamics and Tectonics

Mean global sea level rose at an average rate of1.8 mm/year for the past 80 years. At this rateit will rise 2 cm in the next dozen years. Thisforecast based on 8 decades of data, is accurateto only 30%. If we are to forecast future sealevel trends to, say, 20% for the next fewdecades, we should need to measure sea level to

35

an accuracy of approximately 1 mm/year. Thisaccuracy is almost an order of magnitudebeyond what has been achieved in the past.

In measuring sea surface elevation directly weneed to take account of the effects of thermalvariations in the ocean that are an indirectconsequence of global change, and whichmodify our interpretation of the significance ofa sea level rise. The TOPEX mission willcontribute substantially to our understanding ofocean dynamics, but with a 3 cm accuracy orbetter will contribute little to forecasting futuretrends in sea levelat yearly periods unless therate of rise increases by an order of magnitude.

The direct measurement of sea level at acontinental coastal site often provides a noisyestimate of global sea level. Island sites arepreferable. The precision of tide gauges istypically of the order of 1 mm but seasonalsignals local to the tide gauge are generated bylocal wind stress, onshore currents,temperature and salinity variations. Some ofthese error signals perturb sea level in a randomsense (e.g. residual errors in estimating localsalinity or temperature). However, others mayintroduce systematic departures from sea level,as do, for example, long period fluctuations inonshore current direction which invite thedevelopment of trapped waves and onshoresea-surface slope.

In experiments designed to monitor differentialtectonic vertical motions, tide gauges indicatethat sea level can vary by up to 1 cm/year overdistances less than 20 km even though longperiod energy may be coherent over manyhundreds of km. This suggests that sub-cmresidual errors in tide gauge measurements maybe removed only by averaging a dense network.

The potential accuracy of the new VLBI/SLRglobal vertical datum is approximately 3 mm.Seven-mm vertical accuracy from a globalnetwork of continuously operating GPS

receivers based on weekly averages has beenreported. Recent developments in theapplication of radiometer corrections to GPSdata suggest that GPS vertical repeatabilities of2-3 mm may be achievable. Further work isneeded to demonstrate whether these results areuncontaminated by systematic corrections, andif so, GPS measurements near tide gauges havethe potential to provide an adequate verticalmeasure of tide gauge elevation.

It is envisaged that GPS can contribute in threeways to monitoring sea level: (1) links betweenVLBI sites (i.e. the international verticaldatum) and coastal sites; (2) continuousmeasurements at selected tide gauges (a GPSvertical datum); and (3) measurements betweenshoreline and offshore buoys. The first two ofthese measurements would be applied to theGLOSS network (global sea level system), anda start on this has already been made.

In addition to the GLOSS sites, measurementsto tide gauges on tectonically unstablecoastlines yield both the tectonic signal and thesea level signal when they are related to theVLBI datum. Absolute-g and space geodesydeliver comparable vertical measurementprecision. Combinations of absolute-g andspace geodesy, with their similar error budgets(but different systematic sources of error) arelikely to provide important tests of systemaccuracy.

4.6 Precise Mapping andNavigation in Oceanography

GPS technology can advance the study of theocean dynamics problems by means analogousto those for ice dynamics. Unlike land basedoperations where locating point measurementsis the dominant mode of operations, the oceanand ice issues are best addressed by remote

36

sensing tools. These tools include buoys,aircraft, ships and satellites.

The crucial contribution of GPS is to accuratelyposition these instruments, which then makespossible a wide variety of measurements, suchas Doppler velocity measurements of the watercolumn and precise measurements of the oceanor ice surface from an aircraft or satellite. Thebuoys must be positioned to 100 m, the aircraftto better than 1 m, the satellite to severalcentimeters and the ship to several meters.

Traditionally, oceanographers have relied uponpoint measurements of ocean currents usingmoored mechanical and electromagneticcurrent meters. However, during the pastdecade it has been possible to measure seasurface height to high accuracy with satellite-borne radar altimeters.

Because the large scale flow of ocean currentscontrol sea level height, with contributionsfrom essentially static variations in gravity inthe underlying solid Earth, it has becomecommon practice to infer currents globally andto repeat the measurements frequently. Theimportance of accurate positioning at sea hasbeen well understood since the Renaissance.However, improvements in navigation havebeen largely incremental during the pastseveral centuries. Precise positioning at sea hasbeen extremely expensive, requiringspecialized equipment and highly trainedpersonnel. The advent of GPS has changed thispicture completely; positioning with accuraciesof 10 m is essentially instantaneous andinexpensive. Though time consuming, precisepositioning to centimeter levels at sea is nowpossible.

4.6.1 Mapping of Global Sea Height

The TOPEX/Poseidon spacecraft is currentlymapping sea surface height in coordinationwith the World Ocean Circulation Experiment

(WOCE). The spacecraft is the only instrumentbeing applied in WOCE to provide globalcoverage of ocean circulation. The spacecraftitself is tracked by a variety of methodsincluding laser and ground-based techniques(SLR/DORIS) and the payload includes a GPSreceiver. The baseline SLR/DORIS orbits areestimated to have an rms accuracy of 5–6 cm.The SLR/DORIS orbits are now limited bymodeling errors that will be difficult to reducefurther.

In contrast, the GPS-based orbits for TOPEX/Poseidon now have an accuracy approaching2 cm, which represents an extraordinaryincrease in accuracy. Moreover, the accuracy ofthe GPS-based orbit solutions is much lesssensitive to the accuracy of models ofgravitational and non-gravitational forces onthe satellite than the SLR/DORIS solutions andthe GPS orbits have the smallestgeographically correlated orbit errors. Thisfeature makes the GPS-based orbit solutionsuniquely suited for determining the absolutegeostrophic circulation (the net flow throughthe full water column), a quantity thatoceanographers have no other means fordetermining globally.

The GPS-based orbits for the TOPEX/Poseidon spacecraft have the potential to beimproved to the 2 cm or better range in theradial direction. Any improvement in theaccuracy of the orbits translates directly intomuch reduced uncertainty of the oceancirculation on all scales for both time variableand time average motions. Reduced uncertaintyin estimates of the circulation in turn translatesdirectly into improved model initialization and,therefore, smaller errors in forecasts of futureclimate states.

If 2 cm rms errors can be achieved, the orbituncertainty will no longer be the dominanterror source, as it is now over much of theEarth. Such accuracies could, for example,

37

permit the computation of the flux divergencesof heat to and from the atmosphere by the oceanat a level of accuracy approaching that requiredto infer directly a greenhouse warming.

4.6.2 Acoustic Doppler CurrentProfiling

Oceanographers have recently developedacoustic Doppler current profilers (ADCP)which are capable of measuring current profilesin the upper several hundred meters of oceanfrom moving ships. Such currents include theimportant western boundary currents such asthe Gulf Stream. The principle of operationinvolves the measurement of the Doppler shiftin acoustic energy back scattered byheterogeneities in the water column. This newtool permits the routine measurement ofimportant shallow currents when the ship isunderway between stations or accomplishingscientific tasks of an entirely unrelated nature.

GPS provides a reference frame for themeasurement of shallow currents using theADCP. The ship’s heading must be known tohigh accuracy. Lack of knowledge of headinggenerates an enormous error in the inferredcurrents. Fortunately, the measurement of theattitude of a moving platform, including itsheading, can be measured using several GPSantennae located around the ship using phasedifferences between broadcast signals. Whilesuch measurements on oceanographic ships iscertainly not yet routine, GPS provides anelegant solution to this technical problem. Theremaining unknown in the system is the ship’sspeed which, again, can be inferred by ajudicious averaging of velocity determinationsfrom GPS position, velocity, and time (PVT)measurements. The more accurate and frequentthe PVT measurements, the more accurate theinferred shallow currents.

4.6.3 Ocean Warming

An international effort is underway to measurethe temperature of the oceans and to determineif any secular change may be related directly togreenhouse warming. The idea is quite simple;the warmer the ocean, the faster the soundspeed, and the shorter the acoustic travel timebetween a source and a receiver. Mesoscalevariability, longer term variations such as ElNiño, and seasonal variations lead tosignificant variations in temperature, butadequate averaging in time over large spatialscales (4–10 mm) can isolate the effects ofwarming.

An experiment recently conducted using aNavy-leased ship, the M/V Cory Chouest and aNavy source array to transmit coded soundsuccessfully demonstrated the practicality oftransmitting acoustic energy globally (Figure20). The ship was positioned near Heard Islandin the southern Indian Ocean and the signalswere received along the east and west coasts ofthe U.S. as well as at other stations including anarray at Ascension. The Gulf War had justbegun and AS had been turned off to facilitatefixing in the battlefield. Figure 21 plots GPSdetermined position along a ray connecting toAscension with the position integrated from theDoppler measurements at Ascension, 10,000km away. The two positions agreeextraordinarily well.

The object of the long-term measurement(Acoustic Thermometry of Ocean Climate,ATOC), however, is to fix the source andreceiver and measure the travel times over along period of time. In an associatedexperiment (GAMOT) the investigators plan toaugment the permanent array with floatinghydrophone buoys. The buoys will float freelythrough the world’s oceans recording thetransmitted signals every few hours. Theposition of the buoy is to be determined usingGPS. Because the Precise Positioning Service

38

Figure 20. Paths taken by sound in the Heard Island feasibility test (Baggeroer, 1992).

Figure 21. Observed and GPS phases (Baggeroer, 1992). The thick line shows the R/V CoryChouest’s departure from uniform motion along a course of 267° toward Ascension Island on 26January 1991, as determined by GPS. The thin line is the detected phase of the signal at Ascen-sion Island. The source-receiver separation is approximately 9000 km, yet the relative distancebetween them is tracked to within 10 m.

39

(PPS) cannot be installed on an unattendedbuoy, the on-board receiver will have access toonly C/A code with an accuracy on the order of100 m.

The investigators have determined that a 40 maccuracy is the minimum acceptable. In orderto correct the positions, the buoy PVT data willbe compressed and transmitted via the ARGOS(French) satellite to a classified DataCorrection Facility (DCF) at PennsylvaniaState University. The DCF will calculatecorrected positions and these will subsequentlybe used for data analysis. The combination ofthese two acoustic systems, one of which reliesentirely upon the GPS, should be capable ofisolating a greenhouse warming signal in adecade’s time.

4.6.4 Seafloor Mapping

Many important discoveries were made in theoceans during the 1960s and 1970s usingpainstaking mapping methods involving theintegration of single profiles of magnetics andseafloor bathymetry data. Plate tectonics,among the greatest revolutions in 20th centuryscience, followed directly from such efforts.During the past decade, fostered by earlierclassified programs within he U.S. Navy,oceanographic research vessels have begun tomap seafloor bathymetry over large swathsdirectly.

Prior to the advent of GPS, the mapping reliedupon infrequent Transit satellite fixes and deadreckoning. Data collection efforts weregenerally followed by months of painstakingprocessing involving track readjustments andrubber sheet distortions of data to removeinconsistencies in the data collected.Repositioning a ship on the otherwise highlyaccurate topography to install a station, torepeat a depth measurement, or forreoccupation in general, was essentiallyimpossible.

GPS has revolutionized mapping in the oceans.The resolution of the swath mapping systems,given the size of the acoustic footprint, isapproximately 100 m, sometimes a bit better. Inorder to position these footprints on theseafloor in a regional or global reference frame,accuracies better than 100 m are needed (C/Acode). Through efforts within the Navy, theinstallation of high-value GPS receivers on allNavy-owned ships in the oceanographiccommunity is planned.

A trial with an installation on a single ship hasbeen a great success. It is anticipated that whenthis experiment with Navy-owned ships (e.g.,Melville, Knorr, and Thompson) provessuccessful, similar receivers will be deployedon other university-operated oceanographicships (e.g., Ewing and Atlantis II) whereaccurate positioning is important.

4.6.5 Other Shipboard Applications

The measurement of gravity from a movingship has been a challenge since the 1920s. Thefirst measurements were made in submarinesusing a pendulum. Technology has advancedconsiderably and inertial tables are now used tomaintain a standard gravimeter in a verticalposition. However, one of the greatest sourcesof error in seaborne gravimetry involvescorrections of ship’s speed and heading, theEotvos Correction. While crossover errors of1–2 mgal are routine and probably indicative ofmeasurement accuracy, the use of precisepositioning at sea to determine velocity andheading could improve upon these standardsconsiderably. Furthermore, GPS precisepositioning allows the proper registration of thegravity data with the underlying topography.

One of primary economic drivers in thedevelopment of differential GPS techniquesnear continents was petroleum explorationinvolving 3-D migration of multichannel data.In this experiment, the ship, sources, and

40

receiving array is navigated with an accuracyon the order of 2 m. The data is collected in thisprecise grid and then migrated downward inthree dimensions to reveal extraordinary detailsin the subsurface structure. Whileoceanographers have used this 3-D techniquein important surveys near continental marginsand trenches near land, the lack of a similarreference frame in the open ocean precludessimilar experiments on, say, mid-ocean ridges.Perhaps GPS with greatly improved orbitscould be used to post-process position data tofacilitate such experiments.

While centimeter-level positioning on land hasbecome routine using GPS, the same is not truefor the oceans. While long-term averaging ispossible on land, the dynamics of a movingplatform at sea preclude the use of thesemethods. However, differential GPSmeasurements to a float with attached acoustictransponders to fixed monuments on theseafloor, have been successfully used to obtaincentimeter-level accuracy. Such seafloorgeodetic measurements will allow, for the firsttime, the determination of ocean ridgespreading rates and the nature of the episodicityof spreading. A pilot experiment on the Juan deFuca Ridge has been underway for severalyears. The technique depends entirely upon theuse of differential GPS so that open oceanmeasurements appear to be precluded bycurrent technology.

4.7 Atmospheric Sensing

Water, by virtue of its unique physicalproperties, plays an important role inatmospheric processes that act over a widerange of spatial and temporal scales. Watervapor is the most variable and inhomogeneousof the major constituents of the atmosphere.The distribution of water vapor is intimatelycoupled to the distribution of clouds andrainfall. Due to the large latent heats associated

with the water’s phase changes, the distributionof water vapor plays a critical role in thevertical stability of the atmosphere, and in thestructure and evolution of storm systems.

The transport of water vapor and its latent heatby the general circulation of the atmosphere isan important component of the Earth’smeridional energy balance. Water vaporcontributes more than any other component ofthe atmosphere to the greenhouse effect.Conversely, the high albedo associated withclouds limits the solar radiation available toheat the Earth. In light of these considerationsit is clear that atmospheric water vapor plays acrucial role in both weather and climate.

Atmospheric scientists have developed avariety of means to measure the vertical andhorizontal distribution of water vapor. Theradiosonde, a balloon-borne instrument thatsends temperature, pressure, and humidity datato the ground by radio signal, is the cornerstoneof the operational analysis and predictionsystem at the National Meteorological Center,and at similar operational weather forecastcenters worldwide. Space-based water vaporradiometers (WVRs) are used to mapintegrated water vapor (IWV) over the oceans,and ground-based WVRs can provide similardata over land.

None of these systems provide adequatecoverage of the water vapor distribution in bothspace and time. Accordingly there isconsiderable interest in the meteorologicalcommunity in developing new approaches tothe measurement of atmospheric water vaporthat can complement the systems already inuse. One of the most promising emergingtechnologies is GPS meteorology.

GPS meteorology refers to the use of theGlobal Positioning System for remote sensingof the atmosphere in general, and of water

41

vapor in particular. Two major approaches tothis goal have been identified.

1. Ground-based GPS meteorology. This techniqueuses ground-based networks of geodetic GPSreceivers to measure integrated water vapor (IWV)or, equivalently, precipitable water (PW) overlyingeach receiver in the network on a time continuousbasis, with a temporal resolution of 10–30 minutes.In most cases the ground-based approach will pro-vide no direct information on the verticaldistribution of water vapor.

2. Space-based GPS meteorology. GPS receiversbased in low-Earth orbit (LEO) satellites would beused to measure GPS signal delays for sub-horizon-tal signal paths through Earth’s atmosphere duringatmospheric occultation events. These data wouldbe used to estimate vertical refractivity profilesthrough the atmosphere. Given prior knowledge ofatmospheric temperature the refractivity could beused to infer atmospheric humidity, or vice-versa.The spaced-based approach will provide relativelylittle information on the lateral variation of atmo-spheric conditions.

The ground-based and space-based conceptswere developed independently. While theground-based approach exploits techniquesdeveloped by GPS and VLBI geodesists, thespace-based approach has it origins in the radiooccultation technique developed by planetaryastronomers to study the atmospheres of Mars,Venus, and the outer planets. The ground- andspace-based approaches are complementaryrather than competing techniques, and ideallyboth will be implemented as operationalsystems that will support improved weatherpredictions, better regional water vaporclimatologies, and basic research intoatmospheric storm systems and global climatechange.

4.7.1 Ground-Based GPS Meteorology

An important task in the geodetic analysis ofGPS or VLBI observations is estimatingatmospheric propagation delays affecting themicrowave signals recorded by a GPS or VLBIreceiver. These delays are caused both by the

ionosphere and the neutral atmosphere.Ionospheric delays are dispersive and areestimated by acquiring observations at twofrequencies. The delay due to the neutralatmosphere can be partitioned into ahydrostatic delay and a wet delay. Thehydrostatic delay can be predicted givenaccurate surface pressure measurements.

The possibility of using continuously operatinggeodetic GPS networks for remote sensing ofatmospheric water vapor is based upon thedevelopment of “deterministic” least squaresand Kalman filtering techniques in which thezenith wet delay (ZWD) affecting a GPSreceiver is estimated from the observationsrecorded by that receiver. The physical basisfor this measurement is the simultaneousobservation of the signal delays at a givenreceiver from multiple radio sources whichdiffer in their angles of elevation.

Using these techniques it is now possible toestimate, on a routine basis, the ZWD at eachstation in a continuously operating GPSnetwork with less than 10 mm of long-term biasin equivalent excess path lengths, and less than10 mm (rms) of random noise. Since the ZWDat a radio receiver is nearly proportional to thevertically integrated water vapor overlying thereceiver, the possibility arises of usingemerging networks of geodetic GPS receiversfor remote sensing of atmospheric water vapor.

If the vertically integrated water vaporoverlying a receiver is stated in terms ofprecipitable water (PW), that is as the length ofan equivalent column of liquid water, then thisquantity can be related to the ZWD at thereceiver. Thus, in the following equation, ZWDis given in units of length, and thedimensionless constant of proportionality, K, isa function of various physical and chemicalconstants and a parameter, Tm, rathermisleadingly known as the “mean

42

temperature,” which is a function of the verticaltemperature and water vapor distribution.

(1)

As a rule-of-thumb K≈ 0.15, but its depen-dence on Tm, which changes with site, seasonand weather, causes the value of K to vary by asmuch as 10%.

Given access to surface temperatureobservations at a GPS receiver it is possible topredict K with a relative rms error of about 2%,and numerical weather models can used topredict K with a relative rms error of about 1%.The significance of these results is that whenequation (1) is used to transform an estimate ofZWD into an estimate of PW, very little noiseis introduced by the transformation. The errorin the resulting PW estimate would derivealmost entirely from the error in the earlierestimate of ZWD.

For GPS networks with apertures smaller thanabout 500 km both the determinstic (leastsquares) and Kalman filtering techniques usedto estimate zenith delay are more sensitive torelative rather than absolute delays. Thissituation comes about because a GPS satelliteobserved using two or more receivers is viewedat almost identical elevation angles, causing thedelay estimates to be highly correlated. Theestimates of ZWD and hence PW derived froma small network are subject to an unknown biasat each epoch. The value of the bias is constantacross the whole network (i.e. the bias varies intime but not in space).

There are several possible approaches toestimating this bias, a task known as “levering”(one levers the group of biased ZWD estimatesup the ZWD axis until they all have the correctabsolute value). One is to measure PW with aWVR at a single site, then solve for the bias inthe GPS-determined ZWD and apply thiscorrection across the network (this is known as

PW K= ZWD•

“WVR levering”). Other approaches have beenconceived but are not discussed here. The mostattractive approach to eliminating the need foran independent measurement of PWsomewhere within the network is to incorporatea few GPS stations that introduce baselines inexcess of 500 km. Provided that sufficientlygood orbit information is available it will thenbe possible to estimate absolute ZWD andtherefore absolute PW using the GPSobservations alone.

In a recent experiment, a GPS receiver and aWVR were operated at either end of a 50 kmbaseline in Colorado and it was demonstratedthat the GPS-derived and WVR-derived timeseries for the difference in PW at either end ofthe baseline usually agreed to better than 1 mm.

A much larger demonstration experimentcalled GPS/STORM involved collecting GPSobservations at five Wind Profiler sites in Okla-homa and Kansas, and at a site in Platteville,Colorado. A WVR at Platteville was used tolever the ZWD values so as to deliver absoluteestimates of PW at the five remaining sites. Atthree of these sites PW was measured indepen-dently using WVRs. Thirty days of GPS- andWVR-derived estimates of PW were then com-pared (Figure 22). The difference between theWVR and GPS solutions for PW had an rmsvalue of 1.6 mm. Part of this error derives fromerrors introduced by the WVRs.

Even at this early stage in its developmentground-based GPS meteorology has alreadyproved capable of delivering PW measure-ments of practical utility to meteorology.

4.7.2 Future Opportunities in Ground-Based GPS Meteorology

It should soon be possible to develop and dem-onstrate a nearly real-time GPS meteorologycapability using a continuously operating GPS

43

1234 126

128

130

132

134

136

138

140

142

144

146

148

150

152

154

156

12341234 precipitable water vapor in cm 12341234

PW

V F

rom

GP

S, W

VR

, & R

adio

sond

esU

NA

VC

O/N

CS

U 1

993

GP

S S

TO

RM

EX

PE

RIM

EN

T

HA

SK

ELL

HA

VIL

AN

D

LAM

ON

T

PU

RC

ELL

VIC

I

DA

Y O

F Y

EA

R, 1

993

Purc

ell (

28 k

m)

Lam

ont (

1 km

)

Hav

iland

(96

km

)

Precipitable Water in cm

Day

of

Yea

r, 1

993

Pre

cipi

tabl

e W

ater

Fro

m G

PS,

WV

R, &

Rad

ioso

ndes

UN

AV

CO

/NC

SU 1

993

GPS

/ST

OR

M E

xper

imen

t

Vic

i (12

8 km

)

Has

kell

(125

km

)

Figure 22. Precipitable water (PW) is shown for 30 days and five stations (Rocken et al., 1994).Each panel shows the station name and distance from nearest ROAB release site in the lower leftcorner. The small blue dots in the top three traces of each panel are WVR measurements, smallred lines are 30-minute GPS estimates, and the isolated green points, separated by many hours,are determined from RAOBS. WVRs at VICI and Purcell were pointed at the GPS satellites andthe measurements were scaled to zenith. The Lamont WVR observed in the zenith direction only,thus the reduced scatter of the WVR measurements. The arrows in the Lamont panel point atexamples of faulty WVR data—in these cases caused by dew on the WVR window. Other WVR out-liers in the top three panels are typically caused by rain.

44

network being constructed in the central U.S.under the auspices of the NOAA Wind Profilergroup, and possibly using other testbeds inaddition. Once a real-time capability has beendemonstrated it is likely that the meteorologycommunity will push ground-based GPS mete-orology as a major operational initiative.

There are major potential synergies betweenthis activity and GPS networks presentlyplanned by a variety of government agencies.The FAA is planning to install continuouslyoperating GPS receivers at more than 300airports in the U.S. to support precisionlandings during the next few years. The CoastGuard is planning to install 30–50 continuouslyoperating systems at coastal sites in order tosupport DGPS navigation. Many state geodeticsurveys and highway departments areconstructing or have already constructed smallcontinuous GPS networks.

If these efforts could be coordinated, it shouldbe possible to construct a national multipurposeGPS network capable of supporting a widerange of scientific, governmental, andcommercial applications. GPS meteorologycould act as the catalyst in forging such analliance. The National Weather Service alreadyhas stations in most of the airports at which theFAA plans to install GPS, and has many yearsof experience in nearly real-time datagathering. The major problem in organizing anational GPS network will be political ones—itrequires several disparate government agenciesto work in a timely fashion towards a commongoal.

4.7.3 Space-Based Meteorology withGPS

Preliminary studies indicate that accurate,high-resolution atmospheric soundings can beretrieved from the observations obtained whenthe radio path between a low Earth orbiting

(LEO) GPS receiver and one GPS satellitetraverses the Earth’s atmosphere. When thepath of the GPS signal begins to transect themesopause at about 85 km altitude, it issufficiently retarded that a detectable (1 mm)delay in the dual-frequency carrier phaseobservations is obtained by the LEO GPSreceiver. As the signal path descends throughsuccessively more dense layers of theatmosphere, the delay increases toapproximately 1 km at the Earth’s surface.

Thus, the atmosphere creates a unique signalwith over 6 orders of magnitude in dynamicrange. A single LEO GPS receiver couldobserve more than 500 total occultations perday (~250 rising and ~250 setting), withroughly uniform global coverage.

The delay is caused primarily by increasedbending as the ray path descends throughincreasingly dense layers of the atmosphere,and to a lessor degree by the reduction inpropagation velocity. GPS can be used tomeasure the precise positions of the satellitesinvolved, just as the system is used on Earth. Inaddition, the LEO receiver can be used torecord precise phase measurements on theocculted GPS signals. Combined with theprecise positions, these phase measurementsprovide a time series of “excess Doppler” (i.e.,Doppler in excess of that explained by themotion of the spacecraft), which can beinverted to a refractivity profile of theatmosphere.

Measurements of this type could provide newdata for at least two key global changevariables: atmospheric temperature andmoisture distribution. For dry air (typically,measurements above 7 km), the refractivityprofile can be converted to a temperatureprofile with an accuracy of ~1° C up to analtitude of about 45 km. For regions withsignificant moisture in the atmosphere, therefractivity profile can be converted to a

45

moisture profile with an accuracy of roughly10% RH, provided ancillary temperature data,or a model of temperature is available.

An operational system that meets its fullpotential could significantly improve weatherforecasts, particularly over ocean areas, wheremeteorological data is sparse today. A longrecord could provide valuable new climatedata. Such a system could be implementedusing a constellation of 50–100 very small (<20kg/20 W) dedicated satellites. These satellitescould be mass produced and launched 8–16 ata time on small, inexpensive launch vehicles. Itis estimated that the average cost peroperational satellite, delivered in orbit, wouldbe less than $5 million in such a scenario.

To investigate the practical potential of theconcept, UCAR/UNAVCO is developing asystem to collect radio occultation data using acommercial spacecraft and a special GPSreceiver. The project, known as GPS/MET, isbeing funded by NSF, FAA, and NOAA.Partners in the project include NCAR/MMM,Orbital Sciences Corporation, Allen OsborneAssociates, the Jet Propulsion Laboratory, andthe University of Arizona. Data will becollected and published on the Internet forapproximately 6 months beginning shortlyafter launch.

4.7.4 Probing the Ionosphere

The non-neutral part of the Earth’s atmosphereaffects the GPS signals dispersively. Dual-frequency GPS receivers can therefore providea direct measure of the integrated number ofelectrons between the receiver and the satelliteand thus a direct measure of ionosphericactivity. This very precise measurement ofionospheric activity can be used in ionosphericscience.

Precise dual-frequency measurements byground- and space-based GPS receivers will

probe the ionosphere along many paths.Variations in signal phase will revealionospheric density waves caused by seismic,tsunami, massive explosions (as caused bynuclear tests), meteorological, and many otheratmospheric and geophysical events. Thesedata can be used to study the relationshipbetween the source event and the atmosphericeffect, and to trace the flow of energy withinand across different Earth systems.

Measurement of total electron content alongthese paths will be assembled into high-resolution, 3-D electron distribution maps bycomputerized tomography to study irregularand transient mesoscale structures, such as theequatorial bubbles and the mid-latitude trough.Study of the ionosphere with GPS signals willlead to a better understanding ofcommunication problems associated withionospheric disturbances.

4.8 Instrument Arrays andData Networks

UNAVCO scientists are interested incontinuously operating GPS networks in a widevariety of contexts. Some new or emergingapplications of GPS, especially GPSmeteorology, actually require fixed andcontinuously operating networks in order toachieve useful results. Some of theseapplications will involve deploying fixed,continuously running networks for periods asshort as a few months; others (for examplewater vapor climatology) will require verylong-term deployments.

Other applications, like volcano monitoring,can be performed in a campaign mode butbecome far more powerful when implementedusing permanent networks. Most crustalmotion projects involving large numbers ofstations in remote areas are likely to rely

46

predominantly on field campaigns for theforeseeable future, on purely financial grounds,but nevertheless may benefit considerably byincorporating a small permanent subnetwork.Global tracking has much improved over thelast few years, but even now there are someareas of the world (e.g. polar areas and thePacific Ocean basin) where additional trackingstations would be very valuable.

The key concept for promoting continuouslyoperating GPS stations is multipurposejustification. Another but related concept isbuilding up these networks using multiplesources of funding. For example NOAA mightbe interested in funding permanent stations atsome tide gauges in the Pacific. The rationale isonly permanent tracking can average downvertical errors to the point that one candistinguish between absolute and relative sealevel change. Meanwhile, DOE’s ARM projectmight fund several other stations in the regionfor the purposes of water vapor climatology.The existence of these sites might enable acrustal motion researcher to write an NSFproposal to add a few more permanent stationsnearby the DOE and NOAA sites, and end upwith large network that could support crustalmotion research. The crustal motion projectmight add barometers to its stations so that theycould support water vapor studies. One of thetide gauge sites might send data out daily toSIO, JPL and CODE to support improvedglobal tracking.

The point is that few applications can justifythe construction of large numbers ofcontinuously tracking GPS stations in a givenregion, but that such a network might comeabout by seeking multiple applications andmultiple sources of funding and sharing thedata. Each effort leverages the others! Smallinterdisciplinary groups of PIs can set upcollaborations of this kind, but they will need tothink about configuring each station so that itsupports the broadest possible range of

activity—even non-scientific activities such asDGPS navigation where local governmentsupport would be useful.

5 Accomplishments and Ideas

At the time of the workshop which formed thebasis for this report, numerous individuals pro-vided written summaries of their researchbased on the Global Positioning System. Thesesummaries are reproduced here to provide asample of the great variety of new GPS re-search programs spanning virtually the wholeof the geosciences.

5.1 Subsurface Geodesy

Roger Bilham, University of Colorado

Several categories of geological process are notwell monitored using surface geodesy. Forexample, the surface deformation fields ofhorizontal detachments are indistinguishablefrom buried vertical faults. Anothercomplication can occur where several thrustsmay be stacked and intermittently active.Another application where interpretation ofsurface geodesy is ambiguous occurs whereseveral parallel faults are active within a plateboundary. An example is the widened velocityfield found where the San Andreas, Calaveras,and Hayward faults overlap. It is not possible todistinguish the portion of the subsurface slipuniquely as a function of depth from the surfacegeodesy alone.

The inherent ambiguity in these measurementscan be overcome by combining boreholemetrology and GPS. Advances in boreholemetrology permit horizontal deviations of aborehole to be measured to sub-mm accuracy.Random errors of inclinometer measurementsgrow as where L is the borehole length inkm and k is a constant in the range of 0.5 to 1.4.Vertical measurements using fiber rods

k L

47

attached at different depths yield mm accuracy,and micron accuracy can be obtained usinginterferometric methods and optic fibers. Acombination of vertical and horizontalmeasurements permit 3-D subsurface positionsto 1.4 mm accuracy to depths of 2 km.

A corollary of the method is to use boreholes toprovide exceptionally stable control pointseven in regions considered unsuited to preciseGPS geodesy: hillslopes, permafrost, andsedimentary basins.

5.2 Measurement of LocalGeoid Slope

Roger Bilham, University of Colorado

First-order spirit leveling delivers a relativeelevation measurement accuracy of approxi-mately 1 mm in 2 km. GPS over similardistances yields similar accuracies in relativeellipsoidal elevations. The difference in eleva-tions measured by these two techniquesprovides a measure of the slope of a co-geoid atthe local elevation of the Earth’s surface. Atsea-level, the measurement is of geoid slope.

A 2-km line of GPS and double-run levelingcan be measured in a day by a three-personcrew. Thus geoid slope data are relatively easyto acquire. The angular accuracy is in the orderof 1 microradian, similar to qualitymeasurements of the deviation of the vertical.Unlike deviation of the vertical measurementsusing stellar setting, GPS/leveling combinedmeasurements yield a spatial densification ofgeoid undulations, and investigations of thevertical gradient of geoid slope.

Two applications are envisaged for geoid slopedata: verification and spatial densification ofgeoid undulations, and investigations of the

vertical gradient of geoid slope. The firstapplication is important in regions whereleveling lines are rare, for example, parts of theHimalaya. In these locations, spot verificationsof geoid slope would be of great value, aswould an investigation of geoid “granularity.”The second application, useful in mountainranges, is to measure the slope of co-geoids atdifferent elevations. The gradient of the geoidslope provides an indication of gross massdistributions, hitherto accessible only to gravitymeasurements.

5.3 Precise Airborne SurfaceAltimetry Based on GPS

Positioning

D.D. Blankenship, Institute for Geophysics,University of Texas

R.E. Bell and V.A. Childers, Lamont-DohertyEarth Observatory

J.M. Brozena, Naval Research Laboratory

Precise surface elevation measurements arecrucial to understanding the dynamics andmass balance of major ice sheets as well as theform of vertical tectonics. As part of aninterdisciplinary effort to understand theinteraction of the West Antarctic ice sheet withthe underlying lithosphere the CASERTZproject (Corridor Aerogeophysics of theSouthern and Eastern Ross Transect Zone) hasdeveloped the capability to recover precisemeasurements of the ice surface. Thiscapability is a component of a broad spectrumaerogeophysical platform which couplessimultaneous airborne gravity, aeromagnetic,ice-penetrating radar, and laser rangingmeasurements with kinematic GPSpositioning, UHF radio navigation, laser-gyroinertial navigation and high-resolution pressure

48

altimetry. The program has focused on a 97,000square kilometer region along the southernflank of the West Antarctic rift system whichtraverses critical boundaries both in the riftsystem and the ice sheet.

The CASERTZ aircraft includes a high-resolution surface imaging capability using aHolometrix laser altimeter. This laser altimeterhas a ranging accuracy of 0.1 m, a footprint ofabout 0.9 m and is capable of up to 1,000 rangeobservations per second. In general, an averageof 64 ranges is recorded every 10 m alongtrack. We have evaluated the quality of the icesurface measurements derived from theCASERTZ laser altimetry system including thecorrections for aircraft attitude and fulldifferential GPS positioning. The ice surfacealong coincident profiles flown in the samefield season and approximately 70 m apart hasa mean difference of 24 cm with an rmsdeviation of 35 cm. Preliminary results indicatethat the full three-dimensional ice surface willhave an absolute accuracy of approximately0.5m.

5.4 Airborne GeophysicalObservations of Long Valley

Caldera

C.R. Carrigan, Lawrence Livermore NationalLaboratory

J.B. Rundle, University of Colorado

The Long Valley Caldera is a 640 cubickilometer basin formed from the collapse of thecrust over a magma chamber that violentlyerupted its contents some 700,000 years beforethe present. Both seismic and geodetic datasuggest that the chamber underlying the basinmay currently be refilling with a fresh charge ofmagma. At the point of maximum uplift in the

basin near Highway 395, the surface is rising ata rate of about 0.1 m per year. Because it is soactive in terms of its seismicity and inflation, anumber of detailed geodetic surveys have beencarried out in this region in recent years. Thus,it represents an ideal location to employ a new,rapid surveying technique that can be validatedagainst the results of earlier, but more standardsurveys. By combining the elevation data froman ATLAS laser altimeter mounted beneath aT39 aircraft with the highly accurate airplanelocations provided by two onboard commercialGPS receivers, extremely rapid surveys oftopographic elevation can, in principle, beobtained over large areas. During September,1993, such an aerial survey was conducted in ajoint experiment with NASA, UC San Diego,the USGS and LLNL. The flight plan took thelaser altimeter over the region of maximumuplift near Highway 395. It is hoped thataccuracies in elevation of 5 cm can beultimately achieved, besides validating a noveltechnique. The survey results will be used inconstraining geodynamical models of an activevolcanic region. Presently, LawrenceLivermore National Laboratory is using finiteelement models that can utilize this type of datato better understand the nature of magmainjection beneath the caldera.

Possible future applications of this techniqueinclude rapid aerial survey of suspectedunderground nuclear test sites as part of an on-site inspection program intended to verify thatthe conditions of a nuclear treaty are beingobserved by its signators. Such a wide-areasurvey approach could be used to pre-selectmuch smaller areas using topographicsignatures associated with undergrounddetonations. The candidate sites would then besubjected to detailed surface studies—e.g.,seismic aftershock observations and detectionof noble gas transport—by a ground team todetermine both the occurrence and nature of asuspected underground event.

49

5.5 GPS Measurement ofRelative Motion of the Cocos

and Caribbean Plates andStrain Accumulation Acrossthe Middle America Trench

Timothy H. Dixon, University of Miami

GPS measurements in 1988 and 1991 on CocosIsland (Cocos plate), San Andres Island(Caribbean plate), and Liberia (Caribbeanplate, mainland Costa Rica) provide anestimate of relative motion between the Cocosand Caribbean plates. The data for Cocos andSan Andres Islands, both located more than 400km from the Middle America Trench, define avelocity that is equivalent within two standarderrors (7 mm/yr rate, 5 degrees azimuth) to theNUVEL-l plate motion model. The data forLiberia, 120 km from the trench, define avelocity that is similar in azimuth butsubstantially different in rate from NUVEL-l.The discrepancy can be explained with asimple model of elastic strain accumulationwith a subduction zone that is locked to arelatively shallow (20+5 km) depth.

5.6 Constraints onDeformation of the ResurgentDome, Long Valley Caldera,

California from SpaceGeodesy

Timothy H. Dixon, Frederic Farina, andStefano Robaudo, University of Miami

Mar cus Bursik, State University of New York

Susan Kornreich Wolf, Michael Heflin, andFrank Webb, Jet Propulsion Laboratory

Long Valley Caldera near Mammoth Lakes ineast-central California has experiencedincreased levels of seismic activity and surfacedeformation associated with inflation of ashallow crustal magma chamber since 1978.Inflation is concentrated in the vicinity of acentral resurgent dome. We use VLBI and SLRdata from the western U.S. to constrain theregional tectonic setting of Long Valley, andVLBI and GPS between Owens Valley RadioObservatory (OVRO) and sites near(Mammoth) or on (CASA) the resurgent dometo investigate local volcanic strain.

OVRO has significant (10+0.5 mm/yr.) north-northwest motion with respect to stable NorthAmerica, suggesting that OVRO, Mammoth,and CASA all lie west of the Eastern CaliforniaShear Zone. Vertical velocities defined byMammoth VLBI data (1983–1986) and earlyCASA GPS data (1985–1988) have largeuncertainties (1 errors are+14.7 and+5.3 mm/yr., respectively), but suggest no significantuplift relative to OVRO, consistent with otherobservations, suggesting this was avolcanically quiet period. In contrast, GPS datafor CASA for the period 1988–1992 indicateuplift at of 24.4 mm/yr. +2.8 mm/yr. andsouthward motion at 9.0+0.4 mm/yr. relative toOVRO, agreeing with the leveling and two-color laser geodimeter data withinuncertainties. This motion is expected for a siteon the south flank of an inflating dome. TheGPS data are also consistent with the rate ofdeformation observed in geodimeter data aftermid-1989, but are too sparse to define the strainevent independently.

While available space geodetic data provideuseful constraints on contemporarydeformation in Long Valley, the number of sitesis insufficient to define the regional tectonic orlocal volcanic strain fields, and the observation

50

frequency of roughly once per year at CASA istoo low to define the time-varying signalassociated with volcanic deformation. Thechanging observation conditions characterizingthese intermittent experiments have alsocontributed to systematic error. The advent ofpermanent GPS tracking networks, thedeclining costs of GPS receivers, and the needto improve accuracy and avoid temporalaliasing suggest that continuous monitoring ofLong Valley with several GPS sites is bothfeasible and desirable.

5.7 GPS Monitoring ofCrustal Strain in SouthwestBritish Columbia with the

Western Canada DeformationArray.

H. Dragert and X. Chen, Geological Surveyof Canada

J. Kouba, Geodetic Survey of Canada

Contemporary crustal deformation has beenmonitored across the northern part of theCascadia subduction zone since 1981.Measurements of accumulating horizontalstrain have been obtained through repeatedlaser-ranging, GPS surveys, or both of fivegeodetic control networks established invarious locations on Vancouver Island.Although shear-strain rates ranging from 0.05to 0.23 ppm per year have been resolved, theseestimates are spatially and temporally isolatedsince they lack a continuous common fiducialframework that has a precision commensuratewith the internal survey precisions of theindividual networks. To address this problem,the Geological Survey and the Geodetic Surveyof Canada are cooperating in establishing acontinuous automated network of four regional

GPS tracking stations. A configuration of threeRogue receivers has now been operatingcontinuously since August 1992. The fourth isscheduled to come on-line by October 1993.Dual-frequency pseudorange, and phase data,sampled at 30-second intervals, are collecteddaily by an automated process running on a Sunworkstation. Quality checks are performedautomatically by programs which generatestatistical summaries and plots of the past day’sdata for each site. Baseline analysis is carriedout using precise ephemerides provided by theGeodetic Survey of Canada. Results from thepast year of data show a repeatability of 3 to 5mm for baseline lengths of 300 and 600 kmrespectively. This order of precision will allowcomprehensive monitoring of crustal strain onCanada’s west coast.

5.8 GPS Used to MonitorGround Movements at

Augustine Volcano, Alaska

Daniel Dzurisin, U.S. Geological Survey

In June 1992, the USGS Volcano HazardsProgram installed a permanent, radio-telemetered GPS network to monitor grounddeformation associated with future eruptions ofAugustine Volcano, Cook Inlet, Alaska.Augustine is the most active volcano in CookInlet, as exemplified by eruptions in 1935,1963, 1976, and 1986. The island volcano hasno permanent residents, but airborne ash fromfuture eruptions could threaten air traffic and alandslide-generated sea wave could threatenthe coastal city of Homer. The Augustine GPSproject is a cooperative effort between theCascades Volcano Observatory (Vancouver,Washington) and the Alaska VolcanoObservatory (Anchorage/Fairbanks, Alaska).

51

The current USGS network consists of threeAshtech LD-XII dual-frequency, carrier-phaseGPS receivers: one at the base of the volcanonear the eastern shore of the island, one on theupper north flank, and one on the 1986 lavadome near the 1250 m summit. The near-shorestation is held fixed for data processing; dis-tances to the other stations are approximately4.5 km and 4.8 km. Thirty-second epoch dataare transmitted at 300 baud via VHF radio to anIBM compatible PC in Homer, 80 km away. Asecond PC in Homer is networked to the first,enabling users to download data via standarddial-up telephone lines.

Five 18-watt solar panels charging eight 12-volt 90 amp-hour batteries provide power ateach field site. Because this is not sufficientpower to operate the GPS receiverscontinuously (power consumption≈ 20 watts),we control their duty cycle by issuingcommands from the Cascades VolcanoObservatory through the dial-in computer inHomer. These commands are radio-transmittedto the field sites, where a low-power controllerreceives and executes the commands to turn thereceivers on and off. During the first year ofoperation, the receivers were turned on for 2-hour sessions once per week. Ice accumulationon solar panels and fiberglass antenna housingsresulted in data loss for several months duringthe winter. Modifications to the stations weremade in July 1993 in an attempt to mitigate thisproblem.

The Augustine GPS data are processed at theCascades Volcano Observatory by an IBMcompatible 486 PC using Ashtech software.Processing requires only minimal user-intervention. The time required to downloadand process 2 hours of data is approximately 45minutes. Repeatability during the first year ofoperation (excluding data affected by antennaicing) was+5 mm north,+8 mm east, and+35mm up.

In a related effort, we are exploring the possi-bility of monitoring restless volcanoes usingrelatively inexpensive, single-frequencyreceivers such as the Motorola SixGun(approximate price = $3000). During prelimi-nary tests, we processed SixGun data withAshtech software and obtained several centi-meter repeatability over baselines of 3 km and8 km. Our goal is to design a field-ready, porta-ble, telemetered station capable of+1 cmhorizontal repeatability and+3 cm verticalrepeatability over baselines up to 5 km, for ahardware cost of about $5000/station. Suchinstruments could be rapidly deployed duringvolcano emergencies together with portableseismic stations and electronic tiltmeters toprovide real-time monitoring data in support ofhazards assessments and eruption forecasts.Additional tests are planned for 1994.

5.9 Use of AirborneKinematic GPS and LaserAltimetry for Determining

Volume Changes of MountainGlaciers

K. Echelmeyer, W. Harrison and C. Larsen,University of Alaska

Mountain glaciers are sensitive indicators ofclimate change. In addition, it is believed thatthe mountain glaciers of Alaska and Canadaplay an important role in changes of sea level.The key parameter characterizing the effects ofthe glaciers on sea level and their response to achanging climate is glacier volume.

In the past, changes in glacier volume havebeen measured using repeat aerialphotogrammetry or terrestrial surveying. Bothof these methods were expensive and timeconsuming, and therefore existing data on

52

volume changes of glaciers throughout theworld is very limited. With the advent ofcompact laser systems and GPS positioning ithas become possible to monitor changes inglaciers from small aircraft.

We have developed a lightweight, compactlaser profiling system which can be mounted ina small single engine aircraft which is capableof flying at low speed over the winding surfaceof a mountain glacier. This system provideshighly accurate elevation profiles of glaciers inmountainous terrain. Aircraft positioning at thedecimeter level is provided by continuouskinematic GPS. Dual-frequency data isrecorded for the entire flight, plus staticinitialization periods at the beginning and endof each flight. Post processing of the GPS datathen gives the position of the aircraft relative toa fixed base station as a function of time.During the flight the laser range above the icesurface and the gyro-determined attitude(pointing angles) of the laser are measuredcontinuously. Both of these data sets are givenGPS time tags. All the data are then combinedto give a profile of surface elevation for theglacier along a given transect. Comparisonwith previous profiles or with existingtopographic maps then gives an estimate ofvolume change.

Results to date indicate that the profiles areaccurate to 10 to 50 cm, depending on baselinelength, ionospheric activity, altitude above theice surface, and surface slope. This is an orderof magnitude or more better than aerialphotogrammetry. The major source of error,and the most time consuming aspect of the finalprofile, is the kinematic GPS data processing.Improvements in on-the-fly ambiguityresolution and receiver tracking abilities willlead to a reduction in both the overall errorbudget and the time involved in producing afinal profile.

In order to repeat profiles in the future using thelaser system, say on a yearly or 5-year basis, theaircraft must have the capability of real-timenavigation which allows the given ground trackto be repeated within+5 to 10 meters. For suchnavigation we have utilized real-timedifferentially corrected GPS in which thecorrection factors are broadcast from a fixedbase station over a radio modem. The airbornereceiver then uses these correction factors toguide the pilot through a series of waypointsalong the transect.

To date, several profiles have been done onglaciers in various climatic regions of Alaska.Preliminary results indicate that there has beensubstantial thinning of the measured glaciers inArctic and interior Alaska, while the signatureof volume change is variable in the coastalregions.

5.10 Use of GPS Technologyin Glaciology

Keith Echelmeyer, University of Alaska

The study of the dynamics of ice sheets andmountain glaciers requires the determination ofice velocity, strain rates, thickening or thinning(mass balance), and ice thickness. Since 1989we have been applying GPS surveying methodsto these measurements. Examples of this workinclude:

l) Velocity and strain rate determination usingstatic dual-frequency GPS positioning overrelatively long baselines (50 to 500 km) inGreenland, Alaska, and Antarctica. Speeds wehave determined range from 2 m/yr on the WestAntarctic ice sheet, to 20 to 500 m/yr in Alaska(e.g., Ruth, West Fork, Black Rapids, Gulkana,and McCall glaciers) and West Antarctic icestreams (ice stream B), and up to 7000 m/yr on

53

Jakobshavns Isbrae in Greenland. The length ofthe baselines to fixed bedrock requires dual-frequency processing at the 1 to 10 ppm level.

2) Stop-and-go and continuous kinematic GPSmeasurement of surface elevation and mappingof terminus position. A GPS receiver istransported on foot, skis, or snow machinealong the surface of a glacier or around theterminus. Comparison with existingtopographic maps and optically surveyedtransects then allows us to determine howglaciers in different regions have beenchanging over the past few decades. We haveperformed such studies on McCall Glacier inArctic Alaska, Gulkana and Black Rapidsglaciers in interior Alaska, and on glaciersflowing off of Mt. Wrangell, a south-centralAlaskan volcano.

3) Differentially corrected GPS for positioningseismic and ice radar transects. In order todetermine the depth of glaciers we employ bothseismic and monopulse radar techniques.Positioning of the measurement systems andalong-transect surface topography is needed toaccurately determine ice depth, especially in adeep, narrow valley. We utilize post-processeddifferentially corrected GPS to obtain locationsat the 2- to 5-meter level or stop-and-gokinematic to obtain positions at the centimeterlevel.

4) Control networks around glaciers for use inoptical surveying. We routinely use static GPSto carry geodetic control into a glacier becausehigh-order benchmarks do not usually exist inglaciated regions. One ppm accuracy isrequired.

5) Geoid determination in areas of highmountainous relief. Using concurrent GPS andoptical surveying we have determined theheight of the geoid in relation to the ellipsoid inDenali National Park. Accurate geoidinformation did not previously exist in this

region, and departures from existing modelswere larger due to the large mass of Mt.McKinley which dominates the region.

5.11 Solar Hygrometer forGPS Field Use

Henry F. Fliegel, The Aerospace Corporation

Three methods are in common use by which tocalibrate the effect of atmospheric water vapordelay on geodetic field measurements for GPS:(1) surface weather data models, based ontemperature and relative humidity on theground; (2) statistical modeling, in which theGPS data itself is used to provide an estimate ofthe delay as a stochastic variable, and (3) watervapor radiometry. These three methods are notsuitable for all applications, since method (l) isnot accurate, method (2) requires a long span ofdata under stable conditions, and method (3) isexpensive. The solar hygrometer, whichestimates total water vapor content in the lineof sight from absorption bands in the visual ornear infrared spectrum, is a logical candidatefor a calibration instrument midway in cost andaccuracy between methods (l) and (3). Untilnow, two strong objections to the solarhygrometer have been that (i) it must bepointed at the sun, and therefore cannot be usedat night nor aimed toward a GPS satellite, and(ii) it cannot be used in cloudy weather. Twodevelopments are worth exploring.

(I) There is a moderately strong water vaporabsorption band at about 7000 Å, and also acomplex of features in the blue visual spectrum(the so-called “rain bands”). Although thesefeatures are shallow and sloppy compared tothe classic band at 9350 Å, they are accessibleto standard photomultipliers, which have greatsensitivity. Then a properly designedhygrometer can be aimed at stars by night and

54

at the blue sky by day. Since the half height ofthe Rayleigh scattering component of theatmosphere is more than 5 kilometers, and thatof water vapor 1 kilometer or less, the blue skyused with suitable corrections is almost as gooda background for estimation of total watervapor content as the sun.

(II) A sufficiently sensitive instrument can beaimed at Polaris, which is now only about 0.7degrees from the north celestial pole, andoperated day and night. A diaphragm canscreen out most of the skylight, and the rest canbe filtered out. Using a clock drive, almost allbackground light can be excluded, since thediaphragm can be made very small, and sincesky light 90 degrees from the sun is stronglypolarized.

5.12 Non-Equilibrium Forceson GPS Satellites

Henry F. Fliegel, The Aerospace Corporation

Two anomalies in GPS space vehicle (SV)accelerations are manifested in the first fewmonths of operation or during eclipse,especially during the first eclipse season afterlaunch. (l) Theinitial acceleration away fromthe sun of an SV recently placed in orbit isabout 6% higher than the final value, and thendeclines exponentially by a factor of e aboutevery 18 weeks. It finally converges to withinabout 2% of the value predicted by the IERSadopted T20 model. (2) It then exhibits aresidual accelerationhistory of about +2%around the mean value, roughly in phase withthe eclipse seasons. I conjecture that thematerials of the SV are outgassing. Theintegrated impulse for Block II SV’s rangesfrom 20 to 75 kilogram meters/ second. Fourcontributors were examined as possible sourcesof outgassing: (A) the solar panels; (B) the SV

body; (C) the apogee engine; and (D) themultilayered insulation (MLI) on the SV body.I conclude that water vapor from contributor D,the SV insulation, is the most likely primarysource of the outgassing suspected to beresponsible for the anomalous initialacceleration.

5.13 Sub-Daily EarthRotation During Epoch’92

from GPS

A. P. Freedman, R. W. Ibanez-Meier, S. M.Lichten, and J. O. Dickey, Jet Propulsion

Laboratory

Thomas Herring, Massachusetts Institute ofTechnology

Earth rotation data were obtained using GPStechniques during the Epoch’92 campaign inthe summer of 1992. About 10 days of datawere acquired during the last week of July andfirst week of August from 25 globallydistributed stations and a constellation of 17GPS satellites. These data were processed toestimate corrections to a nominal UT1 series at30-minute intervals using several strategies.Earth orientation data during Epoch’92 werealso obtained by several VLBI groups, andwere processed together to yield VLBIestimates of UT1 at several time resolutions.One can see that the high-frequency behaviorof both data sets is similar, although driftsbetween the two series of ~0.1 ms over 2–4days are evident. The diurnal and semidiurnalvariations of both series are largely attributableto changes in UT1 induced by non-equilibriumocean tides, and are probably affected bydiurnal atmospheric variations also. Tidallyinduced UT1 from both theoretical tide modelsand empirical (VLBI) measurements overmany years were compared with the GPS and

55

VLBI series. In addition, estimates ofatmospheric angular momentum (AAM) at 6-hour intervals generated by severalmeteorological centers have been comparedwith the geodetic data. These comparisonsindicate that much of the GPS signal in thediurnal and semidiurnal bands can be attributedto known physical processes. At both longerand shorter periods, the GPS results may beaffected by systematic effects such asunmodeled orbit or other errors.

5.14 Atmospheric Excitationof Rapid Polar Motions

Measured by GPS

Richard S. Gross and Ulf J. Lindqwister, JetPropulsion Laboratory

Daily polar motion values determined fromobservations taken during the GIG’91 GPSmeasurement campaign conducted duringJanuary 22, 1991 to February 13, 1991 arecompared to twice-a-day polar motionsinduced by atmospheric angular momentum(AAM) excitation functions from the JapanMeteorological Agency (JMA). AAM-inducedpolar motion series are formed from just thewind term, just the pressure term (with andwithout assuming the inverted barometerapproximation for the response of the oceans tosurface atmospheric pressure changes), as wellas from their sum in order to study the relativeimportance of these terms in exciting polarmotions. It is found that the wind and pressureterms are of comparable importance to excitingthe observed polar motions during this period,and that somewhat greater agreement with theobservations are obtained when using thepressure term computed under the invertedbarometer approximation. Treating the polarmotion series as complex valued, correlationsas high as 0.88 are obtained between the

observed and AAM-induced series, and asmuch as 74% of the variance of the observedseries can be explained by the AAM-inducedseries. It is therefore concluded thatatmospheric wind and pressure fluctuations arelargely responsible for the polar motionsobserved to have occurred during the GIG’91measurement campaign.

5.15 Large ScaleCombination of GPS and

VLBI Data

Thomas Herring, Massachusetts Institute ofTechnology

In recent years large volumes of both GPS andVLBI data have been collected and analyzed.In our research we have been developingmethods for combining these two data setstogether to produce velocity fields and to studythe non-steady state motions of the sitesobserved using these two systems. Theformulation we use for these combinations isbased on the Kalman filter. The basic inputs forthese combinations are solutions from GPS andVLBI data analyses in which all of the sitepositions, satellite orbital elements,extragalatic radio source positions and a varietyof tidal and Earth rotation parameters(currently only the VLBI analyses) are looselyconstrained. These loose constrained solutionscan be combined with additional constraintsapplied using the Kalman filter formulation.

The unique aspects of our implementation ofthese combination analyses are several fold.Within the GPS data analysis, we have theability to “bias fix” using a constrained GPSsolution and then to remove the constraintswhile maintaining the fixed biases. Thisprocedure overcomes the problems ofinstabilities in the bias fixing algorithms if the

56

solution is unconstrained when biases are fixed.Some other features of the combination systemthat we use include:

(1) The ability to combine all stationinformation to produce position and velocitysolutions while not heavily constraining anyparameters in the analysis. In a post processingstep, we are then able to constrain theadjustments so selected parameters are thesame, to force specified parameters to their apriori values, or both. The former of these stepsallows specification of survey ties betweenmonuments without constraining the locationof the ensemble of sites and to ensure thatmonuments in the same area move with thesame velocity without constraining the value ofthe velocity. More importantly, as eachconstraint is applied we evaluate the change inχ2 due to the imposition of the constraint andthus we are able to test the consistency betweenthe parameter estimates and the constraints.

(2) As the Kalman filter is incremented witheach new set of data, the change inχ2 due to theaddition of the new set is evaluated and can bechecked for consistency between the currentdata set and the accumulated results from allprevious data sets. Theχ2 increments arerigorously computed using the completecovariance matrix and thus fully account forcorrelations and possible rank deficiencies inthe analysis. By selection of parameters to beestimated, and the methods used to eliminateunwanted parameters in the data sets, we canpartition the increments inχ2 between satelliteorbits, regional stations, and global trackingstations, thus allowing the origin of large χ2increments to be isolated.

(3) The full solution and covariance matrixfrom an analysis can be saved and then latercombined other data sets which themselvesmay be combinations. The advantage of thiscapability is that it reduces considerably thetime required to combine all data set together.

The typical fashion in which these features areused is to first combined the results fromindividual sessions within a GPS campaignlasting for several days to weeks. In thiscombination, satellite orbits may be integratedfor several days and treated as eitherdeterministic or stochastic parameters. Each ofthe field experiments are combined in thisfashion, and then in later processing only thecombined analyses are used. We often will thencombine multi-year GPS data into a singleanalysis. In this case, station velocities will beestimated and often variations in Earth rotationparameters are also included. A similarcombination can made for the VLBI data setand GPS data sets such as the global trackingnetwork.

A resultant velocity field in California wasobtained. This analysis uses 1,452 VLBlexperiments with 738,470 group delaymeasurements included and 263 GPS days ofdata with 5,437,265 double differenceionospheric-delay free carrier-phasemeasurements. In all, the positions andvelocities of 252 stations were estimated. Oneof the interesting features is that only the VLBIvelocity field information was needed to obtainan accurate GPS velocity field in California. Inthe standard velocity analysis, no VLBIpositional information was needed, and thuspossible problems with tying between theVLBI and GPS reference points avoided.

5.16 Antarctic Search forMeteorites (ANSMET) Project

Ralph Harvey, Bill Cassidy, and JohnSchutt, University of Pittsburgh

ANSMET (the Antarctic Search forMeteorites) has conducted field research inAntarctica since 1976 on a yearly based. In

57

general our work consists of detailed small-scale searches for meteorites on remote blue-ice areas of the East Antarctic ice sheet, whichare recovered and returned to the U.S. fordistribution to interested scientists throughoutthe world. One important facet of our researchis our continuing efforts at understanding howmeteorite concentrations develop in Antarctica.This involves detailed, fine scale mapping ofthe areas where the meteorites are found, and oflonger term studies of regional ice movementand history. During the early years of theproject, meteorite locations and surroundingbedrock and glacial features were surveyedusing theodolite and EDM techniques. Theslowness of these techniques and the harshweather of Antarctica made these effortslaborious to set up and difficult to complete. Inaddition, it required that each meteorite bevisited twice: once, when initially located, andagain when surveyed in from an establishedbase station.

During the past three seasons we have beenexperimenting with GPS as a way of replacingour antiquated methods for surveying meteoritelocations. We are equipped with four Magellan5000 handheld GPS receivers, which we use indifferential mode from established basestations to survey meteorite locations. Theseinstruments are limited to the use of carrierwave signal only, and cannot autonomously logall visible satellites. Thus our accuracy remainsrelatively limited (perhaps 10 meters at best),and we still must dedicate special days tosurveying, when one field party member mustremain in camp, logging a subset of theavailable satellites and relaying those satellitedesignations by radio to the remote units. Todate, the use of GPS has improved our ability tosurvey meteorite locations by giving us all-weather capability and improving our range (tothe limit of handheld radio communications).Our immediate goal for continued exploitationof GPS technology is to acquire (either on loanor by direct purchase) a low-end carrier phase

GPS base station capable of autonomoussatellite logging (something like an ASHTECRanger). Acquisition of such a unit would havea dramatic effect on our field work. With anautonomous data logger left running back incamp, we would have the ability to surveymeteorite locations immediately when they arefound and collected, eliminating a second tripto the meteorite and reducing the amount oftime currently required for each meteorite inhalf. Ultimately we hope to use more sensitivebase stations to directly measure small-scalemovement of the meteorite icefields.

5.17 Landslide Hazard andGPS

M.E. Jackson, P. Bodin, and E. Nel,University of Colorado

W. Savage, U.S. Geological Survey

GPS has been used to map the horizontalvelocity field of a rapidly moving earthflow toa greater spatial and temporal resolution thanavailable with conventional survey methods.Our work suggests that the most active portionof the Slumgullion slide is moving at 1.5 cm/day and a portion of the slide assumed to beinactive may be moving at 3 mm/day.

The Slumgullion earthflow in the San JuanMountains of southwestern Colorado, has beenmoving almost continuously for the last 300years. The slide consists of an inner, active flownestled within an inactive older slide. The oldslide dammed the Lake Fork of the GunnisonRiver approximately 700 yrs b.p. resulting inthe ponding of Lake San Cristobal. Traditionalcontrol surveys initiated over the last 20 yearssuggest that the upper portion of the active slideis moving at average velocities of ~0.5 cm/day,the middle or most active portion is moving at

58

1.6 cm/day, while the lower toe of the slide ismoving at 0.4 cm/day. The Slumgullionlandslide poses a significant risk to a smallresort community along the shore of Lake SanCristobal because houses are built on old slidematerial and the town is directly downslopefrom the rapidly moving younger slide.

The GPS survey was initiated in June, 1993 todetermine (1) if satellite geodesy could be usedto rapidly map the slide velocity field, (2) thespatial and temporal distribution of velocity,and (3) if the inactive slide is moving relative toa stable benchmark well off the slide. A total of7 GPS stations (baselines <5 km) wereinstalled, 5 on the active slide, 1 on the inactiveslide, and 1 on the stable portion of the slideheadscarp. Station benchmarks consist of 1 msections of capped steel pipe driven flush withthe surface. These benchmarks probably have along term stability of ±3 cm which is negligiblegiven the rapid velocities encountered on theslide. Relative coordinates for the networkwere carried from the Pietown fiducial trackingsite in northern New Mexico to a precision of 3,3.5, and 10 cm in the north, east, and upcomponents, respectively. Each station wasmeasured for at least two 4–6 hr sessions withmost stations being occupied for four 4p6 hrsessions. The baseline between station SEI1and GP01, located on the fastest movingportion of the active slide, was measured fornine consecutive 4–6 hour sessions.

The 1.5–1.2 cm/day velocities measured byGPS relative to a station on the inactive slideare similar to those estimated from traditionalsurvey techniques and long-term creep studiesbut were acquired in a fraction of the time.Moreover, on short time scales the GPS datasuggest the station velocities are constant.Although the spatial and temporal distributionof measurements are limited, it appears theinactive slide may be moving at ~2.5 mm/dayrelative to a stable station on the slideheadscarp. If the older slide is indeed moving,

the community built on slide material downvalley may be at greater risk than was oncethought.

5.18 Crustal Deformation atConvergent Plate Boundaries

James Kellogg, University of South Carolina

Global plate motion models describe therelative motion of the major tectonic platesaveraged over the last few million years. Thesemodels are based primarily on data fromdivergent and transcurrent margins, as theorientations of trench slip vectors tend to besystematically biased whenever the plateconvergence is oblique. The direction of slip intrench earthquakes tends to be between thedirection of plate motion and the normal to thetrench, suggesting active deformation withinthe overriding plate. The convergent boundaryzone between the Nazca, Caribbean, and SouthAmerican plates is particularly complex withactive seismicity and deformation occurringover a broad area.

In January 1988, scientists from over 25organizations in 13 countries established theCASA (Central and South America) network,the first major GPS network to measure crustaldeformations in the northern Andes and thewestern Caribbean. The CASA 1988experiment was the first civilian effortimplementing a global GPS satellite trackingnetwork. The repeatability of long baselines(400–1000 km) was improved by up to a factorof two on the horizontal vector baselinecomponents by using tracking stations in thePacific and Europe to supplement stations inNorth America. The results encouraged theestablishment of a permanent civilian globaltracking network in 1991.

59

Repeat measurements of the CASA network in1990 and 1991 have confirmed global platemotion model predictions of rapid subductionat the middle America (76 mm/yr), Ecuador (71mm/yr), and Columbia (54 mm/yr) trenches.However, intriguing differences occur betweenthe measurements and the global models at the95% confidence level. The northern Andeanmargin appears to be shearing northeastwardrelative to stable South America. Preliminaryresults also suggest Caribbean–North Andeanconvergence and an independent North Nazcaoceanic plate. Perhaps the most surprising ofall, the CASA results suggest the existence of arigid Panama–Costa Rican Microplate that ismoving northward relative to the stableCaribbean plate and colliding with the northernAndes. Planned additional occupations of theCASA GPS network will help resolvequestions of active slip positioning at complexconvergent margins.

5.19 The GPS FlightExperiment on Topex/

Poseidon

William G. Melbourne, Jet PropulsionLaboratory

TOPEX/Poseidon, a joint NASA/CNESmission to study ocean circulation bymeasurement of sea surface topography, waslaunched on 10 August 1992. After a decade ofpreparation, the GPS precise orbitdetermination (POD) experiment on TOPEX/Poseidon is now yielding definite results. Awide range of orbit consistency and accuracytests indicate that GPS is routinely providingsatellite altitude with an rms accuracy of about3 cm. Recent tests indicate better than 5 cmaccuracy in the along-track and cross-trackcomponents.

Scientific Rationale.The principal limitation ofpast satellite altimetry missions to study oceancirculation has been the error in determiningthe geocentric radial position of the altimeter,which was provided by POD with tracking datafrom ground-based Doppler or laser rangingsystems (SLR). The best SeaSat orbits, forexample, had a radial accuracy of about 40 cmrms. Although that can suffice for regional andocean variability studies, for global circulationa radial accuracy of 10 cm or better is needed.To reach 10 cm, the TOPEX/Poseidon Projectraised its planned orbit altitude from 800 to1336 km, reducing drag and gravityperturbations, and selected two operationalprecise tracking instruments—SLR and theCNES-sponsored DORIS Doppler system. Italso supported extensive development of theground segments of those tracking systems,and of spacecraft dynamic models. The projectalso chose to carry a GPS flight receiver as anexperiment and to support the implementationof a global tracking network.

The GPS Tracking System. The GPSexperiment on TOPEX/Poseidon wasconceived in the early 1980s by a group at JPLseeking alternatives to conventional Dopplerand SLR tracking. The motivation was simple:with the high precision of altimetry, oceanscience would benefit from centimeter levelorbit accuracies. But traditional PODtechniques, which depend on precise models ofsatellite forces to recover the orbit, are limitedby imperfections in the models, and theseeffects are exacerbated at lower altitudes. JPLturned to geometrical techniques, which areless sensitive to dynamical limitations, andsoon realized that the enveloping coveragegiven by GPS offered an almost ideal solution.By the mid-1980s, a strategy known as reduceddynamic tracking emerged that sought tocombine the best elements of dynamical andgeometrical positioning to minimize overallorbit error. This technique has been applied to

60

TOPEX/Poseidon to yield the above cited PODperformances.

The GPS tracking system consists of foursegments: the GPS constellation, the flightreceiver, a global network of GPS groundreceivers, and a central monitor, control andprocessing facility. The POD strategy requirescontinuous tracking of the visible GPSsatellites by ground and flight receivers. Datafrom all receivers are brought together andprocessed in a grand solution in which theTOPEX/Poseidon orbit, all GPS orbits,receiver and transmitter clock offsets, and otherparameters are estimated. Simultaneoussampling at all receivers (which may beachieved by later interpolation) eliminatescommon errors, such as clock dithering fromGPS selective availability. In the end, TOPEX/Poseidon position and velocity are determinedin a reference frame established by key sites inthe global GPS tracking network, known toabout 2 cm in the International TerrestrialReference Frame. The accuracy of the GPSsatellite ephemerides and terrestrial referenceframe transformation parameters are alsosignificantly improved over current resultsfrom the ground network by using TOPEX/Poseidon flight data.

The Future. Innovations in data processingstrategies and modeling that are planned forimplementation over the next year shouldimprove TOPEX/Poseidon POD accuracies bya factor of two or three. Although the GPSflight receiver on TOPEX/Poseidon has a fewlimitations (it only tracks six satellitesconcurrently and with a limited field of view),it is not unreasonable to expect that a GPStracking system on future remote sensingplatforms will provide 1 cm orbital accuracies.It is important to note that this accuracy islargely insensitive to orbital altitude over therange of 200 to 3000 km; thus, the orbitalaltitude of future altimetry missions may be

drastically lowered to save mission costswithout incurring POD penalties.

5.20 Use of GPS for TetheredSatellite PositionDetermination

Bjorn Johns and George W. Morgenthaler,University of Colorado

In the late 1990s, tethered satellites mayprovide an effective tool for probing the Earth’supper atmosphere. Recently, Martin Marietta,Alenia, the Italian Space Agency (ASI), andNASA completed the design and fabrication ofthe first Tethered Satellite System (TSS-l). Aproposed future Atmospheric VerificationMission (AVM) could follow TSS-l. While inan Earth orbit, the space shuttle would reel outthe sub-satellite up to 120 km toward the Earthinto a 90 km x 260 km elliptical orbit. In situdata collected at these altitudes would be veryvaluable in characterizing the lowerthermosphere and also for determining sub-satellite aerothermodynamics.

GPS is considered to be the best system todetermine the position of a tethered satellitewith respect to the space shuttle. A relativepositioning simulation was performed applyingthe position correction method to post-processed experiments, using only C/A codecorrelating receivers. The simulation providedsimultaneous GPS data over a 118 km baseline,designed to simulate the space shuttle–tetheredsatellite separation. After performingdifferential corrections, the data was passedthrough a low-pass filter which suppresses thehigher frequency components of the data. Sincethe tethered satellite dynamics are expected tobe of a low frequency, higher frequency data isprimarily from receiver noise.

61

The following 2σ errors were obtained:

The above results were obtained using TrimbleTANS C/A code correlating receivers. The dataexceeds by a large margin the minimumaccuracy requirements provided by MartinMarietta. Future work is proposed to determinethe specific receiver type and hardware bestsuited for an actual space based application,and develop a real-time system to determineboth position and velocity accuracies. Properfiltering schemes need to be evaluated, and theenvironmental effects on both the receiver andcommunication link operating at orbitalvelocities in the upper atmosphere must beaddressed.

5.21 Status of GPS/AcousticMeasurements of Seafloor

Strain Accumulation Acrossthe Cascadia Subduction Zone

G. H. Purcell, Jr. and L. E. Young, JetPropulsion Laboratory

F. N. Spiess, D E Boegeman and R. M.Lawhead, Scripps Institution of

Oceanography

H. Dragert, M. Schmidt, G. Jewsbury,Geological Survey of Canada

M. Lisowski, U. S. Geological Survey

D. C. DeMets, University of Wisconsin

Un-Filtered Filtered

Azimuth (deg) 0.00203 0.000463

Elevation (deg) 0.00577 0.00124

Range (m) 4.48 1.48

A long-term experiment was conducted using ahybrid GPS/acoustic system to determine strainrates across the Cascadia subduction zone. InMay–June of 1991, long-lived acoustic tran-sponders were installed on the sea floor on bothsides of the Cascadia subduction zone. Mea-surements from a surface buoy, equipped withthree GPS antennas and a precision acoustictransducer, located seafloor reference points inthe coordinate frame of land-based GPS receiv-ers. These measurements were repeated inSeptember, 1993, and two additional transpon-ders were set in place. Periodic measurementsover five to ten years are expected to yield sitevelocities with sub-cm/year accuracy, sufficientto improve estimates of fault locking depth andnet convergence velocity. Currently, determi-nations of these parameters rely on onshoregeodetic measurements and plate motion mod-els. This experiment also serves as anengineering test of the new GPS/acoustic sys-tem, and is expected to lead to substantialimprovements in experiment design and analy-sis techniques.

5.22 GPS in Volcanology:Monitoring of Icelandic

Volcanoes

Freysteinn Sigmundsson, University ofIceland

Volcanoes in Iceland deform at a rate of up toseveral cm per year during repose periods andcrustal movements on the meter scale occurduring eruptions. The deformation reflectspressure changes in the roots of the volcanoesthat can be caused by a number of processessuch as: magma movements, heating of magmafrom below, partial melting of country rock,and focusing of regional strain near magmachambers. Deformation monitoring providesimportant constraints on these and other

62

processes occurring at volcanoes. Theadvantages of GPS over conventional geodetictechniques are of great importance in volcanomonitoring. The simultaneous horizontal andvertical control from GPS is important, as bothtype of motion are prominent at volcanoes.Adverse weather conditions prevail at manyvolcanoes in Iceland and the almost weatherindependence of GPS facilitates measurementsgreatly. Rugged topography at volcanoesprevents line-of-sight between many controlpoints that can be measured with GPS but notwith a geodimeter. Regional GPSmeasurements have been conducted in Icelandsince 1986 as an international collaborativeproject. In addition to providing importantconstraints on plate boundary deformation,these measurements have also providedinformation on deformation of volcanoes,including a recording of magma chamberdeflation associated with the 1991 Heklaeruption in South Iceland. The small number ofGPS points near the Hekla volcano limited theinformation about the deformation.Recognizing the need for dense local networkson volcanoes, the Nordic VolcanologicalInstitute has now initiated a program to monitordense GPS networks on volcanoes in Iceland.In 1993 five such networks have beenmeasured, each consisting of 20–30 controlpoints with a typical spacing of a few km. Thenetworks will be partly or wholly remeasuredperiodically of intervals 1–2 years. Some goalsof the measurements are to: (1) monitorinflation and deflation of magma chambers andto observe possible precursors to eruptions, (2)search for anomalies for Mogi point sourcedisplacement fields at volcanoes, e.g.,displacement anomalies associated withcaldera faults or extension across fissureswarms transecting volcanoes, and (3) be ableto use GPS data to resolve possiblesimultaneous magma chamber deflation anddike injection associated with future eruptionsin Iceland. The local networks are connected tothe regional GPS network in Iceland, allowing

the monitoring of the interplay between local(volcanic, seismic) and regional (plateboundary) deformation.

5.23 GPS Geodesy andSeismology

Bob Smalley, Memphis State University

An application of combined GPS andseismology is the study of short- to medium-term (days to years), post-seismic deformation.One area to investigate is any relationshipbetween aftershock activity and a) co-seismicslip on the fault (obtained from seismic dataand, if it exists, geodetic, especially GPS data)or b) postseismic deformation in the near field(principally from GPS geodetic data). A greatearthquake on a strike slip fault (length ofrupture zone is typically much larger thanwidth) where slip information is available fromsurface rupture and seismology would providethe simplest and most well-constrained“experiment” of this type.

As modern VBB seismometer response (100–300 seconds) and GPS response, the timenecessary to obtain mm accuracy overrelatively short baselines (now at 300-600seconds) converge, one can combine a VBBseismometer with a GPS receiver, resulting inan instrument for installation in the near fieldwith an effective frequency response offractions of a second to years, at least in thehorizontal. A combined network ofseismometer-GPS instruments could obtainlong records of spatially dense post-seismicdeformation. Practically, development of aformal relationship with the IRIS program tocombine GPS and seismic aftershockmonitoring efforts would be beneficial to bothgroups, both scientifically and logistically (forexample, the PASSCAL program is presently

63

acquiring GPS receivers to provide a time baseat each site).

Two applications that would benefit from real-time continuous deformation monitoring arevolcanology, specifically assisting in predictionof eruptions based on the active deformationimmediately preceding an eruption, and real-time earthquake warning systems. Both ofthese applications would depend on a new formof real-time differential that would provide thepositions of all the receivers in the network inreal time. This real-time differential operationwould operate in a manner similar to a real-time seismic network in that the GPS datawould be collected in a central location andprocessed sufficiently rapidly to be useful forreal-time response. This depends on certaincomponents beyond what is usually meant byreal-time differential. First it will require theuse of phase-plus-code receivers and a high-quality, inexpensive communication method tocollect the pseudo-range and phase data in realtime at a central location where networksolutions can be automatically processedrapidly enough that they are useful for theshort-term prediction. The technical questionis: How fast could one unequivocallydetermine that a movement of x cm hasoccurred on a fault, where x is a threshold fordeclaring an event has occurred. For this towork, we would have to be able to detect aposition change of say 20 cm for a magnitude6.5 event within one or two seconds. Largerevents would produce larger slips, but the timefor detection, determined by the time for theseismic waves to arrive at the threatenedpopulation center, would remain the same.Incorporation of such a GPS component intothe early warning seismic networks being builtin several places (SCEC, Japan, and Taiwan)would add significant robustness to the eventdeclaration process. This technology could alsobe applied to the study of deformation ofvolcanoes before eruptions, landslides, slumps,glacier surges, and other geologic phenomena

where there is motion on the order of mm to mper day. The volcano application is interestingbecause the surface deformation can be used toestimate that rate at which magma is movingwithin a volcano. Combining this data withseismic data, which can be used to delimit themagma chamber, one can obtain a betterunderstanding of the plumbing of volcanoes.These applications depend on a significantdecrease in the price of GPS receivers. As GPStechnology enters the consumer market, pricesshould fall significantly, making networks oflarge numbers of continuously operatingreceivers possible.

5.24 Microplate VersusContinuum Descriptions of

Active Tectonic Deformation

Wayne Thatcher, U.S. Geological Survey

Whether deformation on continents is moreaccurately described by the motions of a fewsmall rigid plates or by quasi-continuous flowhas important implications for lithosphericdynamics, fault mechanics, and earthquakehazard assessment. Actively deforming regionsof the western U.S., central Asia, Japan, andNew Zealand show features that agree for bothstyles of movement, but new observations arenecessary to determine which is mostappropriate.

Geologic, geodetic, seismic and paleomagneticmeasurements tend to sample complementaryaspects of the deformation field, so an inte-grated observation program can utilize thestrengths of each method and overcome theirseparate spatial or temporal biases. Providedthe total relative motion across each region isknown and the distribution of active faults iswell mapped, determination of fault slip ratescan provide potentially decisive constraints.

64

Reconnaissance geological studies supply use-ful estimates but precise values depend upondetailed intensive investigation of individualsites. Geodetic survey measurements can deter-mine the spatial pattern of contemporarymovements and extract slip rate information,but the sometimes elusive effects of cyclic elas-tic strain buildup and relief must be accountedfor in relating current movements to the long-term deformation pattern. Earthquake cata-logues can be applied to determine seismicstrain rates and relative velocities, but must beaveraged over large regions and are usuallylimited by the inadequate duration of historicalor instrumental seismicity catalogues. Paleo-magnetic determinations of vertical axisrotations provide estimates of block rotationrates, but are often locally variable and aver-aged over many Ma.

Which of the two descriptions of continentaltectonics is more nearly correct depends on thelocal rheological stratification of the lithos-phere, especially the strength and thickness ofthe elastic crust relative to the ductile lithos-phere, and dynamical models can providecontrasting forecasts of observable featureswith testable consequences. Within a givenregion, earthquake hazard assessment method-ology should differ considerably depending onwhether deformation is concentrated on a fewmajor faults or distributed widely on manyequally important structures.

5.25 GPS Detects VerticalSurface DisplacementsCaused by Atmospheric

Pressure Loading

Tonie vanDam, NASA/Goddard Space FlightCenter

Geoffrey Blewitt and Michael B. Heflin, JetPropulsion Laboratory

Tide gauge measurements only give sea-levelvariations relative to the crust, whereas thecrust itself may be deforming vertically due tovarious factors. If vertical crustal deformationis ignored, then erroneous conclusions may bedrawn with regard to the causes and predictionsof global sea-level change. However, if tidegauge measurements were sufficiently well tiedto a stable Earth-centered reference frameusing space geodetic techniques, sea-levelchange would be observed in an absolute sense.We demonstrate that GPS is an excellentcandidate for addressing the geodetic aspects ofglobal sea-level studies by looking at a physicalphenomenon that should have a detectable andcalculable effect on station heights. Temporalvariations on the geographic distribution ofatmospheric surface pressure deforms thesurface of the Earth at the several millimeterlevel. Starting with daily no-fiducial solutionsof the global GPS network, we show a newmethod of deriving a time series of individualstation heights (an improvement over the moreusual geodetic baseline analysis). Atmosphericloading is modeled using Farrell’s GreensFunctions, and using pressure values from theNCAR database on 2.5¡ grid. A sample plottaken from Yellowknife, Canada, showsvertical displacement where both the modeledand GPS-estimated station heights aresmoothed using a seven-day sliding window.Out of the seven stations investigated so far, sixhave decreased rms height distribution aftercalibrating for modeled atmospheric heightloading. One of the questions currently underinvestigation is how to model the heightvariations of coastal sites (which are importantfor tide-gauge measurements). For coastalsites, the crustal response is a function of theextent of the inverted barometer effect, inwhich the ocean tends to act as a high-pass filter(t = several days) in transmitting atmosphericpressure to the crust.

65

5.26 Ionospheric MonitoringUsing GPS

Anonymous, Universitat Hannover

GPS satellites provide signals on twofrequencies in order to enable the user tocorrect for signal delays due to ionosphericrefraction. The difference in propagation timebetween the L1 and L2 GPS signal’s code orphase contains information on the electroncontent integrated along the signal path. Thisinformation cannot only be used to observe thetotal ionospheric electron content, but also todetect and study ionospheric disturbances—e.g., phase scintillations caused by small-scaleirregularities in the electron content of theionosphere. These small-scale irregularitiesmay have severe effects on radio propagation.Precise GPS applications are affected in such away that the continuous satellite tracking isfrequently interrupted. The resulting carrierphase cycle slips need to be fixed in the post-processing. Under disturbed conditions, eachfixing of a cycle slip requires an increasedeffort due to the present irregularities.Therefore, ionospheric scintillation should beavoided for precise GPS applications. In orderto be able to predict and avoid scintillationoccurrence, detailed observation and long-termstudies have to be performed on small-scaleirregularities in the ionosphere. Here, GPSplays an important role. Dual-frequency GPSphase measurements allow a detaileddescription of the occurrence of phasescintillation. At the Institut fur Erdmessung(Institute of Geodesy) of the UniversitatHannover in Hannover, Germany, we havedeveloped algorithms which enable us to detectphase scintillations in GPS phase data.

5.27 Atmospheric Noise inAirborne Gravimetry

Jim Johnson, Christian Rocken, FredSolheim, Teresa Van Hove, and Randolph

Ware, University Navstar Consortium

DGPS measurements are being widely used inairborne gravimetry to correct for aircraftvertical acceleration errors. Accurate airbornegravimetry has scientific as well as mineral andpetroleum exploration applications. However,variations in GPS signal delay resulting fromatmospheric water vapor, or wet delay,introduces apparent accelerations that limit theaccuracy of airborne gravimetry. In order toevaluate GPS wet delay error, we observed wetdelay using a water vapor radiometer (WVR)and compared apparent vertical accelerationsresulting from GPS wet delay to GPS multipathand receiver noise.

We operated Trimble 4000 SSE 8-channeldual-frequency C/A and P-code receivers atboth ends of a 10-meter baseline on the roof ofa laboratory in Boulder, Colorado. Data wererecorded at 1-second intervals. Each GPSantenna was mounted in a dished, l-meterdiameter slab of microwave absorber, 0.5-meter thick. The absorber reduced observedcarrier multipath amplitude by a factor of two.For GPS data analysis we used the Kinematicand Rapid Static Software (KARS) developedby G. L. Mader at the National GeodeticSurvey. One-second vertical GPS solutionswere calculated. Fifty-second filter solutionsand apparent vertical accelerations calculatedas the second time derivative of the filteredsolutions were also obtained. The apparentvertical acceleration resulting from GPSmultipath and receiver noise is 1.5 mgal rms.Multipath clearly dominates. We anticipate thatthe multipath signal can be further reduced bya factor of three or more through the use ofchoke ring antennas, internal suppression in the

66

receiver, or through mapping and subsequentremoval in processing.

We measured zenith wet delay with aRadiometrics WVR (manufactured byRadiometrics Inc., Boulder, CO). The WVRobserved sky brightness temperatures at 23.8and 31.4 GHz and converted them into wetdelay. The WVR is calibrated using an ambientblackbody target, a noise diode, and a tippingcurve. The sampling interval was one minute.Wind speed during the WVR observations wastypically several m/sec. In order toapproximate observations from a fixed-wingaircraft, we compressed the time scale by afactor of 24. Assuming a wind speed of 2 m/secat the WVR sites this approximates an airspeedof 100 mph. Fifty-second filter solutions andapparent vertical accelerations calculated as thesecond time derivative of the filtered solutionswere obtained. The apparent verticalacceleration resulting from variations in WVR-measured GPS wet delay is 0.5 mgal rms.

The climate in Colorado is relatively dry, witha typical zenith wet delay of 5 to 15 cm. Inmore humid regions zenith wet delay may be aslarge as 40 cm, and the resulting apparentvertical accelerations resulting from wet delayvariability may be larger than the 0.5 mgal seenin Colorado. If the GPS multipath noise issignificantly reduced, GPS wet delay willbecome the dominant error source. To correctfor GPS wet delay, WVRs could be operated atthe GPS ground station and on the airborneplatform. In order to avoid errors resulting fromchanges in WVR observation angle aboard theaircraft, attitude correction would be required.With a combination of multipath reduction andwet delay correction, fixed-wing airbornegravimetry may be able to achieve 1 mgalresolution over distances of 1 km. Thiscapability would be valuable for scientificresearch applications, and also for rapid, cost-effective exploration for minerals andpetroleum.

5.28 Improved GPS VerticalSurveying

Chris Alber, Jim Johnson, ChristianRocken, Fred Solheim, Teresa Van Hove,and Randolph Ware, University Navstar

Consortium

GPS signal delay caused by atmospheric watervapor, or wet delay, is a major source of error inhigh-accuracy vertical GPS surveying.Scientists at UNAVCO have been working toimprove vertical GPS surveying accuracy. Thiswork has been supported by grants from theNational Science Foundation and the U.S.Geological Survey. UNAVCO scientistsrecently reported results from a GPS surveyusing water vapor radiometers (WVRs) pointedtoward GPS satellites. The WVRs measureintegrated water vapor which can be convertedto wet delay for correction of GPSobservations. Agreement of wet delaysmeasured by two WVRs separated by 30 m is 3mm rms, indicating the ability to correct forGPS wet delay at this level of accuracy. Thetripod-mounted WVRs were designed for usein GPS surveying, and were loaned byRadiometrics, Inc., of Boulder, Colorado. GPSreceivers and WVRs collected data at both endsof a 50-km baseline from Boulder to Platteville,Colorado. Vertical repeatability was 2.6 mmwith pointed WVR corrections. Precision wasdegraded by a factor of two using Kalman filteror least squares methods commonly used tocorrect for wet delay. The pointed WVR resultsare more than a factor of three better thanprevious results. and a factor of two better thanKalman or least squares results. Improvedvertical GPS surveying has applications ingeodesy, earthquake studies, volcanology, sealevel measurements, airborne and ground-based gravimetry, atmospheric sensing, andGPS orbit determination.

67

5.29 Sensing the Atmospherewith an Orbiting GPS

Receiver

Mike Exner, Christian Rocken, BillSchreiner, Fred Solheim, Chuck Spaur, and

Randolph Ware, University NavstarConsortium

Mikhail Gorbunov and Sergei Sokolovsky,Institute of Atmospheric Physics (Russia)

Ken Hardy, Lockheed

Ben Herman, University of Arizona

Tom Meehan and Bill Melbourne, JetPropulsion Laboratory

The National Science Foundation, the NationalOceanic and Atmospheric Administration, andthe Federal Aviation Administration are jointlysponsoring the $3 million proof-of conceptGPS/MET program. It is being conducted byscientists at the University Corporation forAtmospheric Research (UCAR), UNAVCO,the National Center for Atmospheric Research(NCAR), JPL, and the University of Arizona.The goal of the program is to demonstrateactive limb sounding of the atmosphere by anorbiting GPS receiver. The private sector is alsoparticipating in GPS/MET. Orbital SciencesCorporation is providing launch and spacecraftservices under its innovative “data sale”contractual agreement with UCAR. AllenOsborne Associates, Inc. is supportingdevelopment of special receiver equipmentneeded for GPS sounding.

Phase shifts in GPS signals transecting theatmosphere will be detected by the orbitingGPS receiver. These data will be converted intovertical refractivity, temperature, and humidityprofiles. One-km vertical resolution from 50km to the surface, and sub-Kelvin temperatureaccuracy in the lower stratosphere are

expected. The GPS/MET data set will be madeavailable for weather, climate, and otherresearch. Currently, weather forecasting reliesheavily on radiosonde data obtained from 0:00and 12:00 Z launches from 600 sites. Most ofthe sonde data are collected over land in thenorthern hemisphere. A single orbiting GPSreceiver can observe 600 longitudinallydistributed soundings per day (latitudinaldistribution depends on orbit inclination). Mostof the GPS soundings will be obtained overoceans where sonde data are sparse. Improvedweather forecasts over oceans may have valuefor commercial aviation.

5.30 GPS Sensing ofAtmospheric Water Vapor

Chris Alber, Jim Johnson, ChristianRocken, Teresa Van Hove, Fred Solheim,and Randolph Ware, University Navstar

Consortium

GPS signals are delayed by atmospheric watervapor. This wet delay can be measured usingGPS receivers and analysis software developedfor high accuracy geodetic surveying. Wetdelay converts easily into perceptible watervapor (PWV). More widespread PWV datacould be used to improve precipitationforecasts. Currently, weather forecasts arebased on radiosonde data collected twice dailyfrom 600 sites worldwide. Soon, GPS-sensedPWV measured by permanent GPS arrays maybe used to improve weather forecasts.

GPS sensing of PWV was recently reported byscientists at UNAVCO in Boulder, Colorado.UNAVCO is a national facility that assistsuniversity investigators using GPS technologyfor geosciences research. UNAVCO scientistswere working to improve the vertical accuracyof GPS high-accuracy geodetic surveying.

68

They found that half-hour GPS and water vaporradiometer (WVR)-sensed PWV agreed tobetter than 1 mm for several weeks ofobservations. Portable high-accuracy WVRswere provided for this experiment byRadiometrics, Inc., Boulder, Colorado.

Scientists from UNAVCO and North CarolinaState University are exploring the use of GPStechnology in weather forecasting. In June1993 they gathered GPS and WVR data fromNOAA Wind Profiler sites and a Department ofEnergy Atmospheric Radiation Monitoring(ARM) site in Oklahoma and from sites inColorado. Improvement in weather forecastingusing GPS-sensed PWV from these sites willbe explored. Scientists from NOAA’s ForecastSystems Laboratory in Boulder, Colorado, planto collect GPS-sensed PWV data and evaluateits use for weather forecasting. They willoccupy Wind Profiler and ARM sites in Kansasand Oklahoma. UNAVCO scientists haveagreed to cooperate during the implementationand help analyze the GPS data and convert itinto PWV.

Large numbers of high-accuracy GPS receiversare now in continuous operation worldwide forgeodesy, surveying, orbit determination, andother applications. Data from most of thesesites could be accessed in near real time via theInternet or telephone. This presents a new, cost-effective opportunity to improve precipitationforecasts using GPS sensing of water vapor.

5.31 GPS in Antarctica

Ian Whillans, Ohio State University

The stop-and-go GPS technique was used todetermine horizontal strains and relativevertical velocities of markers in the WestAntarctic ice sheet with a precision and scopenever achieved before. The strain rates were

mathematically inverted into stresses and usedin a force budget analysis to describe thestresses acting at depth, the objective being tounderstand the flow of the giant anomalous icestreams. The vertical velocities were used todemonstrate that most of the flow formsstanding waves, presumably associated withbasal disturbances. Other surface features aremigrating against glacier flow, an astonishingand unexplained result.

GPS is also being used to determine the rate ofthickening or thinning of selected sites on theAntarctic ice sheets. Special considerationsmust be taken to allow for snow compactionand ordinary glacier motion. The success of themethod relies on both the long-distancecapability and precision of GPS. The resultswill be “fiducial” sites for the calibration andinterpretation of repeat satellite or aircraft radaror laser altimetry.

In the past ice velocities at remote sites weredetermined by repeat Doppler satellitetracking. A TRANSIT receiver would track for24 hours at each site in each of two years. Thisentailed 2 aircraft visits each year to deploy andrecover the receiver. Precision positions couldnot be determined after the field season ended.Now we have cut field costs to less than half.Sufficient data are acquired with 20 minutes oftracking, while the aircraft waits. Moreover,relative positions can be calculated at the tentcamp later that day, and problems identifiedwhile there is still an opportunity to repeat thework.

The advent of low-noise receivers and full L1and L2 capability will enable important newprojects to be undertaken at reasonable cost.For example: (1) Quick determination of thepoint mass-balance of the ice sheets. It willenable the contribution of the current ice sheetsto the present-day rise in sea level to be reliablyassessed. (2) Determination of the crustalrebound rate toward isostasy to assess the size

69

of the former ice sheets. This will bear on therole of various ice sheets and the cause of thegreat glaciations. (3) Airborne repeat GPScontrolled laser altimetry to monitor ice sheetgeometry. (4) Monitoring the inflation anddeflation of volcanoes and other magmachambers (e.g. Rebus). (5) Intracontinentalstrain due to plate-tectonic action. (6)Intracontinental flexure due to changing iceload and mantle motion or phase changes. (7)Search for episodic seafloor spreading rates. (8)Refinement of techniques for ice strain and icevelocity determination.

5.32 Fast Technique forComputing Precise Aircraft

Acceleration from GPS Phase

T. P. Eunuch and C. A. Raymond, JetPropulsion Laboratory

R. E. Bell, Lamont Doherty Earth Observatory

D. D. Blankenship, University of Texas

Experience with GPS in aerogravity surveyshas revealed two fundamental facts: (1) theaircraft vertical acceleration may be accuratelyextracted from differential GPS carrier phasemeasurements, and (2) the extraction ofvertical acceleration by traditional GPSmethods can be costly and time-consumingwhen it relies on algorithms devised for staticpositioning rather than airborne gravimetry.Here we present a simple and robust techniquewhich avoids full phase connection andambiguity resolution, which are not requiredfor precise acceleration measurements.

GPS has been widely used in aerogeophysicalsurveys. Airborne gravimetry requiresmoderately accurate position and velocity andhighly accurate vertical acceleration. For

1 mgal gravimetry tracking, requirements are:height above the geoid to <3 m; lateral velocityto 5-30 cm/s; and vertical acceleration to <1mgal. Initially, GPS was used only to determineposition and velocity, though more recently ithas been adopted to recover verticalacceleration as well. Traditional phase-basedGPS analysis techniques have required thatphase lock be maintained throughout the flightand that all integer cycle phase ambiguitiesbetween flight and ground receivers beresolved.

While in principle that approach can achieveposition accuracies of a few centimeters (farbetter than needed), in practice it has provendifficult to meet the conditions required forsuccess. In the usual analysis technique, theentire flight, which may last 6-10 hours, mustbe processed as one continuous event. Ideally,the flight receiver would maintain continuouslock on all visible GPS satellites, with no cycleslips. In practice this is improbable since,during turns, satellite visibility may be blockedor large bursts of multipath may cause cycleslips. In addition, if a satellite rises and setsduring the flight its ambiguities may beunresolvable. If the data set is processedindependently in both directions (as if the flighthad been flown forward and backward),unflawed processing would yield an identicaltrajectory both ways. This in fact is standardcheck. In reality, because of remaining phasebreaks and faulty ambiguity resolution, towardand backward processing give typical altitudediscrepancies of several meters. The irony isthat data problems that cause thosediscrepancies almost invariably occur duringturns or transits, when the data are of nointerest.

The technique described here processes onlythe straight segments, each independently, in anautomatic and almost foolproof procedure. Theresult is a vertical acceleration measurementthat may actually be more accurate than that

70

obtained with current techniques—and only afew tenths of milligal off the ideal.

Carrier phase is highly precise, but has anunknown phase bias. Pseudorange is about 100times less precise, but is absolute—that is, ithas no bias. Apart from the bias in carrierphase, the two data types measure the samequantity: the range plus the time offset betweenthe GPS satellite and the receiver. This allowsus to estimate the bias of the phasemeasurement by comparing phase andpseudorange measurements made at the sametime. With phase and pseudorangemeasurements at one time point, our estimateof the phase bias is just the difference betweenthe two and it will have the precision of 1-2 m.If we have phase and pseudorange at many timepoints, and phase continuity is maintained overthe interval so that the bias is common to allphase points, then we can estimate the bias byaveraging the phase-minus-pseudorangedifferences over the full interval. Thiseffectively slides a phase curve to the mean ofthe phase of the unbiased pseudorange curve.Note that the phase biases are estimatedseparately as real-valued parameters for eachreceiver-satellite pair: No effort is made todetermine the integer-cycle ambiguity in thedifferenced observables between flight andground receivers.

The typical measurement interval for airbornesurveys is 1 second, which yields 1200measurements in 20 minutes. If the errors on allmeasurements were independent, a 20-minuteaverage would reduce the error on the biasestimate by more than a factor of 30, to a fewcentimeters. In reality, because of the lowfrequency content of the multipath error, thetypical error reduction for a 20-minute averageis more like a factor of 5 or 10, giving a typicalbias error of 20 cm. Taking into accountobserving geometry (PDOP, or positiondilution of precision) this becomes a bias in thealtitude measurement of typically 60-80 cm.

which translates into a bias of about 0.2 mgal inaccelerations computed over the interval. Thishas been confirmed in limited tests with GPSflight data collected in Antarctica.

This technique still requires phase continuityduring the (comparatively short) straight lineflight segments over which the averages arecomputed. Fortunately, phase breaks areextremely rare under such steady conditionswith the highest quality GPS receivers. Byavoiding the problematical turns, keeping thedata segments relatively short, and dispensingwith troublesome cycle ambiguity resolution,processing headaches are removed. A programto estimate the bias averages for a full 6-hourflight will execute in a few minutes or less, andthe procedure can be easily automated withrobust quality checks. Moreover, there arefurther computational gains that come fromdealing with 20–30 minute data sets, ratherthan a full 6-hour set. This technique will workwell with the data from any high-quality, dual-frequency GPS receiver that produces bothphase and pseudorange data, and thus can beused to improve results from existing data sets.

6 Technology

This section provides some technical back-ground on the Global Positioning System: howit works, the sources of error, and the accuracyrequirements.

6.1 GPS Methodology

6.1.1 Introduction

In this section the various methods, theaccuracy, and the applications of GPStechniques are discussed. The discussionranges from the applications of staticpositioning, to dynamic positioning, toapplications of the ancillary products that are

71

produced in GPS analyses (such as estimates ofwater vapor delays). The types of GPSreceivers available are also discussed althoughthe discussion will concentrate on the high-accuracy applications of GPS. First, the basicapplications of GPS and the types of receiversavailable are reviewed.

The basic observables of a GPS receiver are thepseudo-ranges (defined as the difference intime between the transmission and reception ofa signal as determined by the clocks in thesatellite and receiver) to the satellites visible,and optionally the phase of the GPS receiversignal relative to a local oscillator phase. Thislatter quantity is often referred to as the carrierbeat phase. The signals from the GPS satellitesare transmitted on two frequencies: L1(1.57542 GHz,λ=190 mm) and L2 (1.2276GHz, λ=244 mm). Two pseudo-random codesare written on the L1 signal using p phasechanges. These two codes, the C/A (courseacquisition) code and the P (precise) code, aremodulated on the carrier signal with differentamplitudes, with the C/A code being thestrongest. The modulation rate for the C/A codeis 1.023 MHz, i.e., the code may change phaseat about 1 msec intervals; for the P-code themodulation rate is 10.23 MHz. The C/A codealso repeats at about 1 msec intervals where asthe P-code does not repeat for a full week.When the GPS satellites are in Anti-Spoofing(AS) mode, there is an additional modulation ofthe P-code by the so-called W-code. The W-code modulates the P-code at a rate of about 20KHz. Either the C/A code or the P-code (withW-code modulation when AS is on), but notboth, are written on the L2 carrier signal.Nearly always is the P-code written on the L2channel.

In the simplest of the GPS receivers, only thepseudo-range on the L1 signal is measured bycross correlating the C/A generated in thereceiver with the incoming signals. Because ofthe weaker modulation of the P-code on L1, the

existence of P-code can be ignored during thiscorrelation at the cost of a slight increase in thenoise of the measurement. Within this class ofreceivers are navigation receivers and hand-held receivers. The accuracy of thepseudorange measurements obtained from thisclass of receiver varies greatly, but is on theorder of 100 m. Whenever four or moresatellites are visible, these pseudorangemeasurements can be used to determine theposition of the receiver and the differencebetween the receiver clock time and GPS time.These time and position determinations arestrongly affected by Selective Availability(SA), which primarily is implemented by“dithering” the fundamental oscillator on thesatellite by amounts which introduce timingerrors of up to about 0.2 msec (at the currentlevel of SA). The effects of SA can be removedalmost totally through the use of differentialtechniques, either by real-time transmission ofrange corrections or by postprocessing data andusing the difference between stations of therange measurements.

The most accurate applications of GPS usemeasurements of the phase of the reconstructedcarrier (i.e., the carrier signal after the p phasechanges introduced by the codes are removed).In these accurate applications, the phase andpossibly the range at the two frequenciestransmitted by GPS are used. Instruments thatmeasure phase come in several varieties. Earlymodel geodetic quality GPS receivers (e.g., theTI4100 circa 1980) made phase and rangemeasurements at both L1 and L2 assumingknowledge of both the C/A and P codes.However, uncertainty about civilian access tothe codes in the future prompted thedevelopment of “code-less” receivers thatcould measure phase (or half-cycle phase)without any knowledge of the codes other thanthat the encoding scheme was bi-phase (e.g.,Macrometer). These receivers had practicallimitations because they could not decode thesatellite ephemeris measurements and could

72

not re-synchronize their clocks because theymade no range measurements at all. By themid-1980s it was clear that the access to the C/A code would not be denied and receivers weredeveloped (e.g., MiniMac, Trimble SST, andAshtech II) that used the C/A code to make L1range and phase measurements and used one ofthe code-less schemes to make measurementsof the phase. By the late 1980s it was clear thatit would be some time before access to the P-code would be denied (effectively bysuperimposing on it the W-code) and aresurgence of instruments that would use the P-code to measure phase and range at L1 and L2appeared. These latter generation instruments(e.g., the Rogue, Turbo Rogue, Trimble SSA,and Ashtech IIP) could revert to codelesstracking of L1 and L2 if AS was turned on. Thetechniques used to recover L2 phase and rangewere refined in this generation of instrument.Although not on the market yet, instrumentsthat could take advantage of the lowmodulation rate of the W-code are now beingdeveloped, and these instruments offer asignificant signal-to-noise (SNR) advantage forL2 measurements over those currentlyavailable. All modern geodetic quality GPSreceivers appear to be able to measure phasewith an rms error of < 1 mm under good SNRconditions and range with an rms error ofbetween 0.05 and 1 m, with most instrumentsoperating in the 0.1–0.2 m range.

6.1.2 Static Receiver

The most accurate applications of GPS to datehave been those that have used GPS to monitortectonic deformations and measure Earthrotation variations. There are several types ofdeployment of receivers used in staticapplications, with the choice dependingprimarily on accuracy requirements andlogistics. The most accurate long-term GPSresults are likely to be obtained frompermanent deployment of GPS receivers suchas those currently in operation associated with

the International GPS Service (IGS), whichbecame a permanent service in January, 1994.There are currently 45–50 stations around theworld in continuous operation and thesestations report their data to several datacollection facilities around the world with afew day lag time. Averaged over a week or so,the positions of these stations are currentlydetermined relative to each other withaccuracies and precisions of about 10 mm inthe horizontal components and 20-30 mm inthe vertical component. The continuouscollection of data will allow the models used inthe GPS analyses to be improved and in allexpectation much better results will beobtained in the future and from the existingdata through reprocessing (see Error Budgetbelow).

For many applications the permanent operationof stations is not possible and in these cases,campaign-style GPS occupations are made. Inthis scenario, receivers are located overgeodetic monuments for a relatively shortperiod of time (usually 8 hours to several days)and then moved to other locations so that amuch greater density of points may bemeasured over a campaign lasting for severaldays to several weeks. A large number of suchcampaigns have been carried out with many ofthem being located in the western UnitedStates, but a large number also being carriedout in various regions around the world.Similar accuracies are obtained in thesecampaigns as in the permanently deployedstations, although since the intersite distancesare usually much smaller for campaign-stylemeasurements, the relative accuracy can bevery good and often as low as a fewmillimeters. For reasons to be discussed in theinfrastructure section, these campaign-stylemeasurements are becoming increasinglyaccurate.

73

6.1.3 GPS Ocean Applications

GPS oceanographic applications have beendiscussed in detail elsewhere in this report. Inthis section the accuracy of GPS solutions forthese ocean applications will be discussed.These applications include ship navigation andheading, buoy location, monitoring of iceextent and location, and monitoring of thevertical position of tide gauges. Accuracyrequirements for ship locations are dependenton the application. General navigationrequirements can usually be satisfied byposition accuracy of a few tens to a fewhundred of meters. There are applicationswhere a few meters or better ship position isdesired. Examples of this are harbor and rivernavigation, positioning of a ship to determineseafloor baselines, and ship and buoypositioning to determine the position of anacoustically tracked probe as it descendsthrough the water column. In this last example,the position history of the probe relative to aship and buoy yields the velocity of oceancurrents with depth.

Achievable position accuracy for ships canvary from a few tens of meters to a few hundredmeters for direct application of GPS,depending on whether or not SA is activated.Dif ferential GPS can provide ship positions toa few meters or better in a post-processingmode even if SA is activated. Differential GPScan be used in near real time if a system wereavailable to solve for ephemeris and clockerrors for each GPS satellite and broadcast thisinformation to the user. Wide area differentialGPS (WADGPS) is a scheme that would allowsuitably equipped users to perform differentialGPS solutions for their positions in real timeover a much larger region.

The WADGPS network comprises a masterstation, a number of local monitor stations, anda communications link. The local monitorstations report data from all GPS spacecraft in

view to the master station. The master stationestimates ionospheric time delay parameters,plus the satellite ephemeris and clock errors.These corrections are then transmitted to theuser via any convenient communications link.A test of this concept is scheduled by the FAAfor the U.S., but the concept could be extendedworldwide with a few master stations and a fewhundred monitor stations. Such a network hasthe potential to provide real-time navigationcapability for land, air, and marine applicationsof a few meters worldwide.

Ship heading can be determined to better thanone degree by the use of two GPS receiverslocated on a few meter baseline. However, shipheading accuracy generally will be limited bythe multipath environment which can varysignificantly depending on antenna location ona given ship.

While for most applications vertical errors inship location are not important, there are manybuoy applications where the vertical is themajor component of interest. This includesmeasurement of sea surface height formonitoring of sea level, measuring sea surfaceslope to yield gravity vertical deflections at seafor a more accurate marine geoid, or forcalibration of satellite altimeter rangemeasurements. Differential GPS can provideocean buoy position accuracy of a fewcentimeters to a few meters depending on datatype and the baseline length to the fiducial site.For baseline lengths less than 100 km, and in apost-processing mode, kinematic buoy positioncan be determined to about a centimeter in thehorizontal and about two centimeters in thevertical using carrier phase data in a differentialmode.

Major sources of error for the determination ofthe vertical are tropospheric refraction,multipath, and errors in the vertical position ofthe fiducial station. Path delay errors and errorsin the vertical station location map almost one

74

for one into vertical errors in the buoy position.Tropospheric delay can be handled byestimation, by the use of water vaporradiometers, or by using radiosonde dataacquired in the vicinity of the GPS receiversites. The most convenient method of handlingtroposphere refraction is by estimation of thepath delay. This also is probably the mostaccurate method, although recent water vaporradiometer results have been encouraging.

The use of GPS buoys has been shown to yieldexcellent results for calibrating the altimeterrange measurement on the TOPEX/Poseidonsatellite. For an experiment off the Californiacoast at the Texaco offshore platform, Harvest,agreement between a GPS buoy and tidegauges for ocean surface height above thereference ellipsoid was approximately onecentimeter. The TOPEX/Poseidon project alsohas installed a GPS receiver on the Harvestplatform. This receiver has been used todetermined the vertical position of the Harvestplatform to better than one centimeter rms.Knowledge of the vertical component of theplatform location to better than a centimeter iscrucial to the accurate calibration of theTOPEX/Poseidon altimeter height bias.

The above comments regarding accuracycapability for ship and buoy locations alsoapply to ice monitoring although someproblems may be more difficult such as thoseassociated with coverage of the GPSconstellation—i. e. no high latitude satellitesand the ionospheric correction. Accuracycapability of differential GPS for iceapplications therefore should range in theneighborhood of a few centimeters to a fewmeters depending on atmospheric conditions,GPS data type, and receiver location relative tofiducial sites.

6.2 Ancillary Data Output

There are three primary outputs of GPS that canbe used by others: (1) timing, (2) ionosphericdelay calibrations and thus estimates of theTotal Electron Content (TEC) of theionosphere, and (3) atmospheric delays. Directtiming applications are severely compromisedby SA which effectively limits the quality ofreal-time output to time to 0.2 msec. When SAis not on, time tags accurate to severalnanoseconds can be obtained from GPS.

Accurate estimates of the ionospheric delay aredifficult to obtain from GPS because ofinterchannel biases between the L1 and L2frequency channels and the differential delay inthe satellites between the L1 and L2 channels.The former can be obtained from carefulground receiver calibration but will always bedifficult for precise phase measurements due tothe unknown number of cycles. The latter iscalibrated in the GPS satellites before they arelaunched but the exact values when thesatellites are in orbit are not known. However,GPS does provide an excellent method forstudying the time variable behavior of theionosphere, with the largest complication herearising from the need to separate spatial andtemporal variations because of the relativelyslow motion of the GPS satellites across thesky. (This characteristic can be contrasted withTRANSIT system measurements in which thesatellite would pass from horizon to horizon inabout 20 minutes, thus providing a single slicethrough the ionosphere at effectively one time.)

The applications of GPS for meteorologicalmeasurements is in its infancy. It seems clearthat obtaining estimates of perceptible water to1 mm is possible and with this accuracy it willbe a very useful product for meteorologicalmodels.

75

6.3 Infrastructure

Of critical importance to routine applications ofGPS is the global tracking network establishedby the International GPS Service (IGS) whichis sponsored by the International Association ofGeodesy (IAG). The routine operation of anetwork of stations of between 20 and 50receivers for the past two years has made theprocessing of regional GPS campaigns mucheasier because of the availability of accurateorbits a week or so after the data has beencollected. In addition, this service alsomaintains quality checks on both the receiverswithin the network and the satellites. Theexisting data set is also extremely useful fortesting model improvements to the analyses ofGPS data.

Another important function which must beaddressed now is the archiving of GPS data.Although most data is distributed in the RINEXformat, in the long run this format (unlessrevised) has severe limitations for maintainingthe integrity of GPS data. As discussed below,there is no means for storing amplitudeinformation adequately in this format and,when “clean” data is distributed, there is nomeans for saving the changes made to the data.The error flagging of data is so limited in theformat that most centers need to delete datatotally from the clean RINEX file for datawhich was not used in their analysis (e.g., dueto low elevation angle) even though these dataare likely to be error-free and could proveuseful in the future when models for theatmospheric delay and multipath are improved.The current situation is that the raw data filesfrom the receiver, raw RINEX files, andcleaned RINEX files all will need to stored inan archive for future use.

6.4 Error Budget

The error budget for GPS can be broken downinto a number of components some of whichare reasonably well understood and otherswhich are not so well understood. In nearly allcases the full error spectrum is not known andhere we lay out and give examples of thoseparts of the error spectrum which areunderstood. We also attempt to bound that partwhich is not well understood. For eachcomponent of the error we will attempt toquantify its magnitude and the spectralcharacter of the error (i.e., the magnitude of theerror as a function of frequency).

6.4.1 Data Noise

Data noise from instrumental effects should bethe most quantifiable of the GPS error sources,and is likely to be one of the smallercontributions to the total error budget. The datanoise can be theoretically computed from thegain of the GPS antenna, the GPS signalstrength, and the thermal noise characteristicsof the receiver and its environment. Data noiseis expected to be elevation angle dependent(due to both the gain of the GPS antenna andthe loss of signal strength due to greaterattenuation of the Earth’s atmosphere) andaveraging time in the GPS receiver. For L1observations that decode either the C/A code orthe P-code, the integration time required toachieve data noise of about 1 mm is only about1 msec. Similar integration times can be used atL2 if the P-code is known but the integrationtimes need to be increased almost a thousand-fold (to about 1 second) if codeless techniquesare used at L2. (This increase in integrationtime compensates for the loss of SNR incodeless tracking systems and therefore thenoise level does not appear that much better forcodeless tracking. In dynamic applications, theincreased integration time can and often does

76

cause frequent losses of the lock on the GPSsignal.)

One common method for experimentallydetermining the noise in GPS receivers is toconnect two receivers to the same antenna andlook at the differences of the measured phaseand range values. However, these types ofexperiments need to be performed verycarefully because reflections from the receiversback through the splitter used to divide thesignal from the antennas between the tworeceivers can affect the output of the receivers.As an example of the results obtained from asplit signal experiment, we show the doubledifferenced phase residuals from a TrimbleSST and SSE connected to the same antenna inFigure 23. Examples of similar typeexperiments with a variety of receivers havebeen carried out. Generally tests of this typehave shown that the noise in the L1 and L2phase measurements varies between 1 and 5mm and often show some systematic trends.Also large trends can develop from driftsbetween the L1 and L2 phase measurementsthat are common to all channels within areceiver, but differ from receiver to receiver.This type of error does not affect the doubledifference observable because it cancels in thedifferencing operation, but it does affectestimates of the ionospheric delay. Betweendifferent receiver types, the error in ionosphericdelay can be affected by up to 400 mm. Thuswhile instrumental errors are small, these errorsources should be better understood than theycurrently are.

6.4.2 Multipath

Multipath for GPS receivers is defined as thecontribution to the phase and rangemeasurements from reflected signals. One ofthe major compromises that exists in designingomni-directional antennas of the type used inGPS is maintaining relatively high gain at lowelevation angles while minimizing the gain for

elevation angles less than zero. Multipath isdifficult to characterize in general since itsamplitude and phase depend on manyparameters, although many of these parametersare sufficiently constant that multipath can beseen to repeat day-to-day because of therepeating orbits of the GPS satellites. Ingeneral, an object near a GPS antenna thatappears smooth to the 190- and 240-mmwavelength GPS signals will act as a source ofmultipath. The most common of these arenearby buildings, trees, and the ground itself.The ground is generally a strong source ofmultipath, and if the ground is treated as aninfinite level plane, the difference in the phaseof the signal and the reflected signal is given by2hsin e, whereh is the height of the antennaabove the ground in wavelengths, ande is theelevation angle. We show in Figure 23 a typicalexample of multipath. Here the oscillatorybehavior of the double difference phaseresiduals is most likely due to the full rotationof the (weaker) reflected signal affecting thephase determinations. In extreme cases thereceiver can temporarily lock onto the reflectedsignal and this may explain the single cycleslips that some receivers in certaincircumstances experience.

One of the best diagnostics of the presence ofmultipath is the amplitude of the receivedsignals, which will also show an oscillatorybehavior in the presence of multipath.Unfortunately, the current internationalstandard for exchange of data (RINEX) doesnot allow the amplitude to be reported withsufficient significant digits to allow it to beuseful in detected and correcting multipathcorruption of a signal.

While the multipath shown in Figure 24averages effectively over the long interval ofobservations shown, for short periods of time(up to 2030 mins depending on h and the rate ofchange of elevation angle) multipath does notaverage to zero, and this can seriously effect the

77

Figure 23. Double difference carrier phase noise obtained from a Trimble SSE and Trimble SSTreceiver connect to the same antenna (contributed by Herring). The solid line shows the differ-ences between phase measurements to PRN 03 and PRN 07 at the L1 frequency; the dashed curveis for the L2 frequency. At L1 the mean and rms scatter (about the mean) of phase differences are-1.5 mm and 1.0 mm; at L2 the mean and rms are 1.6 mm and 2.1 mm. The stippled lines show theelevation angles of the satellites and clearly show that the noise is elevation angle dependent. Thelarger scatter at the L2 frequency is mostly likely due to the codeless tracking of the L2 signal inthe SST. Although the means of these residuals are not zero; the differential position estimate fromthe two receivers was 0. x mm north, X m east and X mm in height.

-10.0

-5.0

0.0

5.0

10.0

15.0

Pha

se d

iffe

renc

e (m

m)

Elevation angle (m

m)

10

20

30

40

50

60

70

80

90

12 13 14 15 16 17

UTC (hrs)

18 19

PRN 17

PRN 3

results from kinematic and rapid static surveyswhere site occupation times can be as short as afew minutes. In these cases errors of tens ofmillimeters may occur because of multipath.

6.4.3 Receiver Timing

While the first-order effects of satellite andreceiver clocks cancel during the doubledifferencing operation or by estimating white-

noise clock correction, there is a dependence ofthe geodetic results on the actual time at whichmeasurements are made due to the non-linearity of the geodetic problem. Themagnitude of the errors associated withincorrect timing of GPS signal sampling can becomputed from the Doppler shift of the GPSsignals, and it is typically 1 mm permicrosecond of timing error. While such timingrequirements should be easily met by a GPSreceiver, examples can be found in data in

78

Figure 24. A typical example of multipath (contributed by Herring).

-150

-100

-50

0

50

100

LC P

hase

res

idua

l (m

m)

Elevation angle (deg)

10

20

30

40

50

60

70

80

90

19 20 21 22 23UTC (hrs)

which this is not the case. These examples arereflected as single outliers in the doubledifferenced LC observable, but with noperturbation to the ionospheric delay estimateof the wide-lane observable (computed usingthe range and phase measurements). Suchbehavior is consistent with the receiver makingits measurements at a time different to thatreported. The overall effect of this type of erroris small except in those cases where thereceiver does not correctly resolve themillisecond ambiguity in the C/A code range,in which case it becomes almost impossible todetermine where the measurements were made.When this occurs, that data almost always needto be discarded. Ideally, this type of errorshould be detected in the field so thatmeasurements can be repeated.

6.4.4 Atmosphere

Modeling the delay in the GPS signals due tothe Earth’s atmosphere is considered to be oneof the limiting error sources in GPS primarilybecause of the natural variability of the delayand difficulty in developing a practical meansfor calibrating the delay other than throughestimation techniques using the GPS datathemselves. The full atmospheric delay canreach about 13 m at 10 degree elevation angle,but much of this is predictable frommeasurements of surface pressure. Theremaining part of the delay primarily due todipole component of water vapor refractivityand gradients in the atmosphere can not be soeasily modeled. Recent attempts to calibrate

79

the water vapor component using instrumentssuch as water vapor radiometers have beenencouraging. However, these newer results arevery limited in location and time span.

Current analyses would indicate that residualerrors in modeling the atmospheric delay couldbe tens of millimeters and may have annualsignatures.

6.4.5 Second Order Ionospheric Delay

While the dual-frequency ionospheric delaycalibration is likely to be accurate to a fewmillimeters in most cases there is theoreticalevidence that the neglected contribution fromthe magnetic field in the dual-frequencycorrection could introduce errors of up to 30mm. However, with some simplifyingassumptions, this error appears to be reducibleto a few millimeters in even the most extremecases by modifying the method used tocompute the effective frequencies of the GPSsignals. The most serious consequence of theionospheric delay may be the loss of lock onGPS satellites during periods of rapidionospheric variations. This loss of lock is aserious problem for codeless receiversoperating in polar and equatorial regions, andbecome a serious problem in these regions forfull-coded receivers when AS is turned on.

6.4.6 Satellite Motions and SatelliteAntenna Characteristics

The GPS constellation consists of 26broadcasting satellites. The satellite orbits canbe generally characterized by their Kepler orbitelements: semimajor axis = 26000 km (orbitalperiod = 43082 sec, half of a sidereal day),eccentricity < 0.01, and inclination = 55 deg(Block II), plus three additional elements thatdepend on the specific satellite. The additionalthree orbit elements are: ascending nodelocation (point where the satellite crosses the

equator from Southern hemisphere to theNorthern hemisphere, known as the “RightAscension of the Ascending Node,” withrespect to a celestial reference frame), theperigee location (known as “argument ofperigee”), and the time when the satellite waslast at perigee. The orbital semimajor axis andeccentricity describe the shape of the orbit interms of a conic section and the inclinationdescribes the orientation of the orbit plane withrespect to the Earth’s equator. The GPSconstellation consists of six orbit planes, with 4or more satellites in each plane, evenlydistributed within the plane. If the Earth were asphere with constant, uniform mass density, thesix orbit elements would be constant. Since theEarth rotates, the longitude of the locationwhere the satellite crosses the equator changeswith time even if the orbit elements areconstant.

Various forces act on the satellites that producechanges in the orbit elements with time. Theseperturbing forces are dominated by thegravitational component associated with theEarth’s oblateness, which produces a lineartime variation in both the orbit plane ascendingnode location and the location of the perigee,with respect to an inertial reference frame. Thenode location makes one revolution in about 25years in response to this perturbation; however,the time required for a complete revolution ofperigee is about 50 years. These rates of changeare much slower than the changes experiencedby a low altitude satellite. Although the sun andmoon produce a similar linear drift in theseorbit elements, the effect is about a factor offive smaller than the oblateness effect.

In addition to the linear (or secular) variation inthe node and perigee locations, the Earth grav-ity and luni-solar gravitational perturbationsproduce periodic variations in all orbit ele-ments. The perturbation spectrum includesperiodicities ranging from approximately onehour to 18.6 years. Over a one year interval of

80

time, these periodic perturbations are domi-nated by the luni-solar forces that produce“long-period” changes in the orbit elementswith periods of 14 days (lunar) and 183-days(solar semi-annual). At the GPS altitude, thegravitational forces on the satellite due to theEarth, moon, sun, and planets are well-modeledand understood. Furthermore, errors in themodeling of the mass variations in the Earth arediminished at GPS altitude compared to a muchlower altitude. On the other hand, the non-grav-itational forces experienced by the GPSsatellites are complicated and prone to mis-modeling of the forces. The nongravitationalforces are produced by solar radiation, Earthradiation, and thermal radiation from the satel-lites, to name the major contributors. Theseforces produce periodic variations in the orbitelements similar to gravitational forces. Themis-modeling of the nongravitational forces isone of the major contributors to GPS orbiterror.

6.4.7 Impact of SA and AS

Selective Availability (SA) denies precisepositioning by corruption of the GPS signalstructure. It is composed of two components (1)corruption of broadcast navigation messageand (2) rapid “dithering” or oscillations of thefrequency standards in the satellites. The firstcomponent of SA is of little consequence toscientific users because in precise millimeterapplications the orbits of the GPS satellites arecomputed from carrier measurements muchmore accurately than even the preciseephemeris available from the DoD. Also thedynamics of the GPS satellites are well enoughunderstood that these orbits can be integratedforward in time by several days with accuraciesbetter than the (uncorrupted) broadcastephemeris except when there are thrusterfirings on the GPS satellites. The secondcomponent currently also has little effect whendifferential (either real-time or post-processing) techniques are used. We show in

Figure 25 an example of the dithering of theGPS frequency standards obtained from theanalysis of the data collected with the globalGPS tracking network. In this particularexample, SA is turned off the satellite atapproximately 18:00 UTC. The results ofdithering have been converted to meters ofrange error in this figure. Prior to SA beingturned off, the rms error in the rangemeasurements is 22 m; when SA is turned off,the range error rms drops to about 0.25 m.(These results are obtained from the analysis ofcarrier phase measurements, and thereforenearly all of the rms is due to dithering and thenatural drifts in the satellite and ground stationclocks.) While the effects of SA are almosttotally eliminated in differential positioning,the rapid fluctuations in satellite clocks docomplicate the process of removing slips in thenumber of carrier phase cycles accumulated bythe receiver and SA makes it nearly impossibleto use averaged phase measurements (normalpoints) with a much lower sampling rate. Inreal-time navigation problems, SA limitsaccuracy of navigation, although for aircraftthis is not a major problem since the position ofan aircraft can generally only be controlled towithin about 100 m.

Anti-spoofing (AS) is meant to stop falsesignals from corrupting military receivers, butas a consequence of the system used, AS deniesaccess to the P-code. AS is implemented bymodulating the P-code with an additional code(the W-code), and because only the P-code(with the W-code modulation) is superimposedon the L2 frequency, some type of codelesstracking of L2 is required when AS is turnedon. In Figures 26a and b we show the effect ontwo different receivers when AS is turned on atapproximately 17:00 UTC. The degradation ofthe performance of the receiver in Figure 26a isimmediately obvious, and severely limits theeffectiveness of the automatic data processingsystems. In Figure 26b, the degradation of theperformance of the receiver is not so obvious

81

Figure 25. Example of Selective Availability (contributed by Herring). In this example, SA wasturned off at about 18:00 UTC and impact of SA can be clearly seen. In (a), the range error toPRN 01 is shown. The rms error in the range errors when SA is on is 22 m, consistent with deny-ing point positioning accuracies better than 100 m. In (b) an expanded view of the time intervalwhen SA is not on is shown after a quadratic polynomial is removed. These results were obtainedusing carrier phase measurements. The rms range error for this interval is 0.25 m, and indicatesthat 1 meter accuracy point positioning would be possible with existing satellites and receivers ifSA were not on.

-100.0

-50.0

0.0

50.0

100.0

16 17 18 19 20 21 22UTC (hrs)

(a)

SA turned off

-1.0

-0.5

0.0

0.5

1.0

Ran

ge r

esid

uals

(m

)

18 19 20 21 22UTC (hrs)

(b)

82

although careful analysis of these data indicatethat the effectiveness of automatic dataprocessing systems is also compromised whenAS is turned on. (It is also interesting to notethe general stability of the receiver shown inFigure 26b is not as good as that for Figure 26areceiver when AS is off, possibly indicating thereceiver manufacturer has put most of theefforts into handling AS rather than developinga stable receiver when AS is off.) The primeconsequence of AS is loss of accuracy in bothrange measurements and phase measurements.For aircraft applications, this loss of accuracy(about 1000-fold for L2 tracking) isparticularly severe because of the dynamics ofthe aircraft and the usually high multipathenvironment on the aircraft. In particular,results of the quality shown in Figure 26b canonly be obtained by coherently averaging theL2 signal for about 1 second. This is anacceptable compromise for a static receiver, butin an aircraft leads to frequent loss of lock onthe L2 signal.

7 References

Figure 1. Contributed by R. Ware, internalUNAVCO publication, 1994.

Figure 2. Feigl, K., D. Agnew, Y. Bock, D.Dong, A. Donnellan, B. Hager, T. Herring, D.Jackson, T. Jordan, R. King, S. Larsen, K. Lar-son, M. Murray, Z. Shen, and F. Webb,SpaceGeodetic Measurement of Crustal Deforma-tion in Central and Southern California,1984–1992,J. Geophys. Res., 98, 21,677–21,712, 1993.

Figure 3. Contributed by R. Ware, internalUNAVCO publication, 1994.

Figure 4. From K. Hudnut and M. Murray,U.S.G.S., January 25, 1994.

Figure 5. Donnellan, A., B. Hager, and R.King, Discrepancy Between Geological andGeodetic Deformation Rates in the VenturaBasin,Nature, 366, 333–336, 1993.

Figure 6. Contributed by Y. Bock, ScrippsInstitution of Oceanography, 1994.

Figure 7.Bock, Y., D. Agnew, P. Fang, J. Gen-rich, B. Hager, T. Herring, K. Hudnut, R. King,S. Larsen, B. Minster, K. Stark, S. Wdowinski,and F. Wyatt, Detection of Crustal Deforma-tion from the Landers Earthquake SequenceUsing Continuous Geodetic Measurements,Nature, 361, 337–340, 1993.

Figure 8. Yuan, L., R. Anthes, R. Ware, C.Rocken, W. Bonner, M. Bevis, and S.Businger, Sensing Climate Change Using theGlobal Positioning System,J. Geophys. Res.,98, 14,925–14,937, 1993.

Figure 9. Tapponnier P., G. Peltzer, A. Y.LeDain, R. Armijo, and P. Cobbold,Propagat-ing Extrusion Tectonics in Asia: New Insightsfrom Simple Experiments with Plasticine,Geology, 10, 611–616, 1982.

Figure 10.Dokka, R. K., and C. J. Travis,Roleof the Eastern California Shear Zone inAccommodating Pacific–North AmericanPlate Motion,Geophys. Res. Lett., 17, 1323–1326, 1990

Figure 11. Royden L.,Late Cenozoic Evolu-tion of the Pannonian Basin,Am. Assoc.Petrol. Geol., Mem., 45, 27–48, 1988.

Figure 12. Oral, B., GPS Measurements inTurkey (1988-1992): Kinematics of the Africa-Arabia-Eurasia Plate Collision Zone, doctoralthesis, MIT, 1994.

Figure 13.Calais, E., Y. Bock, et al., ScrippsInstitution of Oceanography, in preparation,1994.

83

Figure 26. Example of the effects of Anti-spoofing (contributed by Herring).

-8.0

-6.0

-4.0

-2.0

0.0

2.0

4.0

6.0

8.0

Wid

e-L

ane

(cyc

les)

14 15 16 17 18 19 20UTC (hrs)

AS Turned on

(a)

-8.0

-6.0

-4.0

-2.0

0.0

2.0

Wid

e-L

ane

(cyc

les)

4.0

6.0

8.0

14 15 16 17UTC (hrs)

18 19 20

AS Turned on

(b)

84

Figure 14.Contributed by R. Smith, Univer-sity of Utah, 1993.

Figure 15.Owen, S., P. Segall, J. Freymueller,A. Miklius, R. Denlinger, T. Arnadottir, M.Sako, and R. Bürgmann,Rapid Deformation ofthe South Flank of Kilauea Volcano, Hawaii,submitted toScience, 1994.

Figure 16.Contributed by C. Meertens and R.Smith, 1994.

Figure 17. Bürgmann, R., P. Segall, M.Lisowski, and J.L. Svarc,Strain developmentsubsequent to the 1989 Loma Prieta earth-quake, U.S.G.S. Professional Paper 1550D (inpress), 1994.

Figure 18.Courtesy Herring, T., CSDEI Sci-ence Plan, 1993.

Figure 19. Waschbusch, P. and M. McNutt,Yellowstone: A continental midplate (hot spot)swell, Geophys. Res. Lett., 21, 1703–1706,1994.

Figure 20.Baggeroer, A. and W. Munk, TheHeard Island Feasibility Test,Physics Today,45, 22–30, 1992.

Figure 21.Baggeroer, A. and W. Munk, TheHeard Island Feasibility Test,Physics Today,45, 22–30, 1992.

Figure 22. Rocken, C., T. Van Hove, J.Johnson, F. Solheim, R. Ware, M. Bevis, S.Businger, and S. Chiswell,GPS/Storm—GPSSensing of Atmospheric Water Vapor for Mete-orology, submitted to J. Atmos. OceanicTechnol., 1994.

Figure 23.Contributed by T. Herring, MIT.

Figure 24.Contributed by T. Herring, MIT.

Figure 25.Contributed by T. Herring, MIT.

Figure 26.Contributed by T. Herring, MIT.