JOURNAL OF GEOPHYSICAL
RESEARCH, VOL. 102, NO. B9, PAGES 20,269-20,285, SEPTEMBER 10, 1997
A seafloor longsbaseline tiltmeter
Gregory Anderson,StevenConstable,Hubert Staudigel,and Frank K. Wyatt
Institute of Geophysicsand Planetary Physics, ScrippsInstitution of Oceanography
University of California, San Diego
Abstract.
Long-term monitoring of seismicity and deformation has provided
constraints on the eruptive behavior and internal structure and dynamics of
subaerial volcanoes,but until recently, such monitoring of submarine volcanoeshas
not been feasible. Little is known about the formation of oceaniccrust or seamounts,
and we have therefore developed a stand-alone long-baselinetiltmeter to record
deformation on active seafloorvolcanoes. The instrument is a differential pressure,
two-fluid sensor adapted for use on the seafloor, combined with an autonomous
data loggerand acousticnavigation/release
system.The tiltmeter can be installed
without use of remotely operated vehiclesor manned submersiblesand, to first order,
is insensitive to noise driven by temperature or pressuregradients. We recorded
65 days of continuousdata from one of these tiltmeters on Axial Seamount on the
Juan de Fuca Ridge during a multidisciplinary experiment that included ocean
bottom seismographs,magnetotelluric instruments, and short-baselinetiltmeters.
After instrument equilibration the 100-m-ionõ tiltmeter provided a record with
long-termdrift ratesof 0.5-5 /•rad day-• and higherfrequency
variationsof the
order of 5-10/•rad. Comparison with recordsof subaerial volcanic tilt showsthat
this instrument can discriminatevolcanic deflation events,though none occurred
during our deployment, a conclusionsupported by nearby short-baselinetilt and
bottom pressurerecordings. The short- and long-baselinedata constrain volcanic
inflationof AxialSeamount
to bebelow0.5-1/•radday-1 duringmid-1994.Analysis
of the long-baselinetilt data in conjunction with electric field, temperature, and
short-baselinetiltmeter data showsthat high-frequencysignalsare largely driven
by ocean currents. Improved coupling between the tiltmeter and seafloorshould
reduce this noise, improve stability and drift, and further enhance our ability to
record
tilt
related
to active
submarine
volcanism.
Introduction
Monitoring of subaerial volcanoesrelies heavily on
two complementary techniques: seismologyand surface deformation.
Each of these has its own distinct
that has been erupted at the surface versusemplaced
in shallow dike systems. Continuoussurfacedeformation measurements,however, are limited in their ability to pinpoint the location of intrusionsor the force
of magmatic injection becauseof the small size of the
signals,the existenceof environmentalnoise, and the
difficulty of uniquely inferring the sourceof the signal.
Thus complementarymonitoring techniquesand verification of measurementsthrough redundancyare key in
volcano monitoring.
Use of such combined techniqueson major subaerial
volcanoeshas provided key details on the geometry of
magmafeedersystems,temporal and spatialbehaviorof
magma influx from depth into shallowreservoirs,redistribution of melt within a volcano, and eruptive behav-
strengths, and the successof volcano monitoring decreasessignificantly if either technique is used in isolation. Seismologyis unmatched in early detection of
intrusive activity and accuracyin locating brittle deformation events. During volcaniccrises,harmonic tremor
is crucial for pinpointing major magma flow. However, magma flow itself does not cause earthquakes,
and other techniques must be used to detect aseismic
magma redistribution within a volcano. Surface deformation measurementsare unique in providing constraints on the temporal evolution of melt accumula[e.g.,Klein, 1984;Decker,1987;Klein et
tion, the behaviorof magmaredistributionthrough vol- ior of volcanoes
al.,
1987].
These
techniqueshave thus been extremely
canic feeder systems,and the fraction of available melt
successfulin improving our understandingof how volcanic systemswork and in predicting the timing and
Copyright 1997 by the American GeophysicalUnion.
types of eruptions,but all major advancesin this field
have comefrom studieson a few very well-instrumented
Paper number 97JB01586.
volcanoes[e.g.,Decker,1987]. Until recently,technical
0148-0227/97/97JB-01586509.00
20,269
20,270
ANDERSON ET AL.' SEAFLOOR LONG-BASELINE TILTMETER
are extreme. Coldtemdifficultieshave precluded meaningful long-term moni- averageconditionsthemselves
peratures
(near
0øC)
and
high
confiningpressures
(1
toring of the most abundant types of volcanoesexposed
in the submarine environment, so there is much still MPa per 100 m depth) can causecreepof instrument
parts,contactwith seawaterwill causecorrosion
if care
is not taken in selectingmaterialsfor usein a seafloor
Significanteffort has been expendedto fill this gap in tiltmeter, and couplinginstabilitiescan result in noise
our understandingof volcanoes,focusingon seismology drivenby seafloorcurrents.Furthermore,the condition
and surface deformation. Marine seismologyhas long of a candidate site on the seafloor is barely known in
beenan established
discipline[e.g.,Bradnatat al., 1965; comparisonwith a subaerialsite, and it is often imposJohnsonat al., 1977; Willoughbyat al., 1993],but the sible to determine before deployment whether or not a
developmentof seafloordeformation measurementshas particular area is primarily mud, rubble, or relatively
begunonly recently[e.g., Staudigalat al., 1990; Chad- stable bedrock. For these and other reasons it is quite
wall at al., 1995; Wyatt at al., 1996]. In this paperwe difficult to achieve the same level of performancewith
shall discussthe developmentof a seafloorlong-baseline a seafloor instrument as with subaerial instruments.
Is it feasibleto build a tiltmeter capable of recording
tiltmeter, including the results of a successfultest degeophysically
relevantsignalson the seafloorgiventhe
ployment at the Juan de Fuca Ridge.
to be learned
about
how ocean crust or seamounts
are
formed.
extreme conditions there? Tilt driven by relative plate
Tiltmeters
for
Submarine
Volcano
motion on continentstypically showsrates of 0.1-0.3
/•rad yr-1 [Wyatt at al., 1988],well belowthe mini-
Monitoring
mum tilt rates that can be detected by current seafloor
tiltmeters. However, tilt driven by eruptive or intrusive
Surface deformation can be measuredin many different ways, including tilt, strain, and geodetictech- volcaniceventscanhavemuchhigherrates[e.g.,Dvorak
niques;of these, tilt is usedmost frequentlyfor volcano at al., 1986],whichmay be within the reachof current
monitoring. The Westphal short-baselineinstrument technology.Simplemodelsof volcanicdeflationprovide
[Wastphalat al., 1983]is the mostcommonlyusedtilt- bounds on tilt rates that might be observedduring a
meter, but long-baselinedesignshave alsoplayeda ma- volcanic event.
Eruptions of Kilauea Volcano on Hawaii give some
jor role at somesubaerialobservatories
[e.g., Wyatt at
al., 1984;Davis at al., 1987]. Short-baseline
tiltmeters constraintson the rates and volumesof magma injec(SBTs), that is, with baselineof the orderof _<1m, have tion during intrusive events at a hotspot volcano. Inthe main advantageof being relatively inexpensiveand jection rates during Kilauean volcaniccrisesare often
easy to install. However, SBTs are sensitiveprimarily greaterthan 106m3 day-1 andcanbe asmuchas 108
to short-wavelengthtilt and noise,which we expect to m3 day-1 [e.g.,Pollardat al., 1983;Dvorakat al., 1986].
be highly heterogeneousin fragmented volcanic envi- When we compute the surface displacementsand tilts
ronments. Also, SBTs are subject to creep instability producedby deflatinga 2-km-radiusMogi source[Mogi,
in the instrument itself and can be stronglyinfluenced 1958]centered
5 km belowthe surfaceat ratesof 106,
by very localizedfluctuationsin temperature,rain, and 107,and 108m• day-1, wefind tilt ratesof about1, 10,
other such effects.
and 100 /•rad day-1 overa distanceof 2-10 km from
Long-baseline
tiltmeters (LBTs), with baselineof the
order of _>10 m, have severalpotential advantagesover
the SBT design. By virtue of their greater baseline
length they are lesssensitiveto short-wavelength
noise
maskingthe broader-scaledeformationfield. They are
also inherently much lesssusceptiblethan SBTs to the
effectsof creepinstability in the instrumentcomponents
or ground coupling. However,LBTs typically are more
the volcano(Figure la).
Measurementsof dike dimensions
in ophiolites[e.g.,
Baragarat al., 1987;Rothary,1983]and data fromeruptionson Iceland[e.g.,Sigurdsson,
1987]and alongthe
CoAxial segmentof the Juan de Fuca Ridge [Dziak at
al., 1995]canbe combinedto estimatemagmainjection
rates that might occur during dike formation along an
activesegmentof the mid-oceanridge(MOR); a reason-
expensive and much more difficult to install than SBTs
and are susceptibleto different sourcesof noise.
Seafloorenvironmentalconditionspresentboth major
ablerangeis 106-108m3 day-1. On the basisof the axial magmachamber(AMC) observations
of Sinhaat al.
[1997]we modelan AMC alonga segmentof the MOR
advantagesover subaerial conditions and seriouschal- as a prolate spheroid with a 20-km semimajor axis, a
lenges to the successfuluse of tiltmeters for subma- 2-km semiminor axis, and a 4-km centroid depth. We
rine volcanicmonitoring. The primary advantageof
then usethe methodof Yangat al. [1988]to compute
the seafloor is that
the surface displacementsand tilts produced by deflat-
environmental
conditions such as
temperature are much more stable than those on land;
for example, typical seafloordiurnal temperaturefluctuations are +0.1øC in comparisonwith about +15øC
experienced by the LBT on land at Pition Flat Geo-
ing our modelAMC at ratesof 106, 107,and 108m3
day-1 and find tilt ratesof about 1, 10, and 100/•rad
day-1 overa distanceof 2-10 km from the ridgeaxis
acrossstrikeand about0.5, 5, and 50/•rad day-1 over
physicalObservatory[Wyatt and Bargat,1980]. How- a distanceof 10-20 km along strike from the center of
ever, while fluctuationsin theseconditionsare low, the
the AMC (Figureslb and lc).
ANDERSON ET AL.' SEAFLOOR LONG-BASELINE TILTMETER
30
25
/! \\
I
\%
I
ß
o
10
5
o
15
20
25
30
Distance (km)
30
,
20,271
We do not wish to imply that a particular volcano
or ridge segment would behave as shown, but merely
that these models give a reasonableorder-of-magnitude
bound on tilt rates that might be observedin situ. It is
also important to remember that such intrusive events
might occur only episodically on an annual to decadal
scale on any given seamount or MOR segment. Given
these caveats, however, it is clear that tilt rates during
volcaniccrisesare 3-5 ordersof magnitude higher than
tilt rates related to relative plate motions on continents.
Thus, however unlikely it is that seafloortiltmeters can
be constructed to make meaningful measurementsof
slow tectonic tilt, constructing instruments with performance good enough to make them useful for our
understanding of submarine volcanoesis a much more
tractable problem.
/
/
25
A Seafloor Long-Baseline Tiltmeter
/
/
/
Most of the techniquesusedin installing high-quality
/
land instruments[Agnew,1986] either cannot be ap-
/
/
plied to, or are prohibitively expensivein, the seafloor
environment. For example, even installing a level instrument, a requirement for nearly all of the better
tiltmeter designs,is extremely difficult without the use
of remotely operated vehiclesor manned submersibles,
and free fluid surfaces,suchas those in highly accurate
/
/
/
/
/
5
/
/
/
o
5
10
15
20
25
30
Distancealong ridge (km)
30
25
Michelson-GaleLBTs [Wyatt et al., 1984],are impractical under seafloorconditions. However, LBT designs
that measurethe pressurein fluid-filledtubes[e.g.,Beavan and Bilham, 1977;Agnew,1986]are feasibleon the
seafloor, and this is the approach we have chosen to
pursue. Our development work has resulted in the design illustrated in Figure 2. It is a center-pressureinstrument folded back on itself, with a sensorthat mea-
suresthe differentialpressurebetweentwo tubes(arms)
filled with fluids of differing densities,which terminate
with fluid reservoirs(pots) that are sealedby compliant membranes. This design has several advantages.
First, becausethe fluid reservoirsopen to environmen\
tal pressureare side by side, the potentially significant
seafloorlateral pressuregradients need be neither compensated for nor measured. Also, we require only one
30
10
15
20
25
pressuresensor, which makes for simplicity and reliaDistanceacrossridge(km)
bility in both the designand deploymentof the instruFigure 1. (a) Estimatedtilt for modelsof a deflating ment. Additionally, with an appropriate choiceof fluid
Mogi source 2 km in radius, centered 5 km below the properties this LBT design can be made temperature
surface in an elastic half-space with Lam• constants• compensatingto first order, as discussedbelow.
/• -- 30 GPa. Solid curve is tilt rate that would result
6
3
1
The confining pressure of the seafloor environment
from removing 10 m day- of magma from the source,
the short-dashed curve is for 107 m a day-, z and the presentsa significantchallengein measuringtilt with
long-dashed
line is for 10s ma day-z. (b) Estimated this type of instrument. If the relative heights of the
along-ridge tilt for models of a deflating Yang source fluid reservoirsand the pressuresensorchangeby Ah,
with a semimajor axis of 20 km, a semiminor axis of the correspondingpressurechangeis
2 km, and a centroid depth of 4 km. Curve notation
as in Figure la. (c) Estimatedacross-ridge
tilt for the
same Yang source as in Figure lb. Curve notation as
in Figures la and lb.
Ap = (pz - p•)gAh
(1)
where px and p• are the fluid densities in the two
20,272
ANDERSON ET AL.' SEAFLOOR LONG-BASELINE TILTMETER
/
/
/
/
\
\
\
\
\
•
I
•
I
I
/
/
/
/
/
/
/
\
\
/
ANDERSON ET AL.' SEAFLOOR LONG-BASELINE TILTMETER
20,273
tiltmeter arms. Our current instrument uses ethylene
fluid reservoirs,respectively.However,if the two fluids
glycoland isopropanol;
the densitydifference
between are chosensuchthat the product of densityand thermal
these fluids under seafloor conditions is approximately expansivity(p/•) is the sameor nearly sofor eacharm,
1131- 811 - 320 kg m-3. To measurea 1-pradtilt, the first term in (2) is greatly reduced,and the entire
corresponding
to a relativeelevationchangeof 0.1 mm
over an instrument baseline of 100 m, we must therefore be able to measure differential pressureto better
than 0.3 Pa. At a deploymentdepth of 1000-4000m our
instrument has to measure differential pressureswhile
instrument is temperature compensating.
Our seafloor instrument has arms 100 m long, made
of 0.22-inch ID soft alloy copper tubing with 6-inch
ID aluminum end reservoirs(a permanentobservatory
installation
would
use all
stainless
steel
or titanium
subjected
to a common
modepressure
107-108times construction). One arm is filled with ethylene glylarger.Also,sincewe intendthe instrumentto be use- col, and the other is filled with isopropanol; see Table I for the relevant properties of these materials.
ful on slopingterrain, we require an operating range
sufficientto handle these slopes;our design currently
has a dynamic range of about +40 kPa, which corresponds to 11 m height difference over a 100-m baseline. One commercial pressuresensorthat fulfills our
requirementsis a variable reluctance gauge manufactured by Validyne (ValidynemodelDP215-36-N1S4A).
Although this transducer will tolerate large internal
propanol arm. The result is a temperature-driven dif-
common mode pressures, we operate it in a pressure-
ferentialpressureof only 0.0536Pa, or, using(1), 0.171
By using (2) and assuminga 1-m height differencebetween the ends of the tiltmeter, each 1øC averagetemperature increase along the tiltmeter producesa pressure of -6.3403
+ 0.7732 - 0.0251 =
-5.5922
Pa for
the arm filled with ethylene glycol and a pressureof
-6.4363
+ 0.8085 - 0.0180 = -5.6458
Pa for the iso-
compensatedoil bath to avoid creep of the component prad. As expected,while the first term in (2) domimaterials, which would otherwisetend to introducedrift nates for each fluid individually, the resultant differeninto the measurements.
tial pressureis quite small; thus the instrument should
be temperature compensatingto first order.
Thermal
Noise
Since it is difficult to estimate the product p/• accurately
under seafloorconditions, we have also chosento
On land, thermal fluctuations are the largest sources
of noise for tiltmeters, and although seafloortemperature variations are much lower, their effects are still
important. BeavanandBilham[1977]havedetailedthe
effect of thermal noise on fluid tube tiltmeters; here we
give only a brief review. If a fluid tube tiltmeter is
perturbed by a changein temperature along its length,
measure
thequantity
f0sAT(s)ds,
thealong-tiltmeter
temperature, directly. A standard four-conductor,28gauge copper telephone cable runs along the tiltmeter
with the conductorsconnectedin series, giving a total
length of 400 m. The resistancetemperature coefficient
of this lengthof cableis 0.34 f• K-l; we supplya con-
AT(s), there will be a corresponding
pressurechange stant current and measure the resulting voltage as our
estimate of integrated temperature. Using this record,
we can estimate the LBT temperature coefficient and
given by
AP[AT(8)]
use it to reduce the thermal
•0
$AT(8) sinO(s)ds
--
K1
+
K2
AT(s) ds
+
K3
AT(y) dy
Deployment
(2)
where
K1
--
--(Pl/•1 -- p2/•2)g
noise in the final data.
(3)
Ideally, we would use remotely operated vehicles or
manned submersiblesto deploy a long-baselineinstrument; however, suchequipment is costly to operate and
requires accessto a specialized ship. To facilitate our
designand test program, we choseto use the dynamic
deploymenttechniquewhich Webbet al. [1985]de-
K2
-- [P1(/•1
--3/•t)
--p2(/•2
--3/•t)]
•pp
g (4)
veloped for long electromagneticsensors. At sea the
sensoris releasedinto the water from a reel, starting
with the end pot assembly,with the ship traveling at
K3
=
about8 m s-1. The dragforceassociated
with the ship
At
--(Pl -- P2)g•p
(5)
motion keepsthe tiltmeter tubing nearly level (imporand /91, p2, /•1, and /•2 are the densities and thermal tant if the pressuregauge and end pots are not to be
expansivities
of the two fluids;At, Ap,/•t, and/•parethe overpressured)and preventscrimping the tubing durcross-sectionalareas and thermal expansivities of the ing deployment. When the full length of the tiltmeter
tubing and fluid reservoirsrespectively;$ is the length is nearly paid out, we make the electrical and physiof the tiltmeter; O(s) is the angle from the horizontal cal connections
betweenthe tubing/sensorpackageand
of the tube element ds; and Y is the depth of fluid in the data logger and heavy anchor assembly. The anthe fluid reservoir.The first term in (2) is usuallythe chor is connected to an electromechanical cable such as
largest, whereasthe other terms are second-ordereffects
related to thermal expansionof the tiltmeter tubing and
a conductivity-temperature-depth cable, which is itself
terminatedby an acoustictransponder/release
system,
20,274
ANDERSON ET AL.: SEAFLOOR LONG-BASELINE
TILTMETER
Table 1. Physical Properties of Tiltmeter Materials
Material
Ethylene glycol (25 øC)
Ethylene glycol (0 øC)
Ethyleneglycol (0 øC, 15 MPa)
Isopropanol(0 oC)
Isopropanol(0 øC, 15 MPa)
pa
1.1099
1.1257
fib
0.6341
0.6431
0.6463
3.3*
1
8.09
0.6483
8.4
1
0.6561
7.9*
0.81101
122 alloy copper tubing
Aluminum end pots
Cc Source
5.713
1.13131
0.8014
pfi
3.7
3.5*
1
1
1
8.94
0.177
2
2.73
0.226
2
Sources:1, Washburn[1930];2, Boyer and Gall [1985];
a Density(10s kg m-s)
b Thermalexpansivity
(volumetric
for fluids,linearforsolids)(10-4 K-1)
c Compressibility
(10-4 MPa-1)
* Estimate from data at 25-100 øC and 0.1, 50 and 100 MPa
I Estimateusingcompressibility
allowingus to monitor the height abovethe seafloorduring deployment(i.e., a standarddeep-towedinstrument
arrangement).We then lowerthe loggerand tubing assembly to near the seafloor,using the cable, and when
the entire package(the full length of the tiltmeter) is
periment describedin this paper is currently being constructedin-housefor about $7500 in parts.
The total cost of parts, machining, and batteries
amountsto lessthan $15,000 per LBT, which compares
favorablywith short-baselineinstruments,as the logger
and release components are essentiallythe same and
the sensorcosts are comparable. The deep-towedde-
about 15 rn from the seafloor,we releasethe instrument
to settle under its own weight.
Acoustic transpondersthat operate at 12 kHz are at- ployment schemerequiresmore time (about 5 hours
tached to the end pot assemblyand the loggerassembly per instrument) and a somewhatmore capableship
(i.e., the two endsof the tiltmeter) and allowusto mea- than does simply dropping instruments from the sursure the deployedlength and the depths of the ends of face (which requiresabout 2 hours per instrumentif
the tiltmeter by an acoustic survey. The instrument they are trackedto the seafloor),but in either casethe
then operates on the seafloorfor the duration of the ex- time required and costsinvolvedin getting to, conductperiment. At the end of the recordingperiod an acous- ing bathymetric surveysof, and returning from the retic command to the data logger initiates releasefrom searcharea usually greatly exceedthe time and costs
the logger anchor, and the insulated wires connecting involvedin deployingthe instruments.
The loggerand releasesystemare recoveredand are
the sensorblock (whichis securedto the anchor)to the
reusable.
For the deploymentsthat were part of the deloggerare broken, allowingthe loggerpackageto rise to
the surfaceunder its own buoyancy.
velopment
cycle(andthe experiment
wedescribe
below)
the anchorand sensorassemblyare left on the seafloor.
Costs
It is relevant to discussthe coststo deploy such an
instrument. There are three main components to the
system: the sensorand anchor assembly,the data logger, and the acousticnavigation and releasedevice.
Parts, machining and labor for the sensor compo-
We haveexperimentedin providingextra buoyancyand
lifting the sensorbackto the surface,but the longtube
assemblytends to foul and catch on rocks,endangering our ability to recover the data. However, we have
also experimentedwith providingan underwatermatable electricalconnectoron the tiltmeter sensor,allowby using
nents cost about $5000, includingthe Validyne gauge, ing a workingtiltmeter sensorto be reoccupied
to connecta seconddata logger.
which amounts to about one sixth of this cost. Sensor a submersible
cost depends somewhat on the materials used; copper
tubing and aluminum parts can be employedfor short
Juan de Fuca Ridge Experiment
deployments, while more costly stainlesssteel or titanium parts would be needed for observatory installaAxial Seamountis a volcanoof the Cobb-Eickelberg
tions. Acousticnavigation/releasesystemscan be pur- Seamount Chain, located at the intersectionwith the
chasedcommerciallyfor around $10,000; we build our Juan de Fuca Ridge (JDF) approximately450 km west
own devices in-house for about one third this cost. Low-
of the Washington-Oregon
border (seeFigure 3 inset)
[Johnson,
1993].The volcanohasan elongated
summit
year or more are currently not commercially available. caldera measuringabout 3 by 8 km, with its long axis
A more modern versionof the loggerusedfor the ex- orientedN160ø [Embleyet al., 1990]. Its highestelepower, seafloordata loggerscapableof recordingfor a
ANDERSON ET AL.: SEAFLOOR LONG-BASELINE TILTMETER
20,275
TILT-OBS
ß
48'
ß
48'
..i•111i!!!4•i•?
........
.8.
_I
x
Magnetotelluric
46o00 '
44'
44'
receiver
Long-baseline
tiltmeter
.Sharyn............................
Depth, m
1780
45o58 '
:::::::::::::::::::::::::::::::::
..
:.i:i:i:•:i:M:!:i:i:i:i:i:i:i:i"-
======================================
:i;•':
i 1660
:"- '
45o56 '
1620
1580
1540
1 km
1500
.................
..................
=======================================
':...:::::.
45o54 ,
...................
,,,,,,:•,,,,,,,,,•
-130o03"
-130o00 '
-129o57
'
Figure 3. Map of the summit of Axial Seamountshowingdeploymentlocationsfor the 1994
experiment. Rhonda, Lolita, Noddy, and Quail are LBTs; Janice, Phred, Judy, Lynn, Karen, and
Sharyn are TILT-OBS; and Pele, Macques, and Ulyssesare electric field instruments. The solid
circle denotes the location of the PMEL bottom pressurerecorder. Scale is shown in the lower
left corner, and bathymetry is indicated by the shading, with lighter shadesdenoting shallower
depths. Inset map showsthe location of Axial Seamountand the Juan de Fuca Ridge in relation
to the northwest coast of North America. Axial Seamountis denoted by the black circle labeled
AS; other abbreviationsare BlancoFractureZone (BFZ) and SovancoFractureZone (SFZ).
20,276
ANDERSON ET AL.' SEAFLOOR LONG-BASELINE
TILTMETER
39950
a I
I
I
I
2.5
....---'-•
I
39900
39850
il .•.•J•
....
ij.•..•• '•'"
-2.1 .....
I
-O.6
39800
_
I
I
39750
190
180
200
210
220
230
240
250
20
10
-10
-20
200
210
220
230
240
250
6.05
6.00
5.95
5.90
5.85
I
200
I
210
•
•
220
230
I
240
•
250
Time (UTC day of !994)
Figure 4. (a) Raw tilt data from RhondaLBT. Signalprior to day 195 is largely rapid postdeploymentinstrument settling. Offsetsat days 191 and 228 are likely due to settling or other
instrument instability. Drift rates are indicated with dashed lines above and below tilt record:
5.7, 2.5, -2.1, and-0.6/•rad day-1, respectively.
Higher-frequency
(approximately
tidal) noiseof
about +6/•rad is evident. Vertical dotted lines indicate the range of data shown in Figures 4b
and 4c. (b) Residualtilt data from RhondaLBT after correctionof the offsetsat days 191 and
228 and removal of a best-fit spline to eliminate the low-frequencyvariation. Data prior to day
195 (which are contaminatedby deployment-related
noise)have beendropped. Noiseat about
+6 /•rad is visible. (c) Raw temperaturerecordfrom RhondaLBT. Variationsof the order of
+0.05
K are evident.
vation is on the western edge of the caldera, where the
volcanorises approximately 700 m abovethe surrounding ridge axis.
We deployedsix short- and four long-baselinetilt-
meters on Axial Seamount(see Figure 3) from June
29 to September 8, 1994; these instruments collected
approximately 65 days of continuousdata on five shortbaseline instruments and one long-baselinetiltmeter.
Minor but correctableproblems(an incorrecthard disk
SCSI ID, a blown fuse, and a faulty releasesystem)
son at al., 1996]. In total we collected30 channels
of short-baselinetilt, one channelof long-baseline
tilt
and the associatedtemperaturerecord, nine magnetic
field records,six electricfield records,five hydrophone
recordings,and 15 channelsof seismicdata. Here we
present analysisof data pertinent to stability and performanceof our long-baselinetiltmeter. More detailson
the short-baselineinstrumentsare presentedby Wyatt
at al. [1996]and Tolstoyet al. (submittedmanuscript,
1997).
prevented us from collectingdata from the other three
Results
and Discussion
long-baselineinstruments.We alsodeployedthree electric and magneticfield recordersfor a magnetotelluric
For the Juan de Fuca Ridge experimentwe usedthe
survey;thesedata havebeenanalyzedelsewhere[Hain-
loggingsystemdescribed
by Constable
and ½'ox[1996],
ANDERSON ET AL.' SEAFLOOR LONG-BASELINE TILTMETER
which utilizes Onset Corporation'sCPU8088 microcom-
20,277
- - -DT
ST.
puter alongwith voltage-controlled
oscillators
and 16bit countersto effect 16-bit analog-to-digitalconversion
with a least count of 0.1 Pa or 0.33 •urad. Data were
collectedat a samplingrate of 1 Hz and, duringlater
processing,
wereblockaveraged
to hourlysamples.Figure 4a showsthe raw hourly block-averaged
data from
thelong-baseline
tiltmeterRhonda;
signals
withseveral
0.5
differentcharacteristictimescales,rangingfrom DC to
minutes,areevident.First, thereis a DC levelof about
40 mrad, whichcorresponds
to a heightdifference
of 4
m betweenthe end pots and the pressuresensor;this
value is in excellentagreementwith an acousticsurveyof the instrument
madeafterits deployment.
After
about 10 daysof rapid instrumentsettlinga long-term
variationcan be seenwith a tilt rate of 5 •uradday-1
near the beginningof the record,decreasing
to about
0.5 /•rad day-1 by about day 240. This decreasing
drift signalis likely due to continuedinstrumentsettling, althoughinstrumentdrift or volcanictilt could
also cause the final observed variation.
0
100
10
Frequency(cpd)
103
(z)102
There are two
E
abrupt offsetsin the recordwith amplitudesof 56 and
19 •urad(5.6 and 1.9 mm relativeheightchangeover
the 100-minstrumentbaseline),respectively.
Theseoff-
104
sets occur over a timescale of a few minutes and are
thus distinct from volcanic deflation events, which typ-
10-•
100
Frequency(cpd)
icallyhavetimescales
of hoursto days;we believethat
thesesharpdropsare causedby settlingof the endpots
or sensorblock, as might be expectedon unconsolidated oceaniccrust. However,they alsomight possibly
3 c
be associated with mechanical or electronic instrument
tares. Finally, there is an approximatelytidal variation
presentthroughoutRhonda'srecording.To analyzethis
last signalmoreclosely,we haveremovedthe long-term
drift, usinga best-fitdepleted-basis
B-spline[Constable
andParker, 1988].The resultingresidualtilt is shown
in Figure4b anddisplaysa meanamplitudeof about+6
•urad,muchhigherthan the 0.2 •uradexpectedtidal tilt
(Tolstoyet al., submitted
manuscript,
1997);weexplore
•
2
•
Frequency(cpd)
severalexplanationsto explainthesehigh-frequency
sigbetweenRhonda,
nals, sinceit is clearthat they are not primarilytidal Figure 5. (a) Rstimateof coherence
LBT tilt and temperature(solidline). Short-dashed
tilt.
level,and long-dashed
line
Rhonda'stemperaturerecordis shownin Figure 4c; line is the 95% significance
oscillations of the order of -F0.05 K are evident.
We
is the 99%significance
level.Notecoherence
significant
at the 99% level in three broad bands, surroundingthe
computeestimatesof the coherencebetweentilt and diurnalandsemidiurnal
tides(lineslabeledDT andST,
temperatureusingboth conventionalWelch'soverlap- respectively)
andthe 4-dayoscillation.(b) Amplitude
ping segmentaveraging[Percivaland Walden,1993] spectrumof the transferfunctionfrom temperatureto
of the transferfunction.Note
and multitaper techniques[Park et al., 1987; Vernon tilt. (c) Phasespectrum
et al., 1991];wenotenosignificant
differences
between the linearity from about 0.5 to 3 cpd.
the estimates.Figure5a showsthe coherence
estimate
andthe 95%and99%significance
levelscomputed
nonWe estimate the temperature-to-tilt transfer function
deterministically
by usinga methodadaptedfromthat
of Kuehneet al. [1993].Highlysignificant
coherenceto give an upper boundon the apparenttemperatureco-
is evidentin threebroadfrequencybands(0.2-0.5, 0.8-
efficientof our tiltmeter (seeFigure5b); our computed
1.3,and1.5-2.4cycles/day
(cpd)),whichindicates
that valueis 70 •-30 •uradK -1 from 8 hoursto 10 days
thereisa component
to thetilt datathat iseitherdriven period. The phase of the transfer function estimate,
bytemperature
or drivenalongwithtemperature
by a shownin Figure 5c, is linear from 5 to 48 hoursperiod;
third phenomenon.
a linear phaserelationshipindicatesa simpletime off-
20,278
ANDERSON
20
ET AL.' SEAFLOOR LONG-BASELINE
TILTMETER
I
-20
200
210
220
230
240
250
200
210
220
230
240
250
3.30
3.25
3.20
•_.3.15
3.10
0.10
0.05
o.oo
-0.05
-0.10
•
•
•
•
•
•
200
210
220
230
240
250
Time (UTC day of 1994)
Figure 6. (a) Bottompressure
recording
fromthePMEL BPR. (b) Bottomtemperature
record
fromthe PMEL BPR. Note oscillations
of the sameorderas thoseon the LBT temperature
record,but with a differentmeanvalue. (c) LBT temperature
record(thickline) and BPR
temperature
data (thinline). Notetheexcellent
agreement
between
therecords
exceptfor a time
offsetof approximately
8-12 hourswith the LBT temperature
leading.Bothsignals
havehad
their computed means removed.
set betweenthe two signals[Priestley,1981],and from between the DC tilt offset and the acoustic survey of
the slope of the transfer function phasewe find that tilt
leads temperature by about 40 min.
Our tiltmeter's "temperature coefficient"is much lar-
the instrument indicates that the tiltmeter was operating correctly, and that the coherence between temperature and tilt would be much higher and broadband
ger than we expect from (2), and we have considered (sincea fluid tube tiltmeter with one arm closedis essentiallya thermometer). The final and perhapsmost
several possibilitiesto explain this. First, if p• for the
tiltmeter fluids under high pressureis sufficientlydifferent from p• under atmosphericconditions,the temperature compensationin our design could be lost. However, since published values of p• for fluids similar to
those used in our tiltmeter show pressureeffectsof less
compelling argument against temperature-driven "tilt"
is that the phase of our transfer function estimate indicatesthat tilt leadstemperatureby 40 min, precisely
the opposite of what one would expect if temperature
drove tilt. While this lag could be explained by dif-
than 3%, and we computefrom (2) that p• wouldhave
ferences in the rates of heat diffusion
to change by over 400% to explain the observations,
we reject this pressureeffect as an explanation for the
high "temperature coefficient." Another possibilityis
that one of the tiltmeter tubes tubes crimped closed
fluids and the plastic insulation on the telephone cable, simplediffusioncalculationsusingpublishedvalues
of thermal conductivityfor the materials involvedshow
in the tiltmeter
that the maximum expectedlag would be 50 s, with
during deployment.We reject this on the groundsthat temperature leading tilt.
the design of our tiltmeter makes it unlikely that one
We concludefrom the abovethat temperature is not
tube could crimp closed,that the excellentagreement driving tilt but that both temperatureand differential
ANDERSON
ET AL'
SEAFLOOR LONG-BASELINE
TILTMETER
20,279
20
15
•
10
LU -10
-15
I
-20
210
200
I
220
I
I
230
250
240
Time (UTC day of 1994)
103
103
4D
102
102
DT
ST
104
104
o. 100
o. 100
10-1
10-4
........
O ........
1o
104
Frequency(cpd)
10-4
10-1 .......... 100
104
Frequency(cpd)
Figure 7. (a) Residualelectricfield data from Ulysses,Macques,and Pele after removalof a
best-fit spline and rotation to a common orientation. Thick line is north component, and thin line
is eastcomponent.Variationsof the orderof +3/•V m-• are evident.Note similarityof records
acrossinstruments. (b) Power spectrumof Macqueselectricfield convertedto oceancurrent
speed. Thick line is north, and thin line is east. Note strong peaks at 1, 1.5, 2, and 3 cpd and
broad peak centered at 0.25 cpd. The peaks at 0.25, 1, 1.5, and 2 cpd are due to ocean currents.
(c) Powerspectrumof measuredcurrentspeedsnear Axial Seamount.Peaksat 0.25 cpd and the
frequenciesof the diurnal and semidiurnal tides are denoted 4D, DT, and ST respectively. Note
the similarityto Macqueselectricfield in Figure7b. After Cannonand Pashinski[1990].
pressure(and thus tilt in our instrument)are being our data. However,from theseBPR data and our LBT
driven by a third commonphenomenon.Sinceour es- record, we can estimate that the commonmode pressure acceptanceof our tiltmeter is no higher than one
part
in 50,000.
showpeaks at the periodsof the diurnal and semidiWhile
the bottom pressuredoes not explain the apurnal tides (frequencies
of 1.003and 1.932cpd, respecparent
temperature
coefficient,the BPR and LBT temtively), we examinethe possibilitythat tidal absolute
oceanheightchangesare the drivingforce. Data from perature data showsomeinterestingsimilarities. Figa bottompressurerecorder(BPR) on Axial Seamount ure 6b showsthe temperaturerecordfrom the BPR, and
[Fox,1990,1993]showa strongtidal signaturewith an Figure 6c showsthe BPR temperaturerecordoverlain
amplitudeof about +15 kPa (Figure6a). Becausethe with our LBT temperature record (both signalshave
BPR data are effectivelytidal (spectralamplitudes7-8 had their computedmeansremoved).The temperature
ordersof magnitudehigherin tidal bandsthan outside waveformsmatch very well, except for an apparent time
thosebands),and our LBT temperatureand tilt data offset of 8-12 hours with Rhonda's temperature leading.
are much broader band than the BPR data, we con- This waveformagreementis an important confirmation
clude that bottom pressureis not the driving forcefor that our temperature measurementmethod works.
timates of the power spectra for tilt and temperature
20,280
ANDERSON ET AL.' SEAFLOORLONG-BASELINETILTMETER
Ocean currents could drive the observedtemperature
and tilt signals. Currents could drive temperature fluctuations either by modulating the rate of heat lossfrom
the volcanoor by moving parcelsof relatively warm or
cool water over the tiltmeter; either mechanism could
explain the time offset between the LBT and BPR tem-
perature records.A spuriousapparenttilt signalcould
result from current-drivenwobblingof the bulky end
pot assemblyor wiggling of somepart of the tiltmeter
tubing.
Seafloorelectricfield variationsin the frequencyband
the semidiurnal tidal band; this weak coherencemight
be partially governedby the 2-km physicalseparation
between
Pele and Rhonda.
We conclude
that
ocean
currents are at least partially responsiblefor apparent
tilt as recorded by our long-baselinetiltmeter but that
other as yet undetermined forces are also driving the
observed tilt.
Comparison
Seafloor
With
Short-Baseline
Tiltmeters
of interest (up to the semidiurnaltide) are primarily
If currentsare at least partially responsiblefor appardriven by barotropic ocean motionscouplingthrough ent tilt in our long-baselinetiltmeter, currents should
the Lorentz force (E-v x B) [Filloux, 1987] so we have an appreciable effect on the short-baseline tiltmay use the electric field recordswe collectedduring
our experiment as a proxy for ocean currents to test
the current-forcing hypothesis. Figure 7a showsthe
electric fields recordedby Ulysses,Macques, and Pele
after a best-fit spline is removed and after rotation to
a commonorientation (rotation anglesdeterminedby
meters we deployed. Figure 8a showsthe residual tilt
from these four instruments after a best-fit spline has
convertedto currentspeedsof about5-10 cm s-1. Mul-
instruments.
titaper estimates of coherenceand cross spectra show
high coherencebetween instruments with zero relative
phase, which indicates that all three instruments were
recordingthe same signal.
Figure 7b shows representative electric field spectra
sistentwith the interpretationthat currentsare rocking
the tiltmeters, as Karen would be relatively sheltered
been removed (Tolstoy et al., submittedmanuscript,
1997), with the long-baselinetilt data from Rhonda
shown for comparison. On three of the short-baseline
tiltmeter records, tidal oscillations of the order of 4-5
Heinsonet al. [1996];channeli is northpositive).The /•rad are evident. The fourth instrument, Karen, was
residual electric field showsa tidal signaturewith typi- deployed inside the summit caldera and showsa noise
cal variationsof the orderof 4-3/•V m-1, whichcanbe signature much reduced in relation to that of the other
The lower noise levels on Karen
inside the caldera
are con-
and thus would be moved less than
the other tiltmeters.
Figure 8b showsrepresentativepower spectra for the
(from Macques),while Figure 7c showsa spectrumof short-baseline
tilt data from Judy. Both spectra(X
currentvelocitydatafrom CannonandPashinski[1990] and Y component)exhibit distinct diurnal and semidfor comparison. Both spectra exhibit similar peaks at iurnal tidal peaks, as one might expect if theseinstrufrequencies of about 1, 1.5, and 2 cpd with a broad
peak centered near 0.25 cpd. The peaks at I and 2
cpd correspond to tidal currents at the periods of the
diurnal and semidiurnal tides, respectively,while the
peak at 1.5 cpd correspondsto Coriolis-coupled wind-
ments were recordingearth tides. However,the inertial
peak at 16.7 hours period and the 4-day oscillationare
also evident in the spectra from the short-baselinetiltmeters that
were installed
on the caldera rim.
These
two peaksare current-related[Cannonand Pashinski,
driven currents. Cannon and Thomson[1996]identify 1990], and given that these instrumentsare not perthe broad peak at about 0.25 cpd as delineatingcurrents fectly coupled to the seafloor,it is reasonableto condriven by local weather; these currentsinteract with the clude that somecurrent-drivenwobblingis contained
rough topography of the JDF to create northward propagating ridge-trapped waves,which travel at about i m
in the short-baseline
tilt records.
If currents are rocking both the short- and longs-1. The peak at 3 cpd is due not to currentsbut to baselinetiltmeters, the time seriesrecordedby theseinthe geomagneticsolar daily variation [Filloux, 1987]. strumentsshouldbe coherentat frequenciesoutsidethe
This comparison confirms that our electric field mea- known tidal bands. Unfortunately, the short-baseline
surementsreflect the ocean currentsin our deployment tiltmeter nearestRhondais Janice,whichdid not give
area at periods longer than 8 hours.
usefulrecordings,and the next nearest,Phred,recorded
We test for ocean current forcing by estimating co- too few data to give reliable coherenceestimates. We
herences between the temperature and tilt data from are reducedto usingdata from the next nearestinstruRhonda and the electric field records from the nearest
ment, Judy, which is 3.5 km away. The coherencebeelectromagneticinstrument, Pele. We observemoder- tweenJudy'stilt and Rhonda'stilt dependson compoately strong coherencebetween the electric field and nent and is not very strong,but it is significantat the
LBT temperature record in the semidiurnal tidal fre- 99% levelin the 4-day band and the inertial band, as
quency band as well as significant, but lower, coherence in the diurnal tidal, inertial, and 4-day oscillation
bands. We also note significant,though weaker,coherence betweentilt and the electricfield, particularly in
well as the diurnal and semidiurnal tidal bands. There
are qualitativesimilaritiesas well; Figure8a showsthat
the increased
noiseon Rhonda'srecordfromaboutdays
198-207 is also evident on Judy and Lynn. It is thus
ANDERSON ET AL.' SEAFLOOR LONG-BASELINE TILTMETER
20,281
50
40
30
20
10
0
-10
-20
-30
-40
-50
200
210
220
230
240
250
Time (UTC day of 1994)
103
102
• 00
0-1
10-2
10-1
i i iiiiiiI i i ii....
1
100
101
Frequency(cpd)
Figure 8. (a) Residualshort-baseline
tilt datafromPhred,Judy,LynnandKarenafterremoval
of a best-fit spline.Thick line is X component,and thin line is Y component.Note that Karen's
noise level is about 10 times lower than the levels of the other instruments, a decreasethat may
reflectthat instrument'srelativelyshelteredlocationin the caldera.RhondaLBT data are shown
for comparison.
(b) Powerspectrafor the short-baseline
tiltmeterJudy.Thicklineis X, andthin
line is Y. Both spectradisplaythe 4-day and 1.5 cpd oscillations,whichare currentdriven.
reasonableto suggestthat currents are rocking both ent (lower)rate. Eventually,both the short-and longthe short-baselinetiltmeters and Rhonda, contributing baselinetiltmeters are likely to reach some low stable
high-frequencynoiseto the data recordedby both kinds tilt rate.
of instruments.
It is apparent from Figure 8a that our long-baseline Comparison With Subaerial Volcanic
tilt data have a high-frequencynoise level that is only Tilt Observations
about a factor of 2 larger than that of the short-baseline
data; how do the long-term drift rates compare? TolThe long-baselinetilt data from Rhonda exhibit long-
stoy et al. (submittedmanuscript,1997) estimatethe term drift ratesof 5/•rad day-1 nearthe beginningof
to lessthan 1 /•rad day-1 near
drift rates (after the initial instrumentsettling)for the the record,decreasing
short-baselinetiltmeters to range from about 1 to 10 day 240, as well as a higher-frequencynoise level of
/•rad day-1; thesevaluesare comparableto the drift about :t:6/•rad. Given these drift rates and noiselevels,
ratesin our long-baseline
data (0.5-5/•rad day-1). In it is important to ask whether or not we would be able
both casesthe drift rates decline with time. However,
the character of the decline is fundamentally different:
the SBT data stabilize quasi-exponentially, while the
LBT data tend to exhibit stable tilt rates, punctuated
by abrupt offsets,after which tilt continuesat a differ-
to detect known volcanicsignalswith this instrument;
we may appeal to long time series of tilt recorded on
active subaerial volcanoesto addressthis question.
The best suchlong-term recordis that from the openended fluid tube tiltmeter operated at Uwekahunavault
20,282
ANDERSON ET AL.' SEAFLOOR LONG-BASELINE TILTMETER
18oo
I
I
I
I
I
16oo
_
12oo
_
I
1000
2000
0
4000
6000
8000
10000
12000
Time (days)
30
I
i
i
i
i
b
20
•
10
•
0
_
_
n- -10
-20
_
I
-30
0
I
2000
4000
I
6000
I
8000
10000
12000
Time (days)
8OOO
40
6OOO
30
'
I'1
E 2O
E 4000
z
z
lO
2OOO
-lJo
o
Tilt Rate (•rad/day)
lJo
o- 100
-50
0
50
Tilt Rate (prad/day)
loo
Figure 9. (a) Recordof tilt from Uwekahunavault at HVO. The breakin the data at about day
9100 is due to operator intervention. Note steady long-term inflation punctuated by rapid tilt
rates near eruptions.(b) Daily first differences
of the data in Figure9a. Dashedline is shownat
4-6 •rad to reflect the typical high-frequencynoiselevel observedon Rhonda. (c) Histogramof
Figure 9b. (d) As in Figure9c, exceptthat only binsfor tilt ratesin excessof twice the standard
deviationare shown.Note the high tilt rate eventscorresponding
to eruptiveinflation/deflation
events.Thereare morethan 30 eventswith deflationtilt ratesin excess
of 20 •rad day-1. Dotted
lines are shown at 4.6 •rad for comparison.
(about 300 m from Kilauea Caldera)by the HawaiiVolcano Observatory. Figure 9a showsa 32-year record
from the Uwekahuna LBT. Steady rises over several
months, reflecting slow inflation of a shallow magma
chamber, are plainly evident. These rises are punctuated by abrupt drops in tilt, which correspondto
catastrophic deflation of the volcano during eruption
or drainageof the central magma chamberinto a flank
or rift zone reservoir.
We have computed daily first differencesof the Uwekahunatilt recordto givean estimateof the background
noiseobservedwhile operating a long-baselinetiltmeter
on an active volcano,as well as to quantify the tilt rates
associated with eruptive or intrusive deflation events
(Figure 9b); histogramsof these first differencesare
shownin Figures 9c and 9d. The mean daily variation
observedin the Uwekahunarecordis 0.02 •rad day-1,
with a scatterof about 3.2 •rad day-1, whichreflects
ANDERSON ET AL.: SEAFLOOR LONG-BASELINE
TILTMETER
20,283
backgroundgroundand instrumentnoise. Closerex- measure, and bottom pressuresensorsof the type used
aminationof the tilt rates (Figures9b and 9d) shows on the BPR would help us to compensatefor the effects
many high tilt rate events, primarily deflations,corre- of these sources of environmental noise.
There are fluids with densitiesgreater than ethylene
sponding to eruptive or intrusive episodesat Kilauea.
The largest of these events have tilt rates that can ex- glycol(althoughthey aregenerallyharderto workwith)
ceed100/•rad day-1, with manymoremoderateevents that would provide larger pressuresignalsfor a given
displaying
tilt ratesof greaterthan 10/•radday-x [e.g., tilt. While it is not yet clear what the optimal length
Duffieldet al., 1976;Mooreet al., 1980;Decker,1987]. of the tiltmeter is, longer instrumentsare likely to proSmaller eventsalmost certainly grade smoothly into the vide a better signal-to-noiseperformance. The maximumpractical
instrument
lengthis limitedl•yseafloor
backgroundnoise on the record.
roughnessbut in any caseprobably will not exceedthe
In Figures 9b and 9d we have indicated the +6/•rad
high-frequencynoise seen on our LBT record as a vi- water depth (usuallyin the rangeof 1 to 4 km).
It is currently technically infeasible to construct a
sual measure of the ability of our instrument to record
that will be
these eruption signatures. While the backgrounddaily seafloortiltmeter (short- or long-baseline)
tilt on Kilauea is roughly half that of our current instru- capable of recording tilt events with rates typical of
ment, it is clear that moderate to large volcanic events continental-style plate boundary tectonics; suchsignals
at Kilauea have tilt rates well above the drift rates and
are just too small in relation to instrument and environnoiselevelswe observein our record of long-baselinetilt. mental noise. However, tilt rates associatedwith erupThe fact that we did not record any eruptive signature tive or intrusive events at subaerial volcanoes are sevat Axial Seamountsuggests
that no activeeruptionwas eral ordersof magnitude larger than the rates observed
occurring at the time of our deployment;data from the at active continental plate boundaries and are within
short-baseline tiltmeters, nearby independent bottom the reach of current technology.Potentially greater inpressurerecorders,and the U.S. Navy's OregonSOSUS stability of volcanoes on relatively thin oceanic crust
(SOund SUrveillanceSystem)arrays are in agreement and similarities in magma emplacementrates between
with this interpretation. By combiningour long- and Hawaiian and MOR volcanoesmay lead one to expect
short-baseline
tilt records we can also bound the rate of
that MOR volcanoeswould display deformation behavinflation or deflation on Axial Seamount to be less than
ior similar to that of subaerialvolcanoes,but this theory
still hasto be verifiedthroughdirect observation.These
0.5-1 prad day-x duringmid-1994.
argumentsand the data presentedin this paper lead us
to
concludethat the long-baselineinstrument described
Future
Directions
in this paper is capable of recordingrapid volcanictilt
Although the necessityof developingand testingthe on seamountsor active MOR segments.Long-duration
new tiltmeter by actual deployment has made its devel- monitoringexperimentsusingboth improvedtiltmeters
opment a long and difficult process,we have achieved and ocean bottom seismographsare necessaryto pronotable success in that we have an instrument that not
vide order-of-magnitudeboundson the deformationat
only works but is capablein principle of measuringthe and the eruptive behavior and internal structure and
signal of interest. The cost and performance of the new dynamics of active submarine volcanoes.
tiltmeter are comparable to short-baseline tiltmeters,
Acknowledgments.
Tom Deaton and JacquesLemire
and becauseof its ability to measure a broad deformaprovided
technical
support.
We thank the captain and
tion field rather than local instabilities,a long-baseline
crew
of
R/V
Wecoma
for
support
in both the deployment
tiltmeter is clearly a desirable instrument.
and recovery cruisesand Chris Fox and Harold Mofjeld for
Given our understandingof the in situ instrument providing pressure and temperature data. Deployment of
performancegainedby the analysispresentedin this pa- the short-baseline tiltmeters was conducted in collaboration
per, we would expect to achieve better noise and drift with Maya Tolstoy and John Orcutt, who also gavehelpful
levels in future deploymentsof the tiltmeter, particu- advice and criticism. We thank Roger Denlingerfor supplylarly those that are longer than the 2-month equilibra- ing the UwekahunaLBT tilt record. Paul Davis kindly provided computer codesfor computing the prolate spheroidtion time of the instrument. As ocean currents appear driven deformation field. We thank Roger Bilham, David
to be a noise source in our tiltmeter, we could do two Clague,Megan Flanagan,and Alan Linde for insightfulcritthings to improve the seafloorcouplingand reducethe icismswhich improved this paper. This work was supported
current-drivennoise:(1) the currentgenerationof data by NSF grants OCE-8911428 and OCE-9200879.
loggersare about one third the size of the onesused on
Axial Seamountand so will presenta smallercrosssec-
tion to oceancurrents,and (2) the constructionof the
end pot assemblycan be rationalized to make the as-
References
Agnew, D.C., Strainmetersand tiltmeters, Rev. Geophys.,
2•, 597-624, 1986.
semblysmaller,heavier,and radially symmetric.Data
from ancillary sensorssuch as current meters or elec- Baragar, W.R.A., M.B. Lambert, N. Baglow,and I. Gibson,
Sheeteddykesof the Troodosophiolite, Cyprus, in Mafic
trodeson the tiltmeter to measurewater motion,therDyke Swarms, edited by H.C. Halls and W.F. Fahrig,
mistors at the pot and sensorends to give point esSpec.Pap. 3•, 257-272, Geol. Assoc.Canada, St. Johns,
timates of temperatureto complementthe integrated Newfoundland, 1987.
20,284
ANDERSON ET AL.: SEAFLOOR LONG-BASELINE TILTMETER
Beavan, J., and R. Bilham, Thermally induced errors in fluid
tube tiltmeters, J. Geophys.Res., 82, 5699-5704, 1977.
Boyer, H.E. and T.L. Gall, (Eds.), Metals Handbook,Am.
Soc. for Metals, Metals Park, Ohio, pp. 49-52, 7-2, 1985.
Bradher, H., J.G. Dodds, and R.E. Foulks, Investigation
of microseisin sourceswith ocean bottom seismometers,
Geophysics,30, 511-526, 1965.
Cannon, G.A., and D.J. Pashinski, Circulation near Axial
Seamount, J. Geophys.Res., 95, 12,823-12,828, 1990.
Cannon, G.A., and R.E. Thomson, Characteristicsof 4-day
oscillationstrapped by the Juan de Fuca Ridge, Geophys.
Res. Left., 23, 1613-1616, 1996.
Chadwe!l, C.D., F.N. Spiess,J.A. Hildebrand, L. Prawirodirdjo, L.E. Young, G.H. Purcell Jr., and H. Dragert, Juan
de Fuca Geodesy Project: Strain and plate motion measurement using GPS and acoustics, EOS Trans. A GU,
76(46), Fall Meet. suppl.,F412, 1995.
Constable, C.G., and R.L. Parker, Smoothing, splinesand
smoothing splines: Their applications in geomagnetism,
J. Cornput. Phys., 78, 493-508, 1988.
Constable, S., and C.S. Cox, Marine controlled-sourceelectromagnetic sounding,2, The PEGASUS experiment, J.
Geophys.Res., 101, 5519-5530, 1996.
Davis, P.M., P.A. Rydelek, D.C. Agnew, and A.T. Okamura,
Observationof tidal tilt on Kilauea Volcano,Hawaii, Geophys. J. R. Astron. Soc., 90, 233-244, 1987.
Decker, R.W., Dynamics of Hawaiian volcanoes: An overview, in Volcanism in Hawaii, edited by R.W. Decker,
T.L. Wright, and P.H. Stauffer, U.S. Geol. Surv. Prof.
Pap., 1350, 997-1018, 1987.
Duffield, W.A., D.B. Jackson,D.B., and D.A. Swanson,The
shallow, forceful intrusion of magma and related ground
deformation at Kilauea Volcano, May 15-16, 1970, in Proceedingsof the Symposiumon Andean and Antarctic VolcanologyProblems, edited by O.G. Ferran, pp. 577-597,
Int. Assoc. of Volcanol. and Chem. of the Earth's Int.,
Rome, 1976.
Klein, F.W., Eruption forecasting at Kilauea Volcano, Hawaii, J. Geophys.Res., 89, 3059-3073, 1984.
Klein, F.W., R.Y. Koyanagi, J.S. Nakata, and W.R. Tanigawa, The seismicityof Kilauea's magma system,in Volcanism in Hawaii, edited by R.W. Decker, T.L. Wright,
and P.H. Stauffer, U.S. Geol. Sur. Prof. Pap., 1350, 10191185, 1987.
Kuehne, J., S. Johnson, and C.R. Wilson, Atmospheric excitation of nonseasonalpolar motion, J. Geophys.Res.,
98, 19,973-19,978, 1993.
Mogi, K., Relations between the eruptions of various volcanoes and the deformationsof the ground surfacesaround
them, Bull. Earthquake Res. Inst. Univ. Tokyo, 36, 99134, 1958.
Moore, R.B., R.T. Helz, D. Dzurisin, G.P. Eaton, R.Y. Koyanagi, P.W. Lipman, J.P. Lockwood, and G.P. Puniwai,
The 1977 eruption of Kilauea Volcano, Hawaii, J. Volcanol. Geotherrn. Res., 7, 189-210, 1980.
Park, J., C.R. Lindberg, and F. L. Vernon III, Multitaper
spectral analysis of high-frequencyseismograms,J. Geophys. Res., 92, 12,675-12,684, 1987.
Percival,D.B., and A.T. Walden, SpectralAnalysisfor Physical Applications: Multitaper and Conventional Univariate Techniques,pp. 289-295, CambridgeUniv. Press,New
York, 1993.
Pollard, D.D., P.T. Delaney, W.A. Duffield, E.T. Endo, and
A.T. Okamura,Surfacedeformationin volcanicrift zones,
Tectonophysics,9d, 541-584, 1983.
Priestley, M.B., SpectralAnalysis and Time Series, pp. 664,
68!-685, AcademicPress,San Diego, Calif., 1981.
Rothery, D.A., The baseof a sheeteddyke complex,Oman
ophiolite: Implications for magma chambers at oceanic
spreadingaxes, J. Geol. Soc. London, 1•0, 287-296, 1983.
Sigurdsson,
H., Dyke injectionin Iceland: A review,in Ma.fic
Dyke Swarms, edited by H.C. Halls and W.F. Fahrig,
Spec. Pap. 3d, 55-64, Geol. Assoc. Canada, St. Johns,
Newfoundland, 1987.
Dvorak, J.J., A.T. Okamura,T.T. English,R.Y. Koyanagi, Sinha, M.C., D.A. Navin, L.M. MacGregor, S. Constable,
J.S. Nakata, M.K. Sako, W.T. Tanigawa, and K.M. YaC. Peirce, A. White, G. Heinson,and M.A. Inglis, Evimashira, Mechanical response of the south flank of Kidencefor accumulatedmelt beneath the slow-spreading
lauea Volcano, Hawaii, to intrusive eventsalong the rift
mid-Atlantic ridge, Philos. Trans. R. Soc. London,Set.
systems, Tectonophysics,12d, 193-209, 1986.
A, 355, 233-253, 1997.
Dziak, R.P., C.G. Fox, and A.E. Schreiner,The June-July Staudigel,H., D. Agnew,S.C. Constable,C. Cox, J. Orcutt,
1993 seismo-acousticevent at CoAxial segment,Juan de
G. Sasagawa,
S. Webb,andF. Wyatt, In situ monitoring
Fuca Ridge: Evidence for a lateral dike injection, Geoof microseismicity
andtilt at mid oceanridges:A progress
phys. Res. Left., 22, 135-138, 1995.
report, EOS Trans. AGU, 71(43), Fall Meet. suppl.,1601,
1990.
Embley,R.W., K.M. Murphy, and C.G. Fox, High-resolution
studiesof the summit of Axial Volcano,J. Geophys.Res., Vernon, F.L., J.B. Fletcher, L. Carroll, A.D. Chave, and
95, 12,785-12,812, 1990.
Filloux, J.H., Instrumentation and experimental methods
for oceanicstudies,in Geornagnetisrn,
edited by J.A. Jacobs, pp. 143-248, Academic Press, San Diego, Calif.,
1987.
Fox, C.G., Evidence of active ground deformation on the
mid-ocean ridge: Axial Seamount, Juan de Fuca ridge,
April-June 1988, J. Geophys. Res., 95, 12,813-12,822,
1990.
Fox, C.G., Five years of ground deformationmonitoring on
Axial Seamount using a bottom pressurerecorder, Geophys. Res. Left., 20, 1859-1862, 1993.
Heinson, G., S. Constable, and A. White, Seafloormagnetotelluric soundingaboveAxial Seamount,Geophys.Res.
Left., 23, 2275-2278, 1996.
Johnson, H.P., Processesassociatedwith ocean crustal formation: The Juan de Fuca Ridge, Geophys.Res. Left.,
œ0, 1847-1850, 1993.
Johnson, R.V., C.R.B. Lister, and B.T.R. Lewis, A direct recordingoceanbottom seismometer,Mar. Geophys.
Res., 3, 87-102, 1977.
E. Sembera, Coherence of seismic body waves from local
events as measured by a small-aperture array, J. Geophys.
Res., 96, 11,981-11,996, 1991.
Washburn,E.W. (Ed.), International Critical Tablesof Numerical Data, Physics, Chemistry, and Technology,vol. 3,
pp. 27-42, McGraw-Hill, New York, 1930.
Webb, S.C., S.C. Constable, C.S. Cox, and T.K. Deaton, A
seafloor electric field instrument, J. Geornagn. Geoelectr.,
37, 1115-1130, 1985.
Westphal, J.A., M.A. Carr, W.F. Miller, and D. Dzurisin,
Expendable bubble tiltmeter for geophysicalmonitoring,
Rev. Sci. Instrurn., 5d, 415-418, 1983.
Willoughby, D.F., J.A. Orcutt, and D. Horwitt, A microprocessor-based ocean-bottom seismometer, Bull. $eisrnol. Soc. Am., 83, 190-217, 1993.
Wyatt, F., and J. Berger, Investigations of tilt measurements using shallow borehole tiltmeters, J. Geophys.Res.,
85, 4351-4362, 1980.
Wyatt, F., R. Bilham, J. Beavan, A.G. Silvester, T. Owen,
A. Harvey, C. Macdonald,D.D. Jackson,and D.C. Agnew, Comparing tiltmeters for crustal deformation mea-
ANDERSON ET AL.: SEAFLOOR LONG-BASELINE TILTMETER
surement: A preliminary report, Geophys.Res. Left., 11,
963-966, 1984.
Wyatt, F.K., S.-T. Morrissey, and D.C. Agnew, Shallow
borehole tilt: A reprise, J. Geophys.Res., 93, 9197-9201,
1988.
Wyatt, F.K., J.A. Orcutt, G. Sasagawa,H. Staudigel, and
P. Zimmer, Toward in situ monitoring of active submarine volcanoes:A progressreport, Fifth Circum-Pacific
Energy and Mineral ResourcesConference Transactions,
edited by G.P. Salisbury and A.C. Salisbury,pp. 757-773,
Gulf, Houston, Tex., 1996.
Yang, X.M., P.M. Davis, and J.H. Dieterich, Deformation
from inflation of a dipping finite prolate spheroid in an
20,285
elastic half-spaceas a model for volcanicstressing,J. Geophys. Res., 93, 4249-4257, 1988.
G. Anderson,S. Constable,H. Staudigel, and F. K. Wyatt, Institute of Geophysicsand Planetary Physics, Scripps
Institution of Oceanography, La Jolla, CA 92093-0225.
(e-mail: [email protected];
[email protected];
[email protected];
[email protected])
(ReceivedNovember22, 1996; revisedMay 15, 1997;
acceptedMay 23, 1997.)