Nov 10, 1990 - Berkeley, and the Center for Computational Seisinology, Earth. Sciences Division .... later episode of deformation as a single stage of overall volumetric ..... The only departure from a Gaussian distribution is the increase in ...
JOURNALOF GEOPHYSICAL RESEARCH, V()L. 95,NO. BI2. PAGES19.839-19,856, NOVEMBER10,1990
Inversionfor Sourcesof CrustalDeformationand Gravity Change at the Yellowstone
Caldera
D. W. VASCO1 Geophysics Laboratory,Hanscorn Air ForceBase,Massachusetts
R. B. SMITH Department of Geology andGeophysics, University of Utah,SaltLakeCity,Utah
C. L. TAYLOR Geophysics Laboratory,Hanscorn Air ForceBase,Massachusetts
The 3100km2 Yellowstone caldera, Yellowstone NationalPark,wasformedin thelatest of three
explosive eruptions of rhyolites andashflowtuffstotalling 3700km3 at 2, 1.2,and0.6m.y.before
present. Its youthfulvolcanichistory,widespread hydrothermal activity,intenseseismicity,and extremely highheatflow,/n excess of 30 timesthecontinental average, markstheYellowstone volcanic system asa giantcaldera at unrest.Orthometric heightincreases of thecaldera of upto 76 cm,measured fromprecise levelingsurveys from 1923to 1975-1977, were/averted to determine volumeexpansion sourcemodelsfor the caldera-wide deformation.For the 1923to 1977uplift episode,two regionsof
expansion werefound: (1)in thenorthern partofthecaldera neartheSourCreekresurgent domeof -0.37
krn 3,and(2)inthesouthern part ofthecaldera, near theMallard Lake resurgent dome of -0.41km3. Both bodies occurin theuppercrustfromnear-surface depths to 6.0kin,butthelargest volumeexpansions were foundin the3.0-6.0kmdepthrange.Thesouthern caldera source volume, neartheMallardLakedome,may extenddownto 9.0 kin. The dataare,however,unableto resolveif thesetwo resurgentdomes,separated
laterallyby-40 km, areconnected. From1976to 1987,nearlysimultaneous measurements of elevation andgravitychanges weremadeona profileacross thenorthern caldera duringa periodof netuplift Models
of thetemporal gravityvariation inferthatthevolume increase for thenorthern caldera source mustlie
above 9.0kmandinvolved a density perturbation greater than +0.002 g/cm 3. Themodeled volumetric
sources arein thesamegeneral locations asbodies of lowP wavevelocities, highseismicattenuation, and largenegative Bouguer gravityanoma•es.In view of theintense Quaternary volcanism, theanomalously highheatflow, andthecorrelation of thesource volumes withtheregional geophysical anomalies, it is likelythatthemodeled volumetric increases werecaused by migration of megmss and/ortheintroduction of largevolumesof hydrothermal fluidsinto the uppercrust.
timesthecontinental average)andtheremarkable crustaluplift of up to 1 m from 1923 to 1985 of 'theYellowstonecaldera, of 6 cm from 1985 to 1987,that sets Yellowstone National Park, in northwestern Wyoming followedby subsidence of (Figure1), is the siteof oneof the world'slargestandmost Yellowstoneapartas a giantcalderaat unrest. Summaries activehydrothermal systems (geysers, hot springs, fumeroles, the geologyand geophysicsof the Yellowstonevolcanic etc.)encompassing the3100km2 Yellowstone Plateau andis systemby Smith and Christiansen[1980], Christiansen thesiteof a largeQuaternary volcanic system.In thepast2 [1984],SmithandBraile [1984], andNationalAcademyPress for furtherreading. m.y.threecatastrophic volcaniceruptions haveexpelledmore [ 1987] arerecommended INTRODUCI•ON
than3500 km3 of rhyoliticashflow tuffs formingthree
In the mostrecentcaldera-formingeruption,600,000 years
calderas [Christiansen, 1984].An additional 3000km3 of ago,drainageof themaLnmagmachamber released roofsupport it to collapse andproduced the45 km wideby 75 km rhyoliteflowsandashflow tuffsweredeposited betweenthe causing caldera"(Figure2) [Christiansen,1984]. explosive eruptions. Intense earthquake activity,including the long "Yellowstone resurgence thenformedtwo structural domeswithin Intermountain region'slargesthistoricearthquake, the nearby Mag-marie
1959,ML 7.5, HebgenLake,Montanaeventandextensivethe caldera:(1) the Sour Creek dome in the northeastcaldera earthquake swarmswithinthe Yellowstone calderareflect that beganits resurgenceat the time of the main eruption contemporary tectonicdeformation.However,it is the 600,000 years ago, and (2) the Mallard Lake dome in the
extremely highheatflowof-1500mWm'2 (inexcess of-30
southern caldera, that began doming at 150,000 years [Christiansen, 1984]. The Quaternary tectonics of the Yellowstone Plateau is 1Nowat Seismographic Station, University of California,dominatedby structures relatedto two recenttectonic/magmatic Berkeley, andthe Centerfor Computational Seisinology, Earth events. The oldernormalfault systems,the GallatinandTeton Sciences Division,Lawrence BerkelyLaboratory, Berkeley. fault zones, are of Late Tertiary Basin-Range origin and
Copyright 1990bytheAmerican Geophysical Union. Papernumber90IB01402. O148-0227/9 0/90JB-O 1402505.00
generallytrendnorth-south at the northand southmarginsof the plateau,respectively(Figures1 and 3). Segmentsof these large north to northwest trending normal fault systems 19,839
19,840
VASCOET AL.: SOURCES OFCRUSTAL DEFORMATION ATYELLOWSTONE !!1 ø 15' W
! IOøOO' W 45ø!5'N
HOT SDœ/#GS ß
YELLOW•TONE
,
FA I I'HFl./L
i
44ø00'N
o
QUATE
• SuRFIclAL
rHYou,te
fo
R NAR"i'
,• DEPOSITS
5
TERTIARY
-.1-'7• '----'• "-'•FLows
eAsAuT
weuoeo AsH FLOWS
F,.'] ,nTermeolate VOLCANIc rOCKs
PR
E-TER
TIAI•Y
::::• crYSTALLINE SœDIMœNTARY
AND ROCKS
Fig. l. Generalgeo]oõicmap oœthe Yellowstone?lateau(modLfiedfrom Christiansen []954, ]989] and Claw•o• el •1., 1989) sho•n• •c relationshipof the caldera,•uatcma• resurscmdomes,and locationsof •e ]•ve]•8 Hncs. [, 2-m. y. caldera•unda•; •, ].2-m. y. calderabounda•, and• = 6•,000 yearold "•c•owstonecaldera"•unda•.
presumablyexist in an area now coveredby the Quaternary active crustal deformation,high topography,and extremely volcanicsof the YellowstonePlateau. high heat flow. These aspectsof the Yellowstoneregionare An additionalareaof young faultingoccurson the southeast also reviewed by Smith and Christiansen [1980] and Smith and side of the caldera where north-trendingnormal faults, with Braile [1984]. A variety of geophysical evidence, when Holocene displacements, extend along the east side of interpretedwith the Quaternaryvolcanic history, suggeststhe Yellowstone Lake (Figure 3). At the northwest side of the presenceof magmas, partial melts, and hydrothermal activity plateau,the seismicallyactive Hebgen Lake fault zone extends at midcrustalto upper crustaldepths[Smith et al., 1974, 1977; -30 'Ionwest-northwestbeyond the caldera. Regionalnorth to Eaton et al., 1975]. Measurements of crustal deformation of northwest trending epicenters and the general north to the caldera in the mid-1970s highlighted the dynamic northwestalignment of volcanic vents in the western caldera propertiesof the Yellowstone volcanic system. Evidencefor [Christiansen, 1989] may be related to the buried, but still the unusual crustal deformation of the Yellowstone Plateau was activeQuaternaryfaults. A circularfault systemaccommodated shownby Pelton and Smith [1979, 1982] for up to 76 cm of the collapse of the 600,000-year-old Yellowstone caldera caldera-wideupld.t baseduponrepeatedleveling of benchmarks (Figures1 and 3) and is seenas the calderarim-boundingfault in 1975-1976-1977 that were originally establishedin 1923 zone and the inner-calderarim fracturesystem. The youngest (Figure 2). From the mid 1970s to 1984, leveling surveys normal faults of the caldera are thoseassociatedwith central revealed an additional 25 cm of uplift, but by 1987 the grabensof the resurgentdomesand northeasttrendingfaults deformationreversed to subsidenceof .-6 cm [Dzurisin and alongthe top of the ElephantBackMountainthatoffsetglacial Yamashita, 1987]. Precisiongravity measurements,initially
and alluvial material. The normal faults along the top of made at leveling benchmarksin 1977 [Evoy, 1977], and ElephantBack Mountain appear to have accommodatedthe observed again in 1987 [Hollis et al., 1987], showed a net uplift of this structuralhigh. decrease of up to 30 IlGal. The areaof gravitydecrease occurred The Yellowstone Plateauis markedby intenseseismicity, in the same general location as the crustal uplift maxima.
lowuppercrustal P wavevelocities, lowuppercrustal densities,Togethertheseobservations suggestconcomitant volumetric
VASCO ETAL.: SOURCES OFCRUSTAL DEFORMATION ATYELLOWSTONE
19841
•. r,.•x.•.YELLOWSTONE NATIONAL PARK./"•
I
45000' N-•-
I
m
iiioo0,w •
i•Oo•O'W k...\
oo
i •2•
,
WEST
YELLOWSTONE_
.
'
I
\
'""0.
:
• '
•
"•-"---"•'"'••
DOM
: .
• KI2 /I----•MA
,•'
oO
k '-.
-
.g•o'u+k "-
,oo•
• •,.oo. wk '-.
/
':.
:
..•'
"lOW
I
o
o
Fig. 2. Map of 1923to 1975-1976-1977surfaceuplift of the Yellowstonecaldera[fromPeltonand Smith,1982]. Contours are in millimeters relative to a referencestation,K12, eastof the calderaand assumedleast influencedby volcanic or tectonic activity. Inset showslocationsof level lines by numbersreferencedin the text. Dottedlinc showsinner cdgcof calderarim fracturezone. EBFZ, ElephantBack fault zone;MA, MammothHot Springs;TF, Tower Falls; WT, West Thumb; MJ, Madison Junction;NI, Norris Junction;CJ, CanyonJunction;FB, FishingBridge; LB, Lake Butte.
section we discuss the procedures involved in analyzing the leveling data from the two time intervals: 1923 to 1975-1977 volumetric increase and modeled the orthometric height and 1977 to 1987. Recognizing the reversal to subsidencein increaseand gravity decreaseto constrain the mass-volume 1985 [Dzurisin and Yamashita, 1987], we modeled the net ratio. The primary focusof this paper is thus a discussionof uplift from 1977 to 1987 as a single episodeincorporatingthe the crustal deformation and an inversion of the 1923 to 1975constraints of temporal gravity changes. Because of the 1976-1977uplift and the 1977 to 1985 net uplift and gravity differencesin data distribution, two diverse approacheswere decreaseof the Yellowstone caldera. We also discussplausible taken to draw conclusions from the data.
and massperturbationsin the crust. Note that we treated this later episode of deformation as a single stage of overall
source mechanisms that could account for the deformation of
First, we consider the 1923 to 1975-1977 interval in which
the caldera.
it was possible to derive a model of net volumetric increase from the leveling data (Figure 2). Most of the leveling data used in this study were discussedextensively by Pelton and CRUSTAL DEFORMATION OFTHEYœLLOWSXONE C^L• Smith [1982]. The earlier leveling campaign of 1923 was conductedin order to establish elevations for the highways LevelingData within Yellowstone National Park [Pelton and Smith,1982]. the routesfollow the majorroadwaysof the park Repeatedleveling surveyshave providedthe most direct Consequently, evidence for the historic deformation of the Yellowstone (Figure 2). This second-ordersurvey,carried out by the U.S. caldera(Figure2) with nearly1 m of upliftbetween1923and Coast and Geodetic Survey between late July and early 1985at ratesof 1.5-2.5cm/yr[Peltonand Smith,1979,1982; November 1923, was predominantlysingle run. The more Smith et al., 1989a, b; Dzurisin and Yamashita, 1987]. This recent 1975, 1976, and 1977 level line observations were all unusualuplift, however,has beenepisodicas Dzurisin and doublerun and were conductedby the TopographicDivision of Yamashita[ 1987] haveshownthat the deformation reversedto the U.S. GeologicalSurvey. The field work was donebetween subsidenceof more than 6.0 cm from 1985 to 1987. •
this
mid-August andmid-September 1975andbetweenmid-Julyand
19,842
VA$CO ETAL.: SOURCES OFCRUSTAL DEFORMATION ATYELI•WSTONE 10'
ENATIONAL•
PARK
45'
/'/,6o 50'
40'
30'
Malla Lake
Dome
20'
lo
KM
I .........
I
44*
40'
30'
20'
lO'
lllø
50'
40'
30'
20'
10'
110 ø
50'
Fig.3. Mapof Quaternary faultsof theYellowstone Plateau (faultsfromChristiansen [1989])withoutlineof source volume grid usedin the data inversion.
mid-October1976 and 1977, accordingto first-order,classH specifications [Pelton and Smith, 1982; Federal Geodetic
are essentiallyall positive, indicatingnet uplift of the region over the approximately 53-year interval, except for a small
Control Committee, 1975].
area of subsidence around Norris Junction, in the northwest
The errors in the leveling measurements can be estimated given certain assumptions.In particular,we treat the random error at eachsargonsetupas a zero-mean,independent variable with unknownvariance. These principalslead to a statistical modelin which the measurements are normallydistributedwith a variancegiven by [Peltonand Smith, 1982].
calderathatmay be relatedto a MS 6.1, 1975, earthquake.The largest vertical height increases(greater than 700 ram) were alongthe centralaxis of the caldera,tendingtowardzero at the peripheryfor mostlevel lines (Figure2). Between the 1923 and the !975-1977 surveys, two
O•= fl2Ll/n
(1)
significant earthquakes occurred.The largestwastheML 7.5, 1959 HebgenLake event, whosescarp is located-25 km northwest of the caldera.
Footwall
subsidence associated with
thiseventmay haveeffecteda few of the benchmarksnearthe northwestboundaryof the caldera fReilinger et al., 1977]. Closerto the calderawas the 1975, Ms 6.1, Norris luncdon earthquake,with up to 200 mm of subsidence, locatednear the northern caldera boundary [Pitt et al., 1979]. Many other
where I• is a constantdependingon the equipment and techniquesused in the survey, L i is the distancefrom the leveling line base station to the ith station, and n = 1 for single-runlevelingandn = 2 for double-runleveling. For the 1923 surveys,,8= 3.59 and for 1975-1977surveys,/I =1.34. There is also the possibilitythat systematicerrorsor blunders are presentin the data,e.g., an elevationdependent bias. Such possibilitieswere examinedin detail by Pelton and Smith [1982],whopresented upperboundson severaltypesof errors. In the analysisthat followswe have includedestimatesof both
Norris event. on our inversion results for the caldera sources.
randomand systematicerrors.
Firstthenegativebenchmark valueswereleft in thedataandan
smaller, but neverthelesssignificant, earthquakeshave occurredin the region, but the net effect of these eventsis unknown.
We examined the possible effect of the coseismic deformationfrom the nearestlarge earthquake,the M S 6.1
In thisstudy,thelevelingsurveysfrom 1975through1977 inversionwas run. No appreciableeffect was noted. Second, were treated as a single epoch and examined relative to the we arbitrarilyadded-200 mm to the two levelingpointswith baseline 1923 data. The differences in elevation between these
large -200 mm subsidencevalues and reran the inversion.
two epochsare shownin Figure2. Note that thesedifferences Againwe foundno significant influenceon the sources of the
VASCO ETAL.:SouRcES OFCRUSTAL DœFORMAT•ON ATYELLOWSTONE upliftwithinthecaldera.Onlythevolume expansions in the
19,843
elevation differences between successive stations, i.e.,
vicinityof the epicentralarea, outsidethe caldera,were changesoccurringbetween the 1923 survey and the 1975, affected. 1976, 1977 surveys.Figure4 showsthe changesin pairwise Myers and Hamilton [1964] interpretedthe surface elevationdifferencesfor the two levelingcampaigns,plotted deformation of theML 7.5 HebgenLake,Montana,eventand at the midpointsbetweenthe stationpairs. Thesedifferences notedthatthe hangingwall subsidence of up to 5 m occurred are well over30 cm for somepairsof stations. nearlycoincident with the fault scarp,-30 km northwest of the The datumof interestfor a givenepochis thedifference in caldera. They noted little or no deformation from this elevationbetweentwo successive stationson a particularline.
earthquake beyond 30 km,i.e.,withintheareaof leveling data.
The error associated with this measurement contains both a
On this basis, we have not included a test of the effect of
randomas well as a systematiccomponent. The random deformation associatedwith this earthquakeand hencenot component is described by equation(1) whereœ represents the includeda correction for it. distance betweenthestationpair. Whenthe elevationbetween
Theleveling dataweresubdivided intosevenleveling lines two stationsat epoch! is subtractedfrom the differenceat a (1923-1977)coveringthe caldera(Figures2 and 4). In later epoch,/c , the standarderror,is given by [Bevington, addition, an east-northeast line across the Central Plateau
1969].
(1955-1977) in thenorthern caldera wasconsidered especially (2)
importantbecauseof its proximity to a possiblefluid or magmatic zoneconnecting the SourCreekandthe MallardLake
domes.This levelline encompassed a differentperiod(1955- wherei andj denotethestations in thedifferenced pair and•k 1977)andwasscaledto thetimeintervalof interest (53 years) and I denotethe particularleveling campaigns. If the by assuminga constantvelocityof uplift. parameters • andn areconstant for a givenepoch,theformula Because the cumulativestationelevationalonga levelline maybe written,usingequation(1), dependson the observationsat adjacentstations,errorsin elevation accumulate andthemeasurements of thisquantityare correlated. Therefore, strictly speaking,the errors are not independent randomvariables. To avoidthis difficulty,we instead considered the differences in elevation between
where L ij is thedistance between theithandjth station.The
valuesof/•/ usedaregivenabove.The subscripts k successive stationsas thefundamental datum(Figure4). This particular and I denote the values of • for the kth and /th surveys, quantity is not dependent on anypreviousvalues,restoring statistical independence. Furthermore, it is insensitive to any respectively. constant offsetof all stations, suchas a broad-scale upliftof In addition to the random error there are sourcesof inherent thecaldera.Thuswe areinterested in temporal changes in the systematicerror. Two sources of systematic error were
YELLOWSTONE
DIFFERENCES
t 20. O0
--{-45'00' N
11o' lO' w
o .I 90. O0
60. O0
5
o
Z
--{-44-20' N
11o' lO' w
30. O0
0.2M
m
m
0.00 -10.00
20.00
50.00
80.00
1! 0.00
EAST (KM) Fig.4. Differences inupliftforadjacent station pairs onthegiven leveling lines.Thesizes ofthesquares, plotted atthe
midpoints between thestation pairs,areproportional to thechanges between station differences from1923to 197:5-1977. Solidsquares indicate positive changes; opensquares indicate negative changes.
19,844
VASCO ETAL: SOURCES OFCRUSTAL DEFORMATION ATYELLOWSTO•
considered to havesignificanteffectson the observations: (1) rod scaleerror whichresultsfrom non-linearchangesin rod lengthwith changesin temperature,and (2) refractionerror which resultsfrom the slight curvaturein the line of sight causedby the gradientin the air temperature neartheground.
leveling surveys,and in 1987, 20 more gravity stationswere establishedcoincidentally with the furst observationin 1987
of GlobalPositioning Surveys(GPS) [Smithet al., 1989a, b] at sites along the park roadways; giving a total of 214
precision gravity stations. During August and September For the 1923 and 1975-1977 surveys, the rod scale error 1987, most of the gravity stationsoriginally establishedin in accumulated at respectiveratesof 3 and 2 mm/km from the 1977 and a portion of the backcountrystationsestablished bench mark at station K12, chosen as the base station because the summerof 1984 wereobserved [Hollis, 1988;Holliset al., thesingle it was consideredto be locatedas far away from tectonicand 1987;Smithet al., 1989a,b]. We onlyconsidered volcanicdisturbances as possible[Peltonand Smith,1982]. profile of precisiongravity data (Figure 6) from Canyon Refraction error,Rij, fromstation i tostation j isprimarilyJunction to Lake Butte (Figure 2) for our discussionhere dependent uponthe sightlengthbetweensuccessive stations becauseit was the only repeatedprofile coincidentwith the (lines I and 2) for this period. and is describedby the equation[Pelton and Smith, 1982; levelingobservations The 1977 and 1987 gravity observationswere made in Strange, 1981]. Septemberof eachyear to minimize gravitationaleffectsfrom seasonal variations in groundwater levels. LaCoste and Rg=- 4.0 x 10'9? LO .• •Sh(T2soTso) RombergmodelG andmodelD gravimeterswereusedfor these surveyswith at least two and as many as threeinstruments for HereT is a theoreticEly determined constant calculated to be
eachsurvey. Fieldprocedures includedspecialladderlooping
80.7,L ij is the sightlengthbetween stations, 3h is the
techniques to reduceinstrumentdrift. An averagestandard error of 7 gGal for the outer caldera and backcountrystationsand and T$O are temperatures (Cø) at points250 and50 cm above the ground, respectively. For this investigation, the 5 gGal for inner calderastationswas achieved[Hollis et al., temperaturedifferenceis estimatedto be -1.4øC [Pelton and 1987; Smith et al., 1989 a,b]. Gravity measurements(Ag) along the CanyonJunctionto Smith, 1982]. The resulting error estimates for the data Lake Butte profile in the northerncaldera(Figure 6) indicated includingrandomandsystematic arepresented in Figure5. For most differencesthere is good signal relative to the estimated that gravity decreasedby up to 30 gGal from 1977 to 1987 [Hollis et al., 1987]. This decreasereflects both increasesin errorsasmay be seenfrom a comparison of Figures4 and5. orthometricheight and the possiblemigrationof massinto the upper crust [Hollis et al., 1987]. Note that the area of PrecisionGravity Surveys gravitationaldecreaseand orthometricheight increasecoincide A 139-stationprecisiongravity networkwas establishedin closely (i.e., the area of maximum orthometricheight increase to the maximum gravity decreaseof 30 1977 coincident with leveling benchmarksof Yellowstone of 17 em corresponds to a Ag/Ah ratio of -1.7 gGal/cm. [Evoy, 1977] to provide an economicaland rapid method to gGal), corresponding As an example of similar measurementsin anotheractive assesscrustaldeformation. In 1984, an additional55 precision gravity stations were establishedin areas inaccessibleto volcanic system, Johnson [1987] discussed simultaneous elevationdifferencebetweenstationsin half centimeters,T250
YELLOWSTONE
DIFFERENCE
ERRORS
20. O0
-,•45•woo' N
-,•45 øO0'N
ll ø 00'
90. O0
ltO' tO' W
ß
..
n
I
ii
II ß
60. O0
ii II
ii II
¸
z
-•-4•,øW'20' N
it ß
30. O0
110 ø 10
I O. O0 -10.00
20. O0
0.2M
50. O0
80. O0
1 t O. O0
EAST (KM) Fig.5. Estimated standard errors of thepairwise differences in elevation change.
VASCO ETAL.: SOURCES OFCRUSTAL DEFORMATION ATYELLOWSTONE
19.845
NW
200
•
SE
1977-1987
-
-D20 -
'•-.040
