VOL. 82, NO. 28 JOURNAL OF GEOPHYSICAL RESEARCH SEPTEMBER 30, 1977

Seismology on Mars

DON L. ANDERSON,1 W. F. MILLER, 1 G. V. LATHAM,2 Y. NAKAMURA,2 M. N. ToKsl)z,3 A. M. DAINTY,3 F. K. DUENNEBIER,4 A. R. LAZAREWICZ,4 R. L. KOVACH,5 AND T."C. D. KNIGHT6

A three-axis short-period seismometer has been operating on the surface of Mars in the Utopia Planitia region since September 4, 1976. During the first 5 months of operation, approximately 640 hours of high­ quality data, uncontaminated by lander or wind noise, have been obtained. The detection threshold is estimated to be magnitude 3 to about 200 km and about 6.5 for the planet as a whole. No large events have been seen during this period, a result indicating that Mars is less seismically active than earth. Wind is the major source of noise during the day, although the noise level was at or below the sensitivity threshold of the seismometer for most of the night during the early part of the mission. Winds and therefore the seismic background started to intrude into the nighttime hours starting on sol 119 (a sol is a Martian day). The seismic background correlates well with wind velocity and is proportional to the square of the wind velocity, as is appropriate for turbulent flow. The seismic envelope power spectral density is proportional to frequency to the -0.66 to -0.90 power during windy periods. A possible local seismic event was detected on sol 80. No wind data were obtained at the time, so a wind disturbance cannot be rulec! out. However, this event has some unusual characteristics and is similar to local events recorded on earth through a Viking seismometer system. If it is interpreted as a natural seismic event, it has a magnitude of 3 and a distance of 110 km. Preliminary interpretation of later arrivals in the signal suggest a crustal thickness of 15 km at the Utopia Planitia site which is within the range of crustal models derived from the gravity field. More events must be recorded before a firm interpretation can be made of seismicity or crustal structure. One firm conclusion is that the natural background noise on Mars is low and that the wind is the prime noise source. It will be possible to reduce this noise by a factor of 1O" on future missions by removing the seismometer from the lander, operation of an extremely sensitive seismometer thus being possible on the surface.

INTRODUCTION weight and complexity penalty of such an operation; thus an on-board location was dictated that immediately increased the Because the primary emphasis on landed Viking science was noise level because of lander and wind activity (by at least 3 on biology, organic chemistry, imagery, and meteorology, orders of magnitude). The data constraint required severe data other areas such as surface chemistry, petrology, and geophy­ compression using an on-board data processing capability and sics were mostly relegated to future missions. The exceptions thus also a weight and power penalty. The most severe con­ were the inorganic analysis experiment, the magnet experi­ straints on the Viking seismic experiment were limited data ment, and the seismometer. However, these studies were lim­ allocations and the on-board location of the seismometer. ited to reconnaissance measurements. These various constraints and trade-offs led to the design of The ultimate goals of a seismic experiment are to determine a short-period three-component seismometer with on-board the dynamics and internal structure of a planet. These are data compaction and triggering to optimize the data return relevant to the composition and evolution of the interior. (Anderson et al.. 1972]. The objectives were (I) to characterize Before a comprehensive seismic experiment can be considered, the seismic noise environment at the landing sites, (2) to detect however, it is necessary to establish the background noise level local events in the vicinity of the lander, and (3) to detect large of the planet, the level, nature, and location of the seismicity, e"ents at teleseismic distances. and the nature of the seismic signals. Estimates of some of Under optimal conditions it would also be possible to deter­ these parameters have been obtained by the Viking seismome­ mine the following: (I) the approximate distance of events by ter. The ultimate goal, however, can only be accomplished the separation of the various seismic phases, (2) the direction with a network of sensitive seismometers deployed so as to of events with a 180° ambiguity, (3) the attenuation and scat­ minimize artificial sources of noise and wind-induced vibra­ tering properties of the crust to determine if the crust is lunar­ tions. like or earthlike in these characteristics (which are related to The Viking seismic experiment design was constrained by the volatile content), and (4) an estimate of crustal thickness if strict weight, power, and data allocations and perturbed by the crustal and reHected phases can be identified. The Viking 1 con~icting demands of the- other on-board experiments. The seismometer failed to uncage, and no useful data were re­ weight constraint precluded an ultrasensitive seismometer of turned. The Viking 2 seismometer, emplaced on the surface of the Apollo class or a broadband seismometer. The original Mars in the Utopia Planitia region, 47.9°N, 225.9°W, success­ desire to offload the seismometer was sacrificed because of the fully uncaged and has been operating nominally since that time [Anderson et al., 1976]. 'Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, California 91125. STATE OF STRESS JN THE MARTIAN INTERIOR 2 University of Texas, Galveston, Texas 77550. 3 Massachusetts Institute of Technology, Cambridge, Massachu- The regional topography of the earth is in approximate setts 02139. isostatic balance, and regions of partial compensation are •University of Hawaii, Honolulu, Hawaii 96822. generally the most seismically active. Except for the nearside 'Stanford University, Palo Alto, California 94305. mascons the moon is also in equilibrium. Both bodies are •Martin Marietta Aerospace Corp., Denver, Colorado 80201. seismically active, although the moon is many orders of mag­ Copyright © 1977 by the American Geophysical Union. nitude less active than the earth. Large areas of Mars seem to Paper number 7S0408. 4524 ANDERSONET AL.: $EISMOLOGYON MARS 4525

be only partially compensated[Phillips and Saunders,1975] resultsmost relevant to the seismicexperiment. The only direct and could be seismicallyactive. These are the youngerareas evidenceconcerning the internal structureof Mars prior to the which includethe Tharsisplateau and the adjacentlow areasof Viking landings was the mean density, moment of inertia, Chryse and Amazonis. Stressesimplied by lack of com- topography, and gravity field. It is possibleto calculatemodels pensationand dynamicstresses in the earth and the moon are of the Martian interior with plausible chemical models and in the 10- to 100-barrange. Stress drops in earthquakesare temperature profiles that satisfy these few constraints. How- also in this range. Laboratory measurementson crustalrocks ever, the processis highly nonunique. at modestpressure give short-termstrengths of kilobars,but The mean density of Mars, corrected for pressure,is less rockscreep at high temperaturesat relativelylow stresslevels. than that of earth, Venus, and Mercury but greater than that Stressesin the crust of a planet thereforetend to decreasewith of the moon. This implies that Mars has a smaller total Fe-Ni time unlessthey are rejuvenatedby plate tectonic motions, content than do the other terrestrial planets and more than convectivedrag, or thermoelasticstresses caused by heatingor does the moon. Plausible models for Mars can be constructed coo!i.ng.Phillips and Tiernan [1977] argue that the Tharsis which have solar or chondritic valuesfor Fe [Anderson,1972]. gravityanomaly represents a long-wavelength stress supported Mars is the only terrestrial planet, including the moon, which by the lithosphereor the asthenospherefor 108-109years, could have chondritic abundances of iron, and it is therefore implying stressesin the mantle of 100 bars or stressesin the likely to contain a larger proportion of volatilessuch as water-, lithosphereof a fewkilobars. They favor at leastpartial dy- sulfur-, and potassium-bearingphases. The densitiesand com- namic supportof Tharsisby mechanismssuch as mantle con- positions of the planets are consistentwith the view that the vection or viscous and thermoelastic stresses associated with volatile content increaseswith distance from the sun [Ander- magmaticactivity. The existenceof largestresses in the Tharsis son, 1971]. This can be understoodin terms of a primitive solar region,at leastat the time of formation,is alsoimplied by the nebula in which the decreasingtemperature away from the sun pattern of lineaments,including apparent grabens and the controlled the composition of the condensateswhich were great canyon, alignedgenerally radially with respectto the available for incorporation into the various planets. In this plateausummit [Blasius and Cutts, 1976].The depthof com- view, Mars should be more volatile rich than the other terres- pensationof the Martian topographyis of the order of 150km, trial planets. V olatiles in this context include FeS, hydrous and about 3 km of Tharsisis presentlyuncompensated [Phil- minerals, water, K•O, and Na•O. Low condensation temper- lips and Saunders,1975]. It is likely that the uncompensated atures also imply a higher oxidation state for the available areas of Mars are seismicallyactive, since the stressesare iron; i.e., more Fe should be available as FeS, FeO, FelOn, and similar or greater than stressesin the earth and moon. This is Fe•O4than is the casefor the primitive earth. The core of Mars true whetherthe planet is strivingtoward a state of isostatic therefore is probably richer in FeS, and the mantle is richer in equilibrium or whether the stressesare being rejuvenated. FeO than is the case for the earth, where more free Fe is Most of the seismicactivity on earth is associatedwith plate available. A large Fe-FeS core, an FeO-rich mantle, and a margins,although moderate to large (but infrequent)earth- thick crust (Na•O, K•O, and hydrous minerals) are therefore quakesalso occurin plate interiors of both oceanicand conti- expected.On the other hand, the relative proportions of CaO nental plates. Plate motions on the moon and Mars are cer- and AI•O•, componentsof crustal minerals, are expectedto be tainly lesspronounced than on earth,so we mightalso expect less, and the total crustal thickness may be buffered by the Mars to be less seismicallyactive than earth. On the other availability of these constituents. However, SiO• and low- hand, the largerstresses required to supportthe topography melting-point FeO-bearing compoundsprobably dominate the and the gravity field imply the potentialfor seismicactivity crustal composition. even if creep and flow occur. In the earth.and in the labora- With such broad chemical constraints, mean density (h = tory, deformationproceeds by both slow aseismicprocesses 3.96 g/cm'), and moment of inertia factor (C/MR s = 0.365, (creep) and rapid seismicprocesses (earthquakes). The latter Reasenberg[1977]) and under the assumptionof a differenti- dominateat low temperaturessuch as exist in the lithosphere. ated planet it is possibleto trade off the size and density of the The variationsof gravity and topographysuggest that the core and density of the mantle [Anderson, 1972; Binder and seismicityof Mars is not uniform and that it is likely to be Davis, 1973; Johnstonet al., 1974; Johnstonand Toksb'z,1977]. highestnear Tharsisand its environsand similaryoung un- Most of these models favor an FeO enrichment of the Martian compensatedconstructs. The Viking 2 landeris approximately mantle relative to the mantle of the earth. Anderson [1972] 110ø from ChrysePlanitia (VL-1 site) and 25ø from Elysium concluded that Mars has a total iron content of about 25 wt %, M ons. Marsquakesat theselocations would have to be quite which is significantlyless than the iron content of earth, Mer- largeto be detectedby theViking 2 seismometer.In summary, cury, or Venus but is closeto the total iron content of ordinary we expect Mars to be a tectonicallyactive planet, but the and carbonaceouschondrites. The high zero-pressuredensity seismicitymay be localized. of the mantle suggestsa relatively high FeO content in the silicates of the Martian mantle. The radius of the core can MODELS FOR THE INTERIOR OF MARS rangefrom as smallas one-thirdthe radiusof the planetfor an Interpretation of Martian seismicdata requiresan a priori iron core, or a core similar in compositionto the earth's core, model of the planet's interior. Such a model must be calculated to more than half the radius of the planet if it is pure FeS. With on the basis of available direct data, cosmochemical consid- chondritic abundances of Fe-FeS the size of the core would be erations, equations of state, and theoreticalmodeling of the about 45% of the planet's radius, or about 12% by mass. A planetary interior and its thermal evolution. Such efforts have small dense core would imply a high-temperatureorigin or been made by a number of investigators[Urey, 1952; Ander- earlyhistory because of the highmelting temperature of-Fe- son, 1972; Anderson and Kovach, 1967; Kovach and Anderson, Ni, while a larger light core, presumablyrich in sulfur, would 1965; Binder, 1969; Binder and Davis, 1973; Johnstonet al., allow a cooler early history, sincesulfur substantiallyreduces 1974;Johnston and Toks&, 1977; Ringwoodand Clark, 1971; the melting temperature.A satisfactorymodel for the interior Toks& and Johnston,1977]. Here we will briefly review the of Mars can be obtained by exposingordinary chondritesto 4526 SEISMOLOGY modesttemperatures, melting the Fe-FeS, and removingmost 90 ø of it to the core [Anderson,1972]. This would make the Mar- MARS tian core substantiallymore S rich than the terrestrialcore and would favor the larger core radius. Alternatively, Mars could R be a mixture of carbonaceous and ordinary chondrites (or hyperstheneor bronzite chondrites)and satisfythe mean den- sity and moment of inertia with partial removal of Fe-Ni-S from the silicate phaseto the core. The trade-offs betweencore size, composition, and mantle density are given by Anderson[1972] and Johnstonand Toks& o 18o [1977]. Under the assumptionof an uncertaintyof +0.005 in the moment of inertia factor the core radius in these models could vary between about 1300 and 1800 km, and the mantle density,corrected to 0øC and 1-barpressure, between p,• of 3.4 and 3.6 g/cm a. It is obviousthat a direct determinationof the size of the core by seismicmeans could constrain the overall compositionof the planet. The topography and gravity fields of Mars indicate that parts of Mars are grosslyout of hydrostatic equilibrium and that the crust is highly variable in thickness.If variationsin the Fig. 1. Seismic ray paths for compressionalwaves through the gravity field are attributed to variations in crustal thickness, interior of a theoretical Martian model that satisfiesthe mean density with a constant density ratio between crust and mantle, then and moment of inertia. The 'mantle' is homogeneousin composition, reasonablevalues of the densitycontrast (<0.8 g/cm a) imply the olivine-spinel transition causing a discontinuity at a 1100-km that the average crustal thicknessis at least 30 km (W. M. depth.Note the shadowzone at the surfacecaused by the decreasein Kaula, personal communication, 1976). This minimal bound velocity in the core. The core is assumedto be molten. Parametersfor the model are shown in Figure 2. is based on the assumption of zero crustal thicknessin the Hellas basin (R. Phillips, personalcommunication, 1976). An impact large enoughto excavatethe Hellas basin would easily [Anderson,1973] can yield a muchgreater crustal thickness. remove a 30-km-thick crustal layer (T. J. Ahrens, personal The surfacereflectance spectra of Phobosand Deimos and communication,1976). This minimal averagecrustal thickness the low densityof Phobosare alsoconsistent with the fact that on Mars gives a crust/planet mass ratio that is more than 5 these bodies are similar to carbonaceous chondrites. Even if times the terrestrial value, indicating a well-differentiated theseare captured bodies, the suggestionis that Mars may planet. have accretedfrom relatively volatile rich material. The state The crust of the earth is enriched in CaO, A12Oa,K20, and of water on Mars is unknown, but the evidence for running Na•O in comparisonto the mantle. Ionic radii considerations water at the surfaceearly in its history and the inability of and experimentalpetrological results suggest that the crust of substantialamounts of water to escapefrom the atmosphere any planet will be enrichedin theseconstituents. A minimal indicate that there should be free or bound water in the average crustal thicknessfor a fully differentiated chondritic interior. planet can be obtained by removing all of the CaO possible, Atmosphericanalyses suggest that Mars is a lessoutgassed with the available AlcOa, as anorthite to the surface.This gives planetthan earth [Owenand Biemann, 1976], but the high ratio a crustal thickness of about 100 km for Mars. Incomplete of 4øArrelative to the other inert gasessuggests that the Mar- differentiation and retention of CaO and AlcOa in the mantle tian crust is enriched in 4øK in relation to the earth's crust. will reducethis value, which is likely to be the absoluteupper The volatile content of a planet's interior along with the bound. temperatureof the interior are the two mostimportant factors The average thicknessof the crust of the earth is 15 km, that control the attenuation of seismic waves. The attenuation which amounts to 0.4% of the mass of the earth. The crustal propertiesof the M artJanmantle may be similarto thoseof the thickness is 5-10 km under oceans and 30-50 km under older earth'smantle with the possibleexception of the low Q (high continental shields. The thickness of the crust increases with attenuation) zones of the earth's asthenosphere.Thus we age. The volume of the crust increasesonly slightlyif presently would expectseismic waves to propagatethrough the Martian active subductedregions are taken into account.The situation interiorwith similaror greaterefficiency than they do on earth, on the earth is complicated,since new crust is constantlybeing particularly if the lithosphereis as thick as is implied by created at the midoceanicridges and consumedat island arcs. isostaticcalculations [Phillips and Saunders, 1975]. Compres - It is probable that some of the crust is being recycled. If the sional wave ray paths are shown in Figure 1 for a typical presentrate of crustalgenesis was constantover the age of the Martian model based on an FeS core composition.Note the earth and none of the crust was recycled,then 17%of the earth largeshadow zone due to the low-velocitycore. It is important would be crustal material. to point out that these models are nonuniqueand actual The moon apparently has a mean crustal thicknessgreater seismicdata are required to determinethe internal structure than that of the earth. If the averagecomposition of the moon and compositionof Mars. The velocitiesand densitymodels is similar to chondrites minus the Fe-Ni-S, then the crust could usedto calculatethe ray pathsin Figure 1 are shownin Figure be as thick as an averageof 62 km. This is about the thickness 2. of thecrust determined by seismicexperiments on thefront- CRUSTAL STRUCTURE OF MARS side of the moon [Toks& et al., 1974] but lessthan the average thickness inferred from the gravity field [Bills and Ferrari, It is clearfrom the topographyand gravity field of Mars that 1977a]. An alternative model based on a Ca-Al-rich moon the crustal layer is highly nonuniform.Attempts have been ANDERSON ET AL.: SEISMOLOGY ON MARS 4527

Model AR

339 878 1018 1355 189q 2033 2372 2710 30q9 3388 DEPTH ( KM )

Fig. 2. Theoretical velocitiesand density in the interior of Mars. This model satisfiesthe known constraintsof mean densityand momentof inertia.The coresize can be increasedor decreasedif the densityof the corerelative to that of the mantleis decreasedor increased,respectively. Data usedin thismodel are from Anderson[1.972], Okal andAnderson [1977], and Johnstonet al. [1974]. made to infer variations in crustal thickness[Phillips et al., is quite thin in comparison to the rest of the planet. If the 1973; Phillips and Tiernan, 1977; Bills and Ferrari, 1977b],but crustal thicknesscould be determined at a singlesite and if the the process is nonunique unless it can be tied to a direct densitycontrast is constant,then we could immediately deter- measurementof the crustal thicknessmade by seismicmeth- mine the average crustal thicknessof the planet and, for ex- ods. Interpretation of the Bouguer gravity field involvesboth ample, the crustal thicknessunder the Tharsis plateau. the crustal thicknessand the crust/mantle densityratio. Usu- An average crustal thicknessof even 30 km would indicate ally, the mean crustal thicknessand the density contrast are that Mars is an extremely well differentiated planet having a assumed,and the Bouguer anomaliesare interpreted in terms crust/planet mass ratio of over 5 times that for earth. The of deviations of the crustal thicknessfrom the mean. Figure 3 average thickness would be less if the density contrast were givesa crustalthickness (isopach) map of Marsfor an as- greater. There is evidencethat the Martian mantle is denser sumedmean crustal thicknessof 40 km and a densitycontrast than the terrestrial mantle [Anderson, 1972], but there is of 0.6 g/cm 3 [Bills and Ferrari, 1977a]. Note that the thickest no direct information on the densityof the crust. If ice, water, crust, 75 km, is under Tharsis at the head of Vailes Marineris and hydrated minerals are abundant, then the crustal density and the thinnest crust, 10 km, is under Hellas. A secondcrustal may be quite low. On the other hand, a hydrousmantle could map for an assumed mean crustal thickness of 100 km is result in a well-differentiatedplanet becauseof the effect of shown in Figure A-1 of the microficheappendix. • water on melting temperaturesand viscosity. One way to bound the problem is to assumezero crustal thicknessunder the Hellas basin. If this large circular feature is DESCRIPTION OF THE VIKING SEISMOMETER of impact origin, deep excavation can be expected which The Viking seismometerincludes sensors, amplifiers, filters, would remove most, if not all, of the crustal material. For a and electronics for automatic event detection, data com- density contrast of 0.3 g/cm 3 the mean crustal thicknessis paction, and temporarydata storage.The instrumentpackage, about 65 km, and the thicknessat the Viking 2 site would be shownin Figure 4, is 12 X 15 X 12 cm and weighs2.2 kg. It is 10-20 km. A higher densitycontrast would decreasethe mean locatedon top of the lander'sequipment bay near the attach- thickness and decrease the variation in crustal thickness. The ment of leg 1 (Figure 5). The nominal power consumptionis maximum crustal thicknessis 140 km for ,5,p= 0.3 g/cm a and 3.5 W. The useful frequencyrange is 0.1-10 Hz with a min- 90 km for At> = 0.6 g/cm a. These calculationsare due to R. imum ground amplituderesolution of 2 nm at 3 Hz and 10 nm Phillips (personal communication, 1976). He has demon- at 1 Hz. stratedthat the crust in the vicinity of the Viking 2 landing site Figure 6 comparesthe Viking seismometerdisplacement x Supplementis available with entire article on microfiche.Order responsein each of its operatingmodes with the displacement from American GeophysicalUnion, 1909 K Street, N. W., Washing- responseof a typical station in the U.S. Geological Survey ton, D.C. DocumentJ77-004; $1.00. Paymentmust accompanyorder. World-Wide Standard SeismographNetwork (WWSSN). The 4528 SEISMOLOGY ANDERSONET AL.: SEISMOLOGYON MARS 4529

The seismic sensors are three matched, orthogonally mounted (one vertical and two horizontal) inertial seismome- ters fitted with velocity transducers.One sensoris shownsche- matically in Figure 8. Each sensoroccupies a volume measur- ing 7.5 X 3.8 X 3.8 cm. A 16-gmass-coil assembly is supported on twin booms by two Bendix Free-Flex elastic hinges such that the flat transducercoil is poisedbetween the facing poles of two channelm.agnets arranged in series.Motion of the ELECTRONICS frame causes the transducer coil to move in the field of the magnetsand generatesa signal which is in proportion to the relative velocity of the coil with respectto that of the magnet.

SENSORSUSASSEHSLY The undampednatural frequencyof each instrumentis 4 Hz, SUBASSEMBLY the coefficientof damping is 0.6, and the generatorconstant is 177 V/(m/s). The natural undamped frequencyof the sensors was chosen such that the instrument would meet the con- straints of weight and volume and to insure that the sensors would operate over the largestexpected tilt of the lander (15ø ) without the use of any mechanicalzeroing adjustments.With this design the horizontal units will tolerate up to 23ø of tilt, SEISMOMETER ASSEMBLY and the vertical up to 35ø . Fig. 4. Seismometerpackage. The dimensionsof the unit, ;nclud- The inertial mass of each sensor is individually caged to ing sensorsand electronics,are 12 X 15 X 12 cm with a weightof 2.2 protect it from the shock and vibration encounteredduring kg. The nominal power consumptionis 3.5 W. launch, separationof the lander from the orbiter (the greatest shock peaking at 1200 g), and the landing. The caging is peak magnificationis 218,000 at 3 Hz when the returned clata providedby spring-loadedplungers which hold the massfirmly are plotted at a scaleof 0.44 mm per digital unit of seismome- against a stop seat. The plungersare securedby a palladium- ter output. Figure 7 comparesthe accelerationsensitivity of aluminum fusewire. Uncaging is accomplishedby causingthe the Viking seismometerwith that of conventionalaccelerome- fuse wire to fail nonexplosively when it is electrically heated. ters, the U.S. Geological Survey WWSSN, and the Apollo All three axesfailed to uncageon the Viking 1 lander. Because lunar instruments.The Viking seismometerhas a maximum the uncagingsystem was triply redundant except at the inter- sensitivityequivalent to that obtainable at a relatively quiet face with the lander and because the seismometer functioned site on earth. properly in all other aspects,it is believedthat the failure was

VIKING LANDED SCIENCE CONFIGURATION

S BAND HIGH-GAIN ANTENNA (DIRECT)

MAGNIFYINGMIRROR, CAMERATEST TARGET & MAGNET

RADIO SCIENCE PATTERN METEOROLOGYSENSOR S CAMERA(2)

SEI SMOMETER

UHF ANTENNA

METEOROLOGY BOOM ASSEMBLY

TEMPERATURE SENSOR

LEG 2

INTERNALLYMOUNTED: ,CMS PROCESSOR Biology VIEW MIRROR (2) GCMS X RayFluorescence BIOLOGY PROCESSOR Pressure Sensor X RAY FLUORESCENCE BOOM FUNNEL

,COLLECTORHEAD

MAGNETS

Fig. 5. The Viking landershowing the locationof the seismicpackage and otherinstruments. The distancefrom footpad to footpad is about 2.5 m, and the lander massis 605 kg. 4530 SEISMOLOGY

at the interface.The Viking 2 landerseismometer uncaged in a nominal fashion. Each inertial sensoris equippedwith a calibrationmecha- nism by which the massmay be magneticallydeflected ap- proximately 4 #m. The deflection and releaseof the massin each direction produce a pair of pulse doublets,shown in Figure 9, from which the operationand characteristicsof the instrumentmay be ascertainedand which may be used to determinethe attitudeof the seismometerpackage relative to the local verticalby measuringthe asymmetriesof the calibra- tion doublet which are a function of the tilt of the instrument. This comparisonof the Viking 2 seismometercalibration dou- blet with prelaunch calibration data showed the instrument attitude to be 9.5 ø + 2.8 ø down at an azimuth of 278 ø + 17ø from north. This is in agreementwith the lander's inertial referencesystem, which determinedthe tilt of the lander to be 8.2 ø down at an azimuth of 278 ø from north. A block diagramof the instrumentis shownin Figure 10.

FREQUENCY (HZ) Each axishas an amplifierwith a bandwidthof 0.2-1.2 Hz and an amplification,which is selectableupon earth command, 5EISMBMETER MBBNIFIœBTION from 6 X 10a to 4 X 105in sixincrements. After amplification, Fig. 6. Magnification of the Viking seismographin each of the prefiltering,and analogmultiplexing, the seismicsignals are operatingmodes. The magnificationsare basedon the assumptionthat convertedto a 7-bit plus signdigital word at the rate of 121.21 the digitized data are plotted at scalesof 0.44, 0.51, and 0.76 mm?du samplesper secondby a dual-slopeintegrating analog-to- for the high data rate, event, and normal modes, respectively.A digitalconverter. The subsequentdigital processing of the data typical U.S. Geological Survey short-periodinstrument used in the WWSSN is shown for comparison. includesfiltering, averaging, compression, event detection, and buffermemory storage. These functions, as well'asdigital multiplexing,command decoding,timing, and control, are implementedin customlarge-scale-integrated (LSI)circuitry.

0-2 VIBRATION ANALYSIS ACCELEROMETERS

0-3

/

0- 4 ACCELEROMETERSINERTIALGUIDANCE

0-5

--- 0-6, Z • VIKING SEISMOMETER o

< 0-7 Et• WORLD-WIDENETWORK ___1 SHORT-PERIOD LLI SEISMOMETER • 0-8 ...... WORLD-WIDE NETWORK LONG -PER IOD SEISMOMETER 0-9 .... LONG-PERIOD SEISMOME[ER WITH DISPLACE MENT 0_10 TRANSDUCER

.... APOLLO LONG-PERIOD SEISMOMETER 0-•1 APOLLO SHORT-PERIOD SEISMOMETER

O-12 EARTHSEISMIC NOISE I05 I04 I03 I02 I0 I0 0 I 01 PERIOD (SECONDS) SENSITIVITIES OF ACCELERoMETERS AND SEISMIC INSTRUMENTS

Fig. 7. Comparisonof the accelerationsensitivity of the Viking seismometerwith the sensitivitiesof conventional accelerometers,theU.S. Geological Survey WWSSN, and the Apollo lunar seismometers. Themaximum sensitivity ofthe Viking seismometeris equivalentto that of a relativelyquiet short-period seismometer on earth. ANDERSONETAL.: SEISMOLOGY ONMARS 4531

•-MASS the frequencyresponse of the inertial sensor.The equivalent MASSCLAMP • / noisebandwidth at eachfilter cutoffis shownin Figure 11.The • / /--CALCOILTERMINALS rectangular response represents an ideal band-passfilter with the samepower and peak value as thoseof the total system COIL--, I • / I • I • • CALl frequencyresponse. F...... •- x• • • BRATECO•L After filtering,the absolutevalue of the data is then passed througha low-passfilter to obtainthe 12.7-srunning average of the microseismiclevel which is in turn sampledat the rate of one sampleevery 14.85 s on eachof the threeaxes. The digital low-passfilter may be fixedor automaticallystepped through each cutoff frequencyat the rate of 2 min per step. POLEMASS CLAMP• OUTPUT SCREW Et;ent or triggered mode. This mode refers to the data PLUNGER compactionmode of operation, where the envelopeof the seismicsignal and the numberof zero crossingsrather than the SEISMOMETER SCHEMATIC signal itself are sampled at the moderate rate of 1.01 sample/s/axis.To producethe envelope,the absolutevalue of Fig. 8. The Viking seismicsensor. As the mass-coilassembly the seismicsignal is smoothedby passingit throughthe digital movesin relation to the frame, a signalis inducedin the coil as it filter operatingat the 0.5-Hz cutofffrequency. Simultaneously, passesthrough the field of themagnets. The signal produced ispropor- tionalto thevelocity of thdrelative motion. a runningcount of the positiveaxis crossings (a measureof the dominant frequencyof the signal)is sampledat the samerate. This combinationof samplingthe envelope(7-bit word) and Theinstrument may operate in anyof threedata processing axis crossing(5-bit word) resultsin a 12.3to 1 reductionin the modes. data requiredto encodethe originalsignal over the high data Highdata rate mode. Eachchannel is digitallyfiltered, and rate. The effectsof smoothingthe envelope,demonstrated by a the 7-bit plus signword is temporarilystored in one of two computersynthesis of its operations,are shownin Figure 12. 2048-bitrecirculating memories to await servicingby the Recall that the resultingenvelope is sampledonce per second. lander'sdata acquisition and processing unit (Dapu). The data Of particularnote are the delay in the start of the signaland rate is 20.2 samples/s/channel. the rise time, nominally 1 s each. In addition, there are effects The filter is a digital implementationof a sixth-ordermax.- due to transientsand incompletesmoothing at frequencies imally flat (Butterworth)low-pass filter with command-select- below 0.4 Hz. able cutoff frequenciesof 0.5, 1.0, 2.0, and 4 Hz. The event mode may be initiated by earth commandor Normal mode. The normal mode is the lowest data rate automaticallyby enablingan event detectoron any one or mode,operating at 4.04 samples/min/channel.Its purposeis combinationof the three axes.The purposeof the detectoris to investigatethe averagelevel and spectralcontent of the to recordseismic events which are transientin nature(Mars- microseismicbackground. A form of 'combfiltering' is per- quakesor meteorimpacts) at the higherdata rate of the event formedby usingthe digitallow-pass filter in conjunctionwith mode while it is conservingdata by monitoringthe seismic

, Ii + Displacement ll sec.J ea

Displacement +Release +Cal - -' - Cal

SEISMOMETER CALIBRATION PULSE SEQUENCE Fig.9. Seismometercalibration sequence. The deflection and release of themass by the calibration coils produce a pair of pulsedoublets from which the operationalcharacteristics and the attitude of the seismometercan be ascertained. Calibrationsequences are run in the highdata rate mode. 4532 SEISMOLOGY

x AX•S activitieswould produceexceptional lander vibrations. When the data allocation for the seismometer is more than 1700 06 12 1824 30 36 db YAXIS -blULTIPLEXER• buffers, the excess data are recorded in the high data rate mode.

INSTRUMENT PERFORMANCE

• cou•ERI- T- T Followingtouchdown the azimuthsof thehorizontal (Y and Z) componentseismometers were determined with regardto the lander inertial referencesystem. The positiveoutput from TRIGGERT RESHOLD ENABLEIMULTIPLIER the Y componentcorresponds to landermotion toward the northwest(N31øW), and the positiveoutput from the Zcom- NORi•ALEVENT•.•DATA_ I' •OIS•PLES/•IN/ CHANNELI AVERACE •267-SEC • q •LUE I- ponentcorresponds to lander motion toward the southwest I OI SA•PLES/SEC/CHANNEL(EVENT •ODE) (S59øW).The polarity of the vertical(X) componentfollows HIGH•• 202_lSA•PLES/SEC/CHANNEL DATA (HIGHDATARATE) the usual convention of positive output for upward dis- •ITRANSFERBUFFERI •l •E•ORIESI - '• D•PU placement.The 'vertical'is 8.2ø from the true verticaltoward FUNCTIONALBLOCK DIAGRA• OF SEISmOmETERELECTRONICS N82øW. Fig. 10. Block diagram of seismometerelectronics. The analog The initial calibration pulse amplitudeswere in good agree- amplifier, shown for the Z axis only, also includesa low-passfilter to ment with predictedvalues. A gradual increasein calibration prevent aliasing. pulseamplitudes (18%) occurredduring the first 140 sols,but the relative sensitivitiesremained closelymatched throughout this period (see Figure A-2 of the microficheappendix). The background in the normal mode. The trigger level of the instrumentresponses were matchedto within approximately detector is a multiple of the long-term (1081 s) averagemicro- 5% during prelaunchadjustment. The gradualchange in pulse seismic level. The multiple is selectedby ground command amplitudesis a consequenceof the decreasingtemperature of from values times 4, 8, 12, 16, or 20. When the signal level the lander and thus the lowering of the temperature (resist- drops below the averagebackground, the instrument will re- ance)of the transducerand calibrationcoils. The differencein vert to the normal mode. To reduce the number of data taken calibrationpulse amplitudesamong the three componentsis from false triggering from lander thermal 'pops,' the event causedby the tilt of the instrument. detector on time beyond the normal off time is controlled to be It was noted early in the experimentthat signalamplitudes proportional to the time that the level of the event is above the were consistentlysmallest on the Y component(10-20% less trigger level. This time increasevaries from a minimum of 2 s than those on X) and largest on the Z component(20-50% to a maximum of 1 min for the longest event. greaterthan thoseon X) even thoughthe evidencefrom cali- This autotrigger mode was used during the conjunction bration data showedthat the componentsensitivities were well period when the total number of Viking sciencedata were matched. In addition, the three componentswere nearly al- limited to those which could be stored on the lander tape waysin phase,with a high degreeof coherencebetween chan- recorder during the 44 days of conjunction. This mode is also nels during the brief intervals of high data rate operation in used in the automatic mission, which is designedto continue lander operations in the event that the ability to load new commandsinto the lander is lost. The instrumentwill produce 6170 bits/h in the normal (or backgroundmonitoring) mode, 1.47 X 105 bits/h in the event mode, and 1.77 X 106 bits/h when it is operating in the high data rate mode. The instrument contains dual 2048-bit data buffer memo- ries. Two are necessaryso that data flow continueswhile the full buffer is serviced by the lander. The data are in turn transferred along with engineering and scientific data from other experimentsto the lander'smagnetic tape recorder.Nor- mally, the data are transmitted by the lander to an orbiter for relay to earth; however, they can be transmitted directly to earth at a much lower data rate. The seismometer will transfer 69 buffersper day if it is operatingwholly in the normal mode, 1716 buffersper day in the event mode, and 20,700 buffersper day in the high data rate mode. The number of data buffers that the seismometertransfers each day is controlled by the lander and is part of the overall mission planning. It varies from zero to about 3500 buffersper day but is nominally about 1000-1500 buffers per day. The tape recorder on the lander cannot physically hold more than about 18,000 buffers of seismic data. Normally, the operating time of the instrument is divided among the three modes so as to maximize the amount of time Fig. 11. Equivalent noise bandwidth at each filter settingand of in the event mode with a sampling of high data rate mode the event mode. The rectangular responserepresents an ideal band- during the quietestpart of the sol and for calibration. The pass filter with the same area and peak value as the total system normal mode is used only to fill in or when other lander frequencyresponse. ANDERSONET AL.: SEISMOLOGYON MARS 4533

_•E:C.

i•.H H7.

IE,. IS H7.

ENVELOPE DETECTOR RESPONSE

Fig. 12. Eventmode envelope detector response. The effectsof smoothinglow-frequency rectified signals with a 0.5-Hz low- passfilter are shown.The resultingsignal is then sampledonce per second.The input signalsare zero beforetime zero. which phasecomparisons can be made. This observationcan day. During the primary mission, covering the first 62 sols, only be explained by rectilinear motion of the lander in a planning of commands to be sent to the lander was divided preferreddirection. In-phase correlation requires that the pre- into 6-day cycles, each cycle covering three command se- ferreddirection of motion be up to the westand down to the quences(uplinks) to be sent to the lander. Planning for each east,i.e., along the directionof tilt of the lander. This direction cycle started 16 days before the first command uplink was to is consistentwith a rockingmotion of the landerabout the line occur. As the time for a particular uplink approached, the connectingfootpads 1 and 2 (seeFigure A-3). The seismome- sequenceof commandsbecame less flexible, and more justifi- ter is located approximately 1 m above ground level. The cation was required to incorporate a change.With these re- remainingdimensions and anglesare shownin Figure A-4. strictions each command could be tested before it was exe- From this the expectedamplitude ratios for a rockingmotion cuted on the lander, thus insuring against disastrouserrors. about the line connectingfootpads 1 and 2 are as follows: During the conjunctionperiod when Mars was behind the sun and totally out of communicationwith the earth (sols63-108), Ax:Ay:Az = 1:0.83:1.44 the lander operated in a reducedmode, executingcommands These values are in approximateagreement with those ob- uplinked beforeconjunction. During the extendedmission (sol served. 109 to the end of the mission), planning cyclescover 2 weeks, Postlandingimagery revealed that footpad3 is balancedon and commandsare sent to the lander once per week. During a rock. Thus a rocking motion in the directioninferred from the primary mission, data were returned to earth (downlink) seismicdata is a reasonableexpectation. once per day, while downlinks during the extended mission A related observation is that wind-induced vibrations of the averageabout twice per week. landerare highestfor a givenvelocity for windsfrom the east. The seismometer commands are stored in the lander com- We would expect that the rocking motion describedabove puter together with the time to be spent in each command. wouldbe mosteasily excited by easterlyand westerlywinds. When the end of the command sequenceis reached,the se- quenceis restartedautomatically. Uplinks needto be sentonly SUMMARY OF OPERATIONS to changea command or command sequence.Each command On earth,operations such as change of gainand filter setting must specifythe mode of operation, vertical gain, horizontal are usuallytrivial. During the Viking mission,however, these gain, filter frequency, trigger state, trigger threshold, calibra- and other operations(such as changein operatingmode and tion state, and time to be spentin this configuration.In addi- trigger levels)had to be plannedmany daysin advanceof the tion, the number of buffers to be recorded and the number of actualexecution of the command.This advanceplanning was commands to be used can be specified. Until sol 141, two necessaryto avoid conflictswith other experimentson the tables of 12 commands each plus a secondarytable of three lander, to insure the safety of the lander, and becausecom- commands could be specified.After sol 141, two tables of 34 mandscould be sent to the lander generallyonly everyother entries each became available. 4534 S EISMOLOGY

II/•œI ANDERSONET AL.' SEISMOLOGYON MARS 4535

TABLE 1. Examples of Time Lines

Period Beginning Period End

Sol Time, LLT Sol Time, LLT Mode, Other Comments

Primary Mission, Sol 17 Downlink 16 0845: 48 16 0947:39 Relay link, no data collected 16 0947:39 16 1053:59 Event mode 16 1053:59 16 1422:34 Direct link, normal mode 16 1422:34 16 2300:31 Event mode 16 2300:31 16 2317:31 Normal mode, filter stepping 16 2317:31 16 2320:31 Calibration, 30-dB attenuation 16 2320:31 16 2424:20 High data rate mode,4.0 filter 16 2424:20 17 0843:16 Event mode 17 0843:16 Poweroff, 2187 bufferdumps Conjunction,Daily Sequence N 0400: 23 N 1900:01 Normal mode, 4.0 filter N 1900:01 N + 1 0300:01 Normal mode,4.0 filter, X triggeron, threshold X 12 N + 1 0300:01 N + 1 0330:01 Event mode N + 1 0330:01 N + 1 0330:17 Calibration, 30-dB attenuation N + 1 0330:17 N + 1 0359:07 Event mode N + 1 0359:07 Power off, 152 buffer dumps (seetext) Extended Mission, Sol 129 129 0000:00 129 0200:02 Event mode 129 0200:02 129 0200:36 Calibration, 30-dB attenuation 129 0200:36 129 0700:01 Normal mode, 4.0 filter 129 0700:01 129 1800:01 Normal mode, 4.0 filter, 6-dB attenuation 129 1800:01 129 2012:31 Normal mode, 4.0 filter 129 2012:31 130 0000:00 Event mode, 500 buffer dumps Extended Mission, Sol 1.56 156 0000' 00 156 0159:48 Event mode 156 0159'48 156 0200:18 Calibration, 30-dB attenuation 156 0200' 18 157 0000:00 Event mode, 1720buffer dumps

Horizontal and vertical attenuations are the same and are 0 dB unless they are otherwise noted. No filter means that the low-pass filter was set to 0.5-Hz cutoff.

The planning proceduresand structurefor the lander oper- ignored except for inhibitions due to the relay link and direct ation have been describedby Lee [1976]. Readers are referred link, sincethese occurred on a regular basisduring the primary to this paper for a more completedescription of the uplink mission and were generally of short duration. procedures.Although a large amount of time was spent by LANDER NOISES each team in these planning sessions,the necessityof the systemcannot be denied. Since the seismometeris located on top of the lander, it is Before touchdown the commands loaded into the tables important to understand all internal sources of noise, the were to surveythe ambient seismicnoise at variousgain levels, responseof the lander to wind, and the effect of the lander on with approximately 500 buffers of data allocated per sol. The external seismicsignals. A catalog of these noise sourcesand data returnedfrom this phaseof the operationdemonstrated representativesignatures of each are given in the microfiche that the seismometercould be operated at all times at max- appendixto this report. Noises generatedinside the lander are imum gain. After sol 9, high enoughdata rates were available generally of short duration, infrequent, and distinctive in na- to use the event mode most of the time. During conjunction, ture and occur at predictable times. There is therefore little 152 buffers per sol were available to the seismometer.Only 1 likelihood that they would be confusedwith natural events, hour of event mode per sol was scheduledwith 8.5 hoursof the but they do complicate data processing,and it has taken remaining 23.5 hours of normal mode set aside for trigger- considerableeffort to identify and catalog them. The internal enabledoperation. If the triggerwas not activated, 141 buffers sourcesidentified are tape drive, X ray, sample dump, an- would be taken; thus 11 buffers were set aside for triggered tenna tracking, camera motions, operationsof the soil sam- event mode. In the extended mission a more flexible schedule is pler,and electricaltransients from other •xperiments. The followed. On occasion, 1700 buffers are available, continuous known vibrational frequenciesof these activitiesare in agree- event mode recording thus being permitted. At other times, ment with the frequenciesmeasured by the seismometer.These fewer data, down to 500 buffersper sol, were available. are generallygreater than 4.7 Hz. When the soil samplerarm is To date, the seismometerhas been operatednearly continu- in the extendedconfiguration, frequenciesof 2-4 Hz are ob- ously except for two periods: sols 103 and 104 and sols served,the dominant frequencydepending on the extension. 141-146. One hundred and seventy-onesols of data have been The Q of the arm is about 200. Wind-inducedvibrations of the collected,consisting of 87 solsof normal mode data, 75 solsof lander are also more pronounced when the arm is extended. event mode data, and 63 hours of high data rate mode data Fortunately, the arm is usually stowed.Before sampledelivery (the balanceis losttime dueto playbackperiods of the lander the collector head is usually vibrated at 8.8 or 4.4 Hz for a tape recorder for relay to the orbiter). Some detailed examples short period of time. These frequenciesare clearly observed of data collection are given in Table 1. Periods when the by the seismometerat the appropriatetimes. seismometerwas inhibited by other experiments have been There is no firm evidencethat any high Q lander resonances 4536 SEISMOLOGY

!

o o o o o

.. ANDERSONET AL.:SEISMOLOGY ON MARS 4537

•,i/59

<1.) (:D .

<3<

..

9 4538 SEISMOLOGY

SOL 33 VL-2 Figure 17 showsthe high degreeof correlationof the ampli- tude of the seismicsignals recorded by the seismometerin the

• 20 normal mode (Z component) with the wind speedsrecorded & 0 by the meteorologyexperiment. The data shownin Figure 17 o 20 J 0 indicate that the seismic amplitude is proportional to the • •0 squareof the wind speed,as is expectedfor turbulent flow. 0 Detailed correlations are difficult, sincethe seismicdata repre- sent averagesand the wind data are widely spacedpoint sam- ples.A similar figure for X componentdata in the event mode is shown in Figure A-19. In the high data rate mode of oper- • 0 _ ation, where no averaging is involved, seismicactivity peaks 13 57 24 14 O0 33 14 03 41 and wind speed peaks are also reasonablywell correlated. LOCAL LANDER TIME Figure A-20 showsthe correlation betweenZ component am- Fig. 16. Correlationof windspeed with seismicnoise on sol33. In plitudesand wind speedsin the high data rate mode on sols35 thiscase, wind gusts of 7-8 m/s producednoise levels of 16-20du. The and 36. Data observedin the high data rate mode also show wind detection threshold for seismic noise is about 3-4 m/s. that the seismicamplitude increasesas the squareof the wind speed.Data plotted in Figure 17 alsoshow the correlationof are excitedby any landeractivity. There is a suggestionof a seismic noise for different wind directions. The data are in resonanceat 7.6 Hz, seenduring the initial slew of the high- good agreementwith the relation A = 0.15V •, where A is the gain antennaeand occasionallyduring a samplingarm oper- mean seismicamplitude in digital units (event mode) and V is ation, and also a resonanceat frequenciesof 14-15 Hz based the meanwind speedin metersper second.The scatterof data on infrequenthigh-amplitude impulsive events of possible at low speedssuggests a thresholdof about 3 m/s. thermalorigin. High Q landerresonances were designed to be During the period of conjunction(sols 60-90) the back- outsidethe usefulfrequency band of the seismometer.There is ground noise level was extremelylow from 1800 hours each also no evidencefor high Q wind-inducedlander resonances. eveningto 0700 hoursthe followingmorning, the background WIND-GENERATED NOISE level increasingslightly starting between0630 and 0700 hours. The afternoon gusty wind period began between 1018 and The main sourceof external noiseis the wind blowing on the 1230 hours each day, usually closeto local noon. Peak gusts lander. Generally speaking,the wind noise backgroundre- usually occurrednear 1430 and 1520 hours. Typically, 8-12 peatsfrom day to day, but the backgroundis quitevariable wind gustswere observedbetween 1200 and 1800hours lasting when it is looked at in detail. Some days are very noisy 8-15 min and separatedby intervalsof roughly 40 min. The (windy), and somedays are very calm. Martian windsand gustsproduced seismic amplitudes ranging from 10 to 20 du in hencethe backgroundnoise level vary distinctlywith the sea- the normal mode. son. Prior to conjunction,during summerin the northern Northwestwinds typicallyoccur in the afternoonfrom 1300 hemisphere(sols 0-60), the averagewind speeds were observed to 1800 hours. Northeast winds generally occur in the early to be somewhathigher in the morningthan in the afternoon, morning hours, changing to southeastwinds from 0700 to but individualwind gustshad higherpeak speeds in the after- 0900 hours and to southwest winds from 0900 to 1200 hours noon [Hesset al., 1977].This is reflectedin the seismicdata [Hesset al., 1977]. During the postconjunctionperiod from (Figure 13). The highestseismic background typically oc- sols 110 to 118 the winds were generally lessthan the seismic curred between 1300 and 1700 hours (Mars local time), when threshold (•3 m/s) from about 1800 hours to sometimebe- the seismicamplitudes ranged from l0 to 40 digitalunits (du). Noise burstswere typically 1-3 min in duration, separatedby time intervals of 10-50 min. The seismic background level i i i i i i i I i dropsbetween individual bursts of wind; this drop indicates VL-2 Sols 70-90,119,129, 1:51,141 , that there is little direct transfer of energyinto the ground by • 40 []NE WIND ...... --/ 131 the wind and that the principal effecton the seismometeris F"'l(l• 30 '"xSESW WIND 14, producedby wind-inducedvibrations of the lander. õ • •0 o. w,. Typically,the quietestrecording interval occurs from about 1800 hours local lander time (LLT•j (2 hours before sunset)to ß•.aolo 0 ,,• - about 0400 hours the following morning (Figure 14). During this time interval the wind speeddecreases to lessthan 1-2 m/s, and the seismicnoise background level falls to between1 • 0 - •Z 4 x •o and 2 du. (In the event mode at maximumsensitivity, I du • x x ooo• _ correspondsto 2 nm of instrumentdisplacement at 3 Hz.) • xXx • o _ Around Martian sunrisethe backgroundnoise level generally increasesto about 2-3 du (Figure 15). x • &xOo• x I • I I I I I I II An exampleof the correlationbetween seismic noise activity 2 3 5 I0 in the eventmode and a wind gustrecorded on sol 33 is shown Meon wind speed, m/sec in Figure 16. In this casea 7- to 8-m/s wind gustproduced a seismicsignal of 16- to 20-du amplitude.A similarfigure is Fig. 17. Correlationbetween wind speed and normal mode seismic shown in the microficheappendix (Figure A-18) coveringa amplitude(Z component)for windsfrom differentdirections. The solidline correspondsto the slopeexpected if seismicamplitude is differentperiod of time.The thresholdof winddetectability of proportionalto the squareof the windspeed, as is appropriatefor the seismometer, as seen in the event mode at maximum turbulentflow. The scatterof pointsbelow wind speeds of 3 m/s is due sensitivity,appears to be about 3-4 m/s. to noise from other sources than wind. ANDERSON ET AL.: SEISMOI.OGYON MARS 4539

70- tween 2 and 3 Hz, closeto the natural period of the seismome- SOL1:58 ter system.Also shown are data from a noiseburst on sol 138. 20 37 2319- 60- 2039 52 45 Figures A-21 and A-22 show histogramscompiled for a 15- SOL 39 min interval during an unusuallywindy period near Martian I-- .50- midnight on sol 133 and a similar 5-min interval on sol 139. As ZI,J.I NW12 53WIND 39- 136- 08. 8 43 m/s NW17' 34,24WIND -61m/s 7 37' 03 before, the main frequency appears to be concentratedbe- tween 2 and 3 Hz. There are no obvious lander resonances or LL 40-- 0 preferred excitation frequencies.The shape of the histograms reflectsthe responseof the instrument.

WIND STRESSES ON THE LANDER

20- Wind stresses on the lander are of the order of 0.002-0.2 dyn/cm2, depending on wind velocity. Although the wind guststhemselves are of generally low frequency,large changes in amplitude occur of the order of secondsor less.The lander

l! ' itself can introduce a high-frequencymotion due to turbulent I00 4 6 lB, r-, I0 12_, 14,--',, 16 .•] I0, 0 2 4 6 8 i0 shedding at preferred frequencies.The latter are related to FREQUENCY, Hz wind velocity, the characteristicdimensions of lander com- Fig. 18. Axis-crossinghistograms for windy periods.The number ponents,and the Strouhal number via of events(ordinate) is a measureof the amplitude at the corresponding frequency.If high Q lander resonanceswere excitedby winds, these f- sV/h histogramswould showsharp peaks at the resonantfrequencies. Such peaks are not observed,and the shapeof the histogramsreflects the where V is the wind velocity, h is the dimension of the lander frequencyresponse of the seismometer. or extendedcomponents, and s is the Strouhhlnumber, which is about 0.2 for cylindrical or blunt obstacles.Taking 1 m as the characteristicdimension of the lander, the sheddingfre- tween 0600 and 1030 hours LLT. Winds generally averaged quency is 4 Hz for 20-m/s winds and 0.4 Hz for 2-m/s winds. from 2.5 to 4.9 m/s from 1000 to 1800 hours LLT, with Therefore vortex sheddingfrequencies are in the seismicband. maximum wind peaks from 6.4 to 11.5 m/s. The seismicdata There is a slight indication from the seismicdata that in- were characterizedby quiet nights,relatively quiet mornings, creased frequency correlates with increasedamplitude (and and increasedactivity in the afternoon. wind velocity), but the passbandof the instrument is too Unusually high seismicbackground was recordedon sol 119 narrow to make a detailed correlation at this stagein the data beginning at 1100 hours LLT. The noise level was continuous processing. and unlike the usual afternoon windy periods.Average wind The vibrations of the lander due to wind action are propor- speedsin this period were unusually high even after 1800 tional to the square of the wind velocity V. Wind forcesr on hours, when it is usually quiet. The wind speedsaveraged the lander can be written greater than 7.5 m/s during the time interval from 1200 to 1800 hours, with peak gustsof 11.6, 13.2, and 15.1 m/s. Quite r e pAV • surprisingly,the wind speedsfrom 1800 hours to midnight where A is the cross-sectionalarea of the lander and o is the ranged from 4.7 to 6.4 m/s for successivel•-hour-long mete- density of the atmosphere. The seismometer threshold to orology samples.Wind speedpeaks of 11.4, 10.7, 9.5, and 8.9 winds is about 3 m/s. The location of the seismometer on the m/s were observedthroughout this normally quiet period. top of the lander magnifiesthe wind-inducedstrains, and the Continuous, although lower-level,activity also occurredin responseof the lander to wind stressescan be written the early morning hours on sols 118, 121, 128, 129, and 130. Exceptionallyhigh activity was observedon sol 131 lasting r = Me = Md/l until 0200 hoursLLT on sol 132. The mean seismicZ ampli- tude was 37.1 du with a maximum of 56 du in the normal mode where M is the effective modulus of the lander system,• is the at 0 dB. Extrapolating the relation found earlier (Figure 18) strain, d is the displacement,and I is the distance of the suggestsmean wind speedsof 16-18 m/s, a value higherthan seismometerfrom the contact betweenthe legsand the surface. any previouslyreported. Unfortunately, becauseof a tempo- In order to reduce wind noise therefore, an optimum seismic rary malfunction,there were no meteorologydata duringthis experiment would minimize A and I and maximize M. The period. Imagery resultsshow that at some time betweensols cross-sectional area of the lander is about 104 cm •'. For an 110and 120the optical depthapparently increased by an order offloaded seismometer this could be reduced to 10•' cm•', an of magnitude. Average temperatureswere also observedto immediate decreasein wind stressesof a factor of 10•' being drop rapidly on sols 119 and 120, 128, and 131. All of these given, raisingthe thresholdwind speedto 30 m/s. The modu- observationsare suggestiveof an unusualmeteorological dis- lus M of a compact surface instrument could probably be turbanceor the onsetof a new meteorologicalregime due to increasedby about an order of magnitude,and the height of the change of seasons. the packagecould be decreasedby an order of magnitude,an Beforelanding, there was somequestion about whetherthe improvement of another 2 orders of magnitude thus resulting wind would excitehigh Q resonancesin the lander. To exam- in the signal-to-noiseratio. In principle therefore a 104 im- ine this question, zero-crossing(frequency) histogramsfor provement in seismometersensitivity could be usefully em- different times of day and wind conditionswere compiled. ployed for a Martian seismometer,and direct wind-induced Figure 18 shows a zero-crossinghistogram for a moderate noise would be no problem for winds as high as 300 m/s. northwesterlywind during sol 39 at two times during the Future missionsto Mars could have seismometersof Apollo- afternoon windy period. The predominantfrequency is be- like sensitivity.This is important not only to increasethe range 4540 SEISMOLOGY

Table 2 summarizesthe spectralslopes for 34 samplesof I I I windnoise. The slopesare substantiallylower for thevertical componentthan theyare for the horizontal.The slopesvary from 0.58 to 0.98 with a mean of 0.91 + 0.10 on the Z axis.

TRANSIENT EVENTS Many transientseismic signals have been detected by the Vikingseismometer. However, as is described elsewhere inthis paper,wind-induced vibrations of the landergreatly com- plicatethe problem of establishingthe sources of theobserved signals.Winds were light during th• midsummerto latesum- merperiod of operationat site2; thusabout half of eachday was left relativelyfree of wind disturbances.With the ap- proai:hof fall at site2, however,winds began to increaseto the pointwhere wind disturbances occasionally persisted through- out the Martian day. In the absenceof additionalstations for intercomparisonour onlyrecourse is to searchthe records for signalsof unusualcharacter and to determineif thesesignals correlate with local-wind-generatedor lander-generated events.We have seenthat seismicsignals do correlatequite closelywith variationsin wind speedand directionand that the patternsof thesevariations can be quitecomplex. Onsets of wind gustscan be abrupt, and isolatedgusts can occur duringotherwise quiet periods. Unfortunately, wind data are not obtainedby the meteorologyexperiment on a continuous

I I I basis,so correlationwith seismicdata is not alwayspossible. -2.000 -1.000 0.0 1.000 2.000 For example,during the conjunctionperiod, wind data were LOG FRE0 (CYCLES/MIN) only availablefor about 17%of the total time. From the seismicdata availablethrough sol 144, eight sig- Fig. 19. Powerspectrum of eventmode amplitude envelope data for a 21•-minperiod starting at 1429:18on sol 16.Two majorwind nalswere selectedfor further analysisprior to receiptof wind gustsduring this period provide the energyfor thisspectrum. The data. Theseare shownin FiguresA-24 to A-29 and Figures20 slopeof thespectrum at frequencieshigher than 1 cycleper minute is and 21. Selectionwas basedprimarily on their relativelyim- about -0.9. pulsiveonsets and relativeiy low axis-crossing counts. Follow- ingreceipt of winddata, six of theseevents were found to of detectabilitybut becausethe numberof smallevents is correlatewith wind gusts. Wind datawere not obtained during undoubtedlymuch greater than the numberof largeevents. the time intervalsof the remainingtwo signals.The times, Magnitude-frequencyrelationships indicate that for eachor- maximumwinds, and average signal frequencies for eachevent derof magnitudeincrease in instrumentsensitivity, there will aregiven in Table3. Signalfrequencies are average values for be an order of magnitudeincrease in numberof detected thevertical component over 40-s samples centered at thetime events.An optimalseismic network on Mars thereforecould of peaksignal amplitudes. On the basisof thesedata, there usefullyemploy magnitude zero events to definethe seismicity appearsto be sometendency toward increasing signal fre- and internal structure of the planet.

NOISE SPECTRA TABLE 2. Seismic Noise (Wind) Spectra, 55-Min Samples, The spectraof wind-inducedseismic noise are not only FrequencyRange 0.05-0.5 Hz, and PSD With MeanRemoved usefulfor micrometeorologicalpurvoses but arepotentialiy usefuldiscriminants for identifyingnatural events. Most of the -d log PSD meteorologicaldata have been taken at samplingintervals of Axis d log f a Samples 4 s or greater,although occasionally wind speed is sampled at Morning 1.2-sintervals. Higher samplingis necessaryto definethe X 0.58 0.04 9 turbulent characteristicsof the lower Martian atmosphere. Y 0.68 0.08 9 Figure19 shows the power spectrum of theamplitude enve- Z 0.88 0.10 9 iopefor a 21.5-minsection of event mode data (Z axis)starting .4fternoon at 1429:18 hours on sol 16. There are two major wind gustsin X 0.78 0.05 5 Y 1.00 0.14 5 thisperiod. At periodsof lessthan 1 rainthe powerspectrum Z 0.98 0.11 5 has a meanslope of -0.9. Data at periodsof lessthan 15 s maybe aliased by high-frequency changes in theenvelope and X 0.73 Night 0.06 3 should be ignored. Y 0.80 0.00 3 Z O.9O 0.00 3 FigureA-23 gives the power spectra of all threeaxes of the seismometerfor a high-wind period of sol 131 startingat .411Day x 0.66 0.10 17 2015:34 hoursand extendingfor 55 min. The nearlymonoto- Y 0.79 0.17 17 nic decreaseof powerwith increasingfrequency with a slope z 0.91 0.10 17 slightlyless than unityis typicalof the windnoise spectra on Mars. PSD is power spectraldensity. ANDERSONET AL.' SEISMOLOGYON MARS 4541

SOL 53 13:32 10 m/sec

ß ß ß Bin m I ßIm m_ ß

ß li

n, ii ß

[dh,.k.i.l-,..,lkd.mb.bn•

.,mdh#di!lldh,,_L[bndl.___L.__lhl_._nt_lJ!d I I I I I I I i I I

Fig. 20. Sol 53 1332LLT eventrecorded in eventmode. Although it was thoughtto be a possibleMarsquake when it was first observed,this event was later correlatedwith a 10-m/s gust of wind. quency with increasingwind speed.However, the frequency the spectraof any of the other candidatesignals except that of content of the sol 80 signal, for which there are no wind data, the sol 42 signal. The signal character in the time domain is is unusuallyhigh in relation to the remainderof the data set. quite differentin thesetwo cases,however. The high-frequency Other unusualcharacteristics of the sol 80 signal are worthy of character of the sol 80 and sol 42 signals is also distinctly note: (1) It is the only signalof the group that occurredduring different. Histogramsgiving the distributionof zero crossings the normallyquiet period of the early morning,(2) both ampli- for the sol 80 event and three wind eventsare given in Figure tude and frequencyvariations occur at intervalsof about 10 s, 23. The dominant frequencyon the horizontal axesis 3-4 Hz and (3) the amplitude variationsare similar to those of two in all cases, close to the natural period of the seismometer local eventsrecorded by the Viking seismometersduring oper- system. The same is true for the vertical component of the ational testsin southernCalifornia. Regardingthe first point known wind events.The vertical (X) componentfor the sol 80 an isolated wind gust was detected during the late evening event has peaks at 2, 4, and 7 Hz. A peak at 6 or 7 Hz also hours of sol 82, so suchgusts, though they are infrequent,can showsup in the horizontal componentsof the wind gusts.The occur. sol 80 event is distinctly bimodal in comparisonto the other The 10-speriodicity in the amplitudevariation of the sol 80 events. signalis clearlyseen in the spectrumof the signalenvelope in The sol 80 signal is comparedin Figure 24 with two terres- Figure 22. A pronouncedspectral peak of 10 s is not seenin trial signalsrecorded with a test instrument operated at the 4542 SE•SMOLO•¾

California Institute of Technologycampus in Pasadena.The approximately 13 s. This places the epicentral distance at uppermostsignal is an aftershockof the 1971 San Fernando about 110 km for a surface focus event if crustal velocities earthquake, recordedat a range of 65 kin, the middle signal normal for the earth are assumed.The maximum amplitudeof was generatedby 102 metric tons of chemicalexplosive deto- the shear wave arrival (peak 1, average of Y and Z com- nated in a rock quarry at a distanceof 69 kin, and the bottom ponents) is 188 nm at a frequencyof about 5 Hz. The trace signal is the sol 80 event recordedon Mars. amplitude on a standardWood-Anderson seismograph (To of Obviously, a strongcase for seismicorigin for any Martian 0.8 s and magnificationof 2800) would be 0.53 ram, yielding a signal cannot be made until a strong signal is obtained at a Richter magnitude of 2.8 for the event. time when wind speedsare known to be low. However, on the The magnitude-duration relation determined for southern basisof the similaritiesbetween the signalsrecorded from local California is given by shallow sourcesin southernCalifornia and the sol 80 signal, let Mt = -0.68 + 1.75 log t + 0.0014A us see if the amplitude-duration relations are consistentwith the hypothesisthat the sol 80 signalwas generatedby a Mars- where t is the signal duration in secondsand A is the distance quake similar in crustal structure to that of the terrestrial in kilometers. If t = 70 s and A = 110 km are used, this examples.Interpreting the amplitude and frequencyvariations expressionyields Mt = 2.7, in good agreementwith the magni- as P and S arrivals, as shown in Figure 25, the S-P time is tude estimate from maximum signal amplitude. The duration

SOL 80 03:00

ß Ill I I

I

I

Fig. 21. Sol 80 0300 LLT event. This event is the most likely candidatefor a Marsquake recordedto date. Becausewind data werenot recordedat the time of thisevent, it is possiblethat thisevent could also have been generated by the wind. ANDERSONET AL.: SEISMOLOGYON MARS 4543

TABLE 3. Times, Average Frequencies, and Wind Speeds for SOL 80 SOL 42 SOL 53 SOL 23 Eight Seismic Signals

Local Lander AverageSignal Maximum Wind Sol Time, hours Frequency,Hz Speed,m/s

12 1650 3 48 no data 23 1515 3 6O 12 42 1554 3 9O 11 49 1345 373 12 40 49 1435 2 98 8 49 1548 3 5O 10 53 1332 3.45 10 20 All 80 0300 4.80 no data

0 0 I0 0 I0 0 I0 0 I0 of the sol 80 event is therefore consistentwith the amplitude Frequency, Hz for a Marsquake at a distance of 110 km in a crust having Fig. 23. Histogram of the sol 80 axis crossingscompared with scattering and attenuation properties similar to those of the histogramsfrom three other events. The sol 80 event is noticeably earth's crust. richer in high frequencies. There are at least two amplitude peaks (peaks 2 and 3) following the main peak at intervals of 10 s (peak 1 to peak 2) stressed.Aside from possiblestructural interpretationsof the and 9 s (peak 2 to peak 3). One way to explain this pattern is type describedabove, much can be learned from signalcharac- by shearwave reflectionsin a surfacelayer. Many investigators teristicsalone. For example,if the sol 80 eventis a Marsquake, have recognizedsuch reflectionsin interpretation of seismo- the similarities in signal characteristicsbetween it and the grams and have usedthem in determinationof crustal struc- terrestrialcases would indicatethat the sourceparameters and ture [e.g., Gutenberg,1944]. If the two-way travel time for a transmission characteristics of Mars and earth are similar. shear wave traveling vertically in a surface layer is taken to This would be in marked contrast to the lunar case, in which equal 9 s and the layer shearwave velocityto equal 3.5 km/s, signalsare prolonged,with little coherencebetween the three this interpretationgives a discontinuityat a depth of approxi- componentsof ground motion. This is ascribedprimarily to mately 16 km beneath the region of landing site 2. The peak the very high Q of the outer shell of the Moon and a sharp amplitude ratios and assumedQ of the layer can be used to increasein velocity with depth in the superficial(regolith) calculate the reflection coefficientof the discontinuity. How- zone. A moonquakewith magnitudeand range equal to those ever, further speculationconcerning the signalis of little value of the casesdiscussed above would have generateda seismic unlessadditional signalspossessing similar characteristicsare obtained. signal lasting several hours instead of 1 min. If similarity between Martian and terrestrial signalscan be established,we The importance of positive identification of even a single can say immediately that a lunarlike regolith is unlikely on event as being of internal (Marsquake) origin cannot be over- Mars and that absorption of seismicwave energy is much higher in the region of Mars traversedby the observedsignal.

_ The high Q in the case of the moon can be attributed to the 1.Oj•. I I • •• I• I ISOL I 80 I I I IlL MPONENT absenceof volatiles in the outer layers. In summary, during the first 146 sols of operation of the Viking 2 seismometer,only one event has been detectedthat might be a local Marsquake. No eventsthat can be interpreted as large teleseismicMarsquakes have been identified with cer- tainty. Theseconclusions are basedon visual inspectionof the 0.1 records.Only limited digital processinghas been done to date. It should be pointed out that the seismometerhas only been operatingat highestgain and in the optimal modes(high data rate or event mode) during quiet periods,uncontaminated by lander activities or winds, for a cumulative total of about 30 days. This is too short a period of time to draw definitive conclusionsabout the seismicityof Mars. It should also be pointed out again that the seismicityof Mars is likely to be 0.01 nonuniform both in time and space,as it is on the earth and the moon, and that the Viking 2 landingsite is remotefrom the possiblymost activeregions of the planet inferredfrom photo- geology and gravity information.

DETECTION CAPABILITY, SEISMICITY,AND METEOROlD IMPACTS The detection of Marsquakes by the Viking seismometer 0.0010.01 I II iii111 0.1 I II l l•jt 1.0 FREQUENCY, Hz system depends on the size and distance of the event, the magnification (gain level) of the instrument, and the back- Fig. 22. Spectrumof the event mode amplitudeenvelope for the sol 80 event. The 10-s periodicity observedin the time domain is ground noise level. The noise level is discussedextensively clearly visible in the spectrum. elsewherein this report. Furthermore, the amplitude of the 4 544 SEISMOLOGY

tude 6 over half the surfaceof the planet(A = 90ø). It is likely that an event of magnitude6.5 or greater occurringanywhere SYLMAR AFTERSHOCK on the surface of the planet can be detected by the Viking M3.0, 65Km seismicinstrument when the winds are relatively low. To evaluate the level of seismicactivity on Mars, it is neces- sary to analyze the available data, taking into account the operation modesof the instrument.The Marsquake detection capability of the Viking seismographsystem has not been uniform throughout this period of observationbecause of the variability of wind interference, other lander activities, and •', , , ,.--; : : ...... limited data allocations. The detection capability is highest when the instrumentis operatedeither in the high data rate or event mode, the level of wind-generatednoise is lessthan 1-2 Iil 2El 3El qla Kl• I::;• "lla Ella 9t• IEI• ii0 120 du, and there are no other interfering lander activities.Under theseconditions, seismic signals of amplitudesgreater than 25 du should easily be identified.These conditionswere met for a total of 640 hours during the first 146 sols of operation, or about 18% of the total time of operation. On the basis of these data the comparative seismicityof Mars relative to that of the earth and moon can be estimated. On the earth, there are about 45 earthquakesper year of magnitude6.5 or greater. Most earthquakesare at shallow depths.If the averagenumber of Marsquakesper unit areaof the Martian surfacewere the sameas that on the earth, with the detectioncapability of the Viking seismometerone would expectto detect 13 Marsqua.kesof magnitude6.5 or greater per year, or approximatelyone per month. No suchevents have yet been recognizedduring the 640 hours of the high- sensitivitylow-noise recording period of the Viking mission. From this limited sampleit appearsthat Mars is seismically less active than the earth. A more quantitative estimate of comparative seismicityof Mars relative to that of the earth and moon can be made by taking into account smaller events. If we assumethat Mars- quakes above a certain magnitude, like earthquakes,can be Ill 2•1 ':lB qll œGI GGI '7GI Bi• 91• Iia• Ilil 121il I TIME (5ECEINI)5)

EVENT MODE COMPRESSED DATA 5BL BIZi EVENT , ; : : : : ; ; : : : : : : : : : : Fig. 24. The sol 80 event comparedwith similar eventsrecorded in LIB 1 southernCalifornia by a Viking seismometerlocated at the California Institute of Technology.All eventswere recordedin the event mode. < ZB 2 3 rr x ground motion producedat the Viking site from a given M ars- quake dependson the velocity structure and the attenuation I I characteristics of the Martian interior. The internal structure modelscalculated on the basisof availableconstraints predict the presenceof a core shadowzone starting at about 90ø- 120ø. The seismicray paths for one suchMars model are shownin Figure 1. The event detection capability of the Viking seismic in- strument has been determined by combining the actual test data obtained in the earth and the ray theoretical amplitudes of seismicbody waves calculated for a theoretical model of Mars. An averagemantle Q value of 2000 was assumed.The detection threshold as a function of distanceis shown in Fig- ure 26. This curve is based on the condition that an event can ..,,.:.•: ...... i B 2B :3B LIB be identified if it producesa ground motion equivalentto 25 TIME (SECEINDS) du or greater in the output seismic data. This is indeed a conservativecase, since during wind-free periods the back- Fig. 25. The sol 80 eventas recordedon all three axes.The P and S ground noise rarely exceeds3 du. Figure 26 shows that the refer to the possiblearrivals of compressionaland shear waves.The Viking seismometershould detect Marsquakes of magnitude3 peaks labeled 2 and 3 are possiblereflections from the base of the at a distanceof 200 km, magnitude 5 at 1000 km, and magni- crust. ANDERSON ET AL.' SEISMOLOGYON MARS 4545

have detectedthree high-frequencytransient events(possible shallow moonquakes at large distances [Nakamura et al.,

VIKING 1974]) and one meteoroid impact during the 6 years of oper- DETECTION ation. Multiplying by a factor of 4 to accountfor the difference 5 in the surface areas of the moon and Mars, we would thus expectto have two Marsquakes of detectablesizes annually if Martian seismicitywere the sameas lunar seismicityand if the Q of the Martian interior were as high as that of the lunar mantle, which is about 4000 [Dainty et al., 1976; Nakamura et al., 1976]. If we usestatistics similar to thoseabove, the chance that no Marsquakes are observedto date when an average22 Marsquakes are expectedper year (i.e., activity an order of magnitudehigher than that on the moon) would be 20%. Thus

.1 I J I Ill for a Mars of low attenuationcomparable to the Moon we can 10 100 1000 10,000 DISTANCE, (krn) say with 80% confidencethat the seismicityof Mars cannot be greaterthan an order of magnitudeabove the lunar seismicity. Fig. 26. HypotheticalMarsquake detection threshold values based With the presenceof volatiles the Q of the Martian interior on an averagemantle Q of 2000 and a detectionthreshold amplitude most likely is lower than that of the moon. If we assumea Q of of 25 du. It is likely that a Marsquake of magnitude 6.5 or larger 500, the detectionof moonquake-sizedevents will be limited to occurringanywhere on Mars could be detected. within 500 km of a station. Then if we assume uniform areal distributionof Marsquakes,the expectednumber of detectable representedby a random Poissonprocess, the probability of Marsquakes would be reduced to 1 in about 50 years if the observingn eventsover a period interval T is Martian seismicitywere the same as the lunar seismicity. P[n, T] = [(aT)"/n!]e-"r In summary,the Viking seismicobservations to date are not sufficient to determine whether Mars is more active than the and the probability of observingno eventsis moon. P[0, T] = e-•r We next examine the question of the expectedrate of detec- tion of meteoroid impacts on Mars. Approximately 300 mete- where a is the mean rate of occurrenceof events.Substituting oroid impacts are detectedon the lunar surfaceeach year by the above mentioned values into this equation, we obtain the instrumentsof the Apollo SeismicNetwork [Duennebieret e[0, 640] = 0.39 al., 1976]. These fall in the massrange from about 0.5 kg to 1 metric ton. Most estimatesof meteoroid flux in the vicinity of Thereforealthough it is moreprobable that Mars is lessactive Mars are equal to or larger than those in lunar space, so we than the earth, there still is a 39% probability that Mars can be might reasonablyask, Where are the impact signalson Mars? as active as the earth if the landing site happens to be an Three factorsreduce the likelihoodof impact detectionori aseismicregion of Mars. If we considerevents of magnitude4 Mars in relation to the moon. First, accordingto the theoreti- or greater and if we assumethat Marsquakes are uniformly cal resultsof Gault and Baldwin [1970] the Martian atmosphere distributedin space(thus assumingthe nonexistenceof aseis- is surprisinglyeffective as a shield againstmeteoroid impacts. mic regions), the number of expected Marsquakes will be Their results indicate that the number of meteoroids of 10-kg increased to about 40 per year at earth rate. Therefore the mass reaching the surface would be only 10% of the number probability of not detecting any event during 640 hours of incident on the atmosphere, and the velocity of those that observation under ideal conditions becomes reach the surface would be reduced to several hundred meters per second.In fact, the impact velocity of objectsas large as 1 P[0, 640] = 0.05 metric ton will be significantlyreduced. The predicted shield- Thus with the data collected so far and with the assumptions ing from objectsin the smallto intermediatemass range is stated, the seismicityof Mars per unit surfacearea is lessthan consistentwith the angular appearanceof the surfacefeatures that of the earth with a probability of 0.95. The additional data of Mars comparedwith the muchmore subdued lunar mor- that will be collected during the next 2 years clearly will place phology. A second factor bearing on relative rates of mete- more definite limits on Martian seismicity.It is also probable, oroid detection is instrument sensitivity.The peak sensitivity as we discussedearlier, that the seismicity of Mars is not of the Viking seismometeris about 70 times lower than that of uniformly distributed in spaceand that the Viking 2 site may the short-period Apollo seismometers.If the Apollo in- be in a relatively aseismicpart of the planet. strumentshad been operating at the sensitivityof the Viking A comparison of Martian seismicitywith lunar seismicity seismometers,only one impact would have been recorded can be made by usingthe data obtained from the seismometers during the entire period of operation of the Apollo network (7 deployed on the lunar surface during the Apollo mission yr). [Latham et al., 1973]. The four seismicstations that are cur- Finally, the influence of seismicwave absorption must be rently in operation are detecting moonquakesand meteoroid considered.The Q of the outer shell of the moon has been impact signalsat a rate of about 2000 per year. The high rate estimated to be about 6000 [Latham et al., 1973]. Typical of detection, however, is not due to the high seismicityof the terrestrial values for shallow depths range from 200 to 300. moon but is due primarily to the high sensitivityof the in- This differenceis ascribedprimarily to the quantity of water struments,the low background noise level, and the low atten- contained in the medium: reduced water content resulting in uationcharacteristics of the lunarinterior. ' higher Q. On this basiswe might expect Q valuesin the outer Correcting for the differencein instrumentresponse, a Vik- shell of Mars to be much lower than those found in the moon ing seismometerdeployed at the Apollo 14 landing site would but possiblyhigher than terrestrial values. 4546 SEISMOLOGY

If we assume a Mars Q of 500 and a moon Q of 6000, the Anderson, D. L., F. K. Duennebier, G. V. Latham, M. N. Toksoz, attenuation of a seismic wave at a range of 2000 km will be 40 R. L. Kovach, T. C. D. Knight, A. R. Lazarewicz, W. F. Miller, Y. Nakamura, and G. Sutton, The Viking seismic experiment, Science, times greater for Mars than for the moon. 194, 1318-1321, 1976. If we take these three factors (atmospheric shielding, in­ Bills, B. G., and A. J. Ferrari, Mars topography harmonics and strument sensitivity, and attenuation) together, the likelihood geophysical implications, J. Geophys. Res., 82, in press, 1977a. of detecting a meteoroid impact on Mars with the Viking Bills, B. G., and A. J. Ferrari, A lunar density model consistent with seismometers during the planned lifetime of the experiment is topographic, gravitational, librational, and seismic data, J. Geophys. Res., 82, 1306-1314, 1977b. extremely small. Binder, A. B., Internal structure of Mars, J. Geophys. Res., 74, 3110-3117, 1969. IMPLICATIONS FOR FUTURE MISSIONS Binder, A. B., and D. R. Davis, Internal structure of Mars, Phys. The Viking experiment, even though it was highly curtailed Earth Planet. Interiors, 7, 477-485, 1973. by mission priorities and by the inability to uncage the first Blasius, K. R., and J. A. Cutts, Shield volcanism and lithospheric structure beneath the Tharsis plateau, Mars, Proc. Lunar Sci. Con/ seismometer, has provided and will continue to provide valu­ 7th, 3, 3561-3574, 1976. able information for the planning of any future missions to Dainty, A. M., M. N. Toksoz, and S. Stein, Seismic investigation of Mars. First, it established that the seismic background noise the lunar interior, Proc. Lunar Sci. Con/ 7th, 3057-3075, 1976. on Mars due to winds and atmospheric pressure fluctuations is Duennebier, F., Y. Nakamura, G. Latham, and H. Dorman, Mete­ oroid storms detected on the moon, Science, 192, 1000-1002, 1976. very low. The Viking seismometer, mounted on a relatively Gault, D., and B. Baldwin, Impact cratering on Mars-Some effects of compliant spacecraft with a large surface area exposed to the atmosphere (abstract), Eos Trans. AGU, 51, 343, 1970. winds, could still operate at maximum sensitivity at least half Gutenberg, B., ReAected and minor phases in records of near-by of the time during the nominal mission, a situation that in­ earthquakes in southern California, Bull. Seismal. Soc. Amer., 34, dicated that seismometers with much greater sensitivities can 137-160, 1944. Hess, S. L., R. M. Henry, C. B. Leovy, J. A. Ryan, and J.E. Tillman, be operated on the planet. Emplaced by penetrators or de­ Meteorological results from the surface of Mars: Viking I and 2, ployed as small packages, seismometers more sensitive than J. Geophys. Res .. 82, this issue, 1977. the Viking instrument by a factor of at least I Q3 can operate on Johnston, D. H., and M. N. Toksoz, Internal structure and properties the planet without being affected by typical Martian winds. of Mars, Icarus. in press, 1977. Although the data on Martian seismicity are still prelimi­ Johnston, D. H., R. T. McGetchin, and M. N. Toksoz, Thermal state and internal structure of Mars, J. Geophys. Res., 79, 3959-3971, nary, the indications are that Mars is probably less active than 1974. the earth. Greater sensitivity is a must for any future seismic Kovach, R. L., and D. L. Anderson, The interiors of the terrestrial instruments on the planet. planets, J. Geophys. Res., 70, 2873-2882, 1965. Another important consideration is the deployment of a Latham, G. V., M. Ewing, F. Press, J. Dorman, Y. Nakamura, M. N. Toksoz, D. Lammlein, F. Duennebier, and A. M. Dainty, Passive network of instruments. With a well-placed network of five seismic experiment, Apollo 17 Preliminary Science Report, NASA very sensitive Apollo seismographs it was possible to study the Spec. Pub/ .. 330, 1-9, 1973. moon and determine its structure, even though most moon­ Lee, B. G., Mission operations strategy for Viking, Science, 194, quakes are very small. A similar approach of deployment of a 59-62, 1976. network of highly sensitive instruments is needed for Mars Nakamura, Y., J. Dorman, F. Duennebier, M. Ewing, D. Lammlein, and G. V. Latham, High frequency lunar teleseismic events, Proc. exploration. Since Mars is a larger planet than the moon and Lunar Sci. Con( 5th, 3, 2883-2890, 1974. shows much greater geologic and tectonic diversity, both Nakamura, Y., F. Duennebier, G. V. Latham, and J. Dorman, Struc­ global and regional seismic networks are needed to understand ture of the lunar mantle, J. Geophys. Res., 81, 4818-4824, 1976. Martian structure and tectonics. It is important that future Okal, E., and D. L. Anderson, Seismic models for Mars, submitted to Icarus. 1977. missions deploy such networks. Owen, T., and K. Biemann, Composition of the atmosphere at the surface of Mars: Detection of argon 36 and preliminary analysis, Acknowledgments. The Viking seismometer was designed and Science. 193. 801-803, 1976. tested by F. L. Lehner and W. Miller at the California Institute of Phillips, R. J., and R. S. Saunders, The isostatic state of Martian Technology. Flight hardware was designed and fabricated by J. topography, J Geophys. Res., 80, 2893-2897, 1975. Lewko, D. Gibson, M. Van Dyke, T. Gaffield, and D. LaFeniere of Phillips, R. J., and M. Tiernan, Martian stress distributions: Argu­ Bendix Aerospace. The large-scale integration circuitry was fabricated ments against the static support ofTharsis, submitted to J. Geophys. by American Microsystems, Inc. Wyatt Underwood has been in Res. 1977. charge of downlink operations. George Sutton and Ken Anderson Phillips, R. J., R. S. Saunders, and J.E. Conel, Mars: Crustal structure were involved in various aspects of the experiment. We thank W. inferred from Bouguer gravity anomalies, J. Geophys. Res .. 78, Kaula, R. Phillips, B. Bills, A. Ferrari, and E. Okal for useful dis­ 4815-4820, 1973. cussions and for some of the calculations given in this paper. The Reasenberg, R., The moment of inertia and isostasy of Mars, J. cooperation of the Meteorology Team headed by S. Hess is gratefully Geophys. Res .. 82. 369-375, 1977. acknowledged. This work was supported by NASA contract NASl- Ringwood, A. E .. and S. P. Clark, Internal constitution of Mars, 9703. Contribution 2910 of the Division of Geological and Planetary Nature. 234. 89-92, 1971. Sciences, California Institute of Technology, Pasadena, California Toksoz, M. N .. and D. H. Johnston, The evolution of the moon and 91225. the terrestrial planets, in The Proceedings of the Soviet-American REFERENCES Conference on the Cosmochemistry of the Moon and Planets, in press, 1977. Anderson, D. L., Implications of the inhomogeneous planetary accre­ Toksoz, M. N., A. M. Dainty, S. C. Solomon, and K. R. Anderson, tion hypothesis, Comments Earth Sci. Geophys., 2, 93-98, 1971. Structure of the moon, Rev. Geophys. Space Phys., 12, 539-567, Anderson, D. L., Internal constitution of Mars, J. Geophys. Res., 77, 1974. 789-795, 1972. Urey, H. C., The Planets, Their Origin and Development, Yale Univer­ Anderson, D. L., The moon as a high temperature condensate, Moon, sity Press, New Haven, Conn., 1952. 8. 33-57, 1973. Anderson, D. L., and R. L. Kovach, The composition of the terrestrial planets, Earth Planet. Sci. Leu .. 3, 19-24, 1967. Anderson, D. L., R. L. Kovach, G. Latham, F. Press, M. N. Toksoz, (Received March 30, 1977; and G. Sutton, Seismic investigations: The Viking Mars lander, revised May 9, 1977; Icarus, 16, 205-216, 1972. accepted May 9, 1977.)