JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 98, NO. Ell, PAGES 20,871-20,889, NOVEMBER 25, 1993 An Improved Gravity Model for Mars' Goddard Mars Model 1 D. E. SMITH, 1 F. J. LERCH,1 R. S. NEREM, 1 M. T. ZUBER,1,2 G. B. PATEL,3 S. K. FRICKE,4 AND F. G. LEMOINE5,6 Doppler tracking data of three orbiting spacecraft have been reanalyzed to develop a new gravitational field model for the planet Mars, Goddard Mars Model 1 (GMM-1). This model employs nearly all available data, consistingof approximately 1100 days of S band tracking data collected by NASA's Deep Space Network from the Mariner 9 and Viking 1 and Viking 2 spacecraft, in seven different orbits, between 1971 and 1979. GMM-1 is complete to spherical harmonic degree and order 50, which correspondsto a half-wavelength spatial resolution of 200-300 km where the data permit. GMM-1 represents satellite orbits with considerably better accuracy than previous Mars gravity models and showsgreater resolutionof identifiablegeologicalstructures.The notable improvement in GMM-1 over previous models is a consequenceof several factors: improved computational capabilities, the use of optimum weighting and least squarescollocation solution techniques which stabilized the behavior of the solution at high degree and order, and the use of longer satellite arcs than employed in previous solutions that were made possible by improved force and measurement models. The inclusion of X band tracking data from the 379-km altitude, near-polar orbiting Mars Observer spacecraftshouldprovide a significantimprovement over GMM-1, particularly at high latitudes where current data poorly resolve the gravitational signature of the planet. 1. INTRODUCTION Knowledge of the gravitational field, in combination with surface topography, provides one of the principal means of inferring the internal structure of a planetary body. By removing the gravitational signal of the topography, the distribution of internal density anomalies associated with thermal or compositional differences can be addressed. Gravity can also be used to understand the mechanisms of compensation of surf'acetopography, providing information on the mechanical properties and state of stress of the lithosphere. The earliest global gravitational field models of Mars were derived from Doppler tracking data of the Mariner 9 spacecraft [Lotell et al., 1972, 1973; Born, 1974; Jordan and Lotell, 1975; Reasenberg et al., 1975; Sjogren et al., 1975]. These models provided estimates of low-degree spherical harmonic gravity coefficientsthat yielded information on the oblateness and rotation vector orientation of Mars. Later Balmino et al. [1982] resulted in what was then the highest resolution Martian gravitational model to date: an 18th degree and order field with half-wavelength resolution of approximately 600 km. That field, which is characterized by a spatial resolution comparable to what was then the highest resolution (16 x 16) topographic model [Bills and Ferrari, 1978], was utilized in analyses of the state of stress of the Martian lithosphere and the isostatic compensation of surface topography [Sleep and Phillips, 1979, 1985; Banerdt et al., 1982, 1992; Willemann and Turcotte, 1982; Esposito et al., 1992]. However, the resolution and quality of the current gravity and topographic fields (particularly the latter) are such that the origin and evolution of even the most prominent physiographic features on Mars, namely, the hemispheric dichotomy and Tharsis rise, are not well understood. The resolution of the Balmino et al. [1982] gravity field was limited not by data density, but rather by the computational resources available at the time. Because this restric- tion is no longer a limitation, we have reanalyzed the Viking models that incorporated data from Mariner 9 and the Viking 1 and 2 orbiters [Gapcynski et al., 1977; Christensen and and Mariner data sets and have derived a new gravitational field, designated the Goddard Mars Model 1 (GMM-1). The Balmino, 1979; Christensen and Williams, 1979] resolved higher degree gravity coefficients, showingthe higher power objectives of this study were (1) to develop the best possible in the Martian gravity field compared to Earth's, and the a priori gravitational field model for orbit determination of strong correlation of long wavelength gravity with topogra- the Mars Observer spacecraft in support of the radio science phy. The subsequentinclusionof additional Doppler data by and Mars Observer laser altimeter investigations, (2) to validate analysis techniques to be implemented in Mars •Laboratoryfor TerrestrialPhysics,NASA GoddardSpace Observer gravity modeling studies, and (3) to improve scientific interpretations of geophysical and geological data Flight Center, Greenbelt, Maryland. 2Alsoat Department of Earth and PlanetarySciences, Johns collected in previous missions to Mars. Hopkins University, Baltimore, Maryland. GMM-1 is complete to spherical harmonic degree and 3Hughes STX Corporation, Lanham,Maryland. order 50. The corresponding half-wavelength resolution, 4RMSTechnologies, Inc., Landover,Maryland. 5Colorado Centerfor Astrodynamics Research, Universityof which occurs where the coefficients attain 100% error, is Colorado, Boulder. 200-300 km where the data permit. In contrast to previous 6Now at Laboratoryfor TerrestrialPhysics,NASA Goddard models, GMM-1 was solved to as high a degree and order as Space Flight Center, Greenbelt, Maryland, and Astronomy Pronecessary to nearly exhaust the attenuated gravitational gram, University of Maryland, College Park. signal contained in the tracking data. This was possible Copyright 1993 by the American Geophysical Union. mainly due to the use of optimum weighting and least squares collocation solution techniques [Lerch et al., 1979], Paper number 93JE01839. 0148-0227/93/93 JE-01839505.00 which stabilized the behavior of the solution at high degree 20,871 20,872 SMITH ET AL.' GODDARDMARS MODEL 1 TRACKING DATA • APRIORI MODELS: MEASUREMENT, SOLIDMARS FIXED TIDES, RELATIVITY, /' •._•,.. THIRD BODY EFFECTS GM, DRAG, SOLAR RADIATION, COMPUTE ORBITS f ADJUSTED IGRAVITY, STATE VECTORS ! SELECTNEW I WEIGHTS ' i i G EN ERATE NORMAL EQUATIONS COMBINE NORMAL I EQUATIONSAND / _ _ GRAVITY FIELD COMPUTE SOLUTION ! I _ SOLUTION GMM- 1 RESIDUAL ANALYSIS Fig. 1. and order where correlation ORBIT ERROR PREDICTIONS (DUETO GRAVITYFIELDERRORS) Flow chart of the procedure used to derive GMM-1. and data sensitivities become problematic. As discussedbelow, the extension of the model to high degree and order significantly reduced errors resulting from spectralleakage comingfrom the omitted portion of the gravitational field beyond the limits of the recovered model. GMM-1 has a higher spatial resolution than preliminary versions of this model [Smith et al., 1990a; Zuber et al., 1991] and, in addition, is fully calibrated to give a realistic error estimate from the solution 1.1. ing gravitational model may be input back into GEODYN for residual analyses. 1.2. Representation of the Gravitational Potential The gravitational potential at spacecraftaltitude, V M, is represented in spherical harmonic form as GM covariance. In the following sections we discuss the development of GMM-1 and make a detailed comparison of the field with the previous model of Balmino et al. [1982]. We also include an error analysis and a discussionof the implications of GMM-1 for Martian geophysics and for navigation and precision orbit determination in support of the upcoming Mars Observer ERROR COVARIANCE Mission. General Approach of Gravitational Field Recovery Figure 1 shows a flow chart of the procedure used in the recovery of the gravitational field model. The data were processed using the GEODYN/SOLVE orbit determination/ estimation programs [Putney, 1977]. These programs have previously been used in the derivation of a series of gravitational Goddard Earth Models (GEM) [e.g., Lerch et al., 1979;Marsh et al., 1988, 1990] and have been adaptedfor the analysis of planetary tracking data [Smith et al., 1990b; Nerem et al., 1993]. GEODYN provides orbit determination and geodetic parameter estimation capabilities, and numerically integrates the spacecraftCartesian state and the force model partial derivatives employing a high-order Cowell predictor-corrector method. The force modeling includes a spherical harmonic representation of the gravity field, as well as a point mass representation for the Sun, Earth, Moon, and the other planets. Atmospheric drag, solar radiation pressure, measurementand timing biases, and tracking station coordinates can also be estimated. The least squares normal equationsformed within GEODYN may be output to a file for inclusion in error analysesand parameter estimations. The SOLVE program then selectively combines the normal equationsformed by GEODYN to generate solutions for the gravity field and other model parameters. The result- VM = • + - • r • /=2 m=0 film(Sin c•)(Clm COS mA q- S!m sin mA) (1) where r is the radial distance from the center of mass of Mars to the spacecraft, •b and A are the areocentric latitude and longitude of the spacecraft, r M is the equatorial radius of Mars, GM is the product of the universal constant of gravitation and the mass of Mars, P!m are the normalized associatedLegendrefunctionsof degreeI and order m, C!m and S !m are the normalized spherical harmonic coefficients which were estimated from the tracking observations to determine the gravitational model, and N is the maximum degree representing the size (or resolution) of the field. The gravitationalforce due to Mars which acts on the spacecraft correspondsto the gradient of the potential, V M. 2. REFERENCE SYSTEMS FOR DEEP SPACE NETWORK TRACKING OF MARS SPACECRAFT Spacecraft orbiting Mars are tracked from Earth through the NASA deep space network (DSN) tracking stations at Goldstone (California), Canberra (Australia) and Madrid (Spain). The adopted planetary ephemeris for Earth, Mars, and the other planets was the JPL DE-96 system [Standish et al., 1976]. The inertial coordinate system for the Earth is defined by the direction of the Earth' s rotation axis and the location of the vernal equinox. The orientation of the Mars rotation axis is specified by the right ascension and declination of the Martian pole, as given in Davies et al. [1986, 1992]. The z axis of the Mars inertial coordinate system is the instantaneous Mars rotation axis. The direction of the x axis (the SMITH ET AL ßGODDARDMARS MODEL 1 TABLE 1. Planetary and Astronomical Constants Parameter Value GravitationalconstantGM 42828.28 Equatorial radius of Mars 3394.2 Spinrateof Mars Solid tide amplitude k2 Flattening of Mars Unit km3 s-2 km 350.891983 degd-1 0.05 1/191.1372 Speedof light 2.99792458 x 108 Astronomicalunit m s-1 1.49597870660x 10TM m International Astronomical Union (IAU) vector) is defined to be the intersection of the instantaneousMars equator with the mean Earth equator of the appropriate ephemerisepoch. For this analysis, 1950.0 was chosen as the epoch, and thus the IAU vector was defined with reference to the Earth mean equator and equinox of 1950.0(EME50). The prime meridian of Mars is defined by Davies et al. [1989]. At the beginning of this analysis, tests were performed using a reference date and planetary ephemeris of 2000.0, but results showed that the 1971 Mariner 9 data and 1976-1978 Viking data were better satisfied with a reference date of 1950.0. Various reference system constants that were used are given in Table 1. 3. DATA SUMMARY AND ORBITAL SENSITIVITY OF GRAVITY 3.1. Satellite Orbital Characteristics The Mariner 9 (M9), Viking Orbiter 1 (VOl), and Viking Orbiter 2 (VO2) spacecraft were in highly eccentric orbits with periods of approximately 1 day for VOl and VO2 and one-half day for M9. The satellite orbit characteristicsare summarized in Table 2. The orbital periods are nearly commensurate with the rotational period of Mars (24.623 hours) which produce dominant resonant perturbations [Kaula, 1966] for all orders rn of the Viking spacecraft (24-hour period) and for the even orders of M9 (12-hour period). The resonant periods range mostly from about 1 to 50 days for shallow resonant terms and also include deep (very long) resonant periods. The beat period, or fundamental resonant period, identifies the shift (or "walk") in successive ground tracks and is useful in mapping the orbital coverage over Mars (a plus sign represents an eastward "walk" and a negative sign for a westward "walk"). The beat period changes after each maneuver of the VOl and VO2 [Snyder, 1979]. For example, an orbit maneuver by VO2 on March 2, 1977, produced a very slow walk to synchronize with the Viking Lander (VL2). Subsequently, on April 18, 1977, another maneuver resulted in a walk around the planet in 13 revolutions, producing a beat period of 12 days. Strong resonant perturbations of long period were produced on VOl, commencingon December 2, 1978, to provide a slow walk around the planet. Mission events such as leakages, attitude control jetting, and other phenomena that cause variations in the orbital periods are described by Snyder [1977, 1979]. The 300-km periapsis altitude orbits of VOl and VO2 provide the strongestcontributionof data to the solutionfor the higher degree terms, particularly, the VO 1 low orbit with a range of 180øfor the argument of periapsis (co)as compared to 25ø for the VO2 low orbit. The observing period for the 20,873 VO 1 low orbit shown in Table 2 covers almost 2 years from March 12, 1977, to January 27, 1979, and the periapsis point varies in latitude from + 39ø to -39 ø during this period. The VO 1 low orbit provides about a 9ø ground track "walk" per revolution for the 1.5-year period from July 1, 1977 to December 2, 1978. This correspondsto a "near repeat" of the ground track for a 39-day period. After the 39-day near-repeat period, the orbital ground track shifts by 1.8ø from the previous repeat track, which correspondsto a deep orbital resonance with a period of about 200 days. This produces, over a 200-day coverage, a global grid (_+39ø latitude) with approximately 1.8ø ground track separations and provides for a high-resolutionrecovery of the gravity field. 3.2. S Band Doppler Tracking Data Used in the Solution 3.2.1. Data summary. The data set consisted of 265 orbital arcs, representingover 1100 days of S band Doppler tracking data from the M9, VOl, and VO2 collected by the DSN between 1971 and 1978. These data, grouped by satelliteperiapsisaltitude and inclination, are summarizedin Table 3. In total, over 230,000 observations were included in the GMM-1 3.2.2. solution. Data characteristics. The data consist of two- way S band (2.2 GHz) Doppler measurementscompressed to 60 s (1-min data points). Data far removed from periapsis (approximately greater than 12,000-km altitude) were compressed over 10-min intervals. All observations were collected by three DSN sites located at Goldstone, Madrid, and Canberra. They were processed in the differenced-range Doppler formulation, taking into account relativistic bending due to the Sun [Moyer, 1971]. Observationsnear satellite periapsisare most valuable for determining the gravity field, and periapsis is generally observableby at least one of the DSN sites except when occulted by Mars. The data distribution and coverage per satellite orbit are discussedin section 3.4. The signal is significantly degraded in precision during solar conjunction due to the solar plasma effects when its path comes within 5ø of the Sun from Mars. This occurred for a period of about 1.5 months beginning November 7, 1976, for seven VOl and VO2 arcs. The data were downweighted in the solution for this period. 3.3. Spectral Sensitivity of Gravity Signal The spectral sensitivity of the gravity field is analyzed for the M9, VO 1, and VO2 orbits. Sensitivity for the high-degree terms (>30) is the main area of interest, and these are comparedwith a thresholdlevel correspondingto the precision of the DSN signal. The signal, when compressed to 1-mindatapoints,hasa precision of 1 mms-•, andapproximately0.3 mm s-• for 10-mindatapoints.A sensitivity study for the above Mars orbiters has been made by Rosborough and Lemoine [1991] and Lemoine [1992] for terms through degree 20. Analysis for the high-degree terms is discussedin detail by Lerch et al. [1993]. A brief summaryis given here. The orbit perturbations were studied using linear perturbation theory and through numerical integration by GEODYN. The gravity signalfor sensitivity analysis employed a formof Kaula'srule, 13 x 10-5//2, for termsof degreel, 20,874 SMITHETAL.:GODDARD MARSMODELI TABLE 2. SatelliteOrbit Characteristics IncludingBeat Periodsand GroundTrack Walks Periapsis Periapsis Altitude, km Epoch* IncliOrbit Nodal Beat Walkper nation, Eccen- Period, Argument, Rate Rate Period, Revolution, deg tricity hours deg deg/d deg/d days deg Mariner 1500 Nov. 16, 1971 9 64 0.62 11.81 -24 -0.02 -0.18 18.3 64 0.60 11.98 -24 -0.01 -0.16 19.5 New orbit Trim maneuver End of data Dec. 31, 1971 April 19, 1972 Comments Viking 1 1500 June21, 1976 Sept.13,1976 Sept.24, 1976 Jan.22, 1977 March12, 1977 March24, 1977 July1, 1977 Dec. 2, 1978 300 38 0.75 39 0.80 24.63 21.87 24.63 22.99 21.92 23.50 23.97 24.85 Jan. 27, 1979 47 56 60 79 99 102 120 264 0.17 -0.13 0.27 -0.21 >1000 8 > 1000 14 8 21 38 129 <1 44 <1 25 43 17 10$ 3 10 > 1000 13 15 35 <1 29 25 270 Synchronous 9 rev/walk Over Lander Near Phobos Low periapsis New walk Dual station Near synchronous End of data Viking 2 1400 Aug.7, 1976 Aug.28, 1976 55 0.76 27.31 24.63 72 73 0.05 -0.09 1500 800 Oct. 2, 1976 Dec. 21, 1976 75 80 0.80 0.80 26.79 26.48 68 62 750 March5, 1977 24.73 55 215 2 700 April 18, 1977 Sept.26, 1977 22.72 24.29 51 34 12 78 29 5 600 300 Oct. 9, 1977 Oct. 25, 1977 July 25, 1978 24.20 23.98 33 32 2 60 39 6 9 -0.05 -0.08 -0.04 -0.03 -0.10 -0.04 New orbit Synchronous New inclination Low periapsis/ inclination change Slow walk 13 rev/walk Near Deimos Near Deimos Low periapsis End of data *Start epochof the orbit parameterscited. ?Walk for two 12-hour revolutions for Mariner 9 orbits. $Secondary walk is about2øover approximately a 200-dayduration. to Table 2, for the VOl orbit from July 1, 1977, through December2, 1978, the near repeat of the groundtrace (to paredwith the noiseof the DSN Dopplerdata. Both the within 1.8ø) after 38 revolutionsproducesa perturbationat analyticaland numericalstudiesconfirmthe importanceof order 38 with a period of about 200 days. The analyticalvelocityspectrumby degreeis presentedin the resonanceperturbationsin determiningthe satellite Table 4 for both M9 and the 300-, 800-, and 1500-km VOl sensitivityto the Mars gravity field. The resonanceson the Viking spacecraftfall into three andVO2 orbits.The analyticalvelocity spectrumis obtained classes: (1) resonancesat low orders (characterizedby by computingthe Kepler element perturbationsusing periodsof up to 40 days),(2) long-period resonances (periods Kaula's [1966] theory, and then mappingthese to velocity greater than 50 days) at specifichigher orders, and (3) space.Sincewe are interestedin the satellitesensitivityto intermediateresonancesat the other orders.The long-period the gravityfield over the periodsof the arc lengthsof data resonancesresultfrom a nearrepeatof the groundtraceafter used in the GMM-1 solution, perturbationswith periods an integernumberof spacecraftrevolutions.Thus,referring greater than 40 days were excluded.In addition,those which was obtainedby Balmino et al. [1982]for the power spectrumof Mars. The velocity perturbationswere com- TABLE 3. Summaryof DSN TrackingData Used in GMM-I Arcs Satellite Viking 1 Viking I set 2 Viking 2 Mariner Total 9 Average Average Input rms Altitude, Inclination, Arc Length, Residuals, km deg Total days cm/s 1500 300 1400 1500 38.2 39.1 55.4 75.1 80.1 80.2 64.4 29 95 12 11 54 37 32 4.2 4.8 3.8 4.7 3.8 4.2 4.3 0.446 2.844 0.256 0.507 0.210 6.282 0.212 778 300 1500 270 Average Observations 1082 673 990 952 655 793 1559 Total Observations Total Days 31,393 63,977 11,878 10,467 35,375 29,355 49,878 122 425 46 52 204 155 138 232,323 1142 SMITH ET AL..' GODDARDMARS MODEL I TABLE 4. 20,875 Theoretical Harmonic Analysis of Velocity Perturbations by Degree Altitude, km VOl 300 VO2 300 VO2 800 VOl 1500 M9 1500 Epoch Jan. 15, 1978 Dec. 17, 1977 Apdl 19, 197 Feb. 5, 1977 Apdl 10, 1992 144.452 39.761 16.519 7.771 4.248 2.498 1.617 1.197 1.118 1.523 2.686 1.195 1.350 0.211 0.048 0.012 0.004 0.000 0.000 0.000 0.000 0.000 0.000 0.000 64.324 36.417 13.981 5.659 2.270 1.051 0.579 0.304 0.158 0.093 0.056 0.031 0.019 0.010 0.002 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 828.128 100.049 55.852 24.812 12.463 6.003 2.669 1.135 0.490 0.260 0.167 0.105 0.061 0.002 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 2 3 4 5 6 7 8 9 10 11 12 13 14 18 22 26 30 34 36 38 40 42 46 50 566.175 353.495 207.188 107.358 66.205 40.783 28.610 18.054 13.176 9.049 6.808 4.711 3.858 1.390 0.581 0.317 0.260 0.370 0.507 0.468 0.373 0.241 0.265 0.322 653.411 218.135 106.629 59.610 34.485 21.119 13.923 9.347 6.677 4.736 3.555 2.649 2.053 0.780 0.381 0.249 0.110 0.047 0.034 0.028 0.029 0.036 0.038 0.027 A powerruleof 13 x 10-5 l 2 is used.DataarefromMarine9 andViking1 and2 sampled orbits. Periods >8 days wave their amplitudes multiplied by 8/period. Periods >40 days have been excluded. Values are in units of centimeters per second. perturbationswith periods between 8 and 40 days have been prorated to 8 days by the factor 8/period. The VOl 300-km orbithasa sensitivity in excessof 1 mms-• (theaccuracy of the S band data) out to degree 50. In contrast, the VO2 300-km orbit is sensitiveonly to terms out to approximately degree 30. As the periapsisaltitude is raised, the sensitivity in degree is diminished. The limit is degree 18 for the VO2 800-km orbit, and degree 11 for the VO 1 1500-km orbit. The Iooo M9 orbit has stronger perturbations than the VOl 1500-km orbit by virtue of its closeraveragedistanceto Mars, with its twice per day revolution and smaller orbital eccentricity. The sensitivity was also evaluated through numerical integrationusingthe GEODYN programfor the VO 1 300-km orbit. The spectral rms velocity perturbation by order is shown in Figure 2 for different arc lengths. The results show Gravity Perturbations sensitivity greaterthan 1 mm s-• for the highdegreeand by Order for V•']qh• Arc Lengths Using Power Rule 13x.10-'/[l_' _-_ . Vtktni-I 300 • Periapsis Orbit (epoch 75-01-15) • lo0 0.001 0 10 20 30 40 50 60 order Fig. 2. Spectral sensitivity of gravity signal (by order). GEODYN, numerically integrated perturbations' HAP, harmonic analysis of perturbations from analytical theory; rss, root sum square. 20,876 SMITHETAL.' GODDARD MARSMODEL1 (•0=269') (•0=175') Velocity RSS' Perturbations 392339 124383 28804 8027 2909 1297 675 400 268 229 203 187 179 186 199 VKGI: 2 *** 6 ****** 10 14 18 22 26 30 34 36 38 40 42 46 50 * RSS taken Velocity In .001 cm/sec DEG RSS' 78-01-15 Arg. of periapsis 8 day arc length 469****** 54'****'156 6 52***843 28 3 10138592237 8 1 2 7161289 76 1 1 2 20137143 0 0 1 2 33104 0 0 0 1 13 69 0 0 0 0 4 39 0 0 0 0 1 20 0 0 0 0 0 9 0 0 0 0 0 1 0 0 0 0 0 0 2 6 10 over all 14 18 orders, 22 3 29 77 86 79 63 43 14 3 2 13 29 48 62 67 49 22 1 2 1 9 2 1 22 7 1 0 41 19 5 1 0 73 59 29 9 1 69 84 67 34 11 26 30 34 not just 36 581650 72711 = 175 deg 38 40 42 0 0 46 0 50 sampled orders. 22705 9766 4959 2771 1631 984 604 483 378 316 244 162 110 Perturbations In .001 cm/sec DEG VKGI: 2 *** 6 ****** 10 14 18 22 26 30 34 36 38 40 42 46 50 ***494336 ******627 ******274 762***554 148420411 120 47192 176108 38 133175148 157131 40 33 63 81 112106 60 67 69 52 32 37 35 ORD: 2 78-12-20 Arg. of periapsis 8 day arc length 6 10 * RSS taken over all 3 86 37 62 91 65 45 24 46 6 18 19 1 6 11 2 13 5 15 6 10 2 3 14 18 0 1 2 1 2 1 2 2 2 1 22 orders, 0 0 1 0 0 0 0 0 0 26 0 0 0 0 0 0 0 0 30 not just 0 0 0 0 0 0 0 34 - 269 deg 0 0 0 0 0 0 36 0 0 0 0 0 38 0 0 0 0 40 0 0 0 42 0 0 46 0 50 sampled orders. Fig. 3. Signalsensitivityby degreeandorderfor VOl low orbit. order termsfor arc lengthsgreaterthan 3 days. For arcsof 3 days the limit in sensitivity is approximatelyorder 30, whereas for the 1-day arcs the limit in sensitivityis approximately order 20. The increasein sensitivityresultsfrom the samplingof the medium-periodresonanceperturbations. Although these results suggestit would be beneficial to processthe VOl 300-kmdata in batchesof 8-16 days, this was not possiblebecauseof insufficienttracking coverage and errors in the nonconservative force models. separationbetweengroundtracksfor the orbitaldata setsis indicatedin the plates by the term "walk." Also the data spanis givenalongwith the walk to depictthe extentof the coverageoverall dataof thistype asoriginallygivenin Table 2. In theseplateswe can seethe extent of periapsiscoverage of the VOl low orbit (---39ø latitude), which is well complemented in the northern hemisphereby the VO2 low orbit. Theseplatesshowreasonablygoodglobaldatacoveragefor the VOl low orbit, VO2 low orbit, VO2 800-km orbit, and alsofor M9. The globalcoverageprovidesfor good separa- For the highly eccentric Viking orbits the sensitivityof sphericalharmonic coefficientsdependsnot only on the bility of the lower degreeterms of the gravity field and periapsisaltitude,but alsoon the locationof the argumentof possibly out through degree 30, consideringthe strong periapsis.The GEODYN spectrumof rms velocity pertur- sensitivities to these terms. Plate 2 shows the combined bationssampledby degreeand order are givenfor an 8-day coverageof the low orbits of VOl and VO2 from observaarc for two VOl 300-km orbits in 1978. In the first arc, beginningJanuary 15, 1978, periapsisis located near the equator (to --- 175ø). In the second test arc, beginning December 20, 1978, the periapsisis located near 39øS(to •269ø).When tois near 180ø, the orbit tendsto be sensitiveto termsof highdegreeand highorder, whereaswhen tois near 270ø, the orbit is sensitiveto terms of high degreeand low order (see Figure 3). tions with altitudes less than 5000 km and coverage by 2- 1.0 -- Another importantcharacteristicof theseeccentricorbits is that significantsensitivityto the high-degreeterms exists over a broad rangeof altitudes.As a demonstration,for the VOl arc described in Figures 2 and 3 (epoch January 15, 1978),the perturbationsfor termsof order25 (degrees31-50) are shown over a revolution in Figure 4. For these terms, significantsensitivityis apparentup to an altitudeof 10,000 km, covering half of an orbit revolution. In summary, the high degree sensitivity of the Viking Orbiter tracking data to the gravity field of Mars is deter- 2 - 0.1 -- minedby the periapsisaltitude,locationof the argumentof periapsis,and the length of the arcs used to processthe tracking data. 40000 3.4. Distribution of Observational Coverage 20000 0 20000 40000 altitude (kin) The groundtrack for each of the Doppler observationsis Fig. 4. Velocityperturbationfor VOl low orbit due to 25thplotted in Plates l a and lb for a completeset of ground orderharmoniccoefficientsfor degrees31-50 (curvea) and degrees tracks coveringall major data setsusedin the solution.The 41-50 (curve b). , V 0 II mm • • Id F,. Z 0 ß II • I I I I I I i I ß A ? II ! SMITHET AL.: GODDARDMARSMODEL 1 20,879 o o o v ß + '.' ! i I 20,880 SMITH ET AL.: GODDARDMARS MODEL 1 different levels of altitude over 300 km. This low-altitude data coverage of the observation points along the ground tracks for the VO 1 low orbit is seento be complementedby the VO2 low orbit, particularly in the northern hemisphere. However, the lack of complete data coverage near periapsis indicates that separability will not be complete for the high-degreeterms (30-50), as noted above in the argumentof periapsis coverage for VO2. Nevertheless, the result from Plate 2 indicates that there is great sensitivity to the higher degree terms for altitudes less than 2000 km for the low-altitude Viking orbits. Hence we may expect that the ground track coveragefor the combined VOl and VO2 low orbits, particularly, for the observed coverage with altitude less than 500 km over a wide area, will provide for good resolution of localized geophysicalfeatures in the vicinity of these ground tracks. 4. 4.1. MODELING Physical Model Because the Viking and Mariner data do not provide uniform spatial coverageof Mars, the applicationof a priori constraintswas critical to the development of a high degree and order solution. The Viking and Mariner data were initially processedusing the gravity model of Balmino et al. [1982]. However, in the final iteration to produce GMM-1, an intermediate solution, MGM-635, was used as the a priori 4.2. Method of Solution 4.2.1. Least squares with a priori constraints. The methodof solutionis a modifiedleast squaresprocess[Lerch et al., 1979; Schwartz, 1976, 1978] which minimizes the sum Q of signal and noise as follows: + E •fk Q= EC;m +2•;m •t ri2• l,m irl irk (2) where the signal is given by Elm, Slm, which are the normalized spherical harmonics comprising the solution coefficients. Theparameter irl = 13 x 10-5/I2 isthermsof the coefficients of degree l (a priori power rule) and is introduced to permit solutions to degree and order 50. This expression, which is based upon Kaula's [1966] rule, has been obtained by Balrnino et al. [1982] and representsthe power in that gravity model. The noise given by r ik is the observationresidual (observed-computed)for the ith observation of satellitetracking data set type k, irk is the rms of observationresidualsof data type k (generally significantly greater than the a priori data precision), and fk is a downweightingfactor to compensatefor unmodeled error effects for each data type k (ideally, fk = 1 for pure noise). The optimum weighting method estimates the combined weights directly, namely, wk = model. The gravitational effects of the Martian solid body tide obsi 2 irk (3) were includedin the satelliteforce model, and a value of k2 = 0.05 was adopted [Christensenand Balmino, 1979]. The effects of atmospheric drag were incorporated into the When minimizing Q above, using the least squaresmethod, the normal matrix equation and error covariance are ob- satellite force model using a spherical model for the satellite body and the atmosphericdensity model developedby Culp and Stewart [1984]. The coefficientof drag CD was adjusted once per data arc, except for the VO 1 and VO2 low-periapsis orbital arcs, where CD was adjustedonce per day. Solar radiation forces were calculatedusinga sphericalmodel for the spacecraft body and adjusting a reflectivity coefficient, CR, once per data arc. The solar flux at Mars at a given time was scaledfrom the N• = R Earth value for that date to the actual distance of the tained as follows: (4) where • is the solution, N is the normal matrix, R is the vector of residuals, and V = N-1 N = E wkNk (5) is the approximate form for the variance-covariance error matrix which must be calibrated by adjusting the weights. Nk is the contribution for each satellite data set k to the normals, where k = 0 corresponds to the normal equations for the satellitea priori coefficientconstraintsfor which No is the matrix of Kaula constraintsand the weight w0 is fixed at unity for the constraints. The process of minimizing both signal (by application of the Kaula power rule constraints) plus noise in (2) is also spacecraftfrom the Sun. One range rate bias was estimated for each tracking station per arc. The measurementswere corrected for tropospheric effects using the Hopfield [1971] model. The tracking data records did not contain meteorological data or troposphericcorrections;thus the corrections were computed assumingstandard temperature, humidity, known as collocation [Moritz, 1978]. The constraints bias the and pressure, scaled to reflect the station height above sea coefficientstoward zero, where they are poorly observed. level. With the conventional least squares approach (noise-only Third-body gravitational perturbationson the spacecraft minimization) there is a problem of separability due to the were computed from the point massgravitationalforces due strong correlation between many of the high-degreecoeffito the Sun, the Earth-Moon system, the other planets, and cients. The absence of collocation (w0 = 0 in (5) for Phobos, one of Mars' natural satellites. In addition, GEO- GMM-1) results in excessivelylarge power in the adjustment DYN was modified to read an ephemerisfor Phobosand to of the potential coefficientsas in Figure 5. Hence we see the add the point mass gravitational acceleration due to Phobos benefit of the constraints which permit resolution of the to the total spacecraftacceleration. The ephemerisof Pho- high-degree terms wherever the data permit and provide bos was prepared at Goddard Space Flight Center (GSFC) control of aliasing in the solution. Biasing of the coefficients by processingoptical measurementsobtainedfrom the M9, is unimportant in those areas where data density is high. VOl, and VO2 images of Phobos [Duxbury and Callahan, 4.2.2. Data weighting and error calibration. The 1988, 1989]. weightingtechniqueand error calibration [Lerch et al., 1988; SMITH ET AL.' GODDARDMARS MODEL 1 20,881 (10) represent extremes in estimation: complete independence and complete dependence. • GtfIK-1 without Belmino Constraint Truncated to 35x35 lOOO 5. 5.1. • loo RESULTS OF THE GMM-1 SOLUTION Description of Solution GMM-1 is a 50 x 50 spherical harmonic gravity model. There are a total of 5250 estimatedparameters:2597 gravity coefficientsplus GM, and approximately 2650 orbit and spacecraft parameters. The GMM-1 gravity coefficients throughdegreeand order 18 are shown in Table 5. The full lO set of coefficients can be obtained from Smith et al. [ 1993] or fo........ Fig. 5. 'io........ Harmonic •'o'....... Degree directly from the authors. Calibrated accuracy estimatesof ¾o ........ the coefficients Gravity model uncertainties. 5.2. have also been obtained for the model. Gravity Model Tests Using Orbital Observation Residuals Orbital arcs have been selectedfrom the seven major data Lerch, 1991] of the solution (2)-(5) is based upon subset solutions.The subsetsolution(Ck) is formed by deletinga major data set k from the completesolution(C). The weight wk is adjustedas in (8) below by requiringthat sets summarized in Table 3 and used to test the orbital accuracy of the model by fitting the DSN Doppler data. Observationresidualshave been computed from our 50 x 50 model and compared with the prior best available 18 x 18 gravity model of Balmino et al. [1982]. Table 6 is a compi- lation of orbit tests for 14 arcs. Each arc is fit using the Balmino et al. field and GMM-1. For all 14 test arcs, the rms where K• is an error calibrationfactor which ideally should residual fits are significantly smaller, sometimes 5-10 times equal to unity' smaller, when computedusing GMM-1 than when using the (6) Balmino l acll -- (c- }1/2 2 (7) 5] et al. field. Table 7 summarizes the results of orbit predictiontestsfor a subsetof the arcsin Table 6. The orbits obtainedin the orbit accuracytests are projected forward in time 2 or 3 days. Then rms residualfits are comparedfor the data in the predicted time periods. Again, the fits are significantly smaller when computed using GMM-1 than when usingthe Balmino et al. field. The improvement in the andwhererr• andrr2 arethevariance of thesubset andthe fits is not entirely due to the increasedresolution of GMM-1. complete solutions,respectively. The sum in (7) is over all the coefficients,and the scale factor K k is needed for the errors, since the error covariance in (5) is only an approxi- An 18 x 18 version of GMM-1 outperformed the 18 x 18 model of Balmino et al. in all cases except the 300-km VOl and VO2 orbits, for which the performance was comparable. mation [Lerch, 1991]. The new weights w•: shouldbe adjustedso that each K• convergesto 1 for all k, and the new weights are computed from Wk w•,= 2 K• (8) 5.3. Analysis of the Gravity Coefficients Figure 6 is a plot of the degreevariance of the coefficients (power) and error variance of GMM-1 per degree. Also plotted are the power of the 18 x 18 gravity field [Balmino et al., 1982]and a powerrule (13 x 10-5//2) takenfrom Balmino et al. [1982], which is the basis of the constraint (10) matrix usedin GMM-1. The plot showsthat for degreesless than 15, the power spectrumof GMM-1 and the Balmino et al. field are about the same. However, above degree 15, GMM-1 drops below, while the Balmino et al. field rises above the power spectrumof Balmino's rule. The upward turn of the Balmino et al. field is undoubtedly due to aliasing. Aliasingadverselyaffectsthe performanceof a gravity field with respect to orbit fits from independentdata, orbit predictions, and other geophysical information which are derived from the gravity coefficients.The truncation level of GMM-1 at degree 50 is high enough that the high-degree gravity signal is not significantly aliased into the lower as indicated by (7). This meansthat in our case the covariance between the square of the difference of the two estimates of the coefficients is equal to the difference of the variancesof the subsetand complete solutions.Thus (9) and The power spectrumof GMM-1 dropsbelow the valuesof the power rule for high degrees.Above degree22, the errors of the coefficientsare larger than the coefficientsthemselves. This drop-off in the power spectrumoccursbecausethe drag The processis iterated by forming a new completesolution and subsetsolutionsfrom the new weights, and this process may continue until the weights converge. In a case where two solutionsare based upon independent data, then (in the above notation) for a single coefficient parameter the two estimatesgive E(C- C)2= •2 + rr2 (9) whereas in (2) the data for the subset solution is wholly embeddedin the complete solution in which case E(C- C)2= •2_ or2 degree terms. 20,882 SMITH ET AL.: GODDARDMARS MODEL 1 TABLE 5. GMM-1Normalized CoefiScients Through DegreeandOrder18in Unitsof 10-lø Index Index n m Value 2 8 14 20 26 32 38 44 50 0 0 0 0 0 0 0 0 0 -8759770 1075 1627 932 -417 -303 - 103 -2 15 2 3 4 5 5 6 6 7 7 8 8 9 9 9 10 10 10 11 11 11 12 12 12 13 13 13 13 14 14 14 14 15 15 15 15 16 16 16 16 17 17 17 17 17 18 18 18 18 18 1 1 1 1 5 1 5 1 5 1 5 1 5 9 1 5 9 1 5 9 1 5 9 1 5 9 13 1 5 9 13 1 5 9 13 1 5 9 13 1 5 9 13 17 1 5 9 13 17 36 37473 42612 6140 -44655 18281 16382 13088 -1266 -178 -29578 2743 -20839 -11064 11266 2582 -15370 -11144 14309 -3755 -11655 6147 6846 -1541 15 9134 3215 3403 -1812 3125 9560 4884 -10076 -2844 -780 -1952 -4603 -3031 -47 -1891 5100 -2758 -2304 1294 928 2770 -659 -2052 -2537 Index Index n rn Value n rn 3 9 15 21 27 33 39 45 0 0 0 0 0 0 0 0 -119062 -1924 3462 -1254 -572 -57 44 32 4 10 16 22 28 34 40 46 0 0 0 0 0 0 0 0 GMM-1 65 252926 37090 20365 37260 -14489 16726 -1926 -12343 5158 -17748 -439 -14007 -6448 -4987 -12538 -15510 3564 11617 -3912 -4769 3962 3922 5749 -1555 9571 9523 4961 -8041 8702 21239 -1130 -3830 -1616 7182 -1748 3784 -9907 -5937 -6095 3698 -1126 -5787 3664 -1651 -2105 2574 -2866 -10316 2 3 4 5 2 2 2 2 6 6 7 7 8 8 9 9 2 6 2 6 2 6 2 6 10 10 10 11 11 11 12 12 12 13 13 13 2 6 10 2 6 10 2 6 10 2 6 10 14 14 14 14 15 15 15 15 16 16 16 16 17 17 17 17 2 6 10 14 2 6 10 14 2 6 10 14 2 6 10 14 18 18 18 18 18 2 6 10 14 18 Value n Index Index rn Value n rn Value n m Value 0 0 0 0 0 0 0 0 -17635 -4064 1112 395 -73 -129 -50 -5 6 12 18 24 30 36 42 48 0 0 0 0 0 0 0 0 13340 1038 -4292 -927 -23 129 66 13 7 13 19 25 31 37 43 49 0 0 0 0 0 0 0 0 13025 -7098 -367 603 289 72 -10 -17 GMM-1 Zonals 51491 5 7098 11 2885 17 1274 23 611 29 107 35 -34 41 -36 47 Tesserals and Sectorials - 846829 490611 -159844 - 10546 -41571 83160 -89776 -13689 3 4 5 3 3 3 351325 64742 33602 255554 - 1728 3147 4 5 4 4 2973 -45993 8244 16668 6 3 8376 2989 6 4 9447 27894 8686 -128554 -35011 26941 29707 -5593 7 3 11115 -3998 7 4 26621 -4873 -5563 15193 -9586 12058 8637 -19244 6245 -18195 1543 4988 7 8 8 9 9 7 3 7 3 7 4224 -14284 -5040 -8436 -4347 -17858 -14281 16982 -9009 8088 8 8 9 9 4 8 4 8 10334 -3135 9281 12605 741 -2319 15600 -1645 475 6192 -9821 2122 10 10 3 7 -7646 2874 3178 -5952 10 10 4 8 -18250 4783 -214 7701 3 7 8530 -7097 11 11 4 8 -10855 -9343 -4991 7947 11 3 7 11 -8925 7927 -513 -14660 1595 7044 1764 -5332 8870 -15713 3878 5169 -8478 12 12 12 13 13 13 4 8 12 4 8 12 -5403 -16783 -11 5277 -1440 -15838 1344 -6155 -1882 9094 2831 -3279 -3370 8017 -2557 -1785 3840 -29 -3514 5315 -469 -196 403 -11720 -228 19515 8259 -16113 12839 2873 -7846 -7620 11 11 11 12 12 12 13 13 13 8078 -507 -2433 -5192 -797 -13363 14 14 14 3 7 11 5749 -6641 -8531 1038 1808 1998 14 14 14 4 8 12 713 8070 -2997 -9173 1451 -4818 15 15 15 3 7 11 -3327 8499 -7197 -2122 3714 6605 15 15 15 4 8 12 -6995 8817 9103 -8219 7175 8887 1212 -2598 -1913 -127 -267 11 3 7 -2883 1901 -5631 -8491 -5447 4891 -4057 15 15 -4849 -4860 3250 -5903 -74 1094 5287 -3407 -3784 -1966 862 3735 -9887 3650 -3206 3532 -4601 16 16 16 16 17 17 17 17 3 7 11 15 3 7 11 15 -2155 4516 -780 -4941 -821 2686 4904 -5352 3452 2623 -6243 1661 2035 -1516 2276 2974 16 16 16 16 17 17 17 17 4 8 12 16 4 8 12 16 -5471 2082 3041 1691 2372 -5777 58 10094 1725 -280 5799 1144 1746 -4332 593 5379 -2322 1470 4920 -4524 3176 -4761 663 -135 18 18 18 18 3 7 11 15 823 -3103 2969 2471 -2615 1644 -2634 8800 18 18 18 18 4 8 12 16 5812 1927 1647 6505 1093 -288 -1097 3824 4125 -11150 2455 3926 parameters (once per day values) are absorbingpart of the gravity signal. Also, the high degree terms are highly correlated, and hence the effect of the power rule constraintin the solution is quite strong, which further explains the small power for these terms. However, because the a priori constraint does not have a major effect on the solution, the terms do contain information on the short wavelength gravity field in the vicinity of the spacecraftperiapses.While the power in the field falls below that predicted by the power rule at high degrees,the field is a better representationof the true gravitational signature of the planet at those wavelengths than would be the case if the field were solved to SMITH ET AL.' GODDARDMARS MODEL 1 TABLE 6. 20,883 Orbit Accuracy Tests: rms of Orbital Fits Arc Arc Satellite Inclination, Arc deg Epoch No. of Observations Balmino GMM-1 Length 18 x 18, 50 x 50, days cm/s cm/s 1 M9 64 Jan. 15, 1972 1896 4 0.456 0.090 2 3 VOl 1500km VO2 1500km 39 55 Aug. 22, 1976 Sept. 17, 1976 1326 1511 6 6 0.687 0.387 0.097 0.196 4 5 VO2 1500 km VO2 800 km 75 80 Oct. 26, 1976 Jan. 2, 1977 1350 682 6 4 0.649 0.434 0.340 0.143 0.522 0.173 6 7 8 VOl 300 km VOl 300 km VOl 300 km 39 39 39 Nov. 22, 1977 Feb. 10, 1978 June 4, 1978 568 754 538 9 9 2 5.07 6.44 2.42 1.04 1.22 0.74 9 10 VOl 300 km VOl 300 km 39 39 Aug. 11, 1978 Sept. 4, 1978 387 1025 2 8 1.43 8.58 0.09 1.24 4.79 0.87 VO2 VO2 VO2 VO2 80 80 80 80 Nov. 17, 1977 Dec. 17, 1977 May 16, 1978 May 26, 1978 1114 688 791 705 6 2 8 4 1.52 0.73 7.67 0.60 1.02 0.11 1.78 Average Average 11 12 13 14 300 km 300 km 300 km 300 km Average 0.15 0.77 2.63 For arcs 1-5 the arc parametersadjustedare position,velocity,Cr, and stationbiases.For arcs6-14 the arc parametersadjustedare position,velocity, Cr, and Ca per arc. Arc 13 comesin as two separatearcsin GMM-1. lower degreeand order and all of the high degreeand order coefficients were constrained error of 1 cm s-1 while the VOl 300-kmdata groupis assigned anerrorof 0.71cms-1 (wt = 1/(0.71) 2 = 2). The to zero. Figure 7 shows the coefficient differences between GMM- 1 and the 18 x 18 gravity field of Balmino et al. While the rms differencesper degree are about the size of Balmino's rule for terms above degree 10, the differencesbetween larger errors (indicatingdownweighting)for the data setsof VO 1 at 1500 km and VO2 at 1500 km (55øinclination) are due to the synchronous(repetitive)nature of the orbits (over the Viking Landers)as shownin Table 2 and in Plates 1a and 1b particular(C•m, S--•m) pairsevenfor lowerdegreetermsare for the data distribution. The calibration factors (k) given in seento be quite large. In fact, the rms differencesfor lower degreeterms are over an order of magnitudegreaterthan the error estimates of GMM-1 as given in Figure 6. The coefficient discrepancybetween these models reflects the large differences seen in the orbital residuals for the two models, as shown in Table 6. 5.4. Table 8 indicatethat the model is reasonablywell calibrated, where a factor of k = 1 indicates perfect calibration. 5.5. Error Analysis The calibrated error covariance matrix from GMM-1 was used to project the orbital errors in satellite position and velocity. Table 9 showsthe projectederrors for M9, VO 1 at 300 km, VO2 at 300 km, and Mars Observer (MO) for a 6-day arc length. The results for the MO orbit are of special interest, since they indicate the expected error in MO's positionif the GMM-1 model were used in the orbit determination. Figures 8 and 9 give the error spectrumby degree Calibration of Gravity Model Errors The calibration of the gravity model error estimates is based upon the method described in section 4.2.2 and is developedin greater detail by Lerch [1991]. In the application of this method, weights of basic observationsets from different orbits are adjustedbased on subsetsolutions.The and order for the radial and along-track position compodata are separatedinto sevengroups(see Table 7), yielding nents,respectively,of Mars Observer(cross-trackerrors are sevensubsetsolutionsfor the weight adjustment.In Table 7 similar to the radial errors). Note the largest error is shown eachgroupis assignedan a priori data weight which is based for resonant order 25, indicating that a field complete to at on our experience in computingprevious gravity solutions. least degree30 is required to reasonablymodel these coefFor example, the VOl 1500-km data group is assignedan ficientsbasedupon the error spectrumby degree.The radial TABLE 7. Orbit Prediction Tests' rms of Prediction Fits Predicted Arc 1 2 3 4 5 Satellite M9 VOl VO2 VO2 VO2 1500 km 1500 km 1500 km 800 km VOl VO2 300 km 300 km Inclination, deg Arc Epoch No. of Observations 64 Jan. 13, 1972 39 55 Aug. 22, 1976 Sept. 17, 1976 1308 840 720 75 80 Oct. 26, 1976 Jan. 2, 1977 697 367 39 80 Sept. 4, 1978 Nov. 17, 1977 353 597 Average 6 7 Period, days Balmino GMM-1 18 x 18, 50 x 50, cm/s cm/s 5.61 4.20 1.62 8.04 21.40 8.17 105.6 102.9 0.36 2.58 0.68 2.38 0.73 1.34 13.2 30.1 20,884 SMITH ET ^L.' GODD^}•DM^}•S MODEL 1 Coefficient andErrors X 108 100000 10000 1000 • Rule 100 Balmino's 18 x 18 10 Sigma's GMM- 1 GMM- 0 5 10 15 20 25 30 35 1 40 45 50 degree Fig. 6. The rms of Mars gravitymodelCoefficients and standarddeviationsper degree. cases the estimate of GM is largely uncoupled from the remainingcoefficientsof the Mars gravity field. orbit error is expected to be the largest contributorto the uncertaintyof the topographyof Mars derivedfrom the laser altimeter [Zuber et al., 1992]. 6. 5.6. Recovery of GM GEOPHYSICAL IMPLICATIONS along with the other coefficientsof the Mars gravity field. Plate 3 shows free air gravity anomaliescomputed from GMM-1 complete to degree and order 50, and Plate 4 displaysaccompanyinggravity anomaly errors computed The value of GM determined from the error covariance matrix. As illustrated in Plates 1 In the GMM-1 solution, the GM of Mars was adjusted in the solution was 42828.36 -+ 0.05 km3 s-2. Lemoine[1992]analyzeda smallerset of and 2, the satellitesused in this study are characterizedby a complicateddistributionof low-altitude data. The shortest wavelengthsresolved (200- to 300-km half-wavelength)occur within the latitudinal band of -+40ø, correspondingto the (42828.32-+0.13km3 s-2 [Null, 1969]),Mariner6 (42828.22 data coveragefrom the VOl low orbit. This region also -+ 1.83 km3 s-2 [Andersonet al., 1970]),and Mariner9 includesthe periapsiscoverage(0o-30ø latitude) of the VO2 (42828.35-+0.55km3 s-2 [O'Neillet al., 1973])arein close low orbit as seenin Plate 2. That plate also showsthat above agreementwith the GMM-1 value. The Mariners 4, 6, and 9 40øN latitude there is still strong coverage of periapsis, values are especially interesting, since they were derived extendingto 63øNlatitude, particularlyfor the VO2 800-km from tracking of spacecraft from flybys of Mars. In these altitude orbit. This coverage is reflected in the gravity Viking and Mariner 9 Doppler data as well as Viking Orbiter rangedata and determineda value of GM of 42828.40-+ 0.03 km 3 s-2. The estimates of the Mars GM from Mariner 4 8 Difference X 10 Balmtrio Rule 3250 1444 812 52O 361 265 203 160 130 107 90 77 66 58 51 45 40 rms degree 51 32 61 71 93 80 90 76 77 78 93 58 62 48 57 49 44 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 76 31 2 115 96 96 39 54 116 165 116 58 142 48 88 118 69 28 27 35 61 44 65 48 60 132 37 64 44 2 27 101 225 113 133 123 87 64 65 64 64 42 21 43 49 100 94 22 12 40 74 38 44 47 60 55 48 26 55 102 78 83 28 81 75 26 13 8 20 21 57 120 78 84 50 58 45 22 101 42 45 15 22 62 52 169 77 66 38 61 59 20 41 71 38 101 130 54 62 69 25 28 51 114 14 99 171 59 42 14 85 54 45 10 9 29 90 45 59 46 88 67 14 19 54 37 32 88 58 3 76 23 25 40 37 24 84 26 21 67 61 25 12 68 3 4 5 6 7 8 9 10 11 12 13 68 84 57 41 56 49 17 4 34 7 106 24 42 69 61 50 8 0 56 49 67 6 62 90 27 27 25 71 59 73 49 65 30 43 0 1 2 66 49 31 80 34 57 19 11 71 77 19 44 36 69 14 15 16 17 18 52 35 62 40 69 order 70 80 54 99 Fig. 7. 56 58 63 69 Coefficientdifferencesfor Balminominus GMM-1. 20,885 SMITH ET AL..' GODDARD MARS MODEL 1 TABLE 8. TABLE 9. Calibration and Data Weights of GMM-1 Projected Gravity Orbit Error From GMM-1 Covariances Subset Solution Data Set Removed A Priori Satellite GMM- GMM-1 Arc 1 Calibration Length, days Inclination, deg Weights, cm/s Weights, cm/s Factors k* 55 75 80 80 39 39 1.0 1.0 0.71 0.71 1.0 0.71 1.0 4.1 1.5 0.72 1.0 3.5 0.8 2.0 1.2 0.81 0.81 1.16 0.96 1.05 0.99 Satellite Along Track Radial Cross Track Total Orbit Position Error, rn VO2 VO2 VO2 VO2 VOl VOl M9 1500 km 1500 km 800 km 300 km 1500 km 300 km Mars Observer Mariner 9 VOl 300 km VO2 300 km 6 6 6 6 67 2 26 26 757 4 69 83 90 2 31 12 765 5 80 87 Orbit Velocity Error, cm/s Here,k = [E(c - •)2/•(•.2 _ o.2)]1/2, w' = w/k2, o.o= 1/wl/2, o.b= 1/(w')'/2. Mars Observer Mariner 9 VO 1 300 km VO2 300 km 6 6 6 6 66.3 0.07 1.9 2.0 8.0 0.03 0.8 0.9 8.0 0.02 0.2 0.1 67.1 0.08 2.1 2.3 *For complete convergence, k = 1. Long-period terms excluded. anomaly error map, which shows more longitudinal structure and better resolution than in the corresponding southern hemisphere beyond the region of 40øS latitude. Plate 4 also shows some artificial "striping," which is a consequence of the distribution of orbit tracks, along which the gravity anomaly errors are smallest. In Plate 3, the free air anomalies are overlain by contours of topography. The topographic field is a spherical harmonic expansion, also complete to degree and order 50, of the Mars Digital Elevation Model (DEM [Wu et al., 1986]). Our spherical harmonic topographic model is plotted with respect to the GMM-1 geoid, so that all topographic values are relative to an equipotential surface. Thus the spherical harmonic model and DEM have similar hypsometric distributions; however, the reference levels differ. It is worth noting that the range of topographic highs and lows is reduced by 2 km as a result of referencing the topography to the geoid rather than to an ellipsoid. Plate 3 shows that, as in previous studies, the gravity anomalies correlate well with principal features of Martian topography, including volcanic shields, impact basins, and the Vailes Marineris. Most major features exhibit anomalies with considerably higher magnitudesthan in previous models. GMM-1 also showsgravity anomaliesin associationwith some observed structures that were not previously detected. For example, GMM-1 resolves all three Tharsis Montes, while the model of Balmino et al. [1982] fails to resolve the central volcano in the line, Pavonis Mons. GMM-1 also shows considerably more detail associated with the Vailes Marineris, and for several of the major impact basinsinclud- ing Isidis and Argyre. However, it is important to interpret short-wavelength features resolved in the model with cau- Total RMS tion, as the individual coefficients associated with the highest degree and order terms are 100% in error. One of the most prominent physiographic features on Mars is the hemispheric dichotomy, which is characterized by a 2-km elevation difference between the northern and southern hemispheres of Mars. Plate 3 indicates that the dichotomy does not have a distinct gravitational signature associated with it. This observation indicates GMM-1 fail to correlate with observed surface Asforthegravity anomaly representatic•n ofthefield,the geoid from GMM-1 as shown in Plate 5 exhibits a higher dynamic range of power (2300 m versus 1950 m) than the model of Balmino et al. [1982]. The distribution of geoid errors shows a similar pattern to the gravity anomaly errors. 7. SUMMARY Reanalysis of Doppler tracking data from the M9, VOl, and VO2 spacecraft has led to the derivation of a 50th degree = 67m RMS per Order 4O a2 15 20 '• -- 15 5 5 0 0 0 5 10 15 20 25 30 35 Harmonicdegree Fig. 8. 40 45 50 features. These include a 300-mGal negative anomaly on the western edge of Tharsis (200øE longitude, 30øN latitude) and a 200-mGal positive anomaly in Utopia (105øE longitude, 50øN latitude). Both of these areas are in the northern hemisphere and may have been resurfaced. The latter feature was also present in the field of Balmino et al. [1982], but the anomaly was smaller in magnitude. RMS per Degree 30 that the dichot- omy boundary is isostatically compensated at the resolvable wavelengths of GMM-1, perhaps due to a change in crustal thickness across the boundary, as suggested in previous studies [Phillips and Saunders, 1975; Lambeck, 1979; Phillips and Lambeck, 1980; Phillips, 1988]. It is significant that several prominent anomalies in I 0 ,,,, I ,,., 5 [ ,.,, 10 i ,,,, I ,,,, 15 20 I ,,,! I ,,,, 25 30 I ,,,, I ,,,, 35 40 I ,,,, I i 45 50 Harmonicorder Projected radial position error on Mars Observer from GMM-1 gravity covariances. 20,886 SMITH ET AL..' GODDARDMARS MODEL 1 Total RMS = 757m RMS per Degree RMS per Order 600 • '•700 ,•300 o 0 ................ 0 5 0 10 15 • 25 :50 :35 40 45 50 0 5 10 Harmonicdegree Fig. 9. 15 • 25 :50 :35 40 45 •3 Harmonicorder Projected along-track position error on Mars Observer from GMM-I gravity covariances. and order gravitational model for Mars. The model has a maximum (half-wavelength) spatial resolution of 300 km where the data permit, which represents a factor of 2 improvement over that attained by the previous field of Balmino et al. [1982], which utilized essentially the same data. Probable reasons for the improvement include increasedcomputationalcapabilities, the application of collocation and optimum data weighting techniques in the least squares inversion for the field, and the use of longer arcs (days versus hours) than used previously for Viking lowaltitude data made possibleby improved force and measurement models. Error analyses based on the observation data, derived power spectrum, and comparison with topography demonstrate that this field representsthe orbits with considerably better accuracy and shows greater resolution of identifiable Mar s F ree Ai r geological structures than previous models. The model also shows a greater dynamic range of power in both the gravity anomaly field and the geoid. The inclusion of X band tracking data from the 379-km altitude, near-circular, polar orbit of Mars Observer will allow significantimprovement of the Martian gravitational field [Smith et al., 1990b; Esposito et al., 1990; Tyler et al., 1992], the greatest refinement occurring at high latitudes far removed from the M9, VOl, and VO2 periapsis latitudes. That gravitational field, in combination with topography data from the Mars Observer laser altimeter [Zuber et al., 1992], will allow detailed analyses of Mars' internal structure, state of lithospheric stress, and mechanismsof isostatic compensationof surface topography. Note added in proof. Grav i t y Anoma I i es Since the loss of the Mars Ob- Compu ted F rom GMM- 1 ,! Contour 60 -- interval --------,.• .•..__..._• • : Anomol ies = 50 mgals /• , Topogrophy ,• .---.. ,. -.-'-'.. = 1 km • 20 •2 ( 0 v C•' ' -•• •7•0 20 0 • ' eO •0 •00 120 •&O •0 • • • ,• ' •80 200 ß 220 2 0 • 260 -.• ' 280 o - 500 , 320 340 Longlfude Plate 3. Free air gravity anomalies computed from GMM-1 (colors). The contour interval is 50 mGals. The anomaliesare plotted relative to a reference ellipsoid with an inverse flattening of 191.1372. The line contours in the plot represent a 50 x 50 spherical harmonic expansion of the Mars DEM [Wu et al., 1986], with topographic values referenced to the GMM-1 geoid. The contour interval is 1 km. 360 SMITH ET AL.: GODDARDMARS MODEL 1 Mars Oravi Jy Anomaly Contour Errors Inferval = 2.0 20,887 from GMM-1 rngals 60 50: 40 - $0 20- 0- -20 - -,30 - -$0 - -60 - ;. 0 2o •o •o so •oo ,20 14o 16o ,80 Longi 48 50 52 54 56 Plate 4. Contour •-'• •o ß"-'",,-" '.".'•,.'N -ø o• 82 64 Geo i d Interval :--'"-'--/ . .)(\ -•o- 60 66 68 70 72 220 Surf : 74 ace Hetght$ 76 78 80 82 84 2so 86 88 280 90 ' ',• , -- 50 m. , 3oo s2o 3•o 360 92 g4 9E 98 100 , 1 km c' • 2-,••., . oo -, '."-L.«,• ; -, ,.,__.----:'• • -.-'v.•,-----, .... .•,o.o.-•o•, -..to- ,, -- ,.., ,'----'%k. ",.-- • '• ,- ..,';,• ••..,,.•, GMM-1 Topography ,, -..o---. ß , (-• -, From Compu t,ed (Z} /-'"-•..-/2 '" •,"".,%,"• • (_ c,2..../o - •o- 240 Gravity anomaly errors from GMM-1. The contour interval is 2 mGals. Mar s so 58 200 rude / _.•• ' • • o ,. --.•',.•. ß k,k' .,' c ' ,-•. c_._.• ;-..& , '" _ ø -50 - -60 - •'•• ... -65 o •o •o •o so •oo l•o •o J8o •o =oo 2•o •o •o •o •oo a•o •4o k ong i • ude .:: . Plate 5. Geoid anomaliesfrom GMM-1. The contour interval is 50 m. Also shown are contours of topography, produced as in Plate 3, with a contour interval of 1 km. •o 20,888 SMITH ET AL.: GODDARDMARS MODEL 1 server spacecraftin August 1993, planning has been initiated for a recovery mission with a probable launch in November 1996 and arrival at Mars in September 1997. While the details of the mission are still being defined at the time of this writing, the base line favors a spacecraft(or multiple spacecraft) with X band tracking capability and an orbit as close as possible to the original Mars Observer orbit. If this occurs, the conclusions regarding orbit and gravity field determination based on GMM-1 can be expected to apply for the Mars Observer Recovery Mission. Esposito, P. B., S. Demcak, and D. Roth, Gravity field determination for Mars Observer, AIAA Rep., 621-626, 1990. Esposito, P. B., W. B. Banerdt, G. F. Lindal, W. L. Sjogren, M. A. Slade, B. G. Bills, D. E. Smith, and G. Balmino, Gravity and topography,in Mars, edited by H. H. Kieffer et al., pp. 209-248, University of Arizona Press, Tucson, 1992. Gapcynski, J.P., R. H. Tolson, and W. H. Michael, Jr., Mars gravity field: Combined Viking and Mariner 9 results, J. Geophys. Res., 82, 4325-4327, 1977. Hopfield, H. S., Tropospheric effect on electromagneticallymeasuredrange: Prediction from surface weather data, Radio Sci., 6, 357-367, 1971. Jordan, J. F., and J. Lorell, Mariner 9: An instrument of dynamical science, Icarus, 25, 146-165, 1975. Kaula, W. M., Theory of Satellite Geodesy, 124 pp., Blaisdell, Acknowledgments. We are grateful to George Balmino for proWaltham, Mass., 1966. viding the DSN tracking data used in the solution and for suggestions and analysesassociatedwith the developmentof our solution, Lambeck, K., Comments on the gravity and topographyof Mars, J. including comparisonsof gravity models with his JEEP software. Geophys. Res., 84, 6241-6247, 1979. Thisstudyhasalsobenefited significantly fromcarefulreviewsby Lemoine, F. G., Mars: The dynamics of orbiting satellites and Robert Reasenbergand Nicole Rappaport, from suggestionscongravity model development, Ph.D. thesis, 413 pp., Univ. of Col., cerning the development of the model from Bill Sjogren, Pat Boulder, 1992. Esposito, and Ed Christensen, and from discussionswith Bruce Lerch, F. J., Optimum data weighting and error calibration for Bills, Herb Frey, and Walter Kiefer. Finally, we thank Dave estimation of gravitational parameters, Bull. Geod., 65, 44-52, Rowlands, Despina Pavlis, Scott Luthcke, and John McCarthy for 1991. their analysis with the GEODYN interplanetary orbit program, and Lerch, F. J., S. M. Klosko, R. E. Laubscher, and C. A. Wagner, Joe Chan, Doug Chinn, Huseyin Iz, Rich Ullman, and Sheila Gravity model improvement using GEOS 3 (GEM 9 and 10), J. Kapoor for analytical and computer supportin the preparation of the Geophys. Res., 84, 3897-3915, 1979. solution and material of this document. Lerch, F. J., J. G. Marsh, S. M. Klosko, E. C. Pavlis, G. B. Patel, D. S. Chinn, and C. A. Wagner, An improved error assessment for the GEM-T1 gravitational model, NASA Tech. Memo., REFERENCES 100713, 1988. Anderson,J. D., L. Efron, and S. K. Wong, Martian massand Lerch, F. J., D. E. Smith, J. C. Chan, D. S. Chinn, and G. B. Patel, Earth-Moon mass from coherent S-band tracking of Mariners 6 High degree gravitational sensitivity from Mars orbiters for the and 7, Science, 167, 227-229, 1970. GMM-1 gravity model, NASA Tech. Memo., in press, 1993. Balmino, G., B. Moynot, and N. Val/•s, Gravity field model of Mars Lorell, J., G. H. Born, E. J. Christensen, J. F. Jordan, P. A. Laing, in spherical harmonics up to degree and order eighteen, J. W. A. Martin, W. L. Sjogren, I. I. Shapiro, R. D. Reasenberg, Geophys. Res., 87, 9735-9746, 1982. and G. L. Slater, Mariner 9 celestial mechanics experiment, Science, 175, 317-320, 1972. Banerdt, W. B., R. J. Phillips, N.H. Sleep, and R. S. Saunders, Thick shell tectonics on one-plate planets: Applications to Mars, Lorell, J., G. H. Born, E. J. Christensen, P. B. Esposito, J. F. J. Geophys. Res., 87, 9723-9733, 1982. Jordan, P. A. Laing, W. L. Sjogren, S. K. Wong, R. D. ReasenBanerdt, W. B., M.P. Golombek, and K. L. Tanaka, Stress and berg, I. I. Shapiro, and G. L. Slater, Gravity field of Mars from tectonics on Mars, in Mars, edited by H. H. Kieffer et al., pp. Mariner 9 tracking data, Icarus, 18, 304-316, 1973. 249-297, University of Arizona Press, Tucson, 1992. Marsh, J. G., et al., A new gravitational model for the Earth from Bills, B. G., and A. J. Ferrari, Mars topography harmonics and satellite tracking data: GEM-T1, J. Geophys. Res., 93(B6), 61696215, 1988. geophysicalimplications, J. Geophys. Res., 83, 3497-3508, 1978. Born, G. H., Mars physical parametersas determinedfrom Mariner Marsh, J. G., et al., The GEM-T2 gravitational model, J. Geophys. Res., 95, 22,043-22,071, 1990. 9 observations of the natural satellites and Doppler tracking, J. Geophys. Res., 79, 4837-4844, 1974. Moritz, H., Least squarescollocation, Rev. Geophys., 16, 421-430, 1978. Christensen,E. J., and G. Balmino, Developmentand analysisof a twelfth degree and order gravity model for Mars, J. Geophys. Moyer, T. D., Mathematical formulation of the double precision Res., 84, 7943-7953, 1979. orbit determinationprogram (DPODP), JPL Tech. Rep. 32-1527, Christensen, E. J., and B. G. Williams, Mars gravity field derived Jet Propul. Lab., Pasadena, Calif., 1971. from Viking-1 and Viking-2: The navigation result, J. Guidance Nerem, R. S., B. G. Bills, and J. B. McNamee, A high-resolution Contr., 3, 179-183, 1979. gravity model for Venus: GVM-1, Geophys. Res. Lett., 20, 599-602, 1993. Culp, R. D., and A. I. Stewart, Time-dependent model of the Martian atmosphere for use in orbit lifetime and sustenance Null, G. W., A solution for the mass and dynamical oblateness of studies, J. Astronaut. Sci., 32, 329-341, 1984. Mars using Mariner IV Doppler data, Bull. Am. Astron. Soc., 1, Davies, M. E., V. K. Abalakin, J. H. Lieske, P. K. Seidelmann, 356, 1969. A. T. Sinclair, A.M. Sinzi, B. A. Smith, and Y. S. Tjuflin, O'Neill, W. J., J. F. Jordan, J. W. Zielenbach, S. K. Wong, R. T. Reports of IAU Working Group on cartographiccoordinatesand Mitchell, W. A. Webb, and P. E. Koskela, Mariner 9 navigation, rotational elements of the planets and satellites, Celestial Mech., JPL Tech. Rep. 32-1586, Jet Propul. Lab., Pasadena, Calif., 1973. 29, 309-321, 1986. Phillips, R. J., The geophysicalsignatureof the martian dichotomy, Davies, M. E., V. K. Abalakin, M. Bursa, G. E. Hunt, J. H. Lieske, Eos Trans. AGU, 69, 389, 1988. B. Morando, R. H. Rapp, P. K. Seidelmann, A. T. Sinclair, and Phillips, R. J., and K. Lambeck, Gravity fields of the terrestrial Y. S. Tjuflin, Report of the IAU/IAG/COSPAR working groupon planets: Long wavelength anomalies and tectonics, Rev. Geocartographicand rotational elements of the planets and satellites: phys., 18, 27-86, 1980. 1988, Celest. Mech. Dyn. Astron., 46, 187-204, 1989. Phillips, R. J., and R. S. Saunders, The isostatic state of martian Davies, M. E., R. M. Batson, and S.S. C. Wu, Geodesyand topography, J. Geophys. Res., 80, 2893-2898, 1975. cartography,in Mars, edited by H. H. Kieffer et al., pp. 321-342, Puthey, B. H., General theory for dynamic satellite geodesy, The University of Arizona Press, Tucson, 1992. National Geodetic Satellite Program, NASA Spec. Publ., SP-365, 319-334, 1977. Duxbury, T. C., and J. D. Callahan, Phobos and Deimos astronometric observationsfrom Viking, Astron. Astrophys., 201, 169- Reasenberg,R. D., I. I. Shapiro, and R. D. White, The gravity field 176, 1988. of Mars, Geophys. Res. Lett., 2, 89-92, 1975. Duxbury, T. C., and J. D. Callahan, Phobosand Deimos astrono- Rosborough,G. W., and F. G. Lemoine, Sensitivity studiesof Mars metric observations from Mariner 9, Astron. Astrophys., 216, orbiters for Mars gravity recovery, J. Astron. Sci., 39, 327-358, 284-293, 1989. 1991. SMITH ET AL.: GODDARDMARS MODEL 1 Schwartz, K. P., Least-squares collocation for large systems, Bull. Geod. $ci., 35,309-324, 1976. Schwartz, K. P., On the application of least-squares collocation models to physical geodesy, in Approximation Methods in Geodesy, edited by H. Moritz and H. Sunkel, pp. 89-116, H. Wichmann-Verlag, Karlsrfihe, Germany, 1978. Sjogren, W. L., J. Lorell, L. Wong, and W. Downs, Mars gravity field based on a short arc technique, J. Geophys. Res., 80, 2899-2908, 1975. Sleep, N.H., and R. J. Phillips, An isostatic model for the Tharsis province, Mars, Geophys. Res. Lett., 6, 803-806, 1979. Sleep, N.H., and R. J. Phillips, Gravity and lithospheric stresson the terrestrial planets with reference to the Tharsis region of Mars, J. Geophys. Res., 90, 4469-4489, 1985. Smith, D. E., F. J. Lerch, R. S. Nerem, G. B. Patel, and S. K. Fricke, Developing an improved higher resolution gravity field model for Mars, Eos Trans. AGU, 71, 1427, 1990a. Smith, D. E., F. J. Lerch, J. C. Chan, D. S. Chinn, H. B. Iz, A. Mallama, and G. B. Patel, Mars gravity field error analysis from simulated tracking of Mars Observer, J. Geophys. Res., 95, 14,155-14,167, 1990b. Smith, D. E., F. J. Lerch, R. S. Nerem, M. T. Zuber, G. B. Patel, and S. K. Fricke, An improved gravitational model for Mars: Goddard Mars Model 1 (GMM-1), NASA Tech. Memo., 104584, 50 pp., 1993. Snyder, C. W., The missions of the Viking orbiters, J. Geophys. 20,889 Tyler, G. L., G. Balmino, D. Hinson, W. L. Sjogren, D. E. Smith, R. T. Woo, S. Asmar, M. J. Connally, and R. A. Simpson, Radio scienceinvestigationswith Mars Observer, J. Geophys. Res., 97, 7759-7779, 1992. Willemann, R. J., and D. L. Turcotte, The role of membrane stress in the support of the Tharsis rise, J. Geophys. Res., 87, 97939801, 1982. Wu, S.S. C., R. Jordan, and F. J. Schaefer, Mars global topographic map, 1:15,000,000 scale, NASA Tech. Memo., 88383, 614-617, 1986. Zuber, M. T., D. E. Smith, F. J. Lerch, R. S. Nerem, G. B. Patel, and S. K. Fricke, A 40th degree and order gravitational field model for Mars (abstract), Lunar Planet. $ci., XXII, 1581-1582, 1991. Zuber, M. T., D. E. Smith, S.C. Solomon, D. O. Muhleman, J. W. Head, J. B. Garvin, H. V. Frey, J. B. Abshire, and J. L. Bufton, The Mars Observer laser altimeter investigation, J. Geophys. Res., 97, 7781-7797, 1992. S. K. Fricke, RMS Technologies, Inc., Landover, MD 20785. F. G. Lemoine, F. J. Lerch, R. S. Nerem, and D. E. Smith, Laboratory for Terrestrial Physics, NASA Goddard Space Flight Center, Greenbelt, MD 20771. G. B. Patel, Hughes STX Corporation, Lanham, MD 20706. M. T. Zuber, Department of Earth and Planetary Sciences, Johns Hopkins University, Baltimore, MD 21218. Res., 82, 3971-3983, 1977. Snyder, C. W., The extended mission of Viking, J. Geophys. Res., 84, 7917-7933, 1979. Standish, E. M., Jr., M. S. W. Keesey, and X. X. Newhall, JPL development ephemeris no. 96, JPL Tech. Rep. 32-1603, Jet Propul. Lab., Pasadena, Calif., 1976. (Received March 11, 1993; revised July 6, 1993; accepted July 7, 1993.)
© Copyright 2024 Paperzz