The Evolution of Large Shield Volcanoes on Venus

Transcription

1 1 The Evolution of Large Shield Volcanoes on Venus Robert R. Herrick Lunar and Planetary Institute, Houston, Texas now at Geophysical Institute, University of Alaska Fairbanks Josef Dufek University of Chicago, Chicago, Illinois now at University of Washington, Seattle, Washington Patrick J. McGovern Lunar and Planetary Institute Submitted to: Journal of Geophysical Research Planets January 23, 2006 Correspondence Author: Robert R. Herrick Geophysical Institute University of Alaska Fairbanks 903 Koyukuk Dr. Fairbanks, AK

2 2 Abstract. We studied the geologic history, topographic expression, and gravity signature of twenty-nine large Venusian shield volcanoes with similar morphologies in Magellan SAR imagery. While they appear similar in imagery, sixteen have a domical topographic expression and thirteen have a central depression. Typical dimensions for the central depression are 150 km wide and 500 m deep. The central depressions are probably not calderas resulting from collapse of a shallow magma chamber, but instead are the result of a corona-like sagging of a previously domical volcano. The depressions all have some later volcanic filling. All but one of the central-depression volcanoes have been postdated by geologic features unrelated to the volcano, while most of the domical volcanoes are at the top of the stratigraphic column. Analysis of the gravity signatures in the spatial and spectral domains shows a strong correlation between the absence of postdating features and the presence of dynamic support by an underlying plume. We infer that the formation of the central depressions occurred as a result of cessation of dynamic support. However, there are some domical volcanoes whose geologic histories and gravity signatures also indicate that they are extinct, so sagging of the central region apparently does not always occur when dynamic support is removed. We suggest that the thickness of the elastic lithosphere may be a factor in determining whether a central depression forms when dynamic support is removed, but the gravity data is of insufficient resolution to test this hypothesis with admittance methods.

3 3 1. Introduction There are numerous large shield volcanoes on Venus whose flows cover an area more than 500 km in diameter. Venusian volcanoes of this size are generally accepted as forming from hot spots that are or were located over mantle upwellings [e.g., Head et al., 1992; Solomon et al., 1992; Phillips and Hansen, 1994; Smrekar et al., 1997]. In contrast, a variety of explanations have been put forth for the origin of the volcano-tectonic structures known as coronae. Coronae are quasicircular structures that typically have a raised rim superimposed by an annulus of concentric fractures or ridges. The largest coronae are comparable in size to the large shield volcanoes. Explanations for coronae origins include: coronae result from runaway partial melting in the uppermost mantle initiated by modest spreading [Tackley and Stevenson, 1991, 1993]; coronae are caused by plumes originating from a mid-mantle layer [Stofan et al., 1992]; coronae are caused by breakup of a mantle plume head [Stofan et al., 1995]; coronae form from plumes interacting with thin lithosphere and volcanoes from plumes interacting with thick lithosphere [McGill, 1994, 1998; McGovern and Solomon, 1998]; coronae are caused by detached diapirs, perhaps followed by retrograde subduction and or delamination [Janes et al., 1992; Sandwell and Schubert, 1992; Koch and Manga, 1996; Smrekar and Stofan, 1997; Jellinek et al., 2002]; and coronae are formed by small, long-lived plumes from mid-mantle depths that evolve to delaminate the lithosphere, while shield volcanoes result from large, hot plumes from the core-mantle boundary [Smrekar and Stofan, 1999]. More recent work [Stofan et al., 2001] has refined the corona database and expanded it to include 'Type 2' coronae, features with minimal surface fracturing but a topographic signature typical of coronae. A series of papers also analyzed the distribution of coronae in the expanded

4 4 database and their gravity signatures [Glaze et al., 2002; Smrekar et al., 2003; Smrekar and Stofan, 2003]. Those papers attribute the variety of corona morphologies and topographic shapes primarily to variations in crustal and lithospheric density and the complex interaction between varying lithospheric properties and a transient plume. The presence or absence of a depleted mantle layer and possible delamination have significant effects on coronae morphology and evolution. Initial plume diameters for coronae are interpreted as confined to a narrow range, and elastic lithospheric thickness is not interpreted to be an important factor in determining the presence or evolution of coronae. Other recent work [Herrick, 1999; Johnson and Richards, 2003] has taken a more holistic approach towards explaining the presence of coronae in relation to a global convection scenario. The basic idea first advanced in Herrick [1999] and elaborated on in Johnson and Richards [2003] is that in the Venusian convective regime transient small-scale upwellings coexist with larger-scale, longer-lived upwellings and downwellings. The small-scale upwellings avoid major downwelling areas and are 'swept up' if they occur near major upwellings. In this view mantle upwellings range in form from small, transient, diapiric features to large, long-lived, 'firehose' features. Typical coronae result from the former and regions with one or more large shield volcanoes result from the latter. Recent physical modeling experiments have indicated that an absence of plate tectonics, or the prevalence of stagnant lid convection, reduces the typical viscosity contrast across the base of the mantle thermal boundary layer and makes transient corona-forming thermals more common on Venus than on Earth [Jellinek et al., 2002]. Other possibilities arise if both compositional and thermal plumes coexist on a planet. Given the lack of geochemical data available to test the idea of compositional plumes on Venus, we do not consider them within this paper.

5 5 A key issue in determining the origin of coronae is whether or not coronae and shield volcanoes of similar size are related features. Do the two classes of features result from variations in a single process, for example plume strength or lithospheric thickness; or do coronae and volcanoes have separate origins, such as plumes from different boundary layers? There are a number of features that are listed in compiled databases for both coronae [Stofan et al., 2001] and volcanoes [Crumpler et al., 1997]. Most of these dually-listed features have extensive volcanic flow aprons emanating from a central location (hence the volcano designation) but are topped with a wide, flat plateau that is surrounded by a raised and/or faulted annulus (hence the corona designation). This dual classification could result from the grouping together of features with different origins by the database-compiling groups. Alternatively, these dually classified features could represent transitional forms between shield volcanoes and coronae [e.g., McGill, 1994, 1998; McGovern and Solomon, 1998]. Research we have conducted over the past few years has led to development of the following hypothesis: During the extinction phase of volcano formation, the central region of a large shield volcano sags when dynamic support from a long-lived mantle plume is removed; the width of the sagged region reflects the width of the former plume stem. This hypothesis was developed in part based on results of a regional geological-geophysical study of the Kunhild- Ereshkigal region on Venus [Herrick and McGovern, 2000]. Kunhild and Ereshkigal are similar in morphology to Sif Mons and Sappho Patera, respectively, two structures previously interpreted to be located over broad mantle upwellings and sites of geologically recent (or active) volcanism [Grimm and Phillips, 1992; Senske et al., 1992; McGill, 1994]. However, unlike Sif and Sappho, a low gravity signature and photogeologic mapping indicate that the region is currently extinct and not located over a mantle upwelling [Herrick and McGovern,

6 6 2000]. The terminal stage of both features involves a sagging of the central region. For Ereshkigal the process is recognizable as an example of the final stage of corona formation, and the central collapse is attributed to loss of dynamic support resulting from the spreading and dissipation of a hot thermal diapir [e.g., Squyres et al., 1992; Koch, 1994; Musser and Squyres, 1997]. The abundant volcanism associated with Ereshkigal has led to Ereshkigal's classification as both a volcano and a corona. McGill [1998] made a similar interpretation that Sappho was a shield volcano that evolved into a corona, and surmised that it is in its waning stages of activity. Kunhild's appearance and overall topographic shape indicate it is a typical large Venusian shield volcano, yet its terminal stage involves a central sagging that appears very similar in nature to what occurred at Ereshkigal. Because of its unusual size and morphology, it is unlikely that Kunhild's central depression is a caldera in the traditional sense of the word. Calderas approaching 100 km in diameter on Earth are associated with explosive ignimbrite eruptions; examples of basaltic volcanoes with comparable-sized calderas are the largest Martian shield volcanoes. For a typical volcanic caldera on the Earth or Mars, the collapse of the surface over a shallow magma chamber produces circumferential steep-sided faults with significant throw [e.g., Lipman, 2000]. In contrast, Kunhild's depression is surrounded by a gradual upslope to a radially-fractured annulus, and in places it appears that preexisting flows now drape over this annulus. This suggests that the cause of Kunhild's central depression was deep and that the surface response was a distributed downwarping of the summit region. A similar appearing radially fractured rim embayed by interior volcanic deposits has been observed in coronae of similar size [e.g., Eastern Eistla Regio in Smrekar and Stofan, 1999]. While sag features, rather than discrete faults, have been observed at some terrestrial volcanoes, in those cases the diameter of the sagged region is at

7 7 most a few tens of kilometers and still consistent with shallow withdrawal of magma [e.g., Walker, 1984]. Here we test the hypothesis of Herrick and McGovern [2000] with a survey of large shield volcanoes on Venus. Our approach is to analyze the imagery, topography, and gravity signatures of a number of similar-appearing Venusian shield volcanoes. In its basic form, our hypothesis predicts that extinct shield volcanoes should have a central sagged region while active volcanoes should not. Thus, shield volcanoes with a broad central low would be expected to correlate with the presence of unrelated postdating geologic features and with a gravity signature consistent with isostatic compensation. Active volcanoes are expected to be at the top of the local stratigraphic column and have a gravity signature consistent with some level of dynamic support from the mantle. 2. General Characterization On the basis of their appearance in Magellan imagery, we selected 29 shield volcanoes for detailed analysis. Our goal was to select a population that are generally similar in appearance. The volcanoes were subselected from the volcano database of Crumpler et al. [1997] because they appear to be broadly domical, are axisymmetric, and have radial flows exceeding 500 km in diameter. Our data set includes a few volcanoes listed as between 300 and 500 km in diameter in the Crumpler et al. [1997] for which we determined that the flow apron diameter had been underestimated. Our survey represents about half of the 500+ km diameter volcanoes in the Crumpler et al. [1997] database. The selected volcanoes do not appear to have any steep slopes or sharp peaks, and they are not cut by major rifts (although they may superpose a rift). Table 1 identifies the 29 volcanoes along with their basic characteristics. Location and flow apron

8 8 diameter are taken from Crumpler et al. [1997]. Height is listed relative to the typical elevation of the surrounding terrain. Topographic width is characterized by two measures: the width of the volcano at half of its maximum elevation, and the width of the volcano at an elevation 500 m above the surrounding terrain. All topographic measurements were made from perpendicular profiles across each volcano using the gridded Magellan altimetry data (GTDR). 2.2 Topography Using Magellan radar altimetry (~10 km resolution) the volcanoes were characterized by the presence or absence of a central depression of width greater than 100 km (Table 1). Sixteen of the features had no significant topographic depression and a broad domical shape (labeled "d" for domical in Table 1, column 4). Ten of the volcanoes had a broad central depression (labeled "c" for central depression). Three of the features appear to have broad central depression with volcanic edifices of significant size in their interiors (labeled "ce" for central depression with a large interior edifice). Figure 1 shows some examples of the different categories of volcanoes with topographic profiles. 2.3 Postdating Features There is no available information that can be used to estimate an absolute age for a Venusian volcano. Furthermore, there are too few craters on Venus to use crater counting techniques to evaluate the relative ages of the volcanoes. However, a crude comparison of the relative ages of the different categories of volcanoes can be made by observing what features, if any, postdate each of the volcanoes. If members of one group of volcanoes consistently have a number of geologic features that postdate the volcano, then it is reasonable to interpret this group as containing many extinct volcanoes. In contrast, if members of the group are all at the top of the

9 9 stratigraphic column, then the possibility exists that many of these volcanoes are currently or recently active. We have used basic photogeologic mapping techniques to determine if there are any geologic features that postdate the volcano that are unrelated to its formation. Our photogeologic analysis was performed using the full resolution Magellan imagery (resolution ~100 m). The nature of the features that postdate the volcano are as follows (Table 1, column 8): impact craters (C), embayment by flows from another volcano (E), wrinkle ridges (R), fractures (F), and lineaments (L) of unknown origin. Figure 2 shows examples of these relationships. Impact craters with radar-dark floors are likely to be partially flooded by volcanic flows [Herrick and Sharpton, 2000], so only impact craters with radar-bright floors and ejecta blankets that appeared intact were considered as a feature post-dating the volcano. 2.4 Isostatic Anomaly Using the spherical harmonic gravity field of Konopoliv et al. [1999] and the topography of Rappaport et al. [1999] we have calculated the global isostatic anomaly. The resolution of the Magellan-derived gravity varies across the planet, but most of the globe can be considered reliable to degree and order 75 based on the degree strength map of Konopoliv and Sjogren [1996]. The isostatic anomaly was calculated globally (to degree and order 75) by subtracting the gravity signal attributable to the Airy-compensated topography from the free-air gravity anomaly. We assumed a crustal thickness of 30 km and a crustal density of 2900 kg m -3 [Grimm and Hess, 1997]. Table 1 gives the value of the anomaly at the center of the structure and whether it appears correlated with the location of the volcano, and Figure 3 shows sample profiles of the isostatic anomaly for Sif (correlated) and Kokyanwuti (uncorrelated). A

10 10 significant, correlated positive anomaly (tens of mgals) can be interpreted as indicative of partial support of the volcano by heated mantle material. Significant flexural support of the volcano is a possible alternative explanation. A strong correlation between absence of postdating features and a significant isostatic anomaly favors the interpretation that the isostatic anomaly results from dynamic support. We interpret an anomaly with absolute magnitude near zero as indicative of crustal support by Airy isostasy (Pratt isostasy is also an acceptable interpretation). The results are insensitive to the exact choices of crustal density and thickness. 2.5 Central depressed regions A typical size for the central depressed region is on the order of 150 km across and a few hundred meters deep (Table 2). All of the depressed central regions appear to have a geologic history similar to that described for Kunhild in Herrick and McGovern [2000]. The central depressed region in the studied volcanoes appears to have formed through sagging of the topography. Circumferential normal faulting occurs in only a few of the volcanoes, and the throw is minimal. Several volcanoes have finely spaced radial fractures along the topographic rim of the depression, and in some cases volcanic flows can be seen that drape over the rim. Partial volcanic filling has occurred in all of the volcanoes with a depressed central region. A spectrum of volcanic edifice types and sizes appear in the interiors of the depressed regions, ranging from a few small shields to features that are large enough to fill almost the entire depressed region. There are end member volcanoes that have topographic breaks in slope and morphologic remnants of a depressed central region, but no significant central depression currently exists. An interpretation we favor is that a central sagged region existed that has been nearly filled by post-sagging edifices, and these volcanoes are labeled "ce" in Table 1.

11 Discussion The statistical measurements of elevation and width of the volcanoes combined with our general observations confirm that we have selected a family of reasonably similar volcanoes. There are some rather obvious correlations between the different columns of Table 1. Table 3 shows the mean and median isostatic anomaly for the different volcano types. Volcanoes with no post-dating features are clearly most likely to have a large isostatic anomaly indicative of dynamic support, and volcanoes with abundant post-dating features generally have gravity signals consistent with isostatic support of the edifice. This correlation is expected and is consistent with the idea that a large isostatic anomaly indicates dynamic support from a mantle plume beneath a currently or recently active volcano. All of the volcanoes with central depressions are consistent with isostatic support, and all but one have clear post-dating features. Most of the domical volcanoes have large isostatic anomalies and no post-dating features, but several have post-dating features and are consistent with isostatic support. What is not easily quantifiable for placement in a data table is the pervasiveness of postdating features. For domical volcanoes a postdating feature is likely to be a few fractures or folds deforming some of the distal flows, while the volcanoes with central depressions are often permeated by tectonic deformation. Nevertheless, it can be argued that at least some of the domical volcanoes are extinct features. 3. Two-layer dynamic inversion 3.1 Technique To place the gravity signature associated with each volcano in a form that allows more

12 12 quantitative interpretation, we employed both spatial and spectral domain approaches. In this section we discuss results from a spatial-domain inversion performed using the technique described in Herrick and Phillips [1992]. The approach is to use two constraints, the gravity and topography, to solve uniquely for surface densities on two subsurface layers. Stated a different way, it is always possible to match the free-air gravity signal by compensating the topography with sheets of mass anomalies placed at two depths. In the particular model we use here, we assume that the topography is supported by Airy isostasy at the base of the crust and by upper mantle convective flow. If a nominal crustal thickness is assigned, then Airy isostasy at the base of the crust can be represented by a sheet of varying surface density placed at a depth equal to the assigned crustal thickness. It is reasonable to assume that the effect of mantle convection on the gravity and topography is dominated by the flow, and corresponding mass anomalies, in the upper mantle. We can represent these anomalies by a sheet of varying surface density placed at a depth somewhat below the base of the conductive lid (i.e., the thermal lithosphere) for the convective system. If a viscosity structure for the mantle is specified and we assume that mantle convection is being driven by that mass sheet, then we can solve for the surface effect on the gravity and topography using the technique described in Richards and Hager [1984; for further description see also Hager and O'Connell 1979, 1981; Hager and Clayton, 1989]. We make all of the same model assumptions used in Herrick and Phillips [1992]. Here we restate the most important features of the model. The crustal density is assumed to be 2900 kg m -3, and the mantle density just below the crust is assumed to be 3300 kg m -3. We have assumed that the mantle is an incompressible Newtonian viscous fluid with a no-slip condition at the surface and a free-slip condition at the core-mantle boundary. We assume an isoviscous mantle. The base of the crust is assumed to be at 30 km depth, and the perturbing shell in the mantle is

13 13 placed at 200 km depth. The model parameters are the favored values found in Herrick and Phillips [1992]. That work experimented with a variety of shell depths and viscosity models. Our assumed nominal crustal thickness of 30 km is also consistent with typical estimates of the Venusian crustal thickness [e.g., Grimm and Hess, 1997]. A depth of 200 km for the lower, convective shell is consistent with the most important mantle perturbations being located in the upper mantle below a conductive lid of 100 km thickness, a typical estimated thermal lithosphere thickness [e.g, Phillips and Malin, 1983; Phillips et al., 1997; Simons et al., 1997; Brown and Grimm, 1999]. We carried out the inversion through harmonic degree and order 90, with a cos 2 taper of the coefficients from degrees 60 through 90. Because we are comparing features in the spatial domain, it is important to perform the inversions using the same number of harmonics. The degree strength of the gravity field is greater than 60 for all but one of the features (Table 1), so we feel we are not introducing significant noise into the inversions. The choices of parameters for the inversion are certainly debatable and largely unconstrained. However, they are adequate for the purposes of broad comparisons and general hypothesis testing. 3.2 Results Figure 4 shows typical results of the inversions. Table 4 shows the surface density for each shell over the center of each volcano and identifies whether the volcano is distinguishable from its surroundings on the given surface. Figure 5 graphically summarizes the results. The domical volcanoes with no post-dating features all had modest to significant negative surface densities on the lower, mantle convection shell that are consistent with dynamic support by upward moving buoyant material in the mantle. A number of additional assumptions must be made to translate

14 14 surface densities into meaningful quantities. For example, if we wish to interpret the densities in terms of a thermal plume, we must make assumptions about the thickness of the perturbing layer and the coefficient of thermal expansion. If we assume a 200 km thick perturbing layer and a nominal upper mantle density of 3300 kg m -3, then a surface density anomaly of kg m -2 translates into a density change in that layer of 5 kg m -3, or about 0.15%. If we assume a coefficient of thermal expansion of 3 x 10-5 K -1, then this density anomaly represents a 50 K temperature anomaly. Most of the domical volcanoes with no post-dating features have surface density anomalies on the mantle convection shell that have maximum negative values of -2 x 10 6 to -6 x 10 6 kg m -2. Using the above assumptions, this implies density changes of 0.3% to 0.9%, or central plume temperatures K above the ambient mantle. These values are similar to estimates from seismic tomographic studies for Earth at comparable resolutions over locations of suspected mantle plumes [see review by Nataf, 2000], and they are consistent with estimates of excess plume temperatures derived from geochemical data [e.g., Schilling, 1991; also see review by Sleep, 1992]. Using these assumptions, Figure 6 shows an example of the surface densities on the lower subsurface shell converted to temperature for Ushas Mons. Most of the volcanoes have similar modest negative anomalies,-1 x 10 6 to -4 x 10 6 kg m -2, on the upper subsurface shell (representing perturbations of the crust-mantle boundary) that are consistent with a crustal root beneath the volcano [Table 4]. For example, if the crustal density is 2900 kg m -3, then isostatic compensation of a 1 km high volcano would require a surface density of -2.9 x 10 6 kg m -2 on the upper subsurface shell. Some of the domical volcanoes with no post-dating features have positive anomalies on the upper subsurface shell. This may represent some amount of crustal thinning over a putative plume. Alternatively, it could

15 15 represent some level of flexural support. Consistent with the idea that the volcanoes with postdating features are geologically inactive and no longer located over an upwelling plume, most of these volcanoes have surface densities on the lower subsurface shell that are small in magnitude. Some of these features, however, do have significant positive anomalies on the lower subsurface shell. This can occur if the nominal crustal thickness is assumed to be deeper than its true value. For these volcanoes, the positive anomaly on the lower subsurface shell is eliminated or substantially reduced if the upper shell is placed at 15 km depth instead of 30 km [Figure 7]. Alternatively, there could be a real positive density anomaly in the upper mantle beneath some of the volcanoes. Possibilities include a denser than normal lower crust or a thermal downwelling. The positive anomaly is also essentially eliminated if the higher order harmonics are not used [Figure 8]. Among the volcanoes with postdating features, it appears that those with a depressed central region are more likely to have a positive density anomaly on the lower subsurface shell that correlates with the location of the volcano, but there are enough exceptions that we cannot confidently assert that this is a characteristic of those volcanoes. For example, from the data one could argue that a positive lower-surface anomaly that correlates with the volcano always occurs, but it is only observable in areas where the gravity field has a degree-strength greater than ~85; i.e., the positive anomaly is only observed if there is no noise in the data for harmonic degrees 60 to Spatio-spectral localization 4.1 Technique Utilizing the method developed and presented in Simons et al. [1997], we determined the RMS amplitude of the gravity, topography, and admittance (ratio of gravity to topography)

16 16 spectra for areas centered on each of the 29 volcanoes. The technique is to multiply the spherical harmonic representation of the planet by a spherical cap centered on the location of each volcano, and then analyze the resulting spectra. This is the spherical harmonic version of multiplying a one-dimensional function by a boxcar (or some other finite, symmetric function) and then evaluating the power spectra. The technique is a convolution with a broad filter in the spectral domain, and the resulting spectra can be considered valid only where the filter is convolved with valid data. Consequently, the localized spectra contains usable data only over the spectral range of the initial data minus the width of the filter. The localized spectra can also be thought of as smoothed over the width of the filter. There is a trade-off between the width of the cap in the spatial domain and the width of the filter in the spectral domain. A narrow cap, which better localizes a feature, generates a smoothed spectra with usable data over a small range of harmonics. Because the ultimate goal is to interpret the spectra in terms of a surface loading model, one can restate the problem as smaller caps providing less data to use in creating a robust spectra. The Simons et al. [1997] paper suggested two simple choices for caps. The choice they used was a cap whose width was a set number of "wavelengths" for each harmonic degree. In the spatial domain the cap varies in width for each harmonic degree, meaning the spectra is more localized for high-degree harmonics versus low-degree harmonics. In the spectral domain this results in convolution with a filter that broadens for higher harmonic degrees (once again, a trade-off between spatial and spectral resolution). Another suggested choice was a spherical cap of a fixed horizontal width equal to one wavelength of zonal harmonic degree L win (width 2 π R planet / L win ). This results in convolution with a fixed width filter in the spectral domain. A familiar analog is that multiplying a 1-dimensional function by a boxcar represents convolution

17 17 with a sinc function in the spectral domain. The first choice, the varying-width cap, is wellsuited for evaluating global patterns of variations in admittance (the focus of the Simons et al. [1997] paper), while the fixed-width cap is conceptually easier to visualize when comparing the spectra of individual features. The fixed-width cap was successfully employed by McGovern et al. [2002] to evaluate the admittance spectras of selected features on Mars, and it was also used in the investigation of the Kunhild-Ereshkigal region on Venus [Herrick and McGovern, 2000]. In this work we use a fixed-width cap, and the usable range of harmonic l in the windowed fields is L win l (L obs L win ), where L obs is the maximum degree of observation. For admittances on Venus the gravity data is the limiting factor, and we use the degree strength of the gravity field [Konopoliv and Sjogren, 1996] as an estimate of L obs. It is also possible to evaluate the spectra in terms of specific models of support of the topography by Airy isostasy and elastic flexure. We utilize the model for loading of a thin elastic shell described in McGovern et al. [2002], which is based on models presented by Turcotte et al. [1981] and Kraus [1967]. The general technique is to generate a global free-air gravity model using the global topography compensated with a specified set of parameters for the crust and lithosphere. The gravity model is then windowed in an identical manner to the actual gravity to produce a model curve for admittance in a given region. We chose simple models with top-loading only, a load density and crustal density of 2900 kg m -3, and an upper mantle density of 3300 kg m -3. The elastic lithospheric thickness and the nominal crustal thickness were varied in the models from 0 to 40 km. 4.2 Results and Discussion In our initial spectral localization we used a value of L win = 15 for a cap width of 2500 km.

18 18 We feel this is a reasonable compromise between isolating the large volcano and measuring a broad spectral range. Figure 9 shows several examples of the admittance spectras, and admittance and correlation values at several different harmonic degrees are shown in Table 5. The plots in Figure 9 also show model admittances for the topography compensated with crustal thicknesses of 20 and 40 km, and elastic lithospheric thicknesses of 10 and 30 km. There are a few general patterns in the data that we feel confident in identifying. Admittances for the domical volcanoes without any postdating features are significantly higher than the admittances for the volcanoes with depressed central regions up thru degree 40. Also, for degrees 30 through 50 the correlation between gravity and topography is significantly lower for volcanoes with depressed central regions than domical volcanoes with no post-dating features. A decorrelation between the gravity and topography coefficients can be thought of in the spatial sense as different features within the cap requiring different depths of compensation. Alternatively, one can envision density anomalies deep enough that their effect on surface topography is minimal. It is more difficult to assess whether there are significant differences between the different classes of volcanoes that have postdating features. We calculated means for a variety of quantities after dividing the volcanoes into three groups: domical volcanoes without postdating features, domical volcanoes with postdating features, and all other volcanoes. These summary plots are shown in Figure 10. Because the volcanoes of Bell Regio are in close proximity to each other but in different groups, they are not included in the summary plots. Hathor and Innini are in close enough proximity that their spectras were almost identical, and only Hathor's spectra is shown in the summary plots. In neither the summary plots nor the individual plots can we see a substantive difference between the spectra of the different types of volcanoes with postdating features.

19 19 Our general interpretation of the spectral data can be summarized as follows. As a family, the domical volcanoes without postdating features are partially dynamically supported by a mantle plume whose location correlates with the location of the volcano. The result is high admittances, particularly at longer wavelengths (up to degree 40), and a strong correlation between gravity and topography. The remaining volcanoes have admittances at long wavelengths that are modestly higher than can be accounted for by lithospheric compensation, but the nonlithospheric component contributes minimally to the regional gravity and topography signature. Without regional gravity and topography being dominated by a single plume, gravity-topography correlation drops off significantly above about degree 30 or 40. No clear distinction can be made in the spectral domain between different types of volcanoes with postdating features. The spectral data are not interpretable for the purpose of discerning the nominal thickness of the crust or elastic lithosphere in the regions encompassing the volcanoes. 5. Conclusions There are some conclusions that we feel can be made confidently, and there are others that are more tenuous. Our results strongly suggest that shield volcanoes with central depressed regions are not currently active features and are not receiving dynamic support from a mantle plume. These volcanoes are consistently post-dated by unrelated geologic features, and there is no indication in the gravity data that they are underlain by a mantle upwelling. We feel confident that many of the domical volcanoes are recently active and are located over a mantle plume. They are at the top of the stratigraphic column, and they have gravity signatures consistent with dynamic support in both the spatial and spectral domains. The morphologies of the depressed central regions indicate that the central depression forms from sagging of the center of a domical

20 20 volcano, and modest volcanism then later partially fills the depression. Based on these observations, it is tempting to conclude that the evolutionary sequence of large shield volcanoes on Venus always involves a terminal stage of corona-like sagging of the interior when the mantle plume feeding the volcano dies out. This interpretation would suggest that plume duration and the associated amount of constructional volcanism is an important factor in determining whether a corona or a shield volcano forms at a given location. However, there are some volcanoes that are domical that are best interpreted as extinct features. They are postdated by unrelated geologic features and have gravity signatures consistent with static support within the lithosphere. If shield volcanoes always evolve to having a central depression, then the extinct domical volcanoes must represent an intermediate stage. A possible evolutionary sequence that parallels an evolutionary sequence proposed for coronae [Smrekar and Stofan, 1997, 1999] would be that the central depression actually results from delamination, and extinct domical volcanoes represent a phase between upwelling and later delamination. This, however, is a rather ad hoc explanation and testing its plausibility through numerical modeling is beyond the scope of this investigation. We feel the morphologic evidence is compelling that the volcanoes with central depressions were domical at some point in their history. If the extinct domical volcanoes are at an end stage, then there must be something different that causes the terminal stage of a shield volcano to vary. McGovern and Solomon [1998] showed that elastic lithospheric thickness at the time of formation may be an important factor in determining whether a volcano or corona forms, with coronae forming over thin lithosphere and shield volcanoes over thick lithosphere. Volcanoes with central depressions were suggested to represent a transitional state between the two. The absence of coronae and dominance of shield volcanoes over present-day regions of large mantle

21 21 upwellings (where the lithosphere is likely to be thinnest) argues against the elastic lithospheric thickness being the only differentiator between corona versus volcano formation [Herrick, 1999; Johnson and Richards, 2003]. We suggest here that, if other factors are similar, the thickness of the elastic lithospheric is important in determining whether a shield volcano has a terminal stage of central sagging. Unfortunately, we do not feel the gravity data are of adequate resolution for our sample set to be able to distinguish the elastic lithospheric thickness for the different types of volcanoes with postdating features. As we discuss above, we do not favor the interpretation that the central depressions are calderas. However, if one chooses to view the central depressions as unusually large calderas, extinction of the volcano still occurs when the underlying plume goes away, but the depression results from the physical withdrawal of magma rather than the deeper removal of thermal support. A key difference in the volcano's evolution is that the structure of the central depression may form early in a volcano's history, but be kept filled until the volcano becomes extinct. The appeal of this interpretation is its ability to account for the partial to complete interior embayment of central depressions. In this case, the genetic cause for the central depression in a large shield volcano is different from that associated with a corona. In summation, our observations indicate that sagging of the central region is a common occurrence in the late stages of shield volcano formation on Venus, but this sagging apparently does not always occur. We suggest that both plume duration and elastic lithospheric thickness are important factors in determining whether the final product of a mantle upwelling is a corona, a domical shield volcano, or a shield volcano with a central sagged region. Acknowledgements. The authors thank Michael Manga and an unidentified reviewer for useful comments. This is LPI Contribution 1226.

22 22

23 23 References Brown, C. D., and R. E. Grimm, Recent tectonic and lithospheric thermal evolution of Venus, Icarus, 139, 40-48, Crumpler, L. S., and 5 others, Volcanoes and centers of volcanism on Venus, in Venus II, eds. S. W. Bougher, D. M. Hunten, and R. J. Phillips, pp , U. of Arizona Press, Tucson, Glaze, L. S., E. R. Stofan, S. E. Smrekar, and S. M. Baloga, Insights into corona formation through statistical analyses, J. Geophys. Res., 107(E12), 5135, doi: /2002je001904, Grimm, R. E., and P. C. Hess, The crust of Venus, in Venus II, eds. S. W. Bougher, D. M. Hunten, and R. J. Phillips, pp , U. of Arizona Press, Tucson, Grimm, R. E., and R. J. Phillips, Anatomy of a Venusian hot spot: Geology, gravity, and mantle dynamics of Eistla Regio, J. Geophys. Res., 97, 16,035-16,054, Hager, B. H., and R. W. Clayton, Constraints on the structure of mantle convection using seismic observations, flow models, and the geoid, in Mantle Convection: Plate Tectonics and Global Dynamics, edited by W. R. Peltier, pp , Gordon and Breach, New York, Hager, and R. J. O'Connell, Kinematic models of large-scale flow in the Earth's mantle, J. Geophys. Res., 84, , Hager, and R. J. O'Connell, A simple global model of plate dynamics and mantle convection, J. Geophys. Res., 86, , Head, J. W. L. S. Crumpler, J. C. Aubele, J. E. Guest, and R. S. Saunders, Venus volcanism: Classification of volcanic features and structures, associations, and global distribution from Magellan data, J. Geophys. Res., 97, 13,153-13,199, Herrick, R. R., Small mantle upwellings are pervasive on Venus and Earth, Geophys. Res. Lett., 26, , Herrick, R. R., and P. J. McGovern, Kunhild and Ereshkigal, an extinct hot-spot region on Venus, Geophys. Res. Lett., 27, , Herrick, R. R., and R. J. Phillips, Geological correlations with the interior density structure of Venus, J. Geophys. Res., 97, 16,107-16,034, Herrick, R. R., and V. L. Sharpton, Implications from stereo-derived topography of Venusian impact craters, J. Geophys. Res., 105, 20,245-20,262, Janes, D. M., and 6 others, Geophysical models for the formation and evolution of coronae on Venus, J. Geophys. Res., 97, 16,055-16,067, Jellinek, A. M., The influence of interior mantle temperature on the structure of plumes: Heads for Venus, tails for the Earth, Geophys. Res. Lett., 29, /2001GL014624, Johnson, C. L., and M. A. Richards, A conceptual model for the relationship between coronae and large-scale mantle dynamics on Venus, J. Geophys. Res., 108(E6), 5058,

24 doi: /2002je001962, Koch, D. M., A spreading drop model for plumes on Venus, J. Geophys. Res., 99, , Koch, D. M., and M. Manga, Neutrally buoyant diapirs: A model for Venus coronae, Geophys. Res. Lett., 23, , Konopliv, A. S., and W. L. Sjogren, Venus Gravity Handbook, JPL Publication 96-2, Pasadena, Konopliv, A. S., W. B. Banerdt, and W. L. Sjogren, Venus gravity: 180 th degree and order model, Icarus, 134, 3-18, Kraus, H., Thin Elastic Shells, John Wiley, New York, Lipman, P. W., Calderas, in Encyclopedia of Volcanoes, ed. H. Sigurdsson, pp , Academic Press, San Diego, McGill, G. E., Hotspot evolution and Venusian tectonic style, J. Geophys. Res., 99, 23,149-23,161, McGill, G. E., Central Eistla Regio: Origin and relative age of topographic rise, J. Geophys. Res., 103, , McGovern, P. J., and S. C. Solomon, Growth of large volcanoes on Venus: Mechanical models and implications for structural evolution, J. Geophys. Res., 103, 11,071-11,101, McGovern, P. J., S. C. Solomon, D. E. Smith, M. T. Zuber, M. Simons, M. A. Wieczorek, R. J. Phillips, G. A. Neumann, O. Aharonson, and J. W. Head, Localized gravity/topography admittance and correlation spectra on Mars: Implications for regional and global evolution, J. Geophys. Res., 107(E12), 5136, doi: /2002je001854, Musser, G. S., and S. W. Squyres, A coupled thermal-mechanical model for corona formation on Venus, J. Geophys. Res., 102, , Nataf, H.-C., Seismic imaging of mantle plumes, Annu. Rev. Earth Planet. Sci., 28, , Phillips, R. J., and V. L. Hansen, Tectonic and magmatic evolution of Venus, Ann. Rev. Earth Planet. Sci., 22, , Phillips, R. J., and M. C. Malin, The interior of Venus and tectonic implications, in Venus, edited by D. M. Hunten, L. Colin, T. M. Donahue, and V. I. Moroz, pp , University of Arizona Press, Tucson, Phillips, R. J., C. L. Johnson, S. L. Mackwell, P. Morgan, D. T. Sandwell, and M. T. Zuber, Lithsopheric mechanics and dynamics of Venus, in Venus II, eds. S. W. Bougher, D. M. Hunten, and R. J. Phillips, pp , U. of Arizona Press, Tucson, 1997 Rappaport, N. J., A. S. Konopliv, and A. B. Kucinskas, An improved 360 degree and order model of Venus topography, Icarus, 134, 19-31, Richards, M. A., and B. H. Hager, Geoid anomalies in a dynamic Earth, J. Geophys. Res., 89, ,

25 Sandwell, D. T., and G. Schubert, Flexural ridges, trenches, and outer rises around coronae on Venus, J. Geophys. Res., 97, 16,069-16,084, Schilling, J., Fluxes and excess temperatures of mantle plumes inferred from their interaction with migrating mid-ocean ridges, Nature, 352, , Senske, D. A., G. G. Schaber, and E. R. Stofan, Regional topographic rises on Venus: Geology of Western Eistla Regio and comparison to Beta Regio and Atla Regio, J. Geophys. Res., 97, 13,395-13,420, Simons, M., S. C. Solomon, and B. H. Hager, Localization of gravity and topography: constraints on the tectonics and mantle dynamics of Venus, Geophys. J. Int., 131, 24-44, Sleep, N. H., Hotspot volcanism and mantle plumes, Annu. Rev. Earth Planet. Sci., 20, 19-43, Smrekar, S. E., and E. R. Stofan, Corona formation and heat loss on Venus by coupled upwelling and delamination, Science, 277, , Smrekar, S. E., and E. R. Stofan, Origin of corona-dominated topographic rises on Venus, Icarus, 139, , Smrekar, S. E., and E. R. Stofan, Effects of lithospheric properties on the formation of Type 2 coronae on Venus, J. Geophys. Res., 108(E8), 5091, doi: /2002je001930, Smrekar, S. E., W. S. Kiefer, and E. R. Stofan, Large volcanic rises on Venus, in Venus II, eds. S. W. Bougher, D. M. Hunten, and R. J. Phillips, pp , U. of Arizona Press, Tucson, Smrekar, S. E., R. Comstock, and F. S. Anderson, A gravity survey of Type 2 coronae on Venus, J. Geophys. Res., 108(E8), 5090, doi: /2002je001935, Solomon, S. C., and 10 others, Venus tectonics: An overview of Magellan observations, J. Geophys. Res., 97, 13,199-13,256, Squyres, S. W., and 6 others, The morphology and evolution of coronae on Venus, J. Geophys. Res., 97, 13,611-13,634, Stofan, E. R., and 6 others, Global distribution and characteristics of coronae and related features on Venus: Implications for origin and relation to mantle processes, J. Geophys. Res., 97, 13,347-13,378, Stofan, E. R., S. E. Smrekar, D. L. Bindschadler, and D. A. Senske, Large topographic rises on Venus: Implications for mantle upwelling, J. Geophys. Res., 100, 23,317-23,327, Stofan, E. R., S. E. Smrekar, S. W. Tapper, J. E. Guest, and P. M. Grindrod, Preliminary analysis of an expanded corona database for Venus, Geophys. Res. Lett., 28, , Tackley, P. J., and D. J. Stevenson, The production of small Venusian coronae by Rayleigh- Taylor instabilities in the uppermost mantle (abstract), Eos, Trans. AGU, 72, 287, Tackley, P. J., and D. J. Stevenson, A mechanism for spontaneous self-perpetuating volcanism on the terrestrial planets, in Flow and Creep in the Solar System: Observations, Modeling and Theory, eds. D. B. Stone and S. K. Runcorn, pp , Kluwer Academic Publishers, 25

26 Netherlands, Turcotte, D. L., R. J. Willemann, W. F. Haxby, and J. Norberry, Role of membrane stresses in support of planetary topography, J. Geophys. Res., 86, , Walker, G. P. L., Downsag calderas, ring faults, caldera sizes, and incremental caldera growth, J. Geophys. Res., 89, ,

27 27 Figure Captions Figure 1. Imagery and topographic profiles for several shield volcanoes (trailing letter indicates volcano type in table 1): (a) Sif Mons d, (b) Innini Mons d, (c) Kunapipi Mons d, (d) Chloris Mons ce, (e) Kokyanwuti Mons - c, and (f) Atanua Mons - c. Figure 2. Profiles of the isostatic anomaly for (a) Sif Mons d, and (b) Kokyanwuti - c. See Figure 1 for locations of profiles. Figure 3. Examples of the different types of features that postdate Nzambi Corona (top) and Mielikki Mons (bottom). Figure 4. Examples of inversions for surface densities on two layers for (a) Ushas Mons - d, (b) Hathor Mons d, and Innini Mons - d, (c) Tuulikki Mons - d, and (d) Atanua Mons - c. Upper layer is isostatic compensation at a depth of 30 km, and lower layer is a mass sheet driving mantle flow placed at a depth of 200 km. Inversion was carried out through harmonic degree and order 90, with a cos 2 taper from degrees 60 to 90. Contour interval is 10 6 kg m -2. Scale bar at image bottom is 500 km in length. Figure 5. Surface densities on upper and lower subsurface shells (units 10 6 kg m -2 for): domical volcanoes without postdating features (squares), domical with postdating features (circles), volcanoes with a depressed central region (triangles), and volcanoes with depressed region that has a large interior edifice (stars). Figure 6. Profile of topography and estimate of temperature anomaly on the lower subsurface shell for Ushas Mons. Temperatures are calculated from the surface density on the lower shell for the two-layer inversion. The surface density, centered at 200 km depth, is converted to temperature by assuming a perturbing layer of thickness 200 km and a coefficient of

28 28 thermal expansion of 3 x 10-5 K -1. Figure 7. Same as Figure 4d (Atanua Mons), but with upper subsurface shell placed at a depth of 15 km. Figure 8. Same as Figure 4d (Atanua Mons), but inversion is only carried out to degree and order 60, with a cos 2 taper from degrees 40 to 60. Figure 9. Spatio-spectral localizations for several volcanoes using L win = 15 (2500 km window). Thick black line is admittance (mgal/km) and thin dotted line is gravity-topography correlation. Thin lines are the admittance of the windowed topography compensated with elastic lithospheric thicknesses of 10 km (solid lines) and 30 km (dashed lines), and crustal thicknesses of 20 km (lower line of pair) and 40 km (upper line). Sif Mons, Hathor Mons, Unnamed (71.5 N, E, and Uretsete Mons are type d in Table 1; Nagavonyi Corona is type ce; and Nzambi Corona, Uti Hiata, and Mielikki Mons are type c. Figure 10. Summary of spatio-spectral localizations with mean (solid line) and standard deviations (dashed line) for domical volcanoes with no postdating features (thick line), domical volcanoes with postdating features (medium line) and nondomical volcanoes (thin line).

29 29 Tables Table 1. General properties of studied volcanoes. Width (km) Isostatic Anomaly Volcano Name Latitude Longitude Type Height (km) Half-Width 500-m Width Postdating mgal Separable Ushas Mons d x 60 y Sif Mons d x 46 y Tepev Mons d x 45 y Dzalarhons Mons d x 45 y Idunn Mons d x 40 y Unnamed d F,R 29 y Innini Mons d x 28 y Hathor Mons d x 28 y Unnamed d x 25 y Renpet Mons d C 11 n Var Mons d x 6 y Api Mons d R 3 n Uretsete Mons d E,F 0 n Furki Tholus d R,F -10 n Kunapipi Mons d C,F,R -12 y Tuulikki Mons d E,R -26 n Nyx Mons ce E 65 y Chloris Mons ce x R,F 16 n Nagavonyi Corona ce F -11 y Nefertiti c C 24 n Kokyanwuti Mons c C,E,F 3 n Nzambi Corona c C,E,R,L 2 n Unnamed c C,R,F -4 n Uti Hiata c E,F,R -5 n Kunhild c C,R -10 n Atanua Mons c x -11 n Mielikki Mons c C,L,E,F -12 n Atira Mons c E,R -14 n Ituana Corona c E,C,R -18 n Latitude is degrees north, longitude is degrees east. Type characterizes the shape as having a broad domical shape with no significant topographic depression (d), containing a broad central depression (c), or containing a broad central depression with volcanic edifices of significant size in their interiors (ce). Height is given relative to the surrounding terrain. Width is given as both the width at half of the maximum height of the volcano and at an elevation 500 m above the surrounding terrain. Postdating features unrelated to the volcano are impact craters (C), embayment by flows from another volcano (E), wrinkle ridges (R), fractures (F), and lineaments (L) of unknown origin. Isostatic anomaly is calculated assuming a 30-km compensation depth and a crustal density of 2900 kg m -3, and last column indicates whether the anomaly is clearly separable and associated with the volcano.

30 30 Table 2. Dimensions of central depressed regions Volcano Name Depth (km) Width (km) Nefertiti Kokyanwuti Mons Nzambi Corona Unnamed (38N, 323E) Uti Hiata Kunhild Atanua Mons Mielikki Mons Atira Mons Ituana Corona Table 3. Summary statistics of isostatic anomalies Mean (mgal) Std. deviation Median Domical (d), all Domical without postdating features Domical with postdating features Depressed central region (c ) Edifice fills depression (ce)

31 31 Table 4. Summation of results for 2-layer inversions Upper Surface Lower Surface Name Type Postdating 10^6 kg/m^2 separable 10^6 kg/m^2 separable Ushas Mons d x 0.5 n -7.0 y Sif Mons d x 0.5 n -5.1 y Tepev Mons d x 0.7 n -7.1 y Dzalarhons Mons d x 0.6 n -5.5 y Idunn Mons d x -1.8 y -1.9 n Unnamed (71.5N, 256E) d F,R -3.9 y -1.7 y Innini Mons d x -0.6 y -3.8 y Hathor Mons d x -2.1 y -4.1 y Unnamed (2.5N, 45.5E) d x -0.6 n -1.9 y Renpet Mons d C -1.7 y 0.3 n Var Mons d x -0.7 n -0.1 y Api Mons d R 1.1 y -1.0 n Uretsete Mons d E,F -1.9 y 0.1 n Furki Tholus d R,F -0.4 y 1.3 n Kunapipi Mons d C,F,R -3.4 y 2.2 n Tuulikki Mons d E,R -2.5 y 2.5 y Nyx Mons ce E -1.4 y -5.9 y Chloris Mons ce R,F -0.6 n -1.8 n Nagavonyi Corona ce F -2.4 y 0.0 n Nefertiti c C -1.1 y -2.8 n Kokyanwuti Mons c C,E,F -0.4 y -0.3 n Nzambi Corona c C,E,R,L -1.8 y 0.3 n Unnamed (38N, 323E) c C,R,F -0.6 y 1.9 n Uti Hiata c E,F,R -2.3 y 1.5 y Kunhild c C,R -3.2 y 1.8 y Atanua Mons c x -2.3 y 2.0 y Mielikki Mons c C,L,E -3.8 y 1.0 n Atira Mons c E,R -2.4 y 3.4 n Ituana Corona c E,C,R 0.0 y 2.7 y Volcanoes are ordered as in Table 1. Surface density values are located over the center of the volcano in units of 10 6 kg m -2. Columns labeled "separable" indicate whether a surface density anomaly is centered over the volcano and clearly separable from nearby features.

32 32 Table 5. Summation of spatio-spectral localizations RMS gravity (mgal) Admittance (mgal/km) Correlation Name Type Postdating Degree Ushas Mons d x Sif Mons d x Tepev Mons d x Dzalarhons Mons d x Idunn Mons d x Unnamed (71.5N, 256E) d F,R Innini Mons d x Hathor Mons d x Unnamed (2.5N, 45.5E) d x Renpet Mons d C Var Mons d x Api Mons d R Uretsete Mons d E,F Furki Tholus d R,F Kunapipi Mons d C,F,R Tuulikki Mons d E,R Nyx Mons ce E Chloris Mons ce R,F Nagavonyi Corona ce F Nefertiti c C Kokyanwuti Mons c C,E,F Nzambi Corona c C,E,R,L Unnamed (38N, 323E) c C,R,F Uti Hiata c E,F,R Kunhild c C,R Atanua Mons c x Mielikki Mons c C,L,E Atira Mons c E,R Ituana Corona c E,C,R

33 33 Figure 1. Topography (km) N- S profile (km) Topography (km) W-E profile (km) a. Sif Mons

34 34 Fig 1 cont. Topography (km) N- S profile (km) Topography (km) W-E profile (km) b. Innini Mons

35 35 Fig 1 cont. Topography (km) N- S profile (km) Topography (km) W-E profile (km) c. Kunapipi Mons

36 36 Fig 1 cont. Topography (km) N- S profile (km) Topography (km) W-E profile (km) d. Chloris Mons

37 37 Fig 1 cont. Topography (km) N- S profile (km) Topography (km) W-E profile (km) e. Kokyanwuti Mons

38 38 Fig 1 cont. Topography (km) N- S profile (km) Topography (km) W-E profile (km) f. Atanua Mons

39 39 Figure 2. a. Isostatic anom (mgals) N-S profile (km) Isostatic anom (mgals) W-E profile (km)

40 40 Fig 2 cont. b. Isostatic anom (mgals) N- S profile (km) Isostatic anom (mgals) W-E profile (km)

41 Figure 3. 41

42 42 Figure 4. a. Ushas Mons

43 43 Fig. 4 cont. b. Hathor Mons and Innini Mons.

44 44 Fig. 4 cont. c. Tuulikki Mons

45 45 Fig 4 cont. d. Atanua Mons

46 Figure 5. 46

47 Figure 6. 47

48 Figure 7. 48

49 Figure 8. 49