4.1. Distribution and Characteristics of Impact Melt-Rich Lithologies
 Probable impact melt-rich lithologies were identified using well-established morphologic criteria. These include smooth, nearly level surfaces of deposits filling local depressions, cooling and tension cracks, flow features (e.g., flow lobes, leveed channels, flow lineations), cracked veneer over irregular surfaces, and typically fairly low albedo [Shoemaker et al., 1968; Guest, 1973; Howard and Wilshire, 1975; Hawke and Head, 1977; Bray et al., 2010; see also Strom and Fielder, 1970].
 As expected, the largest continuous concentrations of melt-rich material can be found on the floor of Kepler. Photogeologically, the melt-rich material forms two distinct units that grade into each other. The smooth floor material unit (Figure 5) forms a level, smooth, fractured, and fairly dark unit mainly in the northern and western parts of the crater floor, covering altogether about 35 km2 (Figure 4). Where stereo imagery is available (mainly the eastern part of the crater), the fractures in most cases are associated with topographic gradients (Figure 5). This relationship applies to all generally more or less level melt-rich units.
Figure 5. Perspective view across the southeastern crater floor. The typical fairly low-albedo smooth floor unit at the lower right grades into brighter and more hummocky floor material toward upper left. The bold black-in-white arrow points to a deep collapse pit (see Figure 15c), and the narrow white arrows mark a step-like change in topography, which without stereo imagery would appear merely as fractures on the surface. The inset shows the same feature from a slightly different perspective. The image strip is about 1 km wide, and the deep pit is located at about 7.95°N, 322.16°E. Parts of NACs M135439816LE and M135433031RE.
Download figure to PowerPoint
 The hummocky floor material unit (Figures 5–7) is more extensive than the smooth floor unit, covering ∼78 km2 (Figure 4). It is reminiscent of the smooth floor unit, but as the name suggests, the surface of the unit is not level. Hummocks of higher albedo material are typically tens to a few hundred meters in diameter, and the material between these individual hummocks is similar to the smooth floor material. However, stereo imagery reveals the hummocky floor unit has also longer wavelength topographic irregularity, quite distinct from the smooth floor unit. In many places, best seen in the eastern part of the crater, there is a distinct topographic bench (Figure 7) near the contact between the hummocky floor and the lowermost part of the terrace zone, with heavy fracturing of the melt-rich floor material where the topographic gradient is highest.
Figure 6. Perspective view toward south across the northeastern crater floor and the easternmost minor peaks of the central uplift complex. The fractured hummocky floor unit in the foreground onlaps the central uplift material. This part of the uplift rises ∼85 m above the surrounding floor, with fairly shallow slopes reaching ∼10°–12°. Note that the central uplift material is covered by a large number of boulders ∼2–12 m in diameter. The image strip is ∼1 km wide, and the center of the most prominent part of the central uplift in the foreground is located at about 8.15°N, 322.12°E. Parts of NACs M135439816LE and M135433031RE.
Download figure to PowerPoint
Figure 7. A heavily fractured topographic bench (middle) between lower eastern terrace zone (right) and subsided hummocky floor (left). The center of the image is located at about 8.12°N, 322.26°E. A part of NAC M104755664LE.
Download figure to PowerPoint
 In addition to fractures, collapse pits some tens of meters in diameter are seen in the hummocky floor unit. Prominent ledges in some of the fractures and pits offer possibilities for estimating the thickness of the visible melt-rich material. An apparently deep pit (Figure 5) near the southern edge of the floor shows signs of granular flows on its eastern wall, implying the maximum slope is at the angle of repose, or ∼31° [e.g., Quaide and Oberbeck, 1968]. Assuming the wall slope is 31°, the shadow length gives a maximum depth of ∼30 m for the pit. Shadow measurements of ledges of melt-like material in other pits and fractures, where slopes give the impression of being shallower than the angle of repose, imply that the thickness of the directly observable melt-rich material in both hummocky and smooth floor units is commonly on the order of 5–10 m.
 Assuming the smooth floor unit has 80 volume-% melt and a thickness of 10 m, and the hummocky floor material is 5 m thick and composed of 50 vol.-% melt, the total combined observed melt rock volume is only ∼0.5 km3. However, as the observed ledges form only the uppermost crust of the melt-rich units, and some melt also must have flowed down to the fractures in the crater floor, the amount of impact melt derived from this rough approximation is a minimum estimate.
 Terrace ponds are isolated and relatively small patches of melt-rich material on top of the terrace zone and massive slump material. Definite flow features related to terrace ponds are rarely observed (although Kaguya TC images show flows from a pond on the southeastern terrace zone to the crater floor). However, there are characteristics indirectly indicating the flow of material, such as narrow radial depressions (Figure 8) interpreted as erosional channels caused by impact melt-rich material flowing downward and then pooling in local depressions (similar to, e.g., Aristarchus [Mustard et al., 2011] or Thales [Bray et al., 2010]). In the geomorphologic sketch map (Figure 4) only the larger fractured pools are included, but probably many of the smaller patches of dark material are similar ponds of melt-rich rock.
Figure 8. Perspective view across the eastern terrace zone and melt ponds. Impact melt-rich material was probably draped over most of the terrace zone, and it flowed down in narrow radial channels to form ponds in local depressions, which fractured when cooled. The arrows indicate probable flow directions. The width of the image in the middle section is about 2.7 km, and the center of the fractured pond is located at about 7.84°N, 322.31°E. Parts of NACs M104762819LE and M104755664LE.
Download figure to PowerPoint
 Wall lobes are lobate, mutually overlapping sheets of smooth, usually fairly dark material on the upper crater wall (Figure 9). In the current (until 15 March 2011) NAC coverage, they are best seen on the western wall, and are much harder to distinguish on other parts of the crater wall. Shadow length measurements imply that their thickness varies from ∼0.8 m to perhaps slightly more than 5 m. Due to their mutually overlapping layered nature, the length of the individual lobes is hard to define, but seems to be typically a couple of hundred meters. Instead of individual lobes, the wall lobes should be considered as a zone of the upper crater wall, where this characteristic flow morphology of melt-rich material occurs. It is noteworthy that the “stubby” and sheet-like nature of the wall lobes clearly distinguishes them from long, narrow, digitate clastic debris flows that commonly occur on the crater wall, partially covering some of the wall lobes.
Figure 9. Wall lobes of melt-rich material on the upper western wall. Note how melt flows down from the rim crest (upper edge of the image, i.e., the center of the crater is toward the bottom of the page) forming overlapping lobes. Individual lobes are typically ∼2 m thick. The center of the image is located at about 8.10°N, 321.53°E. A part of NAC M104769976RE.
Download figure to PowerPoint
 Rim veneer covers most of the circumference of Kepler (Figure 4). It is smooth but typically highly fractured, and gives the impression of forming a thin coating on the rim (Figure 10). The veneer often seems to be the source for the wall lobes, and it grades into the exterior ponds, although distinct flow channels (Figure 10) are rarely seen. In the geomorphologic sketch map (Figure 4) ponded material that appears to be continuous with the veneer has been included in the rim veneer unit. The leveed flow in Figure 10 exhibits flow fronts (black-in-red arrows) on the side of the main flow, where the flow was diverted toward a small local depression. The flow and the associated pond are also fractured perpendicular to the flow direction.
Figure 10. The morphology of melt-rich deposits on the western rim crest and flank. (a) The rim crest (on the right, i.e., the center of the crater is to the right) is covered by a thin melt veneer, which is typically heavily fractured. (b) Melt flowed downslope (∼5° slope) to the left, partially in a leveed flow channel (yellow arrows in Figure 10c), and forms a dark, level, and slightly fractured pond. Fractures on the leveed flow as well as on the pond are perpendicular to the flow direction. (c) The black dotted lines delineate the most prominent patches of ponded and fractured melt. Note the overlapping flow fronts (black-in-red arrows ) on the side of the flow where the melt was diverted toward a small local depression, indicating that the flow of the melt was a relatively long-duration process of multiple stages. The white box in Figure 10a shows the location of Figures 10b and 10c. The center of the leveed flow in Figure 10b is located at about 8.12°N, 321.45°E. Parts of NACs M104769976RE (Figure 10a) and M148414768RE (Figures 10b and 10c).
Download figure to PowerPoint
 The larger areas mapped as rim veneer on the southwest, southeast, and north sides of the crater (dotted in Figure 4) are difficult to map, as the surface is covered with many small patches of slightly darker, apparently ponded but poorly delineated material. There are some indications of flow, but the characteristic fracturing or ledges of the rim veneer are generally lacking. We interpret these areas as being covered by a very thin veneer of impact melt-rich material. Thus, although the mapped areas are large (and some data imply that, e.g., in the northeast they should perhaps be even larger), the volume of the melt rock in these deposits is very small.
 Exterior ponds (Figure 10) with their generally level, slightly fractured dark surfaces are similar in appearance to the terrace ponds, and many of them are closely related to the rim veneer. The most striking feature of the exterior ponds is their non-symmetric distribution around Kepler (Figure 4). There are large ponds north of the crater and numerous small ponds close to the northwest rim crest. Some small and poorly defined ponds are present south and particularly southeast of the crater (dotted in Figure 4), but in general they are rare or lacking on the eastern and southern to southwestern sides of the crater. As the larger ponds can readily be identified in Kaguya TC and MI or LRO WAC imagery, the asymmetric pond distribution is not due to uneven high-resolution NAC coverage or unfavorable illumination geometry. Instead, it appears to be a true phenomenon related to the initial distribution of the melt-rich material.
4.2. Melt-Poor Interior Deposits
 Geologic units with a low apparent melt content were also mapped. These include the central uplift, terrace zone, massive slumps, and the exposed wall. All of these, and particularly the central uplift and massive slump units, include small deposits of melt-rich material that, mostly due to their limited size and very poorly discernible boundaries, are not depicted in Figure 4.
 The central uplift material in Kepler does not consist of a single peak, but instead forms a cluster of rounded hills with typical slopes of ∼15°, although steeper slopes occur locally. The most prominent part of the central peak cluster is slightly north from the crater center, and the majority of the central uplift forms a crudely crescent-shaped structure open to the southeast (Figure 4). The highest of the hills crossed by LOLA profiles rise ∼550 m above the surrounding crater floor. This closely matches the predictions from the scaling relationships by Wood  and Hale and Grieve , but is ∼65 m higher than Wood and Andersson's  and ∼150 m lower than Pike's  scaling laws predict. The surface of central uplift material has a higher albedo than the surrounding materials and typically has more boulders on the surface than the surrounding crater floor (Figure 6). No layering or other structures can be seen in the central uplift.
 Ledges of material with melt-rich characteristics can be seen onlapping the lower parts of the uplift, and melt lithologies are also present in some of the troughs between the hills, making the transition to floor material sometimes difficult to discern. The easternmost separated and fairly low (∼85 m above the surrounding floor) parts of the central uplift (Figures 3, 4 and 6) may instead be extremely large bedrock clasts embedded in the hummocky floor material.
 The collapse of the wall of Kepler resulted in two types of materials, classified as terrace zone material of moderate albedo and massive slump material. However, in most cases the difference between the two is not clear-cut. In general, the surface expression of the terrace zone material is smoother than that of the fairly rugged massive slumps. Importantly, the terrace zone material hosts linear troughs that run roughly parallel to the crater rim, interpreted to be normal faults responsible for the terrace- or stair-like appearance of some parts of the crater wall (Figures 3 and 8), typical of a crater modification zone. Terrace ponds of melt-rich material can be found in many of the troughs (Figure 8). The blocks defining those troughs appear to have moved down fault planes as coherent blocks, rather than as rock avalanches, which is typical for the massive slump material. A sketch of the main faults can be seen in Figure 3. In addition to the faults and the terrace ponds, other prominent features of the terrace zone are narrow radial grooves running between terraces, or from a terrace to the crater floor (Figure 8).
 The faults are mainly present on the western half of the crater wall (Figure 3). Based on LOLA point data, approximate minimum vertical displacements (throws) of these terrace-forming normal faults vary from ∼1000–1600 m for the fault closest to the final crater rim, whereas the inner faults have vertical displacements of only ∼40–60 m. The southeastern wall has few faults, and it seems that massive slumping covers at least some probable faults on the eastern and northwest terrace zones. The extensive image coverage shows that the observed east/west difference in the fault distribution is not influenced by lighting geometry, but it is a true distinction in the structure of the crater wall.
 The formation of massive slumps is a major contributor to the scalloped, partially polygonal plan view of Kepler (see Öhman  for a discussion). Terrace ponds and other smaller and poorly defined (hence, not depicted in Figure 4) occurrences of melt-rich material are in many places present on top of the massive slump material. The massive slump material is best exposed on the north-northeast lower wall of the crater (Figure 3), whereas the slump deposits on the southern wall are minor, and their nature can properly be seen only in images having a low solar angle. The massive slump material has an irregular, sometimes grooved surface, and particularly on the lower north-northeast wall it can be seen to form a large bulge, extending to the crater floor. On the south-southwest wall of the crater, the upper wall hosts a small bulge of similarly textured massive slump material, below which is the terrace zone. However, below the terrace zone there is another slump-like feature, which protrudes far onto the crater floor and central uplift region. Despite some morphological differences, this material has been classified as massive slump material in Figure 4.
 Relatively high-albedo material comprising the exposed wall unit occurs above the terrace zone or the massive slump material. The best exposures of this material are on the northern and particularly the northwestern wall (Figure 11). Although spatially fairly limited, a characteristic feature of this unit is the presence of distinct horizontal layering (Figure 11; also observed by Zanetti et al. [2011a]), traces of which can be seen throughout the upper crater wall. This layering is visible in a Kaguya TC morning image (Figure 3) as well as in Apollo 12 oblique imagery, but NAC images provide the most detailed view. The layers are visible partially due to their tendency to form steeper parts of the wall, thus casting shadows. However, based on scarce LOLA point data, even the steepest slopes are only ∼40°–47°. Another factor in the visibility of layers is their color. Particularly in smaller exposures on the southern and northeastern walls, the darker color of the layers is emphasized by debris flows of dark material originating from the layered rocks. Yet, some of the outcrops of layered rocks have a brighter albedo than the surrounding crater wall. A more detailed assessment about the origin of this variation, and the relative importance of factors like composition, maturity, and illumination geometry lies beyond the scope of this study.
Figure 11. Layering exposed at the upper northwestern wall. The most prominent layered section is ∼65 m high, and has a slope of ∼45°. Note also the less pronounced section of layered material near the crater rim. The top of these layers is ∼400 m above the lowermost layers seen in the image. The center of the image is located at about 8.53°N, 321.69°E. A part of NAC M107128381RE.
Download figure to PowerPoint
 The most prominent layered section on the northwestern wall is ∼100 m high, the best exposed part being ∼65 m (Figure 11). Less striking layers are present ∼400 m higher, close to the rim crest (Figure 11). The elevation of the layers varies somewhat in different parts of the crater wall, but the best exposed sections in the northwest are ∼760–860 m below the lunar reference elevation, i.e., several hundred meters higher than the surrounding maria (Figure 2). The exposed section of the layers in the northern and northwestern wall is ∼2500 m in length, with ∼1200 m of practically continuous outcrop. Individual layers have a maximum thickness of ∼3–5 m. There are also a few boulders in different parts of the terrace zone which display weak traces of probable layering.
 The upper boundary of the exposed wall material is often difficult to decipher, as the illumination geometry strongly affects differentiating the wall lobes from the exposed wall (particularly other than the distinctly layered parts). It should also be noted that not all of the material classified as exposed wall material in Figure 4 is literally exposed bedrock. Commonly it is covered by what appear to be a thin scree deposit and long, narrow, digitate clastic debris flows forming talus deposits. In addition, debris fans deposited on the units below are common on the lowermost part of the exposed wall.
4.4. Physical Properties of Melt-Rich Lithologies
 We estimated the physical properties of the impact melt, particularly its yield strength (τy) and bulk effective viscosity (ηe). The yield strength (in Pa) estimates were based on shadow length measurements (in meters) of the wall lobe thicknesses [Hulme, 1974; Moore et al., 1978] using equation
Yield strength based on lobe widths [Moore et al., 1978] and levee widths [Hulme, 1974; Moore et al., 1978] were calculated with equations (2) and (3), respectively:
In equations (1)–(3), ρ is the density of the melt (taken, based on noritic composition and MELTS calculations, to be 2665 kg/m3 at 1200°C), g is the surface acceleration of gravity (1.62 m/s2), θ is the downhill slope in degrees, H is the thickness of the lobe (in meters), wf is the width of the flow, and wb is the width of the levee (∼17 m in the most prominent case).
 Yield strength measurements based on flow thicknesses (equation (1)) were made from the upper western wall, where wall lobes are ubiquitous and favorably illuminated in NAC M104769976RE (1.2 m/px; Figure 9). The slope of the upper wall is ∼29°. The measured lobe thicknesses are generally on the order of a few meters, varying from ∼0.8 m to possibly slightly more than 5 m.
 Levee widths were measured only from one channel close to the lobes (visible on the same NAC image, as well as on M148414768RE (0.6 m/px)), but situated on the outside crater rim flank (Figure 10). The slope of this part of the rim flank is ∼4°–5°.
 Given the assumptions and measurements described above, equation (1) implies that the yield strength of the wall lobes is typically ∼4–5 kPa (Figure 13), varying from ∼2 to ∼10 kPa. For the same lobes, equation (2) gives unrealistically low strengths. This is due to the very large width of the wall lobes, i.e., equation (2) is applicable to more channelized flows, not to a wide zone of flowing melt. The yield strength of the leveed flow is somewhat lower than that of the wall lobes, being ∼1 kPa, in agreement with the melt flow morphology. The wall lobes with higher yield strength are relatively short (typically on the order of a 100 m, although their layered nature makes it difficult to define the starting points of individual lobes) and wide, whereas the leveed flow is longer (altogether ∼1700 m) and narrow. Both seem to have their origin in the melt layer deposited on the rim, now seen as thin veneer covering most of the rim crest (Figure 4). The change in temperature or composition has only a relatively minor effect on the yield strength (or the density) of the liquid phase (Figure 13).
Figure 13. Calculated yield strength (kPa, dots) and liquid density (kg/m3, triangles) of a cooling 2 m thick impact melt flow on a 29° slope (similar to wall lobes), having an average anorthosite (reddish curves on the left), norite (blueish curves in the middle), and Apollo 12 basalt (greenish curves on the right) composition. The kinks in the curves result from the melt reaching liquidus temperature and the consequent crystallization of plagioclase and olivine.
Download figure to PowerPoint
 Yield strength enables estimation of the flow thicknesses for cases where it cannot be measured based on shadow lengths. According to Hulme , the flow thickness (in meters) at the surface of the centerline of the flow (H0) is given by
and the critical thickness (Hc), which defines the minimum thickness (in meters) of the flow for downhill movement to occur, by
or, for a leveed flow [Hulme and Fielder, 1977], by
where θG is the slope expressed as a gradient (i.e., the difference in elevation divided by the distance, θG being equal to tan θ). The width of the leveed flow (Figure 10) on the western flank is not easy to measure, but it is ∼100 m. Assuming a yield strength of 1 kPa, according to equation (4) the thickness of the flow at the centerline is ∼5 m. Conversely, assuming the leveed flow is ∼5 m thick, equation (2) gives a yield strength of ∼1 kPa for a ∼100 m wide flow. For the critical thickness (Hc), assuming a measured slope (θ) of 5° and therefore a gradient (θG) of 0.875 for the leveed flow (wb = 17 m), equations (5) and (6) give approximately Hc = 2.6 m and Hc = 3.0 m, respectively. The good agreement between the results from equations (4), (5), and (6) implies that at least for this leveed flow, a thickness on the order of ∼5 m and yield strength of ∼1 kPa are reasonable.
 The viscosities calculated with MELTS subroutine Adiabat 1_ph assume that no clasts or crystals are entrained in the melt. However, impact melt rocks typically consist of melt matrix with a variable amount of relic lithic clasts, and crystals growing from the cooling melt. Therefore, we calculated the effective viscosity (ηe) of the bulk melt by applying equations relating the change in relative viscosity to particle (clast or crystal) concentration [e.g., Petford, 2009]. In addition to particle concentration, particle shapes also have a significant effect on the viscosity [e.g., Mueller et al., 2011]. However, compared to the large uncertainties in estimating the concentration of particles and their effect on the impact melt viscosity, the effect of particle shapes is probably minor, and was not considered in the following calculations. The equations for effective viscosity we applied are
where ηl is the viscosity of the liquid, Φ is the volume fraction of solid particles, and Φmax is the maximum packing fraction. Equation (7) is the classic Einstein–Roscoe (referred to as E–R) equation [Einstein, 1906, 1911; Roscoe, 1952; Shaw, 1965] and it is mainly applicable for melts where particles comprise less than about 20–30 vol.-% of the total volume [e.g., Petford, 2009]. Equation (8) is suitable for higher concentrations of solid particles, at least up to ∼55 vol.-% (referred to as G–N–A; Gay et al. ; Pinkerton and Stevenson ; Williams et al. ). Equation (9) (referred to as K–D; Krieger and Dougherty ) is another well-known equation [e.g., Stickel and Powell, 2005; Petford, 2009] used for higher particle concentrations. It is used here as an additional estimate of the effective viscosity, because in lower particle concentrations and temperatures it predicts values that are intermediate between those given by equations (7) and (8), and in higher concentrations it predicts notably lower viscosities than equation (8). The maximum packing fraction (Φmax) varies between ∼0.6 and 0.74 [Petford, 2009], but compared to the effects of the other unknown parameters, its influence is minor. In equations (8) and (9), we use Φmax = 0.64 for disordered dense packing [Petford, 2009].
 As discussed by Stickel and Powell  and Petford , for example, different effective viscosity equations generally agree with each other in low particle concentrations, but they differ substantially at higher concentrations. The equations used here were chosen to portray those uncertainties, thus providing a reasonable estimate for the possible range of effective viscosities encountered in Kepler impact melts.
 The results (Figure 14) imply that a superheated (∼1700°C) melt without clasts has a viscosity of ∼5 Poise ( = 0.5 Pa·s), whereas melt entraining 30 vol.-% clasts has a higher viscosity of ∼19 Poise (equation (7)). At a lower melt temperature (∼1275°C) closer to the liquidus, the viscosities are ∼120 and ∼430 Poise for a clast-free melt and a melt including 30 vol.-% clasts, respectively. At high temperatures, equations (7) (E–R) and (8) (G–N–A) produce similar results in the 30 vol.-% clasts case, whereas at lower temperatures, equation (9) (K–D) is a closer match to equation (8) (G–N–A). For 50 vol.-% clast contents, equation (8) (G–N–A) gives very high viscosities, particularly for temperatures below the liquidus. For comparison and setting reasonable minimum and maximum viscosity estimates for pure melt in Kepler, Figure 14 also includes clast-free viscosity curves for the average composition of basalts from the Apollo 12 landing site, as well as for average anorthosite (Table 1) [Haskin and Warren, 1991]. These compositions can be considered the end-members of the possible target lithologies, and, therefore, also of the impact melt.
Figure 14. Calculated impact melt viscosity as a function of temperature. The lines without symbols represent clast-free melts from possible target compositions at Kepler, with average Apollo 12 basalts and average anorthosite as possible end-members, and average norite probably closer to the bulk target composition. The lines with symbols are based on the average norite composition with 30 (triangles) or 50 volume-% (crosses) of clasts in the melt. The Einstein–Roscoe (E–R, equation (7)) equation is suitable for clasts contents up to 20–30 volume-%, whereas the Krieger–Dougherty (K–D, equation (9)) and Gay–Nelson–Armstrong (G–N–A, equation (8)) equations are thought to give more reliable estimates for higher clast contents. The kinks in the curves result from the melt reaching liquidus temperature and the subsequent crystallization of plagioclase and olivine. See text for further details.
Download figure to PowerPoint
 According to Moore and Ackerman  and Wilson and Head , viscosity (in Pa·s) and yield strength (in Pa) are related by an empirical formula
where Q = 6·10−4. For the leveed flow (τy = ∼1 kPa; Figure 10), equation (10) results in a viscosity estimate of ∼9.5·104 Poise (9500 Pa·s), and for the wall lobes (τy = ∼4.5 kPa; Figure 9) ∼3.5·106 Poise (3.5·105 Pa·s). In our model calculations with a noritic composition (Figure 14), such viscosity values are applicable to melts below the liquidus with clast contents of ∼30–55 vol.-% (equation (9), K–D). It is worth noting, however, that the E–R equation (7) does not predict such high viscosities at all, and using the G–N–A equation (8) with ∼50 vol.-% clasts contents, these viscosity values are encountered even above the liquidus.
 In addition to effective viscosity and yield strength, we estimated the flow velocities of the impact melt-rich lithologies. The mean flow velocity (U) of a laminar flow of a Newtonian liquid is given by the Jeffreys  equation
and the velocity of a turbulent flow by an equation from Wilson and Mouginis-Mark 
where f is a friction factor, taken by Wilson and Mouginis-Mark  to be ∼0.01. Note that in a turbulent flow the viscosity is irrelevant unless it is significantly higher than about 10000 Poise [Wilson and Mouginis-Mark, 2001]. In the present study, the limit of significantly higher viscosities than 10000 Poise is generally reached at temperatures below ∼1250°C, i.e., in melts that are starting to crystallize, although this is strongly dependent on the clast content and particularly on the equation used (Figure 14).
 As an example, we assume a 2 m thick flow on a 25° slope (similar to the wall lobes on the upper part of the crater wall; Figure 9) with 30 vol.-% clasts at a temperature of 1200°C. Such a melt is starting to crystallize and has a viscosity of ∼2200 Poise (equation (7) and Figure 14). If it behaved as a Newtonian fluid and flowed in a laminar fashion (equation (11)), it would have a velocity of ∼11 m/s (∼40 km/h). However, impact melts are Bingham fluids, i.e., a yield stress must be applied before any flow can take place. Depending on, for example, local topography, impact melts also at least partially are likely to be turbulent, so a velocity of ∼17 m/s (∼60 km/h; equation (12)) is a better estimate. In this case (near liquidus), equations (11) and (12) are in a good agreement, although at higher temperatures (and, therefore, lower viscosities) the velocity estimates provided by the two equations can differ by two orders of magnitude.
 To first order, the individual wall lobes may have been formed within tens of seconds from the time the melt was emplaced on the rim. The leveed flow would have initially formed in a few minutes, but the flow likely continued for a slightly longer time to produce the exterior pond at the end of the flow. As can be seen in Figure 10, the melt flowed in several episodes, as there are several flow fronts on the side of the main flow where a small local depression diverted part of the flow. Similarly, the wall lobes overlap each other (Figure 9), indicating that the formation of the entire zone of wall lobes was a multiphase process.