SECOND WORKSHOP ON MARS VALLEY NETWORKS 51

COMPARING DISSECTION PATTERNS OF MARS AND EARTH: PALEOCLIMATE IMPLICATIONS. W. Luo1 and T. F. Stepinski2, 1Department of Geography, Northern Illinois University, DeKalb, IL 60115, [email protected], 2Lunar and Planetary Institute, 3600 Bay Area Blvd., Houston, TX 77058, [email protected].

Introduction: Martian valley networks (hereafter models (DEMs) of planetary surface. In this paper we referred to as VN), discovered in 1971 by the Mariner start to address the problem of VN origin from the 9 spacecraft, are geomorphic features on Mars exhibit- global perspective by deriving the planet-wide map of ing some visual resemblance to terrestrial river sys- dissection on Mars and comparing it to a similar map tems. The VN are present mostly in the Martian high- derived for the Earth. We use global-scale, similar lands and date to the Noachian. They point to a possi- resolution DEMs of Earth and Mars to extract valleys bility that Noachian Mars was warmer and wetter than over the entire planetary surfaces. These data sets are present-day Mars, perhaps capable of maintaining pre- used to derive maps of dissection density and their cipitation and some form of life. A geomorphic analy- zonal statistics. sis of VN should, in principle, be able to establish a predominant style of erosion leading to their forma- Data sources: The DEM data for Mars comes tion. Geomorphic features indicative of runoff erosion from the Mars Orbiter Laser Altimeter (MOLA) Mis- signals precipitation and a warmer climate whereas sion Experiment Gridded Data Records (MEGDR) features indicative of erosion by groundwater sapping [20]. The MEGDR DEM has a spatial resolution of allow for cold and dry conditions during formation of 1/128 degrees/pixel or about 463 m/pixel at the equa- VN. Alas, the geomorphic evidence is inconclusive tor. Note that MEGDR is an interpolated DEM, as an and the origin of VN is still a topic of active research. across-the-track spacing of the original MOLA mea- Some features, such as branching, dendritic pat- surements is ~ 1000 m at the equator [21] so some of terns, origin near dividing ridges [1,2,3,4], consistency the grid cells lack direct measurement values. This with crater degradation [5,6], and significant erosion notwithstanding, the MEGDR is suitable for our study, requiring aquifer recharge [7,8,9], favor runoff ero- as we are only interested in deriving large-scale prop- sion. Other features, such as widely spaced tributaries erties of Martian dissection for the purpose of statis- with alcove-like terminations [10,11], constant valley tical analysis. Only the portion of the MEGDR located width downstream [12], short, stubby tributaries [13], between 50˚ N and 70˚ S latitudes is used because our flat longitudinal profiles [14], and U-shaped cross- focus is on the Noachian terrain. sections, favor erosion by groundwater sapping. A For the Earth we use the Global Land One-km common feature of all aforementioned studies is their Base Elevation (GLOBE) 30-sec/pixel (~ 1 km/pixel at reliance on detailed examination of individual valley the equator) DEM developed by the National Geo- systems. physical Data Center (NGDS, 2008), which is the best Several hypothesis pertaining to the origin of VN available 30-arc-second global digital elevation data were proposed in an attempt to reconcile the conflict- set, compiled from various sources. ing geomorphic evidence: episodic melting of snow accumulating during high obliquity epochs [15], high Methods: Automated delineation of valleys and/or intensity rare storms (like in the Atacama desert, see streams from a DEM is a rather standard task in terre- [16]), and episodic, multi-year intense rainfall events strial hydrology. The established methods follow the due to basin-scale impacts or intense volcanic erup- flow along local directions of the steepest descent (i.e., tions [17]. they are based on the idea that water flows downhill). In order to test these hypothesis, a global analysis Streams extracted by the steepest descent method tend of dissection patterns must supplement the detailed to be spatially uniform; the method is not capable of local analysis of VN geomorphic features. This re- mapping spatial variations of dissection [22,23] and it quires a global, sufficiently detailed map of VN. How- should not be used to map surfaces where non-uniform ever, the only currently available global map of VN dissection pattern is observed and/or expected. This [18] is based on the Viking Orbiter images and con- includes the Noachian terrain on Mars because VN tains only ~ 800 networks. The lack of more detailed lack spatial integrity; dissected areas are commonly global maps is due to the cost of their acquisitions by separated from each other by undissected areas of sim- means of visual interpretation of images. Recent ad- ilar or bigger size. Even on Earth, where dissection vances in machine extraction of valleys from topo- tends to be uniform on small spatial scales, the steepest graphic data [19] make it feasible to acquire global descent method should not be used for assessing the maps of VN by computer parsing of digital elevation SECOND WORKSHOP ON MARS VALLEY NETWORKS 52

density of dissection on spatial scales on which its groundwater sapping as a viable mechanism, but get- non-uniformity becomes evident [23] ting the right answer would be challenging without In this work we use a terrain morphology-based ex- additional data. As it happens, high values of DD on traction method [19] that identifies concave upward Earth are found in the orogenic belts where high topo- (U-shape) topographic features as valleys. The con- graphic relief (caused by the orogenic uplift) combined cave upward features are associated with the cells in a with the rainfall-fed runoff erosion give rise to the DEM having positive curvature. This method has been high values of DD. Thus, the pattern of high DD re- previously applied to extract VN in the entire MC22 gions on Earth is explained by combination of an ero- [24] and MC19 [25] quadrangles, yielding maps of VN sional style (runoff) with a separate physical pheno- that are much more detailed than the Carr map. The menon (tectonics). method was also applied to a large site in the Cascade On Mars, the distribution of DD is directly oppo- Range, Oregon, USA [23], demonstrating its ability to site of what is found on Earth - rather homogeneous on identify non-uniform dissection in the terrestrial con- large spatial scale, but very inhomogeneous on smaller text. For the present study we have applied the mor- scales. As indicated by latitudinal zonal statistics, the phology-based extraction technique to the MEGDR average values of DD are highest in the equatorial re- and GLOBE grids. Masks were applied to the resulting gions, north of ~ 30 S˚, and decrease rapidly south of data sets of valleys to filter out ocean floor on Earth ~ 30 S˚ latitude. Although the mean values of DD are and non-Noachian terrain on Mars. A maps of dissec- relatively high throughout the equatorial belt, there is a tion density, DD, is calculated by applying a circular strong local variability. This reflects the well-known moving window to the corresponding map of valleys. lack of spatial integration of VN. No systematic varia- A given cell is assigned a value of DD corresponding tion of an average value of DD with elevation is ob- to a ratio of the total length of valleys in the window to served on Mars. On the other hand, a systematic de- the area of the window. The results presented here crease of the value of DD with slope is detected. We were obtained using a window with a radius of 100 have also calculated a relationship between zonal mean km, but the results are not overly sensitive to the size of DD and the crater density on Mars. The crater den- of the window. sity was derived from the Catalog of Large Martian Impact Craters [27] using a circular moving window of Results: Figure 1 shows the maps of DD for Earth radius 150 km. A given cell is assigned a value of cra- and Mars. As expected the values for Mars (0-0.12 ter density corresponding to a ratio of the total count of km-1) are lower than for the Earth (0-0.32 km-1), but craters in a window to the area of the window. Figure the ranges are in the same order of magnitude. The 3(left) shows the results; on average the value of DD derived values of DD for Earth are considerably small- increases with the crater density. In other words, the er than ~ 1-100 km-1 quoted in literature [26]. This is spatial distribution of craters and VN are positively because of relatively coarse resolution of the DEM. correlated. Nevertheless, the comparison of DD patterns on the two planets is fair because DEMs of similar resolu- Discussion. The aforementioned features of the tions are used. The maps on Figure 1 reveal striking global distribution of DD on Mars impose constrains differences in spatial distributions of high DD regions on possible scenarios of VN formation. The strong on the two planets. In order to quantify these differ- local variability of DD argues against global-scale and ences we have calculated zonal statistics of DD with persistent precipitation, but the planet-wide extent of respect to elevation, latitude, and regional (33 km- the region where average values of DD are high re- base) slope. The result of the zonal statistics calcula- quires some sort of a global-scale mechanism. That tion is a function showing a dependence of DD (aver- mechanism cannot depend on elevation, but may be aged over the entire zone) on a variable of interest. weakly dependent on regional slope. It must be corre- Figure 2 shows the comparison of zonal statistics, the lated with the process of impact cratering. top row corresponds to Earth whereas the bottom row We submit that a VN formation mechanism best corresponds to Noachian Mars. satisfying the restrictions imposed by the global distri- On Earth, the highest values of DD are found in bution of DD is the impact cratering itself. In this sce- geographically disparate locations that, however, are nario, as originally proposed by [17] and later elabo- all characterized by high elevations and steep slopes. rated by [28] and [29], impactors as small as 10 km in Within such locations the DD is uniformly high. Could diameter result in a global warming for 10-20 seasons we infer the mechanism of valley formation on Earth and precipitation of 0.2 m total liquid water (TLW). having only the topography data and the global map of Larger impacts results in longer periods of warming DD at our disposal? We could definitively eliminate and more precipitation. In addition, impact-induced SECOND WORKSHOP ON MARS VALLEY NETWORKS 53

hydrothermal activity [30] can force local groundwa- ries of alterations to an otherwise cold and dry global ter sapping. Because impacts were ubiquitous in Noa- climate. These climatic diversions create episodic and chian, dissection density due to VN carved by impact- short-lived precipitation events and groundwater sap- induced erosion should be homogenous on large-scale ping activities that are responsible for the patchy cha- but variable on spatial scale of the order of the largest racter of VN. They also explain mixed and inconclu- craters. This is in agreement with the map of DD sive geomorphic evidence and observed lack of ma- shown on Fig. 1. The only constrain not fulfilled by turity [35] in the development of VN. We call upon the impact-induced erosion is the relative lack of val- terrain softening to explain paucity of VN in the mid- leys (low values of DD) southward of ~ 30 S˚. This to-high southern latitudes. seems incompatible with ubiquity of impacts. Acknowledgements. We acknowledge support by We propose that observed deficiency of valleys NASA (WL under grant NNX08AM98G, TFS under southward of ~ 30 S˚ is not a primordial feature but grant NNG05GM316). A portion of this research was rather the effect of subsequent surface modification. conducted at the Lunar and Planetary Institute under Indeed, the terrain located southward of ~ 30-40 S˚ is contract CAN-NCC5-679 with NASA. This is LPI known for its “softened” appearance. This attribute of Contribution No. 1436. mid-to-high latitude terrain has received most attention in connection with craters [31,32] but it affects all sur- References: [1] Milton, D. J.,(1973) JGR 78 (20), 4037- face features. It is related either to the viscous relaxa- 4047. [2] Irwin, R. P., III, and A. D. Howard, (2002) JGR, 107(E7), tion due to presence of ground ice [33] or to mantling 10 1-10 23. [3] Craddock, R. A., and A. D. Howard, (2002) JGR, [34]. We propose that originally VN were distributed 107(E11). [4] Hynek, B. M., and R. J. Phillips (2003) Geology, uniformly over the Noachian, but those located south- 31(9), 757-760. [5] Craddock, R. A. et al. (1997) JGR 102, 13,321– ward of ~ 30-40 S˚ were softened beyond recognition 13,340. [6] Forsberg-Taylor, N. K. et al. (2004) JGR 109. [7] resulting in a presently observed distribution as encap- Goldspiel, J.M., and S. W. Squyres, (1991) Icarus, 89, 392-410. [8] sulated by the map of DD shown on Fig. 1. Grant, J. A. (2000) Geology, 28 (3), 223-226. [9] Gulick, V. C. To further test this hypothesis, we extracted valley (2001) Geomorphology, 37, 241-268. [10] Sharp, R. P., and M. C. depth information from the topographic data for both Malin (1975) Geol. Soc. Am. Bull., 86, 593-609. [11] Pieri, D. C., planets by using average of the cells surrounding val- (1980) Science, 210, 895-897. [12] Goldspiel, J. M. et al. (1993) ley cells to approximate their pre-incision elevation Icarus, 105(2), 479-500. [13] Baker, V. R., and J. B. Partridge and calculating the difference between this pre- (1986) JGR, 91 (3), 3561-3572. [14] Aharonson, O. et al. (2002) incision elevation estimate and the current elevation. Proc. Nat. Acad. Sci., 99, 1780-1783. [15] Carr, M.H. and Head, J. The latitudinal zonal statistics of the depth estimates W. III, (2004), LPSC XXXV abstract #1183. [16] Irwin, R. P., III, et are shown in Figure 3(middle and right). For Earth, al. (2006) LPSC XXXVII #1912. [17] Segura, T. L. et al. (2002) there is great variability in the depth with latitude but Science, 298, 1977-1980. [18] Carr, M. H. (1995) JGR 100 (E4), no systematic trend, indicating latitude independent 7479-7507. [19] Molloy, I., and T. F. Stepinski (2007) Computers & tectonic activities. For Mars, the depth does decrease Geosciences, 33(6), 11. [20] Smith, D., G. et al. (2003) NASA with increasing latitude, consistent with the scenario that VN were created by episodic and short-lived pre- Planetary Data System. [21] Neumann, G. A. et al., (2001) JGR, cipitation events and groundwater sapping activities 106, 23,753-23768. [22] Tarboton, D. G. and D. P. Ames, (2001) in induced by impact cratering and thus initially uniform- World Water and Environmental Resources Congress, Orlando, ly distributed over Noachian, but later relaxed at high- Florida, May 20-24, ASCE. [23] Luo, W. and Stepinski, T. F. er latitude regions, making the valley depth shallower (2008), Geomorphology, 99, 90-98. [24] Luo, W., and T. F. Stepins- there. ki, (2006) GRL 33(L18202, 1-4. [25] Stepinski, et al. (2007) LPSC XXXVIII abstract #1205. [26] Knighton, D., (1998) Fluvial forms Conclusion: Global maps of dissection can now be and processes: A new perspective, Edward Arnold, London. [27] constructed by means of computer parsing of digital Barlow, N. G. (1988) Icarus, 75, 285–305. [28] Colaprete A. et al. elevation models. This new technique offers an oppor- (2005) in Role of Volatiles and Atmospheres on Martian Impact tunity to test hypothesis of VN formation from a global Craters, LPI, Laurel, Maryland. [29] Segura, T. L. et al. (2008), perspective, a welcome addition to the tests based on LPSC XXXIX abstract #1793. [30] Pierazzo, E. and Ivanov B. A. local geomorphic features. Dissection patterns on (2008), LPSC XXXIX abstract #1155. [31] Squyers, S. W. and Carr Earth and Mars are completely dissimilar. Terrestrial M. H. (1986) Science, 231, 249-252. [32] Stepinski, T. F., and E. R. case reveals that climatic data by itself is not sufficient Urbach, (2007) 7th Int. Conf. on Mars. Pasadena, California. [33] for a correct interpretation of dissection pattern. Dis- Parmentier, E. M. and Head, J. W. (1981) Icarus, 47, 100-111. [34] section pattern on Mars points to the impact cratering Soderblom, L.A., et al. (1973) JGR, 78 (20), 4117-4122. [35] Ste- as an ultimate source of VN. The impacts cause a se- pinski, T. F., and A. P. Stepinski (2005) JGR 110(E12), E12S12. SECOND WORKSHOP ON MARS VALLEY NETWORKS 54

Figure 1: Global maps of dissection density (DD) on Earth (left) and Mars (right). Blue-to-red gradient corresponds to low-to-high values of DD. The same colors on the two maps do not corresponds to the same values of DD. The gray areas on Mars correspond to the regions that are not classify as Noachian, or to the regions for which DD is not defined.

Figure 2: Zonal statistics of dissection density (DD, km-1) based on elevation (left), latitude (middle), and regional slope (right).

Figure 3: Zonal statistics of dissection density (DD, km-1) based on crater density (left) and zonal statistics of valley depths based on latitude for Earth (middle) and Mars (right).