Updated Global Map of Martian Valley Networks: Implications for Hydrologic Processes

Home , Amazonian (Mars), Valley network (Mars)

SECOND WORKSHOP ON MARS VALLEY NETWORKS 39

UPDATED GLOBAL MAP OF MARTIAN VALLEY NETWORKS: IMPLICATIONS FOR HYDROLOGIC PROCESSES. B.M. Hynek1-2, M. Beach1-2, and M.R.T. Hoke1, 1Laboratory for Atmospheric and Space Physics, 2Dept. of Geological Sciences (392 UCB, Univ. of Colorado, Boulder, CO 80309) [email protected]

Introduction: The valley networks of Mars have greater limitations. Particularly, the resolution of long been viewed as some of the best evidence that MOLA data is too coarse to resolve many of the water flowed across the surface under climatic condi- smaller features. The size of the smallest valley the tions that were much different in the past. However, algorithm can detect is limited by the resolution and many questions remain, including the timing and completeness of the gridded MOLA data at a given mechanisms of valley formation. To address these latitude: At the equator, one MOLA pixel is ~500 m on questions, Carr [1] and Carr and Chuang [2] produced a side. The smallest valley that can be reliably re- a comprehensive Martian valley network map from solved is ~3-km-wide (6 pixels). However, this as- Viking-based images covering ±65° of latitude. This sumes that there is complete coverage. Due to the or- product contained over 900 networks with over 8,000 bit of the Mars Global Surveyor, approximately 60% branches that generally appeared immature in plani- of the pixels in the MOLA gridded data at the equator metric form and contained large, undissected regions are interpolated, which often eliminates the signature between networks. The authors used these attributes of a few-km-wide valley. Indeed, Molloy and Stepin- to suggest that only limited surface water was neces- ski [6] found that their MOLA-based algorithm identi- sary for their formation. Mars Global Surveyor data- fied on average 69% of the valleys they could manu- sets have revealed many valleys that were not evident ally identify in THEMIS images. Missing ~1/3 of the in the Viking images and correspondingly more mature valleys, particularly all the smaller ones, would result drainage systems [3]. The Thermal Emission Imaging in an incomplete picture of fluvial processes on Mars. System (THEMIS) experiment [4] on the Mars Odys- Further, MOLA data have higher spatial resolution and sey spacecraft provides an even higher resolution look more complete coverage toward the poles and this at valley networks. In particular, the 100 m resolution would bias the valley detection. daytime IR data is ideal for identification and analysis Results: of valleys, which are generally a few kilometers wide Drainage Density: As expected, many more val- and hundreds of meters deep. Coregistered, ~0.5-km- leys were seen in THEMIS data compared to those resolution Mars Orbiter Laser Altimeter (MOLA) data detected in Viking images. Figure B shows an exam- [5] are also useful in mapping and studying the val- ple of valleys mapped by Carr [1] compared to this leys’ characteristics. study. In the original mapping, a few, often uncon- Methods: We used the THEMIS daytime IR and nected, valleys without many tributaries could be seen MOLA datasets to remap valley networks on a global (Fig. 1a). Figure 1b shows that with higher resolution scale in an effort to better understand their distribution data, we can now resolve these valleys into a dense, and to test hypotheses regarding their origin (Fig. A- mature drainage network. In this example, Carr global map). The valleys were manually mapped us- mapped 16 valley segments totaling 1,492 km in ing similar determining characteristics to those of Carr length; giving a corresponding drainage density of [1] for ease of comparison (i.e. sublinear, erosional 0.034 km-1. In contrast, our updated mapping shows channels that form branching networks and increase in 462 valley segments totaling 6,031 km in length; giv- size downstream). We remapped valleys primarily ing a corresponding drainage density of 0.14 km-1. In with the THEMIS global daytime IR mosaic, which this case, the higher resolution mapping reveals a fac- covers ~90% of the planet, with colorized MOLA data tor of 4 increase in stream length and drainage density. to aid in detection of valleys and determine downhill The characteristics of the originally mapped valleys direction. In areas of poor THEMIS coverage, we are consistent with formation by groundwater proc- substituted Viking images. Our mapping concentrated esses. However, the updated map indicates that sus- in the region within ±70° of latitude, since very few tained precipitation and surface runoff were likely re- valleys have ever been noted at the highest latitudes. quired due to the dense drainage, high stream order, However, we conducted a cursory assessment at polar and many small tributaries that reach right up to drain- latitudes for completeness. Manual mapping of val- age divides. leys is a tedious and somewhat subjective process that In total, the updated global map shows ~4 times as can be influenced by albedo variations and image qual- many valleys identified, totaling a summed length ~2 ity. While a number of automated routines to map times greater than Carr [1] (Table 1). Drainage densi- global valleys have been attempted with MOLA grid- ties of networks are almost always much higher and ded data and other products [6], these have even SECOND WORKSHOP ON MARS VALLEY NETWORKS 40

a greater complexity of drainage is evident. Carr and Chuang [2] did not identify a single valley greater than 4th order in the Strahler system (higher order equals more maturity and water) [7] yet we have found one 7th order, six 6th order and twenty-five 5th order networks. The average drainage density on Noachian-age units is nearly 4 times greater than previously determined. Many smaller tributaries are seen that were not evident with Viking data as well as structure within the valleys including braided channels and terraces indica- tive of sustained flow. Age Distribution: Figure 2 shows that in terms of age, roughly 84% of newly mapped valley segments lie entirely within Noachian terrains (>3.7 Ga ago), 10% cross into or are entirely contained within Hespe- rian-age surfaces (3.7-3.0 Ga) and 6% occur on Ama- zonian terrain (<3.0 Ga) (Fig. 2, Table 1). These per- centages are shifted to younger ages relative to those reported from Viking-based mapping [1] but still argue for relatively rapid climate change at the end of the Noachian. This is supported by the rapid drop in drainage densities on Martian volcanoes [10] and other terrains at this time boundary. The valleys on Noa- chian surfaces occur across most of the Southern High- lands although a few regions are still relatively undissected (such as west of Hellas basin, Fig. 2). We Figure 1. Comparison of valleys mapped by Carr [1] speculate that in these regions additional resurfacing from Viking data (a) to those identifiable in THEMIS has obscured the signature of valleys and this is sup- data (b) near 3°S, 5°E; both on a THEMIS basemap. ported by the heavily degraded and infilled craters in

Figure 2. Global comparison of valleys identified by Carr [1] from Viking data (top) and our preliminary updated map using THEMIS data (bottom). In the new map, denser concentrations of valleys are seen almost everywhere and some areas with few identified valleys from Viking data show significant dissection (e.g., south of Hellas Basin; 55°S, 70°E). Colors on the bottom map represent inferred valley ages determined by the youngest terrain unit they incise (red = Noachian, purple = Hesperian, blue = Amazonian) (unit ages from [8-9]). Figure 1 is a higher resolution example comparing the two maps. SECOND WORKSHOP ON MARS VALLEY NETWORKS 41

Valley Elevations: Figure 3 shows the relative Table 1. Comparison of global Martian valley network distribution of Mars valley networks across the three characteristics identified from Viking imagery and recently- defined major geological epochs. As previously un- acquired topographic and IR data. derstood, most of the valleys occur on ancient terrains, which are typically high in elevation [1-2]. These val- Carr 1995 [1] This study leys have a rough Gaussian distribution centered # valleys 11,336 40,005 around 1500 m (Fig. 3). The younger valleys, while total length 398,935 797,083 comprising a small percentage of the total, developed (km) roughly evenly across all elevations (Fig. 3). These highest 4 7 valleys formed after climatic conditions favored pre- stream order cipitation and stable surface water on a global scale. highest net- The younger valleys probably originated from proc- work drain- ~0.02 km-1 0.14 km-1 esses such as hydrothermal circulation on volcanoes age density [10] or from asteroid impacts that could force climate (length/area) change locally [13]. The somewhat uniform distribu- ~3.7- ~3.7- tion of these valleys across all elevations may support >~3.7 <~3.0 >~3.7 <~3.0 age break- 3.0 3.0 these ideas. Ga Ga Ga Ga down Ga Ga Comparison to Other Datasets: The updated 90%, 5%, 5%, 84% 10% 6% global valley network map was compared to other Noachian datasets including global databases of fluvial features units’ avg. and mineralogical/compositional data. No strong cor- 0.0041 km-1 0.015 km-1 drainage relations exist between valley density and elemen- density tal/mineralogical data such as the K/Th ratio from GRS [14] or water-altered minerals from OMEGA and CRISM [15 and 16, respectively]. Figure 4 shows these regions. A number of the larger valley systems newly-mapped valley drainage density, mathematically have been age-dated with crater density measurements defined as the density of linear features within a set [11-12]. These authors suggested that most of these radius of each output raster cell. Overlaid on this map valleys formed toward the end of the Noachian or start are global locations of putative paleolake and fan-delta of Hesperian epoch. In some cases, systems were re- deposits [databases from 17-18, respectively]. There is activated after a period of up to 100s of millions of a strong correlation between density of valleys and the years and this may indicate that the climate was not identified fluvio-sedimentary deposits. That is, most continuously warm and wet, but instead more episodic [11]. Valleys are now identified in terrains thought to be undissected from previous work (Fig. 2). Particu- larly, many valleys are now recognized south of Hel- las basin, which exist on Hesperian-aged terrain. The remainder of Hesperian-aged valleys are peppered throughout the cratered highlands with concentrations near Hellas Basin and on volcanoes (Fig. 2). Many of these are on the gentle flanks of the volcano Amphi- trites Patera. The Amazonian-aged valleys occur predominantly on the flanks of volcanoes, including the Early Ama- zonian-aged Tyrrhena, Hadriaca, and Alba Paterae, but also on the eastern slopes of Hellas basin. We mapped the valleys and crater-age dated each channel-covered volcano [10]. While pre-Amazonian valleys show strong characteristics of precipitation and surface runoff, most valleys on Amazonian-age volcanoes are more consistent with a hydrothermal origin. This is true of the youngest valleys on Mars, which are iso- Figure 3. Distribution of ages and elevations of all identified lated to the Tharsis volcanoes. valleys on Mars. Results show that valley formation shifted to more distributed elevations as time progressed. SECOND WORKSHOP ON MARS VALLEY NETWORKS 42

Figure 4. Valley density on Mars in relation to putative paleolake deposits (blue, [17]) and fan-deltas (red, [18]). There is a strong correlation between thigh drainage density and reported fluvio-sedimentary deposits.

of the suggested paleolake deposits and fan-deltas oc- vide the best evidence for water on the surface of cur in regions of higher drainage density (Fig. 4). Mars, but the magnitude and duration of precipitation- Conclusions: An updated global map of valleys fed valley formation is only now being constrained. on Mars has been completed and will be released to References: [1] Carr, M. H. (1995) JGR, 100, the community once we have finished cleaning and 7479–7507. [2] Carr, M. H. and F. C. Chuang. (1997) validating the map. The 100-meter-resolution global JGR, 102, 9145-9152. [3] Hynek, B. M. and R. J. Phil- THEMIS day IR data, combined with topography, lips. (2001) Geology 29, 407-410. [4] Christensen, P. allow an unprecedented look at these features related R., et al. (2004) THEMIS Public Data Releases, PDS to surface and near-surface water. Many more valleys node, ASU, http://themis.asu.edu. [5] Smith, D. E. et are seen globally than previously and drainage density al. (2001) JGR, 106, 23689-23722. [6] Molloy, I. and measurements are correspondingly higher. Some val- T. F. Stepinski. (2007) Comp. and Geosci., 33, 728- leys are seen with interior anatomizing channels and 738. [7] Strahler, A. N., (1958) Geol. Soc. Am. Bull., terracing. These new results are consistent with a 69, 279-300. [8] Scott, D. H., and K. L. Tanaka (1986) warm and wet early climate that incised the crust. U.S. Geol. Surv. Misc. Invest. Ser., Map I-1802-A. [9] Hesperian and Early Amazonian valleys also show Greeley, R., and J. E. Guest, (1987) U. S., Geol. Surv. signs of formation by precipitation and surface runoff, Misc. Invest. Ser., Map I-1802–B. [10] Bowen T. A. although at much decreased levels. The youngest val- and Hynek B. M. (2008) LPSC XXXIX, abs. 2393. [11] leys on the planet are confined to volcanic constructs Hoke, M. R. T. and Hynek B. M. (2008) J. Geophys. and most likely had a hydrothermal origin. Areas on Res. (in review). [12] Fassett, C. I. and Head, J. W. Mars that have higher valley density correlate with (2008) Icarus, 195, doi: 10.1016/j.icarus.2007.12.009. fluvio-sedimentary deposits that have been noted by a [13] Segura, T. L. et al. (2008) LPSC XXXIX, abs. number of workers. This supports the idea that many 1793. [14] Taylor, G. J. et al. (2006) J. Geophys. Res., Mars valleys and associated deposits were generally 111, doi:10.1029/2006JE002676. [15] Bibring, J. P. et formed by long-lived surface runoff from precipita- al. (2006) Science, 312, 400-404. [16] Mustard, J. F. et tion. However, recent results show that in some places al. (2008) Nature, 454, 354-359. [17] Fassett, C. I. and valley formation may have been limited to short, in- Head, J.W. (2008) Icarus, in press. [18] Di Achille, G. tense episodes. In summary, valley networks still pro- et al. (2008) this conference.