Presentation given at Smithsonian National Air and Space Museum October 2006
The Science and Art of Mapping Martian Valley Networks Using a Computer Algorithm Tomasz F. Stepinski Lunar and Planetary Institute Ian Molloy Purdue University (Computer Science) Wei Luo Northern Illinois (Geography)
Valley networks Valley networks in the MC-21 Iapygia quadrangle (MDIM 2.1)
Global map of valley networks (circa late XX century) Carr (J. Geophys. Res. 100(E4), p7479., 1995 Carr and Chuang (J. Geophys. Res. 102(E4), p9145., 1997 About 800 networks Networks lack spatial integration Drainage density of Noachian terrain ~0.007 km -1
Local maps of valley networks (circa early XXI century) Hynek and Phillips, Geology, 29(5), p407., 2001
A test site in the Western Cascades of Oregon East West Uneven drainage, the western part is more dissected than the eastern part
DEM, Flooding, and Slopes Original DEM Slope 128 x 187 = 23,936 pixels 216 (0.9%) pixels are modified by flooding Mean flood < 1 meter Maximum flood 30 meters DEM modified by flooding
Drainage directions (D8) Drainage direction is calculated for every pixel D8 network defines individual drainage basins D8 network is not a drainage network Drainage network is a subset of D8 network Individual basins (45) Stream order (1 to 7)
Drainage directions (D8) original DEM Even small DEM imperfections lead to large changes in D8 network Imperfections increase number of basins Drainage Imperfections lower the maximum stream order Individual basins (79) Stream order (1 to 5)
Delineating drainage network using contributing area A drainage (contributing) area Threshold = 10 pixels Q mean annual discharge Q ~ A
Delineating network using contributing area threshold Threshold = 10 pixels Threshold = 50 pixels
Delineating drainage network using stream order >= 3 rd order
Delineating network using stream order threshold >= 3 rd order >= 4 th order
Delineating network using contributing area and slope a A S > C a = 2 C = 0.5 This algorithm provides a mechanism for spatially variable drainage density Montgomery and Dietrich, Science, 255, p826, 1992 It results in feathering of the drainage network in steeper areas, while omitting drainage networks in less steep valleys.
Trouble with Mars Evros Vallis West. site 31% modified by flood A>=500 pixels network
Trouble with Mars (imperfect solution) Evros Vallis West. site 18% modified by flood <=200 meters Valley mapped by our new method A>=500 pixels network
Terrain morphology-based valley mapping algorithm Steps 1 and 2 - digital terrain analysis methods Steps 3 and 4 - image processing methods Molloy and Stepinski, Computers & Geoscience, submitted, 2006
Step 1.2: Calculating tangential curvature Pollack test site Tangential curvature is measured in a direction of tangent to contour. Blue: positive curvature (convergent) Red: negative curvature (divergent)
Step 1.3: Thresholding curvature map Map of tangential curvature Only pixels with curvature > 0.003 (1/m) are shown in blue
Image processing steps
Final product of our algorithm Map of VN derived using our algorithm Map of VN derived using our D8 Algorithm with A>=200 pixels threshold
Test sites
Automatic vs. manual mappings (visual comparison)
Automated vs. manual mappings (quantitative comparison) Manually mapped network Automatically delineated network
Conclusions about our algorithm D8 algorithm is not a good base for automatic mapping of drainage network. Different channelization criteria can be applied to D8 network to yield an actual drainage network, but they all lead to approximately constant drainage density, in contrast to what is seen on Mars. Our algorithm is not based on drainage directions. It maps valleys where they are seen. It produces maps that are in good agreement with manual mapping. Our maps are consistent and topographically correct in contrast to manually derived maps. Our algorithm is fast enough and precise enough to be employed in mapping of VN over large areas of Martian surface.
Application to the Mare Tyrrhenum quadrangle Luo and Stepinski, Geophysical Research Letters, 33, L18202, 2006
Drainage density in the Mare Tyrrhenum quadrangle
Continuous map of drainage density in the Mare Tyrrhenum quadrangle
Mare Tyrrhenum results An order of magnitude higher drainage density in Noachian terrain than the values inferred from a global map based on Viking images. Drainage density in Noachian terrain comparable to the values inferred from the precision manual mapping of selected focus sites. High dissection of all Noachian units. Variation in dissection of Noachian terrain on scales from 1000 km to 100 km. Lack of correlation between degree of dissection and terrain parameters. Omnipresence of dissection suggests precipitation.
Bonus: Terrestrial applications for our algorithm J. Taylor Perron PERMEABILITY, RECHARGE, AND RUNOFF GENERATION ON MARS Workshop on Mars Valley Networks Kohala Coast, Hawaii August 11-15, 2004
D8 algorithm fails in Oregon Cascades
Our algorithm succeeds in Oregon Cascades
Map of drainage network Bigger picture Map of drainage density Map of simplest geology tertiary quaternary