Investigation of small-scale polygonal networks on Mars using models of terrestrial fracture and ice-wedge networks.

Polygons formed by closely spaced (tens to hundreds of meters) interconnected troughs, visible in Mars Orbiter Camera images, are qualitatively similar to ice- and sand-wedge patterns in lowland Arctic and Antarctic terrain on Earth. The spacing and relative orientation between troughs in Mars networks varies between polygonal networks. Terrestrial networks, which form by recurrent opening of tension fractures in perennially frozen ground during periods of rapid cooling in winter, also display broad variations in the characteristic spacing, width and intersection angles of ice- and sand-wedges. Hypothesized causes for variations between terrestrial networks include variability in magnitude and orientation of maximum cooling-induced tensile stress, in substrate-dependent strength and heterogeneity, and in limits to downward propagation of fractures owing to a temperature-dependent brittle/ductile transition at depth. To investigate mechanisms for variability in Mars and terrestrial networks and to test if properties of some or all measured Mars networks fit within the range of terrestrial variability, we explore the response of a recently-developed computational model for terrestrial networks to changes in substrate strength and heterogeneity, maximum tensile stress, and fracture depth. The model treats initiation, propagation and arrest of fractures in a tensile stress field perturbed by neighboring fractures, and includes the growth of ice or sediment wedges along fracture paths. Modeled networks are compared to 20 1x1 km network regions from MOC images of Utopia Planitia using two methods. In the first method, joint distributions of relative orientation and spacing between troughs are used to characterize mean spacing and orthogonality of networks. In the second method, regions of a pixelated image of a network are used to predict the pixel pattern of displaced regions with a nonlinear spatial forecasting algorithm that operates on pixel brightness. Prediction error as a function of algorithm parameters is used to assess the degree of nonlinearity and stochasticity in processes that generate the patterns. To test whether patterns on Mars form by single fracture episodes or through recurrent fracturing, measured patterns are compared with those generated by simulations of a single fracture episode and with recurring fracture episodes. Supported by the Natural Science and Engineering Research Council, the Andrew W. Mellon Foundation, and the Canadian Institute for Advanced Research