COMPLETION OF THE FIRST AUTOMATIC SURVEY OF CRATERS ON MARS

Introduction: Advances in surveying impact craters present in data gathered by remote sensing of planetary surfaces have not kept up with advances in data collection. As a result, there is a deluge of planetary data but no means for its comprehensive analysis. If left to manual surveys, the fraction of cataloged craters to the craters actually present in the available and forthcoming data will continue to drop precipitously. Therefore, we submit that automating the process of crater detection is the only practical solution to a comprehensive surveying of craters. As the first step to automating the surveying process we have developed [1,2] a robust crater detection algorithm (CDA) capable of identifying and characterizing most craters over the entire surface of Mars having diameter D ≥ 3 km. The limit on the size of the crater is set by the resolution of topographic data used for crater detection. Presently, the global topographic coverage is given with the resolution of 1/128 degree by the MOLA Mission Experiment Gridded Data Record (MEGDR) [3]. Methods. Our CDA is a two stage system that first identifies round depression on Mars and then uses machine learning technique to separate true craters from false positives. Its application to the entire surface of Mars requires subdividing the surface into 356 overlapping tiles. The craters are identified and measured at each tile separately and the results from individual tiles are concatenated into a single catalog from which duplicate detections are eliminated. In its present “beta” version the catalog lists coordinates of the center of each crater, its diameter, depth, and an underlying geological unit. Results. The catalog lists 75,919 craters ranging in size from 1.36 km to 347 km. Fig.1 shows the exceedance probability of crater diameter, D, for all craters in the catalog. Exceedance probability, P(D>X), is a probability that a randomly chosen crater has diameter larger than X. It represent a convenient way of displaying distribution of crater sizes. For comparison, Fig. 1 also shows exceedance probability of crater diameter in manually collected Barlow [4] catalog that lists 42,283 craters. The graph reveals that the new catalog is statistically “complete” down to craters having size of about 3 km, whereas the Barlow catalog is statistically complete down to the size of 5 km. Fig. 2 shows that density of craters in the new catalog. The density is calculated using a moving circular window of radius 250 km. It reflects the combined effect of true distribution of craters and the existence of some biases in the methodology due to different rate of automatic detection between different surfaces.