Mapping Scalloped Terrains on Mars at the Meter Scale

Abstract Scalloped terrains are a kind of periglacial landform, shaped by permafrost evolution that involves degradation of near‐surface excess ice, and are mostly distributed in the mid‐latitudes of Mars. Their evolution mechanism remains debated, including whether liquid water was involved (thermokarst‐lake vs. sublimation) and what the ice‐rich precursor conditions were (pre‐existing vs. post‐deposited volatiles). Therefore, studies on scalloped terrains can provide important clues to the distribution and abundance of local water ice, as well as the climatic evolution of late Amazonian Mars. However, their hemispheric‐scale distribution has remained poorly constrained. In this study, we develop a deep‐learning semantic segmentation framework to map scalloped terrains across the northern hemisphere of Mars using the global CTX mosaic. The model achieves high accuracy, enabling reliable pixel‐level identification of isolated, coalesced, and degraded forms. On this basis, we generate and publicly release a 5‐m‐per‐pixel hemispheric‐wide map of scalloped terrains and analyze their distribution, density and orientation. The results reveal a concentration of scalloped terrains in Utopia Planitia and a systematic latitudinal trends in depression types, most notably a transition from isolated depressions at 42°–45°N to increasingly coalesced and boundary‐less terrain beyond 48°N, reflecting the distribution of near‐surface ice‐rich deposits. Orientation analysis further reveals the influence of geomorphologic context on the growth and degradation of scalloped terrains in addition to insolation condition. These findings provide new quantitative constraints on the evolution of Martian periglacial landforms and establish a scalable framework for automated planetary geomorphologic mapping.