Rocky Mountain Section - 75th Annual Meeting - 2025

Paper No. 9-7
Presentation Time: 8:00 AM-5:30 PM

SNOW, SLOPE, AND STONE: MACHINE LEARNING REVEALS KEY PREDICTORS OF ROCK GLACIER FORMATION IN THE WASATCH RANGE OF UTAH, USA


MORRISS, Matthew, Utah Geological Survey, 1594 West North Temple, Suite 3110, Salt Lake City, UT 84116, ANDERSON, Leif, Department of Geology and Geophysics, University of Utah, Salt Lake City, UT 84112, OLSON, Matthew, Department of Earth Science, Utah Valley University, 800 W University Pkwy, Orem, UT 84058 and WASICKO, Zane, Natural and Environmental Sciences Department, Western Colorado University, 600 N. Adams St, Gunnison, CO 81231

Rock glaciers are common periglacial landforms in arctic and alpine settings that act as critical water stores, geohazards, and biological refugia. While previous studies focused on regional inventories or detailed analyses of individual rock glaciers, less attention has been given to the factors controlling their presence or absence. Our study addressed this gap using a machine learning (ML)-based random forest classifier to explain rock glacier occurrence in the Wasatch Range of northern Utah, USA. We developed a comprehensive inventory of rock glaciers and associated periglacial landforms (e.g., protalus lobes and protalus ramparts) through detailed geologic mapping in the Wasatch Range. We outlined 141 periglacial landforms, including 68 rock glaciers, 37 protalus ramparts, 35 protalus lobes, and 1 heavily debris-covered glacier, as part of the most detailed survey of rock glaciers and associated landforms in the Intermountain West. To better constrain geomorphic and topographic controls at play for each landform, we also mapped their rooting zones and contributing headwall catchment areas.

Analysis of our inventory revealed that periglacial features occur only above 2400 m a.s.l., where temperatures are low and precipitation rates are high. However, modern climate variables alone do not explain rock glacier occurrence. We demonstrated that the presence of rock glaciers and other periglacial landforms is strongly correlated with low-slope areas at high elevations. In the central Wasatch, where low sloped cirque floors are abundant, periglacial landforms occupy up to 6% of the land area. In Little Cottonwood Canyon, they cover more than 20% of the land surface. Importantly, our ML-based decision tree analysis shows that the Normalized Difference Snow Index is the strongest predictor of rock glacier presence of the 10 climatic or topographic metrics analyzed. This indicates that rock glaciers form preferentially and persist in areas with heavy snowfall and perennial snowfields. Our research provides crucial insights into the spatial patterns and controlling factors of rock glaciers in the western U.S., with implications for all periglacial landscapes. These findings are vital for assessing future water resources and potential geohazards in alpine environments under a changing climate.