Snow distribution is a function of interactions among static variables, such as terrain, vegetation, and soil properties, and dynamic meteorological variables, such as wind speed and direction, solar radiation, and soil moisture that occur over a range of spatial scales. However, identifying the dominant physical drivers responsible for spatial patterns of the snowpack, particularly for ephemeral, shallow snowpacks, has been challenged due to the lack of the high-resolution snowpack and physical variables with high vertical accuracy as well as inherent limitations in traditional approaches. This study uses an Unpiloted Aerial System (UAS) lidar-based snow depth and static variables (1-m spatial resolution) to analyze field-scale spatial structures of snow depth and apply the Maximum Entropy (MaxEnt) framework to identify primary controls over open terrain and forests at the Thompson Farm Research Observatory, New Hampshire, United States. We found that, among nine topographic and soil variables, plant functional type and terrain roughness contribute up to 80% and 76% of relative importance in MaxEnt to predicting locations of deeper or shallower snowpacks, respectively, across the landscape. Soil variables, such as organic matter and saturated hydraulic conductivity, were also important controls (up to 70% and 81%) on snow depth spatial variations for both open and forested landscapes suggesting spatial variations in soil variables under snow can control thermal transfer among soil, snowpack, and surface-atmosphere. This work contributes to improving land surface and snow models by informing parameterization of the sub-grid scale snow depths, downscaling remotely sensed snow products, and understanding field scale snow states.