Modeling Late-Winter Forest Snow Depth Using Stand Structure, Terrain, and a Thermal Forcing in a Hemiboreal Mountain Catchment

Winter snow accumulation in forested ecosystems plays a critical role in hydrological processes, ecosystem functioning and seasonal water availability in cold regions. However, direct measurement of snow distribution beneath forest canopies is challenging because of strong spatial variability and limited long‑term field observations. This study develops and validates an empirical framework to estimate forest snow depth using commonly available structural, topographic and meteorological variables. Forest snow depth was measured at permanent plots (n = 364) over four winters (2011, 2012, 2024 and 2025) in a cold‑region forest in central Hokkaido, Japan. A Snow Retention Index (SRI) was defined as the difference between plot‑level forest snow depth and the seasonal maximum snow depth recorded at a nearby meteorological station. Mixed‑effects models were first used to quantify the relative contributions of spatial controls and interannual variability, followed by the development of a transferable fixed‑effects model in which interannual variability is represented through a physically interpretable thermal forcing metric, positive degree days (PDD). The fixed‑effects model was trained using data from 2011–2012 and independently validated using data from 2024–2025. Validation results showed strong predictive performance (R 2 = 0.77, RMSE = 34.2 cm), indicating that the model reliably captures within‑winter spatial variability in forest snow retention. Elevation, total basal area and PDD emerged as dominant predictors, reflecting the combined influence of topography, canopy interception and late‑winter melt processes. By linking forest snow depth to standard forest inventory data, terrain information and routine meteorological observations, the proposed approach provides a practical and transferable method for estimating snow availability in forested headwaters under data‑limited conditions.