Decision Support for Snow Removal Dispatch Integrating Meteorological Information and Fixed-camera Images

learning-based predictionSnow removal dispatch decisions during winter nights require not only the accurate prediction of dispatch necessity but also the intuitive presentation of road conditions and decision-relevant information to human operators.In this study, we propose a snow removal dispatch decision support method that integrates fixed-camera images and meteorological information to both predict dispatch necessity and visualize information for operational support.The proposed method employs a machine learning model with feature selection, integrating meteorological observations, short-term forecasts, snow depth measurements, and an imagederived road condition indicator called the snow coverage ratio (SCR).From a sensing perspective, the proposed method integrates heterogeneous environmental sensing modalitiesincluding fixed-camera imaging, snow-depth sensing, and meteorological sensing-to represent road surface conditions relevant to dispatch decision-making.By visualizing road conditions together with prediction results, the system provides operators with objective and consistent situational awareness to support decision-making.Experimental results confirm that SCR effectively reflects temporal changes in road conditions and contributes to dispatch prediction.Furthermore, evaluation using data collected under similar winter meteorological conditions shows that a dispatch prediction model with L1 regularization tends to outperform human operators' decisions.Additional evaluation applying models trained on past-year data to subsequent years demonstrates that training with multi-year datasets improves generalization under varying winter conditions.These results indicate that the proposed method provides practical and reliable support for snow removal dispatch decisions in real operational environments.