Estimating Annual Average Daily Traffic on Local Roads: Integrating Spatial Insights with Machine Learning

Abstract Obtaining street-level annual average daily traffic (AADT) is crucial for environmental assessments, infrastructure planning, and sustainable transport policy. However, comprehensive traffic data collection remains resource intensive and spatially sparse. We present a scalable machine learning (ML) framework that integrates over 900 contextual features (e.g. demographics) with engineered spatial statistical features (e.g. eigenvector spatial filtering) to predict AADT at over 19,000 locations across England and Wales. The framework is evaluated using two spatially informed cross-validation (CV) strategies: sampling-intensity-weighted and spatial block CV, achieving a test R² of 0.79 and 0.67, respectively. We systematically benchmark spatial feature contributions and show they enhance generalisation. Residual analysis confirms that the model effectively captures spatial dependencies. By integrating spatial statistical theory with scalable and interpretable ML pipelines, our framework addresses spatial autocorrelation and heterogeneity while remaining computationally efficient. The framework is broadly transferable to other spatial prediction tasks in environmental modelling, urban studies, and regional science.