Fine-Grained Urban Air Quality Mapping from Sparse Mobile Air Pollution Measurements and Dense Traffic Density

Urban air quality mapping has been widely applied in urban planning, pollution control and personal exposure assessment. maps are traditionally derived using measurements from fixed monitoring stations. Due to high cost, these stations generally sparsely deployed a few representative locations, leading highly generalized map. In addition, varies rapidly over short distances (<1 km) is influenced by meteorological conditions, road network traffic flow. These variations not well represented coarse-grained generated conventional fixed-site methods but have important implications for characterizing heterogeneous exposures identifying localized hotspots. Therefore, fine-grained indispensable. this context, supplementary low-cost mobile sensors make promising alternative. Using sparse collected various contextual factors, especially flow, we propose context-aware locally adapted deep forest (CLADF) model infer the distribution of NO2 100 m 1 h resolution mapping. The CLADF exploits construct local each cluster consisting nearest neighbor feature space, considers flow as an feature. Extensive validation experiments were conducted 17 postal vans equipped with operating Antwerp, Belgium. experimental results demonstrate that achieves lowest RMSE advances accuracy correlation, compared benchmark models, including random forest, extreme gradient boosting support vector regression.