Local-Scale Pollution Source Apportionment in Complex Port Environments Using High-Density Low-Cost Sensor Networks and Statistics Analysis

Conventional air-quality observations and model simulations often break down at the subkilometer scale of busy ports and industrial hubs, leaving a critical gap in identifying and apportioning local emission sources. Bridging this gap is essential for effective site-specific pollution control. Our study addresses this need by leveraging the fine spatial and high temporal resolution of low-cost sensor (LCS) networks capable of capturing pollutant variations at meter scales across complex, emission-intensive areas. During the 2019 China International Import Expo (CIIE), we deployed a dense network of LCSs across one of Shanghai’s major port areas to monitor key pollutants. Mean nitrogen-oxide (NOx) concentrations in the port were 3.4 times higher than urban background levels, with a distinct midnight–06:00 peak tied to port logistics. Fast Fourier Transforms (FFT) analyzed the frequency domain of the high-resolution time series, revealing periodic patterns unique to port operations and distinguishing them from urban influences. Principal Component Analysis (PCA) further decomposed spatial variability, grouping areas by dominant pollution patterns, and clarifying the propagation dynamics. Combining these statistical approaches with bivariate polar plots enabled us to trace high NOx levels primarily to heavy-duty trucks on internal roads, rather than to ships or nearby urban sources. Spatial clustering delineated source, susceptible, and low-impact zones, revealing how emissions spread from core transport routes into surrounding operational areas. Targeted truck-traffic restrictions during the CIIE effectively cut NOx levels, validating both the source apportionment strategy and the intervention strategy. Our results demonstrate that LCS networks, coupled with advanced statistical analysis, offer a scalable, real-time framework for managing complex microenvironmental pollution.