Abstract: Air quality evaluation is a crucial component of urban planning and control, closely related to human health and society development. The rational deployment of air quality monitoring stations is vital for predicting the spatial variations of air quality. This study aimed to evaluate the minimum sample size of air monitoring points to represent the optimal prediction accuracy of air NO₂ and to establish reliable inputs for predictive models. We designed a methodological framework for spatial optimization of air quality monitoring networks based on spatial simulated annealing algorithm, with the ground-level NO₂ monitoring stations in the southwest Fujian of China as an example. The layout of monitoring stations was improved based on the spatial simulated annealing algorithm with the minimum Kriging variance, enhancing the precision and effectiveness of air pollution monitoring. During the optimization process, a perturbation algorithm was used to change the layout of monitoring stations. The minimum Kriging variance (MKV) was employed for evaluation (without considering the total financial budget limit), seeking the highest optimization efficiency in sampling quantity. By optimizing the spatial layout of ground-level NO₂ air pollution monitoring stations, we found that after adding 160 monitoring stations, the MKV of NO₂ decreased from 0.09 μg·m^-3 to 0.02 μg·m^-3, and the prediction accuracy was increased by 71.46%. This optimization is likely to improve the coverage and sampling density of the monitoring system, enhancing the accuracy and reliability of monitoring data. This is crucial for timely identification and response to pollution sources, monitoring air quality trends, and formulating environmental policies. Moreover, this method could potentially be applied to monitor other forms of pollution (water, soil, and noise). This study provides a reference for the future optimization of air monitoring station deployment and management.

Key words: monitoring station location selection, spatial simulated annealing algorithm, empirical Bayes Kriging, NO₂ monitoring, layout optimization, point optimization