A Comparative Analysis of Interpolation Methods to Produce a Spatially Continuous Map of PM2.5 over Thailand

Particle pollution in Southeast Asia has been aggravated as a consequence of increasing vehicular and industrial activity. Accordingly, PM2.5, has also become a matter of great concern globally due to its adverse effects on public health and environment. With this respect, monitoring and regulating PM2.5 has become essential in mitigating influences of increased PM2.5 in many countries. Thailand has suffered from particle pollution and attempts have been made to develop air quality monitoring systems. For example, the Pollution Control Department (PCD), the authority to control and monitor pollution-related problems, has established 91 stations for real-time measurement of air quality. However, since the spatial coverage of fixed ground monitoring stations is limited and establishing more stations is problematic due to their high installation and maintenance fees, alternative approaches to monitor PM2.5 concentration are necessary. Therefore, we investigated models to interpolate PM2.5 data from ground stations and produce a spatially continuous map of PM2.5 over Thailand. In this framework, inverse distance weighted (IDW), ordinary kriging (OK), regression kriging (RK), and random forest (RF) approaches were analyzed and compared, and data from the Sentinel-5P satellite were supplied to the model as spatially continuous predictors. We also attempted to integrate meteorological and industrial factors to enhance the performance of the model. We present the results from our findings and the comparative analysis of the 4 interpolation methods. Consequently, we demonstrate the advantages of producing a spatial continuous map of particulate matters over isolated point ground data.