An efficient spatiotemporal data calibration approach for the low-cost PM2.5 sensing network: A case study in Taiwan.

The rapid growth of Internet of Things has provided a new aspect to air quality monitoring system. In Taiwan, over 5000 PM2.5 sensors have been installed in the last two years. The greatest asset of low-cost sensors is possibly mapping spatiotemporal air pollution with finer resolution. But the data quality of low-cost sensors is the most common question that how to take proper interpretation of the measurements. This study proposes an efficient calibration approach based on generalized additive model which is further applied to a particular low-cost PM2.5 sensor in Taiwan. The study carried out a field calibration that collecting both measurements of low-cost sensors and the regulatory stations, and investigated the space/time bias between the low-cost sensors and regulatory stations. Results show that the proposed approach can explain the variability of the biases from the low-cost sensors with R-square of 0.76. In addition, the present calibration model can quantify the uncertainty of the low-cost sensors observations and the average standard deviation is about 13.85% with respect to its adjusted levels. This operational spatiotemporal data calibration approach provides an useful information for local communities and governmental agency to face the new era of IoT sensor for air quality monitoring.