Natural Hazards Research Summit 2024: Improving Low-Cost Air Quality Monitoring Network through Probabilistic Spatiotemporal Modeling: Enabling Smart Cities in Coastal Regions of Environmental and Industrial Change

In this poster, we present a new data-driven approach that integrates spatiotemporal information to increase the data quality of post-processed particulate matters 2.5. This method assimilates multiple data channels from nearby monitoring stations using a probabilistic spatiotemporal model that describes a two-dimensional advection and diffusion process with stochastic uncertainties. The assimilated data consists of two types of measurements: highly accurate readings at distant stations and fairly accurate readings at the ambient locations in the area of interest. We apply the method to post-process monitored PM2.5 concentrations from low-cost sensors (SCI-608, SailBri Cooper Inc.) deployed near Corpus Christi, Texas. We test the prediction performance by comparing the post-processed low-cost sensor data with co-located regulatory-grade sensor measurements. Our findings demonstrate that integrating heterogeneous and multi-modal datasets in the post-processing of sensor data can improve data quality and accuracy of the monitoring network while enabling uncertainty-aware predictions.