Detection of PM_2.5 plume movement from IoT ground level monitoring data

In this study, we analysed a data set from 10 low-cost PM_2.5 sensors using the Internet of Things (IoT) for air quality monitoring in Mae Sot, which is one of the most vulnerable areas for high PM_2.5 concentration in Thailand, during the 2018 burning season. Our objectives were to understand the nature of the plume movement and to investigate possibilities of adopting IoT sensors for near real-time forecasting of PM_2.5 concentrations. Sensor data including PM_2.5 and meteorological parameters (wind speed and direction) were collected online every 2 min where data were grouped into four zones and averaged every 15 min interval. Results of diurnal profile plot revealed that PM_2.5 concentrations were high around early to late morning (3:00-9:00) and gradually reduced till the rest of the day. During the biomass burning period, maximum daily average concentration recorded by the sensors was 280 μg/m³ at Thai Samakkhi while the minimum was 13 μg/m³ at Mae Sot. Lag time concentrations, attributed by biomass burning (hotspots), significantly influenced the formation of PM_2.5 while the disappearance of PM_2.5 was found to be influenced by moderate wind speed. The PM_2.5 concentrations of the next 15 min at the downwind zone (MG) were predicted using lag time concentrations with different wind categories. The next 15 min predictions of PM_2.5 at MG were found to be mainly influenced by its lag time concentrations (MG_Lag); with higher wind speed, however, the lag time concentrations from the upwind zones (MS_Lag and TS_Lag) started to show more influence. From this study, we have found that low-cost IoT sensors provide not only real-time monitoring information but also demonstrate great potential as an effective tool to understand the PM_2.5 plume movement with temporal variation and geo-specific location.