Data science-based tool to reduce measurement errors in atmospheric monitoring systems (#1185)

Authors

González, Jonathan

Quintero, Eladio

García, Felisindo

García, Antony

Sáez, Yessica

Collado, Edwin

Abstract

The growing concern about air pollution and climate change, derived from global industrial growth, has generated environmental disturbances with unpredictable consequences for human health. Among the most harmful polluting agents are particles suspended in the air, known as particulate matter, whose rapid dispersion over large areas aggravates the problem. With the aim of addressing this problem, an urban atmospheric monitoring system was developed based on the Internet of Things that allows measuring the concentrations of PM2.5 and PM10 particulate matter. This system is mainly composed of a Nova PM SDS011 sensor to take measurements of PM10 and PM2.5, a DHT22 sensor to record temperature and humidity levels, and an ESP32 microcontroller programmed in MicroPython, which was responsible for data collection and transmission. via WiFi. The system was validated using the AEROQUAL S500 as reference equipment, where it was concluded that the proposed measurement station satisfactorily measures the variables studied. However, during the validation period it was observed that part of the data collected presented errors, either because the measurement was corrupted, the system did not work correctly and/or an unwanted value was recorded. Therefore, this work proposes to develop a tool based on Data Science to adequately process the data and reduce the amount of errors in the collected measurements. Specifically, this tool uses Python to apply outlier cleaning technique based on Interquartile Range. This allows erroneous readings caused by external factors to be identified and corrected. This seeks to generate accurate information on air pollution levels to support decision making. The developed prototype demonstrates that, through the use of the Internet of Things and simple Data Science techniques, it is possible to provide an urban atmospheric monitoring system with high precision, low cost and reliable data, allowing a better understanding of the risks to the health to support the adoption of environmental protection measures.