Abstract: Introduction: Environmental pollution is the leading cause of death worldwide. Added to this problem is the large concentration of people living in urban areas. This condition directly impacts the population’s human health due to the exposure. Another aspect is the high cost of acquiring and maintaining conventional monitoring stations, which makes networks with adequate spatial distribution economically unviable. To deal with this issue, low-cost sensors (LCS) have been pointed out as an alternative, have gained notoriety, and have been used worldwide for several purposes. Objective: In this direction, the main objective of the EnvCity project (FAPESP, process 2020/05135-5) is to study the LCS, their performance, and the correction methods of data and integrate them, making the developing of devices (architecture, integration of components, and transmission) and of the systems to operate providing data in real-time. Method: The literature on LCS was revised using the main scientific databases and several intercomparison tests between LCS and reference stations or equipment performed. As a tool to correct the dataset, machine learning methods were evaluated and used to study and improve the performance of the measurements done by the devices developed. The system’s architecture, the firmware, the printed circuit board, and the case of devices were also developed. Findings: The results, for air quality are summarized: i) the LCS for air quality sensors are sensitive to do measurements and suffer from the influence of temperature and relative humidity, and to minimize these effects, they must be in a box with an adequate size and air circulation, besides the corrections needed; ii) although some LCS are calibrated in the factory, the measurements should be corrected in the local; iii) the performance of LCS for particulate matter (PM) present agreement among them; iv) machine learning methods for data processing are the most viable alternatives identified to correct the measurements; v) the cost of DustAI developed for PM is around 320 USD, while the station (CO, NOx, O3, SO2, PM) is around 850 USD; vi) the devices developed are plug and play, which makes it a more accessible and more economical way to change the sensor; vii) the use of low-cost air quality and particulate matter stations can be useful in places with low or lack of reference stations, which together with other ways of obtaining environmental data contribute and are recommended for the study of air pollution and health. For water quality, we also found interesting results, which can be highlighted here that few studies evaluated the sensors against reference equipment in the applications and that an effort should be made to perform long-term studies for multiple comparisons in the laboratory and field, as well as, studies that support the development of calibration and correction protocols low-cost water quality. We also suggest the concept for water quality LCS. Finally, the network architecture with several ways to visualize the dataset was developed, including the EnvCity app. for mobile, with various features and tools from real-time data visualization to interactive mapping capabilities and detailed reporting.

Keywords: low-cost sensors, air quality, internet of things, low-cost devices, particulate matter