Statistical Modelling of low cost PM2.5 sensor data in Dar es Salaam City

The rapid urbanization and industrialization in many parts of the world have made air pollution a global public health problem. Exposure to air pollutants has both acute and chronic impacts on health. In low and middle income countries like Tanzania have experienced accelerated population growth and urbanization in which air quality is generally poor and there is a lack of long-term reliable air quality monitoring.

Among the major air pollutants, particulate matter 2.5 (PM_2.5) is the most harmful, and its long-term exposure can impair lung functions. Low-cost sensors (LCS) are becoming increasingly popular for measuring the level of particulate matter (PM) in the air. However, issues with reliability require calibrations before the sensors can be used in regulatory settings. The aim of this paper was to develop a statistical model for determining the accuracy of low-cost sensor network data. Considering PM2.5 adverse health impacts, especially on people’s respiratory systems. We therefore, developed a low-cost PM_2.5 sensor calibration model for measuring PM2.5 concentrations using maximum likelihood method. Moreover, we used the model to predict the PM_2.5 with its driving forces that is temperature and humidity, and estimated its parameters. The PM2.5 data used for the developed model were collected from LCS network of five stations from Dar es Salaam City including Kigamboni, Vingunguti primary, Jangwani, Ubungo, and Buza recorded from April 2022 to May 2023. The data was fitted to a regression model using Maximum-Likelihood (MLR). Descriptive and trend analysis was also performed using Mann-Kendall Trend analysis to describe the pollutant characteristics and identify significant trends in the selected stations in Dar es Salaam. The model performed well with high accuracy and performance with root mean of 3.58 and mean squared errors of 12.846, a coefficient of determination of 0.967, and mean absolute errors of 2.8.The results for MLR showed a high value of coefficient determination (R²=0.82) and low error measure.

Our results will aid in improving the accuracy of low-cost sensors for measuring PM2.5 concentrations, thereby providing cost-effective solutions for enhancing people’s health and well-being in Tanzania.