Short-term analysis of urban noise dynamics during covid-19 lockdown using a machine learning approach in Mashhad, Iran

Noise pollution is an important environmental issue that has a significant effect on human health. This study investigates the short-term impact of human activities on noise pollution in the Mashhad, Iran, and how these patterns changed during the COVID-19 pandemic. By leveraging a dual strategy, a comprehensive analysis was initiated. First, the equivalent continuous sound level (Leq) measurements collected before (March 21 to April 20, 2019) and during (March 20 to April 19, 2020) the COVID-19 quarantine period at four key intersections in Mashhad were compared. The non-parametric Wilcoxon signed-rank test was employed to evaluate the statistical significance of the observed changes. The results showed a statistically significant reduction in Leq at all four intersections during the quarantine period. Next, a predictive modeling algorithm named random forest (RF) was developed to predict noise pollution levels by considering time factors such as month, day, hour, and cumulative hour. The RF model achieved a high R-squared (R²) value (0.914), representing a strong correlation between predicted and actual Leq. The predictive power of this model was demonstrated by the root mean square error (RMSE) of 0.967 and the mean absolute error (MAE) of 0.620, indicating reasonable accuracy. This case study demonstrates evidence that human activities are the main cause of noise pollution in Mashhad. The findings highlight the potential benefits of urban planning strategies that reduce traffic and noise generation. Furthermore, developing of a noise prediction model using a random forest approach provides a practical solution for future noise management efforts at the local scale.