A Mobile Sensing Based Stochastic Model to Forecast AQI Variation of Pollution Hotspots on Urban Neighborhoods

Abstract

Due to massive population migration, most Indian cities have experienced fast urbanization, resulting in a significant increase in construction activity, traffic pollution, and uncontrolled expansion. Some of these cities also have a high concentration of polluting industries, significantly worsening air quality. Pollution hotspots exist in certain cities, with levels well surpassing the authorized mark. Air pollution is generally classified as extremely hyper-local, which signifies that the pollution index decreases as we travel away from hotspots. Since the pollution data collected from traditional sources is occasionally inadequate, the extended consequences of such hotspots on neighboring communities remain unidentified. If the flux in pollution values in neighboring locales is efficiently mapped for locations encountered travelling further from identified hotspots, AQI levels for these areas can be forecasted and projected. Knowledge from monitoring these levels will aid the city administrations and government in drafting suitable proposals for susceptible establishments like hospitals and schools. In this research work, the Air Quality Index (AQI) data was accurately gathered at an identified pollution hotspot and its immediate neighborhood over a defined period along a specific route and a mathematical model was developed to forecast how AQI varies with distance for best results. Stochastic models such as ARMA and ARIMA were used to create the predicted model. Its reliability and performance were measured using various forecasting error calculation methods such as MPE (Mean Percentage Error), MAP (Mean Absolute Percentage), MAD (Mean Absolute Deviation), RMSE (Root Mean Square Error), and MSE (Mean Square Error).