• Home
  • FT
  • GE4007
  • Predicting Air Pollutant Levels using ML models and GIS

Jha Satyamkumar Manoj

GE4007

Predicting Air Pollutant Levels using ML models and GIS

This study forecasts PM2.5 air pollution in Mumbai using Random Forest and LSTM models, trained on hourly data from 2021 to 2024 and validated with observations from the SAFAR portal. Data from 20 air quality monitoring stations across Mumbai was used to capture spatial and temporal variations. The LSTM model achieved high accuracy (R² = 0.88). Scenario testing showed that a 20% reduction in vehicular emissions can significantly improve air quality. Ward-level analysis revealed 739,836 residents, 379 of 803 hospitals, and 56 of 147 schools are in high-risk zones, highlighting the need for targeted urban interventions.

View Additional Work

Report Content

Understanding Air Pollution: Global Crisis and Local Impact

Why Mumbai? Urban Growth, Pollution Sources, and Strategic Importance

Aim, Objectives, and Methodology

PM2.5 Patterns: Seasonal Trends and Influencing Factors

Random Forest Modeling: Feature Selection and Model Optimization

Visualizing Predictions and Testing Scenarios

LSTM Model: Capturing Spatiotemporal Patterns in Air Quality

Prediction Map, Model Validation and Impact on Sensitive Groups

Uncovering the Drivers: Human Activity and Environmental Factors

Key Findings, Limitations, and Future Directions