The rapid expansion of industrialization and continued reliance and nonrenewable energy have significantly increased emissions including air pollutants in the atmosphere. Accurate forecasting of atmospheric pollutant concentrations is critical to inform effective environmental and climate mitigation strategies. This study presents a deep learning approach using the Temporal Fusion Transformer (TFT) to predict the levels of atmospheric pollutants in Indonesia, specifically in the Jakarta and Banten area. The research collected air pollutants values using the Sentinel-5P satellite from diverse regions characterized by varying socio-economic factors such as population density. The TFT model, renowned for its ability to capture both short-term and long-term temporal dependencies and provide interpretable forecasts, analyzes historical emission patterns and predicts future trends at a granular spatial resolution. The performance of the model is evaluated using the mean absolute error (MAE), the root mean square error (RMSE), and the coefficient determination (R 2 ). The results it achieves the best performance with the lowest Mean Absolute Error (MAE) of 0.0016, the lowest Root Mean Squared Error (RMSE) of 0.0020, and the highest coefficient of determination (R 2 ) score of 0.9622. These results indicate a superior ability of the TFT model to capture complex pollutant dynamics, providing valuable insights for environmental monitoring and policy decision making. Furthermore, the study emphasizes the relevance of air quality forecasting within the framework of the United Nations Sustainable Development Goals (SDG), particularly SDG 3 (Good Health and Well-being) and SDG 11 (Sustainable Cities and Communities). Accurate pollutant predictions support public health initiatives and urban planning efforts at reducing pollution exposure and enhancing urban resilience. This integration of advanced deep learning techniques with satellite-derived pollutant data represents a promising direction for scalable, data-driven environmental management in Indonesia and similar contexts globally.