Industrial heat island and carbon emissions: machine learning-based environmental impact assessment in Cilegon, Indonesia

Abstract

Industrialization has become a major environmental challenge. It intensifies urban heat island effects, degrades vegetation, and increases carbon emissions, particularly in rapidly developing regions. However, quantifying and mitigating these impacts remains a significant challenge, especially in cloud-prone areas where remote sensing data is often compromised. This study employs an advanced machine learning-based approach to assess the environmental impacts of industrial activities in Cilegon, Indonesia. Multi-source remote sensing datasets from 2014 to 2022 were analyzed. Integrating Landsat-8 for Land Surface Temperature (LST) and vegetation health (NDVI), GPM for precipitation, and ODIAC for carbon emissions. A hybrid filter approach enhanced data quality by reducing cloud-induced noise. An XGBoost model was developed to reconstruct LST, enabling a spatiotemporal assessment of industrial heat island (IHI) dynamics. The results reveal a staggering increase in carbon emissions, with Industrial Area 2 emitting 107 times more carbon than Industrial Area 1 due to coal-fired power plants. The industrial heat island effect extends to 1500 m, while carbon emissions significantly influence areas within 1000 m, exacerbating environmental stress. NDVI analysis indicates an 81.36 % decline in healthy vegetation in Industrial Area 2 between 2014 and 2019, emphasizing the severe ecological impact of industrialization. Seasonal trends show that La Niña-induced precipitation partially aids vegetation recovery, but its effects cannot counterbalance industrial degradation. Strong negative correlations were observed between NDVI and LST (-0.928) and NDVI and carbon emissions (-0.739), reinforcing the crucial role of vegetation in mitigating environmental damage. This study presents a data-driven framework to assess industrial environmental impacts using machine learning and enhanced satellite processing. The findings emphasize the urgent need for stricter emissions control, urban greening, and sustainable industrial planning to mitigate environmental degradation. Given the global rise in industrial emissions and urban heat stress, similar mitigation strategies are essential to enhance climate resilience and ecosystem sustainability.