The impact of industrial activities on the surrounding environment based on hybrid filter and machine learning

Industrial development has emerged as a significant driver of environmental degradation and urban heat island (UHI) formation. However, studies explicitly addressing the long-term spatial impact of heavy industries—particularly in tropical, cloud-prone regions—remain limited due to persistent data gaps and noise in satellite observations. This study addresses that research gap by analyzing the environmental effects of industrial activities in Cilegon City, Indonesia—one of the nation's largest industrial zones—using monthly Landsat-8 time series data from 2014 to 2022. A hybrid filtering approach was applied to reconstruct high-quality data by removing cloud and cloud shadow interference. The reconstructed NDVI and LST were then used as multivariate input features to model Land Surface Temperature (LST) using the XGBoost algorithm, with 30-m spatial resolution. The predicted LST was subsequently analyzed alongside NDVI to examine spatio-temporal trends and quantify industrial heat island (IHI) effects. Results show that industrial heat extends up to 1.5 km from core industrial zones, with IHI intensity reaching 5.58 °C in 2022. Vegetation health showed severe decline, with NDVI values dropping by 81.36 % in industrial cores and 29.25 % in adjacent areas. LST exhibited a positive trend of 0.23 °C/month in highly industrialized subdistricts and maintained a strong negative correlation with NDVI (r = −0.95). These findings highlight the amplified environmental impact of industrial activities in cloud-prone tropical cities and emphasize the urgent need for sustainable land management and the implementation of green infrastructure to mitigate local warming and protect surrounding ecosystems.