Predictive modelling of fine-mode combustion aerosol events near Indonesian coal-fired power plants using AE-filtered Himawari-8 satellite data.

This study develops a predictive framework to identify fine-mode combustion aerosol events around coal-fired power plants (CFPPs) in Java, Indonesia, using Himawari-8 satellite data filtered using the Ångström Exponent (AE). Although an AE > 1.2 indicates fine-mode particles characteristic of combustion processes (including biomass combustion, traffic, and industrial emissions), the persistent nature of CFPP operations allows for the characterization of aerosol signals in their vicinity. Analysis of 41,119 daily observations (2015-2025) across 13 major CFPPs reveals that 8.52% of days meet the criteria for fine-mode combustion events (AE > 1.2 and AOD > 0.4) near these plants. With substantial spatial variability, Suralaya exhibits the highest mean AOD (1.01) and 51.38% very high pollution days, while Pacitan shows the cleanest conditions (mean AOD = 0.46). Three machine learning models-random forest, XGBoost, and LSTM-were trained for binary event classification, achieving near-perfect performance with random forest and XGBoost attaining perfect scores across all metrics (accuracy, precision, recall, F1-score, AUC-ROC = 1.000). Using a modified Monte Carlo approach incorporating seasonal patterns, temporal autocorrelation, and extreme events, we project fine-mode combustion aerosol events near CFPPs from 2026 to 2060. The modified Monte Carlo predictive modelling predicts 9.3% more aerosol event days than the standard Monte Carlo ( p < 0.001), with pronounced spatial heterogeneity: Pelabuhan Ratu shows the largest increase (+62.3%), while Cilacap exhibits a substantial decrease (-38.1%). Superior uncertainty quantification is evidenced by wider confidence intervals (212 vs. 172 days) and larger average standard deviations (82.7 vs. 43.1 days), indicating conventional methods produce falsely precise projections by ignoring natural variability. This model provides robust evidence for prioritizing emission control strategies and designing targeted interventions at specific power plant locations.