DIAN KURNIAWATI. Atmospheric Condition Modeling Related to Three Aircraft Accident Cases in Indonesia. (supervised by Prof. Dr. Halmar Halide, M.Sc).
Background: Aircraft accidents in tropical Indonesia often occur under complex and highly dynamic atmospheric environments characterized by strong weather variability, making atmospheric analysis essential in aviation safety studies. Previous studies were largely descriptive and did not integrate multi-parameter reanalysis data with quantitative predictive modeling approaches. Objective: This study aims to analyze atmospheric dynamics in three aircraft accident cases in Indonesia and to identify pre-event atmospheric patterns using the Random Forest algorithm. Methods: ERA5 reanalysis data were utilized to examine horizontal wind shear, two-meter air temperature, mean sea level pressure (MSLP), total column water vapor (TCWV), and vertically integrated moisture divergence (VIMD). Modeling was conducted using data from six hours prior to each event through Stratified Cross Validation and evaluated using confusion matrix, Heidke Skill Score (HSS), Out-of-Bag (OOB) error, Receiver Operating Characteristic (ROC), and Area Under the Curve (AUC). Results: The findings indicate atmospheric instability characterized by significant wind shear, high moisture content, and moisture convergence across all cases. The model demonstrated good discriminative performance with AUC values ranging from 0.87 to 0.93, although sensitivity to accident events remained limited as reflected by relatively low HSS values. Conclusion: The integration of atmospheric dynamics analysis and machine learning effectively identifies pre-event atmospheric patterns in an objective and measurable manner; however, it cannot determine accident causation and should be interpreted as an environmental risk indicator for aviation safety.
Keywords: wind gradient; moisture transport; stratified validation; ensemble learning; integrated vapor content; forecast skill score
Â
