Bayesian Uncertainty Quantification and Sensitivity Analysis for the H₂ + OH → H₂O + H Reaction: A Comprehensive Comparison of Ten Kinetic Studies

Abstract The reaction H₂ + OH → H₂O + H is fundamental to hydrogen combustion, atmospheric chemistry, and energy systems. Despite numerous experimental studies, comprehensive statistical comparison with modern uncertainty quantification has been lacking. This study presents a systematic analysis of ten independent kinetic investigations (1981-2021) spanning 200-3044 K, employing multiple complementary approaches: (1) comparative statistical analysis of modified Arrhenius expressions from diverse experimental methods; (2) Bayesian uncertainty quantification determining posterior distributions with decomposed measurement and inter-study variability; (3) parameter sensitivity analysis; (4) machine learning validation; and (5) temperature-zone optimization. Bayesian analysis reveals 14.6% average uncertainty with excellent agreement (CV 10-20%) at combustion conditions (800-2000 K). Sensitivity analysis demonstrates that activation energy (Ea) dominates at low temperatures (T < 700 K, sensitivity >5), while the temperature exponent (n) becomes critical above 1500 K. We provide application-specific recommendations: atmospheric chemistry (Atkinson et al., 2004; Sutherland et al., 1996), combustion modeling (Yang et al., 2021), and high-temperature applications (Hong et al., 2010). Machine learning models achieve R² > 0.999 for interpolation but confirm modified Arrhenius expressions remain superior for extrapolation. All analysis code, datasets, and Bayesian posteriors are provided open-source. This comprehensive evaluation establishes robust uncertainty bounds for a critical elementary reaction and demonstrates the value of Bayesian methods in chemical kinetics. Keywords: Chemical kinetics, hydrogen combustion, Bayesian analysis, uncertainty quantification, sensitivity analysis, Arrhenius equation