Accuracy and Risk Assessment of Rail Temperature Estimation Using Ground Observations and KMA Gridded Air Temperature

Although wide-area monitoring of rail temperature (T<sub>R</sub>) is increasingly needed to address the heightened risk of track buckling under climate change, existing point-based analysis remain limited in spatial scalability. This study quantitatively evaluated the error characteristics of T<sub>R</sub> estimation and the associated limitations in risk assessment when using spatially gridded Air Temperature (T<sub>a</sub>) data provided by the Korea Meteorological Administration (KMA). Measured T<sub>R</sub> from two study sites, Namsan Junction and Osong Railway Comprehensive Test Track, were used as benchmarks to compare T<sub>R</sub> estimates derived from point-based T<sub>a</sub> and gridded T<sub>a</sub>. A probabilistic risk assessment using cumulative distribution functions was performed with a critical T<sub>R</sub> threshold of 50 ℃, informed by actual accident cases. The analysis showed that point-based T<sub>R</sub> estimates overestimated T<sub>R</sub> by up to +9.6 ℃ during summer daytime, resulting in a calculated probability of exceeding the 50 ℃ that was 2 to 8 times higher than the actual risk. In contrast, grid-based T<sub>R</sub> estimates consistently underestimated T<sub>R</sub>, primarily due to scale-mismatch effects. This produced an opposite error pattern, with most cases indicating a 0% exceedance probability even when actual TR values surpassed the critical threshold. Based on these fundamental error characteristics, future research should focus on developing more reliable wide-area T<sub>R</sub> estimation methods by validating and fusing diverse spatial datasets, including satellite observations and high-resolution numerical models.