This study aimed to enhance the practical utility, in real operating contexts, of warning sounds designed to prevent declines in train drivers’ arousal. We analyzed subjective ratings of audibility and arousal for warning sounds presented at 75 dB under varying running noise levels (from 65 dB to 85 dB), considering age as the grouping factor. Experiment 1 (general participants, five noise levels × five warning types (four arousal-designed, one monotone)) showed that increasing noise decreased audibility and, in turn, arousal ratings; moreover, the ≥40s group reported higher arousal than the 20s–30s group. Under high-noise conditions, lower-frequency arousal-designed warning sounds were rated lower than the monotone, such that ratings could fall to levels at which neither audibility nor perceived arousal was attained. Experiment 2 (train drivers, two noise levels × three warning types showed the same qualitative tendency, providing ecologically relevant support for the Experiment 1 finding that low-frequency arousal-designed warnings perform poorly under high noise; no clear age-group difference was observed. Experiment 3 (in-service vehicles) lends support to the findings of Experiments 1 and 2. Overall, under high running noise, warning sound effectiveness depends on frequency, suggesting that designing accordingly can improve practicality even for arousal-designed sounds.