Association of observed ramp-electrification status with APU runtime and turnaround performance: A staggered difference-in-differences analysis at a tropical ASEAN hub

Ramp electrification can reduce aircraft auxiliary power unit (APU) use during turnaround, but operational evidence from tropical airports remains limited and sensitive to treatment-timing and implementation definitions. This study presents a replacement reanalysis of 7122 de-identified turnarounds across 20 stands at a tropical ASEAN hub from 1 March to 28 June 2025. Fourteen stands exhibited an observed transition to active treatment status, six remained never observed treated, and five first-observed treatment-date groups were identified. Because official commissioning records were unavailable, treatment timing was defined by the first observed active-treatment timestamp rather than independently certified go-live dates. A cohort-by-event interaction regression incorporated stand and calendar-date fixed effects and adjusted for aircraft body class, carrier, time of day, wet-day status, temperature, relative humidity, and rainfall. Supported post-treatment cohort-event coefficients were aggregated using cell-count weights, yielding a turnaround-weighted post-treatment estimand across the realized distribution of observed post-treatment turnarounds. Inference used CR1 stand-clustered covariance and a 9999-replication restricted-residual wild-cluster bootstrap- t . Observed treatment status was associated with a 5.874 min/turn reduction in APU runtime (95% CI −6.586 to −5.162; p < 0.001; wild-bootstrap p = 0.0001), approximately 32.0% of the eventually treated pretreatment mean. At 6.5 kg CO 2 e/min, this corresponds to 38.18 kg/turn of gross avoided APU-related CO 2 e, a deterministic transformation rather than a net-emissions estimate. Average turnaround time was +0.657 min/turn (95% CI −0.539 to 1.853; p = 0.281; wild-bootstrap p = 0.353), with variation across group-specific estimates, including +6.15 min/turn for the single-stand second treatment group; these differences are interpreted cautiously because post-treatment support and follow-up varied across groups. Joint pretreatment tests were not statistically significant for APU runtime ( p = 0.157) or turnaround time ( p = 0.528), although the earliest pooled APU lead was individually positive. The findings support a consistent association between observed electrification status and lower APU runtime while requiring caution regarding counterfactual comparability, causal interpretation, operational variation, and net-carbon interpretation.