Estimating corridor-level bus running-time variability from identifier-free segment GPS in a data-scarce city: evidence from Addis Ababa
Most bus reliability studies rely on high-frequency automated vehicle location (AVL) data, leaving data-scarce cities without systematic running-time evidence. This study develops and stress-tests a reproducible data-fusion framework for estimating corridor-level bus running-time variability from identifier-free GPS segment records. Applied to 65,543 Sheger Bus records from Addis Ababa, endpoint clustering identifies eight candidate hubs, and 10,326 records (15.8%) enter the corridor analysis. The buffer time index (BTI) for Mexico-to-Kolfe/Jemmo 1 movements is 49.78 (95% CI: 44.75–54.68), 18.3 percentage points above the Megenagna-to-Bole Arabsa corridor. This contrast persists under alternative hub buffers, outlier rules, spatial-linkage restriction, trip-length control, and calendar equalization, although absolute BTI levels depend on the outlier rule. Network-wide screening identifies 24 candidate low-speed locations, four prioritized for field validation. Scheduled GTFS running times differ up to twofold within the same corridor, confirming the feed is a spatial reference, not a timetable.
Authors
- Frezewd Lemma (ORCID: https://orcid.org/0009-0009-5030-9480)
- Junseok Hwang (ORCID: https://orcid.org/0000-0003-2415-1711)
- Ashenafi Duguma Ido (ORCID: https://orcid.org/0009-0001-3533-5922)
Institutions
- Seoul National University (KR)
- Adama Science and Technology University (ET)
Publication Details
- Journal
- Transportation Letters
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1080/19427867.2026.2765312
- Primary Topic
- Transportation Planning and Optimization
- Type
- article
- Field-Weighted Citation Impact
- 0.00