Validating GTFS route geometry with one smartphone: on-board GPS logs, national bus-route GIS data and published shapes.txt on the community buses of Agui and Higashiura, Japan

This preprint documents two open workflows for generating the GTFS route-geometry file shapes.txt for a Japanese community bus, Agupii-go of Agui Town, Aichi Prefecture: Workflow A from on-board GPS ride logs recorded with a browser-based logger, and Workflow B as stop-to-stop shortest paths over the MLIT N07 bus-route polylines with an optional via-point correction. It pre-registers a validation protocol and reports its first stage, a survey on the neighboring Higashiura town community bus “u-ra-ra”, whose operator publishes shapes.txt. On two full-length rides logged with one handset, the published shapes lay within 25 m of the ride log over 97.8% and 99.8% of their length, and the network-based shapes over 88.0% and 70.3%, the shortfall arising on roads realigned after the N07 survey year (fiscal year 2022). An operator’s shapes.txt can be checked by a single rider, and a one-handset ride reproduces it; where routes have changed since the national bus-route data were surveyed, the network-based method cannot follow them. Code, generated shapes, the complete corrected Agui feed (feed_version 5.0), the Higashiura scripts and the two ride logs (calendar dates replaced by 2026-04-01) are archived at https://doi.org/10.5281/zenodo.23876967 (v3.0; concept DOI 10.5281/zenodo.20579190). The pre-registered protocol frozen before the survey is at https://doi.org/10.5281/zenodo.23579824 (concept DOI 10.5281/zenodo.21803782). An earlier version of the manuscript is on SSRN (https://ssrn.com/abstract=7105058). AI disclosure: The scripts, the logging application and this manuscript were drafted and edited with Claude (Anthropic); earlier versions of the scripts also used Google Gemini. All figures, numbers and verification results were produced by running the code. The author has checked them and takes responsibility for the content.