SIG-Bus has a well-defined purpose: read GTFS feeds inside QGIS, cross-reference them with per-stop boarding records, and deliver synthetic reports per line, with no external dependencies beyond QGIS itself. The project started as a PIBIC undergraduate research initiative at CEFET-MG in 2020 (with the application filed in 2019) and was presented at the institution’s Science and Technology Week in 2025.
The integration problem
QGIS has no native GTFS support beyond basic geometry loading. Analyzing a bus network means cross-referencing three pieces of information that live in separate sources: route geometry, scheduled trips, and real passenger demand. Without a dedicated tool, bringing them together is manual, repetitive work.
Boarding data is available on the City of BH’s open data portal as CSV files with their own coordinate system and line codes that don’t map directly to GTFS identifiers. Building a reliable spatial link between the two datasets, without generating false matches with nearby parallel routes, was the core technical challenge.
How the plugin works
SIG-Bus organizes the workflow into six sequential steps:
- GTFS Validation: checks feed integrity and generates synthetic calendars when
calendar.txtis missing. - GTFS Import: converts files to GeoPackage, creating indexed spatial layers for stops, route shapes, and trips.
- Demand Import: loads the boarding CSV and spatially links demand points to GTFS stops, restricted to each line’s dominant service pattern to prevent false positives.
- Route and time period selection: the user picks a line and a time slot (full day or a specific hour, 0h to 23h).
- Demand allocation: generates the
tramos_demandalayer with accumulated load and trip counts per segment. - Visualization: graduated styles are applied automatically, highlighting the highest-volume segments.
The departure schedule layer
When filtering a line, the plugin automatically generates the horarios_paradas layer, a point layer with scheduled times for all trips on the selected route.
For each stop, the layer records:
arrival_timeanddeparture_timestop_sequenceandstop_nametrip_id,shape_id, anddirection_id
The layer is built from a direct join across the stop_times, trips, and stops tables in the GeoPackage database, indexed by trip_id. It makes it straightforward to visualize service distribution throughout the day, identify stops with irregular headways, or check hourly coverage for a specific line.
Like all SIG-Bus layers,
horarios_paradasis memory-based and must be regenerated after reopening QGIS.
Segment demand allocation
The tramos_demanda layer is the plugin’s main analytical output. The load on each segment is calculated using flow conservation: boardings at stop i add to the load from the previous segment.
Since the fare system records only boardings, without alighting data, the method assumes all passengers alight at the final stop. This produces an upper bound on actual occupancy, appropriate for:
- Relative comparison between segments on the same line
- Identifying the most loaded sections by time period
- Analyzing hourly demand distribution
The METHODS.md file in the repository details the assumptions, edge cases, and interpretation guidelines.
Science and Technology Week 2025 and current state
The project was presented at CEFET-MG’s Science and Technology Week in 2025. After that version, two additions were incorporated: cluster-based categorization of demand segments with automatic graduated styling, and PDF report generation with two maps (accumulated load and cluster grouping) overlaid on the route geometry.
The full repository is available on GitHub. All testing was done with the real GTFS feed for the city of Belo Horizonte, and the docs/gtfsfiles.zip file contains that same data for anyone who wants to reproduce the analyses.
If you’d like to try the plugin and don’t have the file at hand, I can send the .zip directly. Just ask by email (eng.dcamargo@outlook.com.br) or WhatsApp (+55 31 3565-3353).