Platform
The engine behind every comparison.
Farleda combines live European timetables, road and place data, historical or modelled hourly grid profiles and published emission factors, then ranks, grades and signs every journey. This page documents what the product computes and how.
From open data to a signed decision
Three layers, one pipeline per journey.
Data in
- Timetables and realtime from open GTFS feeds
- Roads, places and airports from OpenStreetMap and OurAirports
- Grid intensity for 31 countries, with historical or modelled hourly profiles
- Seven emission factors, plus zero for walking and cycling
- Weather, air quality and destination knowledge
Farleda engine
- Connection ranking
- Emissions at the hour of departure
- Data-quality grade per journey
- Plausibility filter
- Disruption check against realtime
- Calibration from real trips
- Signed calculation records
Results out
- Ranked journeys with transfers and times
- Emission ranges with sources and version
- Trip Carbon Passport with signature
- Public JSON endpoints for grid, modes and verification
- Corridor reports for pilot partners
Data sources
Every layer the product queries, what it is used for and how fresh it is.
| Layer | Source | Used for | Freshness |
|---|---|---|---|
| Timetables and realtime | Transitous (MOTIS) from open GTFS and GTFS-Realtime feeds of European operators | Train, coach and ferry connections, transfers and delays | Queried live per journey, cached up to 6 h |
| Road routing | OSRM on OpenStreetMap | Car and access legs with real road geometry | On demand |
| Places and points of interest | Nominatim and Overpass (OpenStreetMap) | Geocoding, stations, accommodation and sights | On demand |
| Airports | OurAirports | Distances for flight comparisons | Downloaded on demand, cached |
| Grid intensity, annual | Ember European Electricity Review 2024, EEA 2023 | Electric rail in 31 countries when no hourly value exists | Embedded snapshot |
| Grid intensity, hourly | electricityMaps history or Farleda model | Electric-rail estimate for the departure hour in UTC; not a travel-date forecast | Recent history where available, otherwise modelled; cached 1 h |
| Emission factors | UK DESNZ 2026, revised 31 July 2026 | Seven emission factors, with a range where the source gives one | Updated with methodology releases |
| Weather and air quality | Open-Meteo | Trip-day forecast and air quality at destinations | Forecast up to 14 days |
| Destination knowledge | Wikipedia and Wikivoyage | Destination summaries in discovery | On demand |
| Booking hand-offs | ÖBB, WESTbahn, DB, SBB, FlixBus | Deep links into the operator's booking | Built per selected connection |
| Accommodation search | Booking.com, Ecobnb, OpenStreetMap | Search links for stays at the destination; a Booking.com affiliate ID may be attached | Built on request |
Licences and attribution for every source are listed in the data inventory. Open the data inventory →
01 · Ranking
Connection ranking
Every public-transport option is scored on how well it serves the traveller, not on modal share alone. A long, fragmented itinerary cannot beat a practical one just because it stays on rail. A second stage keeps the Pareto-optimal routes and orders them by the traveller’s priority.
- Preferred mode presentRail, coach or ferry, as chosen
- 25
- Share of riding time in that modeWalking excluded
- 30
- Duration against the fastest optionFastest = full score
- 25
- Simplicity1 ÷ (1 + 0.45 × transfers)
- 15
- Access qualityLess time in local access legs
- 5
- Tie-break
- Shorter total duration, then original order
- Second stage
- Pareto filter on duration, changes and emissions; priority weights emissions/duration/changes: balanced 45/35/20 · time 15/70/15 · carbon 72/18/10
02 · Emissions
Hourly emission estimates
Farleda estimates electric-rail emissions using a fixed energy-use assumption and a grid value for the departure hour in UTC. Where configured, the hourly profile is derived from recent provider history; otherwise it is modelled around the country’s annual average. These are calculation inputs, not measured emissions for a particular train or a forecast for the travel date.
Same train, different hour
Modelled day profiles for three grids, computed exactly as the product does when no live data is available. In this model, solar-heavy grids have lower intensity at midday and higher intensity in the evening; this is not a forecast.
DEGermany
ATAustria
ESSpain
- Energy use
- 0.09 kWh per passenger-km for electric intercity rail
- Modelled profile
- Morning, midday, evening and night factors around the annual average, per country
- Range
- −15 % to +30 % around the expected value
- Cross-border legs
- Average of the two endpoint countries
03 · Provenance
Data-quality grade per journey
Each journey shows how much of it rests on scheduled services and specific data, weighted by kilometres. A modelled result is never presented as a measured one.
Journey score from 0 to 1, weighted by kilometres
- Route source
- Scheduled service 1.0 · road routing 0.95 · great circle 0.85 · detour estimate 0.5 · unknown 0.35
- Emission grade
- Route option graded A 1.0 · B 0.86 · C 0.66 · D 0.42; otherwise live grid 1.0 · modelled hourly 0.78 · annual average 0.65 · published factor 0.7
- Transit legs
- 65 % route · 20 % schedule match · 15 % emission grade
- Other legs
- 80 % route · 20 % emission grade
04 · Filter
Plausibility filter
Every mode is classified for the distance as best, reasonable, a stretch or absurd. Absurd options are hidden, stretches are dimmed and the greenest plausible mode is suggested first.
- Rail
- Best up to 2,500 km
- Short-haul flight
- Reasonable between 300 and 1,500 km only
- Walking and cycling
- Best up to 8 km and 40 km
05 · Realtime
Disruption check
Saved journeys are re-queried against realtime data at their original departure times. Accepting an alternative replaces the leg and re-issues the passport.
- Delayed
- Same service, realtime departure 5 min or more late
- Alternate
- A different service now serves the leg
- Unknown
- No answer for the leg; the traveller is asked to re-check
06 · Learning
Calibration from real trips
When travellers report what actually happened, Farleda compares it with its estimate and derives a correction factor per mode group: rail, coach, road and air, with a global fallback. The median keeps a single unusual trip from moving the result, and the factor is always limited to 0.5–2.0.
- Below 10 tripsNo effect
- 10 to 30 tripsEffect ramps in linearly
- 30 trips and moreFull factor
07 · Integrity
Signed calculation records
Every Trip Carbon Passport is signed with HMAC-SHA-256 over its content and methodology version. Any change to the record invalidates the signature. Once verification is released, anyone can check it on the verification page.
- Signed content
- Every leg with route, mode, distance and emissions, the grid countries used, the flight comparison and the methodology version
- Verification
- Public endpoint and verify page, no account needed
Go deeper
The methodology explains every factor and limit; the data inventory lists sources and licences; the API documentation shows the public endpoints.
Next stop
Start with one corridor.
Twelve weeks, one agreed use case and a documented basis for your decision. Tell us which corridor you want to understand better.
