> ## Documentation Index
> Fetch the complete documentation index at: https://code.dcycle.io/llms.txt
> Use this file to discover all available pages before exploring further.
# Transport Type Inference (do_not_know)
> How Dcycle automatically infers the transport mode and distance when transport_type is set to do_not_know
# Transport Type Inference
When a transport section is created with `transport_type: "do_not_know"`, Dcycle automatically infers the most appropriate transport mode and calculates the distance. This page explains the exact decision logic.
## Why use do\_not\_know?
Use `do_not_know` when you know the origin and destination of a shipment but not the transport mode — for example, when processing supplier invoices that don't specify how goods were shipped. Dcycle applies a standardized inference algorithm to ensure the most realistic emission factor is used.
`kms_manual` is **not allowed** for `do_not_know` sections. Because the emission factor depends on the inferred transport mode (which itself depends on geography), manual distances cannot be meaningfully combined with automatic mode selection.
## Inference Algorithm
The inference runs at section creation time and follows this decision tree:
```
Given: origin address, destination address
│
▼
1. Geocode both addresses → get coordinates + country codes
│
▼
2. Check for sea crossing
Same country AND different landmass?
(e.g. mainland Spain → Canary Islands, mainland Italy → Sicily)
│
┌─────┴──────┐
YES NO
│ │
▼ ▼
Use maritime Same country AND road-reachable via Google Maps?
distance (status=OK, distance > 0 km, not a default fallback)
(haversine │
or searoute) ┌───┴───┐
YES NO
│ │
▼ ▼
Use road Use air distance
distance (haversine great-circle)
(Google Emission factor:
Maps) "aircraft, unspecified"
travel_method → truck
```
## Decision Rules
### Rule 1 — Sea crossing (same country, different landmass)
**Condition:** Origin and destination share the same country code, but their coordinates resolve to different landmasses (e.g. a continental mainland vs. an island).
**Outcome:**
* **Distance:** maritime (navigable-water via searoute, or haversine fallback)
* **Emission factor:** air — specifically the "aircraft, unspecified" concept
This special case exists because island-to-mainland routes involve a sea crossing where ferry emission factors are unavailable in ecoinvent. Air unspecified provides the most conservative and available estimate.
**Examples:**
* Madrid, Spain → Las Palmas, Gran Canaria (Spain) → **maritime kms, air EF**
* Rome, Italy → Palermo, Sicily (Italy) → **maritime kms, air EF**
* Lisbon, Portugal → Funchal, Madeira (Portugal) → **maritime kms, air EF**
### Rule 2 — Same country, road-reachable
**Condition:** Origin and destination are in the same country, no sea crossing detected, and Google Maps returns a valid driving distance (status OK, distance > 0, not a hardcoded default).
**Outcome:**
* **Distance:** road (Google Maps Distance Matrix)
* **Emission factor:** road freight
* **travel\_method** is automatically set to `truck`
**Examples:**
* Madrid, Spain → Barcelona, Spain → **road, truck**
* Paris, France → Lyon, France → **road, truck**
* Berlin, Germany → Munich, Germany → **road, truck**
### Rule 3 — Different countries (or road unreachable)
**Condition:** Everything else — different countries, or same country where Google Maps cannot return a driving route.
**Outcome:**
* **Distance:** air (haversine great-circle between geocoded coordinates)
* **Emission factor:** air — "aircraft, unspecified"
* **travel\_method** stays `null`
**Examples:**
* Madrid, Spain → Paris, France → **air** (different countries)
* Madrid, Spain → Damascus, Syria → **air** (different countries)
* Copenhagen, Denmark → Reykjavik, Iceland → **air** (no road route available)
## What stays as do\_not\_know in the database
The `transport_type` field on the section **always remains `do_not_know`** in the database, regardless of which mode was inferred. The inferred mode only determines:
1. Which geographic distance record is used for `kms`
2. Which emission factor concept is matched during calculation
This means the `kms_source` field will be `null` for `do_not_know` sections in the current API response, since the distance was stored under the inferred transport type's geographic distance record, not under `do_not_know`.
## Emission factors used
| Inferred scenario | Emission factor concept |
| -------------------------------- | -------------------------------- |
| Same-country sea crossing | aircraft, unspecified (air) |
| Same-country, road-reachable | road freight, lorry, unspecified |
| Cross-country / road unreachable | aircraft, unspecified (air) |
The "aircraft, unspecified" factor covers all haul distances without requiring a distance band, making it the natural choice when the exact route characteristics are unknown.
## Comparison with explicit transport types
| | `do_not_know` | Explicit type (e.g. `maritime`) |
| ---------------------- | ----------------------------------------- | ---------------------------------------------------- |
| Distance calculation | Automatic per inference rules | Based on declared type |
| `kms_manual` allowed | No | Yes |
| `transport_type` in DB | Always `do_not_know` | The declared type |
| `travel_method` | Set automatically (truck if road) or null | Provided by user |
| Emission factor | Inferred from geography | Matched from declared type + travel\_method + detail |
## Frequently asked questions
The inference algorithm prioritises the lowest-complexity mode that can be unambiguously determined from geography alone. For cross-country routes, air is the safest default because:
* Maritime requires knowing the actual port-to-port routing, which is not available for all country pairs.
* Rail is not universally available and depends on railway network connectivity.
* Air operates between any two points on the globe.
If you know the actual transport mode, you should always declare it explicitly rather than using `do_not_know`.
Dcycle uses a landmass detection algorithm based on coordinates. A sea crossing is detected when origin and destination coordinates fall on different landmasses, even if they share the same ISO country code. This correctly identifies:
* Mainland Spain → Canary Islands
* Mainland Italy → Sicily or Sardinia
* Mainland Portugal → Azores or Madeira
* Mainland Denmark → Bornholm island
It does **not** trigger for tunnels or bridges (e.g. Copenhagen ↔ Malmö via Øresund Bridge), because the road routing check passes first.
Yes — use an explicit `transport_type` instead of `do_not_know`. If you know the shipment was maritime, set `transport_type: "maritime"`. If it was road, set `transport_type: "road"` and optionally `travel_method: "truck"`. You can also set `kms_manual` to override the calculated distance for any explicit type.
The `kms` value for a `do_not_know` section is taken from a geographic distance record stored under the inferred type (road, air, or maritime). Because the join that populates `kms_source` matches on `transport_type`, it doesn't find the record stored under the inferred type. This is a known limitation — the distance was correctly calculated; only the source label is unavailable in the API response.
## Related
Full data model, transport types, and distance calculation methods
How to create a route with do\_not\_know sections
All valid transport type + method + detail combinations
Reading kms, kms\_source, and inferred transport fields from a route