Recently, I worked on a project to map global freight transport emissions. The transportation of freight by land, sea and air underpins the global trade in physical commodities. Greenhouse gas emissions from freight transportation are a significant component of global emissions, however the inclusion of freight transport in emissions accounts and environmental impact studies is often incomplete. Both data availability and the difficulty in allocating freight emissions to specific commodity trades contributes to this.
It is not possible to connect every origin-destination country pair directly using every mode, as some routes are infeasible. For example, a country may not have a seaport and therefore cannot trade directly by sea, or there may be no land border between two countries meaning direct road or rail connections are impossible. Furthermore, freight statistics contain the transport mode as it appears to the reporting country, i.e. the mode used to cross that country’s border. However, this record may represent only the first hop (for exports) or the final hop (for imports) of a multi-hop journey. For example, where a commodity is reported as arriving by sea from a partner country, it may have first travelled by road to an intermediate country with a sea port and then onward to the destination country by ship. To handle all these cases, multi-hop trade routes can be modelled between trading partners.
To perform this modelling of trade routes, an adjacency matrix is created that defines feasible direct connections between countries. A number of data sources are used to determine feasible direct trade, by mode, and the distance each route traverses. For sea freight, the WFP Geonode dataset of world sea ports is used to determine whether both origin and destination countries have sea ports and the sea distance between them given by the CERDI database. The CERDI database considers actual shipping routes between countries and shortcuts between landmasses, for example through the Panama Canal. For rail and road freight, countries must share a land border for overland routes to be feasible (with some exceptions, such as Denmark-Sweden, and Singapore-Malaysia). For rail freight adjacency, the additional condition is imposed that both origin and destination countries have internal rail networks. For air, road and rail freight the distance between the origin and destination capital cities is used (calculated using the Haversine distance between two points on a sphere). We assume a path exists between all countries with sea ports.
The adjacency matrix is available here: https://github.com/spottedquoll/cargo-journeys
Resources
WFP 2017, Global ports Geonode, World Food Programme, United Nations
Bertoli, S., Goujon, M., and Santoni, O. 2016, The CERDI-seadistance database
ITF, 2013, Key Transport Statistics, 2012 Data, International Transport Forum
World Bank, 2019, World Bank Open Data
