Global Food Twin

The Global Food Twin is a model that estimates the most likely transportation routes for food flows between producing and consuming regions worldwide. By combining food balance sheets, production maps, trade data, transportation networks, and optimization algorithms, we've created the best guess of how food moves across the globe—from producing to consuming regions—accounting for the complex multi-modal transportation systems. Visit a proof of concept application visualizing this model and global food flows here.

A few facts about the Global Food Twin:

• Comprehensive coverage: 82 food groups across 3,787 subnational administrative regions globally
• Multi-modal transport: Maritime, land-based (rail, road, inland waterway), and combined transportation routes
• Detailed routing: Actual geospatial routes including specific ports, rail lines, and road networks

Acknowledgements

This project was led by Cameron Kruse at Earth Genome working with Zia Mehrabi, Ginni Braich, and Kushank Bajaj Better Planet Laboratory on the model; Development Seed on the front end application; the model built on prior work by Jasper Verschuur Assistant Professor at Delft University.

Methods

This repository contains the data produced by the model and used by the Global Food Twin Application. More details about how this data was developed including the code used can be found in this repository GitHub repository.

the Data

This repository is organized into inputs and outputs. Below is a diagram of the folder structure. the all_files.md document contains a full file structure.

1├── 📁 inputs/
2│   ├── 📁 downscaled_exports_imports/
3│   │   ├── 📁 demand_allocated/
4│   │   └── 📁 supply_allocated/
5│   │       └── 📁 foodgroup1/
6│   ├── 📁 infrastructure/
7│   ├── 📁 production_demand/
8│   │   ├── 📁 demand_rasters/
9│   │   └── 📁 production_rasters/
10│   └── 📁 transport_cost/
11│       ├── 📁 land_transport/
12│       └── 📁 maritime_transport/
13└── 📁 outputs/
14    ├── 📁 additional_files/
15    ├── 📁 land/
16    │   ├── 📁 land_domestic/
17    │   └── 📁 land_reexports/
18    ├── 📁 sea/
19    │   ├── 📁 sea_domestic/
20    │   └── 📁 sea_reexports/
21    └── 📁 within_country/
1├── 📁 inputs/
2│   ├── 📁 downscaled_exports_imports/
3│   │   ├── 📁 demand_allocated/
4│   │   └── 📁 supply_allocated/
5│   │       └── 📁 foodgroup1/
6│   ├── 📁 infrastructure/
7│   ├── 📁 production_demand/
8│   │   ├── 📁 demand_rasters/
9│   │   └── 📁 production_rasters/
10│   └── 📁 transport_cost/
11│       ├── 📁 land_transport/
12│       └── 📁 maritime_transport/
13└── 📁 outputs/
14    ├── 📁 additional_files/
15    ├── 📁 land/
16    │   ├── 📁 land_domestic/
17    │   └── 📁 land_reexports/
18    ├── 📁 sea/
19    │   ├── 📁 sea_domestic/
20    │   └── 📁 sea_reexports/
21    └── 📁 within_country/

Using this data

This data is licensed under the Creative Commons Attribution 4.0 International License.

If you use this data in accordance with the CC 4.0 license, please cite: Better Planet Laboratory and Earth Genome. (2025). Global Food Twin. Harvard Dataverse. https://doi.org/10.7910/DVN/BLVPFU

Reach out to Cameron Kruse with questions.