food-opt Documentation

A global food systems optimization model that explores trade-offs between environmental sustainability and nutritional outcomes using linear programming.

Dominant crop group and land-use intensity under increasing trade friction. The animation sweeps from nearly free trade (0.25× baseline transport costs) through the baseline (1×) and costly trade (4×) to near-autarky (100×). As trade becomes more expensive, production disperses from comparative-advantage regions toward local self-sufficiency, and total land use rises.

---

Global scope and spatial resolution

The model divides the world into sub-national optimization regions and connects them through hub-based trade networks. Within each region, high-resolution geophysical data — crop yield potentials, land cover, irrigation infrastructure, water availability — are aggregated from gridcell-level datasets to drive the optimization. More than 60 crops are represented, each with spatially explicit yield potentials derived from the GAEZ framework. See Data Sources for the full list of input datasets.

Optimization regions (here 250) created by clustering administrative units. Each region has its own land endowment, crop yields, water budget, and dietary requirements.

Example input data: wheat rainfed yield potential (tonnes/ha) from GAEZ v5. See Crop Production for yield maps of other crops, water resources, and multi-cropping.

Supply chain representation

The optimization covers the food supply chain from primary resources to human nutrition: land allocation and crop production, livestock systems with grazing and feed-based pathways, food processing with co-products, waste, and international trade, and finally nutritional requirements that ensure diets meet caloric and food-group constraints for every country’s population.

High-level topology of the model. Commodities flow from primary inputs (land, water, fertilizer) through crop and animal production, processing and trade, to final consumption — with emissions tracked at each stage. See Model Framework for the mathematical formulation.

Environmental impacts

The model tracks greenhouse gas emissions from multiple sources — including land-use change, rice cultivation, livestock, and fertilizer application — all spatially resolved and converted to CO₂-equivalents. The figure below shows one component: annualised emission factors from land-use change, derived from satellite-based biomass and soil carbon data. These and other emissions can be priced into the objective function or capped as constraints, allowing the optimizer to find production patterns that reduce environmental pressure. See Environmental Impacts for details.

Annualised land-use change emission factors used in the optimisation. Left: CO₂ released per hectare of cropland expansion. Right: CO₂ sequestered per hectare of existing cropland spared and allowed to regenerate.

Diet and health

Dietary constraints ensure that each country’s population meets nutritional requirements across food groups. The model integrates epidemiological data from the Global Burden of Disease study to quantify how dietary patterns affect disease burden, measured in years of life lost. This makes it possible to optimize jointly for environmental sustainability and public health. See Nutrition, Current Diets, and Health Impacts.

Baseline diet-attributable chronic disease burden (years of life lost per 100,000 population) by health cluster, derived from Global Burden of Disease data. Countries are grouped into epidemiological clusters that share similar disease profiles.