A major new multimodel ensemble (MME) study assesses how global agriculture could look by 2050 under a coordinated “food systems transformation” package. The research builds on AgMIP-style modeling protocols designed to increase comparability across economic models while exploring uncertainty through scenario design and harmonized inputs.
The ensemble spans ten global economic models—AIM, CAPRI, ENVISAGE, FARM, GCAM, GLOBIOM, IMAGE, IMPACT, MAGNET, and MAgPIE—each representing food, environmental, and socioeconomic processes with different levels of detail. Although all are global, their internal structures vary in how they translate changes in diets, productivity, and losses into production, trade, prices, land use, and emissions.
Scenarios were jointly authored by MME coordinators and model teams, then iteratively tested through six submission rounds from late 2023 to mid-2025. A BAU (“current trends”) case serves as the counterfactual, while the centerpiece EL2 scenario mirrors key components of the 2025 EAT–Lancet report: movement toward a healthy reference diet, faster agricultural productivity growth, and reduced food loss and waste (FLW).
To keep the models aligned, teams standardized key drivers using SSP2 v.3 projections for population and GDP, shared productivity assumptions, and applied climate-related impact shocks consistent with RCP 7.0. These shocks include crop impacts (soybean, maize, wheat, rice), livestock impacts, and climate-driven labor productivity declines, then mapped onto each model’s native crop and livestock commodity sets.
Diet changes are implemented as consumer preference shifts toward the EAT–Lancet reference diet by 2050, with calories held via constant caloric coefficients. For fruits and vegetables, the target is treated as a floor, while animal-source foods operate more like a ceiling—populations can converge down toward the goal if they exceed it, but remain guided by regional trends if below.
For productivity, baseline yield growth rates are taken from long-running IMPACT updates informed by expert consultation and FAOSTAT-based trend analysis. EL2’s productivity improvement scales BAU growth by differences in per-capita GDP between SSP1 and SSP2, yielding roughly a 10–15 percentage-point yield growth increase over 2020–2050.
Model outputs are converted to relative changes against 2020 to enable cross-model comparison despite different native reporting formats. A reporting template captures 13 macro-regions and 13 sector types across crops and livestock, focusing on producer prices, production, land, employment, GHG emissions, water withdrawals, and nitrogen fertilizer use, acknowledging that not every model reports every variable.
To ground the simulation in history, the study uses FAO data (1961–2020) for production, area, yields, and value of production, then applies scenario-relative changes to FAO mean values for 2019–2021. Land-use context comes from HYDE v.3.3, and producer-value estimates are reconstructed from model producer prices multiplied by model production changes.
Subject of Research: Food systems transformation and global agricultural reshaping using a multimodel ensemble (MME) approach
Article Title: Food systems transformation would reshape global agriculture
Article References: Gibson, M., Sundiang, M., Mason-D’Croz, D. et al. Food systems transformation would reshape global agriculture. Nature (2026). https://doi.org/10.1038/s41586-026-10775-2
Keywords: multimodel ensemble; food loss and waste; EAT–Lancet diet; SSP2; RCP 7.0; agricultural productivity; land use; greenhouse gas emissions
Tags: agricultural productivity growthcross-model comparability in agricultural researchfood security and climate resiliencefood system scenario modelingfuture of global agricultureglobal food systems transformationimpact of climate change on cropsintegrated economic and environmental modelingland use and emissions projectionsmultimodel ensemble analysispolicy implications for food systemssustainable diets and food waste reduction
