Your current location: home page
>News>News

AI Large Model for Agri-Food Economic Policy Introduction

Date of issue:2026-06-08 Number of visits: information sources:全球食物经济与政策研究院

Background and Rationale

Agri-food economic policymaking is inherently complex. Decision-makers face fragmented information sources, incomplete and lagging data, and analytical tools that demand specialized expertise. The central challenge is how to transform dispersed, dynamic information into computable, analyzable intelligence that genuinely supports decisions. The AI Large Model for Agri-Food Economic Policy was created to bridge this gap by integrating fragmented information, simplifying analytical tools, and enabling a shift from experience-based judgment toward data-driven, scientific decision-making.

Positioning

The model is an intelligent decision-support tool serving multiple stakeholders, including government, researchers, enterprises, farmers, and the public. Built on a complete “Data, Knowledge, Decision, Service” chain, it integrates four components: the China Agricultural University Agri-Food System Model (CAU-AFS), large language models, multi-source agricultural economic and policy data, and an intelligent report-generation and visualization platform. Four features define it. Cloud-based computing lowers the barrier to using professional models. AI-powered analysis automatically generates structured reports. Multi-objective coverage spans food security, economic efficiency, nutrition and health, and resource and environment. Multi-scale analysis spans global, national, provincial, and county levels down to a grid scale.

Components

The AI Model for Agri-Food Policy mainly consists of two modules: a professional modeling module and an intelligent decision-support service module. The first module, the professional modeling module, deploys and upgrades the existing CAU-AFS Model on the cloud. It establishes a cloud-based simulation laboratory that is not constrained by time or location. Users can select research questions online, design different scenarios, run the model with one click, and view result charts and analytical reports. In this way, the professional model can significantly lower the technical barrier for users and improve the efficiency of policy analysis. The second module, the intelligent decision-support service module, is designed to meet practical application needs. It includes functions such as agricultural product price early warning, investment cost-benefit analysis, dietary nutrition analysis, and international hotspot tracking. This module can support rapid data queries and early warnings based on data, and it can also be integrated with the professional model to conduct trend analysis, policy simulation, and risk assessment.

Module One: Professionalized Modeling (CAU-AFS Model)

The CAU-AFS Model was developed by the AGFEP, China Agricultural University. The core models are three economic models: the China Agricultural Spatial Equilibrium Model, the Global Agricultural Partial Equilibrium Model, and the China Dynamic Computable General Equilibrium (CGE) Model. The model integrates agriculture, food, economy, nutrition, and environment dimensions in a systematic way. Under a multi-objective framework, it can be used to assess scenarios such as agricultural support policy reform, dietary transition, diversified food supply, and low-carbon transformation. The CAU-AFS Model has been applied in annual flagship reports such as the China and Global Food Policy Report from 2022 to 2025, as well as in a number of policy advisory studies.

China Agricultural Spatial Equilibrium Model (Grid level). This model is mainly used to describe the spatial distribution of agricultural production and the relationship between supply and demand across different regions. At present, the model covers 31 provinces, supply and demand balance sheets for 58 agricultural products, and grid-level production data at a resolution of 50 km by 50 km. This model can not only observe the overall national results, but also further analyze production changes across different provinces, counties, and grids. It can be used to simulate the impacts of climate change, technological progress and adjustments in planting structure on the spatial distribution of agricultural production in different regions. This is highly important for conducting regional economic policy evaluations of agricultural products.

  China Agricultural Spatial Equilibrium Model (National level). One key feature of this model is that it is built on the concepts of a broad food perspective and a broad resource perspective. It covers 64 types of agricultural products, processed products, and by-products, including various food resources from cultivated land, forests, grasslands and water. The model not only captures the linkages among different markets, but also establishes the relationships among different food resources. It traces the crop, feed, and livestock chain as well as the production, processing, and consumption chain. Based on partial-equilibrium theory with 2023 as the base year, this model can be used for simulating the impact of policy, technology, disasters, trade, and consumption-preference changes.

  Global Agricultural Partial Equilibrium Model. The greatest strength of the Global Agricultural Partial Equilibrium Model is its broad coverage of countries and products, as well as its inclusion of a bilateral trade module. With 2022 as the base year, it links 42 countries and regions across 35 major agricultural and processed products through supply-demand balance sheets, bilateral trade matrices, and tariff data. The model simulates how tariff adjustments, trade agreements, economic and trade frictions, and climate shocks reshape global production, consumption, prices, and trade patterns. A world-market-clearing mechanism, where total exports equal total imports, determines equilibrium world prices, while a cross-entropy bilateral trade module decomposes country-to-country trade flows. It therefore provides strong support for international market assessment and trade policy analysis.

  China Recursive Dynamic Computable General Equilibrium Model (CGE). The CGE model is designed to analyze the entire agricultural value chain, including the broader agri-food system, and to examine the interactions between agriculture and other sectors of the national economy. Built on a Social Accounting Matrix derived from the 2023 China input-output table, it covers 95 economic sectors, including 26 agricultural sectors, 15 food processing sectors, input sectors such as fertilizers, pesticides, and agricultural machinery, as well as catering services and related sectors, 6 production factors, and 4 economic agents. It places the agri-food system within the wider national economy to evaluate the direct and indirect effects of macro policies on GDP, sectoral output, consumption, investment, trade, employment, and household income.

Module Two: Intelligent Decision Services

  Price Early-Warning. Integrating domestic and international weekly and monthly prices and stock-to-use ratios, the system identifies periods of excessive price volatility and issues tiered Red, Orange, and Green warnings. The methodology builds on the IFPRI Food Security Portal’s excessive-price-volatility early-warning approach, adapted to our data structure, supporting precise price queries, historical-trend analysis, and anomaly flagging.

  Investment Analysis. Drawing on cost-benefit survey data from 2018 to 2023, it analyzes cost structures, net profits, and return on investment, enabling cross-variety profitability comparison and price-cost simulation.

  Nutrition Analysis. Using the China Food Composition Table and dietary-quality assessment methods, it estimates nutrient intake, evaluates dietary quality, and generates personalized improvement recommendations.

  International Hotspot Tracking. Connecting to public sources such as FAO, the World Bank, IFPRI, OECD, and USDA, it automatically produces Chinese-language weekly briefs and thematic reports, monitoring international perspectives, identifying external shocks, and drawing on global experience.

Core Capabilities

The model delivers four integrated capabilities.

  • Scenario simulation offers quantitative, multi-dimensional assessment of policies and external shocks with automated reporting.

  • Price early-warning provides real-time monitoring and risk alerts.

  • Trend analysis forecasts medium-to-long-term changes in supply-demand, trade, GDP, employment, income, and consumption.

  • Knowledge integration draws on policy documents, international reports, and cost-benefit data. Together these enable rapid information extraction, strong analysis, stakeholder coordination, and the broader transformation of agricultural governance.

Relevance to “AI and Trade”

This model is closely aligned with the theme of AI and trade. Through its Global Agricultural Partial Equilibrium Model, it offers rigorous, quantitative assessment of how tariffs, trade agreements, and trade frictions reshape global agri-food production, prices, and trade flows. By combining advanced AI with established economic modeling, it shows how AI can enhance market transparency, anticipate disruptions, and support evidence-based trade policy. In doing so, it contributes a Chinese perspective on agriculture to the multilateral trading system.

Outlook

Looking ahead, the model will connect with the Shennong Large Model to build an integrated production, market, and policy decision system. Plans include strengthening data support through Shennong’s professional knowledge base, constructing realistic multi-scenario simulations via agricultural intelligent agents, enhancing model linkage with cultivation, pest-recognition, and meteorological-warning capabilities for cross-disciplinary whole-chain analysis, and promoting inclusive applications that deliver intelligent decision services to governments, researchers, and producers alike.

Contact

Zhang Yumei, Professor, AGFEP, China Agricultural University

zhangyumei@cau.edu.cn

https://afep.cau.edu.cn/