Ceres

Summary

Ceres is a country and commodity observatory tracking the full food value chain across roughly 195 economies. Four engines cover production and farms, markets and trade, consumption and nutrition, and policy and fiscal stance. Fifty-seven async collectors pull from FAOSTAT, USDA, World Bank, OWID, WFP, AMIS, NASA POWER, OECD, and IPC into a commodity aware warehouse with read only analytics endpoints.

What it is

Ceres is a continuously updated food and agricultural economics observatory built for ministries of agriculture, multilateral institutions, commodity desks, and policy researchers who need a single, reconciled view of global food systems. The platform organizes evidence around four analytical engines. The production engine tracks yields, harvested area, fertilizer and pesticide use, livestock counts, and arable land. The markets engine compiles commodity prices, bilateral trade flows, stocks to use ratios, and AMIS market outlooks. The consumption and nutrition engine covers diet composition, food balance sheets, undernourishment, stunting, wasting, and IPC and Cadre Harmonise acute food insecurity classifications. The policy and fiscal engine assembles producer support estimates, tariffs, subsidy programs, and public expenditure on agriculture.

Fifty seven asynchronous collectors run on staggered UTC schedules and write into a SQLite warehouse with a commodity aware schema that links series to HS codes, CPC codes, food groups, and per kilogram calorie and protein content. A FastAPI service exposes read only endpoints for individual series, country panels, cross country comparators, and derived statistics including AR(1) persistence, half lives, year over year growth, volatility, and shock detection. Every observation traces back to a named primary source, with collection logs preserved per run so analysts can audit provenance, freshness, and reconciliation against alternative providers.

Methodology

• Four engine decomposition covering production and farms, markets and trade, consumption and nutrition, and policy and fiscal stance
• Fifty seven asynchronous collectors on staggered UTC schedules with structured collection logs and retry logic
• Commodity aware warehouse linking series to HS codes, CPC codes, food groups, and per kilogram calorie and protein values
• Reconciliation across overlapping sources, for example FAOSTAT, USDA PSD, and World Bank, with provenance preserved per observation
• Derived analytics including AR(1) persistence, half life, year over year growth, volatility, and shock detection
• ISO3 and UN M49 normalization with reference tables for roughly 200 countries and territories
• Read only API surface with country panels, cross country comparators, and long format multi series extracts
• No mock data policy enforced by an architecture test that blocks synthetic values from entering the warehouse

Data sources

• FAOSTAT (production, fertilizer, fisheries, livestock, pesticides, food balance, land use, GIEWS)
• USDA PSD, NASS livestock, and Drought Monitor
• World Bank WDI, agriculture, climate, and trade indicators
• Our World in Data (crop yields, hunger, diet, land use, fourteen series)
• WFP VAM and HungerMap
• AMIS market monitor and outlook
• NASA POWER agroclimate
• OECD producer support estimates and IPC and Cadre Harmonise acute food insecurity

Deliverables when used in engagements

• Country and commodity dashboards spanning production, markets, consumption, and policy
• Read only data API for series, country panels, comparators, and long format extracts
• Derived indicators for persistence, volatility, shock detection, and compound growth
• Operator diagnostics with per collector run history and freshness checks
• Briefing templates for AMIS style market outlooks and food security situation reports