Food and Agriculture Organization

What You're Getting

FAO (FAOSTAT) is the primary global source for agricultural statistics — crop production by commodity and country, food supply/demand balances, trade in agricultural commodities, land use, fertilizer consumption, food prices, and food security indicators. Covers 245 countries/territories × 200+ commodity codes × 1961–present. Also includes AQUASTAT (water), FishStat (fisheries), and Forestry data in separate portals.

Ingestion Strategy

Use the FAOSTAT API for targeted queries or download bulk ZIP files for full dataset loads. The bulk ZIP approach is more reliable for initial loads — each domain (e.g. "Production", "Trade") is a separate ZIP containing a CSV of all data + metadata files.

import requests, zipfile, io, pandas as pd

# Bulk download of Production dataset
url = "https://bulks-faostat.fao.org/production/Production_Crops_Livestock_E_All_Data.zip"
r = requests.get(url, stream=True)
z = zipfile.ZipFile(io.BytesIO(r.content))
df = pd.read_csv(z.open("Production_Crops_Livestock_E_All_Data_NOFLAG.csv"), encoding="latin-1")

Rate Limits & API Details

• No authentication required for API or bulk downloads
• API endpoint: fenixservices.fao.org/faostat/api/v1/en/data/{domain_code}
• Bulk ZIP downloads: bulks-faostat.fao.org/production/ — updated after each annual release
• Domain codes: QCL (crops), TP (trade), FBSH (food balance), FBS (food balance new), ASTI (R&D)

Schema Stability

FAO commodity codes (FCL codes) are stable but FAO is migrating to CPC 2.1 classification — some series have methodology breaks around 2010. Item codes and element codes are documented in the metadata CSV inside each ZIP. Country codes follow M49 standard (numeric) — map to ISO 3166 in your geography dimension.

Data Quality Gotchas

• Flags column: FAO embeds data quality flags (A = official figure, E = estimated, I = imputed, M = missing) — always store and filter on this
• Two food balance series: FBS (2010+) and FBSH (pre-2010) use different methodology — joining them requires adjustment factors
• Encoding: bulk CSVs use Latin-1 encoding, not UTF-8 — specify encoding="latin-1" in your parser
• Crop year vs calendar year: production data uses crop years that don't align with calendar years for all countries

Tool Compatibility

• Python: faostat PyPI package; or direct bulk ZIP download + pandas
• R: FAOSTAT CRAN package
• dbt: works well after initial load; build a commodity dimension from the metadata CSVs
• Airbyte: no official connector; bulk ZIP + cloud storage is the standard pattern

Benchmarking Approach

Global wheat production should be ~780M tonnes for 2022. Total country rows in Production_Crops_Livestock: expect ~3M+ rows. After annual release (typically Q4), verify new year data is present for top-10 wheat producers. Track flag distribution — if estimated/imputed flags exceed 30% for a country, flag as low-quality coverage. Spot-check against USDA PSD for major crops.