Eighty Percent of Foods with No Footprint Data

Environmental impact assessments of food systems rely on Life Cycle Assessment (LCA) databases that contain data for fewer than 20% of food products consumed globally. Existing LCA databases (ecoinvent, Agri-footprint, AGRIBALYSE) have strong coverage for European and North American commodity crops and livestock but are nearly empty for tropical staple crops, smallholder farming systems, wild-caught fisheries, and processed foods. Without accurate LCA data, claims about the environmental footprint of diets are unreliable, sustainable food labeling schemes lack scientific grounding, and food system policies designed to reduce environmental impact are based on extrapolations that may not hold across production systems.

Food systems account for approximately 26–34% of global greenhouse gas emissions, 70% of freshwater use, and 40% of land use. The EAT-Lancet Commission's dietary recommendations — now adopted by dozens of governments and institutions — rely on LCA data that systematically underrepresents the diets and production systems of the Global South, where most food system growth is occurring. A cassava farmer in Nigeria and a rice-shrimp polyculture in Vietnam have environmental footprints that existing databases cannot estimate to within a factor of 3. This data gap means that well-intentioned dietary shift policies may achieve far less environmental benefit than expected — or may even increase impacts if they redirect consumption toward foods with poorly characterized but high actual footprints.

The Global LCA Data Access Network (GLAD) was established to harmonize regional LCA databases, but harmonization cannot fill data gaps where no primary data exists. Meta-analyses (e.g., Poore & Nemecek, Science 2018) aggregate available LCA studies but inherit the geographic and methodological biases of the underlying literature — 85% of studies cover Europe, North America, and Oceania. Satellite-derived estimates of agricultural inputs (fertilizer, water) can partially fill gaps for crop production but cannot capture post-harvest processing, storage, transport, or waste — which often dominate the total footprint. Smallholder farming systems are particularly resistant to LCA because they are heterogeneous (mixed cropping, variable inputs, informal markets), making a single LCA "representative" value misleading.

Hybrid LCA approaches that combine sparse field data from smallholder systems with satellite-derived input estimates and statistical models could produce defensible first-order estimates for underrepresented food products. Standardized rapid-LCA protocols designed for low-data environments — where a 2-day farm-level survey plus remote sensing data yields a credible footprint estimate — would enable orders-of-magnitude faster data generation than traditional LCA studies (which take 6–18 months per product). Open-access databases designed for incremental, community-contributed data (modeled on GenBank for genomics) could distribute the data generation effort.

A student team could select a specific underrepresented food product (e.g., teff from Ethiopia, jackfruit from Southeast Asia, or tilapia from West African aquaculture) and construct a first-order LCA using publicly available data (FAO statistics, satellite-derived crop extent, literature values for input intensities), then systematically document which parameters dominate uncertainty and where primary data collection would have the greatest impact. Alternatively, teams could design a rapid-LCA survey instrument for smallholder systems and pilot it with available agricultural extension datasets. Relevant disciplines: environmental engineering, agricultural science, data science, sustainability science.