Machine learning-optimized composting strategies can enhance nutrient recycling and transform food system waste into a net carbon sink

Abstract

Composting organic waste offers a circular solution to recycle nutrients and restore soil health, but nitrogen (N) and carbon (C) losses during the process undermine its agricultural and climate benefits. Here, using a machine learning approach, we analyse 848 global manure, food waste and sewage sludge composting experiments, identifying 19 key management parameters that drive ammonia (NH3), nitrous oxide (N2O), methane (CH4) and carbon dioxide (CO2) emissions. We estimate annual global losses during the composting process at 747 kt NH3-N, 81 kt N2O-N and 592 kt CH4-C, equivalent to ~61 MtCO2e yr−1. By optimizing organic waste processing strategies, such as adjusting aeration, additives and C/N ratios, the composting chain could transition from a net emitter (40.1 MtCO2e) to a C sink (−15.1 MtCO2e), while conserving nutrients for crops. Under an optimal reduction scenario, China, Brazil and the United States emerge as the top three C sinks for composting, collectively accounting for 65% of total emission reductions. Organic waste mismanagement can degrade soils and increase pollution. A machine learning approach using data from 848 global composting experiments reveals 19 key management parameters and strategies to transform composting from a net emitter to a carbon sink.