Machine learning monitors intensive animal agriculture

A machine learning application that can map intensive animal agricultural facilities in the United States faster and more efficiently than manual surveys is reported in Nature Sustainability this week. The application, which identified an additional 589 of these facilities in North Carolina, may help with tracking environmental violations in the food industry.

Concentrated Animal Feeding Operations (CAFOs) produce an estimated 40% of US livestock and generate around 335 million tonnes of waste per year. An estimated 60% of CAFOs in the US are not registered and do not have appropriate permits to dispose of waste, which can have serious impacts on food safety and water and soil pollution. The current legal landscape in the US has made it difficult for government agencies to monitor such facilities. There is currently no accurate data about the number, size or location of CAFOs.

Daniel Ho and Cassandra Handan-Nader use machine-learning techniques and high-resolution images to identify CAFOs in North Carolina, USA, and compare their results to a manual inventory. This enabled the authors to identify 589 additional poultry CAFOs compared to previous manual surveys - an increase of 15%.

The authors suggest that this method can facilitate the monitoring of CAFO compliance with environmental law by identifying non-permitted sites or those that pose a particular environmental risk.

doi: 10.1038/s41893-019-0246-x

영어 리서치 하이라이트