Reconstructing the historical expansion of industrial swine production from Landsat imagery

In the USA, historical data on the period over which industrial swine farms have operated are usually only available at the county scale and released every 5 years via the USDA Census of Agriculture, leaving the history of the swine industry and its potential legacy effects on the environment poorly understood. We developed a changepoint-based workflow that recreates the construction timelines of swine farms, specifically by identifying the construction years of swine manure lagoons from historical Landsat 5 imagery for the period of 1984 to 2012. The study focused on the Coastal Plain of North Carolina, a major pork-producing state in the USA. The algorithm successfully predicted the year of swine waste lagoon construction (+ /- 1 year) with an accuracy of approximately 94% when applied to the study area. By estimating the year of construction of 3405 swine waste lagoons in NC, we increased the resolution of available information on the expansion of swine production from the county scale to spatially-explicit locations. We further analyzed how the locations of swine waste lagoons changed in proximity to water resources over time, and found a significant increase in swine waste lagoon distances to the nearest water feature across the period of record.