Crop-livestock integration (CLI) systems combine the growing of crops in rotation or consortium, especially grain crops such as soybeans, corn and sorghum, and forage plants used to feed cattle and pigs, with the raising of livestock, typically beef cattle. The crops provide most of the cash income, while the livestock has food available during the dry season and facilitates seed management.
CLI improves soil fertility, raises yields and helps rehabilitate degraded areas while reducing the use of pesticides, mitigating the risk of erosion and the seasonality of production, and lowering operating costs. It makes farming more sustainable: crops benefit livestock and vice-versa; the environmental impact of agricultural activity falls; and the carbon footprint is reduced.
In a study reported in the journal Remote Sensing of Environment, researchers affiliated with the Brazilian Agricultural Research Corporation (EMBRAPA) and the State University of Campinas (UNICAMP) have developed a method based on artificial intelligence (AI) to identify CLI areas by analyzing satellite imagery. According to the authors of the article, this knowledge can benefit Brazilian agriculture in several ways.
“The main aim of the project, which was an international collaboration to address issues relating to sustainable agriculture, was to promote the integration of remote sensing data with satellite images using AI, precision agriculture and biogeochemical models to understand and create models of the dynamics of this type of system,” said Inácio Thomaz Bueno, first author of the article. A forest engineer, Bueno conducted postdoctoral research on the monitoring of CLI systems using remote sensing data and satellite imagery with high spatiotemporal resolution.
The team used deep learning techniques to process satellite imagery time series and extract patterns pointing to areas where CLI was being practiced. The study sites were areas in the states of São Paulo and Mato Grosso. Object-based image analysis was performed at intervals of 10 and 15 days in four steps: CLI data acquisition via Planetscope, a constellation of satellites that capture high-resolution images of Earth’s surface, showing changes in the areas over time; training of algorithms to recognize patterns associated with CLI; mapping of CLI areas; and assessment of the model’s accuracy by comparing automatic results with previous knowledge.