Remote sensing for detecting freshly manured fields

The application of manure on fields during the rainy season can pollute the environment by leaching nitrates into nearby bodies of water. To reduce and prevent this problem, closed periods are declared to stop the use of manure as a fertilizer during these seasons. However, compliance is difficult to verify due to the impracticality of monitoring all the fields in a region or country through on-site or satellite surveys. To address this problem, a method to automatically monitor freshly manured fields using machine learning and satellite imagery is proposed. This paper evaluates the Sentinel-2 (S2) satellite bands as well as 51 of the most common vegetation indices (VI) found in the literature for precision agriculture through experiments based on several feature selection methods. To train the models, a new dataset of freshly manured plots verified by on-site investigations is generated and made publicly available for future research. The proposed method is able to fully detect all freshly manured plots in the test dataset obtaining an F1-Score close to 90% of the tested area.

PEDRAYES Oscar;  USAMENTIAGA Rubén;  TRICHAKIS Ioannis;  BOURAOUI Faycal; 

2023-02-03

ELSEVIER

JRC131039

1574-9541 (online),   

https://www.sciencedirect.com/science/article/pii/S1574954123000353,    https://publications.jrc.ec.europa.eu/repository/handle/JRC131039,   

10.1016/j.ecoinf.2023.102006 (online),