Satellite imagery dataset of manure application on pasture fields

Abstract

Applying manure to pasture fields is a very common method of fertilization. However, rainfall can cause the manure to leach into water bodies near the field, contaminating the water and damaging the environment and the animals living in it, ultimately affecting human life. This paper presents a dataset consisting of images of 30 plots after manure application, verified by on-site investigations. This involved visiting 38 different plots, of which 8 were discarded because they were not suitable, either because of their small size, the lack of a specific manure application date, or the images being too cloudy in that period. The imagery is collected through Google Earth Engine using the satellite Sentinel-2, which offers 13 hyperspectral bands in the range of ultraviolet and near-infrared wavelengths including the visible spectrum. From these 13 bands, the most common hyperspectral indices in the literature for precision agriculture are calculated and added into the images as channels. 51 hyperspectral indices are calculated, summing up to a total of 64 channels per image when adding the raw bands from Sentinel-2. No normalization has been performed on any of the channels. The data can be used for further research of automatic classification of manure application to control its use and prevent contamination.

Keywords

• Fertilizer
• Land
• Slurry
• Crops
• Semantic segmentation
• Classification