An Experiment on Extended, Satellite-Based Traceability of Organic Crops in North-Western Italy

In this work, we investigate how time series of the Normal-ized Differential Vegetation Index (NDVI) can provide use-ful clues to enhance the traceability of organic food, and dis-cuss the possibility to use machine learning in this context. Crop rotation, non-chemical weed control operations such as 'green mulching', fertilization, water supply management, all reflect into variables that are observable from space and may help reassuring the consumer that the traceable food they are purchasing matches the declared standards of sustainability and organic compliance. In this study we address detection of green mulching and weeding, based on experiments on a set of rice fields in North-ern Italy. Our findings suggest that the cover crops associated with green mulching can be confirmed and weeding can be detected using data from the Sentinel-2 satellite constellation, whereas fertilization is far more difficult to detect correctly. The cost associated with procuring training data seems to dis-courage the use of machine learning at this stage