Automated model reveals the condition of agricultural crops
The current state of crops and trends of their development over time can be determined by a new automated model based on machine learning methods. SK-CROP-net - (Satellite based crop classification model) can identify the current use of agricultural plots on a wider regional scale.
"In our study, we implemented a domain generalization approach, which aims to train a model on multiple source domains, often with different characteristics, so that it can be applied to new target domains without the need for further adjusting or adaptation to a new domain. We mainly use 5-year series of Sentinel products and reference data, which makes it possible to explore more deeply the potential of Sentinel-2 data in interannual crop mapping," said Tomáš Rusňák of the Institute of Landscape Ecology SAS.
Since the model is sufficiently robust and tested on multi-year time series, it is possible to determine not only the current state of crops, but also the trends of their development over time. Along with other environmental influences, such as climate indicators or soil parameters, it is then possible to comprehensively evaluate the impact of climatic extremes, sowing practices or the supply of nutrients on the condition of agricultural crops.
"We carried out the research as part of the European Space Agency's ESA PECS (Plan for European Cooperating States) project call in cooperation with YMS a.s., which transferred this knowledge to a service for farmers in the form of the so-called space translator. Among other things, it enables the monitoring of the condition of the crops in the fields during their growing season," added Tomáš Rusňák.
Spracovala: Monika Tináková
Foto: Unsplash/No one cares