Scientists can pair satellite imagery with machine learning to collect and analyze land cover information, thus enabling a better understanding of our environment.
While clouds often degrade satellite images, recent advances in machine-learning models allow the scientists’ algorithm to identify 19 types of crops with 88 per cent accuracy, an April release from Germany’s Helmholtz Centre for Environmental Research (UFZ) said.
“If we can determine the cultivated crop for each agricultural field, we can draw conclusions not only about nutrient requirements, but also about the nitrate load of surrounding waters,” Sebastian Preidl said in the release. He’s a scientist in the landscape ecology department at UFZ.
“We can only protect a region's biological diversity effectively if we have a clear picture of the spatial land cover distribution.”
Researchers can now fill in data gaps caused by clouds. The scientists use customized algorithms that do not need large-scale cloud-free images to do this work, the release said.
Scientists can use the data to assess species-specific vulnerability to extreme events and climatic stress.
The study is published in the April edition of the journal Remote Sensing of Environment.