As the global population increases by 2 billion, it is essential we adapt to eco-friendly practices for a greener tomorrow. When we talk about a greener tomorrow, we cannot ignore the current climate crisis. Organizations in AI, ML, and artificial neural network technologies have welcomed this challenge and aim to solve this issue. Many solutions such as using drone tech for planting seeds, or helping farmers tackle climate change, to name a few, are on the move to empower small and large-scale farmers globally
One such solution to the fields of agriculture and forestry is offered by Earth Observing System Data Analytics (EOSDS). The company collects satellite data about crops and forest cover on a specific piece of land.
Then, it converts the data into valuable insights. This can help farmers and the forest departments to plan their actions in advance. After the conversion, it trains the neural networks to categorize the data into as many classes as possible.
https://twitter.com/ESA_EOHelp/status/1531976405776576512?s=20&t=biAK0gR7nYOtKpQFHSfOEg
Satellite data to crop data
EOSDS uses the convolution neural network (CNN) model to classify crops into different categories. The model will then predict the possibility of expansion to the agriculture market. CNNs are hierarchical, highly adaptable, and make good predictions. It can also differentiate the object from the background noise. They also learn to segregate the true images from their deep fakes.
Studying crops in Ukraine, India, and Brazil
Since its founding in 2015, EOSDS has collected crop and forest data from satellites such as Sentinal-1, Sentinel-2, and others. The first project was to classify the wheat crops into categories like summer wheat, winter wheat, barley, and many others in Ukraine. Over a period of seven years, the company was successful in retrieving the ground truth data and now provides insights to farmers every 10 days.
Director of Account Management for the Agriculture and Forestry at EOSDS, Lina Yarysh, shared that they “successfully detected the harvesting status for 6,500 fields with a total area of 47,000 ha. Data scientists provided reports on the harvesting status every 10 days.” In India and Brazil, the crop under study was sugarcane. Users will see the results in a colored format on their screens.
“For the ground-truth data collection, specialists were making trips twice a year, during summer and winter, to map spring crops that grow during these seasons,” said Yarysh. EOSDS’ analysis of the land shows the state of harvest, whether the land has been fully harvested, partially harvested, or hasn’t been harvested at all.
The company studied sugarcane in Maharashtra, India, and Sao Paulo, Brazil. They used the CNN Long Short Term Memory model for this study. They received an accuracy of 94% with this model. To study the crop data in time and space, EOSDS used satellite data and NDVI, a time series technology used to detect the state of the harvest.
The state of deforestation
For the forestry industry, the EOSDS doesn’t analyze the forest using NDVI technology. The state of deforestation is displayed in form of a piece of landmass where the forest covers can be seen, similar to how one would see it on Google Maps. The Forest Monitoring tech is described as the “youngest child” of the company, as it was released in 2021. This tech also makes use of the CNNs and takes extreme care in separating the data from the clouds in the sky. For this, they studied the Tasmania island in Australia. The company’s deforestation models work on a monthly and quarterly basis.
YouTube: How EOSDA uses neural networks to transform #SatelliteData into actionable insights
Photo credit: The feature image is only symbolic and has been taken by DGLimages. All other images are owned by EOSDS and have been provided for press usage.
