What do you think of when the Sahara is mentioned to you? You probably imagine an arid desert and sand ad infinitum. Well, you don’t. The millions of trees in the Sahara change its landscape over thousands of kilometers. These are not forests, but solitary trees.
An area of 1.3 million km2 was surveyed in northwest Africa. It includes Algeria, Mauritania, Senegal and Mali. The scholars counted in total about 1.8 billion trees.The author of the study is Martin Brandt of the University of Copenhagen. He mentions that “most are in the Sahel, there are hundreds of millions in the Sahara itself.” “It is about one tree per hectare on average in the hyper-arid Sahara. That’s more than one might have thought,” he noted.
Teaching computers
The area investigated represents only 20% of the Sahara and Sahel. “The total tree count must be much higher.” The work was done by accessing high-resolution satellite imagery reserved for military or industrial uses.
To find the trees, they used a type of artificial intelligence known as deep learning. A computer was taught to identify trees. They counted canopies with an area greater than three square meters. The work showed that, on average, the canopies were about 12 square meters.
The millions of trees in the Sahara do not compensate for the huge problem of deforestation. “Trees in arid areas have always been there. Knowing their number and location is important. But it is not equivalent to growing new trees,” they say. “They are crucial to livelihoods. They fertilize the soil, leading to higher yields, and provide shade and shelter for humans and animals. They generate income and are crucial for nutrition,” he listed.
Counting the trees
Experts believe that their tracking system could serve as a basis for finding trees in other ecosystems. But the conditions are not yet in place to be able to count all the trees on the planet. “We need to create more robust models that can be used on a variety of different satellite systems. They need to have different spatial resolutions,” Brandt said.
“Also, if you apply it to forests, it’s often difficult to identify individual trees. If it’s difficult for our eye, it will be difficult for the model.”
