A New Method for Ozone Depletion Detection Over Antarctica by Deep Convolutional Neural Network

The amount of ozone in the isothermal layer was decreasing gradually with five percent every decade since the 1970s around the earth. With the extensive use of modern industry chlorofluorocarbon, the speed of ozone content decreasing is getting faster and faster, which leads to an ozone hole in some rarefied ozone areas like Antarctica. Some national space administrations have monitored this environmental change issue since the 1980s by launching several meteorological satellites. However, such ozone hole data from the satellite is usually manually recorded by scientists, which is quite inefficient. Aiming at this problem, this paper proposed a new method for identifying the ozone depletion over Antarctica with the deep convolutional neural network. We collect the meteorological satellite images of ozone monitoring and tag a label on each of them. Then, we construct a deep convolutional neural network (CNN) to train these images with the label. Finally, we could apply the well generalized neural network to detect the new meteorological satellite images of ozone monitoring over Antarctica, which outputs the detailed information of the ozone depletion area. Experiment results on the new ozone monitoring dataset indicate that our proposed method achieved a satisfying detection accuracy with 99.23%, which is the first attempt in ozone monitoring problems over the world.