To monitor the reduction of vast oceans over time, researchers are utilizing artificial intelligence to swiftly detect them in satellite images. This AI-driven process is completed in under 0.01 seconds, a stark comparison to the conventional method of manually tracing icebergs in images, which necessitates human intervention for each identification, making it 10,000 times slower.
Anne Braakmann-Folgmann, the lead researcher at the University of Leeds in the United Kingdom, emphasizes the significance of locating oceans and monitoring their dimensions to estimate the volume of glaciers they discharge into the sea.
A recent report from the British Antarctic Survey warns of accelerated melting of massive ice sheets in Antarctica over the next decade, inevitably leading to a global rise in sea levels. Notably, A68a, one of the largest known icebergs, melted in the South Atlantic Ocean after a five-century journey from the Antarctic Peninsula, where it calved in 2017. This colossal iceberg, measuring over 100 miles in length and 30 miles in width, released 1 trillion tons of freshwater into the sea, altering the local ecology significantly.
The research team utilized satellite imagery to track the movement and shrinkage of A68a. However, pinpointing the precise location of the iceberg is challenging due to its translucent nature. To streamline this process, scientists trained a neural network to differentiate between the Antarctic coastline and actual icebergs, significantly reducing the time and effort required for analysis.
By employing images from the Sentinel-1 satellite of the European Space Agency, equipped with radar technology capable of penetrating cloud cover and darkness, researchers trained the AI to identify massive icebergs accurately. The AI system exhibited a remarkable 99% accuracy rate in detecting satellite images of glaciers, except for instances where the icebergs exceeded the size range used for training.
Andrew Shepherd, a co-author of the study from Northumbria University, highlights the transformative potential of machine learning in enabling real-time monitoring of remote regions. The AI tool’s precision in mapping iceberg extents swiftly and accurately offers a promising avenue for monitoring large glacier areas effectively.
