GeoAI for Climate Science

You can apply AI and ML techniques to just about any domain. My favourite domain is space: AI in space, AI for space, AI using data from space to help us understand climate change. So, do I get excited when I see GeoAI papers? Yep!

In Real-time GeoAI for High-resolution Mapping and Segmentation of Arctic Permafrost Features, Wenwen Li , Chia-Yu Hsu , @Sizhe Wang, Chandi Witharana , and Anna Liljedahl developed a deep learning model called SparseInst that identifies and delineates permafrost landforms like ice-wedge polygons nearly twice as fast as previous methods.

The Importance

Fast and accurate permafrost mapping will help scientists monitor the stability of Arctic infrastructure and habitats. The authors' new GeoAI method also has practical value for indigenous and government agencies stewarding the environment and industries like oil and gas that operate in permafrost regions. Real-time automated mapping improves situational awareness and enables more predictive risk management.

The Research

Automated permafrost mapping with GeoAI can rapidly process huge volumes of satellite imagery to track changes across the vast Arctic region. The new model achieves real-time performance using a lightweight neural network architecture optimized for speed. The authors show that their model, SparseInst, matches or exceeds the accuracy of other deep learning models like Mask R-CNN while running much faster.

The Impact

This research highlights the power of GeoAI to analyze changes over massive geographical regions at scale. Automating complex geospatial analytics with deep learning holds promise for many urgent sustainability and climate challenges. The authors' paper is a perfect example of the benefits applied geospatial and AI research can bring.

Read the Paper

As always, please don't stop with my review. Go read the paper! Thank you to the authors for their work.