IceGraph

IceGraph will tackle the substantial uncertainty in sea level rise projections due to polar ice mass loss by creating a new
emulator for ice sheet and glacier models. Using MeshGraphNets, an emerging type of graph neural network, the project
will develop emulators that can tackle the complexity and computational burden of existing ice flow models. The proposed
research builds upon recent graph neural network developments by specifically addressing the need for irregular and
adaptive mesh representations in ice sheet modeling. The primary goal is to develop a computationally efficient emulator
framework that enhances accuracy and efficiency of ice sheet and glacier models, thereby reducing uncertainty in sea
level rise projections. The resulting emulators will improve our understanding of current and future ice dynamics, aiding
targeted monitoring and modeling efforts (e.g. for coastal adaptation and mitigation strategies).