CDI-Type I: GPU-Accelerated Interactive Supercomputing for Climate Studies in the Northern Environment

点击翻译按钮获取中文摘要

This project exploits modern GPU-acceleration techniques to design a new visually interactive supercomputing technique for human-in-the-loop investigation of the postglacial evolution of the Earth?s surface since the Last Glacial Maximum via numerical modeling. The complex coupling between Earth and climate has imposed a large and growing impediment to scientific investigation. This complex system embraces many non-linear physical models. An exhaustive examination of all parameters and algorithms of numerical models is computationally prohibitive for such complex systems. Investigators aim to solve this challenge by designing a new interactive framework that exploits modern Graphics Processing Units (GPUs) to enable scientists to visually drive the simulation process at run time based on the feedback of high-resolution visualization of intermediate simulation results. A multi-disciplinary team will apply the GPU-accelerated capability to explore the coupled ice sheet-viscoelastic system that underlies Earth surface evolution over a large range of temporal and spatial scales.The new interactive computation framework allows the user to visualize and monitor an ongoing simulation in real-time, and also to change computing models, parameters, and algorithms at runtime without restarting the simulation from the beginning. It is a critical tool for scientists to better understand complex climate and Earth systems. The software will be made publicly available to the geosciences community. Furthermore, by integrating commonly used software tools such as MATLAB and COMSOL and by creating visually oriented computational processes, we will make high performance computer simulation available to an entirely new audience. We will outreach to K-12 and undergraduates via two NSF-funded projects, led by the PIs of this project, to promote computational thinking and inquiry-based learning.