SIParCS 2021 - Thomas Johnson III
Employing Machine Learning Models for CESM Timing Data
The Community Earth System Model (CESM) is a software package that is critical to research in climate science and ecological study. The CESM community has been growing over the years and contributing more to the software package and proposing new features to expand the current features available in CESM. One such proposal has been the implementation of machine learning techniques to expand the current versatility of CESM. In the end goal, these machine learning methods would be utilized to be able to determine complex matters such as estimations of model performance based on timing file data. To look into the capabilities of machine learning methods on timing file data, said machine learning methods are employed to determine the classes that exist within generated sample data. The idea being that if the timing data can be used to address much simpler tasks such as determining what system the timing data corresponds to, the inverse problems of higher complexity can be explored as well in later work. In exploring simpler cases and moving to more complex scenarios, the usage of machine learning methods can be evaluated and insights can be obtained from the timing data. Furthermore, a general workflow can be established for machine learning methods that can be employed for CESM timing file data for future use.
Mentors: Sheri Mickelson, Brian Dobbins, & John Dennis
Slides and poster