Columbia University, Dept. of Applied Physics and Applied Mathematics
NASA Goddard Institute for Space Studies
Using Pixel-Level Satellite Data to Improve a Coarse-Resolution Global Climate Model
This talk will begin with an overview of recent development in the NASA GISS General Circulation Model (GISS-E3 GCM, currently about to be submitted to CMIP6). Global satellite products are important references for continual GCM development and evaluation, and I will discuss recent effort to use machine learning (ML) to optimize a multitude of GCM free parameters in a way that leads to agreement with numerous NASA satellite climatologies. Even though ML can constrain a far greater number of free parameter settings using more observational references than are typically used in non-ML efforts, GCM discrepancies at cloud and convection lifecycle scales (i.e., process-level scales) remain. Since satellite products often provide data at coarse temporal and/or spatial resolutions, thus making it difficult to conceptualize and evaluate process representations in a GCM, I will discuss the various approaches I am taking to average global satellite retrievals in new ways to derive composite high-resolution evolutions clouds and the nearby environment (with a focus on deep convection lifecycles in this talk). These depictions will not only inform future parameterization development, but they are also crucial for ensuring that recent improved mean-state representations are not the result of errors cancelling at the process level.
613 Seminar Series Coordinators
Please do not forward the meeting invitation to outside recipients as the presentation, discussions, and questions may contain proprietary or confidential information.
Meeting number (access code): 199 051 6388
Meeting password: PNijn3wR?27
Wednesday, August 19, 2020
1:15 pm | Eastern Daylight Time (New York, GMT-04:00) | 1 hr 30 mins
Join by phone
+1-929-251-9612 USA Toll 2
+1-415-527-5035 US Toll
Global call-in numbers
Join from a video system or application
Join using Microsoft Lync or Microsoft Skype for Business