The Department of Atmospheric Sciences at the University of Washington is seeking a Postdoctoral Scholar for the development of a deep-learning earth-system model and its application to the study of sub-seasonal to seasonal predictability and low-frequency atmosphere-ocean variability.  Background in deep learning and in working with large geophysical datasets is essential. Specific familiarity with convolutional, recurrent, or graph neural networks, PyTorch, and ERA5 reanalysis or satellite data is also desirable. 

Project Summary

Deep-learning-based weather forecasting has advanced at a remarkable pace in the past few years, generating skillful forecasts on timescales of days to weeks that are nearly as skillful as physics-based models but considerably cheaper to run. This project will explore the benefits of deep-learning-based models for prediction on seasonal-to-decadal timescales, where coupling with the ocean and land surface is important. To do so, the postdoc will help incorporate land-system components into the DLESM-HPX (Deep Learning Earth System Model using the HEALPix Mesh) model. The realism of the resulting deep-learning earth-system model’s atmospheric responses to ocean and land surface anomalies will then be compared against high-resolution climate simulations, which have been shown to have larger and more realistic responses to surface anomalies than coarse-resolution climate models. In this way, this project will explore the benefits of deep learning to unlock the benefits of high-resolution modeling at a fraction of the computational cost.

Duties and Responsibilities

• Contribute to the development of the land components of a deep-learning earth-system model (DLESM) and couple it to the existing atmospheric and ocean components of the DLESM-HPX model
• Explore strategies for training the DLESM based on ERA5 reanalysis, satellite data, and high-resolution climate model output
• Perform idealized experiments in the DLESM to understand how it responds to different SST anomaly patterns, and compare to the response found in high- and low-resolution climate models

Appointment Details

The position is a twelve-month appointment at 100% FTE with the opportunity to extend to a second year subject to approval and availability of funding. A Postdoctoral Scholar is an academic appointment that requires evidence of a conferred Ph.D. by the appointment start date. The salary range is between $5,705 and $5,916.67 per month, commensurate with experience and qualifications, or as mandated by a U.S. Department of Labor prevailing wage determination. Other compensation associated with this position may include a moving allowance. The anticipated start date is April 1st, 2024.

University of Washington postdoctoral scholar appointments are for a temporary, defined period not to exceed five years/60 months, including any previous postdoctoral experience. Postdoctoral scholars are represented by UAW 4121 and are subject to the collective bargaining agreement, unless agreed exclusion criteria apply. For more information, please visit the University of Washington Labor Relations website.

Candidates must have an earned doctorate in atmospheric sciences or a closely related field (at the time of appointment), experience with machine learning, and strong programming skills.

Desired

It is preferred that candidates have: experience with one or more of the following: climate modeling; machine learning applied to atmospheric data; physical understanding of atmospheric flows and atmosphere-ocean interactions; analysis of satellite data, reanalysis, and/or climate model output; and have a strong verbal and written communication skills.

To apply, please submit the following application materials:

• A cover letter outlining current research accomplishments and a statement of how your experience meets the requirements for the postdoctoral position.
• Curriculum Vitae
• Names and contact information for two professional references.

Applications will be reviewed as they arrive.

Consideration of applications will begin immediately and continue until the positions are filled. For questions about the positions, please contact Professors David Battisti (battisti@uw.edu) and Dale Durran (drdee@uw.edu) by email.