Skip to main content

Postdoctoral Scholar: Machine learning of sub-grid microphysical and turbulent processes for next-gen climate modeling

We are no longer accepting applications for this recruitment. Browse open recruitments

Position description

Postdoctoral position available in Machine Learning of sub-grid microphysical and turbulent processes for next-gen climate modeling.

The Department of Earth System Science at the University of California, Irvine is seeking a motivated atmospheric scientist / machine-learning (ML) enthusiast to engage in collaborative work between our research group at the University of California, Irvine and an exciting group of DOE collaborators at the Pacific Northwest National Laboratory.

Position perks:

  • Annual salary: $55k-70k USD + quality University of California medical, dental, and retirement benefits.
  • Flexible start date, through Fall 2021.
  • Remote work optional; international possible with approval.
  • One year initially but renewable for a second, pending good progress & funding.
  • Supportive department: excellent administrative support, postdoctoral community, early-career Slack.

Your qualifications:

  • An expected or recent Ph.D. in Atmospheric science and familiarity with parameterization issues.
  • Technically proficient in python, managing large datasets, some machine learning (ML).
  • Comfortable working on remote clusters, shell scripting, general data wrangling, etc.
  • Independent, collegial and open to multi-institutional collaborations.

Perks of our group:

  • Computing: Millions of CPU-hours, tens of thousands of GPU-hours, as needed.
  • Collaboration: Internally with 3 PhD students and 4 senior staff plus external networking with friendly colleagues working on similar themes at Columbia, UW, DLR, and MIT.
  • Technical training opportunities:
  • (By Professor Michael Pritchard): High-resolution climate modeling, high-performance computing, cloud superparameterization, coupling neural networks to climate models, overview of ML for climate science.
  • (By the group): Basic ML learning workflows for training and tuning feed-forward neural networks and variational autoencoders, incorporating physical constraints.
  • (via PNNL collaborators): Next-gen aerosol and microphysics parameterization, large-eddy simulation (LES).
  • (via my UCI CS collaborator Professor Stephan Mandt): Co-mentoring on ML methods such as variational auto-encoding (VAE) for generative stochastic parameterization, data compression and associated latent space inquiry / anomaly detection.  

Project description:

We are seeking an ambitious early career scientist interested in pushing frontiers of data-driven parameterization. The idea is to inform strategies for representing sub-4-km physics within a next generation of global cloud resolving models, focusing especially on microphysics-turbulence interactions. Some examples of specific problems of interest include:

● How many statistical moments are enough for higher order closure (HOC) parameterization? This appears objectively testable via ML auto-encoding of Large Eddy Simulation (LES) data, given enough of it.

● A timely “interpretable AI” opportunity: Can clustering the latent space of variational autoencoders trained to compress such data provide insight to the conventional heuristic parameterization developer about which sub-regimes deserve formal separation?

● To what extent can ML-based closure schemes fit the most challenging pressure perturbation covariances that rely on especially ill-defined closure in HOC parameterization schemes?

The work is to be in close collaboration with a group of established experts in microphysics, higher-order closure parameterization, and large-eddy simulation, based at or affiliated with the US Department of Energy’s Pacific Northwest National Laboratory, Sandia National Laboratories, University of Wisconsin-Milwaukee, University of Washington, University of Arizona, and Texas A&M University. Funding is through the auspices of DOE’s EAGLES project (Enabling Aerosol-cloud interactions at Global convection-permitting scales). Novel ML training data exist through this effort, including large eddy simulation ensembles optionally coupled to spectral bin microphysics.

Please apply online at https://recruit.ap.uci.edu/apply/JPF06685 with a cover letter that also describes your immediate and long-term research goals, a curriculum vitae including publications list, and names for three letters of reference (please do not solicit letters).

UC Irvine is located about halfway between Los Angeles and San Diego in coastal Southern California. It’s a nice place to live, not far from nice beaches, and in a Mediterranean climate.

The Earth System Science Department at UC Irvine is a highly interdisciplinary environment comprising ~ 25 faculty with expertise across many components of the Earth System, including atmospheric and climate dynamics, land surface processes, terrestrial and marine biogeochemical cycles, ice sheets, and human systems.

Qualifications

Basic qualifications (required at time of application)

An expected or recent Ph.D. in Atmospheric science and familiarity with parameterization issues.

Application Requirements

Document requirements
Reference requirements
  • 3-5 required (contact information only)
Apply link: https://recruit.ap.uci.edu/JPF06685

About UC Irvine

The University of California, Irvine is an Equal Opportunity/Affirmative Action Employer advancing inclusive excellence. All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, age, protected veteran status, or other protected categories covered by the UC nondiscrimination policy.

As a University employee, you will be required to comply with all applicable University policies and/or collective bargaining agreements, as may be amended from time to time. Federal, state, or local government directives may impose additional requirements.

Job location

Irvine, CA