ENSO prediction using Deep Learning

[gmanuch]

Title
ENSO prediction using Deep Learning

Summary
We propose a project to develop a framework for ENSO prediction using a range of machine learning approaches and benchmark their skill against other methods. Identify observables that are most relevant for ENSO predictability at a particular timescale.

Personnel
Georgy Manucharyan and Scott Martin, University of Washington, School of Oceanography

Specific Tasks
To begin with, we will try using the large ensemble of climate model simulations https://www.cesm.ucar.edu/projects/community-projects/MMLEA/. Then we will proceed with transfer learning using reanalysis data, like ERA5. Then benchmark against other techniques. Finally, identify relevant predictors using relevance propagation techniques.

Reading

• ENSO as an Integrating Concept in Earth Science https://doi.org/10.1126/science.1132588
• Deep learning for multi-year ENSO forecasts https://www.nature.com/articles/s41586-019-1559-7

[anaray23]

Hello! This project seems like something I'd like to work on! I have some ML experience in the climate science field, mainly methods like logistic regression, trees, and ANNs but would love to get more into deep learning. I am also interested exploring explainable/interpretable AI.

[bartonnoah]

I'm also interested in working on this project! In particular, the idea of identifying relevance predictors seems helpful in terms of understanding the dynamics and "why/how" ENSO affects climate as opposed to solely the "what". I'd love to discuss this further after the Git tutorial.

[smartin98]

Glad there's interest in the project! Here is another interpretable ML resource that could be helpful: https://github.com/slundberg/shap . Looking forward to discussing this more after the git tutorial. @anaray23 let me know if you want me to try zooming you into the discussion (should be around 4pm PDT), no worries if not though!

[anaray23]

@smartin98 I might not be able to make it at 4pm PDT, might be 15-20 min late. If people are still there, I'd love to join through zoom. If not, no worries!

[anaray23]

@smartin98 do let me know if there's still a discussion happening, should be free to join now, thanks! If not, no worries

[smartin98]

@anaray23 oops, sorry I missed this! I don't think you missed anything particularly different to what we discussed on zoom earlier though

[anaray23]

No worries!

[CassiaCai]

I'm also interested in working on this project! I want to learn about ENSO prediction.

[LisaLi1224]

I'm interested in this project! I'm excited about using machine learning to learn about the effects of climate change on ocean

[syvvu]

This project stood out to me the most! Looking forward to apply deep learning to climate forecasts.