Project Summary

The purpose of ENSO_forecast is to propose new methods based on machine learning predicting El Niño-Southern Oscillation (ENSO) with time and computation efficiency and Pearson correlation can achieve greater than 0.6. Traditional theory-based models are too computationally expensive for predicting ENSO. Our forecasting system not only can maintain skillful prediction with a Pearson correlation above 0.6 for long-range forecast but also can let clients run our forecasting system quickly and on a local computer when out in the field, and can allow for a quick visualization of the comparison with instrumental data and theory-based models.

For more detail of the project, please see and download the PDF version: Project Report

Planned the project

Gantt Chart

For more details see Download Gantt Chart of ENSO_forecast.

Milestone

• Electing a project manager
• Deliver the sildes for the 1st meeting
• Meet with our instructor (1st meeting)
• Create and deliver the sildes for the 2nd meeting
• Meet with our instructor (2nd meeting)
• Meet with our instructor (3rd meeting)
• Meet with our instructor (4th Meeting)
• Deliver the rehearsal video
• Final meeting
• Assemble and publish the code
• Report submission

Deliverables

• Arrange tasks for each person for the 1st meeting
• Create and beautify the sildes
• Resolve Datetime format error
• Test the code on different environment
• Put experimental results on figshare

Identified Wastes

• Skills: The user needs to learn how to run our codes in the terminal and understand Github’s tutorial.
• Waiting: The user needs to wait for a couple of minutes to get the result.
• Over-processing: The user needs to input parameters from the our README.md in the terminal each time when they use our system.
• Inventory: Multiple folders in the local, so the user might need to find the results in those folders.

Usage of ENSO_forecast

See our GitHub README.md

Used Packages

• Pytorch
• Scikit-learn
• Xarray

Used Dataset for ENSO_forecast

Download the data through the link mentioned below directly on a local computer that used for the ENSO_forecast project

• COBE-SST2
• Nino3.4 Index
• MPI
• CNRM-CM5
• Reforecast Data (GRIB format)
• Reforecast Data (NetCDF format)

Contact Infromation of Our Team

• Zerui Xie: zeruixie@usc.edu
• Jieqiong Pang: jieqiong@usc.edu
• Kuan-Hui Lin: kuanhuil@usc.edu
• Yunyi Liao: yunyilia@usc.edu
• Jinhong Lei: leijinho@usc.edu
• Feilong Wu: feilongw@usc.edu