Ocean/Atmosphere Time Series Analysis: Theory and Practice – 2021

When:
April 26, 2021 @ 13:00 – May 7, 2021 @ 18:00
2021-04-26T13:00:00+02:00
2021-05-07T18:00:00+02:00
Where:
Virtual (via ZOOM)

Responsible: Joe LaCasce / UiO
International lecturer: Jonathan Lilly (Theiss Research)
When: April 26–May 7, 2021, 13:00–18:00 Oslo Time
Where: Online via Zoom
Max. no. of participants: 20
Credit points: 5 ECTS
Register with this form here. (Deadline: 9 April) (registration closed)


Description of activity:

This course will introduce students to classical as well as cutting-edge techniques for analyzing timeseries in oceanographic, atmospheric science, and climate applications. Following the course “Fundamentals of Ocean/Atmosphere Data Analysis” in 2019 and 2020 by the same instructor, this course will focus on more advanced methods.

Beginning with a solid understanding of the link between time-domain and frequency-domain analyses, we will proceed from simple smoothing, to Fourier spectral estimation, to time-frequency methods such as the continuous wavelet transform, to stochastic modeling using random processes.

The chosen techniques are those that experience has shown to be the most useful in dealing with timeseries from the ocean and the atmosphere. Emphasis will be given to hands-on, practical application of methods, as well as to understanding the theory behind various methods. Extensive course notes may be found at http://www.jmlilly.net/course.html, which will be updated prior to the course.

In particular, we will cover:

— multi-taper spectral analysis
— rotary spectral analysis
— spectral confidence intervals
— wavelet analysis
— instantaneous moment methods
— stochastic modeling

Time pending, we will cover one or both of the following topics
— wavelet ridge analysis
— objective mapping with local polynomial fitting

Students will bring with them a dataset of their choice that they would like to investigate in detail. The dataset may consist of model output or observations, but must be in the form of a time series; image data or spatially distributed point data would not be suitable. Datasets containing several or many timeseries are encouraged.

The final project will consist of applying the methods taught in the course to their dataset, and interpreting the results. The students will receive personalized feedback, tailored to their specific datasets, through one-on-one meeting sessions with the instructor.

The course will be taught using Matlab, primarily with the instructor’s freely available jLab toolbox, http://www.jmlilly.net/software.html. All students must have an up-to-date copy of Matlab, as well as jLab, installed locally on their computer at the start of the course.

An option exists for students to do some of their work in Python, using a partial translation of the course materials into that language. Students interested in this option must communicate this to the instructor (jonathanlilly@gmail.com) at the time they register.

The course will be limited to 20 students, with preference given to those who have previously taken the fundamentals course.

Structure:

The class will consist of 5 hours of coursework a day, with 2–2 1/2 hours of lectures and the remainder being laboratory and group work sessions.

Outcomes:

At the end of the course, students will be well-prepared to begin efficiently analyzing any timeseries they might encounter through a range of techniques, while avoiding common pitfalls. Students will gain practical experience through hands-on demonstrations and exercises, and are expected to make substantial progress in analyzing their chosen dataset.