Chapter: DMD
aaren opened this issue · comments
- General plan is good, just need to get on paper.
- Where possible make remarks about key physical features.
- Possibly worth a focus on the lower level structures.
- Mention EOF earlier on.
- emphasise advantage over EOF immediately
- 'approx state space trajectory' - how does EOF do this? (maximises variance)
- make sure explanation clear to an experimental reader
- Translational structures
- EOF can represent with pairs of modes with similar variance. These can show sinusoidal translation. Classic example is in MJO.
- Can we see translational structures in DMD modes? (KH billows)
- Can we order modes differently?
- length scale of features?
- variance at low levels?
- Do any of the modes show a low level signal?
- Do the larger scale (or any scale) modes have events propagating to the surface?
- Can we take first two DMD / EOF modes and recompose into a timeseries? Can we still see KH?
- Wheeler Hendon diagrams
- Can we do this for DMD?
- Should be trivial (Im vs. Re for mode) -> should give spiral in phase space
- Can we compare with EOF?
- Multiple runs
- Just look for consistency between runs
- Can we feed in e.g. momentum flux? Same result?
- Conclusions
- How do EOF / DMD describe known structures (KH billows)?
- What do we learn about low levels and other things? Length of head etc.