the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Dynamic Mode Decomposition of Extreme Events
Abstract. Most data-driven methods, among them Dynamic Mode Decomposition (DMD), focus on analysing and reconstructing the average behavior of a system. However, the primary interest often lies in the anomalous behaviour, known as extreme events. This is especially the case in climate research, where extreme events have significant economic and societal costs. Therefore, we extend a DMD method to account for extreme events by adding a penalisation term. This extension allows us to not only better reconstruct the extreme events, but also extract the spatio-temporal structures related to those extreme events. DMD was originally developed by Schmid and Sesterhenn (Schmid and Sesterhenn, 2008) to enable the fluid dynamics community to identify spatio-temporal coherent structures (called modes) from high-dimensional data. In its essence DMD uses most relevant modes to filter the noise and reconstruct the original signal. We ask "Is the noise really noise"? Or can we attribute some of these dynamic modes, that result from the DMD, to extreme events? We applied this new method to the climate system, well known for its high-dimensionality. We examined two heatwaves that occurred in Europe (HW 2003 and HW 2010). In both cases we were able to improve the accuracy of the reconstruction. This novel variation of the DMD, can also be applied to other dynamical systems across many disciplines, in which extreme events are of interest.
- Preprint
(9065 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 09 Oct 2025)
- RC1: 'Comment on egusphere-2025-3505', Anonymous Referee #1, 07 Sep 2025 reply
-
RC2: 'Comment on egusphere-2025-3505', Anonymous Referee #2, 11 Sep 2025
reply
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-3505/egusphere-2025-3505-RC2-supplement.pdf
-
RC3: 'Comment on egusphere-2025-3505', Anonymous Referee #3, 15 Sep 2025
reply
The study presents an extension of Dynamic Mode Decomposition (DMD) by incorporating a regularization term specifically designed for extreme events, thereby enhancing the reconstruction accuracy of extreme climate phenomena. The proposed method successfully identifies extreme event mode that facilitate targeted analysis of extreme events, offering both scientific and practical value. However, the following aspects require further refinement to strengthen the manuscript:
Detailed Recommendations
The Abstract could include quantitative comparisons (e.g., relative error reduction) of extreme event reconstruction performance between the proposed method and standard DMD.
The Conclusion section requires refinement to clearly delineate and distinguish the novel contributions from existing theoretical frameworks.We can delineate three novel contributions: a) Regularization design: integration of an extreme-event penalty term into the DMD objective function. b) Mode selection: Automated identification of extreme-relevant mode. c) Discovered climate science insight.
While the manuscript mentions sensitivity analysis regarding the selection of rank values, it does not present the corresponding results or their implications.
A comprehensive sensitivity analysis could be conducted to validate the rationality of extreme event selection criteria. Additionally, I’m curious about whether the selection ofextreme events outside the modal space would impact the results.
The physical interpretation of the identified modes needs deeper exploration, particularly through integration with atmospheric dynamics principles to substantiate the selection of extreme event modes.Is it overlooked in previous studies? If not, we should add some references.
(Optional) Improving code and data accessibility would enhance research reproducibility and academic exchange, aligning with current academic best practices.
Citation: https://doi.org/10.5194/egusphere-2025-3505-RC3
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
1,355 | 37 | 10 | 1,402 | 41 | 51 |
- HTML: 1,355
- PDF: 37
- XML: 10
- Total: 1,402
- BibTeX: 41
- EndNote: 51
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
See attached PDF.