Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-scale decomposition of sea surface height snapshots using machine learning

Jingwen Lyu, Yue Wang, Christian Pedersen, Spencer Jones, Dhruv Balwada

TL;DR

These challenges of decomposing instantaneous SSH into BMs and UBMs can be addressed by using zero-phase component analysis (ZCA) whitening and data augmentation; making this a viable option for SSH decomposition across scales.

Abstract

Knowledge of ocean circulation is important for understanding and predicting weather and climate, and managing the blue economy. This circulation can be estimated through Sea Surface Height (SSH) observations, but requires decomposing the SSH into contributions from balanced and unbalanced motions (BMs and UBMs). This decomposition is particularly pertinent for the novel SWOT satellite, which measures SSH at an unprecedented spatial resolution. Specifically, the requirement, and the goal of this work, is to decompose instantaneous SSH into BMs and UBMs. While a few studies using deep learning (DL) approaches have shown promise in framing this decomposition as an image-to-image translation task, these models struggle to work well across a wide range of spatial scales and require extensive training data, which is scarce in this domain. These challenges are not unique to our task, and pervade many problems requiring multi-scale fidelity. We show that these challenges can be addressed by using zero-phase component analysis (ZCA) whitening and data augmentation; making this a viable option for SSH decomposition across scales.

Multi-scale decomposition of sea surface height snapshots using machine learning

TL;DR

These challenges of decomposing instantaneous SSH into BMs and UBMs can be addressed by using zero-phase component analysis (ZCA) whitening and data augmentation; making this a viable option for SSH decomposition across scales.

Abstract

Knowledge of ocean circulation is important for understanding and predicting weather and climate, and managing the blue economy. This circulation can be estimated through Sea Surface Height (SSH) observations, but requires decomposing the SSH into contributions from balanced and unbalanced motions (BMs and UBMs). This decomposition is particularly pertinent for the novel SWOT satellite, which measures SSH at an unprecedented spatial resolution. Specifically, the requirement, and the goal of this work, is to decompose instantaneous SSH into BMs and UBMs. While a few studies using deep learning (DL) approaches have shown promise in framing this decomposition as an image-to-image translation task, these models struggle to work well across a wide range of spatial scales and require extensive training data, which is scarce in this domain. These challenges are not unique to our task, and pervade many problems requiring multi-scale fidelity. We show that these challenges can be addressed by using zero-phase component analysis (ZCA) whitening and data augmentation; making this a viable option for SSH decomposition across scales.
Paper Structure (14 sections, 6 equations, 5 figures, 3 tables)

This paper contains 14 sections, 6 equations, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Results
  4. Discussion and Outlook
  5. Data and Code Availability
  6. Comprehensive Overview of Dataset
  7. Zero-phase Component Analysis (ZCA)
  8. UNet Architecture and Training Specifications
  9. Visualization of Prediction Outcomes
  10. Gaussian Filter (GF)
  11. Error Distribution
  12. Power Spectral Density (PSD) Analysis
  13. Gradient Loss
  14. Scale-wise $R^{2}$

Figures (5)

  • Figure 1: Top: Schematic representing model pipeline and architecture. Bottom: Visualization of UBM and BM for our best performing test sample using AugZCA-UNet. A1-A5: BM; B1-B5: UBM. Black squares are a visual aid.
  • Figure 2: Comparison of PSD and $R^2$ across various models. (a) Mean PSD over test samples for UBM (left) and BM (right). (b)$R^2$ for UBM (left) and BM (right).
  • Figure App.1: A snapshot of the whole region SSH, BM and UBM
  • Figure App.2: Effect of ZCA whitening on the original UBM snapshot (a) and its PSD (b)
  • Figure App.3: Visualization of two test samples based on correlation coefficients between AugZCA-UNet predicted and true BM gradients: (a) Sample with median correlation (0.936); (b) Sample with lowest correlation (0.232). A1-A5: BM; B1-B5: UBM; C1-C5: BM Gradient; D1-D5: UBM Gradient.