Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Advancing Marine Heatwave Forecasts: An Integrated Deep Learning Approach

Ding Ning, Varvara Vetrova, Yun Sing Koh, Karin R. Bryan

TL;DR

The paper tackles global marine heatwave forecasting and aims to improve multi-month lead times using a unified deep-learning strategy. It introduces an integrated framework that combines graph-based spatial modeling, imbalanced regression losses to emphasize extremes, and temporal diffusion to enable long-range forecasts. It also provides a novel graph construction method to prevent isolated nodes and releases a public SSTA graph dataset, along with an empirical evaluation across hotspot regions. The results show improved forecast skill over traditional numerical models in several regions and demonstrate the potential of temporal diffusion to replace sliding-window approaches for forecasts up to six months ahead.

Abstract

Marine heatwaves (MHWs), an extreme climate phenomenon, pose significant challenges to marine ecosystems and industries, with their frequency and intensity increasing due to climate change. This study introduces an integrated deep learning approach to forecast short-to-long-term MHWs on a global scale. The approach combines graph representation for modeling spatial properties in climate data, imbalanced regression to handle skewed data distributions, and temporal diffusion to enhance forecast accuracy across various lead times. To the best of our knowledge, this is the first study that synthesizes three spatiotemporal anomaly methodologies to predict MHWs. Additionally, we introduce a method for constructing graphs that avoids isolated nodes and provide a new publicly available sea surface temperature anomaly graph dataset. We examine the trade-offs in the selection of loss functions and evaluation metrics for MHWs. We analyze spatial patterns in global MHW predictability by focusing on historical hotspots, and our approach demonstrates better performance compared to traditional numerical models in regions such as the middle south Pacific, equatorial Atlantic near Africa, south Atlantic, and high-latitude Indian Ocean. We highlight the potential of temporal diffusion to replace the conventional sliding window approach for long-term forecasts, achieving improved prediction up to six months in advance. These insights not only establish benchmarks for machine learning applications in MHW forecasting but also enhance understanding of general climate forecasting methodologies.

Advancing Marine Heatwave Forecasts: An Integrated Deep Learning Approach

TL;DR

The paper tackles global marine heatwave forecasting and aims to improve multi-month lead times using a unified deep-learning strategy. It introduces an integrated framework that combines graph-based spatial modeling, imbalanced regression losses to emphasize extremes, and temporal diffusion to enable long-range forecasts. It also provides a novel graph construction method to prevent isolated nodes and releases a public SSTA graph dataset, along with an empirical evaluation across hotspot regions. The results show improved forecast skill over traditional numerical models in several regions and demonstrate the potential of temporal diffusion to replace sliding-window approaches for forecasts up to six months ahead.

Abstract

Marine heatwaves (MHWs), an extreme climate phenomenon, pose significant challenges to marine ecosystems and industries, with their frequency and intensity increasing due to climate change. This study introduces an integrated deep learning approach to forecast short-to-long-term MHWs on a global scale. The approach combines graph representation for modeling spatial properties in climate data, imbalanced regression to handle skewed data distributions, and temporal diffusion to enhance forecast accuracy across various lead times. To the best of our knowledge, this is the first study that synthesizes three spatiotemporal anomaly methodologies to predict MHWs. Additionally, we introduce a method for constructing graphs that avoids isolated nodes and provide a new publicly available sea surface temperature anomaly graph dataset. We examine the trade-offs in the selection of loss functions and evaluation metrics for MHWs. We analyze spatial patterns in global MHW predictability by focusing on historical hotspots, and our approach demonstrates better performance compared to traditional numerical models in regions such as the middle south Pacific, equatorial Atlantic near Africa, south Atlantic, and high-latitude Indian Ocean. We highlight the potential of temporal diffusion to replace the conventional sliding window approach for long-term forecasts, achieving improved prediction up to six months in advance. These insights not only establish benchmarks for machine learning applications in MHW forecasting but also enhance understanding of general climate forecasting methodologies.

Paper Structure

This paper contains 11 sections, 2 equations, 3 figures, 17 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Marine Heatwave Forecasts
  4. Deep Learning Methods
  5. Data
  6. Methodology
  7. Experiments
  8. Graph Construction and Loss Functions
  9. Prediction for Hotspot Locations
  10. Prediction for Longer Terms
  11. Limitations and Future Work

Figures (3)

  • Figure 1: Overview of the diffusion-GraphSAGE with imbalanced losses for MHW forecasts.
  • Figure 2: The SEDI (left) and CSI (right) maps for MHW prediction: a one-month-ahead forecast with the BMSE (first line), a four-month-ahead forecast with the WMSE trained in temporal diffusion (second line), and a six-month-ahead forecast with the WMSE trained in temporal diffusion without sliding windows (third line). All used the $m=25$ graph construction method.
  • Figure 3: The two highest SEDIs with the corresponding graph construction methods and loss functions at the 12 selected locations. Abbreviations are explained in the Appendix text.