A Study on Monthly Marine Heatwave Forecasts in New Zealand: An Investigation of Imbalanced Regression Loss Functions with Neural Network Models

2025-02-19

AI Summary

This study addresses the regression imbalance problem in monthly-scale marine heatwave (MHW) forecasting for waters surrounding New Zealand, arising from the sparsity of extreme temperature anomalies. We propose a novel scaled-weighted mean squared error (SW-MSE) loss function and introduce balanced MSE—previously used in classification—for the first time in long-term MHW regression forecasting. Within a fully connected neural network framework, we systematically compare seven loss functions: MSE, MAE, Huber, weighted MSE, focal-R, balanced MSE, and SW-MSE. Results show that short-term (1-month) forecasts significantly outperform medium- and long-term (3-/6-month) forecasts; moreover, specialized loss functions improve F1-scores for both confirmed and candidate MHW events by over 40%. This work demonstrates that loss-function customization is a critical pathway to enhancing predictive capability for rare extreme events, offering a new methodological approach for modeling oceanic extremes in climate-sensitive regions.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Search and Optimization: Metareasoning and Metaheuristics
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Marine heatwaves (MHWs) are extreme ocean-temperature events with significant impacts on marine ecosystems and related industries. Accurate forecasts (one to six months ahead) of MHWs would aid in mitigating these impacts. However, forecasting MHWs presents a challenging imbalanced regression task due to the rarity of extreme temperature anomalies in comparison to more frequent moderate conditions. In this study, we examine monthly MHW forecasts for 12 locations around New Zealand. We use a fully-connected neural network and compare standard and specialized regression loss functions, including the mean squared error (MSE), the mean absolute error (MAE), the Huber, the weighted MSE, the focal-R, the balanced MSE, and a proposed scaling-weighted MSE. Results show that (i) short lead times (one month) are considerably more predictable than three- and six-month leads, (ii) models trained with the standard MSE or MAE losses excel at forecasting average conditions but struggle to capture extremes, and (iii) specialized loss functions such as the balanced MSE and our scaling-weighted MSE substantially improve forecasting of MHW and suspected MHW events. These findings underscore the importance of tailored loss functions for imbalanced regression, particularly in forecasting rare but impactful events such as MHWs.

Problem

• Forecasting marine heatwaves in New Zealand
• Addressing imbalanced regression in neural networks
• Improving prediction accuracy with specialized loss functions

Innovation

• Neural network models
• Specialized regression loss
• Monthly marine heatwave forecasting