Accessing Convective Hazards Frequency Shift with Climate Change using Physics-Informed Machine Learning

TL;DR

This paper introduces a physics-informed machine learning approach to predict the frequency shift of convective hazards like heavy rainfall and hail due to climate change, enhancing forecast accuracy and understanding regional risks.

Contribution

It presents a novel neural network architecture that integrates physics-based features and climate data to improve extreme weather event prediction under climate change.

Findings

Outperforms baseline models in accuracy and reliability

Identifies regions with increased hazard frequency due to climate change

Provides insights into landscape impacts on hazard distribution

Abstract

In this paper we discuss and address the challenges of predicting extreme atmospheric events like intense rainfall, hail, and strong winds. These events can cause significant damage and have become more frequent due to climate change. Integrating climate projections with machine learning techniques helps improve forecasting accuracy and identify regions where these events become most frequent and dangerous. To achieve reliable and accurate prediction, we propose a robust neural network architecture that outperforms multiple baselines in accuracy and reliability. Our physics-informed algorithm heavily utilizes the whole range of problem-specific physics, including a specific set of features and climate projections data. The analysis also emphasizes the landscape impact on the frequency distribution of these events, providing valuable insights for effective adaptation strategies in response to climate change.