Meteorologically-Informed Adaptive Conformal Prediction for Tropical Cyclone Intensity Forecasting

TL;DR

This paper introduces a physically-inspired adaptive conformal prediction framework that improves tropical cyclone intensity forecasts by dynamically adjusting uncertainty estimates based on cyclone process information, leading to more accurate and interpretable predictions.

Contribution

It presents a novel covariate adaptive conformal prediction method that incorporates cyclone process data, enhancing probabilistic forecasting of tropical cyclone intensities.

Findings

Outperforms state-of-the-art models in point prediction accuracy

Provides physically consistent and interpretable forecast intervals

Establishes a process-aware framework for extreme weather prediction

Abstract

Rapid intensification (RI) of tropical cyclones (TCs) poses a great challenge due to their highly nonlinear dynamics and inherent uncertainties. Conventional statistical dynamics and artificial intelligence prediction models typically rely on static parameterization schemes, which limits their ability to capture the non-stationary error structure in the intensity evolution. To address this issue, we propose a physically-inspired covariate adaptive conformal prediction framework that dynamically adjusts uncertainty quantification by incorporating process information such as intensity and evolutionary stage. Our approach not only surpasses state-of-the-art models in point prediction accuracy, but also delivers physically consistent and interpretable forecast intervals, establishing a more process-aware framework for probabilistic prediction of extreme weather events.