LASSE: Learning Active Sampling for Storm Tide Extremes in Non-Stationary Climate Regimes

TL;DR

This paper introduces LASSE, an active learning method that efficiently identifies extreme storm tide events in large, non-stationary climate datasets with minimal hydrodynamic simulations, improving risk assessment accuracy.

Contribution

LASSE presents a novel online learning framework that adaptively samples storm data, significantly reducing the number of simulations needed to predict destructive storm tides in changing climates.

Findings

Achieved 100% precision in detecting rare destructive storms with less than 20% of simulations.

Surrogate models effectively generalize to new climate scenarios.

The method is scalable and efficient for large storm catalogs.

Abstract

Identifying tropical cyclones that generate destructive storm tides for risk assessment, such as from large downscaled storm catalogs for climate studies, is often intractable because it entails many expensive Monte Carlo hydrodynamic simulations. Here, we show that surrogate models are promising from accuracy, recall, and precision perspectives, and they "generalize" to novel climate scenarios. We then present an informative online learning approach to rapidly search for extreme storm tide-producing cyclones using only a few hydrodynamic simulations. Starting from a minimal subset of TCs with detailed storm tide hydrodynamic simulations, a surrogate model selects informative data to retrain online and iteratively improves its predictions of damaging TCs. Results on an extensive catalog of downscaled TCs indicate 100% precision in retrieving rare destructive storms using less than 20% of the simulations as training. The informative sampling approach is efficient, scalable to large storm catalogs, and generalizable to climate scenarios.