PACT: Peak-Aware Cross-Attention Graph Transformers for Efficient Storm-Surge Emulation

2026-05-09

AI Summary

This study addresses the computational expense of high-fidelity hydrodynamic models, which hinders their application in multi-scenario storm surge simulations. To overcome this limitation, the authors propose an efficient modeling approach based on graph structures and the Transformer architecture. Atmospheric forcing fields are represented as graphs, with spatial features encoded via GraphSAGE; information is aggregated through learnable site-specific queries and cross-attention mechanisms. A Transformer encoder–decoder framework enables multi-step temporal forecasting. The method further introduces a novel peak-aware learning strategy—comprising a lightweight auxiliary head, a tail-focused loss, and a horizon-slope regularization term—to significantly enhance extreme event modeling. Evaluated on tide gauge data along the U.S. Northeast Coast, the model outperforms strong baselines in RMSE, MAE, and peak fidelity, achieves a single-year prediction in just 3.5 seconds, and demonstrates robust transferability under CMIP6 scenarios.

Abstract

Accurate and efficient storm-surge emulation is essential for coastal hazard assessment, yet high-fidelity hydrodynamic models remain too expensive for large scenario ensembles and rapid evaluation under heterogeneous climate forcings. We present PACT, a peak-aware cross-attention graph transformer for efficient station-level storm-surge prediction from atmospheric forcing fields. PACT represents each forcing patch as a graph, encodes spatial structure with GraphSAGE, and uses a learned station query to aggregate node information through cross-attention rather than uniform pooling. A Transformer encoder models temporal dependence across the forcing history, and a horizon-query decoder generates lead-specific forecasts from a shared temporal memory. To better capture extreme events, we introduce a peak-aware learning strategy that couples a lightweight auxiliary peak-aware head with a tailored training objective, including a tail-focused loss on peak-dominated samples and a horizon-wise slope regularizer to encourage coherent multi-step evolution. Across multiple tide-gauge stations along the US Northeast coast, PACT outperforms a strong spatio-temporal graph neural network baseline in both RMSE and MAE. Diagnostics show improved peak fidelity and tail preservation for reanalysis and most CMIP6 datasets. PACT is also computationally efficient, requiring about 3.5~s to generate a full winter-season surge trajectory for one year after training. Under distribution shift across five CMIP6 forcings, PACT transfers well within the CMIP6 family but degrades markedly when transferring from reanalysis to climate-model forcings, highlighting a persistent reanalysis--GCM gap.

Problem

• storm-surge emulation
• efficient prediction
• extreme events
• climate forcings
• distribution shift

Innovation

• peak-aware learning
• cross-attention graph transformer
• storm-surge emulation
• extreme event modeling
• spatio-temporal graph neural networks