CTP: A hybrid CNN-Transformer-PINN model for ocean front forecasting

TL;DR

CTP addresses ocean-front forecasting by integrating CNN-based spatial encoding, Transformer-based temporal modeling, and PINN-based physics constraints. The approach yields state-of-the-art accuracy, F1, and temporal stability across single-step and multi-step horizons in SCS and KUR, outperforming LSTM, ConvLSTM, CLP, and AttentionConv. Ablation studies show that each component—CNN, Transformer, and PINN—contributes, with PINN providing the largest gains by enforcing Navier–Stokes–like dynamics within the loss. The framework promises practical utility for real-time marine forecasting and can be extended to 3D processing and multi-source satellite data.

Abstract

This paper proposes CTP, a novel deep learning framework that integrates convolutional neural network(CNN), Transformer architectures, and physics-informed neural network(PINN) for ocean front prediction. Ocean fronts, as dynamic interfaces between distinct water masses, play critical roles in marine biogeochemical and physical processes. Existing methods such as LSTM, ConvLSTM, and AttentionConv often struggle to maintain spatial continuity and physical consistency over multi-step forecasts. CTP addresses these challenges by combining localized spatial encoding, long-range temporal attention, and physical constraint enforcement. Experimental results across south China sea(SCS) and Kuroshio(KUR) regions from 1993 to 2020 demonstrate that CTP achieves state-of-the-art(SOTA) performance in both single-step and multi-step predictions, significantly outperforming baseline models in accuracy, $F_1$ score, and temporal stability.

Figures (9)

• Figure 1: Overall architecture of the proposed CTP framework, including CNN encoding, temporal attention and physical loss integration
• Figure 2: Performance of different number of CNN layers
• Figure 3: Performance of different number of Transformer layers
• Figure 4: Performance of different loss functions
• Figure 5: Results of ablation studies