LangYa: Revolutionizing Cross-Spatiotemporal Ocean Forecasting

TL;DR

Compared to existing numerical and AI-based ocean forecasting systems, LangYa uses 27 years of global ocean data from the Global Ocean Reanalysis and Simulation version 12 for training and achieves more reliable deterministic forecasting results for OSVs.

Abstract

Ocean forecasting is crucial for both scientific research and societal benefits. Currently, the most accurate forecasting systems are global ocean forecasting systems (GOFSs), which represent the ocean state variables (OSVs) as discrete grids and solve partial differential equations (PDEs) governing the transitions of oceanic state variables using numerical methods. However, GOFSs processes are computationally expensive and prone to cumulative errors. Recently, large artificial intelligence (AI)-based models significantly boosted forecasting speed and accuracy. Unfortunately, building a large AI ocean forecasting system that can be considered cross-spatiotemporal and air-sea coupled forecasts remains a significant challenge. Here, we introduce LangYa, a cross-spatiotemporal and air-sea coupled ocean forecasting system. Results demonstrate that the time embedding module in LangYa enables a single model to make forecasts with lead times ranging from 1 to 7 days. The air-sea coupled module effectively simulates air-sea interactions. The ocean self-attention module improves network stability and accelerates convergence during training, and the adaptive thermocline loss function improves the accuracy of thermocline forecasting. Compared to existing numerical and AI-based ocean forecasting systems, LangYa uses 27 years of global ocean data from the Global Ocean Reanalysis and Simulation version 12 (GLORYS12) for training and achieves more reliable deterministic forecasting results for OSVs. LangYa forecasting system provides global ocean researchers with access to a powerful software tool for accurate ocean forecasting and opens a new paradigm for ocean science.

Figures (5)

• Figure 1: Overview of the LangYa Forecasting System Architecture and Its Key Module. (A) Overall system pipeline integrating Time Embedding and deep feature extraction, (B) LLM-based Time Embedding module, (C) Asynchronous Cross-Iterative Random Sampling Strategy, (D) Ocean self-attention module based on cosine attention, and (E) Adaptive loss function for thermocline forecasts. W-MSA means window-based multi-head self-attention module. SW-MSA means shifted window-based multi-head self-attention.
• Figure 2: Performance Comparison of LangYa, XiHe, XiHe-AR, and Numerical Models Across OSVs. A: RMSE of the u-component of ocean currents for 1-7 day forecasts. B: RMSE of the v-component of ocean currents for 1-7 day forecasts. C: RMSE of the ocean temperature for 1-7 day forecasts. D: RMSE of the salinity for 1-7 Day Forecasts.
• Figure 3: Global RMSE Distribution of LangYa Forecasts. Temperature and Salinity at 1, 4, and 7-day forecasts. A-C: RMSE distribution of temperature at 1, 4, and 7-day forecasts. D-F: RMSE distribution of salinity at 1, 4, and 7-day forecasts.
• Figure 4: RMSE of LangYa Forecasts for Temperature, Salinity, and Currents Across Eight Ocean Basins (1-7 Days). A-D: u-, v-component of ocean currents, temperature, and salinity forecast RMSE at different ocean basins.
• Figure 5: Depth-Dependent RMSE Distributions of LangYa's Temperature Forecasts at 1, 3, 5, and 7 Days. A: 1-day forecast result. B: 3-day forecast result. C: 5-day forecast result. D: 7-day forecast result.