Invited Talk
Long-term Marine Ecosystem Forecast based on Large Model
Jin QI, Zhejiang University
qijinjesse@zju.edu.cn

Biography

Dr. Jin Qi is a researcher at Zhejiang University's Department of Earth Sciences, specializing in artificial intelligence applications in oceanography. His work focuses on integrating machine learning techniques with geospatial analysis to address complex environmental challenges. Dr. Qi has contributed to the development of spatiotemporal neural network models for predicting key oceanic variables such as chlorophyll-a concentration, dissolved oxygen, and sea surface temperature. He also explores how physics-informed deep learning can improve the reconstruction of subsurface ocean profiles using sparse satellite and Argo observations. His research aims to enhance the accuracy and interpretability of marine environmental assessments and forecasts, particularly in dynamically active coastal and marginal seas, providing refined tools for climate-related ocean analysis.

Elements such as temperature, salinity, wave height and chlorophyll concentration in the surface layer of the ocean profoundly affect the health of marine ecosystems by regulating processes such as ocean biochemical cycling and air-sea energy exchange. Traditional methods are limited by the difficulty of solving the set of equations for the evolution of nonlinear ocean processes at the global scale, and are often restricted to the task of forecasting a single variable or a localized region. With the explosive growth of ocean observation data, this research proposes a multivariate synergistic forecasting large model based on the global ocean, which achieves efficient modeling of spatio-temporal heterogeneity and nonlinear coupling characteristics of global marine ecological elements by fusing multisource ocean data and embedding ocean spatio-temporal constraints for optimization, and greatly enhances the forecasting efficiency and accuracy, as well as the stability in the long-time-series forecasting task. By forecasting multiple elements including chlorophyll a concentration, this research provides important data support for marine conservation decision-making, as well as a new means for monitoring and forecasting the health of marine ecosystems, which further enriches the reliable technological chain from ocean observation to marine conservation.