Speaker: 孙杨, 清华大学

Inviter: 熊世峰

Title: An Introduction to Data Assimilation in Atmospheric and Oceanic Systems: Concepts, Algorithms, and Parameter Estimation

Language: Chinese

Time & Venue: 2026.04.16 10:00-11:00 思源楼 813

Abstract: This report provides an introduction to data assimilation (DA) in atmospheric and oceanic systems, focusing on fundamental concepts, algorithmic implementations, and parameter adjustment techniques. The basic principle of DA is first introduced as a statistical framework for optimally combining model forecasts with observations to estimate the system state. The Ensemble Kalman Filter (EnKF) is then presented as a practical and widely used sequential DA method. To address common challenges in high-dimensional nonlinear systems, two essential extensions are discussed: variance inflation, which counteracts ensemble underdispersion and filter divergence, and localization, which reduces spurious long-range correlations by applying a Schur product to the covariance matrix. Finally, the report demonstrates how the EnKF framework can be applied to parameter estimation in state-space models. We propose a hybrid data assimilation framework that combines EnKF with Bayesian optimization, enabling efficient sequential updates of both the state and parameters jointly.