A Coupled Model Data Assimilation Study with Particle Filter

Due to its advantage on nonlinearity of analysis process, particle filter (PF) has risen as a promising advanced data assimilation method in the atmospheric and oceanic studies. However, how to apply the PF idea into a coupled climate model to implement coupled data assimilation (CDA) is still undiscovered. This study uses a simple coupled model to explore the PF CDA algorithm. With the model that shares the fundamental nature of the climate system, focusing on the characteristics of climate signals on multiple timescale interactions with nonlinearity in long time evolution, a particle filter is designed with characteristic weighting strategy and observational window for CDA. Validation shows that due to addressing the climate system nature this PF CDA algorithm outperforms over traditional ensemble Kalman filter, and thus it also has a good result to be applied to an intermediate coupled ocean-atmosphere-land model to produce improved climate analysis and prediction initialization.