Code Simulation Conclusion

"This study develops and validates a typhoon-centric adaptive OI system that improves SST data assimilation by applying quadrant-dependent error covariance in a storm-relative framework, resulting in more physically consistent and robust performance under severe weather conditions."

This study successfully developed and validated a Typhoon-Centric Adaptive Optimal Interpolation (Adaptive OI) Data Assimilation System. By reorienting fixed grids into a 'Heading-up relative coordinate system' aligned with the storm's motion vector, the framework dynamically scales the observation error covariance matrix (R) based on the physical asymmetry of the typhoon.

Validated against a 10-year dataset (2014–2023) and independent cases from 2024, the comprehensive conclusions are as follows:

Demonstrated Statistical Robustness and Scalability: The quadrant-specific uncertainty matrices established from historical data consistently improved the RMSE reduction rates across all sectors when applied to completely validation data (2024)
Fluid Weight Allocation: The framework successfully executed a dynamic balancing act. it applied a reduced Kalman gain in the highly unstable RR quadrant to reduce the influence of observation data, while leveraging the more reliable forward and left quadrants (FR, FL, RL) to assimilate high-quality satellite signals and facilitate overall domain convergence.

As a result, through this simulations, we confirmed that the proposed adaptive covariance matrix precisely captures the distinct SST error structures that markedly depart from normal years due to severe weather patterns like 'Cold Wake' generated by the Typhoon.

Under severe weather conditions, when the atmosphere–ocean system departs from its climatological reference state, conventional static observation error covariance matrix (Fixed R) tend to make the bias by indiscriminately incorporating observational noise. In contrast, the adaptive error covariance matrix (Adaptive R) reduces overfitting by incorporating typhoon-specific physical structures and non-climatological SST anomalies within a statistically consistent framework. This provides a robust and scalable approach for assimilating satellite remote sensing data in operational numerical weather prediction during high-impact events.