​Project Overview

• The project systematically evaluated the structure and characteristics of observation errors in sea surface temperature (SST) products under extreme tropical cyclone conditions. Adopting the theoretical framework of Janjić et al. (2018) and leveraging the high-resolution datasets from He (2024), we successfully decomposed total observation errors into specific sub-components of representation error without the prohibitive computational costs of direct numerical weather prediction (NWP) simulations.

 

Major Findings

• The project confirms that the dominant sources of representation error exhibit a distinct spatial asymmetry depending on the typhoon quadrants. Specifically, pre-processing or quality-control errors, alongside unresolved-scale errors, were significantly amplified by localized physical mechanisms such as storm-induced cold wakes, intense precipitation, and vigorous convective activities, particularly in the right-rear quadrant of the cyclone. These high-frequency dynamic processes introduce substantial non-linear noise, which conventional data assimilation systems fail to adequately capture.

​

Validation of the Research Hypothesis

• The empirical results validate our initial hypothesis: the traditional assumption of a static, isotropic observation error covariance is fundamentally inadequate for extreme meteorological events. In fields where dramatic physical gradients occur over short spans, a uniform error assignment inherently leads to miscalculated weights in data assimilation, degrading overall analysis accuracy.

​

Implications for Future Data Assimilation Systems

• Consequently, this project highlights the critical necessity of implementing location- and situation-dependent, adaptive R covariance matrices. By dynamically adjusting error thresholds based on real-world dynamic properties, future data assimilation systems can significantly mitigate representative discrepancies, thereby yielding more robust initial fields and enhancing short-term numerical forecast skills for severe weather phenomena.