Coupled Multiparameter Estimation Based on Instantaneous Parameter Sensitivity: Simulations with an Intermediate Coupled Model

Coupled parameter estimation (CPE) optimizes model parameters by assimilating observational constraints to reduce systematic biases in coupled models, thereby enhancing simulation accuracy and predictive skill. However, its effectiveness in correcting multiparameter biases is constrained by pronounced differences in parameter sensitivities and mismatches between sensitivity analysis and data assimilation timescales. To overcome these limitations, this study proposes an instantaneous parameter sensitivity (IPS)-based strategy in which the IPS quantifies state variable transient responses to parameter perturbations and enables hierarchical coupled multiparameter estimation (CMPE) framework construction for coupled systems. By addressing the sensitivity-related challenges inherent in coupled models, this approach provides an effective pathway for implementing CPE, particularly in systems characterized by processes operating across multiple timescales. The IPS-based framework further advances CMPE methodologies. The key findings reveal the following: (a) Simultaneous perturbation of all the parameters produces the strongest model response, whereas perturbation of a subset of parameters with dominant IPS can achieve a comparable response. (b) A strong positive correlation exists between the IPS of state variables with respect to parameters and optimization performance; consequently, the simultaneous estimation of parameters with dominant IPS achieves optimal parameter estimation. (c) Parameters obtained using the optimal IPS-based optimization scheme sustain effective predictive skill over longer lead times, thereby improving climate analysis and prediction initialization.