降雨径流模型是海绵城市、韧性城市、智慧城市建设的重要工具，准确模拟降雨径流是城市现代化管理的关键支撑。以往的模型参数识别研究中，模型的参数率定和验证通常仅考虑不同下垫面的影响，较少系统探讨降雨条件对参数的动态影响。已有研究表明，模型参数可能表现出显著的降雨强度依赖性。该文以中国海绵城市示范城市典型排水分区为例，基于Infoworks ICM模型，通过灵敏度分析与参数率定验证，系统揭示小、中、大雨情景下产汇流参数的动态响应规律。研究发现，固定径流系数随降雨量增大呈阶跃式增长，而汇流参数和初期损失参数则随雨强增大呈系统性减小的特点。这种动态响应印证了强降雨条件下地表饱和产流效率提升的物理机制。通过构建降雨条件驱动的动态参数调整机制，模型在所有降雨情景下均表现出较好的模拟验证效果(NSE均大于0.8)，显著优于不考虑降雨量影响情况下模型的表现(NSE小于0.5)。研究证实降雨强度是驱动模型参数变化的关键因子，建立的动态参数适配框架有效解决了传统方法因忽略降雨条件影响导致的“参数失真”问题，为城市雨洪模型的工程化应用提供了可靠的技术路径。

Objective: Rainfall-runoff models are vital for the planning and management of sponge, resilient, and smart cities, enabling accurate hydrological simulations to support urban modernization. However, previous research on model parameter identification, calibration, and validation has predominantly focused on static factors such as land cover and soil permeability, neglecting the dynamic influence of rainfall intensity on parameter behaviors. Addressing this gap, this study systematically investigates parameter identification under light, moderate, and heavy rainfall scenarios using a typical urban drainage catchment within a Chinese Sponge City Pilot Project. Methods: The research employs the Infoworks ICM hydrodynamic model to explore dynamic relationships between rainfall conditions and parameters regulating runoff generation and concentration. Incorporating detailed terrain data (12.5 m resolution), land use classifications, and drainage network geometries (48 pipe segments), the model ensures high physical realism. The methodology integrates a structured methodology combining scenario-based simulations, sensitivity analysis, and multistage calibration to evaluate how parameters change with rainfall intensity. Rainfall scenarios are divided into three categories to simulate urban storm conditions, namely light (< 10 mm), moderate (10-25 mm), and heavy (>25 mm). The assessed parameters include the fixed runoff coefficient, initial loss parameters, and convergence parameters such as Manning's roughness coefficient. A sequential calibration framework integrates a revised Morris sensitivity analysis with gradient-based optimization, ensuring robust validation through independent datasets. Results: Results reveal significant rainfall-dependent parameter variations. The fixed runoff coefficient increases by 15%-25% with rising rainfall intensities, while convergence and initial loss parameters decrease by 30%-40% under heavy rainfall compared to light events. These trends align with physical mechanisms: high-intensity rainfall induces surface sealing (reducing infiltration via pore clogging) and accelerates flow velocities through steeper hydraulic gradients. Although physically consistent, such dynamics had not been systematically quantified in prior studies. Earlier research attributed parameter variations primarily to static catchment characteristics, overlooking the dynamic feedback between rainfall intensity and hydrological processes. For instance, the observed inverse relationship between rainfall intensity and initial loss parameters addresses a longstanding contradiction in urban hydrology, where conventional models artificially increased initial losses for heavy storms despite evidence of reduced infiltration. Similarly, the rainfall-dependent convergence parameter reduction challenges historical assumptions of static flow routing behaviors across varying storm magnitudes. These findings demonstrate rainfall intensity as a hidden driver of parameter dynamics, previously overlooked by calibration frameworks treating rainfall as a fixed boundary condition. Validation outcomes demonstrate substantial improvements in model performance with rainfall-adaptive parameterization, with dynamically calibrated models achieving Nash-Sutcliffe efficiency (NSE) values exceeding 0.80 across all scenarios. These models outperform static parameter models by 25%-35% in accuracy. Conclusions: The findings emphasize the crucial role of incorporating rainfall intensity as a dynamic variable in parameter calibration frameworks, particularly for storm events where conventional static models fall short. The study advances urban hydrology by quantifying rainfall-parameter relationships and introducing a scenario-specific calibration framework. By demonstrating how neglecting rainfall-intensity dependencies leads to systematic prediction biases, particularly during extreme events, the findings provide critical insights for optimizing sponge city infrastructure. Improved simulations of rainfall-runoff interactions under climate variability support the design of effective measures such as permeable pavements and retention basins.