A Parameter Genetic Algorithm for Evolutionary Calibration in Coupled Dynamics Models of Geological Hazard Chain

Objectives: Complex terrain and geological conditions in the western mountainous areas of China have resulted in frequent hazard chains. The simulation assessments of the landslide-barrier lake-outburst flood hazard chains are an effective tool for integrated hazard management, and reasonable model parameters are the primary prerequisite to ensure the reliability of simulation assessment results. However, the coupling relationships among the multi-stage mechanism models of hazard chain are complex. The traditional parameter calibration methods for single-hazard models seldom consider the chain relationships among different hazards, and the multi-parameter calibration calculations are complicated and the error accumulation problem is prominent during the cooperative interaction operation of the hazard chain model groups. Methods: In response to the demand for tightly coupled modeling and global calibration of nonlinear cascade relationships, we proposes a dynamic data-driven multi-parameter genetic calibration method for landslide-weir- flood hazard chain coupled dynamical model. Firstly, a landslide hazard chain coupled dynamical model coupled by granular flow model, drift flux model, and sediment scour model is established. Secondly, a multi-parameter genetic evolutionary under dynamic observation data constraint is studied The calibration method is used to obtain the parameter set of the optimal population by iterative genetic evolution operations such as selection-crossover-variation. Results: The proposed method was validated using the 2018 Baige landslide hazard chain experiment, and the simulation results of the optimal parameter set were found to be in good agreement with the measured data after 400 iterations. The maximum fitness of the population was 0.932, with a deviation of less than ±5%, demonstrating the feasibility and accuracy of this parameter calibration method. The advantages of this method over traditional calibration methods in terms of effectiveness, stability, time consumption, and efficiency were discussed from multiple perspectives. Conclusions: This method effectively utilizes temporal observation data to dynamically calibrate model parameters and accurately describes the spatiotemporal variation process of the hazard chain, providing support for the comprehensive management of geological hazard chains. Further research will focus on knowledge-guided reliable simulation and evaluation methods for hazard chains, enabling reliable analysis of hazard risks under incomplete data conditions.