Nonlinear 1D Time-Domain Electromagnetic Inversion using Adaptive Differential Evolution: Application to Volcanic-Hydrothermal System

<title>Abstract</title> <p> Time-Domain Electromagnetic (TDEM) subsurface resistivity inversion is a highly nonlinear, ill-posed optimization problem hindered by parameter coupling, layer equivalence, and local minima trapping. To address these challenges, this study presents an advanced stochastic inversion framework based on Improved Linear Population Size Reduction Success-History Based Adaptive Differential Evolution (iL-SHADE). The algorithm incorporates dynamic memory-driven parameter adaptation and linear population size reduction to balance global exploration and local exploitation, eliminating the need for manual tuning of mutation and crossover parameters. Evaluated over 30 independent stochastic runs across four distinct 5-layer synthetic earth models (Ascending, Bell, Bowl, and Descending), iL-SHADE achieved superior solution quality, reaching median misfits of 10 ^-8 to 10 ^-9 , outperforming standard DE and SHADE by 2 to 4 orders of magnitude. Non-parametric Wilcoxon rank-sum hypothesis testing confirmed that these performance enhancements are statistically significant ( <italic>p</italic> &lt; 0.01). Noise robustness assessments demonstrated consistent convergence to noise-floor-bounded minima under Gaussian noise levels up to 20%. Field data validation at Station UZ01 on Unzen Volcano, Japan (1.4 km grounded electric wire transmitter, 2.0 km offset) successfully resolved the 5-layer subsurface structure extending to a depth of 5.6 km. These results demonstrate that iL-SHADE provides a highly robust, parameter-adaptive optimization framework for reliable electromagnetic imaging in complex volcanic-hydrothermal systems. </p>