We present a new inverse modeling procedure to characterize the hydraulic properties of partially saturated soils from soil moisture measurements during a natural transient flow experiment. The inversion of the governing one-dimensional Richards equation is carried out using the Global Multilevel Coordinate Search optimization algorithm in sequential combination with the local Nelder-Mead Simplex algorithm (GMCS-NMS). We introduce this optimization method in the area of unsaturated zone hydrology since it is adapted for solving accurately and efficiently complex nonlinear problems. Several numerical experiments have been conducted to evaluate the proposed inversion method using synthetic error-free and error-contaminated data for different textured soils. Inversion of the simulated error-free data and examination of the related response surfaces demonstrated the uniqueness of the inverse solution and the suitability of the GMCS-NMS strategy when identifying four key parameters of the hydraulic functions described by the Mualem-van Genuchten model. Inversion of the error-contaminated data proved further the good stability of the inverse solution that is consistent with the needs required by real experiments.