Determining planetary surface composition via remote spectroscopy frequently requires the use of inverse modeling, as the surface presents a complex mixture of materials which cannot be directly identified from the spectra. Depending on the complexity of the radiative transfer model (RTM) used, however, the inverse problem can be nonlinear and very high-dimensional, and the computational cost of traditional optimization methods becomes prohibitive. We demonstrate the utility of a multi-step metaheuristic approach for the inversion of high-dimensional RTMs, using the example of Pluto.