Direct radiative forcing is a major impact of atmospheric dust aerosols. Mineral dust is an important aerosol component that interacts with both incoming and outgoing radiation, modulating radiative fluxes on Earth and its atmosphere. Dust radiative impact directly depends on the particles physico-chemical characteristics (e.g. mineralogy, shape, size) which are the main source of uncertainties. Spectral signatures of dust particles in the visible/short-wave infrared (V/SWIR) and long-wave infrared (LWIR) are linked to mineral composition and variability. Obtaining the precise spectral signature and size distribution of dust particles can result in accurate derivation of refractive indices which are used as inputs to model radiative forcing. In this work, V/SWIR and LWIR reflectance spectra of heterogeneous dust samples from the United States, East Asia, and Middle East were analyzed for their mineral abundances. For this study we used global well-characterized soil samples with comparable mineral compositions to windblown dust. The soil samples cover a wide range of mineral compositions and represent both arid and semi-arid regions [J P Engelbrecht et al., 2016]. Our preliminary analysis used linear spectral mixing for both V/SWIR and LWIR reflectance spectra. This approach is the simplest method to determine mineral abundances from reflectance spectra. While this resulted in a very low RMSE for the fit between the sample and modeled spectra, modeled spectra did not match band centers and strengths for all features. We also converted reflectance spectra to continuum removed (CR) and mean optical path (MOPL) which have the potential to eliminate nonlinear effects (e.g. multiple scattering) in spectral mixing. These approaches, which modify the reflectance hull, significantly weakened or removed the calcite absorption features at 2375 nm. Because these samples are very fine grained (< 38 µm [J P Engelbrecht et al., 2016]) multiple scattering effects are expected to be important for both V/SWIR and LWIR spectral ranges and as our initial results show linear mixing is insufficient to produce reasonable mineral abundances. Our next efforts will include full radiative transfer models of the measured spectra which we will present at the meeting.