X-ray absorption spectroscopy is often used for assessing the chemical speciation of metals in complex environmental matrices. The approach consists of performing spectral decompositions that rely on the selection of a suite of reference spectra representing the various coordination environments of the metal in the sample. From an experimental perspective, the quality of the data collected (that is the level of noise compared to the number of counts from the selected element) plays an important role in assessing the confidence of the speciation results. We present a method for the selection of an appropriate set of reference spectra based on the quantitative assessment of their interdependency. This method is illustrated in the case of the selection of reference spectra for performing nickel speciation.