Photoelectron Energy Loss Spectroscopy (PEELS) is a highly valuable non destructive tool in applied surface science because it gives access to both chemical composition and electronic properties of surfaces, including the near-surface dielectric function. An algorithm is proposed for real materials to make full use of experimental X-ray photoelectron spectra (XPS). To illustrate the capabilities and limitations of this algorithm, the near-surface dielectric function s(hco) of a wide range of amorphous carbon (a-C) thin films is derived from energy losses measured in XPS, using a dielectric response theory which relates 6( roi) and the bulk plasmon (BP) loss distribution. Self-consistent separation of bulk vs surface plasmon excitations, deconvolution of multiple BP losses and evaluation of Bethe-Born sensitivity factors for bulk and surface loss distributions are crucial to obtain several material parameters: (1) energy loss function for BP excitation, (2) dielectric function of the near-surface material (3-5 nm depth sensitivity), (3) inelastic mean free path,),,lambda(p) (E0), for plasmon excitation, (4) surface excitation parameter, (5) effective number NEFF of valence electrons participating in the plasma oscillation. This photoelectron energy loss spectra analysis has been applied to a-C and a-C:H films grown by physical and chemical methods with a wide range of (sp3/sp2 + sp3) hybridization, optical gap and average plasmon energy values. Different methods are assessed to accurately remove the photoemission peak tail at low loss energy (0-10 eV) due to many body interactions during the photo-ionization process. The sigma + Tr plasmon excitation represents the main energy-loss channel in a-C; as the C atom density decreases,),,p (970 eV) increases from 1.22 nm to 1.6 nm, assuming a cutoff plasmon wavenumber given by a free electron model. The Tr-Tr* and o--o-* transitions observed in the retrieved dielectric function are discussed as a function of the average (sp3/sp2 + sp3) C hybridization and compared with literature results. (C) 2016 Elsevier B.V. All rights reserved.