This paper presents a statistically more likely distribution (SMLD) approach for the evaluation of the presumed probability density function (PDF) in flamelet progress variable (FPV) models for non-premixed supersonic combustion. The numerical simulation of the NASA Langley Research Center supersonic H-2-Air combustion chamber is performed using two approaches: the first one is a standard FPV model, built presuming the functional shape of the PDFs of the mixture fraction, Z, and of the progress parameter, Lambda; the second approach employs the SMLD technique to presume the joint PDF of Z and Lambda. The standard and FPV-SMLD models have been developed using the low Mach number assumption. In both cases, the temperature is evaluated by solving the total-energy conservation equation, providing a more suitable approach for the simulation of supersonic combustion. By comparison with experimental data, the proposed SMLD model is shown to provide a clear improvement with respect to the standard FPV model, especially in the auto-ignition and stabilization regions of the flame. Copyright (C) 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.