Straightforward semi-phenomenological models have been developed for highly porous polyurethane foams to predict the macroscopic nonacoustic parameters involved in the classical Johnson-Champoux-Allard model (i.e., porosity, airflow resistivity...) from microstructure properties (i.e, strut length, strut thickness and reticulation rate). These microstructure properties are measured using sophisticated optical methods (i.e., optical microscope, SEM) and a large variability can be observed due to great complexity of the 3D microstructure; variability which also depends on the precision of the measurement device. This work investigates how the variability associated with the model inputs affects the model outputs (i.e., non-acoustic parameters, surface impedance and sound absorption coefficient). The sensitivity analysis is based on the Fourier Amplitude Sensitivity Test (FAST). It helps quantify the correlation between the input parameters and identify the parameters contributing the most to output variability, thus requiring precise measurement. This study illustrates the preponderant impact of the reticulation rate (i.e, open pore content) on acoustic performances and guides the user on the required optical measurement device.