Recently a new x-ray absorption spectroscopy (XAS) called inverse partial fluorescence yield (IPFY) has been developed that is bulk sensitive and free of saturation effects, which also provides a direct measure of total absorption coefficient. However, IPFY was originally formulated for smooth bulk samples, but XAS is often performed on rough samples. To test the applicability of IPFY on rough surfaces, a model is presented and the calculations based on this model are compared to the experimental results measured on NdGaO3. It is shown that the correspondence between calculated and experimental intensities of IPFY is sufficient to corroborate this model a means of estimating the maximum allowable surface roughness size and the optimal detection geometry.