A method to obtain quantitative information about the chemical composition spatial distribution on rough (Ra between half a micrometer and tens of micrometers) surfaces is developed by using a Raman microscopy technique. It is tested on a wrinkled polyesterurethane-acrylate powder coating surfaces with pronounced starpatterned (mountains and valleys) structures. Raman spectra at different features on the surface are analyzed quantitatively by using a least-squares ® tting method based on coating components spectra, taking into account chemical reactions which occurred during cure. This method is able to extract the same information as qualitative analysis could from component Raman peaks when large composition heterogeneity occurs. In addition, it provides more subtle information when chemical composition distribution is less marked. This feature is particularly useful for establishing the cause of sophisticated surface structures. The origin of the surface structures is attributed to spinodal decomposition frozen by chemical reactions during cure.