This paper is concerned with the propagation characteristics of electromagnetic (EM) waves propagating along inhomogeneous random rough surfaces (RRSs) with power spectral density functions of Uniform, Gaussian and Exponential types of spectra. According to the convolution method, RRS generation is performed by the convolution of Gaussian pulses with a weighting function derived from the Fourier transform of an appropriate amplitude spectral density function. Since the weighting functions for the above mentioned three types of spectra are derived analytically, we can generate inhomogeneous RRSs only through changing the standard deviation of height dv and the correlation length cl arbitrarily from point to point. Using the least squares method (LSM) based on the 1-ray model, we estimate numerically the modification factor α and propagation order of distance β with respect to EM wave propagation along inhomogeneous RRSs. We also show some numerical examples in order to demonstrate the influence of the inhomogeneity of RRS to the propagation characteristics.