Inland navigation is an important part of traditional transport and traffic. The security of inland navigation depends to a large extent on inland river communication. This paper presents an analysis of shadowing fade correlation and proposes an accurate statistical model for inland river radio channel. The main obstacles in inland river communication environment are considered, i.e., huge bridge, building, and large ship. Our results show that decorrelation distances are 100-150 wavelengths at 1.4 GHz in bridge and building scenarios and 500 wavelengths at 5.9 GHz in large ship scenario, respectively. Moreover, we propose a triple Gaussian mixture distribution (TGMD) to characterize the inland river radio channel, considering the transition between line of sight (LOS), obstructed LOS (OLOS) and non-LOS situations. Rician K-factor, root mean square (RMS) delay spread and RMS Doppler spread are respectively modeled by TGMD with a high goodness of fit. The results can provide a meaningful reference to the research of inland river radio channel.