All previous equations presented for hydraulic geometry of stable channels were mainly based on uniform flows and thus are not suitable for using in natural rivers and channels. The main objectives of this study were evaluating distribution of local shear stress and Shields parameter in nonuniform cross sections in stable channels and also developing a model for prediction hydraulic geometry of stable cross sections using statistical nonlinear multiple regression. For this purpose, shear velocity was estimated using hydrodynamic methods, field measurements of velocity distribution profiles, bed sediment characteristics and geometric characteristics of channel cross sections. Suitable equations to predict width, depth and longitudinal slope of the channel were produced using hydraulic parameters such as shear stress, flowrate and morphological parameters such as sediment particle diameter. This study was conducted in two sections. For the first part, four reaches of a 20 km length in Kaaj River, located in Chahar Mahal Bakhtiari province, were selected. The second part including 12 reaches of Sheikh Bahaei landscape channels were chosen. In general, of the total studied reaches, geometric characteristics of 76 cross sections, 338 velocity profile (including 5000 point velocity) and characteristics of bed materials and surveying data were recorded. By using preliminary calculations, morphological and hydraulic characteristics were determined. Afterwards, shear velocity of each profile was estimated using data acquired from velocity profile and boundary layer characteristics method. Eventually, shear stress and shields parameter were determined and their distribution patterns in non-uniform cross sections were studied. Results indicated that distribution pattern of these two parameters is a useful index for determination of stability in the studied cross sections. In order to predict hydraulic geometry of stable channels, including width, depth and slope, a model using nonlinear multiple regression was produced. These equations were estimated separately for river and channel cross sections. Results revealed that depth predicting models used in two parts of this study were significantly effective. Presenting proper equations with acceptable correlation coefficients to anticipate width and slope of channels failed.