The successful prediction of the stream or river water quality is gaining the attention of various governmental agencies, and pollution control boards worldwide due to its useful applications in determining watershed health, biodiversity, ecology, and suitability of potable water needs of the river basin. The physically based computational water quality models would require large spatial and temporal information databases of climatic, hydrologic, and environmental variables and solutions of nonlinear, partial differential equations at each grid point in a river basin. These models suffer from estimability, convergence, stability, approximation, dispersion, and consistency issues. In such a problematic modeling scenario, an artificial neural network (ANN) modeling of 22 stream water quality parameters (SWQPs) is performed from easily measurable data of precipitation, temperature, and novel land use parameters obtained from Geographic Information System (GIS) analysis for the Godavari River Basin, India. The ANN models are compared with the more traditional, statistical linear, and nonlinear regression models for accuracy and performance statistics. This study obtains regression coefficients of 0.93, 0.78, 0.83, and 0.74 for electrical conductivity, dissolved oxygen, biochemical oxygen demand, and nitrate in testing using feedforward ANNs compared with a maximum of 0.45 using linear and nonlinear regressions. Principal component analysis (PCA) is performed to reduce the input data dimension. The subsequent modeling using radial basis function and ANNs is found to improve the overall regression coefficients slightly for the chosen four water quality parameters (WQPs). A closed form equation for electrical conductivity has been derived from MATLAB simulations. The successful modeling results indicate the effectiveness and potential of ANNs over the statistical regression approaches for estimating the highly nonlinear problem of stream water quality distributions.