Real-time monitoring of river water quality is at the forefront of a proactive urban water management strategy to meet the global challenge of vital freshwater resource sustainability. The concentration of dissolved oxygen (DO) is a primary indicator of the health state of the aquatic habitats, and its modeling is crucial for river water quality management. This paper investigates the importance of the choices of different techniques for preprocessing and stochastic modeling for developing a simple and reliable linear stochastic model for forecasting DO in urban rivers. We describe several methods of evaluation, preprocessing, and modeling for the DO parameter time series in the Credit River, Ontario, Canada, to achieve the optimum data preprocessing and input selection techniques and consequently obtain the optimum performance of the stochastic models as an effective river management tool. The Manly normalization and standardization (S_td) methods were chosen for preprocessing the time series. Modeling the preprocessed time series using the stochastic autoregressive integrated moving average (ARIMA) model resulted in very accurate forecasts with a negligible difference from sole normalization and spectral analysis (S_f) methods.