Forecasting Monthly Streamflow of Spring-Summer Runoff Season in Rio Grande Headwaters Basin Using Stochastic Hybrid Modeling Approach

Monthly streamflow forecasting during spring-summer runoff season using snow telemetry (SNOTEL) precipitation and snow water equivalent (SWE) as predictors in the Rio Grande Headwaters Basin in Colorado was investigated. The transfer-function noise (TFN) models with SNOTEL precipitation input were built for monthly streamflow. Then, one-month-ahead forecasts of TFN models for the spring-summer runoff season were modified with SWE using an artificial neural networks (ANN) technique denoted in this study as hybrid TFN + ANN. The results indicated that the hybrid TFN + ANN approach not only demonstrated better generalization capability but also improved one-month-ahead forecast accuracy significantly when compared with single TFN and ANN models. The normalized root mean squared errors (NRMSE) of one-month-ahead forecasts of TFN, ANN, and TFN + ANN approaches for spring-summer runoff season were 0.38, 0.30, and 0.25. These findings accentuate that the presented stochastic hybrid modeling approach is an advantageous option to improve one-month-ahead forecast accuracy of monthly streamflow in spring-summer runoff season in the Rio Grande Headwaters Basin. DOI: 10.1061/(ASCE)HE.1943-5584.0000322. (C) 2011 American Society of Civil Engineers.