Forecasting streamflow for Colorado River systems

Abstract

The research reported herein was aimed at improving current forecasting procedures used in the State of Colorado. In this study, single input-single output and multiple input-single output periodic transfer function models have been developed for forecasting monthly stream flow. First, the monthly stream flow is deseasonalized and filtered by a periodic autoregressive (PAR) model. then, a transfer function model, in which the deseasonalized snow water equivalent is the input, and the streamflow residual from the PAR model is the output, has been formulated. The building of the transfer function model has been carried out by using spectral analysis and the non-linear least squares method was used for parameter estimation. The models have been applied to forecast monthly flows of the Rio Grande watershed system in southern Colorado.