Identification of Streamflow Stochastic Models

The identification of streamflow stochastic models is explored. The identification of the type of stochastic model is made, based on a conceptual physical representation of natural watershed. The identification of the form of the model is made, based on the recently developed R-functions and S-functions. ARMA is the autoregressive moving average. For a ARMA(p,q) precipitation input, the ground-water storage is an ARMA(p+1, q) process, and the streamflow is an ARMA(p+1, q+1) process. In general, such ground-water and streamflow processes belong to the class of restricted ARMA processes in the sense that their parameter space is a subspace of that corresponding to the general ARMA models. The form or order of the ground-water and streamflow ARMA processes for given historical time series can be uniquely identified by using the R-functions and S-functions. An example is given as a application of such techniques.