Discussion of “Comparative assessment of time series and artificial intelligence models to estimate monthly streamflow: A local and external data analysis approach” by Saeid Mehdizadeh, Farshad Fathian, Mir Jafar Sadegh Safari and Jan F. Adamowski

Abstract This discussion extends published findings on the use of artificial intelligence models for time series modeling/monthly streamflow forecasting at the Port Elgin station on the Saugeen River, Canada. Published results are applied and run in autoregressive (AR) and moving average (MA) models as well as hybrids of these with autoregressive conditional heteroscedasticity (ARCH), namely AR-ARCH and MA-ARCH. The hybrid solutions are concluded to be superior to both linear and nonlinear modeling approaches. However, common nonlinear methods including neural networks have a recognized defect in time series forecasting known as “inappropriate time series modeling inputs”. The present study addresses this significant source of error in nonlinear modeling by referring to time series components via suitable time series preprocessing. Of particular interest in this discussion provides the novel vision for time series modeling using nonlinear approaches. The nature of the hydrological variables in time series modeling has great impact on the predicted output and should thus be considered in the modeling procedure. An appropriate preprocessing technique must also be considered carefully in order to attain reliable nonlinear modeling results.