Forecast models using historical time series data for temperature and flow are constructed by means of parametric time series models recently proposed by G. E. P. Box and G. Jenkins. An exposition of the new methods and two examples of data illustrative of the iterative procedures necessary to construct a good forecast model are given. The daily water temperature and Ohio River discharge at Wheeling, West Virginia, for January 1, 1963–December 31, 1968, provide the raw material for the models. The methods are easily extended to other problem