The limitations and usefulness of streamflow generation methods: a case study
Probabilistic models have become important hydrologic tools. However, increasing model complexity makes the connections between the model and the physical world more and more vague. This can lead to a de-emphasis of engineering judgment, since model validity is easily assumed when even partial verification must await future occurrences. A simple autoregressive model was used to generate stochastic flow sequences for the dam and reservoir being constructed on the Red Deer River in Alberta. The results from this model were compared with those obtained from a more complex autoregressive moving average (ARMA) model. Both models have similar deficiencies. It is concluded that since stochastic generation can never represent future conditions with certainty, the common practice of basing the hydrologic design of reservoirs on actually recorded data is usually the most valid procedure. However, stochastic streamflow generation can be used to give valuable probabilities of reservoir storage failure.