Using ANN-SSA coupled with uneec for uncertainty estimate of daily rainfall-runoff transformation

Uncertainty has always been inherent in water resources engineering and management. For example, in river flood defenses it was treated implicitly through conservative design rules, or explicitly by probabilistic characterization of meteorological events leading to extreme floods. Incorporating prediction uncertainty into deterministic forecasts (or called point forecasts) helps enhance the reliability and credibility of the model outputs. In this article, point forecasting of daily rainfall-runoff is first estimated by artificial neural network with singular spectrum analysis model (ANN-SSA), and then uncertainty estimation based on local errors and clustering method (UNEEC), which is based on model errors, is employed for uncertainty analysis of point prediction with the bootstrap method as comparison. Results indicate that UNEEC is capable of making appropriate uncertainty predictions in terms of prediction interval coverage probability (PICP). However, occurrence of some negative lower prediction limits implies that performance of UNEEC can be further enhanced by improving the ANN-SSA model. Compared with the bootstrap method, UNEEC performed better in locations of the low runoff whereas the bootstrap method proves to be better for estimates of prediction intervals in locations of the high runoff.