Real-time assessments of stream flows under ice cover, using two distinct objective approaches, were compared over a 5-year period. Approaches based on artificial neural networks, defining mathematical relationships between stream flow, water level, and air temperature, and on a deterministic hydrological model were applied at eight gauged sites located in southern Quebec. Good results were obtained using both approaches, when no snowmelt contributes to the rise of the inflows. In the other hydrological situations, the neural network results were the best, but results of both approaches were sensibly poorer. Nevertheless, the potential for increasing the skills of the deterministic model seems high. Otherwise, a preliminary analysis showed that both approaches lead to stream-flow estimations that are not radically worse than the ones performed by the team of experts.Key words: discharge measurement, hydrology, ice-affected stream flow, hydrological modeling, neural network.
Predicting the Determinants of Dynamic Geometry Software Acceptance: A Two-Staged Structural Equation Modeling — Neural Network Approach
