3A.6 Forecasting Flood Hazard on Real Time: Implementation of a New Surrogate Model for Hydrometeorological Events in an Andean Watershed

Monday, 8 January 2018: 3:15 PM
Room 17A (ACC) (Austin, Texas)
Maria Teresa Contreras, Univ. of Notre Dame, Notre Dame, IN; and C. Escauriaza, A. Taflanidis, and J. Westerink

In recent years, the occurrence of flash floods and landslides produced by hydrometeorological events in Andean watersheds has had devastating consequences in urban and rural areas near the mountains. Two factors have hindered the hazard forecast in the region: 1) The spatial and temporal variability of climate conditions, which reduce the time range that the storm features can be predicted; and 2) The complexity of the basin morphology that characterizes the Andean region, and increases the velocity and the sediment transport capacity of flows that reach urbanized areas.

Hydrodynamic models have become key tools to assess potential flood risks. Two-dimensional (2D) models based on the shallow-water equations are widely used to determine with high accuracy and resolution, the evolution of flow depths and velocities during floods. However, the high-computational requirements and long computational times have encouraged research to develop more efficient methodologies for predicting the flood propagation on real time.

Our objective is to develop new surrogate models (i.e. metamodeling) to quasi-instantaneously evaluate floods propagation in the Andes foothills. By means a small set of parameters, we define storms for a wide range of meteorological conditions. Using a 2D hydrodynamic model coupled in mass and momentum with the sediment concentration, we compute on high-fidelity the propagation of a flood set. Results are used as a database to perform sophisticated interpolation/regression, and approximate efficiently the flow depth and velocities in critical points during real storms.

This is the first application of surrogate models to evaluate flood propagation in the Andes foothills, improving the efficiency of flood hazard prediction. The model also opens new opportunities to improve early warning systems, helping decision makers to inform citizens, enhancing the reslience of cities near mountain regions.

This work has been supported by CONICYT/FONDAP grant 15110017, and by the Vice Chancellor of Research of the Pontificia Universidad Catolica de Chile, through the Research Internationalization Grant, PUC1566 funded by MINEDUC.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner