Author(s): Javier Fernandez-Pato; Estela Nadal-Romero; Noemi Lana-Renault; Pilar Garcia-Navarro

Linked Author(s): Javier Fernández-pato

Keywords: GPU simulation; Hydrology; Mountain catchments; Shallow water equations; Central Pyrenees

Abstract: Environmental sciences have evolved in the last decades due in part to the introduction of simulation techniques at catchment scale. These strategies make it possible to evaluate and predict the behavior of water in any type of basin. A properly calibrated an efficient computational model represents a very useful tool to provide insight into the catchment dynamics at hydrological level. Furthermore, the use of distributed models allows to obtain very detailed results thanks to the spatial discretization of the terrain in computational cells. The greatest counterpoint of this type of numerical model is its high computational cost, since a large number of operations are necessary in every time step. This is especially relevant when considering long-term hydrological events (e.g. 1 year). In order to overcome this issue, the numerical model used in this work is solved by taking advantage of all the power of the latest generation graphics cards (GPUs), achieving simulations hundreds of times faster than with a normal processor (CPU). In particular, three mountain catchments (Arnás, Araguás and San Salvador) are considered for the simulation of long-term storm events. The basins domain is discretized by means of a locally refined unstructured triangular mesh and distributed rainfall and infiltration processes are taking into account in order to provide an adequate input data. The numerical results show a good agreement between numerical and observed outlet hydrographs.

DOI: https://doi.org/10.3850/IAHR-39WC252171192022270

Year: 2022