Improvement of low flows simulation in the SASER hydrological modeling chain

Omar Cenobio-Cruz, Pere Quintana-Seguí, Anaïs Barella-Ortiz, Ane Zabaleta, Luis Garrote, Roger Clavera-Gispert, Florence Habets, Santiago Beguería

Research output: Indexed journal article Articlepeer-review

3 Citations (Scopus)

Abstract

The physically-based, spatially-distributed hydrometeorological model SASER, which is based on the SURFEX LSM, is used to model the hydrological cycle in several domains in Spain and southern France. In this study, the modeled streamflows are validated in a domain centered on the Pyrenees mountain range and which includes all the surrounding river basins, including the Ebro and the Adour-Garonne, with a spatial resolution of 2.5 km. Low flows were found to be poorly simulated by the model. We present an improvement of the SASER modeling chain, which introduces a conceptual reservoir, to enhance the representation of the slow component (drainage) in the hydrological response. The reservoir introduces two new empirical parameters. First, the parameters of the conceptual reservoir model were determined on a catchment-by-catchment basis, calibrating against daily observed data from 53 hydrological stations representing near-natural conditions (local calibration). The results show, on the median value, an improvement (ΔKGE of 0.11) with respect to the reference simulation. Furthermore, the relative bias of two low-flow indices were calculated and reported a clear improvement. Secondly, a regionalization approach was used, which links physiographic information with reservoir parameters through linear equations. A genetic algorithm was used to optimize the equation coefficients through the median daily KGE. Cross-validation was used to test the regionalization approach. The median KGE improved from 0.60 (default simulation) to 0.67 (ΔKGE = 0.07) after regionalization and execution of the routing scheme, and 79 % of independent catchments showed improvement. The model with regionalized parameters had a performance, in KGE terms, very close to that of the model with locally calibrated parameters. The key benefit if the regionalization is that allow us to determine the new empirical parameter of the conceptual reservoir in basins where calibration is not possible (ungauged or human-influenced basins).

Original languageEnglish
Article number100147
JournalJournal of Hydrology X
Volume18
DOIs
Publication statusPublished - 1 Jan 2023

Keywords

  • Distributed modeling
  • Genetic Algorithm
  • Hydrology
  • Land-Surface Model
  • Low Flows
  • Parameter Regionalization

Fingerprint

Dive into the research topics of 'Improvement of low flows simulation in the SASER hydrological modeling chain'. Together they form a unique fingerprint.

Cite this