Évaluation hydrologique de jeu de données de précipitations et températures journalières issues du modèle régional de climat "CROQ"

Abstract

The low density of the meteorological station network in remote areas and the need of reliable meteorological data for hydrological studies push researchers to seek alternative datasets to complement or even replace traditional data from stations. The objective of this research is to analyze the capability of CROQ regional climate model dataset to reproduce the hydrology of the different watersheds of Quebec. The CROQ regional model dataset comes from a high-resolution simulation of the Canadian regional climate model driven by the ERA-Interim reanalysis. First, the quality of CROQ temperature and precipitation data is evaluated in relation to those of different datasets as "NRCan" interpolated database on the grid, "ERA-Interim" reanalysis and those derived from another simulation from the same regional climate model "MRCC5", but in a different configuration. Subsequently, the hydrological validation of CROQ datasets is performed. To do this, the meteorological data of each dataset are used as input data for two global conceptual hydrological models of different complexity (HSAMI and GR4J) in order to simulate river flows. The potential of the two models to represent river flows is then compared. In order to discover the spatial and temporal diversity between the datasets, climatic and hydrological validation is carried out by various metrics with annual and seasonal time steps and by comparing the hydrological simulations carried out on 117 watersheds in Quebec. These results were then refined by comparing 20 watersheds in southern and northern Quebec. In general, the results of the comparison show that CROQ systematically overestimates precipitation, and its temperatures are colder, especially in the spring months over whole Québec. In southern Quebec, CROQ data are less biased than other datasets and show a lower inter-basin variability. The CROQ dataset allows good performance in hydrological modeling on southern watersheds for both hydrological models. Hydrological modeling results are comparable, although slightly lower than for other datasets. However, in Northern Quebec, the biases associated with the CROQ data give less successful result in hydrological modeling, particularly with the GR4J model. This model presents few adjustable parameters during its calibration and is unable to adjust to existing biases in the precipitation and temperature data from CROQ. In general, ERA-I reanalysis, despite its coarser resolution, gives superior results (especially in the North) in hydrological modeling compared to CROQ. The CROQ data are generally more biased than its driver (ERA-I), indicating that the regional model adds its biases to those of its driver. In hydrological modeling, the increased CROQ resolution does not seem to compensate for these additional biases as well as the increased freedom of the regional model within its computational domain. This increased freedom results in trajectories of meteorological systems potentially slightly different from those observed, which results in a daily correlation of observed and simulated meteorological variables and an inevitable decline in performance in hydrological modeling.

Date	30 Nov 2017
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	François Brissette (Supervisor) & Sébastien Biner (Co-supervisor)