Programme  Poster session 1  abstract 575

Modelling runoff in semi arid mountainous watersheds using the simple conceptual model GR4J – Examples in the High Atlas mountains

Author(s): Vincent Simonneaux (2), L. Hanich(1), G. Boulet(2), S. Thomas(2)
(1) Faculté des Sciences et Techniques Marrakech, Maroc (2) CESBIO (UMR 5126 CNES-CNRS-UPS-IRD), Toulouse, France.

Keyword(s): Runoff, Modelling, Watershed budget, Drought

Article: abs575_article.pdf
Poster: abs575_poster.pdf
Get Adobe Reader

Session: Poster session 1
AbstractThe Tensift watershed (20,000 km2), located around Marrakech (Morocco) is experiencing

water resource shortage resulting from the strong increase of water consumption due to both urban and tourist

development, and more and more frequent drought periods which may prefigure the ongoing climatic changes. Water

management in this area would benefit from tools based on a reliable knowledge and modelling of the water fluxes.

The SudMed project is aimed at a better understanding of the local hydrological cycle, through collaboration with the

local watershed agency.
Considering that the High Atlas Range (maximum altitude 4167 m) is from far the major

water source in the Tensift Bassin, the hydrological modelling of these mountains is necessary to quantify the inputs to

the system and run scenarios for the future. After initial trials using the complex semi-physical distributed hydrological

model SWAT, the lack of detailed physical parameters needed to run such models and the poor quality of the input

(rainfall) or verification (runoff) data lead us to consider the use of simpler but more robust conceptual global models

like GR4J (CEMAGREF), which is based on only four parameters.
This model was applied to four mountainous

sub-watershed (Rheraya, Ourika, Zat, Rdat) over periods ranging from 16 to 30 years, and giving Nash coefficients

between 0.45 and 0.55. It appeared that this poor results were mainly due to the low quality of rainfall / runoff time

series. Nevertheless, in the Rheraya catchment, after calibrating GR4J over 16 years, the validation on the 90/91

hydrological year — considered as the best one regarding data quality — gave a Nash value similar to the one found

by calibrating the SWAT model over this same year (0.83). This encouraging result shows that GR4J calibration

over long periods seems to manage a good trade-off between errors and to provide with a good tuning of the model.

It was shown that taking into account the snow in GR4J through a simple temperature base snow generator, and a

degree day melting module, was improving only slightly the runoff simulation, which is in good agreement with the

SWAT experiments. We also observed that the evapotranspiration term had to be very low, which is in agreement

with the previous SWAT sensitivity experiments, and also with the characteristics of the soil surface (very few

vegetation and high stone coverage). Finally, GR4J was able to simulate correctly the low water level observed for

the last years due to drought periods, and more generally the interannual variations of streamflow. This model seems

therefore well adapted to poor quality data assuming long enough time series are available, and it is able to depict

correctly very different hydrological years, which could make it a candidate tool for simple climate change scenario


  Return up