300
I. Thiaw et al.
98% (DC1); 27% and 63% (DC2); and 4% and 11% (DC3), respectively. Model
IPSL-CM5A-LR, predicts a 20% increase in DMAX, and a chronic decrease in
DCC_10j (57%), DCC_20j (86%), DC1 (96%) and DC2 (67%), by 2050. Overall
averages with RCP8.5 (multi-model-RCP8.5) predicts a decrease of 41% (DMAX),
66% (DCC_10j), 80% (DCC_20j and DC1), 51% (DC2) and 7% (DC3). The models
also predict by 2050 an increase of 23%, 32% and 33% of the DC4, DC5 and median
flows, respectively.
For low flow characteristic rates (DC7, DC8, DC9, DC10, DC11, DCE_20j,
DCE_10j, and DMIN), models predict an average increase of 63% by 2050, under
RCP8.5. In sum, the hydrological model submitted to the outputs of regional climate
models projects a decrease in DMAX, DCC_10j, DCC_20j, DC1, DC2 and DC3,
and an increase in Diarha low flow characteristic rates under both emission scenarios
(RCP4.5 and RCP8.5). The hydrological model only predicts an increase in DMAX
with the outputs of the IPSL-CM5A-LR model under both RCP scenarios.
4 Discussion and Conclusions
Predicting the variation of water volumes in a watershed is essential because it enables
planners to assess the correlation between water demand and water availability in
order to anticipate and manage potential conflicts among users and sectors. In this
study, two hydrological models, GR4J and SAC-SMA, of the RS Minerve were
calibrated based on existing rainfall data for the periods 1975–1992 and 1998–2003.
Results show that the GR4J model is better correlated with actual observed flow
rates of the Diarha over the calibration (1975–1992) and validation (1998–2003)
phases. The parameters X1, X2, X3 and X4 were determined based on the calibration
period and used to simulate flow rates over the missing periods. This allowed for
the extension of data to represent discharges continuously from 1961 to 2012. In
addition, randomness verification tests show the absence of stability on Diarha flow
chronicles, confirming the sensitivity of climate variables to climate change. (Stanzel
et al. 2018). The work of Thiaw (2017) confirms that rainfall series in the Diarha
catchment are marked by high interannual variability and long-term fluctuations,
from a very wet period between 1921 and 1967 to a very dry period between 1968
and 1990, followed by a slight return of rainfall starting in the 2000s.
At the same time, it should be noted that flow rates do not depend solely on climate.
Several factors related to the hydrological structure of the watershed (geology,
pedology, land use, etc.) may also impact model outputs. Unfortunately, the GR4J
model, though calibrated in semi-distributed conceptual mode, does not integrate
land use dynamics in the flow simulation. In addition, several studies have shown
that this model does not fully account for the relationship between groundwater
and surface water (Fabre et al. 2014, Bodian et al. 2018). Therefore, depending on
the physical characteristics of the basins modeled, the GR4J model overestimates
or underestimates low flows and peak flows. There are other hydrological models
that better take into account the interaction between groundwater and surface water
I. Thiaw et al.
98% (DC1); 27% and 63% (DC2); and 4% and 11% (DC3), respectively. Model
IPSL-CM5A-LR, predicts a 20% increase in DMAX, and a chronic decrease in
DCC_10j (57%), DCC_20j (86%), DC1 (96%) and DC2 (67%), by 2050. Overall
averages with RCP8.5 (multi-model-RCP8.5) predicts a decrease of 41% (DMAX),
66% (DCC_10j), 80% (DCC_20j and DC1), 51% (DC2) and 7% (DC3). The models
also predict by 2050 an increase of 23%, 32% and 33% of the DC4, DC5 and median
flows, respectively.
For low flow characteristic rates (DC7, DC8, DC9, DC10, DC11, DCE_20j,
DCE_10j, and DMIN), models predict an average increase of 63% by 2050, under
RCP8.5. In sum, the hydrological model submitted to the outputs of regional climate
models projects a decrease in DMAX, DCC_10j, DCC_20j, DC1, DC2 and DC3,
and an increase in Diarha low flow characteristic rates under both emission scenarios
(RCP4.5 and RCP8.5). The hydrological model only predicts an increase in DMAX
with the outputs of the IPSL-CM5A-LR model under both RCP scenarios.
4 Discussion and Conclusions
Predicting the variation of water volumes in a watershed is essential because it enables
planners to assess the correlation between water demand and water availability in
order to anticipate and manage potential conflicts among users and sectors. In this
study, two hydrological models, GR4J and SAC-SMA, of the RS Minerve were
calibrated based on existing rainfall data for the periods 1975–1992 and 1998–2003.
Results show that the GR4J model is better correlated with actual observed flow
rates of the Diarha over the calibration (1975–1992) and validation (1998–2003)
phases. The parameters X1, X2, X3 and X4 were determined based on the calibration
period and used to simulate flow rates over the missing periods. This allowed for
the extension of data to represent discharges continuously from 1961 to 2012. In
addition, randomness verification tests show the absence of stability on Diarha flow
chronicles, confirming the sensitivity of climate variables to climate change. (Stanzel
et al. 2018). The work of Thiaw (2017) confirms that rainfall series in the Diarha
catchment are marked by high interannual variability and long-term fluctuations,
from a very wet period between 1921 and 1967 to a very dry period between 1968
and 1990, followed by a slight return of rainfall starting in the 2000s.
At the same time, it should be noted that flow rates do not depend solely on climate.
Several factors related to the hydrological structure of the watershed (geology,
pedology, land use, etc.) may also impact model outputs. Unfortunately, the GR4J
model, though calibrated in semi-distributed conceptual mode, does not integrate
land use dynamics in the flow simulation. In addition, several studies have shown
that this model does not fully account for the relationship between groundwater
and surface water (Fabre et al. 2014, Bodian et al. 2018). Therefore, depending on
the physical characteristics of the basins modeled, the GR4J model overestimates
or underestimates low flows and peak flows. There are other hydrological models
that better take into account the interaction between groundwater and surface water