Impact of future climate change on water supply and irrigation demand in a small mediterranean catchment. Case study: Nebhana dam system, Tunisia

This study evaluates the impacts of climate change on water supply and demand of the Nebhana dam system. Future climate change scenarios were obtained from five general circulation models (GCMs) of CMIP5 under RCP 4.5 and 8.5 emission scenarios for the time periods, 2021–2040, 2041– 2060 and 2061–2080. Statistical downscaling was applied using LARS-WG. The GR2M hydrological model was calibrated, validated and used as input to the WEAP model to assess future water availability. Expected crop growth cycle lengths were estimated using a growing degree days model. By means of the WEAP-MABIA method, projected crop and irrigation water requirements were estimated. Results show an average increase in annual ETo of 6.1% and a decrease in annual rainfall of 11.4%, leading to a 24% decrease in inflow. Also, crops’ growing cycles will decrease from 5.4% for wheat to 31% for citrus trees. The same tendency is observed for ETc. Concerning irrigation requirement, variations are more moderated depending on RCPs and time periods, and is explained by rainfall and crop cycle duration variations. As for demand and supply, results currently show that supply does not meet the system demand. Climate change could worsen the situation unless better planning of water surface use is done. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/wcc.2019.131 om http://iwaponline.com/jwcc/article-pdf/11/4/1724/830438/jwc0111724.pdf er 2021 M. Allani (corresponding author) R. Mezzi A. Sahli Laboratoire des Sciences Horticoles (LSH), Institut National Agronomique de Tunisie (INAT), Université de Carthage (UC), 43, Avenue Charles Nicolle 1082, Tunis-Mahrajène, Tunisia E-mail: mohamed.allani@gmail.com A. Zouabi R. Béji F. Joumade-Mansouri Commissariat Régional au Développement Agricole de Kairouan (CRDA), Cité Sidi Layoun, Kairouan 3100, Tunisia M. E. Hamza Institution de la Recherche et de l’Enseignement Supérieur Agricoles (IRESA), 30, Rue Alain Savary 1002, Tunis Belvédère, Tunisia


INTRODUCTION
A combination of climate change and increase in irrigation usage is likely to substantially decrease fresh water availability (by 2-15% for 2 C of warming), especially in Mediterranean regions, which would lead to the largest decreases in the world (Cramer et al. ). In fact, these regions are likely to warm at a rate about 20% greater than the global annual mean surface temperature (Lionello & Scarascia ). For these countries, water management is facing major challenges due to increasing uncertainties caused by climate change. Policies need to mitigate these risks and consider adaptation options; however, decisionmakers currently lack adequate information, particularly in the south Mediterranean regions where impact assessment studies are limited (Cramer et al. ).
In these regions, the use of water management tools, taking into consideration water supply and demand variability in an integrated way, are needed. Water Evaluation and Planning software (WEAP) is a modelling decision support system that helps assess climate, hydrology, land use, infrastructure and management priorities at regional levels (Yates et al. a, b). WEAP has been widely used around the world with differents objectives, such as the management of surface and groundwater resources, adaptation strategies and policy changes. giving better simulation of the climate conditions (Liu et al. ; Reshmidevi et al. ). In this study, as one statistical downscaling method, a stochastic weather generator, namely, LARS-WG 6.0 (Semenov et al. ; Semenov & Barrow ), was used to extract the local-scale future daily rainfall temperature and radiation from five GCMs of the CMIP5 available in the software incorporating the daily observed climate data under two representative concentration pathways (RCPs) (4.5 and 8.5). The LARS-WG weather generator has been widely used in impact assessment studies because it is not only computationally inexpensive, it delivers climate scenarios that match the statistical properties of observed weather and provide daily meteorological variables while preserving statistical interrelationships between variables (Semenov & Stratonovitch ). All climatic models' outputs involve a number of biases that, if not corrected, can lead to significant errors in impact assessments.
Therefore, bias correction of models' outputs is necessary before their use in impact analyses (Ahmed et  there is a continuing debate in the hydrologic modelling research area on whether physically based distributed models better capture recorded streamflow than the conceptual lumped models approach does. Therefore, a better understanding of the hydrological processes in impact studies should not necessarily translate into the use of more complex models (Blöschl & Montanari ).
Lumped models are the most widely used tools for operational applications because they offer simplified catchment-scale representations of the precipitation's transformation into river discharge (Coron et al. ). When dealing with climate change impact on water supply and demand systems and the general evolution of the overall water availability, the use of monthly time step hydrological models is accepted (Lespinas et al. ) The specific objectives of this research are to: (1) analyse the impact of climate change on rainfall and reference evapotranspiration; (2) evaluate the variability of the water supply to a system through the inflow received to a reservoir; (3) study the impact of climate change on crop growth cycle; (4) calculate crop water demand and irrigation water demand; (5) examine the expected future change in irrigation demand of the whole system.

STUDY AREA
The study area, referred to as the Nebhana dam system, is located in the centre of Tunisia, downstream of the Nebhana watershed, and is composed of six public irrigated districts ( Figure 1). The reservoir is supplied by two main rivers, Maarouf and Nebhana. Through a system of pipelines, water is delivered to supply the six irrigated districts within the study area and also for irrigation and domestic use outside the watershed. The Nebhana dam system has semi-arid weather with a cool wet season from November to March and a dry hot season from May to September. The average annual rainfall (1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004) is 309 mm and the reference evapotranspira-

Data collection and characterization of the study area
The collected data were relative to irrigation and crop water requirements. They focused on climatic data, land use and cultivated area, crop and soil characteristics and volume of supplied water to the Nebhana dam system.

Climate characterization
Daily meteorological data were obtained from the Tunisian National Institute of Meteorology (INM) from Kairouan climatic station located 30 km from the study area. The climate data include maximum temperature (T max , C), minimum temperature (T min , C), maximum relative humidity (RH max , %), minimum relative humidity (RH min , %), sunshine duration (SD, hrs) and wind speed (U 2 , m/s). Daily rainfall (P, mm) data were collected from a rain gauge located within the study area in a radius of 20 km from the farthest irrigated district. These data were obtained for the 1985-2004 period. Based on the collected data, the reference evapotranspiration (ETo) was calculated using the Penman-Monteith FAO 56 method (Allen et al. ).   with the lowest inflow reached during July and the highest inflow during March with 0.4 and 3.6 Mm 3 month -1 , respectively. Meanwhile, the average annual outflow from the dam is 21.3 Mm 3 year -1 . Since the primary use of the Nebhana dam is irrigation, outflow from the dam has the higher value during spring and summer with a maximum of 2.6 Mm 3 achieved during July. This is a critical issue for the study area where the Nebhana dam system barely ensures the balance between demand and supply.

Land use characteristics
Land use and land cover patterns in a region are the prerequisites for planning and implementing effective land use policies and schemes for sustainable regional development (Abdal & Suleiman ). In order to assess climate change impact on the water resources of the system, a field   (Table 1).
It is important to note that the created crop characteristics' database was validated during workshop sessions with the participation of stakeholders from all levels, including farmers as well as engineers from the local extension service and regional representatives of the ministry of agriculture.

Soil characteristics
The soil water content at saturation, field capacity and wilting point are used to calculate soil water holding capacity.
When missing, they can be estimated using pedotransfer functions and soil texture characteristics ( Jabloun & Sahli ; McNeill et al. ). In our case, mean particle size   In addition to the transformed NSE, the model mean cumulative error (CE) was also used to assess its efficiency: where CE measures the ability of the model to correctly reproduce the total water volume observed over the simulation period. When it equals 100%, the water balance is perfectly simulated and when greater or lower than 100% it is overestimated or underestimated, respectively.
The obtained values X1 and X2 for the GR2M model are 10.341 for X1 and 0.200 for X2, respectively. Table 3 presents the performance of the model for the calibration and the validation periods.  Figure 5 shows the modelled and observed runoff series of the calibration and validation periods.   1986-1989 1990-2004 Nash√ ( The generated climatic data concern three periods of time, 2021-2040, 2041-2060 and 2061-2080 (     Under the RCP 4.5 emission scenario, annual inflow is reduced by 28% in the first period, 14% in the second and   The authors explain this sensitivity of runoff by the combination of increasing ETo and decreasing precipitation.

Reference evapotranspiration and rainfall changes
Monthly inflow to the dam shows similar changes to the annual inflow (Figure 8).
From the baseline data, it is important to notice that the winter season represents the period of the year with the highest inflow to the dam (40%) followed by autumn (30%), spring (25%) and summer (5%).
For the first period, 2021-2040, the reduction in monthly inflow is projected to range from 20% during summer to 40% during autumn under the RCP 4.5 emission scenario. In contrast, a slight increase of 3% is expected during spring. Under RCP 8.5, inflow is projected to decrease by 19% during autumn, 17% during winter, 1% during spring and 44% during summer (Figure 8(a)).
During the second period, 2041-2060, the decrease varies from 2% during winter to 48% during summer under RCP 4.5 and from 10% during spring to 35% during summer under RCP 8.5 (Figure 8(b)).
For the third period, 2061-2080, the decrease is substantially identical under the RCP 4.5 and 8.5 scenarios with 41% in autumn and 39% in winter. During spring, the decrease reaches 16% under RCP 4.5 and 24% under RCP 8.5 (Figure 8(c)).

Lespinas et al. () report similar results on small
Mediterranean coastal rivers where climate change scenarios led to a lower reduction during winter and a higher discharge reduction during summer than spring and autumn seasons.
Finally, in regards to the global water balance of the system, it appears that the water resources currently do not meet the demand system requirement. The situation could get worse under climate change, unless the water demand outside the study area is supplied from another source.
The integrated modelling approach presented here provides a meaningful framework to analyse the status of a