Integrated water management under climate change through the application of the WEAP model in a Mediterranean arid region

In arid and semi-arid environments, water plays a crucial role in sustainable development, including poverty reduction. The Mediterranean region, due to its geographical location, is experiencing frequent droughts (Fniguire et al. 2017; González-Hidalgo et al. 2018; Tramblay & Hertig 2018; Zittis 2018; Paniagua et al. 2019; Seager et al. 2019; Tognetti et al. 2019). The succession of drought years makes water resources increasingly less available and difficult to access. Likewise, surface water availability is significantly decreasing (Marchane et al. 2017; Caloiero et al. 2018), which, in turn, results in an overexploitation of groundwater resources (Hssaisoune et al. 2020). The demand for irrigation on one side, and for drinking and industrial water on the other, continue to increase due to demographic development, land use change and economic development (Droogers et al. 2012; Le Page et al. 2012). Moreover, the discharge of saline wastewater by various industries degrades water quality and makes it unusable directly for potable water (via desalination) and industrial applications (Panagopoulos 2021a, 2021b, 2022).

In Morocco, water is a vital issue. The available resource is among the lowest in the world when compared to the population. According to Gassert et al. (2013), Morocco is experiencing a worrying water situation: in fact, it is one of the countries most threatened by water scarcity and occupies the 23rd place out of 165 countries exposed to water risks. Worse still, the country's water resources are estimated at 650 m³ per capita per year. This volume places it in a situation of acute distress. Also, future scenarios predict that consumption needs will increase manifold while resources will become less and less available (Hellegers et al. 2013; Pascual et al. 2017; Le Page et al. 2020). In addition, in Morocco, which is highly exposed to climate change impacts (Brahim et al. 2017; Hadria et al. 2019; Ouatiki et al. 2019), the water availability challenges might be accentuated in the future. The country is likely to experience annual decreases in precipitation, change in their distribution, change in extreme events occurrence and increases in temperature (Driouech et al. 2010; Hadri et al. 2020). Furthermore, all these changes can be reflected to the decrease in the surface runoff (Seif-Ennasr et al. 2016; Marchane et al. 2017; El Khalki et al. 2021).

The agricultural sector accounts for 13% of the total Moroccan gross domestic product (GDP). However, its socio-economic impact is much larger, as it provides more than a third of national employment and nearly 74% of jobs in rural areas. Furthermore, since the implementation of the Green Moroccan Plan (GMP), many regions have known a strong expansion of irrigated areas with the proliferation of new entrepreneurial types of farms, large and specialized in market gardening and arboriculture, driven by subsidies allocated by the country's Agricultural Development Fund (FDA). Consequently, this situation of intensive exploitation is leading to substantial pressure on groundwater and a large depletion of the water table (Malki et al. 2017; Bahir et al. 2021). Indeed, the groundwater abstraction exceeds the annual renewable groundwater volume, which has produced, over a long time period, a continuous decline in the piezometric level (20–65 m in the past 30 years) (Gleeson et al. 2016; Hssaisoune et al. 2020). In the entire country, an annual groundwater deficit of about 862 MCM is recorded (Hssaisoune et al. 2020). The overexploitation is reaching a rate of 248% in some aquifers like Saiss (Benabdelfadel 2012) and is creating an annual average decline in groundwater level up to 2 m in Souss (Bouchaou et al. 2011) and up to 1.8 m in the Mejjate plain (Hadri et al. 2021). Additionally, the drastic imbalance between groundwater extraction and recharge and the improvements in drilling techniques push farmers to solicit deep aquifers, where the groundwater renewal time needs several decades (Lapworth et al. 2015).

Accordingly, it is necessary to establish sustainable withdrawals and supply of water and increase water-use efficiency through an improvement of water resource management and resolve problems resulting from the imbalance between water demand and its availability (Ragab & Prudhomme 2002; Bahir et al. 2021). For these reasons, the water sector is now a primary concern for public authorities. The latter try to adopt policies and strategies to ensure water conservation and management, within the framework of socio-economic development (Molle & Tanouti 2017; Kadi & Ziyad 2018). Several planning tools and structures intending to help the management of regional water resources and simulated by the Moroccan Water law 36-15 have been implemented: Master Plan for Integrated Water Resources Management (PDAIRE), aquifer contracts, Participatory Water Management Contract (CGPE), Basin Comity, etc.. However, these institutional tools remain unsatisfactory due to a lack of a deep knowledge of water resource potentials and demands and the need of water management tools, considering, with an integrated approach, water supply and demand variability and surface and groundwater abstractions.

In the last two decades, digital software dedicated to basin modelling has undergone major development and usage. Researchers have proposed several tools, sometimes called decision support systems (DSSs), to help managers and decision-makers in water resource management (Hadded et al. 2013). The limits of these tools lie in their considerable data inputs, their calibration and validation. The Water Evaluation And Planning (WEAP) model has emerged as a powerful tool for decision support (Yates et al. 2005; Le Page et al. 2012). It is commonly used to model the management of a watershed with its water resources, infrastructure and management rules. It allows us to establish a conjunctive management of surface and groundwater resources and provides a representation of the physical and spatial dimensions of water resources, climate and water demand management (Esteve et al. 2015). WEAP was also used by several researchers in Morocco to explore the impact of climate change and land use change on water supply, demand and management (Johannsen et al. 2016; Ayt Ougougdal et al. 2020) and to implement the water decision support system aiming to compare spatially and temporally sectorial water demands in regard to available surface and groundwater resources (Le Page et al. 2012).

In order to assess the impact of climate change and the different scenarios on the availability of water resources, the regional climate models (RCMs) are an added value of these studies especially for data-sparse regions where the observed precipitation and discharge cannot be available, such as Morocco (Tramblay et al. 2013; El Khalki et al. 2018; El Khalki et al. 2020). These models can provide gridded regional averages over large basins (Tramblay et al. 2018). In the case of the absence of long series of observed data, the RCM simulations can be a solution to study the climate system of the Mediterranean region (Romera et al. 2015). The assessment of future water availability is strongly linked with water supply (precipitation) and the atmospheric demand (evapotranspiration). In the case of the absence of discharge data, the runoff could be estimated through the difference between precipitation and evapotranspiration (Alessandri 2015). This method offers an alternative to assess future projections of water resources.

Given that the agricultural water needs, based essentially on crop water demand, are one of the important inputs of the water resource management and water allocation models, the control of the evolution of this demand in the future is an important prerequisite. Several studies have addressed this evolution by limiting the analysis by considering the effect of the increase in temperature on the reference evapotranspiration and consequently the increase in future crop water requirements throughout the 21st century (Gorguner & Kavvas 2020; Haider & Ullah 2020; Kobuliev et al. 2021). However, the consideration of the impact of the increase in CO₂ and the change in air temperatures, foreseen by the various RCPs, on the length of crop growing period and on the crop water demand should not be neglected (Chattaraj et al. 2014; Saadi et al. 2015; Zheng et al. 2020; Funes et al. 2021). Several authors have observed that the increase in reference evapotranspiration (ET₀) can be compensated by the reduction in the plant's development cycle. Nevertheless, a warmer climate will influence phenological dates, stomatal closure and crop water requirement but may also affect plant productivity and crop yields (Hatfield & Prueger 2015), which can disturb farmers' incomes and food safety.

Several software programs have been used to calculate crop water needs under various climate conditions. Aquacrop, developed by Water Unit at the FAO as a revised version of CropWat, is a useful tool to predict crop response and quantity of water required over time under changes of climate conditions (RAES et al. 2012). AquaCrop has already been used in many impact projection studies (Abedinpour et al. 2014; Bird et al. 2016; Akumaga et al. 2018). In the western regions of Morocco, AquaCrop simulates canopy cover, actual evapotranspiration, total soil water content and grain yield reasonably well over two growing seasons for winter wheat (Toumi et al. 2016). Toumi et al. (2016) argued that the AquaCrop tool could be considered as a potentially useful tool for planning irrigation schedules in arid and semi-arid regions.

In this study, a DSS for water resource management in the Chichaoua-Mejjate region was developed by combining WEAP and Aquacrop. This system has allowed us to (i) simulate several management and adaptation scenarios under climate change, (ii) consider the impact of the increase in temperature on crop growth cycle and then on crop water demand and (iii) assess the groundwater and surface water interaction, and the impact of limited groundwater on meeting water needs.

The specific objectives of this study are to (i) formulate a response to the growing water scarcity resulting from the conflict of use between irrigated perimeter and the drinking water supply (DWS) of surrounding municipalities in a semi-arid basin qualified as a data-scarce zone and (ii) contribute to establishing a framework and a methodology for sustainable and reliable water resources management in the future, in an understandable way for policy-makers.

The Chichaoua-Mejjate area is located at 70 km southwest of Marrakech. It covers a surface of 5,193 km² and is composed of three separate topographic areas: the High Atlas Mountains that are considered as a water tower of the surrounding regions (Bouamri et al. 2018), the foothill zone and the plain area considered as a water consumer (Figure 1). This zone is a vast expanse of moderate relief which becomes more accentuated towards the south in the vicinity of the Atlas range. It is in the form of a succession of east–west to north–east–south–west anticlinal ridges, separated by flat-bottom synclines (Ambroggi & Thuille 1952).

There is a rapid development in this study area, especially agriculture which is the main component of economic activity with over 38,000 ha of irrigated perimeters. It provides permanent employment for more than 40,000 employees as a wage and family workforce. After the arrival of the GMP and the granting of subsidies to farmers about 12 years ago, this region has shown a strong expansion of the irrigated area with the proliferation of new entrepreneurial type farms, large and specialized in market gardening and arboriculture, driven by subsidies allocated by the country's Agricultural Development Fund (FDA).

The main oueds (watercourses) of the study area are Oued Chichaoua and Oued Assif ElMal. They originate from the Western High Atlas and are ephemeral. The discharge is conditioned by the presence of rain events in the high altitudes of the Western High Atlas (Boudhar et al. 2009). The climate in the study area is semi-arid continental. It rains especially in the autumn and winter periods and the precipitation is very irregular in time and space. Groundwater flows into the Plioquaternary formations (clay sands, lacustrine limestones, pebbles and sandy-clay matrix gravels). These formations are relatively thick to the east of the Assoufid flexure, but decrease in thickness west of this flexure, in the plain of Mejjate. These formations are deposited on the bedrock of Miocene marls or of Visean schists (Chouikri et al. 2016).

Administratively, the study area falls under the provinces of Chichaoua and El Haouz and has two municipalities (Chichaoua and Imintanout) and 29 rural communes. The number of residents in the study area was obtained from the RGPH (Recensement Général de la Population et de l'Habitat (General Population and Housing Census)). A total population of 320,749 inhabitants was recorded in 2014 with approximately 15% of the population lived in the urban area. Between 2004 and 2014, the average annual growth rate was 0.86%, even lower than that of the two previous decades.

As the population data are updated every 10 years in Morocco, the number of residents was put into the WEAP model with 2014 as the base year. After 2014, the annual growth rate recorded in each municipality and commune for the 2004–2014 period was applied.

The data of land use types in the study were derived from the work of Hadri et al. (2021). The land use types were consolidated into nine classes, namely urban area (housing structures), plantations, irrigated agriculture, rainfed agriculture (non-irrigated agriculture), forest, shrub (combination of sparse grassland and shrubby vegetation), grassland (contains areas of apparently non-productive arable fields), bare soil/rock (areas without vegetation) and water body, as shown in Figure 2. Within irrigated areas, we can distinguish:

• PMH (small and medium hydraulic) areas which are irrigated by gravity from surface water. Water is taken by the diversion from oueds through water intakes (Seguias) located on the banks of the oued. The most important PMH in the study area are (i) PMH Douirane, located on the oued of Seksaoua (tributary of Chichaoua oued), northeast of the Imintanout city. It starts from the Illoudjane village and then runs along the Seksaoua oued on both banks. It spreads over a total area of approximately 5,460 ha. This perimeter contains two sub-perimeters called Douirane and Douirane Crue (where flood water is used to irrigate farmers plots), (ii) PMH Tagnaouite located just downstream of the PMH Douirane and made up of Tagnoauite and Tagnaouite Crue sub-perimeters. It covers an area of about 2,176 ha, (iii) PMH Chichaoua which extends along the Chichaoua oued, on an area of approximately 3,706 ha and (iv) PMH located on Assif AlMal oued with an area of 4,733 ha.

• The private irrigation (IP) perimeters are located in the Mejjate plain. These perimeters are developed with the arrival of the GMP and the granting of subsidies to farmers. They are stretched over an area of about 21,000 ha and have only groundwater for irrigation. The irrigation method used in these perimeters is the drip (El Faïz 1999; Simonneaux et al. 2009; Ruf 2017; Ruf & Kleiche-Dray 2018)

According to Agricultural Department census and land use analysis carried out as a part of the present work, nine crop categories were identified as the most practised by local farmers. These crops are wheat and barley, olive, citrus, almond tree, pomegranate, apricot, vegetables (tomato and potato), fava bean, alfalfa, sweet melon and watermelon (Figure 3).

Precipitation is the key input data for the net precipitation calculation, for the evaluation of the natural replenishment of aquifers and for the estimation of the irrigation water requirements (IWRs). Daily datasets were obtained from the Tensift Hydraulic basin Agency (ABHT), which were recorded from 1971, 1970 and 1989 at the stations of Chichaoua, Abadla, Iloujdane and Sidi Bouathmane, respectively (Table 1). The average annual rainfall at these stations ranges from 173 to 355 mm. The wet season is from November to April, with 81% of the annual rainfall. However, the dry season extends from May to October and represents only 19% of the annual rainfall amount.

For the future projections, five simulations of RCMs derived from the Med-CORDEX (Mediterranean Coordinated Regional Climate Downscaling Experiment) initiative (Ruti et al. 2016) were used (Table 2). Med-CORDEX provides a high-resolution regional climate model (with a horizontal resolution of 50 km) and was evaluated by many authors in the High Atlas Mountain of Marrakech (Tramblay et al. 2013; Marchane et al. 2017; Zkhiri et al. 2019). The simulations are in daily time step, and they are converted to monthly time step to better study different scenarios of water management over the study area. The complex topography over the High Atlas of Marrakech can give a discrepancy between the models. However, RCMs that modulate the climate signal in finer scales, unlike GCM, are able to capture the effects of local forcing and reproduce this complex influence of topography on climate (Giorgi et al. 2009; Filahi et al. 2017). In North Africa, many authors pointed out the need for climate projections at the regional scale with a fine resolution for climate change impact studies, due to the discrepancy between models related in particular to the complex orography of the domain (Gleckler et al. 2008; Kotlarski et al. 2014; Romera et al. 2015) In the present study, the five RCMs, derived from the Med-CORDEX project, have similar skills in reproducing the climate characteristics of the study area.

The mean value of the different models was considered. Indeed, the five RCMs are considered together in a multi-model ensemble mean which is consistent with the aproaches of other multi-model studies (Sheffield et al. 2013; Sillmann et al. 2013; Dong et al. 2015; Marchane et al. 2017). The ensemble mean generally outperforms individual models because part of the systematic errors of the individual models is offset in the multi-model mean (Gleckler et al. 2008; Wilby & Dessai 2010; Filahi et al. 2017; Tramblay et al. 2018). The projections are concerning one period (2020–2050) under RCP4.5 and RCP8.5 scenarios.

Under the absence of the observed temperature, we referred to the study of Marchane et al. (2017) where they found that the historical period reproduces the shape of the observed evapotranspiration in the High Atlas of Marrakech, for this reason; in this study we compared the projected changes from the five climatic models of the historical period (1970–2005).

The WEAP model requires a lot of data and parameters. The use of projected scenarios of precipitation and discharge in the area needs observed data of discharge and precipitation to be evaluated. In our case, the observed discharge data are not available, which lead us to follow the approach proposed by Tramblay et al. (2018) that consists of using the historical period of the climatic models as reference and to calculate the net precipitation. This approach is a solution to assess the future impacts of temperature and precipitations on the runoff under climate change scenarios RCP4.5 and RCP8.5. Results of this method were integrated directly as inputs to the WEAP model.

The WEAP model is a decision support system (DSS) that was developed by the Stockholm Environment Institute (SEI) to simulate and assess water resources management and planning under different scenarios. Several (i) water demand systems such as drinking water, irrigation systems, livestock watering, industrial demand, hydropower energy production, environmental flow, etc. and (ii) water supply systems such as reservoirs, river flow, groundwater, wastewater reuse, rainfall, etc. could be modelled in WEAP.

WEAP is a robust tool that allows the reproduction of current water management practices and the simulation of different ones. In different zones across the world, WEAP has been largely used to assess the water resources management and planning systems and practices. Allani et al. (2019) used WEAP to evaluate the impact of future climate change on water supply and irrigation demand in the Nebhana dam system in Tunisia. Adgolign et al. (2016) have applied WEAP to simulate surface water resources allocation in Didessa basin in West Ethiopia. This software was also employed by Asghar et al. (2019) and Hassan et al. (2019) to establish an integrated modelling system for assessment of water demand and supply under socio-economic and IPCC climate change scenarios in Indus basin in Pakistan. In Greece, WEAP was employed to present a holistic hydro-economic framework for sustainable water resources management in data-scarce areas. Salomón-Sirolesi & Farinós-Dasí (2019) have created, using WEAP, a water governance model aimed at supply-demand management for irrigation and land development in the Mendoza river basin in Argentina. WEAP was also used by several researchers in Morocco to explore the impact of climate change and land use change on water supply, demand and management (Johannsen et al. 2016) and to implement water decision support systems (Le Page et al. 2012).

In our study, the WEAP software was employed as a key tool to assess the water balance under climate change impacts. The schematic of the Chichaoua-Mejjate system representing all catchments and establishing water supply and demand nodes was constructed as shown in Figure 4.

The simulations were run on a monthly basis. The time frame was set up from 2010 to 2050. 2010 was considered as the current account/baseline year and the simulation was carried out from 2011 to 2050 as reference scenario. The data introduced to the model from 2010 to 2019 are mostly the observed data. All socio-economic and climatic scenarios were developed with water demand and supply for the 2021–2050 period.

The current and future water requirements were assessed for different sites' demands in the study area. Water demand analysis was performed for all the sectors using the disaggregated based approach in the WEAP model. The major demand sites in the study area are irrigation and domestic water demands.

The AquaCrop modeling tool was employed to calculate crop water needs under various climate conditions and in the different irrigation perimeters for the 2011–2050 period. Developed by Water Unit at the FAO as a revised version of CropWat, AquaCrop is a useful tool to predict crop response and quantity of water required over time under changes of climate conditions (RAES et al. 2012). According to Araya et al. (2010), the AquaCrop model relates its soil–crop–atmosphere components through its soil and its water balance. This model uses six input files for simulation: climate file (minimum and maximum air temperature, ET₀, effective rainfall (the portion of total rainfall which is useful directly and/or indirectly for crop growing at the site where it falls (Dastane 1974)) and CO₂), crop file (time to emergence, maximum canopy cover, start of senescence, and maturity), soil file, management file, irrigation file, and initial soil water conditions.

Adjustments were introduced to the model by changing some AquaCrop pre-set crops to reflect local practices and conditions. The generated model outputs for the current situation were compared to the known crop water demand commonly used by the Agriculture Department. This comparison allowed us to calibrate and improve AquaCrop input and parameters.

AquaCrop has already been used in many impact projection studies (Abedinpour et al. 2014; Bird et al. 2016; Akumaga et al. 2018). In the same region, where our study area is located, Toumi et al. (2016) showed that for winter wheat, AquaCrop simulates canopy cover, actual evapotranspiration, total soil water content and grain yield reasonably well over two growing seasons. He argued that the AquaCrop tool could be considered as a potentially useful one for planning irrigation schedules in the arid and semi-arid regions.

Domestic water demands are driven by population. There are two major urban areas which are deriving water from reservoirs and from groundwater for domestic use: Chichaoua and Imintanoute cities. The domestic demand sites include these two cities and villages belonging to 29 rural communes.

The official daily water demand per capita adopted by ONEE (National Office of water and Electricity) and the population growth rate for each commune and city were used to evaluate the domestic water demand, which covers domestic, industrial and administrative buildings demands. For the rural villages, ONEE adopts 50 l per person per day which is bigger than the minimum of 20 l per person per day required by the WHO to ensure that basic needs are met (WHO 2003).

Priorities of water allocations are user-defined for the WEAP model. They range from 1 to 99 depending on demand and supply preferences. In general, the highest priority was given to domestic water demand followed by irrigation demand. For the supply preferences, priorities are given to surface water in order to preserve groundwater.

The supply sources of the existing and planned demands in Chichaoua-Mejjate region are fed from surface water resources (Chichaoua and Assif AlMal wadies) and from groundwater. Currently, there is one reservoir called Abu AlAbbas Essebti, located in Assif AlMal Wadi, as a supply source for domestic water and irrigation in the study area. The second reservoir Boulaouane, located in Chichaoua Wadi, is under construction and is planned to be operational by 2026 (Table 4).

A volume elevation curve was used for calculating evaporation losses depending on reservoirs' storages. A decrease in storage capacity of the reservoirs by 0.51 Mm³/year and by 0.2 Mm³/year due to siltation phenomena was simulated in the model for Abu AlAbbas Essebti reservoir and Boulaouane reservoir, respectively. This sedimentation rate was already observed in the Abu AlAbbas Essebti reservoir over the last 6 years and witnessed a lack of soil protection actions that could delay the erosion processes upstream of the reservoir.

Groundwater is represented with a simple model called Reservoir, which is integrated into the WEAP model. The concept of this model is to consider that the aquifer system works as a reservoir characterized by inputs and outputs, which are quantified to extract the variation of the reservoirs over a given period. The advantage of this model is that it is limited to variations in inputs and outputs without worrying about the intrinsic parameters of the reservoir.

Groundwater storage is filled by river bed infiltration, by the excess of the irrigation supply, by the underground recharge from the mountains area and by the rainfall infiltration. The losses to groundwater in the irrigation canals (Seguias) for surface irrigation were also considered in the WEAP model.

All data related to those terms are linked and fed from the precipitations, discharge and irrigation data contained in the model. The amount of water entering from the mountains area is calculated as 9% of the precipitation falling in this area (ABHT 2016). The rates of rainfall infiltration and river bed infiltration are taken from ABHT (2016).

The main crop characteristics of the study area were obtained from the FAO 56 database (Allen et al. 1998), from field survey and from other researches carried out in the same region (Er-Raki et al. 2007, 2009, 2010). These characteristics (crop coefficients, date of planting or bud burst, duration of each development stage, crop variety, etc.) are used by Aquacrop to model crop water demand for each culture.

For each crop, the average required CGDD for the crop cycle was obtained (Figure 5).

The WEAP model permits multiple scenario analyses (Yates et al. 2005). The baseline scenario and several management and climate scenarios were run according to their specific input data as follows:

• – Scenario0 (Sc0): Business as usual: Boulaoune reservoir is active, crop superficies are maintained and water needs are covered from wadies intakes, groundwater, Boualaoune and Abu AlAbbas Sebti Reservoirs.

• – Scenario1 (Sc1) Reference scenario: Business as usual, Boulaoune reservoir is not active, crops areas are maintained and water needs are covered from wadies intakes, groundwater and Abu AlAbbas Sebti Reservoir.

• – Scenario2 (Sc2): Increasing (i) plot efficiency by adopting drip irrigation in 50% of PMH area and (ii) canals of irrigation network efficiency by concreting 70% of traditional Seguias.

• – Scenario3 (Sc3): Adopting alternative less water-demanding crops. 50% of watermelon and sweet melon is replaced by winter wheat.

• – Scenario4 (Sc4): Limiting the groundwater available for abstraction from aquifer to the renewable volume thereof.

All these scenarios were processed for both RCP4.5 and RCP8.5 climatic models. Consequently, a total of 10 scenarios were simulated.

Figure 6 shows the projected monthly evapotranspiration and precipitation from the ensemble of the five models under two RCP scenarios. The result shows that the observed evapotranspiration is overestimated when using the Hargreives–Samani equation contrary to Oudin's equation which is in accordance with Marchane et al. (2017), who have found that Oudin's equation provides lower values of evapotranspiration than the Hargreives–Samani formula. However, the projected values show an increase of evapotranspiration between 1989–2005 and 2020–2050. The increase is more pronounced using Oudin's formula during winter with +20% under the RCP4.5 scenario and +26% under the RCP8.5 scenario. The Hargreives–Samani formula shows an increase of +14% under the RCP4.5 scenario and +20% under the RCP8.5 scenario. This indicates that in the future, the winter season will be less cold than the actual period. The projected precipitation shows a pronounced decrease in autumn season around −36% under the RCP4.5 scenario and −32% under the RCP8.5 scenario. This increase in precipitation amount is due to the increase of precipitation under the scenario RCP8.5 during autumn season which can be related to the increase of flood events (El Khalki et al. 2021).

The impacts of these results on water resources are shown in Figure 7 which indicates the projected changes in net precipitation during the period 2020–2050. As indicated above, the impact of the underestimation of evapotranspiration by Oudin's formula is figured out in this result with an overestimation of net precipitation compared to the Hargreives–Samani result. The projected results using the two formulas show an important decrease of net precipitation with an average of −43.44 and −36.26% under the RCP4.5 scenario and −55.17 and −50.55% under the RCP8.5 scenario using Oudin and Hargreives–Samani formulas, respectively. The major decrease will be during autumn and winter with a maximum decrease of −55.78 and −58% under the RCP4.5 scenario and −51.02 and −52% under the RCP8.5 scenario using Hargreives–Samani and Oudin formulas, respectively. These results are in coherence with Tramblay et al. (2018) and (El Khalki et al. 2021). However, according to Tramblay et al. (2018), in the basins that are located in the same longitude, the net precipitation using Hargreaves–Samani will decrease with around −34% under the RCP4.5 scenario and −41% under the RCP8.5 scenario.

According to the crop cycle CGDD and due to the rise in air temperature expected under both RCP scenarios, the crop's cycle length (CCL) is expected to shorten. A decrease of this cycle is observed for all studied crops ranging for the scenario RCP4.5 from −1 to −37% for vegetables and citrus, respectively. The results show a systematic reduction of CCL for both RCP4.5 and RCP8.5 scenarios at 2050. This reduction is more pronounced for the olive and citrus crops and the wheat CCL, for instance, would decrease by −7% (Table 5). Indeed, the shorter CCL is explained by premature loss of leaf chlorophyll due to heat stress that accelerates mobilization of photosynthate to newer leaves and thus triggers early maturity of the whole plant (Leshem et al. 1986; Moore et al. 2021).

These findings are close to the work of Bouras et al. (2019), who have stressed that, as a consequence of increases in air temperature, the length of the wheat season in the Tensift region would be reduced by 10% under the RCP4.5 scenario at 2050. The same trends but with stronger reductions are observed by this author for the other scenarios and horizons. For the more extreme scenarios at the end of the century (RCP8.5 2090), the decrease of the length of the crop season could reach 32% (50 days). Similar results were obtained by Saadi et al. (2015) who have shown a decrease of 9.6 and 6% in tomato and wheat length cycle, respectively, under the A1B emission scenario in Mediterranean regions.

Our results are also consistent with Pathak & Stoddard (2018), who have pointed out a significant decrease in the number of days between transplanting and maturity of tomato in California, with an expected harvest of 2–3 weeks earlier than normal under current conditions. Moreover, Chattaraj et al. (2014) projected a shortening of the wheat growing season by 5–15% in India due to increase of future temperature.

Furthermore, it is worth mentioning that the phenological stages of crops are impacted differently even if the total decrease of CCL is similar for all sowing dates, which would have repercussions on the temporal pattern of irrigation, with a possible shift of the season of the crops requirement water peak (Bouras et al. 2019). Interestingly, the choice of crop varieties should take into consideration this projected shortening of the CCL in order to have an agriculture more resilient to climate change in the future (Boote et al. 2011).

Table 6 shows the mean crop water demand (ET_c), rainfall, and IWD for each crop over the 2021–2050 period under RCP4.5 and RCP8.5 emission scenarios.

For the actual situation, the crop water demand ranges from 2,414 m³/ha for tomato to 10,155 m³/ha for almond. In the future, a decrease in the crop water demand is foreseen for all crops except sweet melon and vegetables. This decrease is expected coming up on average to 4% for olive and wheat under RCP4.5. Fava bean is expected to have the highest decrease reaching 13% under the same emission scenario. Under the RCP8.5 scenario, this decrease is more pronounced for most crops and would attain −5, −4, −8 and −15% for wheat, citrus, pomegranate and alfalfa, respectively.

Contrary to what is a common presupposition, the crop water demand is expected to be lower due to climate change. The main reason for this decrease is the shortening of crops' season length following the increased temperatures expected under the two RCP emission scenarios. Despite the annual increase of the referential evapotranspiration, the changes in crops' cycle patterns will induce a decrease of the crop water demand. This finding is in agreement with the conclusions of Allani et al. (2019) in Tunisia who reported a decrease in ET_c for many crops (olive, citrus, apricot, wheat, barley, summer and winter vegetables). According to this author, the most affected crops were winter vegetables, mainly fava bean and green peas, as well as olive with a decrease of 16.9% under the RCP4.5 scenario and 18.9% under the RCP8.5 scenario.

Likewise, the work of Bouras et al. (2019), which aimed to quantify the impact of climate change on the grain yields of irrigated cereals and their water requirements in the Tensift region of Morocco (the region where our study area is located), found that the crop water requirements are expected to decrease by 13–42%, mainly in response to the reduction of the growth cycle. This study reported that the water requirement peak is coming 2 months earlier than under current conditions (Bouras et al. 2019). Under optimal conditions, Saadi et al. (2015) suggest a decrease of the net water demand of tomato and winter wheat by 11% at the 2050 horizon in the southern Mediterranean region. Further studies demonstrated the same tendency for boro rice in Bangladesh (Acharjee et al. 2017), for maize in North China (Chen et al. 2010) and for citrus (Fares et al. 2017) across major producing areas in the world.

Nevertheless, even if the crop water demand could decrease following the shortening of crops' season length under climate change, high temperatures may affect plant productivity and crop yields (Hatfield & Prueger 2015), which can disturb farmers' incomes and food safety. Indeed, faster phenological development of crops results in shorter productive duration, which would significantly lower yield potential. More precisely, extreme high temperature, when it occurs during the reproductive stage, will affect pollen viability, fertilization, and grain or fruit formation (Hatfield et al. 2011).

It is also interesting to note that other studies suggest that the reduction of the crop growth cycle could not compensate for the rise of ET₀ associated with the rising temperatures, and consequently the crop water requirements will increase under climate change effect (Lovelli et al. 2010; Dettori et al. 2017).

Even with the decrease of water crop demand, the pattern of future IWD varies from one crop to another during the period of 2021–2050 (Table 6).

In the 2050 horizon, the majority of crops will undergo an increase in IWD, under the RCP4.5 scenario, varying from 1% (almond and apricot) to 15% (olive). An important decrease in IWD of fava bean is expected for the scenario RCP4.5 due to the important decrease in the water demand of this crop. For wheat, almond tree, vegetables and sweet melon, the IWD will maintain the same rate as the 2010–2017 situation under RCP8.5. The olive IWD will witness the highest increase compared to the actual situation by registering an increase of 15 and 22% under RCP4.5 and RCP8.5 scenarios, respectively. Likewise, the citrus, which is one of the most important crops in the study area, will record an important relative increase in IWD, unlike the case of other fruit trees (pomegranate, almond, and apricot).

Compared to the actual situation, the change of the future IWD remains low under different climate change scenarios. These results could be explained by the behavior of climate change impacts on precipitation that differ from one season to another, coupled with the length and period of the growing seasons that present a contrast between crops. The future IWD changes are impacted by the uncertainty of the projected rainfall amount and distribution. These results are in agreement with the findings of Acharjee et al. (2017) who found that, for boro rice, despite the reduction of total crop water demand, the net irrigation demand can reduce or increase depending on changes in precipitation. The same future variation behaviors were suggested by Allani et al. (2019), in the Nebhana dam system in Tunisia, who has studied more than 15 crops' IWR change. For citrus, Fares et al. (2017) indicated that predicted irrigation requirements show significant spatio-temporal variations across major producing areas in the world.

The use of WEAP for the simulation of the 10 development scenarios by 2050 in the study area has made it possible to analyse the extent of satisfying the water demand of the different demand sites, to establish the future evolution of the total water demand, to evaluate the spatial and temporal variations in water shortages and to monitor the variations in water storage in the reservoirs and in the aquifer.

The mean annual current water demand in the study area is 180 Mm³ /year. The agricultural sector acts as the major water user, which consumes 92% of total water use. The domestic sector thus is consuming only 8% or an equivalent of 14 Mm³/year (Figure 8). The IP requires much more water than other irrigated perimeters. Indeed, water abstraction is generally free for farmers in this region. A negligible minority pay rudimentary fees to the Hydraulic Basin Agency of Tensift (ABHT) against the pumping of groundwater. Therefore, many farmers take water for granted and do not use it effectively. On the other hand, some farmer practices tend to provide water quantities much greater than the crop's water demand (over-irrigation).

By 2050, water demand is projected to rise for scenario0, scenario1 and scenario4 due to population growth and to the slight increase of IWD. Based on these scenarios, the increase of the annual water demand would reach 4.9 and 10.1 Mm³ under RCP4.5 and RCP8.5 scenarios, respectively (Figure 8). The drinking water demand would increase by about 42% until 2050. The agricultural sector will still be the largest user in the Chichaoua-Mejjate area, capturing 90% of the total water consumption. The scenarios2 and 3, as they envisage actions of reducing IWD, show a decrease in the annual water demand that reach −8% for scenario3 (RCP4.5) and −4% for scenario2 (RCP4.5). Consequently, the water management actions planned in these two scenarios could compensate for the increase in total water demand projected by 2050.

For the scenarios Sc0–Sc3 in both RCP4.5 and RCP8.5 scenarios, the coverage of drinking water demand remains almost unchanged (demand will be satisfied at 100% in 40% of the years and at more than 85% in 60% of the years of the studied period) because the scenarios mainly target the management of agricultural water. This high satisfaction rate is due mostly to the priority given to drinking water in the model. Contrariwise, the coverage of drinking water demand is significantly impacted when the available groundwater is limited at the level of the renewable volume (scenario4). In this case, drinking water demand could not be met within more than 38% of the simulated years (Figure 9). However, it is worth noting that the installation of the two dams will relieve the DWS which relied solely on groundwater resources.

The results for meeting IWRs can be found in Figure 10. The results corresponding to the Sc0, Sc2, Sc3 and Sc4 are compared to the business-as-usual scenario (Sc1). There is unmet demand in all demand sites supplied from surface water resources, although there are differences in the level of non-satisfaction. Demands supplied from groundwater or a mixture of surface and groundwater will be fully met under the four scenarios Sc0–Sc3.

The following results are noted for the Sc0, Sc2, Sc3 and Sc4 scenarios compared to scenario1.

Scenario0 (Impact of the construction of Boulaouane Reservoir):

• – The annual demand coverage upstream the Boulaouane reservoir would be improved by about 54% (7.4 Mm³) under the RCP4.5 scenario.

• – The amount of the water requirement met would be clearly improved by about 185 Mm³ over the 2026 (planned year of construction of the dam)–2050 period, for the Tagnaouite, Tagnaouite Crue, Douirane and Douirane Crue irrigated perimeters. This might occur due to the regulatory role that the reservoir will play and the attempts to adequately serve all demands.

• – The reservoir construction would induce some losses from the system compared to scenario1 due to water evaporation. In addition, DWS to the Imintanoute city will be covered from the reservoir in this scenario instead of groundwater used in scenario1.

• – The siltation phenomena will reduce the storage capacity of the reservoir by 0.2 Mm³/year and will therefore lower its role as flow regulator.

Scenario2:

• – The increasing of plot and canals of irrigation efficiency by adopting drip irrigation and concreting 70% of traditional Seguias is still not enough to cover completely the water demand of irrigated perimeters.

• – The shortage of water to meet the water demand would be reduced by 73 and 74% in the Douirane perimeters, under RCP4.5 and RCP8.5 scenarios, respectively.

• – In the Tagnaouite perimeters, the annual unmet demand in the RCP4.5 scenario would reach about 1.2 Mm³, against an unmet demand of more than 3.9 Mm³ under the scenario1.

• – For the RCP4.5 scenario, the annual water demand would decrease significantly from 18 to 14 Mm³ and from 23.4 to 19 Mm³ in the Assif AlMal and Chichaoua perimeters, respectively, over the 2021–2050 period.

• – Since the irrigation efficiency is improved in this scenario, less water percolates to the aquifer in the plots and Seguias, which will reduce the annual renewable volume of the aquifer. However, the aquifer storage in this scenario remains better compared to scenario1, because of its lower water demand.

Scenario3:

• – By replacing 50% of watermelon and sweet melon area and 50% of alfalfa area by winter wheat and by sorghum, respectively, about 18 and 16% of the water demand of IP would be saved, under RCP4.5 and RCP8.5 scenarios, respectively.

• – This action would reduce the pressure on groundwater by about 12.7 Mm³/year for the RCP4.5 and 11.8 Mm³/year for the RCP8.5.

Scenario4:

• – By limiting the maximum withdrawal from aquifer to the renewable volume, the ratio of requirement water met in the IP decrease significantly and fell below 20 and 16% over 60% of simulated years under the two emissions scenarios RCP4.5 and RCP8.5 scenarios, respectively.

• – The total annual use of groundwater largely exceeds the annual renewable volume of the aquifer. On average, only 20% of water consumed is replenished each year. Therefore, the overexploitation of groundwater attains 500%, which testifies to an exorbitant imbalance.

In general, almost similar impacts would be recorded for the two RCP scenarios, which can be explained by the short time horizon (2050) chosen for this study and by the small differences in precipitations predicted by these two climatic scenarios in the study area. A similar finding was reported by Wang et al. (2016) in China where IWD would not differ much between RCP4.5 and RCP8.5 scenarios.

In the scenario1 (business as usual), the study area would witness a growing unmet demand in the future period (2021–2050) compared to baseline period (2010–2017). Despite the future projected decrease of crops' water demand and because of the climate change impacts on water supply, the annual unmet demand would increase at the 2050 horizon by about 13 and 22% under RCP4.5 and RCP8.5 scenarios, respectively (Figure 11). The same trend was observed by Allani et al. (2019) in Tunisia where the Nebhana dam system is expected to cover only 25% of the water demand for the 2061–2080 period. In the Tensift region, the unmet demand of the Rherhaya basin could attain 70% under the B2 scenario if no mitigation measures are adopted (Rochdane et al. 2012).

As reported by Le Page et al. (2020), even though water demand is slightly impacted by climate change, this latter might reduce the water offer as a result of runoff decreasing followed by a reduction of water available for irrigation (Le Page et al. 2020). Furthermore, groundwater offer would also be impacted due to the decrease of rainfall and runoff predicted by the two emission scenarios RCP4.5 and RCP8.5. Indeed, aquifer storage, which relies on the amount of water free for infiltration in the wadies beds and in the whole plain area, will be affected (Johannsen et al. 2016).

However, the unmet demand would experience a decrease compared to the reference period (2010–2017) for the three scenarios Sc0, Sc2 and Sc3. The construction of the Boulaouane dam (Sc0) will reduce the average annual uncovered demand by 14% (RCP4.5) and by 3% (RCP8.5) in the 2050 horizon. Improving irrigation efficiency (Sc2) will reduce water shortage by 41 and 39% for the two climate change scenarios. On the other hand, as the adoption of crops with high water consumption (such as watermelon and citrus) in this arid zone remains a serious issue, the change in crop types (Sc3) will lead to a net reduction in unmet demand of 17 and 6% for RCP4.5 and RCP8.5 scenarios, respectively. Therefore, it is possible to conclude that the decision to intensify efforts towards improving irrigation efficiency and using less water-consuming crops could be efficient solutions to improve water management in the Chichaoua-Mejjate area.

The monthly distribution of the unmet demand shows that the shortages would be more pronounced from February to June for all simulated scenarios and more intensified under the RCP8.5 scenario. The monthly unmet demand could reach 22 Mm³ in April for the scenario4 under the RCP4.5 scenario. For this scenario, the unmet demand decreases quickly from May to December contrary to the other scenarios where the uncovered demand curve flattens between May and September and between September and December (Figure 12).

Interestingly, the reduction of the various losses at the distribution system level (Seguias) and at the plot level (deep percolation) has the greatest impact on the improvement of the water demand met. The efficiency improvement (scenario2) has more benefits to the reduction of unmet water demand than the scenario0 (dam construction) and scenario3 (replacing crops). According to Le Page et al. (2020), irrigation water requirement can vary from single to more than quadruple between the ideal best and the worst efficiency configuration. In fact, the construction of the dam will not by itself reduce the water deficit in the downstream area but must be accompanied by measures aiming to improve the irrigation efficiency and modernize agriculture. In addition, it is interesting to note that the storage capacity of the reservoirs is endangered by the important rate of the upstream soil erosion (Klose 2009) which significantly reduce their service life.

For the four scenarios Sc0–Sc3, water deficit of the aquifer would fluctuate between 2.2 and 2.7 MMm³ over the 2021–2050 period under the RCP4.5 scenario, and between 2.4 and 2.9 MMm³ under the RCP8.5 scenario (Figures 13 and 14). The groundwater would show a drastic declining trend that averages to 92 Mm³/year in scenario0 (for example) over the 2021–2050 period, which supports the findings of many authors who have studied the groundwater overexploitation within the whole Haouz Mejjate aquifer (Le Page et al. 2012; Jarlan et al. 2015).

In the IP perimeters, the extension of market gardening and citrus fruits that requires a lot of water is inadequate to the aquifer potential and constitutes a threat to the groundwater resources sustainability. Furthermore, in periods of surface water shortages, farmers use groundwater intensively to compensate the missing river inflow and the stopping supplies from reservoirs (Heidecke et al. 2010). Therefore, the Haouz Mejjate aquifer is experiencing a continuous intensive use that does not give the aquifer time to recover. Taking into account the mean permeability of the aquifer, the Mejjate plain would experience significant declines of about 1.97 m/year, which is in agreement with Hadri et al. (2021) who analysed the piezometric levels monitoring in the same plain and found a decline of about 1.8 m/year over the last 10 years.

Faced with this worrying situation, it is essential to maximize water-use efficiency and promote demand-management instead of supply-management. Indeed, to address the demand-supply imbalance, the quantity of water demanded has to be reduced considerably (Hellegers et al. 2013) by adopting better irrigation techniques and by cultivating less water-consuming crops. The financial subsidies of Agricultural ministry are important to encourage farmers to consider these solutions, which are not motivating farmers because of their financial risk (Ghallabi et al. 2011). Likewise, other agricultural practices that require less investment should be considered by farmers, like cropping, cultivation structure adjustments and rainwater harvesting.

By combining different tools, models, and data, and by considering the climate change impacts on both water demand and water supply, this study provides a consistent methodology to establish integrated water management in a data-scarce basin. The WEAP software is used as the main tool of a DSS to simulate the current and the projected future water balance. Following a quantitative approach, the projected crop water demand and IWD were analysed taking into account the climate change impacts on precipitation, reference evapotranspiration and crop cycle lengths.

The results are very useful for stakeholders and give many indicators that significantly help with the decision-making in the Chichaoua-Mejjate area and in other similar semi-arid basins in Morocco. They show a decrease in net precipitation and thus in surface water availability under both climate change scenarios, RCP4.5 and RCP8.5. Therefore, groundwater recharge will also be impacted due to its interaction with the surface water. The water demand, for most studied crops, is expected to be lower due to climate change. The main reason for this decrease is the shortening of crops' season length following the increased temperatures expected under the two RCP emission scenarios. However, the behaviour of future IWD would vary from one crop to another due to the change in the future rain pattern.

According to different simulated water management scenarios, the unmet water demand in the study area is likely to increase significantly, especially when no mitigation measures are adopted. For the ‘business as usual scenario’, compared to the baseline period (2010–2017), the annual unmet demand would increase at the 2050 horizon by about 13 and 22% under RCP4.5 and RCP8.5 scenarios, respectively. These findings are of interest to decision-makers to define optimal management scenarios and to provide adequate answers to questions such as: which crops to adopt and encourage in the study area; how to identify adaptation measures and what would be their potential impacts; what is the impact of changing the water allocations priorities; how to deal with the unmet water demand; how would climate change exacerbate the water balance; what data are needed and with what precision, etc.

Overall, traditional approaches no longer allow us to meet growing needs and to preserve water resources. For this purpose, several measures would be necessary and imperative to implement a sustainable development of the agricultural sector: the choice of adapted crop types, enhancing irrigated efficiency, water-saving irrigation techniques, and the prevention of the reservoir sedimentation by implementing upstream soil protection measures. Moreover, more encouragement of public participation and local awareness of water issues are becoming crucial and unmissable to establish a sustainable development of the agricultural sector.

The integrated modelling approach of the DSS presented in this work is scalable and replicable in other arid and semi-arid basins and would provide a useful tool to decision-makers and water planners for achieving successful water management strategies and for designing and prioritizing mitigation actions and interventions. Further research could be conducted by developing a groundwater hydrodynamic model connected to the current DSS and by simulating the potential impacts and limitations of other adaptation measures scenarios.

All relevant data are included in the paper or its Supplementary Information.