The experiment was conducted in Taroudant, located in the Souss-Massa region of Morocco, about 80 km east of Agadir city (Figure 1). The Souss-Massa basin spans approximately 27,800 km² and is bordered to the north by the Tensift basin, to the east and south by the Draa basin, and to the west by a 200 km stretch of the Atlantic coast. In the Souss-Massa basin, the water balance indicates strong pressure on groundwater resources, with an exploitable potential estimated at 369 Mm³/year compared to a mobilized volume reaching 646 Mm³/year, resulting in an overexploitation of 277 Mm³/year. This situation has led to a decline of approximately 30 meters in piezometric levels over the past thirty years, based on data covering the period 1933-2015 (Hssaisoune et al., 2020). Agriculture accounts for 93% of total water consumption, leaving only 7% for domestic and industrial uses (ABHSM, 2007). These variations are driven not only by climate variability but also by human activities, highlighting the complex interplay between natural and anthropogenic factors shaping water availability in the region (Bouchaou et al., 2011; Abahous et al., 2017; Malki et al., 2017; Marieme et al., 2017; Attar et al., 2022).

The Souss-Massa region is a vital agricultural zone in Morocco, known for its significant production of vegetables, cereals, and fresh fruit for both domestic consumption and export. In this region, maize is predominantly cultivated as a fodder crop for livestock feeding. The study area features fine clay loam soil and is characterized by a semi-arid Mediterranean climate with distinct dry conditions. Annual rainfall ranges from 250 mm in the plains to 600 mm in the mountainous areas, with a rainy season occurring from November to March (Mansir et al., 2021). The region benefits from over 300 days of sunshine per year, with average temperatures ranging from 14 to 22°C in January and 23 to 38°C in July. However, due to limited water resources and considerable groundwater overexploitation, water levels have declined by 0.5-2.5 meters annually over the past four decades (Choukr-Allah et al., 2017; Hssaisoune et al., 2017).

The AquaCrop model was calibrated and validated using seventeen field data from Al Moutmir database (https://www.almoutmir.ma/fr). Although the dataset spans three growing seasons (2022-2024), each field was monitored during only one season. The calibration fields are indicated as C1-C7, while the validation fields are identified as V1-V10. All fields were cultivated with long-cycle maize varieties sown in March and harvested in July. Soil water content at field capacity (FC) and permanent wilting point (PWP) was measured using a Pressure Plate Extractor following the methodology of (Klute, 1986). Three undisturbed soil samples were saturated by capillarity with water up to two-thirds of the ring height and, after saturation, subjected to matric potentials of 0 kPa (saturation), −33 kPa (FC), and −1,500 kPa (PWP). The hydraulic conductivity was determined in the field using the method suggested by (Reynolds and Elrick, 1991). For fields where direct measurements of hydraulic properties were not available, soil water retention parameters were estimated using pedotransfer functions (Saxton and Rawls, 2006) based on measured soil texture. A summary table of soil physical and hydraulic properties for each field are provided in the Supplementary Material (Supplementary Table S1). Soil fertility was assessed using the following soil chemical properties: pH, organic matter, P2O5, K2O, and electrical conductivity (EC). Soil pH was determined using a pH probe with a 1:5 soil-to-deionized water ratio, as described by (Bouray et al., 2024). Electrical conductivity (EC) was determined in a 1:5 soil-to-deionized water suspension using an EC meter (SevenCompact, Mettler Toledo, USA). Exchangeable K₂O was extracted with 1 M ammonium acetate and analyzed by Atomic Absorption Spectroscopy (Agilent 200 Series AA, Santa Clara, USA) following the NF X31–108 method. Available phosphorus (Olsen P) was determined using the Olsen method (Olsen et al., 1954). The fields studied differed in sowing dates and the amounts of water and fertilizer applied. The irrigation doses were set according to crop water requirements estimated from reference evapotranspiration (ETo) and crop coefficient (Kc) adjusted to local climatic conditions. The sowing dates, plant densities, seeding rates, and fertilizer amounts for each field are detailed in Table 1.

Fertilizer application rates were determined individually for each field according to crop nutrient requirements and the results of soil analyses. All fields studied were irrigated using a drip irrigation system. Nitrogen fertilizer is typically supplied as one-third of the total amount before sowing, with the remaining two-thirds applied progressively through fertigation in four split applications over the growing season. Soil fertility in the AquaCrop model is expressed as a percentage, corresponding to classes ranging from non-limiting (100%) to very poor (20%), as shown in Table 2.13a of (Raes et al., 2022). This is quantified using the Brel index, which represents the ratio of biomass produced under fertility stress to that under optimal fertility conditions. Lower Brel values indicate stronger fertility stress (Van Gaelen et al., 2015; Akumaga et al., 2017). Weed and disease management reflected actual farmer practices, with recommended post-emergence herbicides and fungicides applied by the farmers.

Meteorological data for the maize growth periods from the 2021/22 to 2023/24 maize growing seasons were systematically collected using a weather station situated near the experimental area. The dataset comprised daily averages of solar radiation(W/m²), maximum and minimum temperatures(°C), wind speed(m/s), relative humidity (%), and rainfall (mm). Daily reference evapotranspiration (mm) values were derived using the Penman-Monteith equation, as described by (Allen et al., 1998). Figure 2 illustrates the temporal variation in maximum and minimum daily air temperatures (Tair), ETo, and rainfall across the three growing seasons.

The soil‐water profile was monitored with a PR2/6 (Delta-T Devices Ltd, 2015), which uses moisture sensors embedded in a sealed rod that can convert an analogue DC voltage signal into volumetric soil water content (VWC) with a stated accuracy of ± 3%. One access tube per plot was installed mid-row, and measurements were taken at depths of 0.10, 0.20, 0.30, and 0.40 m.

To account for site-specific bias, probe readings were calibrated directly using gravimetric VWC obtained from paired core samples, dried at 105°C for 72 h. The resulting depth-specific regressions were applied to all probe data. Root-zone water storage was then calculated by integrating the calibrated profile. Given that maize planting in the region typically follows a significant pre-sowing irrigation event, farmers start sowing when soil moisture levels are favorable for seed germination. As a result, the initial soil moisture parameter used in the AquaCrop model was close to field capacity (Raes et al., 2022). Farmers schedule the irrigation events so that the applied water depth covers the crop’s water requirement. This requirement is estimated using the formula (ETc=ETo*Kc), where ETo represents the reference evapotranspiration obtained with the FAO Penman-Monteith equation from meteorological data recorded at the nearest weather station (30°33′15.20″ N, 8°40′0.05″ W), and Kc is the maize crop coefficient adjusted to local climate conditions following (Allen et al., 1998).

Field sampling was conducted during the growth period to measure aboveground biomass (B). Five uniformly growing plants were randomly selected from each field and cut at the base of the stem with scissors. The leaves, stems, and fruits were placed in archival bags to measure their fresh weights. The dry matter weight was then determined using the oven drying method. All plant samples were heated at 105°C for 30 minutes, followed by drying at a constant temperature of 80°C until they reached a constant weight, and finally weighed using a balance.

In this research, data from the Copernicus Sentinel-2 mission (Gascon et al., 2017) were utilized. Launched in March 2017, this mission involves two satellites (S2A and S2B) providing 10 m spatial resolution products and a 5-day temporal resolution at the equator (ESA Sentinel, 2014). These data were processed to estimate NDVI (Normalized Difference Vegetation Index), using L2A atmosphere-corrected products. Each acquisition date yielded images capturing reflectance in the red (Band 4) and near-infrared (Band 8) spectra. Rigorous screening procedures were applied to identify and exclude images affected by clouds or shadows over the areas of interest. Additionally, measures were taken to minimize border effects by excluding pixels from field perimeters adjacent to roads and neighboring fields (Khaliq et al., 2019; Sozzi et al., 2020). The operations described were conducted on the Google Earth Engine platform. To address missing data, we employed a linear interpolation method. Additionally, the NDVI time series were smoothed using a Savitzky-Golay filter (Chen et al., 2004), which effectively preserves vegetation signal information. Leaf area index (LAI) was measured 8 to 10 times during the growing season with the SunScan canopy analysis system (Netzer et al., 2009; Oguntunde et al., 2012; Ariza-Carricondo et al., 2019) The resulting values allowed estimating canopy cover (CC) using Equation 1 (Hsiao et al., 2009):

The CC values were also determined using the Normalized Difference Vegetation Index (NDVI) obtained from Sentinel-2 imagery. This estimation used a modified equation from the (Tsakmakis et al., 2021) study, where the relationship between NDVI and canopy cover (CC) is expressed as Equation 2.

Figure 3 illustrates the relationship between observed canopy cover (CC Observed) and canopy cover estimated using Sentinel-2 data (CC Sentinel-2) in C1-C4 fields. A trend line is fitted to the data points, indicating a positive correlation between the observed and estimated values (r =0.87), which indicates an accurate estimate of the observed canopy cover across the study area.

At maturity, five 1^x1 m² quadrats of maize plants were randomly selected from each plot to measure plant density, height, weight, average number of cobs, number of kernels per cob, and 1000-kernel weight. Grain yield was derived based on yield per plant and the planting density for each field.

The AquaCrop model is a water-driven decision support tool developed by FAO to simulate the impact of environmental conditions and management practices on crop production and support food security. The model integrates four core modules: meteorology, crop, soil, and field management to simulate dynamic interactions within the soil-plant-atmosphere system. It estimates crop growth and yield based on daily weather inputs (rainfall, temperature, evapotranspiration, CO₂ concentration), crop characteristics (such as phenology, canopy cover, root depth, water productivity, and stress responses), management practices (irrigation, fertilization, planting density), and soil properties, using a daily soil water balance to quantify the impact of water stress on productivity (Raes et al., 2009; Steduto et al., 2009a).

AquaCrop refines the classical water-yield relationship described by (Doorenbos and Kassam, 1979) through a more physiologically grounded modeling approach. While the previous method relied on empirical production functions that relate yield loss proportionally to reductions in evapotranspiration, AquaCrop improves this framework by explicitly separating evapotranspiration into soil evaporation and crop transpiration, based on the premise that only transpiration directly contributes to biomass accumulation. Biomass is subsequently converted into yield via the harvest index, with water productivity as a key parameter linking transpiration to biomass production. These relationships are represented by Equations 3, 4.

where HI is the harvest index, and B is the accumulated above-ground biomass (kg m^-2). WP represents the normalized water productivity (kg m^-2 mm^-1), Tr is the daily crop transpiration (mm day^-1), and ETo is the reference evapotranspiration (mm day^-1). A comprehensive description of these concepts and equations is provided by (Hsiao et al., 2009; Raes et al., 2009; Steduto et al., 2009a).

The AquaCrop model was calibrated using data collected from seven fields and validated on ten fields during the 2021–2024 growing seasons, all managed under drip irrigation. The parameters selected for calibration were determined based on the literature and the results of the sensitivity analysis (Supplementary Figures S1 in the Supplementary Material). As illustrated in the sensitivity analysis, the canopy growth coefficient (CGC), the time from emergence to maturity, and the maximum crop transpiration coefficient (Kc_Tr,x) were identified as the most influential parameters driving the final maize biomass.

Crop development in AquaCrop was driven by cumulative growing degree-days (GDD) calculated using a base temperature of 8°C and an upper temperature of 30°C, as recommended by (Hsiao et al., 2009; Raes et al., 2018) for maize. The initial canopy cover (CC₀) was determined from measured plant density and seed characteristics. With a row spacing of 0.7 m and an intra-row spacing of 0.13 m, the average density reached approximately 11 plants per m2, corresponding to an initial CC₀ of 0.71%.

Calibration of the model involved iterative adjustment of parameters to minimize discrepancies between observed and simulated canopy cover (CC), soil water content (TWC), and above-ground biomass. Initially, phenological development was calibrated by modifying the growing degree-day requirements for emergence, maximum canopy cover, flowering, onset of senescence, and maturity until the simulated crop calendar accurately reproduced field observations. Subsequently, parameters controlling canopy expansion and decline, including the canopy growth and decline coefficients, were fine-tuned to match the observed canopy dynamics. In parallel, parameters regulating maize response to water stress and transpiration were optimized by adjusting the thresholds for leaf expansion, stomatal conductance, and canopy senescence, as well as the shape of the stress-response functions and the maximum transpiration coefficient (KcTr,x). In addition, the normalized crop water productivity (WP*) was set to the default value recommended for maize in AquaCrop, ensuring consistency across all calibration fields. The final set of calibrated conservative and non-conservative parameters is summarized in Supplementary Table S2 of the Supplementary Material. Model validation was performed using an independent dataset from ten separate fields, and model performance was evaluated by comparing simulated and observed CC, TWC, and biomass through standard statistical indicators, confirming the reliability of the model under diverse field conditions.

To evaluate the performance of AquaCrop model, five statistical indicators were employed in this study. The coefficient of determination (R²), root mean square error (RMSE), normalized root mean square error (CV(RMSE)), Nash-Sutcliffe efficiency (NSE), and mean bias error (MBE). Coefficient of determination (R2) quantifies the proportion of the variance in the observed data that is explained by the model (Equation 5). It ranges from 0 to 1, with values close to 1 indicating a high level of agreement. The root mean square error (RMSE), defined in Equation 6, measures the discrepancy of simulated values around observed ones (Jacovides and Kontoyiannis, 1995; Legates and McCabe Jr, 1999).

The normalized root mean square error (CV(RMSE)), defined in Equation 7, expresses the relative difference between simulated and observed values as a percentage. Model performance is classified as excellent (CV(RMSE)< 10%), good (10-20%), fair (20-30%), or poor (>30%) (Jamieson et al., 1991). The Nash-Sutcliffe Efficiency (NSE), given in Equation 8, assesses model accuracy by comparing residual variance to observed data variance. NSE ranges from −∞ to 1, with 1 indicating perfect agreement (Nash and Sutcliffe, 1970; Moriasi et al., 2007).

The mean bias error (MBE), defined in Equation 9, evaluates the average tendency of the model to overestimate or underestimate the observed values, where Pi and Oi refer to the simulated and observed values, respectively.

In the AquaCrop model, soil water stress is quantified based on root zone water depletion relative to total available water (TAW). Irrigation is triggered when depletion reaches a predefined threshold, typically expressed as a percentage of the crop’s readily available water (RAW). After calibration and validation, the model was used to simulate different scenarios, as described in Supplementary Table S3 (included in the Supplementary Material), including irrigation scheduling under three threshold levels: 120%, 100%, and 70% of RAW, with each event refilling the soil to field capacity. Additionally, three deficit irrigation (DI) regimes were tested, supplying 100%, 75%, and 50% of the seasonal crop evapotranspiration under standard conditions (ETc). These simulations generated outputs including soil moisture status, actual evapotranspiration (ET_{c act}), transpiration (Tr), drainage (Pr), and water productivity (WP), enabling a comparative assessment of irrigation management strategies relative to farmer practices under uniform edaphoclimatic conditions. These scenarios were tested to explore practical options for improving water use under limited availability. Varying RAW thresholds assess the impact of irrigation timing, while deficit irrigation scenarios represent supply constraints commonly encountered in water-scarce regions.

In addition to irrigation management, the AquaCrop model was also used to assess how shifts in maize sowing dates influence crop development and yield. Three scenarios were tested: early, delayed, and standard planting, with time adjustments of approximately 20 to 40 days. The objective was to determine whether better alignment between crop growth stages and climatic conditions, particularly rainfall distribution and temperature patterns, could enhance water use efficiency and yield stability. These adjustments may enable crops to capitalize on favorable environmental conditions during critical growth stages, especially under increasing climate variability and water scarcity.

Furthermore, the model was also used to simulate the influence of various mulching strategies on soil moisture dynamics and maize performance and to determine which type of mulch provides optimal soil microclimatic conditions to support maize growth and yield, particularly under conditions of limited water availability. Three treatments were examined: organic mulch, synthetic (plastic) mulch, and a no-mulch control. Mulching practices were selected based on their proven ability to reduce soil evaporation, conserve moisture, and moderate soil surface temperatures-factors that directly influence water availability in the root zone and crop physiological responses.

Additionally, the different outputs of the AquaCrop model were used to calculate the water productivity (WP) as the ratio of yield and water consumption. Various ways referring to the amount of water consumption were chosen to calculate different forms of WP: WP_Tr referred to transpiration (Tr), WP_ETcact to actual evapotranspiration, WP_ET+Pr to the sum of evapotranspiration and percolation (Pr) (Molden, 1997; Molden et al., 2001). The calculations were as follows:

This section presents the simulation results from the calibrated AquaCrop model, used to test different crop management scenarios in the Souss-Massa region. It focuses on evaluating irrigation schedules, mulching practices, and variations in planting dates to provide practical recommendations for local farmers.

The performance of the AquaCrop model in simulating canopy cover (CC) revealed a tendency to slightly underestimate peak development stages. This discrepancy could be linked to AquaCrop’s tendency to assume optimal physiological responses when explicit stress factors are not fully parameterized. Similar underestimations under combined abiotic stress have been reported by (Afrooz et al., 2025). In this study, canopy development was calibrated using observed LAI-derived CC curves, which enabled adjustment of key parameters including the time required to reach maximum canopy cover (CCx), as well as the canopy growth coefficient (CGC) and canopy decline coefficient (CDC) based on data from the calibration fields to represent the overall canopy development pattern of maize in the study area. In contrast, the validation fields were excluded from calibration and simulated using the calibrated parameters to maintain an independent model evaluation. Despite these refinements, minor mismatches persisted, indicating that AquaCrop may still overlook certain limiting factors affecting canopy expansion under field conditions. These findings are consistent with (Sandhu and Irmak, 2019), who reported underestimation of CC across all irrigation treatments, especially under higher water stress. Likewise (Birhan et al., 2025), observed similar patterns in irrigated maize, attributing the model’s bias to low initial CC and an underestimated CGC, which slowed simulated canopy development. In addition, AquaCrop’s limited sensitivity to early-season water stress likely contributed to the observed divergence between simulated and measured values.

The model accurately represented soil water content throughout the season, although it consistently exhibited a slight underestimation. Similar results have been reported in previous studies that found AquaCrop tends to underestimate SWC in maize systems (Biazin and Stroosnijder, 2012; Mebane et al., 2013). This bias could result from several factors, including soil heterogeneity, which is not explicitly represented in AquaCrop. The application of uniform soil parameters across diverse field conditions can introduce uncertainty in SWC simulations (Ahmadi et al., 2015). Additionally, AquaCrop’s simplified water balance module only initiates drainage when moisture exceeds field capacity, disregarding preferential flows through macropores or cracks that can lead to deep percolation even below field capacity (Hsiao et al., 2009; Zeleke et al., 2011). Underestimated or non-calibrated crop coefficients (Kc) may also reduce simulated evapotranspiration, resulting in an overestimation of retained soil moisture (Paredes et al., 2014). Conversely, AquaCrop has occasionally overestimated SWC immediately following irrigation events, likely due to mismatches between simulation output time steps and the timing of field observations. This discrepancy was noted by (Nyakudya and Stroosnijder, 2014), who reported post-irrigation overestimations when soil moisture measurements failed to coincide with peak water content.

Above-ground biomass was simulated with high accuracy, although a modest overestimation bias was observed, particularly in higher-performing fields. This could be linked to AquaCrop’s limited sensitivity to nutrient constraints, especially in the absence of detailed soil fertility calibration. Katerji et al. (2013) noted that AquaCrop tends to slightly overpredict final biomass when nutrient limitations are present but not explicitly modeled, particularly in late growth stages. However, the model may not fully reflect the combined effects of soil fertility, micro-nutrient availability, and subsoil variability, which could explain the residual deviations observed. In addition, Early-season biomass overestimation likely stems from AquaCrop driving canopy closure too fast when canopy parameters (CCx, CGC) aren’t tailored to the local conditions and when soil-water depletion thresholds (P_upper/P_lower) postpone the onset of modeled water stress (Geerts et al., 2009; Hsiao et al., 2009). The bias can be amplified by calibration data and water-management context, which make early stress appear milder in the model. Later in the season, AquaCrop’s simplified treatment of stress-induced senescence and limited recovery after rehydration tends to flip the error to underestimation (Katerji et al., 2013; Vanuytrecht et al., 2014b; Ahmadi et al., 2015).

The simulation results corroborate the positive influence of mulching on water-use efficiency, in alignment with previous studies that underscore its agronomic and hydrological advantages. For example (Pacetti et al., 2025), reported that organic mulching can reduce irrigation requirements by 8-50% and enhance water productivity by 5-50%. Similarly (Masasi et al., 2025), demonstrated that straw mulching substantially increased simulated yields and water-use efficiency across multiple vegetable crops when compared to conditions without mulch. In a comparable context, Abera et al. (2025) observed that the combination of mulching and deficit irrigation in furrow-irrigated, raised-bed maize systems significantly improved both simulated water productivity and final yield. In addition to hydrological benefits, numerous studies have highlighted the thermoregulatory functions of plastic mulch. It has been shown to improve soil moisture conservation and elevate soil temperature, thereby facilitating seedling emergence, enhancing transpiration, promoting biomass accumulation, and ultimately increasing crop yield (Cook et al., 2006; Peltonen-Sainio et al., 2009; Dong et al., 2014; He et al., 2017). These thermal effects have been associated with improved physiological responses such as increased evapotranspiration, enhanced spike formation, and superior grain quality. These results align with our simulations, which indicated that using mulch especially synthetic mulch led to significantly higher transpiration and dry biomass compared to scenarios without mulch. Nonetheless, the conceptual simplifications incorporated within the AquaCrop model must be acknowledged. Specifically, the model represents the effects of mulching primarily through its impact on reducing soil evaporation, thereby omitting other important biophysical interactions. For instance, it does not fully account for air temperature dynamics, which can critically affect thermal stress responses, pollination processes, and biomass partitioning (Raes et al., 2009). Plastic mulch raises the near-surface air temperature more than bare soil, potentially increasing reference evapotranspiration (Wang et al., 2012). These limitations indicate that, although AquaCrop reliably simulates the hydrological impacts of mulching, its outputs should be interpreted with caution in cropping systems where thermal effects significantly influence crop performance. Plastic mulching can increase yields, improve product quality, and enhance water-use productivity, especially with drip irrigation (Biswas et al., 2015; López-López et al., 2015), but its economic feasibility remains uncertain since revenues may not always offset the costs of purchase, application, and disposal (Tiwari et al., 2003). Although water savings and reduced labor for weed and pest control offer economic benefits (Ingman et al., 2015; Jabran et al., 2015), overall profitability depends on crop type, labor costs, market opportunities, and irrigation conditions, making its use advantageous only in specific contexts (Steinmetz et al., 2016). For smallholder farmers, high upfront and labor costs can further limit adoption, underlining the need for policy support such as subsidies, training, or improved recycling systems to ensure broader economic feasibility.

Advancing the sowing date by about 3 to 6 weeks increased biomass production and improved both evapotranspiration- and transpiration-based water productivity. The simulations show that earlier sowing exposes the crop to lower early-season evaporative demand, which reduces soil evaporation and increases the share of water used for productive transpiration. It also lengthens the vegetative phase and allows earlier and longer canopy development, leading to greater cumulative transpiration and consequently higher simulated biomass. These findings are in agreement with previous studies reporting significant yield improvements associated with early sowing in rice, maize, and other field crops under climate-adaptive management strategies (Barati et al., 2024; Tesfay et al., 2024). Early sowing has also been linked to enhanced water management and transpiration efficiency (Achli et al., 2025), notably in maize, where it facilitates rapid canopy development and earlier ground coverage, thereby reducing soil evaporation and overall crop evapotranspiration. Similar effects have been observed across various agroecological zones, where shortened growth cycles contribute to reduced water consumption (Islam et al., 2012; Bassu et al., 2014). By shifting planting dates, farmers can make more efficient use of soil moisture and reduce exposure to erratic rainfall (Graef and Haigis, 2001). Increasing climatic variability has made the timing and distribution of seasonal precipitation increasingly unpredictable, thereby limiting the effectiveness of traditional planting calendars in guiding optimal sowing decisions (Kijazi and Reason, 2009; Recha et al., 2012; Nyagumbo et al., 2017). A diversified sowing strategy identifies the optimal planting windows each season to reduce yield losses. By exposing crops to varying light, temperature, and moisture regimes during successive growth stages, it influences their development and productivity (Li et al., 2021). Still, identifying the precise start of these varied dates remains a critical and inherently challenging task (Folberth et al., 2012). Furthermore, optimizing sowing dates tends to homogenize field conditions, resulting in improved irrigation planning, cutting water use by over 40% (Heng et al., 2007; Toumi et al., 2016; Belaqziz et al., 2021; Dewenam et al., 2021; Taaime et al., 2022).

Establishing relative root-zone water depletion thresholds to trigger irrigation is essential for designing irrigation schedules that maximize crop yield and water productivity under a constrained water supply. These findings resonate closely with (Zhang et al., 2023), where applying irrigation only once depletion reached approximately 110-120 % of RAW maximized both wheat grain yield and water productivity, non-productive soil evaporation declined by nearly 40%, and wheat yields remained above 95% of the full-irrigation control, all without increasing total irrigation depth. Similarly (Mostafa et al., 2023), found that 120% RAW optimized water productivity in rice. According to (Ket et al., 2018), irrigation triggers ranging from 0 to 200% RAW were evaluated. Thresholds near 130-150% RAW achieved up to 22% water savings with less than 5% biomass loss, thereby boosting water productivity again, underscoring the benefit of delayed irrigation initiation around or above 120% RAW. Fereres and Soriano (2007) confirms that applying water below full evapotranspiration demands can increase crop water productivity by 10-30% through reduced unproductive evaporation and improved transpiration efficiency. When the deficit becomes too severe, yield losses can offset these gains, emphasizing the importance of moderate stress thresholds.

The results indicate that moderate deficit irrigation improves water productivity with minimal impact on biomass production. Similar findings have been reported in previous studies, where applying irrigation at 70–90% of crop evapotranspiration (ETc) enhanced water use efficiency while maintaining yield stability (Sharafkhane et al., 2024). Further reductions in irrigation have been associated with higher water use efficiency and potential economic gains when supported by appropriate pricing policies (Asmamaw et al., 2023). Moreover, imposing 50% ETc deficits specifically during the mid-to-maturity growth stages sustained grain yields of 8 842 kg ha^-1 while achieving water-use efficiency up to 2.11 kg m^-3, thereby illustrating the advantages of phenological targeting (Gebreselassie et al., 2015). In contrast, integrated SWAT-MODFLOW-AquaCrop simulations predicted that a uniform 50% water supply reduction would reduce maize yield by 17.3%, whereas season-long 50% ETc stress in this study imposed a biomass penalty exceeding 30% (Hu et al., 2025). Moreover, coupled DSSAT CERES Maize AquaCrop modeling demonstrated that a mild 15 % soil-water reduction maximizes grain yield, while a 60% depletion optimizes water-use efficiency at net depths of 60-134 mm, indicating the critical role of irrigation depth and environmental context in deficit irrigation strategies (Painagan and Ella, 2022).

This work provides a reliable modeling approach for assessing maize cultivation practices under Souss-Massa conditions. However, some limitations need to be considered. The nutrient cycle and its dynamics are not explicitly considered during the simulation process in AquaCrop; instead, the model uses aggregated stress coefficients (Ks) that adjust canopy expansion, maximum canopy cover (CCx), and normalized water productivity (WP*), making fertilization scenarios broad proxies that cannot reflect fertilizer dose, type, or timing (Steduto et al., 2009b). Additionally, the model’s parameterization could present limited transferability across crop cultivars, as phenological development, canopy growth, and harvest index often require calibration tailored to specific genotypes or environmental conditions (Coudron et al., 2023). The model is less accurate under severe water stress, especially during late growth stages or senescence (Heng et al., 2009). In addition, calibration and validation rely on field data that are subject to measurement and record uncertainties, which could limit the precision of management-specific insights, while maintaining the general observed trends. Beyond these factors, the AquaCrop model presents a notable limitation in that it does not account for the impacts of pests and diseases, which may lead to an overestimation of crop yield (Adeboye et al., 2021). Despite their usefulness, crop models such as AquaCrop inherently face challenges in accurately estimating crop growth due to the complex interactions between plants and their environment, as well as the combined effects of biotic and abiotic stressors that cannot all be fully represented (Puig et al., 2025). Moreover, AquaCrop does not simulate the allocation of assimilates to individual plant organs or represent plant architecture, thereby restricting its capacity to capture organ-level growth dynamics (Steduto et al., 2009a).

Considering the limitations mentioned above, Future efforts should consider integrating a mechanistic nutrient cycling module to explicitly simulate nutrient dynamics and refine model calibration to account for cultivar-specific characteristics and sensitivity under severe water stress conditions. In addition, future research should primarily focus on validating the simulated scenarios such as deficit irrigation, mulching, and sowing dates under real field conditions to strengthen the credibility of model outputs. Integrating regional climate change projections will also be essential to evaluate the long-term performance of these strategies under evolving climatic conditions. Moreover, future research should further consider socio-economic aspects, including the cost-effectiveness and adoption potential of the proposed practices, to enhance their relevance for smallholder farmers. In addition, adopting participatory approaches involving local farmers and local advisory services will be crucial to ensure the practicality and local acceptance of the recommended strategies.