Assessment of the Impacts of Climate Change Scenarios on Maize Yield and Irrigation Water Using the CropSyst Model: An Application in Northern Greece

Abstract

In the coming decades, crop production in regions such as the Mediterranean Basin is expected to be influenced by climate change. This study evaluates the impacts of climate change on maize yield and irrigation water in Northern Greece for the mid-21st century and late 21st century using CropSyst, a cropping systems simulation model. Data from a two-year field experiment with maize, in 2016 and 2017, were used to calibrate and validate CropSyst. RCP4.5 and RCP8.5 climate change scenarios were employed, derived from three Regional Climate Models (RCMs), for two future periods (2030–2050 and 2080–2100) and the baseline period (1980–2000). The RCMs used in this study were derived from the Rossby Centre regional atmospheric model (RCA4), which downscaled three General Circulation Models (GCMs), CNRM-CM5, CM5A-MR, and HadGEM2-ES, as part of the Coordinated Regional Climate Downscaling Experiment (EURO-CORDEX) for the European domain. Results indicate that changes in climate variables will exert potential pressure on full irrigation water requirements, leading to both increases and decreases in irrigation amounts, with varying magnitudes of change. Yield impacts vary depending on the climate change scenario and climate model, with CropSyst predictions indicating both positive and negative effects on maize yield under full irrigation. The combined effects of increased temperatures, reduced precipitation, and elevated CO₂ concentrations under the high-emission scenario RCP8.5 by the late 21st century resulted in substantial declines in maize yields. The study identifies the key factor influencing maize yield in future periods as the combined changes in climate variables under CO₂ concentration enrichment, which lead to alterations in full irrigation water requirements, highlighting the multiparameter nature of impact assessment on agricultural production in Northern Greece under various future climate scenarios.

1. Introduction

Climate plays a critical role in crop growth rates and yield; thus, a global concern is the potential effects of climate change on crop production. Agricultural crop production is influenced by climate variables as photosynthetically active radiation, air temperature, and water are key drivers of crop growth [1,2,3]. Most plants exposed to elevated atmospheric CO₂ levels exhibit an increased rate of photosynthesis, resulting in higher biomass accumulation [4]. However, there is uncertainty regarding whether this increase in photosynthetic rate also occurs in the case of C4 plants. The agricultural sector is expected to experience a range of impacts due to climate change with consequent development of potential pressures on agricultural systems. Therefore, understanding the potential impacts of climate change is a challenge for researchers to develop both adaptation strategies and actions for reducing climate change risks.

Maize (Zea mays L.) is a grain with a high germ content that is widely used for animal feed and starch production, making it one of the main crops worldwide. Arable crops in Greece cover almost 53% of cultivated land [5], with maize representing 10% of arable crop cultivated area in Greece [5]. In Greece, maize is planted at the beginning of April and harvested in September and October, and it is intensively irrigated. The maize yield in Greece increased from 3016.5 kg ha⁻¹ in 1970 to 10,680.7 kg ha⁻¹ in 2019, growing at an average annual rate of 2.87% [6]. Numerous research efforts have been undertaken to assess the potential climate change impacts on maize yields at various locations worldwide due to maize’s critical importance to the sustenance and livelihood of many human systems.

Maize is one of the primary crops investigated in climate change impact assessments, leading to different results depending on the region’s climate conditions, global and regional climate models, emissions scenarios used, and the climate change period studied. According to Parry [7] and Meza et al. [8], the responses of maize under conditions of climate change were accelerated development rates, decreased grain number, and reduced grain unit weight. Jones and Thornton [9] studied maize performance across various Latin American locations, projecting yield reductions of approximately 10%, and Magrin et al. [10] simulated maize growth in Argentina, showing potential yield reductions of about 16% under climate change conditions. Studies by Alexandrov and Hoogenboom [11] in Bulgaria and by Cuculeanu et al. [12] in Romania concluded that maize yield would decrease by 5–10% in Bulgaria and by 10% in Romania, respectively. In maize yield assessments in irrigated agriculture in Mediterranean climates such as Spain [13] and Italy [14], the results showed that biomass accumulation is projected to decrease, with a negative impact on yields on the order of 16% and 20%. Voloudakis et al. [15] conducted a study in different areas of Greece under different climate change scenarios, which showed small changes in maize yields, resulting in both reductions and increases according to the scenario used.

In climate research, possible climate change impacts on crop yield in the future can be determined by using crop growth simulation models under different climate scenarios. Crop growth simulation involves a complex interaction among weather parameters, soil properties, plant characteristics, and management practices, all of which affect a crop’s response to different water and nutrient inputs [16,17]. A growing number of models have been used for specific applications and scales, incorporating various input variables and mechanisms of crop growth [18,19,20]. One of the most significant challenges is the models’ ability to adequately simulate crop growth and development under different management practices. CropSyst (a cropping systems simulation model) is a multi-year, multi-crop simulation model designed to assess the impacts of cropping systems management on productivity and the environment [21,22], applied to simulate the growth of various crops, with reliable results [23] across different global regions. Thus, employing the CropSyst model can enhance insights into crop responses under varying environmental conditions and agricultural management practices [24,25].

This study aims to evaluate climate change effects on maize yield and irrigation water, altering the full irrigation water according to the CropSyst auto-irrigation option for the mid (2030–2050) and late (2080–2100) 21st century under different climate change scenarios. The specific objectives of this study are to (i) analyze future projections of climate parameters, including CO₂ concentration changes; (ii) simulate maize yield responses to climate change, considering the effects of temperature, precipitation, and CO₂ concentrations under full irrigation; and (iii) determine changes in full irrigation water amounts in the future. In this regard, data were obtained from three Regional Climate Models (RCMs), derived from the dynamical downscaling of GCMs CNRM-CM5, HadGEM2-ES, and IPSL-CM5A-MR, under two Representative Concentration Pathway (RCP) scenarios, RCP4.5 and RCP8.5, for the future climate change periods (2030–2050 and 2080–2100) and the baseline period (1980–2000). The predictions were based on the implementation of the CropSyst growth simulation model for simulating the maize yield under the different climate scenarios (RCP4.5 and RCP8.5) according to the RCMs used during the periods of climate change. CropSyst model calibration, validation, and evaluation were conducted using experimental data from maize crop cultivated in Northern Greece over two years (2016 and 2017 growing seasons).

The study contributes to research conducted in Northern Greece, as there have been only a limited number of model-based studies examining the impacts of climate change on maize and irrigation water demands in this region. Unlike previous research, our study utilizes simulations based on the latest generation of climate models, incorporating the most up-to-date future scenarios (RCPs) and employing CropSyst, a dynamic crop growth simulation model that integrates interactions among crop, water, soil, and weather, while also considering different irrigation practices and management options. Additionally, this study addresses scientific issues related to climate change impacts on agricultural systems at small spatial scales with fine resolution. This approach enhances the applicability of our findings for local producers who require adaptation strategies, thereby addressing a significant gap in the current literature. The anticipated results are expected to provide valuable insights for developing effective irrigation management and promoting sustainable agricultural practices.

2. Materials and Methods

2.1. Study Area

In this study, data were obtained from a field experiment conducted in 2016 and 2017 at the experimental field of the University farm of Aristotle University of Thessaloniki, in Northern Greece (40°32′ N and 23°00′ E) (Figure 1a). The experimental field was located about 11 km from Thessaloniki in an area characterized by a typical Mediterranean climate, with annual precipitation distributed mostly in autumn and winter and with hot summers and cool winters. The climate variables (Figure 1b) used in the study (maximum and minimum temperature, maximum and minimum relative humidity, wind speed, solar radiation, and precipitation) were taken from the meteorological station of the Hellenic National Meteorological Service of Greece (HNMS), which is located about 3 km from the experimental site (40°31′ N latitude and 22°58′ E longitude), and include the period 1980–2000. The experimental field had sandy clay loam soil for the top 40 cm with organic matter of 1.58%, pH of 7.99, and EC of 0.48 dS m⁻¹ with respect to the topsoil (0–20 cm). The main soil physical and chemical properties of the field are presented in

Table 1. Physical and chemical properties of the soil at the site of the experiment.

Depth (cm)	Texture	Sand (%)	Clay (%)	Silt (%)	pH (1:2.5)	EC (dS m−1)	OM (%)	CEC (meq/100 gr)
0–20	SCL	53	25	22	7.99	0.48	1.58	29.86
20–40	SCL	59	23	18	7.97	0.41	0.86	27.88
40–60	SL	66	19	15	8.05	0.41	0.37	27.85
60–80	SL	68	16	16	8.08	0.38	0.19	27.73
80–100	SL	66	17	17	8.10	0.43	0.17	26.77

SCL: Sandy clay loam; SL: sandy loam; EC: electrical conductivity; OM: organic matter; CEC: cation exchange Capacity.

.

2.2. Description of the Field Experiment

To simulate maize yield under different climate change scenarios, a field experiment was carried out during the 2016 and 2017 growing seasons with maize (Zea mays L.) cultivar AGN 720. During the preceding growing seasons, the experimental field was left fallow. The experimental design included one treatment with full irrigation applied in four replications (plots) in both years. Each experimental plot was 6.4 m × 6 m with eight rows, and the total size of each plot was 38.4 m², including 38 plants per plot. The experimental field was prepared before planting and then sown with 2–3 seeds per position, with plant spacing of 0.80 m between rows and 16 cm within each row, resulting in a final density of approximately 7800 plants/ha in both years. Sowing occurred on the 21st and 22nd of April, and harvesting took place on the 5th and 9th of September, with yields of 13,450 kg ha⁻¹ and 14,965 kg ha⁻¹ for 2016 and 2017, respectively. Plants were thinned approximately 30 days after sowing to obtain one plant per position. Weed control was achieved through hand weeding and tilling when necessary. Nitrogen fertilizer was divided into three applications and manually applied as ammonium sulphate (21%) in two applications and as ΚΝO₃ (13.5-0-46.2) in one application during the 2016 growing season. Regarding the year 2017, nitrogen fertilizer was applied in the form of ammonium nitrate (12-12-17) and (21-5-10) in two doses. The applied fertilizer amounts were adequate to establish optimum growth in both years.

To improve irrigation management, an irrigation scheduling approach was adopted. This approach integrates meteorological, crop, and soil data to estimate the soil water balance within the effective root zone on a daily basis, providing a dynamic approach to irrigation scheduling under full irrigation conditions. This ensures that irrigation fully meets the crop’s water requirements while avoiding water stress or wastage. Measured weather data were used to calculate daily reference evapotranspiration using the FAO Penman–Monteith equation [26]. Crop coefficient (k_c) values, adjusted for the Greek region, were based on FAO 56 [26]. The crop evapotranspiration was calculated as 528 mm for the 2016 cultivation period and 519 mm for the 2017 cultivation period. To monitor soil water content, two soil moisture sensor probes were installed at depths of 60 cm and 100 cm in the experimental field (measuring soil water content at 10 cm intervals throughout the soil profile, from a depth of 0 to 60 cm and 0 to 100 cm, respectively).

A drip irrigation system was used, with drippers spaced at 50 cm intervals and each dripper having a flow rate of 4 l h⁻¹. In the year 2016, full irrigation of 340 mm was applied, composed of 9 irrigation water applications; in 2017, there were 11 irrigation water applications with a total water amount of 390 mm.

2.3. Climate Projections

2.3.1. Regional Climate Models (RCMs)

Climate change impact assessments are conducted by developing region-specific climate scenarios, which are then used to drive crop models in the sector of interest. In climate research, climate scenarios are represented through General Circulation Models (GCMs) and Regional Climate Models (RCMs) to predict potential future climate change. The aim of these scenarios is not to predict the future but to explore both the scientific and real-world implications of how the future may evolve [27]. In 2013, the Intergovernmental Panel on Climate Change (IPCC) introduced Representative Concentration Pathways (RCPs), a set of four new pathways designed for long-term and near-term future projections. These pathways provide internally consistent, time-dependent forcing projections that consider climate change mitigation policies aimed at limiting emissions [28]. The RCPs are named based on the approximate radiative forcing levels relative to the pre-industrial period, achieved by the year 2100 or stabilized shortly thereafter [29]. RCP2.6 is a mitigation scenario that is characterized by a radiative forcing of about 2.6 W m⁻² and requires CO₂ concentrations to be limited to approximately 430 ppm. RCP4.5 and RCP6 represent medium stabilization scenarios, with radiative forcings of about 4.5 W m⁻² and 6 W m⁻², respectively, with CO₂ concentrations projected to reach approximately 580 ppm and 670 ppm. RCP8.5, the very high baseline emission scenario, anticipates a radiative forcing level of about 8.5 W m⁻², with CO₂ levels projected to reach approximate 1000 ppm [30].

GCMs (General Circulation Models) are the most advanced tools for representing the global climate system’s response to increased greenhouse gas concentrations in the atmosphere [28,30,31]. A limitation of GCMs is their relatively coarse horizontal resolution, typically ranging from 100 km to 300 km. This resolution restricts their utility for providing detailed information in climate change impact studies related to hydrology, ecosystem services, and other landscape and agriculture-related issues [31]. Future projections based on Regional Climate Models (RCMs), which operate over limited-area domains with higher horizontal resolution (12.5–50 km), have been extensively used and offer a valuable advancement for climate change impact assessments. RCMs provide more reliable regional-scale results compared to General Circulation Models (GCMs) [32]. These models are developed by nesting a secondary model within one or more of the grid spaces of a GCM. RCMs more accurately represent land-use data, elevation, rainfall events, and soil conditions, and they may simulate certain processes, such as convective cloud behavior, more effectively than GCMs, which currently cannot be simulated satisfactorily.

In this study, climate parameters from three GCMs (General Circulation Models) were used, which were dynamically downscaled by the Rossby Centre Regional Climate Model (RCA4) and available in the context of the Coordinated Regional Climate Downscaling Experiment (EURO-CORDEX) for the European domain with a finer resolution of 0.11 degree (EUR-11, ~12.5 km). The Rossby Centre, part of the Swedish Meteorological and Hydrological Institute (SMHI), has created the most recent version of the Regional Climate Model (RCA4). RCA4 generated a large ensemble of simulations for Europe, using boundary conditions from nine (or five) different global climate models from the Coupled Model Intercomparison Project Phase 5 (CMIP5) under two forcing scenarios at 50 km (12.5 km) resolution [33]. In a European context, EURO-CORDEX [34] has recently produced two extensive ensembles of RCM simulations with grid spacings of approximately 12.5 and 50 km, with significant contributions from RCA4 simulations. The EURO-CORDEX RCM ensembles represent the most comprehensive to date and are currently widely used for climate change impact research efforts. While the downscaled climate data have a resolution of 0.11° (~12.5 km), the study area does not exhibit significant spatial climate variability, making this resolution appropriate. Additionally, the use of observational data from an HNMS meteorological station in close proximity to the study site ensured that the downscaling process captured local climate conditions. This minimizes potential discrepancies between the model outputs and field-scale conditions.

In the fifth phase of the Coupled Model Intercomparison Project [35], many global climate change scenarios have been produced and are extensively used in the Fifth Assessment Report (AR5) from the IPCC (Intergovernmental Panel on Climate Change) [28]. In

Table 2. List of EURO-CORDEX RCA4 models and CMIP5 GCM driving models used in the present study.

RCM Model	GCM Driving Model	GCM Modeling Center	Resolution	Historical Run	Future Run
					RCP4.5	RCP8.5
RCA4(CNRM-CM5)	CNRM-CM5	Centre National de Rech. Météorol.	0.11° × 0.11°	1970–2005	2006–2100	2006–2100
RCA4 (IPSL-CM5A-MR)	IPSL-CM5A-MR	Institut. Pierre-Simon Laplace.	0.11° × 0.11°	1970–2005	2006–2100	2006–2100
RCA4 (HadGEM2-ES)	HadGEM2-ES	Met Office Hadley Centre	0.11° × 0.11°	1970–2005	2006–2100	2006–2100

, the EURO-CORDEX RCMs used in the study, along with the CMIP5 GCMs that provided boundary conditions for the RCA4 runs, are presented. In this study, the three Regional Climate Models (RCMs) are referred to by the name of the corresponding GCM driving model from which they were derived.

2.3.2. Downscaling Method: Linear Scaling Bias Correction

Although Regional Climate Models (RCMs) are generally preferred over General Circulation Models (GCMs) [32] because they provide more reliable results for climate change impact studies on regional and local scales [36], their outputs can still exhibit significant systematic biases, potentially leading to inaccurate conclusions [37,38,39]. Biases in RCMs are described as long-term average deviations between model simulations and observed data [40]. These biases arise from several sources [32], including boundary errors inherited from GCMs, inadequacies in representing surface properties due to spatial resolution, and errors related to numeric resolutions and internal model parameterization [41,42]. Therefore, post-processing of RCMs outputs through bias correction (BC) methods is a prerequisite step for most climate change impact analyses. Bias correction involves adjusting biased simulated data to align with observations [43]. Many bias correction methods have been developed to correct biases in RCM outputs, ranging from simple scaling approaches to more sophisticated distribution mapping techniques.

In the present study, climate change impacts on maize yield were estimated by developing future scenarios based on future climate predictions. This involved repeating the historical climate over a 21-year cycle (baseline conditions) and variations with respect to two Representative Concentration Pathways (RCPs): RCP4.5 and RCP8.5. The climate change regimes were projected using three Regional Climate Models shown in Table 2. Climate data from each climate model (RCP4.5 and RCP8.5) were extracted at the station location and adjusted using the linear scaling approach of bias correction methods (Equations (1)–(4)). Precipitation was calculated according to Equations (1) and (2), while the calculation of mean, minimum and maximum temperature, mean, minimum and maximum relative humidity, and wind speed was based on Equations (3) and (4). No process was used to project solar radiation, so average annual values from historical data were used in future climate change scenarios.

$${\mathrm{P}}_{\mathrm{h}\mathrm{i}\mathrm{s}\mathrm{t}}^{\mathrm{B}\mathrm{C}}(\mathrm{d})={\mathrm{P}}_{\mathrm{h}\mathrm{i}\mathrm{s}\mathrm{t}}(\mathrm{d})\times \left[{\displaystyle \frac{{\mathsf{\mu }}_{\mathrm{m}}\left({\mathrm{P}}_{\mathrm{o}\mathrm{b}\mathrm{s}}(\mathrm{d})\right)}{{\mathsf{\mu }}_{\mathrm{m}}\left({\mathrm{P}}_{\mathrm{h}\mathrm{i}\mathrm{s}\mathrm{t}}(\mathrm{d})\right)}}\right],$$

(1)

$${\mathrm{P}}_{\mathrm{s}\mathrm{i}\mathrm{m}}^{\mathrm{B}\mathrm{C}}(\mathrm{d})={\mathrm{P}}_{\mathrm{s}\mathrm{i}\mathrm{m}}(\mathrm{d})\times \left[{\displaystyle \frac{{\mathsf{\mu }}_{\mathrm{m}}\left({\mathrm{P}}_{\mathrm{o}\mathrm{b}\mathrm{s}}(\mathrm{d})\right)}{{\mathsf{\mu }}_{\mathrm{m}}\left({\mathrm{P}}_{\mathrm{h}\mathrm{i}\mathrm{s}\mathrm{t}}(\mathrm{d})\right)}}\right],$$

(2)

$${\mathrm{Z}}_{\mathrm{h}\mathrm{i}\mathrm{s}\mathrm{t}}^{\mathrm{B}\mathrm{C}}(\mathrm{d})={\mathrm{Z}}_{\mathrm{h}\mathrm{i}\mathrm{s}\mathrm{t}}(\mathrm{d})+\left({\mathsf{\mu }}_{\mathrm{m}}\left({\mathrm{Z}}_{\mathrm{o}\mathrm{b}\mathrm{s}}(\mathrm{d})\right)-{\mathsf{\mu }}_{\mathrm{m}}\left({\mathrm{Z}}_{\mathrm{h}\mathrm{i}\mathrm{s}\mathrm{t}}(\mathrm{d})\right)\right),$$

(3)

$${\mathrm{Z}}_{\mathrm{s}\mathrm{i}\mathrm{m}}^{\mathrm{B}\mathrm{C}}(\mathrm{d})={\mathrm{Z}}_{\mathrm{s}\mathrm{i}\mathrm{m}}(\mathrm{d})+\left({\mathsf{\mu }}_{\mathrm{m}}\left({\mathrm{Z}}_{\mathrm{o}\mathrm{b}\mathrm{s}}(\mathrm{d})\right)-{\mathsf{\mu }}_{\mathrm{m}}\left({\mathrm{Z}}_{\mathrm{h}\mathrm{i}\mathrm{s}\mathrm{t}}(\mathrm{d})\right)\right),$$

(4)

where P is the precipitation (mm); Z is the mean, max, and min temperature (°C), max and min relative humidity (%), and wind speed (m s⁻¹); d stands for daily; µ_m is the long-term monthly mean; BC represents bias correction; hist is the raw RCM historical data; obs is the observed data; sim is the raw RCM future data; ${\mathsf{\mu }}_{\mathrm{m}}\left(\mathrm{P}(\mathrm{d})\right)$ is the long-term monthly mean of the observed or raw RCM historical precipitation data (mm); and ${\mathsf{\mu }}_{\mathrm{m}}\left(\mathrm{Z}(\mathrm{d})\right)$ is the long-term monthly mean of the observed or raw RCM historical data: mean, max, and min temperature (°C), max and min relative humidity (%), and wind speed (m s⁻¹).

2.4. CropSyst: Crop Growth Simulation Model

2.4.1. Model Description

Climate change impacts on crop yield can be estimated using experimental data or crop growth simulation models, which present common approaches for predicting future impacts on crop yields. The most common method for analyzing climate change impacts involves running crop growth simulation models under climate change scenarios, which are mainly derived from precipitation and temperature outputs of the climate models used. To serve as an effective tool for assessing climate change impacts, a crop model must accurately predict yield as a function of weather variables and require a relatively small set of parameters and variables. The cropping systems model CropSyst [22] is a process-based, multi-year, multi-crop, daily time-step cropping systems simulation model that has seen increasing use in recent years and has been applied to a wide range of crops across various regions [14,44]. CropSyst has evolved to address new demands in agro-ecosystem simulation, including the combined cycling of carbon and nitrogen, the carbon footprint of the agricultural sector, enhancement of water-use efficiency, differences in physiological processes between C3 and C4 plants, and the assessment of agricultural responses to future climate change [45].

Considerable attention has been given to the development of a user-friendly interface, which provides a weather generator, utility programs, and links to GIS software [20]. CropSyst simulates the soil–water budget, soil–plant nitrogen budget, crop phenology, canopy and root growth, biomass production, crop yield, residue production and decomposition, soil erosion, and soil salinity [22]. The above processes are influenced by weather, soil characteristics, crop characteristics, and cropping system management options, such as crop rotation, cultivar selection, irrigation scheduling, nitrogen fertilization, water salinity, tillage operations, and residue management. The simulation of the water budget includes precipitation, irrigation, runoff, interception, water infiltration and redistribution in the soil profile, crop transpiration, and evaporation. The nitrogen budget involves nitrogen application, nitrogen transport and transformations, ammonium absorption, and crop nitrogen uptake. A significant feature of CropSyst is its incorporation of carbon dioxide (CO₂) dynamics for assessing the impacts of increasing atmospheric CO₂ levels on crop growth and yield. By simulating the interaction between CO₂ concentration and plant physiological processes, the model can evaluate how increased CO₂ influences photosynthetic and transpiration rates, water-use efficiency, and overall biomass production. CO₂ was taken into account through the CropSyst model, considering it as both an initial concentration and an annual rate for each simulation run. Specifically, CO₂ concentration values for the historical period 1980–2000 range from 338.36 to 368.87 ppm. For the periods 2030–2050 and 2080–2100, CO₂ concentrations under RCP4.5 are projected to be between 435.05 and 486.54 ppm and between 531.14 and 538.36 ppm, respectively. The RCP8.5 scenario indicates CO₂ concentrations of 448.84 to 540.54 ppm for 2030–2050 and significantly higher values of 758.18 to 935.87 ppm for 2080–2100. The above concentrations of CO₂ were derived from projections based on RCP scenarios. Figure 2 presents a flowchart describing the methodology employed in CropSyst for crop yield simulation.

Daily biomass accumulation is calculated using two approaches: (i) radiation-dependent biomass growth as a function of photosynthetically active radiation intercepted by the crop, and (ii) transpiration-dependent biomass growth based on potential crop transpiration. Therefore, potential daily biomass production according to the first approach can be calculated as [22,46]:

$$B_{\mathrm{P}\mathrm{T}}={\displaystyle \frac{{\mathrm{K}}_{\mathrm{B}\mathrm{T}}\times {\mathrm{T}}_{\mathrm{p}}}{\mathrm{V}\mathrm{P}\mathrm{D}}},$$

(5)

where B_PT is the crop potential transpiration dependent on aboveground biomass production (kg m⁻² day⁻¹); K_BT is an aboveground biomass-transpiration coefficient (kPa); T_P is the crop potential transpiration (kg m⁻² day⁻¹); and VPD is the daily mean atmospheric vapor pressure deficit (kPa).

The second approach of unstressed biomass production is based on Monteith [47]:

$$B_{\mathrm{I}\mathrm{P}\mathrm{A}\mathrm{R}}=RUE\times \mathrm{I}\mathrm{P}\mathrm{A}\mathrm{R}\times {\mathrm{T}}_{\mathrm{l}\mathrm{i}\mathrm{m}},$$

(6)

where B_IPAR is the daily aboveground biomass production dependent on the intercepted photosynthetic radiation-PAR (kg m⁻² day⁻¹); RUE is the radiation use efficiency (kg MJ⁻¹); IPAR is the daily amount of crop-intercepted photosynthetically active radiation (MJ m⁻² day⁻¹); and T_lim is the temperature limitation factor.

The increase in leaf area throughout the vegetative phase, expressed as leaf area per unit soil area (leaf area index—LAI), is estimated based on biomass accumulation [22]:

$$\mathrm{L}\mathrm{A}\mathrm{I}={\displaystyle \frac{\mathrm{S}\mathrm{L}\mathrm{A}\times B}{1+\mathrm{p}\times B}},$$

(7)

where LAI is the green leaf area index (m² m⁻²); B is the accumulated aboveground biomass (kg m⁻²); SLA is the specific leaf area (m² kg⁻¹); and p is a leaf/stem partition coefficient (m² kg⁻¹).

Crop yield can be determined by multiplying the total cumulative biomass at physiological maturity (BPM) by the unstressed harvest index (HI).

$$\mathrm{Y}=B_{\mathrm{P}\mathrm{M}}\times \mathrm{H}\mathrm{I},$$

(8)

$$\mathrm{H}\mathrm{I}={\displaystyle \frac{\mathrm{H}\mathrm{a}\mathrm{r}\mathrm{v}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{b}\mathrm{l}\mathrm{e} \mathrm{Y}\mathrm{i}\mathrm{e}\mathrm{l}\mathrm{d}}{\mathrm{A}\mathrm{b}\mathrm{o}\mathrm{v}\mathrm{e} \mathrm{G}\mathrm{r}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{d} \mathrm{B}\mathrm{i}\mathrm{o}\mathrm{m}\mathrm{a}\mathrm{s}\mathrm{s}}},$$

(9)

where Y is the crop yield (kg m⁻²) and B_PM is the total biomass accumulated at physiological maturity (kg m⁻²). The harvest index, derived from experimental data, can be adjusted by users during the calibration process considering sensitivity to water and nitrogen stress during flowering and/or grain filling.

2.4.2. CropSyst Calibration and Validation Procedures

The CropSyst model was calibrated and validated using the data obtained from the field experiments in 2016 and 2017, respectively, for full irrigation. Calibration and validation were carried out through a repetitive process that utilized the measured crop growth variables, observed phenological stages, parameters estimated from available data, derived growing coefficients, and parameters from previous studies that tested the model [48]. Initially, soil, weather, and irrigation input files were prepared. Subsequently, measured and estimated crop parameters were input into the CropSyst model. The final calibration and validation phase included refining other parameters to ensure that the simulated values fit well with the observed data. The parameters were manually adjusted within a range of default values to achieve the best possible fit with the measured data.

Input files required by the CropSyst model for the location and maize crop (including crop growth, morphology, and phenology) were prepared and used to run the model. Crop growth parameters included the aboveground biomass-transpiration coefficient, light to aboveground biomass conversion factor, optimum mean daily temperature for growth, maximum water uptake, actual to potential transpiration ratio that limits leaf area growth, actual to potential transpiration ratio that limits root growth, etc. Crop morphology parameters included initial green LAI, maximum expected LAI, fraction of LAI at physiological maturity, specific leaf area, maximum rooting depth, stem/leaf partitioning coefficient, leaf duration, leaf duration sensitivity to water stress, extinction coefficient for solar radiation, and crop k_c at full canopy. Total biomass and grain yield at the end of the cultivation period, along with leaf area index (LAI) and aboveground biomass (AGB) during the growing season, were utilized to calibrate and validate the model. Crop phenology parameters involved degree days at emergence, peak LAI, flowering, and maturity, as well as base and cut-off temperatures and the phenological sensitivity coefficient for water stress. Growing degree days for specific phenological stage were estimated based on the dates for each stage.

The crop input parameters’ values were either obtained from the CropSyst manual [49] and other research studies [50] or adjusted to match those observed in the experiments. The calibration and validation involved minor adjustments to the maize input parameters to ensure accurate simulations. The parameters calibrated and validated were specific leaf area (SLA, m² kg⁻¹), stem/leaf partition coefficient (SLP, m² kg⁻¹), unstressed harvest index (HI, 0–1), aboveground biomass–transpiration coefficient (K_BT, Pa), mean daily temperature that limits early growth (°C), leaf duration (degree days), and radiation -use efficiency (RUE, g MJ⁻¹).

2.5. Model’s Statistical Evaluation Criteria

The CropSyst model was evaluated by comparing the observed/measured and simulated/predicted values of Leaf Area Index (LAI) and aboveground biomass over the two growing seasons, which were also used for the model’s calibration and validation. Specifically, the performance of the model’s simulation was assessed through different statistical criteria (Equations (10)–(15)), including the coefficient of determination (R²), the root mean square error (RMSE), the mean absolute error (MAE), the mean bias error (MBE), Willmott’s index of agreement (d), and modeling efficiency (EF).

The coefficient of determination (R²) describes the extent to which the observed dispersion is described by the simulation. R² values range from 0 to 1, where a value of 1 indicates that the variance in what is measured is described by the model’s simulation, while a value of 0 indicates no correlation between the observed and simulated values.

$${\mathrm{R}}^2={\displaystyle \frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}\left({\mathrm{O}}_{\mathrm{i}}-\overline{\mathrm{O}}\right)\left({\mathrm{S}}_{\mathrm{i}}-\overline{\mathrm{S}}\right)}}{\sqrt{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{O}}_{\mathrm{i}}-\overline{\mathrm{O}}\right)}^2}}\sqrt{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{S}}_{\mathrm{i}}-\overline{\mathrm{S}}\right)}^2}}}},$$

(10)

where n is the number of observations; O_i and S_i are the observed and simulated values; and $\overline{\mathrm{O}}$ and $\overline{\mathrm{S}}$ are the observed and simulated mean values, respectively.

The root mean square error (RMSE) indicates the variance of errors and ranges from 0 to positive infinity [51], with model performance improving as RMSE approaches zero.

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{{\displaystyle \frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{O}}_{\mathrm{i}}-{\mathrm{S}}_{\mathrm{i}}\right)}^2}}{\mathrm{n}}}},$$

(11)

Mean absolute error (MAE) expresses the average magnitude of the errors between simulated and observed values, irrespective of their direction.

$$\mathrm{M}\mathrm{A}\mathrm{E}={\displaystyle \frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}\left|{\mathrm{O}}_{\mathrm{i}}-{\mathrm{S}}_{\mathrm{i}}\right|}}{\mathrm{n}}},$$

(12)

The mean bias error (MBE) measures the overall bias error or systematic error between simulated and observed values and indicates the under or overestimations by the model being below or above zero, respectively.

$$\mathrm{M}\mathrm{B}\mathrm{E}={\displaystyle \frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}\left({\mathrm{S}}_{\mathrm{i}}-{\mathrm{O}}_{\mathrm{i}}\right)}}{\mathrm{n}}},$$

(13)

The index of agreement (d) [52] represents the degree to which the observed data approached the simulated data with values ranging from 0 to 1. A value of 1 indicates perfect agreement between observed and simulated data, whereas a value of 0 signifies no agreement.

$$\mathrm{d}=1-{\displaystyle \frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{O}}_{\mathrm{i}}-{\mathrm{S}}_{\mathrm{i}}\right)}^2}}{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left(\left|{\mathrm{S}}_{\mathrm{i}}-\overline{\mathrm{O}}\right|+\left|{\mathrm{O}}_{\mathrm{i}}-\overline{\mathrm{O}}\right|\right)}^2}}},$$

(14)

A Nash–Sutcliffe model efficiency coefficient (NSE) value of 1 indicates a perfect model performance, while a value of 0 shows the observed mean value is a good predictor of the model. Negative NSE values suggest that the average of the observed values is a better predictor than the model.

$$\mathrm{N}\mathrm{S}\mathrm{E}=1-{\displaystyle \frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{O}}_{\mathrm{i}}-{\mathrm{S}}_{\mathrm{i}}\right)}^2}}{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{O}}_{\mathrm{i}}-\overline{\mathrm{O}}\right)}^2}}},$$

(15)

3. Results

3.1. Projected Response of Climate Parameters

To develop future projections of climate variables for the climate change periods 2030–2050 and 2080–2100, climate data taken from three RCMs under the RCP4.5 and RCP8.5 climate change scenarios were used. Due to systematic deviations (biases) between the observed and RCM climate variables, bias correction was applied to the RCM climate data. Specifically, linear scaling was selected for bias correction and applied to all climate variables except solar radiation. Indicatively, the datasets of precipitation and mean temperature for the historical period before and after bias correction are presented in Figure 3. Figure 3 shows that the simulated CNRM-CM5 data related to daily precipitation were overestimated compared to observed data during January, February, March, September, October, and December and underestimated during the remaining months. The simulated precipitation related to RCMs IPSL-CM5A-MR and HadGEM2-ES was underestimated for most months of the year. Simulated RCMs underestimated mean temperature for all months except July and August regarding CNRM-CM5 and IPSL-CM5A-MR and June, July, and August for the HadGEM2-ES model. After applying bias correction to both precipitation and temperature, the simulated means were improved, showing a good fit with the observed data.

The observed changes in mean annual temperature as well as during the cultivation period, after bias correction, under RCP4.5 and RCP8.5 based on the CNRM-CM5, IPSL-CM5A-MR, and HadGEM2-ES models during the 2030–2050 and 2080–2100 periods relative to the baseline period (1980–2000), are shown in

Table 3. Differences in annual and cultivation period mean temperature (°C) according to CNRM-CM5, IPSL-CM5A-MR, and HadGEM2-ES under RCP4.5 and RCP8.5 during the 2030–2050 and 2080–2100 periods in relation to 1980–2000.

Mean Temperature (°C)	Historical	RCP4.5		RCP8.5
	1980–2000	2030–2050	2080–2100	2030–2050	2080–2100
		Climate model: CNRM-CM5 (after bias correction)
Tmean (°C)	15.31	16.20	17.20	16.38	18.86
ΔTmean (°C)		0.89	1.89	1.07	3.55
Tmean (°C) (cultiv. period)	21.76	22.81	23.96	23.05	26.08
ΔTmean (°C) (cultiv. period)		1.06	2.21	1.30	4.32
		Climate model: IPSL-CM5A-MR (after bias correction)
Tmean (°C)	15.31	16.96	18.04	17.19	21.07
ΔTmean (°C)		1.65	2.73	1.88	5.76
Tmean (°C) (cultiv. period)	21.76	23.83	24.99	24.26	28.39
ΔTmean (°C) (cultiv. period)		2.07	3.23	2.50	6.63
		Climate model: HadGEM2-ES (after bias correction)
Tmean (°C)	15.31	17.28	17.35	17.52	19.59
ΔTmean (°C)		1.97	2.04	2.21	4.28
Tmean (°C) (cultiv. period)	21.76	24.44	24.58	24.64	26.99
ΔTmean (°C) (cultiv. period)		2.68	2.83	2.88	5.24

. Projections suggested a positive shift in the future thermal regime of the study area, both at annual and cultivation period scales. However, the magnitude of the temperature change was different depending on the climate model, RCP scenario, and climate change period.

Predicted changes in mean temperature are expected to be higher during the 2080–2100 future period in relation to 2030–2050 under the RCP4.5 and RCP8.5 scenarios, with the latter showing a greater increase between the two periods. This difference is because RCP4.5 represents a moderate stabilization pathway, in which emissions peak around 2040 and then stabilize by 2100, while RCP8.5 describes a continuous increase in emissions throughout the 21st century. Analysis of the Regional Climate Models used showed that HadGEM2-ES projections were the warmest among the three models, with the exceptions of RCP4.5 and RCP8.5 during the 2080–2100 period, while projections from the CNRM-CM5 model showed the smallest temperature increase. The medium scenario RCP4.5 predicts an increase in mean annual temperature varying from 0.89 °C to 1.97 °C during the 2030–2050 period according to RCMs CNRM-CM5 and HadGEM2-ES and from 1.89 °C to 2.73 °C during the 2080–2100 period for CNRM-CM5 and IPSL-CM5A-LR, respectively. Increases under high-emission scenario RCP8.5 are expected to be 1.07 °C (CNRM-CM5) to 2.21 °C (HadGEM2-ES) for the climate change period 2030–2050 and 3.55 °C (CNRM-CM5) to 5.76 °C (IPSL-CM5A-LR) for the 2080–2100 period. In terms of mean temperature during the cultivation period, the predicted increases of the models for both the 2030–2050 and 2080–2100 periods are expected to be higher compared to the annual temperature. The highest value with respect to an increase in cultivation period temperature, as estimated during the 2080–2100 period under RCP8.5, is 6.63 °C, according to the IPSL-CM5A-MR model.

In

Table 4. Differences in annual and cultivation period mean precipitation (mm) according to CNRM-CM5, IPSL-CM5A-MR, and HadGEM2-ES under RCP4.5 and RCP8.5 during the 2030–2050 and 2080–2100 periods in relation to 1980–2000.

Precipitation (mm)	Historical	RCP4.5		RCP8.5
	1980–2000	2030–2050	2080–2100	2030–2050	2080–2100
		Climate model: CNRM-CM5 (after bias correction)
Pr (mm)	419.28	515.16	524.69	466.33	465.45
Pr change (%)		22.87%	25.14%	11.22%	11.01%
Pr (mm) (cultiv. period)	116.07	138.29	143.46	131.39	123.43
Pr change (%) (cultiv. period)		19.51%	23.60%	13.21%	6.35%
		Climate model: IPSL-CM5A-MR (after bias correction)
Pr (mm)	419.28	442.15	352.98	458.12	376.68
Pr change (%)		5.45%	−15.81%	9.26%	−10.16%
Pr (mm) (cultiv. period)	116.07	103.02	91.22	114.51	58.11
Pr change (%) (cultiv. period)		−11.24%	−21.41%	−1.34%	−49.93%
		Climate model: HadGEM2-ES (after bias correction)
Pr (mm)	419.28	416.91	415.90	468.64	392.13
Pr change (%)		−0.57%	−0.81%	11.77%	−6.47%
Pr (mm) (cultiv. period)	116.07	95.21	101.39	117.06	80.46
Pr change (%) (cultiv. period)		−17.97%	−12.65%	0.86%	−30.68%

, the varying changes in annual and cultivation period mean precipitation (%) under RCP4.5 and RCP8.5, according to the CNRM-CM5, IPSL-CM5A-MR, and HadGEM2-ES models during the 2030–2050 and 2080–2100 periods relative to 1980–2000, are depicted. It can be noted that the projected changes in annual precipitation vary among the models, resulting in both decreases and increases with different magnitudes of change. The CNRM-CM5 model predicts a rise in annual precipitation for both future periods ranging from 11.01% under RCP8.5 to 25.14% under RCP4.5 during the 2080–2100 period. With respect to IPSL-CM5A-MR, annual precipitation is predicted to increase (5.45–9.26%) in the middle of the century (2030–2050) and decrease (10.16–15.81%) during the late 21st century (2080–2100). The HadGEM2-ES model projections show decreases except in the case of the RCP8.5 scenario for 2030–2050, where annual precipitation is expected to increase by 11.77%.

In terms of mean precipitation during the cultivation period, the IPSL-CM5A-MR and HadGEM2-ES models project decreases for almost all cases, while CNRM-CM5 shows a rise in precipitation ranging from 6.35% to 23.60%. Greater decreases in precipitation are predicted by the IPSL-CM5A-MR model compared to HadGEM2-ES (except for RCP4.5 during the 2030–2050 period), with the highest decrease of 49.93% observed under RCP8.5 for the late 21st century. HadGEM2-ES projections for mean precipitation during the cultivation period range from reductions of 12.65 to 30.68% to a slight increase of 0.86% according to RCP8.5 for the 2030–2050 period.

Analysis of projected atmospheric CO₂ concentrations, as shown in

Table 5. Projected changes in atmospheric CO₂ concentrations (ppm) according to RCP4.5 and RCP8.5 for the 2030–2050 and 2080–2100 periods relative to 1980–2000.

CO2 Concentrations (ppm)	Historical	RCP4.5		RCP8.5
	1980–2000	2030–2050	2080–2100	2030–2050	2080–2100
CO2 (ppm)	353.07	460.80	534.16	491.31	845.70
CO2 change (%)		30.51%	51.29%	39.16%	139.53%

, indicates a substantial increase under both RCP4.5 and RCP8.5 scenarios relative to the historical period (1980–2000). Under RCP4.5, CO₂ concentrations are projected to increase by 30.51% during the 2030–2050 period and by 51.29% by the late 21st century (2080–2100). With respect to RCP8.5, the projected increase is more pronounced, reaching 39.16% for 2030–2050 and continuing to escalate, with concentrations rising to 845.70 ppm by 2080–2100. This corresponds to an increase of approximately 139.53% relative to the historical baseline. The results highlight the significant divergence between the two scenarios, with RCP8.5 reflecting a trajectory of continued high emissions, while RCP4.5 suggests a more moderate stabilization of CO₂ levels in the latter half of the century.

3.2. Calibration and Validation of CropSyst

Calibration and validation were conducted using field data, with values iteratively assigned to specific CropSyst variables to successfully simulate the main crop growth parameters. Crop parameter values for simulating maize growth and development by CropSyst were derived by estimation from the experimental field data, the recent bibliography, using default values of CropSyst, direct field measurements and observations, and calibration of the model. The model’s calibration and validation steps involved the input of several parameters, including crop growth, phenology, morphology, transpiration, and root growth.

For the evaluation of the goodness of fit between observed and simulated values of the above parameters, the statistical criteria of the coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE), mean bias error (MBE), Willmott’s index of agreement (d) and the modeling efficiency (EF) were used. In Figure 4, the observed and simulated values of the leaf area index and aboveground biomass are presented.

The leaf area index progression is influenced by the SLA, SLP, and AGB parameters. In accordance with the AGB and LAI experimental field measurements during the cultivation periods of 2016 and 2017, the SLA and SLP values were calibrated and validated. CropSyst-simulated LAI values during the cultivation period showed strong agreement with the experimental data (Table 5). R² values were relatively high (0.97), indicating a good correlation between measured and predicted data. RMSE values for LAI were 0.39 m² m⁻² and 0.40 m² m⁻² for 2016 and 2017, respectively, while MAE values were 0.27 m² m⁻² and 0.37 m² m⁻². The index of agreement (d) between the observed and simulated green leaf area index confirmed the model’s performance.

The mean field-measured aboveground biomass-transpiration coefficient (K_BT) and the average measured radiation use efficiency (RUE) were calibrated and validated. The RMSE and MAE for aboveground biomass estimation were 1.96 tn ha⁻¹ and 1.53 tn ha⁻¹, respectively, during the 2016 cultivation period. For the 2017 cultivation period, RMSE and MAE values were 0.94 tn ha⁻¹ and 0.77 tn ha⁻¹, respectively. Willmott’s index of agreement was 0.99 (2016) and 1.00 (2017), showing an accurate agreement between measured and predicted data, with R² values near 1.00 (

Table 6. Statistical comparison between the observed and simulated leaf area index and aboveground biomass for the calibration and validation periods.

Cultivation Period	Crop Parameters	Statistical Criteria
		R2	RMSE	MAE	MBE	d	NSE
2016	Leaf Area Index	0.97	0.39 1	0.27 1	−0.16 1	0.99	0.97
	Aboveground Biomass	0.99	1.96 2	1.53 2	−1.53 2	0.99	0.96
2017	Leaf Area Index	0.97	0.40 1	0.37 1	0.08 1	0.99	0.97
	Aboveground Biomass	0.99	0.94 2	0.77 2	0.55 2	1.00	0.99

R²: coefficient of determination; RMSE: root mean square error; MAE: mean absolute error; MBE: mean bias error; d: Willmott’s index of agreement; NSE: modeling efficiency; ¹ m² m⁻²; ² tn ha⁻¹.

). CropSyst underestimated observed aboveground biomass by 1.53 tn ha⁻¹ during the calibration year and overestimated it by 0.55 tn ha⁻¹ during the validation year. Nash–Sutcliffe model efficiency coefficient (NSE) values, which approached 1 for both green leaf area index and aboveground biomass, indicated a very good model performance for both calibration and validation years.

The observed and simulated total aboveground biomass (kg ha⁻¹) and yield (kg ha⁻¹) at the end of the growing seasons of 2016 and 2017 are shown in Figure 5. For the calibration year 2016, the observed yield was 13,450 kg ha⁻¹ and the observed total biomass was 27,654 kg ha⁻¹, while the simulated yield was 14,357 kg ha⁻¹ and the simulated total biomass was 29,577 kg ha⁻¹. The percentage differences between the observed and simulated total biomass and maize yield were both 7% for the 2016 calibration year. For the 2017 validation year, the observed yield was 14,965 kg ha⁻¹ and the observed total biomass was 31,250 kg ha⁻¹, with the simulated yield at 14,007 kg ha⁻¹ and the simulated total biomass at 28,818 kg ha⁻¹. The percentage differences between observed and simulated total biomass and yield were 8% and 7%, respectively, with the simulated values being lower compared to the observed.

3.3. Simulation of Yield Response to Climate Change Under Full Irrigation

The CropSyst model was run to predict maize yield for the baseline period (1980–2000) and the 2030–2050 and 2080–2100 climate change periods under two RCP scenarios (RCP4.5 and RCP8.5) using three RCMs. Table 6 presents the differences in mean maize yield for the two future periods relative to the baseline period. The simulated results show the climate change impacts on mean maize yield in the future, changing the full irrigation amounts through the auto-irrigation option provided by the CropSyst model according to the plant water requirements to achieve full irrigation under future conditions.

The RCMs CNRM-CM5 and IPSL-CM5A-MR produced yield increases for both the future climate change periods relative to the baseline period of 1980–2000, while negative estimates were produced by the above models under RCP8.5 for 2080–2100 (Table 6). The HadGEM2-ES model projected yield decreases for the climate change periods, except for RCP4.5 during the 2030–2050 period, where a 3.1% increase was projected. The highest reduction in maize yield was projected by the HadGEM2-ES model under the high-emission scenario (RCP8.5) during the long-term period (2080–2100).

Regarding CNRM-CM5, a positive effect on yield was estimated under RCP4.5, giving an increase of 2.3% during the 2030–2050 period and 4.0% during the 2080–2100 period, while under the RCP8.5 scenario, there was a higher increase (9.6%) for 2030–2050. For the IPSL-CM5A-MR model, the projections showed crop yield increases (except for RCP8.5 for 2080–2100), with the highest increase of 5.8% observed under RCP4.5 during the 2030–2050 future period. The HadGEM2-ES model produced yield decreases in most cases, ranging from 0.1% to 15.8% under RCP8.5 for 2030–2050 and 2080–2100, respectively.

According to

Table 7. Differences in maize yields (kg ha⁻¹) according to CNRM-CM5, IPSL-CM5A-MR, and HadGEM2-ES under RCP4.5 and RCP8.5 during the 2030–2050 and 2080–2100 periods in relation to 1980–2000.

Historical		RCP4.5				RCP8.5
1980–2000		2030–2050		2080–2100		2030–2050		2080–2100
Irrigation (mm)	Yield (kg ha−1)	Irrigation (mm)	Yield (kg ha−1)	Irrigation (mm)	Yield (kg ha−1)	Irrigation (mm)	Yield (kg ha−1)	Irrigation (mm)	Yield (kg ha−1)
		Climate model: CNRM−CM5 (after bias correction)
466	16,542	460	16,915	461	17,197	459	18,137	417	15,425
Change (%)		−1.3%	2.3%	−1.1%	4.0%	−1.6%	9.6%	−10.5%	−6.8%
		Climate model: IPSL−CM5A−MR (after bias correction)
466	16,542	473	17,499	487	16,947	469	17,014	558	14,150
Change (%)		1.4%	5.8%	4.3%	2.5%	0.5%	2.9%	19.7%	−14.5%
		Climate model: HadGEM2−ES (after bias correction)
466	16,542	473	17,059	501	15,933	452	16,519	581	13,933
Change (%)		1.5%	3.1%	7.5%	−3.7%	−3.1%	−0.1%	24.7%	−15.8%

, irrigation amounts, as provided by CropSyst model, will decrease in the case of the CNRM-CM5 model for both scenarios and climate change periods. These decreases are slight in most cases, while a 10.5% decrease is recorded under RCP8.5 during the long-term period. With respect to the IPSL-CM5A-MR model, irrigation is projected to increase, with the greatest increase of 19.7% under RCP8.5 during the 2080–2100 period. Increases in full irrigation amounts are shown according to HadGEM2-ES, except for RCP8.5 during the 2030–2050 period, where a decrease of 3.1% is projected. The greatest increase in full irrigation amounts, at 24.7%, among the three models is provided by the HadGEM2-ES model under RCP8.5 for the long-term climate change period of 2080–2100.

Box plots of maize yield change (kg ha⁻¹) under RCP4.5 and RCP8.5, as simulated by the three different RCMs, are presented in Figure 6. In each box plot, the midline indicates the median values and the lower and upper boundaries denote the 25th percentile and 75th percentiles, respectively. The whiskers of the box plots, at the top and bottom, indicate the maximum and minimum observed values, respectively. Variations in the heights of the rectangles across the box plots reflect differences in data spread among the three RCMs, indicating significant variation. The decrease in maize yield, predicted by CNRM-CM5 and IPSL-CM5A-MR under RCP8.5 for 2080–2100 and by HadGEM2-ES under both scenarios for 2080–2100 is obvious according to the box plots as their means and medians have negative values. The largest spread in the mean yield distribution is observed in the cases of CNRM-CM5 and IPSL-CM5A-MR models under RCP8.5 during the 2030–2050 and 2080–2100 periods, respectively. According to CNRM-CM5 under RCP4.5 for both periods, the medians are generally close to the means (dotted lines), with the mean of each box plot positioned in the middle of the rectangle. Furthermore, the relatively short box plots for each dataset indicate a high level of agreement. The box plots show that the maximum yield value is observed according to CNRM-CM5 under RCP8.5, while the minimum yield value is observed according to HadGEM2-ES under RCP4.5 during the 2030–2050 climate change period. With respect to the 2080–2100 period, the maximum observed values are given by the CNRM-CM5 and IPSL-CM5A-MR models under RCP4.5 and the minimum values are given by IPSL-CM5A-MR under the RCP8.5 scenario.

Monthly irrigation amounts for maize during the mid-term and long-term climate change periods under the RCP scenarios according to the three RCMs are presented in Figure 7. Maize irrigation amounts were predicted to change due to future climate alterations, with variations contingent upon the month. As documented, the uncertainty in monthly irrigation increases further into the future, showing a broader range of possibilities influenced by the selected Regional Climate Model (RCM) and RCP scenario. Maize irrigation amounts are highest in July and August, with the lowest requirements recorded in May and September. Irrigation amounts are projected to increase in June and July for both scenarios and climate change periods, while a decrease in monthly averages is expected for August. Furthermore, irrigation increases are observed in May under RCP8.5 during the 2080–2100 period, as projected by the three RCMs. However, there is no shift in the month (July) where the peak irrigation amount is observed. Changes in irrigation amounts, whether increases or decreases, depend on the selected RCM and RCP scenario. These changes are more pronounced during the 2080–2100 period compared to 2030–2050, reflecting greater temperature increases and more pronounced precipitation changes simulated by the CropSyst model for the long-term climate change period. Irrigation shifts are greater moving further into the future, with the greatest changes observed according to the high-emission scenario RCP8.5 for 2080–2100. In most cases, uncertainty increases further into the future, as indicated by the wider range between the first and third quartiles and more extended whiskers/outliers.

Figure 8 illustrates the annual percentage changes (%) in maize yield under RCP4.5 and RCP8.5, as projected by (a) CNRM-CM5, (b) IPSL-CM5A-MR, and (c) HadGEM2-ES, during the near future (2030–2050) and distant future periods of climate change (2080–2100). As shown in the figure, variations in maize yield changes are evident, with projections indicating both positive and negative impacts within each climate change period under full irrigation. Furthermore, there were differences in crop yield change among the three Regional Climate Models, showing great variation between the two different climate change scenarios and future periods. As projected by the CNRM-CM5 model, maize yield is projected to increase for the majority of the 21 years during the medium-term period, whereas in the long-term period, the number of years with increases and decreases is similar. The greatest yield increase, at approximately 32%, is projected for 2030 under RCP8.5, while the greatest decrease, at about 30%, is expected in 2097 under the high-emission scenario. With respect to IPSL-CM5A-MR, more years show yield increases under both scenarios during the 2030–2050 period and under RCP4.5 during the 2080–2100 period. The above model predicts negative impacts on maize yield for most of the 21 years of the 2080–2100 period under RCP8.5. According to IPSL-CM5A-MR, the greatest increases, exceeding 20%, were recorded in 2030 (RCP8.5) and 2032 (RCP4.5), while the highest decline (approximately 60%) was recorded in 2098 (RCP8.5). For the HadGEM2-ES model, more years show yield increases than decreases during the 2030–2050 period under RCP4.5, whereas in the long-term period, yield is projected to decrease for most years under both emission scenarios. The above model projects the highest yield increase (28%) in 2030 under the RCP4.5 scenario and the greatest decline, at approximately 45%, in 2097 under RCP8.5.

Figure 9 illustrates the projected annual full irrigation change (%) under RCP4.5 and RCP8.5, as estimated by the three RCMs, during the 2030–2050 and 2080–2100 periods. The future irrigation changes show variations within each climate change period, with the CNRM-CM5 model giving the greatest variations with respect to the magnitude of change. The greatest increase, at 36%, is recorded for 2030, while the highest decreases, at approximately 50%, are expected in 2092 and 2099 under RCP8.5. According to IPSL-CM5A-MR, full irrigation is projected to increase for most of the 21 years for both scenarios and future periods. For the HadGEM2-ES model, most years in the long-term period under both scenarios show increases, with the greatest increase in irrigation, at 48%, projected for 2096 under RCP8.5.

4. Discussion

4.1. Projected Response of Climate Parameters

In this study, the projected response of mean annual temperature to climate change, according to the RCP scenarios, is an increase during both the mid-century and the end of the century. The slightest increase (0.89 °C) is recorded under RCP4.5 during the 2030–2050 period, while the highest increase (5.76 °C) is projected according to RCP8.5 for 2080–2100. According to the RCPs, the global mean temperature is predicted to rise by 1 °C to 3.7 °C by the late 21st century (2081–2100) compared to the 1986–2005 baseline period, as reported in the IPCC Fifth Assessment Report (AR5) [28]. A study by Voloudakis et al. [15], conducted in different areas of Greece, predicted temperature increases ranging from 1.5 °C to 5.0 °C, depending on the climate change scenario and the climate change period. Todaro et al. [53] also identified similar trends in temperature increases under the RCP4.5 and RCP8.5 scenarios. Their study indicated that temperatures have exhibited an upward trend in Mediterranean regions, with warming being particularly pronounced under the RCP8.5 scenario. For Greece, temperature increases are projected to range from 1.0 °C to 2.2 °C under RCP4.5 and from 1.22 °C to 4.29 °C under RCP8.5 in the short- and the long-term, respectively. These findings are consistent with those of Zittis et al. [54], who projected a temperature increase of 1.0 °C to 5.0 °C in the Mediterranean region by the end of the 21st century.

In terms of precipitation, RCMs predict both decreases and increases, with the CNRM-CM5 model giving the greatest increase of 25.14% under RCP4.5 during the 2080–2100 period and the IPSL-CM5A-MR model giving the highest decrease of 15.81% by the end of the century according to RCP4.5. Both decreases and increases in annual precipitation were predicted in areas of Greece, with the highest reduction of approximately 40% projected during the 2071–2100 period, although increases were projected in some cases [15]. Politi et al. [55] found that, under both RCP4.5 and RCP8.5 scenarios, precipitation is projected to decrease over Greece, with more pronounced reductions under RCP8.5 towards the end of the century. Similar results were found in another assessment in Greece by Georgoulias et al. [56], who indicated a general decrease in precipitation, particularly under the RCP8.5 scenario, with an average reduction of 16% by the end of the century. A study by Faggian [57] highlighted that precipitation changes exhibit complex patterns at the local level, with some areas experiencing increases and others showing decreases, depending on local conditions, posing a challenge for water management and agricultural planning, especially regarding irrigation strategies.

The projected increase in atmospheric CO₂ concentrations under both RCP4.5 and RCP8.5 scenarios highlights a significant divergence in future emission pathways. While RCP4.5 suggests a stabilization of CO₂ levels, RCP8.5 predicts a continued rise due to high emissions. This aligns with global projections that emphasize CO₂’s critical role in driving climate change. The IPCC [28,58] emission scenarios for estimating future CO₂ emissions and their impacts on climate change underscore the importance of limiting CO₂ emissions to avoid severe climate impacts. In particular, high CO₂ emissions persist in Europe, making the region vulnerable to extreme weather events. The findings emphasize that slow progress in mitigation efforts highlights the urgency of strengthening policies and technologies to reduce emissions. Immediate and coordinated action is essential, particularly in vulnerable regions like Europe, to mitigate the risks associated with climate change.

4.2. Calibration and Validation of CropSyst

The CropSyst model was calibrated and validated using experimental data from the maize crop during the 2016 and 2017 growing seasons, respectively. Statistical comparisons between the observed and simulated leaf area index (LAI) and aboveground biomass for the calibration and validation periods confirmed the model’s reliable performance. The percentage difference between the observed and simulated total aboveground biomass (kg ha⁻¹) and yield (kg ha⁻¹) was 7% for the calibration year (2016). For the validation year (2017), the differences between observed and simulated values of total biomass and yield were 8% and 7%, respectively. The CropSyst model has been corroborated for maize at different locations, showing reasonably accurate simulations. Stockle et al. [59] found good agreement between simulated and observed biomass and yield of maize grown at two locations, as indicated by several statistical indicators. A study by Umair et al. [60] suggested that the CropSyst model can be used with a considerable degree of accuracy to simulate the maize grain yield. CropSyst was able to simulate maize yield reasonably well for long-term rotations at two Italian locations, as reported by Donatelli et al. [48].

4.3. Simulation of Yield Response to Climate Change Under Full Irrigation

(a)

Projected maize yield response

The results show changes in maize production in the future due to changes in climate variables, resulting in both reductions and increases depending on the regional climate model and the climate change scenario used. The observed positive impacts of climate change on maize yield recorded in some cases, under full irrigation where water requirements are met, are consistent with the fact that maize, as a C4 plant, has a competitive advantage over C3 plants. This advantage is particularly pronounced under drought, nitrogen, and/or carbon dioxide limitation conditions. Authors such as Maroco et al. [61] have documented the beneficial impacts of elevated ambient CO₂ concentrations on biomass production and yield. Negative responses to climate change were predicted in this study, with varying magnitudes of yield reductions depending on the magnitude of temperature increase. Reductions in maize yield were recorded for temperature increases exceeding 2.8 °C during the cultivation period, reaching decreases of up to 15.8%. Since crop growth is primarily a function of temperature when water is available to optimum satisfaction, an increase in temperature may also affect photosynthesis, resulting in maize growth reduction. The impact of temperature increase on maize’s developmental rate is important, as an increase in ambient temperature, within a temperature range, accelerates development rates, enabling maize to complete its phenological stages in shorter periods [8]. Beyond the upper limit of the optimal temperature range, and up to a maximum temperature, maize exhibits adverse responses, with a decrease in the developmental rate as temperatures rise [62]. Meza et al. [8] concluded that climate change could adversely impact maize yield, with the extent of reductions depending on the intensity of climate change. The relationship between temperature and developmental rate implies that climate change will lead to shorter crop durations. This accelerated developmental rate partially explains the observed changes in crop productivity. Furthermore, it is documented that elevated CO₂ concentrations could lead to increases in water use efficiency [8], primarily due to reduced stomatal conductance, which in turn decreases evaporation rates [63].

The findings of our study indicate that the combined effects of rising temperatures, decreasing precipitation, and elevated atmospheric CO₂ concentrations are intensified, resulting in more substantial yield reductions under the high-emission scenario (RCP8.5) by the late 21st century. Castaño-Sánchez et al. [64] assessed maize yield responses to CO₂ enrichment using three process-based crop models and reported positive effects on crop growth. However, while elevated CO₂ levels may initially enhance crop development, these benefits are expected to be counterbalanced by the concurrent rise in temperature and alterations in precipitation patterns. The overall impact of increased CO₂ and climate change on crop yields will ultimately depend on local conditions [14]. For instance, in northern temperate regions, higher spring–summer temperatures may extend the growing season, potentially benefiting crop yields. Conversely, in Mediterranean climates, where elevated summer temperatures and water stress already limit crop production, further warming is likely to have negative effects [65].

(b)

Projected irrigation water response

Both increases and decreases were projected regarding full irrigation water amounts in the future. Full irrigation amounts decrease in cases where precipitation during the cultivation period is projected to increase, whereas full irrigation increases following the precipitation reductions. Different amounts of full irrigation are required in the future according to the magnitude of temperature and precipitation changes compared to the baseline period. The above changes in full irrigation are of different magnitudes according to the magnitude of precipitation and temperature changes. The results align with the findings of the Fifth Assessment Report, which provides various projections regarding the impact of future climate on crop water requirements. These predictions indicate that, due to climate change, the water required to achieve a specific yield in both rainfed and irrigation systems is expected to rise in many areas. There is high confidence that by the 2080s, irrigation water requirements will increase in numerous regions, although some areas may experience minor reductions, attributable to increased precipitation. Irrigation demand increases of 7% to 21% by the 2080s were predicted by Wada et al. [66], depending on the climate change scenario used. Conversely, minor global reductions in crop water deficits by the 2080s for irrigated and rainfed regions are predicted by Zhang and Cai [67], associated with a reduced difference between daily maximum and minimum temperatures. Gondim et al. [68] concluded that irrigation water requirements are projected to increase (8–9%) by the mid-21st century due to precipitation decreases. Increases in average irrigation requirements were found by Döll [69] and Fischer et al. [70], while global decreases in irrigation water demand were simulated by Pfister et al. [71] and Konzmann et al. [72] for the end of the century. The impact of precipitation and temperature changes on irrigation needs could be exacerbated by higher CO₂ levels, particularly under scenarios of continued high emissions like RCP8.5.

Moreover, this study projected that monthly irrigation amounts would change due to climate change, with increases observed in June and July and decreases in August during the two climate change periods compared to the baseline period. Increases in maize irrigation demand in May, June, and September and decreases in July and August were observed by Woznicki et al. [73], with the changes more pronounced in the late 21st century due to the greater temperature increases.

In conclusion, the primary factor affecting maize yield during future periods seems to be the combined impact of climate variables changes, which alter irrigation water requirements. The negative impacts on crop production, due to the reduction in precipitation during the growing season of maize observed in some cases, seem to be modulated by increasing the full irrigation amounts. However, full irrigation increases cannot mitigate the negative impacts of climate change on crop productivity when substantial temperature increases and precipitation decreases are observed, resulting in significant maize yield reductions. Therefore, it is essential to integrate both climate and CO₂ projections into agricultural planning to ensure the resilience of crop systems under future climate conditions.

5. Conclusions

This study examines the impacts of climate change on maize yield and irrigation water amounts in Northern Greece for the mid-21st century (2030–2050) and late 21st century (2080–2100), using climate projections from the CNRM-CM5, IPSL-CM5A-MR, and HadGEM2-ES climate models under the RCP4.5 and RCP8.5 pathways. Results indicate an increase in temperature during both future periods, with higher increases projected by the end of the century under the RCP8.5 high-emission scenario. Precipitation projections vary among the models, with CNRM-CM5 indicating increases, while the IPSL-CM5A-MR and HadGEM2-ES models project both increases and decreases. Atmospheric CO₂ concentrations are projected to increase significantly under both RCP4.5 and RCP8.5 scenarios, with higher increases under the latter scenario during the late 21st century. Maize yield responses also vary depending on the climate change scenario and climate model, ranging from increases of up to 9.6% (CNRM-CM5) to reductions of 15.8% (HadGEM2-ES). Projected increases in atmospheric CO₂ concentrations may further influence maize yield, with potential benefits from rising CO₂ being counterbalanced by the negative impacts of higher temperatures and altered precipitation patterns. While C4 photosynthetic mechanisms may enhance maize yield under full irrigation, temperature increases beyond 2.8 °C (during the cultivation period) negatively impact maize yield. Our findings suggest that the combined effects of increased temperatures, reduced precipitation, and elevated CO₂ concentrations are amplified under the RCP8.5 high-emission scenario by the late 21st century, resulting in more substantial declines in maize yields. Irrigation water amounts are projected to both increase and decrease with varying magnitudes of change depending on temperature and precipitation changes.

In conclusion, maize yield in future periods will be influenced by the combined changes in climate variables, impacting irrigation water requirements. This study enhances our understanding of climate change’s role in agricultural productivity and provides insights into maize production in Northern Greece under future climate scenarios. The findings support the development of crop- and region-specific agricultural adaptation strategies and sustainable irrigation management practices in the context of climate change. Practical applications include informing local farmers and policymakers about effective water management strategies for mitigating the adverse effects of climate change and enhancing maize production sustainability at the farm level.