A Scrutiny of plasticity management in irrigated wheat systems under CMIP6 Earth system models (case study: Golestan province, Iran)

Global wheat production has faced, and will persist in encountering many challenges. Therefore, developing a dynamic cultivation approach generated through modeling is crucial to coping with the challenges in speci�c districts. The modeling can contribute to achieving global objectives of farmers’ �nancial independence and food security by enhancing the cropping systems. The current study aims to assess the effects of cultivars and sowing windows intricately on irrigated wheat production using the two models from Coupled Model Intercomparison Project Phase 6 (CMIP6), including ACCES-CM2 and HadGEM31-LL under two shared socioeconomic pathways (SSP245, and SSP585). A two-year on-farm experiment was conducted for parametrization and validation of the APSIM-Wheat model at two locations. The model reasonably simulated the days to anthesis, maturity, biomass production, and yield within all cultivars. The normalized root-mean-square error (RMSE) of the phenological stages was simulated and measured values were 5% and 2–4%, while the index of agreement (IOA) was in the range of 0.84–0.88 and 0.95–0.97. An acceptable agreement of the simulated biomass (RMSE = 5–7% and 0.91 − 0.78) and yield (RMSE = 6–11% and IOA = 0.70–0.94) was identi�ed in the model. Afterward, the LARS-WG model generated the baseline (2000–2014) based on the weather data at the sites and projected the models for the near (2030–2049) and remote future (2050–2070). The models revealed that not only the average maximum and minimum temperatures will rise by 1.85°C and 1.62°C which will exacerbate the reference evapotranspiration (ET 0 ), but also the precipitation and solar radiation will reach + 58%, and + 0.25 Mj m − 2 . Our results clearly showed that precipitation volume over the growing seasons would elevate approximately two times as much as the baseline in the future, while there is a signi�cant decrease in water productivity (WP) and yield from the intensive ET 0 . Based on the wheat simulation, the short-duration cultivar (Kalate) combined with the postponed planting (16-Dec) was determined as a practical alternative; nonetheless, both WP and yield signi�cantly decreased by 40% and 7%, respectively (p < 0.05). In conclusion, identifying and analyzing future farming conditions (e.g., agro-climate, soil and crop management data) would provide a perception of the forthcoming scenarios. When applied, this knowledge can potentially mitigate the adverse impacts of climate change on global wheat production.


Introduction
Wheat (Triticum aestivum L.) is recognized as a staple cereal globally, producing approximately 771 Mg ha − 2   to support the global communities (FAO, 2021), which is crucial for food security.In Iran, wheat is extensively cultivated under different agronomic management and weather conditions.Although irrigated wheat farms comprise around 33% of all wheat lands, their production surpasses rainfed counterparts (Ministry of Agriculture-Jahad, 2021).
Agricultural models are developed to simplify the complexities of crop production prediction and boost our understanding of the environment, particularly climate change effects.With these platforms' ongoing expansion and enhancement, a growing and widespread demand for their integration into farming systems has become apparent.Regarding crop growth models, APSIM (Agricultural Production Systems sIMulator) is one among a range of models widely associated with simulating crop development and production.APSIM relies on inputs such as weather, crop and soil data (Chaki et al., 2022;Vogeler et al., 2023).Kumar et al. (2023) recently evaluated APSIM-Wheat, focusing on location, year and management practices.Their results demonstrated commendable outcomes in simulating phenological stages, with a realistic accuracy between simulated and observed biomass production, grain yield, and N uptake.
Additionally, the Long Ashton Research Station Weather Generator (LARS-WG), a well-recognized downscaling model, and its bene ts were underlined compared to other models (Semenov et al., 2002).Several studies have highlighted the simulation of metrological factors using LARS-WG in the simulation of temperature and precipitation worldwide, such as in Sudan and China (Chen et al., 2013;Duan et al., 2014).Further, it was shown by Bayatvarkeshi et al. (2020) that LARS-WG is an effective tool for accurately predicting climatic variables and the outputs facilitate understanding different scenarios to monitor the impacts of climate change on the reference evapotranspiration (ET 0 ).
For years, investigations have been conducted to promote farming systems and increase Iran's wheat grain yield (Houshyar & Esmailpour, 2020;Roozbeh & Rajaie, 2021).Despite the successful enhancement of grain yield through genetic or agronomic management, the effects of climate change that will pose a challenge and potentially hamper these achievements have not been considered.Albeit new wheat cultivars have better yield performance through improved strategies such as enhanced grain-lling rates (Wu et al., 2018) in various climatic conditions.Given the favorable adaptation of modern cultivars to warmer climates, achieving optimal wheat yield in Iranian environmental conditions is feasible (Mohammadi et al., 2015).
The sowing date is another in uential agronomic practice on the growing-degree days (GDD) accumulation, nutrient acquisition, yields and yield components (Liang et al., 2019;Ren et al., 2019).Considering this, climate change is a critical overarching concern confronting cropping systems (Zhao et al., 2017).Nevertheless, food security and nutrition are already threatened by climate, change and these challenges are expected to worsen in the future (IPCC, 2020) through unpredictable rainfall patterns and climatic extremes (Collins et al., 2021;Das et al., 2020).
Wheat farming systems face formidable challenges from high temperatures and restricted precipitation (Del Pozo et al., 2016).A study in Portugal was carried out to reveal the impact of genetic and agronomic adaption on rainfed wheat.The outcomes indicated that utilization of cultivars with the earlier anthesis improved grain yields (26-38%).Additionally, the study suggested a 6-10% grain yield improvement achieved through earlier sowing dates (Yang et al., 2019).Consequently, different breeding and farm management practices can contribute to resilient and robust wheat grain production in future climatic scenarios.
Although several studies acknowledging climate change effects on wheat productivity in Iran (Deihimfard et al., 2018;Koocheki et al., 2022;Moghaddam et al., 2023), none have introduced the Coupled Model Intercomparison Project Phase 6 (CMIP6) in climatic scenarios alongside genetic, and agronomic practices.
Therefore, the current study aims to assess the APSIM-Wheat across different years and regions, evaluate the outcomes of the CMIP6 model on irrigated wheat, monitor weather variables, and identify suitable responses of cultivars and planting windows under future climate conditions in the northern part of Iran.
2 Materials and Methods

Experimental Site
The required data for the crop model was collected from two different sites, the Gorgan and Gonbad Agricultural Research Stations, located at 36 54 , and 37 16 N latitude, and 54 25 , and 55 13 E longitude, respectively.A two-year on-farm experiment was carried out at either site, employing a spilt-plot design and a randomized complete block design (RCBD) with three replications over 2017 and 2019.The main plots featured wheat cultivars, namely "Tirgan," "Meraj," and "Kalate."Sub-plots were designated for different sowing times on 1 November, 16 November, 1 December, 16 December, and 31 December for each growing season or site.Seed density of wheat was set at 350 (plant m − 2 ).Nutrient requirements were applied before planting, namely Urea (100 kg ha − 1 ), Potassium sulfate (50 kg ha − 1 ), and Super-phosphate (100 kg ha − 1 ).An additional Urea dose (50 kg ha − 1 ) was applied at the beginning of the stem elongation phase.Supplementary irrigation was applied at 40 (mm) on 3 April, and 6 May during the rst growth season, and 9 April during the second.Phonological stages were determined based on Zadok's growth scale from emergence to maturity (Zadoks et al., 1974).The wheat produced in each plot was harvested, and samples were transported to the laboratory to measure wheat biomass and grain yield.

Description of APSIM-Wheat Model
Processes of development and growth in crop are simulated by APSIM, a dynamic crop model, which can also be utilized to model relationships among crop and livestock agroecosystems (Holzworth et al., 2014).APSIM-Wheat has expanded to calculate biomass through daily radiation intercepted where water de cit in uences biomass accumulation.Key factors in uencing crop growth in APSIM-Wheat include the canopy leaf area index, which directly correlates with light interception.The increase in leaf area includes the proliferation of leaf number and size each day following the emergence phase.Crop development is primarily affected by daily thermal time, which depends on the daily average of minimum and maximum temperatures.The cardinal temperature required for phenological progress falls between 0 and 34°C, with the exibility to adjust this threshold in APSIM-Wheat.Vernalization is another crucial factor contributing to the development of phenology from emergence to oral initiation.The duration of vernalization concludes if the daily average of the maximum temperature surpasses 30°C.The total vernalization is calculated by summing daily vernalization from germination to oral initiation, with the condition that it is less than 10.This intricate interplay of environmental and biological features within APSIM-Wheat allows for a comprehensive simulation of crop growth dynamics.

Parametrization and Validation of APSIM-Wheat
The calibration and validation of APSIM-Wheat were carried out using APSIM version 7.10 (The APSIM initiative (AI), 2018) for three common wheat cultivars in Golestan province, Iran.The calibration process involved using the R package apsimx (Miguez, 2021), which speci cally adjusted the Radiation-use E ciency (RUE).RUE is a critical parameter representing the e ciency of converting intercepted radiation into biomass in crops.This process is linked to various environmental and agronomic variables, involving converting light energy and CO 2 into biomass (Huang et al., 2016;Sinclair & Muchow, 1999).The calibration process utilized phenological stages, wheat biomass, and grain yield measurements from the 2017-2018 eld experiments.Subsequently, the 2018-2019 on-farm experiment contributed to the validation process.Genetic coe cients speci c to each wheat cultivar were extracted during the parametrization phase, as detailed in Table 1.The data, including average daily minimum (Tmin) and maximum (Tmax) temperature, precipitation (Pre) and solar radiation (Rad), calculated using the sunshine duration method (Ågnström, 1924), were collected from the Gorgan and Gonbad meteorological observatory sites.According to the downscaling method, the GCMs were used to project the near (2030-2049) and remote future (2050-2070) climate under SSP245 and SSP585 scenarios by the delta change factor (DCF) process.This was expanded based on the future and baseline weather data (Semenov & Barrow, 2002).The monthly climatic data utilized for future projections was Pre, Tmin, Tmax, and Rad from the GCMs, as well as the data available at: https://cds.climate.copernicus.eu/cd=dsapp#!/dataset/projections-cmip6?tab=overview which is extracted in ArcGIS 10.3 by using Multidimension instrument.

Calibration and Validation of LARS-WG
LARS-WG v 6.0 was used for calibration and validation.Each site's baseline data (2000-2014) calibrated the model parameters.The performance of LARS-WG 6.0 was evaluated using three statistical criteria introduced in the SITE ANALYZE module, speci cally examining p-values obtained from a similarity test with a 0.05 signi cant level between the generated and observed climatic data.The rst test includes the distributions of seasonal wet/dry series (WDSeries), daily distributions of precipitation (PreD), monthly average of precipitation (PMM), daily distributions of minimum temperature (TminD), monthly average of minimum temperature (TminM), daily distributions of maximum temperature (TamxD), and daily distributions of solar radiation (RadD).These were computed between the generated and observed climatic data by the Kolmogorov-Smirnov (K-S) test.The second t-test analyzes the t between the generated and observed weather variables such as the monthly average of precipitation (PMM), minimum (TminM) and maximum (TmaxM) temperature and solar radiation (RadM).The third test includes an F-test, which computes the monthly variances for the generated and observed precipitation were computed.Following the calibration and validation process, the GENERATOR module, associated with the observed weather data projected future climate change scenarios based on the GCMs' extractions.

Estimation of the Reference Evapotranspiration (ET 0 )
The FAO Penman-Monteith equation (Allen et al., 1998) is the standard method for calculating ET0, but its application requires a broad range of data.In cases where there is a lack of adequate and accurate data, an alternative method is the Hargreaves-Samani (HS) equation (Eq. 1) within the CROPWAT8 software.Studies conducted in Iran, such as those by Rajabi & Babakhani (2018) and Raziei & Pereira (2013), have veri ed the acceptability of the HS method.
Where T represents the average temperature (°C), TR is the difference between Tmin and Tmax (°C), and Ra shows the extraterrestrial radiation (MJ m − 2 day − 1 ).

Scenario Description
Agronomic adaptation strategies were implemented considering different planting dates: standard planting date (16 November), early planting date (1 November), and late planting dates (16 December).These dates were determined based on the baseline and projected climatic scenarios under SSP 245 and 585.Both seed density, and nutrient management were similarly run with on-farm experience, while 80 (mm) irrigation automatically was added at the anthesis phase to avoid water stress, which has drastically contributed to yield loss in APSIM-Wheat (Monteleone et al, 2023).Simulations were conducted for three different wheat varieties to ensure a robust evaluation of the wheat cropping system under the projected weather conditions.

Statistical Analyses
The measured local wheat parameters were statistically evaluated by using the coe cient of determination (R 2 ), root mean square error (RMSE), and minimum normalized RMSE (NRMSE).RMSE, and NRMSE are calculated as follows: Where , and represent the observed and simulated values, respectively.The is the number of observations, , and are the average of simulated and observed values.

Performance of LARS-WG Model
The performance of the LARS-WG model was evaluated at two different stations using three statistical criteria: the Kolmogorov-Smirnov (K-S) test, t-test, and F-test, as shown in RadD showed no statistical difference among results in both sites, the average rows were 0.5 and 1.5 for PMV and RadM (Table 2).Monthly long-term generations and observations of Pre, Tmax, Tmin and Rad are indicated in Fig. 2. Minimal standard deviation and high overlapping of generated and observed average outputs con rmed the rational performance of LARS-WG for both Tmin and Tmax in either site (Fig. 2).Even though the PMV was statistically signi cant in December at Gorgan site, where the average generated and observed precipitation were 52.43 and 68.93 mm (Fig. 2), there was no statistically signi cant difference in PreD and PMM (Table 2).Regarding RadM, the differences between the average generated and observed solar radiation were 23.43 and 22.66 Mj m − 2 in June and 7.54 and 7.89 Mj m − 2 in December at Gonbad, as well as Gorgan; however, the difference was 20.73 and 21.95 Mj m − 2 solely in June (Fig. 2).
Relationships between the simulated and observed biomass, and yield were represented in Figs. (Fig. 5).The APSIM-Wheat accurately calculated yield during calibration, in which the R 2 , RMSE and NRMSE were in the range of 0.86-0.96,292.57-405.77kg ha − 1 and 6-8%, respectively.Thus, the model generally provided a reasonable yield calculation, with the Meraj cultivar exhibiting the lowest performance (IOA = 0.70).

Projecting weather parameters
Meteorological changes (Pre, Rad, Tmin and Tmax) were applied to simulate wheat production under future climate as an input of the APSIM-Wheat model, and the results are shown in two different GCM models (ACCESS-CM2 and HadGEM3) for each location (Figs. 7 and 8).A higher Pre increase than the baseline is anticipated for 2030-2049 (2040s).Moreover, it is expected to observe a more pronounced Pre rise during the spring (115%) and winter (64%) over 2040-2060 (the 2060s) than in the 2040s (Figs. 7 and 8).Henceforth, in anticipation of future Pre, monthly positive shifts resulting in elevated Pre levels can be observed throughout the year, except for July, August, September, and October at the designated stations.Overall, the expected increase in Pre is 56% and 58% at the Gonbad site and 54%, and 64% at the Gorgan site, respectively, based on the ACCESS-CM2 and HadGEM3 models.Regarding the temperature, both Tmax and Tmin will face a rise, with the most minimal and maximal shifts projected under the SSP245 and SSP585 scenarios (Figs. 7 and 8).
Although the highest heat wave average changes in Tmax and Tmin are anticipated to be 3.93 and 3.34°C in October under HadGEM3 SSP585 at the Gorgan site, these changes are projected at 3.32 and 2.34°C in October under ACCESS-CM2 SSP585 at this site (Fig. 7).Similarly, at Gonbad site, Tmax and Tmin considerably rose by 2.86 and 2.12°C in October under ACCESS-CM2 SSP585.However, March and August experienced the most remarkable average changes in Tmax and Tmin under ACCESS-CM2 SSP585, respectively (Fig. 8).Generally, temperature variations on average were more notable in summer, fall, winter and spring (Figs. 7 and 8).
According to Rad, the average changes are expected to remain relatively stable under the ACCESS-CM2 model at both sites (Figs. 7 and 8).However, the HadGEM3 model projects increased variations in Rad at both sites.
April and September are expected to experience the most considerable increases, with 1.78 and 0.56 Mj m − 2 at the Gonbad and Gorgan sites, respectively (Figs. 7 and 8).An overall increase in Rad can be observed seasonally, with more considerable changes occurring in spring, summer, winter, and fall.

Impacts on yield and ET 0
The calculated 14-year baseline and 40-year wheat yield prediction for three cultivars are showed at both sites in Fig. 9.Although the Tirgan, which is a long-duration cultivar, had the highest yield (5836 kg ha − 1 ) surpassing the baseline at Gonbad site, the highest yield was obtained by the Kalate (5708 kg ha − 1 ) at Gorgan site (Fig. 9).
The projected weather outputs demonstrated a consistently signi cant negative impact on wheat yield overall.The extent of this decrease in average yield varied depending on the CGM models, cultivars and locations.
Across all cultivars, the average yield declined in future periods, with the most substantial reduction observed in HadGEM3 (-9%), as shown in Fig. 9.For 2030-2049, the GCMs showed different responses, with a gradual decline in average yield from SSP245 to SSP585 under ACCESS-CM2 but more severe under SSP245 in HadGEM3.For the remote future (2060s), SSP245 and SSP585 projected a gradual decrease in wheat yield for all cultivars, with a more intense reduction under SSP585.In the baseline scenario at the Gonbad site, the optimal cultivar was Tirgan, with a yield reduction percentage ranging from − 11% to -14% in projections, and these declines were statistically signi cant (p < 0.05), as shown in Fig. 9. Similarly, Kalate experienced a slightly lesser decrease in yield, uctuating between − 9% and 14%.
The change of reference evapotranspiration (ET 0 ) in the baseline and climate projection at the sites is illustrated in Fig. 10.The ET0 is expected to increase over the years steadily, but the changing trend is expected to encounter a slightly more signi cant rise under SSP585 than SSP245ea.Under the SSP245 scenario, the ET 0 which was 3.17 mm day − 1 for the baseline at the Gorgan site will increase to 3.62 mm day − 1 and 3.67 mm day − 1 for the near and remote future in ACCESS-CM2.Under the other scenario, there will be increases in the ET 0 of 3.64 mm day − 1 and 3.71 mm day − 1 for the two projected timescales (2040s and 2060s).The ET0 will be in uenced by GCMs, and the highest levels of ET0 will be distributed in HadGEM3 (Fig. 10).In GCM models, the proportion increase of the ET 0 will range from 3.7-6.3%under SSP245-2040s and SSP585-2060s in ACCESS-CM2 at Gonbad site; nonetheless, it will reach 4.9% and 8.3% in the GCM counterpart.

Impacts on phenology and WP
The simulation dates of phenological stages in the baseline and projected future periods at Gonbad (Table 3) and Gorgan (Table 4) locations are demonstrated.At both locations, all cultivars' anthesis and maturity dates will statistically decrease (p < 0.05) compared to the baseline.Notably, under SSP585, the period reduction was more pronounced, with a decrease of 14 days, as opposed to SSP245, which exhibited a slightly smaller reduction of 12 days (Tables 3 and 4).Considering both scenarios, the anthesis phase was calculated to occur 15, 14 and 15 days earlier for Meraj, Kalate, and Tirgan under SSP245-2060s, and 19 days earlier for all varieties under SSP585-2060s at Gonbad site (Table 3).In addition, GCM models affected the growing season length under both scenarios (Tables 3 and 4).HadGEM3 simulated an earlier occurring maturity stage (20 days for all cultivars in SSP585-2060s) than ACCESS-CM2 (11 days for all varieties in the same period) (Table 4).

Impacts of sowing windows and cultivars on yield and WP
The effect of management in the planting dates combined with cultivars on wheat yield is shown in Fig. 11.
The highest yield calculation of Kalate (5917 and 5952 kg ha − 1 at Gonbad and Gorgan sites, respectively) and Meraj (5687 and 5598 kg ha − 1 at Gonbad and Gorgan sites, respectively) achieved on the 16-Dec planting.In contrast, the early planting resulted in the highest yield of Tirgan (5607 and 5858 kg ha − 1 at Gonbad and Gorgan sites, respectively) (Fig. 11).Although the calculated average yield of all cultivars confronted a statistically signi cant reduction (p < 0.05) in the future, a delay in planting can assist achieving better yield in all cultivars.In the latest planting window, the yield of Kalate under SSP585-2060s uctuated between 5165 kg ha − 1 and 5753 kg ha − 3 , while it was 4996-5484 kg ha − 1 , and 4999-5553 kg ha − 1 for Meraj, and Tirgan, respectively (Fig. 11).
The details regarding the calculation of WP based on the cultivars and sowing dates are shown in Fig. 12.
Similar to the historical period, the highest WP in all cultivars was attainable through planting on 16-Dec.The WP for all cultivars and sowing date in future periods signi cantly declined (p < 0.05) compared to the baseline; Kalate achieved the highest WP level with sowing on 15-Dec (1.81-0.91 kg m − 3 ) (Fig. 12).In HadGEM3, lower WP levels than the other GCM were achieved through all management practices (Fig. 12).

Discussion
The comprehensive assessment of the LARS-WG 6.0, employed for downscaling GCM models, yielded a satisfactory outcome.However, it is essential to note that a statistical difference was indicated by PMV and RadM, preventing them from attaining a reasonable performance (Table 2).These uncertainties often occurred in months with higher Pre and Rad values.Given the model performance, a higher accuracy level in LARS-WG was obtained when the model thoroughly projected Tmin and Tmax (Bayatvarkeshi et al., 2020).The standard deviation between the generated and observed monthly Pre signi cantly increased during months with higher Pre levels.This rise could potentially lead to a less accurate assessment of the model (Khalaf et al., 2022).
Furthermore, the model's inherent uncertainty in simulating Pre is anticipated to pose a challenge in accurately computing the average and standard deviation, particularly in months with substantial rainfall (Sha et al., 2019).The ndings of this study were in line with (Kavwenje et al., 2022;Mohammed et al., 2022) that the model had a satisfactory level of con dence in synthesizing a wide range of weather variables in different districts.The LARS-WG performance used for downscaling in Northern Iran reasonably generated Tmin, Tmax, and Pre through the K-S test (Roshani & Hamidi, 2022).
The APSIM-Wheat successfully simulated three cultivars' main phenological development stages over calibration and validation (Figs. 3 and 4).This con rms the accuracy of the data collected for determining phenological stages, a rming that the model could reliably calculate the cumulative GDD across diverse sowing windows and locations.Kumar et al (2023) reported that APSIM-Wheat achieved a robust performance for simulating the developmental stages of wheat in Denmark regarding the several locations (7), years (5), sowing data (2) and N management (7)(8)(9)(10)(11)(12)(13).In addition, the model reasonably simulated biomass production and yield compared to the collected data (Figs. 5 and 6).Evaluating different scenarios used in this study indicated that the amount of Pre and Rad, and either Tmax or Tmin are expected to increase.However, GCMs showed these trends differently, which may be driven by the inconsistencies in the models (Figs. 7, and 8).The projection of Pre predicted that higher changes are expected in spring and winter.It is also reported that there will be a decrease in Pre for the vast majority of locations in The impact of climate change on wheat developmental stages is presented in Tables 3 and 4, indicating a statistically signi cant reduction in the growing season for all cultivars compared to the baseline.This reduction is attributed to the increase in both Tmax and Tmin which will be able to supply the required GDD in shorter periods than usual (Figs. 7 and 8).A 4-5 days reduction in the length of the growing season is expected when both Tmax and Tmin experience a 1°C increase (Gao et al., 2018;Xiao et al., 2020;Zhang et al., 2022).The total water input is simulated to increase which can be driven by the Pre projection (Figs. 7 and 8).
This augmented trend is more intensive during the 2060s than 2040s (Tables 3, and 4).Conversely, at its peak level in the baseline, WP is expected to undergo a statistically signi cant decrease across all cultivars (Tables 3, and 4).Particularly, Kalate stands out as the best cultivar exhibiting effective water consumption throughout the entire timeframe and across all sites, owing to its genetic characteristics, such as the potential grain-lling rate and grain per gram stem (as detailed in Table 1).These variables contribute to the proper grain yield, indirectly enhancing the WP.The integral role of biomass accumulation after the anthesis phase for obtaining high wheat grain production was revealed under water stress conditions (  3 and 4), wheat yield decreased (Figs. 8 and 9), which is attributed to the intensi cation of ET0 (Fig. 10).
The timing of cultivar planting had a diminishing impact on the wheat yield and WP as illustrated in Figs.

Conclusion
The LARS-WG, utilized for downscaling GCMs, exhibited reasonable agreement in projecting climatic variables (e.g., precipitation, radiation, maximum and minimum temperatures) at both sites.Thus, it remains a valuable tool, complementing other methods in management practices.Parametrization and validation processes were conducted in two environmental conditions with different sowing dates and three wheat cultivars: 'Kalate,' 'Tirgan,' and 'Meraj'.The APSIM-Wheat reasonably calculated the crop variables (e.g., the phenological stages, biomass production, and grain yield).This model enabled us to drop the future weather data (2030-2049 and 2050-2070), which the LARS-WG projected into the APSIM-Wheat as an ago-climatic input.The projection results implied that all climatic components will likely increase, affecting the ET 0 , even the Pre.This rise in ET 0 , which will be more severe under SSP585 than SSP245 will cause a statistical reduction in the length of developmental stages, WP, and yield.A suitable management strategy regarding the projected weather was implemented to achieve a satisfactory yield and optimal WP when the short-duration variety with a higher grain-lling rate (Kalate) was sown in the late planting calendar (16-Dec) in northern locations of Iran.Lastly, crop yield, in uenced by environmental, agronomic and genetic traits should be precisely simulated by modelling platforms to unwind the obstacles and concerns for stable wheat production prospect in each location combined with available management techniques.

Figure
Figure 3

Figure 7 Monthly
Figure 7

Table 1
Extracted genetic coe cients of three wheat cultivars in the APSIM

Table 2 .
The numerical series indicate the number of tests that resulted in signi cantly different results at the 5% signi cance level out of the total number of tests (Table2).Higher numbers suggest lower model performance, while smaller numbers indicate more reliable model performance.Although the WDSeries, PreD, PMM, TminD, TminM, TamxD, TamxM, and

Table 2
Performance of the LARS-WG model based on the statistical tests to compare the simulated and observed weather variables for either station TamxD and RadD are the inspected, generated, and observed daily distributions of precipitation, minimum and maximum temperature and solar radiation by the Kolmogorov-Smirnov (K-S) test, respectively; PMM, TminM, TamxM and RadM are the inspected, generated, and observed monthly average of precipitation, minimum and maximum temperature and solar radiation by the t-test, respectively; PMV is the inspected, generated, and observed monthly average variances of precipitation.
Number of null assumption rejections is illustrated at the 5% signi cance level by the numerals.WDSeries is the inspected, generated, and observed distributions of seasonal wet/dry series by the Kolmogorov-Smirnov (K-O̅ S) test; PreD, TminD,
An acceptable accuracy in calculating crop production (biomass and yield) is vital and indispensable, as it enables users to implement decision-making techniques through various scenarios.The model exhibited reasonable predictive outputs for yield in Morocco, characterized by a semi-arid weather condition (RMSE = 0.13, NSE = 0.95, and d = 0.98) (Briak & Kebede, 2021).The acceptability of APSIM-Wheat in simulating wheat production has been con rmed under various environmental conditions and management strategies by several studies (Bana et al., 2022; Berghuijs et al., 2021; Devkota et al., 2023; Eyni-Nargeseh et al., 2020; Silungwe et al., 2018).
(Yang et al., 2020)0), 2020;., 2015)ltivars is projected to decrease signi cantly compared to the baseline yield in the future (Figs.8 and 9).The Kalate cultivar exhibited the highest yield across different time frames and under various climate scenarios.Crop genetics have been manipulated to enhance yields, particularly in semiarid regions, susceptible to uctuations in meteorological variables.Zhang et al. (2009)highlighted that the dry matter, accumulating during the wheat's owering and grain-lling stage makes up 60-80% of yield.A weak connection between grain lling length and grain weight was reported as the synthesis rate of stored products is a more vital factor than its length(Wang et al., 2009;Xie et al., 2015).Moreover, the negative impacts of climate change on ET 0 are represented in Fig.10, whereas ET 0 was exacerbated in the future periods based on the baseline by implementing different GCMs.These results were also reported in Iran(Bayatvarkeshi et al., 2020; Rajabi & Babakhani, 2018), and the USA(McEvoy et al., 2020).The prediction of ET0 signi cantly increased under the highest GHG emissions (SSP585), which is consistent with previous research(Yang et al., 2020).Despite a considerable increase in rainfall over the growing season (Tables (Li et al., 2022)Yang et al., 2021), promoting the WP in prone areas(Li et al., 2022).
11and 12. Adjusting sowing windows is tied to speci c cultivars, particularly in the baseline.According to statistical tests, delaying the sowing window could effectively sustain wheat yield at the Gonbad site.Similarly, higher yields were observed with a late planting date (16-Dec) at the Gorgan site, showing a statistically signi cant difference in yield between the baseline and the near and remote future.This nding is consistent (Kheir et al., 2021;Paymard et al., 2019)Hussain et al., 2020)an et al., 2021)7;Yanan et al., 2021).On the other hand, others reported that early sowing dates can enhance wheat yield in the future(Dubey et al., 2020;Gunawat et al., 2022;Hussain et al., 2020).This implies that climatic variables vary according to geographical differences can signi cantly in uence the crop calendar.Pre and Rad levels are projected to increase considerably during winter and spring, whereas this elevation will be observed in Tmax, and Tmin during summer and fall seasons (Figs.7 and 8).These changes may compel a delay in the sowing of cultivars, even for those initially planted during the early window in the baseline (e.g.Tirgan).According to the results, the optimal wheat yield and (WP) are associated with the late sowing date (16-Dec) and the Kalate cultivar in the projected weather (Figs.11 and 12).However, both of these factors considerably decreased in the future compared to the baseline counterparts, which results from the excessive level of ET 0 .Previous studies con rmed these ndings(Kheir et al., 2021;Paymard et al., 2019).