Effect of Temperature on Sowing Dates of Wheat under Arid and Semi-Arid Climatic Regions and Impact Quantification of Climate Change through Mechanistic Modeling with Evidence from Field

: Rising temperature from climate change is the most threatening factor worldwide for crop production. Sustainable wheat production is a challenge due to climate change and variability, which is ultimately a serious threat to food


Introduction
Wheat is a staple food in Pakistan and contributes to food security locally and globally. Pakistan ranks 6th in wheat production and 8th in the harvested wheat area but 59th in terms of yield per unit area worldwide [1]. Wheat production largely depends upon weather conditions, especially adequate rainfall and temperature during the crop growing season. The production of wheat was higher in 2013-2014 compared with 2012-2013 due to sufficient rainfall and favorable temperature in Punjab-Pakistan [2]. Quantification of climate change impacts on wheat is an important step to develop adaptations and boost wheat grain yield at the field and national level.
The elevated temperature has threatened crop production around the globe during the last few decades. Wheat yield is predicted to fall by 6% worldwide with a rise of 1 °C [3]. Climate models predicted a mean increase in temperature by 1-3.7 °C, with a higher likelihood of increased frequency of heatwaves during the mid-to late-21st century [4]. Grain yield of wheat would decline 6-9% with 1 °C elevation in temperature of semi-arid and arid regions of Pakistan [5]. Ahmad et al. [6] reported that a 2 °C rise in temperature would reduce grain yield by 15.2% of irrigated wheat during mid-century 2040-2069 under representative concentrations pathway (RCP) 8.5 in Punjab-Pakistan. Elevated temperature negatively affects wheat growth and development, such as small grains, at 25-35 °C due to shorter grain filling duration and low photosynthesis rate at more than 30 °C [7]. Damaged vegetative structure negatively affects the reproductive structure and grain yield [6]. Additionally, warm temperature (>30 °C) significantly inhibits plant photosynthesis under water stress conditions (shortage of rainfall) [7]. Therefore, understanding the warm temperature effects on wheat growth and development is crucial in establishing a sustainable production system that can offset climate change effects.
There are different methods to analyze the high-temperature impacts, such as growth chambers, radiant tunnels, temperature free-air controlled enhancement, but field experiments and crop simulation modeling are the most useful methods [8]. Crop models are efficient tools for quantifying the climate change impacts on crops [3,9]. Crop growth models are being used to demonstrate the effect of crop management practices such as sowing dates and genetics of plants and even the interaction of these two entities with environmental conditions [10][11][12][13][14][15][16]. In many studies, crop models were utilized to assess the climate change impacts on crop production with changing temperature, elevated carbon dioxide, and rainfall patterns [9,10,[13][14][15]. The Decision Support System for Agrotechnology Transfer (DSSAT) is a complete system analysis approach for evaluating crop management, including sowing dates, irrigation, nutrients, and many more [17][18][19][20][21].
It has the potential to simulate crop development, growth, and yield by the interaction of soil, plant genetics, and atmospheric variables such as temperatures, precipitations, and solar radiation [10]. The CSM-CERES-wheat in DSSAT has been widely used for crop management decision-making for sustainable wheat production [18][19][20][21]. A comprehensive study of quantifying temperature rise and its impact on wheat in the region is limited, especially in Pakistan. This research is a combined effort of experiment and model simulation in the semi-arid and arid region of Pakistan to measure the high-temperature impacts on wheat morphology and grain yield through field experimentation and simulation modeling, respectively.

Experimental Sites and Environmental Conditions
Field trials were conducted during 2013-2014 and 2014-2015 wheat growing seasons in two contrasting environmental regions, one belonging to a semi-arid climatic region (31.4303° N, 73.0672° E) and the other belonging to an arid climatic region (31.1338° N and 70.5761° E) (Figure 1). Mixed cropping is dominant in both study regions. However, the wheat crop is dominant during the winter seasons, while there is winter fodder also grown in both regions for domestic purposes. The study regions, being a semi-arid and arid climate, experienced noticeable changes in air temperature of day and night and even significant fluctuations during wheat crop growing seasons, especially during February to April. Rainfall during this season is also highly variable and did not coincide with the production technology of wheat crops; rainfall either happened during sowing time or during harvest timing, which severely affects cultural practices. Daily weather data during field experimentation and the historical weather data were collected from the Pakistan Meteorological Department (PMD). Weather variables included daily minimum (Tmin), maximum (Tmax) air temperature and precipitation, solar radiation, humidity, and wind speed. Physiochemical soil properties of both study regions are presented in Table 1.

Experimental and Crop Management Details
The experiment was laid out in a randomized complete block design (RCBD) with a split-plot arrangement having four replications. The experiment consisted of eleven sowing dates from 15 October to 16 March with an interval of 14-16 days in the main-plot, with spring wheat cultivars (Galaxy-2013, Punjab-2011, and Lasani-2008) in sub-plots. The net plot size was 2 × 8 m. The wheat crop was sown on the different sowing dates with the help of a single row hand drill, keeping the R x R distance of 22 cm, using a seed rate of 100 kg ha −1 . Nitrogen, phosphorus, and potassium were applied in all plots at the rate of 120, 85, and 60 kg ha −1 , respectively, in the form of urea, diammonium phosphate, and sulphate of potash, respectively. There was no application of another chemical. All cultural practices such as weeding and intercultural practices were kept uniform for all the experimental treatments. The crop was irrigated with canal water on the physical appearance of crop and with soil to keep non-stress water conditions. At maturity, ten selected plants were taken from each plot of each replication to measure the plant height, spike length, and spikelets per spike. Collected data were analyzed statistically by employing Fisher's analysis of variance technique, and differences among treatment means were compared by using Tukey's multiple range test at a 5% probability level.

Temperature Calculations
The temperature was calculated to quantify the impacts on wheat morphology. The mean seasonal temperature (MST) of each sowing date was calculated from the average of the maximum and minimum temperature during crop sowing to maturity, and the mean air temperature from heading to maturity was calculated between the heading stage to maturity stage using the following equation.
Mean Temperature stage to stage = (∑Tmax/∑Days) + (∑Tmin/∑Days)]/2 ∑ Days = total number of days for the duration between two stages ∑ Temp = total maximum/minimum temperature during two stages

Climate and Crop Modeling
Historical climatic data  of two environments, Faisalabad and Layyah, were simulated by general circulation models (GCMs) to compare with future simulated data for the duration of 2051-2100. Observed data at Faisalabad were available for the calibration and evaluation of simulated data of GCMs. Historical data for the arid study site (Layyah) were not available, which was produced by following AgMIP (Agricultural Model Inter-comparison and Improvement Project) procedures [22,23].
Future climatic data were simulated using 29 GCMs available around the globe using AgMIP procedures and protocols through R Script [23][24][25]. These future weather data were distributed by a stretched distribution approach among five quadrants of climatic conditions (Hot/Dry, Hot/Wet, Cool/Dry, Cool/Wet, and Middle), as shown in Figure 2.
There was variation in the number of GCMs for each quadrant of climatic conditions during the distribution of 29 GCMs based on a change in temperature and rainfall as compared with the baseline. Among different GCMs from each quadrant, the GCM closest to the median value (the A dot in each quadrant shows the median value) of the quadrant was selected for simulation of future weather data to be used as the input data set in the crop model. A well-calibrated and evaluated CERES-wheat model under DSSAT v4.7 was used to simulate the grain yield data with baseline climate data and future climate data [8]. Calibration was the first process of simulation to adjust the parameters of models to the local conditions. It was also necessary to find the genetic coefficients of the three wheat cultivars grown in the region. Soil parameters of CERES-wheat were also adjusted to parameterize the model. The genetic coefficient of cultivars (Lasani-2008, Punjab-2011, and Galaxy-2013) was obtained by calibrating the CERES-wheat with optimum planting date of 15 November 2013. General characteristics of wheat cultivars are mentioned in Table 2.
Overall calibration performance of CERES-wheat was reliable for all parameters (days to anthesis, days to maturity, grain yield, biomass, and leaf area index). First, days to anthesis for cultivar Lasani-2008 were close to observations with a percent difference (PD) of −3.85%, whereas, days to anthesis of cultivars Punjab-2011 and Galaxy-2013 were matched to observations. Days to maturity of cultivars Lasani-2008, Punjab-2011, and Galaxy-2013 were simulated with the PDs of −2.05%, 1.41%, and 2.10% respectively. While in the case of grain yield, maximum PD was calculated at 3.58%, and the minimum difference was observed in Galaxy-2013 (−1.28), while the model underestimated the grain yield of Punjab-2011 −2.42%). Biomass was also well predicted with PDs −3.75%, −1.31%, and −2.89% of cultivars Lasani-2008, Punjab-2011, and Galaxy-2013, respectively, with the observed data. Maximum LAI was underestimated by CERES-wheat for Lasani-2008 (−5.88) but overestimated for Punjab-2011 (2%) and Galaxy-2013 (3.92%). Generally, the model performed reasonably well during calibration and evaluation for all studied parameters.

Current Weather Conditions during the Experimental PERIOD
With regard to the climate of the study area, Faisalabad (semi-arid) and Layyah (arid climate) have dry, hot, and humid summers and dry, cold winters. The temperature remains lower from October to February while being higher from March to September. The minimum temperature for November 2014 was higher (19.3 °C) as compared with November 2013 (12.1 °C); likewise, the maximum temperature was also higher during November 2014 as compared with November 2013 at Faisalabad. December 2013 was 2-3 °C hotter in terms of minimum temperature, while the maximum temperature was the same at the start of December. Minimum temperatures were 3.4 °C and 5.9 °C during January 2014 and January 2015, respectively. During the last fifteen days of February, the minimum temperature of 2015 was 4 °C higher than in 2014, but in terms of maximum temperature, the difference was only about 10 °C. The mean monthly temperature of March was 17.5 °C. The minimum temperature ranged from 17 to 19 °C, while the maximum temperature ranged from 26 to 29 °C, with a cold to hot trend from the start to the end of April during both years. At the end of January 2015, there was a first rainfall of 12 mm. In the arid region (Layyah), the maximum average air temperature ranged from 24 to 26.5 °C for November 2013, and the minimum temperature ranged from 10.2 to 11.9 °C during 2014. The difference in mean temperature was about 4 °C between the first and second half of December. The minimum average temperature of two years during January was 4.5 °C, while the maximum temperature was 18 °C. February was hotter than January during the last two weeks. The minimum temperature was higher during the last fifteen

Future Climatic Conditions
Capital letters in different quadrants (Figure 2) represent the 29 GCMs based on changes in precipitation and temperature from baseline. Hot/Wet, Cool/Wet, Cool/Dry, Hot/Dry, and Middle climatic scenarios were predicted by the six, three, six, four, and ten GCMs respectively for Layyah. A maximum of ten GCMs predicted that Middle climatic conditions would prevail in the future while the minimum of the three models agreed on chances of Cool/Wet climatic conditions. For Faisalabad, ten GCMs predicted the chances of Middle climatic conditions followed by Hot/Wet (7 GCMs), while Cool/Wet, Cool/Dry, and Hot/Dry conditions were predicted by four GCMs. Figure 3A

High Temperature and Wheat Morphology
Increasing temperature severely affects wheat morphology. Plant height (Figure 4) was significantly different from changes in environments, cultivars, and sowing dates but non-significant between locations (Table 3)   The results depict that the maximum spike length (8.85 cm) was observed during 2013-2014, while the minimum from the environment occurred in 2014-2015 (8.52 cm), as given in Figure 5. Sowing dates also affected spike length significantly. The spike of 1 November sowing and 1 March were the lengthiest (13.54 cm) and shortest (5.48 cm) under MTHM 16.23 °C and 25.41 °C, respectively. No spike was produced from 15 March sowing with MTHM 27.2 °C. Table 4 shows the relationship strength between temperature and spike length. The interactive effect of E × L, E × PD, E × G, L × G, E × L × G, and E × PD × G on grains per spike depicted significant differences.  The maximum spikelets per spike were recorded during the environment 2013-2014 (12.26), while the minimum was observed during 2014-2015 (11.77) ( Figure 6 and Table  3). The results for sowing dates depicted maximum spikelets per spike for 15 October sowing (16.75) and minimum (5.46) for 1 March sowing. Slope and R 2 were the slopes of fitted lines, and coefficients of determination between wheat spikelets per spike and temperature are given in Table 4. The interactive effect of E × L, E × PD, L × PD, E × L × PD, E × G, and E × PD × G on spikelets per spike depicted significant differences (Table 3). However, on spikelets per spike L × G, E × L × G, PD × G, L × PD × G, and E × L × PD × G interactions were not statistically significant.

Simulated Grain Yield for Future Climate
The grain yield reduction was different under different climatic conditions as compared with the baseline climatic conditions of Faisalabad (average 4035 kg ha −1 ), as shown in Figure 7B. There was a reduction of grain yield in all climatic conditions except some years of Cool/Dry and Cool/Wet climatic conditions, as shown in the thirty-year data distribution in box-whisker. Maximum grain yield reduction was under Hot/Dry climatic conditions, with an average of 2502 kg ha −1 due to the hottest and driest conditions, while the minimum reduction was under Cool/Dry climatic conditions with 3235 kg ha −1 . The average grain yield of wheat was 2988 kg ha −1 , 3088 kg ha −1 , and 3203 kg ha −1 under Hot/Wet, Middle, and Cool/Wet climatic conditions, respectively.
At Layyah (Figure 7A), climatic conditions were severe, which increased the chances of severe yield reduction due to the arid region and climate. The simulated grain yield of the Layyah baseline (3805 kg ha −1 ) was lower than the simulated grain yield of Faisalabad under baseline climatic conditions (4035 kg ha −1 ). The reduction of grain yield was the highest under Hot/Dry climatic conditions, with an average value of 2177 kg ha −1 , and the lowest under Cool/Dry climatic conditions, with a simulated average grain yield of 2864 kg ha −1 . The different climatic condition was the cause of yield reduction at Layyah except for some years of Cool/Wet and Cool/Dry climatic conditions. The average simulated grain yield was 2386 kg ha −1 , 2452 kg ha −1 , and 2792 kg ha −1 under Hot/Wet, Cool/Wet, and Middle climatic conditions.

Future Climate Projections
Prediction of representative climate scenarios of any site is the first important step of climate change impact estimation. The stretched distribution of 29 GCMs among five quadrants-Hot/Wet, Hot/Dry, Cool/Wet, Cool/Dry, and Middle-based on the change in temperature and rainfall shows the chances of occurrence of each climatic condition. This approach gives us a comprehensive overview of all plausible climatic conditions and their chances of occurrence [6,23]. The selection of representative GCMs from each quadrant increases the reliability of simulations for future climatic data and saves the resources for computation [22,24,25]. The maximum number of GCMs in the quadrant of Middle climatic conditions shows the highest chances (35.5%) of prevailing in the future at both Faisalabad and Layyah. The lower number of models in the quadrant shows lower occurring chances of climatic conditions such as in the Cool/Wet quadrant [22][23][24][25][26]. In general, chances of Cool/Wet conditions in the future are meager, while the GCM distributions also predicted the lower chances (10.3% at Layyah and 13.7% at Faisalabad) of such climatic conditions. A different number of models in different quadrants also showed that there is uncertainty in simulations of GCMs [24,25].
The output of the most representative GCMs from each climatic condition was used to calculate the changing climate temperature during the second half of the century. Among five climatic conditions, Hot/Dry was the hottest and driest condition, which showed that these conditions could have drastic effects on the wheat crop as compared with Cool/Dry, Cool/Wet, Middle, and Hot/Wet climatic conditions at Faisalabad and Layyah. The predicted hotter climatic conditions of Layyah than Faisalabad was illustrated from baseline data. The primary trend of the season could remain the same as a baseline, e.g., the hotness can increase from November to April. Iqbal et al. [27] reported that temperature could rise by 3 °C and 6 °C during mid-century and end-century under RCP 8.5, respectively. Islam et al. [28] showed concerns about the imminent warming of the winter season as compared with past climatic data. Similarly, Ahmad et al. [6] reported a 2 °C rise in temperature during mid-century scenarios for the wheat crop growth season. The final relative change in temperature showed a maximum rise during March at Faisalabad (6.35 °C) and Layyah (7.05 °C). Iqbal et al. [27] found March the most warming month of the year. Overall, this rise in temperature in the future can be a grave threat to wheat production if the current climate warming trend continues.

Elevated Temperature and Wheat Morphology
Wheat morphology is affected by changing environment, management, and genotypes. Variation in environment and management, especially sowing date, change the temperature and photoperiod. Wheat genotypes have a different response to temperature and photoperiod. Adaptation of wheat to different environments on variable sowing dates occurs due to low-level sensitivity to photoperiod [29]. Temperature is an important factor affecting wheat morphology and ultimately affecting wheat yield. Therefore, in this study, we focused on quantifying the impacts of temperature on morphology. The elevated temperature reduced the plant height, which is mainly associated with reduced and shorter crop growth duration [30,31]. During high temperature, plant growth is accelerated, but the duration is shortened, which mainly cause small plant structure. Liu et al. [32] also found that warming increased the plant height, which might be caused by the lesser difference of day and night temperature. In our study, warming was the same in day and night temperature, which is associated with reduced plant height. Meanwhile, timely sown crop under warmer climates is taller due to the lower number of tillers, as reported in [31]. Increasing temperature with delay in sowing causes a reduction of spike length as well as of plant height. Baloch et al. [33] reported that spike length is significantly affected by sowing time. Sowing wheat on 16 October and 30 October produced spikes of 11.04 cm and 10.91 cm, respectively, owing to longer crop duration and low temperature. Shorter spikes (9.4 cm) were recorded in December or later sown crop [34][35][36][37], being short duration crops. The significant differences in spike length between the two environments were due to temperature (Table 4). During 2013-2014, climatic variables such as temperature, precipitation, and solar radiation remained optimum as compared with 2014-2015. A significant variation for spikelets per spike between the two environments was due to variability in solar radiation and air temperature. The optimum conditions during 2013-2014 caused good spikelets per spike. A variable trend for spikelets per spike was recorded for semi-arid and arid study regions. A decrease in spikelets per spike for the last dates was due to high temperature and heat stress as well as to shrinking crop duration encountered by the crop during the life cycle. As the number of spikelets determined during two particular phenological phases (terminal spikelet and grain filling), both phases are temperature sensitive, as high temperature particularly at floret formation causes sterility and hence reduces grain yield.

Future Climate and Wheat Grain Yield
Quantification of the future climate of semi-arid and arid environments showed an alarming situation of climate impact on wheat yield. The average temperature is not suitable for the grain formation stage of wheat, as data show that March temperature can be 6.3 °C and 7 °C hotter than baseline climate at Faisalabad (semi-arid) and Layyah (arid), respectively, under Hot/Dry climatic conditions (harsh climatic scenario). Grain formation of wheat occurs during March at both study regions, under either semi-arid or arid climatic conditions. Usually, mid-March high temperature currently has a negative impact on grain size and ultimately on grain yield, as it has also been depicted in previous studies under rainfed climatic conditions [38][39][40][41][42]. In the future, March could have worse effects than the current conditions on the final grain yield of wheat. Overall, at Faisalabad during the second half of the century, grain size reduction can be 38%, 27.37%, 23.45%, 20.86%, and 19.82% under Hot/dry, Hot/Wet, Middle, Cool/Wet, and Cool/Dry climatic conditions, respectively. Ahmad et al. [6] and Sultana et al. [5] reported a 12.5% grain yield reduction with a 2 °C increase in temperature and a 6-9% yield reduction with a 1 °C rise in temperature, respectively. The maximum reduction would be in Hot/Dry climatic conditions due to the highest temperature (harsh climatic scenario). Because wheat is irrigated in the semi-arid and arid environments, it is not significantly affected by dryness or wetness of season; consequently, the percentage grain yield reduction under Cool/Dry and Cool/Wet conditions is almost equal. Yield reduction was even 1% higher under Cool/Dry climatic conditions. Climatic warming has been reported in different studies ranging from 10% to 50% by Hussain et al. [15] depending on time, location, and wheat cultivars. The reduction of grain yield at Layyah (42.76%) under Hot/Dry climatic conditions could be worse than Faisalabad (38%) due to its high temperature in the future. Hot/Wet, Middle, Cool/Wet, and Cool/Dry conditions would cause a yield reduction of 37.92%, 35.54%, 26.6%, and 24.72%, respectively. These reductions are higher in the arid climate than in the semi-arid climate due to warmer climatic conditions in the arid region [43][44][45][46][47].

Conclusions
Middle climatic conditions would prevail in the future with 3.52 °C and 3.84 °C rises in temperature at Faisalabad (semi-arid) and Layyah (arid) under RCP 8.5 during the second half of the century. The grain yield of wheat would reduce by 23.5% at Faisalabad (semi-arid) and 35.45% at Layyah (arid climate) due to rising temperature till the end of the century. Although there are different responses to each GCM due to their behavior and class, there is still a predicted yield reduction. The mean seasonal temperature and mean temperature from heading to maturity ranged from 16 to 27.3 °C and from 12.9 to 30.4 °C, respectively, for eleven different sowing dates. The spikes were not formed at both locations and during both seasons when the crop was sown after 1 March due to abnormal climatic conditions, especially the temperature for the critical growth phases. Sowing dates proved to be very useful for the quantification of temperature and its impact on wheat crops. These findings would be helpful for the wheat-growers to adjust the sowing time accordingly when there is a climate shift and when climate variability is a threat to sustainable wheat production. Further, these predictions would be helpful for decisionmaking by the policymakers to ensure food security in the studied region. Furthermore, future studies are required, especially to develop new wheat genotypes that have phenotypic plasticity to adjust according to temperature shift and climate variability under future climate conditions. Data Availability Statement: Data will be available on special request from first and corresponding authors.