Influence mechanism of climate change over crop growth and water demands for wheat-rice system of Punjab, Pakistan

Climate plays a critical role in determining the crop growth rates and final yield, thus directly impacting the agriculture sector and food security (Martins et al. 2019). However, the recent climate change phenomenon, which is supported by undeniable scientific evidence, is responsible for the rapid diminishing of snow/ice caps, sea-level rise and frequent natural calamities such as floods and droughts (IPCC 2014; Kang et al. 2017). The climate change trends are severely threatening the agriculture-dependent economies like Pakistan, which is one of the worst climate-struck and food-insecure nations because of its over-reliance on irrigated agriculture, constantly diminishing river flows, swift population growth and low adaptive capacity (Abid et al. 2016; Khan et al. 2016).

Wheat and rice are the staples, accounting for up to 50% of daily calorific intake (Arshad et al. 2017) and their availability/accessibility dictates the food security of Pakistan. According to the World Food Program, despite having a good wheat harvest in 2014, up to 47% of the country's population was food insecure led by widespread malnutrition, uneven food distribution and water shortages (Kirby et al. 2017). Considering the future climate warming in Pakistan, yields of both the crops will drastically decline, accompanied by substantial water shortages (Sultana et al. 2009; Kirby et al. 2017; Ahmad et al. 2019; Ahmad et al. 2020).

Previously, the climate change threats against the agriculture sector of Pakistan regarding crop yield and water consumption were discerned as a function of temperature and rainfall perturbations (Pakistan National Communication 2000; Sultana et al. 2009; Rasul et al. 2012; Ahmad & Choi 2018a). The climatic interactions with crop evapotranspiration (ET_c) and the subsequent irrigation water requirement (IWR) are not solely temperature-dependent since other components such as solar radiation, wind speed and vapour pressure deficit also control the ET_c rates (Irmak et al. 2012). Recently detected declines in sunshine hours and solar radiations can counteract the warming-driven ET_c increments (Liu et al. 2014). An increased ambient temperature often corresponds to a shortened crop growth period (Asseng et al. 2015); thus reducing the cumulative seasonal ET_c and IWR (Karimi et al. 2018; Ahmad et al. 2019, 2020). A combination of the contracted growth period and less solar radiation could reduce the seasonal cumulative ET_c and IWR, but the daily ET_c and irrigation/rainwater consumption rates would increase simultaneously (Saadi et al. 2015; Ye et al. 2015; Ahmad et al. 2019, 2020). Hence, the future ET_c and the associated IWR should be predicted by considering the anticipated reductions in the crop growth period and solar radiation.

This study primarily aims at investigating the primary mechanism of climate change on the crop growth periods and the corresponding water demands of the wheat-rice cropping system in Punjab, the largest agricultural province of Pakistan. Statistically bias-corrected climate change projections from eight general circulation models (GCMs) produced the future climate under the medium and extreme representative concentration pathway scenarios. The future climate was projected during two time periods, 2030s (2021–2050) and 2060s (2051–2080), against the baseline duration of 1980–2010. The future crop growth periods were projected by utilising the growing degree day (GDD) concept and the associated crop water demands were estimated by considering all the climate variables which could influence the ET_c and IWR. The study outcomes would facilitate the policy-makers and stakeholders to optimize the irrigation schedules and regulate canal water supplies according to the future crop water demands for mitigating climate stresses on the wheat-rice system of Punjab, Pakistan.

Punjab, the bread basket of Pakistan, has two distinct cropping seasons: winter and summer lasting from November–April and May–October, respectively (Ahmad & Choi 2018b). The wheat-rice system of Punjab covers approximately 1.1 million hectares (Mha) primarily irrigated by the Upper Chenab Canal (UCC) that has a gross command area of 0.64 Mha and cultivable command area of 0.59 Mha of which up to 60% is mostly allocated for the wheat and rice cultivation during the winter and summer season, respectively (Shakir et al. 2010).

This study focuses on the UCC command area (Figure 1) where the canal water is available only during the summers and the rice production is partly, and the wheat production is entirely, dependent on marginal quality groundwater. Seasonal temperature and rainfall fluctuations are intensely characterized by short and dry winters followed by long and hot summers, which also receive the major proportion of annual rainfall as monsoon rainfalls. The average winter and summer temperatures vary in ranges of 8–19 and 20–42 °C, and approximately 60% of annual cumulative rainfall (994 mm) occurs as monsoon rainfalls in July–August. Moderately fine to medium textured soils prevail in most parts of the UCC command area (Jehangir et al. 2002, 2007).

Sialkot is the only weather station in the vicinity of the study area with reliable long-term weather data records. A climate dataset for the baseline period of 1980–2010, including daily maximum temperature (T_max), minimum temperature (T_min), number of sunshine hours (n), net radiation (R_n), rainfall (P) and monthly average values of relative humidity (RH) and wind speed (u₂), was collected from the Pakistan Meteorological Department. The daily RH and u₂ values were also retrieved from the National Centers for Environment Prediction (NCEP) Climate Forecast System Reanalysis (CFSR) global climate data set (https://globalweather.tamu.edu/#pubs). The CFSR data had a horizontal resolution of 38 km and daily T_max, T_min, RH, u₂, and R_n data were available from 1979 to 2014 (Saha et al. 2010; Fuka et al. 2013; Dile & Srinivasan 2014) which were downloaded for the grid containing the Sialkot station.

A simple linear-scale change-factor bias-correction procedure was applied on the CFSR daily dataset by adding (multiplying) the difference (ratio) between the long-term mean monthly values of a specific climate variable in observed and CFSR datasets. Additive corrections were made for the T_max and T_min and multiplicative corrections were made for the RH, u₂ and R_n. Underlying principles, merits and demerits of the bias correction approach can be cited from the literature (Ines & Hansen 2006; Chen et al. 2011; Teutschbein & Seibert 2012; Kum et al. 2014; Miao et al. 2016). After removing the possible biases (Figure 2), it was assumed that the CFSR daily RH and u₂ time series could efficiently represent the climate of the study area. The study area's daily baseline climatology was devised by combining the observed T_max, T_min, P and R_n time series with the basic-corrected CFSR reanalysis time series of RH and u₂.

GCMs are the numerical models which are widely used to simulate the Earth's climate dynamics hundreds of years in the future (Ehret et al. 2012; Anandhi et al. 2016). A scenario defines the plausible future state of the world after incorporating the anthropogenic and natural variabilities in the climate and socioeconomic conditions (Anandhi et al. 2016). In the fifth phase of the coupled model intercomparison project (CMIP5) four new climate scenarios called the representative concentration pathways (RCP) were included. The RCPs were categorized based on a rough estimate of the radiative forcing in 2100 relative to the preindustrial era. For example, in the extreme and medium emission scenarios of RCP 8.5 and 4.5, the radiative forcing increases throughout the 21st century, finally reaching a level of 8.5 and 4.5 W/m², respectively, at the end of the century (Taylor et al. 2011).

In this study, eight GCMs (Table 1) from CMIP5 archives were chosen based on their ability to represent the South Asian monsoon climate after carefully reviewing the latest studies aiming at selecting the suitable GCMs to project future climate-change trends for Pakistan and/or neighbouring countries such as India and China (Biemans et al. 2013; McSweeney et al. 2015; Sharmila et al. 2015; Sooraj et al. 2015; Hasson et al. 2016; Lutz et al. 2016; Su et al. 2016; Chaudhuri & Srivastava 2017; Ruane & McDermid 2017). Daily GCM outputs of T_max, T_min, RH, u₂, P and R_n forced under the RCP 4.5 and 8.5 during two future time periods: 2030s (2021–2050) and 2060s (2051–2080) were included in the study. RCP 8.5 was selected to discern the extreme climate-induced risks because it covered higher radiative forcing and temperature change. RCP 4.5 was preferred over RCP 6.0 because of its relatively better prediction capability of annual average CO₂ emission growth rates during 2005–2012 (Peters et al. 2013; Sanford et al. 2014). We exclude RCP 2.6 as the CO₂ mitigation rate proposed by the RCP is unlikely to prevail in the near future in Pakistan (Khan & Koch 2018).

Coarse-resolution of the GCMs outputs lacked indispensable fine-scale, sub-grid information for agricultural impact assessment. Inherent unavoidable biases originating from various sources also distort the intensity and frequency of extreme events like heatwaves and dry/wet spells in the GCMs outputs. The literature suggests a variety of techniques and tools to tackle these biases ranging from complex and computationally extensive dynamical-downscaling techniques to simpler, readily adoptable statistical bias-correction techniques. Among the latter group, quantile mapping (QM) is a commonly followed procedure where the biases in the GCM historic simulations are removed by mapping the cumulative distribution function (CDF) of a certain variable in GCM time series over the CDF of the same variable in the observed time series. The same empirical CDFs are then applied over the GCM-projected future time series. Details about the underlying principles, limitations and extensions of QM and its comparison with other approaches has been extensively discussed in the literature (Ehret et al. 2012; Ahmed et al. 2013; Kum et al. 2014; Rockel 2015; Miao et al. 2016; Sippel et al. 2016; Eum & Cannon 2017). In this study, the daily GCM outputs of T_max, T_min, RH, u₂, P and R_n during the historic and future runs were subjected to QM based on daily observed baseline data.

For medium duration wheat and rice cultivars, the T_base values were assumed to be 5 and 10 °C, respectively. In Punjab, the seasonal cumulative GDDs for the wheat and rice crops lie in the ranges of 1,200–2,250 and 1,650–2,750 °C, usually accumulated over 120–270 days and 108–125 days, respectively (Hanif Qazi et al. 1997; Pakistan National Communication 2000). We assume average optimum GDD of 1,800 and 2,200 °C for the medium duration cultivars of wheat and rice, respectively (Hanif Qazi et al. 1997; Pakistan National Communication 2000). The crop growth periods were computed as the number of days required to accumulate the respective crop GDDs during baseline and future time periods to gauge the climate-driven impacts on crop growth rates.

The single crop-coefficient methodology proposed by the FAO was adopted to estimate the crop water requirements. The daily reference crop evapotranspiration (ET_o) was estimated by the FAO Penman–Monteith equation and the crop evapotranspiration (ET_c) was approximated from the product of ET_o and crop coefficients (K_c) (Allen et al. 1998). The 10-daily K_c values for the UCC wheat-rice system were referenced from Ullah et al. (2001). The difference between the effective rainfall (ER) and ET_c was the IWR that have to be applied through irrigation. The ER was calculated using the United States Department of Agriculture method (Luo et al. 2015; Ye et al. 2015; Tukimat et al. 2017; Zhou et al. 2017).

In Punjab, rice is grown under constant flooded conditions with ponding depths (W_p) of 50–75 mm maintained during most of the season, whereas the water usage (W_lp) for land preparation through puddling varies in the range of 250–900 mm (Soomro et al. 2015). We used W_p of 75 mm and W_lp of 450 mm as the representative values in Equation (2) by following the previously conducted field studies (Aslam et al. 2002; Bhatti et al. 2009; Soomro et al. 2015). Percolation loss (W_p) through rice fields is mainly a function of soil texture and it varies from 2 mm/day for heavy clay soils to 6 mm/day for sandy soils (Chapagain & Hoekstra 2011). More than 60% of farms located in the UCC command area had heavy clay loam as dominant soil texture (Hussain et al. 2012). Ullah et al. (2001) used a percolation rate of 1.5 mm/day for rice consumptive water demand estimation in various canal commands of the Indus Basin. We considered an average percolation rate of 2.5 mm/day as suggested by Chapagain & Hoekstra (2011) for heavy clayey soils.

Growing season lengths of 150–180 and 120–150 days were considered for the wheat/winter season and rice/summer season, respectively. The wheat sowing and harvesting dates were set as November 1 and April 31 whereas for rice these were set as June 1 and October 30 with July 1 being the transplantation date. All the calculations were performed for a 10-day time step during both growing seasons under the baseline and future time slices.

During the baseline (1980–2010) period, annual average T_max, T_min and annual cumulative P outputs from the selected GCMs before and after bias correction and the respective biases are shown in Figure 3. All GCMs underestimated the observed temperature (both T_max and T_min) and P except HadGEM2-ES which showed a temperature over-estimation tendency. After bias correction, means, medians and distribution of extreme heat events were significantly reduced. Individually, inmcm4 and MIROC5 showed the highest tendency of duplicating the observed T_max and T_min time series during the baseline (Figure 3(a) and 3(b)).

The GCMs capacities to reproduce baseline observed P data remained limited even after the QM application. The mean and median mismatches between observed and GCM P time series were adjusted but the misrepresented extreme P events were not fully eliminated (Figure 3(f)). This issue was resolved by taking an ensemble of bias-corrected P data, and the biases in the mean, median and frequency or distribution of intense P-events were successfully removed (Figure 3(c)). Thus, we assume that bias-corrected GCM-ensemble P outputs can fairly project the study area's future rainfall patterns.

The projected changes in GCM-ensemble T_max, T_min, R_n and P, after bias correction, under RCP 4.5 and 8.5 during the 2030 and 2060s compared to the baseline (1980–2010) are shown in Figure 4. An intense and hotter shift in the thermal regime was projected with tangible evidence of climate warming (both T_max and T_min) at seasonal and annual scales irrespective of the GCM, RCP (4.5 and 8.5) or time period (the 2030 and 2060s). The projected climate warming featured various noteworthy trends: T_min > T_max, winter season > summer season, RCP 8.5 > 4.5 and 2060s >2030s.

The area was also projected to receive higher annual P, particularly due to intense monsoon P during summer/rice season regardless of the GCM, RCP or time slice. For both RCPs, the winter/wheat seasonal P showed an unnoticeable rising trend during the 2030s and a declining trend during the 2060s. Only the positive monsoon-P shifts during the summer/rice season were actively contributing towards the annual P increments. Climate projections suggested that the area may experience cycles of hot/dry winters followed by hot/wet summers, which can also result in frequent episodes of drought and floods, particularly during the 2060s. The winter season was more susceptible to enduring worst climate change impacts than the summer season.

ET_o, being only a function of climate variables, is a reliable reflector of the climate change impacts on ET_c rates (Ahmad & Choi 2018b). Figure 5 presents the projected change in cumulative seasonal ET_o during the future time slices relative to the baseline. Future climate warming increased the cumulative seasonal and annual ET_o. The winter/wheat season ET_o-change remained consistently higher than that of the summer/rice season. The ET_o increment under RCP 8.5 was more pronounced than that of under RCP 4.5 regardless of the time slices; whereas the positive ET_o shifts during the 2060s were higher as compared to the baseline or 2030s at both seasonal and annual time scales.

Considering the future climate warming, the magnitude of ET_o increment was unusually small. The study area was projected to face a 132 MJm⁻² (–2.1%) and 213 MJm⁻² (–3.4%) reduction in annual average R_n by the end of the 2030 and 2060s, respectively. Hence, the warming-driven ET_o increments would be counteracted by the declining R_n in future. This perception was further elaborated when analyzing the monthly ET_o data. Figure 6 shows the GCM-ensemble monthly ET_o variations relative to the baseline during the 2030 and 2060s. Monthly ET_o remained unchanged or even declined in the few cases, despite the temperature rise predictions. Only the November and December monthly ET_o values displayed a rising tendency in the future which were eventually translated as the seasonal and annual ET_o rises. The IQR and whiskers of boxplots in Figure 6 also signify the expectancy of persistence and a lesser range of variability of future monthly ET_o as compared to baseline.

Ambient temperature is a key driver of crop growth rate, consequently determining the growing season length. The changes in wheat and rice growing season lengths under climate warming of the 2030 and 2060s were gauged using Equation (1) and the results are presented in Figure 7. During the baseline, wheat and rice had an average growth period of 154 and 124 days against average optimum GDDs of 1,800 and 2,200 °C, respectively.

The anticipated climate warming shortened the median growth periods of both crops with wheat being the primary target. During the 2030s, a moderate decline in growth periods of both the crops was shown under the two RCPs whereas during the 2060s, the growth period shortening was alarmingly high under RCP 8.5. The wheat and rice showed a potential decline of 11–13 and 8–11 days during the 2030s, and 21–30 and 15–21 days during the 2060s, respectively. These results suggested higher probabilities of future crop failures due to early maturity; specifically, sustaining wheat production in this area could be a challenging task. However, careful re-adjustments of sowing dates in the context of climate warming for optimum crop yields could be a viable option to negate these climate change impacts.

Figure 8 shows the projected changes in wheat and rice seasonal cumulative ET_c, ER and IWR during the 2030 and 2060s. Figure 9 compares the GCM-ensemble monthly time-averaged ET_c, ER and IWR during the baseline and future time periods. A potential decline in future ET_c and IWR of both the crops was detected and the depressions were higher during the 2030s than during the 2060s. RCP 8.5 differentiated from RCP 4.5 by presenting comparatively steeper decline in ET_c and IWR of both crops irrespective of the time period. Rice seasonal cumulative ET_c, compared to wheat, showed a higher (lower) tendency to decline during the 2030s (2060s), relative to their counterparts during the baseline.

The future seasonal cumulative ER featured a slight and substantial increase during the winter/wheat season and summer/rice season, respectively. The ER increments were prominently higher during the 2060s than the 2030s and particularly noticeable under RCP 8.5. As the rice IWR assimilated water requirements for both ET_c and percolation losses, it was expected that a shorter growth period would cause a greater decline in rice IWR than that of wheat. However, the wheat IWR had a steeper decline rate than that of rice despite the minor ER increments due to early maturity. These trends emphasized the higher climatic vulnerability of wheat.

Projected changes in the monthly distribution of future ET_c, ER and IWR according to GCM-ensemble climate conditions were also examined (Figure 9). An ample ER-reduction was noted during the future wheat seasons while the monthly ET_c and IWR remained unchanged or decline a little, except at the end of the growing season where a small increase was shown. For rice, the monthly ER distribution was characterized by an increase during the early months followed by a decrease approaching the end of the growing season (Figure 9(c) and 9(d)). Rice monthly ET_c and IWR exhibited an increasing tendency right from the beginning of the growing season and then took a dramatic downturn at the end of the season due to early maturity.

The mentioned trends were produced without altering the sowing dates. This means that the same and/or higher irrigation amounts would be required during a shortened growth period; suggesting that the temperature rise would certainly accelerate the daily ET_c rates in the future. The future crops would mature earlier yielding an overall decline in seasonal cumulative ET_c but daily ET_c rates would also increase simultaneously, thus making the crops more irrigation-dependent and temperature-sensitive. Due to uneven distributions, the projected rise in monsoon P may not be able to compensate for the intensive monthly ET_c rates; thus, calling for intensive irrigation applications.

Distinctively strong future climate warming in terms of both T_max and T_min accompanied by minor P increments was projected for the study area. The temperature rise was particularly evident for the winter/wheat season and the plausibility of experiencing warmer future winters was supported by the historic temperature rise trends across Pakistan (Rasul et al. 2011; Ahmad et al. 2014; Iqbal et al. 2016). Ahmad & Choi (2018b) also confirmed that climate of the same area had been gradually warming up since the 1980s due to a temperature rise phenomenon particularly associated with the winter months.

The T_max (T_min) had higher variability during the future winter (summer) season. This implied that winter/wheat season T_max and summer/rice season T_min could be the key determinants controlling the future warming of the study area and hotter days/nights could be expected during the winter/summer seasons. Rasul et al. (2012) also reported that the nighttime temperature is rising at a higher rate than that of daytime across Pakistan. T_min is usually associated with nighttime and it tends to rise during cloudy nights marked by greater near-surface longwave radiations; whereas, the T_max is typically associated with daytime and it declines or remains unchanged as the cloudy sky blocks the incoming solar radiations (Irmak et al. 2012). This phenomenon gave rise to asymmetrical climate warming which seemed to occur in our study area under future climate change projections.

Annually, the future P increments were mainly caused by a sharp rise in magnitude and/or intensity of monsoon P in summer/rice season, especially during the 2060s. The GCM ensemble envisaged an increment of 10–13 and 13–23% in summer/rice season P during the 2030 and 2060s, respectively. The seasonal and annual temperature and P trends predicted in our study were in good agreement with previous studies focused on projecting future climate-change trends in Pakistan (Kazmi et al. 2015; Sharmila et al. 2015; Sooraj et al. 2015; Amin et al. 2016; Su et al. 2016).

Under the projected climate, the wheat-rice system in the command area of UCC could face multifaceted consequences. Due to higher monsoon P, flood conditions and drainage problems may arise in the summer/rice season. Because of its non-perennial irrigation system, wheat production in this area is completely groundwater dependent. Higher temperature would drive daily ET_c at higher rates, which if not compensated by enough ER, would result in intensive groundwater exploitation for irrigation purposes, exerting extra pressure on dwindling groundwater resources.

The baseline (1980–2010) average seasonal cumulative ET_o values for wheat and rice were 395 and 875 mm, respectively. The wheat season ET_o increased by 10–13 and 26–37 mm; and rice season ET_o increased by 5–22 and 5–25 mm during the 2030 and 2060s, respectively. The magnitude of ET_o increment was markedly small, particularly for the summer/rice season, in the context of climate warming threats. Ahmad & Choi (2018b) examined the historic trends of climate variables in this area and identified downward (upward) ET_o trends for the summer/rice season (winter/wheat season) since 1980. Their study attributed these ET_o trends to a steep statistically significant downturn historic trend of R_n and n. We also projected a potential decline in seasonal and annual scaled R_n totals during future time periods. As the study area's ET_o is highly sensitive to R_n variations (Adnan et al. 2017; Ahmad & Choi 2018b), its decline ceased the rapid ET_o-increase and counteracted much of the inverse temperature rise influences on ET_o.

Historically, dropping rates of pan evaporation and ET_o, despite a clear climate warming, have been reported from all around the globe; most probably triggered by decreased R_n and/or wind speeds. Cloud cover, air pollutants, and manmade aerosols greatly affect the n and R_n. An increased concentration of manmade aerosols and other air pollutants in the atmosphere could describe the recent depressions observed in n and R_n (Irmak et al. 2012; Liu et al. 2014; Vicente-Serrano et al. 2014; Jhajharia et al. 2015).

During the baseline, wheat seasonal cumulative ET_c and IWR were in the ranges of 207–260 and 90–212 mm; and for rice these were in ranges of 519–600 and 950–1,278 mm, respectively. Shakir et al. (2010) estimated average wheat and rice ET_c values in the UCC canal command to be 250 and 523 mm, respectively. According to the local agriculture office in this area, the consumptive water demands for wheat and rice were 225 and 1,600 mm, respectively (unpublished data). Bhatti et al. (2009) estimated a basin wise average ET_c of 400 mm for wheat and 950 mm for rice in the Indus basin of Pakistan. After conducting field experiments during 2001–2003 at some selected farms located near to our study area, Jehangir et al. (2007) reported that wheat ET_c could vary in the range of 251–368 mm and rice ET_c could vary in the range of 537–627 mm. Similarly, IWR values of both crops were also supported by findings of the survey and/or experimental studies previously conducted in this area (Aslam et al. 2002; Hussain et al. 2012; Soomro et al. 2015).

The future ET_c and IWR were projected to decline mainly because of a shortened growth period. A potential decline of 11–13 and 15–21 days for wheat; and of 8–11 and 21–30 days for rice was predicted during the 2030 and 2060s, respectively. These results partly coincide with the findings of the Pakistan National Communication report on climate-change impact assessment over the agriculture sector. The projected shrinkage in crop growth periods was confirmed, whereas the predicted declines in ET_c and IWR were contradicted. The report used the Hargreaves method for ET_o estimation excluding the R_n-influence and yielding exaggerated ET_c estimations (Hanif Qazi et al. 1997; Pakistan National Communication 2000). Bhatti et al. (2016) also identified higher probabilities of reduction in crop water demands due to probable shortening of the growth period for the wheat-cotton system of Punjab. Other studies, which include temperature rise effect over growth rate, also support our findings. An anticipated decline in ET_c and IWR originating from the potential shrinkage of the growing season has been repeatedly reported in the literature (Seung-Hwan et al. 2013; Ye et al. 2015; Acharjee et al. 2017a, 2017b; Tukimat et al. 2017).

Our study had few limitations and assumptions which should be addressed before finalizing various future water management strategies such as identification of suitable planting dates, re-adjustment of crop calendars and irrigation schedules, etc. Due to a lack of detailed soil hydrological characteristics and groundwater-depth data, these two important factors were not included in our analyses, which can significantly shape soil water balances and finally IWR of the crops. In the case of rice, the percolation rates and water usages for nursery and land preparation could be extremely heterogeneous at farm levels due to contrasting soils and farming practices. A generalized version of field conditions could be portrayed by using some representative values but actual farm-level situations may vary dramatically according to crop variety, irrigation method, water quality and soil management practices. Incorporation of such information when devising climate-change mitigation strategies would play a pivotal role in offsetting the adverse climate change impacts.

Although climate change could reduce the ET_c and IWR of both crops, it may also negatively affect the yields. Sultana et al. (2009) reported a 6–10% wheat yield decline for ten representative sites spread across Pakistan, due to a shortened growth period caused by a 1 °C temperature rise over the whole season. In this study, the yield reduction influences of climate change were not included but should be an integral part of future proposed studies. A trivial short-term strategy to cope with climate change would be to alter sowing dates to achieve optimum yield and water consumption levels according to the future temperature and rainfall trends. The development of suitable crop varieties to resist climate change may be the long-term and reliable solution to the problem.

We were only able to predict an overall decline in the crop growth periods under the projected thermal regimes. The magnitude of warming at various growth stages could have varying impacts over the final yields. Thermal fluctuations during growth periods have the ability to prolong or shorten certain growth stages, ultimately affecting the yields. Therefore, temperature rise influences on crop development process should be examined at different crop growth stages. Apart from climate variables, crop physiological response to increased future CO₂ concentrations could also play a decisive role in ET_c estimation and final yield levels. High uncertainties associated with projected rainfall amounts and intensities require careful monitoring to pinpoint the direction of change in the future.

Despite the shortcomings, the study still provides valuable information regarding the response of the wheat-rice system of Punjab to climate change. Stakeholders can employ our results to improve management of the UCC irrigation system under current climate conditions and to meet any challenges that may arise because of climate change in the future. This study also provides key information to irrigation managers for better allocation of limited water resources according to crop water requirements.

In this study, the future climate change threats were discerned concerning the growth periods and water demands for the wheat-rice system of Punjab, Pakistan. Distinctively strong warming rates distinguished the future wheat- from the rice-season, whereas the T_min remained markedly higher than the T_max. The possibilities of receiving more annual rainfall were projected primarily because of a sharp rise in erratic monsoon rainfalls during the future rice season; while the wheat season rainfalls were unchanged. Hence, the area may experience cycles of hot/dry winters followed by hot/wet summers, which could also result in frequent episodes of drought and floods.

The temperature rise induced positive changes in reference crop evapotranspiration (ET_o) featuring consistently higher increase rates of the wheat season. The ET_o change magnitude in the context of climate warming was unnoticed as the projected solar radiation decline counteracted much of the inverse temperature rise influences.

The future ET_c and IWR declined because of the shortened growth periods instigated by the temperature rise. Steeper recessions in growth spans and the associated ET_c and IWR for the wheat were attributed to higher climate-warming rates inherent to the future wheat season. A prominent negative shift in wheat IWR was projected despite the marginal ER increments, whereas, for rice, it was less noticeable beside substantial ER increments.

The early maturing wheat crop may manifest significant yield losses, thus rendering it more vulnerable to climate change than rice, which may have some more resilience against future climate warming. Moderate negative shifts in rice IWR pointed out that the beneficial contribution of projected rice seasonal ER would be limited due to the intense/erratic nature of the future rainfall events.

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