Evaluating the impact of projected CO₂, temperature, and rainfall change on groundwater resources in a rice–wheat dominated cropping region of northwestern India

In the past few decades, alteration in climatic parameters due to global warming has affected the availability of surface and groundwater resources worldwide, particularly in water-scarce arid and semi-arid regions. For instance, in the northwest Indo-Gangetic plains of India, high water-consuming rice–wheat crop rotation with inadequate canal water supply has resulted in an alarming decline in groundwater levels (Ministry of Agriculture – MOA 2013). Since the emission of greenhouse gases (carbon dioxide (CO₂), methane (CH₄), nitrous oxide (N₂O), and others) is increasing due to industrialization and changing human lifestyle, rise in temperature and changed precipitation patterns are directly or indirectly influencing local and regional hydrological and agricultural ecosystems. Under the representative concentration pathways (RCPs), the global average temperature is projected to increase by 1.7, 2.6, 3.1, and 4.8 °C under RCP2.6, RCP4.5, RCP6.0, and RCP8.5, respectively (IPCC 2013). The atmospheric CO₂ concentration is projected to increase to 430, 530, 670, and 936 ppm from the base level of 330 ppm under RCP2.6, RCP4.5, RCP6.0, and RCP8.5, respectively, by the end of the 21st century (IPCC 2013).

While the increase in air temperature leads to a direct rise in evaporation (Nand et al. 2021), the CO₂ concentration influences plant transpiration in two ways: diminishing of stomatal conductance with a decrease of stomatal opening (Hong et al. 2019; Bhargava & Mitra 2021) and a rise in leaf area index (LAI) and cell expansion leading to enhanced crop water requirement during the early growing period (Ficklin et al. 2010; Kirschbaum & McMillan 2018). Morison (1987) reported a 40% linear reduction in leaf stomatal opening with the doubling of CO₂ concentration from 330 to 660 ppm. Similarly, elevated CO₂ concentration resulted in an increase in LAI and biomass (Mndela et al. 2022), a reduction in transpiration rate and a consequent increase in water use efficiency (Eamus 1991; Field et al. 1995; O'Grady et al. 2011; Islam et al. 2012b; Hatfield & Dold 2019), reduction in crop evapotranspiration (Islam et al. 2012a, 2012b; Priya et al. 2014), increase in global mean runoff, and decrease in groundwater recharge (Ficklin et al. 2010; Andaryani et al. 2022). Reinecke et al. (2021) reported large uncertainties in modeling groundwater recharge as projected changes strongly vary among the different global hydrological models (GHMs)–global circulation models (GCMs) combinations due to the complicated process feedbacks between vegetation and water (transpiration changes due to available water together with vegetation productivity) and complex feedbacks between changes in CO₂, temperature, and precipitation which affect vegetation. In other words, the projected increase in CO₂ concentration under different climate change scenarios is likely to influence crop evapotranspiration, irrigation requirement, groundwater recharge, and groundwater extraction in agriculturally dominant regions.

Although the impacts of climate change on evapotranspiration, irrigation requirements, water footprints, and groundwater draft have been studied in India (Chattaraj et al. 2014; Abeysingha et al. 2016; Kambale et al. 2016; Dar et al. 2017; Mali et al. 2021), the interaction of atmospheric CO₂ concentration with different climate and crop parameters has not been accounted for. Furthermore while, the integrated impact of increasing levels of temperature and CO₂ concentrations on crop evapotranspiration, water use efficiency, and crop yield has been assessed (Jalota et al. 2009; Priya et al. 2014; Mohanty et al. 2015; Pandey et al. 2017; Lenka et al. 2020), the combined effect of elevated CO₂ and climate change on water budget components including groundwater behavior has not been studied. Kumar et al. (2022) modeled the impact of climate change on groundwater and adaptation strategies for its sustainable management for the Karnal district of the northwestern region of India, but the interaction of projected CO₂ concentrations with the anticipated changes in temperature and rainfall was not accounted for. It is hypothesized that elevated atmospheric CO₂ concentration levels under changing rainfall and air temperature regimes of different climate change scenarios would influence water budget components and consequently groundwater extraction and water table in a groundwater irrigated region.

Groundwater flow models are commonly used to study complex interactions in groundwater systems for evaluating recharge, discharge, aquifer storage processes, and sustainable yield, and predicting effects of management measures, thus enabling water managers and decision makers to make informed decisions for sustainable development and management of groundwater resources (Zhou & Li 2011; Rossman & Zlotnik 2013). Over the years, several groundwater flow models such as the United States Geological Survey (USGS) three-dimensional finite difference model (Trescott 1975), the USGS MODFLOW (McDonald & Harbaugh 1988), FEMWATER (Lin et al. 1997), HST3D (Kipp 1997), SEEP2D (Jones 1999), SUTRA (Voss & Provost 2002), and FEFLOW (Diersch 2014) have been developed. These simulation models use different numerical discretization schemes, i.e., finite difference (FD), finite element (FE), and finite volume (FV) for solution of the groundwater flow equation. The FD-based MODFLOW uses structured rectangular grids to represent the model domain and maintains mass-balance but fails to simulate complex geological features, such as angled faults and steep hydraulic gradients such as rewetting/drying cells (Kumar 2019). On the other hand, the FE-based commercially available FEFLOW model uses a triangular mesh to represent the model domain, can incorporate complex geologic features and also simulate wetting/drying of cells, but local mass conservation is not guaranteed (Kumar 2019). The choice of a particular model depends on the complexity of model setup and development, ease and accuracy of representing aquifer systems, availability of data related to aquifer properties and geology of the porous media, the ease with which the model input and output files could be prepared, availability of graphical user interfaces (GUIs), pre- and post-processing modules, project time-frame and cost involved, future use of the model, etc. The MODFLOW model has been widely used worldwide because of its extensive testing, flexible modular structure for integration of additional simulation capabilities, complete coverage of hydrogeological processes, public-domain free availability, and continued institutional support, and is regarded as the industry standard in groundwater modeling (Zhou & Li 2011; Rossman & Zlotnik 2013). There are several Windows-based GUIs, post-processors for MODFLOW such as Processing Modflow (Chiang & Kinzelbach 2001), Visual Modflow (Waterloo Hydrogeologic 2019), Groundwater Vista (Rumbaugh & Rumbaugh 2017), MODPATH (Pollock 2016), etc., which makes the modeling process much easier and saves time in analyzing model results. Other approaches like machine learning and statistical models are also being used to simulate groundwater levels globally (Mohanty et al. 2013; Adombi et al. 2022), but they work on the statistical relationship between predictors and predictand variables and do not consider associated physical mechanisms. Furthermore, they forecast the long time–series data at a point scale, while MODFLOW has the capability to account for spatial variability in fluxes owing to land use pattern, lithology, physical interaction of surface water bodies, and aquifer systems, in addition to interaction with physical boundaries while simulating the groundwater flow at the regional scale. Thus, this model accounts for all factors and boundary conditions prevailing in the study region, and provides the opportunity to simulate very near to actual conditions. MODFLOW has been successfully applied to study groundwater behavior at different locations of India (Eastern Gangetic Plains of India, Western Uttar Pradesh, western region of Chhattisgarh, Odisha (Mahanadi Delta)) including the study region (Alam & Umar 2013; Sahoo & Jha 2017; Mali et al. 2021; Gobinath et al. 2022; Kumar et al. 2022).

In this study, an attempt has been made to study the combined effect of anticipated changes in rainfall and air temperature on different water budget components (crop evapotranspiration, irrigation requirement, deep drainage, runoff) of a groundwater irrigated rice–wheat dominated cropping region, under two different scenarios of CO₂ concentration, namely, (i) changing CO₂ concentration with time corresponding to different RCPs and (ii) keeping CO₂ concentration constant at a reference (default) level of 369.41 ppm. The AquaCrop and MODFLOW models were used for simulating water budgeting components and their effects on the groundwater of the study region. While the AquaCrop model, capable of accounting for the effect of CO₂ level in estimating daily water balance components, simulates the yield response to water and is mostly used for field-scale irrigation management and water resources planning (Foster et al. 2017; Xing et al. 2017), the numerical MODFLOW includes complex interactions in groundwater systems (Khadri & Pande 2016; Hughes et al. 2022) and is used to simulate spatiotemporal variation in groundwater condition in the study region of northwestern India.

Weather data on air temperature, rainfall, wind velocity, sunshine hours, and relative humidity for the period 1981–2010 was collected from the meteorological observatory situated at ICAR-Central Soil Salinity Research Institute (CSSRI), Karnal. Since rice–wheat is the dominant cropping system of this region, these two crops were grown in two consecutive crop calendars (2014–2015 and 2015–2016) at the research farm of ICAR-CSSRI, to generate field data (Kumar et al. 2022) for parameterization and validation of the AquaCrop model.

For assessing climate change impacts on groundwater levels, the climate change projections of the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC 2013), based on RCPs were used in this study. Bias-corrected and spatially disaggregated (BCSD) monthly projections of rainfall and temperature for the period 1950–2099 at 0.5° × 0.5° resolution were derived from the World Climate Research Program's (WCRP's) Coupled Model Inter-comparison Project phase 5 (CMIP5) multi-model dataset (https://gdo-dcp.ucllnl.org/downscaled_cmip_projections/dcpInterface.html#Projections:%20Complete%20Archives). The CMIP5 multi-model datasets have been used in several climate change impact studies due to their ability to provide reliable and robust results (Sonali et al. 2017; Zhang et al. 2019; Abeysingha et al. 2020; Doulabian et al. 2021; Kumar et al. 2022).

Since outputs of different GCMs for some regions vary considerably, there has been growing interest in using multi-model ensembles from multiple GCMs to account for the uncertainty associated with individual GCMs in impact assessment studies. In this study, the hybrid-delta method (Hamlet et al. 2010; Islam et al. 2012a, 2012b) was used to generate multi-model ensemble climate change scenarios from multiple GCM projections for four different RCPs. The delta change method, the most commonly used method for generating future climate scenarios, does not consider variability or change in time series behavior in the future. On the other hand, the hybrid-delta method considers inter-annual variability for each month (Hamlet et al. 2010; Islam et al. 2012c; Tohver et al. 2014) using different scaling factors to each month of the historic time–series based on where it falls in the probability distribution of monthly values (Dickerson-Lange & Mitchell 2014). In this method, BCSD monthly GCM data (historical as well as future) were first disaggregated into individual calendar months. BCSD monthly GCM data (temperature and precipitation) were derived for historical (1981–2010, base period) and for the future periods of 2010–2039 (early century), 2040–2069 (mid-century), and 2070–2099 (end-century). The cumulative distribution functions (CDFs) were then developed for each month for historical and future time-periods (2020s, 2050s, and 2080s). For creating an ensemble of multiple GCMs/runs, data from multiple GCMs/runs were used for developing historical and future CDFs. Similarly, the CDFs for the observed time series data (1976–2005) were also developed, and the non-exceedance probability for each of the observed data was computed. Quantile mapping (Wood et al. 2002) was applied to re-map the observations onto historical and future CDFs for each month to obtain the historical and future GCM projected rainfall and temperature data at non-exceedance probability of the observed data. A step-by-step procedure for generating an ensemble of multiple GCMs using the hybrid-delta method is described in Tohver et al. (2014).

The amounts of groundwater withdrawal to meet the water requirements of different sectors including agriculture and replenishment (deep drainage) were treated as upward and downward fluxes, respectively, in the numerical model for simulation of groundwater levels. More than 85% of the study area was covered under the rice–wheat cropping system, while the remaining portion was under plantation, built-up, water sources, and other land uses. The groundwater irrigation (draft) and replenishment (deep drainage) from the rice–wheat cropping system were simulated using the AquaCrop model. The draft and replenishment components for other land uses were estimated using the standard procedures and methodology (GEC 2009; Narjary et al. 2021; Kumar et al. 2022). The general methodology used for calculating upward and downward flow components is briefly described below.

The upward flux from the experimental rice–wheat fields was estimated using the AquaCrop model considering weather, soil, and crop data and the water application criterion (Raes et al. 2022). Similarly, the draft for the domestic sector was calculated by considering human water consumption of 200 litres day⁻¹ per person (Shaban & Sharma 2007) and the population of the area derived from the census data (www.census2011.co.in/district.php). The upward flux from the plantation area was estimated considering the water consumption of 1,500 mm year⁻¹ for the Eucalyptus plantation (Minhas et al. 2015), which was the major tree planted in the study area (Kumar et al. 2022). It was assumed that groundwater pumping will be zero in the remaining parts of the study area because either land was barren or under water bodies.

In relation to crop seasons, a comparatively higher increase in maximum and minimum temperatures was projected during the wheat-growing season (November–April) under all RCPs. The mean temperature, maximum temperature, and minimum temperature are projected to increase in the range of 1.3–1.6 °C, 1.2–1.5 °C, and 1.4–1.7 °C, respectively, during the early century under different RCPs (Figure 4(b)). Similarly, the mean, maximum and minimum temperatures are projected to increase in the range of 2.0–5.2 °C, 2.0–5.2 °C, and 2.0–5.3 °C, respectively, during the end-century under different RCPs (Figure 4(b)). Similar to rice, the wheat-growing season also recorded a greater increase in minimum temperature than that of maximum temperature during all the future periods. During the wheat-growing season too, there is a marginal increase (≤10%) in rainfall under all RCPs. The increase in rainfall varied in the range of 5.3% (RCP2.6)–8.4% (RCP6.0), 3.3% (RCP8.5)–5.8% (RCP6.0), and 3.1% (RCP8.5)–8.4% (RCP2.6), during the early, mid-, and end-century periods, respectively. Though there is an increase in rainfall during the wheat-growing season under RCP8.5, this increase in rainfall gradually decreases from the early century (6.9%) to mid-century (3.9%) and to the end-century periods (3.1%) (Supplementary Table S2).

The field water budgeting components (crop evapotranspiration, irrigation water requirement, deep drainage, and runoff) under different climate change scenarios were simulated and the changes were analyzed with reference to the mean value of the particular component for the base period.

The average (1981–2010) wheat crop evapotranspiration (ET_cw) was estimated as 368.8 mm. An increase in ET_cw over the base period is projected for both scenarios under all the RCPs and future periods, however, with ET_cw being higher in Scenario-II than in Scenario-I. In Scenario-I, the increase in ET_cw varied in the range of 43.8%–48.3% under different RCPs in the three future periods (Figures 5(b) and 6(b)) (Supplementary Table S2), whereas in Scenario-II, the increase in ET_cw varied in the range of 45.4%–61.4%. Furthermore, the increase in ET_cw is projected to be higher during the end-century period as compared with the mid- and early-century periods. Of all RCPs, the maximum increase of 48.3% and 61.4% in ET_cw from the reference level (368.8 mm) is projected under RCP4.5 (Scenario-I) and RCP8.5 (Scenario-II), respectively, during the end-century.

The average rice irrigation requirement (IR_r) of the base period (1981–2010) was estimated as 1,146.2 mm. A marginal (<10%) decrease in IR_r is projected during future periods under all RCPs for both scenarios, except for an imperceptible increase of 0.3% under RCP2.6 and Scenario-II during the end-century period (Figures 5(a) and 6(a)). IR_r under Scenario-I is projected to decrease in the range of 0.7%–6.0% considering all future periods and RCPs, the corresponding range for Scenario-II being 0.3%–1.6% (Figures 5(a) and 6(a)). Hence, IR_r is projected to decrease 2.3%–6.3% more in Scenario-I than in Scenario-II. The average wheat irrigation requirement (IR_w) was estimated as 308.3 mm for the base period. Similar to the wheat crop evapotranspiration, there is an increase in IR_w during the future periods, and it has a similar pattern as that of crop evapotranspiration. The wheat irrigation requirement during the future periods is projected to increase in the range of 51.1%–57.7% and 52.5%–73.8% under Scenario-I and Scenario-II, respectively, as compared with the base period under different RCPs and future periods (Figures 5(b) and 6(b)) (Supplementary Table S2). Hence, an increase in irrigation requirement is projected to be 1.4%–16.1% higher in Scenario-II than in Scenario-I. It was also found that the change in IR_w is maximum under RCP2.6 and minimum under RCP6.0 in Scenario-I. However, in the case of Scenario-II, the highest and the lowest change in IR_w is projected under RCP8.5 and RCP2.6, respectively.

Since deep drainage (D_d) was negligible in the wheat crop, D_d from the rice field only is discussed here. In the case of the rice field, the increase in D_d is projected to be less than 5% under all RCPs and future periods. The increase in D_d was found to be higher (0.4%–4.0%) in Scenario-II than in Scenario-I (0%–2.4%). Simulation results also indicated that D_d would be the highest under RCP8.5 and the lowest under RCP2.6 for both the scenarios of CO₂ concentration (Scenario-I and II).

As there was no runoff during the wheat season, runoff from the rice crop field is discussed here. Simulation results showed an increase in runoff under different RCPs in the three future periods with slightly higher increase when CO₂ concentration was kept constant (Scenario-II) than in the case of increasing CO₂ concentrations (Scenario-I). Runoff is projected to increase by 27.9%, 56.7%, and 62.2% vis-à-vis for the base period (14.7 mm) in Scenario-I and by 30.6%, 58.4%, and 64.4% in Scenario-II under RCP2.6 during the early, mid-, and end-century periods, respectively (Figures 5(a) and 6(a)) (Supplementary Table S1). The maximum and minimum increase in runoff is projected under RCP8.5 and RCP2.6, respectively, in both the scenarios (Scenario-I and Scenario-II) for the simulation periods. Irrespective of climate scenarios, the increase in runoff was the maximum during the end-century and the minimum during the early century period.

Our results indicated an increase in maximum and minimum temperature during future periods, and this increase is projected to be higher during the wheat-growing season (November–April) than in the rice-growing season (June–October). These results are in confirmation with the finding of a study conducted in central India where 1–5 °C increase in maximum and minimum temperature was reported for the winter season coinciding with the wheat-growing period as compared with 1–3 °C increase during the summer season (Kundu et al. 2017). Our results are also in close proximity to the reported projected rise in mean temperature of 3.11 and 5.46 °C under RCP4.5 and RCP6.0, respectively, for the end-century period (Dar et al. 2019). However, the projected rise in mean temperature (2.5–4.4 °C) reported by Bal et al. (2016) for India was less than the projection made in the present study. Sanjay et al. (2020) also reported a mean temperature rise of 4 °C under RCP8.5 at the end-century period for most parts of India and still higher (>5 °C) for semi-arid northwest and north India. It is therefore clear that though the magnitude of changes in projected temperature can vary from place to place depending upon the areal extent considered, the rise in temperature is imminent in the coming future. Our study also projects an increase in rainfall during future periods, up to 21% during the rice-growing monsoon season and <10% during the wheat-growing winter season. Bal et al. (2016) also projected an average increase in rainfall in the range of 15%–24% for all of India. Dar et al. (2019) also projected a similar, though slightly higher increase in rainfall, while Kulkarni et al. (2020) reported a lower (10%) increase in mean annual precipitation under RCP4.5 for the mid-century period, and more than 20% increase for the end-century period over northwest India. The present study and results of the other studies conducted elsewhere revealed an increase in rainfall, though with varying magnitude, during future periods due to climate change effects. The present study also revealed that an increase in temperature during the rice as well as the wheat-growing season is likely to increase evapotranspiration, and thereby crop water demand and irrigation requirement. But, an increase in rainfall during the rice-growing season may offset the effect of increased temperature on the rice irrigation requirement. However, the wheat irrigation requirement is projected to increase in future owing to a rise in crop water demand due to rising temperature because only a slight increase (<10%) in rainfall is projected.

Simulation results showed lower crop evapotranspiration of rice (ET_cr) in Scenario-I as compared with that of Scenario-II during the future periods. The lower increase in ET_cr during the future periods may be attributed to the steady increase in CO₂ concentration under Scenario-I as compared with Scenario-II, where CO₂ concentration remained constant (i.e., 369.4 ppm) during the future periods. The increased CO₂ concentration results in reduced stomatal opening and thereby reduced ET (Yang & Lei 2022). Some previous studies also showed that rising CO₂ concentration decreases the stomatal conductance, which causes lower transpiration and thus crop evapotranspiration (Bernacchi et al. 2007; Lenka et al. 2020; Liao et al. 2021). Allen et al. (2003) also reported that an increase in CO₂ concentrations from 350 to 700 ppm reduced soybean evapotranspiration. Ficklin et al. (2010) also reported a decrease in daily alfalfa ET (−0.3 and −0.9 mm day⁻¹) under increasing temperature (+1.1 and 6.4 °C) and CO₂ (550 and 970 ppm) concentration. This confirms that rising CO₂ concentration impacts crop evapotranspiration, though the magnitude is variable for different crops. Furthermore, in Scenario-II (constant CO₂ concentration), rice evapotranspiration (ET_cr) is projected to increase during the future periods under all RCPs with different magnitudes. Our results project an increase of 1–5 °C in air temperature, which might have increased the drying capacity of the atmosphere, thus leading to more evapotranspiration. Abeysingha et al. (2016) also reported a 3%–9.6% increase in rice evapotranspiration (ET_cr) at the end of the 21st century under rising temperature with a constant CO₂ scenario that concurs well with the present study for Scenario-II. Our study also revealed that ET_cr would increase steadily during the future periods in Scenario-I under RCP2.6 in which the CO₂ level is projected to increase until the end of the early century and have a slightly decreasing trend thereafter. This premise is that the increasing temperature might be the dominant factor for increase in ET_cr under RCP2.6. In contrast, ET_cr is projected to decrease gradually under RCP8.5 during the mid- and end-century periods. A substantial increase in projected CO₂ concentration under RCP8.5 during the mid- and end-century periods might have negated the effect of increasing temperature on ET_cr (Ficklin et al. 2010). Allen et al. (2003) also reported that a doubling of CO₂ concentration results in a reduction in ET of rice by 9%. Priya et al. (2014) also reported that the effect of a 2.5 °C rise in temperature in Varanasi, India, was offset by an increase in CO₂ levels up to 660 ppm.

Results of our study also projected increasing wheat crop evapotranspiration (ET_cw) during the future periods in both scenarios (I and II) under all RCPs. However, an increase in ET_cw will be slightly lower in Scenario-I than Scenario-II, probably due to the dominance of increasing CO₂ concentration over lower temperatures as wheat in this area is grown during the winter season where most of the time, the maximum temperature varies between 5 and 30 °C. Results of this study are in line with the finding of Allen et al. (2003) who reported that the increase in CO₂ concentration from 350 to 700 ppm reduced the crop evapotranspiration in the temperature range of 18–28 °C, while there was little effect in the temperature range of 30–40 °C. Shimono et al. (2013) also observed reduction in evapotranspiration because of elevated CO₂ and affected by higher temperatures. This suggests that elevated CO₂ effects might be moderated by higher ambient temperature. A considerable increase in ET_cw over the simulation period was simulated in Scenario-II also, which indicates an effect of rising temperature during a wheat-growing period (Nand et al. 2021). The finding of the present study confirms the trend of previous findings (Chattaraj et al. 2014; Abeysingha et al. 2016; Dar et al. 2017; Qu et al. 2019), which revealed an increase of 7.8% to 16.3% in ET_cw at the end of the 21st century under different rising temperature scenarios. Hence, the overall results of the present study clearly indicate that an increase or decrease in future ET is driven both by rising temperature as well as variable CO₂ concentration; the dominant factor being dependent on the level of increase in temperature and CO₂. Savabi & Stockle (2001) reported that ET_c of corn decreased when the mean temperature increased by 1.2 °C and CO₂ increased to 480 ppm. However, when the average temperature increased beyond 1.2 °C and CO₂ exceeded 480 ppm, ET_c rose marginally. In a soybean crop, increase in CO₂ concentration from 330 to 350 ppm and increase in daily average temperature by 2.8 °C resulted in an increase in ET_c by 4%, while in the case of constant temperature and rising CO₂ scenario, ET_c reduced considerably.

Our study showed a decrease in irrigation requirement of the rice crop (IR_r) during future time periods in both the scenarios (I and II) under all RCPs due to an increase in rainfall. Furthermore, the projected steady increase in CO₂ concentration under Scenario-I also resulted in reduced evapotranspiration and rice irrigation requirement compared with constant CO₂ concentration (Scenario-II). Hence, the effect of rising temperature on IR_r is probably moderated by the increasing CO₂ concentration effects on crop evapotranspiration (Allen et al. 2003; Priya et al. 2014; Lenka et al. 2021; Yang & Lei 2022). The highest decrease in IR_r is projected for the end-century period under RCP8.5 in Scenario-I as the lowest ET_cr and the highest rainfall were projected during the end-century period under RCP8.5. However, in Scenario-II, increasing rainfall has a singular impact on IR_r under constant CO₂ concentration resulting in a lower decrease in IR_r than in Scenario-I. Our results confirm the observations of Abeysingha et al. (2016), who projected a <5% decrease in irrigation requirement of rice in north India due to an increase in rainfall at the end of the year 2099 and also projected a decline in IR_r for future periods. Dar et al. (2017) also projected a decline in rice irrigation in future under changing climate.

Our results showed an increase in irrigation water requirement of the wheat crop (IR_w) also during the future periods under all RCPs in both the scenarios (I and II). This may be attributed to a substantial increase in ET_cw with rising temperature and negligible (<10%) increase in rainfall during the wheat-growing period. Similar to ET_cw, simulated IR_w is slightly higher in Scenario-II than in Scenario-I. The present study also projects a significant reduction in IR_w in Scenario-I under RCP8.5 due to a lesser ET_cw because of a considerable increase of CO₂ concentration. Our estimates of increase in IR_w during the future time-periods are also in agreement with the findings of other studies of variable changes in wheat irrigation requirements under projected variability (rise or fall) in rainfall amount (Abeysingha et al. 2016; Xiao et al. 2020).

Our study projected negligible deep drainage (D_d) from the wheat field during the future periods. This might be due to the fact that a limited rainfall amount is expected during the winter season that could be effectively utilized by the crop or stored within the root zone. In contrast, an increase (though less than 5%) in D_d is projected from the rice field during future periods that could be attributed to projected higher rainfall and irrigation during the rice-growing period. The irrigation criterion to apply water at field capacity (Kumar et al. 2019) led to higher irrigation and consequent enhanced D_d from the rice fields.

The study further projected higher D_d in Scenario-II than in Scenario-I during the future periods. The lower irrigation associated with lower ET_cr under elevated CO₂ in Scenario-I resulted in lower D_d. In Mediterranean climatic conditions, van Ittersum et al. (2003) also reported a decrease in deep drainage from a wheat field under increasing temperature and elevated CO₂ concentration. Ficklin et al. (2010) also reported a decrease in deep drainage under elevated CO₂ and temperature, and projected a decline in cumulative groundwater recharge for alfalfa, almonds, and tomato crops.

Our simulation results showed an increase in runoff from rice fields during the future periods under all RCPs in both the scenarios of CO₂ concentrations (Scenario-I and II) due to the projected increase in rainfall. However, a negligible difference in simulated runoff under Scenario-I and Scenario-II indicates that rising temperature and CO₂ concentration may not have a substantial effect on runoff from the irrigated cropped land, and increasing rainfall might be the dominant factor affecting the runoff during the rice-growing monsoon season. Furthermore, the present study area is an irrigated land having almost flat topography where fields have ∼10 cm effective bund (embankment) height. The flat topography and on-farm rainfall storage, facilitated by field bunds, might be the reason for little effect of climate change on runoff during the simulation period. This is in contrast to a reported increase in runoff at the watershed level under elevated CO₂ and higher temperature and resulting reduction in crop evapotranspiration (Leipprand & Gerten 2006; Niu et al. 2013; Butcher et al. 2014). Islam et al. (2012c) also reported that changes in temperature had a relatively lesser effect on the magnitude of annual and seasonal stream flow as compared with rainfall changes in the Brahmani River basin. Our results also project the highest and lowest runoff under RCP8.5 and RCP2.6, respectively. Under RCP8.5, an increase in rainfall of 6.5%, 13.1%, and 21.2% resulted in higher runoff than the runoff resulting from 7.8%, 9.0%, and 7.6% rises in rainfall under RCP2.6 during early, mid-, and end-century, respectively. Working on Dez basin in Iran, Ehteram et al. (2018) reported that reduction in precipitation amounts in the future period (2011–2030) is the major reason for decreased runoff volume.

Our study showed a lower GWL decline under Scenario-I (increasing CO₂ concentration) than Scenario-II (constant CO₂ concentration) during the future periods. Hence, it is clear that the GWL decline is associated with CO₂ concentration as it influences crop evapotranspiration and consequently irrigation demand. Our study projects almost identical changes in GWL for the early century period in both the scenarios of CO₂ concentration (Scenario-I and II) due to negligible variation in irrigation demand under all RCPs. However, a remarkably lower GWL decline during the mid- and end-century periods is projected under elevated CO₂ (Scenario-I) than constant CO₂ (Scenario-II). The variation in GWL decline during the mid- and end-century periods may be due to the fact that in Scenario-I, elevated CO₂ might have offset the effects of rising temperature on crop ET, resulting in reduced groundwater withdrawal, which might not have occurred in Scenario-II due to the constant CO₂ concentration (Ficklin et al. 2010). This could be the reason for the highest projected increase in CO₂ concentration under RCP8.5 resulting in the lowest decline in GWL among all RCPs from the reference level.

In agricultural dominant groundwater irrigated regions, crop water requirement has profound impact on groundwater resources. It is expected that any change in temperature, precipitation, and elevated CO₂ concentration owing to climate change is likely to influence field water budgeting components (crop ET, irrigation, runoff, and deep percolation) and groundwater resources. Therefore, it is necessary to study the integrated effect of increasing CO₂ concentrations along with the change in rainfall and temperature regimes on field water budgeting components and consequently on the groundwater resources during future periods. The present study indicates that irrespective of RCPs, the rice irrigation requirement would decrease slightly (0.7%–6%) during future periods under both increasing (Scenario-I) and constant (Scenario-II) CO₂ concentration scenarios. In contrast, the wheat irrigation requirement is projected to increase significantly (53%–74%) under Scenario-I and Scenario-II. Furthermore, under RCP8.5, GWL is projected to decline by 9.2, 20.5, and 24.4 m from the reference level (18.85 m) at the end of the early, mid-, and end-century periods, respectively, in Scenario-I, whereas under Scenario-II, it is projected to decline by 10.7, 26.9, and 40.5 m from the reference GWL at the end of the early, mid-, and end-century periods, respectively. This suggests that the effect of rising temperature on irrigation requirements and groundwater decline may be moderated to some extent by elevated CO₂ concentration. Overall, the study emphasizes that CO₂ concentration, temperature, and rainfall must be considered in an integrated way while simulating the impact of climate change on groundwater and for devising sustainable mitigation and adaptation strategies in irrigated agricultural systems.

Though this study provides valuable information on the integrated effects of CO₂ concentration, temperature, and rainfall changes due to climate change on water budgeting components and groundwater, there may be uncertainty in the estimated components due to the model structure/model algorithms, and model parameters. Multi-crop models, rather than a single crop model like AquaCrop, may provide more realistic results in simulating the effect of increasing CO₂ concentrations, temperature, and rainfall change on water balance components (including groundwater) of irrigated agricultural areas. The median or average of the simulated water budgeting components of multi-crop models may be used in future studies for simulating the unified impact of climate change parameters on groundwater resources in an irrigated ecosystem and for formulating sustainable strategies for its optimal management. Further, in the present study, the estimated fluxes from different models/sources were loosely coupled with Visual MODFLOW Flex for groundwater simulation. The MODFLOW model used in this study considers constant fluid density, and does not simulate water flow in dry cells (unsaturated zone) and stream flow–aquifer interaction, although the model limitation of considering constant fluid density does not have much relevance in this study as fluid quality in the aquifer was very similar in the entire study area. However, better handling of water flow in dry cells can improve accuracy and reliability of the model results (Hunt & Feinstein 2012). A recent version of USGS Modflow (Modflow 6) simulates the water flow in dry cells as well as stream flow–aquifer interaction (Langevin et al. 2017). Thus, its use in future studies to represent aquifer systems more precisely may improve groundwater simulation further, and aid in developing robust groundwater management strategies for groundwater irrigated rice–wheat dominated cropping regions.

The authors acknowledge the Director, ICAR-CSSRI, Karnal (Research article/41/2022) for extending logistics support during the execution of this study.

This article does not report on or involve the use of any animal or human data or tissue.

This article does not contain data from any individual person.

All the authors have given their consent to publish the manuscript.

S.K.: conceptualized the methodology, supervision of work, and manuscript editing; V.N.: calibration and validation of MODFLOW and AquaCrop models; B.N.: data analysis and writing manuscript; P.K.H.: modeling and GIS work; A.I.: GCM projections and climate change projection, manuscript editing; R.K.Y.: field study; S.K.K.: refinement of methodology and editing results and discussion.

This research was funded by the Indian Council of Agricultural Research (ICAR) through the National Innovations in Climate Resilient Agriculture (NICRA) program.

Data that support the findings of this study are available from the corresponding author upon reasonable request.

The authors declare there is no conflict.