Alleviating water scarcity by alternative cropping systems in the North China Plain

Globally, groundwater supports the irrigation of crops responsible for approximately 13% of total food production, supplying over 40% of the water used in agricultural irrigation and serving as a vital water source for nearly two billion people¹. The use of groundwater boosts agricultural production and strengthens farmers’ ability to withstand climate shocks and water variability^2,3,4. Yet, the intensifying overuse of groundwater has emerged as a critical issue, largely fueled by the rapid expansion of irrigation systems and escalating extraction rates in response to population growth and climate-related pressures⁵. The emergence of groundwater over-exploitation threatens agricultural sustainability in many food-producing areas, including the High Plains Aquifer, Central Valley California, Middle Egypt, Indo-Gangetic Plain, and the North China Plain (NCP)^6,7. Therefore, better conservation and highly efficient use of groundwater are urgently needed to sustain groundwater irrigation for crop production.

The North China Plain (NCP) is characterized by a complex Quaternary alluvial aquifer system, which is the hydrogeological foundation of its agricultural productivity⁸. This system generally consists of a shallow, unconfined aquifer that is highly responsive to seasonal precipitation and irrigation return flow, and a deeper, semi-confined to confined aquifer system where severe, long-term depletion has occurred due to historical over-extraction. In the central alluvial plain, where our study site is located, the primary source of natural groundwater recharge is infiltration from precipitation, with lateral recharge from the piedmont areas being less significant⁹. Agricultural irrigation, dominated by the water-intensive winter wheat-summer maize rotation, is the largest consumer of groundwater, creating a substantial deficit between crop water demand and available rainfall. This over-extraction has also led to the upward movement of deep brackish and saline water in some areas, further complicating water management. While large-scale interventions like the South-to-North Water Diversion (SNWD) project have been implemented to augment regional water supplies, their primary contribution has been to urban and industrial sectors, with limited direct allocation to agricultural irrigation¹⁰. Consequently, despite a suite of groundwater governance measures, achieving groundwater neutrality, where sustainable pumping rates match long-term recharge, remains a formidable challenge.

Water-saving measures, including breeding for drought resistance, micro-irrigation, plastic or straw mulching, and water rights trading, have been vigorously promoted in water-shortage regions^11,12,13. While such strategies have contributed to higher crop yields and enhanced water use efficiency, they have not succeeded in reversing the rapid depletion of groundwater, primarily because extraction rates continue to surpass natural aquifer recharge^3,12,14. In the NCP, groundwater tables have dropped significantly—by approximately 1.14 m annually—mainly as a result of a persistent 300 mm deficit between annual rainfall and the water requirements of the intensively cultivated winter wheat–summer maize (WM) rotation^15,16. Winter wheat cultivation in the region heavily depends on groundwater irrigation, as merely 30% of total yearly rainfall is received during the winter months, and surface water availability is limited¹³. Thus, optimized irrigation regimes for wheat tend to occur during critical stages (two times at pre-sowing and jointing stage) or even minimized irrigation (only pre-sowing irrigation) with some yield penalty^16,17,18,19. These strategies have reduced the rate of groundwater decline but have not been able to maintain groundwater neutrality^20,21. Consequently, further adaptive measures are required to curb the ongoing depletion of groundwater resources while ensuring stable agricultural output.

A more profound strategy that has yet to be fully assessed is an alternative cropping system, which involves adjusting cropping system compositions to match crop production with local water resources^3,5,15. Global-scale crop redistribution of 14 major food crops based on a process-based crop water model showed that optimized crop configuration increased crop production and reduced blue and green water consumption by 14% and 12%, respectively². Both model and field experiments show that shifting the cropping system from rice to alternative cereals (maize, millet, or sorghum) with the lowest water footprint would save irrigation water while enhancing crop productivity and human nutrient supply in India³. Moreover, seasonal fallow (i.e., reducing cropping frequency) is often used in arid and semiarid regions to conserve rainwater and reduce crop failure²². Such approaches have sparked growing attention toward developing and assessing alternative cropping strategies aimed at conserving water in the NCP^13,23, particularly by lowering cropping intensity, i.e., reducing the frequency of harvests^14,15,24. However, current research has largely overlooked the full implications of such systems on both yield outcomes and water consumption, with limited attention given to net groundwater use metrics—defined as irrigation water applied minus natural recharge^14,20,21. The dual advantages of alternative cropping—namely, mitigating groundwater depletion and supporting food security—have yet to be widely acknowledged or rigorously quantified in existing literature.

Although fine-tuned irrigation practices within cropping systems have proven effective in reducing water consumption while sustaining yields, most research to date has concentrated on the conventional WM rotation. Despite the fact that altering cropping systems leads to significant changes in the spatial and temporal dynamics of water demand and availability, limited research has explored how irrigation influences both yearly crop output and groundwater usage in diversified systems^5,21,25. Thus, the irrigation strategies of alternative cropping systems need to be adjusted simultaneously to optimally match and time irrigation applications to meet the needs of individual crops^5,7. However, optimized irrigation strategies regarding the shifting of cropping systems have scarcely been investigated and quantified, preventing a systematic evaluation of alternative cropping systems on groundwater conservation.

In this study, a multi-decade (1994–2023) simulation was performed using the Agricultural Policy/Environmental eXtender (APEX) model, integrating historical climate data to evaluate irrigation strategies across various alternative cropping systems. The objectives of this research are threefold: (1) to evaluate crop yields and water productivity under different irrigation regimes within alternative cropping systems; (2) to quantify how optimizing cropping patterns influences evapotranspiration (ET) and groundwater dynamics; and (3) to identify alternative cropping configurations capable of achieving groundwater neutrality without compromising agricultural output.

Equivalent yield and water stress days

The traditional wheat-maize (WM) cropping system achieved the highest yield (14.6 Mg ha⁻¹) under full irrigation (FI) with almost no water stress days (Fig. 1 and Fig. 2). However, with a reduced cropping frequency, the winter wheat→WM, summer maize→WM, spring maize→WM, and spring peanut→WM rotations reduced yields by 25%, 25%, 27%, and 16%, respectively. Further decreasing the cropping frequency to single cropping resulted in more severe yield reductions (41–53%). Although the water stress days for WM increased from 4 d under FI to 17 d under critical irrigation (CI), no significant yield loss was observed (14.7 versus 14.0 Mg ha⁻¹). Similarly, reduced irrigation from FI to CI did not result in yield loss in alternative cropping systems (Figs. 1 and 2). However, further reduction in irrigation from FI to minimum irrigation (MI) caused a 14% yield loss in the WM system. Surprisingly, we found that the yield gap between alternative cropping systems and conventional WM decreased with reduced irrigation. For instance, the yield reduction of the three harvests in two years systems compared to WM was 23% under FI and 10% under rainfed conditions. The spring peanut→WM system could maintain the same yield level (11.8 Mg ha⁻¹) as the WM system under MI conditions (Fig. 1). Moreover, the yield of the WM system was reduced by 24% when the irrigation strategy shifted from FI to rainfed (RF), which only had a slight yield advantage over spring peanut→WM, summer maize→WM, and spring maize→WM, but was still 37% higher than the average yield of single cropping systems. The variation in yields increased with the reduction of irrigation due to the increased water stress days across cropping systems (Fig. 2). For example, the coefficient of variation (CV) for yields increased from 5% under FI to 19% under RF as the water stress days increased from 4d to 84 d. Overall, we found that the yield gap between alternative cropping and conventional WM strongly decreased with reduced irrigation.

Evapotranspiration and water productivity

The simulated annual evapotranspiration (ET) decreased with the reduction in irrigation and cropping frequency (Fig. 3). The WM system had the highest ET values of 778.6, 728.3, 647.5, and 549.7 mm under FI, CI, MI, and RF, respectively. Alternative cropping systems with three harvests in two years reduced ET by 27% under FI and 14% under RF. The spring peanut→WM rotation exhibited significantly lower evapotranspiration than the conventional WM system and showed slightly reduced or comparable ET values relative to the other cropping systems with three harvests over two years (Fig. 3). Furthermore, single cropping systems reduced ET by 46%, 43%, 38%, and 31% compared with the WM system under each irrigation strategy. The single summer maize system had the lowest ET among the nine cropping systems regardless of irrigation strategy, with 380.6, 373.9, 364.0, and 344.4 mm under FI, CI, MI, and RF, respectively.

The yield and ET responses of alternative cropping systems to different irrigation strategies resulted in varied ET-based water productivity (Fig. 4). The water productivity of the nine cropping systems under FI ranged from 1.5 to 2.3 kg m⁻³, with spring peanut→WM, single spring peanut, single summer maize, spring maize→WM, and summer maize→WM having higher water productivity than WM (1.9 kg m⁻³). In contrast, single spring maize, single wheat, and wheat→WM reduced water productivity (Fig. 4). Additionally, the response of water productivity to irrigation was more pronounced for summer maize→WM, WM, and single summer maize, with the highest differences among irrigation strategies being 10%, 6%, and 5%, respectively, and less pronounced for spring peanut→WM and single spring peanut, with a difference of 1% for both. The spring peanut→WM system achieved the highest water productivity among the nine systems, regardless of irrigation strategy, which was 14% higher than that of the WM system even under RF (Fig. 4). The lowest water productivity was observed in single spring maize across all four irrigation strategies, with values between 1.5 and 1.6 kg m⁻³.

Net water use and groundwater level

Net water use generally decreased with the decline in cropping frequency, and the differences among cropping systems narrowed with the reduction in irrigation (Fig. 5). With a lower cropping frequency, the three harvests in two years systems with irrigation reduced net water use by 53–62%. Among these cropping systems, spring peanut→WM and spring maize→WM had relatively low net water use regardless of irrigation strategy. Further reducing the cropping frequency to single cropping systems resulted in a significant decline in net water use. For example, single wheat saved nearly half of the groundwater under FI and CI. The net water use of single spring maize and single spring peanut was both below 35 mm. Moreover, the net water use of single summer maize under FI was less than zero (−97 mm), meaning that deep percolation was 97 mm greater than the irrigation depth, and the residual water eventually contributed to groundwater reserves. The net water use was reduced by 97% in single spring peanut and 30% in spring peanut→WM, representing the maximum and minimum reduction in net water use when the irrigation strategy shifted from FI to CI (Fig. 5). The single spring peanut system started to replenish groundwater (30 mm annually) with further irrigation reduction to MI. Under RF, all cropping systems could replenish groundwater, with less conservation effect for WM (13 mm) and more conservation effect for single summer maize (99 mm).

Under full irrigation (FI), excessive groundwater extraction associated with the conventional WM system led to a continuous decline in groundwater levels, amounting to a total drop of 19.8 m between 1994 and 2023—equivalent to an average annual decline of 0.66 m (Fig. 6). Irrigation reduction for the WM system mitigated the groundwater level decline. Still, the groundwater level under CI and MI continued to decline at rates of 0.48 and 0.24 m yr⁻¹, respectively, except for the rainfed WM system, which is categorized as a groundwater-neutral cropping system. Additionally, reducing the cropping frequency, especially for wheat, dramatically decreased the declining trend of groundwater levels across the four irrigation strategies (Fig. 6). The changes in groundwater level for spring maize→WM, spring peanut→WM, summer maize→WM, and wheat→WM under FI during 1994-2023 were −5.3, −5.9, −8.4, and −10.9 m, with annual decline rates of 0.18, 0.20, 0.28, and 0.36 m, respectively. Notably, the groundwater level stopped declining during 2009–2014 for most alternative cropping systems, even under FI.

For single cropping systems under FI, the groundwater level changed within 3 m by the end of the simulation. The declines in groundwater level induced by the three harvests in 2 years systems decreased on average to 5.1 m under CI and to 1.8 m under MI (Fig. 6). Single summer maize and single spring peanut under CI raised the groundwater level by 4.0 and 1.9 m, respectively. Under MI, the single summer maize, single spring peanut, single wheat, and single spring maize raised the groundwater level by 5.6 m, 3.8 m, 2.2 m, and 1.9 m, respectively. Single summer maize, single spring peanut, single wheat, single spring maize, spring peanut→WM, spring maize→WM, summer maize→WM, and wheat→WM under RF raised the groundwater level by 7.8 m, 6.7 m, 4.9 m, 4.5 m, 3.1 m, 3.0 m, 2.9 m, and 2.8 m, respectively.

The findings suggest that implementing alternative cropping systems in conjunction with efficient irrigation management can substantially reduce groundwater extraction while sustaining agricultural productivity in the North China Plain (NCP). With water shortages becoming more serious, farmers may have to reduce cropping frequency when pumping is not economically feasible or when groundwater is no longer available⁵. A study based on high-resolution satellite and census data indicated that cropping intensity decreased by 68% due to groundwater depletion in over-pumped aquifers in India⁶. In California, transitioning to crops with lower water requirements has been shown to decrease agricultural water use by as much as 93%⁵. Therefore, alternative cropping should be a top priority in national policymaking for water savings, crop production, and human health¹³.

In particular, we found that the yield of alternative cropping systems became more competitive under reduced irrigation compared to traditional WM. In other words, with the limited annual precipitation (560 mm), efforts to increase cropping intensity with water-intensive crops are futile due to water stress without groundwater irrigation. In contrast, with more crop diversification and reduced frequency, the spring peanut→WM system had the minimum yield reduction (from 16% under FI to 0.4% under RF) across irrigation strategies. This result is well-supported by Davis et al. ², who advocated that global food security and water sustainability could be achieved through optimized crop distribution, especially by increasing the production of crops such as peanuts while substantially moving away from crops such as rice and wheat². These findings suggest that the spring peanut→WM system has great potential for replacing the current WM system with minimal yield loss, especially in a water-scarce future.

Reducing consumptive water use (ET), is the only pathway to achieve ‘real’ water savings at the basin scale, as non-consumed water (e.g., runoff and deep percolation) remains within the local hydrological system and can be used downstream or contribute to recharge⁵. The ETs of the WM system ranged from 779 mm under FI to 550 mm under RF annually, which exceeded the long-term average annual precipitation by 88 mm even under MI. Thus, the conventional WM cropping system is implausible to maintain a groundwater neutral status unless the irrigated crop shifts to rainfed^19,22. Besides incremental changes in the current WM system, we found that alternative cropping systems substantially decreased annual ET, thus providing a higher potential for water savings^5,7.

In particular, with reduced cropping intensity, the annual ETs of single cropping systems are about 400 mm averaged across four irrigation strategies, 276 mm lower than the annual ET of WM and 160 mm lower than annual precipitation; this residual water will ultimately replenish the aquifer²⁴. However, the significant yield loss (~44%) from shifting WM to single cropping systems will jeopardize farmers’ benefits and food security, making it unlikely to be widely promoted currently in the North China Plain, except in severe groundwater funnel areas^19,26. In contrast, shifting from WM to three harvests in 2 years has minimal yield reduction, substantial water savings, and neutral or slightly positive long-term downward water flux, thus well balancing the trade-off between food security and groundwater sustainability^4,7,19. This reduction in consumptive use is critical, as it directly lessens the pressure on the heavily exploited shallow aquifer, which is most responsive to agricultural pumping. Remarkably, the spring peanut→WM system consumed less than 550 mm of water, saving 30% of water under FI, showing great promise for more sustainable groundwater management.

Achieving “more crop per drop” remains a fundamental objective in irrigated agriculture²⁷. Substantial variation in water use efficiency among crop species led to notable differences in the annual water productivity of the respective cropping systems⁵. Our results indicate that cropping system selection exerted a greater influence on water productivity than the choice of irrigation strategy. Modified WM-based rotations with three harvests in two years increased water productivity by 20% (up to 2.3 kg m⁻³ for the spring peanut→WM system under FI) compared to the WM system, while single-species continuous cropping resulted in a 20% decrease (to 1.5 kg m⁻³ for single spring maize under MI) in water productivity. Although monoculture systems can lower groundwater consumption, they underutilize available solar radiation and thermal resources, ultimately leading to diminished crop yields²¹. Among the nine cropping systems evaluated, the spring peanut→WM rotation—with three harvests over two years and continued inclusion of winter wheat—demonstrated the greatest potential for groundwater conservation and maximizing water productivity. Moreover, the spring peanut→WM system had higher yield stability than the WM system (CV of 8% vs. 11%), indicating the long-term advantages of diversified cropping systems. Evidence from field experiments showed that the legume pre-crop effect, such as the legacy effect of N and the disease-breaking effect, benefits the yield of the following crop^28,29,30. Additionally, the changes from shallow to deep root systems (peanut to wheat) enable effective rainwater preservation and complementary use of soil moisture, significantly improving their ability to adapt to climate variability^4,25. Furthermore, the cultivation of spring peanut has been fully mechanized in many parts of the region, which significantly reduces labor requirements and enhances its appeal to farmers. Therefore, it is worth considering the spring peanut→WM cropping system as superior to conventional WM for groundwater conservation and crop production in the future.

Our study quantifies the direct impact of agricultural management on the water balance at the root zone, which is a primary driver of changes in the shallow, unconfined aquifer. It is crucial to differentiate the dynamics of this shallow aquifer, which exhibits inter-annual fluctuations in response to precipitation and pumping, from the deeper, confined aquifers that are experiencing more severe, long-term depletion due to historical over-extraction^8,9. While our model estimates the rate of decline under different scenarios, defining a truly ‘sustainable pumping rate’ for the entire region is more complex. It would require a comprehensive basin-wide water budget that accounts for all inflows (natural recharge from precipitation, lateral flow from mountain fronts, inter-aquifer leakage) and outflows (pumping for all sectors, discharge to rivers, evapotranspiration from natural vegetation). Furthermore, large-scale projects like the SNWD primarily supplement urban and industrial water supplies, with limited direct allocation to agriculture in many areas¹⁰. Therefore, while they alleviate overall water stress, on-farm management remains the most critical lever for controlling agricultural groundwater abstraction. Our modeling approach, by focusing on the change in net water use (Irrigation – Deep Percolation), effectively isolates the agricultural contribution to groundwater stress. This provides a clear, quantitative signal of how farming practices alone influence the potential for aquifer recharge or depletion, even if the absolute groundwater level is also affected by other regional factors.

The traditional WM cropping practice in the NCP is facing growing sustainability concerns due to excessive groundwater extraction and the escalating impacts of climate change³¹. We found that the groundwater table declined continuously by 0.7 m yr⁻¹ for the WM under FI, primarily driven by the imbalance between rainfall availability and crop water demand—most notably for winter wheat. The WM system with reduced irrigation mitigated the rapid decline of the groundwater level but at the expense of crop yield. Alternatively, the spring peanut→WM system saved 90–218 mm of irrigation water, mitigating the groundwater table decline while maintaining a higher yield. Additionally, retaining wheat in the spring peanut→WM system avoids significantly decreasing the intensity of main grain ration production and forms adequate coverage to reduce soil erosion³². Moreover, peanuts have great potential to reduce N fertilizer input and associated environmental costs due to its inherent capacity for symbiotic atmospheric N fixation. In economic terms, peanut cultivation offers superior annual returns compared to the WM system, underscoring its potential as a profitable alternative²⁶. Notably, transitioning to water-saving cropping systems can be achieved without the need for extensive technological investments³. Overall, our results indicate that the spring peanut→WM system is a promising alternative for human nutrition production and groundwater conservation, especially under limited irrigation in a more water-scarce future.

Our study has certain limitations that should be acknowledged. First, our model calibration and scenario analysis, consistent with previous work in this specific experimental setting, did not simulate significant surface runoff. This is based on the high infiltration capacity of the local silt loam soils and the common practice of constructing field bunds to retain rainwater in the flat terrain of the NCP. However, we acknowledge that under extreme monsoonal precipitation events, runoff could occur. By largely excluding runoff, our model might overestimate deep percolation and, consequently, the potential for groundwater recharge. This could lead to a more optimistic assessment of the mitigating effects of alternative cropping systems on groundwater decline. Future studies at a watershed scale should explicitly account for runoff generation and routing to provide a more complete picture. Second, this study utilizes a field-scale model (APEX) to simulate processes at a representative point and then extrapolates these findings to infer changes in the regional groundwater table. This approach simplifies the complex, three-dimensional nature of groundwater flow. The actual groundwater level at any given location is the result of an integrated response to spatially variable inputs (recharge) and outputs (pumping) across a wide area, modulated by aquifer properties and hydraulic gradients. Therefore, our calculated changes in groundwater level (ΔGWT) should be interpreted as a direct, localized potential impact of agricultural practices on the water column beneath the field, rather than a precise prediction of the regional piezometric surface. While this method effectively isolates the agricultural signal, it introduces uncertainty. Validating these findings with a spatially distributed, physically-based groundwater flow model would be a valuable next step to enhance the robustness of regional predictions.

This study assessed the effects of alternative cropping systems and irrigation strategies on crop yield, water productivity, and groundwater levels in a typical groundwater depletion region in China. Although a deficiently irrigated wheat-maize (WM) system saved some irrigation water, it lost 5–24% of yield and was not groundwater neutral except when rainfed. We also found that the yield gap between alternative cropping and WM significantly decreased with reduced irrigation frequency, highlighting the importance of alternative cropping systems in future water-deficient scenarios. For instance, the yield and ET reduction of alternative cropping systems with three harvests in two years ranged from 23 to 27% under full irrigation to 11–14% under rainfed conditions compared to the corresponding WM system. Importantly, we found that the spring peanut→WM system achieved the highest water productivity among the nine cropping systems across irrigation strategies and maintained crop yield under minimum irrigation and rainfed conditions. Moreover, spring peanut→WM under critical and minimum irrigation reduced the decline rate of the groundwater table to approximately 0.1 m yr⁻¹, indicating great potential for minimal yield loss and maximum groundwater savings. Although single-species cropping systems considerably decreased ET and net water use, their yield penalty was unacceptable given the pressure to ensure food security. In conclusion, alternative cropping systems with reduced cropping intensity and irrigation showed great potential to achieve groundwater use balance. Specifically, the spring peanut→WM system could be a viable option to balance crop production and groundwater conservation in the North China Plain. Taken together, our study provides strong evidence for the co-benefits of alternative cropping systems in terms of crop production and groundwater conservation. This delivers a scientific basis for designing future sustainable cropping systems, especially in regions facing groundwater over-pumping worldwide.

Study area

The study was carried out in Wuqiao County (37°29′–37°47′ N; 116°19′–116°42′ E), situated in the southern region of the Yundong Plain in Hebei Province, China. As a representative zone of intensive agriculture within the North China Plain (NCP), the area is characterized by fertile alluvial soils and a warm temperate climate conducive to crop production⁸. The region experiences a mean annual temperature of 12.6 °C and a cumulative temperature above 0 °C reaching approximately 4863 °C. Based on meteorological records from 1994 to 2023, mean yearly precipitation is around 560 mm, with the majority (60–70%) concentrated in the summer months from June to August (Fig. 7b). The dominant soil type is Calcaric Fluvisol, exhibiting a silt loam texture as per the USDA classification. Comprehensive soil properties, including bulk density and hydraulic characteristics, are documented in Zhao et al.³³. Agricultural practices in this area are predominantly based on a high-intensity double cropping system, involving winter wheat followed by summer maize. However, limited access to surface water resources has led to over-dependence on both shallow and deep aquifers for irrigation, resulting in significant groundwater over-extraction. Current groundwater levels are recorded at an average depth of approximately 17 m.

APEX model description

The Agricultural Policy/Environmental eXtender (APEX) model, version 1501, was used for this study. APEX is a process-based, farm-to-small-watershed scale model that extends the capabilities of the field-scale EPIC (Environmental Policy Integrated Climate) model. It operates on a daily time step and can simulate long-term crop rotations and their environmental impacts. The model integrates key modules for climate, hydrology, nutrient cycling, crop growth, soil processes, and management operations.

Crop growth is simulated based on daily heat unit accumulation for phenology and interception of photosynthetically active radiation for potential biomass production, which is then adjusted by daily stresses (water, temperature, nutrient, and aeration). Crop yield is calculated from the final biomass and a harvest index, which can be reduced by stress during sensitive growth stages. APEX offers multiple options for simulating key hydrological processes. In this study, potential evapotranspiration was estimated using the Penman-Monteith method, which requires daily inputs of solar radiation, maximum and minimum temperatures, relative humidity, and wind speed. Surface runoff was estimated using the curve number (CN) method. Water percolation through the soil profile is simulated using a storage routing technique, and the version used here (APEX 1501) includes an improved subroutine for water percolation dynamics based on a variable saturation hydraulic conductivity method, enhancing the accuracy of soil water content simulations. A more detailed description of the model’s structure and algorithms can be found in Gassman et al.³⁴.

Model setup, calibration, and validation

The APEX model was parameterized for the specific conditions of our study site using data from long-term field experiments conducted at the Wuqiao Experimental Station. These experiments provided the necessary data for calibrating and validating crop and soil parameters for winter wheat, summer maize, spring maize, and spring peanut.

Data for model setup were sourced from a series of field experiments at the study site, as detailed in our previous publications^23,33. Crop Parameters: Specific parameters for winter wheat (“Jimai 22”), summer maize (“Zhengdan 958”), spring maize (“Jinhai 5”), and spring peanut (“Yuhua 25”) were calibrated. These parameters include those controlling phenology (potential heat units), biomass growth (radiation use efficiency, WA), yield formation (harvest index, HI), and canopy development (maximum leaf area index, DMLA; onset of leaf senescence, DLAI). Initial parameter values were taken from the APEX database and then adjusted manually in a trial-and-error process. The most sensitive parameters adjusted were related to leaf area development, plant population, and harvest index (Supplementary Table S1). Soil Parameters: The soil profile was defined to a depth of 2 m, divided into multiple layers. Key soil properties for each layer, including texture (sand, silt, clay %), bulk density, organic carbon content, pH, and hydraulic properties (field capacity, wilting point, saturated hydraulic conductivity), were derived from field measurements at the experimental station³³. Based on laboratory measurements of field capacity and wilting point, the total plant-available water capacity within the top 2 m of the soil profile is estimated to be approximately 707 mm³⁵. Soil parameters affecting the water balance, such as the runoff curve number weighting factor (Parm15) and the CN retention parameter (Parm16), were also calibrated. Management and Weather Data: Daily meteorological data (1988–2023), including precipitation, maximum and minimum temperatures, solar radiation, relative humidity, and wind speed, were obtained from a weather station adjacent to the study site (located 1300 m from the experimental station). Detailed management operations for each crop, including typical sowing dates, tillage practices (rotary tillage with residue incorporation), and fertilizer application rates, were set to reflect optimal local practices to avoid nutrient stress (Supplementary Table S2).

The model was calibrated and validated against multi-year field measurements of leaf area index (LAI), aboveground biomass (ABIOM), crop yield, ET, and soil volumetric water content (VMC) for the 0–2 m profile. Following standard practice, the datasets were split, using one period for calibration and another for independent validation. For example, for winter wheat and summer maize, data from 2016 to 2017 were used for calibration, and data from 2018 were used for validation²³. For spring peanut, calibration was performed using 2017 data, and validation with 2018 data³³. A 5-year warm-up period (1988–1993) was run before each simulation to allow soil organic matter and nutrient pools to stabilize.

The calibration process involved manually adjusting the most sensitive crop and soil parameters (as identified above) to minimize the differences between simulated and observed data. Model performance was evaluated both graphically and statistically using the coefficient of determination (R²), normalized root mean square error (NRMSE), and the Willmott index of agreement (d). The validation results from our previous work confirmed the model’s reliability in simulating crop growth, yield, and soil water dynamics under the conditions of the NCP^23,33. For instance, the model accurately reproduced the observed VMC dynamics under full irrigation, although some deviations were noted under severe water stress conditions, a common challenge for crop models. Overall, the validated model provided a robust tool for the scenario analysis conducted in this study.

Simulation scenarios

To evaluate strategies for balancing groundwater use and crop productivity, we designed a factorial simulation experiment crossing nine cropping systems with four irrigation strategies. The simulations were run continuously for a 30-year period (1994–2023) using historical weather data. Eight alternative cropping systems were compared against the conventional winter wheat–summer maize (WM) double crop system (Fig. 7). These alternatives included systems with reduced cropping intensity (three harvests in 2 years), such as spring peanut→WM and summer maize→WM, and single-cropping systems (one harvest per year), such as continuous spring maize or single summer maize. Under strict groundwater management policies, growers are already beginning to adopt single-crop-per-year systems (often one fallow and one rainfed season) and alternative “three harvests in 2 years” rotations. The cropping systems simulated in this study are therefore highly relevant for guiding both on-farm practices and regional policymaking. Four stage-specific irrigation strategies were simulated for each cropping system: full irrigation (FI), critical irrigation (CI), minimum irrigation (MI), and rainfed (RF) (Supplementary Table S3). The timing and amount of irrigation were tailored to the specific water requirements of each crop and growth stage. For example, under MI, winter wheat received only a pre-sowing irrigation. Under CI, an additional irrigation was applied at a critical growth stage (e.g., jointing for wheat). FI included a further irrigation at another key stage (e.g., anthesis for wheat). This design allowed us to explore the trade-offs between water input and crop output systematically.

Wheat equivalent yield and water productivity

To address the challenge of yield comparison among different crop species within alternative cropping systems, individual crop outputs were standardized to wheat equivalent yield (EY) using nutritional equivalency coefficients derived from 20 key nutrient indicators, including protein, fat, carbohydrates, minerals, and vitamins³⁶. The annual EY of each cropping system was calculated by Eq. (1):

where n is the number of crops in each cropping system, GYi and fi are the grain yield and coefficient of nutrition equivalent of each crop, and CL is the cycle length of the cropping system.

The annual water productivity of each cropping system was calculated by Eq. (2):

where EY is the annual wheat equivalent yield of each cropping system, and ETi is the evapotranspiration of each cropping season.

Groundwater level change

Due to the absence of surface water resources in the study region, irrigation relies exclusively on groundwater extraction. Thus, we can establish a link between water consumptive use and groundwater replenishment through modeling exercises. Considering the minimal spatial variation in soil characteristics, topography, and geomorphological features across the study area, the entire region was treated as a homogeneous irrigated agricultural landscape for modeling purposes. Based on the simulated water balances, the net water use (NWU) is determined by the irrigation amount and deep percolation:

where I is irrigation (mm/year) and D is deep percolation (mm/year). In the APEX model, water fluxes exiting the base of the 2-m soil profile are classified as deep percolation, which are presumed to contribute to groundwater recharge.

The variation in groundwater table depth (ΔGWT, m yr⁻¹) was estimated using the following equation:

where 0.001 is a unit conversion factor (mm to m), and μ represents the specific yield of the aquifer, set to 0.17 in accordance with Liang et al.¹⁴.

Given that the groundwater table lies at an approximate depth of 17 m, a significant temporal lag exists between root zone percolation and its eventual contribution to aquifer recharge. Yang et al.³⁷ identified a 1-year delay in groundwater response to precipitation, with the influence distributed over 2 years in the piedmont zone of the NCP³⁷. This delayed recharge hypothesis was evaluated in the alluvial plain setting of our study area and found to be consistent with observed dynamics. As previously noted, the prevalent WM rotation is the primary driver of groundwater extraction and consequent declines in water table levels. This modeling approach was validated by comparing the simulated long-term trend of groundwater decline under the conventional WM system with observed groundwater level data from the region (Fig. 8), which showed a close alignment. This confirmed the model’s suitability for assessing the relative impacts of different cropping systems on groundwater dynamics.

Statistics

Differences among alternative cropping systems were determined using a one-way analysis of variance (ANOVA). The equivalent yield, water stress days, ET, water productivity, and net water use were averaged by rotation cycle (2 years) and compared at the cropping system level under each irrigation strategy. A box plot was used to provide the variability and range of these measured indicators. Pairwise comparisons of treatment means were conducted using Fisher’s least significant difference (LSD) test, with statistical significance determined at the 0.05 probability level (α = 0.05). All computations and analyses were carried out using R version 4.0.5³⁸.