The NCP is located in the eastern coastal region of China. It ranges from the Yanshan Mountains in the north to the Yellow River in the south, and from the Taihang Mountains in the west to the Bohai Sea in the east (Cao et al., 2013; Pei et al., 2015). It includes Hebei Plain, the main plain of the cities of Beijing and Tianjin, and the northern plain of Shandong and Henan Provinces. The total area of the NCP is approximately 140,000 km², and the altitude of most of the NCP is less than 50 m above sea level. In this study, NCP is divided into four parts: Northern Hebei Plain (N-HBP), Southern Hebei Plain (S-HBP), Northwestern Shandong Plain (NW-SDP) and Northern Henan Plain (N-HNP) (Figure 1). The study area belongs to a typical temperate continental monsoon climate and 70% of precipitation is concentrated from late June to September. The average annual precipitation is 500-600 mm, and the annual spatial distribution of precipitation in the NCP has a decreasing trend from the southeast to the northwest (Yang et al., 2010).

The NCP accounts for only 1.5% of China’s total land area but supports 10% of China’s total population (Cao et al., 2014). The proliferation of mechanized wells since the 1960s has intensified agricultural production, mainly in the form of winter wheat and summer maize rotations (Shu et al., 2011). The growing season of winter wheat is from early October to mid-June of the following year. During the growing season of winter wheat, annual precipitation is usually less than 200 mm, but the ET of winter wheat is more than 400 mm (Li et al., 2012). Thus, irrigation is necessary to ensure high yields of winter wheat, which mainly relies on groundwater resources. Additionally, with rapid population growth, industrialization and urbanization, water shortages and groundwater over-exploitation have led to a great constraint on sustainable development in the NCP (Li et al., 2008).

The data used in this study mainly included meteorological data, remote sensing data, statistical data, and field observation data. (1) The meteorological data for 2001-2016 was obtained from the China Meteorological Data Sharing Service System (http://data.cma.cn) of 38 national weather stations located within and around the NCP (Figure 1). The meteorological data were mainly used to calculate the reference crop ET. (2) The remote sensing data included the Normalized Difference Vegetation Index (NDVI) of MOD13Q1 for 2001-2016 and yearly land cover data of MCD12Q1 for 2001-2013. The spatial resolution of MOD13Q1 was 250 m, and the temporal resolution was 16 days. The spatial resolution of MCD12Q1 was 500 m. (3) The growth stages of winter wheat in the study area were collected from the agricultural meteorological station (http://data.cma.cn). (4) Statistical data of the planting area of winter wheat for 2001-2015 were derived from the Statistical Yearbook of Hebei Province, Shandong Province, Henan Province, Beijing City, and Tianjin City. (5) The field-observed ET from four sites were as follows: Luancheng station (2007-2016) (Shen et al., 2013b; Zhang et al., 2018), Weishan station (2005-2013) (Lei and Yang, 2010; Lei et al., 2018), Yucheng station (2003-2005) (Yu and Sun, 2006; Yu et al., 2006a, 2006b; Yu et al., 2008; Yu et al., 2013), and Daxing station (2008-2010) (Jia et al., 2012; Liu et al., 2013; Liu and Xu, 2013). Figure 1 shows the location of each experimental station. (6) Groundwater level monitoring data from 1996 through 2008 were obtained from the China Institute of Geo-Environmental Monitoring (CIGEM), and 2010 data were obtained from the published yearbook on groundwater (Gao and Yin, 2011).

To calculate the volume of winter wheat water consumption, net irrigation water consumption and analyze the impacts of the change in winter wheat planting area on groundwater depletion, the planting area of winter wheat was retrieved based on remotely sensed data. In this study, the planting ratio of winter wheat in one pixel was estimated. First, the spatial resolution of the land cover product (MCD12Q1) was resampled from 500 m to 250 m using the nearest neighbor method. Then, the croplands of the NCP from 2001 to 2013 were extracted based on the land cover classification system of the International Geosphere-Biosphere Programme (IGBP). Due to the lack of MCD12Q1 data after 2013, cropland maps during 2014-2016 were the same as that in 2013. Winter wheat is sown in early October and harvested in mid-June of the following year. To acquire the spatial distribution of winter wheat, the maximum NDVI images of the winter wheat growing season were selected. Based on cropland maps, the planting ratio of winter wheat (fc_wheat) was determined according to Gutman and Ignatov (1998) using the maximum NDVI images. The calculation formula was expressed as:

$f{{c}_{wheat}}=(NDVI-NDV{{I}_{soil}})/(NDV{{I}_{wheat}}-NDV{{I}_{soil}})$ (1)

where fc_wheat is the planting ratio of winter wheat in a pixel; NDVI is the NDVI value in a pixel; NDVI_wheat refers to the pixel that is completely covered by winter wheat. NDVI_soil refers to the pixel that is completely covered by soil. In this study, NDVI_wheat is defined as the average NDVI value of the pixels where the cumulative frequency is more than 90%. NDVI_soil is defined as the NDVI value when the cumulative frequency is up to 10%.

In addition, to clarify the characteristic of the change trend of winter wheat planting area from 2001 to 2016, the spatial heterogeneity of the temporal variation in the planting area of winter wheat was analyzed by simulating the trend variability using a linear regression for each pixel. The least squares regression coefficient was calculated as:

$Slope=\frac{\Sigma _{i=1}^{n}{{x}_{i}}{{t}_{i}}-\frac{1}{n(\Sigma _{i=1}^{n}{{x}_{i}})(\Sigma _{i=1}^{n}{{t}_{i}})}}{\Sigma _{i=1}^{n}t_{i}^{2}-\frac{1}{n}{{(\Sigma _{i=1}^{n}{{t}_{i}})}^{2}}}$ (2)

where Slope refers to the change slope of the planting ratio of winter wheat in a pixel; i=1, 2, 3…, n represents the number of years, x_i represents the planting ratio of winter wheat of the ith year, and t_i is the time. The positive slope represents an increasing trend with time, while a negative slope indicates a decreasing trend. The correlation coefficient between the planting ratio and time was used to detect whether the change trend of the planting ratio of winter wheat is significant or not. The equation of the correlation coefficient (r) was expressed as:

$r=\frac{n(\Sigma _{i=1}^{n}{{x}_{i}}{{t}_{i}})-(\Sigma _{i=1}^{n}{{x}_{i}})(\Sigma _{i=1}^{n}{{t}_{i}})}{\sqrt{[n\Sigma _{i=1}^{n}x_{i}^{2}-{{(\Sigma _{i=1}^{n}{{x}_{i}})}^{2}}][\Sigma _{i=1}^{n}t_{i}^{2}-{{({{t}_{i}})}^{2}}]}}$ (3)

where r is the correlation coefficient, i=1, 2, 3…, n is the number of years, x_i is the planting ratio of winter wheat of the ith year, and t_i is the time. When the absolute value of the correlation coefficient is more than r_0.01, the planting ratio has a significant change trend. In contrast, the change trend of the planting ratio is not significant. According to the value of Slope and r, the change trend of the winter wheat planting ratio was divided into four categories: significant reduction (Slope < 0 &|r|≥r_0.01), nonsignificant reduction (Slope<0 & |r|<r_0.01), significant increase (Slope > 0 & |r|≥r_0.01), and nonsignificant increase (Slope > 0 & |r|< _0.01).

Crop ET is affected by weather conditions, crop type, soil moisture conditions and so on. In this study, a model combining the FAO Penman-Monteith equation with the crop coefficient, and remote sensing data was developed to estimate actual crop ET. First, reference crop ET (ET₀) was calculated using the FAO Penman-Monteith equation. Crop ET (ET_c) under standard conditions was calculated by multiplying ET₀ by the crop coefficient (K_c). The NCP is a typical water-stressed area, so the soil water condition is a crucial factor impacting crop ET. The soil moisture condition was taken into consideration when estimating actual crop ET in this study. Actual crop ET was calculated by using the following equation:

$E{{T}_{a}}=E{{T}_{0}}\times {{K}_{c}}\times {{K}_{s}}$ (4)

where ET_a is the actual crop ET; ET₀ is the reference crop ET; K_c is the crop coefficient; and K_s is the soil moisture correlation coefficient.

2.4.1 Reference crop evapotranspiration

Reference crop ET (ET₀) was calculated using the modified Penman-Monteith equation recommended by the FAO. The Penman-Monteith method was developed by defining the reference crop as a hypothetical crop with an assumed height of 0.12 m, a surface resistance of 70 s/m and an albedo of 0.23. The reference surface resembles an extension surface of the green grass of uniform height, actively growing, completely shading the ground and adequately watered (Allen et al., 1998). Reference crop ET is only related to meteorological factors (Shahid, 2011), hence, it can be used in different regions. The FAO Penman-Monteith equation was as follows:

$E{{T}_{0}}=\frac{0.408\Delta ({{R}_{n}}-G)+\gamma \frac{900}{T+273}{{\mu }_{2}}({{e}_{s}}-{{e}_{a}})}{\Delta +\gamma (1+0.34{{\mu }_{2}})}$ (5)

where ET₀ is the reference ET (mm/d); R_n is the net radiation at the crop surface (MJ/m²/d); G is the soil heat flux density (MJ/m²/d); T is the mean daily air temperature at 2 m in height (℃); u₂ is the wind speed at 2 m in height (m/s); e_s is the saturation vapor pressure (kPa); e_a is the actual vapor pressure (kPa); e_s - e_a is the saturation vapor pressure deficit (kPa); is the slope of the vapor pressure curve (kPa/℃); and γ is the psychometric constant (kPa/℃).

2.4.2 Crop coefficient

The crop coefficient (K_c) is defined as the ratio of crop ET under standard conditions to reference crop ET (Shen et al., 2013a) and it is one of the important parameters in estimating actual crop ET. K_c is typically taken from literature values, and it is affected by crop variety and the growth stage (Allen et al., 1998, 2005). The K_c value of winter wheat varies during the six growth stages: the initial stage, the freezing and thawing stage, the overwintering stage, the rapid development stage, the middle stage, and the mature stage. The date of different growth stages was obtained from agricultural meteorological stations. The K_c of winter wheat was determined according to Duan et al. (2004). The K_c values of the initial and overwintering stage were 0.6 and 0.4, respectively, while K_c value of the freezing and thawing stage varied from 0.6 to 0.4. The K_c values of the other growth stages at the typical stations are shown in Table 1.

2.4.3 Soil moisture correlation coefficient

The growth of crops mainly needs sunlight, temperature, water and so on. In the NCP, water is a major limiting factor for crop growth. Owing to the difficulty of obtaining soil moisture data at large spatial and temporal scales, remote sensing data are commonly used. NDVI is a composite indicator of vegetation activity, and it responds to several variables such as canopy, photosynthesis, biomass, and surface moisture and so on (Adegoke et al., 2002). Some studies have linked the spatial and temporal patterns of NDVI with climate and plant ET (Cihlar et al., 1991). Furthermore, NDVI has been associated closely with eco-climatological variables, such as potential ET and soil hydrologic properties (Hayes and Decker, 1996; Yang et al., 1997). In this study, NDVI was used to reflect the soil moisture condition. We assumed that there was a positive linear correlation between the crop growth condition and soil moisture condition at a certain soil moisture range. Since NDVI can reflect the crop growth condition to a certain extent, there was also a positive linear correlation between NDVI and the soil moisture correlation coefficient.

Generally, winter wheat is sown in early October and regenerated in mid-to-late March of the following year. Its canopy coverage is up to the maximum value during the period from April to May, and it is finally harvested in early June. To acquire the soil moisture condition of winter wheat, the maximum NDVI images of the winter wheat growing season were obtained. It was assumed that at least 10% of winter wheat fields in the NCP had soil moisture condition that was very good and there was no water stress (K_s =1). Then, the NDVI value of the pixel when the cumulative frequency was up to 90% was selected as the reference value (K_s = 1). The soil moisture correlation coefficient of other pixels can be calculated according to the ratio between NDVI and the reference value. The equation for estimating K_s was expressed as:

${{K}_{s}}=\left\{ \begin{align} & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 1,\ \ \ \ \ \ \ \ NDVI\ge NDV{{I}_{90}} \\ & NDVI/NDV{{I}_{90}},\ \ \ \ \ \ \ NDVI<NDV{{I}_{90}} \\\end{align} \right.$ (6)

where K_s is the soil moisture correlation coefficient; NDVI is the NDVI value of any pixel; NDVI₉₀ is the NDVI value of the pixel when the cumulative frequency is up to 90%.

The concept of soil water balance provides a framework for studying the hydrological behavior. The equation of soil water balance was as follows:

$ET=P+I-\Delta SWC-R-D$ (7)

where ET is the actual crop ET (mm), P is the precipitation (mm), I is the irrigation (mm), ΔSWC is the soil water change (mm), R is the surface runoff (mm) and D is the deep percolation (mm). It is generally considered that the surface water systems are limited in regions under severe water stress (Moiwo et al., 2011). In fact, since the 1960s, surface runoff in the study area has dropped to a virtually negligible level (Fan et al., 2010; Wu et al., 2011). Under these hydrological conditions and the flat topography in the NCP, surface runoff can be ignored (Cao et al., 2014). Soil water change can be measured through experiments at the field scale, but it is difficult to acquire soil water change at large spatial and temporal scales. In this paper, the average soil water change for multiple years in Luancheng station was used (Zhang et al., 2018). Moreover, D can become the reused groundwater and it is difficult to obtain an accurate estimation (Yuan et al., 2015). Therefore, the equation of the soil water balance can be simplified as follows, and the net irrigation water consumption (I_net) can be calculated according to equation 8:

$I-D={{I}_{net}}=ET-P-\Delta SWC$ (8)

To validate the estimated area of winter wheat, the statistical data were mainly used for comparison. In this study, the planting area of winter wheat at the city-level from the Statistical Yearbook was used. Figure 2 illustrated the correlation between the statistical and estimated planting area of winter wheat. The relative bias was 2.4%, and the root-mean-square error (RMSE) was 7.7×10⁴ ha. A regression analysis between estimated and statistical area was performed, and the Pearson correlation coefficient r was 0.88 (N = 314, p < 0.001). The estimated result had a good match with the statistical planting area. Consequently, the method to retrieve the planting area of winter wheat based on remote sensing data is acceptable.

Figure 3a illustrated that average spatial distribution of winter wheat during 2001-2016. The result showed that the planting ratio of winter wheat in the south was higher than that in the north. Winter wheat was mainly distributed in the piedmont region of the Taihang Mountains and the irrigation district of the Yellow River Basin in Henan and Shandong provinces. In these two regions, the planting ratio of winter wheat was more than 80% because of good irrigation conditions. The planting ratio of winter wheat in Beijing and Tianjin City was less than 40% owing to more construction land, less arable land and the small proportion of primary industry. In coastal areas, planting ratio of winter wheat was less than 20% by reason of the saline soil. The planting ratio of winter wheat in the central plains was also low, mainly due to the sandy soil and high salinity of shallow groundwater. For these two reasons, the central plains are better suited for the cultivation of fruits (e.g., pear trees in Zhaoxian County of Shijiazhuang City, peach trees in Shenzhou City and jujube trees in Cangzhou city), peanuts (e.g., in Daming County of Handan City) and cotton (e.g., in Nangong and Qinghe counties of Xingtai City). Thus, water resources and soil conditions determined the spatial distribution of winter wheat.

To analyze the spatial variation of the winter wheat planting area in the NCP, the change trend from 2001 to 2016 was calculated. Figure 3b showed that the planting area of winter wheat in most parts of the Northern Hebei Plain had a decreasing trend, and especially that in the cities of Langfang, Tianjin, Baoding, and Cangzhou had a significant decreasing trend. According to the field survey in August 2014, it was found that in Langfang and Tianjin, rural township enterprises were rather developed so that many farmers gave up the planting of winter wheat and worked there for a living. Additionally, many wells in these regions had no water, especially in Langfang. The irrigation water fee was so high that the economic benefit of planting winter wheat was low. While in Baoding, some counties have gradually developed nursery (e.g., Wangdu and Boye counties) and herbs (e.g., Anguo County) industries in the last decade. In the east of Shijiazhuang, north of Hengshui and in most parts of the Shandong Plain, especially in Dezhou, the planting ratio of winter wheat had a remarkable increasing trend. This was mostly because farmers gradually shifted from planting single cotton to the double cropping of winter wheat and summer maize in recent years. Dezhou is the main cotton production region in China and has a cotton cultivation history of many years. However, the planting and management of cotton needs more labor than that of winter wheat and summer maize because it is difficult to plant and pick up cotton mechanically. Then, a cotton planter can only be kept in cotton fields all the year round. In addition, the net economic benefit of planting winter wheat and summer maize is higher than single cotton. Figure 4 illustrates the changes of the winter wheat planting area in the four parts of the NCP during 2001-2016. From the regional scale, the winter wheat planting area in the Northern Hebei Plain had a decreasing trend and declined from 84.7×10⁴ ha during 2001-2005 to 46.5×10⁴ ha during 2012-2016, a decrease of 45.1%. While the winter wheat planting area in the Southern Hebei Plain also had a decreasing trend and declined from 151.7×10⁴ ha during 2001-2005 to 139.3×10⁴ ha during 2012-2016, a decrease of 8.2%. While the winter wheat planting area in the Northwestern Shandong Plain had a slightly increasing trend and increased from 118.3×10⁴ ha during 2001-2005 to 134.3×10⁴ ha during 2012-2016. The winter wheat planting area in the Northern Henan Plain had a slightly decreasing trend and decreased by 8.5%.

Winter wheat ET from 2001 to 2016 in the NCP was estimated based on the proposed model. The estimated ET was compared with field-observed ET of the winter wheat growing season at Luancheng, Weishan, Yucheng, and Daxing stations (Figure 5). The relative error of each sample between estimated and field-observed ET varied from 0.2%-12%. The mean absolute deviation between estimated and field-observed ET was 21.5 mm, the relative bias was -2.2% and the RMSE was 25.5 mm. Therefore, the proposed ET estimation model performed well in the study area. Thus, the method for estimating crop water consumption based on the Penman-Monteith method and remote sensing data is acceptable.

Combined with the planting ratio of winter wheat, the spatial distribution of the average ET contribution of winter wheat from 2001 to 2016 in the NCP is shown in Figure 6a. Winter wheat ET was higher in the piedmont regions of the Taihang Mountains and the irrigation district of the Yellow River Basin in Henan and Shandong provinces, and relatively lower in the northern and central parts of the Hebei Plain. Winter wheat ET was more than 350 mm in the piedmont regions of the Taihang Mountains because of good soil texture, climatic conditions, and excellent irrigation conditions, while winter wheat ET in the northern and central Hebei Plain was less than 350 mm because spring maize (e.g., in Langfang), jujube trees (e.g., in Cangzhou) and cotton (e.g., in Xingtai) were the main crops. In addition, winter wheat ET was less than 300 mm in coastal areas because of the saline-alkali land.

Based on the planting area of winter wheat, the amount of water consumption of winter wheat was calculated in the NCP. The average water consumption of winter wheat during 2001-2016 was 20.5×10⁸ m³ in the Northern Hebei Plain, 49.7×10⁸ m³ in the Southern Hebei Plain, 48.5×10⁸ m³ in the Northwest Shandong Plain, and 46.5×10⁸ m³ in the Northern Henan Plain, respectively. Moreover, the water consumption of winter wheat in the Northern Hebei Plain decreased from 29.7×10⁸ m³ during 2001-2005 to 12.2×10⁸ m³ during 2012-2016, a decrease of 56.0%. The water consumption of winter wheat in the Southern Hebei Plain also had a downward trend and decreased from 55.1×10⁸ m³ during 2001-2005 to 45.9×10⁸ m³ during 2012-2016. The reduction in the winter wheat planting area was one of the reasons for the decrease in water consumption of winter wheat in the Northern Hebei Plain and Southern Hebei Plain. Water consumption of winter wheat in the Northwest Shandong Plain had a slight increase, and that of the Northern Henan Plain had a slight reduction.

The above characteristics of winter wheat ET were analyzed (Figure 6a). Based on precipitation and equation 8, the net irrigation water consumption (Figure 6c) in the winter wheat growing season during 2001-2016 was calculated. Figure 6b showed the spatial distribution of average precipitation in the winter wheat growing season from 2001 to 2016. Precipitation in the winter wheat growing season in the NCP increased from 132 mm in the northwest to 215 mm in the southeast. In the piedmont regions of the Taihang Mountains, precipitation was less than 180 mm and these regions need irrigation of approximately 150 mm. In most parts of the Northern Hebei Plain and Southern Hebei Plain, precipitation in the winter wheat growing season was approximately less than 170 mm, and these regions need the irrigation of less than 100 mm. Compared with the Hebei Plain, precipitation in the Northwest Shandong Plain and Northern Henan Plain was relatively higher and ranged from 170 mm to 215 mm. In the irrigation district of the Yellow River Basin in Henan and Shandong provinces, net irrigation water consumption was more than 150 mm.

Based on the planting area of winter wheat, net irrigation water consumption in the four parts from 2001 to 2016 was estimated (Figure 7). The average net irrigation water consumption from 2001 to 2016 was up to 8.0×10⁸ m³ in the Northern Henan Plain, 19.9×10⁸ m³ in the Southern Henan Plain, 17.8×10⁸ m³ in Northwestern Shandong Plain and 17.8×10⁸ m³ in the Northern Henan Plain, respectively. The net irrigation water consumption in the Northern Hebei Plain had a decreasing trend and decreased from 11.0×10⁸ m³ during 2001-2005 to 5.5×10⁸ m³ during 2012- 2016. The net irrigation water consumption in the Southern Hebei Plain decreased from 20.2×10⁸ m³ during 2001-2005 to 19.4×10⁸ m³ during 2012-2016. The net irrigation water consumption in the Northwestern Shandong Plain changed from 15.0×10⁸ m³ during 2001-2005 to 18.3×10⁸ m³ during 2012-2016 and that of the Northern Henan Plain changed from 17.8×10⁸ m³ during 2001-2005 to 13.4×10⁸ m³ during 2012-2016.

During the winter wheat growing season, approximately 4-5 times of irrigations of 60-80 mm from groundwater are used to offset the water deficit (Sun et al., 2010). The drops in the groundwater table in the NCP mainly occurred during the winter wheat growing season due to the extraction of groundwater for irrigation (Luo et al., 2018). Therefore, the relationship between net irrigation water consumption and the change in the groundwater table was performed and groundwater table data were collected in the four parts of the NCP. The distribution of groundwater table monitoring sites is shown in Figure 6c. Figure 8 illustrates the relationship between the average groundwater table and accumulated net irrigation water consumption in the Northern Hebei Plain and Southern Hebei Plain of the NCP. The total net irrigation water consumption of winter wheat during 2001-2016 was up to 12.8×10⁹ m³ in the Northern Hebei Plain and 31.9×10⁹ m³ in the Southern Hebei Plain. The result (Figures 8a and 8c) showed that there was a significant negative correlation between the groundwater table and accumulated net irrigation water consumption in the Northern Hebei Plain and Southern Hebei Plain. In these two regions, with the increase of the accumulated net irrigation water consumption, the groundwater table dropped. This result implied that water consumption for winter wheat production had a significant impact on groundwater resources. During 2001-2010, the average groundwater table in the Northern Hebei Plain declined from 14.1 m to 11.6 m, and the total net irrigation water consumption was up to 9.3×10⁹ m³, while the average groundwater table in the Southern Hebei Plain declined from 15.4 m in 2001 to 11.7 m in 2010, and the total net irrigation water consumption accounted for 19.2×10⁹ m³ during 2001-2010. Before 2001, the rate of decline of the groundwater table in the Northern Hebei Plain was 0.72 m/a, while after 2001, that of the Northern Hebei Plain decreased to 0.25 m/a. Likewise, the rate of decline of the groundwater table in the Southern Hebei Plain decreased from 0.45 m/a to 0.38 m/a. The net irrigation water consumption in the Northern Hebei Plain had a significant decreasing trend and decreased from 16.8×10⁸ m³ in 2001 to 5.0×10⁸ m³ in 2010, while the net irrigation water consumption of the Southern Hebei Plain also had a decreasing trend and decreased from 27.2×10⁸ m³ in 2001 to15.7×10⁸ in 2010.

According to the results mentioned above, the planting area of winter wheat had a decreasing trend from 2001 to 2016 in the Northern Hebei Plain and Southern Hebei Plain, and at the same time, the rate of decline of the groundwater table in these two regions before and after 2001 both decreased to different extents (Figure 8). Thus, in the Hebei Plain, the decrease of planting area of winter wheat can reduce water withdrawals and has a positive effect on the protection of groundwater resources. It was found that the main factor leading to the decline of the groundwater table was the expanding area of winter wheat (Hu et al., 2002). Moreover, Kendy et al. (2004) reported that the only means to reduce net water consumption for agriculture is to reduce crop ET, and this can be accomplished by reducing the crop area. Shen et al. (2013b) had the same suggestion, that shifting the winter wheat-summer maize double cropping to single maize could be expected to reduce water pumping for irrigation significantly. Wang et al. (2014) found that the water resources conserved owing to the reduction in sown area of winter wheat in the Hebei Plain during 1998-2010 were approximately 15.96×10⁸ m³/a. Pei et al. (2015) also suggested that switching crop rotations from double-cropping wheat and corn to single cropping corn with an extended growing season in the NCP might achieve similar yields and would reduce irrigation demand. In addition, the Chinese government has attempted to propose some policies to solve the problem of exploitation of groundwater in Hebei Province since 2014. Accordingly, in the large groundwater depression zones (including Cangzhou, Hengshui, Xingtai and Handan), the ‘one season fallow, one season rain-fed’ policy was proposed. This policy was put forward due to the high irrigation water consumption of winter wheat, and it suggested that winter wheat should not be planted. In addition, spring maize, potatoes, and beans were proposed for cultivation. According to this policy, each hectare of land can save 2700-3000 m³ of water. At the same time, to guarantee the economic benefit for farmers in these experimental regions, farmers can receive a subsidy of 7500 yuan per hectare per year. Through these efforts, the reduction in amount of groundwater extraction was up to 2.2×10⁸ m³ in 2016.

Compared with changes of the groundwater table in the Northern Hebei Plain and Southern Hebei Plain, there was no significant correlation between average groundwater table and accumulated net irrigation water consumption in the Northwestern Shandong Plain and Northern Henan Plain (Figures 9a and 9c). In these two regions, with the increase of accumulated net irrigation water consumption of winter wheat during 2001-2010, the groundwater table had no significant declining trend. The Yellow River is the main irrigation water resource in these two regions; thus, the groundwater table had no continuous decreasing trend. Moreover, the total net irrigation water consumption during 2001-2016 in the Northwestern Shandong Plain and in the Northern Henan Plain was both up to 28.5×10⁹ m³. According to the statistical data for 2003-2010 from the Yellow River Conservancy Commission, the average extracted agricultural water of the Northern Henan Plain from Yellow River was approximately 12.5×10⁸ m³, and this amount of water accounted for 96% of the total extracted water from the Yellow River. In the Northwestern Shandong Plain, the average extracted agricultural water from the Yellow River was about 27.7×10⁸ m³, and this amount of water accounted for 94% of the total extracted water from the Yellow River. Consequently, water resources from the Yellow River can supply irrigation for agriculture and sometimes alleviate the decline of the groundwater table in the irrigation district of the Yellow River Basin in Henan and Shandong provinces. Water withdrawal from the lower reaches of the Yellow River started in the 1970s. Owing to the lack of rainfall, little runoff and human activities (Chen et al., 2004), the lower reaches of the Yellow River have dried up since 1972 (Wu et al., 1998). After 2000, there was no cutoff in the lower reaches of the Yellow River. However, in drought years and peak periods of agricultural irrigation, irrigation from the lower reaches of the Yellow River is not sufficient, and agricultural irrigation mainly relies on groundwater resources. From the long-term development of agriculture, it is also crucial to decrease the planting area of winter wheat in the irrigation district of the Yellow River Basin in Henan and Shandong provinces. In addition, since the Yellow River is the transit water and it cannot meet the needs for agricultural irrigation water in terms of time and amount, some plain reservoirs should be constructed. These reservoirs can retain water in the wet season and provide water in the drought season, then it can greatly alleviate the water stress from agriculture (Bian et al., 2009). At the same time, a scientific and reasonable water price formation mechanism should be established so that farmers’ awareness of water-saving irrigation can be improved (Pan et al., 2010).