Determining Threshold Values for a Crop Water Stress Index-Based Center Pivot Irrigation with Optimum Grain Yield

: The temperature-based crop water stress index (CWSI) can accurately reﬂect the extent of crop water deﬁcit. As an ideal carrier of onboard thermometers to monitor canopy temperature (T c ), center pivot irrigation systems (CPIS) have been widely used in precision irrigation. However, the determination of reliable CWSI thresholds for initiating the CPIS is still a challenge for a winter wheat–summer maize cropping system in the North China Plain (NCP). To address this problem, ﬁeld experiments were carried out to investigate the effects of CWSI thresholds on grain yield (GY) and water use efﬁciency (WUE) of winter wheat and summer maize in the NCP. The results show that positive linear functions were ﬁtted to the relationships between CWSI and canopy minus air temperature (T c − T a ) ( r 2 > 0.695), and between crop evapotranspiration (ET c ) and T c ( r 2 > 0.548) for both crops. To make analysis comparable, GY and WUE data were normalized to a range of 0.0 to 1.0, corresponding the range of CWSI. With the increase in CWSI, a positive linear relationship was observed for WUE ( r 2 = 0.873), while a signiﬁcant inverse relationship was found for the GY ( r 2 = 0.915) of winter wheat. Quadratic functions were ﬁtted for both the GY ( r 2 = 0.856) and WUE ( r 2 = 0.629) of summer maize. By solving the cross values of the two GY and WUE functions for each crop, CWSI thresholds were proposed as being 0.252 for winter wheat, and 0.229 for summer maize, corresponding to a T c − T a threshold value of 0.925 and 0.498 ◦ C, respectively. We conclude that farmers can achieve the dual goals of high GY and high WUE using the optimal thresholds proposed for a winter wheat–summer maize cropping system in the NCP.


Introduction
Irrigation plays a fundamental role in promoting agricultural production. With the development of irrigated farming, agriculture is the largest consumer of anthropic water on the planet, in which 70% of the global fresh water is consumed by the agricultural sector [1]. Precision irrigation is regarded as a promising technique to save water resources and promote agricultural sustainability [2]. In the North China Plain (NCP), sprinkler and drip irrigation methods are commonly used in precision irrigation. However, these techniques do not adapt well to a winter wheat (Triticum aestivum L.)-summer maize (Zea mays L.) cropping system due to the differences in row spacing, plant height, and planting densities between the crops [3]. A center pivot irrigation system (CPIS) is considered adaptable in the mechanized operation for both wheat and maize due to its adaptability to different crops [4]. In addition, due to the over-depletion of aquifers, the NCP has become the severest groundwater depression zone in the world [5]. This especially expedites the development of water-saving techniques, making precision irrigation an inevitable trend for future agriculture in this area.
Precise diagnosis of water stress is the basis of precision irrigation [6]. Up to now, most efforts have focused on soil water balance, or water potential-based approaches [7]. 1). The cropping system is a winter wheat-summer maize crop rotation. The study area has a continent temperate monsoon climate. The annual mean temperature is 14.7 °C, the annual mean precipitation is 698 mm and the annual sunshine is 2280 h. Meteorological data were obtained from a weather observatory near the experimental station, with monitored parameters including air temperature (Ta, °C), wind speed (U2, m s −1 ), solar radiation (Ra, MJ m 2 d −1 ), relative humidity (RH, %), and precipitation (mm). The data were measured at the height of 2.0 m. During the growing season of 2019-2020, total precipitation was recorded as 769 mm, of which 26% fell in the winter wheat season, and the remaining 74% fell in the summer maize season. Reference evapotranspiration (ET0) was 583 and 445 mm for winter wheat and summer maize, respectively ( Figure 2). The soil is a fluvo-aquic soil [35]. The soil bulk density of the 0-60 cm soil layer was 1.35 g cm −3 , total nitrogen content was 1.28 g kg −1 , the total phosphorus content was 1.71 g kg −1 , and organic matter content was 20.3 g kg −1 , respectively. Detailed soil physical properties prior to the experiment were presented by   [36].

Experimental Design
Winter wheat seeds (c.v. Xinmai 26) were sown on 15 October, 2019, and were harvested on 29 May, 2020. The row spacing of wheat was 12 cm, and the basic number of seedlings was 2.25 × 10 6 plants ha −1 . A popular maize hybrid (c.v. Denghai 605) was selected in the experiment. Maize seeds were sown on 5 June 2020 and harvested on 30 September 2020. The plant spacing of maize was 30 cm, and the row spacing was 50 cm, giving a planting density of 6.75 × 10 4 plants ha −1 . Furrow dikes were installed in each inter-row following crop establishment to reduce surface movement of irrigation and precipitation water. Based on soil tests in the 0-100 cm soil profile, base fertilizers including urea (150 kg ha −1 N), diammonium phosphate (180 kg ha −1 P2O5), and potassium sulfate (55 kg ha −1 1). The cropping system is a winter wheat-summer maize crop rotation. The study area has a continent temperate monsoon climate. The annual mean temperature is 14.7 °C, the annual mean precipitation is 698 mm and the annual sunshine is 2280 h. Meteorological data were obtained from a weather observatory near the experimental station, with monitored parameters including air temperature (Ta, °C), wind speed (U2, m s −1 ), solar radiation (Ra, MJ m 2 d −1 ), relative humidity (RH, %), and precipitation (mm). The data were measured at the height of 2.0 m. During the growing season of 2019-2020, total precipitation was recorded as 769 mm, of which 26% fell in the winter wheat season, and the remaining 74% fell in the summer maize season. Reference evapotranspiration (ET0) was 583 and 445 mm for winter wheat and summer maize, respectively ( Figure 2). The soil is a fluvo-aquic soil [35]. The soil bulk density of the 0-60 cm soil layer was 1.35 g cm −3 , total nitrogen content was 1.28 g kg −1 , the total phosphorus content was 1.71 g kg −1 , and organic matter content was 20.3 g kg −1 , respectively. Detailed soil physical properties prior to the experiment were presented by Qin et al. (2019) [36].

Experimental Design
Winter wheat seeds (c.v. Xinmai 26) were sown on 15 October, 2019, and were harvested on 29 May, 2020. The row spacing of wheat was 12 cm, and the basic number of seedlings was 2.25 × 10 6 plants ha −1 . A popular maize hybrid (c.v. Denghai 605) was selected in the experiment. Maize seeds were sown on 5 June 2020 and harvested on 30 September 2020. The plant spacing of maize was 30 cm, and the row spacing was 50 cm, giving a planting density of 6.75 × 10 4 plants ha −1 . Furrow dikes were installed in each inter-row following crop establishment to reduce surface movement of irrigation and precipitation water. Based on soil tests in the 0-100 cm soil profile, base fertilizers including urea (150 kg ha −1 N), diammonium phosphate (180 kg ha −1 P2O5), and potassium sulfate (55 kg ha −1

Experimental Design
Winter wheat seeds (c.v. Xinmai 26) were sown on 15 October, 2019, and were harvested on 29 May, 2020. The row spacing of wheat was 12 cm, and the basic number of seedlings was 2.25 × 10 6 plants ha −1 . A popular maize hybrid (c.v. Denghai 605) was selected in the experiment. Maize seeds were sown on 5 June 2020 and harvested on 30 September 2020. The plant spacing of maize was 30 cm, and the row spacing was 50 cm, giving a planting density of 6.75 × 10 4 plants ha −1 . Furrow dikes were installed in each inter-row following crop establishment to reduce surface movement of irrigation and precipitation water. Based on soil tests in the 0-100 cm soil profile, base fertilizers including urea (150 kg ha −1 N), diammonium phosphate (180 kg ha −1 P 2 O 5 ), and potassium sulfate (55 kg ha −1 K 2 O) were applied prior to sowing. To ensure a non-limited supply of nutrients for crops, additional urea (150 kg ha −1 N) was applied as a top dressing at the jointing and flowering stages for both wheat and maize plants at a ratio of 6:4. All treatment plots received the same amounts of total fertilizer. Weeds and pests were managed according to the local governments' recommendations.
The CPIS consists of a pivot point, truss, overhang, lateral line, sprinkler and other components ( Figure 3). The span (56.5 m) and overhang (3.5 m) are raised and supported on towers above the ground, allowing droppers connected to the lateral line, turning around the pivot point. The irrigated area is 1.13 ha. The flow rate per sprinkler is 0.2 to 0.6 LPS. The rate is controlled by a CPIS control-panel. Water pressure at the pivot point is 3.0 bar (43.5 psi). The spraying system optimizes the combination of sprinklers at different distances from the pivot point to ensure an irrigation uniformity of 90%. The height of the system is 3.0 m, and the height of droppers can be adjusted according to plant height. CPIS is a suitable platform that can transport infrared thermometers (IRTs) over cropped areas [37,38], on which IRTs are mounted to monitor spatial T c data. When measuring T c , rotational speeds can be managed to distribute the coverage of IRTs to all angular positions of the studied area [39]. A precision flow meter (Shanghai Water Meter Manufacturing Co. Ltd., Shanghai, China) was installed at the pump outlet to measure the irrigation volume. K2O) were applied prior to sowing. To ensure a non-limited supply of nutrients for crops, additional urea (150 kg ha −1 N) was applied as a top dressing at the jointing and flowering stages for both wheat and maize plants at a ratio of 6:4. All treatment plots received the same amounts of total fertilizer. Weeds and pests were managed according to the local governments' recommendations. The CPIS consists of a pivot point, truss, overhang, lateral line, sprinkler and other components ( Figure 3). The span (56.5 m) and overhang (3.5 m) are raised and supported on towers above the ground, allowing droppers connected to the lateral line, turning around the pivot point. The irrigated area is 1.13 ha. The flow rate per sprinkler is 0.2 to 0.6 LPS. The rate is controlled by a CPIS control-panel. Water pressure at the pivot point is 3.0 bar (43.5 psi). The spraying system optimizes the combination of sprinklers at different distances from the pivot point to ensure an irrigation uniformity of 90%. The height of the system is 3.0 m, and the height of droppers can be adjusted according to plant height. CPIS is a suitable platform that can transport infrared thermometers (IRTs) over cropped areas [37,38], on which IRTs are mounted to monitor spatial Tc data. When measuring Tc, rotational speeds can be managed to distribute the coverage of IRTs to all angular positions of the studied area [39]. A precision flow meter (Shanghai Water Meter Manufacturing Co. Ltd., Shanghai, China) was installed at the pump outlet to measure the irrigation volume. After each crop sowing, all experimental plots were irrigated with 30 mm water to guarantee the uniform and rapid germination of seeds (Table 1). Otherwise, different irrigation levels were carried out during the growing season of wheat, whereas maize crops were mainly rain-fed. Every irrigation system has an Ea lower than 1, and gross irrigation must be larger than the net irrigation. In this study, net irrigation was also measured by rainfall gauges installed in each plot. Therefore, the irrigation amount calculated in the study was considered equal to net irrigation. Incomplete randomized block design was adopted with six replicates. Four irrigation treatments were set as follows: I1, supplying 45% of crop water requirements, I2, supplying 60%, I3, supplying 85% and I4, supplying 100% of crop water requirements, respectively. Three plots with the same treatments were in the same sector area, where: in zone I, the plot size was 353 m 2 ; in zone II, the plot size was 551 m 2 ; and in zone III, the plot size was 509 m 2 ( Figure 4). Each experimental plot was surrounded by a border plot of the same size to avoid edge effects on the experimental plots. The inclusion of border plots effectively separated irrigation treatments. To induce slight drought and promote the penetration of roots of wheat, the irrigation quota was reduced at the over-winter and jointing stages, whereas more water was delivered at the critical flowering and filling stages [40,41]. The amount of full irrigation (I4) was calculated according to Equation (1), and detailed information of irrigation for different treatments was presented in Table 1. Averaged across the whole growing season, mean CWSI thresholds were 0.45, 0.35, 0.28, and 0.20 for I1, I2, I3, and I4 treatments. In Equation (1), field capacity was used as an indicator for the upper limit of water that can be used by plants, which was determined in the laboratory by measuring the amount of water retained at an applied pressure of -33 kPa, whereas wilting point was determined at an applied pressure of 1500 kPa [42]. In this study, the wilting point in the 0-30, 30-60, 60- After each crop sowing, all experimental plots were irrigated with 30 mm water to guarantee the uniform and rapid germination of seeds (Table 1). Otherwise, different irrigation levels were carried out during the growing season of wheat, whereas maize crops were mainly rain-fed. Every irrigation system has an E a lower than 1, and gross irrigation must be larger than the net irrigation. In this study, net irrigation was also measured by rainfall gauges installed in each plot. Therefore, the irrigation amount calculated in the study was considered equal to net irrigation. Incomplete randomized block design was adopted with six replicates. Four irrigation treatments were set as follows: I1, supplying 45% of crop water requirements, I2, supplying 60%, I3, supplying 85% and I4, supplying 100% of crop water requirements, respectively. Three plots with the same treatments were in the same sector area, where: in zone I, the plot size was 353 m 2 ; in zone II, the plot size was 551 m 2 ; and in zone III, the plot size was 509 m 2 ( Figure 4). Each experimental plot was surrounded by a border plot of the same size to avoid edge effects on the experimental plots. The inclusion of border plots effectively separated irrigation treatments. To induce slight drought and promote the penetration of roots of wheat, the irrigation quota was reduced at the over-winter and jointing stages, whereas more water was delivered at the critical flowering and filling stages [40,41]. The amount of full irrigation (I4) was calculated according to Equation (1), and detailed information of irrigation for different treatments was presented in Table 1. Averaged across the whole growing season, mean CWSI thresholds were 0.45, 0.35, 0.28, and 0.20 for I1, I2, I3, and I4 treatments. In Equation (1), field capacity was used as an indicator for the upper limit of water that can be used by plants, which was determined in the laboratory by measuring the amount of water retained at an applied pressure of −33 kPa, whereas wilting point was determined at an applied pressure of 1500 kPa [42]. In this study, the wilting point in the 0-30, 30-60, 60-100 cm soil depths was 11.4%, 12.3%, and 12.8% v/v, and the corresponding field capacity was 29.3%, 30.1%, and 32.8% v/v, respectively. where Q is the amount of full irrigation per single time (mm); FC is the field capacity (cm 3 cm −3 ); SWC is the soil volumetric water content before irrigation (cm 3 cm −3 ); and D i is the planned wet layer depth (cm).
where Q is the amount of full irrigation per single time (mm); FC is the field capacity (cm 3 cm −3 ); SWC is the soil volumetric water content before irrigation (cm 3 cm −3 ); and Di is the planned wet layer depth (cm).

Canopy Temperature
Infrared thermometers (IRTs) (Shi'ao Instrument Co. Ltd., Wuxi, China) with a bandpass of 5.5-14 μm were adopted to measure Tc. The measurement was implemented after the complete formation of wheat leaves on 15 November, 2019, and maize leaves on 20 June, 2020. Measurements were taken from 10:00 to 14:00 on a daily basis, when Tc typically reached diurnal maxima. The moving IRTs were aboard the CPIS. When measurement started, the CPIS rotated around the center pivot with an angular speed of 3° min −1 , and a wheel moving speed of 2.9 m min −1 . A round of data measurement took almost 2 h. Data were then divided and selected every 15 min per plot. Data were recorded every minute by the datalogger, and then were averaged and stored every 15 min. The IRTs had a field of view of 36°, a sensing window of 12.5 mm, and a resolution of 0.1 °C. The IRT Figure 4. Plot layout of field experiment with the center pivot irrigation system (CPIS).

Canopy Temperature
Infrared thermometers (IRTs) (Shi'ao Instrument Co. Ltd., Wuxi, China) with a bandpass of 5.5-14 µm were adopted to measure T c . The measurement was implemented after the complete formation of wheat leaves on 15 November, 2019, and maize leaves on 20 June, 2020. Measurements were taken from 10:00 to 14:00 on a daily basis, when T c typically reached diurnal maxima. The moving IRTs were aboard the CPIS. When measurement started, the CPIS rotated around the center pivot with an angular speed of 3 • min −1 , and a wheel moving speed of 2.9 m min −1 . A round of data measurement took almost 2 h. Data were then divided and selected every 15 min per plot. Data were recorded every minute by the datalogger, and then were averaged and stored every 15 min. The IRTs had a field of view of 36 • , a sensing window of 12.5 mm, and a resolution of 0.1 • C. The IRT array contained 9 thermometers that were positioned at regular intervals above the canopy. During each time period, 1080 measurements were taken, and a total of 270 T c data were averaged to represent each treatment. The IRTs' view was periodically checked and adjusted. Data can be downloaded from the datalogger to a laptop computer via wireless data transmission.

Soil Water Content and Crop Evapotranspiration
Soil water content (SWC, % v/v) was measured at 10 cm increments to a depth of 100 cm using Insentek sensors (Zhejiang Oriental Insentek Technology Co. Ltd., Hangzhou, China). Sensors were installed in the central sector areas, and measured SWC every 10 min. The average over two hours was stored in a data logger and was transmitted to a local server via wireless network. Our previous study indicated that the Insentek sensor was a reliable tool to represent real SWC values in the field with a root mean square error (RMSE) of 0.927% v/v between the Insentek sensor and oven-dry method (Table S1) [36]. Through checking the year-round data, data of Insentek sensors also had good continuity and stability [5]. Crops around the sensors were not missed and uniform, and were representative of the experimental conditions with the similar sizes and vigor.
Crop evapotranspiration (ET c , mm) was calculated using soil water balance equation [25]: where ET c is crop evapotranspiration (mm) during the growing season of wheat and maize, I is the amount of irrigation water applied (mm), P is the seasonal precipitation (mm) measured at the adjacent weather station, C r is capillary rise (mm), R f is runoff (mm), D p is deep percolation (mm), and ∆S is the change in soil water storage (SWS, mm). In Equation (2), C r was considered to be zero, because the groundwater table was 5.0 m below the surface; R f was also assumed to be zero because furrow dikes were used to control runoff; D p was considered to be zero because SWC below 100 cm did not reach FC on the sampling dates.

Crop Water Stress Index
Daily CWSI was calculated as the average of 15 min CWSI values between 10:00 and 14:00 MST. The measurement time fell in the optimal measurement period proposed by the previous literature [32]. Although we cannot guarantee that every measurement was taken on a sunny and windless day, the data were averaged on a large number of measured values (270 data every time), which help to minimize the discrepancy in data and the adverse effect of weather factors. As an index indicating the severity of drought stress on plants, CWSI was adopted and defined as [43]: where (T c − T a )ll is the lower limit of T c − T a , representing the non-water-stressed lower baseline, and (T c − T a )ul is the upper limit of T c − T a , representing the non-transpiring upper baseline. In this study, (T c − T a )ll from 100% supplying treatment was regarded as the lower non-stressed baseline. (T c − T a )ul was estimated at 5 • C as the upper limit for most cereal crops [44,45].

Grain Yield and Water Use Efficiency
At physiological maturity, an area of 20 m 2 from 24 plots was harvested manually and grain mass was determined. Sampling plots were located at the same distance from a center pivot. The total numbers of plants and ears were counted, and the number of ears per plant was determined. Grain per ear was counted for each ear. Grain weight was determined by oven-drying three samples of 1000 kernels at 80 • C for 72 h to a constant weight. The grains were separated, cleaned, and weighed. Grain yield was expressed at 14% moisture content as determined using a PM-8188 portable moisture meter (Kett Electric Lab., Tokyo, Japan).
Water use efficiency (WUE, kg m −3 ) was calculated as the grain yield (kg ha −1 ) produced per unit of ET c (mm).
where GY is the grain yield (kg ha −1 ) and ET c is the crop evapotranspiration (mm) calculated according to Equation (2). The coefficient (0.1) converts kg ha −1 mm −1 to kg m −3 .
Agriculture 2021, 11, 958 7 of 16 2.3.5. Determining Thresholds of CWSI and T c − T a Data of GY and WUE were normalized to standardized values between 0.0 and 1.0 using min-max normalization [46], corresponding to the range of CWSI. The relationships between CWSI and normalized values of GY and between CWSI and normalized values of WUE were determined using Equation (5) for wheat and Equation (6) for maize. To achieve the dual goal of high GY and WUE, CWSI thresholds for initiating irrigation were determined using the cross values of the CWSI vs. GY functions and the CWSI vs. WUE functions: where NV W and NV M are the normalized values of GY or WUE for wheat and maize, respectively; CWSI is the crop water stress index; and a, b, and c are parameters fitted to the equations.

Statistical Analysis
Analysis of variance (ANOVA) was performed to test the differences in GY, WUE, T c , and ET c among treatments. Correlation analysis was conducted with SPSS 19.0 software (SPSS Institute Inc. USA) to determine the relationships between GY and CWSI, as well as the relationship between WUE and CWSI. Means were compared using Fisher's least significant difference tests at p < 0.05. Graphs presented were plotted using Sigmaplot 12.0 (Systat Software, San Jose, CA, USA).

Grain Yield and Water Use Efficiency
Less water supply induced more drought stress on crops, resulting in lower GY and WUE [47,48]. However, moderate drought during non-sensitive stages was shown to promote root penetration and improve WUE [49]. Through the optimization of irrigation water, deficit irrigation can be an effective way to attain comparable or higher GY while saving water [50]. For both wheat and maize under the I1-I3 treatments, GY and WUE significantly decreased with the increasing CWSI ( Table 2). The I4 treatment with the best soil moisture conditions produced the highest GY of wheat. For maize, the I3 treatment produced the highest GY. However, the difference in GY between I3 and I4 was not significant for both crops. Compared to the I1 treatment, I2, I3 and I4 increased the GY of wheat by 9%, 15% and 20%, respectively. Low GY of wheat with I1 treatment was associated with significant water stress at critical stages [51]. Generally, increasing water supply increased ET c , while decreasing WUE. For wheat, ET c with the I2, I3 and I4 treatments was increased by 10%, 19% and 26%, respectively, compared to the I1 treatment, while for maize crops, ET c was 6-15% higher with the I2-I4 treatments compared to the I1 treatment. The greatest WUE (1.58 kg m −3 ) was observed with the I1 treatment for wheat, whereas for maize crops, I3 produced the highest WUE (2.32 kg m −3 ). On the contrary, WUE with the I4 treatment was the lowest for both crops, implying that excessive water supply lowered WUE [52]. The study indicated that the responses of WUE to various drought levels differed among crop species. Wheat crops were more likely to generate high WUE under high CWSI (0.45), while maize plants generated the greatest WUE under moderate CWSI (0.28). Previous studies reported that mild water stress increased WUE, while high water stress imposed at critical stages decreased the WUE of wheat and maize markedly [28,30,34]. In this study, although precipitation in the maize growing season was adequate, supplemental irrigation of 30 mm was added after sowing. This was because initial SWC was well below the depletion threshold prior to maize sowing. On average, the initial SWC with I1 treatment was only 10.5% v/v, and that of I2-I4 treatments was 13.8% v/v in the 0-50 cm soil depth, which was very close to the wilt point ( Figure S1). Low SWC was also observed during the vegetative period of maize, which finally caused yield losses of 6-10% with deficit irrigation treatments. The study implied that antecedent treatments during the wheat season influenced the initial SWC of maize markedly, and thus caused significant ET c and GY differences for maize crops under rain-fed conditions.

Diurnal Dynamics of Canopy Temperature and Crop Evapotranspiration
Canopy temperature (T c ) can be used to predict water stress, since it was associated with soil water status [53]. For example, a significant decrease in T c usually occurred after irrigation and precipitation (>10 mm) events ( Figure 5A). The lowest T c (-1.1 • C) occurred during the overwintering of wheat, whereas the highest T c (40.4 • C) appeared at wheat maturity. On average, seasonal mean T c with I1, I2, I3 and I4 was 23.5, 22.3, 21.9 and 21.2 • C, respectively. For maize crops, the lowest T c (18.2 • C) appeared at the R6 stage, while the highest T c (37.4 • C) appeared at the V12 stage. Previous studies pointed out that too low T c might predict no stress [54,55]. The findings were also proven by our study in which CWSI was only 0.0-0.25 with significant low T c in over-wintering, which might result in an incorrect irrigation decision [56]. Thus, T c combined with soil moisture measurement may be a better way to quantify the water stress of wheat during the over-wintering period.
Diurnal dynamics of ET c were stable from the sowing to seedling stages of wheat, with a mean value of 1.80 mm d −1 ( Figure 5B). It gradually decreased to 0.09 mm d −1 during overwintering. ET c increased with the increasing T a during re-greening, and peaked in the middle period of the filling stage. The maximum ET c of wheat was 5.95, 8

Diurnal Dynamics of Crop Water Stress Index and T c − T a
Daily CWSI decreased after irrigation or precipitation events and increased during the period between events ( Figure 6A). Generally, CWSI and T c − T a had similar trends among treatments. CWSI dynamics varied inversely with irrigation levels, with an appropriate range of CWSI from 0.1 to 0.3. The result was consistent with the findings of previous studies conducted in China [28,29]. The lowest irrigation amount (I1) generated the highest CWSI of 0.63, which was 46%, 55% and 64% higher than those of I2, I3, and I4 treatments. Water stress significantly increased T c − T a ( Figure 6B). The minimum T c − T a for wheat was −4.5 • C after over-wintering irrigation. It was found that T c − T a values were basically negative the day after irrigation, which was consistent with the results of Chen et al. (2005) [27]. Several low T c − T a values were also induced by heavy precipitation. The Agriculture 2021, 11, 958 9 of 16 maximum T c − T a value (4.71 • C) was observed at the flowering stage of wheat with I1 treatment, which was 87%, 98% and 110% higher than that of I2, I3, and I4 treatments. The maximum T c − T a value was very close to the upper T c − T a limit of 5 • C proposed by Sepaskhah and Kashefipour (1994) [57]. During the maize season, only I1 treatment generated T c − T a > 0 • C, with a maximum value of 1.35 • C, and a CWSI of 0.37, at V12 stage.  Precipitation is presented as ten-day cumulative values. Irrigation is average values among I1, I2, I3, and I4 treatments.

Diurnal Dynamics of Crop Water Stress Index and Tc − Ta
Daily CWSI decreased after irrigation or precipitation events and increased during the period between events ( Figure 6A). Generally, CWSI and Tc − Ta had similar trends among treatments. CWSI dynamics varied inversely with irrigation levels, with an appropriate range of CWSI from 0.1 to 0.3. The result was consistent with the findings of previous studies conducted in China [28,29]. The lowest irrigation amount (I1) generated the highest CWSI of 0.63, which was 46%, 55% and 64% higher than those of I2, I3, and I4 treatments. Water stress significantly increased Tc − Ta ( Figure 6B). The minimum Tc − Ta for wheat was −4.5 °C after over-wintering irrigation. It was found that Tc − Ta values were basically negative the day after irrigation, which was consistent with the results of Chen et al. (2005) [27]. Several low Tc − Ta values were also induced by heavy precipitation. The maximum Tc − Ta value (4.71 °C) was observed at the flowering stage of wheat with I1 treatment, which was 87%, 98% and 110% higher than that of I2, I3, and I4 treatments. The

Correlation between ET c and T c and between CWSI and T c − T a
The relationship between CWSI and T c − T a was fitted to a positive linear equation (r 2 > 0.695) ( Figure 7A,C). The slopes for both wheat and maize crops were highest (0.055-0.059) with I1 treatment, implying that water stress accelerated the increasing rate of CWSI with T c − T a . As for I2-I4 treatments, they had similar slopes for both crops. A similar linear relationship between CWSI and T c − T a was also observed in previous studies [58][59][60]. During the maize growing season, CWSI values fell in a proper range of relieved stress most of the time, which favored maize growth. Similarly, a significant linear relationship was observed between ET c and T c (r 2 > 0.548) ( Figure 7B,D). The slopes increased with irrigation levels, indicating that the increasing rate of ET c with T c increased under well-watered conditions, whereas drought stress reduced the sensitivity of ET c to T c for wheat and maize. Similar results were also observed in spring maize and alfalfa (Medicago Sativa Linn) under a semiarid climate in Northwest China [61].
re 2021, 11, x FOR PEER REVIEW 10 of 17 maximum Tc − Ta value was very close to the upper Tc − Ta limit of 5 °C proposed by Sepaskhah and Kashefipour (1994) [57]. During the maize season, only I1 treatment generated Tc − Ta > 0 °C, with a maximum value of 1.35 °C, and a CWSI of 0.37, at V12 stage. Figure 6. Dynamics of (A) crop water stress index (CWSI) and (B) canopy minus air temperature (Tc − Ta, °C) in a winter wheat-summer maize cropping system in 2019-2020 growing season. Precipitation is presented as ten-day cumulative values. Irrigation was average values among I1, I2, I3, and I4 treatments.

Correlation between ETc and Tc and between CWSI and Tc − Ta
The relationship between CWSI and Tc − Ta was fitted to a positive linear equation (r 2 > 0.695) ( Figure 7A,C). The slopes for both wheat and maize crops were highest (0.055-0.059) with I1 treatment, implying that water stress accelerated the increasing rate of CWSI with Tc − Ta. As for I2-I4 treatments, they had similar slopes for both crops. A similar linear relationship between CWSI and Tc − Ta was also observed in previous studies [58][59][60]. During the maize growing season, CWSI values fell in a proper range of relieved stress most of the time, which favored maize growth. Similarly, a significant linear relationship was observed between ETc and Tc (r 2 > 0.548) ( Figure 7B,D). The slopes increased with irrigation levels, indicating that the increasing rate of ETc with Tc increased under wellwatered conditions, whereas drought stress reduced the sensitivity of ETc to Tc for wheat and maize. Similar results were also observed in spring maize and alfalfa (Medicago Sativa Linn) under a semiarid climate in Northwest China [61].

Determining the Thresholds of CWSI and T c − T a
Determining reliable CWSI thresholds is vital for precision irrigation, since it directly affects GY and WUE [48]. In order to make correlation analysis comparable, GY and WUE data were normalized to a range of 0.0 to 1.0. For wheat, linear decrease and increase relationships were observed for GY (r 2 = 0.915) and WUE (r 2 = 0.873) with increasing CWSI thresholds, respectively ( Figure 8A). Meanwhile, for maize, quadratic equations were established for both GY (r 2 = 0.856) and WUE (r 2 = 0.629) ( Figure 8B). The quadratic relationships between CWSI and GY were also reported in previous studies in terms of citrus (Citrus reticulata Blanco) and maize [62,63]. However, most studies reported a linear relationship between CWSI and GY in wheat, potato (Solanum tuberosum L.), and cotton (Gossypium spp) [27,[64][65][66][67].

Determining the Thresholds of CWSI and Tc − Ta
Determining reliable CWSI thresholds is vital for precision irrigation, since it directly affects GY and WUE [48]. In order to make correlation analysis comparable, GY and WUE data were normalized to a range of 0.0 to 1.0. For wheat, linear decrease and increase relationships were observed for GY (r 2 = 0.915) and WUE (r 2 = 0.873) with increasing CWSI thresholds, respectively ( Figure 8A). Meanwhile, for maize, quadratic equations were established for both GY (r 2 = 0.856) and WUE (r 2 = 0.629) ( Figure 8B). The quadratic relationships between CWSI and GY were also reported in previous studies in terms of citrus (Citrus reticulata Blanco) and maize [62,63]. However, most studies reported a linear relationship between CWSI and GY in wheat, potato (Solanum tuberosum L.), and cotton (Gossypium spp) [27,[64][65][66][67].
In the previous literature, the determination of CWSI thresholds was estimated mainly by experience, which was, to some extent, empirical [68][69][70]. Based on the functions proposed for wheat and maize crops, a distinctly different solution was adopted for the determination of reliable CWSI. In this study, the reliable threshold was defined as the one that can achieve the dual goal of high GY and high WUE. Based on the definition, we considered that the cross values of the two functions was the threshold value that satisfies the dual goal. For example, a normalized value of 0.575 was considered desirable at the CWSI threshold of 0.322 for winter wheat. The threshold value of wheat was very close to In the previous literature, the determination of CWSI thresholds was estimated mainly by experience, which was, to some extent, empirical [68][69][70]. Based on the functions proposed for wheat and maize crops, a distinctly different solution was adopted for the determination of reliable CWSI. In this study, the reliable threshold was defined as the one that can achieve the dual goal of high GY and high WUE. Based on the definition, we considered that the cross values of the two functions was the threshold value that satisfies the dual goal. For example, a normalized value of 0.575 was considered desirable at the CWSI threshold of 0.322 for winter wheat. The threshold value of wheat was very close to that estimated by Zhao (2016) in Hengshui in the NCP [68]. Similarly, an optimal threshold of 0.299 was expected for summer maize, since it obtained a high normalized value of 0.765. The threshold value of maize was higher than that estimated by Wang (2008), who conducted a similar experiment on summer maize in Wuhan, Central China, whose climate was more humid than Xuchang [69]. Compared with a sub-humid climate in this study, a CWSI limit of 0.22 was chosen for maize under the Mediterranean semiarid climate [44], implying that climatic conditions had noticeable effects on CWSI thresholds. In the sub-humid subtropical region of India, it was reported that the threshold of CWSI should be maintained below 0.30 for the optimum yield of wheat [70], which was very close to the threshold in this study.  [68] and Wuhan [69] were adopted to compare the CWSI thresholds for winter wheat and summer maize from Xuchang. Date were reconstructed and reproduced with the permission from Zhao Y., Master's Degree Dissertation; published by Chinese Academy of Agricultural Sciences, Beijing, China, 2016; and Wang S., Master's Degree Dissertation; published by Huazhong Agricultural University, Wuhan, China, 2008. GY and WUE values were normalized before use. Vertical error bars indicate standard deviation (SD). * and ** mean significant correlation at p < 0.05 and p < 0.01, respectively.

Conclusions
In conclusion, the soil water content was depleted at maturity by winter wheat, causing a noticeably low soil water content during the vegetative period of summer maize. This explained why CWSI thresholds in wheat seasons also exerted an influence on maize  [68] and Wuhan [69] were adopted to compare the CWSI thresholds for winter wheat and summer maize from Xuchang. Date were reconstructed and reproduced with the permission from Zhao Y., Master's Degree Dissertation; published by Chinese Academy of Agricultural Sciences, Beijing, China, 2016; and Wang S., Master's Degree Dissertation; published by Huazhong Agricultural University, Wuhan, China, 2008. GY and WUE values were normalized before use. Vertical error bars indicate standard deviation (SD). * and ** mean significant correlation at p < 0.05 and p < 0.01, respectively. Adapted with permission from Ref. [68]. 2016, Zhao Y. Adapted with permission from Ref. [69]. 2008, Wang S.
In addition, through the linear equations between T c − T a (y) and CWSI (x) (r 2 ≥ 0.719), T c − T a threshold can be determined (Table 3). In this study, the thresholds of T c − T a were estimated to 0.925 • C for winter wheat and 0.498 • C for summer maize, respectively. The results confirm the previous conclusion that wheat might be more tolerant to water stress than maize, since it had higher thresholds of T c − T a [27,69]. Several studies believed that T c − T a > 0 • C was the limit value of water shortage for most crops, which was slightly lower than the results from this study [29,57]. The higher thresholds of T c − T a were probably attributable to the deficit irrigation regimes with the CPIS in this experiment. We considered that T c − T a > 0.5 • C for summer maize and T c − T a > 0.9 • C for winter wheat should be used as a reference index for irrigation decision-making in the NCP. Compared with CWSI, T c − T a values are more accessible, making irrigation scheduling much easier than CWSI-based approaches [71][72][73]. Those thresholds help us obtain higher GY with less water. Table 3. Using correlation between T c − T a (y, • C) and CWSI (x) to determine the T c − T a threshold ( • C) for winter wheat and summer maize.

Crop
Linear Function CWSI Threshold T c − T a Threshold r 2

Conclusions
In conclusion, the soil water content was depleted at maturity by winter wheat, causing a noticeably low soil water content during the vegetative period of summer maize. This explained why CWSI thresholds in wheat seasons also exerted an influence on maize plants under rain-fed conditions. The crop water stress index (CWSI) had significant linear relationship with T c − T a . In this study, we defined the reliable threshold as the one that achieved the dual goal of high GY and high WUE. Based on the definition, we considered that the cross values of the two functions were optimal threshold values that satisfied the goal. In this study, CWSI thresholds were proposed as 0.252 for winter wheat, and 0.229 for summer maize, corresponding to a T c − T a threshold value of 0.925 and 0.498 • C, respectively. Compared with traditional soil moisture measurement, the measurement of CWSI and T c − T a was a non-destructive, non-contact, and reliable way for the precise estimation of crop water deficit. We conclude that farmers can achieve the dual goal of high GY and high WUE using the optimal thresholds proposed for a winter wheat-summer maize cropping system in the NCP.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/agriculture11100958/s1, Table S1. Goodness of fit between soil water content (SWC, % v/v) measured using Insentek sensors and oven-dry method across different soil textures in the laboratory at the Xuchang Irrigation Experiment Station in 2017-2018. Figure S1. Seasonal variations in soil water content (SWC, % v/v) under (A) I1, (B) I2, (C) I3, and (D) I4 treatments in a winter wheat-summer maize cropping system during the 2019-2020 growing season.