Evaluation of the Crop Water Stress Index as an Indicator for the Diagnosis of Grapevine Water Deficiency in Greenhouses

Precise irrigation management of grapevines in greenhouses requires a reliable method to easily quantify and monitor the grapevine water status to enable effective manipulation of the water stress of the plants. This study evaluated the applicability of crop water stress index (CWSI) based on the leaf temperature for diagnosing the grapevine water status. The experiment was conducted at Yuhe Farm (northwest China), with drip-irrigated grapevines under three irrigation treatments. Meteorological factors, soil moisture contents, leaf temperature, growth indicators including canopy coverage and fruit diameter, and physiological indicators including SPAD (relative chlorophyll content), stem water potential (φs), stomatal conductance (gs), and transpiration rate (E) were studied during the growing season. The results show that the relationship between the leaf-air temperature difference (Tc-Ta) and the plant water status indicators (φs, gs, E) were significant (P < 0.05), and the relationship between gs, E and Tc-Ta was the closest, with R2 values ranging from 0.530–0.604 and from 0.545–0.623, respectively. CWSI values are more easily observed on sunny days, and it was determined that 14:00 BJS is the best observation time for the CWSI value under different non-water-stressed baselines. There is a reliable linear correlation between the CWSI value and the soil moisture at 0–40 cm (P < 0.05), which could provide a reference when using the CWSI to diagnose the water status of plants. Compared with the Tc-Ta value, the CWSI could more accurately monitor the plant water status, and above the considered indictors, gs has the greatest correlation with the CWSI.


Introduction
Grapevine is one of the important economic fruit trees in China. In recent years, with the adjustment of agricultural industrial structure, the cultivation area of grapevine has increased greatly year by year in China. By the end of 2015, the viticulture area in China had expanded to 800,000 ha, making China the world's largest grape producer [1]. In recent years, with the improvement of grapevine protected cultivation technology, grapevines are the fruit tree species with the largest protected cultivation area in China, with an output that is 3.5 times in greenhouse higher than that of open planting. This method has significant advantages in ensuring the freshness and supply of fruit during the off season. The development of protected cultivation is an effective way to promote the development of rural economies.
have been conducted on diagnosing the plant water status using the CWSI based on the leaf stomatal conductance and transpiration rate.
The main aims of this study were (1) to analyze the relationship between T c -T a and plant water status indicators, (2) to determine the NWSB (non-water-stressed baseline) for grapevine in greenhouse and as well as CWSI diurnal time courses under different stages, (3) to evaluate the feasibility of the CWSI in predicting the canopy coverage, fruit diameter, and SPAD of grapevines, and (4) to clarify the relationship between the CWSI and plant water status indicators. The above objectives are aimed at evaluating the applicability of the CWSI in monitoring grapevine growth and diagnosing the water deficiency status.

Study Area
The experiment was carried out in greenhouse of Yuhe Farm, Shaanxi Province, from March to July 2018 (108.58 • E, 37.49 • N). The annual average rainfall in this area is 365.7 mm, the annual average temperature is 8.3 • C, the annual relative humidity is 69.37%, and the annual average duration of sunshine is 2893.5 h, which is representative of the typical continental marginal monsoon climate of the area. Table 1 shows the meteorological data (cultivation stage averages) recorded over the experimental year. The test soil was an aeolian sandy soil. The chemical properties of the soil were as follows: the soil ammonium nitrogen was 7.48 mg/kg, the nitrate nitrogen was 22.91 mg/kg, the available phosphorus was 4.07 mg/kg, and the available potassium was 163.47 mg/kg. The physical properties of the soil are shown in Table 2.

Experimental Design
Six year-old grapevine (early maturing variety [6][7][8][9][10][11][12] were planted in greenhouse, and grapevines with good growth and similar shapes were selected for the experiments. The entire growth period of grapevines can be divided into four main growth stages: the vegetative stage, the flowering stage, the fruit expansion stage, and the coloring mature stage, the cultivation period was 121 days during the growth season. The greenhouse was oriented east-west and was 70 m long and 9 m wide. The grapevine row spacing was 0.8 m, and the plant spacing was 0.6 m, with 14 grapevines per row. Artificial warming was carried out in greenhouse to ensure the growth temperature of grapevines on 11 March 2018. Drip irrigation was used in the experiment. A single-wing labyrinth drip irrigation belt (produced by Xinjiang Dayu Water Saving Company, Xinjiang, China) was adopted. The inner diameter and wall thickness of drip irrigation belt was 20 mm and 0.18 mm, respectively. The distance between the drippers was 300 mm, the design flow of the dripper was 4.0 L/h, and the laying mode of the drip belt was one row of two pipes.
The experiment was conducted with drip-irrigated grapevines under three irrigation treatments: a full irrigation treatment (T 1 : 100% M) and two regulated deficit irrigation treatment (T 2 : 80% M; T 3 : 60% M), M represents the irrigation quota. There were three treatments in total and three plots per treatment (each plot had a length of 8 m, a width of 4.5 m, and an area of 36 m 2 ), with a random block arrangement. The irrigation dates and irrigation amount is shown in Table 3, the grapevines were irrigated 12 times during the entire growth period. The irrigation quota was calculated by Equation (1). The irrigation time was determined according to whether or not T 1 reached the lower limit of the water quantity, which was 65% of β 1 at the vegetative and coloring mature stages and 70% of β 1 at the flowering and fruit expansion stage. The predicted wet layer depth of the soil was 80 cm. The total amount of fertilization during the entire growth period was 0.84 t/ha 1 , and the proportion of N:P:K was 1.0:0.6:1.2. Fertilization was carried out over three periods: the germination stage accounted for 20% of the total amount of fertilization, the flowering and fruit expansion stages accounted for 60%, and the coloring mature stage accounted for 20%. The drip irrigation and fertilization were controlled by integrated irrigation and fertilization equipment. The total irrigation amount during the entire growth period of each treatment was 3810 m 3 /ha, 3045 m 3 /ha, and 2280 m 3 /ha, respectively.
where M represents the irrigation quota, mm; γ s represents the soil dry bulk density, 1.64 g/cm 3 in 0-40 cm soil depth, 1.46 g/cm 3 in 40-80 cm soil depth; H is the predicted wet layer depth of soil, 0.8 m; P is the designed wet soil ratio, 0.8; β 1 is the field water holding capacity, 13.18% in 0-40 cm soil depth, 17.45% in 40-80 cm soil depth; and β 2 is the lower limit of the soil moisture content, 65% of β 1 at the vegetative and coloring mature stages and 70% of β 1 at the flowering and fruit expansion stage. The air temperature (T a ), relative humidity (RH), and solar radiation (R a ) were recorded automatically every 30 min using a Watchdog micro series (Spectrum Technologies Inc., Chicago, IL, USA) meteorological station in the greenhouse. The vapor pressure deficit (VPD) was estimated by the Horticulturae 2020, 6, 86 5 of 19 relative humidity (RH) and air temperature (T a ) and was calculated by the modified Penman formula. The formula [30] is shown in Equation (2): The soil moisture automatic monitoring system consisted of an EM50 data recorder (Environmental Logging System, Decision Devices, Inc., Salt Lake City, UT, USA) and four ECH 2 O 5 TE sensors (Decision Devices, Pullman, WA, USA). The soil moisture automatic monitoring system was installed 30 cm from the base of the grapevines and perpendicular to the planting row. Three representative grapevines were selected, and three monitoring systems were installed for each treatment. A sensor was installed every 20 cm, the buried depth was 80 cm, the soil volume moisture content was recorded every 30 min. Before the beginning of the growing season, the ECH 2 O 5 TE sensor was calibrated every 20 cm using the drying method, and then the sensor was calibrated every 10 days using the drying method.

Leaf Temperature and Leaf-Air Temperature Difference
The leaf temperature was continuously monitored and measured every 30 min by an ELV-15A infrared thermometer (Optris CSMTL10, Germany, ± 1.5 • C) that was part of the plant physiological and ecological monitoring system. The probe of the infrared thermometer faced the upper part of the grapevine canopy, and the distance between the detection window and the fully expanded leaves of the grapevine canopy was kept at 5-10 cm. The measured data were recorded by a computer, and the mean of three repetitions was used as the result. The leaf-air temperature difference (T c -T a ) is defined as the difference between leaf temperature (T c ) and air temperature (T a ) and can reflect the self-regulation ability of plant leaves that are under water stress to a certain degree [31].

Canopy Cover and Fruit Diameter
From the vegetative stage to the fruit expansion stage, the canopy cover was measured once every 5-7 days. Three representative grapevines with good growth conditions were selected for each treatment. Canopeo is an ACT image analysis tool (developed by the University of Oklahoma) using color values in the red-green-blue (RGB) system to measure the canopy cover. Nadir (i.e., downward-facing) images were taken from areas of experimental plots using "point and shoot" type digital cameras. The camera was kept at about 1.5 m from the top of the canopy using a 1.5 m monopod. Maintain adequate distance from the camera to the top of the canopy.
The fruit diameter was measured in the early stage of fruit growth, three grapevines were selected for each treatment, and a grape bunch was marked for each grapevine. The fruit diameter (transverse and longitudinal) in the middle of the grapevine bunch was measured with a digital vernier caliper (Mitutogo, Japan; the measurement range is 0-150 mm and precision is 0.01 mm) once every 3-5 days until the grapevines matured; the measurements were averaged for each replication.

Determination of Plant Physiological Indicators
Three grapevines were selected for each treatment. From the vegetative stage to the mature stage, a chlorophyll SPAD meter 'SPAD-502 (Konica Minolta, Tokyo, Japan) was used for the precise, rapid, and nondestructive estimation of SPAD value approximately every 15 days. The same healthy leaf was monitored before the fruit expansion stage, and the fifth healthy leaf was monitored at the internode of fruiting branch in the upper part of the plant after the fruit expansion stage. Five values were measured for each leaf, the measurements were averaged for each replication. The pressure chamber (TP-PW-II, Zhejiang Top Cloud-agri Technology Co., Ltd., Hangzhou, Zhejiang, China) was used to measure the stem water potential (ϕ s ) every 5-7 days, and the ϕ s is measured at 9:00 to 10:00 BJS. Three grapevines were selected for each treatment, and one branch under good growth conditions was selected as the sample on the sunny side outside the crown. The sample was put into a plastic bag containing moist gauze and quickly brought into the laboratory. The sample was clamped in a pressure chamber and pressurized by gas (compressed nitrogen), the pressure used for exudation of tissue fluid was observed. At this time, the pressure value was the stem water potential. At the same time, stomatal conductance (g s ) was measured on the same day, but between 09:00-10:00 BJS. An open gas-exchange system Li-6400 XT (Li-Cor Biosciences, Lincoln, NE, USA) with a 2 cm × 3 cm standard chamber was used to measure the g s and leaf transpiration rate (E) under the ambient light (1500 µmol m −2 s −1 ) and the ambient CO 2 concentration (400 ± 5 µmol CO 2 mol −1 ) conditions.

Concept of CWSI and Its Calculation Based on the Idso Method
The empirical form of the CWSI is based on the linear relationship between T c -T a and the VPD of the air for a well-watered crop [11] during the day and under homogeneous conditions and a clear sky. NWSB represents non-water-stressed baseline, which is VPD-dependent. According to the experimental design, the soil moisture content of T 1 was always above 65% of the field water capacity. It could be considered that grapevines under T 1 were in the potential evaporation state during the entire growth period. Therefore, the lower limit baseline of the CWSI was established by using the T c -T a of T 1 . Based on the lower limit baseline and the temperature and relative humidity (RH) measured on that day, the upper limit baseline could be calculated by the formula proposed by Idso et al. [11] and Jackson et al. [12]. The CWSI is calculated as follows: where a and b are the slopes and constants of the fitting equation, e s (T a ) is the saturated vapor pressure at air temperature under specific conditions, and e a is the actual vapor pressure. These formulas follow the following fact that transpiration decreases the leaf temperature relative to the air temperature; the decrease in temperature is greater at relatively high VPD values than at low VPD values. The variation range of the index is from 0 to 1. The closer the CWSI value is to 1, the more serious the water stress is. On the contrary, 0 indicates no water stress.

Data Analysis
The correlation analysis and regression analysis were carried out using SPSS 21.0 software (SPSS Inc., Chicago, IL, USA). Multiple comparisons were performed by least significant difference tests, with a significance level of 0.05. Microsoft Excel 2010 Software was used for processing data. The graphs were created by using Origin 9.1. Correlation analysis was conducted between T c -T a and meteorological and leaf temperature datas every 30 min, and 7 days were selected for each growth stage. The relationships between T c -T a and VPD as well as between T c -T a , CWSI and physiological measurements were analyzed through regression analyses. Linear regression were conducted between CWSI and growth index of grapevine, soil moisture content. In all cases, the coefficient of determination (R 2 ) was used to assess the goodness of fit of the associations among variables. The daily changes in T c -T a before and after irrigation under different stages were V-shaped curves ( Figure 1). It could be observed that the T c -T a value of each treatment was basically negative the day before and after irrigation, which was consistent with the results of Sepaskhah [32]. At the flowering stage, the T c -T a value of each treatment began to decrease after 13:00 and remained stable at approximately 20:00 ( Figure 1c). T c -T a reached a peak under the three treatments during the mature stage, which was −5.45~−7.01 • C ( Figure 1g). T c -T a decreased rapidly between 16:00-19:00. On the one hand, the leaf temperature was affected by R a and ground radiation; on the other hand, the specific heat capacity of water vapor is large, which made the leaf temperature decrease slowly. Thus, T c -T a appeared to decrease.

Results and Discussions
The peak of T c -T a after irrigation is higher than that before irrigation ( Figure 1). Yeşim [33] found that the range of T c -T a during the entire growth period was −4 • C~3 • C, which was slightly smaller than the current experimental findings because T c -T a is affected by the crop variety, canopy resistance, leaf size, and canopy density [34]. T c -T a showed a "double-peak" curve under the three treatments during the fruit expansion stage. The first peak was slightly higher than the second because the leaf stomata were actively closed due to the high temperature ( Figure 1f). When the environmental conditions cooled, the leaf stomata could not quickly recover to their original state. There was a certain lag, and the leaf temperature was still high, resulting in the peak of T c -T a being lower than the first peak. The peak of T c -T a at the mature stage developed at approximately 13:00 (Figure 1g,h). We concluded that the T c -T a of each treatment decreased as the water deficit increased because the leaf stomata tended to close under water stress, and the decrease in transpiration intensity would lead to a small amount of heat loss on the leaf surface; therefore, the leaf temperature was high, and T c -T a decreased [35]. Similar results were obtained by Rǔzica Stričević [36], who found that T c -T a between poplar and sorghum leaves increased as the water deficit increased.

Relationship between T c -T a and Meteorological Factors
The correlation analysis of T c -T a with meteorological factors (T a , R a , RH and VPD) and leaf temperature are shown in Table 4. The positive correlation between T c -T a and R a , T a and VPD was significant (P < 0.05), and the negative correlation with RH was significant (P < 0.05) ( Table 4). The results showed that meteorological factors had a significant effect on T c -T a . King and Shellie [37] obtained similar research conclusions. Moreover, Jensen et al. [38] believed that the main meteorological factor affecting T c was R a . This study indicated that R a had the least correlation with T c -T a , which was mainly because the plastic roofs of greenhouses would block part of the R a . Furthermore, R a at night was zero; thus, the change in T c was no longer affected by R a . The correlation coefficients of T c -T a and R a , T c -T a and RH decreased with the development of growth period, from 0.651-0.734 and 0.711-0.861 to 0.453-0.607 and 0.525-0.621, respectively (Table 4).

Relationship between Tc-Ta and Meteorological Factors
The correlation analysis of Tc-Ta with meteorological factors (Ta, Ra, RH and VPD) and leaf temperature are shown in Table 4. The positive correlation between Tc-Ta and Ra, Ta and VPD was significant (P < 0.05), and the negative correlation with RH was significant (P < 0.05) ( Table 4). The results showed that meteorological factors had a significant effect on Tc-Ta. King and Shellie [37] obtained similar research conclusions. Moreover, Jensen et al. [38] believed that the main meteorological factor affecting Tc was Ra. This study indicated that Ra had the least correlation with Tc-Ta, which was mainly because the plastic roofs of greenhouses would block part of the Ra. Furthermore, Ra at night was zero; thus, the change in Tc was no longer affected by Ra. The correlation

Relationship between T c -T a and Plant Water Status Indicators
The fitting diagrams of T c -T a with plant water status indictors (ϕ s , g s , and E) under the three treatments are shown in Figure 2, and the coefficient of the fitting equation between T c -T a and plant water status index (ϕ s , g s , and E) is shown in Table 5. The relationship between T c -T a and the plant water status indicators showed a correlation when dates were pooled together, and the correlation was significant among all treatments (P < 0.05), which indicates that T c -T a can reflect the water status of grapevines to a certain extent. The ϕ s , g s , and E values decreased with the decrease of T c -T a . It was easy to find that the relationship between ϕ s and T c -T a was a quadratic function curve (Figure 2a).
Gonzalez-Dugo et al. [39] also believed that the relationship between the stem water potential and the CWSI was a quadratic function curve, consistent with the results of this paper. The g s and E values were linearly correlated with T c -T a (Figure 2b,c). Khorsandi et al. [35] reported that there is a significant and linear relationship between T c -T a and g s . In general, the relationships between T c -T a and g s and between T c -T a and E were higher, with R 2 values ranging from 0.530-0.604 and from 0.545-0.623, respectively (Table 5).  Figure 2, and the coefficient of the fitting equation between Tc-Ta and plant water status index ( s ϕ , gs, and E) is shown in Table 5. The relationship between Tc-Ta and the plant water status indicators showed a correlation when dates were pooled together, and the correlation was significant among all treatments (P < 0.05), which indicates that Tc-Ta can reflect the water status of grapevines to a certain extent. The s ϕ , gs, and E values decreased with the decrease of Tc-Ta. It was easy to find that the relationship between s ϕ and Tc-Ta was a quadratic function curve (Figure 2a).
Gonzalez-Dugo et al. [39] also believed that the relationship between the stem water potential and the CWSI was a quadratic function curve, consistent with the results of this paper. The gs and E values were linearly correlated with Tc-Ta (Figure 2b,c). Khorsandi et al. [35] reported that there is a significant and linear relationship between Tc-Ta and gs. In general, the relationships between Tc-Ta and gs and between Tc-Ta and E were higher, with R 2 values ranging from 0.530-0.604 and from 0.545-0.623, respectively (Table 5).  Note: "a" represents the coefficient of the quadratic term; "b" represents the coefficient of the first-order term; "c" represents the constant.

Daily Changes in the CWSI under Different Weather Conditions
The daily changes in the CWSI under the two weather conditions are shown in Figure 3. The fluctuation of the CWSI under T 1 on sunny days was relatively stable (0.078-0.403) (Figure 3). Moreover, Jones [40] found that the leaf temperature of non-transpiration and full-transpiration could be used as the upper and lower limits of the canopy temperature, respectively. The change rule of the CWSI on cloudy days was similar to that on sunny days, but cloudy days showed a significantly larger range in magnitude than sunny days (−0.280-0.556). Gardner et al. [41] concluded that the stable CWSI values could be used to diagnose plant water deficiency, which is consistent with the results of this study. We concluded that the CWSI was more suitable for sunny weather, and the CWSI would be unstable on rainy days, which is also the reason why the CWSI model is limited as a suitable irrigation index [42,43]. Note: "a" represents the coefficient of the quadratic term; "b" represents the coefficient of the firstorder term; "c" represents the constant.

Daily Changes in the CWSI under Different Weather Conditions
The daily changes in the CWSI under the two weather conditions are shown in Figure 3. The fluctuation of the CWSI under T1 on sunny days was relatively stable (0.078-0.403) (Figure 3). Moreover, Jones [40] found that the leaf temperature of non-transpiration and full-transpiration could be used as the upper and lower limits of the canopy temperature, respectively. The change rule of the CWSI on cloudy days was similar to that on sunny days, but cloudy days showed a significantly larger range in magnitude than sunny days (−0.280-0.556). Gardner et al. [41] concluded that the stable CWSI values could be used to diagnose plant water deficiency, which is consistent with the results of this study. We concluded that the CWSI was more suitable for sunny weather, and the CWSI would be unstable on rainy days, which is also the reason why the CWSI model is limited as a suitable irrigation index [42,43].

NWSB and CWSI
There was a significant negative correlation between Tc-Ta and VPD during different stages ( Figure 4). Similar results were obtained by Khorsandi et al., who found that the negative correlation

NWSB and CWSI
There was a significant negative correlation between T c -T a and VPD during different stages ( Figure 4). Similar results were obtained by Khorsandi et al., who found that the negative correlation between T c -T a and VPD was significant [35]. The R 2 value was the highest (0.878) during the fruit expansion stage (Figure 4c), which might be related to the changes in transpiration and photosynthesis during the reproductive stage of grapevines. Furthermore, the absolute value of the slope of the equation during the flowering stage was larger than that during the other stages (Figure 4b), which indicates that the transpiration ability of grapevines changes due to different growth conditions under different stages, thus resulting in different ranges of T c decreases under the same VPD. Bellvert [44] established the upper (ul) and lower limit baselines (ll) of Pinot-noir grapevines as follows: ll = −1.709VPD + 2.534 and ul = 0.465 VPD + 6.125 (R 2 = 0.83). Bellvert [45] also developed the ll equation of grapevines as follows: ll = −1.326 VPD + 3.111 (R 2 = 0.424). However, differences in the crop varieties, management methods, soil moisture, and climate conditions will affect the upper and lower limit equations.
The daily changes in the CWSI under the three treatments during the different stages were inverted "V" curves that increased first and then decreased ( Figure 5). The CWSI of the three treatments reached the maximum at approximately 14:00 during the different stages, and the peaks of the CWSI under the three treatments were the largest during the fruit expansion stage, ranging from 0.51 to 0.81 (Figure 5c). The CWSI under T 1 was significantly lower than that under T 2 and T 3 , indicating that the CWSI increased as the irrigation amount decreased ( Figure 5). The results of this study confirmed the views of Sezen et al. [46] and of Colak and Yazar [47]. There were two peaks of the CWSI during the flowering and mature stages (Figure 5b,d), and the second peak was larger than the first peak. However, some studies [26,48] reported that the CWSI might exceed the range of 0~1.0. This might be caused by inconsistent baseline calculation methods. We concluded that the order of the CWSI for the three treatments during the entire growth period was T 3 > T 2 > T 1 , and 14:00 was the best observation time for the CWSI to reflect the water status of grapevines.
Horticulturae 2020, 6, x; doi: FOR PEER REVIEW www.mdpi.com/journal/horticulturae between Tc-Ta and VPD was significant [35]. The R value was the highest (0.878) during the fruit expansion stage (Figure 4c), which might be related to the changes in transpiration and photosynthesis during the reproductive stage of grapevines. Furthermore, the absolute value of the slope of the equation during the flowering stage was larger than that during the other stages ( Figure  4b), which indicates that the transpiration ability of grapevines changes due to different growth conditions under different stages, thus resulting in different ranges of Tc decreases under the same VPD. Bellvert [44] established the upper (ul) and lower limit baselines (ll) of Pinot-noir grapevines as follows: ll = −1.709VPD + 2.534 and ul = 0.465 VPD + 6.125 (R 2 = 0.83). Bellvert [45] also developed the ll equation of grapevines as follows: ll = −1.326 VPD + 3.111 (R 2 = 0.424). However, differences in the crop varieties, management methods, soil moisture, and climate conditions will affect the upper and lower limit equations. The daily changes in the CWSI under the three treatments during the different stages were inverted "V" curves that increased first and then decreased ( Figure 5). The CWSI of the three treatments reached the maximum at approximately 14:00 during the different stages, and the peaks of the CWSI under the three treatments were the largest during the fruit expansion stage, ranging from 0.51 to 0.81 (Figure 5c). The CWSI under T1 was significantly lower than that under T2 and T3, indicating that the CWSI increased as the irrigation amount decreased ( Figure 5). The results of this study confirmed the views of Sezen et al. [46] and of Colak and Yazar [47]. There were two peaks of the CWSI during the flowering and mature stages (Figure 5b,d), and the second peak was larger than the first peak. However, some studies [26,48] reported that the CWSI might exceed the range of 0~1.0. This might be caused by inconsistent baseline calculation methods. We concluded that the order of the CWSI for the three treatments during the entire growth period was T3 > T2 > T1, and 14:00 was the best observation time for the CWSI to reflect the water status of grapevines. The CWSI value of the T3 treatment showed a zigzag upward trend, while the CWSI of the T1 and T2 treatments increased first and then decreased during the growing season ( Figure 6 and Table  6). The CWSI values of the T1, T2, and T3 treatments were 0.07-0.30, 0.20-0.51, and 0.40-0.89 during the growing season ( Figure 6). Bellvert [44] reported that the CWSI was occasionally negative under full irrigation treatment, which was not found in the results of this study. As can be seen in Figure 6, The CWSI value of the T 3 treatment showed a zigzag upward trend, while the CWSI of the T 1 and T 2 treatments increased first and then decreased during the growing season ( Figure 6 and Table 6).
The CWSI values of the T 1 , T 2 , and T 3 treatments were 0.07-0.30, 0.20-0.51, and 0.40-0.89 during the growing season ( Figure 6). Bellvert [44] reported that the CWSI was occasionally negative under full irrigation treatment, which was not found in the results of this study. As can be seen in Figure 6, the CWSI was different in adjacent time periods, indicating that many external factors, including meteorological factors such as the wind speed, wind direction, clouds, VPD, T a , and R a , affect it [49].
The CWSI values of each treatment decreased after irrigation, and there were different degrees of lag ( Figure 6). The reason is that stomatal conductance does not reach the maximum rapidly as the soil moisture changes after rewatering; rather, it recovers slowly under the influence of early water stress, so the leaf temperature is still high [50]. The seasonal mean CWSI values of the three treatments were 0.181, 0.350, and 0.677, respectively (Table 6). Bellvert and Girona [51] reported that the CWSI values under a full irrigation treatment were lower than 0.5, while those under a severe water deficit treatment were close to 1.0; the findings of this study agree with their results. Yazar et al. [52] evaluated the CWSI values of ergot seedless and flame seedless and suggested that crops should be irrigated when the CWSI values reached 0.30-0.35 in the Mediterranean area.

Relations between the CWSI and Canopy Cover, Fruit Diameter and SPAD
Many studies showed that the CWSI values are closely related to the leaf area index [53], photosynthetic index [54], which can be used as an indirect index to identify the drought tolerance of crops. Based on the above analysis, the correlations of the CWSI values with the canopy cover (CC), fruit diameter (FD), and SPAD of grapevines is shown in Figure 7. On May 3, the peaks of CC under the three treatments were 95.32% for T1, 89.65% for T2 and 85.61% for T3 (Figure 7a). The FD of the grapevines maintained a rapid growth trend before May 20, and there was no significant difference among the treatments (Figure 7c)

Relations between the CWSI and Canopy Cover, Fruit Diameter and SPAD
Many studies showed that the CWSI values are closely related to the leaf area index [53], photosynthetic index [54], which can be used as an indirect index to identify the drought tolerance of crops. Based on the above analysis, the correlations of the CWSI values with the canopy cover (CC), fruit diameter (FD), and SPAD of grapevines is shown in Figure 7. On May 3, the peaks of CC under the three treatments were 95.32% for T 1 , 89.65% for T 2 and 85.61% for T 3 (Figure 7a). The FD of the grapevines maintained a rapid growth trend before May 20, and there was no significant difference among the treatments (Figure 7c). After May 20, the FD under the three treatments increased with increasing irrigation amount and reached 22.45-24.28

Relationship between the CWSI and the Plant Water Status Indicators
Soil moisture is the direct source of plant water. Drought stress will lead to decreases in the soil moisture absorption capacity of the root system and then cause the plant water status to decrease, which is an indirect and adaptive feature of plants to cope with the decrease of soil moisture [58]. To explore the relationship between the CWSI and soil moisture, the correlation between the CWSI and soil moisture at 0-40 cm and 40-80 cm was analyzed. The results showed that the fitting effect of the CWSI and soil moisture (0-40 cm) was sound, and the correlation was significant (P < 0.05) ( Figure  8a). However, the R 2 of the CWSI with soil moisture for the deep soil (40-80 cm) was relatively low (Figure 8b) and between 0.123 and 0.403. This indicates that the CWSI more suitably reflects the water status of shallow soil, which is beneficial to understanding the effect of drought stress on the plant water status to a certain extent. As Figure 7b shows, the correlation of the CWSI with CC reached a very significant level (P < 0.01). The current experimental finding agreed with the results of other crops [55,56]. There was a very significant negative correlation between the CWSI and FD (Figure 7d). However, Orta et al. [57] reported that crop morphological size was limited owing to water deficit, and the decrease in crop assimilation was the main reason for the decrease in fruit quantity and diameter at the fruit development stage. There was a very significant negative correlation of the CWSI with SPAD (Figure 7f), and the determination coefficient reached 0.717. By analyzing the relationship of the CWSI and the growth and physiological indexes, we concluded that the CWSI could effectively reflect the water status of grapevines.

Relationship between the CWSI and the Plant Water Status Indicators
Soil moisture is the direct source of plant water. Drought stress will lead to decreases in the soil moisture absorption capacity of the root system and then cause the plant water status to decrease, which is an indirect and adaptive feature of plants to cope with the decrease of soil moisture [58]. To explore the relationship between the CWSI and soil moisture, the correlation between the CWSI and soil moisture at 0-40 cm and 40-80 cm was analyzed. The results showed that the fitting effect of the CWSI and soil moisture (0-40 cm) was sound, and the correlation was significant (P < 0.05) (Figure 8a). However, the R 2 of the CWSI with soil moisture for the deep soil (40-80 cm) was relatively low (Figure 8b) and between 0.123 and 0.403. This indicates that the CWSI more suitably reflects the water status of shallow soil, which is beneficial to understanding the effect of drought stress on the plant water status to a certain extent. The relationship between plant water status indictors ( s ϕ , gs and E) and CWSI were close ( Figure   9), and the R 2 were high (Table 7). Moreover, the relationships between CWSI and s ϕ , gs and E were significant at the level of P < 0.01, which indicated that CWSI is a suitable indicator for the diagnosis of plant water status. The relationship between gs and CWSI was the closest, the R 2 was between 0.708 and 0.847, whereas the relationship between s ϕ E and CWSI was 0.565-0.730 and 0.569-0.711, respectively (Table 7). When the CWSI is compared to Tc-Ta as a proxy of the water status (by comparing the R 2 and P values of the relationship of both indices to s ϕ , gs, and E, Tables 5 and 7), it can be observed that the CWSI more accurately represents the crop water status. Compared to Tc-Ta, the performance of the CWSI is improved by considering the evaporative demand and the speciesspecific response of the relationship to VPD. These results are similar to the theoretical foundation of CWSI, demonstrating that the change of leaf temperature is mainly determined by the change of gs under water stress [59]. Under water stress, the stem and leaf water potential reached the lowest through stomatal regulation, which also explained that the correlation between CWSI and leaf, stem water potential and transpiration rate was not high. Berni et al. [60] reported that for olive trees, there is a linear relationship between CWSI and gs, the same results were also confirmed in nectarines [61]. The relationship between E and CWSI was weaker than that with gs (Table. 7), although the transpiration rate of leaves depended largely on gs, the reason may be that transpiration rate is not only affected by stomatal conductance but also on the conductivity of the boundary layer [62]. In many previous studies to evaluate the applicability of CWSI, s ψ [39,63] or l ψ [61] is usually the most suitable indicator to determine the applicability of CWSI as an indicator of water status. However, many studies showed that the relationships between CWSI and plant water status indictors were not always linear, which is consistent with the results of this study (Figure 9a,b). For instance, linear and curvilinear relationships were reported for both grapevine [44,45] and peach [63]. The relationship between plant water status indictors (ϕ s , g s and E) and CWSI were close (Figure 9), and the R 2 were high (Table 7). Moreover, the relationships between CWSI and ϕ s , g s and E were significant at the level of P < 0.01, which indicated that CWSI is a suitable indicator for the diagnosis of plant water status. The relationship between g s and CWSI was the closest, the R 2 was between 0.708 and 0.847, whereas the relationship between ϕ s E and CWSI was 0.565-0.730 and 0.569-0.711, respectively (Table 7). When the CWSI is compared to T c -T a as a proxy of the water status (by comparing the R 2 and P values of the relationship of both indices to ϕ s , g s , and E, Tables 5 and 7), it can be observed that the CWSI more accurately represents the crop water status. Compared to T c -T a , the performance of the CWSI is improved by considering the evaporative demand and the species-specific response of the relationship to VPD. These results are similar to the theoretical foundation of CWSI, demonstrating that the change of leaf temperature is mainly determined by the change of g s under water stress [59]. Under water stress, the stem and leaf water potential reached the lowest through stomatal regulation, which also explained that the correlation between CWSI and leaf, stem water potential and transpiration rate was not high. Berni et al. [60] reported that for olive trees, there is a linear relationship between CWSI and g s , the same results were also confirmed in nectarines [61]. The relationship between E and CWSI was weaker than that with g s (Table 7), although the transpiration rate of leaves depended largely on g s , the reason may be that transpiration rate is not only affected by stomatal conductance but also on the conductivity of the boundary layer [62].
Horticulturae 2020, 6, x; doi: FOR PEER REVIEW www.mdpi.com/journal/horticulturae only affected by stomatal conductance but also on the conductivity of the boundary layer [62]. In many previous studies to evaluate the applicability of CWSI, s ψ [39,63] or l ψ [61] is usually the most suitable indicator to determine the applicability of CWSI as an indicator of water status. However, many studies showed that the relationships between CWSI and plant water status indictors were not always linear, which is consistent with the results of this study (Figure 9a,b). For instance, linear and curvilinear relationships were reported for both grapevine [44,45] and peach [63].  Note: "a" represents the coefficient of the quadratic term; "b" represents the coefficient of the first-order term; "c" represents the constant.
In many previous studies to evaluate the applicability of CWSI, ψ s [39,63] or ψ l [61] is usually the most suitable indicator to determine the applicability of CWSI as an indicator of water status. However, many studies showed that the relationships between CWSI and plant water status indictors were not always linear, which is consistent with the results of this study (Figure 9a,b). For instance, linear and curvilinear relationships were reported for both grapevine [44,45] and peach [63].

Conclusions
It can be concluded that according to different NWSBs, the diurnal change in the CWSI among the three treatments was obviously different, and the best observation time of the CWSI value was 14:00 BJS. The CWSI value can be used to predict the canopy coverage, fruit diameter, and SPAD. Moreover, there was a reliable relationship between T c -T a and the plant water status indicators; however, the CWSI could more accurately monitor the water stress and assess the water status variability in grapevines in greenhouses. Stomatal conductance had the closest relationship with the CWSI, outperforming other widely used plant water status indicators such as the leaf water potential, stem water potential, and leaf transpiration rate.
Plant water status can be diagnosed by monitoring leaf temperature data. Meanwhile, CWSI is closely related to soil moisture content, it can be combined with plant water status and soil moisture content to guide crop irrigation based on CWSI. The response of plants to CWSI is affected by many factors under water stress, to carry out a successful irrigation schedule based on CWSI, the future research should also consider the influence of cultivation methods, climate conditions and other factors on CWSI, and the threshold of CWSI is further obtained when plants need irrigation under water stress. At present, CWSI is widely used in satellite remote sensing technology. The study of CWSI measured in the field can provide reference for CWSI based on satellite thermal infrared, the combination of CWSI by the two methods can diagnose plant water deficiency more accurately and predict irrigation schedule in future.
Author Contributions: C.R., X.H., W.W. and H.R. designed the experiments, C.R., T.S., Y.G. performed research and date analysis, C.R. wrote the manuscript with contributions from all authors. All authors have read and agreed to the published version of the manuscript.