Water scarcity is a major factor that limits wheat production in many areas of the world. Many studies have indicated that the water supply for irrigation is expected to decrease owing to climate change and increased competition from other sectors in many countries (Gupta et al., 2020). Thus, the extent to which crops suffer from water stress is gradually increasing. Considering limited water resources, there is an urgent need to understand when plants suffer water stress and to develop effective assessment methods of crop water stress for precision irrigation scheduling (Zhou et al., 2021).

The estimation of soil moisture provides objective criteria for irrigation management. Many methods have been developed to monitor soil water status, including soil water content, soil matrix water potential, and plant water status, including leaf water potential, stem water potential, stomatal conductance, transpiration rate, and photosynthetic rate (Nakhforoosh et al., 2016; Ballester et al., 2013). When plants experience water stress, stomata close, and transpiration decreases, and these changes lead to increasing the leaf and canopy temperatures (Jones, 1999). With the rapid development of high-resolution, non-contact, and high-throughput thermal infrared thermography, the canopy temperature (T_c) has become popular for assessing water stress in plants (Santesteban et al., 2017; Pappalardo et al., 2023). Certain studies have used T_c to calculate the crop water stress index (CWSI) and plan irrigation schedules for different crops and fruit trees (Agam et al., 2013; Morales-Santos et al., 2023).

CWSI can be calculated using an empirical approach defined by Idso et al. (1981). They indicated a positive linear correlation between the canopy and air temperature difference and the vapor pressure deficit (VPD) of a well-watered crop, which can be used to estimate non-water-stressed baselines. Many empirical studies have used these baselines to calculate CWSI. However, non-water-stressed baselines are highly sensitive to changes in weather factors such as net radiation and wind speed (Jackson et al., 1988). Certain studies indicated that there may be different non-water stressed baselines under different growing stages, and the relationships might also be influenced by different experimental locations (Gontia and Tiwari, 2008; Mangus et al., 2016). However, Taghvaeian et al. (2014) indicated that the same set of baselines for morphologically similar maize (Zea mays) species could be used to estimate the CWSI in regions with similar climatic conditions.

In seasons or locations with lower VPD, an empirical approach may not be suitable for evaluating water stress (Yuan et al., 2004). Some researchers have used the temperatures of wet and dry reference surfaces instead of the upper and lower boundaries of the canopy-air temperature difference, and estimated the CWSI (Apolo-Apolo et al., 2020). However, wet reference bodies must be adjusted over time (O’Shaughnessy et al., 2012). The energy balance theoretical CWSI (Jackson et al., 1981) requires estimating net radiation, wind speed, aerodynamic resistance, T_c, air temperature, and VPD. The theoretical method is more reasonable than the empirical approach, because the latter accounts for temperature changes due to radiation and wind speed but requires more parameters (Jackson et al., 1988).

Many studies have used the CWSI as an indicator of plant water status under field conditions and demonstrated its ability to monitor water stress (Ekinzog et al., 2022). A high correlation was found between the CWSI and the soil water content of the topsoil layers (Taghvaeian et al., 2012). In addition, many researchers have found strong relationships between the CWSI and leaf water potential, stem water potential, and stomatal conductance (Kirnak et al., 2019; Mangus et al., 2016; Santesteban et al., 2017). However, O’Shaughnessy et al. (2012) showed that the measurements of leaf water potential taken too soon after rainfall and termination of irrigation would affect the relationship between CWSI and leaf water potential.

Although the CWSI has been widely used to schedule crop irrigation, there is a distinct diurnal change pattern in CWSI values at different times of the day, and the selection of the timing of T_c measurement is very important (Taghvaeian et al., 2014). Typically, measurements are taken near noon on sunny days (O’Shaughnessy et al., 2012). Taghvaeian et al. (2012) indicated that the CWSI reached a maximum value from 12:00 to 13:00, which is considered the optimum time during the day for irrigation scheduling. In contrast, Li et al. (2010) demonstrated that a half-hourly averaged CWSI calculated during midday conditions produced useful estimates of crop water stress. Osroosh et al. (2016) demonstrated that the CWSI averaged over daylight hours was better than the midday CWSI. Since an accurate estimation of T_c is a prerequisite for obtaining the CWSI, the timing of the T_c measurement must be considered when using the CWSI to make irrigation decisions.

Researchers have attempted to capture T_c more frequently and accurately by installing infrared thermometers (IRTs) in the field to detect and diagnose crop water stress statues (Katimbo et al., 2022; Gonzalez-Dugo et al., 2020). Recently, unmanned aerial vehicles (UAVs) equipped with thermal infrared remote-sensing equipment have been widely used to monitor crop T_c and derive CWSI on large scales (Siegfried et al., 2024). Thermal infrared images were processed to determine the spatial distribution of T_c in the field. UAVs are often controlled at different heights over the crop canopy, and T_c may also be measured in different directions. Research has shown that the height and direction of infrared thermography, passing clouds, wind, and other weather factors may also affect CWSI values (Liu et al., 2022; Gadhwal et al., 2023).

Therefore, the objectives of this research were to 1) investigate the influences of atmospheric VPD, T_c obtained at different heights and directions of the infrared thermography, and timing of the day on the performance of CWSI, and 2) demonstrate the relationships between CWSI and soil water content for determining the best timing and measurement methods for obtaining CWSI to provide references for using T_c for crop water status monitoring.

1.   Material and methods

1.1   Study site

A field experiment to compare the CWSI derived from canopy temperatures measured at different timing of day, height, and direction to winter wheat canopy was conducted at Luancheng Experimental Station (37°53′N, 114°40′N; elevation 50 m) in the North China Plain (NCP) during two winter wheat-growing seasons of 2020–2021 and 2021–2022. The experimental area is in a monsoon climate zone with an average annual rainfall of 473 mm. Approximately 70% of the rainfall occurs from June to September, which is the summer. Rainfall during the winter wheat-growing season is approximately 120 mm. Irrigation is important for obtaining a higher grain yield for this crop. The soil at the experimental site is loam soil with an average field capacity of 36% (v/v) and a wilting point of 13% (v/v) for the top 2 m of the soil profile. The soil nutrient content was 80 mg∙kg⁻¹ for available N, 110 mg∙kg⁻¹ for available K, 25 mg∙kg⁻¹ for available P, and 18 g∙kg⁻¹ for organic matter in the top 0–20 cm tillage layer. Detailed soil physical characteristics of the study site can be found in Li et al. (2023).

1.2   Irrigation treatments

Winter wheat was sown in early October and harvested in mid-June of the following year. Cultivar ‘SX633’ was used for the 2020–2022 seasons. Row spacing was set at 15 cm. Seeding rate was 300 viable seeds per square meter. Before planting, diammonium phosphate (DAP) at 300 kg∙hm⁻², urea at 150 kg∙hm⁻² and potassium chloride at 150 kg∙hm⁻² were applied to the soil. An additional 150 kg∙hm⁻² urea was topdressed during the jointing stage in early April. Three irrigation treatments (I1, I2, and I3) were applied to induce different levels of water stress. The amounts and timing of the different irrigation treatments are listed in Table 1. The experiment was arranged in a randomized plot design, with four replicates for each treatment. There were 12 plots, each with an area of 4 m × 4 m. A flow meter was used to record the volume of irrigation water used.

Table  1.  Irrigation treatments used for canopy temperature measurements in winter wheat

Treatment	Irrigation time and amount	Total seasonal irrigation (mm)
One irrigation (I1)	At jointing stage with irrigation amount of 70 mm	70
Two irrigations (I2)	At jointing and anthesis stages with each irrigaiton amount of 70 mm	140
Three irrigations (I3)	At jointing, heading, and grain-filling stages with each irrigaiton amount of 70 mm	210

下载: 导出CSV 

| 显示表格

1.3   Measurements

1.3.1   Weather conditions

Daily meteorological factors, including air temperature, relative humidity, wind speed, radiation, and rainfall were acquired from an automatic weather station approximately 50 m from the experimental site. Daily reference evapotranspiration (ET₀) was calculated with the crop water program developed by the FAO using the Penman-Monteith equation, which represents the concept of grass reference evapotranspiration (albedo = 0.23, height = 0.12 m, surface resistance = 70 s∙m⁻¹) (Allen et al., 1998). Weather factors were used to calculate CWSI.

1.3.2   Canopy temperature

1) Regular canopy temperature measurements

The T_c was automatically monitored using an infrared thermometer (THERMO SHOT F30, Tokyo, Japan) installed on a 10 m tower at the north-middle edge of the experimental site. The thermometer was set at half an hour to take an image, and the images were downloaded from the thermometer and analyzed using the InfReC Analyzer NS9500 Lite software with the emissivity set at 0.98. The average temperatures from different plots were used for the analysis. The soil temperature of the underlying soil surface was manually excluded.

2) Height and directions of the measurements

Measurements of the T_c at three heights (a person standing at 3 m, 5 m, and 10 m height holding the thermometer) and three directions (the thermometer placed separately in the east, south, and north directions to the canopy with a view angle of 45°) were compared to examine whether the height and direction of the infrared thermometer placement influenced the T_c readings. The measurements were conducted for several days for comparison purposes.

1.3.3   Soil water contents

The soil water content of each plot was monitored regularly. An aluminum tube with a depth of 2 m was buried at the center in each plot, and soil volumetric water contents were measured from 0.2 m to 2 m in 0.2 m increments using a neutron probe (503DR, CPN International Inc., USA). Surface soil water content was further validated using the soil core method.

The fraction of available soil water for the main rooting zone of winter wheat was calculated as the available water (difference between the soil water content and wilting point) divided by the total available water (difference between the field capacity and wilting point). The major rooting depths used for winter wheat at different growing stages followed those of Zhang et al. (2004): 40 cm before winter dormancy, 60 cm from recovery to booting, 80 cm from booting to anthesis, and 100 cm during grain filling.

1.4   Calculation of CWSI

According to Idso et al. (1981), the CWSI can be computed using the following equation:

where T_c and T_a represent the canopy and air temperatures, respectively; D₁ is the maximum canopy and air temperature difference for a stressed crop; D₂ is the lower-limit canopy and air temperature difference for the well-watered treatment (the non-water-stressed baseline). The difference in the CWSI between the theoretical and empirical approaches was used in calculating D₁ and D₂.

1.4.1   Empirical approach (CWSI-E)

For the calculation of the D₂ and D₁, an empirical formula is showed as below (Idso et al., 1981; Idso, 1982):

where VPD is the air vapor pressure deficit (Pa); A and B are the intercept and slope of linear regression between the lower-limit canopy and air temperature difference and VPD, respectively; VPG is the difference between the saturation vapor pressure evaluated at the air temperature (T_a) and a temperature equal to T_a+A (Pa).

1.4.2   Theoretical approach (CWSI-T)

For the calculation of the D₂ and D₁, Jackson’s definition can be expressed as (Jackson et al., 1981):

Where R_n and G are the net radiative flux density (W·m⁻²) and the soil heat flux density (W·m⁻²), respectively. VPD is the air vapor pressure deficit (Pa). ρ is the air density. r_s and r_a are the potential canopy resistance and the aerodynamic resistance (s·m⁻¹), respectively. When the wind speed ≤ 2 m·s⁻¹, r_a was calculated by Eq. (7) (Thom and Oliver, 1977), when the wind speed > 2 m·s⁻¹, r_a was calculated using Eq. (8) (Monteith and Reifsnyder, 1974). ∆ is the slope of the saturation vapor pressure-temperature curve (Pa·℃⁻¹). C_p is the air specific heat at constant pressure (J·kg⁻¹·℃⁻¹). γ is the psychrometric constant (65.19 Pa·℃⁻¹). T is the average of T_c and air temperature (℃). Z is the reference height and takes the value of 2 m. Z₀ is the surface roughness length (m), Z₀ = 0.13h. d is the zero plane displacement height (m), d = 0.56h (Legg and Long, 1975). h is the plant height (m), which corresponds to the growing stage of the winter wheat. u is the wind speed at a height of 2 m (m∙s⁻¹). k is von Karman’s constant (0.41). r_min is minimum leaf resistance, r_min = 100 s·m⁻¹. LAI is the leaf index area. R_s and ∆R₁ are the solar radiation and the net long-wave radiation (J·m⁻²·s⁻¹), respectively. R_so is the solar radiation on a clear day (J·m⁻²·s⁻¹), and the value of R_so in this study is the maximum solar radiation during the measurement day. α is the surface albedo, α = 0.22. R_1↓ and R_1↑ are the incoming long-wave radiation and the outgoing long-wave radiation (J·m⁻²·s⁻¹), respectively. β_a is the air emissivity. β is the emissivity of thermal infrared camera, β = 0.98. σ is the Stefan-Boltzmann constant (5.675×10⁻⁸). T_c and T_a represent the canopy and air temperatures (℃), respectively.

1.5   Statistical analysis

A regression analysis was conducted to compare the CWSI derived from different sources of T_c measurements and the correlations between the CWSI and soil water content.

2.   Results

2.1   Changes in fraction of available soil water among the three irrigation treatments

The changes in the fraction of available soil water (FASW) for the main rooting depth under the three treatments for the two seasons were shown in Fig. 1. The soil water content was maintained at relatively stable levels from sowing to the end of winter dormancy owing to the lower ET₀ and smaller canopy. A quick growth and higher ET₀ after the recovery stage of wheat, and a rapid decline in soil moisture were observed, especially for the I1 treatment.

Figure  1.  Changes in the fraction of available soil water (FASW) in the main rooting depth under three treatments during 2020–2022 seasons

I1: one irrigation; I2: two irrigations; I3: three irrigations.

下载: 全尺寸图片 幻灯片

2.2   Establishment of the non-water stressed baselines

Data collected from the three irrigation treatments (I3) in 2020–2021 and 2021–2022 were used to calculate the CWSI values under non-water-stress conditions. As shown in Fig. 2, the canopy and air temperature difference (T_c−T_a) decreased as the VPD increased. Significant negative relationships were found between T_c−T_a and VPD at different growth stages in both seasons. The coefficients of determination (R²) ranged from 0.80 to 0.84 in 2020–2021 and from 0.87 to 0.94 in 2021–2022. The significant relationships indicate that the regression equations can be further calculated using the CWSI-E. The intercept of the T_c−T_a vs. VPD relationship in the middle grain-filling to maturity was higher than that in the jointing to anthesis and anthesis to middle grain-filling stages. The non-water-stress baselines differed from those developed by Agam et al. (2013). It should be noted that the climatic conditions were significantly different from those in the NCP in these studies. These results indicated that it is essential to establish non-water-stressed baselines during different growth stages in different regions.

Figure  2.  Relationship between canopy-air temperature difference (T_c−T_a) and vapor pressure deficit (VPD) at jointing to anthesis, anthesis to middle grain-filling, and middle grain-filling to maturity under I3 (three irrigations) during 2020–2022 seasons

下载: 全尺寸图片 幻灯片

2.3   Effect of VPD on the performance of CWSI

Ideally, the CWSI varies between 0 and 1, where 0 represents a potentially transpiring, well-watered condition, and 1 represents a non-transpiring, water-stressed condition. As shown in Fig. 3, most of the outliers of CWSI-T were concentrated in the region where the VPD was less than 2000 Pa. The CWSI-E fluctuated much more than the CWSI-T throughout the experiment and was significantly affected by environmental factors. These results indicated that a VPD > 2000 Pa is necessary to obtain accurate CWSI values.

Figure  3.  Distributions of crop water stress indexes (CWSI) calculated with an empirical approach (CWSI-E) and with a theoretical approach (CWSI-T) values based on data collected during 11:00−15:00 with the change of vapor pressure deficit (VPD) under I0, I1, and I2 treatments during 2020–2022 seasons

I1: one irrigation; I2: two irrigations; I3: three irrigations.

下载: 全尺寸图片 幻灯片

2.4   Distinct diurnal changes in canopy air temperature difference (T_c – T_a)

Figure 4 shows the distinct diurnal changes in T_c−T_a during the grain-filling stage at VPD < 2000 Pa and VPD > 2000 Pa during 2020–2022 seasons. As shown in Fig. 4, there were large variations in the T_c−T_a values at different hours of the day. T_c−T_a was positive in the morning because the transpiration rate of wheat was relatively low after sunrise; the T_c increased more rapidly than the T_a. Thus, the CWSI value did not reflect the water status of wheat. The transpiration rate increased with increasing radiation dose and VPD. Transpiration removed leaf water, decreased the temperature of the canopy surface, and the canopy-air temperature difference was negative in the afternoon, especially when the VPD was > 2000 Pa. These results suggest that the T_c measurement time is a key parameter for utilizing the CWSI approach.

Figure  4.  Diurnal changes in canopy-air temperature difference (T_c−T_a) during grain-filling stage at vapor pressure deficit (VPD) > 2000 Pa and VPD < 2000 Pa under three irrigation treatments during 2020–2022 seasons

I1: one irrigation; I2: two irrigations; I3: three irrigations.

下载: 全尺寸图片 幻灯片

2.5   Effects of measuring heights and directions on the T_c values

The height of the thermometer used to measure the T_c slightly influenced the T_c values (Fig. 5), especially during the morning. This difference disappeared by the afternoon. Overall, the difference in T_c measured at the three heights was small and could be neglected when calculating the CWSI (Fig. 6).

Figure  5.  Canopy temperature (T_c) of winter wheat under I3 treatment (three irrigations) at different heights and directions during 2021–2022 seasons

下载: 全尺寸图片 幻灯片

Figure  6.  Crop water stress indexes calculated with an empirical approach (CWSI-E) and with a theoretical approach (CWSI-T) of winter wheat under I3 treatment (three irrigations) at different heights and directions during 2021–2022 seasons

下载: 全尺寸图片 幻灯片

The direction of the thermometer used to measure T_c influenced the values (Fig. 5). These influences mainly occurred at midday at the three heights. The lens of the thermometer facing south recorded a higher T_c than the lens facing east or north. The lowest T_c value was obtained when the lens faced north. The influence of the lens direction on T_c at lower height was greater than that at higher height. The difference between the south and north directions could be up to 1 to 2 ℃ at noon. Because the CWSI is usually calculated using the T_c obtained during the middle of the day, the difference in T_c obtained from different directions might need to be considered. Fig. 6 also indicates a large variation in the CWSI-E and CWSI-T scores calculated using the three directions.

2.6   Diurnal and seasonal changes in the CWSI

The three irrigation treatments resulted in a wide range of CWSI values calculated from the T_c. As shown in Fig. 7, the CWSI outliers calculated using the empirical method (CWSI-E) were greater than those calculated using the theoretical method (CWSI-T). Most outliers were concentrated between 7:00 and 10:00. These results showed that midday measurements for estimating the CWSI are more reliable than morning estimates.

Figure  7.  Crop water stress indexes (CWSI) calculated with an empirical approach (CWSI-E) and with a theoretical approach (CWSI-T) during 7:00–18:00 under three irrigation treatments during 2020–2022 seasons

I1: one irrigation; I2: two irrigations; I3: three irrigations.

下载: 全尺寸图片 幻灯片

The seasonal variations in CWSI of wheat at 14:00 on sunny days under different irrigation treatments during 2020–2022 seasons are shown in Fig. 8. The CWSI values decreased with increasing soil water contents by irrigation or rainfall events. However, the decrease of CWSI lagged by 3–5 days. In both analytical forms, most of the CWSI values for the stressed irrigation treatments were higher than those of well-irrigated treatments. However, large differences were observed between the two methods. CWSI estimated via the empirical approach across irrigation treatments was greater than that via the theoretical approach. The CWSI-T values ranged from 0 to 1. However, the CWSI-E values varied greatly throughout the growing season.

Figure  8.  CWSI measurements computed for 14:00 throughout 2020–2022 winter wheat-growing seasons

CWSI-E computed following empirical approach; CWSI-T computed following theoretical approach. I1: one irrigation; I2: two irrigations; I3: three irrigations.

下载: 全尺寸图片 幻灯片

The values of CWSI-E and CWSI-T at the maturation stage were higher than those at the vegetative stage (Fig. 8). This may be because the leaves start to fall off during the late growth stages. The background soil influenced the T_c measurements. With the senescence of leaves, leaf activity decreases, and the reduction in crop transpiration is another reason for the higher temperature readings. These results indicate that the CWSI in the late-growing stage of winter wheat is not a good indicator of plant water stress.

2.7   Relationships between CWSI and physiological indicators

The corresponding physiological indicators were used as references for wheat water status to further compare the performances of CWSI-E and CWSI-T. As shown in Fig. 9, negative relationships were found between the CWSI with transpiration rate and photosynthetic rate. Theoretical methods have steeper slopes and are, therefore, more sensitive to changes in the CWSI. These results indicated that small changes in CWSI values lead to large changes in the predicted transpiration and photosynthetic rates.

Figure  9.  Relationships between transpiration and photosynthetic rates and crop water stress indexes calculated with an empirical approach (CWSI-E) and with a theoretical approach (CWSI-T) during 2020–2022 seasons

下载: 全尺寸图片 幻灯片

2.8   Using CWSI to assess crop water status

One objective of this study was to assess the possibility of using the estimated CWSI for irrigation scheduling. The relationship between the CWSI and fraction of available soil water (FASW) at 12:00–15:00 is given in Fig. 10. Generally, there were negative relationships between the CWSI and FASW, and water stress increased with a decrease in soil water content. Therefore, the CWSI increased as well. The R² between CWSI and FASW at 14:00 was higher than that at 12:00, 13:00, and 15:00, indicating that 14:00 was the optimal time for monitoring the water status of crops using CWSI. A slightly higher correlation was observed for the CWSI-T than for the CWSI-E. This indicated that the CWSI calculated using the theoretical method is more suitable for direct soil water content measurements.

Figure  10.  Relationships between fraction of available soil water (FASW) with crop water stress indexes calculated with an empirical approach (CWSI-E) and with a theoretical approach (CWSI-T) at 12:00, 13:00, 14:00, and 15:00 during 2020–2022 seasons

下载: 全尺寸图片 幻灯片

2.9   Seasonal average CWSI values

As shown in Table 2, grain yield increased with increasing irrigation. The grain yield increased by 8.0% from I1 to I2 and by 2.7% from I2 to I3 in 2020–2021. The values were 13.5% and 1.6% in 2021–2022, respectively. Considering the water shortage in the NCP, I2 (irrigated at the jointing and anthesis stages) is an economical option for minimizing water use and maximizing wheat yield.

Table  2.  Crop water stress index for 14:00 across the season as indicated by empirical (CWSI-E) and theoretical (CWSI-T) approaches and grain yield during 2020–2022 seasons

Season	Treatment	Grain yield (kg∙hm−2)	CWSI-E	CWSI-T
2020–2021	I3	7958.0a	0.07	0.20
	I2	7746.2a	0.23	0.26
	I1	7170.5b	0.48	0.37
2021–2022	I3	8783.1a	0.05	0.19
	I2	8648.4a	0.23	0.25
	I1	7617.7b	0.53	0.33
Different lowercase letters in the same season indicate siginicant differences among different treatments in the same season at P<0.05 level. I1: one irrigation; I2: two irrigations; I3: three irrigations.

下载: 导出CSV 

| 显示表格

In both analytical forms, the average CWSI during the growing season was smallest for I3 and largest for I1. In addition, the average CWSI-E value was higher than the CWSI-T under the most deficit irrigation treatment of I1. In the present study, the I2 treatment resulted in a relative higher grain yield. According to the two-year results, the seasonal average thresholds values of CWSI-E was 0.23 and CWSI-T ranged from 0.25 to 0.26 at 14:00. These could be used as a criterion for irrigation scheduling of winter wheat.

3.   Discussion

T_c has mostly been measured using infrared temperature sensors or thermal images at ground-based platforms to detect and quantify crop water stress (Morales-Santos et al., 2023; Bhatti et al., 2022). Recently, the development of modern remote-sensing technology has offered the possibility of fast, non-contact, real-time, inexpensive, and precise T_c measurements (Siegfried et al., 2024; Zhou et al., 2021). The T_c data are used to compute CWSI. Many studies have indicated that CWSI is an effective indicator for detecting crop water status and triggering irrigation (Bellvert et al., 2014; Luus et al., 2022). However, T_c is not only influenced by meteorological conditions such as wind speed, solar radiation, and cloud cover but also by the height and direction of infrared thermography (Fuentes et al., 2012; Bo et al., 2023; Candogan et al., 2013). Therefore, it is essential to obtain accurate and reliable canopy temperature scans to make irrigation decisions.

Modern remote-sensing technology is widely used to quantify crop water stress by using T_c measurements (Egea et al., 2017; Awais et al., 2022). Previous studies showed that ground-based platforms focusing on the water stress of individual plants, generally maintaining a distance of 0.3–2.0 m from camera to canopy, UAVs obtained thermal images from a distance of 30–100 m above the ground (Zhou et al., 2021). Awais et al. (2022) reported that UAV flights at 60 m altitude provided a more accurate T_c than flights at 25 m and 40 m. However, Gadhwal et al. (2023) indicated that UAVs flying at an altitude of 50 m performed better than flights at 30 m and 70 m. In the present study, the difference in T_c measured at the three heights was small in the afternoon because the height of the infrared thermometer was less than 10 m. In addition, the direction of the thermometer had a significant influence on the T_c values. Soil heat emittance resulted in higher temperature readings, leading to higher CWSI values. In this study, thermal images were collected with the view set to 45° relative to the canopy and shot vertically downward after the jointing stage when the canopy covered the ground. The issue of viewing the underlying soil was avoided by installing an infrared thermometer at an oblique angle.

The results showed that the selection of the T_c measurement time had a greater influence on the CWSI results. Stockle and Dugas (1992) indicated that weather conditions around noon are relatively stable, which is conducive to reducing the impact of the external environment on T_c. Taghvaeian et al. (2014) suggested that 12:00–14:00 was the appropriate data collection time to capture the maximum stress level experienced by the crop. Bellvert et al. (2014) reported that the transpiration rate differed significantly under different water supply conditions. Therefore, T_c at midday can effectively reflect the differences between treatments. Liu et al. (2022) recommended midday (13:00–14:00) as the optimal time to monitor the water status of cork oak (Quercus variabilis). Our study also found that midday measurements for estimating CWSI were more reliable for detecting crop water stress in winter wheat. This finding is similar to the results of a previous research study by Katimbo et al. (2022). However, noon is the optimal time to measure T_c to monitor the water status. The presence of clouds also makes it difficult to obtain T_c. Thus, it is necessary to measure T_c under clear-sky (high VPD, high solar radiation, and low wind speed) conditions (DeJonge et al., 2015; Zhang et al., 2023).

CWSI is an important indicator for assessing drought stress based on leaves or T_c (Santesteban et al., 2017). Both empirical (Idso et al., 1981) and theoretical (Jackson et al., 1981) approaches have been presented to estimate the CWSI. The main advantage of the empirical approach is that only three variables (T_c, T_a, and VPD) must be measured for its application (Candogan et al., 2013). However, the non-water stress baseline (NWSB) used in this method is highly sensitive to changes in climatic variables, such as wind speed and solar radiation (Ekinzog et al., 2022). Certain studies have indicated that there may be different NWSBs under different growing stages, and these relationships might be influenced by different experimental locations (Gontia et al., 2008; Mangus et al., 2016). Our studies also found that the values of the intercept, slope, and R² of the NWSBs were different in the two seasons because of the obvious growth stage and seasonal changes in climate. Yearly differences in NWSBs have also been reported for maize in Colorado (Zhang et al., 2023). Such differences can be attributed to different climatic conditions, water absorption potential, and transpiration rates between seasons (Ekinzog et al., 2022; Zhang et al., 2021). Taghvaeian et al. (2014) indicated that NWSB for CWSI calculations can be utilized in regions with similar climates. Zhang et al. (2023) also indicated that the same NWSB could be successfully used in different maize growing seasons owing to insignificant seasonal changes in maize transpiration rates and climatic conditions.

Ideally, the CWSI varies between 0 and 1, where 0 represents a potentially transpiring, well-watered condition and 1 represents a non-transpiring, water-stressed condition (Kumar et al., 2021). Many studies have indicated that the CWSI-E fluctuated significantly under low VPD conditions (Stockle and Dugas, 1992). In the present study, the CWSI values based on empirical baselines always exceeded the range of 0–1 when the VPD was < 2000 Pa. Additionally, the variations and fluctuations in the CWSI-E were much greater than those in the CWSI-T. Zhang et al. (2023) indicated that a VPD > 1500 Pa was more suitable for determining the CWSI for crop water stress estimation. Bellvert et al. (2016) indicated that a VPD < 2300 Pa negatively influenced the remote estimation of leaf water potential derived from CWSI. Studies have shown that the CWSI-E was a good indicator for monitoring water stress in winter wheat (Alderfasi and Nielsen, 2001). This is because the midday VPD in Alderfasi and Nielsen’s study was often > 2000 Pa (2000–4000 Pa). In the present study, the measured midday VPD was often in the 1000–3000 Pa range during the winter wheat-growing season. Yuan et al. (2004) found that the empirical CWSI may not be appropriate for detecting the water status of winter wheat because of the lower VPD in the NCP.

CWSI has been shown to exhibit good relationships with plant water stress indicators such as stomatal conductance, transpiration rate, and leaf water potential (Egea et al., 2017). Negative relationships were found between the CWSI and photosynthetic and transpiration rates in the present study. Compared with CWSI-E, the relationships between CWSI-T, photosynthetic rate, and transpiration rate were more robust, usually with higher R². Previous studies have also indicated that the CWSI based on theoretical definition is better than the empirical CWSI for monitoring the water stress of winter wheat in the NCP (Yuan et al., 2004). Previous studies have indicated that water stress increased with decreased soil water content; therefore, the CWSI also increased (King and Shellie, 2023). In this study, negative relationships were found between FASW and CWSI. Higher correlation coefficients were observed between the FASW and CWSI-T than that between the FASW and CWSI-E. This further indicates that CWSI is a promising indicator of water availability in the root zone (Osroosh et al., 2016). When the data collection time was considered, the R² between CWSI-T and FASW at 14:00 was higher than that at other times, further indicating that 14:00 was the optimal measurement time of T_c for estimating wheat water stress. King et al. (2021) indicated that irrigation scheduling based on the daily CWSI circumvents the uncertainty associated with estimating the soil water-holding capacity, critical soil water fraction, and effective crop rooting depth.

The CWSI has proven to be an effective indicator for water stress monitoring and irrigation management (Ekinzog et al., 2022). The CWSI started to increase until it peaked at the end of the study period, and this increase was more influenced by the deterioration of stomata as the wheat entered maturity and senescence rather than by the closure of stomata due to water stress (Taghvaeian et al., 2014). Therefore, the CWSI in the late-growing stage of winter wheat is not a good indicator of plant water stress. Lacerda et al. (2022) found that drought stress led to an increase in CWSI. Our study also found that CWSI-E and CWSI-T were the largest for I1 and the smallest for I3 in both seasons. These results were consistent with those reported by Ekinzog et al. (2022). Çolak and Yazar (2017) suggested that irrigation should be provided when the CWSI value reaches 0.2 to alleviate crop stress and ensure seed yields. Candogan et al. (2013) also indicated that a CWSI (calculated by Idso’s method) value of 0.22 could be taken as a threshold value to determine the irrigation time of soybeans (Glycine max) grown under a sub-humid climate of Bursa, Türkiye. In this study, I2 had a higher grain yield during both seasons. Considering the water shortage in the NCP, allowing CWSI-E and CWSI-T to reach the limits of 0.23 and 0.25–0.26 could effectively conserve groundwater resources with a 1.5%–2.7% yield reduction.

4.   Conclusion

The results showed that both CWSI-E and CWSI-T could be used to monitor wheat water stress with significant negative relationships with transpiration rate (R² of 0.3646–0.5725 and 0.5407–0.7213, respectively). Significant negative relationships were also found for both methods with fraction of available soil water at different time, with the highest R² values of 0.5335–0.5853 and 0.6251–0.7005, respectively. However, the empirical CWSI fluctuated significantly at the daily and seasonal levels. The optimal time for thermal image acquisition to eliminate the effects of environmental factors on the performance of CWSI was determined to be approximately midday, and the VPD should be > 2000 Pa. In addition, the lens of the thermometer facing south obtained a higher T_c than those facing east or north, particularly at lower heights. Therefore, T_c measurement using a thermometer lens facing south during midday at low height should be considered when using the CWSI to evaluate crop water stress. According to the results, the seasonal average threshold values at 14:00 of CWSI-E was 0.23, and that of CWSI-T ranged from 0.25 to 0.26, which could be used as criteria for irrigation scheduling of winter wheat. However, the empirical CWSI values exhibited large fluctuations and frequently exceeded the range of 0.1 to 1.0. Thus, the CWSI based on Jackson’s definition is more practical for evaluating winter wheat water stress in the NCP.