Determining water status of walnut orchards using the crop water stress index and canopy temperature measurements

Abstract

Background

Accurately evaluating the water status of walnuts in different growth stages is fundamental to implementing deficit irrigation strategies and improving the yield of walnuts. The crop water stress index (CWSI) based on the canopy temperature is one of the most commonly used tools for current research on plant water monitoring. However, the suitability and effectiveness of using the CWSI as an indicator of the walnut water status under field conditions are still unclear. This paper focuses on walnut orchards in Northwest China using synchronous monitoring of the canopy temperature, meteorological parameters, and water physiological parameters of walnut trees under both full irrigation and deficit irrigation treatments. The aim is to test the effectiveness of the simplified crop water stress index (CWSI_s) and the theoretical crop water stress index (CWSI_t) in tracking the diurnal and daily variations of the water conditions in walnut orchards.

Results

The CWSI_s can reflect the diurnal and daily changes in the water status of walnut orchards. It was found that the CWSI_s at 12:00 local time had the best performance in tracking the daily changes in the water status. Compared to the daily averaged CWSI calculated using the measured transpiration (CWSI_{Tr_day}), the correlation coefficient, index of agreement, and root mean squared error between the CWSI_s and CWSI_{Tr_day} were 0.82, 0.94, and 0.11, respectively. However, due to the calculation errors of the aerodynamic resistance in walnut trees, the CWSI_t was unable to track the diurnal variations in the water status in walnut orchards and the degree of water stress was underestimated. In addition, the variations in minimum canopy resistance in the various growth stages of walnut orchards may also affect the accuracy of the CWSI_t in terms of indicating the seasonal changes in the water status.

Conclusions

The CWSI_s provides a non-destructive, quickly and effective method for monitoring the water status of walnuts. However, the results of this study suggest that the effects of aerodynamic resistance parameterization and variations in minimum canopy resistance in the various growth stages of walnut orchards in the CWSI_t calculation should be noted.

Introduction

Walnut trees (Juglans regia L.) are an important economic and woody oil tree species and are widely planted around the world. The walnut production and planting area in China are the largest in the world and have increased year by year [1]. As a result, the problem of shortages of water for the irrigation of walnut orchards is becoming increasingly severe, especially in the arid areas in northwest China where about 4 × 10⁵ ha of walnut trees are planted. Severe drought may even induce leaf scorch in walnut trees. Generally, regulated deficit irrigation is an effective approach for improving the crop water use efficiency and achieving sustainable agricultural development. It also has a positive impact on the crop yield and quality [2, 3]. To avoid excessive drought in the various plant growth periods, it is important to conduct real-time monitoring of the plant water status during the implementation of a deficit irrigation strategy [4]. Therefore, accurately evaluating the water status of walnut orchards in various growth stages is crucial [5, 6].

Commonly used methods for monitoring orchard moisture include the soil water content (SWC), stem/leaf water potential, and stomatal conductance (g_s) measurement methods [7]. However, the orchard water status is determined by both the SWC and atmospheric evaporation demand, so using only the SWC to indicate the plant water status may produce errors [8, 9]. Moreover, the indexes such as the stem/leaf water potential and g_s rely on professional equipment for small-scale measurements, which has disadvantages such as a poor spatial representativeness, time-consuming, and labor-intensive, making it impossible to apply at the regional scale.

Thermal infrared remote sensing, as a real-time, rapid, and non-destructive monitoring method, has been applied to the water status assessment of many plants [10, 11]. The principle of thermal infrared remote sensing technology is based on the energy balance of the canopy. For example, a water deficit leads to closure of plant stomata and limits canopy transpiration. Thus, the attenuation of the evaporative cooling process is enhanced and the canopy temperature (T_c) is increased [12, 13]. This technology can be applied at different monitoring scales such as the leaf, canopy, and regional scales [6, 14, 15].

The relationships between T_c and the water physiological parameters of plants are not stable [16, 17]. T_c is not only dependent on the plant water conditions but is also strongly influenced by environmental conditions such as the solar radiation and wind speed [7]. In order to reduce the effects of environmental factors on the determination of T_c, Idso et al. [18] proposed an empirical formula for the CWSI based on the empirical negative correlation between the canopy-air temperature difference (dT) and vapor pressure deficit (VPD) under well-watered conditions (Non-water stressed baseline, NWSB). However, there is significant uncertainty in the determination of the NWSB due to lack of consideration of factors such as the wind speed and radiation [7, 10]. Subsequently, Jackson et al. [19] improved the empirical formula based on the canopy energy balance and the Penman–Monteith equation, and proposed the CWSI_t for calculating the CWSI. This method more comprehensively considers environmental factors and has been widely recognized [7, 20,21,22]. Later, Jones [23] simplified the calculation process and proposed the index of CWSI_s, which uses the dry and wet reference surface temperatures [24].

Few studies have been conducted on real-time monitoring of the water status of walnut orchards using thermal infrared technology. Dhillon et al. [25] developed a leaf monitor, which consists of a solar radiation diffuser dome and a wind barrier to ensure consistent light conditions and the reduced wind speed around the leaf. They demonstrated that the leaf temperature can be used to monitor the water status of walnut orchards under controlled environmental conditions. However, the applicability and performance of CWSI indicators to monitoring the water status of walnut orchards under field conditions is unknown. In this study, the canopy temperature and physiological and meteorological parameters of walnut trees under different irrigation conditions were measured. The objectives of this study were (1) to evaluate the applicability of the CWSI_t and CWSI_s methods to indicate diurnal changes in the water status of walnut orchards under field conditions; and (2) to analyze the performances of the CWSI_t and CWSI_s as indicators of seasonal changes in the water status of walnut orchards.

Materials and methods

Site description and irrigation treatments

This study was conducted in a walnut orchard (41°37’N, 80°37’E) in Aksu, Xinjiang, China (Fig. 1). The walnut orchard is located in a northern oasis, approximately 150 km from the Taklamakan Desert. It covers an area of approximately 120 ha, and is planted with the “Wen 185” walnut variety (rootstock variety: “zha343”). The “Wen 185” variety, an excellent native variety, accounts for half of the walnut planting area in Xinjiang. The “Zha343” rootstock is a robust local variety known for its excellent resistance to adverse external conditions and is widely used as the rootstock in the study area. In the studied walnut orchard, the tree spacing is 3 m × 4 m. The tree age was 15 years in 2023, and the average tree height and basal diameter of the walnut trees were 5.98 m and 13.64 cm, respectively. The study area has a typical temperate continental climate, with an average annual temperature of 10.10 °C and an average annual precipitation of only 65.40 mm. All of the walnut trees were planted in a flat plain, and the soil texture in the experimental site was sandy loam. In March 2023, 373.50 kg/ha of N, 373.50 kg/ha of P₂O₅, and 373.50 kg/ha of K₂O were applied once in the orchard, and no further fertilization was conducted in the growing season. The experiment was conducted from June to September 2023, and it was divided into two periods: the nutshell hardening period (Period 1, day of the year (DOY) 158–188) and the oil filling period (Period 2, DOY 189–238).

Fig. 1

Location of experimental site

Two irrigation treatments, i.e., regulated deficit irrigation (RDI) and full irrigation (FI) treatments, were studied and each treatment covered approximately 0.40 ha. In the FI treatment, irrigation was conducted once every three days and the amount of irrigation was calculated as the difference between the tree evapotranspiration and the effective precipitation. The evapotranspiration was determined using the crop coefficient method [26, 27]. In the RDI treatment, irrigation was only conducted when the leaves exhibited slight wilting, so the water status of the trees was changed from sufficient to drought stress through continued evapotranspiration. Throughout the entire experimental period, flood irrigation was performed twice in the RDI treatment (on DOY 171 and DOY 213).

Measurements

Canopy temperature

In the RDI treatment, thermal infrared images of 11 walnut trees were automatically obtained using a thermal infrared camera (FLIR A308, FLIR Inc., Wilsonville, OR, USA). The camera measured the long-wave radiation and converted it to the temperature. The accuracy of the temperature measurement was ± 2% of the reading and the resolution was 320 × 240. In this study, the thermal infrared camera was installed on a 10-m-high tower and the lens faced southeast. The zenith angle of the sensor was 45°. During the measurements, a thermal infrared image was collected every 10 min.

In the RDI treatment, three walnut trees were selected and 10 upper leaves per tree were used to determine the upper and lower limits of the canopy-air temperature difference measurements. For each tree, five leaves were coated with petroleum jelly for use as a dry reference surface, while the other five leaves were sprayed with water for use as a wet reference surface. Then, a handheld thermal infrared camera (Therma CAM S65, FLIR Inc., USA) was used to measure the dry and wet reference surfaces. Measurements were performed every hour on sunny days. For the dry reference surfaces, the measurements were performed after applying petroleum jelly to both sides of the leaf for 10 min. For the wet reference surfaces, the measurements were performed after spraying water on both sides of the leaf for 10 s. In addition, the temperatures of the handheld and tower infrared cameras were calibrated using a thermocouple method. In this study, the temperature threshold method was used for infrared image processing and canopy temperature determination [28]. This includes removing pixels corresponding to tree gaps in the images and determining the temperatures of pure canopy pixels [28].

Meteorological data, plant physiological data, and environment measurements

The meteorological observation system consisted of an anemometer (WindSonic, Gill Inc., Lymington, UK), a temperature and humidity sensor (HMP155, Vaisala Inc., Vantaa, Finland), a net radiation meter (CNR4, Kipp and Zonen Inc., Delft, Netherlands), and a rainfall gauge (TE525MM, Campbell Scientific Inc., Logan, UT, USA). Except that the net radiation meter was installed at a height of 2 m above the canopy, all of the other sensors were installed at the same height as the canopy. All of the sensors were connected to an RR-1008 datalogger (RainRoot Inc., Beijing, China) and output data every 5 min.

In the RDI treatment, soil moisture probes (HydraProbe, Stevens Inc., USA) were installed at depths of 10, 30, 50, and 70 cm. Moreover, a canopy analyzer (LAI-2200 C, Li Cor Inc., USA) was employed to measure the leaf area index (LAI) at an interval of 15 days in the RDI treatment. The LAI was determined by averaging the measurements for the five subplots.

In addition, three walnut trees of similar size in each treatment were measured using a photosynthetic instrument (LI-6800, Li-Cor Inc., Lincoln, NE, USA) in both the FI and RDI treatments. For each tree, the stomatal conductance (g_s), transpiration rate (T_r), and net CO₂ assimilation rate (A_n) were measured for four healthy and sunny leaves every 2 h on sunny days.

Calculations

Crop water stress index determined from measured transpiration rate

According to the method of Gonzalez-Dugo et al. [29], the CWSI determined from the transpiration rate (CWSI_Tr) was calculated as follows:

$$\:{\text{C}\text{W}\text{S}\text{I}}_{\text{T}\text{r}}=1-\frac{{T}_{r}}{{T}_{{r}_{pot}}},$$

(1)

where CWSI_Tr is the actual CWSI from the measured transpiration rate, T_r is the actual transpiration rate measured in the RDI treatment (mmol m^− 2 s^− 1), and T_{r_pot} is the potential transpiration rate measured in the FI treatment (mmol m^− 2 s^− 1). In this study, we used the CWSI_Tr to represent the actual water status of the walnut orchard in order to evaluate the applicability and accuracy of the CWSI_t and CWSI_s.

CWSI determined from dT

The CWSI determined from the dT is [18, 19]:

$$\:\text{C}\text{W}\text{S}\text{I}=\frac{dT-{dT}_{ll}}{{dT}_{ul}-{dT}_{ll}},$$

(2)

where dT_ll is the lower limit of dT, which represents the non-water stressed (well-irrigated) condition (°C), and dT_ul is the upper limit of dT when the leaf is non-transpiring (°C). Theoretically, the CWSI value ranges from zero to one, representing the plant water status under well-irrigated to non-transpiring conditions. According to the different methods of calculating dT_ll and dT_ul, the CWSI can be further divided into the CWSI_t and CWSI_s, which are described below.

CWSI_t

The CWSI_t was introduced by Jackson et al. [19]. The dT_ll and dT_ul were determined by combining the energy balance equation and the Penman–Monteith equation:

$$\:{dT}_{ll}=\frac{{r}_{a}\times\:{R}_{n}}{\rho\:{c}_{p}}\cdot\:\frac{\gamma\:\left(1+\frac{{r}_{cp}}{{r}_{a}}\right)}{\varDelta\:+\gamma\:\left(1+\frac{{r}_{cp}}{{r}_{a}}\right)}-\frac{VPD}{\varDelta\:+\gamma\:\left(1+\frac{{r}_{cp}}{{r}_{a}}\right)},$$

(3)

$$\:{dT}_{ul}=\frac{{r}_{a}{R}_{n}}{\rho\:{c}_{p}},$$

(4)

where r_a is the aerodynamic resistance (s m^− 1), R_n is the net radiation (W m^− 2), c_p is the heat capacity of the air (1013 J kg^− 1 °C^− 1), ρ is the density of the air (kg m^− 3), γ is the psychrometric constant (kPa °C^− 1), and Δ is the slope of the saturation vapor pressure‒temperature relationship (kPa °C^− 1). r_cp is the canopy resistance under the potential transpiration conditions (s m^− 1) and is calculated as follows:

$$\:{r}_{cp}=\frac{{r}_{min}}{{LAI}_{e}},$$

(5)

$$\:{LAI}_{e}=\left\{\begin{array}{c}LAI\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:(LAI\le\:2)\\\:2+\frac{LAI-2}{3}\:\:\:\:\:(LAI>2)\end{array}\right.,$$

(6)

where $\:{r}_{min}$ is the minimum leaf stomatal resistance (s m^− 1); LAI_e is the effective leaf area index [30].

The r_a was calculated using the semi-empirical formula proposed by Thom and Oliver [31], which was also recommended by Jackson et al. [19]:

$$\:{r}_{a}=\frac{4.72{\left\{ln\left[\frac{\left(z-d\right)}{{z}_{0}}\right]\right\}}^{2}}{1+0.54u},$$

(7)

where $\:z$ is the reference height (m), $\:d$ is the zero-plane displacement height (m), $\:{z}_{0}$ is the roughness length (m), and $\:u$ is the wind speed (m s^− 1).

CWSI_s

The CWSI_s was introduced by Jones [23]. In CWSI_s method, the leaves fully wetted with water spray on both sides were used as the natural wet reference surfaces, and the leaves coated with petroleum jelly on both sides were used as the natural dry reference surfaces. The difference between the reference surface temperature and air temperature for the wet and dry reference surfaces represented the lower and upper limits of dT, respectively. Then, the CWSI_s was calculated using Eq. (2).

Clearness index

The clearness index (CI) was used to distinguish between the cloudy and sunny days:

$$\:CI=\frac{{S}_{r}}{{S}_{e}},$$

(8)

$${S_e} = {S_{sc}} \times [1 + 0.033 \times \cos ({360_{td}}/365)] \times \sin \varepsilon $$

(9)

$$\sin {\rm{\varepsilon = sin\varphi }} \times \sin {\rm{\delta + cos\varphi }} \times \cos {\rm{\delta }} \times \cos {\rm{\omega }},$$

(10)

where $\:{S}_{r}$ is the total radiation above the canopy (W m^− 2), $\:{S}_{e}$ is the astronomical radiation (W m^− 2), $\:{S}_{sc}$ is the solar constant (1367 W m^− 2), $\:{t}_{d}$ is the Julian day, $\:\epsilon\:$ is the solar elevation angle, $\:\varphi\:$ is the local latitude, $\:\delta\:$ is the declination of the sun, and $\:\omega\:$ is the time angle. When the daily mean CI is larger than 0.50, the day was defined as sunny; otherwise, it was defined as cloudy [32].

Accuracy evaluation

The accuracy of the CWSI was evaluated by root mean square error (RMSE), coefficient of determination (R²), and index of agreement (IA):

$$\:RMSE=\sqrt{\frac{1}{n}\sum\:_{i=1}^{n}{({O}_{i}-{P}_{i})}^{2}},$$

(11)

$$\:\:{R}^{2}={\left[\frac{{\sum\:}_{i=1}^{n}({O}_{i}-\stackrel{-}{O})({P}_{i}-\stackrel{-}{P})}{\sqrt{{\sum\:}_{i=1}^{n}{({O}_{i}-\stackrel{-}{O})}^{2}}\sqrt{{\sum\:}_{i=1}^{n}{({P}_{i}-\stackrel{-}{P})}^{2}}}\right]}^{2},$$

(12)

$$\:IA=1-\frac{\sum\:_{i=1}^{n}{({P}_{i}-{O}_{i})}^{2}}{{\sum\:}_{i=1}^{n}{(\left|\left({P}_{i}-\stackrel{-}{O}\right)\right|+\left|\left({O}_{i}-\stackrel{-}{O}\right)\right|)}^{2}},$$

(13)

where P_i is the CWSI_Tr or physiological variables, O_i is the CWSI, and $\:\stackrel{-}{P}$ and $\:\stackrel{-}{O}$ are the respective mean values. When R² and IA are close to 1 and the RMSE is close to 0, the CWSI is more accurate.

Results

Temporal changes in environmental factors and physiological variables

Figure 2 shows the seasonal changes in the environmental parameters in the walnut orchard during the experimental period. The trends of the air temperature (T_a) and VPD dynamics were relatively consistent. T_a and VPD reached their first peaks in mid-June, with values of 32.68 °C and 3.86 kPa, respectively, and reached their second peaks in mid-July, with values of 33.98 °C and 3.76 kPa. The relative humidity (RH) and T_a exhibited opposite dynamics during the measurement period. The average RH value was 35.76%, which indicates that the environment in the study region was highly dry. The net radiation (R_n) fluctuated greatly due to the weather conditions, and the minimum and maximum values were 86.59 W m^− 2 and 470.64 W m^− 2, respectively. During the measurement period, the rainfall was only 33.40 mm, and the SWC increased after irrigation and precipitation, with values fluctuating between 0.11 m³ m^− 3 and 0.42 m³ m^− 3. The wind speed (W_s) fluctuated around 2 m s^− 1, with a maximum value of 2.42 m s^− 1.

Fig. 2

Seasonal variations in (a) air temperature (T_a) and relative humidity (RH), (b) vapor pressure deficit (VPD) and net radiation (R_n), (c) soil water content (SWC) and precipitation (P), and (d) wind speed (W_s) in the walnut orchard

Physiological parameters such as T_r, A_n, and g_s can accurately reflect the degree of drought stress of walnut trees. For example, in the oil filling period, the T_r, A_n, and g_s of the walnut trees in the FI treatment fluctuated around 10.40 mmol m^− 2 s^− 1, 14.60 µmol m^− 2 s^− 1, and 0.28 mol m^− 2 s^− 1, respectively (Fig. 3). In contrast to the FI treatment, the T_r, A_n, and g_s of the walnut trees in the RDI treatment were substantially lower when the drought stress increased. For example, on the day before re-irrigation (DOY 213), in the RDI treatment, the walnut trees were most severely affected by the drought stress, and the T_r, A_n, and g_s were only 9%, 7%, and 6% those in the FI treatment. After re-irrigation, all of the physiological indicators of the walnut trees gradually increased and returned to the same levels as in the FI treatment after 14 days (Fig. 3).

Fig. 3

Seasonal changes in (a) transpiration rate (T_r), (b) net CO₂ assimilation (A_n), and (c) stomatal conductance (g_s) of walnut trees in the two irrigation treatments during the oil filling period

Figure 4 shows typical diurnal variations in the physiological parameters of the walnut trees under the two irrigation treatments (DOY 207). In the FI treatment, the T_r, A_n, and g_s of the walnut trees exhibited a single peak at 12:00 and maximum values of 10.45 mmol^− 2 s^− 1, 15.69 µmol m^− 2 s^− 1, and 0.31 mol m^− 2 s^− 1, respectively. In the RDI treatment, the peaks of the T_r, A_n, and g_s of the walnut trees occurred at 10:00, and the physiological activity levels of the trees were relatively low.

Fig. 4

Temporal changes in (a) transpiration rate (T_r), (b) net CO₂ assimilation (A_n), and (c) stomatal conductance (g_s) of walnut trees in the two irrigation treatments

Diurnal variations in CWSI

In this study, we evaluated the ability of the CWSI to represent the diurnal variations in the water status in walnut orchards. According to the theory, the calculated CWSI value will fluctuate around zero under well-watered conditions, while the CWSI of water stressed trees will increase as the atmospheric evaporation demand increases [33]. Therefore, the diurnal variations in the CWSI of water stressed trees will be similar to the transpiration curve of the trees under well-watered conditions, resulting in an inverted U-shaped curve. In the FI treatment, we found that both the CWSI_Tr and CWSI_s remained relatively constant and the temporal fluctuations were small, CWSI_Tr fluctuating around 0.2 and CWSI_s fluctuated around 0.25 (Fig. 5a). In the RDI treatment under water stress, the CWSI_Tr and CWSI_s initially increased and then decreased, and their maximum values were 0.91 and 0.96, respectively (Fig. 5b). The similar temporal changes in the CWSI_s and CWSI_Tr indicate that the CWSI_s is capable of describing the changes in the water status of walnut trees. In contrast, the CWSI_t continuously increased from 8:00 to 18:00, regardless of water abundance or drought (Fig. 5a, b). In the FI treatment, for example, the daily amplitude of the CWSI_t was about 0.50. Similar results were also observed in the RDI treatment, in which the daily amplitude was 0.47. This result indicates that the CWSI_t is unable to describe the temporal changes in the water status in walnut orchards.

Fig. 5

Temporal changes in crop water stress index (CWSI) under (a) well-watered conditions and (b) water stressed conditions. Note: CWSI_s, simplified crop water stress index; CWSI_t, theoretical crop water stress index; CWSI_Tr, calculated crop water stress index using the measured transpiration rate

To clarify the reason why the CWSI_t cannot describe the temporal changes in the water status in walnut orchards, we replaced the upper and lower limits of dT in the CWSI_t with the corresponding upper and lower limits in the CWSI_s. Figure 6 shows that when the lower limit of dT in the CWSI_s is combined with the upper limit of dT in the CWSI_t, the new calculated CWSI (CWSI_h1) still increased gradually during 8:00–18:00. However, the diurnal amplitudes of the CWSI_h1 in the FI and RDI treatments were 0.37 and 0.28, respectively, which were lower than those of the CWSI_t (Fig. 6).

Fig. 6

Temporal changes in CWSI_h1 and CWSI_h2 during 8:00–18:00 under (a) well-watered conditions and (b) water stressed conditions. The CWSI_h1 is the CWSI calculated using the lower limit of the canopy-air temperature difference of the CWSI_s and the upper limit of the canopy-air temperature difference of the CWSI_t. The CWSI_h2 is the CWSI calculated using the upper limit of canopy-air temperature difference of the CWSI_s and the lower limit of canopy-air temperature difference of the CWSI_t

In addition, the upper limit of dT in the CWSI_s and the lower limit of dT in the CWSI_t were used to calculate another new CWSI (CWSI_h2). The result revealed that the CWSI_h2 remained nearly constant under well-watered conditions, but it exhibited an inverted U-shape under water stressed conditions (Fig. 6).

Seasonal variations in CWSI

The daily averaged CWSI_Tr (CWSI_{Tr_day}) between 8:00 and 18:00 on sunny days was used as the actual water status of the walnut orchard [7, 34]. Table 1 shows the relationships between the CWSI_s and CWSI_{Tr_day} and between the CWSI_t and CWSI_{Tr_day} at various times. Two linear relationships between the CWSI_s and CWSI_{Tr_day} and between the CWSI_t and CWSI_{Tr_day} were observed at different times. The R² value initially increased and then decreased. Except 8:00 and 10:00, the linear slope of the relationship between the CWSI_s and CWSI_{Tr_day} at the other times was close to 1 and the intercept was close to 0. The maximum R² and IA and the minimum RMSE of the relationship between the CWSI_s and CWSI_{Tr_day} occurred at 12:00, with the values of 0.82, 0.94, and 0.11, respectively Therefore, the CWSI_s method can describe the daily changes of water status in walnut orchards and 12:00 is the optimal time for measurement. In contrast, the R² and IA values of the linear relationship between the CWSI_t and CWSI_{Tr_day} peaked at 11:00 and 12:00, respectively. The minimum RMSE value occurred at 13:00 and 14:00. In addition, our results show that the slope of the relationship between the CWSI_t and CWSI_{Tr_day} was generally larger than that between the CWSI_s and CWSI_{Tr_day}, and the intercept of the relationship between the CWSI_t and CWSI_{Tr_day} was generally smaller than that between the CWSI_s and CWSI_{Tr_day}. This indicates that the CWSI_t has a relatively large error when describing the water status of walnut trees, so the degree of water stress will be overestimated or underestimated.

Table 1 Relationships between CWSI_s and CWSI_{Tr_day} and between CWSI_t and CWSI_{Tr_day} at various times. Note: CWSI_s, simplified crop water stress index; CWSI_t, theoretical crop water stress index; CWSI_{Tr_day}, the daily averaged crop water stress index calculation using the measured transpiration rate

Figure 7 shows the daily variations in the CWSI_s, CWSI_t, and CWSI_Tr at 12:00. The calculated CWSI_s and CWSI_t dynamics were consistent with the actual water status, which increased during the stress period and decreased after irrigation. The calculated values and temporal changes were similar for the CWSI_s and CWSI_Tr, which suggests that the CWSI_s can satisfactory describe the daily changes in the water status in walnut orchards (Fig. 7a). However, the calculated CWSI_t was significantly lower than the measured CWSI_Tr during the oil filling period (period 2) (Fig. 7b). When the upper and lower limits of dT in the CWSI_t were replaced by the corresponding upper and lower limits of dT for the CWSI_s, there was still a difference between the CWSI_h1 and CWSI_Tr during period 2 (Fig. 8a). Alternatively, the calculated CWSI_h2 was close to the CWSI_Tr during period 2, while it was substantially larger than the CWSI_Tr in period 1 (Fig. 8b). Therefore, the CWSI_s can satisfactory describe the daily changes in the water status in walnut orchards, while the CWSI_t has a relatively large error in quantification of the actual degree of water stress in walnut orchards during the various growth periods.

Fig. 7

Temporal changes in CWSI_s, CWSI_t, and CWSI_Tr calculated at 12:00 on different days. Note: CWSI_s, simplified crop water stress index; CWSI_t, theoretical crop water stress index; CWSI_Tr, calculated crop water stress index using the measured transpiration rate

Fig. 8

Temporal changes in CWSI_h1 and CWSI_h2 obtained by replacing the upper or lower limits of the canopy-air temperature difference. The CWSI_h1 is the CWSI calculated using the lower limit of the canopy-air temperature difference of the CWSI_s and the upper limit of the canopy-air temperature difference of the CWSI_t. The CWSI_h2 is the CWSI calculated using the upper limit of canopy-air temperature difference of the CWSI_s and the lower limit of canopy-air temperature difference of the CWSI_t

Figure 9 shows the relationships among the CWSI_s, CWSI_Tr, SWC, and plant physiological parameters. It can be seen that the CWSI_s had the best correlation with the CWSI_Tr, with R², RMSE, and IA values of 0.79, 0.14, and 0.92, respectively. The next best relationships were the CWSI_s-A_n and CWSI_s-g_s relationships, with R² values of 0.67 and 0.66, respectively. However, the correlation between the CWSI_s and SWC was poor, with an R² value of 0.36. The R² values of SWC-A_n and SWC-g_s relationships were 0.29 and 0.31, respectively. This indicates that the CWSI_s can more accurately and directly reflect the water status of walnut orchards than the SWC index.

Fig. 9

Relationships between CWSI_s and CWSI_Tr, SWC, and plant physiological parameters. Note: CWSI_s, simplified crop water stress index; CWSI_Tr, calculated crop water stress index using the measured transpiration rate; SWC, soil water content; A_n, net CO₂ assimilation; g_s: stomatal conductance

Discussion

CWSI as a water stress indicator

The dT is easily affected by the environmental conditions, so the accuracy of the plant water status determined by directly using the dT is limited [7, 35]. Idso et al. [18] proposed the CWSI method to overcome this deficiency, and the measured dT was normalized using the dT under non-transpiring conditions and non-water stressed conditions. At present, the CWSI is widely regarded as a good indicator of the plant water status at different spatial scales [36,37,38,39]. However, there is no consistent conclusion regarding the performances of the different CWSI calculation methods in terms of monitoring the plant water status. In practical applications, the limitations of the various CWSI calculation methods should be noted [22].

CWSl_s method for walnut trees

The CWSI_s method uses natural or artificial wet and dry reference surface temperatures to determine the dT values under non-transpiring conditions and non-water stressed conditions, respectively. Then the measured actual dT is normalized and the CWSI_s is calculated. At present, commonly used reference surfaces include painting petroleum jelly on leaves, wetted leaves, green paper, green water-absorbent cloth, evaporimeter, and a black metal plate [24]. Maes and Steppe [10] and Prashar and Jones [40] proposed that the differences in the boundary layer conditions between the reference surface and the actual leaf surface should be considered when selecting a reference surface. Moreover, special attention should be paid to the differences in the response times of various reference surface methods, especially when the environment changes. The CWSI_s method does not require additional meteorological equipment; however, the drawback of this method is that the creation and measurement of the reference surface requires manual operation.

It is difficult to distinguish between dry and wet leaves in thermal infrared images to obtain the upper and lower limits of the dT, which makes the CWSI_s method difficult to apply at the large scale [7, 10, 24]. To solve the above question, Meron et al. [41] proposed using a wet artificial reference surface (WARS) instead of wetted leaves. The WARS is a water-absorbent cloth floating on a polystyrence foam board in a water-filled tray. This method has been applied to monitor crops and trees such as cotton [42], soybeans [43], olive trees [44], and grapes [45], and the monitoring accuracy was found to be better than that of the CWSI_t method [46]. However, in the current research on the CWSI_s method, no artificial reference surface can be used to obtain the accurate upper limit of dT at the large scale [24]. Although the upper limit of dT is a function of R_n and r_a [19], it has often been arbitrarily set as 5 °C in the CWSI_s calculation [44, 47], which may cause significant errors, especially for drought tolerant plants in humid regions [7]. In the CWSI_s method utilized in this study, the leaves temperatures were manual measured using petroleum jelly coated leaves and wetted leaves. Then the measured dT was normalized at the field scale. The results revealed that the calculated CWSI_s is able to satisfactory describe the diurnal and daily changes in the water status in walnut orchards (Figs. 5 and 7a). Our results are consistent with the conclusion of Shang et al. [48] who measured the water status of cotton using the CWSI_s method with an unmanned aerial platforms.

Another unknown factor in using the CWSI_s method is that most previous studies wet leaves about 1 min before measurement; whereas we used a time of 10 s. It is not clear whether the timing of the observation affects the accuracy of the CWSI_s results [24]. In this study, when the measurement was performed 10 s after the leaves were wetted, we found that the lower limit of the dT could be effectively determined and the calculated CWSI_s values are similar to the CWSI_Tr measurements (Figs. 5 and 7a). This may be because the upper and lower limits of dT required for the CWSI calculation are not absolute dT limits under completely non-water stressed conditions or non-transpiring conditions, and the current two limits are only used as a temperature indicator [49]. Our results prove that changing the observation time after leaf wetting from 1 min to 10 s improves the efficiency of monitoring of the plant water status when using the CWSI_s method.

CWSl_t method for walnut trees

According to the Penman–Monteith equation and canopy energy balance, the CWSI_t method calculates the dT under both non-transpiring conditions and non-water stressed conditions, and then, the measured dT is normalized. Generally, the accuracy of the CWSI_t method depends on the accuracy of the estimation of the r_a and r_cp. parameters [29, 50]. In this study, we found that within a day, the calculated CWSI_t increased gradually under both the well-watered and severe drought conditions, and it had large errors in describing the diurnal changes in the water status of the walnut orchard (Fig. 5). This result is consistent with the conclusions of Agam et al. [33] and Liu et al. [7], that is, this phenomenon may due to errors in the r_a parameterization method. This is because in the dT upper limit calculation of the CWSI_t, only the parameter r_a is estimated. However, by replaced the upper limit of dT in the CWSI_t method with the measured upper limit of dT, we found that the r_a of the semi-empirical algorithm proposed by Thom and Oliver [31] and recommended by Jackson et al. [19] still yielded a relatively large error (Fig. 6). It should be noted that Jones [12] proposed that the basic assumption of the CWSI_t may only support the monitoring of the plant water status around noon. This is consistent with the results of this study, that is, it is appropriate to use the CWSI_t data at noon to track the water status of walnut orchard. However, for the CWSI_t method, significant underestimation of the actual water stress occurred during the oil filling period (Fig. 7b). Han et al. [50] illustrated that the error in the CWSI_t method may be related to the influence of the crop height on r_a. In this study, there was no significant change in the height of the orchard’s canopy throughout the entire growth period due to tree pruning. By replaced the upper limit of dT, we found that the phenomenon of underestimation of the CWSI_t was caused by the errors in r_a calculation methods (Fig. 8a). This suggests that the method of calculating r_a has a significant impact on the determination of the CWSI_t, and an accurate r_a parameterization method needs to be used in the various plant growth periods [33].

In this study, we found that the performance of the CWSI_t in describing the water status of walnut orchards can be improved by replacing the lower limit of dT with the measured lower limit of dT. Figure 6a shows that, although the diurnal amplitude of the CWSI_h1 decreased compared to that of the CWSI_t, the diurnal variation pattern of the CWSI_h1 did not change (Fig. 6). This result suggests that, although r_a is used in the calculation of the lower limit of dT, its impact on the estimation of the lower limit of dT is relatively small and will not significantly change the diurnal variation pattern of the CWSI_t. In contrast, when the upper limit of dT in the CWSI_t was replaced with the measured upper limit of dT, the results shows the calculated CWSI_h2 described the daily changes in the water status of the walnut orchard well during the oil filling period (Fig. 8b). However, the CWSI_h2 method overestimated the actual water stress during the nutshell hardening period (Fig. 8b). The reason for this phenomenon may be due to the error of the r_cp estimation. Previous studies have shown that the potential transpiration rate of walnut trees in the growth season exhibits the following order: nutshell hardening period < oil filling period [51]. The r_cp in growth season exhibits an opposite pattern: nutshell hardening period > oil filling period. Therefore, if the r_cp in the oil filling period is used for the nutshell hardening period, the r_cp will be underestimated and the lower limit of dT and the CWSI will be overestimated. At present, many studies have used a constant r_cp value or ignored it during the calculation. For example, Zhang et al. [39] used an r_cp value of zero in different growth periods and found that the CWSI_t accurately indicated the changes in the water status of winter wheat. Ekinzog et al. [52] also found that when r_cp is ignored, the CWSI_t method can accurately indicated changes in the potato water status in different growth stages. However, Agam et al. [33] reported that when they ignored the r_cp, their CWSI_t results fluctuated around 0.8 for olive trees even under well-watered conditions. Generally, the r_cp value of woody plants is significantly higher than those of crops and cannot be ignored or used a constant value. For example, Jackson et al. [19] proposed that the r_cp of wheat is 5 s m ^–1, while the r_cp of Chinese cork oak is 90 s m^− 1 [7]. Therefore, the results of our study suggest that when using the CWSI_t method to monitor the water status of woody trees, r_cp should not be ignored, and the variations in r_cp in the various growth stages should also be noted.

CWSI index and soil moisture index

In recent years, several studies have explored the soil moisture status using the CWSI [22, 39, 52, 53]. However, in this study, we found that the CWSI method can effecitively indicating walnut water status, not soil moisture status (Fig. 9). Soil can replenish lost water quickly when precipitation or irrigation occur, but it takes plants a long time to recovery from drought, and the length of time required depends on drought intensity, drought duration, and plant species [54, 55]. In this study, we found that it took at least 14 days for the walnut trees to recover from severe drought to normal levels after re-irrigation (Fig. 3). This result is consistent with that of Agam et al. [33]; in other words, there is a time lag in the recovery of transpiration compared to the recovery of soil moisture. This also explains we found that the correlation between the SWC and plant physiological parameters is weaker than those between the CWSI and plant physiological parameters in this study. In addition, the atmospheric evaporation demand also affects the sensitivity of the response of the CWSI to soil moisture changes. At drought events, the low soil water and high VPD interact with plants through land and air, and a single soil water indicator may cannot be used to accurately evaluate the plant stress degree [8, 9].

Conclusions

In this study, we investigated the water status of a walnut orchard using the CWSI method, the canopy temperature, and meteorological data. The results revealed that the CWSI_s can satisfactory describe the diurnal and daily changes of water status in walnut orchards. We found that the performance of the CWSI_s in indicating the water status of the walnut trees was better than that of the soil moisture index. Moreover, the CWSI_t method was unable to describe the diurnal changes in the water status of the walnut orchard. The calculated seasonal changes in the CWSI_t in the walnut orchard showed that the degree of water stress was substantially underestimated. This was related to the incorrect calculation of the aerodynamic resistance in walnut trees. Additionally, large errors in the estimation of the water status using the CWSI_t method also occurred when a constant value is used for the minimum canopy resistance. Therefore, the variations in the minimum canopy resistance in the various growth stages of walnut trees should be noted when conducting CWSI calculations. Overall, the method of monitoring canopy temperature in this study can effectively determine the water status of walnut orchard and conserve water resources, which is important for production in large-scale commercial orchards.