Study on a Stomatal Conductance Model of Grape Leaves in Extremely Arid Areas

Abstract

Stomata are essential for regulating the exchange of water and energy between plants and the atmosphere. In the context of climate warming, especially in extremely arid regions, the knowledge of stomatal conductance variation patterns is fundamental to the study of crop evapotranspiration, productivity and drought resistance characteristics. The accurate simulation of stomatal conductance in this region is an important prerequisite for the optimal regulation of the crop growth environment. In this study, a two-year field experiment was carried out in vineyards in an extremely arid region. The Jarvis model and BWB model were used to evaluate the daily changes in stomatal conductance. The results showed that stomatal conductance was significantly correlated with environmental factors (temperature difference between leaf and air (ΔT), photosynthetically active radiation and air temperature). The Jarvis and BWB models performed well. However, the response function of the environment factor in the Jarvis model can affect the model performance. The ΔT effectively improved the model, and the modified Jarvis model outperformed the modified BWB model. The R² and model slope b of the modified Jarvis model increased by 45.18–70.37% and 2.51–3.12%, respectively. RMSE and MAE decreased by 38.98–43.12% and 42.69–44.35%, respectively. Overall, the Jarvis3–ΔT model had a good effect on the simulation of the daily change of stomatal conductance during the critical period of grape growth, and the Jarvis3–ΔT model was the best stomatal conductance model in this study. The results of the study are of great significance for further improving the sustainable use of water resources in grapevines in extremely arid regions.

1. Introduction

The stomata are the main channels that control the transport of water and CO₂ between the leaves and the atmosphere [1], which has an important impact on the exchange of matter and energy, plant productivity and water use efficiency in SPAC systems [2,3,4]. Stomatal conductance is a physiological indicator that describes stomatal movement [5], which regulates the water balance between plants and the atmospheric environment [6]. Studies have shown that stomatal conductance varies with external environment factors such as air temperature, photosynthetically active radiation and saturated water vapor pressure difference [7]. Under drought conditions, plants adapt to restrictive conditions by closing their stomata to reduce water transpiration and prevent the disruption of the water balance, which results in decreased stomatal conductance. Therefore, stomatal conductance is an important index to reflect the water status in plants due to plants regulating stomatal conductance to influence physiological functions such as photosynthesis and transpiration [8]. When stomatal conductance is reduced, photosynthesis and plant growth are inhibited. This indicates that the change and sensitivity of stomatal conductance reflect the drought resistance of plants to a certain extent. Understanding the response relationship between stomatal conductance and environmental factors is of great significance for revealing plant water use mechanisms and discovering environmental optimal management. However, the response mechanism of stomata to the environment is complex, and the workload of observing stomatal conductance is large and it is difficult to obtain long-term series data. Therefore, the establishment of a stomatal conductance model is the most effective method to study stomata at present.

The commonly used stomatal conductance models include the model based on the relationship between stomatal conductance and environmental factors [9] and the model based on the relationship between stomatal conductance and photosynthesis [10]. Currently, stomatal conductance models are used for the response of specific plant stomata to environmental factors [11], forest and other ecosystem investigations [12]. However, the accuracy of different stomatal conductance models is affected by plant species, atmospheric environment and study area [13]. Studies have shown that stomatal conductance is mainly affected by environmental factors such as radiation, saturated vapor pressure difference, leaf temperature and soil moisture [14], and the Jarvis model constructed with the response function of environmental factors can better simulate the stomatal conductance. However, when plants are subjected to water stress, the response relationship between stomatal conductance and environmental factors will change, and stomatal conductance models cannot accurately simulate the changes of stomatal conductance. Therefore, scholars have improved the stomatal conductance model by adding factors such as the temperature difference between leaf and air (ΔT) and soil water content, which can reflect the moisture status [15,16]. Wang et al. [17] showed that the improved model of ΔT under water stress conditions can better simulate the relationship between stomatal conductance and environmental factors, while some studies have shown that under continuous water stress, the BWB model has the best simulation accuracy of stomatal conductance [18]. Therefore, the applicability of various stomatal conductance models needs to be determined according to conditions such as crop species and water status.

The grape industry is an important part of Xinjiang’s agricultural economic system [19]. However, the extremely arid area of the Xinjiang Turpan Basin, with its dryness, low rainfall and high temperatures, makes grapes vulnerable to water stress during the growing period. Meanwhile, the region faces problems such as water shortage and low water use efficiency, so it is of great significance to study the drought resistance mechanism of vines to ensure grape yield and improve water use efficiency. Stomatal movement is an important activity to regulate plant water transport, and it is the core link in the study of plant drought resistance mechanisms [20]. At present, the response of the stomatal conductance of grape leaves to environmental factors in the extremely arid area of Xinjiang is not clear, and there are few reports on stomatal conductance models in this area. The accurate simulation of the stomatal behavior of vines is beneficial to reveal the response mechanism of vines to the environment and improve water use efficiency in extremely arid areas.

2. Materials and Methods

2.1. Experimental Site

The experiment was carried out at the test base of the Xinjiang Uighur Autonomous Region Grapes and Melons Research Institution in Shanshan (90.30° E, 42.91° N, altitude 419 m), Xinjiang. The area has an extremely arid climate with an average annual rainfall of 25.3 mm and an average annual evaporation of 2751 mm. The soil texture of the experimental field was loamy with an average soil dry capacity of 1.53 g/cm³.

2.2. Experimental Design

The study was conducted during the 2021–2022 grape growing season, using 6-year-old Thomson seedless white vines with a plant spacing of 2 m and a row spacing of 4 m. The vines used in the study were not grafted. The experimental site was cultivated in a furrow and ridge model with a furrow width of 1.2 m. Irrigation was the main source of water during the growing period of the grapes, and the irrigation method was furrow irrigation. The irrigation time was determined according to the soil’s water content, and the soil’s water content was measured every 5 days using the TRIME-PICO-IPH soil moisture measurement system and corrected with the drying method. The planned wetting layer was 100 cm. Irrigation was carried out when the soil water content dropped to 60% of the field capacity, and the irrigation water quota was 39 mm.

Healthy grape leaves without pests and diseases and with the same maturity were selected from different directions on the grape canopy for monitoring. It was ensured that sunlight could reach the selected leaves, and they were fully expanded but not senescent. Stomatal conductance g_s (mol·m⁻²·s⁻¹), net photosynthetic rate (μmol·m⁻²·s⁻¹) and leaf temperature Tleaf (°C) were observed using a portable photosynthesis system (Li-6400, Li-cor, Lincoln, NE, USA) from 8:00 to 20:00 during the critical growth period of May–July (major stages of canopy growth and fruit growth), with a 2 h test interval. Meanwhile, the environmental factors such as air temperature Ta, relative humidity RH and photosynthetically active radiation (PAR) were recorded using automatic meteorological observation stations. The saturated water vapor pressure difference (VPD) was calculated with Equation (1) [18]. In 2021 and 2022, three days in each month (from May to July) were selected for daily change monitoring. The experimental observation data in 2021 were selected for model correction and parameter fitting, and the observation data in 2022 were selected for model testing.

$$\mathrm{VPD}=0.6108e^{\frac{17.27 \times Ta}{\mathrm{Ta} + 237.3}} \times \left(1-\mathrm{RH}\right)$$

(1)

2.3. Stomatal Conductance Model

2.3.1. Jarvis Model

The Jarvis model considers the influence of environmental factors and assumes that the influences of different environmental factors on stomatal conductance are independent of each other. The specific expression of the model is as follows:

$$g_s=f_1\left(\mathrm{PAR}\right)f_2\left(\mathrm{VPD}\right)f_3\left(\mathrm{Ta}\right)f_4\left(\mathrm{Ca}\right)$$

(2)

where g_s is the stomatal conductance (mol·m⁻²·s⁻¹), and f₁ (PAR), f₂ (VPD), f₃ (Ta) and f₄ (Ca) are the functions of photosynthetically effective radiation, saturated water vapor pressure difference, atmospheric temperature and atmospheric carbon dioxide concentration on the stomatal conductance, respectively. Its function value is in the range 0–1.

The response functions between environmental factors and stomatal conductance have been studied extensively. In this study, the response functions of stomatal conductance to photosynthetic active radiation and saturated water vapor pressure difference were calculated using Equations (3)–(6) [21].

$$f_1\left(\mathrm{PAR}\right)=\frac{\mathrm{PAR}}{{\mathrm{a}}_1+\mathrm{PAR}}$$

(3)

$$f_{21}\left(\mathrm{VPD}\right)=1-{\mathrm{b}}_1\mathrm{VPD}$$

(4)

$$f_{22}\left(\mathrm{VPD}\right)=\frac{1}{1+{\mathrm{b}}_1\mathrm{VPD}}$$

(5)

$$f_{23}\left(\mathrm{VPD}\right)=\frac{1-{\mathrm{b}}_1\mathrm{VPD}}{1+{\mathrm{b}}_2\mathrm{VPD}}$$

(6)

Studies have shown that the influence of air temperature on the stomatal conductance of leaves shows a quadratic curve [22]. The functional equation is Equation (7).

$$f_3\left(\mathrm{Ta}\right)={\mathrm{c}}_1+{\mathrm{c}}_2\mathrm{Ta}+{\mathrm{c}}_3{\mathrm{Ta}}^2$$

(7)

Generally, the response of plant stomata to carbon dioxide concentration ranges from 0 to 250 μmol·mol⁻¹, and the average carbon dioxide concentration during the growth period of grapes in the study area was about 350 μmol·mol⁻¹. This was greater than the upper limit of stomatal response, so the effect of carbon dioxide on stomatal conductance in the Jarvis model was negligible (Equation (8)).

$$f_4\left(\mathrm{Ca}\right)=1$$

(8)

2.3.2. Ball–Woodrow–Berry Model

The BWB model is a semi-empirical model based on the relationship between stomatal conductance and photosynthetic rate (Equation (9)). The model assumes that when the CO₂ concentration and atmospheric humidity are constant, the stomatal conductance has a linear relationship with the net photosynthetic rate.

$$g_s=m\frac{{\mathrm{P}}_{\mathrm{n}}{\mathrm{h}}_{\mathrm{s}}}{{\mathrm{C}}_{\mathrm{s}}}+g_0$$

(9)

where P_n is the net photosynthetic rate (μmol·m⁻²·s⁻¹); C_s is the CO₂ concentration on the leaf surface (μmol·mol⁻¹); h_s is the relative humidity of the leaf surface (%); and m and g₀ are the fitting parameters.

2.3.3. Model Improvements

Leaf water potential is an important indicator reflecting plant water status, which is closely related to soil moisture and meteorological conditions because the temperature difference between leaves and air (ΔT) is strongly related to crop water conditions, and the ΔT can reflect crop water and soil water status well. Therefore, the ΔT was introduced into the model, and the Jarvis model and the BWB model were improved. The relationship between ΔT and stomatal conductance is expressed exponentially (Equation (10)).

$$f_{\left(\Delta \mathrm{T}\right)}=Exp\left({\mathrm{d}}_1\Delta \mathrm{T}\right)$$

(10)

2.4. Model Evaluation

In this study, the coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE) and model slope (b) were used to evaluate the performance of the model. The following equation was used to calculate:

$$\begin{gathered} \mathrm{RMSE}=\sqrt{\frac{{{\displaystyle \sum }}_{i=1}^n{\left(O_i-M_i\right)}^2}{n}} \\ \mathrm{MAE}=\frac{{{\displaystyle \sum }}_{i=1}^n\left|O_i-M_i\right|}{n} \\ \mathrm{b}=\frac{{{\displaystyle \sum }}_{i=1}^n{O}_i{M}_i}{{{\displaystyle \sum }}_{i=1}^n{O}_i^2} \end{gathered}$$

where O_i represents the measured value; M_i represents the analog value; and n is the number of samples. R² was close to 1, indicating that the model simulation is better. The smaller the RMSE and MAE values, the better the simulation. Model slope b reflects whether the model is overestimated or underestimated; when b > 1 it means that the model is overestimated, when b < 1 it means that the model is underestimated, and b being closer to 1 means that the simulation effect is better.

2.5. Data Processing and Analysis

In this study, the correlation between stomatal conductance and environmental factors was analyzed with SPSS26 software. The parameters of the stomatal conductance model and the verification of the simulation effect were analyzed with 1stOpt10.0 software. Origin2022 software was used for the linear fitting of the simulation results and measured values.

3. Results

3.1. Daily Change of Environmental Factors

The daily change curves of environmental factors for the observation days in 2021 and 2022 are shown in Figure 1. The diurnal variation of air temperature, PAR and VPD showed a unimodal curve that first increased and then decreased. The air temperature peaked at 14:00–16:00, except in May 2021. The average daily air temperature (8:00–20:00, the same below) from May to July in 2021 was 25.81 °C, 32.16 °C and 35.07 °C, respectively. The average daily air temperature from May to July in 2022 was 27.94 °C, 33.12 °C and 32.85 °C, respectively. The PAR gradually increased with the advancement of the growth period. The daily average value of PAR during the daytime in 2021 was 761.32 µmol·m⁻²·s⁻¹–943.58 µmol·m⁻²·s⁻¹, and the daily average value of PAR during the daytime in 2022 was 715.01 µmol·m⁻²·s⁻¹–830.37 µmol·m⁻²·s⁻¹. The VPD basically all peaked at 16:00. The daily average value of VPD in May 2021 and 2022 was the smallest at 2.48 kPa and 3.09 kPa, respectively. The difference between June and July was small, with the largest daily average VPD at 4.11 kPa in July 2021 and 3.78 kPa in June 2022. In 2021 and 2022, the daily average air relative humidity was 21.64–31.69%, and the difference in relative humidity between different months was small. Leaf temperature was linearly related to air temperature, and leaf temperature increased with air temperature. The average daily leaf temperature from May to July in 2021 was 29.35 °C, 35.34 °C and 38.31 °C, respectively. The average daily air temperature from May to July in 2022 was 31.28 °C, 36.57 °C and 36.06 °C, respectively.

3.2. Daily Change in Stomatal Conductance

Figure 2 shows the daily change curves of stomatal conductance in May, June and July during the grapevine growth period. The daily variation curve of stomatal conductance did not vary significantly between years. The daily variation in stomatal conductance showed a unimodal curve in May and a bimodal curve in June and July. From 8:00 to 12:00, the air temperature and PAR increased rapidly, the stomatal conductance gradually increased, and the stomatal conductance reached a peak at 12:00. The rapid decrease in stomatal conductance between 12:00 and 16:00 during June and July was due to the high air temperature in June and July, which reduced the stomatal opening of the leaves to prevent excessive water loss from the leaves and the vegetative body. Stomatal conductance gradually increased during the critical period of grape growth (May–July), with the daily average value of stomatal conductance being 0.106 mol·m⁻²·s⁻¹–0.164 mol·m⁻²·s⁻¹ and 0.105 mol·m⁻²·s⁻¹–0.153 mol·m⁻²·s⁻¹ in 2021 and 2022, respectively.

3.3. Correlation between Stomatal Conductance and Environmental Factors

Table 1. Correlation between stomatal conductance and influencing factors, where ** in the table indicates significant correlation, p < 0.01.

	Ta	RH	PAR	VPD	ΔT
2021	0.363 **	0.004	0.570 **	0.258 **	−0.714 **
2022	0.243 **	0.259 **	0.645 **	0.055	−0.686 **

shows the correlation between stomatal conductance and environmental factors. Air temperature, PAR and ΔT all reached significant correlation levels (p < 0.01) with stomatal conductance in 2021 and 2022. Among them, ΔT had the highest correlation with stomatal conductance, with correlation coefficients of −0.714 (2021) and −0.686 (2022), followed by PAR and air temperature. The air in the test area was dry and the relative humidity was low. The analysis results showed that the correlation coefficient of stomatal conductance to RH and VPD was low. There was no significant correlation between RH and stomatal conductance in 2021, while there was no significant correlation between VPD and stomatal conductance in 2022.

3.4. Stomatic Conductance Model Parameter Determination and Performance Evaluation

3.4.1. Jarvis Model

Table 2. Jarvis model parameters.

Model	Parameter							p
	a1	b1	b2	c1	c2	c3	d1
Jarvis1	174.6798	0.1151		0.6136	−0.0382	0.0009		<0.01
Jarvis2	190.3139	−1.7007		−0.1021	0.0285	−0.0017		<0.01
Jarvis3	195.1951	0.0463	−1.8142	−0.6070	0.0805	−0.0031		<0.01
Jarvis1–ΔT	36.7521	0.0979		0.4347	−0.0262	0.0008	−0.1491	<0.01
Jarvis2–ΔT	67.0107	−3.1266		−1.2016	0.1469	−0.0058	−0.1378	<0.01
Jarvis3–ΔT	67.1491	0.0039	−3.2114	−1.3246	0.1595	−0.0062	−0.1376	<0.01

and

Table 3. Jarvis model evaluation indicators.

Model	R2	RMSE	MAE	b
Jarvis1	0.4712	0.0313	0.0260	0.9527
Jarvis2	0.5886	0.0276	0.0230	0.9632
Jarvis3	0.5976	0.0273	0.0227	0.9640
Jarvis1–ΔT	0.8028	0.0191	0.0149	0.9824
Jarvis2–ΔT	0.8676	0.0157	0.0128	0.9882
Jarvis3–ΔT	0.8676	0.0157	0.0128	0.9882

show the Jarvis model fitting parameters and model evaluation indicators. The Jarvis models all passed the significance test (p < 0.01), and the slopes b of the models were all greater than 0.9. The results showed that the Jarvis model could better simulate the stomatal conductance of vines in extremely arid areas. Compared with the Jarvis models of the three structures, the R² of Jarvis3 was 26.82% and 1.53% higher than that of Jarvis1 and Jarvis2, respectively, and the RMSE, MAE and b were better than those of Jarvis1 and Jarvis2. When the ΔT was introduced to improve the Jarvis model, the R² and b of the modified model (−ΔT) improved by 70.37% and 3.12% (Jarvis1), 47.40% and 2.60% (Jarvis2) and 45.18% and 2.51% (Jarvis3), respectively. The RMSE and MAE of the –ΔT model were reduced by 38.98% and 42.69% (Jarvis1), 43.12% and 44.35% (Jarvis2) and 42.49% and 43.61% (Jarvis3), respectively. In the −ΔT model, the evaluation indexes of Jarvis3–ΔT and Jarvis2–ΔT were the same because the structures of Jarvis2 and Jarvis3 were consistent when the parameter b₁ was equal to 0 in the Jarvis3 model. However, the b₁ of Jarvis3–ΔT was 0.0039, which was close to 0. Therefore, the Jarvis3–ΔT and Jarvis2–ΔT simulations were basically the same.

3.4.2. BWB Model

Table 4. BWB model parameters and evaluation indicators.

Model	Parameter			p	R2	RMSE	MAE	b
	m	g0	d1
BWB	0.0497	0.0873		<0.01	0.4259	0.0326	0.0263	0.9486
BWB–ΔT	0.0878	0.0968	−0.3009	<0.01	0.5099	0.0301	0.0244	0.9562
BWB1	2.7493	0.0422		<0.01	0.8438	0.0170	0.0140	0.9860
BWB1–ΔT	2.8595	0.0481	−0.0364	<0.01	0.8497	0.0167	0.0138	0.9866

shows the parameters and evaluation indicators of the BWB model. All models passed the significance test (p < 0.01), and the slope b of the model was greater than 0.9. Because of the low correlation between stomatal conductance and RH in extremely arid areas, the relative humidity factor was removed from the BWB model and noted as the BWB1 model. Compared with the BWB model, the R² of the BWB1 model increased from 0.4259 to 0.8438, the slope b of the model increased by 3.94% and the RMSE and MAE decreased by 47.85% and 46.77%, respectively. Overall, the evaluation indexes of the model were all improved to different degrees after the introduction of the ΔT in the model. The R² and b of the BWB–ΔT model increased by 19.72% and 0.80%, respectively. RMSE and MAE decreased by 7.67% and 7.22%, respectively. However, the evaluation index of the BWB1–ΔT model was improved to a small extent.

3.5. Model Simulation Effect Verification

3.5.1. Model Performance Analysis

The linear relationship between the measured value of the stomatal conductance and the calculated value is shown in Figure 3. The estimate of the –ΔT model was closer to the measured value, and the model performed well. The modified Jarvis model performed better than the modified BWB model (

Table 5. Performance of the stomatal conductance model and correction model.

Model	R2	RMSE
Jarvis1	0.6728	0.0231
Jarvis1–ΔT	0.8479	0.0162
Jarvis2	0.6390	0.0259
Jarvis2–ΔT	0.8846	0.0151
Jarvis3	0.6702	0.0254
Jarvis3–ΔT	0.8847	0.0151
BWB	0.4340	0.0317
BWB–ΔT	0.6119	0.0252
BWB1	0.8178	0.0185
BWB1–ΔT	0.8408	0.0171

). In the Jarvis model, the R² of the Jarvis1 model was maximum, while the R² of Jarvis2 and Jarvis3 was greater than that of Jarvis1 after the introduction of ΔT. Compared with the initial model, the performance of the modified Jarvis model was improved, and R² was improved by 26.03% (Jarvis1), 38.44% (Jarvis2) and 32.01% (Jarvis3). RMSE decreased by 29.87% (Jarvis1), 41.70% (Jarvis2) and 40.55% (Jarvis3). In the BWB model, R² increased from 0.434 to 0.8178 and RMSE decreased from 0.0317 to 0.0185 after removing the relative humidity factor in the BWB1 model. The introduction of ΔT improved the R² of the BWB model by 40.99% and reduced the RMSE by 20.50%, while the R² of the BWB1 model improved by only 2.81% and reduced the RMSE by 7.57%.

3.5.2. Simulation Analysis of Daily Changes in Stomatal Conductance

The Jarvis3–ΔT model and the BWB1–ΔT model with better model performance were selected to simulate the daily change in stomatal conductance (Figure 4). Both models overestimated stomatal conductance during the critical period of grape growth. The estimates of the Jarvis3–ΔT model were 1.89% (May), 7.77% (June) and 2.93% (July) higher than the measured values. The estimates of the BWB1–ΔT model were 5.27% (May), 5.57% (June) and 3.04% (July) higher than the measured values. Overall, the estimated value of the Jarvis3–ΔT model was closer to the measured value than that of BWB1–ΔT model, which could better estimate the daily change in the stomatal conductance of grape leaves in the extremely arid region.

4. Discussion

In recent years, the trend of high temperature and dry weather has been increasing, and the Turpan region in Xinjiang belongs to the extreme drought area. Plants are vulnerable to high temperature and drought stress during the critical period of crop growth. Stomatal sensitivity is an important indicator of crop drought resistance [23]. The accurate simulation of stomatal conductance is essential to study the photosynthesis, transpiration, and water, air and heat exchange capacity of plants within SPAC systems [24]. It has been found that stomatal conductance is affected by a variety of environmental factors, but the response relationship between environmental factors and stomatal conductance is affected by factors such as crop species and climate [25,26]. Studies have shown that stomatal conductance has a good correlation with solar radiation, while the significance of air temperature and stomatal conductance varies between different varieties [27]. In this study, the correlation between the stomatal conductance of grape leaves and ΔT, PAR and air temperature reached a significant level. Leaf temperature is affected by the air temperature and plant water condition. The ΔT can reflect the effective soil moisture and leaf water deficit. Too high or too low leaf temperatures will affect the internal tissue and enzyme activities of the leaves, and appropriate ΔT can ensure the normal physiological function of the leaves [16]. Therefore, ensuring sufficient soil effective moisture and improving leaf moisture status in extremely arid areas is conducive to keeping the stomatal conductance at the optimal level.

A lot of research has been conducted on stomatal conductance models, and the Jarvis model and BWB model are common stomatal conductance models. The structure of the Jarvis model is simple and flexible, which can better reflect the comprehensive influence of multiple environmental factors on stomatal conductance, and the meteorological parameters required by the model are easy to obtain. However, the response functions and parameters of each environmental factor in the model need to be redetermined in different environments [28]. The BWB model is a semi-empirical model which considers the relationship between photosynthesis and environmental factors and stomatal motion, and the parameters have physiological significance; g₀ is the stomatal conductance when the photosynthetic rate is equal to zero [29]. However, the BWB model uses the photosynthetic rate as the input, and it is often time-consuming and labor-intensive to obtain many photosynthetic rates in field experiments, and the BWB model cannot accurately simulate the stomatal conductance when the CO₂ concentration is low due to the complex response relationship between stomatal conductance and environmental factors and the simulation results are affected by many factors. Therefore, the performance of the model varies under different plant species and environmental conditions. Studies have shown that stomatal conductance will be affected by soil moisture when it is subjected to water stress under extreme drought conditions, thus reducing the correlation between meteorological factors and stomatal conductance, which will reduce the accuracy of the Jarvis model compared with the BWB model [18]. However, other studies have also shown that soil moisture and air temperature determine the opening and closing of stomata, which is inconsistent with the premise of the construction of the BWB model, which would result in the Jarvis model outperforming the BWB model [30]. In this study, the model performance of the Jarvis model outperformed the BWB model, because the stomatal conductance was significantly correlated with the environmental factors considered in the Jarvis model (PAR and air temperature), while the correlation between air humidity and stomatal conductance was low.

In extremely arid regions, the stomatal conductance is also regulated by soil moisture and leaf moisture in a hot and dry environment during the growing season [31]. Leaf temperature affects the photosynthetic rate, transpiration rate and respiration rate. High leaf temperature can reduce the photosynthetic rate and stomatal conductance. Meanwhile, the leaf adjusts the leaf temperature through transpiration, which in turn affects the opening and closing of stomata [32]. This indicates that the ΔT is related to environmental factors and leaf water balance. Some studies have shown that the combined effects of drought on stomatal and non-stomatal factors must be considered when simulating stomatal conductance under drought conditions [33]. Therefore, the ΔT was introduced to improve the stomatal conductance model. It has been shown that the modified ΔT model significantly improves the simulation accuracy [16]. In this study, the results indicated that the ΔT significantly improved the model performance in extremely arid areas. R² increased by 0.70–70.37%, and RMSE decreased by 1.76–43.12%. This showed that the influence of water status should be considered in the simulation of stomatal conductance in extremely arid areas. The modified Jarvis model outperformed the modified BWB and BWB1 models. However, the Jarvis model uses a nonlinear relationship between stomatal conductance and environmental factors. The model performance is affected by the response function of environmental factors and stomatal conductance. Therefore, three VPD response functions were selected for analysis. The Jarvis2–ΔT and Jarvis3–ΔT models performed well, and the model evaluation indicators were basically consistent. This is because when b₁ in f₂₃ (VPD) equals 0, the Jarvis2–ΔT and Jarvis3–ΔT models have the same structure.

The results showed that the stomatal conductance model overestimates the stomatal conductance in the critical period of grape growth in the extreme arid region, which may be because the stomatal conductance of leaves is comprehensively regulated and controlled by various environmental factors and soil moisture [34]. However, only environmental factors and ΔT are included in the model of this study. It has been found that when water stress is caused by low soil water content the stomatal conductance of leaves is reduced, and the sensitivity of stomata to water stress is different in different plants [35]. Relevant studies have shown that both Jarvis and BWB models underestimate stomatal conductance under high soil moisture conditions, and overestimate stomatal conductance under low soil moisture conditions [18]. To accurately simulate the stomatal conductance, relevant scholars introduced the water response function in the stomatal conductance model. The introduction of a moisture response function improves the performance of the Jarvis and BWB models, but the accuracy of the model decreases when the soil moisture content exceeds a certain range [36]. Since soil moisture in this study was maintained at between 60% and 90% of the field capacity, the grapevines were not affected by soil moisture stress [37]. At the same time, Li et al. [16] showed that the performance of the two-factor correction model is lower than that of the single-factor correction model. Therefore, this study only considered the ΔT that could reflect plant water deficit, and did not consider the influence of the soil moisture response function on stomatal conductance. However, a wider range of soil moisture should be addressed in the next study, with a view to improving the applicability of the stomatal conductance model.

5. Conclusions

In this study, grapes in extremely arid regions were used as the research object, and the diurnal variation in leaf stomatal conductance during the critical period of grape growth was analyzed. Under the influence of environmental factors and the water status of grape leaves, the daily change in stomatal conductance showed unimodal or bimodal curves during the growth period. ΔT, PAR and air temperature were the environmental factors that had significant effects on stomatal conductance. Among them, the correlation between ΔT and stomatal conductance was the highest. Air humidity had less of an effect on stomatal conductance. The Jarvis model and BWB model can better simulate the stomatal conductance of grape leaves, and the model was modified by ΔT, which significantly improved the performance of the model. Overall, the Jarvis model was more accurate than the BWB model in simulating the stomatal conductance of vines in extremely arid areas. By analyzing the Jarvis model and the BWB model with different structures, the optimal stomatal conductance model (Jarvis3–ΔT) of grapevines in the extremely arid region was established. The results provide a theoretical basis for studying the drought resistance characteristics, the physiological activities of grapevines and the water balance of the SPAC system.