Precision agriculture uses information obtained from temporal or spatial data to address variability within the orchard, improve the productivity and quality of horticultural products, and enhance resource use efficiency and sustainability (Zhang et al., 2002). The application of sensors, remote sensing, information systems, and improved machinery are important for achieving low-input and high-efficiency horticulture (Zude-Sasse et al., 2021; Longchamps et al., 2022). Since irrigation is tightly associated with tree growth and productivity and water is an indispensable limited resource, technological tools that assess tree water status can improve irrigation scheduling and water use efficiency, avoiding over-irrigation or applying undesirable water stress (Gonzalez-Dugo et al., 2013; Blanco and Kalcsits, 2021).

According to the USDA, 2018 Irrigation and Water Management Survey, 22% of growers interviewed in the United States were using any type of sensor for irrigation management. Among all, the most common data-based method to schedule irrigation was soil sensors. Up to 65% more growers use soil sensors to schedule irrigation than those that use crop-water-evapotranspiration data. Even fewer growers use other methods, where soil sensors were used 5 and 14 times more than plant-based sensors or models, respectively (USDA, 2018). Based on this, the next step for growers is adding plant-based indicators to their soil sensors network that accurately provide low-cost and user-friendly information on the tree water status. There is a general agreement that the midday stem water potential is the reference tree water status indicator for fruit trees such as pear trees (Pyrus communis L.). However, its measurements are laborious (Marsal et al., 2000; Blanco et al., 2018). In the same vein, stomatal conductance has been widely reported as a sensitive physiological indicator of tree water status. However, it has not been implemented in commercial orchards because it is also labor intensive and requires highly technical and expensive equipment (Struthers et al., 2015; Noun et al., 2022). Other tree water status indicators such as canopy temperature (Tc) have also been identified (García-Tejero et al., 2011; Gonzalez-Dugo et al., 2012; Ramírez-Cuesta et al., 2022). Tc is strongly related to stomatal conductance by the direct relationship between high stomatal conductance, high transpiration rates, and the cooling effect of the vaporization of water. On the other hand, when stomatal conductance is lower from soil water deficits, transpiration decreases and canopy temperature increases. Despite this strong relationship with stomatal conductance, Tc is also highly affected by environmental conditions such as air temperature, vapor pressure deficit, and radiation (Mira-García et al., 2022). Thus, in order to be used as a tree water status indicator, Tc needs to be normalized to minimize its variation. To achieve that, several thermal indices have been proposed. Among these, the difference between canopy and air temperature (Tc-Ta) and the crop water stress index (CWSI) (Idso et al., 1981; Jackson et al., 1981) are those that have been widely reported as good water status indicators in many fruit trees such as citrus, nuts, and stone fruit (García-Tejero et al., 2011; Bellvert et al., 2018; Lee et al., 2019; Blaya-Ros et al., 2020).

Although pear trees are not drought resistant, previous research has highlighted the greater capacity to overcome water stress compared to other fruit trees such as apples or peaches. In this sense, deficit irrigation has proven to be a successful strategy in pears to control excessive vigor and increase flowering the following season (Marsal et al., 2012a). On the other hand, severe water stress will reduce pear fruit weight, yield, and relative gross revenue, with differences among cultivars, being ‘Blanquilla’/’Spadona di Salermo’/’Krystally’ more resistant to water stress than ‘Conference’ (Marsal et al., 2012a). Therefore, it is important to ensure water indicators that can accurately monitor tree water status and quantify the water stress applied to the trees to take full advantage of these irrigation strategies while avoiding the disadvantages that can cause.

Although new plant-based approaches that continuously assess tree water status have been recently developed (Lakso et al., 2022; Blanco and Kalcsits, 2023; Gonzalez Nieto et al., 2023), they are not as widely implemented in commercial fruit tree orchards as soil sensors, which are a major tool for precision irrigation of fruit trees (Vera et al., 2019). Thus, the combination of soil moisture sensors that continuously measure soil water status and plant-based punctual measurements can be an effective solution to assess tree water status and manage irrigation for commercial production. In this context, the development of low-cost, easy-to-use, portable infrared cameras can provide consistent and effective information on canopy temperature. Moreover, this technology can be incorporated into a smartphone giving growers the opportunity to rapidly include the use of image-based algorithms and thermal indices into their irrigation decision-making process (García-Tejero et al., 2018; Carrasco-Benavides et al., 2020).

The aim of this study was to compare traditional tree water status indicators (stem water potential and stomatal conductance) against the Tc-Ta and the CWSI obtained with thermal cameras for two different pear cultivars, ‘D’Anjou’ and ‘Bartlett’ (also known as ‘Williams’) under semi-arid climate conditions. We assessed if the combination of soil water content sensors and thermal imaging can provide robust information of the tree water status of pear trees.

In pear trees, controlling the excessive vegetative growth is required (Marsal et al., 2002; Carra et al., 2017). Irrigation strategies such as regulated deficit irrigation have shown benefits for many fruit tree species, reducing vegetative growth while increasing fruit quality (Galindo et al., 2018; Martínez-Hernández et al., 2020; Reid and Kalcsits, 2020). These strategies have been successfully applied in pear trees under different environmental conditions (Mitchell et al., 1989; Marsal et al., 2012b; Moreno-Hernández et al., 2017; Venturi et al., 2021). In order to implement these strategies, the impact of the water deficit on physiological responses needs to be quantified.

Midday stem water potential and stomatal conductance have been traditionally used to assess tree water status due to their robust and fast response to soil water limitations (Naor 2000; Van Beek et al., 2013; Conesa et al., 2019). Many studies have reported strong relationships between thermal-based indices, stem water potential and stomatal conductance (García-Tejero et al., 2018; Ahumada-Orellana et al., 2019; Park et al., 2021). Our results are aligned with those studies and show high correlation coefficients between thermal-based indices and traditional water status indicators ( Figure 8B ). Similar correlation coefficients to those obtained for pear trees (r = -0.65) were also reported in olive, citrus, almond, and walnut trees (Ben-Gal et al., 2009; Stagno et al., 2011; Dhillon et al., 2014). The findings reported in this study highlight the usefulness of thermal indices for assessing the tree water status of pear trees.

In the case of stomatal conductance, although deficit irrigation led to lower values and higher canopy temperatures, there was no relationship between the two due to the high influence of air temperature on canopy temperature. Low correlation coefficients between stomatal conductance and Tc were also reported in sweet cherry trees by Carrasco-Benavides et al. (2020) using a similar low-resolution thermal infrared camera, indicating limitations of Tc as a tree water status indicator. As such, it was not possible to identify a threshold for identifying water deficits ( Figure 5 ) because of the high dependency of tree canopy temperature in relation to air temperature and tree phenological stage ( Figure 8 ; Leuzinger and Körner, 2007; Kim et al., 2016). Our results stress the use of the Tc-Ta index over absolute values of Tc. Stomatal conductance and midday stem water potential were strongly related to Tc-Ta for this study. Tc-Ta has been reported to be strongly related to stem water potential in other woody crops such as stone fruits, citrus, nuts, and vines (Wang and Gartung, 2010; Gonzalez-Dugo et al., 2012; Bellvert et al., 2014; Gonzalez-Dugo et al., 2014).

The daily pattern of the canopy temperature and its relationship with stem water potential in ‘D’Anjou’ pear trees was also observed in a previous study (Blanco and Kalcsits, 2021). In pear trees, values of Tc-Ta similar to 2°C were measured during the experiment for DI trees under water restrictions with stem water potential values that ranged from -1.2 to -1.5 MPa. Slightly lower values of Tc-Ta (1.5°C) were reported by Blaya-Ros et al. (2020) for mature sweet cherry trees under similar soil water restrictions and midday stem water potentials of -1.2 MPa, but values of Tc-Ta higher than 4°C were reported in peach trees under deficit irrigation when stem water potential dropped below -2.0 MPa (Wang and Gartung, 2010). Based on these results, 1.5°C might be the threshold value of Tc-Ta proposed for pear trees in order to activate the irrigation as has been reported for ‘Bartlett’ that minimum values of midday stem water potential below -1.4 MPa during the second stage of fruit growth decrease fruit size (Marsal et al., 2002). Tc-Ta for CTL trees were always below 0°C and averaged -2°C, even under maximum air temperatures higher than 40°C ( Figure 6 ). Negative values of Tc-Ta are associated with trees under no soil water content restrictions (Bellvert et al., 2014). Thermal-based sensors have also been reported to be more reliable than stem water potential for assessing tree water status in soils at field capacity (Conesa et al., 2023). Considering that over-irrigated pear trees that have excessive vigor lead to reduced fruit yield, a range of Tc-Ta between -2 and 1.5°C is proposed as the ideal range to avoid penalties for severe water deficit or overirrigation.

When Tc-Ta was compared with the VPD, to calculate the NTB and NWSB, NTB was similar to that proposed by Jones et al. (2018) at 5°C, and higher than that stated by Gençoğlan and Gençoğlan (2018) in ‘Comice’ pear trees at 3 – 4°C. For the NWSB, our results were obtained under a wider range of VPD values, between 0.7 and 4.5 kPa, compared to the lower limit of non-water stressed trees calculated for ‘Comice’. When both baselines were compared at VPD values of 2 kPa, Tc-Ta values were approximately -7°C. However, for lower VPD values, the limits proposed for ‘D’Anjou’ and ‘Bartlett’ trees were more negative than those proposed to ‘Comice’ trees. The values of CWSI calculated in this study were similar to those reported by Egea et al. (2017) in olive trees under severe water stress but lower than those reported in apple trees under no water stress by Mohamed et al. (2021). The difference in the values reported by Mohamed et al. (2021) under similar environmental conditions might be attributed to the calculation of the NWSB, as that study used an approach based on environmental conditions which might overestimate their calculation of the CWSI compared to when non-stressed leaves are directly measured. In our study, CWSI for CTL trees ranged between 0.1 and 0.4, although they increased to 0.5 during the heat dome event of 2021, indicating values of CWSI similar to 0.5 can be measured when environmental conditions are highly demanding. For DI trees, CWSI reached values of 0.8 when the midday stem water potential was as low as -1.5 MPa and the stomatal conductance dropped below 200 mmol m^{- 2}s^{- 1}. Values of CWSI similar to 0.55 have been reported by Gençoğlan and Gençoğlan (2018) in ‘Comice’ pear trees irrigated to satisfy their water needs while the CWSI was similar to 0.8 for pear trees under water deficit. These results suggest a similar pattern for pear trees when combined with our study. Although the response to water deficit was not the same for both pear cultivars, values of CWSI higher than 0.5 were always related to trees under water stress conditions.

‘Bartlett’ pear trees were less stressed than ‘D’Anjou’ trees under the same soil and weather conditions, potentially due to differences in the tree’s vegetative growth. Policarpo et al. (2006) reported that ‘Bartlett’ pear trees tend to have bigger canopies compared to other pear cultivars, which might cause higher transpiration and, consequently, lower canopy temperature compared to ‘D’Anjou’ trees under the same soil water restrictions and environmental conditions. In this sense, CWSI seemed to be more sensitive to identifying water deficit in ‘Bartlett’ trees than in ‘D’Anjou’ trees. These differences between cultivars were also observed in sweet cherry trees, where CWSI was more sensitive for identifying water deficit in the cultivar with the highest stomatal conductance (Carrasco-Benavides et al., 2020). The results reported in the present work show the potential of using CWSI and Tc-Ta for assessing the water status of pear trees and provide threshold values related to the physiological response of the tree. This represents an advancement in the implementation of the use of thermal-based indicators which, jointly with the development of low-cost, easy to interpret continuous sensors, can be useful for irrigation scheduling (Giménez-Gallego et al., 2023).

Soil moisture sensors are still extensively used for making irrigation decisions. This may be due to their sensitivity to water limitations. Although several studies have reported high variability when measuring soil moisture under field conditions compared to other plant-based indicators (Blanco et al., 2018; Domínguez-Niño et al., 2020), in our experiment soil water content detected earlier water deficit than thermal-based indicators. Nine and 15 days after the deficit irrigation was imposed on DI trees in 2021 and 2022 respectively, soil water content was lower for DI treatments compared to the CTL. However, on the same date, there were no differences between irrigation treatments when thermal-based indices were used. These results agree with those described by Mira-García et al. (2022) who reported that thermal indices were not able to detect water stress in lime trees earlier than soil water content. In our study, soil water content was the first water status indicator that detected significant differences between irrigation treatments, followed 7 days later by the midday stem water potential and the stomatal conductance, and 14 days later by the thermal-based indices. Similarly, Struthers et al. (2015) observed in young pear potted trees from the cultivar ‘Conference’ that Tc of water-stressed trees showed significant differences compared with the well-watered trees after 18 days of water stress while other indicators such as stomatal conductance showed significant differences between irrigation treatments after nine days. This can be explained by the physiological response of both pear cultivars to slight/early water deficits, since stomata remain open and maintain high transpiration rates during the first days of water restrictions which consequently, did not lead to increased Tc.

The multiple linear regression presented in this study shows the combination of soil water content and Tc-Ta can be useful for assessing tree water status, particularly to estimate midday stem water potential. Soil water content and environmental conditions have been successfully used in fruit trees to estimate tree water status (Martí et al., 2013; González-Teruel et al., 2022). We propose the use of Tc-Ta compared with CWSI because determination coefficients calculated were similar for both indices and Tc-Ta does not require the calculation of the upper and lower limit baselines. Similarly, although the results of the non-linear regression relationship of CWSI, Tc-Ta and midday stem water potential and the stomatal conductance ( Figure 9 ) showed similar results to those of the multiple linear regression analysis, we propose the use of the combination of Tc-Ta and soil water content, since Tc-Ta alone was not precise enough to detect slight water deficits.

The results presented in this study show that the combination of Tc-Ta and soil water content sensors provides accurate information on tree water status and overcomes the limitations of their individual use. Thus, soil water content sensors detected water deficit earlier than thermal-based indices. However, when they were compared to midday stem water potential, they were not as strongly related compared to thermal-based indices. On the other hand, thermal-based indices were more closely related to tree water status during the season even though it took longer for irrigation treatments to influence this measured variable. The non-limiting baseline for each pear cultivar proposed in this work applied to a wide range of VPD conditions. This stable baseline may help facilitate the use of the CWSI by pear growers for irrigation scheduling. CWSI values higher than 0.7 were always related to severe water stress. The use of the Tc is not recommended, particularly to estimate stomatal conductance. Although Tc increased as a response to the stomata closure, environmental conditions had a greater impact. Measures that best integrate tree water status, soil water availability and atmospheric demand at a physiological scale will continue to be the best indicators of water stress to make precise irrigation decisions in pears.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Conceptualization and design, VB and LK. Methodology, VB, NW, TC, OH, and LK. Investigation, VB, NW, TC, OH, and LK. Writing—original draft preparation, VB, NW, TC, OH, and LK. Writing—review and editing, VB, NW, TC, OH, and LK. Visualization, VB and LK. Supervision, VB and LK. Project administration and funding acquisition, VB and LK. All authors contributed to the article and approved the submitted version.