Agronomic response, transpiration and water productivity of four almond production systems under different irrigation regimes

Scientia Horticulturae 334 (2024) 113335 Available online 30 May 2024 0304-4238/© 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Agronomic response, transpiration and water productivity of four almond production systems under different irrigation regimes Manuel Quintanilla-Albornoz a , * , Joaquim Bellvert a , Ana Pelech´ a a , Xavier Miarnau b a Efficient Use of Water in Agriculture Program, Institute of Agrifood Research and Technology, Fruitcentre, Parc AgroBiotech, Lleida, 25003, Spain b Fruit Production Program, Institute of Agrifood Research and Technology, Fruitcentre, Parc AgroBiotech, Lleida, 25003, Spain ARTICLE INFO Keywords: Almond Production systems Sap flow Remote sensing Crop transpiration Water productivity ABSTRACT In recent years, more intensive production systems have been developed, coinciding with a growing scarcity of water resources. This context underscores the imperative of prioritizing water productivity (WP) as a critical factor in choosing the optimal production system to minimize agricultural water use. This study aims to contribute by evaluating WP in almond orchards under four production systems: open vase with severe pruning (open vase), open vase with minimal pruning (open vase (MP)), central axis and hedgerow. Three irrigation treatments were applied over two consecutive growing seasons: fully irrigated, mild stress and severe stress. Crop transpiration was monitored over the two years using both sap flow sensors and the two-source energy balance (TSEB) model with remote sensing. The severe stress treatment exhibited a notable reduction in kernel yield and nut load of 31.6 % and 34.5 %, respectively, in the second year of water deficit. The hedgerow system tended to have similar kernel yield to the open vase (MP) and central axis systems, and higher compared to the open vase system. Additionally, both transpiration measurement methods revealed that hedgerow exhibited lower transpiration rates across all irrigation treatments. Therefore, the highest WP was observed in the hedgerow system throughout both studied years. Similar findings were derived from the analysis of long-term data. Our findings indicate that the hedgerow production system had the highest WP, averaging 0.43 kg m −3 historically, compared to 0.33 kg m −3 for the open vase, 0.34 kg m −3 for the open vase (MP), and 0.36 kg m −3 for the central axis systems. 1. Introduction In the Mediterranean region, almond tree (Prunus dulcis (Mill.) D. A. Webb. syn. Prunus Amygdalus Batsch.) has traditionally been considered a social and unprofitable crop, mostly cultivated under rainfed conditions and marginal soils (Gradziel et al., 2017). Spain has the largest cultivated area of almond trees in the world but only ranks third in production, with an area of 878,075 ha an average yield of approximately 500 kg ha −1 of in-shell almonds (MAPA, 2023). These values contrast with those obtained in California and/or Australia, which exceed 2000 kg ha −1 of kernel yield. There is little doubt that the main difference between Spain and California/Australia almond yield per hectare lies in irrigation (USDA-NASS, 2022). In Spain, irrigated almond has become more popular during recent decades, rising from 4.7 % of cultivated land in 2005 to 26 % in 2023 (180,562 ha) (MAPA, 2023). This increase in surface area has gone hand in hand with the introduction of new production systems (understood as the interaction between rootstock, cultivar, training system and planting distance) capable of improving agronomic traits such as yield, precocity, and efficiency (Iglesias and Echeverria, 2022; Iglesias and Torrents, 2022; Maldera et al., 2021; Miarnau et al., 2022; Reig et al., 2020). It is well known that orchard intensification provides several advantages, including early bearing and lower labor costs due to mechanization (DeJong et al., 2005; Maldera et al., 2021). Furthermore, alternative novel architectures with higher densities and pruning methods that enhance light interception and distribution into the canopy have recently been developed, ensuring increased orchard profitability under new intensive production systems while reducing expenses (Iglesias and Echeverria, 2022). Nevertheless, one of the main concerns of high-density plantations is the higher initial investment as more trees are needed per hectare. The number of trees per hectare of a traditional almond plantation (6 ×6 m) has been estimated at around 278, whereas for highand super high-density plantations the values can be as high as 666 and 2800 trees per hectare, respectively (Casanova-Gasc´ on et al., * Corresponding author. E-mail address: [email protected] (M. Quintanilla-Albornoz). Contents lists available at ScienceDirect Scientia Horticulturae journal homepage: www.elsevier.com/locate/scihorti https://doi.org/10.1016/j.scienta.2024.113335 Received 23 March 2024; Received in revised form 9 May 2024; Accepted 20 May 2024 Scientia Horticulturae 334 (2024) 113335 2 2019; Miarnau et al., 2022). The transition towards more intensive crop growing is occurring at the same time as water resources are becoming scarcer due to droughts and climate change, especially in the Mediterranean region (Tramblay et al., 2020; Gaona et al., 2022). The gradual depletion of water reserves has emerged as a significant concern in the agricultural sector, prompting regulatory authorities to implement water supply restrictions and rationing measures throughout the Mediterranean region. For example, many irrigation districts located in the Ebro basin (Spain) were forced to impose water restrictions during 2023 without distinction between crops to ensure the survival of fruit tree crops. In the particular case of the Segarra-Garrigues (Lleida, Spain) irrigation district, water rights for almonds in 2023 were established at 1350 m 3 ha −1 , which corresponded to a water restriction of around 80 % (6500 m 3 ha −1 in full water allocation conditions) (ASG, 2023). In this context, the promotion of almond orchards with production systems with a higher agricultural water productivity (WP) is crucial to mitigate the effects of climate change. WP has usually been determined as the ratio between yield and the amount of irrigation water applied (Fereres and Soriano, 2007; L´ opez-L´ opez et al., 2018; Moldero et al., 2021). However, maximizing WP requires both an accounting of water consumed through irrigation and an understanding of the crop yield-evapotranspiration relationship. In this regard, numerous studies have assessed the effect of adopting regulated deficit irrigation (RDI) strategies to improve agricultural WP, showing in general that RDI strategies in almonds could be beneficial for WP maximization (Egea et al., 2010; Goldhamer and Fereres, 2017; L´ opez-L´ opez et al., 2018). Although some studies indicated that the adoption of RDI strategies may result in a 20 % reduction in yield, the substantial water saving of 60 % indicates that RDI remains a promising option in water-constrained areas (Girona et al., 2005). In addition of the adoption of RDI strategies, WP could potentially be improved through the development of new production systems. The crop water requirements are strongly affected by canopy size (Espadafor et al., 2015; Jofre-ˇ Cekalovi´ c et al., 2022). Therefore, among the different parameters involved in a production system, the selection of an appropiate training system and an ideal planting distance probably could have the highest impact on transpiration, and as a consequence, WP. Some studies have compared canopy light interception, as a proxy of transpiration, between training systems in woody crops (Auzmendi et al., 2023). These concluded that open-center or 3D canopy systems had a higher light interception than planar designs such as palmettes and central leader canopy structures (Casanova-Gasc´ on et al., 2019; Iglesias and Echeverria, 2022). This study also described that almonds under an open-center vase system had higher kernel yields in comparison to a super high-density system (Casanova-Gasc´ on et al., 2019). However, other studies have reported that more intense production systems exhibit higher yield per canopy volume, therefore improving yield efficiency due to more efficient use of light (Ben Yahmed et al., 2016; Casanova-Gasc´ on et al., 2019). WP can also be derived as the ratio between yield and unit of water consumed by a crop (transpiration) (Fern´ andez, 2023). Although this methodology is not widely used due to the difficulties of monitoring canopy transpiration, this option seems to be more appropriate to assess differences between crops or production systems, because part of the water supplied by precipitation and irrigation can be lost to drainage and runoff. Among the different techniques used to monitor transpiration, sap flow sensors and remote sensing are the two most promising for the purpose of the present study. Sap flow sensors have significant benefits as they measure transpiration in an easy-to-use, continuous, automated manner with a high temporal resolution (Smith and Allen, 1996; Fern´ andez et al., 2001; L´ opez-Bernal et al., 2010; Noun et al., 2022). For their part, remote sensing thermal-based surface energy balance (SEB) models have also been shown to be favorable for the estimation of evapotranspiration (ET) fluxes in a wide range of ecosystems, monitoring heterogeneous surfaces and generating operational ET products (Allen et al., 2007; Drexler et al., 2004; Kalma et al., 2008; Overgaard et al., 2006; Shuttleworth and Wallace, 1985). Among the different SEB models, the two-source energy balance (TSEB) model (Norman et al., 1995) is the most widely used approach, in part because it can estimate canopy transpiration and soil evaporation separately. Studies have demonstrated the suitability of using this modelling approach with very high-resolution imagery acquired from an unmanned aerial vehicle (UAV) to assess crop ET at field scale (Peddinti and Kisekka, 2022; Gao et al., 2023; Ramírez-Cuesta et al., 2023) or for field-based phenotyping purposes (Bellvert et al., 2021; G´ omez-Cand´ on et al., 2023). To our knowledge, no study has directly assessed the WP of different almond production systems by monitoring canopy transpiration. Only a recent study by Quintanilla-Albornoz et al. (2023) demonstrated that transpiration estimates of almonds through the combined use of a TSEB contextual approach and an adapted version of the Campbell and Norman radiation transmittance model for rectangular hedgerow crops had an overall coefficient of determination (R 2 ) of 0.77 and root mean square error (RMSE) of 0.13 mm h −1 . This study aims to go a step further and determine the WP of 12-year-old almond trees with different training systems and planting distances subjected to diverse irrigation regimes during two consecutive growing seasons. For this purpose, transpiration was monitored through two methodologies in order to confirm the robustness of the results: sap flow sensors and the remote sensing TSEB modelling approach. Additionally, a retrospective evaluation of WP will be undertaken over a long-term period. 2. Materials and methods 2.1. Trial location and design This research was conducted at the experimental station of IRTA (Institute of Research and Technology, Food and Agriculture) in Les Borges Blanques, Spain (41◦30 ′ 31.89 ′ ’N; 0◦51 ′ 10.70 ′ ’E, 323 m elevation) (Fig. 1a,b). The almond orchard, planted in June 2009, features “Marinada” as the scion cultivar grafted onto ‘INRA GF-677 ′ rootstock and incorporates four distinct training systems and planting distance combinations. Henceforth, in this manuscript, the term “production system” will be used to denote the combination of training system and planting distance. The following production systems were assessed: open vase with severe pruning (referred to as open vase in following lines) with a planting distance of 6 ×6 m, open vase with minimal pruning (referred to as open vase (MP) in the following lines) with 5.5 ×3.5 m, central axis with 5 ×3 m, and hedgerow with 4.5 ×3 m (Fig. 1c). The soil texture of the study site was classified as clay-loamy, with a depth ranging between 1.6 and 2 m. The climate in the study site corresponds to Mediterranean, with an average annual rainfall of 360.3 mm and an average annual evapotranspiration of 1090 mm. Table 1 shows the annual cumulative reference evapotranspiration (ETo) and precipitation (PP) from 2011 to 2022, sourced from a weather station located at 500 m from the study site, with code “les Borges Blanques [YD]”, in Catalonia’s official network of meteorological stations (SMC, www.ruralcat.net/web/ guest/agrometeo). Fig. 2 illustrates the daily values of ETo, PP, vapor pressure deficit (VPD), and minimum and maximum air temperature (T min and T max ) throughout 2021 and 2022 seasons. A frost event was recorded on March 20th 2021 and April 5th 2022, with a T min of −3.3 ◦C and −3.9 ◦C, respectively. These events coincided with the full flowering stage of almond trees in both years. The orchard was irrigated using a drip irrigation system. The irrigation system for both the open vase and open vase (MP) systems involved two lateral pipes positioned on each side of the tree at 40 cm, with drippers spaced every 70 cm and a water discharge of 2.2 l h −1 . The central axis and hedgerow had only one lateral pipe along with the row line, with drippers positioned every 60 cm and a water discharge of 3.8 l h −1 per dripper. Prior to 2021, the orchard was irrigated daily seeking to replace crop evapotranspiration (ET C ) as follows: ET C =(ETo x Kc x Kr) – effective rainfall. Kc and Kr represent the crop coefficient and shading M. Quintanilla-Albornoz et al. Scientia Horticulturae 334 (2024) 113335 3 factor, respectively. The Penman-Monteith method was used to determine ETo (Allen et al., 1998). The Kc values were derived from Goldhamer and Snyder (1989), who estimate the following Kc values at different phenological stages in almonds trees under the open vase training system: Kc 1 =0.70 (April), Kc 1 =0.95 (May), Kc 2 =1.09 (June), Kc 3 =1.15 (July), Kc 4 =1.17 (August), and Kc 5 =1.12 (September). Kr was determined as a function of canopy size of the trees, following Fereres et al. (1981). A Kr value of 0.98 was calculated for the entire orchard, considering the size of the trees in the open vase production system. To address potential uncertainties regarding different water requirement among production systems, irrigation was also adjusted to ensure that the stem water potential of fully irrigated trees remained above −1.0 MPa (Girona et al., 2006). Effective rainfall was estimated as half of the rainfall for a single event-day with more than 10 mm of precipitation and otherwise was considered to be zero (Olivo et al., 2009). During 2021 and 2022, each production system was divided into three irrigation treatments: (i) Fully irrigated, where irrigation was aimed at fulfilling tree ET requirements (100 % ET C ) throughout the growing season; (ii) mild stress, with irrigation maintained at 50 % ET C throughout the growing season; and (iii) severe stress, with irrigation limited to 20 % ET C throughout the growing season. A split-block design was used in this experiment. Each block treatment consisted of 6 trees (3 ×2). The amount of water applied to each treatment was measured with digital water meters (CZ2000–3 M, Contazara, Zaragoza, Spain), as Fig. 1. (a) Location of the studied almond orchard located in Les Borges Blanques (Lleida, Spain), (b) RGB orthomosaic of the almond orchard, and (c) Photographs from winter 2011 (top) and summer 2022 (bottom) and description of the different production systems assessed in the trial. Table 1 Cumulative annual reference evapotranspiration (ETo) and cumulative annual precipitation (PP) for the period 2011–2022. Year ETo (mm) PP (mm) 2011 1095.1 294.4 2012 1096.1 303.0 2013 1072.5 435.4 2014 1074.1 425.7 2015 1101.6 185.9 2016 1080.2 352.9 2017 1099.7 290.0 2018 1059.9 535.3 2019 1132.6 377.2 2020 1070.9 475.8 2021 1065.3 288.7 2022 1132.5 317.5 Mean 1090.0 360.3 M. Quintanilla-Albornoz et al. Scientia Horticulturae 334 (2024) 113335 4 detailed in Table 2. 2.2. Physiological and structural measurements 2.2.1. Stem water potential The midday stem water potential (Ψ s ) was measured every two weeks using the protocol outlined by McCutchan and Shackel (1992). In each production system, two trees were measured for each irrigation treatment during the 2021 and 2022 growing seasons. The measured trees were the same where the sap flow sensors were installed in the open vase (MP), central axis and hedgerow systems. Three leaves of each tree were measured to estimate Ψ s . Shaded leaves were selected and placed in a bag covered with aluminum foil for 1 hour before measurement to balance the water potential between leaf, stem and branches. All leaves were measured within 1 hour using a pressure chamber (Plant Water Status Console, Model 3500; Soil Moisture Equipment Corp., Santa Barbara, CA). The integrated stem water potential (Int(Ψ s )) was estimated to account for both the intensity and the duration of water stress. Int(Ψ s ) was calculated using Eq.1, as defined by Myers (1998). Int(Ψs) =      ∑ i=t i=0(Ψi,i+1−Ψmax)n     (1) where Ψi,i+1 represents the average of Ψ s between two consecutive measurement days, n is the number of days in the measurement period, and Ψ max is the maximum Ψ s measured during the season. 2.2.2. Fraction of intercepted photosynthetically active radiation (fIPAR) The diurnal evolution of the fraction of intercepted photosynthetically active radiation (fIPAR) was measured in three trees of each production system, one per irrigation treatment, for the dates May 26th (shell growth period), June 29th (early kernel filling period) and July 29th of 2021 (kernel filling period). The methodology described by Casadesús et al. (2011) was followed. In each tree, fixed 125 cm long sensor bars containing photodiodes every 10 cm (VTB8440BH, PerkinElmer Optoelectronics, Vaudreuil, Canada) were installed. Each photodiode is installed in an aluminum channel with the outer section of 25 mm and covered by a 4.6 mm thick PTFE sheath (TECAFLON PTFE, Ensinger Ltd., Llantrisant, UK) that acted as a light diffuser. The photodiodes have a spectral response ranging from 330 to 720 nm wavelengths, with a peak at 580 nm, corresponding to the photosynthetically active radiation (PAR) region. The sensing bars were positioned parallel to the crop rows, covering the entire distances between trees. Because of the varying planting distances among the production systems, 10 sensing bars were installed in the open vase and open vase (MP) systems, while 5 were installed in the central axis and hedgerow systems. Reference measurements of PAR above the canopy (PAR above ) were obtained by using two sensing bars placed outside the field (uncovered) Fig. 2. Seasonal evolution of daily reference evapotranspiration (ETo), daily cumulative precipitation (PP), daily vapor pressure deficit (VPD), and daily minimum and maximum air temperature (T min and T max ) during 2021 and 2022. Table 2 Seasonal water applied through irrigation in each of the three irrigation treatments in each production system during 2021 and 2022. Production System Irrigation Treatment Irrigation (mm) 2021 2022 Open Vase Fully Irrigated 727 861 Mild Stress 355 404 Severe Stress 140 201 Open Vase (MP) Fully Irrigated 837 1093 Mild Stress 414 464 Severe Stress 177 258 Central Axis Fully Irrigated 741 863 Mild Stress 394 429 Severe Stress 152 201 Hedgerow Fully Irrigated 850 992 Mild Stress 390 475 Severe Stress 175 255 M. Quintanilla-Albornoz et al. Scientia Horticulturae 334 (2024) 113335 5 to measure direct incident PAR. The PAR below the canopy (PAR below ) was measured using bars installed on the ground of the planting system. The PAR data was recorded every 10 min. The fIPAR was calculated as follows: 1 – (PAR below /PAR above ). The daily average fIPAR (fIPARd) was then obtained based on measurements recorded during daylight hours. 2.2.3. Transpiration from sap flow sensors In situ transpiration was monitored in the open vase (MP), central axis and hedgerow production systems using sap flow sensors. Regretfully, open vase sap flow data was not used due to technical issues. In these production systems, two trees of the fully irrigated and severe stress treatments and one tree in the mild stress treatment were monitored with sap flows. Sap flows were installed at 0.5 m above the ground. Sap flow transpiration measurements (T-SF) were corrected for wounding and azimuthal effects (L´ opez-Bernal et al., 2010). The sap flow sensor uses the compensation heat pulse (CHP) and the calibrated average gradient method. The used sap flows were developed by the IAS-CSIC laboratory. The sap flow sensors consist of a 2 mm diameter 4.8-W stainless steel heater and two temperature sensors positioned 10 and 5 mm downstream and upstream of the heater, respectively. Both temperature sensors contain two embedded type E (chromel–constantan wire) thermocouples spaced 1 mm along the needle. The sap flux densities across the trunk radius are calculated based on the heat-pulse velocities at 5 and 15 mm below the cambium. For further specifications, see Villalobos et al. (2009). Sap flow data was recorded every 15 min using a CR1000 datalogger (Campbell Scientifc Inc., Logan, UT, USA). The correction coefficient was obtained from in situ measurements of transpiration obtained from a water balance method using Eq.2 (L´ opez-L´ opez et al., 2018): Twb =PP +IR −ΔSWC −DP −ES(2) where T wb is daily transpiration obtained by the water balance, PP is precipitation, IR is the amount of water applied through irrigation, ΔSWC is the difference in soil water content (SWC) between two consecutive days, DP is deep percolation and E S corresponds to evaporation. The water balance was calculated during three days without PP or IR application. Therefore, the parameters PP, DP and IR were considered null. The soil was covered with a plastic sheet to prevent E S fluxes during sap flow calibration. The SWC was measured using a neutron probe (Campbell Pacific Nuclear Scientific, Model 503) every 20 cm to a depth of 180 cm. At the time of tube installation, soil samples were collected to measure the actual volumetric moisture content (cm 3 of water/cm 3 of soil) for calibration of the neutron probe measurements. Six tubes were installed in one quarter of the planting area in each tree. Tube distribution corresponded to two sets of three tubes (6 tubes in total) placed beneath the emitter, one set at a quarter and the other at half of the distance between rows. Calibration coefficients determined from T wb and T-SF measurements were assumed to remain constant throughout the entire growing season, as demonstrated by Espadafor et al. (2015). The calibrated T-SF was used to calculate cumulative daily and seasonal transpiration. Daily and seasonal transpiration of the open vase trees, the production system without sap flow sensors, and using two trees in each irrigation treatment for 2021 and 2022 was estimated as follows: Transpiration =ETo x Ka. Ka corresponds to the ratio between actual transpiration and ETo. A multi-regression model was used to retrieve the actual crop coefficient (Ka). This model was trained using all the available data from sap flow sensors installed in the open vase (MP), central axis and hedgerow production systems, as well as the fractional canopy cover (fc) and Ψ s measured values. fc was used on the principle that it correlates well with the transpiration coefficient (K T ) (Espadafor et al., 2015). Additionally, the same methodology was applied to retrieve transpiration of all trees during the period 2016 to 2020 and for the remaining trees that had no sap flow sensors during 2021 and 2022. In this case, it was assumed that fc did not change over the six years in question and trees were fully irrigated. Therefore, Ka was calculated using the average fc of all trees during 2021 and Ψ s values of the fully irrigated treatment. 2.3. Transpiration from remote sensing 2.3.1. Image acquisition campaign The image acquisition campaign consisted of five flights carried out in 2021 (March 24th, May 19th, June 3rd and 29th, and July 29th), and four in 2022 (May 27th, June 29th, and August 1st and 19th). The flights were conducted with a UAV Dronehexa XL (DRONETOOLS, Seville, Spain) equipped with a Micasense RedEdge-MX multispectral camera (Micasense, Northlake Way, Seattle, USA) and a FLIR SC655 thermal camera (FLIR Systems, Wilsonville, OR, USA). The Micasense Red-EdgeMX captures images in five spectral bands, with wavelengths of 475 ± 20 nm, 560 ±20 nm, 668 ±10 nm, 717 ±10 nm, and 840 ±40 nm. The FLIR SC655 responds within the 7.5–13 µm spectrum. The flights were carried out at 50 m above ground level in order to obtain multispectral and thermal images with spatial resolutions of 0.03 m and 0.06 m, respectively. All images were subjected to radiometric, atmospheric and geometric correction. The Jaz spectrometer (Ocean Optics, Inc., Dunedin, FL, USA) was used to collect in situ spectral measurements on several ground targets in parallel with the image acquisition for radiometric calibration. The Jaz has a wavelength response from 200 to 1100 nm and an optical resolution of 0.3–10.0 nm. The spectrometer was calibrated using a white reference panel (white color SpectralonTM) before spectral measurements. The thermal sensor was calibrated in the laboratory with a blackbody (model P80P, Land Instruments, Dronfield, United Kingdom). In addition, in situ temperature measurements were collected using an SI-111-SS Apogee infrared radiometer connected to an Apogee AT-100 microCache Bluetooth micrologger (Apogee Instruments Inc, Logan, UT, USA). The mosaicking process and generation of the digital elevation model (DEM) and digital surface models (DSM) from point clouds of the multispectral images were performed with the Agisoft Metashape Professional software (Agisoft LLC., St. Petersburg, Russia), while QGIS 3.4 (QGIS 3.4.15) was used for the geometric and radiometric corrections. 2.3.2. Biophysical traits The canopy area, canopy width, canopy height, fc, canopy volume, canopy temperature (T c ) and soil temperature (T s ) were obtained for all almond trees using a supervised image classification based on the use of the DSM and the soil-adjusted vegetation index (SAVI). SAVI was chosen because it increases the contrast between vegetation and the soil surface by reducing the impact of soil brightness in the red and near-infrared wavelengths (Qi et al., 1994). Pixels were classified as canopy when having DSM values greater than 1.5 m and SAVI values greater than 0.2. Pixels that did not meet the conditions were categorized as pure soil. The canopy area was estimated as the sum of the area of the pixels classified as canopy. Canopy volume was estimated as the product of canopy area and canopy height. In addition, the normalized difference vegetation index (NDVI) (Jackson and Huete, 1991), the normalized difference water index (NDWI) (McFeeters, 1996) and the modified triangular vegetation index (MTVI2) (Xing et al., 2020) were also calculated. These parameters were used as additional inputs of a machine learning approach to estimate the leaf area index (LAI) using a random forest algorithm (scikit-learn Python library) and following the methodology described by Quintanilla-Albornoz et al. (2023) in a study which demonstrated that this approach was able to estimate the LAI with an RMSE of 0.30 m 2 m −2 . The extraction of biophysical traits from high resolution images was conducted using the Python programming language (Python Software Foundation. Python Language Reference version 3.10. available at http://www.python.org). M. Quintanilla-Albornoz et al. Scientia Horticulturae 334 (2024) 113335 6 2.3.3. TSEB model description Transpiration was also estimated for all almond trees with the contextual TSEB model using high-resolution thermal and multispectral imagery (Nieto et al., 2019). TSEB is based on an energy balance approach that assumes that the surface net canopy radiation (R n ) is distributed mainly between sensible heat flux (H), latent heat flux (LE) and soil heat flux (G). Therefore, LE (W m −2 ) can be calculated as a residual of the surface energy equation (Kustas and Anderson, 2009; Norman et al., 1995). To retrieve transpiration, the canopy and soil R n were estimated using the canopy radiative transfer model of Campbell and Norman (1998). To estimate canopy transmittance, a basic clumping index was assumed for the open vase system. Meanwhile, a rectangular hedgerow clumping index was applied to the open vase (MP), central axis and hedgerow systems, following the procedure outlined by Quintanilla-Albornoz et al. (2023). The contextual TSEB approach was applied with direct measurement of T c and T s from high-resolution thermal images. To estimate daily transpiration (T-TSEB) from the instantaneous TSEB estimations, the simulated evaporative fraction variable method was used (Hoedjes et al., 2008; Quintanilla-Albornoz et al., 2024). 2.4. Production and kernel quality parameters Harvesting was conducted by shaking trees with commercial equipment and then picking up almonds from the ground. After that, nuts were dehulled and their in-shell fresh weight per tree measured, calculating the gross yield. To determine the dry kernel weight, a 1 kg inshell nut sample for each replicate/tree was naturally dried until the moisture content in the kernel reached 6 %. The shelling percentage (kernel weight/in-shell weight) was obtained by measuring the dry weight of the in-shell nut and kernel from a sample of 100 nuts. Finally, the kernel yield (kg ha −1 ) was computed by multiplying the shelling percentage by the in-shell nut yield. The estimated kernel yield per canopy volume was calculated by dividing the kernel yield of the tree by its estimated canopy volume. The nut load was measured by dividing the total kernel yield by the average kernel weight. 2.5. Water productivity (WP) The WP was calculated by dividing kernel yield by transpiration. For the 2021 and 2022 growing seasons, WP was assessed using the following two methodologies: (1) seasonal cumulative T-SF data for all production systems, with the partial exception of the open vase which was modelled using a multi-regression model (WP-SF); and (2) using the averaged estimated daily transpiration through the TSEB model of all image acquisition dates in each year (WP-TSEB). The difference between these two methodologies is that in the first only very few trees were monitored with sap flow sensors, while in the second a more representative information was obtained for each treatment/production system since transpiration was estimated for six trees in each. Additionally, WP was also estimated for the period 2016–2020 using historical kernel yield data. In this case, transpiration was obtained for all trees in each production system using the multi-regression modelling approach previously explained to retrieve the Ka. WP was not estimated prior to 2016 due to the lack of information regarding biophysical traits, and therefore it was assumed that canopy volume did not significantly change between the 2016 and 2021. 2.6. Statistical analysis The data collected was analyzed with a three-way analysis of variance (ANOVA) using the JMP Pro-Software (version 16.0, SAS Institute Inc., Cary, NC). A final mixed model including production systems, irrigation treatments and their interaction as fixed factors, and block nested to date (year or day depending on the analysis) as a random factor was used to separate the treatment effects from the data. Tukey’s HSD test was used to compare treatments for all production systems when the source effect was significant, with p-values ≤0.05. 3. Results 3.1. Stem water potential as water status indicator Fig. 3 illustrates the seasonal pattern of stem water potential (Ψ s ) during the 2021 and 2022 growing seasons. The Int(Ψ s ) exhibited a significant response between years, production systems, irrigation treatments, and the interaction between production system and irrigation treatment (PSxTRT) was also significant (Table 3). Overall, Ψ s for the fully irrigated treatment was relatively constant during both growing seasons, with values above −1.0 MPa, with the partial exception of two dates. These two dates correspond to June 19th 2022 (DOY 170, shell growth period) and August 19th 2022 (DOY 231, kernel filling period), where minimum Ψ s values were recorded for all irrigation treatments and production systems. These lower values for the two dates of 2022 explained, in part, the statistical significance between years. The severe stress treatment showed the lowest values throughout both growing seasons. Ψ s started declining at the beginning of the growing season, with values always below −1.3 MPa from June 15th (DOY 166, early kernel filling period), with the partial exception of the open vase production system. Although irrigation was also reduced in the severe stress treatment of the open vase system, Ψ s did not decline as much as in the other systems. The mild stress treatment always maintained intermediate values. Table 4 shows a summary of the Int(Ψ s ) values, comparing differences between the production systems in each irrigation treatment. Open vase was the production system with the lowest values in all irrigation treatments. 3.2. Biophysical traits The biophysical variables fc, canopy height, canopy volume and LAI showed significant differences between years, production systems and in the interaction between production system and year (PSxYear), and PSxTRT (Table 3). Open vase (MP) exhibited the highest fc in all irrigated treatments, which ranged from 55 to 69 % depending on the treatment (Table 4). Conversely, hedgerow had the lowest fc, ranging from 47 to 52 % also depending on the irrigation treatment. Central axis had a higher fc than open vase for all treatments. Canopy height was higher in 2022 than in 2021. Similarly to fc, open vase (MP) had the tallest trees and hedgerow the lowest. On the other hand, open vase had the highest canopy volume in all irrigation treatments, ranging between 78.9 and 102.8 m 3 . In contrast, the central axis and hedgerow systems displayed the smallest canopy volume in all irrigation treatments, with values ranging from 26.8 to 38.4 m 3 . Although LAI was the only variable that did not show significant differences between irrigation treatments, the interaction PSxTRTxYear was slightly significant. This explains that all the production systems under full irrigation had a higher LAI than other irrigation treatments, except for hedgerow. On average, open vase had significantly lower values in comparison to the other production systems. These differences were more evident in the severe stress irrigation treatment. The fIPARd exhibited a significant variation with production system, irrigation treatment and PSxTRT (Table 3). Significant differences in irrigation treatments were observed between open vase and open vase (MP), while no significant differences were detected between central axis and hedgerow across irrigation treatments. Overall, open vase (MP) showed a significantly higher fIPARd, with a daily mean value of 0.62. The open vase system had the lowest fIPARd, with a mean value of 0.54. The other two production systems showed intermediate values (Table 4). The diurnal course of fIPAR for each production system is shown in Fig. 4. The statistical analysis showed a significant interaction between production system and hour (p-value <0.001). The significantly lower values of the open vase system were observed during the M. Quintanilla-Albornoz et al. Scientia Horticulturae 334 (2024) 113335 7 morning, from 7:00 to 11:00 (solar time), and also during the afternoon (from 14:00 to 15:00). It can also be observed how the decline of fIPAR of the hedgerow production system was more pronounced during the morning until reaching its minimum values between 12:00 and 14:00, coinciding with the peak of solar radiation. At this time, however, no significant differences were observed in fIPAR between production systems. During the afternoon, central axis and open vase (MP) exhibited significantly higher fIPAR than hedgerow and open vase. 3.3. Transpiration from sap flow sensors The multi-regression model developed to simulate daily transpiration for the open vase production system yielded the following equation: Ka =1.21 fc - 0.24 Ψ s , where Ψ s is in MPa. The R 2 , RMSE and bias were respectively 0.61, 0.124 and 0.01. The ANOVA analysis to evaluate seasonal cumulative T-SF revealed significant differences between production systems and between treatments, but not in their interaction (Table 3). The seasonal pattern of T-SF for fully irrigated trees had a trend similar to ETo, reaching the maximum values during July coinciding with dates with a higher ETo and canopy growth (Fig. 5). At this time, maximum T-SF values reached 7 mm day −1 for open vase (MP) and central axis, and 5 mm day −1 for open vase and hedgerow. During August and September, T-SF started progressively decreasing in all treatments with ETo and leaf senescence. In contrast, almond trees Fig. 3. Seasonal patterns of midday stem water potential (Ψ s ) for the production systems: (a) Open Vase; (b) Open Vase (MP); (c) Central Axis; and (d) Hedgerow, in the different irrigation treatments (fully irrigated, mild stress and severe stress) during the growing seasons 2021 and 2022. Table 3 Analysis of variance (three-way ANOVA) to assess the effect of year, production system (PS) and irrigation treatment (TRT) on the integrated stem water potential (Int (Ψ s )), the biophysical traits: fractional cover (fc), canopy height, canopy volume, leaf area index (LAI), and daily fraction of intercepted photosynthetically active radiation (fIPARd), seasonal cumulative daily transpiration from sap flow sensors (T-SF) and averaged daily transpiration from remote sensing using the TSEB modelling approach (T-TSEB). Source Year PS TRT PSxYear TRTxYear PSxTRT PSxTRTxYear Int(Ψ s ) 0.005 <0.0001 <0.0001 ns ns 0.0082 ns fc 0.0382 <0.0001 0.0002 0.0276 ns 0.0076 ns Canopy Height <0.0001 <0.0001 <0.0001 <0.0001 ns 0.046 ns Canopy Volume 0.0037 <0.0001 <0.0001 <0.0001 ns 0.0002 ns LAI <0.0001 <0.0001 ns <0.0001 ns 0.0359 0.0063 fIPARd – <0.0001 <0.0001 – – 0.0002 – T-SF ns <0.0001 <0.0001 ns ns ns ns T-TSEB <0.0001 <0.0001 <0.0001 0.0248 ns 0.0039 ns M. Quintanilla-Albornoz et al. Scientia Horticulturae 334 (2024) 113335 8 under the severe stress treatment reached peak T-SF around the second week of May, after which it gradually declined. The T-SF pattern in the mild stress treatment was similar to that in fully irrigated in the open vase and open vase (MP) systems. In central axis and hedgerow, the values were intermediate with the maximum reached around mid-June. Seasonal cumulative T-SF presented significant differences between production systems and irrigation treatments (Table 3). The hedgerow system significantly transpired less under the three irrigation treatments (Table 4), presenting the lowest T-SF with a mean of 441.4 mm year −1 compared to the 532.7 mm year −1 for the central axis system. However, open vase and open vase (MP) had the highest T-SF with 618.1 mm year −1 and 643.2 mm year −1 , respectively. T-SF clearly differed between irrigation treatments, where fully irrigated, mild stress and severe stress exhibited T-SF values of 639.7 mm year −1 , 571.3 mm year −1 and 458.8 mm year −1 , respectively. 3.4. Transpiration from remote sensing The T-TSEB exhibited significant differences between years, production systems and irrigation treatments, and in the PSxYear and PSxTRT interactions (Table 3). The relationship between seasonal cumulative T-SF and the mean T-TSEB for the nine image acquisition dates Table 4 Comparison of integrated stem water potential (Int(Ψ s )) and the main biophysical traits estimated and measured during the image acquisition campaign for the different production systems and irrigation treatments: fractional cover (fc), canopy height, canopy volume, leaf area index (LAI) and daily fraction of intercepted photosynthetically active radiation (fIPARd), and seasonal cumulative daily transpiration from sap flow sensors (T-SF) and averaged daily transpiration from remote sensing using the TSEB modelling approach (T-TSEB). Different letters in each mean indicate significant differences at p-value <0.05 using Tukey’s honest significant difference test and considering the interaction between production system and irrigation treatment. Values corresponds to mean values of the two years. Irrigation Treatment Production System Int(Ψs) fc (%) Canopy Height (m) Canopy Volume (m 3 ) LAI (m 2 m −2 ) fIPARd T-SF (mm year −1 ) T-TSEB (mm day −1 ) Fully Irrigated Open Vase 144.4 b 53 c 5.25 b 102.8 a 3.24 b 0.53 c 641.2 a 5.29 ab Open Vase (MP) 243.4 ab 62 a 5.60 a 67.87 b 3.58 a 0.64 a 765.7 a 6.14 a Central Axis 326.8 a 55 b 4.40 c 36.9 c 3.42 ab 0.61 ab 610.7 a 5.45 ab Hedgerow 330.8 a 47 d 4.22 c 26.8 c 3.38 b 0.55 bc 541.1 b 4.51 b Mild Stress Open Vase 205.7 c 51 b 5.28 a 97.8 a 3.16 c 0.60 b 668.0 a 5.05 ab Open Vase (MP) 562.9 a 69 a 5.61 a 74.94 b 3.52 a 0.66 a 654.2 a 5.80 a Central Axis 652.7 a 57 b 4.46 b 38.4 c 3.23 b 0.59 b 532.9 b 4.91 ab Hedgerow 380.8 b 52 c 4.22 c 33.7 c 3.53 a 0.59 b 430.3 c 4.42 b Severe Stress Open Vase 228.7 b 46 b 4.75 b 78.9 a 2.9 c 0.50 b 509.6 a 4.72 a Open Vase (MP) 761.2 a 55 a 5.17 a 55.3 b 3.29 b 0.58 a 545.3 a 3.99 ab Central Axis 827.7 a 53 a 3.92 d 30.3 c 3.20 b 0.59 a 427.5 b 3.04 b Hedgerow 786.3 a 47 b 4.21 c 28.8 c 3.60 a 0.58 a 352.7 c 3.28 b Fig. 4. Daily evolution of the hourly fraction of intercepted photosynthetically active radiation (fIPAR) for each production system. Each value corresponds to the mean of one tree measured hourly (solar time) in each irrigation treatment for dates May 26th, June 29th and July 29th in 2021. Different letters indicate significant differences between production systems within each hour at p-value <0.05 using Tukey’s test. Grey dashed line corresponds to hourly solar radiation during measured days. M. Quintanilla-Albornoz et al. Scientia Horticulturae 334 (2024) 113335 9 showed a positive regression with an R 2 of 0.49 and a Pearson correlation coefficient (r) of 0.70. Open vase exhibited the highest T-TSEB under the severe stress treatment, whereas the open vase (MP), central axis and hedgerow displayed no significant differences in the severe stress treatment. However, the open vase (MP) production system had the highest transpiration rates in both the fully irrigated and the mild stress irrigation treatments. This confirms the results obtained with TSF. The lowest T-TSEB values under both full irrigation and mild stress conditions were observed in the hedgerow system. Significant differences were observed in the open vase system between years, with mean estimated transpiration values of 4.21 mm day −1 and 5.82 mm day −1 for 2021 and 2022, respectively. Fig. 5. Seasonal pattern of daily transpiration measured with sap flows for the different production systems: (a) and (b) Open Vase; (c) and (d) Open Vase (MP); (e) and (f) Central Axis; and (g) and (h) Hedgerow, in different irrigation treatments (fully irrigated, mild stress and severe stress) during the growing seasons 2021 (left) and 2022 (right). A multiple regression model was used to estimate transpiration in the open vase system. M. Quintanilla-Albornoz et al.