Introduction

Electromagnetic induction (EMI) for soil sensing has been increasingly studied and used in viticulture systems with the surge of precision agriculture technologies owing to its non-destructive and prompt manner in acquiring soil data in commercial vineyards (Morari et al. 2009; André et al. 2012; Rossi et al. 2013; Yu and Kurtural 2020). Soil apparent electrical conductivity (ECa), or its reciprocal soil apparent electrical resistivity (ERa), is the main variable EMI sensors can measure and is utilized as an integrated variable to assess soil physical and chemical properties, including soil texture, water content, and salinity (Bittelli 2011; Brillante et al. 2014; De Clercq et al. 2009). Other soil characteristics, such as soil compaction and soil clay content can be appraised through soil ECa sensing, according to studies conducted by various authors (Hedley et al. 2004; Rodríguez-Pérez et al. 2011; Zarrouk et al. 2012). Previous research work indicated that, depending on the soil types, the severity of water deficits, and the grapevine cultivars, most of the active root zones are frequently distributed within the upper 1.0–1.2 m soil depth in drip irrigated vineyards (Smart et al. 2006; Soar and Loveys 2007; Steenwerth et al. 2008). Thus, the relationship between soil ECa and plant available water can be evaluated through the use of EMI sensors.

In the irrigated vineyards of northern California, when the soil–water salinity is low, the soil moisture content may become one of the main factors influencing soil ECa (Brillante et al. 2015). Among many plant physiological variables, plant water status is one of the most critical factors regulating plant physiological processes and responses to environmental parameters (Ojeda et al. 2001; Acevedo-Opazo et al. 2010). The plant water status is commonly assessed by measuring the hydrostatic pressure in plant tissues with a pressure chamber through leaf or stem water potential measurements (Scholander et al. 1965). The processes is time-consuming and labor-intensive, hence it is not practically applicable at field scale in large commercial vineyards. With the possibility to promptly assess soil ECa at field scale, EMI sensing eases the burden to identify the variability in soil and, potentially, the plant response in terms of plant water status and berry development (Bramley et al. 2011a; Tagarakis et al. 2013). Although the soil water content can be difficult to quantitatively predict from soil ECa assessments, due to the heterogeneity of vineyard soils (Brillante et al. 2014), the soil–plant water continuum can be well explained once the physical and chemical properties are surveyed to make soil ECa by proximal sensing explicitly interpretable (Morari et al. 2009). In viticulture research, the performance of grapevines is always the focus of both scientific research and commercial production. With soil texture playing an extremely critical role in grapevine growth and berry development, by providing nutrients and water, proximal soil sensing has the potential to capture the main sources of variability in the vineyards. A few recent studies have investigated how soil ECa sensing can be integrated into vineyard management (Brillante et al. 2015; Bonfante et al. 2015; Yu and Kurtural 2020). However, in the recent applications of proximal soil sensing in wine grape vineyards, the functional relationships between the soil ECa and the whole plant physiology are lacking thorough investigation and quantification.

Other than soil ECa sensing, an integrator of plant water status was recently investigated by analyzing the carbon isotope ratio of must sugars (van Leeuwen et al. 2010; Brillante et al. 2018). Briefly, ribulose 1,5-diphosphate will deplete more of the heavier carbon isotope 13C when stomata close under water deficits in most of the C3 plants, causing the 13C–12C ratio to increase (Farquhar et al. 1982, 1989). As a result, the assimilated sucrose in leaves will be translocated into berries and converted to glucose and fructose (Dai et al. 2013). The ratio of 13C to 12C was shown to be indirectly related to plant water status and stomatal conductance and net carbon assimilation by earlier studies (Farquhar et al. 1989; Bchir et al. 2016). Based on this mechanism, carbon isotope ratio analysis (δ13C) on berry must sugars can be exploited as an integrated proxy for spatially assessing plant water deficits (Gaudillère et al. 2002). Being able to reveal the accumulated effects of water deficits between veraison and ripening (Bchir et al. 2016), this method can act as a complementary approach to validate proximal soil sensing, without the demand to tracking plant water status throughout the berry maturation, to direct site-specific management for the upcoming season.

Overall, this study aimed at understanding how feasible it was to assess the vine water status through soil ECa sensing in wine grape vineyards. This study was also aimed to confirm the hypothesis that water could be the most contributing factors on soil ECa in wine grape vineyards. Finally, the possibility of using the must δ13C analysis to understand the relationships among grape berry sugar content, plant water status, and leaf photosynthesis would be evaluated. Such relationships may allow using δ13C, among other variables, for managing vineyard variability in Mediterranean climates.

Materials and methods

Vineyard site and weather information

This study was conducted in 2019 in Oakville, California, USA, with Vitis vinifera L. cv. Cabernet Sauvignon grafted on 3309C (V. riparia × V. rupestris). The grapevines were spaced at 1.5 m × 2.0 m (vine × row) and trained as a bi-lateral cordon on a single wire at the height of 1.80 m above the ground. The grapevines were mechanically pruned to a spur height of 0.10 m and treated with mechanical shoot removal at Eichhorn-Lorenz (E-L) stage 17 (Coombe 1995) to meet production demands. Bud break occurred on 7 April, 2019 at the study vineyard. The vineyard was drip irrigated with two emitters per plant with a nominal flow rate of 4 L h−1. Irrigation events were scheduled from fruit-set to harvest to refill about 50% of the potential crop evapotranspiration. About 138.6 mm of irrigation water was applied from fruit set to harvest, which occurred on 20 September 2019. Weather information at the study site was obtained from the automated weather station No. 77 (Oakville, CA) of the California Irrigation Management Information System (CIMIS), which is located about 200 m away from the study vineyard. The amount of irrigation water to apply was calculated multiplying the reference evapotranspiration (ETo), obtained from the local CIMIS station, by an empirical crop adjustment coefficient (Kc) estimated on the basis of the ground shaded area by the vines according to the methodology reported by Williams and Ayars (2005). The growing degree days (GDD) were calculated based on the average air temperature determined from the local CIMIS weather station as:

$$ {\text{GDD = }}\left[ {\frac{{\left( {MaxT - MinT} \right)}}{2}} \right] - Tref $$
(1)

where Max T is daily maximum air temperature, Min T is the minimum air temperature, and Tref is the reference temperature of 10 °C.

Experimental design

A regular 30 m × 30 m grid was used to set equidistant points throughout the vineyard with the size of approximately 0.5 ha where field measurements were collected. The grid allowed to identify 14 experimental units that included 3 grapevines per unit. The geolocations of the center grapevine within each experimental unit were registered with a GPS device (Yuma 2, Trimble Inc., Sunnyvale, CA, USA) connected to a Trimble Pro 6 T DGNSS receiver (Trimble Inc., Sunnyvale, CA, USA) to assist further GIS analysis (Fig. 1).

Fig. 1
figure 1

Location of the experiment site in Oakville, California, USA. a Highlighted area indicates the vineyard location in California, U.S.A. b Map of the experimental block with 14 experimental units marked, yellow squares illustrate the locations of each experimental unit. MHW mechanical high wire. (Fig. 1b previously appeared in Yu and Kurtural 2020) (Color figure online)

Full size image

Vineyard soil physical–chemical characteristics

Soil samples were taken on 29 May 2019, 52 days after bud break (DAB), from two depths in the 14 experimental units, including deep soil (0.75–1.5 m) and shallow soil (0–0.75 m). Soil pH, cation exchange capacity (CEC), soil organic carbon (SOC), total nitrogen (TN), and soil texture were measured according to the soil analysis methods of the North American Proficiency Testing (NAPT) program, western states section. Soil pH was measured with the saturation paste method (S–1.10), CEC was calculated from the extractable cations measured by ammonium acetate method (S–5.10), SOC and TN were measured by combustion method (S–9.30), soil texture was acquired by hydrometer analysis (S–14.10). Soil moisture content, gravel content were determined by the method in Sect. 26 in USDA Handbook No. 60 (Diagnosis and Improvement of Saline and Alkali Soils).

Assessment of spatial and temporal variations of soil features

Soil ECa was measured with an EM38 portable device (Geonics Ltd., Mississauga, ON, Canada), using both the vertical dipole mode (VDM) and horizontal dipole mode (HDM) to obtain ECa at two depths, i.e. 0–0.75 m and 0–1.50 m for shallow and deep soil layers, respectively. The EM38 device was calibrated prior to each soil survey following the manufacturer’s instructions in order to minimize errors. The instrument was placed on a sled consisting of PVC frame and was pulled by an all-terrain vehicle along the inter-rows at a distance of about 0.5 m from the vines to minimize interference phenomena due to metal wires of the trellis system, and about 0.15 m above the soil surface. The soil ECa was first assessed when the vineyard was around field capacity, i.e. 52 DAB. Three more measurements were taken on 94, 129, and 166 DAB using the same EM38 mobile device and settings.

Time domain reflectometry (TDR) sensors were installed at four of the 14 experimental units, including MHW1, MHW4, MHW9, and MHW12, which were selected on the basis of two main plant water status zones that were previously determined across the study orchard within a parallel research work (Yu and Kurtural 2020). Each soil moisture measurement location was equipped with an access tube, TDR sensors for in-situ determination of soil water content and with soil pore water electrical conductivity (ECe) meter to assess the soil–water salinity. The soil at the study site corresponds to the Bale soil series and is classified as fine-loamy, mixed, superactive, thermic Cumulic Ultic Haploxerolls according to the USDA soil classification system. The Bale soil series consists of very deep, somewhat poorly drained soils formed in stratified, gravelly and sandy alluvium from mixed sources. In these soil series, Ap horizon is distributed from 0 to 0.15 m deep, B horizon distributed from 0.15 to 0.61 m, Ab horizon is distributed from 0.61 to 1.12 m, and C horizon is distributed from 1.12 to 1.47 m. Three TDR sensors (TDR-315, Acclima, Inc., Meridian, ID, USA) were installed at 0.375 m, 0.75 m, and 1.45 m at each location to collect soil measurements for 0 – 0.75 m and 0–1.5 m soil depths by assessing horizons B, Ab, and C. However, the Ap horizon was not monitored by TDR due to the mechanical soil tillage periodically conducted throughout the growing season. The TDR sensors function within ± 0.1 vol %, for percent soil volumetric water content between 0 and 100 vol %, and ECe between 0 and 55 dS m−1. All the soil moisture sensors measured a width of about 0.15 m. As such, the sensors installed at 0.375 m and 0.75 m measured the volume corresponding to a depth of 0.75 m by 0.15 width for shallow soil, whereas the sensors at 0.75 m and 1.5 m measured a volume corresponding to a depth of 1.5 m. The values of soil volumetric water content (vol %) from both the analyses conducted on soil samples and those resulting from the TDR sensors were converted to cumulative soil water content (mm). At each location, the TDR sensors were connected to a DataSnap data logger (SDI-12, Acclima, Inc., Meridian, ID, USA) and the measurements were logged automatically every 10 min. The soil moisture data were then averaged at a daily time-step and cleaned following the Tukey rule by removing the outliers whose values were either lower than the first quartile minus 1.5 times the interquartile range (IQR) or higher than third quartile plus 1.5 times the IQR. The soil ECe, which was continuously monitored by TDR sensors at three depths, was consistently stable throughout the season without considerable increase approaching the harvest. Especially for the sensors at 0.375 m deep, consistently yielded ECe values below 2.5 and 3.0 dS m −1. Hence, it was assumed that the soil at this specific vineyard was non-saline and soil salinity did not act as a temporal fluctuating factor which could dominate ECa.

Soil ECa assessed by EM38 was corrected to a reference temperature of 25 °C according to Hayashi (Hayashi 2004) with the following equation:

$$ {\text{EC}}_{{{25}}} = EC_{T} /\left[ {1 + \delta (T - 25)} \right] $$
(2)

where EC25 is the ECa value at a referenced temperature of 25 °C, ECT is the ECa value at a measured temperature T by TDR sensors, and δ is set to 0.0191 °C−1 as the commonly used value based on the EC and temperature relationship of 0.01 M KCl solution.

Measurements of grapevine canopy parameters

Leaf area index (LAI) was measured with a smartphone based program “VitiCanopy” (De Bei et al. 2016) coupled with an iOS system (Apple Inc., Cupertino, CA, United States). The gap fraction threshold was set to 0.75, the extinction coefficient was set to 0.7, and sub-divisions were 25. The device was positioned at about 0.75 m underneath the canopy, with the maximum length of the screen being perpendicular to the cordon, and the cordon being located at the center of the screen according to the user’s instructions.

Normalized differential vegetation index (NDVI) was acquired with Crop Circle ACS-430 (Holland Scientific, Lincoln, NE, USA) at canopy closure on 13 May 2020, and calculated as,

$$ {\text{NDVI = (rNIR}} - {\text{rRED)}} \div {\text{(rNIR + rRED)}} $$
(3)

where rNIR was the reflectance in the near-infrared band, and rRED was the reflectance of the red (visible) band (Kriegler et al. 1969). The measurement was conducted the two sensors mounted on each side of the vineyard vehicle to a height accommodating the trellis. The data was recorded at 1 Hz to a GeoScout datalogger (Holland Scientific Inc., Lincoln, NE, USA) and geolocated with a Garmin 18 × 5 Hz GPS (Garmin Ltd., Olathe, KS, USA).

The spatial homogeneity was tested for LAI and NDVI by Cambardella index (CI), and calculated as,

$$ {\text{CI = [c}}_{{0}} {\text{/(c}}_{{1}} {\text{ + c}}_{{0}} {)]} \times {100} $$
(4)

where c0 is the nugget, c0 + c1 is the sill. When CI is less than 25, there is a strong spatial dependency; between 25 and 75, there is a moderate spatial dependency; more than 75, there is a weak spatial dependency (Cambardella et al. 1994).

Measurements of plant water status: mid-day stem water potential and leaf gas exchange

Mid-day stem water potential (Ψstem) measurements were taken bi-weekly over the course of the 2019 growing season. The measurements were taken on 2 July, 17 July, 1 August, 15 August, 29 August, and 20 September in 2019. From the main shoot axes on the grapevines, three shaded leaves were chosen and concealed in pinch-sealed Mylar® bags up to two hours prior to the measurements in each experimental unit. A pressure chamber (Model 615D, PMS Instrument Company, Albany, OR, USA) was used to take the Ψstem readings (Williams and Araujo 2002).

On the same dates, leaf gas exchange measurements were also taken at mid-day using a portable infrared gas analyzer CIRAS-3 (PP Systems, Amesbury, MA, USA). Three sun-exposed leaves were chosen from the main shoot axes, and three readings were taken on each selected leaf. In 2019, the sunlight was close to saturating conditions when the leaf gas exchange measurements were taken, where the average PARi was 1749.3 ± 203.7 µmol m−2 s−1. And the average vapor-pressure deficit (VDP) was 4.4 ± 1.7 kPa, average leaf temperature was 37.0 ± 4.2 °C. The standard environmental condition in the infrared gas analyzer was set with a relative humidity at 40% and a reference CO2 concentration at 400 μmol CO2 mol−1. Net carbon assimilation rate (AN, μmol CO2 m−2 s−1) and stomatal conductance (gs, mmol H2O m−2 s−1) were directly read by the gas analyzer. Intrinsic water use efficiency (WUEi, μmol CO2 mmol−1 H2O) was calculated as the ratio between AN and gs, according to equation WUEi = AN/gs.

Carbon isotope ratio analysis of berry must sugars (δ13C)

One-hundred berries were randomly collected at harvest from each experimental unit for δ13C analysis. The berries were crushed by hand and 1.5 mL of must was transferred into 2 mL conical tubes, and centrifuged at 4,000 RPM for 15 min. Afterwards, 10 µL of the supernatant from each sample was transferred into tin capsules (Thermo Fisher Scientific, Waltham, MA, USA) and placed in a microplate. The microplates were placed in oven at 80 °C overnight until completely dehydrated. The open tin capsules were then folded into a cubit, and wrapped, folded with another tin capsule to ensure the seal. The samples were analyzed for isotope ratio on a Vario MicroCube elemental analyzer coupled in a continuous flow mode to an isotope ratio mass spectrometer (IsoPrime, Elementar, Ronkonkoma, NY, USA). The δ13C values are reported in parts per thousand (‰) relative to the Vienna Peedeebelemnite-CO2 (VPDB-CO2) international reference.

Data processing and analysis

Geostatistical analysis, including kriging for proximal soil sensing, LAI, NDVI and soil characteristics were performed using ArcGIS (version 10.6, Esri, Redlands, CA, USA). The soil and NDVI were filtered by the speed that the vehicle was driving, which was between 3.2 km per hour and 8.0 km per hour. The specific soil ECa values were extracted from each experimental unit location, and then used for further analysis.

The R software package (R Foundation for Statistical Computing, Vienna, Austria) was used with the ‘dplyr’ package (Wickham et al. 2020) to perform multivariate regression analysis and derive the model between ECa and soil characteristics, including SOC, gravel content, silt content, clay content, soil ECe, and soil water content. Specifically, a multivariate regression model was developed to allow using the relationships between measured soil characteristics as predictors and ECa. An analysis of relative importance was performed with ‘relaimpo’ package (Grömping 2006). Linear regression analysis and graphing were performed using SigmaPlot 13.0 (Systat Software Inc., San Jose, CA, USA). The coefficient of determination between variables was calculated with linear regression analysis, and p values were determined to appraise the significance of the linear fittings, while root-mean-square errors (RMSE) were calculate to evaluate the standard deviation of the unexplained variance.

Results and discussion

Weather at experiment site

The experiment site received 10% of the yearly precipitation during the growing season (Table 1). This amount is typical for the study region and most semi-arid grape production areas. The study site received 970.3 mm of precipitation from October 2018 to April in 2019, whereas 103.1 mm of precipitation occurred from April to September in 2019. During the period from June to September in 2019, only 1.7 mm of precipitation occurred. The highest air temperatures occurred from June to September in 2019, with an average air temperature of 19.8 °C and a maximum air temperature of 29.9 °C. The GDD accumulation was calculated from April 1 and reached 1668 at harvest by late September in 2019. According to the Köppen climate classification (Peel et al. 2007), the experiment site was classified as having dry and warm summer climate and noted as ‘Csb’ and as Region III according to Winkler’s index (Winkler et al. 1974).

Table 1 Weather information at the experiment site as obtained from California Irrigation Management Information System (CIMIS) Station (#77 Oakville, Napa)
Full size table

Spatial distribution of soil physical and chemical properties at experiment site

For the deep soil layer (0.75–1.5 m), the percentages of clay and silt were greater in the southwestern section of the vineyard (Fig. 2a, b). Conversely, the percentage of sand was lower in the southwestern section (Fig. 2c). A lower plant water status from this portion of the vineyard was reported in previous research work conducted on the same study site, where, surprisingly, higher clay and silt contents resulted in more water deficits although the plants were irrigated uniformly at 0.5 ETc (Yu and Kurtural 2020). Although clay and silt are assumed to provide higher water holding capacity and greater soil moisture storage, the results indicated better plant water status in the portion with lower clay and silt contents with deficit irrigation. This was attributed to the fact that water can be more easily taken up by the plants in the sandy soil compared to clay soil when irrigated with water deficit as previously reported (Tramontini et al. 2013). The gravel proportion varied at the experiment site, ranging from 1.65% to 20.3% with the central section of vineyard with less gravel (Fig. 2d). The southwestern corner of the vineyard had greater Ca+, but in general there was not a spatial pattern as significant as for the soil texture (Fig. 2e). Soil pH showed lower values in the southeastern and northeastern section and higher values in the rest of the vineyard, ranging from 4.7 to 7.5 (Fig. 2f). There was a similar spatial pattern in CEC as for the soil texture, where greater CEC was observed in the southwestern section of the vineyard (Fig. 2g). The SOC was relatively uniform except in the northeastern section of the vineyard where slightly higher values were measured (Fig. 2h). Although the central section showed lower TN compared to the rest of the vineyard, as expected, the values were low in the deep soil and the variability was not visible (Fig. 2i).

Fig. 2
figure 2

Soil characteristics for deep depth (0.75–1.5 m). a Clay in %, b silt in %, c sand in %, d gravel in %, e Ca + in millequivalent per 100 g of soil, f pH, g cation exchange capacity (CEC) in milliequivalent per 100 g of soil, h soil organic carbon (SOC) in %, i total nitrogen (TN) in % (Color figure online)

Full size image

In the shallow soil layer (0–0.75 m), the soil texture was relatively homogenous compared to that of the deep soil layer. There was only very small area in the northeastern section of the vineyard showing significant difference for clay and sand, where there was higher clay and lower sand compared to the rest of the vineyard. (Fig. 3a, c), while another small portion of the northeastern section showed lower values in silt content (Fig. 3b). Significant variability in the gravel content was observed, where the northeastern section had higher values compared to the rest of the vineyard, from 19.6% to 9.2% (Fig. 3d). Soil Ca+ showed some variability in the vineyard, where the central section had lower values (Fig. 3e). The pH was relatively homogenous in the shallow soil layer with little variability ranging from 5.5 to 5.9 (Fig. 3f). The CEC was also homogenous in shallow soil, except for one unit in the southwestern section having 36.7 meq. per 100 g of soil, which was significantly higher than the rest of the vineyard (Fig. 3g). The soil organic carbon showed a similar pattern as that observed for the deep soil layer, where the northeastern section of the vineyard had higher organic carbon (Fig. 3h). As expected, there was proportionally higher total nitrogen in the shallow soil layer compared to deep soil layer (Fig. 3i). The northwestern corner and northeastern section of the vineyard showed higher nitrogen content.

Fig. 3
figure 3

Soil characteristics for shallow depth (0–0.75 m). a Clay in %, b silt in %, c sand in %, d gravel in %, e Ca+ in milliequivalent per 100 g of soil, f pH, g cation exchange capacity (CEC) in milliequivalent per 100 g of soil, h soil organic carbon (SOC) in %, i total nitrogen (TN) in % (Color figure online)

Full size image

Spatial distribution of soil ECa across the study vineyard

The soil ECa was periodically measured during the 2019 crop season and showed progressive decreases with time towards the harvest (Fig. 4). For deep soil, 52 days DAB, the northeastern section of the vineyard showed lower ECa values (Fig. 4a). The same spatial pattern was evident at 94 DAB, although there was a smaller portion in the southeastern section of vineyard having relatively higher ECa (Fig. 4b). At 129 DAB, the spatial pattern observed previously became not evident (Fig. 4c) and at 166 DAB, when harvest started, the central section showed lower ECa relative to the rest of the vineyard (Fig. 4d).

Fig. 4
figure 4

Soil apparent electrical conductivity (ECa) sensing by EM38 throughout the growing season in 2019. a DAB 52 deep ECa, b DAB 94 deep ECa, c DAB 129 deep ECa, d DAB 166 deep ECa, e DAB 52 shallow ECa, f DAB 94 shallow ECa, g DAB 129 shallow ECa, h DAB 166 shallow ECa. DAB days after bud-break (Fig. 4d and 4h previously appeared in Yu and Kurtural 2020) (Color figure online)

Full size image

In the shallow soil layer, a small portion of the southwestern section and the northeastern section of the vineyard had lower ECa values (Fig. 4e). The smaller portion in the northeastern section appeared to be confined within one or two vineyard rows. However, the irrigation was not applied to the vineyard before 63 DAB. Thus, the spatial pattern was derived exclusively from the soil conditions prior to the occurrence of the first irrigation. At 94 DAB, lower ECa values kept occurring southwestern section of the vineyard (Fig. 4f). And similar to what was observed in deep soil, the spatial pattern at 94 DAB became unclear on 129 DAB (Fig. 4g). On the 166 DAB, which corresponded with the harvest date, relatively lower ECa values were measured in the northwestern section of the vineyard compared to the rest of the vineyard (Fig. 4h).

Relationship between soil ECa and soil properties

Previous studies have observed a positive correlation between soil water content and soil ECa (or negative correlation with soil ERa) (Brillante et al. 2014). This was possibly among the reasons why the ECa values kept decreasing throughout the season at the study vineyard due to the progressive soil moisture depletion during the course of the hot and dry summer. Also, since the soil ECa is an integrated variable reflecting many individual soil characteristics, some previous research works indicated that other soil parameters can be used to predict its values (E. A. Costantini et al. 2010; Rodríguez-Pérez et al. 2011; Peralta and Costa 2013). A multivariate regression analysis was performed and a model was fitted with six soil variables, including soil organic carbon, gravel content, silt, clay, ECe, and water content, to the soil ECa assessed with EMI (Table 2). Such model could provide a prediction of ECa with R2 = 0.96 and p = 1.288e-13, degree of freedom = 21. The equation of the multivariate regression model is:

Table 2 Multivariate regression analysis of soil electrical conductivity based on soil predictors
Full size table
$$ {\text{EC}}_{{\text{a}}} = 8.669 + 1.844\;{\text{SOC + 1}}{\text{.363}}\;{\text{GR + 1}}{\text{.384}}\;{\text{S + 0}}{\text{.762}}\;{\text{EC}}_{{\text{e}}} - 0.699\;{\text{C + 0}}{\text{.127}}\;{\text{SWC}} $$
(5)

where SOC is soil organic carbon (converted from % to g), GR is gravel (converted from % to g), S silt (converted from % to g), ECe is soil pore water electrical conductivity (mS m −1), C is clay (converted from % to g), and SWC is soil water content (converted from vol % in volume to mm). These soil variables were considered in this model since they are relatively stable in the vineyard soils and there was no need for recurring (or periodic) measurements except ECe, and soil water content. Thus, this model worked well when transferring soil water content to soil ECa. When regressing the values of ECa estimated with this model and the ECa values observed in the field from EM38 readings, an r2 of 0.5770, p value of 0.0002, and root mean square error (RMSE) of 4.8349 mS m−1 were obtained (Fig. 5). When the analysis of relative importance was conducted, it was found that soil water content contributed the most and significantly to soil ECa, at 24% (Fig. 6). Soil texture was the second contributor to soil ECa, including silt with 22.4% and clay with 18.2% relative contributions, respectively. The soil–water ECe had the least contribution towards soil ECa among these six predictors with 5.5%. Some previous studies have shown that ECe might be a direct soil property affecting soil ECa, hence it can be assessed proximally (Corwin and Lesch 2013). This seemed not to be the case in this study due to the low soil–water ECe at the study site in Napa County, California (Weber et al. 2014). However, when ECe is more pronounced in the vineyard soils, there would be the need to adjust the model, as salinity may very likely have greater influence on soil ECa (E. A. Costantini et al. 2010). It was noted that specific soil physical conditions should be properly considered when relating soil water content to ECa (Morari et al. 2009). In this study, results provided evidence that the soil texture acted as the main soil variable on soil ECa, as it has strong influence on the soil moisture content. The proposed multivariate approach allowed taking into consideration various soil properties that are both spatially and temporally variable. The measurement of these parameters may be labor intensive and time consuming, and thus difficult to conduct during the dry and hot summers typically occurring in the study region. Given that the soil texture is stable over time, once the soil texture is determined, then the spatially-distributed plant water status could be easily inferred from the soil ECa with periodic EMI surveys. Furthermore, some physiological responses of the plant that are water-related may also be assessed and interpreted from soil ECa.

Fig. 5
figure 5

Predicted and observed values for the multivariate model with soil apparent electrical conductivity (ECa) from the soil predictors. Grey lines indicate 95% confidence interval, red line indicates the line with a slope of 1 and an intercept of 0 (Color figure online)

Full size image
Fig. 6
figure 6

Relative importance analysis of each predictor for the multivariate model deriving soil apparent electrical conductivity (ECa) by soil physical–chemical properties as predictors, including soil water content, silt, clay, SOC (soil organic carbon), gravel, and soil pore water electrical conductivity (ECe)

Full size image

Relationship between soil ECa and plant water status

The spatial interpolation for LAI and NDVI were performed, and the spatial homogeneity for these parameters was also tested (Fig. 7). The CI value for LAI was more than 75 with c0 of 1.017 and c0 + c1 close to 0; the CI value for NDVI was 90.48, which was also more than 75. Since the grapevine LAI and NDVI did not show spatial dependency, the plant water status can be assumed derived directly from soil. Linear relationships were investigated between ECa and plant water status expressed in terms of Ψstem and gs to evaluate the existence of direct relationships (Fig. 8). It was noticed that deep soil ECa was closely related to Ψstem with an r2 of 0.7214, p value of 0.0038, and RMSE of 0.1694 MPa (Fig. 8a). However, the same relationship was not evident for the shallow soil layer (Fig. 8b). A similarly close relationship was observed between deep soil ECa and gs with an r2 of 0.5007, p value of 0.0221, and RMSE of 56.5192 mmol H2O m−2 s −1. The higher the soil ECa became, the greater stomatal conductance was measured for the leaves (Fig. 8c). However, the relationship between ECa and gs was not as close for the shallow soil layer (Fig. 8d). The Ψstem and gs were the two factors that very well reflected the plant water status, and they were also interrelated. In the case of the study vineyard, the season-long Ψstem and gs integrals were regressed and indicated an r2 of 0.6559, p value of 0.0004, and RMSE of 20.9512 mmol H2O m−2 s −1 (Fig. 9). A conceptual model developed within an earlier study on wheat indicated that the shallow soil was likely to increase the level of root-sourced signals due to most roots distributed within or close to top soil (Whalley et al. 2008). In the study vineyard, the soil ECa measured for the soil depth between 0 and 1.5 m showed strong and positive correlations with Ψstem and gs, which was explained with the fact that most of the active roots were likely distributed within the top 0–1.5 m depth of the soil root zone.

Fig. 7
figure 7

Canopy assessments in a Cabernet Sauvignon (V. vinifera L.) vineyard in Oakville, California, USA. a Leaf area index (LAI), b normalized difference vegetation index (NDVI) (Color figure online)

Full size image
Fig. 8
figure 8

Correlation between soil apparent electrical conductivity (ECa) assessed by EM38, corrected to a reference temperature of 25 °C by time domain reflectometer (TDR) sensors, and stem water potential (Ψstem) and leaf stomatal conductance (gs) for 2019. a Ψstem and deep soil ECa, b Ψstem and shallow soil ECa, c gs and deep soil ECa, d gs and shallow ECa

Full size image
Fig. 9
figure 9

Relationships between stem water potential (Ψstem) integrals and stomatal conductance (gs) integrals in 2019

Full size image

Several other studies have evaluated the relations between the soil ECa (or ERa) and soil properties, which can help explaining plant water availability (Morari et al. 2009; Rodríguez-Pérez et al. 2011; Brillante et al. 2015). Similar relationships were observed in this study as those reported by previously. However, with the ability to measure two soil depths, these relationships can be elucidated with more accuracy. Most of the previous studies have applied soil ECa sensing to test its applicability in delineating management zones in commercial orchards and vineyards (Brevik et al. 2006; Bramley et al. 2011a, 2011b), but without exploring the linkages between ECa and plant physiology. Previous works have shown that the soil ECa (or ERa) assessed with proximal sensing had a positive relationship (or negative with ERa) with grapevine trunk circumference (Trought et al. 2008; Rossi et al. 2013). Similar positive relationships were also reported between soil ECa and grapevine pruning mass determined with active canopy reflectance sensors (Bramley et al. 2011b) and with vegetation indices (Zarrouk et al. 2012). However, those studies did not investigate the relationship between soil ECa and plant water status. Having positive correlation with deep soil ECa, NDVI also showed positive correlations with the applied water amount. Findings from these studies indicated that soil ECa could positively relate to the grapevine vegetative growth, which can vary across the field depending on access to water and nutrients and their uptake by plants, which in turn could be reflected in plant water status. Results presented herein provided evidence that soil ECa could provide relevant information of both grapevine vegetative growth and reproductive growth, including berry maturity in both primary and secondary metabolism (Bonfante et al. 2015; Yu et al. 2020), which can be used for water and nutrient management purposes. The study described in the present article provided evidence of the functional relationship between soil ECa and grapevine water uptake that could possibly be used to predict soil–plant water dynamics in commercial vineyards.

Relationship of δ13C with stem water potential and leaf photosynthesis

The must δ13C was closely and significantly correlated with plant water status, including season-long cumulative Ψstem integrals (Fig. 10a) with r2 = 0.9127, p value < 0.0001, and RMSE = 0.0316 MPa, and also with gs integrals (Fig. 10b) with r2 = 0.6985, p value = 0.0002, and RMSE = 19.6117 mmol H2O m−2 s−1, and also with An (Fig. 10c) with r2 = 0.5693, p value = 0.0018, and RMSE = 1.3489 µmol CO2 m−2 s−1. The relationship between δ13C and WUEi was not significant (Fig. 10d). Two earlier studies have already shown that δ13C may provide reliable predictions of plant water status as ancillary information for managing spatial variability (Martínez-Vergara et al. 2014; Brillante et al. 2016). Results from previous works indicated that soil characteristics and topography may influence δ13C based on modifications in the water continuum between soil and plants (Brillante et al. 2018). When δ13C to soil ECa were regressed in the current study, the relationships were not significant (p > 0.05), with r2 = 0.0015 with deep soil ECa and r2 = 0.1331 with shallow soil ECa (Fig. 11). This was also observed in a previous study where the correlation between δ13C and soil ECa was not significant within the same vineyard (E. A. Costantini et al. 2010). According to previous research, δ13C showed close relations with stomatal conductance, with δ13C values increasing with water deficits. Hence, δ13C can relate to net carbon assimilation rate as well (Farquhar et al. 1989; Brillante et al. 2018). One study showed that low nitrogen availability overruled the dominant effect of plant water status on δ13C (E. A. C. Costantini et al. 2013), so the relationship between plant water status and δ13C could be lost if the photosynthesis was severely impaired due to shortage of nutrient deficits. Bchir et al. (Bchir et al. 2016) showed that δ13C and plant water status were correlated in young and mature leaves as well as berries. Results of this work provided additional evidence of the relationship of δ13C with plant water status, including midday Ψstem and leaf photosynthesis during the berry maturation. Previous studies on δ13C were conducted in regions with relatively cool climate and rain-fed production systems (Bchir et al. 2016; Brillante et al. 2018). The present study can be fundamental for a region-wide calibration, being the first providing evidence that δ13C relates to plant water status and leaf photosynthesis in hot and dry region as California where the majority of grapevine is grown with micro irrigation.

Fig. 10
figure 10

Relationships between carbon isotope ratio analysis (δ13C) with season-long stem water potential and leaf gas exchange measurements. a Stem water potential integrals, Ψstem, b stomatal conductance integrals, gs, c net carbon assimilation integrals, An, d intrinsic water use efficiency integrals, WUEi

Full size image
Fig. 11
figure 11

Correlation between soil apparent electrical conductivity (ECa) assessed by EMI, and carbon isotope ratio analysis (δ13C). a Deep soil ECa, b shallow soil ECa

Full size image

Conclusion

Results presented here-in revealed the spatial variability of soil features across the studied vineyard. Soil ECa measured with proximal sensing in the deep soil layer can be related to grapevine water status. The main hypothesis of this study was confirmed that the soil water content and soil texture were the main factors influencing soil ECa in this study, which would further determine plant water status. Hence, once the vineyard soil is surveyed and its texture determined, the soil ECa can be used to infer spatial variations in plant water status. Additionally, the results provided evidence that δ13C analysis on berry sugar can potentially be used to back-track the spatial variability in plant water status and photosynthesis during berry maturation. These results provide relevant and applicable information to support the use of proximal soil sensing for monitoring soil–plant-water dynamics and changes in commercial production vineyards. Such information may help designing and implementing management strategies that minimize or manage the spatial variability of grapevine productivity and wine grape quality in micro irrigated commercial production vineyards.