Predicting leaf traits in wine grapes with reflectance spectroscopy

Rachel O. Mariani; Marney E. Isaac; Kimberley A. Cathline; Gavin Robertson; Adam R. Martin

doi:10.1371/journal.pone.0336560

Abstract

Estimating crop trait data is critical for predicting crop responses to environmental change, enabling more informed diagnoses of crop performance and the development of on-farm management strategies. Yet, many traditional methods for quantifying plant traits are time-consuming and resource-intensive, limiting sample sizes and study durations. In response, high-throughput phenotyping—specifically reflectance spectroscopy—has emerged as a key element of plant trait research, enabling rapid estimation of plant traits. However, little is known about whether reflectance spectroscopy can detect within-species variation in resource acquisition and plant-water traits, especially variation that exists among different cultivars or genotypes of the same crop. Using wine grapes (V. vinifera subsp. vinifera) as a focal crop, this study aimed to assess the ability of reflectance spectroscopy to quantify intraspecific variation in 12 leaf traits across 12 different cultivars from seven different varieties. We find significant variability in traits across and within cultivars, especially in gas-exchange and hydraulic traits, with cultivars varying along a resource-conservative-to-resource-acquisitive trait axis. Models based on spectral reflectance data were able to differentiate and predict this fine-scale trait variation among cultivars for seven plant traits, with a predictive power range of R² = 0.12–0.57. Models predicting leaf chemical (i.e., carbon and nitrogen concentrations), physiological (i.e., maximum rate of light-saturated photosynthesis), and morphological traits (i.e., leaf dry matter content) were more accurate in their predictions, while models predicting leaf water status were less accurate. Our results indicate that reflectance spectroscopy can capture certain dimensions of the fine-scale trait variation that exists within genetically diverse agroecosystems, though spectroscopic estimates of intraspecific variation in leaf water status are less accurate.

Citation: Mariani RO, Isaac ME, Cathline KA, Robertson G, Martin AR (2026) Predicting leaf traits in wine grapes with reflectance spectroscopy. PLoS One 21(5): e0336560. https://doi.org/10.1371/journal.pone.0336560

Editor: Thomas Roach, University of Innsbruck, AUSTRIA

Received: October 27, 2025; Accepted: April 20, 2026; Published: May 18, 2026

Copyright: © 2026 Mariani et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The full master dataset has been made publicly available without restriction via Borealis: The Canadian Dataverse Repository (doi:10.5683/SP3/XPBSL7) and includes the data used to reach the conclusions drawn in the manuscript and any additional data required to replicate the reported study findings.

Funding: Funding was from a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada, awarded to A.R.M. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Abbreviations: A_max, Maximum photosynthetic rate; C, Carbon; CF314, Cabernet Franc 314; CF327, Cabernet Franc 327; CS29, Cabernet Sauvignon 29; CS412, Cabernet Sauvignon 412; CV, Coefficient of variation; E, Transpiration; g_s, Stomatal conductance; HTP, High-throughput phenotyping; LDMC, Leaf dry matter content; LMA, Leaf mass per area; M181, Merlot 181; M348, Merlot 348; N, Nitrogen; PLSR, Partial least squares regression model; PN828, Pinot Noir 828; PN89, Pinot Noir 89; R171, Riesling 171; R23, Riesling 23; RSME, Root mean square error; SB908, Sauvignon blanc 906; V642, Viognier 642; VIP, Variable importance in projection; WUE_intr, Intrinsic water-use efficiency; WUE_inst, Instantaneous water-use efficiency; δ ¹³C, Stable isotopic carbon fraction; Ψ_pd, Predawn water potential

1 Introduction

In agricultural systems, environmental change driven by deviations in temperature and precipitation patterns is exerting profound impacts on crop growth and productivity [1,2]. While observations and predictions vary widely across crop species and geographies, recent studies have reported anywhere between 7–23% declines in crop yield [1]. Such changes largely represent the impact of fluctuations in precipitation and changes in growing and non-growing season temperatures on crop phenology, ecophysiology, and nutrient uptake [3]. For example, with gradual shifts towards overall warmer winters in many of Canada’s agricultural regions, crops are expected to face reduced cold-hardiness, resulting in increased susceptibility to spring frost damage to early vegetative growth, including primary buds [4,5]. These and other complex interactions between environmental impacts and individual plants scale up to influence crop performance, yield, and quality on farms (e.g., [2–4]).

Globally, wine grapes are one of the world’s most valuable crops [2]. In the province of Ontario, Canada—the region of our study—wine is among the highest-value-added agricultural products, generating a total economic impact of $5.49 billion in 2024−25 [6]. However, the suitability of the world’s wine-growing regions is expected to shift in the coming years and decades. Rising temperatures and prolonged droughts may hinder grape and wine production, making hot and dry regions less suitable for cultivation. Meanwhile, mid-latitude wine-growing regions could face a higher risk of spring frosts and increased precipitation, which are expected to impact plant phenology with negative consequences for production [2]. Wine grapes are also vulnerable to other aspects of environmental change, such as temperature variability, precipitation fluctuations, and extreme weather events, all of which can individually or collectively influence grapevine growth and fruit quality [5,7].

Projected shifts in climate pose significant challenges in viticulture. Predictions suggest that under a 2 °C global temperature increase scenario, up to 56% of all current wine-growing regions will become unsuitable for growing due to changes in grapevine phenology and grape composition, with this number increasing to 85% with a 4 °C increase by the end of the century [2,8,9]. Thus, wine growers now consider adaptation strategies to ensure consistency in production and quantity under variable environmental conditions [2,10]. More specifically, varietal selection and cultivar turnover are potentially key climate change adaptation strategies to ensure vine resilience to changes in growing regions. Cultivar turnover relies on varietal selection, where grapevine varieties are chosen based on their tolerance to certain environmental conditions, and are planted using various combinations of scion, rootstock, and vine training systems [11]. Wine grapes exhibit significant genomic and phenotypic diversity due to years of artificial selection, resulting in cultivars with traits that confer resistance or resilience towards environmental variability.

Plant responses to environmental change drivers can be multifaceted and governed by synchronized and/ or independent leaf, root, and whole-plant stress responses [1]. One key hypothesis in ecosystem science is that inter- and intra-specific variation in leaf functional traits—the characteristics of plants that determine responses to the surrounding environment—play a role in governing plant- and ultimately ecosystem-level responses to climate change [12–14]. Leaf traits related to the exchange of CO₂ and water, as well as a suite of other morphological, chemical, and physiological traits, are especially influenced by both environmental and genetic factors [11,15]. Generally, the literature on functional traits supports the expectation that greater functional trait variation leads to greater ecosystem resistance and/ or resilience to environmental change [16–18]. While earlier research in this area focused on characterizing the causes and consequences of trait variation across species, more recently, research suggests that higher intraspecific trait variation also confers stronger resistance and adaptation of plant communities to change [16].

Quantifying trait variation with crop species is particularly important in agroecosystems, where a limited number of plant species, varieties, or cultivars dominate and ultimately influence ecosystem functioning due to their high abundance [19]. However, many “traditional” methods for characterizing traits, especially those related to leaf-level CO₂ and water fluxes [20,21], are often time-consuming and resource-intensive as they rely on low-throughput tools such as infrared gas analyzers or Scholander-type pressure chambers. For example, pressure-volume curves, used to generate data on plant drought tolerance traits or plant water status, can take multiple hours to complete [16,20,21]. Gas exchange systems used to estimate physiological traits such as maximum photosynthetic capacity, stomatal conductance, or water-use efficiency are likewise time-consuming [22]. In response to such challenges associated with plant trait data collection, high-throughput phenotyping (HTP) has emerged as vital in plant trait research. Compared to “traditional” data collection methods, HTP tools can rapidly quantify several plant traits in a field setting [23].

At the ecosystem level, remote sensing is a key component of HTP. At the plant- or leaf-level, proximal reflectance spectroscopy is now more commonly being used to assess variation in traits across plant species [24,25], particularly those associated with the Leaf Economics Spectrum (LES), such as leaf mass per area (LMA), light-saturated photosynthetic capacity (A_max), and leaf nitrogen concentrations (N). Earlier research in this area primarily focused on employing reflectance spectroscopy to quantify trait variation across species, especially among plants of different functional types with diverse evolutionary histories [26]. Here, reflectance spectroscopy has proven effective in predicting trait variation and values, especially for leaf morphological and physiological traits, suggesting that this technique is suitable for rapid trait assessment in natural and managed ecosystems.

Less is known about the ability of reflectance spectroscopy to quantify finer-scale trait variation that exists within species [18,25,26]. For crops in agroecosystems, both genetic [18] and environmental factors [11] constrain or govern intraspecific trait variation. Characterizing differences in trait expression across cultivars of the same species is especially important in wine grapes, where individual fields and farms often cultivate multiple varieties (e.g., Cabernet sauvignon, Pinot noir) and cultivars of the same varieties (e.g., Pinot noir clones 89 and 828) [7]. A number of studies have quantified the considerable variation that exists among traits across wine grape varieties growing on the same vineyards [5,27–29]; yet, fewer studies have detailed the trait variation that exists among varieties and cultivars growing under otherwise similar environmental conditions. Similarly, research, including our own, has shown that reflectance spectroscopy predicts photosynthetic C assimilation traits, at least across wine grape varieties that differ widely in their domestication histories and, in turn, trait expression [30]. However, to our knowledge, no studies have tested whether or not reflectance spectroscopy can predict the (presumably) smaller differences in trait expression that exist within varieties, specifically among closely related cultivars.

Additionally, uncertainty remains regarding which specific traits can be predicted by spectroscopic techniques. Specifically, while numerous studies have employed reflectance spectroscopy to characterize plant traits related to leaf-level carbon gain, morphology, and stoichiometry—including our previous studies on wine grapes [30]—fewer have used this HTP approach to assess within-species variation in leaf water relations traits [26,31,32]. Leaf physiological traits, including transpiration (E), maximum photosynthetic capacity (A_max), and stomatal conductance (g_s), are central to understanding leaf- and plant water-use efficiency (WUE)—a major focus of this research—as high WUE is associated with enhanced crop performance under water-limited conditions [33,34].

More specifically, to the best of our knowledge, and despite their ecological and agronomic importance, limited studies have evaluated whether or not reflectance spectroscopy can predict leaf-level isotopic signatures associated with intrinsic water-use efficiency (δ ¹³C) or predawn water potential (Ψ_pd): two traits that are critical indicators of plant water relations and water-use strategies [35,36]. Both traits are key in assessing crop drought tolerance by providing measures of stomatal behaviour and plant water status, which regulate carbon assimilation, transpiration, and water-use efficiency under periods of environmental stress. In particular, δ ¹³C reflects long-term stomatal conductance, serving as a proxy to determine water-use efficiency over time [36]. Comparatively, Ψ_pd provides an estimate of baseline plant water status and informs an understanding of water limitation experienced by individual plants [35]. Using reflectance spectroscopy to estimate δ ¹³C and Ψ_pd could therefore be key to continuous monitoring of plant- and farm-level plant-water relations across growing seasons.

This study aims to characterize variation in leaf traits associated with plant-water relations and aboveground resource acquisition across multiple wine grape varieties and cultivars growing on the same farm, to assess the following: 1) Do wine grape varieties and cultivars differ in their leaf traits, specifically in their leaf morphological, chemical, physiological, and water-use traits? If so, 2) Can reflectance spectroscopy quantify this trait variation? We hypothesize that wine grape varieties and cultivars will exhibit significant differences in leaf traits, reflecting genetic variation (i.e., among cultivars), especially in physiological and water-use traits, which tend to vary more widely across varieties than morphological and/ or chemical traits [17,18,33,37,38]. Additionally, we hypothesize that more variable leaf traits, namely those associated with leaf physiological functioning, will be more accurately predicted by reflectance spectroscopy-based models than less variable morphological and chemical traits.

2 Materials and methods

2.1 Study design and trait collection

This study was conducted at the Niagara College Teaching Vineyard in Niagara-on-the-Lake, Ontario (43.146237 °N, −79.156810 °W). The study site is an operational teaching vineyard containing multiple wine grape cultivars (V. vinifera subsp. vinifera) distributed across the 16.2-ha farm in rows separated by either cover crops or unmanaged grasses. The soil at the vineyard is a mineral-rich mix of sand, gravel, loam, and red clay, creating an ideal substrate for wine grape production. At the site, 12 different wine grape cultivars belonging to seven different varieties of both red and white grapes are present, including Cabernet franc clones 327 and 314 (CF327 and CF314), Cabernet sauvignon clones 29 and 412 (CS29 and CS412), Merlot clones 384 and 181 (M348 and M181), Pinot noir clones 89 and 828 (PN89 and PN828), Riesling clones 23 and 171 (R23 and R171), Sauvignon blanc clone 906 (SB906), and Viognier clone 642 (V642). All cultivars are grown on a Selection Oppenheim (SO4) rootstock except Viognier 642, which is grown on the Millardet et de Grasset (101−14) rootstock. The vine training system is primarily a vertical shoot positioning system, most commonly used for varieties that thrive under improved sunlight exposure. The vineyard is not irrigated but tile-drained.

For each of the 12 cultivars, five vines in three different planting rows equally spaced along the northwest side of the vineyard were selected for functional trait characterization. Specifically, a total of n = 178 leaves were sampled from n = 15 vines per cultivar equally across n = 12 cultivars at the vineyard. On each vine, we selected one fully expanded, west-facing upper canopy leaf for reflectance spectroscopy and trait collection, such that all selected leaves were approximately the same size and free from any visible signs of damage. All upper canopy leaves were between 1 and 1.5 m in height at the time of sampling.

For each sample leaf, two main sets of in-field measurements were collected. First, physiological gas exchange data were collected using an LI-6800 portable photosynthesis system affixed with a 6800−03 large light source (LICOR Biosciences, Lincoln, Nebraska, USA). Specifically, for each leaf in our dataset the LI-6800 was used to measure stomatal conductance (g_s; mol m^-2 s^-1), light-saturated photosynthetic capacity (A_max; μmol CO₂ m^-2 s^-1), and transpiration (E; mmol H₂O m^-2 s^-1), under the following leaf chamber conditions: CO₂ concentrations of 420 ppm, saturating light levels of 1,500 μmol of photosynthetically active radiation (PAR) m^-2 s^-1 (with irradiance peaks in blue [453 nm], green [523 nm], and red [660 nm] wavebands), relative humidity (RH) of 50%, leaf vapour pressure deficits (VPD) of 1.7 MPa, and leaf temperatures of 25 °C (Macklin, et al., 2022). All physiological trait measurements were taken between 7:00 and 13:00 to ensure the leaves were not approaching stomatal closure. To ensure the leaves acclimatized to the leaf chamber conditions, we waited approximately 5 minutes after securing the sensor head to the leaf before taking measurements, and measurements were recorded only once leaves had stable A_max readings. Additionally, to ensure stomata were not closed during measurements, no measurements were taken if g_s fell below 0.06 mol^-2 s^-1. Physiological traits measured in the field were used to determine intrinsic water-use efficiency (WUE_intr; μmol CO₂ mol^-¹ H₂O) calculated as A_max/g_s, and instantaneous water-use efficiency (WUE_inst; μmol CO₂ mmol^-¹ H₂O) calculated as A_max/E.

Following gas-exchange measurements, we collected spectral reflectance data for each leaf using an SVC HR-1024i handheld spectroradiometer equipped with an LC-RP Pro leaf clip that included a calibrated internal light source (Spectra Vista Corporation, Poughkeepsie, New York, USA). The SVC HR1024i is a full-range spectroradiometer covering 350–2500 nm, with a spectral resolution of ≤ 3.5 nm (350–1000 nm), ≤ 9.5 nm (1000–1800 nm), and ≤ 6.5 nm (1800–2500 nm). Reflectance spectra were collected on the adaxial side of the leaves with an integration time of 2 s, and reference spectra were taken on a white Spectralon standard before each measurement. All reflectance spectroscopy measurements were taken immediately after predawn water potential measurements.

To determine pre-dawn water potential, all leaves sampled for physiological measurements and reflectance spectroscopy were removed from their vines before sunrise (between 03:30 and 05:00) the morning following gas exchange measurements (i.e., generally within 12–14 hours). Leaves were stored in darkness while maintaining leaf hydration by wrapping each sample in a saturated paper towel, sealing it in a polyethylene bag, and storing it in a cool, insulated bag that blocked all sunlight until they were successfully transported. Predawn water potential measurements were executed using a Portable Plant Water Potential Console (Soil Moisture Equipment Corp., Goleta, CA., USA). For each leaf, a clean blade was used to slice through the petiole across the horizontal axis, and the leaf was inserted into the chamber with a microscope camera aimed at the cut petiole to detect when sap was first exuded, at which point Ψ_pd was recorded.

Following measurements of Ψ_pd, all leaves were transported to the University of Toronto, Scarborough, Canada, for measurement of morphological and chemical traits. First, all leaves were weighed for leaf fresh mass (g) and scanned for fresh leaf area (cm²) using an LI-3100C leaf area meter (LICOR Biosciences, Lincoln, NE, USA). Subsequently, all leaves were dried at 60 °C and reweighed to determine dry mass (g), which was then used to calculate leaf dry matter content (LDMC) as mg of dry mass/ g of fresh mass. Following these steps, leaves were ground into fine powder using a MM40 Retsch ball mill (Retsch Ltd., Hann, Germany), and leaf N and C concentrations were determined on 0.102–0.156 g of leaf tissue using a LECO CN 628 elemental analyzer (LECO Instruments, Ontario, Canada). Then, LMA was calculated as leaf dry mass (g)/ leaf area (m²), and LDMC (mg g^-1) was calculated as leaf dry mass (mg)/ leaf saturated mass (g). Finally, we assessed δ ¹³C on ~1 mg of powdered leaf tissue using a Picarro G2131-i isotope and gas concentration analyser (Picarro Inc., Santa Clara, CA, USA).

2.2 Statistical analyses

We used R statistical software v. 4.4.3 (R Foundations for Statistical Computing, Vienna, Austria) for all data analysis. First, we calculated the summary statistics for all leaf traits and assessed normality using both a Shapiro–Wilk test and visual inspection with histograms and quantile-quantile plots, performed using the base R “stats” package. Then, variance component estimation was performed using the “nlme” R package [39,40] using a nested model approach with the following components: origin (originating from a warm or cool growing region), red/white varieties, cultivar, and planting row. We then performed an analysis of variance (ANOVA) with a Tukey’s Honestly Significant Difference post-hoc test to assess significant trait differences among cultivars.

Following these steps, we then performed analysis designed to test whether or not reflectance spectroscopy predicts leaf trait values using chemometrics, specifically partial least squares regression analysis (PLSR), following published protocols [24]. To do so, we first detected and removed outliers using the interquartile range (IQR) method with a factor of 2, as this approach is used for data with moderate to low variability [41]. Outliers were removed for the following leaf traits: leaf C (n = 1 outlier removed), δ ¹³C (n = 1 outlier removed), leaf area (n = 7 outliers removed), LDMC (n = 7 outliers removed), LMA (n = 7 outliers removed), and WUE_inst (n = 12 outliers removed). Only single trait values, and not entire leaves, were removed from the data set to maintain the largest possible sample sizes for each PLSR model (as per [24]).

We then applied the methodology outlined in [42] and [30] to assess whether reflectance spectroscopy predicts intraspecific trait variation using PLSR models. All PLSR models incorporated reflectance data across the 400–2400 nm wavelength range and were designed to predict raw and/ or log-transformed trait values (as per results from our normality tests). For each PLSR model, trait data were divided into a calibration dataset (80% of the data) and a validation set (the remaining 20%) using the “pls” and “spectrolab” R packages [43, 44]. Since we are especially interested in evaluating whether reflectance spectra can predict the smaller trait variation likely present across grape cultivars, we conducted our PLSR model analysis using two data-splitting approaches. First, we split the dataset by cultivar identity, ensuring that the calibration and validation sets contained approximately equal representation of trait and spectral data for each of the 12 cultivars. Second, we performed a fully randomized split, allowing the proportion of data from each cultivar to vary randomly.

For each calibration dataset, we then used the ‘find_optimal_components’ function in the ‘spectratrait’ R package to identify the optimal number of components used in each PLSR model [45], in conjunction with the minimum calculated prediction residual sum of squares (PRESS) statistic. Then, for each trait calibration dataset, PLSR models were fitted using a leave-one-out (LOO) cross-validation procedure (where the number of iterations is equal to n_cal-1), implemented using the ‘plsr’ function in the ‘pls’ R package [44]. Model performance was evaluated using the validation datasets as external test datasets, by comparing predicted values vs. observed values (again, in the validation data subset only). For each PLSR model fitted to the validation dataset, we used the coefficient of determination (R²), root mean squared error of prediction (RMSE), and percent RMSE (%RMSE) as the primary model evaluation criteria. Then, we analysed the regression coefficients and variable importance in projection (VIP) scores to identify which spectral regions significantly contributed to trait prediction. Additionally, we conducted a jackknife permutation analysis implemented in the “plsr” function in the “pls” R package [44] to assess model uncertainty. The jackknife coefficient estimates were then compared with those from the full model, and we calculated the mean predicted values along with 95% confidence and prediction intervals for each trait in the validation dataset. Lastly, we used a Pearson correlated test to evaluate the relationship between the prediction R² values from PLSR models and the extent of trait variation (as indicated by the CV for each trait).

3 Results

3.1 Trait variation across wine grape varieties and cultivars

Descriptive statistics related to trait variation across our entire dataset are presented in Table 1, alongside variance components attributed to variety origin, red/ white grape cultivars, cultivar identity, and planting row. Almost all traits evaluated in our analysis varied significantly across cultivars (ANOVA p < 0.01 in all cases) except for LMA (ANOVA p = 0.121; Fig 1). Leaf physiological and water-use traits varied most widely across cultivars, as indicated the largest coefficients of variation (CV). More specifically, plant water-use traits including E varied from 0.68–6.14 mmol H₂O m^-2 s^-1 (CV = 34.8%), g_s ranged from 0.1–0.47 mol m^-2 s^-1 (CV = 31.0%), and both WUE_inst and WUE_intr varied widely from 3.6–29.0 μmol CO₂ mmol^-1 H₂O (CV = 33.98%), and 44.90–174.94 μmol CO₂ mol^-1 H₂O (CV = 29.32%), respectively (Fig 1). Comparatively, the leaf chemical and morphological traits exhibited lower CV values. Specifically, traits expressing the lowest extent of variation across our dataset included leaf C (range = 42.57–47.54%, CV = 1.7%) and δ ¹³C (range = −30.86 to −22.66 ‰, CV = 4.4; Table 1. Generally, physiological traits related to leaf-level water fluxes, namely E, g_s, WUE_intr, WUE_inst, and Ψ_pd, varied more widely than morphological and chemical traits, with variance partitioning further indicating that these traits are especially variable/constrained across cultivars (Table 1).

Download:

Table 1. Descriptive statistics and variance partitioning for 12 leaf traits measured across 12 wine grape cultivars (n = 178).

https://doi.org/10.1371/journal.pone.0336560.t001

Download:

Fig 1. Trait variation of 178 individual plants, covering 12 cultivars of wine grapes.

Individual points (circles) represent data for individual plants, the boxes represent mean values per cultivar, and the bars represent standard errors.

https://doi.org/10.1371/journal.pone.0336560.g001

Specifically, for traits related to plant water relations, namely E and WUE_inst, the largest proportions of the variability in the trait data were explained by cultivar origin (0.42 and 0.59, respectively). Otherwise, cultivar identity explained the largest proportion of variation in g_s (0.54), Ψ_pd (0.21), WUE_intr (0.52), leaf C (0.32), and WUE_inst (0.29) (Table 1). Cultivar identity was the second-most important source of trait variation; however, a majority of the traits expressed a significant amount of ‘unexplained variance’, meaning a large portion of the variability of these traits that was not captured in our analysis or by any of our explanatory factors analyzed here. This suggests that trait values emerge from the interaction between cultivar identity and local environmental conditions; in our case, the high unexplained variance may reflect conditions associated with unmeasured soil conditions, water availability, or other environmental parameters.

Across cultivars gas exchange traits, namely E, A_max, and g_s, expressed statistically significant variation (Table 2, Fig 1). Specifically, across cultivars both Cabernet sauvignon cultivars (CS29, CS412) exhibited among the highest physiological trait values, with E rates of 3.5 ± 0.01 and 3.1 ± 0.01 mmol H₂O m^-2 s^-1, respectively; A_max values of 24.2 ± 0.8 and 24.3 ± 0.8 μmol CO₂ m^-2 s^-1, respectively; and g_s values of 0.34 ± 0.02 and 0.37 ± 0.01 mol m^-2 s^-1, respectively, highlighting that these clones generally express traits consistent with high rates of leaf-level resource aquistion. Other cultivars exhibiting high leaf-level resource acquisition, denoted by high values of include Riesling cultivars (R171 and R23) and Sauvignon blanc (SB906). Additionally, Merlot cultivars (M181 and M348), exhibited lower resource acquisition associated with reduced rates of E (2.90 ± 0.16 and 2.67 ± 0.12 mmol H₂O m^-2 s^-1, respectively), and A_max (20.0 ± 0.98 and 19.2 ± 1.07: μmol CO₂ m^-2 s^-1, respectively).

Download:

Table 2. Mean trait values measured across 12 wine grape cultivars with ± standard error (n ≈ 15 per cultivar).

https://doi.org/10.1371/journal.pone.0336560.t002

Chemical traits associated with leaf construction cost (leaf C), resource acquisition (leaf N), and plant-water relations (δ ¹³C) all varied little across cultivars (Fig 1). Here, leaf C exhibited the least variation (CV = 1.7) with cross-cultivar ranges of 47.5–42.6%, and 75% of the data fell between 44.9–46.7% dry mass. Riesling and Cabernet franc cultivars exhibited the highest leaf C concentrations, with values ranging from 45.6–46.8% dry mass. Leaf N exhibited higher variation (CV = 12.7), ranging from 1.6–3.6%, with Merlot cultivars exhibiting the highest mean leaf N concentrations (2.9%), compared to Riesling (cultivar 171), which had the lowest mean concentration of leaf N of 2.4% dry mass (Table 2). Additionally, for δ ¹³C, most cultivar mean values clustered around −28‰, with Riesling clone 23 presenting the lowest average (−26.83 ± 0.55‰), suggesting potentially lower long-term enhanced WUE. Compared to other traits except for leaf C values, δ ¹³C exhibited the most limited variation (CV = 4.4).

Morphological traits, including leaf area, LDMC, and LMA, exhibited greater variation among cultivars. Specifically, both Merlot cultivars (M181, M348) expressed on average the largest leaf area (167.91 ± 9.88 to 180.8 ± 8.61 cm²), while the Viognier cultivar (V642) and one Pinot noir cultivar (PN828) expressed the smallest average leaf areas (121.2 ± 5.3 and 134.99 ± 13.2 cm², respectively; Table 2). Traits associated with leaf construction, namely LDMC, was the highest in the Cabernet sauvignon (CS29 and CS412) cultivars (286.79 ± 23.82 and 280.46 ± 14.20 mg g^-1, respectively), with LMA following a similar trend, being highest in the CS cultivars along with PN828 (range = 73.2 ± 1.4 to 77.3 ± 3.7 g m^-2).

3.2 Reflectance spectra and PLSR models

The spectra quantiles for all 12 wine grape cultivars (total n = 178 leaves) indicated that, with the exception of two leaves (both removed due to measurement errors), leaves selected for our analysis here were free of major damage or underlying structural issues that might compromise leaf health (Fig 2). As presented in Table 3 and visualised in Fig 3, PLSR models parameterized under random data splits yielded, on average, more statistically significant components, which in turn supported PLSR models with stronger predictive capacity. Specifically, of the 12 traits analyzed in our PLSR framework, models fitted on fully randomized datasets were stronger (vs. those balanced across cultivars) in 7 of 11 instances (with one instance of PLSR models not converging).

Download:

Table 3. Partial least squares regression model fits predicting leaf trait variation across 12 wine grape cultivars. Different modelling frameworks denoted in the “Data Split” column, refer to manner in which data was divided into calibration (80%) and validation (20%) datasets for model fitting, where results from the “Cultivar” data split entail calibration/ validation datasets with approximately equal proportions of observations from all 12 wine grape cultivars, and the “Random” approach selected data for the calibration and validation datasets randomly.

https://doi.org/10.1371/journal.pone.0336560.t003

Download:

Fig 2. Spectra reflectance quantiles for 12 wine grape cultivars.

Key wavebands are highlighted by the grey bars. VIS: visible light spectrum (380–780 nm); NIR: near-infrared (780–1400 nm); SWIR1: shortwave infrared radiation 1 (1550–1750 nm); SWIR2: shortwave infrared radiation 2 (2000–2300 nm).

https://doi.org/10.1371/journal.pone.0336560.g002

Download:

Fig 3. Results of partial least squares regression (PLSR) models predicting 11 leaf functional traits in 12 wine grape cultivars.

Panels show the data points used to validate the models fitted to a set of calibration data, based on a fully randomized split (n_cal and n_val presented in Table 3). The black lines represent linear regression model fits between the observed and predicted traits, and the dotted grey lines present a 1:1 relationship. The bottom-right corner of each plot shows both the R² and the %RMSE for each model. Plots G–K present log-transformed trait data as shown in Table 3.

https://doi.org/10.1371/journal.pone.0336560.g003

Specifically, our PLSR models indicated that the traits most well predicted through models trained on fully randomized calibration and validation datasets were LDMC (R² = 0.57, %RMSE = 15.54), A_max (R² = 0.33, %RMSE = 17.22), and E (R² = 0.28, %RMSE = 20.12) (Fig 3). PLSR models with moderate-to-low R² included those fitted for leaf C and N (R² = 0.17, %RMSE = 21.71, and R² = 0.18, %RMSE = 23.05, respectively), Ψ_pd (R² = 0.12, %RMSE = 22.92), and log-Area (R² = 0.13, %RMSE = 25.63). Comparatively, PLSR models based on fully randomized data splits were least successful in predicting g_s and WUE_intr, as indicated by near-zero R² values. Moreover, models with the strongest fits under a randomized data-splitting structure (namely, LDMC, A_max, and E) tended to be traits with high variability (as per Tables 1 and 2), and those functionally interconnected through their roles in carbon assimilation and water use. The PLSR model was not effective in predicting δ ¹³C due to the reduced variability exhibited within-cultivars, and regardless of the data-splitting approach or alternative component selection methods, there was an optimal component output of 0, indicating that there is no latent component that explains the covariance between our predictor and response variables [46].

When calibration and validation datasets contained equal proportions of data from each cultivar, PLSR models predicting leaf morphological traits (log-Area, log-LMA) expressed stronger predictive capacity with R² values of 0.26 (%RMSE = 20.11) and 0.29 (%RMSE = 16.18), respectively. This was also the case for a PLSR model predicting leaf N (R² = 0.35, %RMSE = 21.19). This likely owes to leaf morphological traits varying more widely among cultivars as they are strongly regulated by leaf genetics (for leaf shape and overall size), with environmental factors influencing their plasticity [47]. However, PLSR models based on balanced calibration and validation datasets across cultivars were not stronger for E, A_max, leaf C, Ψ_pd, WUE_inst, and LDMC. Instead, these traits are more robustly predicted by PLSR models when the training and validation datasets were generated using a randomized data split (R² = 0.28, 0.33, 0.17, 0.23, and 0.57, respectively). When comparing model performance across the CVs for each trait, there was a strong inverse correlation with the Cultivar data-splitting approach and a weaker inverse correlation with the Random data-splitting approach. To confirm the influence of our data splitting approaches, we performed a Peason correlated test between prediction R² values and the degree of variation for each trait (CV). For both data splitting approaches, there were negative r values, with −0.55 for Cultivar splits, and −0.22 for the Random split, meaning that as CV increases, R² decreases more strongly for Cultivar splits, which is evident in the more plastic physiological traits.

Based on the VIP scores dervived from our PLSR modelling procedure for traits with moderate R² with randomized data splits, our analysis highlighted the strong influence of wave bands in the the red-edge (680−750 nm) and near-infrared region (780−1400 nm) in predicting physiological traits including E, A_max, as well as for two water relation traits (Ψ_pd and WUE_inst), where VIP scores were > 1, and in multiple instances approaching 2 (Fig 4). We also detected a strong influence of wave bands in the visible light region (400−750 nm) in predicting leaf C concentrations, and LDMC (VIP scores > 1.5), with wavebands in the short-wave infrared-2 region (2,000−2,290 nm) also having a strong influence on PLSR models predicting LDMC and LMA (VIP > 1.3). However, VIPs associated with PLSR models predicting leaf N concentrations did not follow the expected patterns, namely VIP < 1 in the SWIR1 region (1,550−1,750 nm), potentially due to the limited variability in our leaf N data, leading to reduced ability to capture trait variation commonly associated with reflectance in this region. While the predictive power of our PLSR models depended on the data-splitting procedure and parameterization, VIP scores were not sensitive to the configuration of this particular model-fitting procedure.

Download:

Fig 4. Variable importance in projection (VIP) scores associated with partial least squares regression (PLSR) models for calibration data using a “random” split.

The dashed line marks the VIP score threshold of 0.8, indicating the cutoff for identifying the most important wavebands for trait predictions; values ≥ 1 highlight more influential variables.

https://doi.org/10.1371/journal.pone.0336560.g004

4 Discussion

This study contributes to the growing body of research demonstrating the application of reflectance spectroscopy and PLSR modelling for quantifying trait variation that exists among crop cultivars, and within plant species more broadly. Previous work has established this approach to trait estimation across different species [25,26,42], with more recent studies exploring its application in crops, including wine grapes. Specifically, our previous work focused on predicting five leaf traits associated with aboveground resource capture (e.g., photosynthetic carbon assimilation, RuBisCO carboxylation, and electron transport) across varieties, but not cultivars of the same variety [30]. Here, we contribute new findings that traits related to leaf-level physiology (E, A_max, g_s), plant-water relations (Ψ_pd, δ ¹³C, WUE_intr, WUE_inst), stoichiometry (leaf C and N), and leaf construction (leaf area, LMA, LDMC) vary across cultivars within varieties (Fig 1, Tables 1 and 2), and that reflectance spectroscopy can indeed predict this finer-scale trait variation (Fig 3, Table 3). The importance of understanding and quantifying such resolved trait variation as a driver for plant- and ecosystem-level resilience [18,48] has gained increased attention, motivating an interest in using reflectance spectroscopy and PLSR modelling for functional trait characterisation, especially in agroecosystems and crop breeding programs [25]. Here, we explored the extent and sources of functional trait variation across a diverse genetic gradient of wine grape cultivars using both field-based research methods and leaf-level spectral data.

Our findings demonstrate that plant traits varied across cultivars, indicating a strong influence of wine grape genotypes on phenotypic trait expression. This variation was most evident in leaf physiological (E, A_max, g_s) and water-use traits (WUE_inst, WUE_intr), with less pronounced variation in leaf chemical and morphological traits, specifically in leaf C and N concentrations, LMA, and LDMC. Two water relations traits, namely Ψ_pd and δ ¹³C, also exhibited limited variation across cultivars. Similar patterns of more pronounced variability in gas exchange-based physiological traits and water-use traits, vs. more constrained variation in structural and chemical traits, have been documented in other single-species studies [49–52].

Also in line with previous studies on wine grape varieties and other plant species [24,30,39,53,54], we found that the red-edge spectral bands in the ~ 680–750 nm range exert the strongest influence on trait estimation for a majority of traits measured here, including A_max, leaf C and N, Ψ_pd, WUE_inst, LDMC, and leaf area. In all of these cases, VIP scores for red-edge were > 2, such that the red-edge region clearly emerged as one of the more important wavebands for PLSR modelling [24,55]. Although WUE_intr and WUE_inst are not directly optical traits, they are indicative of stomatal relations and therefore covary with other resource-use traits captured by red-edge bands [38]. These findings indicate that reflectance spectroscopy has the potential to capture subtle and cultivar-specific spectral signatures related to stress responses, particularly in traits with strong optical properties (i.e., physiological, stoichiometric, and structural traits). However, VIP scores are sensitive to model performance, meaning their interpretation is viable and biologically-meaningful only if PLSR model fits were “successful” [42].

Consistent with previous studies as well, we find that reflectance profiles in the lower SWIR range wavebands (i.e., SWIR 1 in the 1550–1750 nm range) is critically important in predicting physiological (A_max) and water-use traits, as well as LDMC. The SWIR 1 region contains strong water absorption features, making it highly sensitive to variation in leaf water content [31]. Reflectance values in this region also provide an indirect signal to water-use traits since leaf water status is strongly associated with leaf physiological processes [39,40]. The ability to detect water-stress traits in wine grapes is an essential advancement in applications of reflectance spectroscopy towards crop ecophysiology research and practice, since traditional methods for water-stress analysis remain time-consuming. Detecting water status through proximal sensing, like reflectance spectroscopy, can also ensure closely related cultivars with differing water statuses receive more targeted watering regimes.

In our analysis, we found that PLSR models were most effective in predicting three physiological and morphological leaf traits with a predictive power ranging from 28% to 52%, namely E, A_max, and LDMC. Additionally, our PLSR models were successful, albeit more moderately, in predicting leaf C and N concentrations, Ψ_pd, and leaf area. Compared with other single-species studies [24,30,31] our PLSR models were less precise in their trait predictions. For example, in similar studies, the predictive power of the PLSR models for leaf physiological traits based on gas exchange across only six wine grape varieties was much higher (i.e., 18–62%) [30]. Our findings here highlight broader potential to associated functional traits and physiological status with reflectance spectra signatures, potentially addressing a limitation to the degree of variation that reflectance spectroscopy can estimate. However, in multi-cultivar agroecosystems like vineyards, where multiple cultivars often coexist in a single environment, our results suggest reflectance spectroscopy is a useful tool for detecting certain cultivar-specific stress responses.

One potential explanation for the lower predictive power observed in this study is the reduced variability of several leaf traits across our entire dataset, at the cultivar level. Compared to our previous work [30] the magnitude of trait variation for certain traits was lower in our study here. For example, the CVs for leaf N and LMA in previous studies on wine grapes with fewer cultivars were 13.7% and 12.8%, respectively [30], as compared to CV values of 12.6% and 9.6% for these same traits observed here. One expectation might be that this reduced trait variability reduces PLSR model performance, while the relative trends in predictive power among the different data-splitting approaches remained consistent. In contrast, comparing variability in A_max between studies%, CV values were nearly identical (~34.8). Accordingly, the predictive performance of PLSR models was nearly identical between both our and comparable studies [30].

The predictive power of PLSR models relies on the ability of models to identify covariance structures between predictor variables and response traits. Therefore, if trait variation is more limited (i.e., a lower CV), one might expect the strength of covariance patterns to be weaker, leading to less clear/weaker trait predictions [26]. The patterns found in our study generally support the interpretation that PLSR model performance is influenced by the magnitude of trait variation within a given dataset. However, model performance is not solely constrained by this relationship, and less direct spectral-trait interactions more strongly influence model performance [56].

By extension, the ability of PLSR models to predict traits across wine grape cultivars depended on data-splitting parameterization. Fully randomizing calibration and validation datasets generally led to higher explanatory power (R² and %RMSE values) across 7 of the 11 traits successfully modelled here. When trait variation across a dataset is higher, non-random data splits may not capture the entirety of trait variation, and in turn provide less reliable performance metrics. More specifically, traits that are A) highly plastic across environmental conditions and B) less constrained within genotypes (or other factors)—such as A_max—may be more accurately modelled using a randomized split. This is because a randomized approach would be expected to better capture the entire breadth of trait values, whereas a targeted data split (i.e., balanced across genotypes) is more likely to miss the entire extent of trait distributions. Under this scenario, PLSR models fitted to validation data may be less strong in their predictive ability [56–58]. Thus, our results suggest that ahead of model calibration, data splits need to be assessed based on the general variability of the training data and how this variation is structured [26,42]. Data-splitting strategies can also yield different outcomes depending on the trait being assessed; traits with more pronounced genetic specificity may be more accurately modelled using a cultivar split (i.e., area).

5 Conclusions

Our study contributes new data demonstrating that functional traits vary significantly among and within across wine grape cultivars, with cultivars of the same variety expressing considerable variation in leaf traits related to physiology, plant water relations, leaf tissue chemistry, and leaf morphology. Physiological and water-use traits are particularly variable among closely related cultivars, highlighting their potential use for selecting resilient cultivars for breeding programs and climate adaptation strategies. We also contribute among the first set of results demonstrating that reflectance spectroscopy can detect and estimate this fine-scale intraspecific trait variation, which implies strong potential to scale non-destructive trait monitoring from individual plants to large-scale agricultural systems. However, our results also show that predictive models are sensitive to model parameterization—specifically decisions surrounding calibration and validation dataset construction—in a manner that is associated with trait variation: reflectance-based trait models generally perform better when wider trait variation is present in both training and validation data. Taken together, our analysis highlights the variation in functional traits across genetic gradients and contributes to the growing body of research exploring reflectance-based approaches to quantify intraspecific variation in ecophysiological traits of plants and crops [14,18,30,59].

Acknowledgments

The authors acknowledge Emily Young and Kale Vicario for their assistance with field data collection, and Camila Villa-Escobar and Rachael Harman-Denhoed for their assistance with the sample preparation and processing.

References

1. Rezaei EE, Webber H, Asseng S, Boote K, Durand JL, Ewert F, et al. Climate change impacts on crop yields. Nat Rev Earth Environ. 2023;4(12):831–46.
- View Article
- Google Scholar
2. van Leeuwen C, Sgubin G, Bois B, Ollat N, Swingedouw D, Zito S, et al. Climate change impacts and adaptations of wine production. Nat Rev Earth Environ. 2024;5(4):258–75.
- View Article
- Google Scholar
3. Seleiman MF, Al-Suhaibani N, Ali N, Akmal M, Alotaibi M, Refay Y, et al. Drought stress impacts on plants and different approaches to alleviate its adverse effects. Plants (Basel). 2021;10(2):259. pmid:33525688
- View Article
- PubMed/NCBI
- Google Scholar
4. Qian B, Jing Q, Bélanger G, Jégo G, Smith W, VanderZaag A, et al. Projected changes in risks of winter damage to fruit trees and plant hardiness zones in Canada. Can J Plant Sci. 2025;105:1–17.
- View Article
- Google Scholar
5. Jones GV, Reid R, Vilks A. Climate, grapes, and wine: structure and suitability in a variable and changing climate. In: Dougherty PH, editor. The geography of wine: regions, terroir and techniques. Dordrecht: Springer Netherlands; 2012. p. 109–33.
6. Annual Report 2024/2025. Wine Growers Ontario. https://www.winegrowersontario.ca/annual-reports/
7. Keller M. Climate change impacts on vineyards in warm and dry areas: challenges and opportunities: from the asev climate change symposium part 1 – viticulture. Am J Enol Vitic. 2023;74(2).
- View Article
- Google Scholar
8. Morales-Castilla I, García de Cortázar-Atauri I, Cook BI, Lacombe T, Parker A, van Leeuwen C, et al. Diversity buffers winegrowing regions from climate change losses. Proc Natl Acad Sci U S A. 2020;117(6):2864–9. pmid:31988113
- View Article
- PubMed/NCBI
- Google Scholar
9. Tebaldi C, Debeire K, Eyring V, Fischer E, Fyfe J, Friedlingstein P, et al. Climate model projections from the Scenario Model Intercomparison Project (ScenarioMIP) of CMIP6. Earth Syst Dyn. 2021;12(1):253–93.
- View Article
- Google Scholar
10. Bush E, Lemmen DS. Canada’s changing climate report. Government of Canada; 2019. https://ostrnrcan-dostrncan.canada.ca/handle/1845/143725
- View Article
- Google Scholar
11. Baltazar M, Castro I, Gonçalves B. Adaptation to climate change in viticulture: the role of varietal selection—a review. Plants. 2025;14(1):1.
- View Article
- Google Scholar
12. Guo A, Zuo X, Zhang S, Hu Y, Yue P, Lv P, et al. Contrasting effects of plant inter- and intraspecific variation on community trait responses to nitrogen addition and drought in typical and meadow steppes. BMC Plant Biol. 2022;22(1):90. pmid:35232383
- View Article
- PubMed/NCBI
- Google Scholar
13. Miedema Brown L, Anand M. Plant functional traits as measures of ecosystem service provision. Ecosphere. 2022;13(2).
- View Article
- Google Scholar
14. Violle C, Enquist BJ, McGill BJ, Jiang L, Albert CH, Hulshof C, et al. The return of the variance: intraspecific variability in community ecology. Trends Ecol Evol. 2012;27(4):244–52. pmid:22244797
- View Article
- PubMed/NCBI
- Google Scholar
15. Demmings EM, Williams BR, Lee C-R, Barba P, Yang S, Hwang C-F, et al. Quantitative trait locus analysis of leaf morphology indicates conserved shape loci in grapevine. Front Plant Sci. 2019;10:1373. pmid:31803199
- View Article
- PubMed/NCBI
- Google Scholar
16. Martin AR, Li G, Cui B, Mariani RO, Vicario K, Cathline KA, et al. A high-throughput approach for quantifying turgor loss point in grapevine. Plant Methods. 2024;20(1):180. pmid:39581961
- View Article
- PubMed/NCBI
- Google Scholar
17. Nimmo V, Violle C, Entz M, Rolhauser AG, Isaac ME. Changes in crop trait plasticity with domestication history: management practices matter. Ecol Evol. 2023;13(11):e10690. pmid:38020689
- View Article
- PubMed/NCBI
- Google Scholar
18. Westerband AC, Funk JL, Barton KE. Intraspecific trait variation in plants: a renewed focus on its role in ecological processes. Ann Bot. 2021;127(4):397–410. pmid:33507251
- View Article
- PubMed/NCBI
- Google Scholar
19. Martin AR, Isaac ME. Plant functional traits in agroecosystems: a blueprint for research. J Appl Ecol. 2015;52(6):1425–35.
- View Article
- Google Scholar
20. Bartlett MK, Scoffoni C, Ardy R, Zhang Y, Sun S, Cao K, et al. Rapid determination of comparative drought tolerance traits: using an osmometer to predict turgor loss point. Methods Ecol Evol. 2012;3(5):880–8.
- View Article
- Google Scholar
21. Banks JM, Hirons AD. Alternative methods of estimating the water potential at turgor loss point in Acer genotypes. Plant Methods. 2019;15:34. pmid:30988693
- View Article
- PubMed/NCBI
- Google Scholar
22. Saathoff AJ, Welles J. Gas exchange measurements in the unsteady state. Plant Cell Environ. 2021;44(11):3509–23. pmid:34480484
- View Article
- PubMed/NCBI
- Google Scholar
23. Pabuayon ILB, Sun Y, Guo W, Ritchie GL. High-throughput phenotyping in cotton: a review. J Cotton Res. 2019;2(1).
- View Article
- Google Scholar
24. Ely KS, Burnett AC, Lieberman-Cribbin W, Serbin SP, Rogers A. Spectroscopy can predict key leaf traits associated with source-sink balance and carbon-nitrogen status. J Exp Bot. 2019;70(6):1789–99.
- View Article
- Google Scholar
25. Kim M, Lee C, Hong S, Kim SL, Baek JH, Kim KH. High-throughput phenotyping methods for breeding drought-tolerant crops. Int J Mol Sci. 2021;22(15):8266.
- View Article
- Google Scholar
26. Kothari S, Beauchamp‐Rioux R, Laliberté E, Cavender‐Bares J. Reflectance spectroscopy allows rapid, accurate and non‐destructive estimates of functional traits from pressed leaves. Methods Ecol Evol. 2022;14(2):385–401.
- View Article
- Google Scholar
27. Macklin SC, Mariani RO, Young EN, Kish R, Cathline KA, Robertson G, et al. Intraspecific leaf trait variation across and within five common wine grape varieties. Plants (Basel). 2022;11(20):2792. pmid:36297816
- View Article
- PubMed/NCBI
- Google Scholar
28. Bigard A, Berhe DT, Maoddi E, Sire Y, Boursiquot JM, Ojeda H. Vitis vinifera L. fruit diversity to breed varieties anticipating climate changes. Front Plant Sci. 2018;9(455).
- View Article
- Google Scholar
29. P D, G F, G M, T D. Performance of interspecific grapevine varieties in north-east Italy. Agric Sci. 2013;4(2):91–101.
- View Article
- Google Scholar
30. Cui B, Mariani R, Cathline KA, Robertson G, Martin AR. Reflectance spectroscopy predicts leaf functional traits across wine grape cultivars. Plants People Planet. 2025;7(5):1524–37.
- View Article
- Google Scholar
31. Cotrozzi L, Peron R, Tuinstra MR, Mickelbart MV, Couture JJ. Spectral phenotyping of physiological and anatomical leaf traits related with maize water status. Plant Physiol. 2020;184(3):1363–77. pmid:32907885
- View Article
- PubMed/NCBI
- Google Scholar
32. Novick KA, Ficklin DL, Baldocchi D, Davis KJ, Ghezzehei TA, Konings AG, et al. Confronting the water potential information gap. Nat Geosci. 2022;15(3):158–64. pmid:35300262
- View Article
- PubMed/NCBI
- Google Scholar
33. Lv Y, Gu L, Man R, Liu X, Xu J. Response of stomatal conductance, transpiration, and photosynthesis to light and CO2 for rice leaves with different appearance days. Front Plant Sci. 2024;15:1397948. pmid:39148618
- View Article
- PubMed/NCBI
- Google Scholar
34. Martin AR, Rapidel B, Roupsard O, Van den Meersche K, de Melo Virginio Filho E, Barrios M. Intraspecific trait variation across multiple scales: the leaf economics spectrum in coffee. Funct Ecol. 2017;31(3):604–12.
- View Article
- Google Scholar
35. Charrier G. Extrapolating physiological response to drought through step-by-step analysis of water potential. Plantae. 2020;184(2):News and Views.
- View Article
- Google Scholar
36. Ma WT, Yu YZ, Wang X, Gong XY. Estimation of intrinsic water-use efficiency from δ13C signature of C3 leaves: assumptions and uncertainty. Front Plant Sci. 2023;13.
- View Article
- Google Scholar
37. Martin A, Miller L, Madigan M, Cui B, Cathline K, Robertson G. Functional Ecology. 2025;39:1112–27.
- View Article
- Google Scholar
38. Pan S, Wang X, Yan Z, Wu J, Guo L, Peng Z, et al. Leaf stomatal configuration and photosynthetic traits jointly affect leaf water use efficiency in forests along climate gradients. New Phytol. 2024;244(4):1250–62. pmid:39223910
- View Article
- PubMed/NCBI
- Google Scholar
39. Zhang Y, Wu J, Wang A. Comparison of various approaches for estimating leaf water content and stomatal conductance in different plant species using hyperspectral data. Ecol Indic. 2022;142:109278.
- View Article
- Google Scholar
40. Le TS, Harper R, Dell B. Application of remote sensing in detecting and monitoring water stress in forests. Remote Sensing. 2023;15(13):3360.
- View Article
- Google Scholar
41. Dastjerdy B, Saeidi A, Heidarzadeh S. Review of applicable outlier detection methods to treat geomechanical data. Geotechnics. 2023;3(2):375–96.
- View Article
- Google Scholar
42. Burnett AC, Anderson J, Davidson KJ, Ely KS, Lamour J, Li Q. A best-practice guide to predicting plant traits from leaf-level hyperspectral data using partial least squares regression. J Exp Bot. 2021;72(18):6175–89.
- View Article
- Google Scholar
43. Meireles JE, Schweiger AK, Cavender-Bares J. Spectrolab: class and methods for spectral data. 2025. [cited 2025 Oct 20]. https://cran.r-project.org/web/packages/spectrolab/index.html
44. Liland KH, Mevik BH, Wehrens R, Hiemstra P. pls: partial least squares and principal component regression. [cited 2025 Oct 20]. https://cran.r-project.org/web/packages/pls/index.html
45. Lamour J, Serbin S. Spectratrait: a simple add-on package to aid in the fitting of leaf or canopy scale spectra-trait PLSR models. 2025. https://github.com/plantphys/spectratrait
46. Liland KH, Mevik BH, Wehrens R. Partial least squares and principal component regression. 1999. https://CRAN.R-project.org/package=pls
47. Jessica G, Byrne ME. From genes to climate: a perspective on the importance of leaf shape. J Exp Bot. 2026;77(2).
- View Article
- Google Scholar
48. Martin AR, Isaac ME. Functional traits in agroecology: advancing description and prediction in agroecosystems. J Appl Ecol. 2018;55(1):5–11.
- View Article
- Google Scholar
49. Faralli M, Bontempo L, Bianchedi PL, Moser C, Bertamini M, Lawson T, et al. Natural variation in stomatal dynamics drives divergence in heat stress tolerance and contributes to seasonal intrinsic water-use efficiency in Vitis vinifera (subsp. sativa and sylvestris). J Exp Bot. 2022;73(10):3238–50. pmid:34929033
- View Article
- PubMed/NCBI
- Google Scholar
50. Easlon HM, Nemali KS, Richards JH, Hanson DT, Juenger TE, McKay JK. The physiological basis for genetic variation in water use efficiency and carbon isotope composition in Arabidopsis thaliana. Photosynth Res. 2014;119(1–2):119–29. pmid:23893317
- View Article
- PubMed/NCBI
- Google Scholar
51. Tortosa I, Escalona JM, Opazo I, Douthe C, Medrano H. Genotype variations in water use efficiency correspond with photosynthetic traits in tempranillo grapevine clones. Agronomy. 2022;12(8):1874.
- View Article
- Google Scholar
52. Bota J, Tomás M, Flexas J, Medrano H, Escalona JM. Differences among grapevine cultivars in their stomatal behavior and water use efficiency under progressive water stress. Agricultural Water Management. 2016;164:91–9.
- View Article
- Google Scholar
53. Serbin SP, Singh A, Desai AR, Dubois SG, Jablonski AD, Kingdon CC, et al. Remotely estimating photosynthetic capacity, and its response to temperature, in vegetation canopies using imaging spectroscopy. Remote Sensing of Environment. 2015;167:78–87.
- View Article
- Google Scholar
54. Hunt ER, Rock BN. Detection of changes in leaf water content using near- and middle-infrared reflectances. Remote Sens Environ. 1989;30(1):43–54.
- View Article
- Google Scholar
55. Curran PJ, Dungan JL, Gholz HL. Exploring the relationship between reflectance red edge and chlorophyll content in slash pine. Tree Physiol. 1990;7(1_2_3_4):33–48. pmid:14972904
- View Article
- PubMed/NCBI
- Google Scholar
56. Ji F, Li F, Hao D, Shiklomanov AN, Yang X, Townsend PA, et al. Unveiling the transferability of PLSR models for leaf trait estimation: lessons from a comprehensive analysis with a novel global dataset. New Phytol. 2024;243(1):111–31. pmid:38708434
- View Article
- PubMed/NCBI
- Google Scholar
57. Kothari S, Beauchamp-Rioux R, Blanchard F, Crofts AL, Girard A, Guilbeault-Mayers X, et al. Predicting leaf traits across functional groups using reflectance spectroscopy. New Phytol. 2023;238(2):549–66. pmid:36746189
- View Article
- PubMed/NCBI
- Google Scholar
58. Xu Y, Goodacre R. On splitting training and validation set: a comparative study of cross-validation, bootstrap and systematic sampling for estimating the generalization performance of supervised learning. J Anal Test. 2018;2(3):249–62. pmid:30842888
- View Article
- PubMed/NCBI
- Google Scholar
59. Reich PB, Wright IJ, Cavender‐Bares J, Craine JM, Oleksyn J, Westoby M, et al. The evolution of plant functional variation: traits, spectra, and strategies. International Journal of Plant Sciences. 2003;164(S3):S143–64.
- View Article
- Google Scholar

[ref1] 1. Rezaei EE, Webber H, Asseng S, Boote K, Durand JL, Ewert F, et al. Climate change impacts on crop yields. Nat Rev Earth Environ. 2023;4(12):831–46.
View Article
Google Scholar

[2] View Article

[3] Google Scholar

[ref2] 2. van Leeuwen C, Sgubin G, Bois B, Ollat N, Swingedouw D, Zito S, et al. Climate change impacts and adaptations of wine production. Nat Rev Earth Environ. 2024;5(4):258–75.
View Article
Google Scholar

[5] View Article

[6] Google Scholar

[ref3] 3. Seleiman MF, Al-Suhaibani N, Ali N, Akmal M, Alotaibi M, Refay Y, et al. Drought stress impacts on plants and different approaches to alleviate its adverse effects. Plants (Basel). 2021;10(2):259. pmid:33525688
View Article
PubMed/NCBI
Google Scholar

[8] View Article

[9] PubMed/NCBI

[10] Google Scholar

[ref4] 4. Qian B, Jing Q, Bélanger G, Jégo G, Smith W, VanderZaag A, et al. Projected changes in risks of winter damage to fruit trees and plant hardiness zones in Canada. Can J Plant Sci. 2025;105:1–17.
View Article
Google Scholar

[12] View Article

[13] Google Scholar

[ref5] 5. Jones GV, Reid R, Vilks A. Climate, grapes, and wine: structure and suitability in a variable and changing climate. In: Dougherty PH, editor. The geography of wine: regions, terroir and techniques. Dordrecht: Springer Netherlands; 2012. p. 109–33.

[ref6] 6. Annual Report 2024/2025. Wine Growers Ontario. https://www.winegrowersontario.ca/annual-reports/

[ref7] 7. Keller M. Climate change impacts on vineyards in warm and dry areas: challenges and opportunities: from the asev climate change symposium part 1 – viticulture. Am J Enol Vitic. 2023;74(2).
View Article
Google Scholar

[17] View Article

[18] Google Scholar

[ref8] 8. Morales-Castilla I, García de Cortázar-Atauri I, Cook BI, Lacombe T, Parker A, van Leeuwen C, et al. Diversity buffers winegrowing regions from climate change losses. Proc Natl Acad Sci U S A. 2020;117(6):2864–9. pmid:31988113
View Article
PubMed/NCBI
Google Scholar

[20] View Article

[21] PubMed/NCBI

[22] Google Scholar

[ref9] 9. Tebaldi C, Debeire K, Eyring V, Fischer E, Fyfe J, Friedlingstein P, et al. Climate model projections from the Scenario Model Intercomparison Project (ScenarioMIP) of CMIP6. Earth Syst Dyn. 2021;12(1):253–93.
View Article
Google Scholar

[24] View Article

[25] Google Scholar

[ref10] 10. Bush E, Lemmen DS. Canada’s changing climate report. Government of Canada; 2019. https://ostrnrcan-dostrncan.canada.ca/handle/1845/143725
View Article
Google Scholar

[27] View Article

[28] Google Scholar

[ref11] 11. Baltazar M, Castro I, Gonçalves B. Adaptation to climate change in viticulture: the role of varietal selection—a review. Plants. 2025;14(1):1.
View Article
Google Scholar

[30] View Article

[31] Google Scholar

[ref12] 12. Guo A, Zuo X, Zhang S, Hu Y, Yue P, Lv P, et al. Contrasting effects of plant inter- and intraspecific variation on community trait responses to nitrogen addition and drought in typical and meadow steppes. BMC Plant Biol. 2022;22(1):90. pmid:35232383
View Article
PubMed/NCBI
Google Scholar

[33] View Article

[34] PubMed/NCBI

[35] Google Scholar

[ref13] 13. Miedema Brown L, Anand M. Plant functional traits as measures of ecosystem service provision. Ecosphere. 2022;13(2).
View Article
Google Scholar

[37] View Article

[38] Google Scholar

[ref14] 14. Violle C, Enquist BJ, McGill BJ, Jiang L, Albert CH, Hulshof C, et al. The return of the variance: intraspecific variability in community ecology. Trends Ecol Evol. 2012;27(4):244–52. pmid:22244797
View Article
PubMed/NCBI
Google Scholar

[40] View Article

[41] PubMed/NCBI

[42] Google Scholar

[ref15] 15. Demmings EM, Williams BR, Lee C-R, Barba P, Yang S, Hwang C-F, et al. Quantitative trait locus analysis of leaf morphology indicates conserved shape loci in grapevine. Front Plant Sci. 2019;10:1373. pmid:31803199
View Article
PubMed/NCBI
Google Scholar

[44] View Article

[45] PubMed/NCBI

[46] Google Scholar

[ref16] 16. Martin AR, Li G, Cui B, Mariani RO, Vicario K, Cathline KA, et al. A high-throughput approach for quantifying turgor loss point in grapevine. Plant Methods. 2024;20(1):180. pmid:39581961
View Article
PubMed/NCBI
Google Scholar

[48] View Article

[49] PubMed/NCBI

[50] Google Scholar

[ref17] 17. Nimmo V, Violle C, Entz M, Rolhauser AG, Isaac ME. Changes in crop trait plasticity with domestication history: management practices matter. Ecol Evol. 2023;13(11):e10690. pmid:38020689
View Article
PubMed/NCBI
Google Scholar

[52] View Article

[53] PubMed/NCBI

[54] Google Scholar

[ref18] 18. Westerband AC, Funk JL, Barton KE. Intraspecific trait variation in plants: a renewed focus on its role in ecological processes. Ann Bot. 2021;127(4):397–410. pmid:33507251
View Article
PubMed/NCBI
Google Scholar

[56] View Article

[57] PubMed/NCBI

[58] Google Scholar

[ref19] 19. Martin AR, Isaac ME. Plant functional traits in agroecosystems: a blueprint for research. J Appl Ecol. 2015;52(6):1425–35.
View Article
Google Scholar

[60] View Article

[61] Google Scholar

[ref20] 20. Bartlett MK, Scoffoni C, Ardy R, Zhang Y, Sun S, Cao K, et al. Rapid determination of comparative drought tolerance traits: using an osmometer to predict turgor loss point. Methods Ecol Evol. 2012;3(5):880–8.
View Article
Google Scholar

[63] View Article

[64] Google Scholar

[ref21] 21. Banks JM, Hirons AD. Alternative methods of estimating the water potential at turgor loss point in Acer genotypes. Plant Methods. 2019;15:34. pmid:30988693
View Article
PubMed/NCBI
Google Scholar

[66] View Article

[67] PubMed/NCBI

[68] Google Scholar

[ref22] 22. Saathoff AJ, Welles J. Gas exchange measurements in the unsteady state. Plant Cell Environ. 2021;44(11):3509–23. pmid:34480484
View Article
PubMed/NCBI
Google Scholar

[70] View Article

[71] PubMed/NCBI

[72] Google Scholar

[ref23] 23. Pabuayon ILB, Sun Y, Guo W, Ritchie GL. High-throughput phenotyping in cotton: a review. J Cotton Res. 2019;2(1).
View Article
Google Scholar

[74] View Article

[75] Google Scholar

[ref24] 24. Ely KS, Burnett AC, Lieberman-Cribbin W, Serbin SP, Rogers A. Spectroscopy can predict key leaf traits associated with source-sink balance and carbon-nitrogen status. J Exp Bot. 2019;70(6):1789–99.
View Article
Google Scholar

[77] View Article

[78] Google Scholar

[ref25] 25. Kim M, Lee C, Hong S, Kim SL, Baek JH, Kim KH. High-throughput phenotyping methods for breeding drought-tolerant crops. Int J Mol Sci. 2021;22(15):8266.
View Article
Google Scholar

[80] View Article

[81] Google Scholar

[ref26] 26. Kothari S, Beauchamp‐Rioux R, Laliberté E, Cavender‐Bares J. Reflectance spectroscopy allows rapid, accurate and non‐destructive estimates of functional traits from pressed leaves. Methods Ecol Evol. 2022;14(2):385–401.
View Article
Google Scholar

[83] View Article

[84] Google Scholar

[ref27] 27. Macklin SC, Mariani RO, Young EN, Kish R, Cathline KA, Robertson G, et al. Intraspecific leaf trait variation across and within five common wine grape varieties. Plants (Basel). 2022;11(20):2792. pmid:36297816
View Article
PubMed/NCBI
Google Scholar

[86] View Article

[87] PubMed/NCBI

[88] Google Scholar

[ref28] 28. Bigard A, Berhe DT, Maoddi E, Sire Y, Boursiquot JM, Ojeda H. Vitis vinifera L. fruit diversity to breed varieties anticipating climate changes. Front Plant Sci. 2018;9(455).
View Article
Google Scholar

[90] View Article

[91] Google Scholar

[ref29] 29. P D, G F, G M, T D. Performance of interspecific grapevine varieties in north-east Italy. Agric Sci. 2013;4(2):91–101.
View Article
Google Scholar

[93] View Article

[94] Google Scholar

[ref30] 30. Cui B, Mariani R, Cathline KA, Robertson G, Martin AR. Reflectance spectroscopy predicts leaf functional traits across wine grape cultivars. Plants People Planet. 2025;7(5):1524–37.
View Article
Google Scholar

[96] View Article

[97] Google Scholar

[ref31] 31. Cotrozzi L, Peron R, Tuinstra MR, Mickelbart MV, Couture JJ. Spectral phenotyping of physiological and anatomical leaf traits related with maize water status. Plant Physiol. 2020;184(3):1363–77. pmid:32907885
View Article
PubMed/NCBI
Google Scholar

[99] View Article

[100] PubMed/NCBI

[101] Google Scholar

[ref32] 32. Novick KA, Ficklin DL, Baldocchi D, Davis KJ, Ghezzehei TA, Konings AG, et al. Confronting the water potential information gap. Nat Geosci. 2022;15(3):158–64. pmid:35300262
View Article
PubMed/NCBI
Google Scholar

[103] View Article

[104] PubMed/NCBI

[105] Google Scholar

[ref33] 33. Lv Y, Gu L, Man R, Liu X, Xu J. Response of stomatal conductance, transpiration, and photosynthesis to light and CO2 for rice leaves with different appearance days. Front Plant Sci. 2024;15:1397948. pmid:39148618
View Article
PubMed/NCBI
Google Scholar

[107] View Article

[108] PubMed/NCBI

[109] Google Scholar

[ref34] 34. Martin AR, Rapidel B, Roupsard O, Van den Meersche K, de Melo Virginio Filho E, Barrios M. Intraspecific trait variation across multiple scales: the leaf economics spectrum in coffee. Funct Ecol. 2017;31(3):604–12.
View Article
Google Scholar

[111] View Article

[112] Google Scholar

[ref35] 35. Charrier G. Extrapolating physiological response to drought through step-by-step analysis of water potential. Plantae. 2020;184(2):News and Views.
View Article
Google Scholar

[114] View Article

[115] Google Scholar

[ref36] 36. Ma WT, Yu YZ, Wang X, Gong XY. Estimation of intrinsic water-use efficiency from δ13C signature of C3 leaves: assumptions and uncertainty. Front Plant Sci. 2023;13.
View Article
Google Scholar

[117] View Article

[118] Google Scholar

[ref37] 37. Martin A, Miller L, Madigan M, Cui B, Cathline K, Robertson G. Functional Ecology. 2025;39:1112–27.
View Article
Google Scholar

[120] View Article

[121] Google Scholar

[ref38] 38. Pan S, Wang X, Yan Z, Wu J, Guo L, Peng Z, et al. Leaf stomatal configuration and photosynthetic traits jointly affect leaf water use efficiency in forests along climate gradients. New Phytol. 2024;244(4):1250–62. pmid:39223910
View Article
PubMed/NCBI
Google Scholar

[123] View Article

[124] PubMed/NCBI

[125] Google Scholar

[ref39] 39. Zhang Y, Wu J, Wang A. Comparison of various approaches for estimating leaf water content and stomatal conductance in different plant species using hyperspectral data. Ecol Indic. 2022;142:109278.
View Article
Google Scholar

[127] View Article

[128] Google Scholar

[ref40] 40. Le TS, Harper R, Dell B. Application of remote sensing in detecting and monitoring water stress in forests. Remote Sensing. 2023;15(13):3360.
View Article
Google Scholar

[130] View Article

[131] Google Scholar

[ref41] 41. Dastjerdy B, Saeidi A, Heidarzadeh S. Review of applicable outlier detection methods to treat geomechanical data. Geotechnics. 2023;3(2):375–96.
View Article
Google Scholar

[133] View Article

[134] Google Scholar

[ref42] 42. Burnett AC, Anderson J, Davidson KJ, Ely KS, Lamour J, Li Q. A best-practice guide to predicting plant traits from leaf-level hyperspectral data using partial least squares regression. J Exp Bot. 2021;72(18):6175–89.
View Article
Google Scholar

[136] View Article

[137] Google Scholar

[ref43] 43. Meireles JE, Schweiger AK, Cavender-Bares J. Spectrolab: class and methods for spectral data. 2025. [cited 2025 Oct 20]. https://cran.r-project.org/web/packages/spectrolab/index.html

[ref44] 44. Liland KH, Mevik BH, Wehrens R, Hiemstra P. pls: partial least squares and principal component regression. [cited 2025 Oct 20]. https://cran.r-project.org/web/packages/pls/index.html

[ref45] 45. Lamour J, Serbin S. Spectratrait: a simple add-on package to aid in the fitting of leaf or canopy scale spectra-trait PLSR models. 2025. https://github.com/plantphys/spectratrait

[ref46] 46. Liland KH, Mevik BH, Wehrens R. Partial least squares and principal component regression. 1999. https://CRAN.R-project.org/package=pls

[ref47] 47. Jessica G, Byrne ME. From genes to climate: a perspective on the importance of leaf shape. J Exp Bot. 2026;77(2).
View Article
Google Scholar

[143] View Article

[144] Google Scholar

[ref48] 48. Martin AR, Isaac ME. Functional traits in agroecology: advancing description and prediction in agroecosystems. J Appl Ecol. 2018;55(1):5–11.
View Article
Google Scholar

[146] View Article

[147] Google Scholar

[ref49] 49. Faralli M, Bontempo L, Bianchedi PL, Moser C, Bertamini M, Lawson T, et al. Natural variation in stomatal dynamics drives divergence in heat stress tolerance and contributes to seasonal intrinsic water-use efficiency in Vitis vinifera (subsp. sativa and sylvestris). J Exp Bot. 2022;73(10):3238–50. pmid:34929033
View Article
PubMed/NCBI
Google Scholar

[149] View Article

[150] PubMed/NCBI

[151] Google Scholar

[ref50] 50. Easlon HM, Nemali KS, Richards JH, Hanson DT, Juenger TE, McKay JK. The physiological basis for genetic variation in water use efficiency and carbon isotope composition in Arabidopsis thaliana. Photosynth Res. 2014;119(1–2):119–29. pmid:23893317
View Article
PubMed/NCBI
Google Scholar

[153] View Article

[154] PubMed/NCBI

[155] Google Scholar

[ref51] 51. Tortosa I, Escalona JM, Opazo I, Douthe C, Medrano H. Genotype variations in water use efficiency correspond with photosynthetic traits in tempranillo grapevine clones. Agronomy. 2022;12(8):1874.
View Article
Google Scholar

[157] View Article

[158] Google Scholar

[ref52] 52. Bota J, Tomás M, Flexas J, Medrano H, Escalona JM. Differences among grapevine cultivars in their stomatal behavior and water use efficiency under progressive water stress. Agricultural Water Management. 2016;164:91–9.
View Article
Google Scholar

[160] View Article

[161] Google Scholar

[ref53] 53. Serbin SP, Singh A, Desai AR, Dubois SG, Jablonski AD, Kingdon CC, et al. Remotely estimating photosynthetic capacity, and its response to temperature, in vegetation canopies using imaging spectroscopy. Remote Sensing of Environment. 2015;167:78–87.
View Article
Google Scholar

[163] View Article

[164] Google Scholar

[ref54] 54. Hunt ER, Rock BN. Detection of changes in leaf water content using near- and middle-infrared reflectances. Remote Sens Environ. 1989;30(1):43–54.
View Article
Google Scholar

[166] View Article

[167] Google Scholar

[ref55] 55. Curran PJ, Dungan JL, Gholz HL. Exploring the relationship between reflectance red edge and chlorophyll content in slash pine. Tree Physiol. 1990;7(1_2_3_4):33–48. pmid:14972904
View Article
PubMed/NCBI
Google Scholar

[169] View Article

[170] PubMed/NCBI

[171] Google Scholar

[ref56] 56. Ji F, Li F, Hao D, Shiklomanov AN, Yang X, Townsend PA, et al. Unveiling the transferability of PLSR models for leaf trait estimation: lessons from a comprehensive analysis with a novel global dataset. New Phytol. 2024;243(1):111–31. pmid:38708434
View Article
PubMed/NCBI
Google Scholar

[173] View Article

[174] PubMed/NCBI

[175] Google Scholar

[ref57] 57. Kothari S, Beauchamp-Rioux R, Blanchard F, Crofts AL, Girard A, Guilbeault-Mayers X, et al. Predicting leaf traits across functional groups using reflectance spectroscopy. New Phytol. 2023;238(2):549–66. pmid:36746189
View Article
PubMed/NCBI
Google Scholar

[177] View Article

[178] PubMed/NCBI

[179] Google Scholar

[ref58] 58. Xu Y, Goodacre R. On splitting training and validation set: a comparative study of cross-validation, bootstrap and systematic sampling for estimating the generalization performance of supervised learning. J Anal Test. 2018;2(3):249–62. pmid:30842888
View Article
PubMed/NCBI
Google Scholar

[181] View Article

[182] PubMed/NCBI

[183] Google Scholar

[ref59] 59. Reich PB, Wright IJ, Cavender‐Bares J, Craine JM, Oleksyn J, Westoby M, et al. The evolution of plant functional variation: traits, spectra, and strategies. International Journal of Plant Sciences. 2003;164(S3):S143–64.
View Article
Google Scholar

[185] View Article

[186] Google Scholar

Figures

Abstract

1 Introduction

2 Materials and methods

2.1 Study design and trait collection

2.2 Statistical analyses

3 Results

3.1 Trait variation across wine grape varieties and cultivars

3.2 Reflectance spectra and PLSR models

4 Discussion

5 Conclusions

Acknowledgments

References