Exploring plant responses to abiotic stress by contrasting spectral signature changes

In this study, daily changes over a short period and diurnal progression of spectral reflectance at the leaf level were used to identify spring wheat genotypes (Triticum aestivum L.) susceptible to adverse conditions. Four genotypes were grown in pots experiments under semi-controlled conditions in Chile and Spain. Three treatments were applied: i) control (C), ii) water stress (WS), and iii) combined water and heat shock (WS+T). Spectral reflectance, gas exchange and chlorophyll fluorescence measurements were performed on flag leaves for three consecutive days at anthesis. High canopy temperature ( H CT ) genotypes showed less variability in their mean spectral reflectance signature and chlorophyll fluorescence, which was related to weaker responses to environmental fluctuations. While low canopy temperature ( L CT ) genotypes showed greater variability. The genotypes spectral signature changes, in accordance with environmental fluctuation, were associated with variations in their stomatal conductance under both stress conditions (WS and WS+T); L CT genotypes showed an anisohydric response compared that of H CT , which was isohydric. This approach could be used in breeding programs for screening a large number of genotypes through proximal or remote sensing tools and be a novel but simple way to identify groups of genotypes with contrasting performances.

In this study, daily changes over a short period and diurnal progression of spectral reflectance at the leaf level were used to identify spring wheat genotypes (Triticum aestivum L.) susceptible to adverse conditions. Four genotypes were grown in pots experiments under semi-controlled conditions in Chile and Spain. Three treatments were applied: i) control (C), ii) water stress (WS), and iii) combined water and heat shock (WS+T). Spectral reflectance, gas exchange and chlorophyll fluorescence measurements were performed on flag leaves for three consecutive days at anthesis. High canopy temperature (H CT ) genotypes showed less variability in their mean spectral reflectance signature and chlorophyll fluorescence, which was related to weaker responses to environmental fluctuations. While low canopy temperature (L CT ) genotypes showed greater variability. The genotypes spectral signature changes, in accordance with environmental fluctuation, were associated with variations in their stomatal conductance under both stress conditions (WS and WS+T); L CT genotypes showed an anisohydric response compared that of H CT , which was isohydric. This approach could be used in breeding programs for screening a large number of genotypes through proximal or remote sensing tools and be a novel but simple way to identify groups of genotypes with contrasting performances. KEYWORDS high-throughput plant phenotyping, spectroscopy, spectrometer, spectroradiometer, phenotype, physiological breeding

Introduction
Climate change is intensifying local environmental constraints. For many areas, it is predicted that rainfall will progressively decrease, and extreme heat events will become increasingly common (Dixon et al., 2009;Rebetzke et al., 2012;Hernańdez-Barrera et al, 2017). In addition, the competition for water resources between human and industrial consumption will increase, affecting food security (e.g., grain production) in the coming decades (Trethowan et al., 2005;Tilman et al., 2011;Mohanty et al., 2020).
Bread wheat (Triticum aestivum L.) is one of the most important crops for the human diet, and one of the most widely cultivated in the world (Braun and Payne, 2013). In Mediterranean-climate areas, drought and heat have detrimental consequences during anthesis and grain filling, affecting grain yield (GY) and its components (Thomas and Ougham, 2014;Pinto et al., 2016;Velu et al., 2016;del Pozo et al., 2019). Breeding is one of the pillars for wheat adaptation to the threats imposed by climate change, with phenotyping being considered the bottleneck, in terms of the management of the huge information of the phenotype generated, limiting genetic advance (Araus and Kefauver, 2018;Yudina et al., 2020).
Plant phenotyping using different remote sensing approaches is receiving increasing interest among plant breeders and their users in agriculture. Among remote sensing methods, canopy, or even single leaf reflectance (represented graphically by the spectral signature) is closely related to the ability to absorb and transmit incident radiation under a particular environmental condition (Garbulsky et al., 2011), therefore a useful indicator of the plant physiological status (Peñuelas and Filella, 1998) and the Genotype by Environment (GxE) effect (Vilmus et al., 2014;Garriga et al., 2017). For instance, in a single scan, high-resolution field spectrometers acquire the reflectance of a wide range of wavelengths (350 -2,500 nm), and this spectral signature can be related to various leaf, plant or crop traits using spectral reflectance indices (SRIs; relationships between particular wavelengths or spectrum bands), such as the normalized difference vegetation index (NDVI), which is related to leaf or canopy greenness (Babar et al, 2006a), and the photochemical reflectance index (PRI), associated to the xanthophyll's activity (Gamon et al., 1992;Hernandeź-Clemente et al, 2011). Another approach to using reflectance information is through multivariate regression models to predict physiological or productive traits, that include part or the whole spectral signature (Graeff and Claupein, 2007;Sridhar et al., 2007;Winterhalter et al., 2011;Mullan, 2012;Fallon et al, 2020;Xie et al., 2020), and classification methods for direct identification of the elite genotype group (Garriga et al., 2017). Changes in the spectral signature have been associated with modifications in plant tissue properties like the hydration state or chemical composition of oak leaves (Quercus spp.) (Cavender-Bares et al., 2016), stomatal conductance of cotton plants (Gossypium hirsutum) (Vitrack-Tamam et al., 2020), or leaf photosynthetic traits (Peñuelas and Filella, 1998;Silva-Peŕez et al., 2020) and carbon isotopic discrimination in the grains of wheat (Lobos et al., 2014;Garriga et al., 2021).
Nevertheless, the leafe spectral signature not only describes the biochemical an physiological state of a genotype at a certain or punctual moment, but it also integrates of environmental events throughout the growing season until the day of measurement (Marcinśka et al., 2013;Lobos et al., 2019;Banerjee et al., 2020;Vitrack-Tamam et al., 2020); without ignoring the possible epigenetic effect of environmental characteristics that have impacted the seeds used for the crop (Crisp et al., 2016). Since both the uniqueness of each season's environmental characteristics and the plasticity of the genotype are reflected in the morpho-physiological and physico-chemical traits (Elazab et al., 2012;Jäger et al., 2014;Petrov et al., 2018;Kanbar et al., 2020;Yan et al., 2020). Also, since the hydric state of the plant is closely associated with the ability to satisfy the atmospheric demand for evapotranspiration (Kudoyarova et al., 2011;Zhang et al., 2018), changes in the daily and diurnal vapor pressure deficit (VPD) should be accompanied by changes in the spectral signature (Magney et al., 2016). Thus, genotypes with contrasting tolerance to abiotic stresses should have different responses in their spectral signature to changing environmental conditions. Thus, the aim of this work was to evaluate the daily changes (three consecutive days) in the leaf spectral reflectance, leaf gas exchange, chlorophyll fluorescence and pigment content, of spring wheat genotypes exposed to water stress (WS) and WS combined with heat stress (WS+T). It was hypothesized that genotypes more affected by environmental conditions should exhibit larger differences in spectral signature among the evaluated days. Furthermore, if the above is true, the analysis of its diurnal spectral signature should be consistent with the pattern observed in the day comparison, due to diurnal changes in air temperature and VPD.
2 Materials and methods 2.1 Selection of four genotypes for the study Four spring wheat genotypes of contrasting canopy temperatures (CT) and GY under rainfed conditions (Table 1) were selected from a panel of 384 cultivars and advanced lines from CIMMYT Mexico, INIA Uruguay and INIA Chile. The panel was previously evaluated under rainfed (and high VPD) and irrigated (and reduced VPD) Mediterranean conditions, in (del Pozo et al., 2016; the four genotypes were chosen among a subset of 104 genotypes, having a range of 80 -83 d from sowing to earing, 4.3 -5.6 of leaf area index at anthesis and 93 -100 cm of plant height at anthesis.

Experimental setup
The four spring bread wheat genotypes were assessed in two controlled condition experiments, one in conventional glasshouses at the Universidad de Talca -Chile (UTALCA experiment: 35°24'20" S, 71°38'5" W) in 2018, and the second in plastic growth chambers placed outdoors at the Universidad de las Islas Baleares -Spain (UIB experiment: 39°38′ 17″ N, 2°3 8′54″ E) in 2019. At UTALCA, two glasshouses were used (12 × 9 m and enclosed with alveolar polycarbonate sheets of 6 mm thickness and 86% solar transmission); one was open laterally (only the roof covered to shield plants from rain or fog), representing ambient temperature conditions (At), and the other was fully enclosed, and temperature controlled to~8°C above At (At+8). At UIB growth chambers (1.7 × 5 m; transparent polyethylene film) were set up to provide both the At and At+8 conditions. Temperature and relative humidity (RH) were recorded with HOBO ® Series H8 sensors (Onset Computer Corporation, MA, USA) at UTALCA and HOBO ® UX100-003 sensors at UIB. Temperature and RH data were used to calculate VPD during the experiments (Table 2). Genotypes were exposed to the following growing conditions during anthesis: i) control (C: substrate~75% pot water capacity and ambient temperature); water stress (WS: under At, substratẽ 30 and 50% of pot water capacity at UTALCA and UIB experiments, respectively); and iii) water and heat shock combined (WS+T: water deficit and ambient temperature increased by around 8°C).
For the UTALCA experiment, 40 seeds of each genotype were sown in 20 L plastic containers (radius: 15 cm; height: 50 cm), filled with a 6:1:1 mixture of river sand, organic soil and perlite. For the UIB experiment, 20 seeds of each genotype were sown in 10 L plastic containers (radius: 12.5 cm; height: 19.5 cm), filled with a 1:1 mixture of river sand and perlite. In both experiments, the seeds were distributed in concentric circles, sowing an outer ring to minimize edge effects at the root level (i.e., plants growing in contact with the wall of the pot; 10 and 5 seeds in UTALCA and UIB, respectively) and the inner rings with seeds of plants that were to be measured (30 and 15 seeds in UTALCA and UIB, respectively). Both in UTALCA and UIB, fertilizer was applied weekly, comprising 1,000 ml of fullstrength Hoagland's nutrient solution (Hoagland and Arnon, 1950).

Experimental design and measurements
The two trials were conducted as a completely randomized design, with three replicates (pots) per treatment and genotype. To determine the maximum water holding capacity, each pot was watered until saturation and weighed after 24 h of drainage to determine the amount of water required to reach 100% pot capacity. During the experiments, pots were weighed daily. All pots were treated as control plants (C) until flag leaf blades were fully unfolded (Z41; Zadoks et al., 1974). From Z41 to anthesis (Z68), plants of each genotype were divided in two groups: 1/3 were kept as C and 2/3 were exposed to WS conditions. At Z68, plants under WS were again divided in two groups, half continued in WS while the other plants were exposed to WS +T in the At+8 environment. The leaf spectral signature and gas exchange was determined for each genotype and treatment. In the UTALCA and UIB experiments, once the combined stress (WS+T) was imposed, measurements were performed immediately on three consecutive days (D1, D2, and D3), at solar zenith, just before the plateau of the maximum daily atmospheric demand for water. In addition, to test whether genotypic differences within the day (diurnal progression) were consistent with those in reflectance observed between days, on D3 the spectral reflectance was evaluated at 10: 00,11:30,13:00,14:30,, in the UTALCA experiment. Spectral reflectance and gas exchange evaluations were carried out on the middle third of healthy and sun-exposed flag leaves. At UTALCA, three flag leaves per pot were selected for each assessment, while at UIB, two flag leaves were considered per replicate. Different leaves were considered in each evaluation for both progression between days and within the day.

Spectral reflectance
At UTALCA, the reflectance was recorded with a FieldSpec 3 Jr. portable spectrometer (Analytical Spectral Devices ASD Inc., Boulder, CO, USA) over a spectral range 350 -2,500 nm, and a spectral resolution of 3 nm from 350 to 1,000 nm and 30 nm from 1,001 to 2,500 nm. The fiber was inserted into a contact probe device (ASD Inc., Boulder, CO, USA) equipped with a halogen light (5 W), generating a measuring spot of~5 mm radius. The spectrometer was calibrated every 15 min with a white reference tile (Spectralon ® , ASD Inc., Boulder, CO, USA). The equipment was configured to integrate three samples per scan, and ten scans per leaf (as described in Lobos and Poblete-Echeverrıá, 2017). The data were extracted using View Spec Pro 2008 software (ASD Inc., Boulder, CO, USA). The exploratory analysis of the spectral reflectance data was performed using SK-UTALCA software (Lobos and Poblete-Echeverrıá, 2017).
At UIB, reflectance measurements were performed with a Vis/NIR optical spectrometer (Jaz-EL350, Ocean optics, Dunedin, FL, USA) equipped with a halogen tungsten light source (4.75 W) and a QP600-1-SR-BX optical fiber (Ocean Optics, Inc., Dunedin, FL, USA; spectral range 400 -900 nm), with 0.3 nm spectral resolution. The calibration parameters were adjusted to 10 ms for the integration time, with a light intensity of 2,500 (counts). Before measurement, the spectrometer was calibrated to 100% reflectance using a white reference panel (Ocean Optics, Dunedin, FL, USA), while the dark calibration was obtained with a black panel. The equipment was configured to integrate ten scans per leaf. Because there was not enough space to take measurements in the growth chambers, spectral reflectance was measured in a room (25°C) adjacent to the chambers.

Leaf gas exchange, chlorophyll fluorescence and pigment content
At UTALCA, leaf net CO 2 assimilation (An), the stomatal conductance (gs) and the leaf temperature (T l ) were evaluated using a CIRAS 2 infrared gas analyzer (IRGA) (PP Systems, Amesbury, MA, USA), with a narrow-leaf cuvette (1.7 cm 2 ) at a flow rate of 250 ml min -1 , CO 2 concentration of 400 ppm, cuvette temperature of 25°C, and photosynthetically active radiation (PAR) of 1,500 µmol m -2 s -1 . At UIB, measurements were performed with a Li-6400xt portable photosynthesis system (LI-COR Inc., Lincoln, Nebraska, USA). The environmental parameters of the leaf chamber were adjusted to equal those of the UTALCA experiment, except for the temperature of the cuvette, which was set at room temperature.
In UTALCA experiment, chlorophyll fluorescence (Chl f ) was assessed using a portable pulse amplitude modulated fluorometer (PAM-2500, Walz, Germany). The minimum and the maximum yield of fluorescence under dark conditions (Fo and Fm, respectively), were measured on leaves adapted to darkness for 20 minutes with a Leaf-Clip (Leaf-Clip Holder 2030-B, Walz, Germany). Then, a rapid light curve (RLC) was performed in the same leaf spot. For that, the equipment was programmed to emit 10 pulses of actinic light at different levels of photosynthetically active radiation (PAR), which increased from 0 to 1,982 mmol m -2 s -1 (the time between each PAR level was six seconds). At each PAR, the minimum calculated and the maximum yield of fluorescence under light conditions (~Fo´and Fm´, respectively), were automatically recorded.
Anthocyanin content was estimated using a non-destructive portable chlorophyll meter (Dualex, Force A, France); values are given in relative absorbance units from 0 to 1.5.

Data analyses
The genotypic (G) variability of the spectral signature under each treatment (environment-E) was assessed through direct comparison (ANOVA and p-values < 0.05) of the spectral signatures at the wavelength level (Lobos et al., 2019). Thus, the first reflectance measurements (i.e., measurements of D1 compared with D2 and D3 in case of the daily analysis, and measurement at 10:00 h with each moment of measurement in the diurnal progression analysis) were compared (ANOVA, p-values < 0.05; n = 9 and 6 in UTALCA and UIB experiments, respectively) to the subsequent ones. When significant differences were found (Tukey's multiple comparison test, p-values < 0.05) between the said spectral signatures, the percentage change with respect to the first measurement was calculated and plotted.
To establish statistical differences in the gas exchange performance of the genotypes and environments, multifactor ANOVA (p-values < 0.05; n = 9 and 6 in UTALCA and UIB experiments, respectively) was carried out on T l , gs, and An. When significant differences were found, Tukey's test (p-values < 0.05) was applied. To identify the genotypic variability of biochemical limitations of photosynthesis to the generated environmental conditions, the relationship between An and gs was studied through regression analysis (Medrano et al., 2002;Gago et al., 2020).

Environmental conditions during the experiment
The average temperature and VPD was increased by 8°C and 1.6-2.3 kPa, respectively, at the elevated temperature regime, in both experiments (Table 2). However, temperature at At regimeand, consequently, at all other regimeswas higher in the UIB experiment compared to the UTALCA experiment. During D3 of the diurnal progression study (UTALCA), the ambient temperature and VPD gradient also increased from 09:00 h (At: 14.2°C and 0.65 kPa; At+8: 25.0°C and 2.85 kPa) to a peak between 14:00 and 16:00 h (At: 25.9°C and 2.41 kPa; At+8: 33.2°C and 4.1 kPa).

Determining spectral signature stability throughout daily progression analysis
The spectral reflectance signature of plants growing under C, WS, and WS+T treatments over time (D1 to D3) varied among genotypes according to the CT performance ( Figure 1). The spectral reflectance of low CT genotypes (L CT -H GY and L CT -L GY ) showed greater variation between days (Figures 1A-F) than did that of high CT genotypes (H CT -L GY and H CT -H GY ) (Figures 1G-L). Although somewhat less evident than under stress conditions, this behavior was also reflected in the control condition. The major changes in high CT genotypes were observed on D2 (red lines in Figures 2G, J), while the greatest changes were found in the low CT genotypes in D3 (blue lines in Figures 2A, D). In addition, genotypes showed different patterns throughout the day according to their CT. The mean reflectance signatures of low CT genotypes continued to decrease from D2 to D3, with respect to D1 (Figures 2A-F), while in high CT genotypes, reflectance stopped changing or increased in the direction of their initial condition in D1 (Figures 2G-L). The effects of WS and WS+T treatments on the spectral signature performance was lower in low CT genotypes compared to high CT genotypes, particularly in the UTALCA experiment (Figure 2). In the case of H CT -H GY , greater differences were observed in the patterns of the spectral signature under WS and WS+T. Under WS treatment ( Figure 2H), the reflectance in the VIS and the first part of the NIR increased during D2 and D3 with respect to D1 (negative differences), but at wavelength > 1,400 nm, the reflectance decreased compared to D1 (positive values). While under WS+T conditions, in almost the entire spectra, the reflectance during D2 and D3 increases, compared to D1 ( Figure 2J). In the case of H CT -L GY under WS, the spectral reflectance was reduced on D2, compared to D1, but the opposite was observed in D3 ( Figure 2K). The same tendency was observed for the WS+T treatment.

Determining spectral signature stability through diurnal progression analysis
Spectral signature changes recorded within the day (10:00 -16:30 h) (Figures 3, 4) followed similar patterns to those described for the daily progression analysis. Low CT genotypes, also showed greater differences than high CT genotypes, when comparing the changes in spectral signatures recorded at 10:00 with those recorded later in the afternoon. The spectral signature performance at 12:30 and 16:30 h appears to

Gas exchange assessments
Regarding the initial selection of the genotypes by leaf temperature (Table 1), the results show that these CT characteristics were maintained in the present study when analyzing the gas exchange performance of the genotypes (Figures 5A-C). The ANOVA performed for each day of evaluation, at UTALCA experiment, shows statistical differences between genotypes in gs, An, Fm, Fm´, Fo and Fo´during the three days, meanwhile T l just on D3. In the case of the treatments, each day shows significant differences among them. For the interaction between genotype and treatment, we found significant differences for the gs and the fluorescence variables (Supplementary Table 2). In the case of the UIB experiment, the genotypes showed statistical differences in An on the three days. During D2, both low CT genotypes showed lower T l (4°C minus) than high CT genotypes. in terms of the treatments, similar to UTALCA, showed statistical differences at each day. In D1, the three variables have significative interactions (Supplementary Table 2). In general terms, the high CT genotypes had lower gs and An values across both experiments (Figures 5D-I), than the low CT. In consequence, the leaf temperature of both high CT genotypes was higher compared to low CT genotypes. In all genotypes, gs and An were strongly reduced under WS and WS+T (Figures 5 D-I). The two genotypes with lower CT tend to had higher fluorescence level during D1 and D3. Genotypes under C conditions in the UTALCA experiment showed similar gs and An levels during the three days of the experiments (Figures 5D-I).
The high CT genotypes were the most affected by the combined stress conditions on D1, however gs and An progressively reached the same values of WS conditions on D3 ( Figures 5E, F, H, I). In both trials, H CT -H GY show the greater capacity to overcome the WS+T condition than H CT -L GY , which in the case of the UTALCA experiment achieves almost similar gs values than under C.
The relationship between An and gs indicated that for a given value of gs, An was higher in the UTALCA experiment, in all genotypes ( Figure S3). Although with different IRGAs, it is interesting to note that at the same gs level, the differences in An Statistical differences in spectral reflectance calculated as the percentage change in the wavelength level between days of assessment (D1 to D3: D1-D2 red line, and D1-D3 blue line; at each wavelength, interruptions in the lines indicate sections without statistical differences, p < 0.05) in four spring bread wheat. Genotypes and treatments are indicated in Figure 1. Measurements were performed just before the plateau of the maximum daily atmospheric demand for water (~13:00 h). n = 9. The dashed vertical lines indicate the peaks around 380, 680, 1450, and 1950 nm. between experiments were lesser in the high-yielding genotypes (Figures S3A, C) than in the low-yielding genotypes (L CT -L GY and H CT -L GY ) ( Figures S3B, D). Greater genotypic differences were found at values of gs above 150 mmol H 2 O m -2 s -1 in the UTALCA experiment, while at UIB, genotypic variability was evident at lower gs levels ( Figures. S3E, F). In these sense, two groups were identified according to CT (i.e., Lower and High). Despite the dissimilarities associated with experiment location, the H CT -H GY genotype always had a lower ratio An/gs ( Figures  S3E, F).

Relationship between spectral reflectance indices with environmental and physiological variables
The results of the correlation analysis between the studied SRIs and both the environmental (VPD, HR, air temperature) and foliar (temperature and anthocyanins) characteristics varied according to the index (Table 3). Thus, at compare NDVI and PRI, the first one showed a wide and significance association in L CT -L GY and L CT -H GY . On the other hand, correlations with the same genotypes showed higher variability with PRI.

Consistency between reflectance and fluorescence assessments
A similar trend of changes was observed when comparing daily mean values of reflectance at 380, 680, 1,450 and 1,950 nm (where differences were most evident when comparing D1 versus D2 and D3), with daily mean values of chlorophyll fluorescence variables. In other words, the low TC genotypes showed greater variability at wavelengths 380, 680, 1,450, 1,950 nm and in Fo, Fo', Fm and Fm', from D1 to D3 (Figures 6A-D), whereas in the high TC genotypes smaller changes were observed between days. In fact, H CT -H GY had no change in Fm and Fm' between days, whereas H CT -L GY had the same performance in Fo and~Fo' (Figures 6E-H). Also, higher and significative correlations of gas exchange (T l , gs and An) and fluorescence variables with the wavelength's reflectance (at 380, 680, 1,450 and 1,950 nm) were found (Supplementary Table 1). During D1, the gas exchange and fluorescence variables have higher correlations with the reflectance of the wavelengths especially under WS and WS+T. The physiological traits correlate better with the 380 and 680 nm; gs and An, correlate better with 380 nm, while fluorescence variables with the four wavelengths.

Discussion
The environmental data (Table 2), and specifically the range of ambient temperatures and VPD, suggest that plants at UTALCA under WS+T were subjected to a moderate atmospheric demand (i.e., 1.5 KPa) and moderate heat stress (i.e., 32°C) (Lizana and Calderini, 2013). It has been reported that gas exchange, specifically gs, is reduced by around 17% in anthesis when plants are subject to an environment with heat shock of 32°C and full irrigation (Djanaguiraman et al., 2020). The results of this study showed a reduction of gs and An higher than 30% due to the plants being subjected to two stressor factors (water and heat shock). In the UIB experiment, in addition to higher VPD (> 3.5 KPa) in WS+T condition, the high night temperatures (night mean of 24°C at UIB, versus 13°C at UTALCA, data not shown) were also involved in the plant response to stress environment (i.e., spectral reflectance and gas exchange variables) (Fleitas et al., 2020;Fakhet et al., 2021). Furthermore, plants were exposed to severe environmental stress than in the UTALCA experiment.
Different studies report an anisohydric and isohydric performance of wheat genotypes, although early works classified mainly as anisohydric species (Henson et al., 1989;Tardieu and Simonneau, 1998), both types of stomatal control have been described in wheat genotypes. The anisohydric stomatal response is related to a less conservative water strategy in terms of transpiration (De Boeck et al., 2011;De Boeck et al., 2016). Genotypes with this type of response have higher rates of transpiration and gs, until drought becomes severe and stomatal closure occurs (Galléet al., 2013;Liu et al.,  2018). Different mechanisms are associated with this response such as ABA synthesis at root level (Galléet al., 2013;Saradadevi et al., 2014;Saradadevi et al., 2016;Giusti et al., 2017), osmotic adjustment, the increase of osmoprotective compounds (Smirnov et al., 2020) and the increase of antioxidant capacity (Galléet al., 2008) at leaf level (Smirnov et al., 2020). On the contrary, the isohydric stomatal response has the opposite behavior, more conservative water losses by faster stomatal closure under water stress conditions, due to a higher sensitivity to abscisic acid signals (Blum, 2015). In this study, low CT genotypes maintained higher gs levels in both stress environments, representing a nearly anisohydric stomatal response (Supplementary Figures S4E, F). On the other hand, the high CT genotypes in the WS+T condition, showed an isohydric stomatal regulation, based on a stronger reduction of gs and An at D1, reflecting a greater sensitivity to abscisic acid signals (Blum, 2015). With such stomatal control the dispersion of gs and An across days remains low and constant ( Figures S4E, F); a more conservative strategy to maintain tissues water status, at the cost of low CO 2 assimilation due to stomatal closure (Davies and Pereira, 1992;Maroco et al., 1997;Negin and Moshelion, 2017;Fallon et al., 2020). As in this study, Bayoumi et al. (2015) compared the plant CT of wheat genotypes at zenith, and were able to r e c o g n i z e t h e m o s t c o n t r a s t i n g i s o h y d r i c a n d anisohydric material. Spectral reflectance performance can be related to anisohydric or isohydric performance. For example, measuring at zenith the PRI and NDVI, Vaz et al. (2016) was able to differentiate between anisohydric and isohydric grapevines cultivars under drought conditions. As in this work, isohydric cultivar did not showed significant variations in their spectral reflectance signatures or in the NDVI and PRI pattern compared to the anisohydric cultivar. Similar results were reported by Sobejano-Paz et al. (2020) comparing species with different stomatal control under water deficit; maize (anisohydric) and soybean (isohydric).
In particular, the changes in the PRI are associated with xanthophyll activity which increase under stress (Gamon et al., 1992). Hence, the change of PRI reflects the variation of photosynthetic activity in a certain period (i.e., through the days and hours), as a response to the stomatal and xanthophylls activity under different environments (Gerhards et al., 2018). Likewise, changes in NDVI could be related to tissue hydric status, the photosynthetic activity considering the total canopy green area, or at the leaves level (chlorophylls content) (Gamon et al., 2015;Duan et al., 2017) and also, by stomatal control. In the present work, the low CT genotypes presented higher correlations of SRIs (NDVI higher than PRI) with both the anthocyanin content and the recorded environmental variables, compared to high CT genotypes (Table 3); nevertheless, the interpretation of NDVI is far from easy, and changes in the NDVI could occur by mechanisms other than chlorophylls loss (Atherton et al., 2020).
In addition, the daily behavior of the spectral signature ( Figure 2) and the calculated SRIs (Supplementary Figure S1) were consistent with the pattern during the day (Figure 4 and Supplementary Figure S2); in all tested environmental conditions, increase in VPD and environmental temperature throughout the day resulted in changes in the spectral signature, greater in the low CT genotypes than in high CT ones, mainly in the VIS-NIR region. In fact, Weksler et al. (2020) report the TABLE 3 Pearson correlation analyses (r and p-values) of the relationships between spectral reflectance indexes (normalized difference vegetation index; NDVI and photochemical index; PRI), with the vapor pressure deficit (VPD), relative humidity (RH), air temperature (T a ), anthocyanins content (Anth), and leaf temperature (T l ) in the UTALCA experiment during the 2018 season, for four spring bread wheat genotypes on three days of evaluation for plants growing under control conditions (C), soil water stress (WS) and combined soil water and heat stress conditions (WS+T).  association between changes in the transpiration rate and the spectral signature in pepper plants under different levels of potassium fertilization, evaluated every hour from 07:00 to 17:00 h; leaf spectral reflectance in the morning differed from those in the noon or afternoon, in the VIS-NIR region, as was observed in our results. In the case of the reflectance in the NIR region (700-1300 nm), it has been related to morphological characteristics of the leaf, such as cuticle, intercellular air space, the ratio between palisade mesophyll and spongy mesophyll (Peñuelas and Filella, 1998). In this sense, Willick et al. (2018) determined that, a drought tolerant wheat genotype differs from a susceptible in epidermal characteristics like, higher epicuticular wax density on the adaxial flag leaf surfaces and larger bulliform cells (Uddin and Marshall, 1988). We speculate that the evaluated genotypes in the present work may differ in these types of features, which may partly explain the differences in spectral responses, considering that they originated in different environments (breeding programs Table 1). These structural features have been associated with leaf water status and with several SRIs based on NIR wavelengths (El-Hendawy et al., 2019). Significant peaks were found in the shortwave-infrared range (SWIR,~1,300 -2,500 nm) that allow to differentiate the genotypes according to their CT (Supplementary Figures  S4A−E, S5). The dynamics changes of reflectance peaks at 1,450 and 1,950 nm in the UTALCA experiment (daily and the diurnal progression analysis), could be explained by changes in tissue water content because that wavelength has been reported for estimating plant water status (Curcio and Petty, 1951;Palmer and Williams, 1974;Ihuoma and Madramootoo, 2019;Chandel et al., 2020). regimes. Nevertheless, they also discuss about the low prediction capacity of SRIs when changes in the traits to be evaluated are slight to moderate (Lobos et al., 2014;Romero-Bravo et al., 2019). It is well established that plants that maintain gs and An, under mild to moderate drought conditions (i.e., anisohydric stomatal control) have greater CO 2 fixation, thus is usually associated with a greater tolerance to abiotic stress (Drew, 2006;Sade et al., 2009;Galléet al., 2013;Rashid et al., 2018;Pawłowicz and Masajada, 2019). Nevertheless, the result of this work also suggests that the selection of higher CT (H CT -H GY ) material not always will turn into a low GY genotype.

Genotype
New sources of genotypic variability for GY improvement are difficult to find, and this may compromise food security (Ray et al., 2013). If an isohydric genotype is selected in a high GYbased breeding program for tolerance to drought, the genotype might have the capability to mobilize a higher proportion of carbohydrates to the grain; as in the case of H CT -H GY , probably due to an improved harvest index (HI; Carmo-Silva et al., 2017). Nevertheless, according to this study, it appears that GY would be more associated with differences in the intrinsic water use efficiency between locations (UTALCA vs. UIB) than with the CT pattern; high GY genotypes showed a smaller distance between both An/gs regressions (Supplementary Figures S3A, C) compared to the low GY genotypes (Supplementary Figures  S3B, D). A similar tendency of lower dispersion in spectral signature and gas exchange parameters (gs and An) was found in both high CT genotypes (isohydrics), which could be associated with their lower response to environmental fluctuation (Supplementary Figures S4 , S5).
Finally, the present methodology, based on daily or diurnal progression analysis of spectral signature, appears to be a simple and consistent alternative to evaluate genotypic variability for a particular environmental condition (G x E), with the ability to discriminate between isohydric and anisohydric material; which is a trait used as selection criteria and for planning new crosses (Medina et al., 2019;Feng et al., 2020). This affirmation is reinforced when comparing the change in the daily and the diurnal progression of spectral signature (Figures 2, 4). In particular, the selected wavelengths (380, 680, 1,450 and 1,950 nm, Figures 6A-D), that allow to differentiate the genotypes performance better than Chl f variables (Fo,~Fo´, Fm and Fm´) ( Figures 6E-H). Also, the wavelengths showed higher and significant correlations with the chlorophyll fluorescence and gas exchange variables (Supplementary Table 1). Chlorophyll fluorescence and gas exchange parameters are widely used for genotype or cultivar selection under abiotic condition in different species (Flexas et al., 2000;Flexas et al., 2002;Estrada et al., 2015), however both measurements, at leaf level, are time consuming and therefore the number of measurements that can be performed is limited. In case of the Chl f measurements, methodologies have been developed to estimate sun-induced Chl f through spectral reflectance (proximally or remotely), to solve the time-consuming problem of measurements made at leaf level (Ni et al., 2019), increasing the number of genotypes evaluated in breeding programs (Azam et al., 2015;Bai et al., 2016;Song et al., 2018). Also, the two peaks of Chl f , centered at 685 nm and 740 nm (Corp et al., 2003;Meroni et al., 2009), have similar trends with Chl f variables, especially the one at 740 nm (Supplementary Figure S6). It is interesting to note the shape of Fm and Fm´of H CT -H GY across days, that could be an indicative of PSII photo-activity stability, through a constant capacity for harvesting and transfer of light energy in the mesophyll cells, which is reflected in Fm and Fm´behavior. On the contrary, gas exchange measurements still to be performed at leaf level, for what they are time consuming evaluation. In this sense, different approximation has been probed through modeling procedures using spectral reflectance information, to predict the stomatal conduce, the net CO 2 assimilation and other variables (Garriga et al., 2017;Sexton et al., 2021). One of the advantages of the spectral reflectance measurements it's that they are faster (seconds per leaf or at canopy level), and the protocols of measurements at the field are easy to implement, at the difference of gas exchange, Chl f and leaf or canopy temperature. In the particular case of the last type of measurement, different works have been developed to predict the canopy temperature by spectral reflectance information under abiotic stress conditions (Babar et al., 2006b;Kumagai et al., 2022); these arguments highlight that the methodology presented in this works, based on the interpretation of spectral signature patterns in a particular environment, is valuable to be used in the genotype selection with desirable characteristics, such as the transpiration regulation.

Conclusion
Spectral reflectance analysis showed the possibility of using the spectral signature to differentiate the contrasting responses of wheat genotypes to VPD fluctuations across days and hours. The low CT genotypes were more responsive to the environmental conditions, showing greater differences in their spectral signatures (mainly in the VIS-NIR region) in each environment, when comparing measurements between days and between hours, than the high CT genotypes. Four wavelengths, two in the VIS (380 and 680 nm) and two in the SWIR (1,450 and 1,950 nm) were identified as revealing the highest differences in the low CT genotypes. Spectral reflectance indices were also effective at evaluating the genotype sensitivity to stress conditions. Higher and significant correlations between NDVI and PRI with VPD and RH were obtained in the more sensible low CT genotypes. The spectral signature differences in low CT genotypes were associated with an anisohydric response to WS and WS+T due to higher gs and lower T l , while the higher CT showed an isohydric response.
This study highlights some perspectives in the use of spectral reflectance data for evaluating the plant response regarding to the changes in environmental conditions within a short period of time: i) the simple interpretation of the changes in the spectral signature by itself, due to environmental fluctuation, is a powerful tool for contrasting genotype performance (Lobos et al., 2019) related to isohydric or anisohydric responses, and ii) the wavelengths identified (380, 680, 1,450 and 1,950 nm) are interesting candidates to develop new spectral indices that allow evaluating genotypic sensibility to environmental changes.
The methodology used in this work reveals results that can be subsequently validated as a methodology using remote sensing tools in a large number of genotypes, to perform the fi rst evaluations and segregation of genotypes in breeding programs.

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

Author contributions
GL and FE contributed to the conception and design of the work. BC, CA-R, JG-T, JF, CD, FM-P, DC, IM, AM-E, MG, JA, and AP performed the analysis, and interpretation of data for the work. CD contributed to the experimental design, build of chambers, and measurements performed in UIB experiment in Spain. GL, JF, JA, AP, and FE collaborated to generate and validate the version to be published. All authors contributed to the article and approved the submitted version.