Ability of Modified Spectral Reflectance Indices for Estimating Growth and Photosynthetic Efficiency of Wheat under Saline Field Conditions

Abstract

Hyperspectral sensing offers a quick and non-destructive alternative for assessing phenotypic parameters of plant physiological status and salt stress tolerance. This study compares the performance of published and modified spectral reflectance indices (SRIs) for estimating and predicting the growth and photosynthetic efficiency of two wheat cultivars exposed to three salinity levels (control, 6.0, and 12.0 dS m⁻¹). Results show that individual SRIs based on visible- and near-infrared (VIS/VIS, NIR/VIS, and NIR/NIR) estimate and predict measured parameters considerably more efficiently than those based on shortwave-infrared (SWIR/VIS and SWIR/NIR), with the exception of some modified indices (the water balance index (WABI-1_{(1550, 482)}, WABI-2_{(1640, 482)}, and WABI-3_{(1650, 531)}), normalized difference moisture index (NDMI_{(1660, 1742)}), and dry matter content index (DMCI_{(1550, 2305)}), which show moderate to strong relationships with measured parameters. Overall results indicate that modified SRIs can serve as rapid and non-destructive high-throughput alternative approaches for tracking growth and photosynthetic efficiency of wheat under salt stress field conditions.

1. Introduction

The different components of salt stress (including osmotic stress, ion imbalance, and specific ion toxicities) interact together to constrain all of the physiological variables involved in plant growth and development. Of the multiple physiological variables that result in a considerable decline in biomass production of plant under salt stress, gas exchange mechanisms and the photosynthetic capacity are the most important. Dramatic changes in leaf turgor pressure and a K⁺ content deficit in leaf tissue under salt stress conditions lead to a significant decline in stomatal conductance (Gs), which then adversely affects photosynthesis (Pn) and transpiration (E) rates [1,2]. It is thus necessary to monitor the responses of these physiological variables to salt stress to enable improvements in the salt tolerance of wheat genotypes or apply appropriate agronomic management practices that alleviate the adverse impacts of salt stress, with the ultimate aim of providing sustainable crop production under salinity stress conditions. In addition, it is necessary to monitor these physiological variables in order to detect the modifications in the physiological status of the plant prior the damage of salinity stress become visible.

Using the physiological variables as key indicators of salinity stress or tolerance entails the use of high plant sampling densities and frequencies, which requires equipment and adequate manpower. Although plant sampling techniques provide highly accurate estimations, they are generally destructive and time- and cost-inefficient, and they also fail to provide up-to-date estimations of the physiological status of plants. In addition, it is often not feasible to use such techniques on a large scale.

An alternative solution that can be used to address these limitations is implementation of the hyperspectral reflectance sensing technique, which offers a cheaper, faster, non-destructive, and broader way to conduct a simultaneous and indirect assessment of hundreds of physiological variables and the status of plant stress under the spatial variability of field conditions. Changes in photosynthetic pigments, internal leaf structure, biochemical components, and leaf water content due to salinity stress are determined by substantial changes in the reflectance signatures of the canopy in three parts of the spectrum: visible region (VIS, 400–700 nm), near-infrared region (NIR, 700–1300 nm), and shortwave-infrared region (SWIR, 1300–2500 nm). The close relationship between the physiological status, canopy spectral reflectance, and plant variables means that spectral reflectance information has the potential to be used in plant phenomics applications and enables indirect estimates and quantification of stress-related plant parameters in a faster and non-destructive manner than other techniques [3,4,5]. However, the large amount of data collected by this technique, which hold thousands of general wavebands between VIS and SWIR domains, often limits its high-throughput applications [6]. Furthermore, the features of these collected wavebands are affected by crop types as well as by environmental and atmospheric conditions. However, to overcome these limitations, the specific and effective wavebands in three parts of the spectrum can be used to calculate different vegetation- and water-based indices by employing simple mathematical operations, such as normalized or ratios. Ultimately, these indices can be used to make indirect assessments of stress-related plant parameters.

Several spectral reflectance indices (SRIs) have been proposed as a proxy for biomass production and photosynthetic efficiency under either normal or stress conditions. For instance, the photochemical reflectance index (PRI = (R₅₃₁ − R₅₇₀)/(R₅₃₁ + R₅₇₀)) has been used in several studies as an indicator for photosynthetic efficiency under stress conditions, and significant linear relationships have been found between this index and Pn and Gs in forest tree, olive, tomato, and wheat crops under water stress conditions [4,7,8,9,10,11,12]. It has also been shown that the normalized difference vegetation index (NDVI = (NIR_λ − Red_λ)/(NIR_λ + Red_λ)), which combines different wavelengths (λ) from the red region (620–690 nm) and near infrared region (760–900 nm), is associated with various stress-related plant parameters in different crops [3,13,14,15]. In addition, Rud et al. [16] developed seven vegetation-based indices constructed mainly from wavelengths from the blue (420–470 nm), green (510–580 nm), red (660–680 nm), and NIR (700–800 nm) to differentiate the growth of eggplant plants (Solanum melongena L.) under different salinity levels. Previous studies have also reported that the high spectral reflectance in the blue region (compared to that in the red, red edge, and NIR spectral regions) contributes to the feasibility of using SRIs constructed from the wavelengths from this region to detect the effects of salinity stress on crop growth [17,18,19,20]. Lara et al. [21] also reported that the spectral reflectance regions (500–600 nm and 710 nm) are associated with salt tolerance in lettuce plants, and Maimaitiyiming et al. [22] found that SRIs constructed from wavelengths located between 400–720 nm of spectral regions are significantly correlated with Gs within a vineyard under different levels of water stress.

The salinity stress not only has negative impacts on photosynthetic pigments and internal leaf structure but also leads to significant changes in plant water status due to osmotic stress of salinity. Therefore, the SRIs incorporating water absorption bands in the NIR and SWIR spectral regions could also successively be used to assess the photosynthetic efficiency and growth under salt stress [23,24]. For example, Poss et al. [23] found that the water band index (WBI = (R₉₀₀/R₉₇₀)), which incorporates one of the water absorption bands from the NIR region (970 nm), reflects the response of wheatgrass growth (Agropyron elongatum L.) to salt stress. In addition, El-Hendawy et al. [25] reported that SRIs combining the NIR and SWIR wavebands are suitable for assessing the growth of spring wheat under salt stress, and SRIs incorporating the NIR band (1321 nm) and SWIR band (2264 nm) was found to be important for assessing responses of the growth of cotton plant to salt stress [26].

Most salinity studies have been conducted in uniform growth media under ideal controlled conditions, and studies under natural saline field conditions are limited due to the temporal and spatial heterogeneity of salt concentrations and soil water content within the field, even over short distances. These conditions within a natural saline field make it very difficult to evaluate salt tolerance. Therefore, in this study, experiments were conducted using simulated close-to-field conditions by employing the subsurface water retention technique (SWRT), which provides homogeneity in salt concentrations and the water content within the root zone for all plants. It also provides a large measuring area and a representative sample size, which is important for high-throughput phenotyping technique.

To provide indirect assessments of plant parameters, the variations in environmental conditions, crop types, growth stages, and levels of stress determined in various studies, which might be the main reason why a universal relationship between published SRIs and measured parameters has not been obtained, require validation of published SRIs, or new ones need to be derived. Therefore, the main objective of this study was to evaluate whether SRIs (published and modified) could be used to assess and predict the growth and photosynthetic efficiency of wheat under salt stress conditions

2. Materials and Methods

2.1. Plant Materials, Experimental Site, and Growth Conditions

Salt-tolerant (Sakha 93) and salt-sensitive (Sakha 61) spring wheat cultivars [25,27] were used in this study. The two cultivars were grown during the 2016/2017 and 2017/2018 growing seasons at the Dierab Research Station of the College of Food and Agriculture Sciences, King Saud University, Saudi Arabia (24°25′ N, 46°34′ E; elevation 400 m). The soil texture was sandy throughout its profile, with a field capacity, wilting point, electrical conductivity (EC), and bulk density of 0.101 m³ m⁻³, 0.035 m³ m⁻³, 0.49 dS m⁻¹, and 1.48 g cm⁻³, respectively. The temperature and precipitation ranged from 9.0 to 35.2 °C and 5 to 30 mm, respectively.

2.2. Setup of Subsurface Water Retention Technique (SWRT), Salinity Treatments, Experimental Design, and Agronomic Practices

According to Fageria and Moreira [28], the root biomass of cereal crops with fibrous root systems is concentrated at a soil depth of 0–40 cm. Therefore, polyethylene membranes with a thickness of 0.3 mm were installed in a U-shape with a 3:1 width/depth aspect ratio at a depth of 40 cm under the soil surface. The width of the base of the membrane sheet was 60 cm, and both sides of the membrane were uplifted by 20 cm. To enable drainage when required, the membrane sheet was not used between a distance of 20 cm from the edge of the two sides of the membrane to the soil surface and at a distance of 10 cm between the two SWRTs.

Three water irrigation salinity levels (control, 6.0, and 12.0 dS m⁻¹) were applied in this study. For the control treatment, plots were irrigated with fresh water (~0.35 dS m⁻¹) during all growth stages; and for moderate (6.0 dS m⁻¹) and high salinity levels (12.0 dS m⁻¹) plots were firstly irrigated with fresh water for 25 days to avoid osmotic shock during germination and at the early seedling stage, and then irrigated with artificial saline water containing 3.51 and 7.02 g NaCl L⁻¹, respectively. A surface irrigation system was used. The main line of the irrigation system, which delivers water from plastic water storage tanks (3.0 m³) to each plot, was equipped with a water meter, which was located at the point where the main line makes contact with the plastic tanks, distributed to sub-main hoses at each plot, and was equipped with a manual control valve to enable equal and constant amounts of water to be delivered to each plot. The class A pan was used to guide the irrigation frequency, and irrigation was conducted whenever the amount of water evaporated from the class A pan reached approximately 80 mm. To monitor salt concentrations in the root zone during the entire period of wheat growth, soil samples at a depth of 0–40 cm were collected from the moderate and high salinity treatments. The electrical conductivity of soil samples was measured using the soil water extract method, with a water to soil suspension ratio of 2:1. The EC of the moderate and the high salinity treatments reached 8.5 and 14.1 dS m⁻¹, respectively, at the grain dough stage.

In both seasons, the field experiment was laid out in a randomized complete block split-plot design and replicated three times. The three salinity levels were assigned to the main plots and the two cultivars were distributed randomly in subplots. Each subplot consisted of two SWRT membrane sheets with eight 6.0-m-long rows spaced 15 cm apart (7.2 m² in total area). The seeds of each cultivar were planted at a seeding rate of 17 g m⁻² on 5 December of each season. Nitrogen, P, and K fertilizers were applied at rates of 180, 90, and 60 kg ha⁻¹ as ammonium nitrate (33.5% N), calcium superphosphate (15.5% P₂O₅), and potassium chloride (60% K₂O), respectively. Nitrogen fertilizer was applied in three equal doses at seeding, stem-elongation, and booting stages, whereas entire doses of phosphorus and potassium were applied prior to sowing and at booting stage, respectively. Disease and weeding control were conducted in a timely manner according to the recommended agronomic practices.

2.3. Growth and Pphotosynthetic Parameter Measurements

At the anthesis growth stage, the above-ground shoots of plant were sampled by cutting two 0.5-m consecutive rows after conducting reflectance measurements. The samples were then cut into small pieces before being placed in drying bags. They were then dried in a forced-air oven at 70 °C to a constant weight, and the dry samples were subsequently weighed to obtain the shoot dry weight per square meter.

The three photosynthesis parameters (net photosynthesis rate (Pn), stomatal conductance (Gs), and transpiration rate (E)) were also measured at the anthesis stage on the second fully expanded leaf from the top of 20 plants for each subplot using a portable gas exchange system (Li-6400; Li-COR Inc., Lincoln, NE, USA) between 09:30 and 12:00. During the measurements, the leaf chamber was set to a leaf temperature of 25 °C and a CO₂ concentration of 400 ppm.

2.4. Canopy Hyperspectral Reflectance Measurements

Together with growth and photosynthetic parameter measurements, canopy hyperspectral reflectance was measured within two hours of solar noon under cloudless conditions using a portable ASD spectroradiometer (Analytical Spectral Devices Inc., Boulder, CO, USA). This device captures the solar radiation reflecting from the plant canopy from 350 to 2500 nm using an optical fiber probe with a 25° field of view. The probe was held vertically at approximately 0.80 m above the plant canopy in the nadir orientation to achieve approximately 23.4 cm in diameter sensing area. The spectral reflectance was originally measured at wavelengths of 1.4 and 2.2 nm sampling intervals from 350 to 1000 nm and 1000 to 2500 nm, respectively. However, the entire spectral range (350–2500 nm) was calculated automatically to resample to 1.0-nm continuous bands. A spectralon white reference panel covered with a mixture of barium sulfate (BaSO₄) and white paint (Labsphere, Inc., North Sutton, NH, USA) was used to generate reflected light percentages and to calibrate the spectroradiometer. Calibrations of each subplot were made: five measurements were taken for each subplot at different points on the three central rows while excluding the first meter of the three rows to eliminate border effects. Each measurement used the average of 10 scans, and this was calculated automatically. Finally, the average of five measurements was recorded as the measured spectrum for a subplot.

2.5. Selection of Published and Modified Spectral Reflectance Indices (SRIs)

Thirteen previously-published SRIs obtained from relevant literature, and the nineteen modified in this study are listed with their equations in

Table 1. Full names of published (P) and developed (D) spectral reflectance indices (SRIs) and their equations.

P or D	Full Indices Name and Abbreviation	Formulation
P	Normalized phaeophytinization index (NPQ)	(R415 − R435)/(R415 + R435)
D	Normalized phaeophytinization index (NPQ)	(R482 − R350)/(R482 + R350)
P	Photochemical reflectance index (PRI)	(R570 − R539)/(R570 + R539)
D	Photochemical reflectance index (PRI)	(R580 − R531)/(R580 + R531)
P	Normalized different vegetation index (NDVI)	(R750 − R705)/(R750 + R705)
D	Normalized different vegetation index (NDVI)	(R780 − R715)/(R780 + R715)
D	Blue normalized difference vegetation index (BNDVI)	(R970 − R420)/(R970 + R420)
D	Green normalized difference vegetation index (GNDVI)	(R970 − R482)/(R970 + R482)
D	Red normalized difference vegetation index (RNDVI-1)	(R970 − R710)/(R970 + R710)
D	Red normalized difference vegetation index (RNDVI-2)	(R1100 − R710)/(R1100 + R710)
P	Ratio analysis of reflectance spectra-a (PARS-a)	(R750/R720)
D	Ratio analysis of reflectance spectra-a (PARS-a)	(R780/R720)
P	Ratio analysis of reflectance spectra-b (PARS-b)	R675/(R650 × R700)
D	Ratio analysis of reflectance spectra-b (PARS-b)	R675/(R640 × R705)
P	Ratio analysis of reflectance spectra-c (PARS-c)	(R760/R500)
D	Ratio analysis of reflectance spectra-c (PARS-c)	(R760/R515)
P	Pigment specific simple ratio-a (PSSR-a)	(R800/R680)
D	Pigment specific simple ratio-c (PSSR-a)	(R800/R690)
P	Pigment specific normalized difference-c (PSND-c)	(R800 − R460)/(R800 + R460)
D	Pigment specific normalized difference-c (PSND-c)	(R800 − R482)/(R800 + R482)
P	Red edge chlorophyll index	R750/R710) − 1
D	Red edge chlorophyll index	R760/R710) − 1
P	Water balance index (WBI)	(R1500 − R538)/(R1500 + R538)
D	Water balance index (WBI)	(R1550 − R482)/(R1550 + R482)
D	Water balance index (WBI)	(R1640 − R482)/(R1640 + R482)
D	Water balance index (WBI)	(R1650 − R531)/(R1650 + R531)
P	Normalized difference water index-2130 (NDWI)	(R858 − R2130)/(R858 + R2130)
D	Normalized difference water index-2130 (NDWI)	(R860 − R2270)/(R860 + R2270)
P	Normalized difference moisture index (NDMI)	(R1649 − R1722)/(R1649 + R1722)
D	Normalized difference moisture index (NDMI)	(R1660 − R1742)/(R1660 + R1742)
P	Dry matter content index (DMCI)	(R2305 − R1495)/(R2305 + R1495)
D	Dry matter content index (DMCI)	(R2305 − R1550)/(R2305 + R1550)

. The published SRIs were selected because the wavelengths incorporated within them are influenced by changes in leaf chlorophyll and other pigment contents, leaf cellular structure, photosynthetic efficiency, and/or changes in plant water status, all of which are influenced by salinity stress. The modified SRIs were built by replacing one or two of the wavelengths in the published SRIs with others; however, these wavelengths of both published and modified SRIs are near to each other. The replaced wavelengths in the modified SRIs were selected based on contour maps (Figure 1), which allow the evaluation of all possible dual wavelength combinations from binary, normalized spectral indices, efficient extraction of significant peak-wavelengths, and the extent of effective regions that enable the assessment of each parameter target being studied [29,30,31]. Contour maps for each parameter were compiled for 4,622,500 combinations of pairs of wavelengths in the full spectral region (350–2500 nm) with a sampling resolution of 1 nm (thus 2150 × 2150 possible pairs). The higher R² values in the contour maps are shown as being progressively white and pale yellow (Figure 1). The R package “lattice” from the software R statistics version 3.0.2 (R foundation for statistical computing 2013) was used to draw the contour maps for spectral reflectance data. Based on hotspots of higher R² values in the contour maps, twenty-six single wavelengths (350, 420, 482, 515, 531, 580, 640, 675, 690, 705, 710, 715, 720, 760, 780, 800, 860, 970, 1100, 1550, 1640, 1650, 1660, 1742, 2270, and 2305) were selected, and these replaced either one or two of the published SRI wavelengths to construct the modified SRIs. However, as previously mentioned, the new wavelengths are still close to those of the published SRIs. For instance, in the published ratio analysis of reflectance spectra-a and c (PARS-a and PARS-c) [32], the wavelengths for pigment specific simple ratio-a (PSSR-a), pigment specific normalized difference-c (PSND-c) [33], Cl red edge [34], and dry matter content index (DMCI) [35], wavelengths of 750, 500, 680, 460, 750, and 1495 nm in these published SRIs were replaced by 780, 515, 690, 482, 760, and 1550 nm, respectively, in the modified SRIs. In addition, in the published normalized phaeophytinization index (NPQ), wavelengths of 415 and 435 nm were replaced by 482 and 350 nm, in the modified one, respectively, and for the three water balance indexes (WABI-1_D, WABI-2_D, and WABI-3_D), 1500 nm was replaced by 1550, 1640, and 1650 nm, and 538 nm was replaced by 482, 482, and 531 nm. Furthermore, in the normalized difference moisture index (NDMI), the two wavelengths (1649 and 1722 nm) were replaced by 1660 and 1742 nm, respectively (Table 1).

2.6. Data Analysis

Data for shoot dry weight (SDW) and photosynthetic parameters were tested using the analysis of variance (ANOVA), which is appropriate for a randomized complete block split-plot design, with the salinity level as the main factor and the cultivar as the split factor. Duncan’s test at the 95% probability level was used to compare differences between the mean values of measured parameters between salinity levels, cultivars, and their interactions. Pearson’s correlation coefficient matrix was used to determine the relationship between all parameters for pooled data and for each salinity level in order to determine the close relationship between SDW and photosynthetic parameters as well as to examine the ability of these parameters as screening criteria for evaluating and improving the salt tolerance of wheat genotypes under saline field conditions. The quantitative relationship between SRIs (as independent variables) and measured parameters (as dependent variables) was fitted with linear and non-linear curve-fitting models (Sigma Plot 11.0, Sytat software Inc., Chocago, IL, USA), and the equation with the highest R² was selected as the best model for evaluating the performance of SRI individually for estimating the measured parameters. The relationships between independent and dependent variables were tested across all data and for each salinity level. Six statistical parameters (including the coefficient of determination (R²), root mean square error (RMSE), mean relative error (RE%), Akaike’s information criterion (AIC), Schwarz’s Bayesian criterion (SBC), and the slope of the linear regression) were calculated and used to evaluate the fitness between the observed and predicted values, and also to quantify the performance of each SRI in assessing the measured parameters. Determining effective SRI models requires that the measured parameters have a high R² value and slope and small RMSE, RE, AIC, and SBC values. The six statistical parameters between observed and predicted data were investigated using XLSTAT statistical package (version 2017.4, Excel Add-ins soft SARL, New York, NY, USA).

3. Results

3.1. Growth and Photosynthetic Parameters

There were significant decreases in SDW and in the three photosynthetic parameters (P_n, Gs, and E) in both cultivars with increases in salinity levels. However, the decreases in all parameters, except E, at both salinity levels (6.0 and 12.0 dS m⁻¹) were more pronounced in Sakha 61 (salt-sensitive cultivar) than in Sakha 93 (salt-tolerant cultivar) (

Table 2. Effects of salinity levels, cultivars, and their interaction on shoot dry weight per square meter, photosynthetic rate, stomatal conductance, and transpiration rate at the anthesis growth stage during two growing seasons.

	Season 2016–2017			Season 2017–2018
Cultivars
	Sakha 93	Sakha 61	Mean	Sakha 93	Sakah 61	Mean
Shoot dry weight (g m−2)
Control	2142.93 a	1926.23 a	2034.58 A	1903.50 a	1683.40 b	1793.45 A
6 dS m−1	1533.27 b	1178.43 c	1355.85 B	1310.43 c	1050.77 d	1180.60 B
12 dS m−1	1171.47 c	919.47 d	1045.47 C	1048.13 d	812.37 e	930.25 C
Mean	1615.89 A	1341.38 B		1420.69 A	1182.18 B
Photosynthetic rate (µmol CO2 m−2 s−1)
Control	21.93 a	15.33 b	18.63 A	18.67 a	14.67 b	16.67 A
6 dS m−1	15.23 b	10.37 c	12.80 B	13.07 b c	9.87 d	11.47 B
12 dS m−1	11.36 c	8.50 c	9.93 C	10.67 c d	7.33 e	9.00 C
Mean	16.18 A	11.40 B		14.13 A	10.62 B
Stomatal conductance (mmol m−2 s−1)
Control	284.13 a	240.80 b	262.47 A	270.87 a	251.53 a	261.20 A
6 dS m−1	179.70 c	126.37 d	153.03 B	155.93 b	110.93 c	133.43 B
12 dS m−1	131.47 d	103.13 e	117.30 C	123.13 c	101.47 c	112.30 B
Mean	198.43 A	156.77 B		183.31 A	154.64 B
Transpiration rate (mmol m−2 s−1)
Control	4.05 a	4.22 a	4.14 A	3.38 a b	3.79 a	3.59 A
6 dS m−1	3.16 b c	3.43 b	3.30 B	2.70 c	3.05 b c	2.88 B
12 dS m−1	2.66 d	2.80 c d	2.73 C	2.07 d	2.31 d	2.19 C
Mean	3.29 A	3.49 A		2.72 B	3.05 A

Means in rows within salinity levels as well as means in columns within cultivar followed by the same letter are not significantly different at the 0.05 level according to the Duncan’s test. The lowercase letters are related to the significant differences between means of interaction, while the uppercase letters are related to the significant differences between means of main factors (salinity or cultivars).

). When averaged across the two seasons, the moderate salinity level (6.0 dS m⁻¹) was found to have resulted in decreases in SDW of 29.7% and 38.2%, P_n of 30.3% and 32.6%, and Gs of 39.5%, and 51.8% for Sakha 93 and Sakha 61, respectively, when compared with the control treatment. At the high salinity level (12.0 dS m⁻¹), the reduction reached 45.2% and 52.0% for SDW, 45.7% and 47.2% for P_n, and 54.1% and 58.4% for Gs, in Sakha 93 and Sakha 61, respectively. The values of E for both cultivars were similar at all salinity treatments (Table 2).

When data for the two seasons, salinity levels, and cultivars were pooled together, there were stronger positive correlations between all parameters (r² = 0.67 − 0.95). In addition, SDW, P_n, and Gs showed stronger positive correlations with each other under each salinity level (r² = 0.71 − 0.95), with the exception of the correlation between SDW and Gs, which showed a non-significant correlation for the control treatment. In addition, E showed a non-significant correlation with the other parameters under control and moderate salinity level (

Table 3. Pearson’s correlation matrix of shoot dry weight and photosynthetic parameters for pooled data (n = 36) and for each salinity level (n = 12).

Parameters	SDW	Pn	Gs	E
	Pooled Data
Shoot dry weight (SDW)	1.00	0.94 ***	0.95 ***	0.71 ***
Photosynthetic rate (Pn)		1.00	0.91 ***	0.57 **
Stomatal conductance (Gs)			1.00	0.67 ***
Transpiration rate (E)				1.00
	Control
Shoot dry weight (SDW)	1.00	0.77 ***	0.23 ns	0.55 ns
Photosynthetic rate (Pn)		1.00	0.71 ***	0.21 ns
Stomatal conductance (Gs)			1.00	−0.42 ns
Transpiration rate (E)				1.00
	6 dS m−1
Shoot dry weight (SDW)	1.00	0.92 ***	0.93 ***	−0.53 ns
Photosynthetic rate (Pn)		1.00	0.90 ***	-0.52 ns
Stomatal conductance (Gs)			1.00	−0.47 ns
Transpiration rate (E)				1.00
	12 dS m−1
Shoot dry weight (SDW)	1.00	0.81 ***	0.95 ***	−0.74 ***
Photosynthetic rate (Pn)		1.00	0.89 ***	−0.55 *
Stomatal conductance (Gs)			1.00	−0.56 *
Transpiration rate (E)				1.00

*, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively. ns: not significant.

).

3.2. Contour Map Analysis of Spectral Reflectance Data

A contour map for each parameter was established using the pooled data of seasons, replications, salinity levels, and cultivars (n = 36) (Figure 1). These maps show the coefficients of determination (R²) of relationships between values of each parameter and all possible combinations between one wavelength on the horizontal axis (wavelength-1) and one wavelength on the vertical axis (wavelength-2) as normalized difference spectral indices in the entire spectrum range (350 to 2500 nm).

In general, contour map analyses conducted for SDW, P_n, and Gs showed higher R² values (R² ≈ 0.90) than those for E (R² ≈ 0.65). The hotspot regions for the greater values of R² were located at 640–1900 nm and 2100–2305 nm on the horizontal axis, and at 380–800 nm on the vertical axis (white color in Figure 1).

3.3. Relationships between Measured Parameters and Different Modified and Published SRIs

To evaluate the efficiency of each SRI in the estimation of measured parameters, the relationships between each parameter and individual SRIs were fitted with linear and non-linear curve-fitting, and the equation with the highest R² was selected as the best model. The equations and values of R² for these relationships are summarized in

Table 4. Best regression and coefficients of determination (R²) models for relationships across all data (n = 36) between measured parameters (shoot dry weight per square meter (SDW), net photosynthesis rate (Pn), stomatal conductance (Gs), and transpiration rate (E)) and different published (P) and developed (D) spectral reflectance indices (SRIs).

SRIs	Par.	Equations	R2	SRIs	Par.	Equations	R2
NPQ_P (415, 435)	SDW	y = 1024x2 + 25472x + 2582.4	0.55 **	RNDVI-1_D (970, 710)	SDW	y = 8993.1x2 − 4456.7x + 1509.2	0.85 ***
	Pn	y = 862.8x2 + 227.54x + 24.07	0.53 **		Pn	y = 65.277x2 − 25.52x + 11.04	0.78 ***
	Gs	y = 13834x2 + 3587.2x + 344.96	0.48 *		Gs	y = 1601.6x2 − 851.6x + 219.15	0.85 ***
	E	y = 226.3x2 + 44.612x + 4.93	0.31 *		E	y = 20.343x2 − 13.70x + 4.95	0.47 *
NPQ_D (482, 350)	SDW	y = 2445.5x2 − 2470.5x + 15	0.73 ***	RNDVI-2_D (1100, 710)	SDW	y = 9994.7x2 − 5374.3x + 1684.2	0.86 ***
	Pn	y = 22.63x2 − 23.69x + 13.80	0.69 ***		Pn	y = 81.569x2 − 39.31x + 13.51	0.80 ***
	Gs	y = 357.6x2 − 371x + 184.39	0.64 **		Gs	y = 1840.4x2 − 1066x + 260.68	0.87 ***
	E	y = 7.099x2 − 3.082x + 3.10	0.38 *		E	y = 22.529x2 − 15.92x + 5.46	0.46 *
PR-I_P (570, 539)	SDW	y = 82929x2 + 4069.1x + 984.78	0.79 ***	PARS-a_P (750, 720)	SDW	y = 823.9x2 − 1824.1x + 1931.7	0.85 ***
	Pn	y = 595.16x2 + 45.39x + 9.73	0.70 ***		Pn	y = 5.06x2 − 7.59x + 9.99	0.77 ***
	Gs	y = 12868x2 + 644.91x + 110.03	0.76 ***		Gs	y = 154.21x2 − 372.2x + 329.19	0.84 ***
	E	y = 180.84x2 + 0.621x + 2.48	0.49 *		E	y = 2.365x2 − 7.192x + 8.15	0.47*
PRI-2_D (580, 531)	SDW	y = 35703x2 − 2709.4x + 986.47	0.78 ***	PARS-a_D (780, 720)	SDW	y = 460.9x2 − 841.58x + 1231.6	0.84 ***
	Pn	y = 241.45x2 − 30.44x + 9.85	0.67 ***		Pn	y = 2.86x2 − 2.118x + 6.15	0.77 ***
	Gs	y = 5476.4x2 − 430.27x + 110.78	0.75 ***		Gs	y = 93.29x2 − 210.47x + 218.04	0.84 ***
	E	y = 79.57x2 − 0.467x + 2.47	0.49 *		E	y = 1.481x2 − 4.699x + 6.45	0.45 *
NDVI_P (750, 705)	SDW	y = 6982.3x2 − 3722.9x + 1430.2	0.84 ***	PARS-b_P (675, 650, 700)	SDW	y = 134.35x + 590.98	0.86 ***
	Pn	y = 50.43x2 − 21.58x + 10.91	0.75 ***		Pn	y = −0.07x2 + 2.114x + 3.59	0.73 ***
	Gs	y = 1207.5x2 − 679.08x + 196.57	0.83 ***		Gs	y = 21.196x + 47.22	0.84 ***
	E	y = 13.35x2 − 9.23x + 4.19	0.44 *		E	y = 0.033x2 − 0.265x + 3.25	0.61 **
NDVI_D (780, 715)	SDW	y = 9955.6x2 − 3817.1x + 1312.3	0.84 ***	PARS-b_D (675, 640, 705)	SDW	y = 223.72x + 435.01	0.85 ***
	Pn	y = 71.46x2 − 20.014x + 9.84	0.77 ***		Pn	y = −0.261x2 + 4.413x − 0.22	0.73 ***
	Gs	y = 1791.4x2 − 753.37x + 183.45	0.84 ***		Gs	y = 35.57x + 21.46	0.84 ***
	E	y = 20.92x2 − 11.34x + 4.18	0.44 *		E	y = 0.066x2 − 0.334x + 3.15	0.58 **
BNDVI_D (970, 420)	SDW	y = 47108x2 − 69666x + 26630	0.75 ***	PARS-c_P (760, 500)	SDW	y = 88.863x + 564.44	0.82 ***
	Pn	y = 416.65x2 − 615.02x + 235.37	0.62 **		Pn	y = 0.8233x + 5.434	0.72 ***
	Gs	y = 6587.3x2 − 9690.1x + 3659.9	0.63 **		Gs	y = 13.497x + 47.894	0.74 ***
	E	y = 75.72x2 − 114.2x + 45.55	0.53 **		E	y = 0.0076x2 − 0.053x + 2.82	0.49 *
GNDVI-1_D (970, 482)	SDW	y = 20269x2 − 26671x + 9661.7	0.83 ***	PARS-c_D (760, 515)	SDW	y = 120.9x + 500.38	0.85 ***
	Pn	y = 174.11x2 − 226.02x + 81.709	0.72 ***		Pn	y = 1.1223x + 4.8244	0.75 ***
	Gs	y = 3040.9x2 − 3998x + 1411.7	0.73 ***		Gs	y = 18.58x + 36.569	0.78 ***
	E	y = 31.669x2 − 43.664x + 17.63	0.45 *		E	y = 0.0163x2 − 0.119x + 2.96	0.51 **
PSSR-a_P (800/680)	SDW	y = 6.296x2 − 24.704x + 1010.8	0.86 ***	WABI-2_D (1640, 482)	SDW	y = 14054x2 − 14360x + 4587.2	0.86 ***
	Pn	y = 0.043x2 + 0.040x + 8.72	0.75 ***		Pn	y = 127.42x2 − 129.75x + 41.77	0.74 ***
	Gs	y = 0.924x2 − 2.975x + 112.08	0.80 ***		Gs	y = 1934.8x2 − 1922.2x + 576.8	0.76 ***
	E	y = 0.0196x2 − 0.246x + 3.39	0.57 **		E	y = 26.95x2 − 30.56x + 11.29	0.57 **
PSSR-a_D (800/690)	SDW	y = 8.43x2 − 22.399x + 993.98	0.88 ***	WABI-3_D (1650, 531)	SDW	y = 12840x2 − 8345.3x + 2334.2	0.85 ***
	Pn	y = 0.056x2 + 0.123x + 8.52	0.77 ***		Pn	y = 135.23x2 − 93.52x + 25.72	0.74 ***
	Gs	y = 1.318x2 − 3.503x + 111.35	0.84 ***		Gs	y = 1748.3x2 − 1045.6x + 258.4	0.82 ***
	E	y = 0.028x2 − 0.294x + 3.41	0.58 **		E	y = 26.71x2 − 20.87x + 6.78	0.57 **
PSNDc_P (800, 460)	SDW	y = 23347x2 − 31442x + 11458	0.80 ***	NDWI_P (2130, 858)	SDW	y = 5155.2x2 − 2697.3x + 1214.3	0.55 **
	Pn	y = 194.83x2 − 258.43x + 93.95	0.68 ***		Pn	y = 45.93x2 − 21.81x + 10.28	0.54 **
	Gs	y = 3529.4x2 − 4757x + 1698.5	0.71 ***		Gs	y = 993.66x2 − 620.97x + 195.62	0.51 **
	E	y = 35.358x2 − 49.46x + 19.85	0.44 *		E	y = 2.32x2 − 0.275x + 2.53	0.15ns
PSNDc_D (800, 482)	SDW	y = 18274x2 − 23774x + 8605.3	0.81 ***	NDWI_D (2270, 860)	SDW	y = 6214.8x2 − 3729.7x + 1415	0.56 **
	Pn	y = 155.16x2 − 198.53x + 71.72	0.71 ***		Pn	y = 55.64x2 − 31.028x + 11.96	0.55 **
	Gs	y = 2790.9x2 − 3638x + 1280.8	0.73 ***		Gs	y = 1201.7x2 − 834.2x + 242.56	0.52 **
	E	y = 26.94x2 − 36.663x + 15.05	0.42 *		E	y = 3.016x2 − 0.825x + 2.59	0.16 ns
Cl red edge_P (750/710)	SDW	y = 168.2x2 − 8.77x + 923.89	0.86 ***	NDMI_P (1649, 1722)	SDW	y = −0.0004x2 + 26304x − 2984	0.36 *
	Pn	y = 0.953x2 + 1.82x + 7.65	0.76 ***		Pn	y = −27656x2 + 1988.3x − 19.13	0.35 *
	Gs	y = 30.92x2 − 14.15x + 106.36	0.85 ***		Gs	y = −280929x2 + 24315x − 258.8	0.39 *
	E	y = 0.537x2 − 1.082x + 3.22	0.50 **		E	y = −4245.9x2 + 284.52x − 1.28	0.20 ns
Cl red edge_D (760/710)	SDW	y = 125.38x2 + 39.65x + 900.33	0.85 ***	NDMI_D (1660, 1742)	SDW	y = 920979x2 − 42541x + 1380	0.61 **
	Pn	y = 0.706x2 + 1.947x + 7.52	0.77 ***		Pn	y = 9010.4x2 − 419.68x + 13.14	0.58 **
	Gs	y = 23.625x2 − 5.966x + 102.91	0.85 ***		Gs	y = 247530x2 − 15224x + 338.9	0.64 **
	E	y = 0.420x2 − 0.899x + 3.17	0.50 **		E	y = 1072.6x2 − 57.53x + 3.48	0.22 ns
WABI _P (1500, 538)	SDW	y = 7556.3x2 − 1071.9x + 1141.1	0.24 ns	DMCI_P (1495, 2305)	SDW	y = −42042x2 − 18582x − 458.41	0.13ns
	Pn	y = 136.13x2 − 41.74x + 14.12	0.22 ns		Pn	y = −71.945x + 1.923	0.16 ns
	Gs	y = 1273.4x2 − 186.43x + 132.76	0.26 ns		Gs	y = −5273.3x2 − 2541.8x − 91.30	0.13 ns
	E	y = −0.414x2 + 4.55x + 2.044	0.21 ns		E	y = −44.66x2 − 21.053x + 0.968	0.08 ns
WABI -1_D (1550, 482)	SDW	y = 12469x2 − 10741x + 32	0.84 ***	DMCI_D (1550, 2305)	SDW	y = 34163x2 + 13302x + 2241	0.64 **
	Pn	y = 117.83x2 − 103.06x + 31.32	0.71 ***		Pn	y = 390.82x2 + 161.36x + 25.51	0.65 **
	Gs	y = 1630.6x2 − 1318.7x + 360.49	0.74 ***		Gs	y = 7391.7x2 + 3233x + 461.37	0.63 **
	E	y = 24.696x2 − 24.396x + 8.65	0.60 **		E	y = 36.89x2 + 15.26x + 4.32	0.23 ns

. The relationships were analyzed using pooled data relating to years, replications, salinity levels, and cultivars. Based on the values of R², second-order relationships were found to be the best models for fitting the relationships between SRIs and all measured parameters, with the exception of the following, which showed linear relationships: the published and modified index (ratio analysis of reflectance spectra-c, PARS-c_P and PARS-c_D) for SDW, P_n, and Gs; the published and modified index (ratio analysis of reflectance spectra-b, PARS-b_P PARS-b_D) for SDW and Gs; and the published index (dry matter content index, DMCI_P) for P_n (Table 4). A comparison between all the SRIs showed that the vegetation-based indices, which are related to bands of internal leaf structure, pigment contents and photosynthetic efficiency, and they incorporate VIS/VIS, NIR/VIS, and NIR/NIR wavelengths, generally exhibited higher values for R² with measured parameters than the water-based indices that qualified for monitoring plant water status, and which incorporate SWIR/VIS and SWIR/NIR wavelengths. However, the exceptions were for the three modified water balance indexes (WABI-1_D, WABI-2_D, and WABI-3_D), normalized difference moisture index (NDMI_D), and DMCI_D, which had comparable R² to those of the vegetation-based indices. All the SRIs estimated SDW, P_n, and Gs better than E, except for the published WABI_P and DMCI_P, which showed a non-significant relationship with all measured parameters. With the exception of the three modified WABI, which showed a moderate relationship (R² values ranging from 0.57 to 0.60), none of the water-based indices showed a relationship with E. However, some of the modified SRIs, such as the normalized phaeophytinization index (NPQ_D), performed better than the original published SRIs (Table 4).

3.4. Relationships between Measured Parameters and SRIs under Each Salinity Level

The value of R² and best regression models used to determine relationships between the thirty-two different SRIs and measured parameters for each salinity level are summarized in

Table 5. Best models of regression and determination coefficients (R²) for relationships between difference published (P) and developed (D) spectral reflectance indices (SRIs) and measured parameters (shoot dry weight per square meter (SDW), net photosynthesis rate (Pn), stomatal conductance (Gs), and transpiration rate (E)) for each salinity level. L and Q indicate linear and quadratic fitting models, respectively; C, S1, and S2 indicate control, 60, and 120 mM NaCl, respectively.

Spectral Indices	SDW			Pn			Gs			E
	C	S1	S2	C	S1	S2	C	S1	S2	C	S1	S2
NPQ_P (415, 435)	0.48 Q	0.62 Q	0.38 Q	0.40 Q	0.58 Q	0.38 Q	0.37 Q	0.35 Q	0.40 L	0.37 Q	0.47 L	0.17 Q
NPQ_D (482, 350)	0.50 Q	0.63 Q	0.56 L	0.52 Q	0.57 L	0.50 L	0.54 Q	0.47 L	0.43 L	0.50 Q	0.47 Q	0.10 Q
PRI_P (570, 539)	0.47 Q	0.56 L	0.56 L	0.36 Q	0.73 Q	0.66 Q	0.33 Q	0.68 Q	0.59 Q	0.45 Q	0.29 L	0.10 Q
PRI_D (580, 531)	0.43 Q	0.56 L	0.55 L	0.22 Q	0.72 Q	0.66 Q	0.16 Q	0.66 Q	0.57 Q	0.49 Q	0.28 L	0.14 Q
NDVI_P (750, 705)	0.42 Q	0.67 L	0.63 L	0.18 L	0.71 Q	0.68 L	0.20 Q	0.81 Q	0.57 L	0.11 Q	0.30 Q	0.16 Q
NDVI_D (780, 715)	0.39 Q	0.66 L	0.65 L	0.28 Q	0.69 L	0.72 L	0.38 Q	0.79 Q	0.57 L	0.05 Q	0.29 Q	0.18 Q
BNDVI_D (970, 420)	0.41 L	0.33 L	0.52 L	0.27 Q	0.19 Q	0.52 Q	0.43 Q	0.32 L	0.36 Q	0.46 Q	0.12 Q	0.11 Q
GNDVI_D (970, 482)	0.55 Q	0.60 Q	0.69 Q	0.46 Q	0.58 Q	0.64 Q	0.57 Q	0.55 L	0.49 L	0.57 Q	0.36 Q	0.06 Q
RNDVI-1_D (970, 710)	0.36 Q	0.69 L	0.63 L	0.25 Q	0.71 L	0.70 L	0.28 L	0.79 Q	0.56 Q	0.03 Q	0.21 Q	0.16 Q
RNDVI-2_D (1100, 710)	0.41 Q	0.66 Q	0.71 Q	0.40 Q	0.71 Q	0.72 Q	0.50 Q	0.78 Q	0.61 Q	0.23 Q	0.38 Q	0.17 Q
PARSa_P (750, 720)	0.44 Q	0.65 L	0.65 L	0.26 Q	0.72 Q	0.73 L	0.29 Q	0.81 Q	0.57 Q	0.04 Q	0.28 Q	0.20 Q
PARSa_D (780, 720)	0.31 Q	0.65 Q	0.65 Q	0.29 Q	0.72 Q	0.74 Q	0.46 Q	0.81 Q	0.57 Q	0.09 Q	0.27 Q	0.19 Q
PARSb_P (675, 650, 700)	0.45 Q	0.69 L	0.67 Q	0.02 Q	0.75 Q	0.66 L	0.08 Q	0.84 Q	0.52 Q	0.71 Q	0.20 Q	0.19 Q
PARSb_D (675, 640, 705)	0.51 Q	0.64 L	0.65 Q	0.08 Q	0.69 Q	0.63 Q	0.02 Q	0.74 Q	0.46 Q	0.64 Q	0.16 Q	0.18 Q
PARSc_P (760, 500)	0.42 Q	0.60 Q	0.70 Q	0.45 Q	0.56 Q	0.68 Q	0.61 Q	0.64 Q	0.55 Q	0.63 Q	0.19 Q	0.11 Q
PARSc_D (760, 515)	0.42 Q	0.63 Q	0.69 Q	0.40 Q	0.59 Q	0.69 L	0.58 Q	0.70 Q	0.56 Q	0.55 Q	0.18 L	0.14 Q
PSSRa_P (800/680)	0.63 L	0.61 Q	0.67 Q	0.40 Q	0.66 Q	0.70 L	0.37 Q	0.77 Q	0.60 Q	0.41 Q	0.26 Q	0.17 Q
PSSRa_D (800/690)	0.61 L	0.63 Q	0.65 Q	0.44 Q	0.70 Q	0.68 Q	0.38 Q	0.81 Q	0.59 Q	0.30 Q	0.25 Q	0.17 Q
PSNDc_P (800, 460)	0.37 Q	0.55 Q	0.71 Q	0.29 Q	0.50 Q	0.65 Q	0.51 Q	0.51 Q	0.51 Q	0.55 Q	0.40 Q	0.06 Q
PSNDc_D (800, 482)	0.55 Q	0.58 Q	0.72 Q	0.49 Q	0.56 Q	0.67 Q	0.61 Q	0.55 L	0.53 Q	0.61 Q	0.41 Q	0.07 Q
Cl red edge_P (750/710)	0.43 Q	0.66 Q	0.65 Q	0.20 Q	0.74 Q	0.70 L	0.21 Q	0.84 Q	0.58 Q	0.07 Q	0.27 Q	0.19 Q
Cl red edge_D (760/710)	0.42 Q	0.66 Q	0.65 Q	0.22 Q	0.74 Q	0.71 Q	0.25 Q	0.84 Q	0.58 Q	0.06 Q	0.27 Q	0.19 Q
WABI _P (1500, 538)	0.45 Q	0.25 Q	0.46 Q	0.43 Q	0.50 Q	0.41 Q	0.48 Q	0.47 Q	0.46 Q	0.10 Q	0.23 Q	0.09 Q
WABI -1_D (1550, 482)	0.66 Q	0.55 Q	0.51 Q	0.38 Q	0.51 Q	0.43 Q	0.39 Q	0.47 Q	0.46 Q	0.51 Q	0.08 Q	0.01 Q
WABI -2_D (1640, 482)	0.65 Q	0.58 Q	0.66 L	0.40 Q	0.56 Q	0.58 Q	0.43 Q	0.51 Q	0.51 Q	0.53 L	0.19 Q	0.04 Q
WABI -3_D (1650, 531)	0.56 Q	0.53 Q	0.11 Q	0.48 Q	0.47 Q	0.13 Q	0.51 Q	0.48 Q	0.14 Q	0.20 L	0.03 Q	0.01 Q
NDWI_P (2130, 858)	0.27 Q	0.51 Q	0.59 Q	0.34 Q	0.50 Q	0.59 L	0.37 Q	0.52 Q	0.42 L	0.14 Q	0.44 Q	0.11 Q
NDWI_D (2270, 860)	0.25 Q	0.52 Q	0.58 Q	0.26 Q	0.51 Q	0.59 Q	0.33 Q	0.54 Q	0.42 L	0.08 Q	0.44 Q	0.10 Q
NDMI_P (1649, 1722)	0.02 Q	0.23 Q	0.03 Q	0.19 Q	0.28 Q	0.02 Q	0.68 Q	0.23 Q	0.08 Q	0.13 Q	0.11 Q	0.15 Q
NDMI_D (1660, 1742)	0.14 Q	0.55 Q	0.47 Q	0.26 Q	0.48 Q	0.46 Q	0.51 Q	0.57 Q	0.37 Q	0.05 Q	0.57 Q	0.06 Q
DMCI_P (1495, 2305)	0.50 Q	0.21 Q	0.01 Q	0.62 Q	0.28 Q	0.01 Q	0.61 Q	0.15 Q	0.01 Q	0.42 Q	0.01 Q	0.07 Q
DMCI_D (1550, 2305)	0.44 Q	0.43 Q	0.54 Q	0.58 Q	0.45 Q	0.51 Q	0.50 Q	0.40 Q	0.42 Q	0.06 Q	0.39 Q	0.08 Q

. Most SRIs had curvilinear relationships (a few had linear relationships) with the measured parameters under each salinity level. All SRIs provided better estimations of SDW, P_n, and Gs than of E. The relationships between SRIs and SDW, P_n, and Gs fitted better under moderate (6 dS m⁻¹) and high (12 dS m⁻¹) salinity levels than under the control, but the opposite was true for E. Only 10, 3, 12, and 10 of the thirty-two SRIs showed moderate relationships (0.50 ≤ R² ≤ 0.68) with SDW, Pn, Gs, and E under control conditions, respectively. Under moderate and high salinity levels, most of the vegetation-based indices exhibited higher values of R² (0.50 ≤ R² ≤ 0.84) with SDW, P_n, and Gs than the water-based indices (0.50 ≤ R² ≤ 0.68) (Table 5).

3.5. Validation of Predictive Models for Measured Parameters Based on Different SRIs

The four measured parameters were estimated from the developed predictive models based on the thirty-two different SRIs individually. Six statistical parameters (see Materials and Methods) were used to test the accuracy of the validation of predictive models, and they were calculated from observed and predicted data of measured parameters for the second year. These statistical evaluation parameters are summarized in

Table 6. Validation statistics of predictive models for shoot dry weight per square meter (SDW) and net photosynthesis rate (Pn). R², RMSE, RE, AIC, and SBC represent coefficient of determination, root mean square error, mean relative error (%), Akaike’s information criterion, and Schwarz’s Bayesian criterion, respectively.

Spectral Indices	R2	RMSE	RE	AIC	SBC	Slope	R2	RMSE	RE	AIC	SBC	Slope
	Shoot Dry Weight (SDW)						Net Photosynthesis Rate (Pn)
NPQ_P (415, 435)	0.69	261.2	16.2	202.2	204.0	0.76	0.66	2.75	16.7	38.3	40.0	0.79
NPQ_D (482, 350)	0.83	190.8	11.6	190.9	192.7	0.94	0.80	2.14	13.4	29.2	31.0	0.96
PRI_P (570, 539)	0.69	259.6	16.1	202.0	203.8	0.77	0.61	2.94	18.9	40.7	42.5	0.72
PRI_D (580, 531)	0.68	262.4	16.1	202.4	204.2	0.76	0.60	3.00	18.9	41.5	43.3	0.71
NDVI_P (750, 705)	0.81	202.6	12.4	193.1	194.9	0.91	0.74	2.40	15.0	33.5	35.3	0.90
NDVI_D (780, 715)	0.85	180.2	10.6	188.9	190.6	0.96	0.81	2.08	13.0	28.3	30.1	0.97
BNDVI_D (970, 420)	0.71	253.1	15.8	201.1	202.9	0.81	0.57	3.11	20.9	42.7	44.5	0.74
GNDVI_D (970, 482)	0.78	218.1	12.3	195.7	197.5	0.89	0.67	2.70	17.3	37.7	39.5	0.84
RNDVI-1_D (970, 710)	0.86	175.7	10.0	188.0	189.7	0.97	0.81	2.05	13.2	27.7	29.5	0.98
RNDVI-2_D (1100, 710)	0.87	166.2	9.7	186.0	187.7	0.99	0.84	1.90	12.0	24.9	26.7	1.01
PARSa_P (750, 720)	0.89	154.0	8.7	183.2	185.0	1.01	0.84	1.87	11.9	24.4	26.1	1.01
PARSa_D (780, 720)	0.90	149.1	8.3	182.0	183.8	1.02	0.87	1.70	11.1	21.0	22.8	1.04
PARSb_P (675, 650, 700)	0.93	123.3	7.6	175.2	177.0	1.04	0.80	2.18	12.8	29.9	31.7	0.95
PARSb_D (675, 640, 705)	0.92	130.1	8.1	177.1	178.9	1.04	0.81	2.08	12.2	28.3	30.1	0.97
PARSc_P (760, 500)	0.87	167.8	9.6	186.3	188.1	0.98	0.80	2.31	14.2	32.0	33.7	0.93
PARSc_D (760, 515)	0.91	141.0	8.0	180.0	181.8	1.02	0.82	2.03	12.7	27.3	29.1	0.99
PSSRa_P (800/680)	0.86	175.8	10.7	188.0	189.8	0.96	0.74	2.41	14.6	33.6	35.4	0.89
PSSRa_D (800/690)	0.89	155.7	9.5	183.6	185.4	0.99	0.77	2.25	13.8	31.1	32.8	0.93
PSNDc_P (800, 460)	0.77	226.3	12.9	197.1	198.9	0.87	0.64	2.84	18.3	39.5	41.3	0.81
PSNDc_D (800, 482)	0.77	226.0	13.0	197.0	198.8	0.87	0.66	2.77	17.8	38.5	40.3	0.82
Cl red edge_P (750/710)	0.91	143.1	8.2	180.6	182.3	1.02	0.83	1.94	12.2	25.7	27.4	1.00
Cl red edge_D (760/710)	0.91	139.5	8.1	179.6	181.4	1.03	0.85	1.86	11.9	24.3	26.1	1.01
WABI _P (1500, 538)	0.35	376.1	23.0	215.4	217.1	0.39	0.34	3.86	23.5	50.5	52.2	0.42
WABI -1_D (1550, 482)	0.85	183.7	10.5	189.6	191.3	0.96	0.71	2.57	16.3	35.8	37.6	0.89
WABI -2_D (1640, 482)	0.83	190.8	10.3	190.9	192.7	0.95	0.70	2.60	16.3	36.3	38.1	0.87
WABI -3_D (1650, 531)	0.85	182.7	10.5	189.4	191.2	0.95	0.76	2.34	13.8	32.5	34.3	0.94
NDWI_P (2130, 858)	0.59	298.6	16.9	207.0	208.8	0.67	0.54	3.20	18.8	43.8	45.5	0.66
NDWI_D (2270, 860)	0.61	291.7	16.4	206.2	208.0	0.70	0.57	3.13	18.6	42.9	44.7	0.69
NDMI_P (1649, 1722)	0.39	363.5	20.9	214.1	215.9	0.48	0.42	3.62	20.3	48.2	50.0	0.52
NDMI_D (1660, 1742)	0.69	261.0	16.0	202.2	204.0	0.80	0.62	2.94	19.0	40.7	42.5	0.76
DMCI_P (1495, 2305)	0.19	421.1	25.0	219.4	221.2	0.23	0.23	4.15	25.9	53.1	54.9	0.30
DMCI_D (1550, 2305)	0.70	254.2	15.7	201.3	203.0	0.81	0.64	2.86	18.6	39.7	41.5	0.79

Bold numbers indicate more accurate estimation of measured parameters and fulfill the best values of the six statistical evaluation parameters.

and

Table 7. Validation statistics of predictive models for stomatal conductance (Gs) and transpiration rate (E). R², RMSE, RE, AIC, and SBC represent coefficient of determination, root mean square error, mean relative error (%), Akaike’s information criterion, and Schwarz’s Bayesian criterion, respectively.

Spectral Indices	R2	RMSE	RE	AIC	SBC	Slope	R2	RMSE	RE	AIC	SBC	Slope
	Stomatal Conductance (Gs)						Transpiration Rate (E)
NPQ_P (415, 435)	0.68	39.78	20.16	134.5	136.3	0.65	0.28	0.54	14.40	−20.3	−18.5	0.19
NPQ_D (482, 350)	0.80	31.33	15.01	125.9	127.7	0.77	0.39	0.50	12.88	−23.4	−21.6	0.30
PRI_P (570, 539)	0.67	40.21	21.46	134.9	136.6	0.66	0.32	0.53	13.64	−21.2	−19.4	0.21
PRI_D (580, 531)	0.66	40.95	21.40	135.5	137.3	0.65	0.32	0.53	13.54	−21.2	−19.5	0.21
NDVI_P (750, 705)	0.77	33.65	15.69	128.5	130.2	0.74	0.40	0.49	12.80	−23.7	−21.9	0.30
NDVI_D (780, 715)	0.81	30.26	14.00	124.6	126.4	0.78	0.45	0.47	12.08	−25.1	−23.3	0.34
BNDVI_D (970, 420)	0.63	42.57	24.35	136.9	138.7	0.62	0.49	0.46	11.20	−26.3	−24.5	0.45
GNDVI_D (970, 482)	0.72	37.14	18.61	132.0	133.8	0.71	0.40	0.50	12.70	−23.4	−21.7	0.32
RNDVI-1_D (970, 710)	0.82	29.81	13.82	124.1	125.9	0.78	0.46	0.47	11.89	−25.5	−23.7	0.36
RNDVI-2_D (1100, 710)	0.84	27.95	12.59	121.8	123.6	0.80	0.46	0.47	11.87	−25.3	−23.6	0.36
PARSa_P (750, 720)	0.85	26.77	12.12	120.2	122.0	0.82	0.50	0.45	11.46	−26.9	−25.1	0.39
PARSa_D (780, 720)	0.87	25.77	11.80	118.9	120.6	0.82	0.51	0.45	11.18	−27.2	−25.4	0.40
PARSb_P (675, 650, 700)	0.88	23.89	11.56	116.1	117.9	0.87	0.64	0.38	9.48	−32.9	−31.2	0.53
PARSb_D (675, 640, 705)	0.89	23.50	10.81	115.5	117.3	0.87	0.66	0.37	9.20	−33.5	−31.7	0.55
PARSc_P (760, 500)	0.84	28.31	14.11	122.2	124.0	0.83	0.54	0.43	10.86	−28.1	−26.4	0.43
PARSc_D (760, 515)	0.87	25.48	12.73	118.4	120.2	0.85	0.57	0.42	10.51	−29.5	−27.7	0.46
PSSRa_P (800/680)	0.80	31.39	14.87	126.0	127.7	0.80	0.49	0.45	11.51	−26.6	−24.8	0.38
PSSRa_D (800/690)	0.83	28.73	13.66	122.8	124.6	0.83	0.53	0.44	11.01	−27.9	−26.1	0.41
PSNDc_P (800, 460)	0.70	38.36	19.61	133.2	135.0	0.69	0.40	0.49	12.52	−23.6	−21.9	0.34
PSNDc_D (800, 482)	0.71	38.10	19.29	132.9	134.7	0.69	0.37	0.50	12.98	−22.8	−21.0	0.30
Cl red edge_P (750/710)	0.86	26.22	11.98	119.5	121.3	0.83	0.53	0.44	11.22	−27.8	−26.0	0.41
Cl red edge_D (760/710)	0.87	25.61	11.69	118.6	120.4	0.84	0.53	0.44	11.11	−28.0	−26.2	0.42
WABI _P (1500, 538)	0.35	56.46	28.58	147.1	148.9	0.32	0.41	0.49	11.39	−23.8	−22.0	0.40
WABI -1_D (1550, 482)	0.80	31.18	16.66	125.7	127.5	0.79	0.55	0.43	10.92	−28.6	−26.8	0.48
WABI -2_D (1640, 482)	0.78	32.71	17.05	127.4	129.2	0.77	0.49	0.45	11.46	−26.5	−24.7	0.42
WABI -3_D (1650, 531)	0.81	30.97	15.21	125.5	127.3	0.78	0.61	0.40	10.10	−31.3	−29.6	0.54
NDWI_P (2130, 858)	0.53	48.31	23.86	141.5	143.3	0.52	0.23	0.56	13.84	−19.0	−17.2	0.16
NDWI_D (2270, 860)	0.54	47.45	23.93	140.8	142.6	0.53	0.24	0.55	13.62	−19.3	−17.6	0.17
NDMI_P (1649, 1722)	0.36	56.13	26.60	146.9	148.7	0.31	0.33	0.52	11.16	−21.5	−19.7	0.30
NDMI_D (1660, 1742)	0.63	42.62	23.26	137.0	138.7	0.61	0.36	0.51	12.06	−22.5	−20.7	0.28
DMCI_P (1495, 2305)	0.16	64.26	32.32	151.7	153.5	0.12	0.15	0.59	13.49	−17.3	−15.5	0.16
DMCI_D (1550, 2305)	0.63	42.57	22.50	136.9	138.7	0.61	0.35	0.52	12.06	−22.0	−20.2	0.28

Bold numbers indicate more accurate estimation of measured parameters and fulfill the best values of the six statistical evaluation parameters.

. The models that fulfilled the highest values of R² and slope together with the lowest values of RMSE, RE, AIC, and SBC were selected to make accurate predictions of measured parameters. In general, thirty-two different SRI models provided estimations of SDW, P_n, and Gs that were more accurate than those of E. The vegetation-based indices provided more accurate estimations of the four measured parameters than the water-based indices, with the exception of the three modified WABI, which exhibited values of the six statistical evaluation parameters that were comparable to those of the vegetation-based indices. Of the thirty-two SRIs, eighteen delivered the best values for the six statistical evaluation parameters for SDW, P_n, Gs, and eleven for E. The corresponding R² for these best models ranged from 0.81 to 0.93, 0.70 to 0.87, 0.77 to 0.88, and 0.50 to 0.66 for SDW, P_n, Gs, and E, respectively. Compared to the other models, these models fulfilled the lowest values of RMSE, RE, AIC, and SBC and the highest values of slope. Importantly, some of the SRIs modified in this study provided equal (and sometimes better) estimations of measured parameters than the original published SRIs (Table 6 and Table 7).

4. Discussion

The results indicate that plant biomass production (SDW) has significant associations with the photosynthetic apparatus under salinity conditions, particularly with Pn and Gs (Table 3). In addition, the positive or negative correlation between tested parameters indicates that the photosynthetic parameters along with SDW could be used simultaneously as effective screening tools for evaluating and improving the salt tolerance of wheat genotypes under saline field conditions. The negative correlation between E and other parameters indicates that the various stomatal characteristics such as stomatal size and density could play a vital role in controlling transpiration rate particularly under abiotic stress [36]. However, the relative importance of these parameters as effective screening tools for evaluating salt tolerance depends on the ability to detect them frequently in a large-scale field under realistic simulation saline field conditions and in a fast and non-destructive manner. Real-time detection of these parameters is significantly important to provide useful information that enables a comprehensive understanding of how salinity affects crop growth and productivity and for making decisions regarding implementing suitable agronomic practices to alleviate the adverse effects of salt stress.

There are few studies have investigated the performance of spectral reflectance indices (SRIs) for detecting the variations of these parameters under realistic saline field conditions. El-Hendawy et al. [25] and Hackl et al. [37] reported that, in salinity studies, the growth platform may impact on the use of spectral measurements to assess the phenotypic parameters and in the exploration of the importance of high-throughput phenotyping technique for evaluating salinity tolerance. Hackl et al. [36] found that the relationships between SRIs and phenotypic parameters under simulated close-to-field conditions were more robust compared with the results of pot-grown plants conditions. El-Hendawy et al. [38] also found that the hyperspectral data were successfully used for indirect tracking changes in ion content and leaf water relations of two wheat genotypes differing in their salt tolerance under SWRT-growth platform. Therefore, the growth platform of the SWRT technique, which their advantages have been mentioned in the introduction section, may be able to test the performance of published and modified SRIs for tracking changes in growth and photosynthetic efficiency of wheat under saline field conditions.

Comparison between Published and Modified Spectral Reflectance Indices (SRIs) for Estimating the Measured Parameters under Salinity Conditions

Various changes in biophysical and biochemical characteristics of the canopy (such as leaf chlorophyll and other photosynthetic pigments contents, leaf tissue structure, dry matter accumulation, photosynthetic efficiency, and plant water status) have been found to be change in relation to osmotic and ionic components of salinity stress, and these cause noticeable variability in the spectral reflectance of the canopy in the three parts of the spectrum (VIS, NIR, and SWIR) domains [4,21,25,26]. Therefore, several SRIs that incorporate wavelengths related to the abovementioned main physiological characteristics of plants were modified to indirectly estimate the growth and photosynthetic efficiency of crops under various environmental conditions.

In this study, we tested the accuracy of the most commonly published SRIs in estimating measured parameters under salinity conditions and compared the results with the performances of the SRIs modified in this study. The results of this study found that replacing wavelengths in the modified SRIs improved the accuracy of these SRIs in estimating measured parameters when either all the data of salinity levels and genotypes were pooled together (Table 4) or when salinity levels were analyzed separately (Table 5) compared to the like-published SRIs. These findings were viewed with the NPQ-D, three water balance index (WABI-1_D, WABI-2_D, and WABI-3_D), NDMI_D, and the DMCI_D (Table 4 and Table 5). These findings indicate that replacing some wavelengths with others could improve the accuracy of modified SRIs when estimating phenotypic parameters if used in certain environmental conditions. In addition, some wavelengths that have provided significant results in previous studies cannot be universally applied in all environmental conditions. Different crop types, growth stages, and levels of stress might be the main reasons why a universal relationship between published SRIs and measured parameters has not yet been obtained. Zhou et al. [39] similarly found that replacing the wavelengths 500 and 700 nm in the carotenoid reflectance index (CRI₅₀₀ or CRI₇₀₀) that was constructed by Gitelson et al. [13] by a NIR band 770 nm could improve the estimation accuracy of leaf carotenoids content.

The results of this study also found that the vegetation-based indices, VIS/VIS, NIR/VIS, and NIR/NIR, were much more effective for estimating and predicting the biomass production and photosynthetic properties of wheat under salinity conditions than the water-based indices, SWIR/VIS and SWIR/NIR, with the exception of the three modified WABI indices, the NDMI index, and the DMCI index, which showed moderate to strong relationships with measured parameters (Table 4). These results indicate that decreases in leaf chlorophyll, other photosynthetic pigment contents, and photosynthetic efficiency as well as the damage of leaf structure are logical considerations under salinity stress. Therefore, SRIs used in previous studies that are significantly related to changes in photosynthetic pigments and leaf structure could be modified to detect the growth and photosynthetic properties of wheat under salinity stress. In addition, the wavelengths in the VIS and NIR regions show a considerable potential for use in the development of previously published SRIs to enable them to detect salinity effects. For instance, the sensitivity of the two wavelengths incorporated in the PRI index to both xanthophyll de-epoxidation state and photosynthetic relative pigment contents have been shown in previous studies [40,41,42] and in this current study to provide this index with more ability to efficiently estimate the photosynthetic efficiency (Pn and Gs) and dry matter accumulation under a wide range of environmental stresses (Table 4 and Table 5). Gamon et al. [43] also reported that significant decreases in the values of PRI index under moderate salinity level (50 mM NaCl) indicate the reduced photochemical efficiency and epoxidation of xanthophyll cycle pigments. Previous studies have also reported that the spectral region of 420–470 nm, which is the blue region, is effective for characterizing the absorption features of photosynthetic pigments (such as chlorophyll-a and -b, α- and β- carotenes, and lutein) that have been affected by salinity stress [16,44]. In addition, the blue (480–500 nm), green (520–580 nm), red (640–680 nm), red edge (720–770 nm), and NIR (812–868, 884–809, and 918–930 nm) regions, which are related to photosynthetic pigments absorption features, leaf structure, and water absorption bands in the NIR region, have also been found to be regions sensitive to salt stress in different crops [19,21,45]. Similarly, strong linear and non-linear relationships have been found between vegetation indices formulated from VIS and NIR bands and Pn and Gs in plants subjected to different environmental stresses under field conditions [9,11,12,22,46,47]. Furthermore, Gitelson et al. [48] reported that the two VIS wavelengths (450 ± 20 nm and 550 ± 20 nm) and the two NIR wavelengths (715 ± 20 nm and the band above 750 nm) were sufficient for estimating the total pigment contents of maple, chestnut, and beech leaves, which have a wide pigment content range and composition.

With respect to the above information, our results confirm that (1) the SRIs formulated based on the VIS, red edge, and NIR of the spectrum, which are closely related to changes in chlorophyll and other photosynthetic pigments contents, leaf structure, and water absorption bands in NIR region, may play a distinct role in the estimation of biomass accumulation and photosynthetic efficiency of wheat under salinity stress; and (2) the effectiveness of some published SRIs for estimating the measured parameters in certain environmental conditions could be improved by replacing some of the wavelengths with others that are near to the original wavelengths.

5. Conclusions

Based on quantitative relationships between measured parameters and various SRIs (that incorporate a combination of wavelengths within the VIS-to-SWIR domains and are sensitive to multiple stresses) a number of modified SRIs such as NPQ_{(482, 350)}, WABI_{(1550, 482)}, WABI_{(1640, 482)}, WABI_{(1650, 531)}, NDMI_{(1660, 1742)}, and DMCI_{(1550, 2305)} were derived. These new SRIs are useful for estimating the growth and photosynthetic properties of wheat under saline field conditions, and they show an improved prediction ability compared to original published SRIs. The modified SRIs were derived using wavelengths that are nearest to those of published SRIs.