Estimation of Leaf Water Content of Different Leaves from Different Species Using Hyperspectral Reflectance Data

Yasir QM; Zhang W

Research Article

Ann Agric Crop Sci. 2022; 7(2): 1111.

Estimation of Leaf Water Content of Different Leaves from Different Species Using Hyperspectral Reflectance Data

Yasir QM¹ and Zhang W²*

¹Key Laboratory of Geographical Processes and Ecological Security in Changbai Mountains, Ministry of Education, School of Geographical Sciences, Northeast Normal University, Changchun, China

²Key Laboratory of Digital Earth Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing, China

*Corresponding author: Zhang W, Key Laboratory of Digital Earth Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing, 100094, China

Received: December 27, 2021; Accepted: February 08, 2022; Published: February 15, 2022

Abstract

Water content of individual leaves or vegetation canopies is a significant variable in plant physiological processes. The water content of the vegetation leaves plays a very significant role. The ability of hyperspectral advanced technology to accurately evaluate leaf and canopy water content has improved large-scale measures. Due to the presence of water absorption band in near and SWIR wavelength range, electromagnetic spectrum will allow us to correctly measure the leaf water content. Three different parameters were used to describe the water status: Equivalent Water Thickness (EWT), Gravimetric Water Content (GWC), and Plant Water Concentration (PWC), with leaf multi angular reflectance spectrum, to find sensitive spectral indices, to correctly assess water content of leaves in a wide range of plant species. Using spectral indices derived from multi angular reflectance spectra, we looked into the possibility for predicting leaf water content of six species in the study area. To analyze the status of leaf water, three different forms of hyperspectral indices were evaluated, including the Simple Ratio (SR), Normalized ratio wavelength (ND) and Double Difference ratio (DDn). To look over the possibility of predicting the leaf water status of the species in the study field, we proposed four new indices. The results showed that EWT is comparatively more sensitive to trace leaf water status than GWC and PWC. The best-established EWT indices were (R₉₀₅-R1₇₉₅)/(R1₉₀₅-R1₉₃₅), R1₃₅₀/R1₃₉₀, (R₈₄₀-R1₅₆₅)/(R₈₄₀+R1₅₆₅) and (R₉₂₅-R1₆₂₅)/ (R₉₂₅+R1₆₂₅) and the performance of the proposed hyper-spectral indices surpassed the performance of other indices in this study. The mentioned indices were then further analyzed on LOPEX and ANGERS databases for validation of our suggested indices and we come up with better results. This study indicates that spectral indices can be used and could be more reliable to predict leaf water content, but future studies will need to include more plant species and field data. The newly developed indices can be used to estimate EWT using simple laboratory measurements, making them helpful for agricultural environmental sciences and ecology related studies.

Keywords: EWT; PWC; GWC; LAI; Remote sensing; Hyperspectral indices

Introduction

The water content of the leaves and canopy is essential in many environmental processes because it plays a mjor part in activities such as plant food preparation, evapotranspiration, and total primary productivity [1,2]. Photosynthesis is influenced by the quantity of water in a plant’s leaf; this information may be used to predict water stress during different growth stages. An important sign of early stress in a plant is a rapid decrease in or absence of sufficient water content [3]. As a consequence, knowing how much water is in the leaves may help you to figure out how healthy a plant is [4], drought assessment[5], wildfire hazard prediction [6], and a slew of additional environmental, agricultural, and forestry uses [7], [8]. Physiological development in plants is closely related to water availability, which also needs to be improved [9]. Traditional methods in order to collect high-quality data on leaf water content, on the other hand, have significant drawbacks: They are inefficient and objectionable, and the findings produced for a small research area usually do not properly reflect the spatial variation in leaf water content across different zones.

It has been discovered that optical approaches for determining plant water status are effective [10-16]. A spectral reflectance factor based on theoretical radiative transfer models could be used to extract and identify leaf water [17], [18] and empirical models [15], [19-23]. At the leaf level, experts identified a link between reflected spectral signature and leaf water content, which they subsequently extended to the canopy level. [15,21,22]. However, in recent decades, remote sensing has been emerged as a critical instrument and method for monitoring water condition at different scales [24,25]. So two widespread methods of remote sensing were developed, including model inversion [26] and spectral indices [27]. Based on reflectance data, these two methods were used to investigate leaf water status. The second method, which is spectral indices, is considerably better compared to the model inversion. Despite the fact that it is based on a mix of narrow and wide spectral bands, it is simpler to connect with leaf water status. By establishing the generic spectral index, we may use remote sensing to evaluate vegetation quality and attract more people [25,27].

Since the reflectance factor in one, two, or more wavelength is used to calculate them, spectral indices have been used to analyze the amount of water in leaves and how it is distributed [25,28-31] Spectral indices can offer information on the change of leaf water content with varied degrees of precision at various scales using ground, aerial, and space borne sensors [30,32-34]. Because of continuous advancements in remote sensing technologies, the assessment of leaf water content and other biochemical characteristics is becoming more common. While using the hyperspectral reflectance factor, several spectral indices may be used to enhance estimate of leaf water content from various plant species [25,34-37]. They use the standard watercentered absorption bands at approximately 970, 1200, 1450, 1950, and 2500 (nm).

On the other hand, the increasing application of high spectral and spatial resolution data of leaves or plant multiple layers creates certain issues. There have been several hyperspectral indices suggested, however they have only been used to assess leaf water content in certain plant species or under specific measurement protocols [38,39]. The response of dying leaves responds differently from healthy leaves in several stages of water stresses, and as a result, leaf reflectance tends to increase throughout the dehydration process all across spectrum, 400-2500 (nm) [40-46]. Previous research has demonstrated that the leaf experiments may also provide a dataset with a wide variety of leaf water status and many other biological variables.

Because of the random orientation of individual leaves and variations in light directions, multi angular reflectance has been employed to calculate biochemical properties of leaves [47]. The distribution and amount of directional reflectance factor are regulated by the specular reflectance of a leaf’s surface, which is independent of the leaf’s biochemical characteristics, according to the majority of research [48-51], On the other hand, the majority of leaf reflectance measurements were made in a single direction, such as from the nadir, with a leaf clip, or with an integrating sphere [15,28,38,39]. In these studies, the impact of multi angular reflectance on the estimation of leaf water content is completely disregarded. Due to the anisotropic reflectance of plant cover, researchers observed that view angles affects the values of spectral indices [52,53].

Previous studies have reported that reflectance near 700 (nm) and its ratio to NIR reflectance spectra can track plant water stress [54,55]. Nonetheless, pigments and other variables that directly affect plant water features have a substantial impact on wavebands near 700 (nm), since they do not provide the predicted outcome. These problems are the most significant impediment to the empirical method’s broader application. It is critical to investigate the use of hyperspectral reflection at various viewing angles to improve the use of high spectral and spatial resolution data in the evaluation of leaf water content for ecological, agricultural and forestry applications.

Multi angular spectral reflectance characteristics of different leaves from six plant species were determined in the lab. The main aim of the analysis was to: (1) demonstrate the relationship between the published indices and the species taken for this study and the actual response (2) to evaluate s indices based on different reflected spectra of leaf water status in the dataset. This analysis was conducted using a leaf experiment in which six different species were collected.

Materials and Methods

Experiment involving leaf sampling and dehydration experiment

Leaves of six different plant species, including Prunus padus L., Swida alba Opiz, Epipremnum aureum, Acer saccharum Marsh, Schefflera microphylla Merr, and Pachira aquatic. For calibration, samples were taken from Northeast Normal University’s plant garden in Changchun, Jilin Province, China. As in prior experiments, we only picked healthy leaves with a uniform Colour and no visible evidence of damage [56,57]. Young, aged, and full mature leaves were chosen at random from top to the bottom of plant canopies. Senescing and ageing leaves represent those seen in plants under threat from polluted air, high temperatures, drought, and disease [58,59].

During the 2020 growing season, which runs from April to October, the reflectance parameters of leaves were tested in the laboratory. The leaves were gathered, enveloped in plastic bag with moist paper, and taken to the lab for examination. Then, on adaxial leaf surface, the angular spectral reflection was measured using the Northeast Normal University Laboratory Gonio Spectrometer System (NENULGS) [60]. NENULGS, which is equipped with an artificial light source, an ASD spectroradiometer (Analytical Spectral Devices Field Spec 4, Boulder, CO, USA), and goniometer, was previously described in detail in a prior article [60]. NENULGS has been used in various studies to accurately examine different properties of leaves [61-63]. Fresh weight was measured after taking the reflectance measurement and then air-dried to a stable weight for some time after that sample. Finally, the samples were placed under 80 ^oC for 36 hours in the oven to dry and dry weight was then weighed [64]. On its hemisphere, it has used the spectrum of reflection which ranges from 350 to 2500 (nm) in a variety of inclinations. In this experiment, we employed wavelengths stretch from 400 to 2500 (nm) [58,64].

Due to the structural constraints of NENULGS, the smallest portion angle that could be measured was 8, measurements in the backward scattering direction could not be performed when both viewing and incident angle are same.

While the measurements were collected, leaf sample was put on object stage, which were completely covered with dark black strips of tape. Black background has no influence on leaf reflection since it a wavelength independent reflectance factor of less than 0.05. The reflected radiance (dLSample-lab) from the leaf sample surface is normalized by the reflected radiance (dLReference-lab) from the reference surface (Spectralon) in the same viewing geometry to give the Bidirectional Reflectance Factor (BRF) [65] (Figure 1 and Table 1,2).

Table 1: The Equivalent Water Thickness (EWT), Gravimetric Water Content (GWC) and PWC of the samples used in this study were determined using the geometries and statistics listed below. These samples were applied to create spectral index correlations, whereas EWT, GWC, and PWC basically used them as calibration data.




  
    Name
    Samples
    EWT 

      (g/cm2)
    GWCF

      (g/g)
    GWCD

      �(g/g)
    PWC

      �(%)
  
  
    Prunus padus L.
    20
    Min
    0.0059
    0.4511
    0.8218
    82.18
  
  
    Max
    0.0097
    0.5968
    1.4804
    148.04
  
  
    Mean
    0.0078
    0.5285
    1.1314
    113.14
  
  
    Swida alba Opiz
    20
    Min
    0.0068
    0.5602
    1.2738
    127.38
  
  
    Max
    0.0122
    0.7068
    2.411
    241.1
  
  
    Mean
    0.009
    0.6459
    1.8551
    185.51
  
  
    Epipremnum aureum
    20
    Min
    0.0202
    0.8747
    6.9867
    698.67
  
  
    Max
    0.0293
    0.9462
    17.6017
    1760.17
  
  
    Mean
    0.0253
    0.9117
    10.8147
    1081.47
  
  
    Acer saccharum Marsh
    20
    Min
    0.0063
    0.5869
    1.4209
    142.09
  
  
    Max
    0.0116
    0.7713
    3.3732
    337.32
  
  
    Mean
    0.0094
    0.6928
    2.3796
    237.96
  
  
    Schefflera microphylla Merr.
    20
    Min
    0.0268
    0.8225
    4.6338
    463.38
  
  
    Max
    0.0433
    0.9196
    11.4516
    1145.16
  
  
    Mean
    0.0347
    0.8681
    6.885
    688.509
  
  
    Pachira aquatica
    20
    Min
    0.008
    0.7523
    3.0385
    303.858
  
  
    Max
    0.0141
    0.8614
    6.2192
    621.929
  
  
    Mean
    0.0107
    0.82
    4.7384
    473.842



Table 1: The Equivalent Water Thickness (EWT), Gravimetric Water Content
(GWC) and PWC of the samples used in this study were determined using the
geometries and statistics listed below. These samples were applied to create
spectral index correlations, whereas EWT, GWC, and PWC basically used them
as calibration data.

Table 2: Summary statistics for leaf water content (n = 120).




  
     
    Standard error of mean
    Standard deviation
    Coefficient of variation
  
  
    EWT (g/cm2)
    0.001
    0.01071
    0.67
  
  
    GWCF
    0.013
    0.1397
    0.19
  
  
    GWCD
    0.332
    3.638
    0.78
  
  
    PWC
    33.51
    363.8
    0.78
  
  
    Leaf area
    2.348
    25.72
    0.51 
  
  
    EWT: Equivalent Water Thickness; GWCF:    Gravimetric Water Content on Fresh Leaf Mass Basis; GWCD: Gravimetric Water    Content on Dry Leaf Mass Basis; PWC: Plant Water Concertation.



Table 2: Summary statistics for leaf water content (n = 120).

Figure 1: This figure describes the measuring system in the laboratory’s principal plane. Where φv = 180◦ refers to the primary plane’s forward scattering direction. SZA and VZA correspond to the source zenith angle (θs) and viewing zenith angle (θv) respectively.

    
    
    Figure 1: This figure describes the measuring system in the laboratory’s
principal plane. Where φv = 180◦ refers to the primary plane’s forward
scattering direction. SZA and VZA correspond to the source zenith angle (θs)
and viewing zenith angle (θv) respectively.

For definitions of EWT, GWCF, GWCD, and PWC see Table 3.

Table 3: Intercorrelation among variables in leaf data set (n = 120).




  
     
    leaf area
    EWT g/cm2
    GWCF
  
  
    EWT (g/cm2)
    0.346
    
    
  
  
    GWCF
    0.120
    0.706
    
  
  
    GWCD / PWC
    0.048
    0.703
    0.872



Table 3: Intercorrelation among variables in leaf data set (n = 120).

Leaf water status

Plant water status is measured using a variety of methods, including Equal Water Thickness (EWT), Plant Water Concentration (PWC), and Gravimetric Water Content (GWC). In this study, all three measures for evaluating the status of leaf water were used. The quantity of water content closely connected to the absorption of energy per unit leaf area is referred to Equivalent Water Thickness (EWT) [26]. PWC refers to the dry weight ratio, while GWC will decide on the fresh and dry leaf weight.

The equations for EWT, PWC, and GWC are given below.

and

Fresh weight, dry weight and leaf area are all represented as WF, WD, and LA, respectively.

Published leaf water status indices

Different hyperspectral indices based on multiple ratios, such as a simple ratio or a normalized ratio of different wavelengths within a particular spectrum, have been developed to evaluate plant water status. Nine other already published indices were selected for this analysis to verify their efficiency as to how they react with the data obtained in the leaf dehydration experiment for this study. The indices selected are given in Table 4 for this reason.

Table 4: Published water indices for leaf water status evaluation.




  
    Index
    Formula
    References
  
  
    Simple ratio (SR)
    R1300/R1450 
    [66]
  
  
    Normalized differential Water Index (NDWI)
    (R860-R1240)/(R860+R1240) 
    [67]
  
  
    Normalized Differential Water Index (NDWI)
    (R860-R1640)/(R860+R1640) 
    [68]
  
  
    (R850-R2218)/(R850-R1928) 
    [69]
  
  
     (R850-R1788)/(R850-R1928)
    [69]
  
  
    Moisture stress index (MSI)
     R1600/R820
    [13]
  
  
    Simple ratio water index (SRWI)
    R860/R1240 
    [70]
  
  
    Normalized difference water index centered at    1640 nm( 

      NDWI1640)
    (R858-R1640)/(R858+R1640)  
    [68]
  
  
    Normalized difference water index centered at    2130 nm(

      NDWI2130)
    (R858-R2130)/(R858+R2130) 
    [68]



Table 4: Published water indices for leaf water status evaluation.

Choosing the optimal indices

We implemented three types of measurements: SR, ND and DDn, which are currently using extensively [27]. Plant water concentration (PWC) and Gravimetric Water Content to assess the best indices for Equivalent Water Thickness (EWT) (GWC). The equations are defined as

R_λ1, R_λ2 and R_λ3 represent wavelengths at λ1, λ2 and λ3 respectively. While on the other hand SR, ND and DDn refer to simple ratio, normalized ratio and double difference respectively.

Statistical analysis

A dataset of 120 leaf samples from six various species was used to assess the appropriate hyperspectral indices for tracing leaf water status. Various statistical tests were run on data sets to assist in the definition of new indices and then the best selected indices were used to validate and confirm their robustness.

Further, the regression approach expands into different types, i.e., linear and nonlinear regression. The methods listed were applied to all possible wavelength combinations and wavelength interval was 5 nm using iteration method [64]. The parameters of published indices were set using lowest Root Mean Square Error (RMSE) and highest coefficient of determination (R2). The main goal was to find indices that had the lowest RMSE while having the greatest R² values.

Results

Reflectance factors of leaves: Spectral characteristics and angular distribution

The spectral reflectance factors with multiple indicators at nadir View Zenith Angle (VZA) are shown in Figure 2. Spectral BRF of leaves restricted as the various water indicators extended in the nadir direction in NIR and SWIR wavelengths, due to significant absorption of leaf water at wavelengths greater than 1300 nm. These spectral features are used to estimate water content using various spectral indices. Once the distribution of multi angular reflectance factor has been taken into consideration, it may be used to comprehend the reflection characteristics of leaves from diverse species.

Figure 2: Bidirectional reflectance factor (BRF) of leaves with varying EWT, GWC (fresh and dry), and PWC in nadir direction. These samples are illustrative of research.

    
    
    Figure 2: Bidirectional reflectance factor (BRF) of leaves with varying EWT,
GWC (fresh and dry), and PWC in nadir direction. These samples are
illustrative of research.

Performance of the published indices

Selected published indices’ performance in this study was analyzed individually to assess variation in EWT, PWC, and GWC. The correlation with EWT and the chosen indices were highly significant, and they were checked individually expect two indexes (R₈₆₀-R₁₂₄₀)/(R₈₆₀+R₁₂₄₀) [67] and R₈₆₀/R₁₂₄₀ [70]. The result came out with highest R2 and lowest RMSE. Furthermore, when all species were considered, the results were also good and having highest R2 and lowest RMSE expect two indexes (R₈₆₀-R₁₂₄₀)/(R₈₆₀+R₁₂₄₀) [67] and R₈₆₀/R₁₂₄₀ [70]. Here it is mentioned in Table 5.

Table 5: Using measured datasets, published water indices were evaluated for EWT.




  
    Indices
    Indicators
    R2
    RMSE (g/cm2 or %)
  
  
     R1300/R1450 
    EWT
    0.845
    0.004
  
  
    PWC
    0.575
    239.45
  
  
    GWCF
    0.472
    0.102
  
  
    GWCD
    0.57
    2.395
  
  
    (R860-R1240)/(R860+R1240) 
    EWT
    0.296
    0.009
  
  
    PWC
    0.032
    359.48
  
  
    GWCF
    0.033
    0.138
  
  
    GWCD
    0.032
    3.595
  
  
    (R860-R1240)/(R860+R1240) 
    EWT
    0.853
    0.004
  
  
    PWC
    0.33
    299.07
  
  
    GWCF
    0.271
    0.12
  
  
    GWCD
    0.33
    2.991
  
  
    (R850-R2218)/(R850-R1928) 
    EWT
    0.869
    0.004
  
  
    PWC
    0.401
    282.85
  
  
    GWCF
    0.3
    0.117
  
  
    GWCD
    0.401
    2.829
  
  
     (R850-R1788)/(R850-R1928)
    EWT
    0.916
    0.003
  
  
    PWC
    0.38
    287.67
  
  
    GWCF
    0.34
    0.114
  
  
    GWCD
    0.38
    2.877
  
  
     R1600/R820
    EWT
    0.847
    0.004
  
  
    PWC
    0.368
    290.37
  
  
    GWCF
    0.288
    0.118
  
  
    GWCD
    0.368
    2.904
  
  
    R860/R1240 
    EWT
    0.295
    0.009
  
  
    PWC
    0.031
    359.58
  
  
    GWCF
    0.034
    0.138
  
  
    GWCD
    0.031
    3.596
  
  
    (R858-R1640)/(R858+R1640)  
    EWT
    0.853
    0.004
  
  
    PWC
    0.331
    298.81
  
  
    GWCF
    0.272
    0.12
  
  
    GWCD
    0.331
    2.988
  
  
    (R858-R2130)/(R858+R2130) 
    EWT
    0.781
    0.005
  
  
    PWC
    0.437
    274.15
  
  
    GWCF
    0.289
    0.118
  
  
    GWCD
    0.437
    2.742



Table 5: Using measured datasets, published water indices were evaluated for
EWT.

Newly identified leaf water status indices

Published water indices for EWT were examined using measured datasets. Finally, based on reflectance spectra, the regions with the highest R2 and minimum RMSE value were chosen, and their results are described in Table 6. In general, particular wavelengths were strongly correlated with EWT for original reflectance and had no comparatively substantial connection with PWC, GWCD, and GWCF as shown in Table 6. Specifically the best indices were (R₉₀₅-R₁₇₉₅)/(R₁₉₀₅-R₁₉₃₅) having (R2= 0.939, RMSE= 0.003), R₁₃₅₀/R₁₃₉₀ (R²= 0.914, RMSE= 0.003),(R₈₄₀-R₁₅₆₅)/(R₈₄₀+R₁₅₆₅) (R²= 0.914, RMSE= 0.003) and another index were(R₉₂₅-R₁₆₂₅)/(R₉₂₅+R₁₆₂₅) having (R²= 0.895, RMSE=0.004) with all measurements. These newly identified indices have a significant correlation with EWT, while PWC, GWCD, and GWCF have performed poorly. Overall, these four indices’ performance with original reflectance was the best by looking to R2 and RMSE values. It is mentioned in Table 6.

Table 6: Evaluation of four types of indices with reflectance for leaf EWT, PWC, GWCD and GWCF.




  
    Indicators
    Index
    R2
    RMSE (g/cm2 or %)
  
  
    EWT
    (R905-R1795)/(R1905-R1935)
    0.939
    0.003
  
  
    (R1350/R1390)
    0.914
    0.003
  
  
    (R840-R1565)/(R840+R1565)
    0.914
    0.003
  
  
    (R925-R1625)/(R925+R1625)
    0.895
    0.004
  
  
    PWC
    (R925-R1625)/(R925+R1625)
    0.389
    285.61
  
  
    (R1350/R1390)
    0.617
    226.02
  
  
    (R840-R1565)/(R840+R1565)
    0.433
    275.05
  
  
    R925-R1625)/(R925+R1625)
    0.363
    291.64
  
  
    GWCD
    (R905-R1795)/(R1905-R1935)
    0.389
    2.856
  
  
    (R1350/R1390)
    0.617
    2.26
  
  
    (R840-R1565)/(R840+R1565)
    0.433
    2.751
  
  
    (R925-R1625)/(R925+R1625)
    0.363
    2.916
  
  
    GWCF
    (R905-R1795)/(R1905-R1935)
    0.351
    0.113
  
  
    (R1350/R1390)
    0.546
    0.095
  
  
    (R840-R1565)/(R840+R1565)
    0.355
    0.113
  
  
    (R925-R1625)/(R925+R1625)
    0.290
    0.118



Table 6: Evaluation of four types of indices with reflectance for leaf EWT, PWC,
GWCD and GWCF.

In Figure 3, linear and nonlinear regression were performed to develop linear models and determine the coefficient of determination between different wave length ranges and water indices (EWT, PWC, DWCF and GWCD to compare the efficiency of the model. Nonlinear regression models performed well as compare to linear regression model almost all the studied wave length ranges.

Figure 3: The coefficient of determination of chosen wave length and water indices were calculated using linear and nonlinear regression models (EWT, PWC, GWCF, and GWCD).

    
    
    Figure 3: The coefficient of determination of chosen wave length and water indices were calculated using linear and nonlinear regression models (EWT, PWC,
GWCF, and GWCD).

Linear regression equation

y_i = Β₀ +Β_ix_i (8)

Nonlinear regression equation

y_i = Β₀ +Β₁x_i +Β₂x_i2 (9)

Co-efficient of determination

where, y_i and x_i are dependent and independent variables respectively. Β₀ - Intercept, Β₁, Β₂ - slope and y ^ represent the estimated values of dependent variables.

Validation data sets from different sources

The first set of data comes from the European Commission’s Joint Research Center’s Leaf Optical Properties Experiment (LOPEX), which is composed of 330 leaf samples from 45 different plant species [71]. The ANGERS dataset was prepared in 2003 in France, possesses 275 leaf samples from 43 species of plants [72] (Figure 4).

Figure 4: Spectrum of leaf reflectance from the (a) ANGERS and (b) LOPEX datasets. The dashed red line represent the median spectrum.

    
    
    Figure 4: Spectrum of leaf reflectance from the (a) ANGERS and (b) LOPEX datasets. The dashed red line represent the median spectrum.

In the lab, the directional hemispherical reflectance factors of leaves from the LOPEX and ANGERS datasets were examined using a spectrophotometer with an integrating sphere. Angular effect was not included in the integrating sphere calculations (Table 7).

Table 7: Validation datasets for calculating Equivalent Water Thickness (EWT), Gravimetric Water (GWC) and PWC in this study.




  
    Spectrophotometer 
    LOPEX

      Perkin Elmer Lambda 19 
    ANGERS

      ASD FieldSpec 
  
  
    Measurement 
    Laboratory 
    Laboratory 
  
  
    Spectral range (nm)
    400-2500
    350-2500
  
  
    Number of Samples
    45
    43
  
  
    Mean (g/cm2)
    0.0111
    0.0116
  
  
    Min (g/cm2)
    0.0003
    0.0044
  
  
    Max (g/cm2)
    0.0525
    0.034
  
  
    Reference
    [71]
    [72]



Table 7: Validation datasets for calculating Equivalent Water Thickness (EWT),
Gravimetric Water (GWC) and PWC in this study.

These LOPEX and ANGERS database were further taken into account for the validation of the indices proposed in this study to check out the reliability and generalization. For which we first calculate EWT, GWC and PWC and then find out R2 plus Root Mean Square Error (RMSE) with spectral indices proposed in this study.

In the LOPEX database it exhibits good results in terms of EWT except one index ‘‘d’’ as shown in Figure 5. while three indices i.e. a, b and c shows strong correlation with lowest RMSE.

Estimation of Leaf Water Content of Different Leaves from Different Species Using Hyperspectral Reflectance Data

Abstract

Introduction

Materials and Methods

Experiment involving leaf sampling and dehydration experiment

Leaf water status

Published leaf water status indices

Choosing the optimal indices

Statistical analysis

Results

Reflectance factors of leaves: Spectral characteristics and angular distribution

Performance of the published indices

Newly identified leaf water status indices

Validation data sets from different sources

Discussion

The best indicators for defining the status of leaf water

The stability of the indices suggested

Spectral indices’ development

Leaf properties

Conclusion

Acknowledgment

References