Leaf and canopy water content estimation in cotton using hyperspectral indices and radiative transfer models

International Journal of Applied Earth Observation and Geoinformation 33 (2014) 67–75

Contents lists available at ScienceDirect

International Journal of Applied Earth Observation and Geoinformation journal homepage: www.elsevier.com/locate/jag

Leaf and canopy water content estimation in cotton using hyperspectral indices and radiative transfer models Qiuxiang Yi a,∗ , Fumin Wang b , Anming Bao a , Guli Jiapaer a a b

Xinjiang Institute of Ecology and Geography Chinese Academy of Sciences, 818 Beijing South Road, Urumqi, Xinjiang 830011, PR China Institute of Hydrology and Water Resources, Zhejiang University, Hangzhou, Zhejiang 310058, PR China

a r t i c l e

i n f o

Article history: Received 6 December 2013 Accepted 25 April 2014 Keywords: EWT EWTcanopy PROSPECT-5 model PROSPECT-5 + SAILH model Hyperspectral vegetation indices Cotton

a b s t r a c t In present study some vegetation indices for estimating leaf EWT and EWTcanopy were investigated using simulations and field measurements. Leaf and canopy spectral reflectance as well as leaf EWT and EWTcanopy were measured in cotton during the growing seasons of 2010 and 2011. The PROSPECT-5 model was coupled with the SAILH model to explore the performance of water-related vegetation indices for leaf EWT and EWTcanopy estimation. The vegetation indices evaluated were published formulations and new simple ratio vegetation indices formulated with wavebands at 1060 nm and 1640 nm. The sensitivities of these indices to leaf internal structural N and LAI effects were assessed. Simulation results indicated that all of the water-related vegetation indices were insensitive to leaf internal structural N, with the highest coefficient of determination R2 < 0.15 and the proposed index SR1640 (R1060 /R1640 ) and published index SR2 (R1070 /R1340 ) showed the lowest relationships (R2 < 0.35) with LAI of all the vegetation indices. Furthermore, coefficients of determination between simulated leaf EWT as well as EWTcanopy and vegetation indices tested revealed that the new simple-ratio vegetation indices proposed in this study (SR1060 : R1640 /R1060 and SR1640 ) were found to be significantly related with leaf EWT (R2 > 0.9; P < 0.001) and EWTcanopy (R2 > 0.8; P < 0.001). Results obtained with field measurements were in agreement with simulation results, with the coefficient of determination R2 = 0.5 (P < 0.001) for leaf EWT and R2 = 0.57 (P < 0.001) for EWTcanopy by the new simple ratio indices. This study provides a new candidate for leaf EWT and EWTcanopy estimation using hyperspectral vegetation indices. © 2014 Elsevier B.V. All rights reserved.

Introduction The knowledge of vegetation water conditions can contribute to ˜ detect vegetation physiological status (Carter, 1993; Penuelas et al., 1994; Stimson et al., 2005), to provide useful information in agri˜ culture for irrigation decisions and drought assessment (Penuelas et al., 1993, 1994) and it is important in forestry in determining fire susceptibility (Carlson and Burgan, 2003; Chuvieco et al., 2004; Ustin et al., 1998). Several physiological indicators are used to assess plant water conditions, with stomata conductance (gs ), leaf water potential, fuel moisture content (water content express as percent of dry mass or fresh mass (FMC)), vegetation water content (VWC) and equivalent water thickness at leaf and canopy levels (EWT and EWTcanopy), and so on. Remote sensing techniques provide a non-destructive, rapid, and reliable method for assessing water status. Several water condition indicators have been related

∗ Corresponding author. Tel.: +86 9917823131. E-mail address: [email protected] (Q. Yi). http://dx.doi.org/10.1016/j.jag.2014.04.019 0303-2434/© 2014 Elsevier B.V. All rights reserved.

˜ to spectral reflectance measurements (Cifre et al., 2005; Penuelas et al., 1993; Serrano et al., 2000; Stimson et al., 2005). At leaf level, investigations revealed that the estimation of leaf water content in terms of equivalent water thickness (EWT) expressed in quantity of water per unit area (g/cm2 ) were performed better than water content in terms of moisture content expressed in quantity of water per quantity of fresh or dry matter (%) (Ceccato et al., 2001; Colombo et al., 2008; Datt, 1999; Davidson et al., 2006; Maki et al., 2004). At canopy level, José et al. (2007) also suggested EWTcanopy that expressed in water per unit surface area (g/m2 ) may be more appropriate for predicting vine water status at canopy level. EWT and EWTcanopy have been successfully estimated in agricultural crops, forests, Mediterranean shrublands and savannah woodlands (Ceccato et al., 2002b; Gao and Goetz, 1995; Jacquemoud et al., 1995; Serrano et al., 2000; Ustin et al., 1998; Zarco-Tejada et al., 2003). So in present work, EWT and EWTcanopy were adopted for cotton water content estimation. EWT is defined as a hypothetical thickness of a single layer of water average over the whole leaf area (Danson et al., 1992) and EWTcanopy is defined as EWT multiplied by LAI (the leaf area per unit ground surface area, m2 /m2 ).

68

Q. Yi et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 67–75

EWTcanopy represents a quantity of water per unit surface area at canopy level (Ceccato et al., 2002a). The possibility of estimating water conditions by means of remotely sensed data derives from the fact that water absorbs radiant energy throughout the near-infrared (750–1300 nm) and short-infrared (1300–2500 nm) spectral regions. Leaf and canopy reflectance decreases with increasing tissue water content for wavelengths sensitive to water absorption (Aldakheel and Danson, 1997; Carter, 1991; Ceccato et al., 2001; Hunt and Rock, 1989; Knapp and Carter, 1998; Thomas et al., 1971). Spectral indices have been and are still widely used to retrieve information on vegetation biophysical properties. For wavelengths sensitive to water content (970, 1200, 1450, 1940 and 2500 nm) have been combined in numerous ways to generate vegetation indices related to water ˜ status (Gao, 1996; Hardisky et al., 1983; Penuelas et al., 1993, 1997; Zarco-Tejada et al., 2001). A detailed summary of vegetation indices related to water content can be found in José et al. (2007). A preliminary comparison of several indices from the lecture showed good results in terms of leaf EWT and EWTcanopy retrieval when applied to our experimental dataset. The vegetation indices that were widely used and relatively better related to leaf EWT and EWTcanopy were selected and used in this paper. Adopted indices are summarized in Table 1. Ceccato et al. (2001) showed that both the shortwave infrared (SWIR) and the near infrared (NIR) wavelength ranges are necessary for retrieving EWT at leaf level. The same authors also showed that in the NIR region, variations in reflectance values are exclusively influenced by leaf internal structure N and dry matter content (Cm). Furthermore, in the SWIR region, N and Cm factors also significantly affect reflectance values. Several researches have evaluated and quantified the effects of leaf water content on reflectance data (Aldakheel and Danson, 1997; Bowyer and Danson, 2004; Ceccato et al., 2002a; Dawson et al., 1998; Ustin et al., 1998). However, the relationships between leaf internal structure N and water-related vegetation indices have not been well illustrated. Besides, at canopy level, the reflectance is significantly affected by LAI (Zarco-Tejada et al., 2003). A large variability in LAI may cancel out water-related features in spectral reflectance (Cohen, 1991; Riggs and Running, 1991) and therefore complicates the estimation of EWT at the canopy level (Yebra et al., 2013). Jacquemoud et al. (2009) revealed that the SWIR is highly sensitive to LAI between 1000 nm and 1400 nm and suggested that caution should be taken when using these indices for water retrieval. It is necessary to evaluate the effect of LAI on water-related vegetation indices. Féret et al. (2011) tested the performance of MSI (R1600 /R820 ) for EWT estimation and suggested that other optimal wavelengths could be used to build a better spectral index. The aim of the present study was to evaluate the performance of a set of hyperspectral vegetation indices in EWT and EWTcanopy estimation in cotton using both model simulations and field measurements. The specific objectives were (i) to evaluate the performance of a set of water content related vegetation indices in leaf EWT and EWTcanopy estimation; (ii) to propose a new vegetation indices for EWT and EWTcanopy estimation through sensitive analysis and assess its performance with modeling methods and field measurements.

8–10 ha) and one small water-controlled plot (about 0.1 ha). Every eight big filed plot was consisted of eight small sample sites (about 30 m × 30 m), and other 12 sample sites were set in the watercontrolled plot, for a total of 76 sample sites. The continental arid climate of Xinjiang is characterized by aridity, rich sunlight and rare rainfall, with sharply defined seasons, high annual and diurnal fluctuations in air temperature, and low precipitation. Field data collections were conducted in June–September 2010–2011 for six times from seedling stage until boll stage (the actual dates were 12 June, 14 July, and 8 August, 2010; 24 June, 28 July, and 17 August, 2011). This procedure ensured that the normally occurring variation due to growth stage and measurement factors was included in the indices. Reflectance measurements Canopy reflectance was obtained using an Analytical Spectral Devices, FieldSpec Full Range (ASD FieldSpec FR, Analytical Spectral Devices, Inc., Boulder, CO, USA) that acquires continuous spectra from 350 to 2500 nm. All canopy spectral measurements were taken on clear days with no visible cloud cover between 10:00 am and 14:00 pm (Beijing local time). In each sample site, representative plants were selected for canopy spectral measurement. Leaf reflectance was measured over the spectral region between 350 nm and 2500 nm by coupling a leaf clip (ASD, Inc., Boulder, CO, USA) with the ASD FieldSpec FR. The reflectance was measured in the “reflectance” mode against a black background. Leaves healthy were used for leaf reflectance measurements, for a total of 481 leaf samples. The reflectance of a white Spectralon panel (BaSO4 ) was measured before every reflectance was taken, then the reflectance was calculated as the ratio between energy reflected by the leaf or the canopy and energy incident on the leaf or the canopy. Every reflectance was an average of ten repeated scans that were automatically acquired by the FieldSpec. Leaf sampling and water content measurements Three average-looking plants per plot were pulled out with their roots, placed and sealed in a plastic bag, and then placed in a cool dark container to avoid water loss as much as possible. Upon return to the laboratory, fresh weight (FW) of leaves was recorded immediately using an analytical balance, after which optical properties were measured and leaf photos were taken. Fresh leaves were then put into oven to dry with 105 ◦ C for half an hour and 70 ◦ C till the constant weight were acquired (Saura-Mas and Lioret, 2007). In order to make all measurement simultaneous, four groups worked like a line operation for leaf sampling, weighting, leaf spectra measurement and leaf photo taken. Leaf EWT was calculated for each leaf sample using Eq. (1): EWT =

FW − DW A

(g/cm2 )

(1)

where FW is the leaf fresh weight and DW is the dry leaf weight of all the leaves in the same sample plant, A is the area of fresh leaf (cm2 ), which was obtained by scanning. By multiplying leaf EWT with LAI the canopy water content (EWTcanopy) is obtained: LAI × EWT 10

(kg/m2 )

Materials and methods

EWTcanopy =

Field data collection

LAI was obtained using Li-Cor Plant Canopy Analyzer (model LAI-2000) on the field before collecting. Since the instrument requires diffuse conditions for accurate readings, the measurements were typically collected in the late afternoon hours and an umbrella was used to block the direct solar beam. Five below canopy readings were taken between two above-canopy readings, for a total of 353 valid LAI measurements.

Field experiments The field experiment was conducted in June–September 2010 and 2011 at agricultural belts in Shihezi, Xinjiang, Northwest of China (85◦ 59 E, 44◦ 19 N), where cotton is the dominate crop. The study sites were consisted of eight big filed plots (approximately

(2)

Q. Yi et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 67–75

69

Table 1 Spectral indices selected for assessment of leaf EWT and EWTcanopy. Water indices

Full name

Formula

Reference

NDII NDWI WI SR1 SR2 MSI MSI1 MSI2 SIWSI

Normalized different infrared index Normalized different water index Water index Simple ratio water index Simple ratio water index Moisture stress index Moisture stress index Moisture stress index Shortwave infrared water stress

(R850 − R1650 )/(R850 + R1650 ) (R860 − R1240 )/(R860 + R1240 ) R900 /R970 R858 /R1240 R1070 /R1340 R1600 /R820 R870 /R1350 R1650 /R835 (R858 − R1640 )/(R858 + R1640 )

Hardisky et al. (1983) and Kimes et al. (1981) Gao (1996) ˜ Penuelas et al. (1993, 1997) Zarco-Tejada et al. (2001, 2003) José et al. (2007) Rock et al. (1986) and Hunt (1991) Rock et al. (1986) and Hunt (1991) Rock et al. (1986) and Hunt (1991) Fensholt and Sandholt (2003)

Simulations with PROSPECT-5 and PROSAILH models A model simulation analysis was conducted to assess the sensitivity of water-related indices to Cm and Cw. PROSPECT5 was selected for the leaf-level simulations (Féret et al., 2008; Jacquemoud and Baret, 1990; Jacquemoud et al., 2009). PROSPECT-5 has the capability to independently model the influence of leaf dry matter content (Cm) and leaf water content (EWT) on spectral reflectance between 400 and 2500 nm at 1 nm increments. The simulate leaf directional–hemispherical reflectance and transmittance from the 400 to the 2500 nm spectral region with five input variables: chlorophyll content (Ca + b), carotenoid content (Cx + c), leaf dry matter content (Cm), equivalent water thickness (Cw) and leaf structure parameter (N) (Féret et al., 2008). Nominal values and input parameter ranges used for the leaf modeling are summarized in Table 2. The values of Ca + b and Cx + c were fixed because their absorbance mainly occurs in the visible region and insensitive to the shortwave infrared (SWIR) and the near infrared (NIR) wavelength ranges. Fixed values for Ca + b and Cx + c were set by averaging the field measurements and the variation range for EWT and Cm were set based on the maximum and the minimum field measurement values. At canopy level, the Scattering by Arbitrarily Inclined Leaves (SAIL) model (Verhoef, 1984) incorporating the hotspot effect SAILH model (Kuusk, 1991) was used to simulate canopy spectral reflectance. SAIL is a four-stream radiative transfer model developed by Verhoef (1984). It was later modified by Kuusk (1991) to take the hot spot feature into account. SAILH model was chosen to simulate canopy reflectance since it requires only few input variables, while having a predictive power similar to more elaborated reflectance models, and most importantly, its assumptions match crops well (Jacquemoud et al., 1995, 2000; Bacour et al., 2002). Leaf reflectance and transmittance were simulated with PROSPECT-5. Nominal values and the parameter range used for PROSPECT-5 + SAILH model are summarized in Table 3. Results and discussion

0.01–0.05 g/cm2 for Cw and mean Cm values of 0.006, 0.009 and 0.012 g/cm2 , and a range of 0.003–0.015 g/cm2 for Cm and mean Cw values of 0.01, 0.03 and 0.05 g/cm2 , respectively. The simulations showed that leaf water content Cw has strong absorbance peaks throughout the NIR and SWIR regions of the spectrum (from 800 to 2500 nm), except spectral reflectance around 1060 nm (1040–1070 nm), where leaf dry matter content Cm shows sensitivity. It can be inferred that waveband at 1060 nm was sensitive to Cm and insensitive to EWT. Just as mentioned before, leaf internal structure N also has strong influences on spectrum in NIR and SWIR regions. It is necessary to see if N was sensitive to spectral variations at 1060 nm. Leaf reflectance simulated with PROPSECT-5 considering low N (N = 1.2) and high N values (N = 2.6) for different Cm and Cw are shown in Fig. 2(a) and (b), respectively. It can be seen that the leaf internal structure N also significantly affects the spectral reflectance at 1060 nm. Besides, it can be noticed that the influences of the leaf internal structure N at 1640 nm cannot cancel out the strong influences of the leaf water content at this region (see Fig. 2b). These results indicated that maybe new indices can be formulated with waveband at 1060 nm that is sensitive to the leaf internal structure and leaf dry matter content Cm and insensitive to leaf water content Cw and waveband at 1640 nm that is strongly influenced by the leaf water content. Further analysis was performed on the effects of LAI on canopy reflectance. The canopy reflectance simulated with PROSPECT coupled with SAILH considering low (LAI = 1) and high (LAI = 5) LAI values is shown in Fig. 3. Leaf reflectance and transmittance were simulated with PROSPECT-5 assuming different values of Cm (0.003–0.018 g/cm2 ) (Fig. 3a) and Cw (0.01–0.05 g/cm2 ) (Fig. 3b). As can be seen, the large variability in LAI cannot cancel out the dry matter-related features in spectral reflectance at 1060 nm (Fig. 3a) and water-related features in spectral reflectance at 1640 nm (see Fig. 3b). Furthermore, the spectral reflectance at 1060 nm is insensitive to LAI. Combined with the results obtained from Figs. 1 and 2, it can be inferred that the reflectance at 1060 nm were sensitive to the leaf dry matter Cm and the leaf internal structure N and insensitive to the leaf water content EWT and LAI. So the vegetation indices Table 3 Input parameters for PROSAILH simulations.

Formulation of spectral indices The PROSPECT-5 model was used to simulated leaf reflectance with varying Cw (0.01–0.05 g/cm2 ), Cm (0.001–0.018 g/cm2 ) and leaf structure N (1.2–2.6). Fig. 1 showed the leaf spectral variation derived from the simulations performed had a range of

Model

Parameters

Values

PROSPECT

Leaf structure parameter (N) Chlorophyll content (Cab, ␮g/cm2 ) Carotenoid content (Cx + c, ␮g/cm2 ) Water content (Cw, g/cm2 or cm) Dry matter content (Cm, g/cm2 )

1.2–2.6 60 15 0.01–0.06 0.001–0.018

SAIL

Leaf area index (LAI) Leaf angle distribution (LAD) Fraction of direct solar irradiance Solar declination View zenith angle Time of day (hour) Hot spot Solar zenith angle Observation zenith angle

0.2–8 Spherical 0.8 0◦ Nadir 12:00 0.5 35◦ 0◦

Table 2 Nominal values range of parameters used for PROSPECT-5 simulations. PROSPECT-5 input variables

Value

Unit

Ca + b Cx + c Cw (varied parameter) Cm (varied parameter) N (varied parameter)

60 15 0.01–0.06 0.001–0.018 1.2–2.6

␮g/cm2 ␮g/cm2 g/cm2 g/cm2

Q. Yi et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 67–75

0.6 0.5 0.4

Leaf reflectance

0.6

Cw=0.01 Cw=0.02 Cw=0.03 Cw=0.04 Cw=0.05

R1640

R1060

Cm=0.006 g/cm 2 Chl=60 µg/cm 2

0.3

Cx+c=15 µg/cm 2 N=2

0.2

Cm=0.003 Cm=0.006 Cm=0.009 Cm=0.012 Cm=0.015

R1640

R1060

0.5 Leaf reflectance

70

0.1

Cw=0.01 g/cm2 Chl=60 µg/cm 2

0.4

Cx+c=15 µg/cm 2

0.3

N=2

0.2 0.1

0.0 500

1000

1500

2000

0.0

2500

500

Wavelength (nm)

1000

1500

2000

2500

Wavelength (nm) R1060

0.6

R1640

0.5

Cm=0.009 g/cm 2 Chl=60 µg/cm 2

0.4

N=2

R1640

R1060

0.5

Cx+c=15 µg/cm 2

Leaf reflectance

Leaf reflectance

0.6

0.3 0.2

Cw=0.03 g/cm2 Chl=60 µg/cm 2 Cx+c=15 µg/cm 2 N=2

0.4 0.3 0.2 0.1

0.1

0.0

0.0 500

1000

1500

2000

500

2500

1000

1500 2000 Wavelength (nm)

2500

Wavelength (nm) 0.6

R1060

0.6 Cm=0.012 g/cm2 Chl=60 µg/cm 2

0.5

Cx+c=15 µg/cm 2

0.4

R1060

R1640 Cw=0.05 g/cm2 Chl=60 µg/cm 2

0.5 Leaf reflectance

Leaf reflectance

R1640

N=2

0.3 0.2 0.1

Cx+c=15 µg/cm 2

0.4

N=2

0.3 0.2 0.1

0.0 500

1000

1500 2000 Wavelength (nm)

2500

0.0 500

1000

1500

2000

2500

Wavelength (nm)

Fig. 1. Leaf reflectance simulations with PROSPECT-5 to assess the effects of Cw and Cm on spectral signature.

proposed were formulated as a simple ratio and normalized differences ratio between bands located at 1640 nm and 1060 nm, with Eqs. (3)–(5): SR1060

Performance of vegetation indices using simulations

R1640 = R1060

(3)

R1060 R1640

(4)

SR1640 =

ND1640 =

R1640 − R1060 R1640 + R1060

where R1640 and R1060 are the reflectance at 1060 nm and 1640 nm, respectively.

(5)

The relationships between simulated hyperspectral vegetation indices and leaf EWT and between those vegetation indices and leaf internal structural N are shown in Fig. 4. Coefficients of determination (R2 ) of all the indices studied are provided. Almost all of the water-related vegetation indices tested at leaf level showed good agreement with leaf EWT, with coefficient of determination R2 > 0.85 (P < 0.001). The best relationships were obtained when WI,

Q. Yi et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 67–75 R1640

(a) 0.6

Cm=0.003 Cm=0.007 Cm=0.011 Cm=0.015 Cm=0.019

N=2.6

Leaf reflectance

0.5 N=1.2

0.4 0.3 0.2

0.3 0.2

0.0

2000

N=1.2

0.4

0.0 1500

N=2.6

0.5

0.1

1000

Cw=0.01 Cw=0.02 Cw=0.03 Cw=0.04 Cw=0.05 Cw=0.06

(b) 0.6

0.1

500

R1640

R1060

0.7

Leaf reflectance

R1060

0.7

71

2500

500

1000

Wavelength (nm)

1500

2000

2500

Wavelength (nm)

Fig. 2. Leaf reflectance simulated with PROPSECT-5 considering low N (N = 1.2) and high N values (N = 2.6) for Cm with fixed Cw = 0.03 cm (a) and Cw with fixed Cm = 0.009 g/cm2 (b).

R1060

R1640

0.6

(a)

Cm=0.003 Cm=0.006 Cm=0.009 Cm=0.012 Cm=0.015 Cm=0.018

Leaf reflectance

0.6 0.5 0.4 0.3

LAI=1

0.2 0.1

R1060

R1640 Cw=0.01 Cw=0.02 Cw=0.03 Cw=0.04 Cw=0.05

(b)

0.5

Leaf reflectance

0.7

0.4

LAI=1

0.3 0.2

LAI=5

0.1

LAI=5

0.0

0.0 500

1000

1500

2000

2500

Wavelength (nm)

500

1000

1500

2000

2500

Wavelength (nm)

Fig. 3. Canopy reflectance simulated with PROPSECT-5 + SAILH considering low LAI (LAI = 1) and high LAI values (LAI = 5) for Cm (a) with fixed Cw = 0.03 cm, N = 2, and Cw (b) with fixed Cm = 0.009 g/cm2 , N = 2.

SR1, MSI1 and SR1060 were used, yielding a coefficient of determination around 0.9. Among the three proposed vegetation indices, i.e. SR1060 , SR1640 and ND1640 , SR1060 obtained the closest relationship with leaf EWT, and ND1640 showed the lowest coefficient of determination. Besides, the coefficient of determination between vegetation indices and leaf internal structural N showed that all of these indices were insensitive to leaf internal structural N and with

the highest coefficient of determination R2 = 0.2 (P > 0.01) and the lowest R2 = 0.08 (P > 0.05). The coefficients of determination were very small and uniform. Furthermore, simulations conducted with PROSPECT-5 + SAILH at canopy level were used to assess the relationships between EWTcanopy and water-related vegetation indices and between LAI and these indices (Fig. 5). Results showed that vegetation indices

Fig. 4. Relationships between leaf EWT and vegetation indices and between leaf internal structural N and vegetation indices. Data were simulated at the leaf level with PROSPECT-5 assuming random variation of Cm (0.001–0.018 g/cm2 ), Cw (0.01–0.06 g/cm2 ) and N (1.2–2.6).

72

Q. Yi et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 67–75

Fig. 5. Relationships between EWTcanopy and vegetation indices and between LAI and vegetation indices. Data were simulated at the canopy level with PROSPECT-5 + SAILH assuming random variation of Cm (0.001–0.018 g/cm2 ), Cw (0.01–0.06 g/cm2 ) and LAI (0.2–8).

behaved differently at leaf and canopy levels. The closest relationships between EWTcanopy and vegetation indices at canopy level were obtained by the proposed indices SR1640 , yielding R2 of 0.83. However, SR1060 significantly correlated with leaf EWT at leaf level showed high effects due to LAI and did not obtain high coefficient of determination at canopy level. This also was the case of MSI (R2 = 0.55) and MSI2 (R2 = 0.60). The relationships between simulated vegetation indices and LAI indicated that the proposed index SR1640 showed the highest correlationship with EWTcanopy and the lowest coefficient of determination with LAI. These results indicated that the best relationship between EWTcanopy and SR1640 was not caused by LAI. It can be inferred that SR1640 was a good candidate for EWTcanopy estimation. Besides, it can be learned from Figs. 4 and 5 that the published vegetation index SR2 showed the significant correlation with both leaf EWT and EWTcanopy, and meantime insensitive to LAI. A further comparison of the relationships found in SR2 and proposed indices SR1640 and SR1060 at both leaf and canopy levels is provided in Fig. 6. It can be learned from Fig. 6 that the best relationships between leaf EWT and vegetation indices can be fitted with linear regression, whereas, at the canopy level, the better relationships between EWTcanopy and vegetation indices were found with curvilinear fitting. Performance of vegetation indices using field measurements A summary of statistics for leaf and canopy water content and LAI are given in Table 4. For all measured dataset, leaf EWT ranged from 0.018 to 0.042 g/cm2 , and EWTcanopy ranged from 0.0008 to 0.017 kg/m2 (Table 4). The range of leaf EWT variability (Cv = 0.127) was found to be smaller than that of EWTcanopy (Cv = 0.657). As shown, the larger EWTcanopy variability due to the high variability of LAI. Besides, Cm and leaf EWT were found to be positively correlated, with coefficient of determination R2 = 0.53 (P < 0.001), which was consistent with the result found by Yebra and Chuvieco

Table 4 Basic statistics of Cm, EWT, LAI and EWTcanopy.

Cm (g/cm2 ) EWT (g/cm2 ) LAI EWTcanopy (kg/m2 ) *

n

Max

Min

Mean

Std.

Cv*

481 481 353 353

0.014 0.042 5.586 0.017

0.004 0.018 0.300 0.0008

0.008 0.029 1.681 0.0047

0.002 0.004 1.137 0.003

0.187 0.127 0.676 0.657

Cv is the coefficient of variation.

(2009), and a significant and positive relationship was also found between EWTcanopy and LAI, with R2 = 0.95 (P < 0.0001). The relationships between measured EWT and EWTcanopy and vegetation indices were analyzed (Fig. 7). The coefficients of determination calculated by curvilinear regression at both leaf and canopy levels are shown in Fig. 7. Results showed that vegetation indices behaved differently at leaf and canopy level. The closest relationships between leaf EWT and vegetation indices at leaf level were obtained by proposed indices NDII, SIWSI, SR1640 and SR1060 , with R2 > 0.50 (P < 0.001). However, at canopy level, the best relationships with EWTcanopy was found for SR2, SR1640 and SR1060 , yielding R2 > 0.55. It can be learned that our proposed indices SR1640 and SR1060 showed similar agreement with leaf EWT and EWTcanopy. Other indices such as NDII, MSI2 and SWISI also showed a high coefficient of determination for both leaf EWT and EWTcanopy, but slightly poorer than SR1640 and SR1060 for EWTcanopy. It was also worth noting that most vegetation indices showed higher correlationships with EWTcanopy than with leaf EWT. This is different from the simulation results at leaf and canopy levels, where the relationships between the leaf EWT and simulated vegetation indices were always better than those between the canopy EWTcanopy and simulated vegetation indices. The higher coefficient of determination values at the canopy level than these at the leaf level may be caused by the significant correlationship between the measured EWTcanopy and LAI (R2 > 0.9), that was, most of the variation in EWTcanopy may be due to LAI rather than to leaf EWT. Previous studies showed that water-related indices may be a good measure of LAI (Colombo et al., 2008; Sims and Gamon, 2003). In present study, all spectral indices also provided good estimation of LAI, with the highest coefficient of determination R2 = 0.57 for SR2. These results suggested that a large part of the stronger correlations between EWTcanopy and vegetation indices may be due to LAI, especially for NDII, MSI2 and SWISI (see Figs. 5 and 7). A visual comparison of the relationships between water content and vegetation indices at both leaf and canopy levels is provided in Fig. 8. Only these with high coefficient of determination at leaf or canopy level were shown. It can be learned that the results obtained with field measurements were in agreement with simulation results. The new indices SR1060 and SR1640 were found to be significantly related with leaf EWT and EWTcanopy, with the highest coefficient of determination R2 = 0.50 (P < 0.0001) and R2 = 0.57 (P < 0.0001), respectively. Nevertheless, the significant relationship between EWTcanopy and SR2 was also identified, with R2 = 0.57 (P < 0.0001).

Q. Yi et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 67–75 0.07

(a)

0.05

Leaflevel

(b) EWTcanopy (kg/m2 )

EWT (g/cm2 )

0.06 0.05 0.04 0.03

R² = 0.915

0.02 0.01 0 0.2

0.3

0.4

0.5

0.6

0.7

Canopy level

0.04 Curvilinear: R² = 0.806 0.03

Linear: R² = 0.626

0.02 0.01 0

0.8

0

0.2

0.4

(c)

1

1.2

Canopy level

(d) EWTcanopy (kg/m 2 )

EWT (g/cm 2 )

0.8

0.05

Leaflevel

0.07 0.06

0.6

SR1060 (R1640/R1060)

SR1060 (R1640/R1060)

0.08

73

R² = 0.897

0.05 0.04 0.03 0.02

0.04 Curvilinear: R² = 0.910 0.03

Linear: R² = 0.832

0.02 0.01

0.01 0 1

1.5

2

2.5

3

0

3.5

0

1

2

SR1640 (R1060/R1640)

0.07

5

6

EWTcanopy (kg/m 2 )

R² = 0.919

0.04 0.03 0.02 0.01

Canopy level

(f)

0.06

EWT (g/cm2)

4

0.05

Leaf level (e)

0.05

3

SR1640 (R1060/R1640)

0.04 Curvilinear: R² = 0.779 0.03

Linear: R² = 0.779

0.02 0.01 0

0 1

1.2

1.4

1.6

0.5

1.8

1

1.5

SR2 (R1070/R1340)

2

2.5

3

3.5

SR2 (R1070/R1340)

Fig. 6. Relationships between leaf EWT and simulated vegetation indices at the leaf level and canopy EWTcanopy and simulated vegetation indices at the canopy level. Leaflevel and canopy-level simulations were conducted with PROPECT-5 and PROSPECT-5 + SAILH assuming random variation of Cm (0.001–0.018 g/cm2 ), Cw (0.01–0.06 g/cm2 ), N (1.2–2.6), and LAI (0.2–8), respectively.

Coefficient of determination R2

0.6 0.55 0.5 0.45 0.4 Leaf level Canopy level

0.35 0.3 NDII

NDWI

WI

SR1

SR2

MSI

MSI1

MSI2

SIWSI

SR1060 SR1640 ND1640

Vegetation index Fig. 7. Curvilinear relationships between leaf EWT and vegetation indices and between EWTcanopy and vegetation indices. Results obtained from leaf-level and canopy-level measurements.

74

Q. Yi et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 67–75

0.043

(a)

0.018

Leaf level EWTcanopy (kg/m 2 )

EWT (g/cm 2 )

0.035 0.031

R² = 0.503

0.027 0.023 0.019

Canopylevel

(b)

0.016

0.039

0.014 0.012 0.01

R² = 0.567

0.008 0.006 0.004 0.002 0

0.015 0.3

0.4

0.5

0.6

0.7

0.2

0.8

0.3

0.043

R² = 0.503

EWT (g/cm 2 )

0.035 0.031 0.027 0.023 0.019

0.6

0.7

Canopylevel

0.014 0.012

R² = 0.567

0.01 0.008 0.006 0.004 0.002 0

0.015 1.2

1.5

1.8

2.1

2.4

2.7

1.2

3

1.7

0.043

3.2

3.7

4.2

R² = 0.487

0.031 0.027 0.023 0.019

Canopylevel

(f)

0.016

EWTcanopy (kg/m 2 )

0.039 0.035

2.7

0.018

Leaf level

(e)

2.2

SR1640 (R1060/R1640)

SR1640 (R1060/R1640)

EWT (g/cm2 )

0.5

(d)

0.016

EWTcanopy (kg/m 2 )

0.039

0.018

Leaf level

(c)

0.4

SR1060 (R1640/R1060)

SR1060(R1640/R1060)

0.014 R² = 0.574

0.012 0.01 0.008 0.006 0.004 0.002 0

0.015 1.1

1.2

1.3

1.4

1.5

1.6

1.7

SR2 (R1070/R1340)

1.2

1.4

1.6

1.8

2

2.2

SR2 (R1070/R1340)

Fig. 8. Relationships between leaf EWT and vegetation indices and between EWTcanopy and vegetation indices. Results obtained from leaf-level and canopy-level measurements.

Conclusions In present study, we investigated some vegetation indices for estimating leaf and canopy water content in cotton using simulations and measurements and new vegetation indices were proposed based on simulation reflectance variations at leaf and canopy levels. The following results can be drawn:

(1) Simulation results showed that leaf and canopy reflectance around 1060 nm was the only band that simultaneously sensitive to leaf internal structure N and dry matter content and insensitive to leaf EWT or LAI. Reflectance at 1640 nm was sensitive to both leaf EWT and LAI, but the large variability of LAI at this band cannot cancel out the effects of leaf EWT on it. New index can be formulated with these two bands.

(2) Relationships between simulated leaf EWT and water-related vegetation indices tested and between EWTcanopy and these vegetation indices indicated that the new simple-ratio vegetation indices formulated with wavebands at 1060 nm and 1640 nm were sensitive to both leaf EWT and EWTcanopy and SR1640 (R1060 /R1640 ) was the most robust of all the indices tested at the canopy level. Furthermore, the simulated simple ratio index SR1640 showed the lowest relationship with LAI (R2 = 0.39) compared to other vegetation indices. (3) Results obtained from leaf-level and canopy-level measurements were in agreement with simulation results. The new simple ratio vegetation indices SR1060 and SR1640 that proposed in this study were found to be significantly related with leaf EWT and EWTcanopy (R2 > 0.5; P < 0.001). Nevertheless, this study confirms the robustness of other indices such as the NDII and SWISI (both R2 = 0.51; P < 0.001) for leaf

Q. Yi et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 67–75

EWT estimation and SR2 (R2 = 0.57; P < 0.001) for EWTcanopy estimation. Acknowledgements This work was funded by the following remote sensing programs: National Natural Science Foundation of China (41101430, 51109183, 41171295 and 41371393). The work assistances given by colleagues in Xinjiang Institute of Ecology and Geography Chinese Academy of Sciences are highly appreciated. References Aldakheel, Y.Y., Danson, F.M., 1997. Spectral reflectance of dehydrating leaves: measurements and modelling. Int. J. Remote Sens. 18, 3683–3690. Bacour, C., Jacquemoud, S., Tourbier, Y., Dechambre, M., Frangi, J.P., 2002. Design and analysis of numerical experiments to compare four canopy reflectance models. Remote Sens. Environ. 79, 72–83. Bowyer, P., Danson, F.M., 2004. Sensitivity of remotely sensed spectral reflectance to variation in live fuel moisture content. Remote Sens. Environ. 92, 297–308. Carlson, J.D., Burgan, R.E., 2003. Review of user needs in operational fire danger estimation: the Oklahoma example. Int. J. Remote Sens. 24, 1601–1620. Carter, G.A., 1991. Primary and secondary effects of water content on the spectral reflectance of leaves. Am. J. Bot. 78, 916–924. Carter, G.A., 1993. Responses of leaf spectral reflectance to plant stress. Am. J. Bot. 80, 239–243. Ceccato, P., Flasse, S., Tarantola, S., Jacquemoud, S., Gregoire, J.M., 2001. Detecting vegetation leaf water content using reflectance in the optical domain. Remote Sens. Environ. 77, 22–33. Ceccato, P., Gobron, N., Flasse, S., Pinty, B., Tarantola, S., 2002a. Designing a spectral index to estimate vegetation water content from remote sensing data: Part 1. Theoretical approach. Remote Sens. Environ. 82, 188–197. Ceccato, P., Flasse, S., Gregoire, J.M., 2002b. Designing a spectral index to estimate vegetation water content from remote sensing data: Part 2. Validation and applications. Remote Sens. Environ. 82, 198–207. Chuvieco, E., Cocero, D., Aguado, I., Palacios-Orueta, A., Prado, E., 2004. Improving burning efficiency estimates through satellite assessment of fuel moisture content. J. Geophys. Res. 109 (D14S07), 1–8. Cifre, J., Bota, J., Escalona, J.M., Medrano, H., Flexas, J., 2005. Physiological tools for irrigation scheduling in grapevine (Vitis vinifera L.). An open gate to improve water-use efficiency? Agric. Ecosyst. Environ. 106, 159–170. Cohen, W.B., 1991. Temporal versus spatial variation in leaf reflectance under changing water-stress conditions. Int. J. Remote Sens. 12, 1865–1876. Colombo, R., Meroni, M., Marchesi, A., Busetto, L., Rossini, M., Giardino, C., Panigada, C., 2008. Estimation of leaf and canopy water content in poplar plantations by means of hyperspectral indices and inverse modeling. Remote Sens. Environ. 112, 1820–1834. Danson, F.M., Steven, M.D., Malthus, T.J., Clark, J.A., 1992. Highspectral resolution data for determining leaf water content. Int. J. Remote Sens. 13 (3), 461–470. Datt, B., 1999. Remote sensing of water content in eucalyptus leaves. Aust. J. Bot. 47, 909–923. Davidson, A., Wang, S., Wilmshurst, J., 2006. Remote sensing of grassland shrubland vegetation water content in the shortwave domain. Int. J. Appl. Earth Obs. 8, 225–236. Dawson, T.P., Curran, P.J., Plummer, S.E., 1998. The biochemical decomposition of slash pine needles from reflectance spectra using neural networks. Int. J. Remote Sens. 19, 1433–1438. Fensholt, R., Sandholt, I., 2003. Derivation of a shortwave infrared water stress index from MODIS near- and shortwave infrared data in a semiarid environment. Remote Sens. Environ. 87, 111–121. Féret, J.B., Franc¸ois, C., Asner, G.P., Gitelson, A.A., Martin, R.E., Bidel, L.P.R., 2008. PROSPECT-4 and 5: advances in the leaf optical properties model separating photosynthetic pigments. Remote Sens. Environ. 112, 3030–3043. Féret, J.B., Franc¸ois, C., Gitelson, A., Asner, G.P., Barry, K.M., Panigada, C., Richardson, A.D., Jacquemoud, S., 2011. Optimizing spectral indices and chemometric analysis of leaf chemical properties using radiative transfer modeling. Remote Sens. Environ. 115, 2742–2750. Gao, B.C., 1996. NDWI-a normalized difference water index for remote sensing of vegetation liquid water from space. Remote Sens. Environ. 58, 257–266. Gao, B.C., Goetz, A.F.H., 1995. Retrieval of equivalent water thickness and information related to biochemical-components of vegetation canopies from AVIRIS data. Remote Sens. Environ. 52, 155–162. Hardisky, M.A., Klemas, V., Smart, R.M., 1983. The influences of soil salinity, growth form, and leaf moisture on the spectral reflectance of Spartina alterniflora canopies. Photogramm. Eng. Remote Sens. 49, 77–83.

75

Hunt Jr., E.R., 1991. Airborne remote sensing of canopy water thickness scaled from leaf spectrometer data. Int. J. Remote Sens. 12, 643–649. Hunt Jr., E.R., Rock, B.N., 1989. Detection of changes in leaf water content using nearand middle-infrared reflectance. Remote Sens. Environ. 30, 43–54. Jacquemoud, S., Baret, F., 1990. PROSPECT: a model of leaf optical properties spectra. Remote Sens. Environ. 34, 75–91. Jacquemoud, S., Baret, F., Andrieu, B., Danson, F.M., Jaggard, K.W., 1995. Extraction of vegetation biophysical parameters by inversion of PROSPECT + SAIL model on sugar beet canopy reflectance data. Application to TM and AVIRIS sensors. Remote Sens. Environ. 52, 163–172. Jacquemoud, S., Bacour, C., Poilve, H., Frangi, J.P., 2000. Comparison of four radiative transfer models to simulate plant canopies reflectance: direct and inverse mode. Remote Sens. Environ. 74, 471–481. Jacquemoud, S., Verhoef, W., Baret, F., Bacour, C., Zarco-Tejada, P.J., Asner, G.P., 2009. PROSPECT + SAIL models: a review of use for vegetation characterization. Remote Sens. Environ. 113, S56–S66. José, R.R., David, R., Eli, C., Susan, U., David, R.S., 2007. Evaluation of hyperspectral reflectance indexes to detect grapevine water status in vineyards. Am. J. Enol. Vitic. 58 (3), 302–317. Kimes, D.S., Markham, B.L., Tucker, C.J., 1981. Temporal relationships between spectral response and agronomic variables of a corn canopy. Remote Sens. Environ. 11, 401–411. Knapp, A.K., Carter, G.A., 1998. Variability in leaf optical properties among 26 species from a broad range of habitats. Am. J. Bot. 85, 940–946. Kuusk, A., 1991. The angular distribution of reflectance and vegetation indices in barley and clover canopies. Remote Sens. Environ. 37, 143–151. Maki, M., Ishiahra, M., Tamura, M., 2004. Estimation of leaf water status to monitor the risk of forest fires by using remotely sensed data. Remote Sens. Environ. 90, 441–450. ˜ Penuelas, J., Filella, I., Biel, C., Serrano, L., Save, R., 1993. The reflectance at the 950–970 nm region as an indicator of plant water status. Int. J. Remote Sens. 14, 1887–1905. ˜ Penuelas, J., Gamon, J.A., Fredeen, A.L., Merino, J., Field, C.B., 1994. Reflectance indices associated with physiological changes in nitrogen and water limited sunflower leaves. Remote Sens. Environ. 48, 135–146. ˜ Penuelas, J., Pinol, J., Ogaya, R., Filella, I., 1997. Estimation of plant water concentration by the reflectance water index WI (R900 /R970 ). Int. J. Remote Sens. 14, 1887–1905. Riggs, G.A., Running, S.W., 1991. Detection of canopy water stress in conifers using the airborne imaging spectrometer. Remote Sens. Environ. 35, 51–68. Rock, B.N., Vogelmann, J.E., Williams, D.L., Vogelmann, A.F., Hoshizaki, T., 1986. Remote detection of forest damage. J. Biosci. 36, 439–445. Saura-Mas, S., Lioret, F., 2007. Leaf and Shoot water content and leaf dry matter content of mediterranean woody species with different post-fire regenerative strategies. Ann. Bot. 99 (3), 545–554. ˜ Serrano, L., Ustin, S.L., Roberts, D.A., Gamon, J.A., Penuelas, J., 2000. Deriving water content of chaparral vegetation from AVIRIS data. Remote Sens. Environ. 74, 570–581. Sims, D.A., Gamon, J.A., 2003. Estimation of vegetation water content and photosynthetic tissue area from spectral reflectance: a comparison of indices based on liquid water and chlorophyll absorption features. Remote Sens. Environ. 84, 526–537. Stimson, H.C., Breshears, D.D., Ustin, S.L., Kefauver, S.C., 2005. Spectral sensing of foliar water conditions in two co-occurring conifer species: Pinus edulis and Juniperus monosperma. Remote Sens. Environ. 96, 108–118. Thomas, J.R., Namken, L.N., Oerther, G.F., Brown, R.G., 1971. Estimating leaf water content by reflectance measurements. Agron. J. 63, 845–847. Ustin, S.L., Roberts, D.A., Pinzón, J., Jacquemoud, S., Gardner, M., Scheer, G., ˜ C.M., Palacios-Orueta, A., 1998. Estimating canopy water conCastaneda, tent of chaparral shrubs using optical methods. Remote Sens. Environ. 65, 280–291. Verhoef, W., 1984. Light scattering by leaf layers with application to canopy reflectance modelling: the SAIL model. Remote Sens. Environ. 16, 125–141. Yebra, M., Chuvieco, E., 2009. Linking ecological information and radiative transfer models to estimate fuel moisture content in the Mediterranean region of Spain: solving the ill-posed inverse problem. Remote Sens. Environ. 113, 2403–2411. ˜ D., Zylstra, P., Hunt Jr., E.R., Danson, Yebra, M., Dennison, P.E., Chuvieco, E., Riano, F.M., Qi, Y., Jurdao, S., 2013. A global review of remote sensing of live fuel moisture content for fire danger assessment: moving towards operational products. Remote Sens. Environ. 136, 455–468. Zarco-Tejada, P.J., Miller, J.R., Noland, T.L., Mohammed, G.H., Sampson, P.H., 2001. Scaling-up and model inversion methods with narrow-band optical indices for chlorophyll content estimation in closed forest canopies with hyperspectral data. IEEE Trans. Geosci. Remote Sens. 39 (7), 1491–1507. Zarco-Tejada, P.J., Rueda, C.A., Ustin, S.L., 2003. Water content estimation in vegetation with MODIS reflectance data and model inversion methods. Remote Sens. Environ. 85, 109–124.