J Plant Res (2009) 122:317–326 DOI 10.1007/s10265-009-0215-y
REGULAR PAPER
Predicting vegetation water content in wheat using normalized difference water indices derived from ground measurements Chaoyang Wu Æ Zheng Niu Æ Quan Tang Æ Wenjiang Huang
Received: 24 July 2008 / Accepted: 21 December 2008 / Published online: 26 February 2009 Ó The Botanical Society of Japan and Springer 2009
Abstract Vegetation water content (VWC) is an important variable for both agriculture and forest fire management. Remote sensing technology offers an instantaneous and nondestructive method for VWC assessment provided we can relate in situ measurements of VWC to spectral reflectance in a reliable way. In this paper, based on radiative transfer models, three new normalized difference water indices (NDWI) are proposed for VWC [fuel moisture content (FMC), and equivalent water thickness (EWT)] estimation, taking both leaf internal structure and dry matter content into account. Reflectance at 1,200, 1,450 and 1,940 nm were selected and normalized with reflectance at 860 nm to establish three water indices, NDWI1200, NDWI1450 and NDWI1940. Good correlations were observed between FMC (R2 = 0.65–0.80) and EWT (both at the leaf scale, R2 = 0.75–0.81 for EWTL and at the canopy scale, R2 = 0.80–0.83 for EWTC) at various stages of wheat crop development. Keywords Equivalent water thickness Normalized difference water index Vegetation water content
C. Wu (&) Z. Niu Q. Tang W. Huang The State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing Applications, Chinese Academy of Sciences, 100101 Beijing, China e-mail:
[email protected] C. Wu Q. Tang Graduate University of Chinese Academy of Sciences, 100039 Beijing, China W. Huang National Engineering Research Center for Information Technology in Agriculture, 100089 Beijing, China
Introduction Remote sensing has provided a potential means to overcome the limitations of traditional methods of large-scale vegetation sampling by offering a non-destructive and timely approach at the landscape scale (Chen et al. 2005; Davidson et al. 2006). The fundamental basis of monitoring vegetation water content (VWC) through remote sensing is that reflectance in the near-infrared (NIR, 700–1,300 nm) and short-wave infrared (SWIR, 1,300–2,500 nm) will change with variations in water status (Ceccato et al. 2001; Danson and Bowyer 2004; Seelig et al. 2008). In order to reduce the influences of leaf structural, background soil contamination and of atmospheric effects on a single band reflectance, vegetation indices (VI) composed of two or more bands are derived (Gao 1996; Zarco-Tejada et al. 2003). Recently, researchers have explored methodology for VWC estimation through remote sensing techniques based on radiative transfer models such as PROSPECT (Jacquemoud and Baret 1990) and SAILH (Kuusk 1985). PROSPECT is a leaf-scale model that can simulate reflectance and transmittance from 400 to 2,500 nm by four foliar biochemistry and scattering parameters (leaf structure parameter, N; chlorophyll content, Ca ? b; equivalent water thickness, EWT; and dry matter content, Dm). SAILH (scattering by arbitrary inclined leaves) is a canopy model based on a four-stream approximation of the radiative transfer equation with two direct fluxes (incident solar flux and radiance in the viewing direction) and two diffuse fluxes (upward and downward hemispherical flux). More typically, VWC can be estimated by simple ratio of reflectance values of two bands, including a reference band where the water absorption coefficient is low and a measurement band where water absorption is moderate or
123
318
high (Gao 1996). As most water indices are proposed based on the ratio of two bands, with both water absorption and other features considered (e.g. N and Dm), the selection of wavelengths is critical for the sensitivity of the index to changes in plant water status (Eitel et al. 2006). The influence of VWC is seen mainly in the NIR and SWIR (Gao 1996; Pen˜uelas et al. 1997). Thus, incorporating reflectance in NIR–SWIR will provide a pigment-independent quantitative estimate of VWC, although this method still needs further refinement to account for the observed effects of leaf structure, leaf dry matter, canopy structure and leaf area index (LAI) (Zarco-Tejada et al. 2003). In the SWIR region, specific water absorption features may provide insight for VWC estimation, e.g., wavelengths 970, 1,200, 1,600, 1,950 and 2,250 nm (Bowyer and Danson 2004; Claudio et al. 2006). However, it has also been demonstrated that remote sensing using a single SWIR wavelength range alone is not sufficient in the estimation of VWC as two other leaf parameters (N and Dm) are also responsible for leaf reflectance variations in the SWIR region (Ceccato et al. 2002). In order to find the most appropriate bands for VWC estimation, Ceccato et al. (2002) performed a leaf-level sensitivity analysis of the PROSPECT model (Jacquemoud and Baret 1990) in which it was demonstrated that 1,457 and 860 nm were the wavelengths with the highest and lowest sensitivity to leaf water variation, respectively. Reflectance of 860 nm is deemed an effective reference band for the estimation of VWC due to its relatively better canopy penetration ability. This conclusion was further confirmed by later studies (Eitel et al. 2006; Colombo et al. 2008; Cheng et al. 2008). However, the selection of a measurement band where water absorption is high is much more complicated. First, the band must be well detectable if there is a variation in VWC because reflectance in some bands cannot be detected under slight or moderate water stress (Eitel et al. 2006). This means that bands with strong absorption features are the most appropriate candidates for the measurement band (Jackson et al. 2004; Eitel et al. 2006; Colombo et al. 2008). Second, the measurement band must have a relatively high incoming energy and low level of interference from atmospheric water vapor, indicating that the longer wavelength in the SWIR region may not be suitable for VWC estimation (Sims and Gammon 2003; Claudio et al. 2006). Despite arguments over band selection, different vegetation water indices [normalized difference water index (NDWI), normalized difference infrared index (NDII), maximum difference water index (MDWI), water band index (WBI), R1300/R1450 and T1300/T1450] composed of bands in these absorption peaks have proved to be applicable in the estimation of VWC (Chen et al. 2005; Stimson
123
J Plant Res (2009) 122:317–326
et al. 2005; Claudio et al. 2006; Eitel et al. 2006; Yilmaz et al. 2008; Yebra et al. 2008; Seelig et al. 2008). Traditionally, there are two ways of assessing the VWC of a leaf. The first definition is fuel moisture content (FMC), defined as the ratio between the quantity of water [fresh weight (FW) - dry weight (DW)] and either FW or DW (Chuvieco et al. 1999): FMC ¼ ðFW DWÞ=FWðor DWÞ 100%
ð1Þ
where FW is the fresh weight measured in the field, and DW is the oven dry weight of the same sample. FMC expresses the amount of water in a leaf relative to the amount of fresh weight or dry matter and is related to both leaf water content and leaf dry matter content. The other definition, which is more commonly used in the remote sensing literature, is leaf water content per unit leaf area, or EWT (in units of g cm-2). EWT relates to a hypothetical thickness of a single layer of water averaged over the whole leaf area (Danson et al. 1992) EWTL ¼ ðFW DWÞ=A
ð2Þ
where A is the one-side leaf area and EWTL represents EWT at leaf scale. Equation 2 requires an independent measurement of leaf area. At a canopy scale, EWTC (in units of g m-2) is determined as the product of EWTL and green LAI (Colombo et al. 2008): EWTC ¼ EWTL LAI
ð3Þ
The objective of this paper is to present a study of NDWI in VWC (both FMC and EWT) estimation by incorporating more bands that are responsive to water variation signals. For clarity, we have listed all abbreviations and term used in this study in Table 1, including some notes and descriptions. We first introduced some bands in the NIR and SWIR regions for EWT evaluation based on the PROSPECT model. New NDWIs were proposed by taking account of leaf internal structure and dry matter effects. A validation study in wheat was carried out on four days (17 April, 28 April, 16 May and 29 May) in 2007, corresponding to typical crop growth phases of wheat.
Materials and methods Study sites and materials The study area is located at the National Experiment Station for Precision Agriculture (40°10.60 N, 116°26.30 E) 20 km northeast of Beijing, China. This experimental station has been operational since 2001 and is used for precision agriculture research. The site is located within a warm-
J Plant Res (2009) 122:317–326
319
Table 1 Abbreviations used in this paper
Table 2 Description of wheat cultivar
Abbreviation
Number
Cultivar
1
Laizhou 3729
Erectophile
Dark green
2
Chaoyou 66
Spherical
Dark green
3
Linkang 2
Planophile
Dark green
4
Jing 8
Spherical
Dark green
5
Jing 411
Erectophile
Light green
6
9507
Planophile
Light green
Definition (units)
Water content descriptions VWC
Vegetation water content (%)
CWC
Canopy water content (g m-2) -2
EWTL
Leaf equivalent water thickness (g cm
EWTC
Canopy equivalent water thickness (g m-2)
or cm)
Indices VI
Canopy leaf orientation
Leaf colour
Vegetation indices
WI
Water indices
NDVI
Normalized difference vegetation indices
NDWI
Normalized difference water indices
NDII MDWI
Normalized difference infrared index Maximum difference water index
WBI
Water band index
the canopy surface. Reflectance spectra were derived through calibration relative to a 99% white reference panel (Labsphere, North Sutton, NH). Measurements were conducted systematically at each plot. Leaf and canopy EWT calculation
Model input parameters N
Leaf internal structure
Dm
Dry matter content (g cm-2) -2
Cw
Water content (g cm
Cab
Chlorophyll content ( lg cm-2)
LAI
Leaf area index (m2 m-2)
or cm)
Others FW
Fresh weight (g)
DW
Dry weight (g)
SWD
Specific water density (g cm-2)
SLW
Specific leaf weight (g cm-2)
Twenty individual leaves of wheat were randomly collected in the field and immediately sealed in plastic bags and placed on ice until analysis. In the laboratory, the leaf samples were cut with a knife into pieces approximately 3 cm long. Fresh weight was obtained using an analytical balance and then samples were dried at 80°C in an oven for 24 h to measure the DW values. The leaf areas of all samples were measured with a portable area meter LI-3000A (LI-COR, Lincoln, NE). Leaf area index calculation
temperate zone with a mean annual rainfall of 507.7 mm and a mean annual temperature of 13°C. The plant selected for this study was winter wheat (Triticum aestivum L.), which is one of the most important crops in China. Ground measurements Canopy reflectance acquisition Six cultivars of winter wheat that can be classified into three leaf structural types were used in this experiment (Table 2). Each winter wheat cultivar was cultivated in an area of about 4,000 m2 (about 200 m 9 20 m). Wheat grew in a silt clay soil with sufficient water supply. Data was collected in the morning around 10:30 a.m. local time on four clear days during a typical wheat growth season: 17 April, 28 April, 16 May and 29 May in 2007, corresponding to jointing, heading, flowering and ripening phases, respectively. Canopy radiance data were collected from 380 nm to 2,500 nm (1 nm sampling interval) using a portable spectroradiometer (FS-FR2500, Analytical Spectral Devices, Boulder, CO) with field of view of 25° normal to the canopy located at a distance of approximately 100 cm from
All aboveground plant materials within a 0.6 m 9 0.6 m area were collected immediately following spectral measurements. Leaves of all the sampled plants were collected to determine the LAI. A subsample of plant leaves was used to measure the leaf area in the laboratory with the Li-COR 3100A area meter. The leaf area of the subsample (LAIsub) was used to calculate the LAI of the 0.6 m 9 0.6 m sample area. Sensitivity study Leaf reflectance simulation using the PROSPECT model The PROSPECT model was adopted to analyze leaf reflectance changes due to variations in water content. To examine the effect of water content on reflectance spectra, other parameters were assigned fixed values (these fixed values are determined through in situ measurements) and the EWT was changed from 0.002 to 0.03 g cm-2 in steps of 0.002 (Table 3). The selection of parameter values was based on Hosgood et al. (1995). For the simulation, reflectance was acquired from 600 to 2,400 nm at 10-nm intervals.
123
320
J Plant Res (2009) 122:317–326
Table 3 Parameters used in simulating leaf reflectance using the PROSPECT model Parameter
Values (unit)
N
1.35
Notes Leaf structure parameter -2
Equivalent water thickness
Cw
0.002–0.030 g cm
Dm
0.01 g cm-2
Dry matter content
Ca ? b
35 lg cm-2
Chlorophyll a ? b content
Table 4 Parameters used in simulating canopy reflectance using the SAILH (scattering by arbitrary inclined leaves) model SAILH parameters
Values
Leaf optical propertiesa
N = 1.55; Cab = 35 lg cm-2, Cw = 0.01 g cm-2, Dm = 0.01 g cm-2
Soil reflectance
In situ average measurements at ten spots
LAI
0.5–7 in 0.1 steps
Leaf angle distribution
Spherical
Sun zenith angle
45°
Sensor view angle
0° (nadir)
Fraction specular flux
1
a
Leaf reflectance and transmittance were simulated with PROSPECT with the input parameters shown (see Table 1 for definitions)
Canopy reflectance simulation with SAIL model Canopy reflectance spectra were simulated using the SAIL model (Verhoef 1984), as revised for taking hotspot effect into consideration (Kuusk 1985). We used these models to simulate canopy reflectance and investigate the variation of indices due to changes in LAI. We considered the effect of changing LAI from 0.5 to 7 in steps of 0.1 while other parameters were set to pre-determined values (Table 4). To avoid uncertainty in running the model in the case of Chinese wheat, all parameters was evaluated by determining empirical values and soil reflectance was determined using in situ measurements.
Results Leaf water absorption features Because different cultivars have different canopy leaf orientations, the spectra differed depending on the cultivar. The in situ canopy reflectance of the samples is shown in Fig. 1. In the simulation part, different EWT exerted an evident effect on reflectance from 900 to 2,400 nm, while little effect was seen in the red edge region (see Fig. 2). In this region, there were typically four absorption peaks, centered on 970, 1,200, 1,450 and 1,940 nm. Reflectance of
123
Fig. 1 Spectral curves of the sample wheat from 400 nm to 2,400 nm from data obtained on 17 April 2007. Band regions 1,360–1,380 and 1,780–1,930 nm were excluded because of decreased sensitivity of the ASD spectroradiometer (Analytical Spectral Devices, Boulder, CO) in these regions
these bands showed strong sensitivity to leaf EWT values. The absorption features became much deeper as EWT increased from 0.002 to 0.03 g cm-2, indicating their potential for water content monitoring. Of the main water absorption features (970, 1,200, 1,450 and 1,950 nm), reflectance at 970 nm was not selected as there was no relationship between leaf reflectance and specific water density (SWD) at 975 nm (Danson et al. 1992). Wavelengths longer than 2,000 nm were also excluded due to relatively low incoming energy and high levels of interference from atmospheric water vapor (Sims and Gammon 2003). To see the absorption features clearly, band reflectance was plotted as a function of EWT. Figure 3 shows that reflectance at 860 nm was totally independent of EWT variation, which makes it a suitable candidate for a reference band. For the other three strong absorption bands, reflectance at 1,200 nm exhibited a linear relationship with EWT, while the other two showed certain saturation regions, especially for reflectance at 1,940 nm, which became saturated quickly after an EWT of 0.01 g cm-2, indicating that reflectance at 1,450 and 1,940 nm are very sensitive to low values of EWT and will lose its sensitivity at high EWT values. Derivation of normalized difference water indices Based on combinations of bands where water absorption is high and low, three new NDWIs are proposed, NDWI1200 ¼ ðR860 R1200 Þ=ðR860 þ R1200 Þ NDWI1450 ¼ ðR860 R1450 Þ=ðR860 þ R1450 Þ NDWI1940 ¼ ðR860 R1940 Þ=ðR860 þ R1940 Þ where Rx represents reflectance at x nm.
ð4Þ
J Plant Res (2009) 122:317–326
321
Fig. 2 Water absorption features of leaf spectra (600– 2,400 nm) simulated with the PROSPECT model
Fig. 3 Reflectance variations at 860, 1,200, 1,450 and 1,940 nm with equivalent water thickness (EWT) changes from 0.002 to 0.03 g cm-2. Note saturation of reflectance at 1,940 nm with EWT variations
Fig. 4 Relationship between different normalized difference water indices (NDWIs) and EWT variations at leaf level. Note the rapid rise in NDWI1940 to a value \0.008 g cm-2 and almost complete saturation by 0.016 g cm-2
The three NDWIs were plotted as a function of EWTL with data simulated from the PROSPECT model (Fig. 4). To compare these indices directly, we scaled the sensitivity results from 0 to 1. Similar to the characteristics of the single band, NDWI1200 displayed an almost linear relationship with increasing EWT. But NDWI1940 showed a rapid rise at low EWT values (below 0.008 g cm-2) and quickly became saturated after reaching a value of 0.016 g cm-2. NDWI1450 fell between the two. Figure 4 implies that NDWI1940 is not a suitable index for high EWT leaves but shows a high sensitivity to low EWT, as evident from the larger variation in NDWI1940 in the low EWT range. On the contrary, the nearly linear relationship between NDWI1200 and EWT tentatively suggests that NDWI1200 is suitable for interpreting high EWT levels.
Effect of leaf internal structure and dry matter content on NDWIs with simulated leaf reflectance The effects of leaf internal structure (N) and dry matter content (Dm) were studied using the PROSPECT model by fixing other parameters at constant values. For reflectance simulation with N variations (N = 1, 1.5, 2, 2.5), chlorophyll content (Cab), Dm and water content (Cw) were 0.035, 0.01 and 0.01 g cm-2, respectively. For the case of Dm variation, the parameters N, Cab and Cw were 1.5, 0.035 and 0.01 g cm-2, respectively. Figure 5 depicts the relationships between the NDWIs and N (Fig. 5a) and dry matter content (Fig. 5b). For all NDWIs, variations in leaf structure N and Dm exerted strong effects.
123
322
J Plant Res (2009) 122:317–326
Fig. 5 Relationship between NDWIs and a leaf internal structure (N), and b dry matter content (Dm) variations
As the EWTL of all samples were relatively low (about 60% of all data are below 0.005 g cm-2, Fig. 8a), the index NDWI1940 showed the highest accuracy (R2 = 0.81, n = 24) in EWTL estimation because it was most sensitive to low EWTL values. For the other two indices, the R2 for EWTL estimation were comparable, at 0.78 and 0.75 for NDWI1200 and NDWI1450, respectively. EWTC was obtained by scaling EWTL with LAI. Detailed information of LAI and plant height is shown in Table 5. For EWTC estimation, NDWI1940 still showed the highest accuracy, with R2 = 0.83 (n = 24). Overall, all NDWIs were strongly correlated with EWTC with an obvious increase in R2 compared to that of EWCL. Fig. 6 Relationship between leaf area index (LAI) and NDWIs composed of reflectance simulated with SAILH model
Discussion and conclusions Effect of LAI on NDWIs with simulated canopy reflectance To assess the newly derived NDWIs, a sensitivity study of LAI at the canopy scale with reflectance simulated with SAILH model was also conducted (Fig. 6). Three NDWIs became saturated as LAI increased from 0.5 to 7. NDWI1940 was most sensitive to changes in LAI, with an evident saturate point (LAI = 3.5) as well as a rapid rise at low LAI values (LAI \ 1.5). The saturation problem tended to be alleviated in NDWI1200 as there was no clear saturation until LAI = 6. The NDWI1450 curve lay between the latter two curves. Regression between indices and VWC of wheat The FMC values of wheat ranged from 55% to about 90% for all samples. For FMC of wheat, the three NDWIs all proved to be feasible candidates for FMC estimation, with correlation coefficients R2 ranging from 0.65 to 0.80 (n = 24; Fig. 7). For wheat EWT estimation, the three NDWIs showed a strong correlation with both EWTL and EWTC for all the data collected (R2 = 0.75–0.81 for EWTL and 0.80–0.83 for EWTC, n = 24; Fig. 8).
123
Definition of FMC and EWT The definition of FMC (Eq. 1) suggests that it is dependent largely on two factors: leaf water and Dm. However, previous studies have also pointed out that the leaf internal structure (N), which controls leaf internal scattering, may also be correlated to specific leaf weight (SLW) and thus influence FMC (Jacquemoud and Baret 1990; Danson and Bowyer 2004). Therefore, for the estimation of FMC, we should also take leaf Dm and N into account. The NIR and SWIR regions provide information in terms of N and Dm, respectively. For example, Jacquemoud et al. (1996) proved that, in general, absorption by dry matter increases with wavelength throughout the SWIR region. Bowyer and Danson (2004) demonstrated that reflectance variability due to changes in FMC (EWT and Dm) is expressed most strongly in the SWIR and NIR, respectively. In fact, the retrieval of FMC can be accurate for a specific species only by correlating the indices with field measurements for that species. This is due to the fact that the variations are caused primarily by changes in EWT content but also, to some degree, to changes in dry matter. Therefore, the retrieval of FMC is physically impossible since the wavelengths in the SWIR are sensitive to EWT. As demonstrated by Ceccato et al. (2001), different species
J Plant Res (2009) 122:317–326
323
Thus, FMC is a combination of water and dry matter, both of which are important parameters for assessing fire probability (Levine 1996). This may explain the difficulty in FMC estimation with reflectance data aimed mainly at assessing water variations (Bowyer and Danson 2004; Verbesselt et al. 2007). Problems with VWC estimation Effect of experimental species
Fig. 7 Relationship between NDVIs and FMC for the six cultivars of wheat in the growth cycle
might have the same FMC (expressed in %) but different EWT. The energy in SWIR regions is absorbed physically by the quantity of water per unit area. Similarly, different species might have the same EWT but different FMC. EWT represents the hypothetical thickness of a single layer of water that would cover the surface of a leaf if water contained in the leaf was uniformly distributed over the entire leaf (Danson et al. 1992). This definition is directly related to the water status of a leaf and other factors are excluded. The EWT contains more information about water status, while FMC is more commonly used in the application of fire risk management (Ceccato et al. 2001; Danson and Bowyer 2004; Claudio et al. 2006).
Many studies have demonstrated that the relationship between water indices composed of spectral reflectance and water content is species-specific (Chen et al. 2005; Yilmaz et al. 2008). For example, corn VWC estimation is usually more successful than that for soybean (Yilmaz et al. 2008) due to the very much higher water content in corn. For soybean, which has low leaf water content, satellite vegetation indices are not sensitive enough to detect the status of water by different band combinations (Chen et al. 2005). Other aspects that require our attention are the leaf thickness and height of plant species. A recent study by Seelig et al. (2008) demonstrated that water indices based on light reflected from leaves or transmitted through leaves depend not only on the VWC of leaf cells, which act as an indicator for water deficit stress, but also to a large extent on the thickness these leaves have developed. Compared to leaf thickness, plant height may exert an effect indirectly. Bowyer and Danson (2004) performed a sensitivity study on reflectance to variations in FMC at different scales, and concluded that greater accuracy can be expected for grassland than shrub-land and forests. Besides the different influences of chlorophyll content and leaf internal structure, we can attribute this effect partly to the structure differences caused by LAI. Spatial and temporal variation of LAI will add significant weight to VWC estimation. For taller species, LAI will be more important as it will make the structure difference more evident compared to that of shorter species. Therefore, FMC estimation for species of different heights offers little chance of success, especially for sparsely vegetated areas. Effect of leaf structure parameters and dry matter content The bands measured in the new NDWIs were selected in the SWIR regions, assuming that these regions are responsive only to water changes. However, this is not the case. This can be attributed to the strong influence of Dm on reflectance in the SWIR region and to the correlation of N and SLW (Danson and Bowyer 2004). Furthermore, the NDWIs consist of two bands (a measurement and a
123
324
J Plant Res (2009) 122:317–326
Fig. 8 Relationship between NDWIs and a EWTL and b EWTC of the six cultivars of wheat in the growth cycle
Table 5 LAI values and plant heights of each cultivar of wheat on four experimental days in 2007 Cultivar
17 April 2
28 April -2
2
16 May -2
29 May
LAI (m m )
Height (cm)
LAI (m m )
Height (cm)
LAI (m m )
Height (cm)
LAI (m2 m-2)
Height (cm)
1
2.64
34.2
3.59
55.7
2.62
89.4
2.29
106.5
2 3
2.54 2.89
31.6 32.6
3.80 3.41
57.6 52.9
3.65 3.40
87.3 88.6
2.46 3.22
108.9 112.5
4
2.18
37.4
3.51
58.3
3.36
90.3
2.49
105.3
5
2.51
35.5
3.53
55.3
4.18
91.5
3.48
109.7
6
2.75
34.7
3.43
53.9
3.29
92.8
2.53
110.2
reference band) thought to be responsible only for water variation and other non-water factors, respectively. However, this is also hardly the case because Dm also exerts some effect in the SWIR region from where we selected the measurement bands (Bowyer and Danson 2004). As the leaf structure parameter N is related to the cell structure, it can change with species over different growth stages. Wheat is a monocotyledonous species and the N of wheat in fast growth stages will fall between 1 and 1.5. Therefore, N will not affect NDWI1940 for EWT estimation of wheat, as can be seen in Fig. 5a. However, as wheat become senescent, N will disturb EWT estimation. Unlike
123
2
-2
N, Dm will exert its effect on all three indices, and is much more difficult to dispel (Fig. 5b). Effect of water status The VWC itself may also have some effect on estimation using spectral indices. One specific reason for this is that physiological responses to drought stress happen over a very small interval of VWC and sensors may not be able to detect the change (Jackson et al. 2004). Eitel et al. (2006) carried out a sensitivity analysis of indices to light- and moderate water-stress and their results indicated that no
J Plant Res (2009) 122:317–326
current indices are appropriate for monitoring of slight water stress. Furthermore, different FMC values also affect the sensitivity of spectral indices because a decrease of FMC under certain values could result in a non-comparable reduction in spectral reflectance (Carter 1991). Therefore, water indices may have their own advantages for specific applications. For example, certain indices may have special sensitivity to severe drought conditions (e.g., NDWI1940 in this paper, given its high sensitivity to low EWT). The range of FMC values in this paper (55–90% for wheat) would ensure detectable changes in the reflectance region and thus provide a relatively high accuracy in FMC estimation compared to other studies (Ceccato et al. 2001). This is confirmed by Danson and Bowyer (2004), who stated that estimation accuracy using a ‘global’ range of FMC was very low. For FMC estimation of wheat, reflectance data is collected at a canopy scale and assumes a homogeneous situation. Therefore, the R2 correlation coefficients are acceptable for FMC estimation (around 0.7; n = 24). However, if one aims to estimate FMC across species with satellite data, such indices may result in lower correlation. Effect of scale In the content of VWC estimation, special attention should be paid to FMC because it is a relative descriptor of the amount of water in the vegetation and does not scale from leaf to canopy level. Therefore, the sensitivity of canopy reflectance to changes in FMC will be affected by spatial and temporal variations in LAI (Danson and Bowyer 2004). In addition, scaling-up should also consider the mix of vegetation species and vegetation structures within the instantaneous field of view of an imaging instrument. An interesting aspect of LAI is that the sensitivity of NDWIs to EWTL variations is similar to that of NDWIs to LAI. This indicates a certain relationship between VWC (in terms of EWT) and LAI. This can be explained as follows. Sufficient water supply in a leaf ensures the normal function of photosynthesis and thus results in augmentation of LAI. This potential relationship between EWT and LAI was also confirmed in recent studies by Roberts et al. (2004), who demonstrated that water indices provide better estimates of LAI than NDVI. Although tested in a relatively homogenous situation, our results imply potential ecophysiological relationships that might be common to various types of vegetation. Therefore, these results could provide some insights into remote assessment of VWC in other ecosystems using a wide range of available spectral data. Further research is needed to validate these NDWIs with satellite data and to fully understand the underlying mechanisms of water content (both FMC and EWT) estimation using NDWIs.
325 Acknowledgments We thank Prof. Benoit Rivard and Dr. Feng Jilu for language correction of the paper. We also offer our thanks to anonymous reviewers for constructive suggestions. This work was funded by the China’s Special Funds for Major State Basic Research Project (2007CB714406), the Knowledge Innovation Program of the Chinese Academy of Sciences (KZCX2-YW-313), and the State Key Laboratory of Remote Sensing Science (KQ060006).
References Bowyer P, Danson FM (2004) Sensitivity of remotely sensed spectral reflectance to variation in live fuel moisture content. Remote Sens Environ 92:297–308 Carter GA (1991) Primary and secondary effects of water content on the spectral reflectance of leaves. Am J Bot 78:916–924 Ceccato P, Flasse S, Tarantola S, Jacquemoud S, Gregoire JM (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 (2002) Designing a spectral index to estimate vegetation water content from remote sensing data. Part 1. Theoretical approach. Remote Sens Environ 82:188–197 Chen D, Huang J, Jackson TJ (2005) Vegetation water content estimation for corn and soybeans using spectral indices from MODIS near- and short-wave infrared bands. Remote Sens Environ 98:225–236 Cheng YB, Ustin SL, Rian˜o D, Vanderbilt VC (2008) Water content estimation from hyperspectral images and MODIS indexes in Southeastern Arizona. Remote Sens Environ 112:363–374 Chuvieco E, Deshayes M, Stach N, Cocero D, Riano D (1999) Shortterm fire risk foliage moisture content estimation from satellite data. In: Chuvieco E (ed) Remote sensing of large wildfires in the European Mediterranean Basin. Springer (University of Alcala, Spain), Berlin Claudio HC, Cheng YF, Fuentes DA, Gamon JA, Luo HY, Oechel W, Qiu HL, Rahman AF, Sims DA (2006) Monitoring drought effects on vegetation water content and fluxes in chaparral with the 970 nm water band index. Remote Sens Environ 103:304–311 Colombo R, Meroni M, Marchesi A, Busetto L, Rossini M, Giardin 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 FM, Bowyer P (2004) Estimating live fuel moisture content from remotely sensed reflectance. Remote Sens Environ 92:309– 321 Danson FM, Steven MD, Malthus TJ, Clark JA (1992) Highspectral resolution data for determining leaf water content. Int J Remote Sens 13(3):461–470 Davidson A, Wang S, Wilmshurst J (2006) Remote sensing of grassland-shrubland vegetation water content in the shortwave domain. Int J Appl Earth Obs Geoinf 8:225–236 Eitel JUH, Gessler PE, Smith AMS, Robberecht R (2006) Suitability of existing and novel spectral indices to remotely detect water stress in Populus spp. For Ecol Manag 229:170–182 Gao BC (1996) NDWI—a normalized difference water index for remote sensing of vegetation liquid water from space. Remote Sens Environ 58:257–266 Hosgood B, Jacquemoud S, Andreoli G, Verdebout A, Pedrini A, Schmuck G (1995) Leaf optical properties experiment 93 (LOPEX93) report EUR-16095-EN. European Commission, Joint Research Centre, Institute for Remote Sensing Applications, Ispra, Italy Jackson TJ, Chen D, Cosh M, Li F, Anderson M, Walthall C, Doriaswamy P, Hunt ER (2004) Vegetation water content
123
326 mapping using Landsat data derived normalized difference water index for corn and soybeans. Remote Sens Environ 92:475–482 Jacquemoud S, Baret F (1990) PROSPECT: a model of leaf optical properties spectra. Remote Sens Environ 34:75–91 Jacquemoud S, Ustin SL, Verdebout J, Schmuck G, Andreoli G, Hosgood B (1996) Estimating leaf biochemistry using the PROSPECT leaf optical properties model. Remote Sens Environ 56:194–202 Kuusk A (1985) The hot spot effect on a uniform vegetative cover. Sov J Remote Sens 3:645–658 Levine JS (1996) Introduction. In: Levine JS (ed) Biomass burning and global change. Remote sensing, modeling and inventory development, and biomass burning in Africa, vol 1. MIT Press, MA, pp XXXV–XLIII Pen˜uelas J, Pin˜ol J, Ogaya R, Filella I (1997) Estimation of plant water concentration by the reflectance Water Index WI (R900/ R970). Int J Remote Sens 18:2869–2875 Roberts DA, Ustin SL, Ogunjemiyo S, Greenberg J, Dobrowski SZ, Chen J, Hinckley TM (2004) Spectral and structural measures of northwest forest landscapes at leaf to landscape scales. Ecosystems 7:545–562 Seelig HD, Hoehn A, Stodieck LS, Klaus DM, Adams WW III, Emery WJ (2008) Relations of remote sensing leaf water indices to leaf water thickness in cowpea, bean, and sugarbeet plants. Remote Sens Environ 112:445–455
123
J Plant Res (2009) 122:317–326 Sims DA, Gammon JA (2003) Estimation of vegetation liquid 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 HC, Breshears DD, Ustina SL, Kefauvera SC (2005) Spectral sensing of foliar water conditions in two co-occurring conifer species: Pinus edulis and Juniperus monosperma. Remote Sens Environ 96:108–118 Verbesselt J, Somers B, Lhermitte S, Jonckheere I, Aardt JV, Coppin P (2007) Monitoring herbaceous fuel moisture content with SPOT VEGETATION time-series for fire risk prediction in savanna ecosystems. Remote Sens Environ 108:357–368 Verhoef W (1984) Light scattering by leaf layers with application to canopy reflectance modeling: the SAIL model. Remote Sens Environ 16:125–141 Yebra M, Chuvieco E, Rian˜o D (2008) Estimation of live fuel moisture content from MODIS images for fire risk assessment. Agric For Meteorol 148(4):523–536 Yilmaz MT, Hunt ER Jr, Jackson TJ (2008) Remote sensing of vegetation water content from equivalent water thickness using satellite imagery. Remote Sens Environ 112(5):2514–2522 Zarco-Tejada PJ, Rueda CA, Ustin SL (2003) Water content estimation in vegetation with MODIS reflectance data and model inversion methods. Remote Sens Environ 85:109–124