Remote Sensing of Leaf, Canopy, and Vegetation Water Contents for Satellite Environmental Data Records - Satellite-based Applications on Climate Change - page 333

tested a Leaf Water Stress Index (LWSI, Table 20.2 ), which was derived from the
Beer-Lambert law to be equal to leaf RWC. However, Hunt and Rock ( 1989 ) and
Cohen ( 1991a , b ) showed that LWSI was not practical, because of the requirements
to measure reflectances of fully turgid and dry leaves. Furthermore, Hunt and Rock
( 1989 ) showed there is a log-linear relationship between the Moisture Stress Index
(MSI, Table 20.2 ) with LWC for different species from crops to desert succulents.
Because there is a one-to-one relationship between a ratio index and the cor-
responding normalized difference index, there is a strong relationship between
NDII and CWC (Yilmaz et al. 2008a , b ).
NDII is also an important index for the remote sensing of snow cover and
flooded areas. NDII time series in the northern latitudes show both snow and
vegetation dynamics, so estimating the start of spring for phenological studies
may be difficult (Xiao et al. 2002b ; Delbart et al. 2005 , 2006 ). Xiao et al. ( 2002a )
used NDII to detect flooded rice paddy fields in China.
The design for MODIS defined a band 5 located at 1240 nm (Fig. 20.2 ), thus Gao
( 1996 ) defined the Normalized Difference Water Index (NDWI; Table 20.2 ). Similar
to other NIR wavelengths, radiation at MODIS band 5 is highly scattered by leaf
cellular structure and multiple leaves, so the two bands of NDWI sample the same
amount of canopy (Gao 1996 ). Furthermore, NDWI does not saturate at low LAI as
does NDVI or NDII. Zarco-Tejada et al. ( 2003 ) recommended the Simple Ratio
Water Index (SRWI, Table 20.2 ) which also uses the apparent reflectance at 1240 nm.
The water absorption feature at 970 nm is interesting for remote sensing because
reflectances at this wavelength may be measured using low-cost silicon detectors.
Pe˜uelas et al. ( 1993 , 1997 ) designed the Water Index (WI, Table 20.2 ) and com-
pared it to FMC and measures of plant water stress. To date, WI can only be measured
over larger areas with airborne or satellite imaging spectrometers. There are no current
or planned satellite multispectral sensors with a 970-nm band for calculating WI.
The variety of foliar water indices based on the wavelengths of 970, 1240, and
1650 nm indicates there may not be a single index that outperforms the others for
estimating CWC under all conditions. This situation should be expected simply
from the history of red-NIR indices, such as NDVI and the simple ratio of R 850 / R 680 .
One of the first alternative red-NIR indices was the Soil-Adjusted Vegetation Index
(SAVI, Huete 1988 ), which corrected for numerical offsets between the red and
NIR bands at low LAI. Using the method of Huete ( 1988 ), Cecatto et al. ( 2002a , b )
developed the Global Vegetation Moisture Index (GVMI, Table 20.2 ) to adjust the
NIR and SWIR reflectances so zero vegetation would have zero GVMI (also see
Dasgupta and Qu 2009 ).
New paradigms may be needed for development of effective vegetation water
indices. The Maximum Difference Water Index uses maximum reflectance and
minimum reflectance between 1500 and 1750 nm to estimate the depth of this water
absorption feature (Eitel et al. 2006 ), which were assumed to be at 1500 and
1650 nm in Table 20.2 . Khanna et al. ( 2007 ) created the Shortwave Angle Slope
Index (SASI, Table 20.2 ), which is calculated from the angle between the three
points: (850 nm, R 850 ), (1650 nm, R 1650 ), and (2200 nm, R 2200 ). Ghulam et al.
( 2008 ) created a statistical approach based on a scatter plot of R 850 versus R 1650
called the Vegetation Water Stress Index (VWSI, Table 20.2 ). A trapezoid is