2. REMOTE SENSING APPLICATIONS IN PRECISION
2.2 Evolution of Remote Sensing technologies and their application to agriculture
2.2.7 Selected Vegetation Indices and their use in mapping crop parameters
The concept of using combinations of red and near infrared measurements to estimate biophysical parameters of vegetation was first introduced by Jordan (1969) who used a simple ratio of the canopy transmittance to derive leaf area index.
Table 2.3 gives an overview of selected vegetation indices. These are the most commonly used vegetation indices in the literature, but since manifold empirically derived vegetation indices exist, it is not complete.
Table 2.3: Summary of selected Vegetation Indices
(modified from Sandison, 1999, Elvidge and Chen, 1995)
Index and Source Name Formula
RVI Jordan (1969) Ratio Vegetation Index NDVI Rouse et al. (1973) Normalised Difference Vegetation Index
See (a) below
NDVI = (NIR-R) / (NIR+R)
DVI Tucker (1979)
Difference Vegetation Index
DVI = NIR - RED
SAVI Huete (1988) Soil Adjusted Vegetation Index * See (b) below TSAVI Baret et al. (1989) Transformed Soil Adjusted Vegetation Index ** SAVI2 Major et al. (1990)
Soil Adjusted Ratio Vegetation Index ** WDVI Clevers (1988) Weighted Difference Vegetation Index IPVI Crippen (1990) Infra-red Percentage
Vegetation Index IPVI = NIR / (NIR + R)
PVI Richardson and Weigard (1977) Perpendicular Vegetation Index ** ARVI Qi et al. (1994) Atmospherically Resistant Vegetation Index
MSAVI2 Qi et al. (1994)
Modified Soil Adjusted Vegetation Index Two
TVI
Broge and Leblanc (2001) Triangular Vegetation Index See (c ) below TVI = 0.5[120(R750-R550)-200(R670- R550)]
* L is a soil adjustment factor (in SAVI, it ranges from 0 to 1 and is normally used at .5) ** a and b are rock soil baseline from NIR vs. RED
(a) Normalised Difference Vegetation Index (NDVI)
Perhaps the best known of the vegetation indices is the normalized difference vegetation index (NDVI; Rouse et al. 1974). The NDVI normalizes the difference of the red (R) and near-infrared (NIR) band combination and is therefore a relative measure within the image data (versus absolute measure by single, calibrated bands). The NDVI is often (wrongly) used with the DN values of the red and near infrared band, and not with calibrated reflectance values (Paris, 1998).
NDVI = (NIR-R) / (NIR+R); this index has a range of -1 to +1.
The NDVI is commonly used in multi-temporal mapping of vegetation dynamics based on maximum-NDVI composites (Townshend et al., 1985; Holben, 1986; Gutman, 1989; Wiegand et al., 1991; Viovy et al., 1992; Loudjani et al., 1994), in particular on continental or global scales (Townshend and Justice, 1986; Townshend et al., 1994; Smith, 1994). NDVI values can vary significantly as a function of sensor calibration (Price, 1987; Goward et al., 1991), atmospheric conditions (Deering and Eck, 1987; Singh and Saull, 1988; Kaufman and Tanré, 1992; Myneni and Asrar, 1994), directional surface reflectance effects (Kirchner et al., 1981; Holben 1986; Lee and Kaufman, 1986; Paltridge and Mitchell, 1990; Koslowsky, 1993), and terrain relief (Teillet and Staenz, 1992; Burgess and Lewis, 1994). Special attention has also been paid to soil background effects and soil indices (Richardson and Wiegand, 1977; Baret et al., 1989; Major et al., 1990; Huete and Tucker, 1991; Qi et al., 1994a,b). The NDVI, (like also the RVI), has shown to be sensitive to soil background. The problem
of soil background in vegetation indices was described by Huete et al. (1985). Huete (1988) later developed a formula to account for soils called the Soil Adjusted Vegetation Index (SAVI):
(b) Soil adjusted vegetation Index (SAVI)
SAVI= (1+L)* ((NIR-R)/(NIR+R+L))
The SAVI uses a variable L depending on the amount of vegetation. Huete found that there might be two or three optimal L values for analysing very low vegetation (L = 1), intermediate vegetation (L = 0.5), or higher densities (L = 0.25). However, the adjustment of L = 0.5 offered a spectral index superior to the NDVI for the entire range of vegetation conditions studied (Huete, 1988). Bausch (1993) later tested the SAVI extensively and, like Huete, found SAVI to be more accurate than NDVI. Bausch reported that SAVI was (1) sensitive to a leaf area index (LAI) higher than 3, (2) was excellent in correcting a wet soil surface, and (3) minimized soil background throughout the growing season. Bausch concluded that using the adjustment factor L set at 0.5, the SAVI (1) minimized soil background effects throughout the entire growing season, (2) was independent of planting and effective cover dates, (3) was sensitive to slow and fast plant growth induced by weather anomalies and nutrient deficiencies, and (4) responded to leaf loss caused by hail and by various forms of plant stress induced by insects, disease, and water deficit (Wright et al., 2000).
(c) Broge and Leblanc (2001) developed the Triangular Vegetation Index (TVI), which is meant to characterize the radiant energy absorbed by leaf pigments in terms of the relative difference between red and near-infrared reflectance in conjunction with the magnitude of reflectance in the green region. TVI is determined as the area defined by the green peak, the near-infrared shoulder, and the minimum reflectance in the red region. It is formulated as:
TVI = 0.5*[120*(Reflectance750-Reflectance550) - 200*(Reflectance670-Reflectance550)]
The general idea behind TVI is based on the fact that the total area of the triangle (green, red, near infrared) will increase as a result of chlorophyll absorption (decrease
of red reflectance) and leaf tissue abundance (increase of near-infrared reflectance) (Broge and Leblanc, 2001).
The theoretical foundation of vegetation indices has been well examined (Asrar et al., 1989; Baret and Guyot, 1991; Myneni et al.,1995a,b; Qi, 2001). Vegetation indices are affected by plant and measurement conditions, therefore field validation studies for various plant species, locations, and environmental conditions are needed to derive useful, robust semi-empirical relations. An overview of numerous studies relating spectral vegetation indices empirically by ground measurements to vegetation properties follows.
¾ various above-ground biomass measures (Pearson and Miller, 1972; Kauth and Thomas, 1976; Richardson and Wiegand, 1977; Tucker, 1979; Elvidge and Lyon, 1985; Price, 1992; Steven, 1998; Jago et al., 1999)
¾ fPAR (fraction of Absorbed Photosynthethical Active Radiation) (Asrar et al., 1984; Hatfield et al, 1984; Sellers, 1985, 1987, 1989; Choudhury, 1987; Baret and Guyot, 1991; Goward and Huemmerich, 1992; Myneni and Williams, 1994; Chen, 1996; Inoue et al., 2001)
¾ leaf area index (LAI ) (Holben et al., 1980; Asrar et al, 1984, 1985b; Hatfield et al, 1985; Badhwar et al., 1986; Clevers, 1988, 1989; Spanner et al., 1990; Baret and Guyot, 1991; Chen, 1996)
¾ crop moisture variations (Peñuelas et al., 1995; Russ 1993)
¾ leaf pigment concentrations and chlorophyll levels (Blackburn, 1998a; Blackburn and Steele, 1999; Miller et al., 2002; Jago et al., 1999) ¾ carbon dioxide (Tucker et al., 1986; Cihlar et al., 1992)
¾ biophysical plant canopy properties (Pinty et al., 1993)
¾ assessment of crop or vegetation stress (Blackburn, 1998b; Dawson et al., 1998)
¾ detection of crop phenology (Badhwar and Henderson, 1981) ¾ crop type or species identification (Asner et al., 2000)
¾ land cover characterization (Goetz et al., 1985; Friedl et al., 1994; Lyon et al., 1998; Thenkabail et al., 1999; Thenkabail et. al., 2000)
¾ assessment of carbon fluxes (Fassnacht et al., 1997)
¾ yield (Aase, 1979; Tucker et al., 1985; Bartholome, 1988; Rudorff and Bastista, 1991; Wiegand et al., 1991; Quarmby et al.,1993; Maselli et al., 1993; Cabezon and Taylor, 1994; Smith, 1994; Smith et al.,1995; Murthy et al., 1996; Hamar et al., 1996; Rasmussen, 1996 and 1998; Clevers, 1997; Hayes and Decker,1996 and many more -for further examples see references in Genovese, 1994; Rasmussen, 1998; Moulin et al., 1998).