Vegetation Indices

The spectral properties of vegetation canopies, mentioned above, make it possible to monitor vegetation dynamics and their spatial and temporal variability using various remote-sensing platforms. Satellite data-based detection of vegetation health and stress depends on the strong relationship between simple transforms of reflected red and near-infrared energy and the intercepted or absorbed photosynthetically active radiation (APAR) of the plant canopy. Several such transforms, referred to as vegetation indices (VIs), are based on the unique spectral signature of green vegetation in the red and NIR portions of the spectrum and form the basis for quantitative assessment of vegetation condition using satellite data. According to Jackson and Huete (1991), VIs can be divided into two groups: slope-based and distance-based VIs.

Slope-Based Vegetation Indices

The slope-based VIs are simple arithmetic combinations that exploit the contrast between the spectral response patterns of vegetation in the red and NIR portions of the electromagnetic spectrum. Some of these vegetation indices include the ratio vegetation index (RVI), normalized difference

Spectral Reflectance Examples
Figure 5.3 Examples of spectral reflectance of soil and vegetation at different wavelengths. Note the high spectral contrasts between green vegetation and soil in the red and near infrared spectral regions (from Tucker and Sellers, 1986).

vegetation index (NDVI), and transformed vegetation index (TVI). The RVI, proposed by Rouse et al. (1974) using Landsat multispectral scanner (MSS) imagery, is a simple division of the reflectance values in the NIR band by those in the red band.

Normalized Difference Vegetation Index The NDVI was also proposed by Rouse et al. (1974) as a spectral VI that isolates green vegetation from its background using Landsat MSS digital data. It is expressed as the difference between the NIR and red (RED) bands normalized by their sum:

The NDVI is the most commonly used VI because it has a desirable measurement scale ranging from —1 to 1 with zero as an approximate value of no vegetation. Negative values represent nonvegetated surfaces, whereas values close to 1 have very dense vegetation. The NDVI and RVI have the ability to reduce external noise factors such as topographic effects and sun-angle variations.

Transformed Vegetation Index The TVI, proposed by Deering et al. (1975), modifies the NDVI by taking its square root and adding a constant of 0.50. This creates a VI scale consisting of mainly positive values approximating a more normal distribution. There are no theoretical differences between the NDVI and TVI in terms of quantitative values or

0.50

0.40

0.20

0.10

7 = 0.1233 +16.7980/x chlorophyll absorption

7 = 0.1233 +16.7980/x chlorophyll absorption

0.40

0.20

290 520 780 1040 1300 Total wet biomass (g/m2)

0.70 I—I—I—I—I—I—I—I—I—|— 0.75-D0.80nm r2 = 0.86

y = 0.65 - exp( - 1.2680 - 0.0017 x) no absorption •

290 520 780 1040 1300 Total wet biomass (g/m2)

0.70 I—I—I—I—I—I—I—I—I—|— 0.75-D0.80nm r2 = 0.86

y = 0.65 - exp( - 1.2680 - 0.0017 x) no absorption •

290 520 780 1040 1300 Total wet biomass (g/m2)

0 290 520 780 1040 1300 C Total wet biomass (g/m2)

Figure 5.4 The relationships between total wet biomass of grass canopy (>90 green) and (A) the red (0.63-0.69 /u,m) spectral radiance, (B) near-infrared (0.75-0.80 /u,m) spectral radiance, and (C) the normalized difference vegetation index (from Tucker, 1977).

0 290 520 780 1040 1300 C Total wet biomass (g/m2)

Figure 5.4 The relationships between total wet biomass of grass canopy (>90 green) and (A) the red (0.63-0.69 /u,m) spectral radiance, (B) near-infrared (0.75-0.80 /u,m) spectral radiance, and (C) the normalized difference vegetation index (from Tucker, 1977).

vegetation identification. In fact, all of the slope-based VIs can be shown to be functionally equivalent to each other (Perry and Lautenschlager, 1984).

Distance-Based Vegetation Indices

Perpendicular Vegetation Index The family of distance-based vegetation indices were originally derived from the perpendicular vegetation index (PVI) formulated by Richardson and Wiegand (1977). The principal objective of these VIs is to eliminate the effect of soil brightness over surfaces of incomplete vegetation cover where a mixture of green vegetation and soil background dominates the surface. This is particularly important in detecting the presence of vegetation in arid, semiarid, and subhumid environments. The procedure for deriving the PVI is based on the soil line concept, which describes the typical range of soil signatures in red/near-infrared bispectral plots. The soil line is computed by linear regression of

NIR against red band measurements for a sample of bare soil pixels. Pixels falling near the soil line are assumed to be sparsely vegetated, whereas those farther away, in the direction of increasing NIR and decreasing red, represent increasing amounts of vegetation. Soil lines may be specific to one soil type or more general to a variety of soils within an image or satellite data set. The complexity of the derivation of these distance-based indices has resulted in inconsistencies in their formulation for assessing vegetation status and condition, limiting their applications to regional studies where soil-vegetation characteristics can be clearly segregated. Improvements to the PVI have yielded three other PVIs suggested by Perry and Lautenschlager (1984), and Qi et al. (1994) and referred to as PVI1, PVI2, and PVI3.

Vegetation Indices Based on Orthogonal Transformation Vegetation indices based on orthogonal transformation include the difference vegetation index (DVI) also suggested by Richardson and Wiegand (1977), the green vegetation index (GVI) of the tasseled cap transformation (Kauth and Thomas, 1976), Misra's green vegetation index (MGVI) based on the Wheeler-Misra transformation (Wheeler et al., 1976; Misra et al., 1977), and Principal Components Analysis (PCA) (Singh and Harrison, 1985; Fung and LeDrew, 1988). These indices involve decorrelation of the original bands to extract a new set of components that separate vegetation from other surface materials.

Optimized Indices There is a class of VIs based on semiempirical radiative transfer theory that use both slope- and distance-based properties of spectral data in red and NIR plots. These indices are referred to as optimized indices and include the soil-adjusted vegetation index (SAVI) proposed by Huete (1988). The SAVI aims to minimize the effects of soil background on the vegetation signal by incorporating a soil adjustment factor into the denominator of the NDVI equation:

where L is the soil adjustment factor that takes into account first-order, differential penetration of red and NIR energy through a canopy in accordance with Beer's law. There are modified forms of the SAVI that include the transformed soil-adjusted vegetation index (TSAVI) by Baret and Guyot (1991), and the modified soil-adjusted vegetation index (MSAVI) suggested by Qi et al. (1994), based on a modification of the L factor of the SAVI. All of these modifications are intended to improve correction to the soil background brightness for different conditions of surface vegetation cover.

Optimized indices also include the atmosphere resistant vegetation index (ARVI) by Kaufman and Tanre (1992), the enhanced vegetation index (EVI) by Huete et al. (1994), and the aerosol-free vegetation index (AFRI) by Karnieli et al. (2001). These indices incorporate atmosphere and canopy radiative transfer theory for optimized retrieval of vegetation properties from satellite data (Verstraete and Pinty, 1996).

Ideally, optimized and distance-based VIs are superior to slope-based indices because they attempt to minimize or remove atmosphere and soil brightness noise that may limit a quantitative assessment of green vegetation; however, their robustness over global vegetation conditions remains to be tested. The simplicity of sloped-based indices both in terms of numerical results and interpretation has meant that such indices, such as the well-known NDVI, can be used for vegetation monitoring and drought detection at regional as well as global scales.

0 0

Post a comment