Optimizing spectral information for sensing vegetation properties


One of my research here is about sensing vegetation for image classification. Today, many researcher established formula to optimizing spectral information for this matter. At this moment I will give short formula about basic vegetation indices, normalized indices, and multi-channel indices.

Note that N is for Near Infrared, R is for Red, G is for Green, and B is for Blue. VI represents for Vegetation Index, L, a, b, Beta are constants.

  Formula Advantages/disadvantage Reference
Simple indices      
DVI (Difference VI) N-R Sensitive to illumination condition, slope, etc Tucker (1979)
RVI (Ratio VI) N/R Partially corrects for variations in reflectance,  illumination, especially if using reflectance Birth and McVey (1968)
CI590 (Chlorophyll index) (N880/VIS590) – 1 Apparently more sensitive to canopy N status than is NDVI Gitelson and Merzlyak (1997)
Normalized indices      
NDVI (Normalized Difference VI) (N-R)/(N+R) Good for estimation of LAI (Leaf Area Index); Clouds water and snow tend to negative values (whenR>NIR) Rouse (1974)
GNDVI (Green NDVI) (N-G)/(N+G) Apparently better at higher LAI; particularly good at detecting chlorophyll as it increase over a much wider range of Chl than does NDVI, though the wavelength used varies from 470nm up to amber (580nm) Gitelson (1996)
SAVI (Soil-Adjusted VI) (1+L) (N-R)/(N+R+L) Corrects for varying soil reflectances, where L is a coefficient that varies from 0 at high LAI to 0 at low LAI (often assumed = 0.5) Huete (1988); modifications, e.g. TSAVI/OSAVI Steven (1998)
TSAVI (Transformed SAVI) a(N-aR-b)/(aN+R-ab+X(1+a2)a and b are the constants in the soil line equation: N=a.R+b Corresponds well with LAI Barret (1989)
TVI (Transformed VI) 100x((N-R)/(N+R)+0.5)0.5 Remove negative values, square root stabilizes variance Deering (1975)
PVI (Perpendicular VI) (N-aR-b)/(a2 + 1) Most effective at removing soil effect with low LAI Richardson and Wiegand (1977)
ARVI (Atmospherically Resistant VI) (N-RB)/(N+RB)where RB = R – Beta(B-R) Atmospherically resistant VI, corrects for changes in atmospheric transmission (B corrects R according to differences between R and B) Kaufmann and Tanre in Huete (1997)
SARVI (Soil and Atmospherically Resistant VI) (N-RB)(1+L)/(N+RB+L)where RB = R – B(B-R) Combines ARVI with SAVI (the constant B is normally 1 but can be varied to correct for aerosol (e.g. 0.5 for Sahel dust) Kaufmann and Tanre in Huete (1997)
EVI (Enhanced VI) 2.5(N-R)/(1+N6R-7.5/B) Based on SARVI, and used as the operational index for MODIS products where the toa reflectances are atmospherically corrected Huete (1997)
AFVI (Aerosol-Free VI) (N-0.5p2.1)/N+0.5p2.1) Insentive to aerosols because the mid-IR (e.g. at 2.1 um) is transparent to most aerosols except dust, but surface have similar reflectance to the visible (the factor 0.5 corrects for differences in p at 0.645 um and 2.1 um) Karnieli (2001)
WDRVI (Wide Dynamic Range VI) (alfa*N-R/(alfa*N+R)where 0.1 < alfa < 0.2 Reported to be more sensitive to high LAI then the standart NDVI Gitelson (2004)
Multi-channel indices      
Kauth-Thomas transformation (TCT – Tasseled Cap Transformation) For coefficients for Landsat 71 Derives composite channels related to ‘brightness’, ‘greenness’, and ‘wetness’ Kauth and Thomas (1976); Christ and Cicone (1984)
CAI (Cellulose Absorption Index) 0.5*(p2000-p2200)/p2100 Reported to respond especially to plant dry matter Daughtry (2000)
MTCI (MERIS Terrestrial Chlorophyll Index) (p753.75-p708.75)/(p708.75-p661.25) The wave bands given arethe centres of bands 10, 9, and 8 on MERIS Curran and Dash (2005)
PPSG (Principal Polar Spectral Index) tan-1((PC2-SF2)/(PC1-SF1)) PC1 and PC2 are the values on the 1st and 2nd principal componentsaxes and SF1 and SF2 are the foci where lines of equal vegetation cover converge Moffiet (2010)

1do not  be hesitate to ask me by email about the coefficients for Landsat MSS, 7ETM+, and for QuickBird high resolution imagery

Based on my experinced, the most easiest software to use or to apply the mathematical expression is in ILWIS 3.7.1 environment, it is opensource imagery processing software. You can easely download from this site. Eventhough, ENVI 4.7 and ERDAS Imagine also gives  these utilities in the friendly user interface. But note, it is NOT free of charge to work with these type of quite powerfull software for satellite imagery processing.

Figure below showed us result from WorldView-2 transformation to ‘brightness’, ‘greenness’, and ‘wetness’ of Depok City with small allocation of vegetation covered. Software used is ENVI 4.7. This figure show the spatial distribution of stress level vegetation sorrounding built-up area. Brightness is primary axis calculated as the weighted sum of reflectances of all spectral bands. Greenness is perpendicular to the axis of the Brightness component that passes through the point of maturity of all plants. Wetness is perpendicular to both Greenness and Brightness axis representing senesced vegetation.

TCT Transformation of WorldView-2 Satellite imagery


Jones, H.G., Vaughan, R.A., 2010. Remote sensing of vegetation. Principle, techniques, and applications. Oxford University Press. New York, United States.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s