Elements of Change 1995: Session I Huete

The role and importance of canopy background signals in the global assessment and monitoring of vegetation from satellites are critically examined by Huete. The coupling of soils and vegetation occurs in both a "biome" sense as well as in a radiometric sense. Variations in soil background occur at all spatial scales, including continental variations of dark, organic-rich grassland soils, reddish iron-rich forest soils, and highly variable, geologically exposed arid region soils. Temporally, canopy background signals vary with wetting, presence of snow, and the dynamic behavior of leaf fall and decomposition. Since soils and vegetation co-vary in nature, it is not a simple matter to separate and isolate soil influences on canopy spectral response. However, with greater disturbances and land use changes, patterns of asynchronous soil and vegetation development emerge and it becomes vital to remove radiometric soil influences in operational vegetation monitoring programs.

Operational monitoring of the Earth's vegetative cover currently involves the utilization of vegetation indices (VIs) as a precise radiometric measure of spatial and temporal patterns of vegetation photosynthetic activity. Only with such precise consistency can one utilize VIs for change detection. The normalized difference vegetation index (NDVI), which is the difference of the near-infrared and red bands divided by their sum, has been the most widely used index in global vegetation studies. Its success and accuracy, however, depends on how well it is able to depict actual vegetation differences amidst widely varying soil, atmosphere, and sun-target-sensor variations. Another use of the NDVI is to derive canopy biophysical parameters such as fraction of absorbed photosynthetically active radiation (fPAR), leaf area index (LAI), and % green cover. Many studies have concluded that the translation from NDVI to biophysical parameter is biome dependent and requires knowledge of biome type.

With greater disturbances and land use changes, patterns of asynchronous soil and vegetation development emerge and it becomes vital to remove radiometric soil influences in operational vegetation monitoring programs.

Two distinct forms of soil influences on vegetation indices (VIs) can be examined. The first concerns the spectral properties of bare soils and background surfaces. Since these signals represent non-photosynthetic elements of a canopy, they form a "baseline" against which the "green" vegetation signal is measured. There are two spectral effects on the NDVI, one due to "brightness" variations and the second associated with "color" variations. Brightness variations plot along a "soil line" in NIR-red space, whereas color differences create secondary sources of variation plotting away from the soil line. The NDVI is susceptible to both forms of variation and cause the NDVI to vary from 0 to 0.2 units on a scale of 0 to 1.

The second form of soil influence concerns soil-plant mixtures and radiant transfer processes within the canopy. Two cases of radiative mixing of soils and plants can be compared. In the first case, the leaf elements of a canopy are assumed opaque resulting in a spectrally-independent, linear mixture model. In the second case, the leaves possess transmissive properties (particularly in the near-infrared) and one can utilize Beer's law to model radiant transfer through the canopy. Figure 4.1 shows the high extent of soil back ground-induced variability in the NDVI for both cases. At zero percent cover, one can see the NDVI vary from 0 (snow) to 0.2 (organic, black soil). At intermediate levels of green cover, however, the soil noise problem increases and becomes maximum at 30-40% green cover, before decreasing to zero at 100% green cover. The same results are seen with more complex canopy models (SAIL, Myneni), with both observational ground and satellite data sets.

The soil adjusted vegetation index (SAVI) utilizes a constant "L" to remove the soil back ground noise (Figure 4.2 ). The "L" factor is related to the differential extinction properties between the red and NIR (Huete, 1988). As long as the canopy is actively photosynthesizing, red extinction through a canopy will exceed that of the NIR and a soil correction is necessary. One of the benefits of soil correction is to make the VI more linear with biophysical plant parameters. This benefits vegetation studies by minimizing saturation problems at high levels of vegetation, and allows for more accurate aggregation or scaling of multi-resolution data sets.

The removal of one source of noise will require consideration of the additional influences. ... New methods have been developed to render the VI inherently less sensitive to atmospheric variations.

Numerous ground, air, and simulation studies over a wide variety of canopies have demonstrated large "potential" influences of soil background, atmosphere, canopy architecture, solar zenith, and view zenith angles on the NDVI response. These influences, however, are intricately coupled as, e. g., the proportions of soil and plant viewed or "remotely sensed" varies with sun-target-sensor geometry. The NDVI has benefited greatly in having many of these influences cancel out. For example, in the presence of an atmosphere, soil influences on the NDVI start to become smaller and are nearly removed in turbid at mospheric conditions. The NDVI function is also more linear with LAI in the presence of an atmosphere. The presence of an atmosphere also offsets canopy anisotropic behavior resulting in less variable NDVI differences with satellite viewing angles. Thus, atmospheric correction of satellite data sets will aggravate soil background and sun-view angular problems in the NDVI. The removal of one source of noise will require consideration of the additional influences.

New methods have been developed to render the VI inherently less sensitive to atmospheric variations. These involve incorporation of the blue band, which utilizes atmospheric scattering of blue light to adjust scattering in the "red" band. Kaufman and Tanré (1992) developed the atmospherically resistant vegetation index (ARVI) by including the blue band in an NDVI-like equation. Liu and Huete (1995) incorporated both soil adjustment and atmospheric resistance concepts into a single index, the modified NDVI (MNDVI) (Figure 4.3 ). In Figure 4.3 we see the progressive decline in soil and atmospheric-induced noise with improved VIs which incorporate soil and atmospheric adjustment factors. The MNDVI has been shown to remove smoke plumes and cirrus clouds from Landsat TM imagery, an indication of its ability to minimize atmospheric aerosols on a pixel by pixel basis.

Improved indices will compliment the NDVI in global vegetation monitoring studies in the Earth Observation System (EOS) era to begin in 1998 with the launching of the MODIS sensor (Running et al., 1994). The NDVI will serve as a "continuity" index in order to exploit a 15+ year AVHRR-NDVI global data set. The improved indices, which represent state of the art research, will refine vegetation change detection by physically incorporating soil, atmosphere, and bidirectional effects. Furthermore, there is some evidence that the new indices, such as SAVI and MNDVI are more related to LAI, whereas the NDVI is more correlated to fPAR. This will enable improved methods in extraction of canopy biophysical parameters for global change and land use studies.

The improved indices, which represent state of the art research, will refine vegetation change detection by physically incorporating soil, atmosphere, and bidirectional effects.

Running, S. W., Justice, C., Salomonson, V., Hall, D., Barker, J., Kaufman, Y., Strahler, A., Huete, A., Muller, J. P., Vanderbiblt, V., Wan, Z. M., Teillet, P., and Carneggie, D., 1994, Terrestrial remote sensing science and algorithms planned for EOS/MODIS, Int. J. Remote Sensing, 15:3587-3620.

Huete, A. R., 1988, A soil adjusted vegetation index (SAVI), Remote Sens. Environ. 25:295-309.

Kaufman, Y. J. and Tanre, D., 1992, Atmospherically resistant vegetation index (ARVI) for EOS-MODIS, IEEE Trans. Geosci. Remote Sensing, 30:261-270.

Liu, H. Q., and Huete, A. R., 1995, A feedback based modification of the NDVI to minimize canopy background and atmospheric noise, IEEE Trans. Geosci. Remote Sensing , 33:457-465.