Geomorphologic Distribution of Normalized Difference Vegetation Index
Mitigating the Topographic Effect
Scheneider and Robbins (1995) summarized methods to mitigates topographic effect: 1. Band rationing 2. Image Partitioning, 3. Direct illumination effect using a DTM and 4. Scattering Effects and non-Lambertain assumption. Band rationing similar the NDVI values used in this study. That can efficiency reduce the topographic effect and does not produce artificial variance in the output image. However, Some high relief areas and deeply shaded areas will be unimproved through band rationing.
In order to explore the NDVI distribution without the bias from scattering and diffuse light from inclined surface, images were processed using backwards radiance Correction transformation (BRCT) with Non-Lambertain assumption (Colby, 1991):
L * cos(e) = Ln * cosk (v)* cosk (e)
Where L = reflectance radiance, v = incidence angle, e = slope of the plane, Ln = radiance when v = e-0, k is the Minnaert constant. v can be calculated by Lambertain assumption:
Cos(v) = cos(z) cos(e)+ sin(z) sin(e) cos(|d|),
where z is solar zenith angle and d is the angle between surface aspect and the solar azimuth angle. Minnaert constant k can be estimated from a regression expression which makes logarithm of he original equation: Y = a + k* X, where Y = Log[L * cos(e))], a = Log (Ln) and X = Log[cos(i)*cos(e)]. Original and correct NIR images of Fusan are shown in Fig. 2.
Figure 2: Near Infrared (Band 4) of Fusan Basin
Data Analysis
Influence of Aspect: Comparing with the data of four basins, aspect shows a strong influence in controlling NDVI distribution (Fig. 3). The curves agree with Leprieur and Durand's study (1988) in which the maximum NDVI value happen in the solar orientation (about 140 degree) and a minimum opposite the sun.
Figure 3: Didtribution of NDVI against the aspect
The trends of NDVI against slope and elevation do not show an obvious tendency. It may be induced by the strong influence of aspect. In order to eliminate the orientation effect, data were divided to illuminate and non-illuminate groups for further investigation. Illuminated group includes pixels located on aspect orientation between 50 to 230 degree, perpendicular to 140 degree in where the maximum NDVI occurred.
Influence of Slope : In the illuminate side, NDVI has a smallest value in flat area and then rapidly increase in average value (0.3-0.5) (Fig. 4). The curve gradually change to a maximum value at about 40 degree. Above the point, different basin show diverse trends which may caused by the relatively small amount of data sample and local illumination. In the dark side, NDVI curve are still similar in all four basins but divides into tow branches in slope sleeper than 40 degree (Fig. 5).
Figure 4: NDVI of illuminate aspect against slope
Figure 5: NDVI of non-illuminate aspect against slope
Change of NDVI against slope seems reasonable because of the solar radiance, solar types, and soil moisture and function of slope. Flat areas in valley bottom are always moist and shadowy, steep slope receive more solar energy but relative dry. Such conditions are harmful for vegetation growing.
Influence of Elevation (Fig. 6 and Fig. 7): the pixels in low altitude always has low NDVI and rapidly rise to average values (0.3-0.5). Patterns are same in mountain basin or flat ground, and no difference in bright or dark orientation. It may comes from the reason as same as in flat slope area: pixels in relative low-altitude or flat-slope are usually stream outlet or valley bottom. These area have sufficient moisture but lack of solar radiation, which cause a low biomass and NDVI.
Figure 6: NDVI of illuminate aspect against elevation
Figure 7: NDVI of non-illuminate aspect against elevation
The NDVI curves slowly decrease with elevation. Above the altitude of 3,000 meters, the trend rapidly decrease. IT shows that cooler temperature in higher elevation area turn down biomass, and the effect is much significant in elevation higher than 3000 meters.
