 |
 |
 |
 |
 |
 |
 |
 |
Activiities of ASTER Working Groups |
|
 |
|
 |
 |
|
 |
 |
Chiba Univ.
Takashi Ishiyama |
|
In recent years it became possible to utilize the Normalized Data Set of Vegetation Index (NDVI) obtained by calculating observation data of the visible and near infrared region of NOAA AVHRR or LANDSAT TM. Based on whether or not vegetation exists, as identified from the NDVI, application of this as a method to identify desert and semi-arid areas has been tried as a means to monitor historical changes in the "front line to desert conditions". However, it was found that the NDVI, which is widely utilized as an algorithm to monitor vegetation, also had a problem.
In arid areas such as dry land where there is less vegetation, errors in the Vegetation Index become greater due to the effect of radiation from the background soil. The observed Vegetation Index is reduced in particular in cases where the background soil is red. In order to compensate for this defect in the algorithm, a SAVI that was less affected by reflection from soil was developed. However, on the other hand, this Vegetation Index causes errors when the vegetation density is high.
The author has therefore proposed an Optimized Vegetation Index (OPVI) that is a hybrid algorithm to compensate for both defects, utilizing the shortwave infrared band of ASTER data. This calculates first the ratio of shortwave infrared bands for each pixel. For example, the result is band 5/band 7 for LANDSAT TM and band 4/band 5 for ASTER. A ratio of the two correlates very well to the vegetation density. The Vegetation Index for each pixel of satellite data is calculated from the obtained reflection ratio, where values larger than the threshold value already set are calculated by NDVI, and smaller values are calculated by SAVI.
In order to verify the OPVI, a TM data analysis was conducted covering the area surrounding the oasis in dry land. The targeted area was the area surrounding the Khotan oasis to the south of the Takla Makan desert. This area contains the inner oasis area where a relatively high vegetation density exists, the surrounding area of the oasis where, on the contrary, a low vegetation density exists and desert where vegetation rarely exists.
Figure 1 shows the Vegetation Index Distribution Map of the covered area for the research using NDVI, SAVI and OPVI. It can be seen that almost all the pixels are distributed in NDVI, and that the vegetation index is high. A small part of SAVI is distributed in the bare land within the oasis or in the region where the land is partially changing to desert. In fact, this is also identified by the site survey. Thus, in the area where the density range of the vegetation is large because both agricultural and bare land exists in the area, an evaluation based on the conventional NDVI or SADI only causes errors.
The OPVI evaluation proposed in this research is valuable in this case. If we can obtain in future high-resolution data from ASTER that covers the shortwave infrared band, more detailed monitoring utilizing OPVI would be expected to observe changes to desert conditions.
Please refer to the references listed below for detailed information on SAVI and OPVI.
Huete,A.R(1988): ASoi1-AdjustedVegetation Index (SAVI), Remote Sensing of Enviroment.25:295〜309. Ishiyama,T"et-aL(1996): Vegetation Index A1gorithm for Vegetation Monitoringin Arid and SemiArid Land.Jouma1 of Arid Land Studies,6-1,35-47. |
|
Figure 1
False color images of the area surrounding Khotan oasis, Tien Shan, PRC and Vegetation Index Distribution by NDVI, SAVI
|
 |
False color images
Grey part: Outer circumference of oasis and desert where the vegetation density is low
Red part: Oasis and surrounding area where the vegetation density is relatively high |
|
|
(Eco/Oceanography and Limnology WG ) |
|
 |
|
