1
>
nt,
Landsat
getation
sment of
3K of soil
g of the
and for
p a GIS
ting the
provide
ion, risk
nition of
ARF, in
, based
in the
1 in the
Ji, and
s, Were
s have
within a
s within
> (Pinus
ranging
different
selected
it, soil,
use in
ng data
3. VEGETATION MONITORING
Monitoring the vegetation cover and the regeneration
rate at burned sites is of major importance for land
resource management. In this study an approach is
suggested that uses multitemporal TM imagery. In a first
step, vegetation cover is determined from the near -
infrared and red spectral region of the imagery by
applying a vegetation index. The vegetation cover
estimates for different dates are used in a second step to
determine regeneration rate.
3.1. Vegetation Cover
A dense vegetation cover at a burnt site has normally re-
established itself within two to five years of a fire.
However, the process of regeneration can be impeded by
many factors, such as animal grazing, a paucity of
precipitation, or soil erosion.
Aerial photographs are commonly used to obtain an
estimate of a region's vegetation cover, but there also
exists a strong correlation between vegetation cover and
satellite-based vegetation indices that can provide this
information over much larger areas. One such index is
the Modified Soil Adjusted Vegetation Index (MSAVI) of
Qi et al. (1994), which is given by
2* D uim +1-J@ Por + 1)” =8( Pair = Prep)
2
MSAVI =
where p,, = Reflectance in the NIR spectral region
Po = Reflectance in the red spectral region
For quantitative estimates of vegetation cover, field sites
without any vegetation cover and forested sites with “full
cover were used in the analysis. The selection of
appropriate sites is critical for the accuracy of the
estimates. For non-vegetated areas, sites with bare soil
and rocks were selected. No burnt sites were included.
Aleppo pine forests were used as a reference for "full"
vegetated sites. A comparison of the MSAVI values with
ground-acquired vegetation cover data for the 39
reference sites shows a linear relationship. Therefore, the
MSAVI values were linearly stretched between the mean
values for the non-vegetated and the forested sites.
As an independent quality control, 14 validation sites in
addition to the 39 reference sites were used. A
Comparison between ground cover estimates for this 14
validation sites and their corresponding MSAVI-obtained
estimates is given in Table 1.
241
Table 1: Vegetation Cover for the Validation Sites
Test Site Ground | Estimate | Residual
Area truth (96) (96) (96)
Lavrio 1 85 82 3
2 90 75 15
Pateras 1 75 74 1
2 85 81 4
3 70 86 16
4 90 72 17
5 30 43 19
6 30 38 8
Pendeli-1 1 65 74 9
2 70 92 22
3 40 58 18
Pendeli-2 1 95 80 15
Varnavas 1 95 95 0
2 100 91 8
The results give an overall RMS-Error of +13 %.
3.2. Monitoring regeneration rates
For quantitative and qualitative evaluation of the degree
and extent of regeneration, the difference in vegetation
cover between two image acquisition dates is used.
A prerequisite for this approach is a high precision
overlay of the image data sets. At a minimum, the
different image data sets have to be registered to one
"reference" image. If coordinate information for further
analysis or mapping is required, georeferencing of the
images is necessary. Almer et al. (1991) give an estimate
of the geometric location accuracy as a function of terrain
height and imaging geometry for affine and parametric
geocoding. In this study parametric geocoding was
performed for the image data sets, incorporating a digital
elevation model. By using the RSG (Remote Sensing
Software Package Graz) developed at our institute, we
achieved absolute geometric accuracies better than + 1
pixel for all image data sets. However, for many
monitoring applications, an affine transformation without
a digital elevation model can be sufficient and very cost
effective, especially for monitoring large regions.
By incorporating imagery from several acquisition dates
throughout the monitoring period, the dynamics of the
vegetation development can be studied in detail. In this
study TM imagery acquired in 1984, 1987, 1990 and
1993 were incorporated. Figure 1 shows the dramatic
developments in a part of the Pendeli test area, north-
east of Athens.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
