rnt
ean
iki,
ars
age
isi-
ntal
soil
cifi-
om-
li et
ster-
ased
loss
ires;
fire-
ving
nent
lied
ı the
npo-
ents.
lask-
ness,
| this
gery,
> and
esent
bina-
enta-
cover
tego-
vised
Patrono, Andrea
classification. The selection of these bands was because of their spectral characteristics in relation to vegetation monitor-
ing and detection, and the proportion of overall variance with the total data they represent. These LANDSAT TM band
widths are finely tuned for vegetation discrimination: TM band 3 matches the chlorophyll absorption range that is impor-
tant for discriminating vegetation types, TM band 4 is useful to determine vegetation types, health, and biomass content
and TM band 5, in general, offers also a good contrast between vegetation types (Lillesand and Kiefer, 1987). Natural
colour and false colour composites derived from TM bands 2, 3, 4, and 5 also provided help in the definition of the train-
ing samples.
2.3 Monitoring Vegetation Condition
The status of plant population is commonly measured or estimated in terms of plant cover (the proportion of the substrata
covered) or biomass (the total weight of the living organisms). It is particularly informative because it provides an indica-
tion of the resources produced/consumed and the amount of physiological stress present. Various mathematical combina-
tions of red (R) and near infrared (NIR) channels have been found to be good indicators of the presence and condition of
vegetation/active green biomass (Sabins, 1987). These mathematical quantities are thus referred to as vegetation indices.
The NDVI is a combination of addition, subtraction, and division of R and NIR channels; for the LANDSAT TM sensor
the formula is the following:
IR-R ||. TM4- TM3
NOV = ain TMS
(1)
Vegetated areas generally yield high values because of their relatively high near-infrared reflectance and low real reflect-
ance. In contrast, water has larger visible reflectance than near infrared reflectance thus, such features yield negative
index values. Rock and bare soil areas have similar reflectance in the two bands and result in a vegetation index near to
zero. The normalized index is preferred to the simple index for multitemporal and/or global vegetation monitoring
because it helps compensate for changing illumination conditions, surface slope, aspect, and other extraneous factors.
24 Modelling Vegetation Growth
After a damaging event such as fire, when an area
becomes potentially available for vegetation recov- Y;-F, Y; - Y Y4-Y; || 000000000000 Yn- Yn-1
ery, plants tend to colonize in a series of temporary
stages. Gradually, more permanent plant communi-
ties develop until a mature stage takes over reading
equilibrium with the regional climate and the local
substratum, topography, and water conditions
(Odum, 1993; Whelan, 1995; De Bano et al, Y= NDVI of year "i"
1998). Replanting can speed up the entire process. NR f=] 2 n halysis
In post-fire analysis this is important for two main 2107 7 Years 0° &
reasons: (1) the method presented here may be used (ES CO ICS RAIN dato)
for replanting feasibility studies or (2) for evaluat-
ing ongoing replanting strategies. Due to the tem-
poral limitations of remotely sensed imagery, it is
impossible to predict the time of full vegetation
recovery - which can take several decades - but it is
possible to monitor the first phases of recovery and
activity (Fiorella and Ripple, 1993). Additionally,
this is not applicable as a long term forecast since
Identification of classes with
homogenous growth rates
Mask &
averages
For each growth rate class the
NDVI averages are extracted
For each growth rate class a
growth model is fitted
unpredictable natural or anthropogenic occurrences Y,
may easily modify the studied growth rate. In
recent fires obviously no modelling was performed — Figure 1. Recovery Y Models are applied on Y, and
for lack of data. The entire methodological -growth modelling Ael class masks to forecast NDVI
approach is summarized in the flow-chart of Figure
1.
NDVI images highlight active green biomass, which in a period of high biomass production can be correlated to the pro-
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 1133