oul 2004
re much
joints of
0.01 to
with the
id 1996,
reduced
er of 5).
DVI are
C effects
er types
ves. The
> NDVI
n of the
hat only
ans that
months.
ks were
isplayed
ates that
mber to
mplies a
v during
'sting in
for an
getation
lifficult,
because
»blems.
ifferent
different
>s in the
ver type
‚ance of
ferences
. in the
' to start
nces are
not be
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. [Istanbul 2004
detected with the coarse resolution of this imagery (see
figure 5). The problem is that apparently the time differences in
the beginning of the growing season in each vegetation type
take place within a few days. If all those days fall within the
compositing time of the NDVI week, then they are masked out.
Differences in NDVI between the years for each cover type
polygon, however, do exist for the magnitude of the NDVI in
the order of around 0.05 to 0.1, but differences in the timing of
the growing season do not appear to be evident.
07
0,6 4
NDW
©
Ca
-0,1 T T T T T
n 10 20 week 30 40 50
—s— polygon2 -—— polygon4 polygon& polygons |
Figure 5: NDVI curve of blanket bog polygons, 1997. The
polygons are numbered (from NW to SE). The lower
the number the further north is the polygon.
4.2 Inter-annual comparison of cover types
4.2.1 | Comparison between different cover types:
However, despite limitations in data availability and
unexplained phenomena, the dynamics of the vegetation
response can be seen well within the NDVI dataset. Inter-
annual comparisons of the NDVI curves can be made. Figure 6
shows the NDVI curves for selected cover types exemplary in
the years 1996 and 1997. The curves are displayed in a nearly
consistent order each year: the highest values always occur for
pasture and arable lands which lie around 0.6 units or more.
This is closely followed by broadleaf forest and Nardus/Molinia
(0.5-0.6). These vegetation types display a high seasonal
response with a large NDVI magnitude and range. For arable
land, pasture and broadleaf forest, the onset of the growing
season in spring leads to an abrupt green-up, which can be seen
in the steeply ascending curves. A high NDVI range can also be
seen for montane vegetation, especially in 1995, 1996 and
1998. However, this is also the cover type where the peak in the
growing season, and the start of it, have the lowest NDVI
values. Montane polygons will consist of mixed pixels as this
class is distinguished by the distribution of the vegetation types
in the higher elevation montane mountain areas rather than by
the distribution of specific vegetation types themselves. This
class can also be expected to contain responses from rock
outcrops and soil because vegetation above a certain altitude in
the mountains will be sparser. This would explain the low
NDVI values found. The steep ascent of the curve might be due
to the fact that the sensor response until spring is that of rocks,
soils and potentially snow (= low NDVI). As soon as the first
plants emerge the NDVI will rise abruptly.
Blanket bog/peat and wet heather display similar NDVI curves
in magnitude and dynamics (figure 6, figure 7b). This represents
cither a real NDVI response that is similar for both cover types
or is enhanced by the fact that those cover types co-occur in the
481
same ecosystem. SNH states that these covers mostly occur in
mosaic classes, i.e. mixed in the same arca. Along with dry
heather which has slightly higher NDVI values, they experience
NDVI values of around 0.4-0.55 in summer (compare figure 7).
Their curves are flatter than those of arable land. pasture and
broadleaf forest and have some of the longest peaks of all cover
types, i.e. a long growing season.
The strength of the NDVI to distinguish vegetation at
community level is demonstrated by the response of wet and dry
heather which each display characteristic features.
1996
0,7
0,6 + si
Z x
3 À
0,5 4 /
ff
a
Z 4 e /
94 4 IE
à {€ 7 f X
e i f. \
0,3 4 Y
M
0,2 4
0,1 4 t
0 t + + t
0 10 20 Ves À 40 50
1997
0,7
0,6 4 /
0,5 4 é
us E
/ 7
0,4 4 = m M
z x
= sa ry
034 e res
0,2 -
0,1 +
0 ny : : + t
0 10 20 Werk 20 40 50
dry heather - - wet heather —— blanket bog
Mardus/Molinia = montane vegetation —— pasture
arable land coniferous plantation —— broadleaf forest
Figure 6: Comparison of cover types 1996 and 1997
Another interesting feature that is worth comparison between
cover types, is the growing period. To establish the start of the
growing season is difficult. For example Lloyd, 1990 assigned a
threshold NDVI (0.099) at which the vegetative activity is
assumed to begin. This threshold, however, varies with
vegetation type, soil background. illumination condition, sensor
and calibration. A single threshold is therefore not of much use.
A sudden increase of the curve as a signal of the onset of
significant photosynthetic activity might be more useful. Pasture
and arable land have a steep increase in NDVI from the mid/end
of March onwards (see figure 6, graph of 1997). They reach
their peaks at the same time, around week 21. The curve then
falls slowly for pasture and abruptly for arable land with a deep
trough in the curve around week 35 that reveals the harvest.
Wet heather, dry heather, Nardus/Molinia and blanket bog
begin to rise with a delay of about 2 weeks (between weeks 14
and 18). The rise is also more gradual with nearly level NDVI
values in summer. The rise of montane vegetation starts latest
=
c