International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B8, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
at + 106,91
60.00 4 ;
40.00
20.00
AteaNDVI (kem?)
0.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
AreaNOVIE (km?)
Flow (Ls)
€}
e AreaNDVI s—— Rainfalt
50.00 100.00
$ E
= £
a 5000 =
«+ =
+ 0.00
2000 2004 2002 2003 2004 2005 2006 2007 2008 2009 2010
az # AreaNDVE zSimulsted ArsaNDVI
0.604 Phragmites MR. 0.20
ERE F
a" i
ue A i
& 040 ," Ephemeral |... 947
* ^a i
Nw i
0.20 + 0.44
JR MA Bd LAS QHD
Figure 3: Study of wetland area vegetation extends implemented using the 0.2 NDVI threshold. a) Multiannual values of AreaNDVI
recorded between February 2000 and May 2010 (MODIS-Terra and —Aqua data) and linear trendline. b) Correlation between
Areanpy, and spring flow recorded between 2002 and 2010 for the spring DAAO001 (Figure 1c). White circles indicate years 2007,
2008, and 2009. c) Comparison between 6-months smoothed rainfall and Area NDVI values recorded between 2000 and 2010. d1)
Mean monthly NDVI values calculated from NDVI times traces (2002 to 2010) recorded at 2 selected Phragmites australis and
ephemeral pixels (P1 and E2, Figure 1b) d2) Comparison between the simulated mean monthly Areaypy; reconstructed from the P1
and E2 monthly signal (c1) and the measured mean monthly Areaypy;.
4. DISCUSSION
Satellite imagery provides the synoptic view necessary for
inventory and mapping of ecological systems (Hewitt and
Mason, 1990). The different spatial and temporal resolutions
of sensors are complementary for conducting comprehensive
environmental monitoring. This study demonstrates the
potential of 250-m MODIS Normalized Difference
Vegetation Index for studying long-term dynamics of
Dalhousie Springs wetland vegetation.
4.1 Delineating wetland extent
The scale, heterogeneity and complex spatial patterning of
the dispersed wetland vegetation posed challenges for this
MODIS NDVI study. Selection of pixels representative of
homogeneous vegetation stands was difficult and required
expert knowledge of the environment. However, even if the
number of samples was limited by the landscape (Figure 1b),
their distinctiveness is supported by the Non-Metric
Multidimensional Scaling (Figure 2), as the six DSC cover
types fall into six distinctive clusters. Intra-group variability
can be attributed to sample site differences resulting from
variation between vegetation stands, intermixing of species
and variable influences of soil background: the scale and
heterogeneity of the wetlands made it difficult to find
homogeneous patches that would fill the MODIS
instantaneous ground resolution.
Despite these limitations, our study demonstrates that spring—
fed vegetation photosynthetic activity can be differentiated
from surrounding land responses in this medium resolution
imagery. The higher P90 values recorded in photosynthetic
activity traces are good indicators of wetland vegetation,
when riverine ecosystems are excluded (Figure 2). Similar
results were obtained by White et al. (2011), using NDVI
data derived from a QuickBird image and field vegetation
percentage cover. These authors also found low NDVI signal
for the GAB surrounding dryland vegetation, medium signal
for the ephemeral wetland ecosystems and higher NDVI
values for the perennial wetland samples composed by
Phragmites australis, and Melaleuca glomerata. The distinct
differences in NDVI of the different vegetation types evident
were attributed to the contrasting canopy structure and
growth patterns between the different communities (White et
al, 2011). This difference between non spring-fed and
wetland vegetation greeneess signal allows the establishment
of a NDVI threshold value of 0.2, above which NDVI values
are considered to be indicative of wetland vegetation.
4.0 Temporal dynamics of wetland extent
Our study confirms the potential of photosynthetic wetland
vegetation as an indicator of volume of the spring flow
(Fatchen, 2001; Fatchen and Fatchen, 1993; Mudd, 1998;
White and Lewis, 2011; Williams and Holmes, 1978), as
correlations are demonstrated between multi-annual mean
extent of vegetated spring areas and groundwater discharges
for one large spring at DSC (Figure 3b). The temporal
frequency and archive of MODIS NDVI measurements allow
us to demonstrate this relationship with more observations
over a longer period than previous studies.
Intern:
Howeve
controlle
variabili
historica
variatior
rain mea
wetland
monthly
(Figure
also pla
wetland
Therefor
wetland
longer c
flow m
studies
strong b
context
White :
wetland
satellite
of wetla
50% (F
over the
2010), 1
these re
area, th
increase
While t|
give pr
objectis
area, ai
annual
Observe
phenolc
differer
trends «
suggest
a return
MODIS
with hi
extende
and tri
imple
variatic
for thi
manage
thresho
changii
researc
limits :
will e
dynam
The ai
NDVI
Dalhot
Result:
DSC 1
respon