IX-B8, 2012
mean values for
positions
1 be seen in Table
hese data confirm
rectly correlated
ent, the lower the
an now be done,
really, during the
? lowest degree in
son, with water
inea presence, the
> discretely higher
ninimum is hardly
e is near perfectly
n NDVI response
tes. The fact that
ler response than
f graminea, often
y reflective in the
(SD) is not very
class, as expected.
a high percentage
mporal series and
what can also be
ws to evaluate the
| access and well
' to proceed at the
jbdivision of the
ologous areas.
) 3 Group 4
8 18.09
08 5110.60
6 60.28
23 2613.29
44 9323.46
16 2894.51
97 6453.02
19 1120.87
28 6428.95
1p with different
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
Range
so
Mean
Maximum:
Minimurr
2 2000 4000 6900 3900 10000
Figure 4. NDVI statistics from time-series representing the
mean values for each group with different textural
compositions, based on the amount of fine-grained sediments
4. DISCUSSION
The analysis of the pair of MODIS triplets allied to the 5-year
NDVI time-series analysis and its seasonality parameters show
clearly the possibility of estimating the sediment granulometry
remotely through the behavior of the vegetation along the dry
and rainy seasons. Each granulometric interval of the studied
samples is associated with vegetation that has a particular intra-
annual variability. This allows the proposal of a subdivision of
the Pantanal in geological and environmental homologous sub-
areas. The spatial distribution of these data can be done in pairs
of MODIS images in the dry and rainy seasons, which were able
to meet this demand.
This vegetation analysis seems to be a good method for
providing indirect information about the sediment
granulometry. Since soils have different moisture holding and
drainage capacities, areas with higher amounts of fine-grained
sediments presented higher index values in the dry season,
when water disponibility is more critical, and presented lower
NDVI seasonal amplitude, while areas with predominantly
sandy sediments which dry out deeply during the dry season,
presented a higher variation in NDVI seasonal amplitude, and
lower NDVI values. In the rainy season water is not a limiting
resource, and the NDVI values are higher in all sediment types,
especially between November and begin of April, when the
rainfalls occur.
Higher NDVI values reflect high leaf area and great vigor and
photosynthetic capacity (or greenness) of vegetation canopy,
whereas lower NDVI values for the same time period are
reflective of vegetative stress resulting in chlorophyll reductions
and changes in the leaves internal structure due to wilting (Gu
et al, 2008). These authors found that correlations between
MODIS 500-m satellite indices and the soil moisture index are
highly dependent on both the level of land cover heterogeneity
and soil type. We did not take in consideration the vegetation
heterogeneity, because we worked with groups with different
textural compositions, using mean phenological profiles, which
attenuates the vegetation differences between sample sites.
Probably when working with each sample profile, it will be
Interesting to take into account the vegetation type. Although,
these authors found that NDVI and Normalized Difference
Water Index (NDWI) had high statistically significant
correlations with the 25-cm layer soil moisture, which indicates
that this indices have slightly stronger responses to the soil
moisture variation at this depth, what is consistent with our
findings.
The only sub-areas that did not follow this pattern were the ones
that are inundated most of the year and the sandy areas that
were not affected by inundations. This was due specially
because they presented low NDVI amplitude most of the year,
presenting almost the same behavior along the seasons.
The seasonality parameters showed motivating data, but could
be interesting analyzing each sample separately to understand
the variations between different sub-regions in Pantanal, take
into account the predominant vegetation type in the sampled
area and its associations with the rainfall. The high inter-annual
variability of rainfall in the region results in variability in the
flooding regimes, and possibly in the phenological profiles.
Adami et al. (2008) emphasized that, in the Pantanal, different
kinds of vegetation receive different amounts of rainfall, in
different locations and, consequently, have different spectral
responses. So, the small variations in mean peak time might be
the result of intra-regional rainfall differences, since the rainfall
is significantly higher in some parts of the surrounding uplands
of Pantanal, particularly in the north (EDIBAP unpubl. rep.
apud Hamilton et al, 1997). The general inter-annual behavior,
however, is maintained despite the inevitable variations due to
local rainfall and different vegetation types that occur in the
area, from dense forests to savannas with low rates of leaf area.
The maximum values in NDVI between the groups were not
very different, and it may be due to NDVI limitations.
According to Jensen (2009), since this is an index based on
ratio (non-linear), it can be influenced by the additive effects of
noise, such as additive atmospheric effects like path radiance.
Besides that, the NDVI is very sensitive to the substrate under
the canopy, where the higher values occur with darker
substrates. Nevertheless, the main limitation of the NDVI is the
saturation problem over lands with high biomass such as
tropical forest.
In general, this method showed useful, and the coarse resolution
from MODIS (250-m) was considered adequate for showing
granulometric tendencies in Pantanal, although it has to take in
consideration the great size of this wetland. So, other regional
scale areas need to be investigated to draw stronger conclusions
about the robustness of the methodology. If confirmed for other
regions, the coarse resolution time-series method will enable
low-cost detection and mapping of the granulometry of large
areas. Furthermore, phenological data profiles should be
increased to complete the available data series (2000 to 2011) in
order toreduce the lossof information due to clouds
and atmospheric influences. Other vegetation indices, like
NDWI, Enhanced Vegetation Index (EVI) and Leaf Area Index
(LAI), with wider profiles also might gave significant
improvements in the understanding of the relationship between
the vegetation and the sediments granulometry.
5. CONCLUSION
Results from this study indicate that there is a high relationship
between drought-related changes in vegetation extracted from
NDVI and sediment texture, parameter that plays an important
role in soil moisture, influencing the vegetation response to
droughts. This method showed very good results on accessing
sediment texture from vegetation phenology and improved our
understanding of how phenological profiles vary over different
granulometric sediments over space and time.
In conclusion, 16-day MODIS NDVI time-series may be used
for detecting granulometry tendencies and mapping
granulometric homologous areas with similar behavior in
Pantanal. Although, it is necessary to take in consideration that
