2. METHODS
2.1 Study site
This study was carried out in the Pantanal wetlands of Mato
Grosso and Mato Grosso do Sul states, south-western Brazil, in
an area known as Pantanal Matogrossense. The area is the
largest continuous freshwater wetland in the world, covering
over 140,000 km? (Alho et al., 1988), and is characterized by
low terrain with an annual flooding regime.
The climate of the Pantanal is tropical with a marked wet
season, with most rain falling between November and March.
The vegetation is heterogeneous and influenced by four biomes:
Amazon rainforest, Savanna (predominantly), Chaco, and
Atlantic Forest (Adämoli, 1981). Note that Cerrado and Chaco
are savannas.
According to RADAMBRASIL (1982) the soils are variable but
generally contain more silt and clay in areas subject to riverine
overflow and tend to be sandy on the higher parts of the alluvial
fans, including areas subject to flooding by local rainfall. These
soils are closely related to the nature of the sediments, a
consequence of the source material and the processes of
deposition and sedimentation (Assine and Soares, 2004). For
example, the alluvial fan of Taquari has a predominance of
sandstones in the source area and the sediment granulometry at
the Alta Nhecoländia, Baixa Nhecoländia and Paiaguäs sub
regions are essentially sandy. This variety of source areas
confers contrasting characteristics between the prevalent
sediments in different alluvial fans that filled the depression of
the Pantanal.
2.2 Field data - Sediment site selection and granulometric
analysis
Sediment samples were acquired from 161 ground sites in
August of 2011 (dry period) during 16 days fieldwork covering
over 2,000 km within different Pantanal regions. The spectral
behavior of MODIS during rainy and dry seasons guided the
choice of sampling sites.
We focused our study on MODIS tiles H12V10 and H12V11,
which cover the entire Brazilian Pantanal. There were obtained
two images of the MOD09GQ product, one from the rainy
season (March 5" 2005) and the other from the dry season
(August 3" 2005). These data were converted to NDVI, and
then combined in triplets (NDVI, NIR and red bands) for both
periods to identify vegetation changes in those seasons. The
sampling was carried out using auger Dutch type, collecting the
portion immediately below the soil with presence of roots.
Grain-size analyses were performed. The method for
determining the percentage of silt, sand and clay in a soil was
described by Camargo et al. (2009).
2.3 Design of work hypothesis
The study hypothesis - the presence of a relationship between
the annual behavior of vegetation, indicated by MODIS images
with the granulometry of the soil / sediment - was based on the
statement that this phenomenon cannot be totally explained by
the small relief variations in Pantanal but by the different
sediment granulometries. According to this hypothesis, in areas
where the sediments are sandy the surficial water accompanies
the lowering of water table, impairing the capacity of vegetation
to face the dry season. The idea was tested and confirmed
during a preliminary fieldwork which was guided by drought
and flood MODIS images.
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
2.4 Satellite data and image processing
The MOD09GQ product and the MOD13Q1 NDVI product
from MODIS were acquired for this study. The MOD09GQ
product provides a 250m resolution daily imagery including the
red (620-670 nm) and infrared bands (841—876 nm) and the
MOD13Q1 product is a 250m 16-day composite of the highest-
quality pixels from daily images and includes Normalized
Difference Vegetation Index (NDVI), Enhanced Vegetation
Index (EVI), blue, red, near infrared (NIR), mid-infrared
reflectance and pixel reliability. The images were obtained from
the MODIS website (http://mrtweb.cr.usgs.gov/) as a free
download using USGS MODIS Reprojection Tool Web
Interface (MRTWeb). In this platform, the images were subset
and projected to geographic coordinates using the WGS84
reference ellipsoid. Subsequently, the image processing was
performed using the software Envi 4.7 and TIMESAT.
The subdivision of the Pantanal here proposed is an evolution
of the proposal done in Penatti et al. (2011), in which Pantanal
was divided in 18 areas of homologous behavior, taking into
account the distribution of floods, the behavior of the
vegetation, the differences in contrast and brightness, textures
and patterns between periods of flood and drought (obtained
from MODO09GQ products) and data extracted from digital
elevation model obtained from the SRTM data. The new
vectorization on the triplets generated was done in ArcGIS
improving the limits of each area and analyzing the behavior of
the vegetation between the seasons in each region. Then, we
calculated each sub-area statistics mean, minimum, maximum
and percentage of reduction between seasons for NDVI from
the rainy and the dry periods.
The MODIS-NDVI 16-day composite data (MODI3Ql) was
used to generate time-series from 2006 to 2010 (115 granules)
for pixels associated with locals sampled for granulometric
analyses. The NDVI is based on the relationship between the
red and near infrared wavelength reflectance and responds to
vegetation photosynthetic pigment concentrations (mostly
chlorophyll) and structure. Detailed time series of this data
follow the annual growth cycles — or phenology — of vegetation
found in a pixel (Clark et al., 2010). Since soils act as substrate
for plant communities, the information about its composition
may be important contributions to vegetation studies. The
vegetation analysis can provide indirect information about
the sediment granulometry.
Although MODIS data have been corrected for atmospheric
effects there is remaining disturbance affecting the data quality,
e.g. geolocation errors, angular variations, residual clouds and
atmospheric disturbance (Eklundh et al., 2007). To minimize
these effects the data were smoothed with the Savitsky-Golay
algorithm of the TIMESAT package (Jónsson and Eklundh,
2004, 2011), and 5-year NDVI time-series profiles were
generated for each granulometric sample point. The MODI3
pixel reliability band was used to weight each point in the time
series: value 0 (good data) had full weight (1.0), values 1-2
(marginal data, snow/ice) had half weight (0.5), and value 3
(cloudy) had minimal weight (0.1). Function-fitting parameters
used in TIMESAT were: a Savitzky-Golay filter procedure, 4-
point window over 2 fitting steps, adaptation strength of 2.0, no
spike or amplitude cutoffs, season cutoff of 0.0, and begin and
end of season threshold of 20%.
The output from TIMESAT is a set of files containing annual
seasonality parameters (Length: length of the season; Base
value: average of the left and right minimum values; Peak time:
largest data value for the fitted function during the season; Peak
value: largest data value for the fitted function during the
season; Amplitude: seasonal amplitude, etc), as well as fitted
func
(Fig
Bas:
into
the
cont
sam
Groi
sedi
fine
here
For
valu
iden
goi
and
gen
mea
eacl
The
subi
bui]
ton:
veg
Sin
radi
diff
ulti
reg
In ]
wit]
sea
in t
in |
apr
sed
are
The
var
rel:
at |
are