TEMPORAL RELATIONSHIP OF NDVI AND SEASONAL VEGETATION IN
STRUCTURAL GRADIENT IN PERMANENT PLOTS
H. N. Mesquita Jr M. ^
* M.D. Bitencourt b
+ zh
' beto@ib.usp.br; ’ tencourt@ib.usp.br
Ecology Department, University of Säo Paulo
Phone # 55-11-30917603 - Rua do Matäo, trav. 14, #321
Säo Paulo - SP, Brazil, CEP 05508-900.
Commission VII, WG VII/3:
KEY WORDS: Remote Sensing, Vegetation, Modeling, Sampling, Ecosystem, IKONOS, Landsat, Temporal
ABSTRACT:
The relationship between remotely sensed data and vegetation has been the object of study for experts in both environmental remote
sensing and plant ecology.
One difficulty in establishing relationships between remotely sensed reflectance and vegetation is due to
seasonal dynamics. The objectives of the study were to derive a model for the relationship between the variation of spectral response
of the sensors, climate, and phenology of Cerrado. The site location is the Pé-de-Gigante, Brazil (21037'30" S, 4703730" W). Sensor
spectral responses were analyse
d. in the form of NDVI images from Ikonos-Multispectral, Landsat-TM e Terra-ASTER, l'erra-
MODIS. Two years data of seasonal variation in NDVI images were compared with meteorological data and LAI and
phytossociological parameters o
climatologic parameters. A temporal series of nine Landsat-TM images and one I
btained from 39 plots. The two year time series of fifteen days images were compared with
konos- Multispectral and one Terra-ASTER were
compared with the field data. The structural parameter with the highest overall correlation coefficient was the cylindrical volume.
The best correlation coefficient of NDVI and LAI were obtained to Ikonos and Aster images (r = 0,74). A two year temporal series
NDVI images were correlated with meteorological data to identify the phase lag over the vegetation structural gradient. The seasonal
phase lag variation was lower with less structurally heterogeneous canopy and higher with high floristically heterogeneous canopy.
The final results were achieved by the theoretical models of the seasonal variation of the veg
the climate.
1. INTRODUCTION
The reflectance of the vegetation in the red depends on the
present amount of clorofila in leves and the reflectance of the
vegetation in the next infra-red ray, of the alive foliar structure.
In this way, the IVDN has been mainly related with fitomassa
foliar green, among others parameters of the vegetation. The
vegetation indices are numerical models that relate the spectral
reply of the vegetation with the density of vegetation for area,
in sight of the antagonistic answers of the red and next infra-red
ray.
The Cerrado vegetation is the most seasonal vegetation in Säo
Paulo state and the measurement of seasonal dynamics by
orbital remote sensors was the object of this study. Cerrado
vegetation has a characteristic physiognomic — gradient
determined by the proportion of herbaceous and arboreal
components. The objectives of the study were to derive a model
for the relationship between the variation of spectral response of
the sensors, climate, and phenology of Cerrado.
Amongst the spectral indices of vegetation, the IVDN (index of
vegetation for normalized difference) is more widely used for
the study of the continental vegetation. The first works with the
IVDN presented correlation with biomass of agroecossistemas
and homogeneous vegetations (Tucker, 1979; Anderson et al,
Corresponding author.
setation physiognomies, conditioned by
1993) and with the amount of water in the leaf (Tucker, 1980).
The pioneering works in Brazil with the use of the IVDN had
searched to evidence the existing correlations between
fitomassa of the open pasture and the observed spectral reply in
orbital level through indices of vegetation (Bitencourt-Pereira,
1986: Valeriano & Bitencourt-Pereira, 1988; Santos, 1988).
’articulary in Brazil, Bitencourt-Pereira (1986) and Santos
(1988) had made a calibration of the vegetation index,
respectively, with the cerrado stricto sensu (s.s.). and the campo
cerrado of central plateaus of Brasil. The first author related
green fitomass with the spectral signature gotten by the
Landsat-MSS and with radiometry of field, and found
coefficients of linear correlation of 0,94 and 0,96 respectively.
Beyond the biomass, the fraction of covering of the vegetation
frequently was related to the IVDN (Jasinski, 1990), as much in
opened savannahs (Larsson, 1993), as in denser forests (DeFries
et al., 1997). Another parameter of the structure that frequently
is associated with the IVDN is the index of foliar area (TAF),
that it can be esteem with use of optic sensors, based in the
fraction of light intercepted for the canopy (Chen 1996a and
1996b; White ef al., 1998; Eklundh ef af., 2001). Other
parameters, as density and basal area, also are associates to this
index. D'Arrigo er al. (2000) had related the annual variations
of long temporal series of IVDN with growth's rings of species
of boreal forests.
394
Interne
Some
the IV
the in
paranx
proces:
Chimit
Since |
vegetat
variatic
Sensors
these
tempor
al., 19"
2002).
associa
precipi
in the
some ¢
variatio
for the
1996; [
In Braz
been ca
and rai
cerrado
cerrada
conside
1997. V
The obj
paramet
images
tempora
way, 39
with sa
(30x30n
(4x4m).
21 Stu
The stuc
47037'3(
state, Br.
720 met
area. The
Steepest
draining.
The Pé-c
from can
configuré
Coutinho
