In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
fl 
NDVI (MODIS SENSOR) RESPONSE TO INTERANNUAL VARIABILITY OF 
RAINFALL AND EVAPOTRANSPIRATION IN A SOYBEAN PRODUCING REGION, 
SOUTHERN BRAZIL 
A. Giarolla a , W. E. Baethgen; b , P. Ceccato b 
Institute) Nacional de Pesquisas Espaciais, Centro de Ciências do Sistema Terrestre (CCST/INPE), Av. dos 
Astronautas, 1758, Sào José dos Campos-SP, 12227-010, email: angelica.giarolla@cptec.inpe.br. 
b International Research Institute for Climate and Society (IRI), The Earth Institute, Columbia University, Lamont 
Campus, Palisades, New York, USA, email: baethgen@iri.columbia.edu; pceccato@iri.columbia.edu 
KEY WORDS: soybean (Glycine Max, L. Merr), NDVI, remote sensing, water balance, GIS. 
ABSTRACT: 
This study aimed at evaluating the response of the Normalized Difference Vegetation Index - NDVI (MODIS sensor, TERRA 
satellite) of soybean to interannual variability of rainfall and évapotranspiration in Campos Gerais, a region of the state of Parana in 
southern Brazil. Landsat TM 5 and 7 images were selected for analyzing the spatial soybean field distribution for the region from 
2000/01 to 2006/07 and to identify soybean fields. We then identified 175 pixels (250 x 250m) that contained only soybean fields 
(“pure-pixels”) based on the soybean maps obtained with the Landsat TM images. The next step was to extract the NDVI values for 
these soybean pure-pixels and to analyze the NDVI spectral curves considering the soybean phenology. Data from nearby 
meteorological stations were obtained and used to calculate the soil water balance for soybean fields in 5 locations distributed in the 
Campos Gerais region. To obtain actual évapotranspiration values, the water balance was calculated for each year, and for the same 
period covering the entire soybean growing season. Anomaly values were calculated for each year to verify the interannual rainfall 
variability. Linear regression models were adjusted between NDVI and i) rainfall and ii) actual évapotranspiration for all time series. 
Analysis of the evolution of NDVI values allowed identifying the soybean growing season (November to March) and also the dry 
season for this region according to rainfall anomaly values. Statistical analyses showed that actual évapotranspiration presented best 
agreement with soybean NDVI in relation to rainfall, probably due to the fact that this variable integrates information of rainfall, 
temperature and soil water holding capacity for the entire study period. 
1 INTRODUCTION 
Crops can suffer damage from excess or lack rainfall and 
investigate this variability in its frequency of occurrence over 
a crop season becomes essential for agricultural planning 
systems. Several studies have been made about early warning 
systems, based on seasonal and interannual climate forecasts 
(Guillevic, et al, 2002; Nobre et al., 2006; Meinke and 
Stone, 2005; Ogallo et al., 2008). An objective evaluation 
system of weather variability as well as the improvement of 
crop monitoring capabilities for yield prediction can provide 
knowledge to mitigate weather impacts. Climatic variability 
plays a major role in promoting meteorological conditions 
that deviate substantially from mean conditions, including 
weather extreme events (Seiler et al, 2007). Crop growth, 
development and yield are affected by climatic variability via 
linear and nonlinear responses to weather variables (Semenov 
and Porter, 1995). According to Baethgen (1997) agricultural 
systems which are currently subject to extreme climatic 
interannual variability (drought, flood, storms, etc.) are likely 
to become even more vulnerable under the most commonly 
expected scenarios of climate change (i.e. increased 
temperatures, increased rainfall variability). 
Evapotranspiration (ET) is an essential component of the 
energy and water budgets in grassland and agricultural 
ecosystems (Allen et al., 1998) and understanding the 
processes that affect ET at different temporal scales and 
under a variety of environmental conditions is important in 
crop modeling production. In general, remote sensing 
techniques can not measure evaporation or évapotranspiration 
directly. However, Engman (1995) mentions that remote 
sensing does have two potentially very important roles in 
estimating evapotranspiration: i) to try establish a empirical 
relationships between remote sensing data and ET to larger 
areas, including those areas where measured meteorological 
data may be sparse; and ii) remotely sensed measurements 
may be used to measure variables in the energy and moisture 
balance models of ET (Bastiaanssen, 2000). Tucker et al., 
(2001) report that the use of decadal NDVI time series has 
improved the study of interannual variation in vegetation. 
Currently, Suzuki et al. (2007) examined interannual NDVI 
data and its relation to ET in northern Asia and significant 
positive correlations were found between NDVI and ET 
anomaly interannual variation over vegetated areas. The 
objective of this study was to evaluate NDVI response to 
interannual variability of rainfall and evapotranspiration 
during the entire soybean season for the period 2000-2007 in 
Campos Gerais region, Southern Brazil. 
2. STUDY AREA AND DATA 
2.1 Study area 
Campos Gerais is a region located in southern Brazil, limited 
geographically by the coordinates 23°45'00"; 
25°50'00"South and 49°10'00"; 51°25'00"West. 
Agriculture is the principal land use in the region and the 
three main crops are soybean, maize and dry bean. The 
highest yields for a short duration soybean cultivar are 
obtained when planted in October and November and in the 
southern part of this region, many farmers usually start 
planting in mid-October (Heinemann et al. 2002). Soybean 
cycle lasts 110 days in average and the harvest occurs 
219