In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
221 
grass ETo into crop potential évapotranspiration (ETc), 
which is a function of phenological phases, crop species and 
variety, plant architecture and leaf area. In our research we 
used the soybean Kc values proposed by Allen et al. (1998). 
The main soil types for these locations are: Alfisol (Candido 
de Abreu and Cerro Azul) and Oxisol (Ponta Grossa, 
Jaguariaiva and Telemaco Borba). Soybean water balance 
was performed at daily time steps for each location and each 
year (2000-2007). The water balance was also adjusted 
according to soybean phenology. November 15 th was adopted 
as the main sowing date in this region (Geosafras, 2007). 
Outputs of potential évapotranspiration (PE), actual 
évapotranspiration (AE), soil moisture storage (SME), water 
deficit (DEF) and water excess (EXC) were thus obtained on 
a daily time step. However, in this study, only actual 
évapotranspiration was considered and it was necessary to 
aggregate the above outputs to 16-day means in order to 
compare them with the MODIS NDVI data. 
4. RESULTS 
4.1 Rainfall anomalies 
Figure 1 contains rainfall anomalies for each location and 
year from 2000 to 2007. Although state of Parana generally 
has a seasonal pattern of precipitation, such as, rainy summer 
and dry winter, it is important to point out the significant 
rainfall variation for all time series and for each year 
separately. Due to its location in Southern Brazil, this state is 
influenced by several macro-climatic factors, such as, 
migration of air equatorial and tropical masses from Atlantic 
in summer months; cold air masses (polar front) that come 
from the South of the continent, (Wrege et al., 1999), and dry 
air masses from Midwest Brazil, which can bring warm and 
relative dry weather. So, all this condition leads to high 
rainfall variability, in terms of amount of rainfall 
(millimeters) and number of rainy days (frequency). This 
weather variability influences the agriculture in this region, 
thus affecting crop productivity. Rainfall anomalies revealed 
occurrence of drought episodes from 2000 to 2007, but more 
significantly during 2004/05 year in Candido de Abreu, Cerro 
Azul and Ponta Grossa locations. This situation was reported 
by CONAB (2005) which mention a 7.1% yield loss in 
relation to the previous year yields, in the state of Parana. 
4.2 NDVI temporal analysis 
The different phenological stages of soybeans were identified 
with the NDVI temporal series from 2000 to 2007 for each 
location (Figure 2). Growth starts around mid November 
when soybean is planted and the cycle ends in March when 
the soybean is harvested. A similar pattern was observed for 
all years and locations indicating that the pixels selected for 
the analysis were indeed soybean pure-pixels. In 2004/2005 a 
drop in NDVI values was observed in three locations, 
according to rainfall anomaly values for Candido de Abreu, 
Cerro Azul and Ponta Grossa (Figure 1). Rainfall occurrence 
was low in January and mid February, which coincides with 
the peak of the reproductive stage of soybean in the study 
region. Actual évapotranspiration curves presented high 
values in December, January and mid-February, specially 
during soybean flowering and grain formation phases and it 
seems that NDVI response tends to follow this more realistic 
situation, in terms of biophysical condition, than rainfall 
occurrence. 
i) Candido de Abreu 
Anomalie! Rainfall poou-.'lKU} 
Candido dv Abreu 
iii) Jaguariaiva 
Anomalies - Rainfall (2000-2007) 
jaguariaiva 
— 
•MIA; 
¡own 
2003/04 
2004/05 
2005/00 2006/07 
— 
* 
* 
v) Telemaco Borba 
Figure 1. Rainfall anomalies for each year (2000-2007), 
during soybean growing season and for locations: i) Candido 
de Abreu; ii) Cerro Azul; iii) Jaguariaiva; iv) Ponta Grossa; v) 
Telemaco Borba. 
4.3 Linear regression analyses for all time series 
Analyses indicated that the simple regression of NDVI with 
Actual Evapotranspiration (AE) showed much higher R 2 
values than the regressions of NDVI in relation to rainfall 
(Figure 3). Despite this observed decrease in NDVI values 
during the Jan-Feb 2005 rainfall deficit period, we found 
better agreement between NDVI and AE. May be due to the