Agriculture Reference
In-Depth Information
has been used for applications in vegetation classification of forests, grasslands, and
agriculture areas.
As early as the 1970s, satellite imagery has been used to identify crop types and to
determine if these crops are supplied with water by means of irrigation (Hoffman et
al., 1976). Narciso and Schmidt (1999) investigated the potential of using LANDSAT
TM images acquired using satellite remote sensing for identification and classifica-
tion of sugar cane. The main onboard satellites usually have coarse (greater than
100 m) spatial resolution, such as the AVHRR, SPOT VEGETATION, and MODIS.
Tuppad et al. (2010) derived and quantified irrigated agricultural areas in Texas on
the basis of an unbiased and consistent method of using satellite images and image
processing techniques. A few supervised and unsupervised classification methods
were explored to identify irrigated and nonirrigated winter wheat in a selected
county in the Texas Panhandle. It is apparent that there is a distinguishable differ-
ence between irrigated and nonirrigated wheat according to the NDVI (normalized
differential vegetation index) curves. Irrigated wheat NDVI typically ranges from
0.3 to 0.1, whereas nonirrigated wheat ranges from 0.2 to 0.4. Better classification
was obtained by using a group of pixels as ROIs instead of a single pixel. Gautam
and Panigrahi (2004) applied LANDSAT TM satellite imagery and non-imagery
information to predict the residual soil nitrate content in the Williston research site
from three neural networks: back propagation, radial basis function, and modular
architectures. A root mean square error of prediction of 11.37 (9.09%) was obtained
with the residual soil nitrate prediction model using the modular neural network.
The best correlation coefficient of 0.81 (81%) was also obtained by this model among
those provided by all three neural network models. Hanna et al. (2004) compared
ground-based survey (traditional method) and remote sensing techniques in crop
inventory. The image data acquired from the French satellite SPOT has false color
composites with a 20-m spatial resolution. Its accuracy was checked by using another
LandSat scene. From the study, it was found that the application of satellite images
is cost- and time-effective. The accuracy of 84% was obtained using remote sensing
data in this study. Approximately 25,000 Egyptian pounds (LE; $1 ≈ 6 LE) was used
for the image (60 × 60 km
2
) and its processing, and it took about 2 weeks to complete
the image processing. The number of repeated coverage is mainly influenced by the
availability of ground truth, the weather condition, the accuracy needed, and other
limitations. As a comparison, the ground-based survey (traditional method), which
is performed once every 5 years, costs 13 million LE. Schmidt et al. (2001) derived
a standard vegetation index NDVI for sugarcane in three regions of South Africa on
the basis of images in five channels (two infrared, one red, and two thermal bands)
captured by the NOAA. The growth cycle of each harvest year crop was represented
by the accumulated 10-day NDVI values over periods. The correlations were ana-
lyzed between the resulting index of accumulated NDVI and the observed mill and
farm yields (Schmidt et al., 2001). A close correlation means that the crop estimates
can be improved by operating the NOAA satellite. On the basis of the execution of
this research by operating the NOAA-AVHRR sensor over a period from 1988 to
1998, it was found that there were significant correlations between average sugarcane
yield over a mill area and NDVI for five of the nine sampled areas. AVHRR data,
as the primary source of large area surveys, have been widely used for analyzing
