|The Morrow Plots|| Field
Graduate Research Assistant
Department of Crop Sciences
Broadcast applications of soil- and foliar-applied herbicides have been used routinely for weed control in corn and soybean fields in Illinois. With the evolution of variable-rate fertilizer application technologies using Geographic Information Systems (GIS), there has been an increased interest in mapping weed populations with similar technologies throughout the growing season. Remote-sensing applications may be developed to detect weed populations and species in both corn and soybean fields by measuring a wide spectrum of wavelengths. The development of remote-sensing applications to map weed populations ultimately may allow producers to reduce the total pesticide load on a given acre, track weeds species shifts, and provide both economic and environmental benefits.
|Remote sensing derives information about an object using images acquired from an overhead perspective. Information is based on the reflected radiation of the object and employs modern sensors, data-processing equipment, and information-processing methodology. How does this work? The colors that we see with our eyes are shades of blue, green, and red—colors that are in the visible wavelength range (400–700 nm). This is the same range of radiant energy that is used for photosynthesis. Remote sensing provides a way of seeing what objects look like inside and outside the visible spectrum of light. By being able to visualize plants at wavelengths ranging from 286–1086 nm, subtle differences in reflectance patterns of different plants may be detected. These subtle differences may help us identify different weed species, such as velvetleaf and common waterhemp, from heights of up to 2,000 feet.||
Figure 1. Statistical separability of weeds
from soybean; solid blue regions indicate separability.
In 2000 and 2001, the University of Illinois, in conjunction with USDA-ARS and the NASA Commercial Remote-Sensing Program, set up field studiesto detect weed species from soybeans. The objectives of this experiment were to determine if remote sensing could be used to distinguish weed species from soybeans and determine if remote sensing could be used to differentiate between several grass and broadleaf weed species.
Nine weed species were planted in monospecies blocks measuring 20 by 50 feet, replicated six times. Weed species included barnyardgrass, giant foxtail, giant ragweed, common lambsquarters, smooth pig-weed, shattercane, Pennsylvania smartweed, velvetleaf, and common waterhemp. Soybeans in 30-inch rows were planted through half of these plots. Remotely sensed imagery data were collected weekly by a fixed-wing airplane equipped with a Spectral Vision RDACSH3 hyperspectral sensor with 20 inches spatial resolution collecting 60 bands from 457 to 823 nm, with a 6 nm bandwidth. In addition, radiometer scans (16/plot) were made when weed species were 10 to 15 cm in height with a GER 1500. Scans were collected in the 300 to 1100 nm range (512 spectral bands).
Figure 2. Image data for weed species delineation classifications.
In 2000, radiometer measurements in the visible spectrum (400 to 700 nm) successfully differentiated soybean from shattercane, common waterhemp, velvetleaf, and common lambsquarters (Figure 1). Measurements in the near-infrared region (700 to 1200 nm) separated only two species (shattercane and common waterhemp) from soybean. Radiometer measurements also differentiated several grass and broadleaf weed species. Giant foxtail could not be distinguished from velvetleaf; however, common waterhemp, common lambsquarters, and shattercane could be distinguished from all other species. Remotely sensed imagery collected on June 30 was 100 percent accurate in differentiating soybean from shattercane and common waterhemp (Figure 2). The accuracy of classifying the weeds from soybeans decreased as the soybeans grew.
|Department of Crop Sciences
College of Agricultural, Consumer and Environmental Sciences
University of Illinois Extension
© 2001 University of Illinois
Email site problems to the webmaster