Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: USING REMOTE SENSING AND GIS FOR DETECTING AND MAPPING INVASIVE WEEDS IN RIPARIAN AND WETLAND ECOSYSTEMS Title: Employing broadband spectra and cluster analysis to assess thermal defoliation of cotton

Authors
item Fletcher, Reginald
item Showler, Allan
item Funk, Paul

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: December 1, 2013
Publication Date: March 1, 2014
Repository URL: http://handle.nal.usda.gov/10113/59468
Citation: Fletcher, R.S., Showler, A., Funk, P.A. 2014. Employing broadband spectra and cluster analysis to assess thermal defoliation of cotton. Computers and Electronics in Agriculture. 105:103-110.

Interpretive Summary: Over the last ten years, thermal defoliation has shown promise as a nonchemical alternative for terminating cotton growth and defoliating cotton canopies. The technique involves using propane to heat air that is applied directly to the plant canopy to rapidly kill the leaves. There are several risks associated with incomplete thermal defoliation: (1) staining of the cotton fiber may occur in areas where leaf kill is incomplete, reducing the price grade of the cotton, and (2) juices in green leaves may increase gum build-up on picker spindles, requiring stoppage to clean the spindles and thus increasing harvesting time per field. Rapidly identifying areas not responding to thermal treatment can support producer decision making regarding re-treating these areas and assist them in making adjustments to the defoliator for the next year. ARS scientists demonstrated in a ground-based study that visible and near infrared light reflectance properties of cotton plants could be used for assessing the effectiveness of thermal defoliation, and they used this information with a computer classification algorithm to group cotton plants into treatment response levels. The immediate benefit of this research has been to provide improved information on the application of light reflectance data and computer classification as decision support tools for assessing thermally defoliated cotton fields.

Technical Abstract: Growers and field scouts need assistance in surveying cotton (Gossypium hirsutum L.) fields subjected to thermal defoliation to reap the benefits provided by this nonchemical defoliation method. A study was conducted to evaluate broadband spectral data and unsupervised classification as tools for surveying cotton plots subjected to thermal defoliation. Ground-based reflectance measurements of non-treated and thermally treated cotton canopies were collected at two study sites with a handheld hyperspectral spectroradimeter. The spectral data were merged into eight broad spectral bands: coastal blue (400-450 nm), blue (450-510 nm), green (510-580 nm), yellow (585-625 nm), red (630-690 nm), red-edge (705-745 nm), near-infrared (770-895 nm), and panchromatic (450-800 nm). Additionally, the normalized difference vegetation index (NDVI) was created with the red and near-infrared bands. Two datasets were analyzed: 1) two-class case (original treated and non-treated spectral data) and 2) five-class case (original treated and non-treated spectral data and three additional classes created with the weighted average of the original treated and non-treated data). Data were evaluated with the Mann-Whitney U-test, Kruskal Wallis test, and the Mann-Whitney multiple comparison test with Berferonnic correction. The clustering algorithm referred to as CLUES (CLUstEring based on local Shrinking) was employed to group the data into clusters; cluster validation was determined with the average silhouette width; accuracy was assessed with contingency matrixes. Statistically significant differences (p = 0.05) between treated and non-treated plants spectral properties were observed for the red, red-edge, and near-infrared bands and the NDVI at both sites for the two- and five-class datasets. Clustering analysis worked well in dividing the data into appropriate groups, with the best cluster structure occurring for the NDVI. Specificity and sensitivity values for the NDVI were greater than 0.90, indicating an excellent classification. The immediate benefit of this research has been to provide improved information on the application of broadband spectra and unsupervised classification for surveying thermally defoliated cotton fields and further support studies to evaluate aircraft or high resolution satellite systems to monitor this process for preparing cotton for harvest.

Last Modified: 10/22/2014
Footer Content Back to Top of Page