Skip to main content
ARS Home » Research » Publications at this Location » Publication #281827
Title: Using hyperspectral reflectance data to assess biocontrol damage to giant salvinia

Author
item EVERITT, JAMES - Retired ARS Employee
item Yang, Chenghai
item SUMMY, KENNETH - Texas-Pan American University
item NACHTRIEB, JULIE - Us Army Corp Of Engineers (USACE)

Submitted to: Geocarto International
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/22/2012
Publication Date: 11/6/2013
Citation: Everitt, J.H., Yang, C., Summy, K.R., Nachtrieb, J.G. 2013. Using hyperspectral reflectance data to assess biocontrol damage to giant salvinia. Geocarto International. 28(6):502-516.

Interpretive Summary: Giant salvinia is an exotic, invasive fern that often invades and clogs waterways in subtropical and tropical areas of the world. The salvinia weevil has been used extensively for controlling giant salvinia. Field hyperspectral reflectance data were studied at 50 wavebands (10 nm bandwidth) over the 400 to 900 nm spectral range to determine their potential for distinguishing among giant salvinia plants subjected to four population levels of salvinia weevils to develop feeding damage to the plants. The four populations included a control with no insects and those with low, medium, and high insect populations. The plants were studied in two experiments on each of two dates: October 14, 2010 and July 21, 2011. Two procedures were used to determine the optimum bands for discriminating among treatments: least significant difference (LSD) and stepwise discriminant analysis. LSD comparison test results for both October and July experiments showed that generally the best bands for separating among treatments occurred in the green, red, red-NIR (near-infrared) edge, and NIR regions where three to four treatments could be distinguished. Stepwise discriminant analysis identified four bands in the green, red and red-NIR edge to be significant to discriminate among the four treatments in experiment 1 in October. For experiment 2 in October, discriminant analysis identified five bands in the blue, green, red, and NIR regions to be significant. In experiment 1 in July, five bands in the blue, green, red-NIR edge, and NIR regions were found to be significant. For experiment 2 in July, discriminant analysis identified four bands in the blue, green, and red-NIR edge regions to be significant. These results should be useful to wetland resource managers and weed scientists interested in assessing control of this invasive weed over large and inaccessible areas.

Technical Abstract: Field hyperspectral reflectance data were studied at 50 wavebands (10 nm bandwidth) over the 400 to 900 nm spectral range to determine their potential for distinguishing among giant salvinia (Salvinia molesta Mitchell) plants subjected to four population levels of salvinia weevils (Cyrtobagous salviniae Calder and Sands) in order to develop four levels of vegetative damage to the plants. The four populations included a control with no insects and those with low, medium, and high insect populations. The plants were studied in two experiments on each of two dates: October 14, 2010 and July 21, 2011. Two procedures were used to determine the optimum bands for discriminating among treatments: least significant difference (LSD) and stepwise discriminant analysis. LSD comparison test results for both October and July experiments showed that generally the best bands for separating among treatments occurred in the green (505-595 nm), red (605-635 nm), red-NIR (near-infrared) (695-745 nm) edge, and NIR (755-895 nm) regions where three to four treatments could be distinguished. Stepwise discriminant analysis identified four bands in the green, red and red-NIR edge to be significant to discriminate among the four treatments in experiment 1 in October. For experiment 2 in October, discriminant analysis identified five bands in the blue, green, red, and NIR regions to be significant for distinguishing among the treatments. In experiment 1 in July, five bands in the blue, green, red-NIR edge, and NIR regions were found to be significant to discriminate among the treatments. For experiment 2 in July, discriminant analysis identified four bands in the blue, green, and red-NIR edge regions to be significant to discriminate among the treatments.