|Ge, Shaokui - UC SANTA CRUZ|
|Xu, Ming - RUTGERS UNIVERSITY|
Submitted to: Weed Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: July 12, 2007
Publication Date: November 1, 2007
Repository URL: http://hdl.handle.net/10113/16927
Citation: Ge, S., Xu, M., Anderson, G.L., Carruthers, R.I. 2007. Estimating Yellow Starthistle (Centaurea solstitialis) Leaf Area Index and Aboveground Biomass with the Use of Hyperspectral Data. Weed Science.55:671-678. Interpretive Summary: Yellow starthistle, Centaurea solstitialis, is one of the worst invasive weeds in the state of California, infesting over 22 million acres. It also is a major weed pest in the adjacent states of Oregon, Washington, Idaho and Nevada. It causes losses in forage quality and quantity, reduces livestock production and negatively affects wildlife viability. In severe infestations, it can cause brain lesions in horses that may lead to death. USDA-ARS in cooperation with the California Department of Food and Agriculture and the University of California are working together to develop and implement a variety of integrated management strategies to help control this pest species. One of the primary needs is to make areawide assessment of yellow starthistle infestation levels which is both time consuming and expensive when conducted using ground crews of monitoring personnel. This study reports on the use of hyperspectral remote sensing that was successfully used to estimate yellow starthistle above ground biomass and leaf area index. New quantitative models link specific reflectance patterns to actual ground based measurements, the results showing reasonable detection capabilities for this technology. With additional research on detection within mixed YST and grass canopies, we believe that this technology will be effective in providing land managers with a new tool to assess YST infestations across entire watershed, thus helping in directing larger-scale strategic approaches to its control.
Technical Abstract: Hyperspectral remote-sensed data were obtained via a Compact Airborne Spectrographic Imager-II (CASI-II) and used to estimate leaf-area index (LAI) and aboveground biomass of a highly invasive weed species, yellow starthistle (YST). In parallel, 34 ground-based field plots were used to measure aboveground biomass and LAI to develop and validate hyperspectral-based models for estimating these measures remotely. Derivatives of individual hyperspectral bands improved the correlations between imaged data and actual on-site measurements. Six derivative-based normalized difference vegetation indices (DNDVI) were developed; three of them were superior to the commonly used normalized difference vegetation index (NDVI) in estimating aboveground biomass of YST, but did not improve estimates of LAI. The locally integrated derivatives-based vegetation indices (LDVI) from adjacent bands within three different spectral regions (the blue, red, and green reflectance ranges) were used to enhance absorption characteristics. Three LDVIs outperformed NDVI in estimating LAI, but not biomass. Multiple regression models were developed to improve the estimation of LAI and aboveground biomass of YST, and explained 75% and 53% of the variance in biomass and LAI, respectively, based on validation assessments with actual ground measurements.