Using remote sensing to assess stand loss and defoliation in maize
Abstract
Accurate assessment of hail and wind damage to crops has long been a difficult task. Yield effects of stand loss and defoliation at various crop stages are well known, but getting accurate damage information across a field has been problematic. A quick and accurate method of determining losses such as using remote sensing techniques could lead to more efficient crop management and more accurate insurance claims adjustment. New satellites with spatial and temporal resolution capable of assessing crop damage could provide data for producing maps of field damage. Field plots of maize, Zea mays L. were established in Indiana and Nebraska in 1997, 1998, and 1999 and were subjected to various levels of stand loss and leaf loss during vegetative and early reproductive growth stages. Remote sensing data were collected using a variety of imaging spectroradiometers mounted in airplanes and field spectrometers mounted on boom trucks. Radiometric data compiled by wavelength to produce spectral response curves showed light reflectance level directly related to damage level across a broad range of wavelengths. Using feature selection, a function of the image processing software MultiSpec, the red and near infrared (NIR) wavelength bands provided the most distinction among levels of damage. Classification of images by damage level was accomplished using a combination of red and near infrared bands and using Discriminate Analysis Feature Extraction (DAFE) to utilize the contribution of all sensor bands and attempt high accuracy. Classification accuracies achievable for discriminating four levels of damage inflicted on a particular date ranged from 72% to 100%, with accuracy depending on the band combination used and level of damage. In a field after a storm, remote sensing images could be used to extrapolate information collected by insurance adjusters, and produce damage maps of a field. Fewer ground checks would be needed to get an accurate damage estimate.
Degree
Ph.D.
Advisors
Johannsen, Purdue University.
Subject Area
Agronomy|Remote sensing
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.