Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: January 12, 2007
Publication Date: May 30, 2007
Citation: Detar, W.R., Penner, J.V. 2007. Airborne remote sensing used to estimate percent canopy cover and to extract canopy temperature from scene temperature in cotton. Transactions of the ASABE, Vol. 50(2):495-506. Interpretive Summary: It is well known that cotton not receiving the right amount of water at the right time becomes stressed, resulting in decreased yields and fiber quality. Currently, the methods used to determine water stress are laborious and time consuming. Knowing the degree of plant water stress over an entire field would be beneficial to scientists and engineers who design and operate irrigations systems. When plants are stressed, leaf temperatures rise and the amount of light reflected from the leaf surface changes. The purpose of this three-year study was to determine the best way to calculate the stress level of plants across an entire field. Cameras capable of measuring leaf temperature and the amount and quality of light reflected from a field were mounted in an airplane and flown at an elevation of 5,000 ft over cotton fields experimentally stressed by applying different amounts of water. Data from the cameras and from measurements on the ground were analyzed with sophisticated image analysis and statistical software and a mathematical formula was discovered capable of determining the difference between stressed and unstressed plants. Of particular interest in this study is the solution to the problem of partial canopy where the temperature of hot soil is seen from the air as being mixed in with the cool plant temperature. This information can eventually be used for predicting when irrigation needs to be applied without having to go into a field to take multiple measurements.
Technical Abstract: Airborne remote sensing data, using hyperspectral (HSI) and thermal infrared cameras (TIR), were collected for three seasons, involving three different varieties of Acala cotton and four different experimental fields, for a total of 10 flights, under conditions of partial canopy cover. The TIR camera was used to detect the elevated canopy temperature that occurs when the plant is water stressed. However, when soil is exposed between the plant rows, the resulting TIR data is a mixture of hot soil and cool plants. A procedure was developed for separating these two elements of the scene temperature and it required a measure of the degree of ground cover. Percent canopy cover was measured manually in the fields and correlated to the HSI reflectance data. Linear multiple regression was used to find the best-fitting wavelengths in the range of 429 to 1010 nm for one-,two-, three- and four-parameter HSI models, and produced some new vegetation indices. The best-fitting two-parameter wavelengths were centered at 676 nm and 966 nm, and worked quite well with a coefficient of correlation of 0.92. The normalized difference vegetation index (NDVI) was modified by putting extra weight on the red term, and provided a better fit than an un-modified NDVI. The TIR data was then correlated to the percent canopy cover using covariance. The temperature of the plant canopy was found by projecting all the data points to what they would have been at 100% canopy cover. This covariance procedure also provided the temperature difference between treatments and the statistics needed to show if the difference was significant. Using the difference in the green and red wavelengths, it was possible to make a high-resolution image map of the plant water stress throughout a field. The main finding was that plant water stress in Acala cotton at partial canopy can indeed be detected with airborne remote sensing.