AIRBORNE REMOTE SENSING USED TO ESTIMATE PERCENT CANOPY COVER AND TO EXTRACT CANOPY TEMPERATURE FROM SCENE TEMPERATURE IN COTTON

摘要

The goal of this research was to separate the soil and plant temperatures and create an image map of plant water stress. Data from hyperspectral imagery (HSI) and thermal infrared (TIR) sensors were collected using an airborne platform over three seasons, involving three different varieties of Acala cotton, four different fields, and a total of ten flights. The first step was to measure the percent canopy cover, which ranged from 30% to 100%. Using linear multiple regression, percent canopy cover, measured manually in the field, was found to be closely related to several new vegetation indices, taken from among 60 narrow bands in the wavelength range of 429 to 1010 nm. The highest coefficient of determination (r{sup}2) for a three-parameter hyperspectral model was 0.931, and it included the wavelengths 676, 753, and 773 nm. A two-parameter model using 676 and 966 nm worked especially well. A weighted version of the normalized difference vegetation index (NDVI) was found to relate well to percent canopy cover, but not quite as well as some non-normalized band combinations. Using the two-parameter model, the percent canopy cover was calculated for every part of two experimental fields that had originally been set up to compare the yields from water-stressed versus unstressed treatments. The mean value for scene temperature for each plot was plotted against the mean value of percent canopy cover for each plot. Using analysis of covariance, the scene temperatures were projected to what they would be at 100% canopy. The procedure showed that the canopy for the water-stressed treatment had a significantly higher temperature than the unstressed control, which means that it was indeed stressed. Using analysis of covariance, the green-red difference was found to be an indicator of both percent canopy cover and plant water stress. An image map was produced showing the canopy temperature at every pixel in the field, with a spatial resolution of about 1.0 m. The main finding was that the plant water stress in Acala cotton could be detected with airborne remote sensing under the conditions of partial canopy over a dry soil surface. These results should be useful in selecting filters for multispectral cameras and for selecting the wavebands for HSI sensors when attempting to measure degree of vegetative cover. A straightforward method is presented for separating canopy temperature from soil temperature, and a procedure is given for producing a detailed map of canopy temperature in the field.