Canopy Temperature as Indicator of Water Stress: Image analysis of grapevine canopies.

Canopy temperature has been proposed in field crop and horticulture industries as an indicator of plant water stress, using absolute and/or relative indices. Infrared thermographers are now available at relatively low cost but integrate exposed and shaded leaves in their fields of view. Shaded leaves have a different temperature distribution than exposed leaves. Separation of the exposed and shaded fractions in thermal datasets is critical for subsequent analysis, particularly when canopies display a spherical leaf angle distribution. Various classification and warping algorithms available in the image analysis software ENVI 4.2 were assessed to separate the exposed and shaded fraction of field grown grapevines canopies (Vitis vinifera L. cv Cabernet Sauvignon). The Parallelepiped algorithm with User Defined mix of standard deviation and the polynomial (1st degree) warping algorithm are recommended for the classification and warping of visible images respectively. The mean temperature of the various fractions was significantly different from each other as well as from the combined regions' mean. The Regions of Interests (ROI) defined to classify the images displayed good separability, suggesting the possible automation of the process. An analysis of Red, Green and Blue (RGB) bands and Hue, Saturation and Value (HSV) transformed bands of the visible images showed that HSV images were less sensitive than RGB images to the factors affecting the capture of data by the camera, and should therefore be preferred. These results did not however support the suggestion that spectral libraries or endmember collections may be used to automate the classification procedure. A calibration methodology involving points of known RGB and/or HSV is proposed to reduce the effect of the various factors is discussed for future automated procedures.