DETECTING CROP WATER STRESS IN COTTON (GOSSYPIUM HIRSUTUM) UNDER HUMID GROWING CONDITIONS

Authors: Sassenrath, Gretchen; ALARCON, V - GRI

Submitted to: American Society of Civil Engineers
Publication Type: Abstract Only
Publication Acceptance Date: 2/15/2006
Publication Date: 5/21/2006
Citation: Sassenrath, G.F., Alarcon, V.J. 2006. Detecting crop water stress in cotton (gossypium hirsutum) under humid growing conditions. World Environmental and Water Resources Congress, May 21-26, 2006. Omaha, NE. American Society of Civil Engineers, CD-ROM

Interpretive Summary: Remote imagery offers the potential for rapid, timely, and continuous monitoring of crop status, and has the potential to offer significant benefit to crop production and management. However, errors are introduced in moving from the well-characterized measurements of individual leaves to interpretation of factors operating in an entire production canopy. For example, while the water potential of an individual leaf can be well-characterized through standard measurement procedures, extrapolation of water potential to an entire canopy element is difficult. Moreover, identification of the onset of water limitations in plant tissues is exacerbated by changes in leaf morphology and orientation that alter the apparent reflectance. We are developing a timely, accurate, and simple system for detecting the onset of water stress in cotton (Gossypium hirsutum). Leaf and canopy reflectance in the visible and near to mid-infrared range is correlated with plant physiological parameters and soil moisture levels. While late-season measurements demonstrate differences between reflectance and temperature in the irrigated and non-irrigated canopies, a significant limitation with reflectance and temperature measurements is interference from soil reflectance early season prior to canopy closure. This complicates determination of the onset of water stress, and accurate timing of the first irrigation. To understand the factors contributing to the total canopy reflectance, we have developed a methodology of building synthetic images from reflectance spectra of individual leaves. The individual factors impacting canopy reflectance can then be examined separately. The method initially approximates leaf spectral characteristics through the use of Fourier transformations that adequately describe the spectra and reduces the information required from 512 bands to 20 pairs of harmonics. These spectra from soil and vegetation are then randomly mixed and visualized using an imaging software package. The resultant synthetic spectra are remarkably similar to actual remote images of crop fields obtained from aerial sources. This methodology offers a powerful analytical technique to explore crop reflectance spectra, and build spectral libraries of reflectance patterns with distinct physiological and morphological changes.

Technical Abstract: Good irrigation management requires a timely, accurate assessment of crop water needs. While thermal systems have been used in arid environments, the high humidity in the Mid-South limits the potential evaporative cooling of the crop. Moreover, the frequent cloud cover interferes with image collection from aerial sources. We are developing a system of detecting the onset of water stress in cotton (Gossypium hirsutum). To determine the physiological response of cotton to limiting water, we must first reduce the impact of other factors which interfere with the canopy reflectance. Remote sensing systems generate large data sets from measured hyperspectral reflectance spectrum. Analysis and interpretation of these data sets is hindered by the quantity of data available. We have developed methods of simplifying the large data sets, without losing any of the information contained in the data. The process of Fourier transformation allows condensing the information from 512 bands to 20 pairs of harmonics. These reduced data sets can then be manipulated and analyzed in ways to extract the most significant components. Visualization of the reflectance spectra is aided by a process of synthetic image generation. These synthetic images look very similar to remote images containing much less information. The analytical and visualization tools are powerful methods for robust analysis and interpretation of the reflectance data.