• K. Raja Reddy
• Lee Tarpley
• Gopal Kakani
• Duli Zhao

• John Read and James McKinion
• USDAARS
• Roger King and Lori Bruce
• MSU-ERC
• Alex Thamosson
• MSU-ABE

Characterization and alleviation of crop stresses on a site-by-site basis may be the way to realize the yield potential of crops. Current technologies aiding site-specific management practices include hyperspectral imagery, GIS and GPS systems with variable rate controllers, yield monitors, and real time weather driven GIS-based mechanistic crop expert systems. However, diagnostic remote-sensing tools are needed to identify and quantify potential yield and quality damage as affected by environmental crop stresses (e.g. water, nutrients, growth regulators, UV radiation). Our goal is to develop functional algorithms for operational use in remote sensing applications. Understanding the physical/physiological relationships between leaf/canopy chemistry and reflectance/hyperspectral imagery is essential for the development of reliable spectral algorithms.

• Develop algorithms for early and sensitive detection of nutritional (nitrogen, phosphorous, potassium and CO2) and water deficiencies in major row crops (cotton, soybean, corn and sorghum) grown in Mississippi.
• Establish initial algorithms, for each of the environmental stresses, through intensive monitoring of hyperspectral reflectance at leaf and canopy level, composition and growth of plants under field-like controlled environmental conditions (SPAR and pot-culture facility).
• Test/validate the algorithms for direct detection of foliage components in the presence of multiple stresses in controlled environment and field plot studies.
• Provide functional spectral algorithms for automated analysis of airborne multispectral/hyperspectral images of the visible to SWIP radiance reflected from crop canopies.

1. Soil-Plant-Atmosphere-Research (SPAR) Facility: The SPAR Facility (Fig. 1) will be used to provide an understanding of the relationships among physiological processes, foliage composition and hyperspectral imagery data needed for algorithm development. The SPAR facility allows for growth of plants under a natural radiation environment, but with provision for manipulation of the soil matrix, soil solution, atmospheric composition, and chamber temperature. The plants will be grown in small plots in typical row spacing. Of critical importance to this project, canopy-level photosynthesis and evapotranspiration can be monitored nearly continuously with simultaneous and continuous monitoring of soil and aerial environmental conditions. When the foliage components are also frequently determined, then canopy-level biophysical processes provide a means for defining the critical values of particular components the critical leaf N differs between crop growth and developmental processes and differs among physiological functional groups (C3 vs. C4). These determinations allow us to emphasize the range of concentrations of physiological/economic importance when developing spectral algorithms. We will monitor temporal trends of the visible through SWIR spectrum of reflected radiance of reference, controls, and treatments in the SPAR facility. We expect to obtain a unique dataset of intensively monitored changes in reflected radiation from vegetation, which can be combined with information from near-continuous canopy photosynthesis, transpiration and respiration along with monitoring changes in foliage composition and cover due to stress conditions. As the radiation environment and physiology of the plant change continuously, knowledge of the target independent of radiation or clouds will provide unique and rich information that can be used to understand hyperspectral imagery and physiology, and develop image analysis tools, and to validate the algorithms.
   
   


2. Sand-culture potted-plant array: Plants will be grown in field-like spacing with natural exposure to weather, but good control of pests and diseases and provision of nutrients and water are provided (Fig. 2). Treatments will be imposed by manipulation of computer-controlled fertigation, typically by manipulation of the nutrient solution or the daily extent of nutrient supply. This set-up allows subjection of plants to essentially single nutrient or water stresses, while allowing periodic removal of plants for detailed analysis of growth and development, and further allowing canopy-level collection of hyperspectra and narrow-band images. The canopy-level, and leaf-level spectra will be collected daily to biweekly depending on the particular treatment, and compared to foliage concentration of the particular macronutrient or water potential collected at the same time along with steady-state measurements of leaf photosynthetic gas exchange. Also, a subset of the plants will be destructively harvested weekly for evaluation of growth and development. The potted-plant array will allow us to develop/validate spectral algorithms for accuracy and precision.
   
   


3. Small research plots on the North Farm and in the field: Algorithms developed under the controlled environmental conditions will be tested in the production field with multi-stress environmental conditions (Fig. 3). The treatments will be used are in various combinations to include: several levels of nitrogen and potassium fertility rates, and irrigation and growth regulators. Also, MAFES cotton variety-trial experiments will be used to investigate, if any, cultivar-specific spectral/physiological differences. These treatments will simulate several cultural practices that are being used in the cotton production. Evaluations are being underway to collect weekly near-canopy hyperspectra collection with an ASD FR and collection of foliage samples for analysis of macronutrient and leaf water potential. In addition, post-hoc analysis of hyperspectral/multispectral imagery will be used to evaluate the predictive ability of airborne-platform information to estimate the concentrations of the foliage components obtained near the time of the flyover.
   
   

Data collected will be analyzed to develop algorithms for individual stress factors. A variety of statistical techniques including individual band ratios, partial least squares, stepwise multiple linear regression and wavelet analysis, and published band ratios are being used to derive the algorithms.