REMOTE STORAGE m . Jllliflllllllllllllfllllllllllfllllll l 3 1293 02074 1041 LEBHARY Michigan State University This is to certify that the thesis entitled REMOTE SENSING OF LEAF TISSUE NITROGEN CONTENT AND DISEASE SEVERITY IN CREEPING BENTGRASS AND ANNUAL BLUEGRASS USING NEAR INFRARED SPECTROSCOPY presented by GEOFFREY JORDAN RINEHART has been accepted towards fulfillment of the requirements for M.S. Crop and Soil Sciences 4' degree in Major professor Date 3/11/01) 07639 MS U is an Affirmative Action/Equal Opportunity Institution REMOTE STORAGE RSI: PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE our: DATE DUE IA“! 0 7 21313 2/17 20:: fine Fofiw S/DateDueForms_20I7.indd - pgs REMOTE SENSING OF LEAF TISSUE NITROGEN CONTENT AND DISEASE SEVERITY IN CREEPING BENTGRASS AND ANNUAL BLUEGRASS USING NEAR INFRARED SPECTROSCOPY By GEOFFREY JORDAN RINEHART A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Crop and Soil Sciences 2000 ABSTRACT REMOTE SENSING OF LEAF TISSUE NITROGEN CONTENT AND DISEASE SEVERITY IN CREEPING BENTGRASS AND ANNUAL BLUEGRASS USING NEAR INFRARED SPECTROSCOPY By Geoffrey Jordan Rinehart Site-specific application of nutrients and pesticides based upon the specific needs of turfgrass plants has the potential to save money and reduce the potential threat of polluting the environment. The objectives of this study were to develop a method to determine N content of leaf tissue and disease status of brown patch (Rhizoctonia solani Kuehn) and dollar spot (Sclerotinia homeocarpa Bennett) on creeping bentgrass (Agrostis stolonifera Huds.) and annual bluegrass (Poa annua var. reptans Hausskn) using a direct light visible/near (VIS-NIR) infrared scanning monochromator. Nitrogen was applied at rates of 0, 1.2, 2.4, 3.6, and 4.8 g N/m2 periodically over two growing seasons to creeping bentgrass and annual bluegrass mowed at heights of 5 mm and 14 mm. Absorbance was expressed as “log l/reflectance” between 400 and 2500 nm once color differences were evident. After spectrometer readings were attained, clippings were harvested from each plot and analyzed for N using a dry combustion analyzer. Modified partial least squares regression analysis using the wavelengths from the entire spectrum demonstrated a relationship between leaf tissue N content and canopy reflectance (r2: 0.78-0.92). Wavelengths which illustrated the best association between lab values. for the raw spectrum occurred at wavelengths 670, 1450, and 1930 nm and correspond to chlorophyll a transmission, a primary overtone O-H stretch attributable to water, and an O-H stretch attributable to water, lignin, protein, nitrogen, and starch, respectively. Brown patch and dollar spot are two common diseases of cool season turfgrass in the United States. As governmental and public scrutiny of golf course maintenance practices increases, superintendents are beckoned to balance playability with fewer fungicide inputs. Categorical disease symptom severity ratings of brown patch and dollar spot were made on different turfgrass swards and associated spectra obtained. Discriminant analysis of the data yielded categorical accuracy. In the dollar spot study, 20 out of 193 samples (10.3%) were classified incorrectly using categories associating spectra with diseased areas, areas close to the disease that appeared healthy, and healthy areas away from the disease symptoms. In the brown patch study there were only 29 misses out of a total of 336 samples (8.6%) using three classification categories consisting of severe and medium disease and healthy areas. These results suggest the feasibility of developing a VIS-NIR sensor for the detection of disease severity. Future research should address how various stresses interact to affect the spectral reflectance of the turfgrass plant. These results indicate the potential for developing a real-time remote sensor for site specific nutrient and fungicide applications in turfgrass management. ACKNOWLEDGEMENTS Without the assistance of many pe0ple along the way, this product would not have been possible. For the guidance I received during this project, I express appreciation to my major professor, Dr. James Baird, and committee members, Dr. Joseph Vargas and Dr. Gene Safir. I would like to express thanks to Ron Calhoun, Dr. Baird’s technician; Nancy Dykema and Ron Detweiler for their assistance during the disease project; Frank Roggenbuck for help during the greenhouse experiments, and Joe Kroening at the Tom Company for assistance in the nitrogen analyses. For help concerning the statistical analysis and interpretation much gratitude goes to Dr. John Shenk and Dr. Mark Westerhaus at Infrasoft International; Dr. Syed Dara at the Toro Company; Dr. Oliver Schabbenberger, formerly at Michigan State; and Dr. Carl Ramm, Department of Forestry. Special thanks are certainly in orderto all of my fellow graduate students in the Departments of Crop and Soil Sciences, Botany and Plant Pathology, and Horticulture. I would like to especially recognize graduate students in the turfgrass group and those in Dr. Baird’s lab, Beau McSparin, Ryan Goss, and Susan Redwine. A word of appreciation goes our undergraduate workers, Stephanie Dysinger and Dan Lamb. Finally, and by no means least, I would like to thank my family: my father, Douglas; my mother, Vivian; my sister, Christen; and my brother, Stuart. iv TABLE OF CONTENTS PAGE LIST OF TABLES viii LIST OF FIGURES x CHAPTER ONE: INTRODUCTION AND LITERATURE REVIEW 1 INTRODUCTION 1 LITERATURE REVIEW 4 Nitrogen Uses in the Turfgrass Plant 4 Nitrogen Cycling in the Plant Community 5 Nitrogen Assimilation by the Plant 6 Brown Patch (Rhizoctonia solani) 10 Dollar Spot (Sclerotinia homeocarpa) ‘ 11 Properties of Light 12 Near Infrared Spectrum 16 Spectroscopy 19 General NIR Applications 20 Data Analysis 22 Applications of Spectroscopy to Site Specific 26 Management Applications of Spectroscopy for Disease Sensing 29 CHAPTER TWO:REMOTE SENSING OF LEAF TISSUE NITROGEN ' CONTENT AND IN CREEPING BENTGRASS AND ANNUAL BLUEGRASS USING NEAR INFRARED SPECTROSCOPY ABSTRACT INTRODUCTION MATERIALS AND METHODS Turfgrass Culture Nitrogen Application Spectrometer Measurements Clipping Collection Nitrogen Analysis CENTER and Principal Component Analysis Calibration Equation Development Lab Value Predictions RESULTS AND DISCUSSION Laboratory Reference Values Visible-Near Infrared Reflectance Spectra and Predictions Inter-population Predictions CONCLUSIONS LITERATURE CITED vi 31 31 32 35 35 35 36 37 37 39 39 41 42 42 55 58 62 CHAPTER THREE: REMOTE SENSING OF DISEASE SEVERITY IN CREEPIN G BENTGRASS AND ANNUAL BLUEGRASS USING NEAR INFRARED SPECTROSCOPY ABSTRACT INTRODUCTION MATERIALS AND METHODS Spectrometer Measurements Data AnalySis RESULTS AND DISCUSSION Dollar Spot Study Brown Patch Brown Patch v. Dollar Spot CONCLUSIONS LITERATURE CITED BIBLIOGRAPHY vii 65 65 66 68 69 70 72 72 72 78 78 80 81 TABLE 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.1 3.2 3.3 ' ’ LIST OF TABLES TITLE Laboratory values of nitrogen (N) content in turfgrass clippings. Calibration statistics for ‘Penncross’ green combined over both seasons. Calibration statistics for ‘Penncross’ fairway combined over both seasons. Calibration statistics for ‘Providence’ fairway combined over both seasons. Calibration statistics for Poa annua green combined over both seasons. Calibration statistics for Poa annua fairway combined over both seasons. Calibration statistics for all populations combined over both seasons. Statistics for predicting between turfgrass populations for all data combined. Statistics for predicting subsets with equation developed from remaining global population. Predicted v. actual category classification for dollar spot spectra Predicted v. actual category classification for brown patch spectra using four categories. Predicted vs. actual category classification for brown patch spectra using three categories. viii PAGE 43 46 46 46 47 47 47 56 59 73 74 74 FIGURE 1.1 1.2 1.3 2.1 2.2 2.3 2.4 2.5 2.6 3.1 3.3 LIST OF FIGURES TITLE Schematic representation of the sequence of nitrate assimilation in leaf cells. (Adapted from Marshner, 1995). Model of ammonia assimilation pathways (1,2) Glutamine synthetase-glutamate synthase pathway, with Low NH3 supply (1) and with high NH3 supply (2). (3) Glutamate dehydrogenase pathway. (Adapted from Marshner, 1995). The electromagnetic spectrum. Spectra comparison of low range (<3% N), middle range (3.2-4.0% N), and upper range (>4.5%) average spectra across all measured wavelengths. Histogram of H distances from the mean spectrum for all samples. Actual vs. Predicted scatter plot of all samples in global calibration equation. (Solid line indicates slope = 1.00; dashed line indicates actual slope = 0.998; dotted line indicates 95% confidence intervals). Raw spectrum comparison of the average spectrum of five populations. First derivative comparison of the average spectrum of five populations. Composite graph of loading spectra for the first 6 eigenvector terms used in the global equation. Average spectrum comparison of “dollar spot” and “healthy.” Raw spectra comparison for three brown patch disease categories. First derivative spectra comparison of three categories of brown patch disease. PAGE 13 45 48 50 52 75 76 77 AOTF C: N FAD GIS GPS GOGAT LA] MSR MPLSR NAD MoCo SECV SEC SED SEP SSM TCA UAN USGA VRT LIST OF ABBREVIATIONS Accoustical optical tunable filter Carbon to nitrogen ratio Flavine adenine dinucleotide Geographic information systems Global positioning system Glutamate synthase Infrared Potassium Leaf area index Modified stepwise regression Modified partial least squares regression Nitrogen Nicotinamide adenine dinucleotide Mid infrared Molybdenum cofactor Normalized Difference Vegetation Index Nitrite reductase Near-infrared Near-infrared spectroscopy National Institute of Standards and Testing Nitrate reductase Nitrogen use efficiency Phosphorous Photosynthetically active radiation . Principal component analysis Partial least squares Precision Turf Management Plant Nitrogen Spectral Index Pre-sidedress nitrogen test Reflectance Coefficient of determination Random complete block design Root mean square Standard deviation Standard error of cross validation Standard error of calibration Standard error of the difference Standard error of performance Site Specific Management Tricarboxylic acid Urea ammonium nitrate United States Golf Association Variable rate technology CHAPTER ONE INTRODUCTION Pesticides and fertilizers are an integral part of golf course management today as golfers expect a high level of course maintenance and playability. Accompanying this phenomenon is the increased‘potential for these inputs to have detrimental environmental impact if applied without educated decisions about the needs of the turfgrass ecosystem. As golfers’ expectations increase, golf course superintendents are forced to balance A course playability with environmental considerations. Increasing public and governmental scrutiny will continue to put a premium on a superintendent’s ability to use necessary inputs judiciously. In light of this, it is important that fertilizer and pesticide resources be used responsibly to both reduce environmental impact and maintain a reasonable tiu'fgrass quality. Site specific management (SSM) or Precision Turf Management (PTM) refers to the practice of assessing a property’s variability and adjusting management practices accordingly. Site variability can be affected by a number of factors including soil texture and fertility, terrain, slope and aspect, mowing height, drought stress, disease pressure, turfgrass species and cultivar composition, and by environmental factors such as light quality and intensity and air flow characteristics. The four primary components of SSM involve the global positioning system (GPS), geographic information systems (GIS), sensing, and variable rate technology (VRT). The GPS refers to a collection of 24 orbiting satellites which are oriented circumspherically about the earth and were originally established for military navigation purposes. A GPS receiver communicates via radio signal with appropriate satellites and the distance fiom the satellites. to the reciever is calculated. Using trigonometric principles, the reciver’s exact location can be determined and described in coordinates of latitude and longitude. The precision of the transmitter measurements varies according to sophistication and cost. Current technology allows precision down to millimeter increments. Sub-meter resolution would be required for practical application on golf courses, which require greater precision than production agriculture. Geographic information systems (GIS) refers to any of a number of computer software programs which integrate information about site variability into a visual format, typically in the form of a map. It provides a method by which spatial information may be captured, stored, analyzed, displayed, retrieved and overlaid (Krzanowski et al., 1992). Geographic information systems allow a manager to overlay maps containing information about various parameters of interest and graphically observe relationships that may exist among the parameters. A cost-effective process for acquiring spatial information is currently the most limiting aspect of SSM in the realm of turfgrass science. Real-time sensing is a component of precision management which is necessary in order to collect a large volume of data efficiently, quickly, and relatively inexpensively and is essential to developing the full potential SSM. A sensor based upon reflectance from the canopy could provide a cost- and labor-effective strategy for assessing turf leaf N content and disease symptoms. The information can be geographically referenced with GPS and assimilated with GIS. Based upon the sensor data, a spray vehicle equipped with a manifold of variable- output nozzles can vary the application rate of an input such as a fertilizer or pesticide. The efficient use of chemical inputs on golf courses will help decrease environmental impact. Variable rate technology (VRT) is the process of adjusting the rate of applied inputs according to the assessed needs of the plant. Information acquired in real-time can be processed so that appropriate spray applications are conducted and referenced using GPS and GIS. The goal of sensor-based VRT is “to instantaneously adjust application rates based on sensor measurements of fertility [or other factors] as an applicator travels across the field.” (Stone et al., 1993). Effective use of this technology will sponsor precise applications of inputs needed to retain turfgrass quality and reduce the total amount of inputs needed. The sensing aspect of SSM is the focus of this research and involves scanning turf with a spectrometer which is able to detect reflectance of the turf canopy in the range of 400 to 2500 nm. The objectives of this research were .to: 1) determine if an association exists between leaf N content and reflectance fi'orn the canopy; 2) determine how the relationship is affected by turfgrass species or cultivar, mowing height; and soil type; 3) to establish a spectral signature characterizing the presence of Rhizoctonia solani and Sclerotinia homeocarpa on turf. LITERATURE REVIEW NITROGEN USES IN THE TURFGRASS PLANT Nitrogen (N), potassium(K) and phosphorous(P) are referred to as macronutrients because they are the mineral nutrients required in the greatest amounts for proper plant nutrition, excluding atmospheric elements carbon, oxygen, and hydrogen which are intrinsic to many plant biochemical functions (Marshner, 1995). Nitrogen is required by the plant for the production of amino and nucleic acids, enzymes, and proteins and the proper functioning of chlorophyll (Epstein, 1972). Although 78% of the atmosphere is composed of N, atmospheric N is not available to turfgrass because of the diatomic molecule’s high triple. bond energy. Nitrogen is present in many forms, but nitrate (N03) and ammonium (NH?) are the major sources utilized for plant uptake. These forms of N are produced by aerobic microOrganisms decomposing organic matter or by the input of synthetic fertilizers. Symptoms of N deficiencies include shoot strmting, decreased tillering, and development of chlorosis symptoms in older tissue because N is phloem- mobile (Marshner, 1995). Turfgrass typically contains 3-5% N by dry weight. Turfgrass N requirements depend on soil nutrient holding capacity, natural precipitation or irrigation, mowing height, traffic, and species or cultivar (Beard, 1982). Unlike other nutrients, there is no reliable test for soil N. Although rules of thumb are recognized as guidelines, ultimate N application decisions are subjective and based upon a manager’s experience with a particular turf (Turgeon, 1991; Beard, 1982). Sufficient N should be supplied to maintain density, adequate recuperation and shoot growth and color (Beard, 1982). Excessive N can contribute to excessive thatch, greater disease incidence, a restricted root system, lower recuperative capacity due to energy being allocated to aerial growth, and environmental stress tolerance on account of depleted carbohydrates (Beard, . 1982; Couch, 1995.). NITROGEN CYCLING IN THE PLANT COMMUNITY There are several fates of N applied to turf. Nitrogen can be taken up by the plant, stored in the thatch/soil, volatilized, denitrified or leached. Starr and Deroo (1981) reported that 19-27% of applied N may be immobilized in thatch. Relatively high N levels within thatch can sustain high microbial populations. Leaching (loss of N03'-N through the soil profile) is most prevalent with fast-release fertilizers and sandy soils. Volatilization refers to gaseous phase losses of N as ammonia; these losses increase with higher temperatures and relative humidity. Denitrification involves the reduction of nitrate and nitrite to nitric oxides and N2, The process occurs mainly in waterlogged or anaerobic soil conditions as microbes use nitrate as an electron acceptor instead of oxygen. Mineralization and immobilization are the two dominant processes involving N in soil organic matter turnover and are strongly affected by the carbonznitrogen (CzN) ratio of organic material present in a plant’s rootzone. Mineralization occurs as aerobic heterotrophic organisms conduct arninization and ammonification, converting organically-bound N to NHF’. Ammonification is the process where firngi, bacteria, and actinomycetes transform amino acids from organic matter into ammonia. Mineralization generally increases with increasing temperature and adequate moisture. Conversely, immobilization refers to the conversion of inorganic N to organic N and one of the main factors contributing to this is the ON ratio of the organic matter present. In a high C:N organic matter environment, microbes will use ammonium and nitrate from the soil and effectively immobilize it from use by plants. Subsequently, immobilized N can be mineralized with the addition of high N organic matter (T isdale et al., 1993). NITROGEN ASSIMILATION BY THE PLANT Ammonium assimilation begins with NIL;+ uptake into roots and ends with its incorporation into amino acids, amides, proteins and other nitrogen complexes. Upon plant uptake, either protons are released for charge compensation or anion uptake increases, depending on the soil ionic environment. Accordingly, roots are the primary site of assimilation since they can better dispose of excess protons than shoots. Uptake is optimal in neutral pH soils and decreases with an increase in acidity. Ammonium can be dissociated to ammonia (NH3) or directly assimilated into amino acids and amides in the root and subsequently amino acids in the shoot using carbon skeletons fiom the tricarboxylic acid (TCA) cycle (Marshner, 1995). The nitrate assimilation pathway is cornerstone to incorporating inorganic N into organic compounds. Contrasting ammonium uptake, high nitrate levels correspond with an increase in uptake of organic cations by the roots. Nitrate reduction can occur in roots and shoots. In low concentrations, a greater percentage of nitrate is reduced in the roots and with greater concentrations, more is translocated for reduction in the shoots. Maximum nitrate assimilation occurs when leaf expansion rate is high (Salisbury and Ross, 1992; Marshner, 1995). A As opposed to ammonium, nitrate must be reduced to NIL.+ in order to be incorporated into organic structures. Nitrate assimilation occurs via a specific transport system and involves a two-step reaction which is spatially separated: no; 9 N02' [Eq. 1] N02- + 6e' +8H*9 NH3 [Eq. 2] NO; + 8H“ +8e' -) NH3 + 21120 + OH [Eq. 3] The first reaction [Eq. 1] is catalyzed by Nitrosomonas bacteria and the second step [Eq. 2] is catalyzed by Nitrobacter bacteria. The electron donor in the processes is the compound NAD(P)H. Good correlation has been observed between light intensity and nitrate reduction, but it is unclear whether this is due to the increased light itself or confounded by the fact that there are a greater number of carbon skeletons into which additional fixed N could be assimilated (Marshner, 1995). Nitrate reductase (NR), located in the cell cytoplasm, is a dimer molecule composed of a heme group, FAD, and a molybdenum cofactor (MoCo) and is located in the cytoplasm. Nitrate reductase is regulated by enzyme synthesis and breakdown, reversible inactivation, and the concentration of the substrate present (Solomonson and Barber, 1990). Nitrite reductase (N iR) is located in chloroplasts and proplastids of roots and other non—green tissue (Fig. 1.1). Nitrite rarely accumulates as this step of the reaction is extremely rapid. Ferrodoxin is the primary electron donor in the reaction. Ammonia can be toxic in high concentrations, but is usually rapidly incorporated into organic compounds. Almost all ammonia produced by ammonium oxidation, nitrate reduction, and photorespiration is processed by the glutamate—glutamine synthesis pathway. With the addition of NH3, glutamate synthetase catalyzes the production of glutamine from glutamate. Light stimuli provide the impetus for 2-oxogluterate and glutamate to be exported from the stroma to the cytoplasm, thus aiding nitrate reduction and ammonium assimilation (Woo et al., 1987). Glutamine synthetase and glutamate synthase (GOGAT) are the two primary enzymes involved in ammonia assimilation. Glutamate synthase, facilitated by ferrodoxin or NADPH, catalyzes the transfer of -NH2 from glutamine to 2-oxoglutarate. This results in the production of two glutamate molecules, one of which can be used for maintenance in the cycle and one that can be used for biosynthesis of low molecular weight nitrogen compounds. When high amounts of ammonia are present, both glutamate molecules can accept ammonia molecules (Fig. 1.2). Glutamate and glutamine are used for the synthesis of amides, ureides, amino acids, peptides and high molecular weight compounds such as proteins. Glutamate can be used for amino acid synthesis by transarnination reactions which are catalyzed by arninotransferases located in the cytosol, chloroplasts, and other organelles. Carbon skeletons used for amino acid synthesis are obtained from photosynthesis, the tricarboxylic acid (T CA) cycle, and glycolysis reactions. Proteins are polypeptides constructed from amino acids and coupled by peptide bonds in a condensation reaction in cellular ribosomes. Glutamine and asparagine are the primary low molecular weight compounds produced by the pathway. Amino acids, amines, peptides, and ureides are also produced and are used for transient storage and long distance transport from roots to shoots. Images in this thesis are presented in color. Nitrate reductase Figure 1. 1. Schematic representation of the sequence of nitrate assimilation m leaf cells. (Adapted from Marshner,1995). OOOH ; . T .__/ 301.36 o-ct-I 2-oiuogrmare . ‘ 603" MW ‘ NADH+H" Glutamm W ‘Y'mm “M MD" (GOGAT) 5 COOH (33" $:éo;-,.o0"1 mm , (Damnation Othernttmgen comm Glutamate Figure 1.2. Model of ammonia assimilation pathways (1,2) Glutamine-synthetase- glutamate synthase pathway, with. low NH; supply (1) and with high NH3 supply (2). (3) Glutamate dehydrogenase pathway. (Adapted from Marshner, 1995). _ BROWN PATCH (Rhizoctonia solani Kuehn). Brown patch disease is caused by the fungus Rhizoctom‘a solani. Other species (R oryzae, R. cerealis) are known to be pathogenic to turfgrass as well (Burpee and Martin, 1992).. Brown patch disease occurs on many commonly cultivated turfgrass species. The fungus produces tan to brown mycelium that are 4-15 mi in diameter with constricted dolipore septae and no clamp connections (Couch, 1995). In the absence of optimal growth conditions, the organism survives by dark brown sclerotia produced in the plant tissue, or as a saprophyte, among the soil and thatch. As the ftmgus begins to actively grow at temperatures of 15-20 C, the sclerotia provide a nutrient source as the mycelia resume growth (Vargas, 1994). Hyphal aggregation leads to the formation of appressoria and these infection cushions penetrate the leaf between epidermal cells or through stomates (Shurtleff, 1953). Ultimately, injury can be inflicted upon the plants in two ways, infection of the plant by mechanical pressure and tissue necrosis caused by enzymatic degradation of the cell walls (Couch, 1995). Brown-patch disease symptoms vary with grass type, mowing height, and environmental conditions. Individual leaf blade symptoms are characterized by tan to brown leaf lesions, which can grow to envelop the entire leaf blade turning it light brown and necrotic; lesions sometimes develop reddish-brown margins. Stems, crowns and roots can be infected by the pathogen. Typical symptoms on a given turf sward include foliar necrosis in brown to straw-colored irregular brown patches. A dark purple smoke ring can develop on the leading front of the disease symptoms, especially on low-cut turf <1 3 mm, and can be seen most frequently in the presence of early morning dew. Disease development of the disease is favored by nighttime temperatures >16 C and > 10 h of leaf 10 wetness (Burpee and Martin, 1992). Mycelia begin active growth at 15-20 c and initial infections can occur at 21-26 C (Vargas, 1994). Temperatures between 27- 29 C are optimal for infection by epidermal cell penetration and colonization is most rapid at 29- 32 C accompanied by high humidity. Above 32 C mycelia development is slowed. High humidity and prolonged periods of leaf wetness, as well as high N levels relative to normal levels of P and K can encourage symptom development. Since dew and plant guttation water contain high levels of nutrients favored by the fungus, removing dew by poling or early morning irrigation is recommended (Vargas, 1994). Chemical control is attained with preventative applications of flutolanil, chlorothalonil, iprodione, or azoxystrobulin applied at 14-28 day intervals when favorable environmental conditions persist. DOLLAR SPOT (Scleroa'nia hamoeocarpa Bennett) Dollar‘spot is one of the most prevalent diseases on golf courses in North America, Australia and Japan (Smiley, 1983). Symptoms appear as circular and sometimes sunken bleached straw-colored to brown patches approximately 2-5 cm in diameter (Vargas, 1994). As the disease severity increases, spots can coalesce, blighting large areas of turf. Individual leaves have bleached, water-soaked tan lesions with a reddish-brown margin often appearing as an hourglass pattern. Mycelia appear as grayish white to white and cottony and are especially visible in the presence of morning dew. Under low N conditions, dollar spot symptoms are more prevalent, assuming adequate P and K levels (Couch and Bloom, 1960). The fungus rarely produces apothecia, and if present, they do not contain viable reproductive organs such as ascospores or conidia (Smiley, 1983). It is believed that the 11 pathogen is primarily dispersed via equipment and traffic and survives as dormant mycelia on leaf foliage. Active growth resumes as favorable conditions develop. The pathogen affects the plant by producing a toxin in the foliage, which upon translocation prevents root elongation, causes browning of the roots and encourages root thickening and a decrease in root hairs. Toxin production is optimal between 15.5-26.8 C (Endo, 1964). Cultural management strategies that reduce the duration of leaf wetness such as poling greens, watering after dark and in the early morning to wash off dew and guttation water fi'om leaves ”can alter environmental conditions that are optimal for the disease. Chemical control is attained with applications of triadirnefon, propiconazole, cyproconazole, thiophanate-methyl, benomyl, iprodione, fenarimol, or chlorothalonil when environmental conditions favorable to. disease development persist (Couch, 1995). PROPERTIES OF LIGHT’ The electromagnetic spectrum contains radiant energy deScribed by parameters of “wavelength”, “frequency”, and energy (Fig. 1.3). The entire spectrum covers 20 orders of magnitude from cosmic rays which contain the most energy to radio waves containing the least. In the middle of the spectrum are ultraviolet (200-400 nm range), visible (400- 700 nm), and near infrared (700-2500 nm range) wavelengths (Kemp, 1991). The visible portion of the spectrum is known as “photosynthetically active radiation” since this is the portion utilized by plants for photosynthesis. Light is a unique form of energy in that it exhibits properties of both waves and particles. A light wave is a “transverse electromagnetic wave” in the shape of a sine where electric and magnetic fields are present perpendicularly to the direction of wave 12 .EPsouem 033352620 one .m._ Semi wcogmcmz 25:65: 28:68: .mcgmnom .mco_85_> oEozooE panama/x 8.00 59w; _ _ mo>m>> pegs 5632:: 2pm. 3 tozm .2“. new 06mm 68:2 >9ocm 26.. mice :9: casement. _o_oo Ema; cacao: :9: £92053 tozw mocmcowwm Eszv 5:960: 26;. osmcmwfi 592053 9.6.. 820:2 98m - x mama mEEmo l3 prOpagation (Taiz and Zeiger, 1991). Wave properties are characterized by the wavelength, the distance between two crests of the sine curve (nm); frequency, how many crests occur in a given distance (Hz, /s); and the pattern. The equation c = Av represents the speed of light, 2.998 x 108 m/S,’where A. is the wavelength and v is the frequency; thus, it and v are inversely proportional. Particle (photon) properties of light consist of discrete packets of energy called “quanta.” Energy is explained by the equation E = hv where E is energy in joules, h is Planck’s constant (6.626 x 10'34 J05), and v is the frequency of the radiation (/S or Hz). Subsequently, E = hc/k so a radiation wavelength is inversely proportional to the energy which it contains. Once light strikes an object it may be reflected, transmitted, or absorbed (Woolley, 1971). Reflected light is returned to the atmosphere at a different angle from which it struck the object incidentally. Transmitted light energy passes through the object without being absorbed; transmittance is negligible through turfgrass because of its dense canopy (Trenholm et al., 1999). Energy absorption occurs when incident light energy matches the exact amount of energy needed to move electrons from a ground to excited State. Excitation may be due to translational, vibrational, or rotational changes which occur in the organic molecule. Since electron orbits represent discrete energy levels, electrons require exact amounts of energy for excitement from one to another. The relationship between transmission of energy through the sample and the concentration of the absorbing molecular bonds is described by Beer’s Law. Energy light absorbed is proportional to the molecule or pigment concentration of interest and is expressed as log (l/reflectance) (Shenk and Westerhaus, 1993c). l4 An absorption spectrum illustrates the change in absorption of electromagnetic energy by an object across a range of wavelengths. When transition of a molecule from one energy state to another occurs at a specific wavelength, it corresponds to the energy absorbed at that wavelength. The molecule will only absorb the energy if it is equal to that required for the transition. Due to differences in bond and molecular structure (and the energy required for transition), organic molecules absorb energy difierentially. Highly conjugated molecules such as plant pigments chlorophyll, anthocyanins, carotenoids and xanthophylls absorb at higher energy wavelengths in the visible spectrum. Organic molecule functional groups such as hydroxyls, carbonyls, and amines, absorb at lower energy wavelengths in the near infrared spectrum. Humans have the ability to differentiate light in the visible region from 400-700 nm. Contained in this range is what we traditionally think of as a “spectrum of colors.” (Fig. l .3). All objects absorb light differentially to varying degrees and the human eye perceives an object as a certain color because that color is reflected the most. Likewise, plant pigments absorb differentially across the spectrum so that a plant’s perceived color, or appearance of an object determined by eye response, consists of wavelengths which are absorbed the least. For instance, in examining the absorption Spectrum of chlorophyll one finds that it absorbs the greatest amount of light in the red and blue regions (75-90% absorbance) and absorbs the least in the green region so that when chlorophyll, the dominant pigment is present, plant leaves appear green (<20% absorbance). With an instrument that measures “greenness” one could indirectly measure chlorophyll content. Since nitrogen is an important component of and closely correlated to chlorophyll, measures of “greenness” would give an indication of the nitrogen status of the plant (Thomas and Oerther, 1972). 15 ChlorOphyll produces a green color because it absorbs the least in the green region (~550 nm). When chlorophyll absorbs light, the light energy causes the chlorophyll molecules to be excited to a higher state from its initial “groun ” state. The excited energy contained within the molecule can undergo one of three fates. The molecule may undergo fluorescence where it re-emits the energy as it falls fi'om its lowest excited state back to its ground state. This release is characterized by a phenomenon called the Stokes Shift as the energy is re-ernitted at a wavelength approximately 10 nm longer than that which it was absorbed. Second, the molecule may return to its ground state without re-ernitting energy as a photon, but as heat. Finally, the molecule may activate the plant’s photosystem network, stimulating the electron transport chain in photosynthesis (Taiz and Zeiger, 1991). Near Infrared Spectrum The near infrared (NIR) region of the spectrum ranges from 700-2500 nm. Functional groups such as =CH2 (1090-1167, 1390-1400, 1406-1446, 1616-1626, and 2260-2510 nm), O-H water bonds (984—996, 1010, 1150, 1406-1416, 1788-1796 and 1936-1946 nm), N-H protein bonds (1048-1052, 1508-1516, 2050-2066, 2176-2186, and 2296-2308 nm), and other N-H groups (1464, 1470, 1480-1506, 1518-1536, 1906-1916, 1976-1996, and 2046-2056 nm) and organic molecules absorb energy in the NIR (Winisi, 1999). Absorbance of NIR radiation corresponds to energy required for changes in the internal vibrational frequencies of the molecule and functional groups of organic molecules absorb NIR radiation differentially. A fundamental vibration occurs when the energy supplied is proportional to the energy required to change the dipole moment of the molecule so that the vibrational energy absorbed causes it to change from its ground state to its first excited state (Zabik, 1997). Absorbance by organic ftmctional groups produces 16 characteristic bands in local areas of the near infrared spectrum (Zabik, 1997). Absorption bands can be characterized by three criteria: location, height, and width. Near infi'ared absorption patterns are very complex, existing in a mosaic of overtones, combination bands and repititive bands. Typical NIR spectra exhibit a convolution of Lorentzian and Gaussian distributions and may consist of seven to ten peaks with many “shoulders” (Shenk and Westerhaus, 19930). Band overlapping and composite banding makes it difficult to estimate the three criteria so mathematical functions are needed to provide accurate estimates 0f band locations. Additional confounding may occur due to particle Size multiplicative response, confounding with visible overtones in 1100-1400 region, and confounding with mid-infrared information contained in the 2300-2500 nm region. Reflected light can undergo a scattering effect as it strikes an object. Scatter is a function of the diffuse nature (roughness) of the surface (Shenk and Westerhaus, 19930). Particle size can contribute to scatter, which can cause peak distortion and larger particles make peaks appear higher than they should. Conversely, surface reflectance, or the “shininess” of an object can “squash” peaks to appear lower than they should. Essentially, the information contained in a NIR absorbance spectrum provides useful insight into the physical and chemical composition of a substance (Shenk and Westerhaus, 1999). Every substance has a unique spectral composite “signature” contributed to by scatter, surface reflectance and absorption of chemical bonds (Shenk and Westerhaus, 1999) and diffuse reflectance properties correlate to changes in chemical composition (Morra et al., 1991). Ideally, Since a spectrometer can detect wavelengths over a wide spectrum of electrOmagnetic radiation, a specific band could be used to 17 detect differences attributable to nitrogen status or disease presence in the turf canOpy. However, more practically, a combination of wavelengths would be used to develop a model which characterizes the anomaly of interest. A fundamental absorption may have several overtones, or secondary vibrations which decrease in intensity (amplitude) and energy level, and exist in the range of 700- 1800 nm. Combination bands consisting of two or more overtones of these groups exist in the 1800-2500 nm range. These combination bands indicate rotational and vibrational movements such as stretching, bending, wagging, and rocking of the organic molecule. Stretching vibrations occur at higher frequencies (lower wavelengths) than bending vibrations. Molecular bending can occur in the plane of the molecule or out of the plane. Each deformation absorbs energy of different intensity. Energy striking a compound will NIR region is composed of harmonic overtones of the fimctional groups which absorb primarily in the mid-infiared (MIR). Major bands in the NIR region include second and third overtones of O-H, C-H, and N-H functional groups. Theoretically, peak height of the vibrations diminishes with each successive overtone. Molecular absorptions occur with greater intensity as fundamental bands in the MIR region of the Spectrum because NIR bands are 10-100 times weaker than those found in the MIR. Organic molecule functional groups O-H, C-H, and N-H absorb energy at different wavelengths due to their stretching, bending and deformation vibrations (Shenk and Westerhaus, 1993c). Shifts in the spectrum related to organic molecules can potentially be associated with physiological changes in the plant. Characteristic wavelengths which indicate the presence of these groups include O-H bonds stretches at 1440 and 1900 nm and N-H 18 stretches in ranges fiom 1449-1555 nm and 1800-2080 nm. Within the umbrella of N-H stretches are primary amines (1455-1553 nm), secondary amines (1506-1555), N-H proteins ( 1535- 1614 rim), nitrites (1800-2080 nm), NH; groups (1965-2050 nm) and NH; amines (1449-1538 nm) (Shenk and Westerhaus, 1993c). SPECTROSCOPY As with any spectrosc0pic method, proper assessment of a sample for evaluation is affected by several factors. Instruments used to detect visible and NIR spectra must be accurate and repeatable. Temperature, relative humidity, and spectrometer light source and intensity play significant roles in instrument performance. The ambient light surrounding the stage of the sample will have an effect on how the light reflected, absorbed, and transmitted by the sample will be detected by an instrument. In a laboratory setting, enclOsed spectrophotometers provide for a means of controlling ambient light surrounding a sample. Near infrared detection devices typically consist of several components. A source of radiance, usually a tungsten light bulb, is needed to provide consistent illumination of the sample. In order to process the quality of light, once detected, the light is transmitted through a slit to limit radiation to a narrow band. A lens is used to focus a narrow band of radiation and the energy is sent through a wavelength dispersion device to split the energy into its component parts before passing through a focusing lens. The energy is transmitted through another focusing lens before passing through an exit slit and ultimately a photodetector. The placement of the detectors determines if the instrument initially makes a transmission or reflectance measurement. Signal from the detector is 19 amplified before being converted from analog to digital for computer processing and monitor display. A There are four primary wavelength dispersion devices used in NIR analysis. Filters are used for detection of absorption in specific regions of the Spectrum, disallowing passage of light outside the range(s) of interest. In contrast, light emitting diodes emit light energy only at Specific wavelengths of interest. Accoustical optical tunable filters (AOTF) are used for liquid solution analysis. Wavelength is controlled by the frequency at which a crystal vibrates. A monochromator is a holographic grating which divides light energy into separate wavelengths at a given interval across the range of detection (Shenk and Westerhaus, 1993c). Light striking an object may be detected by reflectance, transmittance, folded transmittance or direct light methods. Normal NIR reflectance and transmittance measurements involve holding the sample in a ring cup, exposing it to a light source at a path length of 1 cm in a closed compartment and detecting how much is reflected or transmitted, depending on the location of the photodetector. Folded transmittance measurements are ideal for materials in solution and use a narrower path length of 0.1 mm. All three of these measurements are made in chambers opaque to outside light. In the direct light method, source radiation is introduced directly upon the sample. The reflected radiation is then transmitted via fiber optic cable to the monochromator and, subsequently, the photodetector. General NIR Applications Near infrared reflectance measurements are used for analysis of a wide range of agricultural and industrial products (Wetzel, 1983). Notable agricultural applications 20 have involved measurement of protein, moisture, fat, oil, and prediction of organic carbon and total nitrogen (Wetzel, 1983; Dalal and Henry, 1986). Near infrared spectroscopy (NIRS) has also been used to measure moisture content in soybeans and fat and moisture in meat emulsions (Ben-Gera and Norris, 1968). The fact that NIR has been used successfirlly for constituent analysis of forages (Norris, 1976; Windham, 1991) lends to its potential effective use in turfgrass analysis. Near infiared spectroscopy is an attractive alternative to traditional laboratory methods that measure crude protein, acid detergent fiber, fats, moisture and other constituents (Wetzel, 1983; Shenk and Westerhaus, 1991). It provides for rapid analysis of plant constituents and requires minimal sample preparation (Couilliard et al., 1997). Near infrared spectroscopy can accurately measure constituents such as water (O-H bonds) and crude protein (N -H bonds) in the micrograrn per kilogram range (Roberts et al., 1991). Near infrared. spectroscopy does not actually measure N, but measures N-H, from which N and protein can be interpolated (Shenk and Westerhaus, 1991a). Fox et a1. (1993) compared reflectance measurements in the NIR region with three other rapid tests for predicting N-supplying capability and grain yield in corn and found that NIRS was as statistically accurate as the pre-Sidedress nitrogen test (PSNT) to predict the soil N- supplying capacity and corn response to N. Prediction equations for forage mixtures and monostands have been developed using NIR (Shenk and Westerhaus, 1991a ). Principally used for detecting plant constituents in agriculture, NIR has also been used for carbon and nitrogen analysis in particle-size soil fiactions (Morra et al., 1991). Near infiared spectroscopy can be useful because it provides a window into biochemical workings of a plant that reflectance in the 21 visible range may not. For instance, changes in leaf area index (LAI) can result in changes in NIR region reflectance without altering the visible region reflectance characteristics (Colwell, 1974). Traditional sample preparation for NIR analysis involves oven-drying the samples to remove moisture before grinding them to insure a uniform particle size. Samples are then packed into a cell for spectral analysis on a laboratory benchtop. model instrument. However, use of NIR technology for real-time analysis will require development of a field unit capable of conducting direct light measurements. Successful attempts to analyze unprocessed sammes have been accomplished for predicting turf soil profiles (Couilliard et al., 1997). Data Analysis Analysis of NIR data is difficult due to factors such as particle size or spectral (particularly water) overtones (Shenk and'Westerhaus, 1993c). Two corrections have been developed to reduce interference caused by differences in particle size. First, de- trend, a multiplicative scatter correction described by Barnes et al. (1989), Shifts the spectra of interest to be. more like a designated “target spectrum”, usually an average spectrum of the spectra of interest. Second, a standard normal variate correction can be used so that the standard deviation of each spectrum is 1.0. Several regression methods may be used to create a prediction equation for using NIR patterns to predict laboratory analysis numbers. Multivariate regression methods such as modified stepwise regression (MSR), neural networks, and partial least squares (PLS) have been used (Shenk and Westerhaus, 1993c). Shenk and Westerhaus (1991b) found that a modified partial least squares regression (MPLSR) had better correlation 22 than MSR in developing constituent calibration equations for diverse forage mixtures. Comparing the MPLSR method to the MSR method, they demonstrated that MPLSR was similar or better than MSR for predicting crude protein, acid detergent fiber, and in vitro dry matter disappearance for two large groups of forage samples. Algorithms CENTER and SELECT were developed to identify Spectra suitable for calibration development by eliminating samples with extreme or similar spectra. These algorithms use the spectral data across a range of wavelengths with absorbance values expressed as Log (UK) and an associated reference value for the constituent(s) of interest. The CENTER function computes a principal components file by full-spectrum single value decomposition, which contains all information needed to calculate sample scores and define H (Mahalanobis) values. Principal component analysis (PCA) identifies patterns (also known as eigenvectors or loadings) in certain wavelength regions which contain the most variation attributable to different laboratory values. Principal component analysis also reduces the spectral information into a smaller number of independent factors. The amount of a pattern present in a spectrum is referred to as a score (Shenk and Westerhaus, 1993 c). Principal component analysis uses a loading-score method to compare spectra in multiple dimensions. Sample loadings are obtained by multiplying the spectral data by the principal component scores (proportion of a pattern present in a specific spectrum) which are associated with the largest eigenvalues. Principal components are linear combinations of NIR data that maximize differences between spectra and are calculated by multiplying NIR data points by linear combinations of the spectra to form new variables. The CENTER function ranks each spectrum according to its H distance from the average spectrum in hyperspace. 23 Principal component analysis iS a technique for limiting the number of intercorrelated spectral data pOints by using the information contained in the spectra to compute independent variables. The first principal component (factor) accounts for the greatest variation in the spectra, the second accounts for the next greatest amount and so on. Afier ranking the spectra, an algorithm is used to eliminate samples that were spectrally similar. The SELECT algorithm identifies spectra with the greatest number of neighbors within a certain proximity (H 3.0 or T-value.[Eq.2] >2.5. H=(x1-xba,)(X’X)"(x1-xbu)’ 1 [Eq. 1] T = (Difference between 2 samples/ standard error of the difference) [Eq. 2] A principal component analysis file was created on the third pass without removing additional files. Calibration Equation Development Software used for all calculations was provided by Infrasoft International, Port Matilda, PA. Using the default setting of the program, ordered files were used for cross validation where one set of samples is used to create a regression equation and the remainder are predicted. All sets are used alternately for equation development until all samples have been used for prediction and have been predicted. Using the best fitted 39 equation as determined by the cross validation procedure, a coefficient of determination (r2) was calculated. According to Shenk and Westerhaus (1993c), r2>0.90 represents acceptable association between spectra values and N values obtained by laboratory NIR instruments during calibration development. Accounting for greater variability in field conditions, r2 values > 0.80 were deemed acceptable for this study. The calibration method used was a modified partial least squares regression (MPLSR) using detrend, and standard normal variate standardization to create a full spectrum regression model (Shenk ‘ and Westerhaus, 1991b; Barnes et al., 1989). Because MPLSR used all 208 wavelengths in the calibration, no calibration equations are shown due to their size and complexity. During cross validation, each sample spectrum has the opportunity to be predicted as if its laboratory reference value were unknown. The standard deviation of these differences between the predicted value of the sample treated as an unknown and the actual laboratory reference value is the=standard error of cross validation (SECV). The SECV values estimate the actual values of the equation when samples are within the global H limits. Using each sample for both calibration and validation of the equation, the lowest model error is used in conjunction with the lowest prediction error to develop an equation with a low performance error. After the equation is created, the difference between the actual N reference values and the predicted N values is calculated. The standard deviation of these differences is the standard error of calibration (SEC). and the SEC describes how well the predicted values fit the regression line. Standard error of calibration will always be lower than SECV since SEC reflects the fitted values; SECV reflects the actual reference values. The standard error of cross validation is a more accurate means of assessing the equation accuracy than the SEC. The 40 SECV indicates acceptable equation accuracy if it is lower than the standard deviaiton of the laboratory analysis. The variance ratio (l-VR) is calculated as 1-SECV2(SD2)'l where SD = the standard deviation of the laboratory values (Couilliard et. al., 1997). The variance ratio is the ratio of the total variance in the population to the variance predicted by the equation and provides an indication of the accuracy of the model since an accurate model will explain a greater amount of the variation that exists. The coefficient of determination (r2) calculation involves actual values, while the variance ratio uses predicted values, but in many instances they are similar. Instances where the unexplained variance, determined by the variance ratio, is greater than the SECV would indicate an unacceptable association between spectral analysis and actual N content. Lab Value Predictions The MONITOR program was used to predict the laboratory values among species/cultivar, soil type and mowing height by using the equation developed from one population to predict the laboratory N values for another as if they were unknown. For prediction evaluation, the bias was calculated as the difference of the two populations’ means and is used as a baseline to adjust the calculated standard error of differences (SED) between spectra. The standard error of differences was expressed as the standard error of performance (SEP) to gauge prediction accuracy in the MONITOR program. The bias confidence limits (0.6 x SEC) were calculated to identify any bias greater than 1.0 x SEC with 90% confidence when using a one tailed Type I error probability = 0.10. SEP(Corrected) limit of 1.3 x equation SEC was used to determine acceptable performance error (Windham et al., 1989). 41 RESULTS AND DISCUSSION Laboratory Reference Values Dry combustion analysis of the turfgrass clippings ranged from 1.47 to 6.28% N for all treatments following N applications ranging from 0 to 4.8 g/m2 (Table 2.1). A representative VIS-NIRS raw spectra comparison of turf that received a range of applied N is shown in Fig. 2.1. Greatest spectral differences in clipping N content were located at 670 nm, 1450 um, 1510 nm and 1950 nm and these absorption bands are associated with chlorophyll a electron transmissions a primary overtone O-H stretch attributable to water, a first overtone N-H stretch attributable to protein and nitrogen, and O-H stretch and deformation attributable to water, lignin, protein, nitrogen, and starch, respectively (Fig. 2.1). Greatest first derivative spectra differences were observed approximately 30 nm higher than raw spectra differences. ’ Visible-N ear Infrared Reflectance Spectra and Predictions The first objective of this research was to determine if a relationship exists between the laboratory reference N values and the VIS-NIR spectra. Calculations of H distance by the program CENTER indicate a right-skewed histogram because the median of H values was lower than the mean (Fig. 2.2). Calibration statistics, estimated through cross validation, for the turfgrass swards and their combination are presented in Tables 2.2-2.7. The r2 (explained variation) and SEC (prediction accuracy) values were 0.92 and 0.25 for the Penncross green, 0.85 and 0.28 for the Penncross fairway, 0.81 and 0.38 for the Providence fairway, 0.80 and 0.45 for the Poa annua green, 0.80 and 0.40 for the Poa annua fairway, and 0.78 and 0.49 for the combination of all turfgrass swards. 42 Table 2.1. Laboratory values of nitrogen (N) content in turfgrass clippings. Number of Treatment Samples 1‘ Mean N(%) Range (%) Std. Dev. 'Penncross' green 83 4.10 2.08-6.28 0.89 'Penncross' fairway 119 3.87 2.15-5.29 0.74 'Providence' fairway 85 4.24 2.37-6.00 0.87 Poa annua green 104 3.38 1.47-5.84 1.00 Poa annua fairway 77 4.07 2.05-6.05 0.91 All treatments 498 3.92 1.47-6.28 0.93 1' Number of samples used in development of global equation 43 .mfiwco_o>a3 confines: =a $88 8.8% omega Gemé Av emcee can: use .§co.v-m.0 emcee 062:. A Z oem vv owns. 32 he scares—Eco «:0on _.N oSmE 3.9.0.033 83. E: 3: mum :3 a .. s t. 8'... ........................... m ................... - m ................ 3...... m m is... m . m m .z 5...»: 4 m ., m u m z guilty m .. m . .......................... 4 ..... . .............. - . . s ......................... J muu; " n a u . u . .U ..................................................... .H mm... L-—--—---—-_--L---- lllllll‘lllllllIlll'iltlll ........................... sued (an) 60': afih 56 .8388 .3 c8 8.58% :38 05 89c moo—836 I he 8833.5 .m.~ 8=wE 8.. m—d aw... . ;;1 ,. . . ti 1;! ;l|l; 1; 1. 1. ; . ; e M M to seldwas ,to JeqwnN 32.820 I N: 45 Table 2.2.Calibration statistics for ‘Penncross’jreen combined over both seasons. Term Wavelengthi SEC: R2§ F-value SECV# l-VRTT 1 686 0.633 0.495 90.29 0.658 0.458 2 1876 0.445 0.75 92.72 0.501 0.686 3 1896 0.405 0.794 19.75 0.464 0.73 4 686 0.359 0.837 24.63 0.525 0.655 5 686 0.302 0.885 37.46 0.744 0.308 6 0.286 0.897 10.59 0.578 0.582 7 0.274 0.905 8.67 0.521 0.66 8 0.267 0.91 5.34 0.43 0.769 9 0.249 0.922 13.55 0.375 0.824 Table 2.3.Calibration statistics for ‘Penncross’ fairway combined over bog seasons. Term WavelengthT sac: R2 § F-value SECV# l-VRTT 1 1896 0.471 0.597 175.87 0.477 0.587 2 1886 0.404 0.704 ' 43.33 0.427 0.669 3 716 0.339 0.792 49.97 0.379 0.739 4 716 0.285 0.852 - 47.84 0.332 0.800 Table 2.4. Calibration statistics for ‘Providence’ fairway combined over both seasons Term Wavelength‘l' sscr R2§ F-value SECV# l-VRTT 1 686 0.647 0.447 68.84 0.663 0.425 2 1876 0.571 0.569 24.66 0.598 0.533 3 686 0.536 0.62 1 1.86 0.558 0.592 4 686 0.429 0.757 46.64 0.481 0.697 5 1886 0.403 0.786 1 1.80 0.463 0.72 6 0.380 0.809 10.52 0.45 0.736 ‘1' Most important wavelength for the first five loading terms used in the equation iStandard error of calibration § Coeffecient of determination # Standard error of cross validation 'H'Explained variance 46 Table 2.5.Calibration statistics for Poa annua green combined over both seasons. Term Wavelength? SEC: R2 § F-value sacwr l-VRTT 1 1396 0.821 0.337 53.27 0.855 0.276 2 1876 0.623 0.618 76.24 0.69 0.529 3 686 0.546 0.706 31.28 0.613 0.628 4 1896 0.49 0.764 25.28 0.566 0.682 5 1396 0.446 0.804 21.49 0.543 0.708 Table 2.6.Calibration statistics for Poa annua fairway combined over both seasons. Term Wavelengthi' sac: R2 § F-value SECV# l-VR‘H' 1 716 0.692 0.416 55.16 0.764 0.303 2 1876 0.587 0.580 30.21 0.707 0.403 3 1876 0.524 0.666 20.07 0.664 0.474 4 686 0.493 0.703 10.26 0.597 0.574 5 686 0.436 0.769 21.28 0.553 0.634 6 0.403 0.802 13.02 0.53 0.664 Table 2.7.Calibration statistics for all populations combined over both seasons. Term Wavelength'l sac: R2 § F-value secw l-VRTT 1 686 0.73 0.386 313.86 0.734 0.381 2 686 0.681 0.465 74.25 0.69 0.452 3 686 0.571 0.624 210.44 0.602 0.583 4 1876 0.525 0.682 91.16 0.552 0.649 5 1876 0.475 0.741 111.98 0.515 0.695 6 0.461 0.756 31.36 0.509 0.702 7 0.448 0.769 29.36 0.499 0.714 8 0.435 0.784 29.20 0.491 0.723 1’ Most important wavelength for the first five loading terms used in the equation iStandard error of calibration § Coeffecient of determination # Standard error of cross validation TTExplained variance 47 .AEZBE 8:20.128 £63 8286.: 0:: eaten wood 1 ego—m 330a 8286:.— QE 8:28 Joe; 1 82m 8:86.: on: Eemv dengue accede—mo 130% E moEEem =c mo 83 .888 38an as 3394 Mam 05w:— va ”N. 83 88% wand €26.35. 595.: v2.6 aseN mam.— .2: Bacon whvd can...” mmvd ~23. (tanlav) uefiomN 48 Prediction accuracy of the global equation is illustrated graphically in Fig. 2.3. Shenk and Westerhaus (1993c) identified r2>0.90 as acceptable for NIRS applications in the laboratory. Accounting for greater variability under field conditions, r2>0.80 was deemed acceptable in this study. Furthermore, the SEC values for all turfgrass swards were lower than the standard deviation values calculated fiom the laboratory N analysis (Table 2.1), thus indicating greater prediction accuracy of N using VIS-NIRS compared to conventional laboratory techniques. The wavelength regions that contributed most to explaining the spectral variation are listed in Tables 2.2-2.7 and shown as both raw and derivatized spectra in Figs. 2.4- 2.5. It should be noted that the derivative treatment causes a shift in the spectra. Using derivatized spectra from the 1,4,4,1 math treatment the wavelength regions used most often in equation development were 686-696 and 716-726 nm in the V18 region and 1870-1890, 1386-1396, 1480-1515, and 2360-2380 nm in the NIR region. Figure 2.6 illustrates comparison spectra of the first six eigenvector loading terms used in the creation of the global equation. The wavelengths in the V18 region correspond to green absorbance and have been associated with chlorophyll content in sweet pepper leaves (Thomas and Oerther, 1972), corn (Walburg et al., 1982), and N content in wheat (Stone et al., 1995). The major absorbance peaks for free water and water lattice occur around 1440 and 1900 nm, and 2200 nm, respectively (Bowers and Hanks, 1965; Hunt and Salisbury, 1970). The major absorbance peaks for N-H occur around 1020 nm, 1510 nm, 1980 nm, 2060 nm, and 2180 nm (Hatchell, 1999). Wavelength areas that contributed most to equation deve10pment were consistent among turfgrass swards indicating the 49 Figure 2.4. Raw spectra comparison of the average spectrum of five populations. 1 — Penncross green 2 — Penncross fairway 3 — Providence fairway 4 - Poa annua green 5 - Poa annua fairway 50 2.200 1930 11m 670 nm , ‘ ‘ 1444 nm \ fi \\\\_fl_/\ .—l 0.433 2.200 1930 nm 670 nm \ 1444 nm é \’//\\ DD 0.433 / 2100 670 nm 1930 nm 1444 nm \V g. /‘\/ 8° / .1 /_J\ 0.433 1930 nm 4 2.200 Log NR 1444 nm / \ AM‘ f\\_/_I_, ,/ \ I- ~..-_ _/ \‘f/r 0.433 1930 nm 5 2200 l 670 nm \ 1444 nm \~. 6: \ \N f” \ / \ :1 fl / \ \ s / V,” .r’ffi/ 0.433 ‘\,,/ 51 Figure 2.5. First derivative spectra comparison of the average spectrum of five populations. 1- Penncross green 2- Penncross fairway 3- Providence fairway 4- Poa annua green 5- Poa annua fairway 52 mom. 111mm «mm 1110“” 9 111mm .0 : .. ”:1... ::: “if WW1“ 53 U1 dosage Eco—w 05 5 wow: 2:8 88?:an o E: 2: 5: 280% maiso— Lo .38» BEOQEoU ed ocsmfi «595.263 we: . an: we: can act . 1 . . awed. awed- E: 0 E r emu. T E: 2.2 E: can; E: oonu Swan 5nd (an) 601 54 potential for development of one sensor that could used to predict N in shoot tissue across a range of turfgrass species and cultivars. mowing heights, and soil types (Figs. 2.5-2.6). Inter-population Predictions The second objective of this research was to determine how the relationship between VIS-NIRS and N in shoot tissue is affected by species or cultivar, mowing height, and soil type. To accomplish this, N from one turfgrass sward was predicted using the equation developed from another turfgrass sward and vice versa. The comparisons evaluated were: Poa annua fairway vs. green on a sandy loam; Penncross creeping bentgrass fairway vs. green on sandzpeat; Poa annua vs. Providence on sandy loam fairway; and Poa annua vs. Penncross creeping bentgrass across soil types. Prediction statistics showing these comparisons are presented in Table 2.8. In general, the ability of one population to predict N from another was very poor and unacceptable for applications in SSM. Coefficients of determination ranged from 0.07 to 0.55 and standard errors of performance (SEP) exceeded the acceptable limited determined as 1.3 times SEC of the equation used for prediction. Poor prediction performance in these experiments indicates that, although there is an association between laboratory N values and spectra patterns, the equations developed in this research were p0pulation-specif1c. These results may due to differences in leaf canopy architecture resulting from different mowing heights, and genetic color differences among species and cultivars. More importantly, however, is the fact that the prediction accuracy inherent in this statistical procedure is optimized by using a broad database of samples. Therefore, it would be difficult to detect an association between two similar but different populations if the 55 E5. 629.5 .. 5.500% :00... 05 808 00:30:. 0.39.2202 0ma.0>< a we... .8 3.00.80 60:08.83.— .o .80 8.655 a 858.80.. 2083...... o. 02. 08:88.... .0 .802 a 00.85.80: .0 .o..0 880.3% + m: cm 2 a... 8... a: 6.862 8.9.. _ ~64 NM 3... a... 2.6... SN mod mm.~ Ed 500 .05 9.6 mm... :85 AU. mmm ma. _ .. a. .N... 29 mmm mud wad :55 35 end... :5-.. #35 m3... 2.»... 3... mm... 2.002 ~_.N QNN Emm .3388 00:023.: 33.8 358 30% $3.88 3.53 gm $3.8 00:02.6... 9503.0... Steam 00.060... cots—2.0m 9.3060... scream 3.060... cows—2.0.. «a mm. 2 48.... 3... a: 6.82;. 8.9.. .3 «a 2 .o 5.: 2.2% on; and 86 who 5.5 .3m Rd 3.: 3......— G. $5 8. _ _.. and... may mmm :6 :d :55 35 3.9.. «.6... #85 36 _ :v and $8 2.32 8.. end Ema $3.8 0.0.05.0... :00.» 08.0.5?— wezozfia 8.33m 60.060... cote—2.0m :00.» 80.0.80; 33...... $8050.. . @5060... .5835 60.060... coca—smog .38....80 «.2. =0 .8 32.22.09 388.... 5030.. $8060... .8 augusm .w.m 030... 56 .0... fl0.2.5 .. 0.0.8000 :00... 0... .00... 02.0.0.0 03000.0..02 0w0.0>< .. 00... .8 00.02.00 .0000....8.0.. .0 .0..0 0.00005 a 0000.58.00 0.0.0000... 0. 000 00000.8...0 .0 0002 a 00:00:85.. .0 .0..0 0.00006 .‘ m5 Now 2 00.0.0200 .00 00.0.0200 .00 0.. 000.03.. 6N6... and... «M 00.... S... 2.20 00... mm... 3... 00.. $00 .0.m 00... on... .00... 6. .mm 3.9.. and... «.0. 05m m... b. .o 0...... 005 h. d- :6 #005 N. .0 00d 00m I...“ 00002 00... 00.0 55m 00...... 00.. ..< 000.005.. =< 000.005.. ..< 00...... 000. .0... $00.00... 00000.5 00.0.00... 0000300.. 9.00.00... 00000.5 00.0.00... 0000.000. 3 mo. 2 8.. a... .0. 00.33. h. .o... mud... «M 00... .0... 2.20 5.. 3... m... 8.. $00.. .0.m and an... .00... .0. 05m «N... 2.... m6. mum 0N... cm... .00... 005 .0... cm... #005 N6 2... m..m mm... 0:002 an. a... 55m :00.» 00...... 00k 9.00.00... 00000.... 003.00. 00...... 00a. 00.0.00... 0000.000. .0300. 00.0.0 00.. 008w 00...... 00.. 0:00.00... 00000.0... 00.0.00... 0000.008. 57 prediction equation was based upon only one population. Developing a calibration equation with only one population’s data provides a local calibration specific to that species/cultivar and mowing height. To illustrate this point, a separate prediction equation was developed for a portion of the global data set, and then used to predict N from the remaining population. The average 1’2 and SEP for the five subsets used in this comparison were 0.65 and 0.58, respectively (Table 2.9). The lower prediction accuracy Compared to the overall global equation (Table 2.7) was most likely due to fewer samples used to develop the equation. CONCLUSIONS These results indicate that a relationship exists between VIS-NIRS and turfgrass leaf N content. The lower prediction accuracy between laboratory N values and VIS- NIRS spectra demonstrated in this study as compared to other research using the same instrumentation and statistical analysis may attributed to a number of factors. Typical NIRS analysis involves uniform grinding of the sample and use of a laboratory benchtop model for spectral acquisition. Although procedures were taken to minimize the variability due to extraneous factors, conducting experiments in the field and analyzing plants in situ lends itself to a veritable plethora of complex influences. Differences in canopy architecture, affected by leaf angle, texture, surface characteristics, mowing height and density, and phenotypic variation among species and cultivars can change reflectance from the plant canopy (Green et al., 1998; Jackson and Pinter, 1986). Since O-H functional group bonding has a considerable affect on spectral absorbance patterns, differences in plant or soil water relations may change the prediction accuracy of N. 58 Table 2.9. Statistics for predicting subsets of the global equation using the remaind global data for equation calibration. 59.009. .ng___ulation Mud 20mm 22000 00000 10 a 'o Subset 2 Global Quation SEPT 0.438 0.674 Means 3.785 3.922 4.012 Biaszl: -0.018 -0.090 Bias Limit 0.440 0.279 SEP (C)§ 0.440 0.672 SEP (C) Limit 0.678 0.454 Std. Dev. 0.790 0.881 0.759 Slope 0.990 0.782 R2 0.760 0.604 Average H# 0.390 0.398 N 102 102 fluation P 'on Egggjpn predicting predicted predicting Glo Subset 4 glppal Eguatign SEPT 0.594 . 0.612 Means 3.950 3.916 3.870 Bias: 0.025 0.040 Bias Limit 0.296 0.280 SEP (C)§ 0.546 0.614 SEP (C) Limit 0.64 0.607 Std. Dev. 0.767 1.028 0.884 Slope 1.123 0.935 R‘ 0.682 0.647 Average H# 0.393 0.408 N 104 104 Equation Mining global Quation SEPT 0.603 Means 3.952 Biasi -0.024 Bias Limit 0.605 SEP (C)§ 0.605 SEP (C) Limit 0.676 Std. Dev. 0.795 Slope 0.865 R2 0.571 Average H# 0.384 N 103 7‘ Standard error of performance 1 Mean of differences due to instrument performance ’ § Standard error of performance, corrected for bias # Average Mahalonobis distance from the mean spectrum 59 The practical use of the association between leaf N and VIS-NIRS depends upon the degree of scrutiny desired. The ability to sense and apply N in SSM by explaining 80% of the variation and with 95% accuracy would be an improvement over standard soil testing practices and blanket applications of N. Using SSM, a turf manager would be able to apply N based on an optimal leaf N range between, for example, 4 to 5%. Further research is needed to determine the optimal range of leaf tissue N for various turfgrass species and under different management conditions. The MPLSR procedure for NIRS has been found to provide greater prediction accuracy compared to other procedures such as stepwise regression (Shenk and Westerhaus, 19913). According to Couillard et al. (1997) and Shenk and Westerhaus (1993c), the success of using spectroscopy and MPLSR analysis to predict plant and soil constituents is highly dependent upon the development of a broad database of samples with known analysis. Accordingly, this research has only begun to develop such a database to accurately predict N in creeping bentgrass and Poa annua. Although MPLSR analysis may be the most useful technique for improving prediction accuracy in NIRS, it is not the preferred technique to analyze how individual factors such as cultivar, mowing height, and soil type affect VIS-NIRS. Therefore, additional analysis is required to determine which wavelengths and wavelength combinations should be used to develop a sensor to detect N or other constituents in different turf environments. Furthermore, this research was conducted under conditions where only one variable was intentionally imposed. To develop an accurate sensor for use in SSM, the influence of other anomalies and their interactions with VIS-N IRS need to be explored. For example, since fungal pathogens affect turf by disrupting phloem translocation and subsequent macromolecule 60 synthesis and assimilation, an interaction between pathogen presence and N content would be expected. 61 LITERATURE CITED Barnes, Dhanoa, Lister. 1989. SNV Transformation and De-trending of Near infrared Diffuse Reflectance Spectra, Appl. Spectros. 43:772-777. Beard, J .B., 1982, T urfgrass Management for Golf Courses, Macmillan Publishing Company, New York. Blackmer, T.M. , J.S. Schepers, and GE. Varvel, 1994, Light reflectance compared with other nitrogen stress measurements in corn leaves. Agron. J. 86:934-938. Bowers, S.A. and R.J. Hanks, 1965, Reflection of radiant energy from soils, Soil Sci. 100: 130-138. Couilliard, A., A.J. Turgeon, J .S. Shenk, and MD. Westerhaus, 1997, Near Infiared Reflectance Spectroscopy for Analysis of Turf Soil Profiles, Crop Sci. 37:1554-1559. Dalal, RC. and R.J. Henry, 1986, Simultaneous Determination of Moisture, Organic Carbon, and Total Nitrogen by Near Infrared Reflectance Spectrophotomeu'y, Soil Sci. Soc. Am. J. 50:120-123. Epstein, E., 1972, ‘Mineral Nutrition of Plants: Principles and Perspectives’, Wiley, New York. Foss NIRSystems, Inc. 1993. Your NIRSystemsTM Instrument Performance Test Guide, Foss NIRSystems, Silver Spring, MD. Fox, R.H., J .S. Shenk, W.P. Piekielek, M.O. Westerhaus, J .D. Toth, and K.E. Macneal, 1993, Comparison of Near-Infrared Spectroscopy and Other Soil Nitrogen Availability Quick Tests for Corn, Agronomy Journal 85: 1049-1053. Gitelson, A., and MN. Merzylak, 1994. Spectral reflectance changes associated with autumn senescence of Aesculus hippocastanum L. leaves. Spectral features and relation to chlorophyll estimation. J. Plant Physiol. 143:286-292. Hatchell, DC, 1999, Analytical Spectral Devices technical guide, 3rd ed., Analytical Spectral Devices Inc., Boulder, CO. Hunt, GR. and J .W. Salisbury, 1970, Visible and near-infrared spectra of minerals and rocks: I Silicate Materials, Mod. Geol. 11283-300. Jackson, RD. and PJ. Pinter, Jr., 1986, Spectral Response of Architecturally Different Wheat Canopies, Remote Sens. Environ. 20: 43-56. Knipling, EB, 1970, Physical and physiological basis for the reflectance of visible and near-infrared radiation from vegetation, Remote Sens. Environ, 1:155-159. 62 Leco Comp., 1994, Instrumentation For: Characterization of Organic and Inorganic Mehods and Microabsorption Analysis, Leco Company, St. Joseph, MI. Nonis, K.H., R.F. Barnes, J .E. Moore, J .S. Shenk, 1976, Predicting forage quality with NIRS , J. Animal Sci. 43:888-897. Schepers, J .S., D.D. Francis, M.F. Virgil, and F .E. Below, 1992, Comparison of corn leaf nitrogen concentration and chlorophyll meter readings, Coomun. Soil Sci. Plant Anal. 23:2173-2187. Shenk, J .S., and MO. Westerhaus, 1991, Population Definition, Sample Selection, and Calibration Procedures for Near Infrared Reflectance Spectroscopy. Crop Sci. 31 :469- 474. Shenk, J.S., and MO. Westerhaus, 1991, Population Structuring of Near Infiared Spectra and Modified Partial Least Squares Regression. Crop Sci. 31:1548-1555. Shenk, J .S., and MO. Westerhaus, 1991, New Standardization and Calibration Procedures for NIRS Analytical Systems, Crop Sci. 31: 1694-1696. Shenk, J .S., S.L. Fales, M.O. Westerhaus, 1993, Using Near Infi'ared Reflectance Product Library Files to Improve Prediction Accuracy and Reduce Calibration Costs, Crop Sci. 33:578-581. Shenk, J.S. and MO. Westerhaus, 1993, NIR Analysis with Single and Multiproduct Calibrations, Crop Sci. 33:582-584. Shenk, J .S. and MO. Westerhaus, 1993, Monograph, Analysis of Agricultural and Food Products by NIRS Reflectance Spectroscopy, PNIS —0120, Infrasoft International, Port Matilda, PA. Shenk, J .S. and MO. Westerhaus, 1999, 181 Windows Near-Infrared Software Winisi Monograph version 4.0, Infrasofi International, Port Matilda, PA. Stone, M.L., J.B. Solie, W11 Raun, R.W. Whitney, S.L. Taylor, J .D. Ringer, 1993, Use of Spectral Radiance for Correcting In-season Fertilizer Nitrogen Deficiencies in Winter Wheat, Transactions of the American Society of Agricultural Engineers, 39(5): 1623- 163 1 . Thomas J .R., and GR Oerther, 1972, Estimating nitrogen content of sweet pepper leaves by reflectance measurements Agron. J. 64:11-13. Walburg, G., M.E. Bauer, C.S.T. Daughtry, and TL. Housley, 1982, Effects of nitrogen on the growth, yield, and reflectance characteristics of corn Agron. J. 74:677-683. 63 Wetzel D.L. 1983, Near-infrared reflectance analysis Anal. Chem. 55:1165A-1176A. Windham, W.R., D.R Martens, and F .E. Barton 11. 1989, 1. Protocol for NRS calibration:Samp1e selection and equation development and validation. p. 96-103. In G.C. Martens et al. (ed.) Near Infrared Reflectance Spectroscopy (NRS): Analysis of forage quality. USDA Agric. Handb. 643 (revised). U.S. Gov. Print. Omce, Washington, DC. Wolfe, D.W., D.W. Henderson, T.C. Hsiao, and A. Alvino, 1988, Interactive water and nitrogen effects of maize: II. Photosynthetic decline and longevity of individual leaves Agron. Jour. 80:865-870. 64 CHAPTER THREE REMOTE SENSING OF DISEASE SEVERITY IN CREEPING BENTGRASS AND ANNUAL BLUEGRASS USING NEAR INFRARED SPECTROSCOPY ABSTRACT Brown patch (Rhizoctonia solam‘ Kuehn) and dollar spot (Sclerotinia homeocarpa Bennett) are two common diseases of cool season turfgrass in the United States. As governmental and public scrutiny of golf course maintenance practices increases, superintendents are beckoned to balance playability with fewer fungicide inputs. The objective of this study was to develop a method of evaluating disease severity using a direct light visible/near—infrared scanning monochromator on creeping bentgrass (Agrostis stolom'fera Huds.) and annual bluegrass (Poa annua var. reptans Hausskn). Categorical disease symptom severity ratings of brown patch and dollar spot were made on different turfgrass swards and associated spectra obtained so that absorbance was expressed as “Log 1/reflectance” between 400 and 2500 nm. Discriminant analysis of the data yielded classification accuracy. In the dollar spot study, 20 out of 193 samples (10.3%) were classified incorrectly and in the brown patch study using three severity categories, accuracy improved greatly as there were only 29 misses out of a total of 336 samples (8.6%). These results suggest the feasibility of developing a visible/near- infrared sensor for the detection of disease severity. Future research should address investigation of how various stresses interact to affect the spectral reflectance of the turfgrass plant. 65 INTRODUCTION Increasing governmental regulation of pesticides and growing public scrutiny of golf course management practices are leading tothe development of improved methods to decrease fungicide inputs on golf courses. As golf courses continue to fill the role of urban green areas and are the subject of increasing public and governmental scrutiny, a premium is placed upon superintendents to balance environmental impact and playability. Although modern chemistry has led to advances on improving fungicide efficacy with lower active ingredient rates, typical management practices involve widespread “blanket” applications of fungicides during periods conducive to disease development. Site specific application of fungicide has the potential to save money, provide an efficient means for effective disease control, and reduce the amount of fungicide applied. Since disease pathogens are dynamic and can infect plants quickly in the presence of optimal growing conditions, a sensor capable of attaining a rapid, real-time assessment of disease status is necessary for incorporation into a site specific management program. Typically, a given stress reduces photosynthetic capability and causes an increase in reflectance in the red and blue portions of the spectrum and decreased reflectance in the NR region due to deterioration of leaf tissue (N ilsson, 1995) and leaf structural changes (Raikes and Burpee, 1998). Several methods of remotely sensing plant disease status have been evaluated in past research. Indices such as-the Leaf Area Index (LAI) ‘ (R reflectance/Red reflectance) and Normalized Difference Vegetative Index (NDVI) [(R-R)/(R+R)] have been correlated with the presence of green biomass and provide a quantitative estimate of general stress on a plant; however, it is often difficult to determine exactly the nature of the stress (N ilsson, 1995). Infrared aerial photographs 66 have been used with moderate success to remotely sense sugar cane rust fungus (Puccinia kuehnii)(Karteris et al.,l980); sugarbeet blackroot disease, one of the causal agents of which is Rhizoctonia solani (Schneider and Safir, 1975); and southern corn leaf blight (Helminthosporium maydis L.) (Safir et al., 1972).. Contrary to others, they found that visible reflectance changes preceded infrared reflectance changes. The objective of this research was to assess disease severity of two common cool- season turfgrass diseases, brown patch (Rhizoctonia solani Kuehn) and dollar spot (Sclerotinia homeocarpa Bennett) using a scanning monochromator capable of measuring spectral reflectance from 400-2400 nm. 67 MATERIALS AND METHODS Two experiments were conducted at the Michigan State University Hancock Turfgrass Research Center (E. Lansing, MI). The first experiment was conducted to assess dollar spot (Sclerotinia homeocarpa Bennett) on swards consisting of mature annual bluegrass (Poa annua var. reptans, Hausskn) grown on a Owosso sandy loam [fine-loamy, mixed, mesic Typic Hapludalfs], ‘Providence’ creeping bentgrass (Agrostis stolonifera, Huds.) grown on a Owosso sandy loam, and ‘Penncross’ creeping bentgrass grown on a 90:10 (v/v) sand:peat mix that conformed to United States Golf Association (USGA) specifications. The former two swards were maintained as fairways and mowed at 14 mm and the latter maintained as a green and mowed at 5 mm. Spectrometer readings were obtained from June 16-19, 1999 from portions of the sward naturally infested with dollar spot. Spectra measurements were categorized qualitatively by visual assessment as diseased (diseased); close to the disease but visually healthy (disease front); and visually healthy within the same sward, but not close to disease symptoms. (healthy). The second set of experiments was conducted to assess brown patch (Rhizoctonia solam' Kuehn) on a mature sward of ‘Penncross’ creeping bentgrass grown on a 90: 10 USGA sand:peat mix maintained as a green and mowed at a height of 5 mm. Spectrometer readings were conducted during September 2-9, 1999 from areas included in a curative fungicide treatment study. Spectra measurements were qualitatively categorized by visual assessment according to disease severity as severe, moderate, and light. 68 Spectrometer Measurements Spectral reflectance from the turf canopy was acquired with a NRSystems (Silver Spring, MD) Model 6500 online scanning monochromator. Spectral data were obtained every 2 nm from 400 to 2500 nm and expressed in absorbance units as the log (l/reflectance). The spectrometer was adapted for field use by mounting onto the rear of a garden tractor. The acquired spectral signal was sent to the spectrometer via a fiber- 0ptic cable that was connected to a 30-cm by 15-cm metal box that was mounted onto four 15-cm diameter wheels. The box was suspended approximately 13 cm above the surface of the turf canopy and collected radiation from a 3.5-cm by 12-cm area. The box was designed to minimize the effects of incident solar radiation by shading the area where reflectance measurements were taken. . Furthermore, direct light was provided from the box to the measured area using a tungsten-halogen bulb. Three measurements were taken from different locations within each plot during each sampling time. Measurements were taken between the hours of 0730 and 1830 h when disease symptoms were present. In order to maintain accuracy and repeatability with the instrument, a reference was attained for each scan and the spectrum for the scan is subtracted from that of the reference. In this regard, the NRS Online 6500 performs similarly to a double beam specu'ometer where a reference and sample spectra are obtained simultaneously and the differences plotted on the output. Diagnostic tests were conducted prior to sample readings for repeatability and photometric accuracy. To insure instrument repeatability, diagnostics are conducted prior to sample readings. A Coors ceramic reference plate, which is 80% reflective was 69 scanned once as a reference and again as a sample to measure repeatability. A noise test was conducted by obtaining 32 scans of the reference and 32 more scans using the reference as a sample. The repeatability noise was plotted as the difference between those two sets. The root mean square (RMS) of noise errors across the entire spectra is used to gauge repeatability. Accuracy tests were conducted with a polystyrene standard with known peaks at 1143, 1681, 2166, and 2305 nm (Foss NRSystems, 1993). Data Analysis Data were analyzed by multivariate discriminant analysis as described by Morrison (1990) using software provided by Infrasoft International (Port Matilda, PA). The three qualitative dollar spot categories were discriminated in the first analysis. In another separate analysis, attempts were made to discriminate among spectra from the three qualitative brown patch disease categories and spectra gathered from a healthy ‘Penncross’ green during a nitrogen assessment experiment. A third analysis combined all levels of disease (excluding “healthy” samples) for each of the two diseases and attempted to discriminate between the two diseases. The variables used for classification assume that each population were characterized by a multivariate normal distribution and has a common correlation variate. Following these calculations, cross validation was conducted as described in Chapter 2 so that each set of spectra was used to develop the prediction equation and was placed into one of the categories. Analysis was conducted using the default settings for the DISCRIMIN ATE program of the Infrasoft Software with a wavelength scanning range from 400-1000 nm and 1100-2100 nm in 4 nm increments and a math treatment of 1,4,4,1 (derivative, gap, smoothing factor 1, smoothing factor 2) without scatter 70 correction (Shenk and Westerhaus, 1999). Eight cross validation groups were used in creating the prediction equation. In addition to the discrimination comparisons described above, an analysis was conducted to discriminate between brown patch and dollar spot. A 10% error rate for prediction of the samples was deemed acceptable in the evaluation of the results. 71 RESULTS AND DISCUSSION Dollar spot study In the dollar spot study, 20 out of 193 samples (10.3%) were classified incorrectly (Table. 3.1). Comparison spectra for the raw and derivatized data are presented in Fig. 3.1. Attempting to identify the spectra obtained from the “disease front” resulted in the highest percentage of misclassified samples. These results indicate the possibility of identifying the disease before symptoms become manifest; however the question still remains whether this is due only to its close proximity to the disease and if the same results would be measured in a symptom-free sward that is on the verge of developing symptoms. A concern that may contribute to confounding is the fact that the scanning view of the spectrometer was often larger than the diseased area for some scans classified as “diseased.” This discrimination suggests the possibility of using information from the VIS-NR portion of the electromagnetic spectrum for a sensor designed to spray variable rates of fungicide preventatively or curatively for the dollar spot disease. Brown Patch Using all four categories, 87 of a total 336 samples (26%) were misclassified (Table. 3.2). Comparison spectra for the raw and derivatized data are presented in Figs. 3.2 and 3.3, respectively. This was most prevalent as “severe” spectra were misidentified as “moderate,” and “moderate” spectra mistaken for “light.” In an effort to improve prediction accuracy at the expense of reduced prediction precision, the “light” and “moderate” categories were combined and the data were analyzed using three categories for discrimination. Prediction accuracy improved greatly as there were only 29 misses out of a total of 336 samples (8.6%) (Table 3.3). It is unclear whether or not these results 72 Table 3.1. Predicted v. Actual Categog Classification for Dollar Smt Smptra Predicted Category % of Close Diseased Healthy Total Total % Error 0 .3 Close 56 6 6 68 35.2 17.6 ‘5 Diseased 5 58 1 64 33.1 9.3 0 Healthy 2 0 59 61 31.6 3.3 V) Totals 63 64 66 193 Misses for Category 7 6 7 Uncertain 22 21 18 73 Table 3.2. Predicted v. Actual Category Classification for Brown Patch Spectra Using Four Categories. Predicted CategorY Light Moderate Severe % of . Healthy disease disease disease Total Total % Error Healthy 68 0 0 0 68 20.2 0.00 E Light disease 0 42 19 1 62 18.5 32.2 g! Moderate disease 0 28 73 14 1 15 34.2 36.5 (g Severe disease 1 ' 3 21 67 112 33.3 22.3 Totals 68 73 113 82 336 Misses for Category 1 31 4O 15 Uncertain 6 41 64 44 Table 3.3. Predicted v. Actual Category Classification for Brown Patch Spectra Using Three Categories. Predicted Category Medium Severe % of Healthy disease disease Total Total ‘70 Error 0 Healthy 68 0 0 68 20.2 0.00 3 CU 8: Medium disease 0 175 1'7 192 57.1 8.86 8 a; Severe disease 1 l l 65 77 22.9 14.3 Totals 68 186 82 336 Misses for Category 1 l l 17 Uncertain 0 27 16 74 Rafi—non: use :2on 3:26.. we cemEmEoo 5.58% ownco>< .mm oczmfi ofiaeoficfi; anew '3— 9: mwm gt a -. -. -. ude m m 2:8:le m .u, ........................... u ................................. - ....................................... 39... H .......................... 3.---- ---------- -- - 53.— 3% 3:2. Iv ---- -----‘--J-------------- -II'III’I'I'I'III III- III IIIIIIIIIIIIIIIIIUIIII.IIII---IIIII'I'Ill’lI-l'I-II- -------L------------- -—-----------.-- ------------- .- I I I I I I I I I I I I I I I I I I I I I I I I I I I. I I I I I I I I I I I I I I I I I I I I I I I I I I L I I I I I I I I I I I I I I I I I I I I I I I I I I -------------------------- A can. — cum.— (all) 501 75 .mocowoao 08.3% :83 5505 025 he cemngoo 88on 25% NM oSwE Sauce-26>) cmvw 3:: ac: mwm new -. awn-e .-_--—_—-_--1 u.--— _---------—--—‘ . o .11littltlllartrrurvr-n.III.:IIIIIIIIIIIIIIIIIIIIII “can s'NJ L new; . .......................... c .......................... mum; (ti/11501 .0386 :23 550.5 he motowfiao 025 .8 sewing—8 380% 0333.06 LEE m.m Semi 959.3263 amvw m2: Na: mum new a s -. _. ac p...- m m ._ m _ m .m ........................... m - -. ........................ m - ......................... m ......................... . 3.....- m r... m m 1, . m . .x _ ,_ / . . u . ‘5 . 4 4 n . “a a --------------,/ I. ------------------ .- --a. .......................... .. -- m3... "A v 215 /\ u n ...................... .U ........................... .n .......................... .H 58:1“: ........ . ...... . 23... m m m 52.5: 1v m m . m 2025 Iv .......................... 2:... 77 eagle/1119c] isi suggest the subjectivity of qualitative severity ratings and subsequent broad overlap of populations classified as “light” and “moderate”. For practical applications, three categories, “healthy”, “light-moderate”, and “severe” may prove sufficient for effective site-specific applications and subsequent savings in fungicide. Brown Patch v. Dollar Spot Combining the three categories of brown patch severity spectra and “diseased” and “front” categories of dollar spot, respectively, analysis was conducted to assess the accuracy of discriminating between the two diseases. Results indicate these populations are significantly different enough to be predicted with 100% accuracy in this particular study; however, the fact that the dollar spot spectra were gathered on 3 different grass swards and the brown patch on only sand-based, green-height creeping bentgrass provides for the strong likelihood of a confounding effect due to grass species, mowing height, and soil type. CONCLUSIONS These results indicate that VIS-N RS is a viable method for assessing brown patch and dollar Spot severity. According to the data presented, the spectrometer can qualitatively categorize disease severity with a suitable degree of accuracy. Unlike previous experimentsinvolving the association of turfgrass disease severity with reflectance at discrete spectral wavelengths, the discriminant analysis described above used continuous portions of the visible and near infrared portions of the spectrum for analysis. Previous research indicates that reflectance values measured at 660-, 710-, 760- , and 810-nm and subsequent mathematical combinations of these provide for the best correlation between spectral and disease severity ratings on brown patch and gray leaf 78 spot (Raikes and Burpee, 1998; Green et al., 1998). The raw data (Fig. 3.2) illustrate spectral differences at these wavelengths and throughout the NR portion of the spectrum, notably at 1448-nm and l932-nm. First derivative results (Fig. 3.3) illustrate the greatest differences between categories at 700-, 1400-, and 1930-um. Because of the various physiological effects produced by pathogens as they degrade leaf tissue, it is difficult to focus on one particular portion of the spectrum for differences in reflectance. For practical integration into a site-specific management regime, threshold levels of disease need to be developed for proper fungicide treatment. One of the caveats of this technology is the limited amount of data that has been collected. Studies such as these have been conducted by focusing on one anomaly of interest and experimental procedures seek to exclude all other extraneous factors that could affect the absorption pattern of the instrument. However, any interaction effect of multiple anomalies (i.e. water stress, disease, insect damage, chlorosis, etc.) on plant reflectance patterns and their subsequent interpretation is relatively unexplored. To further assess the feasibility of VIS-NRS technology in site-specific management, experiments need to be conducted exploring interactions among various anomalies. 79 LITERATURE CITED Foss NIRSystems, Inc. 1993. Your NIRSystemsTM Instrument Performance Test Guide, Foss NRSystems, Silver Spring, MD. Green 11, DE, L.L. Burpee, and KL. Stevenson, 1998, Canopy Reflectance as Measurements of Disease in Tall Fescue. Crop Science 38: 1603-1613. Karteris, M.A., G. Schultink, G.R. Safir, R. Hill-Rowley, 1980, An Evaluation of 70 mm Color Infrared Photography for Sugar Cane Rust Assessment in the Dominican Republic, International Symposium on Remote Sensing of Environment Proceedings. Morrison, DR, 1990, Multivariate Statistical Methods, 3”. ed., McGraw Hill, Toronto. Nilsson, HE, 1995, Remote Sensing and Image Analysis in Plant Pathology. Canadian Journal of Plant Pathology 17:154-166. Raikes, C. and LL. Burpee, 1998, Use of Multispectral Radiometry for Assessment of Rhizoctonia Blight in Creeping Bentgrass Phytopathology 88:446-449. Safir, G.R., G.H. Suits, and AH. Ellingboe, Spectral Reflectance and Transmittance f Corn Leaves Infected with Helminthosporium maydis. Phytopathology 62:1210—1213. Schneider, C .L. and GR. Safir, 1975, Infrared Aerial Photography Estimation of Yield Potential in Sugarbeets Exposed to Blackroot Disease. Plant Disease Reporter 59:627- 631. Shenk, J.S. and MO. Westerhaus, 1999, ISI Windows Near-Infrared Software Winisi Monograph version 4.0, Infiasofi International, Port Matilda, PA. 80 BIBLIOGRAPHY Al-abbas, A.H., R. Barr, J .D. Hall, F.L. Crane, and M.F. Baumagardener,l974 Spectra of normal and nutrient-defecient maize leaves Agron. J. 66:16:20. Association of Official Analytical Chemists, Official Mehods of Analysis of Association of Official Analytical Chemists International, Patricia Cunniff, ed., 16th ed., vol. 1, 1997, AOAC International, Gaithersburg, MD, USA ch. 2 pp. 11-13. Barnes, Dhanoa, Lister.1989. SNV Transformation and De-trending of Near infrared Diffuse Reflectance Spectra, Appl. Spectros. 43:772-777. Bauer, ME. 197 5, The role of remote sensing in determining the distribution and yield of crops. Adv. Agron. 27: 271-304. Beard, J .B., 1982, Turfgrass Management for Golf Courses, Macmillan Publishing Company, New York. Ben-Gera and Norris, 1968, Determination of Moisture Content in Soybeans by Direct Spectrophotometry, Isr. J. Agric. Res. 18:125-132. Blackmer, T.M. , J .S. Schepers, and GE. Varvel, 1994, Light reflectance compared with other nitrogen stress measurements in corn leaves. Agron. J. 86:934-938. Bowers, S.A. and R.J. Hanks, 1965, Reflection of radiant energy from soils, Soil Sci. 100:130-138. Burpee, L. and B. Martin, 1992, Biology of Rhizoctonia Species Associated with Turfgrasses. Plant Disease 76: 1 12-1 16. Carter, GA, 1993, Responses of leaf spectral reflectance to plant stress, Am. J. Bot. 80:230-243. Carter, G.A., 1994, Ratios of reflectance in narrow wavebands as indicators of plant stress. Int. J. Rem. Sens. 15:697-703. Carter, GA. and R.L. Miller. 1994, Early detection of plant stress by digital imaging within narrow stress-sensitive wavebands, Remote Sens. Environ. 50:295-3 02. Carter, G.A., W.G. Cibula, and R.L. Miller, 1996, Narrow-band reflectanceimagery compared with themal imagery for early detection of plant stress, J. Plant Physiol. 128:515-522. Cassman K.G. and RE. Plant, 1992, A model to predict crop response to applied fertilizer nutrients in heterogenous fields Fertilizer Res. 31(13): 151-163. 81 Coleman,S.W., S. Christiansen, J.S. Shenk, 1990, Prediction of Botanical Composition using NRS Calibrations Developed From Botanically Pure Samples, Crop Sci. 30:202- 207. Couilliard, A., A]. Turgeon, J .S. Shenk, and MO. Westerhaus, 1997, Near Infrared Reflectance Spectroscopy for Analysis of Turf Soil Profiles, Crop Sci. 37:1554-1559. Couch, H.B., 1995, Diseases of Turfgrasses, Krieger Publishing Comp., Malabar, FL. Couch, HE. and J .R. Bloom, 1960, Influence of Environment on Diseases of Turfgrasses II, Effect of Nutrition, pH, and Soil Moisture on Sclerotinia Dollar Spot. Dalal, RC. and R.J. Henry, 1986, Simultaneous Determination of Moisture, Organic Carbon, and Total Nitrogen by Near Infrared Reflectance Spectrophotometry, Soil Sci. Soc. Am. J. 50:120—123. Duncan J., D. Stow, J. Franklin and A. Hope, 1993, Assessing the relationship between spectral vegetation indices and shrub cover in the J omada Basin, NM, Int. J. Remote Sensing 14:3395-3416. Dwyer, L.M., A.M. Anderson, B.L. Ma, DW Stewart, M. T ollenaar, E. Gregorich, 1995, Quantifying the nonlinearity in chlorophyll meter response to corn leaf nitrogen concentration Can J. Plant Sci. 75:179-182. Endo, R.M., H.D. Ohr, and EM. Kraussman, 1964, Phytopathology 54:1175-1176. Epstein, E., 1972, ‘Mineral Nutrition of Plants: Principles and Perspectives’, Wiley, New York. Foss NRSystems, Inc. 1993. Your NRSystems" Instrument Performance Test Guide, Foss NIRSystems, Silver Spring, MD. Fox, R.H., J.S. Shenk, W.P. Piekielek, M.O. Westerhaus, J.D. Toth, and K.E. Macneal, 1993, Comparison of Near-Infrared Spectroscopy and Othe er Soil Nitrogen Availability Quick Tests for Corn, Agronomy Journal 85:1049-1053. Gitelson, A., and MN. Merzylak, 1994. Spectral reflectance changes associated with autumn senescence of Aesculus hippocastanum L. leaves. Spectral features and relation to chlorophyll estimation. J. Plant Physiol. 143:286-292. Green 11, DE, L.L. Burpee, and KL. Stevenson, 1998, Canopy Reflectance as Measurements of Disease in Tall Fescue. Crop Science 38:1603-1613. Hatchell, DC, 1999, Analytical Spectral Devices technical guide, 3rd ed., Analytical Spectral Devices Inc., Boulder, CO. 82 Hunt, GR. and J .W. Salisbury, 1970, Visible and near-infrared spectra of minerals and rocks: I Silicate Materials, Mod. Geol. 1:283-300. ISI Windows Near-Infrared Software WINISI Monograph, version 4.0, Infrasoft International Inc., Port Matilda, PA. Jackson, RD. and P]. Pinter, Jr., 1986, Spectral Response of Architecturally Different Wheat Canopies, Remote Sens. Environ. 20: 43-56. Kemp, W., 1991, Organic Spectroscopy, 3rd ed., W.H. Freeman and Company, New York. Karteris, M.A., G. Schultink, G.R. Safir, R. Hill-Rowley, 1980, An Evaluation of 70 mm Color Infrared Photography for Sugar Cane Rust Assessment in the Dominican Republic, International Symposium on Remote Sensing of Environment Proceedings. Knipling, 13.8., 1970, Physical and physiological basis for the reflectance of visible and near-infrared radiation from vegetation, Remote Sens. Environ, 1:155-159. Krzanowski, R.M., C.L. Palylyk, and RH. Crown, 1992, GIS Lexicon In D. Lusch Fundamentals of GIS Workshop, East Lansing, MI.18-l9 Aug.1991, Center for Remote Sensing & Geographic Information Science, Michigan State University. Leco Comp., 1994, Instrumentation For: Characterization of Organic and Inorganic Mehods and Microabsorption Analysis, Leco Company, St. Joseph, MI. Ma, B.L., M.J. Morrison, and L.M. Dwyer, 1996, Can0py Light Reflectance and Field Greenness to Assess Nitrogen Fertilization and Yield of Maize, Agron. J. 88:915-920. Marshner, H., 1995, Mineral Nutrition of Higher Plants, 2nd ed., Harcourt Brace and Company, Toronto. Morra, J .M., M.H. Hall, and LL. F reebom, 1991, Carbon and Nitrogen Analysis of Soil Fractions Using Near Infrared Reflectance Spectroscopy, Soil Sci. Soc. Am. J 55:288- 291. Morrison, D.F., 1990, Multivariate Statistical Methods, 3rd. ed., McGraw Hill. Nilsson, H.B., 1995, Remote Sensing and Image Analysis in Plant Pathology. Canadian Journal ofPlant Pathology 17:154-166. Norris, K.H., R.F. Barnes, J .E. Moore, J .S. Shenk, 1976, Predicting forage quality with NIRS , J. Animal Sci. 43:888-897. Nutter F W, Jr., 1987, Detection of plant disease gradients using a hand-held, multispectral radiometer (Abstract), Phytopathology 77:643. 83 Perry, CR. and L.F. Lautenschlager, 1984, Functional equivalence of spectral vegetation indices. Remote Sensing of Eviron. 14:169-182. Raikes, C. and LL. Burpee, 1998, Use of Multispectral Radiometry for Assessment of Rhizoctonia Blight in Creeping Bentgrass Phytopathology 88:446-449. Roberts, C.A., R.E. Joost, and GE. Rottinghaus, 1997, Quantification of Ergovaline in Tall F escue by Near Infrared Reflectance Spectroscopy, Crop Sci. 37:281-284. Safir, G.R., G.H. Suits, and AH. Ellingboe, Spectral Reflectance and Transmittance f Corn Leaves Infected with Helminthosporium maydis. Phytopathology 62: 1210-1213. Schepers, J .S., DD. Francis, MF Virgil, and FE Below, 1992, Comparison of corn leaf nitrogen concentration and chlorophyll meter readings, Coomun. Soil Sci. Plant Anal. 23:2173-2187. Schneider, CL. and GR. Safir, 1975, Infrared Aerial Photography Estimation of Yield Potential in Sugarbeets Exposed to Blackroot Disease. Plant Disease Reporter 59:62 7- 631. Shenk, J .S., and MO. Westerhaus, 1991, Population Definition, Sample Selection, and Calibration Procedures for Near Infrared Reflectance Spectroscopy. Crop Sci. 31 :469- 474. Shenk, J.S., and MO. Westerhaus, 1991, Population Structuring of Near Infrared Spectra and Modified Partial Least Squares Regression. Crop Sci. 31:1548-1555. Shenk, J .S., and MO. Westerhaus, 1991, New Standardization and Calibration Procedures for NIRS Analytical Systems, Crop Sci. 31: 1694-1696. Shenk, J .S., S.L. F ales, M.O. Westerhaus, 1993, Using Near Infrared Reflectance Product Library Files to Improve Prediction Accuracy and Reduce Calibration Costs, Crop Sci. 33:578-581. Shenk, J .S. and MO. Westerhaus, 1993, NIR Analysis with Single and Multiproduct Calibrations, Crop Sci. 33:582-584. Shenk, J .S. and MO. Westerhaus, 1993, Monograph, Analysis of Agricultural and Food Products by NIRS Reflectance Spectroscopy, PNIS —0120, Infrasoft International, Port Matilda, PA. Shenk, J .S. and MO. Westerhaus, 1999, ISI Windows Near-Infrared Software Winisi Monograph version 4.0, Infrasofi International, Port Matilda, PA. 84 Shurtleff, 1953, Susceptibility of Lawn Grasses to Brown Patch. (Abstract) Phytopathology 43:110. ‘ Smiley, R.W., ed., 1983. Compendium of Turfgrass Diseases, American Phytopathological Society, St. Paul, MN. Solomonson, LP, and M.J. Barber, 1990, Assirnilatory Nitrate Reductase: Functional Properties and Regulation, Annual Rev. Plant Physiol. Plant Mol. Biol., 41 :225-253. Stone, M.L., J .B. Solie, W.R. Raun, R.W. Whitney, S.L. Taylor, J .D. Ringer, 1993, Use of Spectral Radiance for Correcting In-season Fertilizer Nitrogen Deficiencies in Winter Wheat, Transactions of the American Society of Agricultural Engineers, 39(5):1623- 1631. . Taiz, L. and E. Zeiger, 1991, Plant Physiology, The Benjamin/Cummings Publishing Company, Inc., Redwood City, CA. Thomas J .R., and GP. Oerther, 1972, Estimating nitrogen content of sweet pepper leaves by reflectance measurements Agron. J. 64:11-13. Tisdale, S.L., W.L. Nelson, J.D. Beaton, J .L. Havlin, 1993, Soil Fertility and Fertilizers, 5th ed., Macmillan Publishing Comp., New York. Turgeon, A.J., 1991, Turfgrass Management, 3rd ed., Prentice-Hall, Englewood Cliffs, NJ. Trenholm, L.E., R.N. Carrow, R.R. Duncan, 1999, Relationship of Multispectral Radiometry Data to Qualitative Data in Turfgrass Research, Crop Science 39:763-769. Vargas, J .M., Jr., 1994, Management of Turfgrass Diseases, Lewis Publishers, Boca Raton, FL. Walburg, G., M.E. Bauer, CST Daughtry, and TL. Housley, 1982, Effects of nitrogen on the growth, yield, and reflectance characteristics of corn Agron. J. 74:677-683. Wanjura DR and J .L. Hatfield, 1987. Sensitivity of spectral vegetative indices to crop biomass Transactions of the ASAE 30(3):810-816. Wetzel D.L. 1983, Near-infrared reflectance analysis Anal. Chem. 55:1165A—1176A. Windham, W.R., D.R. Martens, and RE. Barton II. 1989, 1. Protocol for NIRS calibration:Sample selection and equation development and validation. p. 96-103. In G.C. Martens et al. (ed.) Near Infrared Reflectance Spectroscopy (NRS): Analysis of forage quality. USDA Agric. Handb. 643 (revised). U.S. Gov. Print. Office, Washington, DC. Winisi, 1999, version 1.02a, Infrasoft International Inc., Port Matilda, PA. 85 Wolfe, D.W., D.W. Henderson, T.C. Hsiao, and A. Alvino, 1988, Interactive water and nitrogen effects of maize: H. Photosynthetic decline and longevity of individual leaves Agron. Jour. 80:865-870. Woo, K.C., U.I. Fluegge, and H.W. Heldt, 1987 , A two—translocator model for the transport of 2-oxoglutarate and glutamate in chloroplasats during ammonia assimilation in the light, Plant Physiology 84:624-632. Wood., CW, DW Reeves, RR Duffield, KL. Edmisten, 1992, Field chlorophyll measurements for evaluation of corn nitrogen status. J. Plant Nutr. 15:487-500. Woolley, J .T., 1971 Reflectance and transmittance of light by leaves. Plant Physiol. 47:656-662. Zabik, M., 1997 , Entomology 893 Course Pack, Mich. State University, E. Lansing, MI. 86 nrcurcaN STATE UNIV. LraRnRrEs 11111111111111111111111111111111111 31293020741041