| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Natural Resources and Environmental Sciences, University of Illinois at Urbana-Champaign, 1102 S. Goodwin Ave., Urbana, IL 61801
* Corresponding author (fermo{at}uiuc.edu).
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Abbreviations: NIRS, near infrared reflectance spectroscopy • SEC, standard error of calibration • SECV, standard error of cross-validation • SEP, standard error of prediction • SEP(C), corrected standard error of prediction • TNC, total nonstructural carbohydrate(s) • 1 –VR, 1 minus the ratio of unexplained variance to total variance
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Although NIRS has been used for several decades for determining N, total protein, carbohydrates, lipids, other organic chemicals, and moisture content in forages and other crop plants (Foley et al., 1998; Masoni et al., 1996; Vazquez de Aldana et al., 1995), its application for measuring carbohydrates in cool-season turfgrasses is not well developed. Although commercially available equations exist in the market for measuring carbohydrates in forages, their use in turfgrasses are not suitable because they were not calibrated for cool-season turf species such as creeping bentgrass.
Dry turfgrass is mainly composed of organic matter. The most common molecular bonds in turfgrass are between hydrogen, carbon, nitrogen, oxygen, phosphorus, and sulfur. The frequency of vibration between these molecules is such these bonds generally absorb light in the near infrared region. The most characteristic wavelengths for monosaccharides are 1457, 2062, 2263, and 2440 nm and for oligo- and polysaccharides 1432, 1931, 2170, 2310, and 2477 nm. These wavelengths, which have been identified from spectral profiles, are the most appropriate for a quantitative study of unknown carbohydrate mixtures (Robert and Cadet, 1998). Therefore, NIR spectral analysis in this range may provide an accurate measure of the concentration of different TNC components in creeping bentgrass clippings. Since NIR can penetrate opaque biological materials, minimum sample preparation is often needed for representative analysis.
An aspect of carbohydrate analysis in turfgrasses, which has not been previously examined, is the effects of different sampling and drying techniques on carbohydrate stability in turfgrass clippings. Carbohydrates comprise the largest amount of substrate for respiration and are the primary substrates metabolized during drying (Parkes and Greig, 1974). When frozen leaves of 24-d-old maize (Zea mays L.) plant were subjected to freeze-thaw treatment, several enzymes, including phosphoenolpyruvate carboxylase (PEPC) and ribulose-1, 5-bisphosphate carboxylase (RuBPC), were rapidly inactivated and degraded (Usuda and Shimogawara, 1994). Evensen and Boyer (1986) reported more rapid decline of reducing sugars and sucrose in sweet corn at 10°C than at 1°C. Lindroth and Koss (1996) stated that, for studies measuring carbohydrates, freeze-drying is a better alternative and should not affect levels of secondary compounds if samples remain frozen during the drying process. Danley and Vetter (1971) found that freeze-drying reduces the loss of readily available organic constituents of fermented forages, such as volatile acids, compared with oven drying.
The objectives of these experiments were (i) to determine any significant changes in TNC and its component concentrations with different sampling and drying techniques; (ii) to develop predictive equations for glucose, fructose, sucrose, fructan, and TNC concentrations in creeping bentgrass clippings using NIR spectroscopy; and (iii) to accurately predict the concentrations of various nonstructural carbohydrates in creeping bentgrass clippings.
|
MATERIALS AND METHODS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Clippings for the sampling study were collected for analysis on 12 June, 7 July, and 7 Aug. 2001. The turf was mowed 4 to 6 d before each clipping collection for analysis depending on the growth rate. In past research, significant variations of measured TNC were reported (Sheffer et al., 1979) due to diurnal fluctuations. Hence, diurnal variations were minimized by harvesting consistently early in the photoperiod (before 1200 h). Since past research has indicated a change in TNC levels with photoperiod, care was taken to collect the clippings at the same time each day, so that the relative differences among different TNC components are not affected during different sampling dates.
Collected clippings from each plot were randomly divided into four subsamples, and subjected to one of the following sampling techniques to investigate their effects on the stability of carbohydrates.
Freeze in the Field—Freeze Dry (FF)
Clippings were instantly frozen in the field in liquid nitrogen and immediately placed in a plastic zip-closure bag, held in a dry ice chamber until all samples were collected, then stored at –20°C until they were freeze-dried. All samples were dried in one run in a freeze drier (The Virtis Company, Gardener, NY). Dried samples were later ground with a cyclone mill (Udy Corporation Inc., CO) to pass through 1.0-mm filter.
Freeze in the Field—Oven Air Dry (FA)
This treatment was similar to FF except samples were oven-dried for 48 h at 71°C, instead of freeze-dried.
Slowly Cooled in the Field–Freeze Dry (AF)
Clippings were collected in the field, held in dry ice, later stored at –20°C, before they were freeze-dried.
Held at Ambient Temperature in the Field— Oven Air Dry (AA)
Clippings were collected in the field, held at ambient temperature for 2 h before transporting them to an oven drier, where they were held for 48 h at 71°C.
Laboratory Analysis of Nonstructural Carbohydrates
Mono- and polysaccharides for all samples were determined by a reference chemical analysis described by Westhafer et al. (1982). The advantage of this technique over other analytical techniques was only a single extraction procedure was necessary and physical separation of carbohydrate fractions based on their solubility characteristics was not required. This described method showed very low experimental standard error with an average variance component equal to 1.19% across three different turfgrass species. Though other techniques exist to measure the concentrations of different carbohydrate fractions, they were not evaluated because of their unavailability.
In this method of reference analysis, 60-mg tissue samples were extracted in 0.1 M phosphate buffer (pH 5.5) overnight (about 12 h). Glucose, fructose, sucrose, and fructans were quantified by means of glucose oxidase and invertase enzymes. Reducing sugars (glucose and fructose) and sucrose were determined by absorbance at 540 nm with a Shimadzu double-beam UV-visible spectrophotometer (Scientific Instrument Corporation, Inc., Columbia, MD). Fructan absorbance was determined at 490 nm. The data presented as TNC is the total of all analyzed carbohydrate fractions. A detailed description of this procedure is provided by Westhafer et al. (1982). Tests for F-test significance and least significant differences among sampling technique treatments were calculated by SAS software (Statistical Analysis System, Inc., Cary, NC).
Scanning and Recording Spectral Data for Calibration
Over 2000 samples were available from a field experiment to evaluate TNC dynamics under mowing stress (Narra et al., 2004). To ensure accurate spectral analysis, all samples were dried and ground by the same method as samples used only for wet chemistry analysis. A NIR spectrophotometer (NIRS Systems 5000, Silver Springs, MD) with a wavelength range of at least 1100 to 2500 nm and "small-ring" sample holder with infrared transmitting quartz window was used for scanning all samples used in the calibration procedure. The computer software WINISI 1.04a (Infrasoft International, Inc., State College, PA) was used for collecting, storing and analyzing spectral information. To mirror absorbance values of the master NIR instrument held at Infrasoft International Inc., a standardization cell or check cell was analyzed before each analysis to confirm that the spectrophotometer was working properly. The spectrophotometer was warmed for 1 h before loading any samples for uniform light output by the light source.
The sample holders were cleaned with a high-pressure air sprayer. Additional cleaning was done with soft tissue paper; glass was made free of fingerprints and any fine foreign material. A sample cup was then loaded with ground clippings. About three-quarters of the cup was filled with the ground tissue and a cardboard back was pushed into the holder to press the sample firmly against the window. The filter of the spectrophotometer was cleaned at regular intervals to remove accumulated dust. The spectrophotometer was maintained at a stable temperature through out the analysis. Diffuse reflectance spectral data were acquired at wavelengths from 1108 to 2492 nm with a bandwidth of 2 nm. In total, reflectance was recorded at 700 different wavelengths.
The Calibration Process
The calibration process involves a search for predictive relationships between spectral data and TNC and its component reference values. Though NIRS calibrations were developed for analyzing TNC concentrations in tropical grasses (Brown et al., 1987), no such calibration equations are available for estimating TNC content in creeping bentgrass. The calibration process was initiated by selecting a representative set of samples. Traditionally, these calibration sets were selected on the basis of chemical data rather than spectral properties, which required samples to be analyzed initially by wet chemistry (Harman et al., 1997). However, recent implementation of advanced spectroscopic and chemometrics software has allowed for the selection of calibration samples entirely on the basis of spectral information.
One hundred twelve representative sample spectra were chosen from a set of 765 samples collected during 1999 and 2000 by WINISI 1.04a software. Samples collected during the 1998 and 2001 growing seasons were not included for developing the calibration equation. The samples were collected from all plots in the study described by Narra et al. (2004), which included samples from eight different cultivars mowed at three different heights. In total, 112 selected samples were used for building and validating calibration equations. Principal Component Analysis (PCA) was chosen for this purpose because of its flexibility and reported good prediction of NIR performances (Cowe and McNicol, 1985; Isaksson and Naes, 1987). The algorithm CENTER was used for the calculation of principle components and Mahalanobis distance (H). Mahalanobis distance is a measure of the difference between a sample and the mean value of a population for the multivariant situation. This information was used for the description of spectral boundaries and detection of outliers (Shenk and Westerhaus, 1995). The spectra of all the samples were ranked according to their Mahalanobis distances from the mean spectra of the file, using PCA. The H values, standardized by dividing them by their average values, are called global "H" (GH) values. Spectra in the file were reordered from smallest to largest GH values. Every sample, which had a GH value less than 2.5, was retained for further analysis by wet chemistry methods.
The reference values of TNC and its components for selected calibration samples were determined by the same method used in the sampling technique study (Westhafer et al., 1982) and data were recorded as milligram per gram of dry tissue sample. Data were merged with the sample spectra for selected calibration samples to create a calibration file. This calibration file was later used to build the calibration equation. Since all the constituents were estimated on a dry weight basis, the constituent concentrations entered in the calibration file were also on a dry weight basis.
To test the validity of the calibration model, the whole sample set (112) was randomly split on the basis of different seed values into two sets, one set for calibration (97) and the other for validation (15). The 15-sample set was used to validate the calibration equations for each carbohydrate component. Likewise, six different 15-sample validation sets were generated randomly from the whole set by setting different seed values to validate different calibration sets.
The final equation for each carbohydrate component was derived after omitting the outliers. An optimum equation for each component was chosen according to the criteria outlined by Windham et al. (1989). Optimum equations were identified as those with high coefficients of determination for calibration (r2) and 1 – VR (Variance Ratio), a low standard error of calibration (SEC), and a low standard error of cross validation (SECV).
Developing Predictive Equation
Calibration equations were developed using full spectral data from 1108 to 2492 nm by modified Partial Least Squares (PLS) regression (Shenk and Westerhaus, 1991). The regression intercepts and coefficients of the predictive equations for each carbohydrate component and TNC are given in Table 1.
|
Predicted vs. Reference Values
The 15-sample validation sets were used to compare predicted and reference values. The optimum equation for each of the constituents was used to predict constituent values of the 15 validation samples. Control limits were set for GH (3.00 and 4.00), NH (0.60 and 1.00), and T (2.50 and 3.00) values. The first-value denotes the critical upper limit and second-value the critical lower limit. Slope was set to 0.9, r2 to 0.6; mean and standard deviation were set at 20% difference between lab and NIR-predicted values.
|
RESULTS AND DISCUSSION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
|
The average fructan concentration decreased significantly by 8.5% in samples analyzed by AA techniques when compared with FF techniques. Differences were not observed in fructan concentration between FF and FA techniques, and AF and FA techniques, while fructan concentrations for the AA sampling technique were at least 5.9% lower than the samples that were instantly frozen in the field. Significant differences existed among all the treatments with respect to TNC concentrations. Highest concentration of TNC was observed in samples instantly frozen in the field and later freeze-dried. In general, instant freezing and freeze-drying increased the mean TNC concentration by 6 to 19% (Table 2). This agrees with the research of Rotz and Abrams (1988), where they found nonstructural carbohydrates to be the major components of respiration losses of crop dry matter in alfalfa (Medicago sativa L.). The effect of each of the processes (instant freezing, storage conditions and drying method) was not determined independently.
Calibration statistics
The calibration statistics for different carbohydrate constituents are summarized in Table 3. Although 97 samples (for which reference values were obtained using wet chemistry) were used in calibration models, the final number of samples used in building the calibration model differed with each carbohydrate constituent on the basis of the number of outliers removed during outlier passes. The final number of samples used in building the calibration equation is listed in Table 3.
|
Comparison between Predicted and Lab Values
The predictive ability of the calibration equations was tested with different validation sets, which were earlier separated from the calibration set. The equations for glucose, fructose, fructan, and TNC showed good predictive ability, while the equation for sucrose was less accurate. The predictive statistics for each constituent are outlined in Table 4.
|
|
|
0.97) for different carbohydrate fractions. This indicates very good predictive ability of different equations developed for automatic assessment of creeping bentgrass carbohydrate concentrations. It can be concluded that the described NIRS technique has high potential to estimate the TNC composition of clippings in different creeping bentgrass samples in a nondestructive way and with high degree of accuracy. Moreover, individual concentrations of TNC components such as glucose, fructose, sucrose and fructans may be determined simultaneously by one measurement in a very short span of time. Therefore, a simple, rapid, and reliable overall determination of TNC concentration may be obtained at low cost. Near infrared reflectance spectroscopy allows for huge savings of labor and costs by avoiding sampling, drying, grinding, and traditional laboratory analyses of TNC and its components. This is particularly advantageous if a large number of samples have to be analyzed. On the basis of the positive results of NIRS, a more comprehensive calibration with larger set of samples is justified. |
ACKNOWLEDGMENTS |
|---|
|
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Received for publication February 5, 2004.
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. L. Mentink, P. C. Hoffman, and L. M. Bauman Utility of near-infrared reflectance spectroscopy to predict nutrient composition and in vitro digestibility of total mixed rations. J Dairy Sci, June 1, 2006; 89(6): 2320 - 2326. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Vadose Zone Journal | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
