Near Infrared Reflectance Spectroscopy Prediction Research Articles

Comparison of In Vitro Dry Matter Digestibility and Cellulase Digestion for Deriving Near Infrared Reflectance Spectroscopy Calibration Equations Using Cool‐Season Grasses

Near infrared reflectance spectroscopy (NIRS) has recently become a potentially valuable and reliable tool for analyses of plant samples in forage‐breeding programs. The success of NIRS is dependent on identification of appropriate and reliable selection criteria and development of reliable calibration. The objectives of this study were to develop and compare NIRS analysis equations for the in vitro dry matter digestibility of cool‐season grasses based on either rumen fermentation (IVDMD) or ceilulase‐solubility (CDMD) methods and to determine if NIRS prediction equations developed from these methods would rank selected genotypes in the same order. Breeding nursery samples of smooth bromegrass (Bromus inermis Leyss.) and crested wheatgrasses [Agropyron cristatum (L.) Gaertner and Agropyron desertorum (Fisch) Schultes] were analyzed for IVDMD and CDMD to develop NIRS analysis equations. Forage samples collected at inflorescence emergence from individual genotypes in the nurseries were used to develop a rank comparison of digestibility methods. The IVDMD and CDMD laboratory values were highly correlated (r ≥0.95). Five to nine wavelengths were required to develop the NIRS analysis equations. The coefficients of determination (R2) from regression of laboratory values on NIRS spectra were 0.96 or higher for all equations. Validation procedures indicated excellent correlations between laboratory and NIRS estimates. Spearman rank order correlations of IVDMD‐ or CDMDderived NIRS digestibility values of genotypes were ≥0.77 for both grasses. At least 15 or more genotypes in the top and bottom 20 were the same with either procedure for both grasses. Near infrared reflectance spectroscopy calibrations for in vitro digestibility based on either IVDMD or CDMD procedures will adequately rank genotypes in cool‐season grass breeding programs in a similar order.

Botanical Composition Definition of Tall Fescue‐White Clover Mixtures by Near Infrared Reflectance Spectroscopy1

Near infrared reflectance spectroscopy (NIRS) was evaluated for determining botanical composition of preformed mixtures containing ‘Regal’ iadino clover (Trifolium repens L.) and ‘Kentucky 31’ lowendophyte tall fescue (Festuca arundinacea Schreb.) as well as the botanical composition of samples harvested from grazing paddocks. Both species were grown in separate plots, harvested at monthly intervals, and hand‐mixed. This provided combinations from 100% grass to 100% legume, resulting in a total of 60 samples that were used for NIRS calibration. The NIRS calibration equations for percentage tall fescue and white clover resulted in standard errors of calibration of 2.98 and 2.80%, respectively. Synthetic mixtures composed of material from grazing paddocks were used for validation of the equations. Species composition was predicted with excellent precision and accuracy showing biases of ‐1.64 and 1.80%, and standard errors of prediction (SEP) of 3.16 and 3.08% for the tall fescue and white clover equations, respectively. High R2 was also obtained when predicting the percentage of both component species. The NIRS prediction of pasture samples, whose actual composition was determined by hand separation, yielded higher SEP and bias. This additional variation (compared to the prediction of synthetic mixtures) was assumed to be made up of error associated with the hand‐separation procedure and the variation of the prediction samples not present in the calibration set. These data indicate that with the proper selection of calibration samples, wavelengths, and data transformation, NIRS can be used to determine the botanical composition of grass‐legume mixed samples.

ANALYSIS OF HAY, HAYLAGE AND CORN SILAGE SAMPLES BY NEAR INFRARED REFLECTANCE SPECTROSCOPY

Separate calibrations for hay, haylage and corn silage were developed to predict chemical composition by near infrared reflectance spectroscopy (NIR). A scanning type of NIR instrument was used to select the best set of wavelengths (λ) while a filter type was used to evaluate the calibrations. Reflectance (R) was recorded as log (1/R). Bias (nonrandom error) was corrected for each set of samples before the NIR analysis. Percent crude protein (CP), acid detergent fiber (ADF), calcium (Ca) and phosphorus (P) were studied in the hay samples. In addition, potassium (K) and magnesium (Mg) were included for the haylage and corn silage samples. Six hundred samples, including calibration (C) and prediction sets (PRE1 and PRE2) were used. PRE1 samples came from the same population as the C samples, whereas PRE2 samples were obtained in a different year. Accuracy of the predictions was assessed by the coefficients of determination (r2), standard error of the estimate (SEE), and coefficients of variation (CV). Crude protein was the parameter best predicted by NIR with r2, SEE and CV ranging from 0.72 to 0.96, 0.43 to 1.17 and 5.6 to 10.4, respectively. The highest SEE for crude protein were associated with the PRE2 samples for haylage and hay samples (1.09 and 1.17), respectively. NIR predictions of ADF had r2, SEE and CV values ranging from 0.21 to 0.92, 1.44 to 2.53 and 5.3% to 7.9%, respectively. Corn silage had the lowest SEE for ADF in both C and PRE1 sets. Predicting mineral contents by NIR gave high CV (10.5%–34.5%) and low r2 values (0.02–0.75). Calcium predictions had the highest variability, and P and Mg predictions the lowest.These results indicate that CP was successfully predicted by NIR. The higher SEE values for ADF may have been due to variation in the wet chemistry values of some samples. Minerals were not adequately predicted by NIR as assessed by r2, SEE and CV values. Key words: Near infrared reflectance spectroscopy, forage, chemical analysis

Quality Prediction of Small Grain Forages by Near Infrared Reflectance Spectroscopy1

A paucity of information exists concerning the utility of near infrared reflectance spectroscopy (NIRS) for prediction of forage quality of annual forage crops and concerning the influence of environmental and plant factors on NIRS prediction equation development and use. Our objective was to ascertain whether NIRS can be relied upon to predict the quality of small grain crop forages at diverse maturation stages. We assayed five forage quality traits in a total of 700 forages comprised of up to four field replicates of two cultivars each of oats (Arena sativa L.), barley (Hordeum vulgate L.), spring wheat (Triticum aestivum L.), and triticale (Triticum durum Desf. ✕ Secale cereale L.) each harvested at six growth stages in two locations in years. Empirical quality prediction equations were obtained by multiple linear regression of known quality values on NIR values from 48 samples obtained from one replicate from one location in 1 year. A scanning monochromator NIR spectro‐computer system was used to develop the prediction equations and to validate them via tests of their applicability to the remaining 652 samples that were grown in other replicates or in other environments. From two to six wavelengths were needed to develop the best prediction equations for crude protein (CP), neutral detergent fiber (NDF), acid detergent fiber (ADF), acid detergent lignin (ADL) or in vitro digestible dry matter (IVDDM). The squared coefficients of multiple determination (R2) of the prediction equations ranged from 0.92 for IVDDM to 0.99 for CP. These calibration equations predicted the quality values for samples from other replicates, locations, and years within standard errors of ± 0.7 to 1.1% for CP, ± 1.8 to 2.3% for NDF, ± 1.0 to 1.6% for ADF, ± 0.3 to 0.4% for ADL, and ± 3.1 to 4.1% for IVDDM. We concluded that NIRS can evaluate forage quality of small grain crops to a degree of accuracy equivalent to or greater than that reported for perennial forage species, and that relative forage quality of small grain crops was adequately determined for several growth environments when prediction equations were developed from forages grown in a single environment.