Estimates of diet selection in cattle grazing cornstalk residues by measurement of chemical composition and near infrared reflectance spectroscopy of diet samples collected by ruminal evacuation

Abstract

Six ruminally cannulated cows (570 ± 73 kg) fed corn residues were placed in a 6 × 6 Latin square to evaluate predictions of diet composition from ruminally collected diet samples. After complete ruminal evacuation, cows were fed 1-kg meals (dry matter [DM]-basis) containing different combinations of cornstalk and leaf and husk (LH) residues in ratios of 0:100, 20:80, 40:60, 60:40, 80:20, and 100:0. Diet samples from each meal were collected by removal of ruminal contents after 1-h and were either unrinsed, hand-rinsed or machine-rinsed to evaluate effects of endogenous compounds on predictions of diet composition. Diet samples were analyzed for neutral (NDF) and acid (ADF) detergent fiber, acid detergent insoluble ash (ADIA), acid detergent lignin (ADL), crude protein (CP), and near infrared reflectance spectroscopy (NIRS) to calculate diet composition. Rinsing type increased NDF and ADF content and decreased ADIA and CP content of diet samples (P < 0.01). Rinsing tended to increase (P < 0.06) ADL content of diet samples. Differences in concentration between cornstalk and LH residues within each chemical component were standardized by calculating a coefficient of variation (CV). Accuracy and precision of estimates of diet composition were analyzed by regressing predicted diet composition and known diet composition. Predictions of diet composition were improved by increasing differences in concentration of chemical components between cornstalk and LH residues up to a CV of 22.6 ± 5.4%. Predictions of diet composition from unrinsed ADIA and machine-rinsed NIRS had the greatest accuracy (slope = 0.98 and 0.95, respectively) and large coefficients of determination (r² = 0.86 and 0.74, respectively). Subsequently, a field study (Exp. 2) was performed to evaluate predictions of diet composition in cattle (646 ± 89 kg) grazing corn residue. Five cows were placed in 1 of 10 paddocks and allowed to graze continuously or to strip-graze corn residues. Predictions of diet composition from ADIA, ADL, and NIRS did not differ (P = 0.99), and estimates of cornstalk intake tended to be greater (P = 0.09) in strip-grazed compared to continuously grazed cows. These data indicate that diet composition can be predicted by chemical components or NIRS by ruminal collection of diet samples among cattle grazing corn residues.

INTRODUCTION

Diet selection among cattle grazing forage can have large impacts on animal performance and efficient conversion of plant tissues to body mass when bulk-fill mechanisms limit caloric intake (National Academies of Sciences, Engineering and Medicine, 2016). Annual forages (e.g., corn residues, simple mixtures of cover crops) often have large differences in nutrient content between different plant components and composition of diets selected by grazing cattle may have large impacts on nutrient digestion (Brunsvig et al., 2017). Selection of plant tissues that facilitate optimal rates and extent of digestion among grazing cattle can ameliorate limits in dry matter intake (DMI). Unfortunately, accurate measures of plant tissues selected by grazing cattle are difficult to obtain in comparison to cattle that are housed individually and fed in confinement. As a result, a number of indirect methods (e.g., clipping forage before and after grazing, transect surveys of plant populations) have been developed to estimate diet selection among grazing cattle; however, indirect methods of diet selection have inherent amounts of inaccuracy because indirect methods of diet selection fail to account for grazing by indigenous herbivores (e.g., nondomesticated ruminants, insects, rodents) and diet selection estimates include forage that decomposes between measurements (Van Soest, 1994) or is removed from the grazing area by wind. Conversely, measures of diet selection from cattle fitted with ruminal or esophageal cannulas allow direct measures of diet selection in grazing cattle. Direct identification of masticated diet samples via histological measures of ingested plant tissue collected from cannulated cattle is difficult and can be imprecise without proper training (Slater and Jones, 1971; Holechek et al., 1982; Alipayo et al., 1992). Several authors (Dove and Mays 1996; Kelman et al., 2003; Boland et al., 2012) have suggested that plant waxes in feces may facilitate accurate estimates of diet selection in grazing cattle; however, attempts to estimate diet selection from fecal recoveries of plant waxes are largely variable (Heublein et al., 2017) and it is unclear if fecal recoveries of plant waxes are complete (Dove and Mays, 2005). Diet samples collected from ruminal evacuation are collected before any appreciable amounts of fermentation of ingested plant tissues can occur (Towne et al., 1986; Wells and Russell, 1996). Thus, it is likely that few nutrients are lost from diet samples collected via ruminal evacuation. Yet, diet samples collected by ruminal evacuation contain nutrients from endogenous sources (e.g., saliva, sloughed epithelia; Van Soest, 1994). Methods that mitigate contamination of diet samples collected from ruminal evacuation may allow for accurate estimates of nutrient composition of diets and increase precision of estimates of diet selection based on chemical composition of diet samples. An increased understanding of diet selection among cattle grazing corn residues and other simple mixtures of annual forages could allow for development of nutritional and management strategies that optimize utilization of these plant tissues by cattle.

MATERIALS AND METHODS

Experiment 1: Validation Study

Animal husbandry and sample collection. All procedures that involved the use of animals in this project were approved by the South Dakota State University Institutional Animal Care and Use Committee (protocol approval No. 17-040A).

Beginning 16 d prior to experimentation, 2 ruminally cannulated Angus cows (531 ± 44 kg; 2.5 ± 0.5 yr) and 4 ruminally cannulated Angus × Simmental cows (615 ± 35 kg; 2.25 ± 0.4 yr) housed in a common drylot pen (0.4 ha) were fed long-stem mechanically harvested corn residue (84.2% DM, 3.6% CP, and 72.3% NDF) to ad libitum and allowed free choice access to pressed mineral and vitamin (Prairie Pride 4% Mineral Block, Ridley Inc, Mankato, MN; 20% Ca, 12% NaCl, 4% P, 1,000 ppm Zn, 100 ppm Cu, 36 ppm I, 36 ppm Se, 143,300 IU/kg Vitamin A, and 35,932 IU/kg Vitamin D₃) and water. Subsequently, cattle were moved to individual pens (1.65 × 2.54 m) in a temperature (19°C) and light-controlled (16 h of light daily) room at the South Dakota State University Ruminant Metabolism Facility and fed the same long-stem corn residues at 0800 h and 2000 h daily with free choice access to the same pressed mineral and vitamin and water. Cows were housed individually for 2 d and then placed in a 6 × 6 Latin square to evaluate predictions of diet intake of cornstalk and leaf and husk (LH) residues via measures of several innate chemical components in diet samples.

Measures of diet selection were accomplished by completely evacuating ruminal contents from cows at 1900 h daily for 6 d. Immediately after evacuation of ruminal contents (Reid, 1965), cows were fed meals (1 kg DM) composed of various combinations of cornstalk and LH residues (Table 1). Cornstalk and LH residues were previously ground (Thomas Wiley Mill Model 4; Thomas Scientific USA, Swedesboro, NJ) to pass a 6 mm screen and each meal was offered in 250 g (DM basis) aliquots to eliminate sorting of each component from meals by cows. Meals fed to cows consisted of different combinations of cornstalk and LH residues in ratios of 0:100, 20:80, 40:60, 60:40, 80:20, and 100:0. Additionally, LH residues were provided as a mixture of leaves (78.7%) and husk (21.3%) in amounts identical to the composition of individual corn plants measured in Exp. 2 and were similar to previous reports of relative amounts of LH commonly observed in corn residues (Stalker et al., 2015). Samples (50 g DM) of each meal were collected daily, composited and ground to pass a 1 mm screen (Thomas Wiley Mill Model 4) for determination of DM, organic matter (OM), NDF, ADF, acid detergent insoluble ash (ADIA), ADL, and N. Replicate measurements of mechanically harvested biomass were determined by quadrat sampling (n = 10, 0.25 m²) after physically unrolling a cornstalk bale. Within the quadrat, plant biomass was collected and subsequently separated into stalk, leaf, husk, and cob prior to DM determination. Corn grain was not included in measures of harvested biomass in cornstalk bales because no corn grain was recovered.

Table 1.

Nutrient composition of cornstalk and LH residues

Item	OM	NDF	ADF	ADIA1	ADL	CP
Exp. 1
Cornstalk	97.09 ± 0.03	78.06 ± 0.95	54.55 ± 0.50	2.39 ± 0.23	9.40 ± 0.11	3.30 ± 0.04
LH	94.65 ± 0.59	72.13 ± 0.27	43.01 ± 0.71	7.91 ± 0.31	4.75 ± 0.59	3.72 ± 0.10
Exp. 2
Cornstalk	94.96 ± 0.78	73.76 ± 5.46	49.66 ± 4.58	2.31 ± 0.70	8.63 ± 0.96	3.08 ± 0.28
LH	87.54 ± 3.14	67.88 ± 1.98	40.19 ± 1.88	8.81 ± 2.78	6.29 ± 2.58	4.82 ± 0.52

¹Acid detergent insoluble ash.

Diet samples from each meal were collected by removal of ruminal contents after 1 h (Harris et al., 1967; Gelvin et al., 2004; Kirch et al., 2007) and ruminal contents removed prior to diet sampling were replaced. Diet samples were weighed and an aliquot (250 g) was dried (55°C for 36 h). Concurrently, another aliquot (250 g) was hand-rinsed by wrapping diet samples in 4 layers of cheesecloth and slowly vortexing samples by hand for a total of 30 revolutions in 4 L of clean tap water prior to drying (55°C for 36 h). After each aliquot of diet samples were partially dried, diet samples were ground to pass a 1 mm screen (Thomas Wiley Mill Model 4; Thomas Scientific USA) and 4 g of each hand-rinsed diet sample was then placed into polyethylene bags (10 × 20 cm, pore size = 50 ± 10 µm; Dacron, Ankom Technology, Fairport, NY) and machine-rinsed 5 times in a commercial washer (Fabric-Matic, Model A511S, Maytag, Newton, IA); each rinse consisted of a 1 min. rinse followed by a 2 min. spin cycle (Vanzant et al., 1998). After machine-rinsing, samples were dried at 55°C for 24 h.

Laboratory Analyses

Samples of unrinsed, hand-rinsed, and machine-rinsed diet samples were analyzed in duplicate for DM, OM, N, NDF, ADF, ADIA, and ADL. Dry matter was measured by drying at 105°C for 16 h, and OM was determined by combustion (500°C for 16 h). Additionally, N content was analyzed by the Dumas procedure (method no. 968.06; AOAC, 2016; rapid Max N exceed; Elementar, Mt. Laurel, NJ). Neutral detergent fiber was measured as described by Van Soest et al. (1991) and included additions of α-amylase and sodium sulfite; ADF was measured nonsequential to NDF (Van Soest et al., 1991), and ADIA was calculated by combustion (500°C for 16 h) of ADF residue. Acid-detergent lignin was measured after thoroughly soaking ADF residue in 72% (wt/wt) sulfuric acid for 3 h and agitating ADF residue in acid each 30 min (Van Soest and Robertson, 1980). Measures of NDF, ADF, and ADL were corrected for ash content which was measured by combustion (500°C for 8 h).

NIRS

Near infrared reflectance spectroscopy (NIRS) was recorded at 2-nm intervals from 1,100 to 2,498 nm with a Foss NIRS 5000 (Foss Inc., Eden Prairie, MN). Predictions of ratios of cornstalk:LH residues were calculated from standards (n = 50) that ranged from 100:0 to 0:100 cornstalk:LH residues; each intermediate standard was created by replacing 2% of cornstalk with LH residues. Each fifth standard was excluded from development of the predictive model to subsequently evaluate accuracy and precision of predictions. A predictive model was developed from the full spectrum using WINISI II (version 1.02a; FOSS INC, Eden Prairie, MN) by regression of modified partial least squares means. Subsequently, NIRS predictions of ratios of cornstalk:LH residues were corrected for particle size using standard normal variate (SNV) and Detrend (WINSI II; FOSS INC). Curve fitting math treatments were analyzed with the math treatment 1, 4, 4, 1 (derivative, gap, smooth, and smooth 2), which generated the most robust equation (SE of calibration equation = 0.40, R² = 0.99, SE of the cross validation = 1.82, coefficient of determination in the cross-validation = 0.99). The predictive model was validated by regression of NIRS predictions of ratios of cornstalk:LH residues with known amounts of cornstalk:LH residues in the 10 cornstalk:LH residue standards removed from model development (slope = 1.01, bias = 0.036, SE of prediction = 2.5, r² = 0.99). Predictions of ratios of cornstalk:LH residues in diet samples were obtained by scanning each diet sample and comparing the spectra to the reference values.

Nucleic Acids and 16S rRNA Gene

Nucleic acids were isolated from samples using the repeated bead beating plus column method, as previously described by Yu and Morrison (2004). Briefly, 86.46 ± 11.86 mg of dried, ground diet sample were lysed by bead beating with 0.4 g of zirconium beads at 3,000 rpm (3 min) in lysis buffer (0.5 M NaCl, 50 mM Tris-HCl, 50 mM ethylenediaminetetraacetic acid, 4% sodium dodecyl sulfate), followed by heat treatment at 70°C (15 min). Lysate was recovered by centrifugation (16,000 × g, 5 min, 4°C). Impurities and SDS were removed by ammonium acetate precipitation (10 M, 20% volume). Deoxyribonucleic acid was recovered from the lysate using isopropanol precipitation and then further purified using column filtration (QIAamp DNA Stool Kit, QIAGEN GmbH, Hilden, Germany). Nucleic acid concentration was determined using the NanoDrop 2000 Spectrophometer (ThermoScientific, Waltham, MA).

Quantitative polymerase chain reactions were carried out with the Stratagene Mx3005P Thermocycler (Agilent Technologies, Santa Clara, CA) in a 20 µL volume containing the following reagents: 10 µL SYBR green 1 Step 2xiTaq (BioRad, Hercules, CA), 2 µL of each primer (40 µM 1114F and 1275F; Denman and McSweeney, 2006), 1 µL template DNA (at 10 ng/µL concentration), and 5 µL H₂0. Polymerase chain reaction amplification was initiated by a hot start at 95°C for 3 min, followed by 40 cycles of 95°C for 10 s and 60°C for 30 s. Cycles to threshold (Ct) were determined by the MxPro-Mx3005 software (V4.10).

Calculations

Dry matter was calculated as partial DM (55°C for 36 h) multiplied by DM measured after drying at 105°C for 16 h. Composition of mechanically harvested bales of corn residues was calculated (DM-basis) as the quotient of each corn residue and total biomass. Differences in concentration of each chemical component between cornstalk and LH residue were standardized by calculating a coefficient of variation to compare effects of greater differences in concentration of chemical components between cornstalk and LH residues on predicted diet composition. The average concentration of each chemical component was calculated across cornstalk and LH residues, and a standard deviation was calculated between the concentration of a chemical component in cornstalk or LH residue. Coefficients of variation were calculated as the quotient of the standard deviation and the average of the nutrient concentration within each chemical component (NDF, ADF, ADIA, ADL, or CP). The absolute value of differences in concentrations of NDF, ADF, ADIA, ADL, or CP in diet samples and cornstalk (∆_cornstalk) or LH (∆_LH) residues were calculated, and proportional intakes of cornstalk and LH residue were calculated as:

$$\mathrm{Pr}oportional cornstalk \mathrm{int}ake=\frac{{\Delta }_{LH}}{{\Delta }_{cornstalk}+{\Delta }_{LH}}$$

$$Proportional LH intake=\frac{{\Delta }_{cornstalk}}{{\Delta }_{cornstalk}+{\Delta }_{LH}}$$

The ratio of cornstalk:LH was calculated as the quotient of proportional cornstalk intake and proportional LH intake using data from treatments containing 20:80, 40:60, 60:40, and 80:20 cornstalk:LH residue. Predicted ratios of cornstalk:LH were regressed on known ratios of cornstalk:LH using the MIXED procedures of SAS (SAS Inst. Inc., Cary, NC). The correlation coefficient of regressions were calculated using the MIXED procedures of SAS as one minus the quotient of the sum of the squared error in the full model and the sum of the squared error in the reduced model (Kuehl, 2000). Initially, effects of greater differences between chemical components on accuracy of predictions of diet composition were evaluated by linearly regressing the correlation coefficient and coefficients of variation. Subsequently, effects of greater differences between chemical components on accuracy of predictions of diet composition were evaluated with a 2-slope straight broken-line model (Robbins et al., 1979) using the NLIN procedure of SAS because the plot of correlation coefficients and coefficients of variation appeared to be asymptotic. Cycles to threshold from each diet samples were calculated relative to amounts of 16S rRNA gene provided from plant chloroplasts by calculating the absolute value of the difference between the diet sample Ct value and the Ct value from plant chloroplasts in each meal (ΔCt). Relative 16S rRNA gene values were calculated as 2^ΔCt times the proportional DNA concentration.

Statistical Analysis

Data from 1 cow in a single period was removed from statistical analysis of measures of nucleic acids and rRNA gene data because it did not appear to be a component of a normal population (studentized residual = 5.7).

Data were analyzed for a Latin square using the MIXED procedure of SAS with the following model:

$${\text{Y}}_{\text{ijkl}}=\text{μ}+{\text{R}}_{\text{i}}+{\text{M}}_{\text{j}}+{\text{RM}}_{\text{ij}}+{\text{C}}_{\text{k}}+{\text{P}}_{\text{l}}+{\text{ε}}_{\text{ijkl}}$$

where Y_ijkl = the dependent variable, µ = the overall mean, R_i = the fixed effect of rinse type (i= 1, 2, 3), M_j = fixed effect of the ratio of cornstalk:LH residues in meal (j = 1, . . . ., 6), RM_ij = fixed effect of the interaction of rinse type and meal, C_k = the random effect of cow (k = 1, . . . ., 6), P_l = the random effect of period (l = 1, . . . ., 6); and ε_ijkl = residual error. Denominator degrees of freedom were calculated by the Kenward and Roger adjustment (Kenward and Roger, 1997). Treatment means were calculated using the LSMEANS option. Effects of different combinations of cornstalk and LH residues in each meal were determined with linear and quadratic contrasts. Effect of rinse type was evaluated with the F-statistic. When the F-statistic was significant (P ≤ 0.05) means were separated using the Student’s t-test with the PDIFF option of SAS.

Experiment 2: Field Study

Animal husbandry and sample collection.

All procedures that involved the use of animals in this project were approved by the South Dakota State University Institutional Animal Care and Use Committee (protocol approval No. 16-083A).

Beginning November 29, 2016 (64 d after grain harvest), a trial was conducted among cows grazing corn residues 6.9-km northeast of Brookings, SD (44°22′23.63′′ N 96°45′46.46′′ W) to evaluate predictions of ratios of cornstalk:LH residue intake with ADIA, ADL, or NIRS. After grain harvest, a corn residue field was divided into 10 paddocks (2.02 ha). Temperature and precipitation data were obtained from a weather station located 4.82 km south of the grazing site. Chemical composition of grazed corn residues (Table 1) was determined by analysis of biomass from 10 quadrats (0.25 m²) stratified across the field and collected 18 d prior to grazing. Quadrat biomass was stored at 60°C and separated into cornstalk, leaf and husk prior to laboratory analysis. All corn residue samples were ground to pass a 1 mm screen (Thomas Wiley Mill Model 4; Thomas Scientific USA). Cornstalk and LH residues were analyzed for DM, OM, NDF, ADF, ADIA, ADL, and CP as described in Exp. 1.

Grazing treatments consisted of continuous (3.5 animal unit days/ha) or strip-grazing (169 animal unit days/ha). Treatments were designed to achieve an identical stocking rate (5.6 animal unit month/ha) in a 48 d grazing period, but allowed comparison of a nearly 50 times greater stocking density in strip-grazing paddocks compared with continuous grazing paddocks. Twenty-one Angus (705 ± 61.3 kg, 5 ± 2 yr) and 19 Angus × Simmental (654 ± 41.1 kg, 5 ± 2 yr) cows were blocked by BW. Subsequently, 4 cows within each BW block were randomly assigned to paddocks. An additional 6 Angus (483 ± 23.2 kg, 2 yr) and 4 Angus × Simmental (544 ± 24.6 kg, 2 yr) cows previously fitted with ruminal cannulas were randomly assigned to each paddock to allow concurrent measures of diet selection. Thus, each paddock contained a total of 5 cows. Strip-grazed cattle were given access to additional corn residues at 0700 h and 1900 h daily, and all cows were individually fed 358 g DM of a common protein supplement that contained (DM-basis) 88.9% soybean meal, 4.5% molasses, 4.5% urea, 2.1% Cr₂O₃ at 0700 h daily and was designed to meet RDP requirements (NRC, 1996). After 7 d, diet samples were collected once, beginning at 0700 h by total ruminal evacuation (Reid, 1965). After removal of rumen contents, cows were allowed to graze for 45 min and rumen contents were removed, weighed and a sample was frozen (−20°C) prior to analysis of DM, ADIA and ADL. Rumen contents removed prior to diet sampling were replaced before cows were returned to paddocks.

Unrinsed diet samples were dried at 55°C for 36 h and then ground to pass a 1 mm screen (Thomas Wiley Mill Model 4; Thomas Scientific USA). An aliquot of unrinsed diet sample was machine-rinsed as previously described in Exp. 1 and subsequently dried at 55°C for 36 h. Measures of ADIA content in unrinsed diet samples and ADL content and NIR spectra were quantified in machine-rinsed diet samples as described in Exp. 1.

NIRS

A NIRS model was developed to predict ratios of cornstalk:LH residues using procedures described in Exp. 1, but the predictive model was developed from cornstalk and LH residues collected from quadrat samples in Exp. 2. The NIRS predictive model used the 2, 10, 10, 1 math treatment with SNV and Detrend (SE of calibration = 1.98, r² = 1.00, SE of cross validation = 2.16, coefficient of determination in the cross-validation = 0.99). Identical to Exp. 1, the NIRS predictive model was validated (slope = 1.01, bias = −0.62, SEP = 2.81, r² = 0.99) by regression of NIRS predictions of ratios of cornstalk:LH residues with known amounts of cornstalk:LH residues in 10 standards removed from development of the predictive model.

Calculations

Dry matter was calculated as partial DM (55°C for 36 h) multiplied by DM measured after drying at 105°C for 16 h. Relative amounts of each corn residue in quadrat samples were calculated on a DM basis as the quotient of amounts of each residue and total biomass. Diet composition of ratios of cornstalk:LH residues were calculated as described in Exp. 1.

Statistical Analysis

Diet samples from 1 cow in a continuous grazing paddock were removed from predictions of diet composition because this cow was observed grazing outside of its paddock and predictions of diet composition did not appear to be representative of a normal population (studentized residual = 5.1). Data from each paddock were analyzed as a randomized block design using the MIXED procedure of SAS with the following model:

$${\text{Y}}_{\text{ijk}}=\text{μ}+{\text{T}}_{\text{i}}+{\text{A}}_{\text{j}}+{\text{TA}}_{\text{ij}}+{\text{B}}_{\text{k}}+{\text{ε}}_{\text{ijk}}$$

Where Y_ijk = the dependent variable, µ = the overall mean, T_i = the fixed effect of grazing management treatment (i= 1 or 2), A_j = the fixed effect of chemical component and NIRS (j= 1, 2, or 3), TA_ij = the fixed effect of the interaction of treatment and effect of chemical component and NIRS, B_k = the random effect of block (k = 1, . . . . 5); and ε_ijk = residual error. Denominator degrees of freedom were calculated by the Kenward and Roger adjustment (Kenward and Roger, 1997). When the F-statistic was significant (P ≤ 0.05) means were separated using the Student’s t-test with the PDIFF option of SAS.

RESULTS AND DISCUSSION

Experiment 1

Corn residues remaining after grain harvest are a heterogeneous mixture of stalks, leaves, husks, and cobs. Stalker et al. (2015) reported that corn residues are comprised predominately (DM basis) of stalks and leaves (40.5% and 35.1%, respectively) and that the remainder is husks (9.6%) and cobs (14.8%). A similar composition of harvested corn residues fed in Exp. 1 (stalk = 50 ± 1.8%, leaves = 27 ± 1.0%, husk = 15 ± 1.0%, cob = 9 ± 1.3%) and in corn residues grazed in Exp. 2 (stalk = 35 ± 3.2%, leaves = 47 ± 2.3%, husk = 11 ± 0.7%, cob = 7 ± 2.4%) was observed compared to the measures of Stalker et al. (2015).

Generally, diets of cows grazing corn residues are comprised of cornstalk and LH residues because grain and cobs do not contribute measurable amounts to the diet (Lamm and Ward, 1981; Gutierrez-Ornealas and Klopfenstein, 1991; Stalker et al., 2015). Perhaps it is not surprising that only small amounts (e.g., <1%; Humberg et al, 2009) of grain remain in corn residue after grain harvest because harvest of grain from corn is often the primary incentive to corn production. Stalker et al. (2015) reported that grain is only a minor component of total corn residues available to grazing cattle, and Gutierrez-Ornealas and Klopfenstein (1991) determined that corn grain remaining in corn residues is rapidly consumed by grazing cattle. Subsequently, these authors (Gutierrez-Ornealas and Klopfenstein, 1991) concluded that corn grain is unlikely to provide measurable amounts of nutrients to cattle grazing corn residues. Similar to grain, cobs apparently do not contribute to diets of cattle grazing corn residues (Lamm and Ward, 1981; Stalker et al., 2015). Cobs contain large amounts of lignin (12%; Pointner et al., 2014), small amounts of CP (2.6% CP; Fernandez-Rivera and Klopfenstein, 1989), and in vitro estimates of DM digestion are small (36%; Fernandez-Rivera and Klopfenstein, 1989). Thus, it is not surprising that cattle grazing corn residue avoid intake of cobs (Lamm and Ward, 1981; Stalker et al., 2015), and that several authors (Lamm and Ward, 1981; Fernandez-Rivera and Klopfentstein, 1989) have limited estimates of diet selection among cows grazing corn residues to that of LH and cornstalk residue alone.

Greater differences in concentrations of a chemical component between each constituent in a binary mixture of plants or plant components may allow for improved predictions of diet composition from diet samples collected from ruminally cannulated cattle. Improvements in direct measures of diet selection from ruminally cannulated cattle may be useful in estimating diet selection among cattle grazing annual forages other than corn. It is reasonable that greater differences in concentrations of a chemical component should reduce impacts of residual variance in analysis on predictions of diet composition from a binary mixture if residual variance in analysis is similar across measures of each chemical component. Differences in concentration of each chemical component (i.e., NDF, ADF, ADL, ADIA, and CP) between cornstalk and LH residues were standardized by calculating a coefficient of variation (CV) within each chemical component. Correlation coefficients were calculated by regression of the predicted cornstalk:LH ratio from each combination of rinsing and chemical component and diet samples with known ratios of cornstalk:LH. Subsequently, the correlation coefficient and differences in concentration of each chemical component between cornstalk and LH residues (differences were standardized across each chemical component by calculating CV) were regressed to evaluate effects of increased differences in concentrations of each chemical component on predictions of diet composition.

The relationship between correlation coefficients and differences in concentration of chemical components did not follow a linear model. Linear regression of correlation coefficients and standardized differences in chemical components between cornstalk and LH residue described only a modest amount of variation (r² = 0.51). Furthermore, when linear estimates accurately describe observations then residuals should be evenly dispersed about each estimate. Residuals from linear regression of the correlation coefficients calculated from predictions of diet samples compared to the known diet composition on CV between concentration of each chemical component were plotted (Fig. 1). Residuals from the linear model formed an uneven distribution (i.e., U-shaped; Fig. 1) and failed the lack-of-fit test (P = 0.03). Therefore, a 2-slope nonlinear approach (Robbins et al., 1979) was used to evaluate improvements in coefficients of determination among estimates of diet composition in response to greater differences in concentration of chemical components in cornstalk and LH residues. The nonlinear model contributed to a nearly 55% improvement in the estimates of variance (pseudo-r² = 0.79) than linear estimates (r² = 0.51).

Figure 1.

Residual plot of the linear regression of the coefficient of variation of chemical components in cornstalk and LH residues and the coefficient of determination of the predicted ratio of cornstalk:LH residues in diet samples.

Predictions of diet composition from chemical components with greater differences in concentration between cornstalk and LH residues improved linearly up to a CV of 22.6 ± 5.4% (Fig. 2), but there was no improvement in predictions of diet composition by use of chemical components that had greater differences in concentration between cornstalk and LH residues. Nonetheless, these data indicate that much of the variation between predictions of diet composition and known diet composition (r² = 0.86 ± 0.11) can be explained when diet composition is estimated from chemical components that have large differences (CV ≥ 22.6%) in concentration between cornstalk and LH residues. It is likely that predictions of diet composition were improved from chemical components with larger differences in concentration, because large differences in concentration between cornstalk and LH residues likely diminished effects of residuals associated with laboratory analyses on predictions of diet composition. However, factors other than residuals from laboratory analyses probably limited further improvements in predictions of diet composition (e.g., variation related to sample collection). It is important to note that the discontinuous first derivative utilized in the 2-slope nonlinear method only provides an approximation of the data. Therefore, caution is suggested when predicting diet composition from chemical components that differ in concentrations near to the point that has been calculated to provide the optimal correlation between predicted and known diet composition unless experimental methods and conditions closely match those presented herein.

Figure 2.

Fitted broken-line plot of coefficient of determination as a function of the coefficient of variation between cornstalk and LH residues.

Nonetheless, these data provide strong evidence that diet composition can be predicted from different chemical components in diet samples from ruminally cannulated cows grazing binary mixtures of forage when there is a sufficient difference between the compositions of the 2 parts in the binary mixture. Brunsvig et al. (2017) previously measured effects of stocking density on diet selection of heifers grazing binary mixtures of cool-season annual forages from NDF content of diet samples. Diet composition predicted from NDF was poorly correlated to known diet composition in this study; however, differences in NDF between cornstalk and LH residues were small (CV = 5.6%) whereas differences in NDF between brassica and grass in the study of Brunsvig et al. (2017) were more than 2 times the amount of difference predicted in this study to allow optimal accuracy in prediction of diet samples among cattle grazing binary mixtures of forage. In this study, ADIA and ADL were the only chemical components that had large enough differences (CV = 75.8% and CV = 46.5%, respectively) in concentration between cornstalk and LH residues to allow for close correlations between predicted and known diet composition.

Effect of Rinsing

A large number of studies investigating nutrient disappearance from ruminally incubated polyester bags have used various rinsing techniques to remove endogenous nutrients and to improve estimates of nutrient disappearance from feed (Balch and Johnson, 1950; Coblentz et al., 1997; Vanzant et al., 1998). Chemical composition of diet samples that were hand-rinsed, machine-rinsed, or unrinsed were measured to determine impacts of endogenous contributions from cattle on chemical composition of diet samples. There was no interaction of rinsing and meal composition (Table 2). Concentration of NDF and ADF increased (P < 0.01; Table 2) with greater rinsing, and chemical composition of ruminally evacuated diet samples more nearly matched chemical composition of feed offered. Similarly, ADL concentration of diet samples (Table 2) tended to increase (P = 0.06) and more nearly matched concentrations of ADL in feed offered after greater rinsing. However, CP concentration in diet samples decreased (P < 0.01; Table 2) with greater amounts of rinsing, and ADIA content was less after machine-rinsing but hand-rinsing had lesser impacts on loss of ADIA from diet samples. Cattle saliva contains a myriad of soluble proteins and mucopolysaccharides (Bailey and Balch, 1961; Boda et al. 1965; Bartley, 1976). Additionally, many epithelia lining the ruminal and esophageal lumen and resident ruminal microbiota contain protein. It is likely that greater amounts of rinsing removed soluble proteins and endogenous components that were loosely (e.g., saliva, epithelia, ruminal microbiota) associated with diet samples (Nocek, 1985; DeBoer et al., 1987; McAllister et al., 1994). Indeed, measures of nucleic acids were 88% and 85% less in machine-rinsed diet samples in comparison to measures of nucleic acids in hand-rinsed and unrinsed diet samples (Table 3). Amounts of 16S rRNA gene relative to amounts for 16S rRNA gene from plant chloroplasts in machine-rinsed diet samples were 98% and 96% less than (P = 0.02) than relative amounts of 16S rRNA gene in unrinsed and hand-rinsed diet samples, respectively (Table 3). Overall, these data suggest that machine-rinsing was able to remove most proteinaceous contribution from both epithelia and ruminal bacteria.

Table 2.

Effects of rinsing on chemical composition in diet samples in Exp. 1

	Cornstalk:LH ratio							P-value
Item, % of DM	0:100	20:80	40:60	60:40	80:20	100:0	SEM	Rinse	Linear	Quadratic	Interaction 3
NDF1,a								<0.01	<0.01	0.78	0.79
Unrinsed	68.5	68.7	70.0	71.8	71.9	73.6	1.04
Hand-rinsed	69.6	70.0	72.9	75.2	74.1	77.1	1.04
Machine-rinsed	76.0	77.4	79.1	80.3	81.1	82.4	1.04
ADF1,a								<0.01	<0.01	0.46	0.95
Unrinsed	43.7	45.9	47.8	49.2	53.0	54.2	1.05
Hand-rinsed	43.6	47.3	49.0	51.7	53.3	55.6	1.05
Machine-rinsed	45.8	48.3	51.0	52.8	55.0	56.9	1.05
ADIA2,b								<0.01	<0.01	0.12	0.47
Unrinsed	7.6	6.7	6.0	4.7	3.6	2.8	0.29
Hand-rinsed	7.4	6.8	5.7	4.9	4.0	2.8	0.29
Machine-rinsed	5.4	4.8	4.1	3.4	2.3	1.5	0.29
ADL1,c								0.06	<0.01	0.85	0.70
Unrinsed	4.9	5.7	6.5	6.5	7.7	8.6	0.26
Hand-rinsed	4.9	5.6	6.6	6.9	7.5	8.2	0.26
Machine-rinsed	5.0	5.9	6.6	7.3	8.1	8.6	0.26
CPd								<0.01	0.01	0.27	0.84
Unrinsed	4.3	4.2	4.2	3.9	4.1	3.9	0.15
Hand-rinsed	3.8	3.9	3.7	3.4	3.8	3.6	0.15
Machine-rinsed	2.8	2.6	2.5	2.5	2.5	2.3	0.15

¹Ash associated with fiber removed from calculation.

²Acid detergent insoluble ash.

³Rinse × cornstalk:LH ratio.

^aUnrinsed < hand-rinsed < machine-rinsed

^bUnrinsed = hand-rinsed > machine-rinsed

^cUnrinsed = hand-rinsed < machine-rinsed

^dUnrinsed > hand-rinsed > machine-rinsed

Table 3.

Effect of rinsing on nucleic acid and 16S rRNA gene in diet samples in Exp. 1

	Cornstalk:LH ratio					P-value
	20:80	40:60	60:40	80:20	SEM	Rinse	Linear	Quadratic
DNA Concentration (ng/mg)a						<0.01	0.11	0.07
Unrinsed	3.4	2.7	2.6	2.9	0.45
Hand-rinsed	5.1	3.4	3.7	3.5	0.45
Machine-rinsed	0.6	0.2	0.4	0.6	0.45
Ct Value1						0.16	0.81	0.18
Unrinsed	16.0	15.2	15.2	16.4	1.3
Hand-rinsed	15.3	13.0	13.5	14.7	1.3
Machine-rinsed	14.8	14.1	13.9	14.0	1.3
Relative 16S rRNA gene2,b						0.01	0.60	0.48
Unrinsed	53.0	131.1	78.6	77.7	41.9
Hand-rinsed	97.0	34.0	29.8	146.3	41.9
Machine-rinsed	3.6	1.6	-0.3	6.8	41.9

¹Cycles to threshold.

²Amount of 16S rRNA gene relative to amounts of 16S rRNA gene provided from chloroplasts in each meal.

^aHand-rinsed > unrinsed > machine-rinsed.

^bUnrinsed = hand-rinsed > machine-rinsed.

Endogenous contributions of cattle to ruminally evacuated diet samples do not contain fiber components (i.e., NDF, ADF, ADL, ADIA). Rinsing increased concentrations of NDF, ADF, and ADL in diet samples, but rinsing decreased concentrations of ADIA. Rinsing with aqueous solutions containing neutral soaps results in losses of silica from plant materials and decreases measures of ADF because silica is not dissolved in solutions that contain acidic soaps (Van Soest, 1994). Perhaps, greater amounts of rinsing prior to measures of ADIA resulted in more extensive losses of silica from diet samples and subsequent under-estimates of ADIA. Measures of ADIA include rinsing of diet samples with acidic soaps prior to measures of ash. It is likely that acidic soaps remove endogenous contributions from cattle to diet samples but do not dissolve silica. Therefore, rinsing prior to measures of ADIA could have contributed to decreased measures of ADIA in diet samples and differences in concentration of ADIA between ruminally evacuated diet samples and ADIA concentration in feed offered were greater with greater amounts of rinsing. Measures of NDF and ADF do not include rinsing with soaps prior to analysis. Greater removal of proteinaceous endogenous contributions from cattle to diet samples prior to analysis by greater amounts of rinsing could then allow for greater concentrations in measures of NDF and ADF that more nearly match concentrations of NDF and ADF in feed offered. Several authors have suggested that mucopolysaccharides in bovine saliva contribute to spurious measures of ADL (Lascano et al., 1970; Theurer, 1970; Olson, 1991). Greater removal of mucopolysaccharides from diet samples by greater amounts of rinsing should have decreased measures of ADL. Indeed, greater amounts of rinsing decreased measures of ADL but to a lesser extent compared to increases in NDF and ADF in response to greater amounts of rinsing. It is possible that endogenous proteinaceous contributions to diet samples from cattle were greater than contributions of mucopolysaccharides.

Analyses of samples with a known composition across a range of applicable concentrations allows for development of predictive equations that account for any variance associated with analytical procedures (i.e., variance estimates for the slope; Youden, 1951; Mandel and Linnig, 1957) and for any constant error in analysis (i.e., variance estimates for the ordinate; Youden, 1951; Mandel and Linnig, 1957) by linear regression. When linear models ideally predict concentrations of analytes then regression of known sample amounts and analyzed amounts results in unity, an ordinate that proceeds through the origin and a model that describes all variances in observations (Youden, 1951). Ratios of cornstalk:LH calculated from NDF, ADF, ADIA, ADL, CP, and NIRS analysis of unrinsed, hand-rinsed, and machine-rinsed diet samples were regressed with ratios of known composition (Table 4). Linear models developed from NDF, ADF, CP, and ADL were poor in accuracy (i.e., slope), precision (i.e., SE of the slope), and failed to account for much of the variation between estimates of diet composition and known diet composition. Concentration of NDF and ADF between cornstalk and LH residues were similar. Poor accuracy of the models developed from NDF and ADF were likely because differences in concentrations of NDF and ADF between cornstalk and LH residues were less than amounts needed for optimal sensitivity of analysis of diet composition. Differences in CP between cornstalk and LH residues were greater than differences in NDF and ADF between cornstalk and LH residues. However, estimates of diet composition from analysis of CP in diet samples were poor.

Table 4.

Evaluation of the diet composition prediction models in Exp. 1

						P-value
Item	Slope	Intercept	SEslope	SEintercept	r 2	Slope	Intercept
NDF
Unrinsed	-0.05	0.28	0.021	0.043	0.54	0.03	0.09
Hand-rinsed	0.10	0.83	0.192	0.371	0.37	0.64	0.27
Machine-rinsed	-0.41	4.58	0.348	0.666	0.42	0.26	<0.01
ADF
Unrinsed	2.62	0.88	1.313	4.419	0.71	0.07	0.85
Hand-rinsed	1.41	1.15	0.488	0.961	0.58	0.01	0.26
Machine-rinsed	2.61	1.07	0.341	0.810	0.86	<0.01	0.21
ADIA1
Unrinsed	0.98	0.05	0.129	0.332	0.86	<0.01	0.91
Hand-rinsed	0.61	0.27	0.078	0.183	0.86	<0.01	0.27
Machine-rinsed	28.86	-18.09	9.024	22.554	0.66	<0.01	0.45
ADL
Unrinsed	2.29	-1.11	1.163	2.602	0.84	0.07	0.68
Hand-rinsed	0.33	0.42	0.081	0.274	0.76	<0.01	0.23
Machine-rinsed	0.68	0.21	0.078	0.234	0.89	<0.01	0.40
CP
Unrinsed	-0.06	1.51	0.698	1.503	0.44	0.93	0.34
Hand-rinsed	0.16	1.75	0.565	1.481	0.56	0.79	0.26
Machine-rinsed	-0.02	1.61	0.032	0.069	0.41	0.46	<0.01
NIRS
Unrinsed	-1.69	5.78	1.940	4.134	0.46	0.40	0.22
Hand-rinsed	2.69	3.14	2.343	6.736	0.59	0.29	0.68
Machine-rinsed	0.95	0.19	0.192	0.442	0.74	<0.01	0.69

¹Acid detergent insoluble ash.

Unrinsed diet samples likely contained large amounts of endogenous protein components (e.g., salivary protein, sloughed epithelium, microbial cells). Greater amounts of rinsing did not increase the accuracy of the estimates from CP. It is possible that greater amounts of rinsing removed proteinaceous components of endogenous origin, but that soluble proteins in cornstalk and LH residues were also removed during rinsing. Additionally, estimates of diet composition from analysis of NDF, ADF, and CP only accounted for small amounts of variation between estimates of diet composition and known diet composition. Differences in ADL between cornstalk and LH were greater than NDF, ADF, and CP, but estimates of diet composition from analysis of ADL were poor. Several authors have reported that salivary mucopolysaccharides and other endogenous components can contribute to measures of artifact lignin (Lascano et al., 1970; Theurer, 1970; Olson, 1991). It is reasonable that greater amounts of rinsing would increase accuracy and precision of estimates of diet samples from analysis of ADL if endogenous components obviated accurate estimates of diet composition. Regression of estimates of diet composition from machine-rinsed diet samples and known diet composition resulted in a slope (0.68) nearer to unity and accounted for a greater amount of variation (r² = 0.89) in comparison to unrinsed (r² = 0.84) and hand-rinsed (r² = 0.76) diet samples. Nonetheless, estimates of diet composition from ADL in machine-rinsed diet samples resulted in lesser accuracy compared to estimates of diet composition from analysis of ADIA in unrinsed diet samples. Generally, measures of ADL have large inherent variation (Porter and Singleton, 1971; Sunvold and Cochran; 1991, Moore and Jung, 2001), and variation in analysis may have limited accuracy of estimates of diet composition from ADL even though differences in concentration of ADL between cornstalk and LH residues were large (CV = 46.5%). Estimates of diet composition from ADIA content in unrinsed diet samples were near to unity (i.e., 0.98) and the ordinate did not differ from the origin (P = 0.91). Regression of estimated diet composition and known diet composition resulted in a large coefficient of determination (r² = 0.86). Analysis of ADIA involves a washing step with an acidic detergent that may have removed any endogenous contaminants and allowed accurate estimates of meal composition from unrinsed ADIA. Furthermore, endogenous components (e.g., salivary protein, sloughed epithelium, microbial cells) do not contain ADIA. Greater amounts of rinsing decreased the accuracy of estimates of diet composition from ADIA. It is possible that greater amounts of rinsing contributed to greater loss of ADIA from cornstalk and LH residues. Overall, these data indicate that measures of chemical components in diet samples can provide accurate estimates of diet composition among cows fed binary mixtures of forage when endogenous contamination is minimal and when differences in concentrations of chemical components are sufficient to allow for adequate sensitivity of analyses.

Near infrared reflectance spectroscopy can allow for rapid, less costly measures of analytes in comparison to measures via wet chemistry and is nondestructive (Stuth et al., 2003). However, NIRS measures reflectance of radiation limited to the near infrared spectra (Shenk and Westerhaus, 1994). Incomplete measures of spectral reflectance fail to separate all unique chemical compounds because multiple molecular species can have identical measures of reflectance in a limited spectral range. Nonetheless, measures of reflectance of radiation across the near infrared spectrum can produce predictive models with high amounts of accuracy and precision when reflectance of near infrared radiation by unique sample matrices are accounted for by development of predictive models based on measures of near infrared reflectance within a unique matrix. Further, comparison of predictions from spectrochemical models (Shenk and Westerhaus, 1994) from known amounts of analyte across a range of concentrations allows for estimates of standard error of prediction to evaluate suitability of predictive models (Stuth et al., 2003). Regression of estimates of diet composition from analysis of machine-rinsed samples by NIRS resulted in a slope near to unity (slope = 0.95) and linear estimates described much of the variation (r² = 0.74) between estimates of diet composition and known diet composition. Estimates of diet composition by NIRS analysis of hand-rinsed (slope = 2.69; intercept = 3.14) and unrinsed (slope = −1.69; intercept = 5.78) diet samples were inaccurate. It is possible that contamination of diet samples by endogenous components reduced accuracy and precision of predictive models by NIRS because endogenous components were not represented in the sample matrix during development of predictive models (Stuth et al., 2003).

Experiment 2

Biomass of corn residues remaining after harvest was 7,230 kg/ha, and cornstalk and LH residue comprised 35% and 65% of corn residue (DM-basis), respectively. Average temperature during the grazing period was −2.0 ± 3.6°C, and wind speeds averaged 18.5 ± 8.0 km/h. It is likely that no biomass accumulated during the grazing period because of low temperatures and a short photoperiod.

Predictions of diet composition (Table 5) were limited to measures of ADIA in unrinsed diet samples and ADL and NIRS in machine-rinsed diet samples. Chemical components used for predictions of diet composition in this study were selected because ADIA and ADL were the only 2 chemical components with adequate differences in concentration between cornstalk and LH residues to allow optimal accuracy in predictions of diet composition (Fig. 2). Additionally, regression of predicted diet composition from ADIA in unrinsed diet samples and ADL or NIRS in machine-rinsed diet samples with known diet composition resulted in slopes nearest to unity in Exp. 1. There was no interaction (P = 0.16; Table 5) between grazing treatment and analyte (ADIA, ADL, or NIRS) used for prediction of diet composition. Nonetheless, estimates of cornstalk:LH ratios in diet samples collected from cows that strip-grazed corn residues were numerically greater than cows that continuously grazed corn residues when diet composition was predicted by ADIA and NIRS; however, predictions of diet cornstalk:LH ratios from measures of ADL were numerically less among strip-grazed cows than continuously grazed cows. Differences between measures of ADL in cornstalk and LH residues were 48% less in Exp. 2 than in Exp. 1(Exp. 1 CV = 46.5% vs. Exp. 2 CV = 22.2%). Further, differences between measures of ADL in cornstalk and LH residues in Exp. 2 were less than the amount of difference determined in Exp. 1 to allow optimal predictions of diet composition. Additionally, measures of ADL within cornstalk and LH residues were more variable in Exp. 2 compared to Exp. 1 (Table 1). Endogenous mucopolysaccharides can impact measures of ADL from diet samples collected by ruminal evacuation (Van Soest, 1994). Measures of concentration of ADL in machine-rinsed diet samples could have been impacted by small losses of soluble proteins from corn residues. Analyses of ADIA include rinsing in acidic soap prior to gravimetric measures of ADIA. Thus, losses of soluble proteins from corn residues likely had no impact on measures of ADIA in diet samples. Small losses of soluble proteins from corn residues could have impacted NIRS measures of machine-rinsed diet samples; however, it is likely that measures of NIRS in diet samples are a composite of several chemical components in corn residues and these composite measures could have diluted effects of small losses of innate soluble components in corn residues.

Table 5.

Estimates of diet selection in cattle grazing corn residues in Exp. 2

				P-value
Item	Continuous	Strip	SEM	Treatment	Chemical component	Treatment × chemical component
Ratio cornstalk:LH				0.02	0.71	0.16
ADIA1	0.21	0.56	0.21
ADL	0.39	0.37	0.21
NIRS	0.11	0.49	0.21

¹Acid detergent insoluble ash.

Overall, estimates of cornstalk:LH ratios were 2 times greater among strip-grazed cows compared to continuously grazed cows (P = 0.02). Several authors (Fernandez-Rivera and Klopfenstein, 1989; Gutierrez-Ornealas and Klopfenstein, 1991) have measured biomass before and after cattle grazed corn residue and concluded that cattle grazing corn residue often select diets with greater apparent digestibility in comparison to the average of all biomass available for grazing. Typically, strip-grazed cattle select diets with less apparent digestibility compared to cattle allowed to continuously graze (Blount et al., 1991). Estimates of total-tract digestibility of diets were not determined in this study, but estimates of cornstalk:LH ratio seem to suggest that strip-grazed cows selected diets with less apparent digestibility (i.e., greater cornstalk:LH ratio) compared to cattle that continuously grazed. Further, estimates of cornstalk:LH ratio indicate a greater intake of cornstalk residue in comparison to indirect measures of diet selection in cattle that continuously grazed corn residue (Fernandez-Rivera and Klopfenstein, 1989; Gutierrez-Ornealas and Klopfenstein, 1991). Indirect measures of diet composition are often unable to account for grazing of plant residues by wild herbivores, and plant residues that decompose during the grazing period (Van Soest, 1994). Additionally, indirect measures of diet selection among cattle grazing corn residues would likely over-estimate intake of LH residues removed by wind.

Overall, diet composition among cattle grazing cornstalk and LH residues can be accurately predicted from chemical components of diet samples collected by ruminal evacuation. Chemical components with greater differences in concentration between each constituent in a binary mixture of forage and that are not found endogenously in cattle are likely to allow improved estimates of diet composition. These data indicate that ADIA from unrinsed diet samples allowed for greater accuracy and precision in estimates of cornstalk:LH ratios among cattle fed corn residues. Additionally, NIRS analyses of diet samples that were machine-rinsed allowed accurate and precise estimates of diet composition. However, caution should be used when applying predictive models from NIRS to other experiments, because NIRS only measures a limited spectral range and likely reflects a composite of chemical components in diet samples. Therefore, predictions of diet composition from NIRS should be calculated from calibration equations developed from corn residues directly available to cattle for grazing. Nonetheless, direct measures of diet samples collected by ruminal evacuation may allow large opportunity for an increased understanding of cattle grazing binary mixtures of annual forage.

Conflict of interest statement. None declared.

This material is a contribution from the South Dakota Agricultural Experiment Station, Brookings, SD 57007