| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ANIMAL GROWTH, PHYSIOLOGY, AND REPRODUCTION |
Department of Animal Science, Iowa State University, Ames 50011
Abstract
Steers from research crossbreeding projects (n = 406) were serially scanned using real-time ultrasound at 35-d intervals from reimplant time until slaughter. Cattle were evaluated for rump fat depth, longissimus muscle area (ULMA), 12th-rib fat thickness (UFAT), and percentage of intramuscular fat (IMF) to determine the ability of ultrasound to predict carcass composition at extended periods before slaughter. Additional background information on the cattle, such as live weight, ADG, breed of sire, breed of dam, implant, and frame score was also used. Carcass data were collected by trained personnel at "chain speed," and samples of the 12th-rib LM were taken for ether extract analysis. Simple correlation coefficients showed positive relationships (P < 0.01) between ultrasound measures taken less than 7 d before slaughter and carcass measures: ULMA and carcass LM area (CLMA, r = 0.66); UFAT and carcass 12th-rib fat thickness (CFAT, r = 0.74); and IMF and carcass numeric marbling score (r = 0.61). The same correlation coefficients for ultrasound measures taken 96 to 105 d before slaughter and carcass values (P < 0.01) were 0.52, 0.58, and 0.63, respectively. Steers were divided into source-verified and nonsource-verified groups based on the level of background information for each individual. Regression equations were developed for the carcass measurements; 46% of the variation could be explained for CLMA and 44% of CFAT at reimplant time, 46% of the variation in quality grade and 42% of the variation in yield grade could be explained. Significant predictors of quality grade were IMF (P < 0.001), natural log of 12th-rib fat thickness (LUFAT, P < 0.001), and ADG (P < 0.01), whereas LUFAT (P < 0.001), ULMA (P < 0.01), live weight (P < 0.001), hip height (P < 0.001), and frame score (P < 0.001) were significant predictors of yield grade. Regressions using ultrasound data taken 61 to 69 d before slaughter showed increasing R2. Live ultrasound measures at reimplant time are a viable tool for making decisions regarding future carcass composition.
Key Words: Carcass Quality • Carcass Yield • Steers • Ultrasound
Introduction
Ultrasound has been used extensively in beef cattle enterprises over the past decade, and its use is continuing to gain in popularity. However, ultrasound technology is not used extensively in the feedlot. Feedlot managers are reluctant to subject animals to added stress, and scanning takes considerably more time to handle the cattle than the typical 90-animal/hr rate in most commercial feedlots. Nevertheless, ultrasound in the feedlot is not a new application.
Brethour (1994
, 2000a,b) assessed marbling and backfat deposition in beef cattle at several stages of growth. It was established that carcass composition could be accurately predicted preslaughter, but its usefulness for animals that were already USDA yield grade 4 was questionable. Predicting carcass composition from animals entering the feedlot as calves proved inaccurate due to differences in preweaning environment (Brethour, 2000b). Rouse et al. (1998; 2000) used similar technology to predict carcass composition at various preslaughter points, finding that a chute-side application just before slaughter was feasible, but predicting carcass composition for extended periods before slaughter would require a more appropriate model.
The first objective of this study was to develop ultrasound-derived prediction equations for quality and yield grade at extended periods before slaughter. A second objective was to investigate whether 100-d, 65-d, or another interval before slaughter was the best time to scan feedlot animals. The final objective was to validate the prediction models in a commercial setting, testing their accuracy.
Materials and Methods
Data for this study were obtained from 406 crossbred feedlot steers in 2000, 2001, and 2002. Angus, Simmental, Red Angus, and Charolais breeds were represented in both sires and dams of these cattle. Crossbred females were mated to purebred bulls from the previously mentioned breeds. All steers were fed as calves and slaughtered at approximately 12 to 15 mo of age based on marketability to a grid paying premiums for quality grade. Ultrasound information was used to make marketing decisions. Steers that experienced continued illness, chronic bloating, or ADG < 0.25 kg/d were removed from the dataset. Steers with missing ultrasound data were deleted from the dataset.
The steers were serially scanned by a Centralized Ultrasound Processing (Iowa State University, Ames) certified field technician in 2000 and 2001 and an Annual Proficiency Testing and Certification (Iowa State University) qualified technician in 2002. All images were interpreted by a certified laboratory technician at the Iowa State University Image Lab. These images were collected using the Classic Scanner 200 (Classic Medical Co., Tequesta, FL) equipped with a 3.5-MHz, 18-cm linear array transducer.
Live animal measurements recorded during the ultrasound period were: 1) live weight (held off feed overnight until after the scan session had taken place); 2) a cross-sectional image was taken between the 12th and 13th ribs to obtain s.c. fat thickness measured at three-fourths the distance from the medial end of the longissimus dorsi muscle (UFAT); 3) longissimus dorsi muscle area (ULMA); 4) a longitudinal image taken between the hooks and pins perpendicular to the shaft of the ileum to measure subcutaneous fat depth over the termination point of the biceps femoris in the rump (reference point) (URFAT); and 5) four independent images collected laterally across the 12th and 13th ribs to estimate the percentage of intramuscular fat (IMF) within the longissimus dorsi. Four independent images are necessary to follow Annual Proficiency Testing and Certification standard format for data submission. Analysis of IMF was based on the average of the four images collected.
In 2000, steers were scanned once in February and once in May (1 wk before slaughter), and all cattle were slaughtered within 7 d. The cattle from the next 2 yr were serially scanned every 30 to 35 d from February to May in 2001 or from January to May in 2002. Steers slaughtered in 2001 and 2002 were marketed at two scheduled dates. Steers that scanned Choice (over 4.0% IMF) or that possessed excessive s.c. fat were slaughtered on the first date. The remaining steers were fed for an additional 30 to 35 d before slaughter.
Data were analyzed separately within each year. Carcass traits most likely did not differ across years due to similarities in genetic base and management. All the steers came from three calf crops of the same breeding project. Additionally, the steers were all fed as calves in a confinement facility; fed a typical high-energy, corn-based diet; implanted once on feedlot entry and again 90 d later; and marketed on a grid emphasizing quality grade. Selection for a grid environment may have helped to make the averages more uniform.
Data were grouped together based on the number of days from the scan date to the corresponding slaughter date of each individual. Cattle from all 3 yr were classified into a period of days from slaughter. The first group was steers scanned 96 to 105 d before slaughter (n = 228). The second group was steers ultrasonically measured 61 to 69 d before slaughter (n = 254).
Cattle were transported to a commercial slaughtering facility, where trained individuals collected routine carcass data. Carcass measurements collected approximately 24 h postmortem were as follows: 1) hot carcass weight (HCW); 2) s.c. fat thickness collected at the three-quarter position between the 12th and 13th ribs (CFAT); 3) area of the longissimus dorsi between the 12th and 13th ribs (CLMA); 4) percentage of kidney, pelvic, and heart fat (KPH); and 5) numeric marbling scores were assigned by a USDA grader. Fat thickness was only adjusted when an obvious deviation in fat thickness was noticeable at the measurement location due to excess fat removal from the hydraulic hide puller. In these situations, the opposite side of the carcass was measured, or the fat depth was adjusted based on a visual estimate of overall fat distribution of the carcass. The USDA yield grade was then calculated using the previously mentioned carcass measurements (USDA, 1997).
Additional cattle background information was also included as possible sources of variation in statistical analysis. Breed of sire, breed of dam, percentage of Black Angus in the individual’s pedigree (PCTANGUS), age of the animal from birth in days, hip height, then converted to a frame score using the Beef Improvement Federation Guidelines equation (BIF, 2002), brand of implant administered, and ADG were calculated for each steer from feedlot entry to harvest. All data were analyzed using CORR, STEPWISE, GLM, and MEANS procedures from version 8.1 of the SAS (SAS Inst., Inc., Cary, NC). Simple correlation coefficients were calculated between carcass measurements and the corresponding ultrasound measure. All regression equations allowed live weight, UFAT, URFAT, ULMA, IMF, and all the abovementioned background information as potential sources of variation for the final prediction model.
Regression analysis with stepwise procedures was used to develop prediction equations for CLMA, CFAT, yield grade, and carcass numeric marbling score from live animal measures. Significance levels for variables to enter a model and to stay in a model were set at P < 0.10. A separate model for each of the scan periods was developed, starting with a model simulating "source-verified" (SV) cattle, where age and PCTANGUS was known. Other variables eligible for inclusion into stepwise regression equations included live weight, ULMA, UFAT, LUFAT, URFAT, IMF, ADG, hip height, frame score, breed of sire, and breed of dam.
Brethour (2000b) reported exponential models best fit serial scans of backfat thickness, giving a final R2 value (0.89). In this study, model R2 values for predicting carcass backfat 100 d before slaughter that included the natural log of backfat thickness were also higher. Thus, the natural log of UFAT was used as a possible source of variation in all of the regression analysis with stepwise procedures. Hip height was not measured on all steers. Therefore, frame score information was also not available on all cattle. As a result, a separate model was developed without using those variables. A model was developed that reflects cattle purchased at an auction, also known as "nonsource-verified" cattle (NSV). These regressions did not use age, hip height, frame score, or PCTANGUS as possible variables. When none of the abovementioned variables were used by the regression employing stepwise procedures, only one model was printed in the table. Environmental factors were intentionally excluded in the regressions to obtain a more robust model.
Results and Discussion
Simple statistics for the live animal measurements and the carcass data collected across all 3 yr are listed in Table 1. Simple correlation coefficients were calculated for ultrasound vs. carcass traits at the 96- to 105-d period, the 61- to 69-d period, and for a preslaughter measurement taken within 1 wk of slaughter. One hundred days before slaughter, UFAT showed moderate to high positive relationship with CFAT and yield grade (r = 0.58 and r = 0.51, respectively), and a stronger relationship within 1 wk of slaughter (r = 0.74 and r = 0.60, respectively). Crews et al. (2002) published residual correlations from a yearling scan of steers for UFAT vs. CFAT of 0.78 and a preslaughter UFAT vs. CFAT correlation of 0.86.
|
, from a growth and development standpoint, rump fat is deposited much earlier than rib fat, but then levels off in the latter stages of the feeding period. Based on the results of this study, it seems that cattle with more rump fat upon feedlot entry often end up with more s.c. rib fat at slaughter. The predictive value of rump fat at extended periods before slaughter needs to be further explored. This result is consistent with Tait (2002), who reported a similar pre-slaughter correlation between URFAT and CFAT (r = 0.54). Realini et al. (2001) found URFAT to be a significant predictor (P < 0.05) of percentage and amount of fat in a carcass.
The correlation between IMF and marbling score remained somewhat constant from 100 d before slaughter to just 1 wk preslaughter (r = 0.63 to 0.61). The preslaughter correlation is similar to that published by Tait (2002; r = 0.63). Results from the 61- to 69-d scanning period were consistent with all other scan sessions (r = 0.62). The deposition of marbling appears to be linear over days of age, suggesting that an IMF measurement taken 100 d before slaughter may be as helpful as one taken much closer to the slaughter date. Brethour (2000a) also used a linear model to predict marbling scores of beef calves. High-quality ultrasound images are easier to obtain on leaner, younger cattle. As a result, it may prove helpful to scan steers at reimplantation time, while management changes in feeding and implanting can still be made.
When hip height is regressed with age to predict a frame score, the correlation of frame score to other measures seems to stay fairly constant, regardless of the scan session. In this dataset, larger framed, Continental-based steers were consistently heavier muscled, leaner, and later maturing, with lower marbling scores. In Table 1
, the relationships between frame score and CLMA, CFAT, and marbling score illustrate these trends (r = 0.29, –0.26, and –0.14, respectively).
Tables 2 and 3 are the stepwise regression to predict CLMA at 96 to 105 d and 61 to 69 d before slaughter, respectively. The first variable to enter all SV models was ULMA, accounting for 27.7 (Table 2) and 30.0% (Table 3) of the variation. Hip height entered next in all of the regressions, accounting for an additional 9.0 and 7.4% of the variation, respectively. Frame score was also included in the SV models, although not in the same order each time, with a partial regression coefficient of 0.060 (Table 2) and 0.025 (Table 3). Marketing decisions may possibly explain other variables inclusion in the models. For instance, IMF entered several regressions with a negative coefficient. Therefore, animals that scanned with more IMF often had smaller CLMA. However, animals that scanned into the Choice grade (IMF > 4.0%) were sent to market earlier. Thus, these steers were usually lighter-weight animals with logically smaller CLMA. In the NSV regressions, ULMA was still important, accounting for 27.5 (Table 2) and 31.9% (Table 3) of the variation. When frame score or hip height was not available, live weight surfaced in all three models, yielding an additional 3.8% and 1.9% of the variation, respectively. Hamlin et al. (1995) also found live weight to be a good predictor of muscling when performing regression of ULMA on live weight in four different biological types of cattle (R2 from 0.65 to 0.78).
|
|
; Abraham et al., 1980). At extended periods before slaughter, the accuracy of projecting CLMA is somewhat low. The root mean squared error (RMSE) for the SV 96- to 105-d prediction model was 6.00 cm2. At 65 d before slaughter, the RMSE drops to 5.52 cm2. Crews et al. (2002) scanned steers at 1 yr of age and achieved a RMSE of 4.78 cm2 and a model R2 value of 0.88 when predicting CLMA. Because USDA yield grade is not heavily influenced by LM area, taking extra time chute-side to trace for ULMA may not significantly help the final prediction of USDA yield grade.
Tables 4 and 5 list the stepwise regression to predict CFAT at 96 to 105 d and 61 to 69 d before slaughter, respectively. The first variable to enter the SV model at 100 d preslaughter was LUFAT (R2 = 0.347). However, UFAT was most important in the 65-d projection (R2 = 0.508). As the number of days before slaughter decreases, the change in fat depth lessens, making a log transformation unnecessary. Crews et al. (2002) used a yearling ultrasound scan of fat thickness on steers, along with year of birth, gender, and age as covariates to get a model R2 = 0.80. Models excluding LUFAT were also computed to be sure log transformation was necessary. Model R2 values were all lower compared with those in Table 4 (R2 = 0.431, 0.419, and 0.405, respectively). A significant predictor of CFAT in the SV models was URFAT, especially at extended periods before slaughter, accounting for an additional 4.8% of the variation at 96 to 105 d before slaughter (Table 4) and 2.8% (Table 5) of the variation at 61 to 69 d before slaughter. Hip height was also significant in SV cattle, accounting for 1.1% additional variation in the 100-d projection and 2.8% in the 65-d projection.
|
|
. Age and PCTANGUS accounted for a small amount of variation in the 100 d projection of rib fat thickness so the model was included in Table 4. In comparison, Hamlin et al. (1995) found age as a source of variation (P < 0.05) for ULMA and UFAT in five measurements in a serial scanning study. In both NSV models, IMF accounted for 1.9 (Table 4) and 0.8% (Table 5) of the additional variation. The genetic relationship reported by the American Angus Association (2002) between s.c. fat and marbling score was only 0.04, with the phenotypic relationship only slightly higher at 0.16.
Accuracy of predicting CFAT is very important with its influence on the USDA yield grade equation. At 96 to 105 d before slaughter, the RMSE of the SV model is 0.20 cm, dropping to 0.17 cm 61 to 69 d preslaughter. Crews et al. (2002) reported a RMSE of 0.14 cm when using a yearling scan on steers to predict CFAT. Brethour (2000b) found that accuracy improved if cattle were given time on feed when projecting the number of days to reach 10 mm of backfat. Evaluation 90 d before slaughter to predict days to 10 mm backfat produced a model accounting for 65% of the variation, with an average error of 33 d. At 43 d before slaughter, the R2 value rose slightly to 0.70, but the average error was 25 d.
When predicting USDA yield grade, LUFAT was the first variable entered in each SV model accounting for 26.3 and 36.3% of the variation 100 and 65 d before slaughter, respectively. Just 1 wk before slaughter, Tait (2002) found an ultrasound measurement of backfat to account for 30% of the variation in percent retail product from the four primal cuts. In this study, the addition of ULMA, live weight, hip height, and frame score added an additional 15.3% of the variation in the 100-d projection.
A measure of carcass muscling, ULMA, entered NSV models, accounting for 4.5% (Table 6) and 2.5% (Table 7) additional variation as cattle progressed towards slaughter. Live weight also entered both projections, reporting a partial regression coefficient of 0.052 in the 65-d projection of NSV cattle. Comparatively, Tait (2002) reported a prediction of percent retail product from the four primals, which included UFAT, ULMA, live weight, and IMF as sources of variation (R2 = 0.49). Also within 1 wk of slaughter, Greiner et al. (2003) included UFAT, URFAT, ULMA, live weight, and ultrasound of body wall thickness to achieve R2 values of 0.61 and 0.67.
|
|
Stepwise regressions to predict carcass numeric marbling score at 96 to 105 and 61 to 69 d before slaughter are listed in Tables 8 and 9. Because only a few variables entered the projection of marbling score at any extended period before slaughter, the results for the SV and NSV groups were identical, and only one result is reported. Percentage of IMF was most important, accounting for 39.3 and 42.7% of the variation in the 100 and 65 d projections, respectively. Included next was LUFAT, accounting for an additional 4.9% and 3.9% of the variation, respectively. With few significant sources of variation, accurately projecting marbling scores in a commercial feedlot facility seems possible. A similar model by Brethour (2000a) projects marbling score using age, PCTANGUS, and the initial ultrasound marbling estimate.
|
|
reported a relative accuracy of 76 to 78% in predicting carcass marbling within one-third of the USDA quality grade. When a power function model was used, cattle that entered with low traces of marbling (Standard00) usually failed to become Choice within feeding period of up to 200 d (Brethour, 2000b). A linear model, like the one used in this study, may tend to regress animals closer to the data set mean. Extensive field testing needs to be performed to assess the error and bias of a commercial model.
To validate the prediction models in this study for accuracy, a group of 63 genetically similar British x Continental crossbred steers was scanned at the Iowa State University Allee Research Farm and slaughtered 95 d later. The correlation between carcass 12th-rib fat depth and predicted CFAT was r = 0.81, and standard error of prediction (SEP) was 0.166 cm. The correlation between carcass LM area and predicted CLMA was r = 0.41, and SEP was 7.807 cm2. Using preslaughter ultrasound measurements, Crews et al. (2002) calculated a SEP of 0.140 cm and 4.49 cm2 when predicting CFAT and CLMA, respectively. The USDA yield grade was calculated from carcass data and compared with the prediction model for yield grade. The raw mean for yield grade across all 63 steers from the carcass data was 2.66; the prediction model for yield grade produced an average value of 2.60. The correlation between yield grade calculated from carcass data and predicted USDA yield grade was r = 0.83, and SEP was 0.339. The average USDA marbling score was 5.70 (Small 70); the mean value from the prediction model for carcass numeric marbling score was 5.77 (Small 77). The correlation between carcass numeric marbling score and predicted marbling score was r = 0.75, and the SEP was 0.663.
The prediction models were also validated on an individual basis. Seven of the 63 steers were stamped USDA Select (4.00 to 4.99). Using the prediction model for carcass numeric marbling score, six of those seven steers projected marbling scores below 5.10, or on the border between Low Choice and High Select (5.0). The USDA yield grade prediction model was also tested on an individual level. Three steers calculated yield grades greater than 3.90 from the carcass measures collected. Those same three individuals were projected to have USDA yield grades greater than 3.90 when the prediction model for yield grade was used. If the average steer in this dataset were scanned at reimplant time and fed for an additional 100 d, a producer could expect a 0.35 cm increase in s.c. fat, a 17.2 cm2 increase in LM area, and a 1.06% increase in percentage of IMF, with daily gains of 1.54 kg.
Implications
Value-based markets pay premiums to producers who can deliver a specific, consistent product, but discount those who cannot. Ultrasound in the feedlot may offer a unique opportunity; it can help the producer recognize which grid environment offers the most profit potential. More importantly, scanning feedlot cattle at an extended period before slaughter still allows sufficient time to make crucial management decisions and decrease carcass discounts. Developing prediction models that are sensitive to differences in genetics and management will be more difficult. Individual feedlots could refine a robust model to best fit their operation, or help them decide how to manage and market a pen of cattle.
1 Correspondence and current address: 714 S. Sycamore, Iola, KS 66749 (phone: 515-294-5275; e-mail: wall{at}allencc.edu).
Received for publication July 8, 2003. Accepted for publication February 20, 2004.
Literature Cited
Abraham, H. C., C. E. Murphy, H. R. Gross, G. C. Smith, and W. J. Franks, Jr. 1980. Factors affecting beef carcass cutability: An evaluation of the USDA yield grades for beef. J. Anim. Sci. 50:841–851.
American Angus Association. 2002. Spring 2002 Sire Evaluation Report. American Angus Association. St. Joseph, MO.
BIF. 2002. Guidelines for Uniform Beef Improvement Programs, Beef Improvement Federation, Athens, GA.
Brethour, J. R. 1990. Relationship of the ultrasound speckle to marbling score in cattle. J. Anim. Sci. 68:2603–2613.[Abstract]
Brethour, J. R. 1992. The repeatability and accuracy of ultrasound in measuring backfat in cattle. J. Anim. Sci. 70:1039–1044.[Abstract]
Brethour, J. R. 1994. Estimating marbling score in live cattle from ultrasound images using pattern recognition and neural network procedures. J. Anim. Sci. 72:1425–1432.[Abstract]
Brethour, J. R. 2000a. Using receiver operating characteristic analysis to evaluate the accuracy in predicting future quality grade from ultrasound marbling estimates on beef calves. J. Anim. Sci. 78:2263–2268.
Brethour, J. R. 2000b. Using serial ultrasound measures to generate models of marbling and backfat thickness changes in feedlot cattle. J. Anim. Sci. 78:2055–2061.
Crews, D. H., Jr., N. H. Shannon, R. E. Crews, and R. A. Kemp. 2002. Weaning, yearling, and preharvest ultrasound measures of fat and muscle area in steers, bulls, and heifers. J. Anim. Sci. 80:2817–2824.
Crouse, J. D., M. E. Dikeman, R. M. Koch, and C. E. Murphy. 1975. Evaluation of traits in the USDA yield grade equation for prediction beef carcass cutability in breed groups differing in growth and fattening characteristics. J. Anim. Sci. 41:548–553.
Crouse, J. D., and M. E. Dikeman. 1976. Determinates of retail product of carcass beef. J. Anim. Sci. 42:584–591.
Greiner, S. P. 1997. The use of real-time ultrasound and live animal measurements to predict carcass composition in beef cattle. Ph.D. Thesis, Iowa State Univ., Ames.
Greiner, S. P., G. H. Rouse, D. E. Wilson, L. V. Cundiff, and T. L. Wheeler. 2003a. Accuracy of predicting weight and percentage of beef carcass retail product using ultrasound and live animal measures. J. Anim. Sci. 81:466–473.
Greiner, S. P., G. H. Rouse, D. E. Wilson, L. V. Cundiff, and T. L. Wheeler. 2003b. The relationship between ultrasound measurements and carcass fat thickness and longissimus muscle area in beef cattle. J. Anim. Sci. 81:676–682.
Hamlin, K. E., R. D. Green, T. L. Perkins, L. V. Cundiff, and M. F. Miller. 1995. Real-time ultrasonic measurement of fat thickness and longissimus muscle area: I. Description of age and weight effects. J. Anim. Sci. 73:1713–1724.[Abstract]
Herring, W. O., D. C. Miller, J. K. Bertrand, and L. L. Benyshek. 1994. Evaluation of machine, technician, and interpreter effects on ultrasonic measures of backfat and longissimus muscle area in beef cattle. J. Anim. Sci. 72:2216–2226.[Abstract]
Realini, C. E., R. E. Williams, T. D. Pringle, and J. K. Bertrand. 2001. Gluteus medius and rump fat depths as additional live animal ultrasound measurements for predicting retail product and trimmable fat in beef carcasses. J. Anim. Sci. 79:1378–1385.
Rouse, G., D. Wilson, D. Duello, and B. Reiling. 1992. The accuracy of real-time ultrasound scans taken serially on small-, medium–, and large-framed steers and bulls slaughtered at three endpoints. 1992 Iowa State Univ. Beef and Sheep Res. Rep. A.S. Leaflet R896, Ames.
Rouse, G., V. Amin, B. Punt, S. Greiner, and D. Wilson. 1998. Chute-side application of real-time ultrasound for feedlot cattle marketing—A pilot project. Iowa State Univ. Anim. Sci. Leaflet R1533, Ames.
Rouse, G., S. Greiner, D. Wilson, C. Hays, J. R. Tait, and A. Hassen. 2000. The use of real-time ultrasound to predict live feedlot cattle carcass value. Iowa State Univ. Anim. Sci. Leaflet R1731, Ames.
Tait, R. G. 2002. Prediction of retail product and trimmable fat in beef cattle using ultrasound or carcass data. M.S. Thesis, Iowa State Univ., Ames.
USDA. 1997. United States standard for grades of carcass beef. Agricultural Marketing Service, USDA, Washington, DC. Available: http://www.ams.usda.gov/lsg/stand/standard/beef-car.pdf. Accessed May 15, 2002.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
