Application of Mathematical Models to Individually Allocate Feed of Group-fed Cattle

A data set of group-fed growing and finishing steers with individual feed access was used to evaluate predictions of required individual DM by 2 mathematical models (Cornell Value Discovery System, CVDS; and beef NRC) to allocate feed of group-fed, commingled cattle. Forty-eight crossbred steers (BW = 296 kg) were assigned to 1 of 6 pens and fed 1 of 4 growing diets formulated to have different energy concentrations for restricted or ad libitum intake regimen for 56 d. The diets were a low-starch diet fed ad libitum, a high-starch diet fed ad libitum, a high-starch with restricted intake, and an intermediate diet fed ad libitum with an average energy intake between ad libitum low-starch and ad libitum high starch diets. On d 57, all steers (BW = 401 kg) were placed on the ad libitum high-starch diet for finishing until d 140. The CVDS was able to account for 61% of the variation in the observed DMI (oDMI) of steers during the growing period, and for 71% of the variation in oDMI during finishing, with an average overprediction of 3.76%. In the same fashion, the NRC model was able to explain 59% of the variation in oDMI after adjustment for known perfonnance during the growing period with no bias (P > 0.10), and 57% of the variation in oDMI during the finishing period, with an average underprediction of 4.40%. Our overall evaluation suggested that the CVDS was more precise and accurate than the NRC model when predicting DMI for individual animals. Both models were sensitive to the previous level of nutrition of the cattle, suggesting that more variables are necessary to increase the prediction precision for cattle growing systems. The results from a risk analysis suggested that an amount of approximately$17.00/animal may be either over- or under-charged in the billing process of a commercial feedlot growing and finishing periods. Therefore, mathematical models could assist commercial feedlots to improve the accuracy of the billing process while maintaining the same income per pen.

Key words: cattle, growth, modeling, requirements, simulation, prediction

INTRODUCTION

Sorting systems have been developed to predict carcass composition of cattle to allow marketing of the feedlot animals at the optimum end point (Perry and Fox, 1997; Brethour, 2000). These systems strive to sort cattle into homogeneous groups for maximization of productivity, enhanced uniformity, and increased economic returns (Tedeschi et al., 2004). In the current beef marketing system, the reduction of nonconforming carcasses can improve the value of a group of cattle dramatically (Bruns and Pritchard, 2005).

Full utilization of these sorting systems in custom feedyards would require commingling of cattle owned by multiple customers, disrupting the billing process. Support systems that can predict individual feed requirements for an observed level of performance might be useful in assigning feed costs to animals of different ownership. Fox and Black (1977a,b,c) devised equations to predict performance and body composition of growing cattle. These equations have been modified to develop the Cornell Value Discovery System (CVDS; Tedeschi et al., 2004), and have been proposed as a support tool for feed allocation (Guiroy et al., 2001). The CVDS has been used to accurately allocate DMI among individual animals fed in pens (Tedeschi et al., 2006) and for genetic selection purposes (Williams et al., 2006).

The NRC (2000) included a computer model that uses information of cattle type, ration components, and environment to predict animal performance (Whetsell et al., 2006). The NRC (2000) model is well accepted and widely distributed and can be used to predict individual intake of cattle when performance level is known. However, the capacity of the NRC (2000) model for the purpose of feed allocation has not been extensively evaluated.

Because nutritional models rely on estimates of energy and nutrient requirements to calculate feed requirements, growing cattle programs that alter growth rate or body composition may influence the results of the applicability of models in practical conditions. To date, model applications have focused on feedlot production without regard to prior plane of nutrition.

The objectives of this study were (1) to evaluate the adequacy of CVDS and NRC (2000) models in predicting individual feed requirements of growing and finishing feedlot cattle; (2) to evaluate model application when growing diets are dissimilar; and (3) to determine the efficacy of model application to the billing process for commingled cattle fed in the same pen.

MATERIALS AND METHODS

Experimental Data

A data set including performance (ADG), DMI, and carcass data from steers (n = 48) fed in individual feeders (American Calan, Northwood, NH) was obtained from an experiment conducted at the Texas A&M University Agricultural Experiment Station in Bushland, TX (Vasconcelos, 2006). Care, handling, and management of steers were approved by the Cooperative Research, Education, and Extension Triangle Animal Care and Use Committee (Texas Agricultural Experiment Station, USDA-ARS, and West Texas A&M University). Briefly, steers (296.0 ± 16.7 kg of BW) were implanted with Synovex-S (20 mg of estradiol benzoate and 200 mg of progesterone; Fort Dodge Animal Health, Overland Park, KS) and individually fed 4 different growing diets for 56 d: a low-starch diet fed ad libitum (ALLS); a high-starch diet fed ad libitum (AL-HS); the same high-starch diet as AL-HS limit-fed to approximate the caloric intake of AL-LS (LF-HS); and a diet fed ad libitum with approximately the midpoint daily energy content between AL-LS and AL-HS (AL-IS). On d 57, all steers (400.6 ± 31.9 kg of BW) were placed on AL-HS diet for an 84 d finishing period.

by Vasconcelos, J T, Tedeschi, L O, Sawyer, J E, Greene, L W
Refer: http://findarticles.com/p/articles/mi_qa4035/is_200708/ai_n19511549