Conversion of High-Frequency Partial Body Weights to Total Body Weight in Feedlot Cattle

Recent technological advances have enabled development of automated data collection systems, e.g., for BW, feed intake, and other performance measures. This study evaluated the use of partial BW data for BW estimation in feedlot cattle. Partial BW (pBW) were collected using the GrowSafe Beef system, consisting of an platform that measures the force applied by the front two hooves while animals are drinking. The animals in this study visited the drinking trough on average 9.2 ± 6.4 times per day with an average visit duration of 69.4 ± 40.2 s. First, variation inherent to traditional chute weighing procedures were assessed by comparing 3610 consecutive-day chute BW for 1763 feedlot cattle ranging from 177 to 748 kg (mean 409 ± 98 kg). The mean difference between consecutive-day BW was 0.1 ± 8.8 kg, setting a lower limit to the precision of any estimation procedure. The next step was to compare single-day chute BW and pBW (averaged for the same day). Body weight and pBW data were collected from 7035 measurements of feedlot cattle (4130 from bulls, 969 from steers, and 1936 from heifers) of different origins and biological types (2316 British breed, 4387 Continental, 332 Wagyu cross). Three models were used to estimate BW from pBW: (1) slope only, (2) slope + intercept, (3) slope + intercept by sex, (4) slope + intercept by type, and (5) a full model including slope + intercept by sex and type, and all interactions. Results indicate that although inclusion of information in addition to pBW (i.e., sex, breed type, origin) were statistically significant (P < 0.001), improvements in precision were not of practical significance (less than 2.3 kg improvement, far lower than the precision of chute BW measurements). Moreover, additional model terms resulted in increased mean bias and slope bias in the error term. Therefore, BW may be estimated as 1.677 x pBW across sexes and breed types with adequate accuracy and precision.

Model	Mean bias	Slope bias	Random error	MSEPa	RMSEPb	R2
1	0.1041	2.717	351.0	353.9	18.81	96.4%
	0.029%	0.77%	99.20%
2	0.00369	0.00636	351.0	351.0	18.74	96.4%
	0.001%	0.002%	99.20%
3	9.755	4.189	320.6	334.5	18.29	96.7%
	2.76%	1.184%	90.60%
4	11.35	3.976	306.0	321.3	17.92	96.9%
	3.21%	1.124%	86.46%
5	11.37	3.978	257.9	273.3	16.53	97.4%
	3.21%	1.12%	72.90%

^aMSEP, mean square error of prediction; ^bRMSEP, root mean square error of prediction.