Precision feeding of feedlot cattle based on phenotypic production profiles that predict performance

Abstract

Feeding feeder calves to the average production with an average diet is the current norm in the South African feedlot industry and needs improvement. The feedlot industry is a low-margin high volume industry. Because of the low financial margins, financial risk and the mitigation thereof, are of utmost importance. Currently, most feedlots feed their calves to the average. The pen is made up of average feeder calves and they are fed with an average diet that has evolved to be the best match. To further advance in the mitigation of financial risk management, feeding diets to classified production groups that unlock their production potential is investigated. This is the problem statement that this research is investigating and presented in this thesis. First, a literature review was done to find that as far back as 1954, attempts have been made to evaluate feeder calves to predict the production in the feedlot. The research focused on different phenotypic traits, like hip height, which is an indication of maturity, length of body, and so forth. The principle was good, but the tools to measure, yardstick, and rulers, were relevant at that time but led to possible less accurate measurements and resulted in different conclusions. The beef cattle phenotype has changed over the years and measurements done years ago might not be relevant today. The exact point from where the measurement was done, was poorly and inconsistently defined. The measurements focused on specific areas of the body, like the forearm, height, length of the body, and head dimensions, to name a few. No data could be found that incorporated a holistic approach in evaluating the phenotype of feeder calves before the feeding period in the feedlot. New data were generated in our first research, where a novel production profile classification system was developed to categorise feeder calves into production groups to precision feed them to their potential. The phenotype of arriving feeder calves, that underwent a 60-day precondition period, were visually categorised into four production profiles (PP). The visual classification enabled a holistic approach and was done before the feeding period, with the prediction of production in mind. The average feeder calf was used as the reference and was called production profile 2. Feeder calves considered to have a better production potential as PP 2 were classified as PP 1, and the feeder calves considered to have the poorest production potential were classified as PP 3. The classification of the feeder calf before the feeding period was based on muscle and skeletal development. Animals considered to have superior, compared to average PP 2, skeletal development and thus the necessary framework for the attachment of the developing muscle in the feedlot environment were considered PP 1. The PP 3 feeder calf represented calves with poorer skeletal and muscle development. Feeder calves were classified into PP before the feedlot period and then fed commingled on different rations and under different management. The mean PP 1 feeder calf gained 29.66 kg more carcass, compared to the mean PP 3 feeder calf, thereby producing heavier carcasses and higher dressing percentages (Table 3.3). The data were analysed and concluded that feeder calves can be successfully categorised into the defined production profiles to predict production in the feedlot. The second and third experiments were designed for precision feeding. The value of the novel production profile classification system was evaluated through a precision feeding experiment, consisting of 2 approaches. In experiment 1 feeder calves that were first classified into PP, were subsequently randomly allocated into 3 different diets, High Producing Diet (HPD), Medium Producing Diet (MPD), and the Low Producing Diet (LPD). The MPD was considered as the control and represented the diet fed in the commercial feedlot where experiment 1 was conducted. Each diet was fed to a pen to which calves were block randomised based on PP classification and entry-weight. In this experiment, interactions between PP and diet were analysed. It was concluded that interactions between PP and the diet exist. The subdued reaction of PP 1 to the HPD called for reformulation and increasing the metabolizable energy (ME) and protein as not being the first limiting. This led to experiment 2, where a similar methodology was followed. This experiment was conducted in an experimental unit, in a different location. In experiment 2 the feeder calves were grouped by PP and fed in pens containing 1 – 4 animals, which allowed for intake and its monetary value to be determined, leading to additional information on the carcass feed conversion ratio (CFCR) and the calculation of carcass feed cost of gain (CFCOG). After analysing the data, it was found that the reformulation done in experiment 2, additional energy and protein not being the first limiting diet formulation, enabled the inherent production potential of PP 1 calves, which was not achieved in experiment 1. The design of experiment 2 had the additional benefit of providing the data to determine feed-to-gain ratio on a carcass basis, as well as the cost to gain a kg of beef. It was found that the PP 2 on the MPD (feedlot equivalent, or “control” diet) had the lowest CFCOG, confirming the concept that South African feeder calves are fed to the average: average calves fed the average diet established over time, provided the best economic outcome. Devolving from the average (PP 2 fed the MPD) resulted in a better carcass growth rate (CADG), carcass gain (kg), and CFCR tended (P=0.06) to be better when PP 1 was fed the HPD. However, the cost of the HPD was still too high, resulting in the PP 1 CFCOG being higher than the control (PP 2 fed the MPD). The results establish the possibility for nutritionists to formulate diets on a cost-effective way to feed feeder calves to their production potential. Artificial Intelligence (AI) with Convolutional Neural Networks (CNN) can replace the current subjective visual classification method. Feeder calves will be digitalised by a 3-D image, after which they will be classified into the different PP classifications, based on the methodology of the subjective method. The digitalisation of the feeder calf is the switch from subjective eye to objective computerised image. It is on the digitalised feeder calf where CNN will be applied as machine learning tool to establish an automated, objective, AI driven classification system for the production profile of incoming feeder calves. Once established, this technology will open the doors to the fine-tuning of precision feeding of feeder calves. The production profiling of feeder calves before the feedlot period, and the subsequent precision feeding by supplying nutrients to the production potential, in a cost-effective way, was successfully established. Further research is called for to find the most cost-effective diet for calves with different production profiles in each feedlot and country. This improved efficiency has the potential to benefit the entire beef value chain, and also the environment, by lowering the cost, as well as the environmental impact of beef production worldwide.