From measuring average body weight of the flock to precision feeding in broilers: A modelling approach to adjust daily feed composition

Precision feeding in broiler production could reduce feed cost and nitrogen and phosphorus excretion. However, in order to adjust daily feed composition, daily requirements of birds according to their actual growth performance must be evaluated. In commercial farms, body weight (BW) is now frequently measured with automatic weighing scales. We propose here to use the data provided by these devices to 1) forecast BW gain to come, 2) estimate associated metabolizable energy (ME), digestible lysine (dLys) and available phosphorus (avP) requirements, and 3) recompose a full diet using two types of pellets with different nutritional characteristics. Firstly, BW data collected during the previous days are used to fit a quadratic function (BW=f(age)). Using this function, BW(d) and BW(d+1) are predicted, and “most probable” BWG for the next day (BWG(d+1)) is then calculated as BW(d+1) – BW(d). This first model was validated by comparing BW and BWG predictions (Y) to measured data (X, n = 814). BW is very well predicted (Y = 1.04 X – 0.03, R² > 99%), while BWG prediction quality is lower but still acceptable (Y = 0.76 X + 0.02, R² = 56%). Secondly, lipid and protein depositions according to actual BW(d) and predicted BWG(d+1) are estimated with allometric relationships developed using literature data. Requirements for ME and dLys are then estimated from these daily depositions. For avP, we considered an equation from the literature, directly using BW(d) and BWG(d+1) values. The requirements sub-model was validated by comparing predictions (Y) to measurements of ME, dLys and avP daily intake (X, n = 325). ME predictions are very good (Y = 1.04 X, R² = 96%) while the predictive quality for dLys and avP is slightly lower but still acceptable (Y = 0.93 X + 139, R² = 75% and Y = 0.99 X + 31, R² = 71% respectively). Thirdly, 1001 blend of A/B pellets are computed (increase of %A in the blend from 0 to 100% with a step of 0.1%), with A being poorer in ME and richer in dLys and avP than B, respectively. For each blend, assuming that feed intake is regulated on an energy basis, daily feed intake is estimated using ME requirement and blend content. dLys intake is then calculated and compared to dLys requirement. The best blend is the first one where dLys intake is above daily requirement. In practice, this blend could be easily prepared and distributed with a commercial weighing/mixing hopper.