Taking NIR beyond feedstuff analysis to enhance turkey production profitability

Hadden Graham (AB Vista Feed Ingredients, Woodstock Court, Marlborough, Wilts SN8 4AN, UK) and Chris Piotrowski (Aunir, The Dovecote, Pury Hill Business Park, Towcester NN12 7LS, UK)

93

With feed costs accounting for 50 to 80% of total variable production costs, nutrition continues to be an area of major focus. The key target for nutritionists is to provide the animal with the correct amount of nutrients to support optimal performance. Both an excess and a lack of nutrients are likely to result in economic losses, through higher costs and/or lower animal performance. Thus, it is important for the nutritionist and raw material purchaser to have correct information on the composition and nutritional value of available ingredients they are using, and not just the industry-given averages or book values. Accurate and regular analysis of feedstuffs and complete feed, to confirm diets are correctly formulated, is a key quality control measure.

To ensure consistency in diets, nutritionists traditionally used proximate analysis from approved laboratories, where ingredients and feeds are analysed for their nutritional contents. Unfortunately, the majority of these analyses are time-consuming and expensive, which not only creates a delay between sampling and receiving results of the analyses, but also restricts the number of samples that can be analysed. Alternatively, a Near Infra- red (1100-2500 nm wavelength) Reflectance spectrometer (NIR) can be used to predict composition. This technology is cost effective and fast meaning that it allows nutritionists to get almost immediate feedback on in- coming ingredients and out-going feed. This in turn enables the analysis of many more samples at a much reduced cost. NIR has many more potential uses in animal production beyond simple proximate analysis.

This paper will discuss the use of NIR in feedstuff analysis and diet formulation, and the opportunities to extend this technology beyond standard analysis to support greater efficiencies in turkey production.

Predicting feed composition
It is common practice for nutritionists to formulate diets with average compositional data for ingredients, taking either a book value or actual analytical data, and often a safety margin based on the expected variability in the data. Safety margin can vary, depending on the formulator and the feedstuff, usually varying between 0 (average data used) and 1 standard deviation from the average. Adjusting the nutritional value of the ingredient based on the standard deviation ensures that the majority of feeds produced will provide the expected level or higher of any nutrient and also meet the requirements of statutory declaration of composition.

NIR can predict chemical and physical properties by relating vibrational spectra obtained on a set of known samples to reference analytical methods performed on the same sample set. The resulting calibration can be used to predict the composition of unknown samples of the same type of materials. NIR offers important advantages over traditional methods, in that it is rapid, non-destructive, requires no chemicals and hence produces no waste. It is easy to operate, once calibrated, and requires minimal sample preparation.

NIR has been used in the feed industry for over 30 years, and is now approved by AOAC to determine moisture, nitrogen (crude protein) and acid detergent fibre (ADF) in feed and forages. However, there is some scepticism across the industry regarding the accuracy of NIR to predict feed composition relative to wet chemistry. There are two main reasons for this; one is the use of poor or inappropriate NIR calibrations, and the other is poor sampling technique. Due to its nature, NIR can only predict the composition of samples similar to those used to develop the calibration, and the variation can never be less than that of the methods used to provide the data set that the calibration is built on. It is common to assume that a wet chemistry result is always better than an NIR result; however, Undersander (2006) reported that when crude protein results differ, a re-run of the wet chemistry agreed with the NIR 80% of the time. This demonstrates that, as might be expected, there is less risk of making a mistake when taking a NIR spectrum than when running a laboratory analysis.

The real advantage of NIR is that it is cheaper and quicker to analyse a number of samples for a range of analyses than to run one wet chemistry analysis, giving the formulator a much more complete real-time picture of the overall composition as well as variation within feed ingredients.

Predicting nutritive value
High phosphate prices, increasing environmental pressures and more effective enzyme products have encouraged feed manufacturers to increasingly replace inorganic phosphates with phytases. However, the extent of phosphorus release by phytases depends to a large extent on the phytate content of the diet. As phytate levels can vary between and within feedstuffs, it is difficult to accurately predict the phytate content of a final feed. While several laboratory methods are available to determine phytate levels in feeds, these are all relatively expensive and time consuming. Recently, NIR calibrations based on an enzymatic laboratory method were developed to give the real-time prediction of the phytate content of feedstuffs and diets, allowing feed manufacturers to maximise phytase inclusion and thus feed cost savings (Santos and Bedford 2012).

Feedstuffs are usually purchased on the basis of parameters such as test weight and crude protein content, both unrelated to a greater or lesser degree to their value in feed. Consequently, Rao (2012) indicated that approximately half of incidences of poor performance in a US commercial broiler company were related to the use of incorrect feedstuff nutritive values. The traditional method of predicting the energy value of feedstuffs or feeds is to use any of a number of published equations to calculate the productive energy from the analysed nutrient content. These equations are usually developed from trials where a diet of known composition was fed to the target animals and the productive value, such as net or digestible energy, determined. The weaknesses of this approach are well known. For example, the assay methods used to develop the prediction equations may be different from those used to analyse the feedstuffs in question, and the feedstuffs or diets used in the animal trials may not represent those used commercially. Further, animal trials are prohibitively expensive and time-consuming. The production advantages of accurate feed formulations, based on NIR analyses rather than book values, in promoting extra broiler performance was demonstrated by Soto et al. (2013).

Starting in 1996, a major research program was initiated in Australia to develop NIR calibrations to predict the nutritive value of commonly used feedstuffs across several animal species including ruminants, pigs and poultry. Close to 4000 cereal grain and protein feedstuffs were surveyed, and over 350 of these were fed to animals to determine available energy and intake index as well as other parameters (Black and Spragg, 2010; Black et al., 2014).

Analysing all incoming feed raw materials by wet chemistry would be both time consuming and expensive, and the delay in receiving results would make this practically ineffective. However, this Australian project has used animal data to develop NIR calibrations which predict energy content and intake index (from 0-100) as well as composition, allowing all incoming raw materials to be quickly analysed and segregated on arrival at the mill. The value of using NIR to determine the composition of in-coming feedstuffs has recently been demonstrated by an integrated UK company. By simply segregating in- coming wheat and soybean meal into either high- and low protein bins for each, this company was able to save over £2 per ton in feed formulations as well as close to £13k per annum on wet chemistry costs. Extending this to other variables such as energy value would save much greater sums.

Delivering NIR services
Today NIR equipment is usually laboratory based and loaded with appropriate calibrations. This presents some challenges: for example, sample delivery to the laboratory can result in delays that eradicate the advantages of speed of analysis. Further, calibrations quickly become out-dated; this requires updated calibrations to be updated on a frequent and on-going basis.

Recent developments in NIR hardware have allowed the production of robust, portable, battery-operated units. This has allowed analysis to be carried out at the point of interest e.g. the grain silo or feed mill intake, rather than in the laboratory, maximising the benefits from the speed of analysis and result delivery enabled through NIR use. In addition, in-line NIR equipment is currently available that allows feedstuffs to be monitored during harvesting or continual analysis of feeds during production in the feed mill.

Software and communications developments have also allowed for the introduction of web-enabled NIR services, where spectra are uploaded to a master machine containing all appropriate calibrations, with instantaneous feedback. This has several advantages: the analyst can pay on an “as-use” basis rather than paying a fixed up-front fee for a calibration, independent of sample numbers. This can also give the analyst access to a wider range of calibrations than would be available on a locally installed solution, all with the added benefit of knowing that the calibrations can be updated regularly and efficiently as they essentially sit on one computer.

Novel use of NIR
Beyond the standard prediction of dietary composition, NIR use has recently been extended into sample identification. Work in other areas suggests that, providing suitable standards are available, NIR can be used to confirm the growing condition of feedstuffs. Another example of an extension of NIR technology is the determination of mixer profiles in feed manufacture. Mixer profiles are usually determined by analysing the variation (% CV) in components such as salt/sodium or protein/nitrogen in five to ten feed samples. However, this approach will include the variation in the assay procedure used to analyse the component chosen, and the result could thus be considered to only apply to the specific component analysed. Thus, if sodium is chosen, the mixer profile will reflect primarily the variability in the dispersion of added salt. This can be overcome by looking at the % CV across the NIR spectra of a series of samples. The % CV as estimated from ten samples of feed taken from a mixer run for 1-5 minutes clearly shows an optimal mixing time of 3-4 minutes, and that the NIR gives the same result as the analysis of specific feed components, but with lower variability (Table 1).

Conclusions
NIR is currently used to analyse feedstuff and feed compositions for quality control within the poultry feed industry. However, developments in hardware and software present the possibility of using this technology to determine the value of incoming raw materials as well as to control in-line and in real-time the accuracy of feed formulations. This is potentially worth several billion pounds in terms of feed cost savings and more predictable animal performance for the worldwide poultry industry. In the future we can expect to see laboratory, hand-held and in-line NIR equipment used widely in the purchase feedstuffs and in feed manufacture.

(From Proceedings of the 9th Turkey Science and Production Conference, Chester, UK).