Feeding by Mary Beth de Ondarza and Ralph Ward Accurate analysis: NIRS versus wet chemistry

ith today’s high grain prices number of samples in the calibra- W Table 1. NIRS reduces variation and tight margins, dairy producers tion, the average nutrient value and need to better understand and con- Nutrient SD*, wet SD*, the standard error of prediction. (% DM) Average chemistry NIRS trol nutrient variation. Accurate and The R-square is the amount of vari- timely forage analysis, therefore, CP 15.8 0.33 0.05 ation explained by the calibration. becomes even more important for ADF 31.4 0.78 0.32 We want low standard errors of pre- reducing feed costs and maximizing NDF 41.9 0.70 0.28 diction and a high R-square value. production. It may also be important Ash 11.3 0.20 0.07 A good R-square value for an NIRS Fat 3.44 0.12 0.02 to look closely at the nutrient varia- nutrient prediction will be over 0.95; tion of commodity feeds and place a *Standard deviation below 0.80 is considered to be a higher value on consistency. questionable prediction. analyses, but one should keep possible Protein is a nutrient with high Accuracy versus precision issues with NIRS mineral analysis in repeatability by wet chemistry and There are two primary approaches mind. If there are potential mineral with significant analytical range in to testing forages and feeds: wet issues with the cows, minerals should most feeds. For the hay protein cali- chemistry and NIR or NIRS (Near be analyzed by wet chemistry. bration there is a large number of Infrared Reflectance Spectroscopy). samples, a low standard error and a Understanding the differences and Developing NIRS calibrations very high R-square. We would expect when to use each is important for The NIRS system used by the lab this calibration to predict well. obtaining the best information for must be calibrated with feeds and While the bakery waste starch the testing dollar spent. forages analyzed by wet chemistry equation has many fewer samples, Wet chemistry methods are the methods. NIRS is fairly accurate if starch is a relatively uniform chemi- most accurate at analyzing feeds the test sample is in the range of the cal entity and should predict well. and forages for nutrient content. calibration sample set in terms of The standard error of prediction is Nutrients are isolated using chemi- forage type, geographic region and low, and the R-square is quite high. cals and heat to break down the growing conditions. It is important On the other hand, the standard forage. For example, NDF (neutral that the sample be accurately iden- error of prediction for vomitoxin in detergent fiber) is the percentage of tified so that the most appropriate distillers is high compared to the aver- fiber in a forage sample that is not NIRS equation can be used. age vomitoxin value, and the R-square solubilized after boiling the sample The quality of an NIRS calibra- value is much lower than for tradi- in neutral detergent solution. Sam- tion for a specific nutrient depends tional nutrients. Both the vomitoxin ples are precisely weighed before on several things. There needs to be and nitrate predication are considered and after a chemical or heat process a spectral relationship between the to be of lower predictive quality. with the difference calculated as near infrared reflectance and the the amount of the nutrient. Unfor- nutrient in question. For water and Nutrient and fiber analysis tunately, wet chemistry analysis is most organic constituents, NIRS With the development of excellent time-consuming, costly and requires will “see” the nutrient quite well. In calibration sets and mathemati- skilled technicians. order to develop a good prediction, cal techniques, NIRS has become a With NIRS, a spectrophotometer there needs to be significant range well-respected method for nutrient is used to analyze the light spec- in the nutrient and good repeatabil- analysis. The NIRS Forage and Feed trum reflected off of a sample when ity of the wet chemistry assay that Testing Consortium has found that, it is exposed to infrared light. Each produces the data for calibration. when the analyzed protein value of nutrient has unique reflection char- The rule of thumb is that the range a feed is different when run by wet acteristics based on its molecular of the nutrient should be 10 times chemistry and by NIRS, the rerun structure (carbon, nitrogen and the error of the assay. wet chemistry result turns out to be hydrogen bonds). The reflectance of Protein in most plant species has the same as the NIRS result 80 per- test samples is compared with that significant range and the analytical cent of the time. For ADF and NDF, of a set of similar samples (calibra- method is generally precise, provid- the wet chemistry reruns turn out tion set) that have been analyzed ing opportunity to develop excellent to be the same as the initial NIRS by wet chemistry. NIRS as an ana- calibrations. By contrast, the starch results about half the time. lytical technique is faster than wet prediction in alfalfa is not nearly as Estimates of NDF digestibil- chemistry methods and requires good because there is little range, ity (NDFD) are obtained by in less labor. However, NIRS requires and the starch assay is less precise. vitro laboratory procedures using sophisticated calibration develop- microbes taken from the rumen of ment, instrument standardiza- Is the calibration good? a fistulated cow. NIRS can be used tion and constant quality control to There can be significant differ- for routine characterization of ensure good results. ences in the quality of nutrient cali- NDFD. The R-square of the NDFD NIRS is a secondary method based brations used by laboratories. NIR 30-hour digestibility prediction for on reference evaluation of nutrients nutrient calibrations are described corn silage is good, and the stan- and by definition will never be more by certain statistics that are used dard error of prediction is reason- accurate than the methods on which to assess calibration quality. Table able. It is probably still wise, how- it is based. However, as an analyti- 2 shows statistics for certain NIRS ever, to check NDFD using in vitro cal tool, NIRS is often more precise, nutrient calibrations of Cumberland laboratory procedures when bench- or repeatable, than wet chemistry Valley Analytical Services. marking quality or troubleshoot- analysis and can be used to reduce For each feed, the table shows the ing problems. analytical variation (Table 1). Minerals do not reflect infrared Table 2. Nutrient prediction better with high R-square values light. So, NIRS mineral analysis is Number of Standard an indirect estimate based on the Feed Nutrient samples Average error R-square typical relationship of minerals with Hay Protein 4,747 15.7 0.56 0.99 other nutrients in the forage. Many Corn silage NDF 1,915 43.5 0.76 0.97 nutritionists do use NIRS mineral Corn silage NDFD (30 hr.) 3,237 61.0 1.14 0.96 The author has a dairy nutrition consulting busi- Bakery waste Starch 359 9.1 0.46 0.97 ness, Paradox Nutrition, LLC, in West Chazy, N.Y.; Hay Nitrate 1,000 0.3 0.08 0.90 and Ward is president of Cumberland Valley Ana- Distillers Vomitoxin 338 4.9 0.79 0.84 lytical Services, Inc., Hagerstown, Md. HOARD’S DAIRYMAN Reprinted by permission from the February 25, 2013, issue of Hoard’s Dairyman. February 25, 2013 129 Copyright 2013 by W. D. Hoard & Sons Company, Fort Atkinson, Wisconsin.