Powered by FOSS
Field research

Don’t discount the value of NIR

In September 2011, our industry lost a very special individual, John Shenk (1933-2011). While many producers and nutritionists may not recognize the name of John Shenk, be assured that your laboratory personnel know him as one of the “fathers” of near-infrared (NIR) analysis.

by Bill Mahanna and Ev Thomas

John ShenkJohn grew up on a farm in Lancaster County, Pa., received his B.S. degree from the Pennsylvania State University, and M.S. and Ph.D. degrees in plant breeding from Michigan State University. John then returned to the Pennsylvania State University where he spent his entire academic career in the agronomy department, retiring in 1998 as a professor emeritus.

Rooted at Beltsville

Shenk first became introduced to NIR in 1972 when he heard of the work of USDA-Beltsville scientist Karl Norris using instrumentation whereby all the important forage constituents could be predicted in minutes rather than the days it required using traditional wet chemistry laboratory methods.

He then went on to make advancing the agricultural application of NIR technology his life work. In 1983, he formed a software company, Infrasoft International, LLC (ISI). Its software products became the standard used by NIR users worldwide. In 2005, John retired from his company, Infrasoft International, and became a consultant to Unity Scientific.

NIR is not new

NIR had been discussed in the literature since 1939. But, it was not until 1968 that Karl Norris and co-workers with the Instrumentation Research Lab of the USDA-Beltsville observed that cereal grains exhibited specific absorption bands of light in the NIR region. They suggested that NIR instruments could be used to measure grain protein, oil, and moisture. Further research proved that absorption of other specific wavelengths could be correlated with chemical analysis of other grains and forages.

Early in 1978, John Shenk and his research team developed a portable instrument for use in a mobile van to deliver nutrient analysis of forages directly on-farm and at hay auctions. This evolved into the use of university extension mobile NIR vans in Pennsylvania, Minnesota, Wisconsin, and Illinois. In 1978, the USDA NIRS Forage Network was founded to develop and test computer software to advance the science of NIRS grain and forage testing. By 1983, several commercial companies had begun marketing NIR instruments and software packages to commercial laboratories for forage and feed analysis.

How does NIR work?

Near-infrared spectroscopy is based on the interaction of the physical matter of feeds with light in the near-infrared spectral region (700 to 2,500nm). Vibrations of the hydrogen bonded with carbon, nitrogen, or oxygen causes molecular “excitement” responsible for absorption of specific amounts of radiation of specific wavelengths. This allows labs to relate specific chemical bond vibrations (spectra) to the concentration of a specific feed component (starch) determined by traditional wet chemistry methods. NIR is possible because molecules react the same way each time they are exposed to the same radiation.

Sample preparation and presentation to the NIR instrument varies widely. Though dried, finely ground samples are often employed, whole grains or fresh, unground can also be scanned. Instruments are increasingly coming to the market that are rugged enough to work in mobile applications such as on-board silage choppers.

What is a “calibration”?

NIR is a rapid, secondary method based on the mathematical relationship (regression) with the accepted wet chemistry method. Consequently, an NIR value can never be more accurate than the traditional method.

Sophisticated software packages are used to perform the mathematical calculations necessary to associate the NIR-produced spectra of specific reference samples with the actual wet chemistry of those samples. These mathematical relationships are termed “prediction models” or “calibrations.”

The robustness of an NIR calibration is primarily determined by:

• the number of samples

• how well they represent the diversity of the feedstuff

• the typical variation observed for the trait being measured

For example, if the goal is to develop a calibration for crude protein in corn grain, samples of corn from diverse genetic and environmental backgrounds must be included in the reference samples being analyzed by wet chemistry. When a particular wet chemistry method does not exist (for prediction of ethanol yield from corn), laboratories may develop an entirely new wet chemistry method upon which to base the NIR calibration.

Cost is a big advantage

The tremendous advantage to NIR over wet chemistry is the cost savings. You can analyze more samples, more often, for the same budget compared to the more expensive wet chemistry. This also helps producers better manage feedstuff nutrient variation by more frequent analysis.

It was not uncommon in the past to hear recommendations to only consider wet chemistry analyses, especially following atypical growing seasons. However, with laboratories using frequently updated calibrations that contain thousands of samples from extremely diverse growing conditions, that advice is no longer credible.

Questions for your lab

As routine users of NIR values, producers and nutritionists should feel comfortable asking laboratories about their NIR statistics. This will help instill confidence in these values similar to the way we have been taught to use statistics (P-values) to determine the confidence we put in research trial results.

Here are three NIR statistics that reputable NIR laboratories should be able to provide:

1) The number of samples in the calibration set. The narrower the range in sample differences, the more difficult it is for NIR (or wet chemistry) to detect those differences. Typically, 80 to 100 samples are required for developing an initial calibration with up to multiple-hundreds of samples in a “mature” calibration.

2) The standard error of calibration (SEC). It defines how well the NIR calibration predicts the wet chemistry values that were used to build the calibration. Low SEC values are desired. For example, if the wet chemistry value is 30 and the SEC is 3, this means approximately two-thirds of the NIR values should fall within the range of 30 +/- 3 and range from 27 to 33.

3) The regression coefficient (R2 or RSQ). It is the “best fit” line when NIR values are plotted against the wet chemistry values. High R2 values are desired. An R2 of 1.0 means 100 percent of the sample variation is being explained by the calibration.

This is article is used by permission from the February 10, 2012, issue of Hoard’s Dairyman.

Copyright 2012 by W.D. Hoard & Sons Company, Fort Atkinson, Wisconsin.

Want to know more about the potential and application of near infrared technology in feed production? Join our fast growing network of experts and professionals from both NIR and feed sectors now.