History of NIRS Near-infrared radiation was first identified in 1800 by William Herschel, who was curious as to whether a certain color was associated to the heat of sunlight. It is this kind of curiosity that has characterized advances in near-infrared spectroscopy (NIRS) in forage and feed testing.
In 1939, NIRS was described in scientific literature but was neglected because it was believed that other spectroscopy was more useful. Then, in 1968, NIRS was first used to describe the characteristics of agricultural products by Karl Norris and associates at the USDA Instrumentation Research Laboratory in Beltsville, Maryland.
This was an important breakthrough, since Norris was the first to combine the right spectrometer, computers and advanced mathematics to develop prediction models for animal feeds. By the 1970s, an array of chemical analysis methods were commonly being used by labs to describe the nutritional characteristics of forages.
In 1976, Karl Norris and his group took the chemical analysis data and used it with NIRS to predict crude protein (CP), acid-detergent fiber (ADF) and neutral-detergent fiber (NDF). Concurrently, John S. Shenk and colleagues at Penn State University were developing advanced software for performing the NIRS analysis.
By the 1980s, hardware and software were available for NIRS testing, and some university and commercial agricultural testing labs had adopted the technology.
However, there were variations in techniques, and in 1990, a group led by John S. Shenk was formed to evaluate the variation among testing labs and determine ways to standardize instrument and software methodology.
In 1992, the group was organized into the NIRS Forage and Feed Testing Consortium (NIRSC) with specific goals to match instrumentation and software and develop standard operating procedures for users. These basic goals of the NIRSC remain today within a much larger realm of testing strategies and animal nutrition characteristics.
The broad scope of the NIRSC aims to leverage and unify the efforts and knowledge of its members and collaborating experts in order to create and promote the basis for rapid and accurate analysis results to return value and benefits to its members, the NIRS community and the public.
How NIRS works
NIRS works by shining light in the near-infrared region onto a prepared sample and recording how the sample reacts to the light in the form of a spectrum. For example, to get to the nutritional values of the sample, the spectrum is compared to a pool of reference analyses.
The reference analyses are sets of results from many samples with laboratory chemical data. The spectra and data of the known samples run with wet chemistry are compared to the spectra of the unknown samples, and through mathematical equations, the nutritional values of the unknown samples are predicted.
Using NIRS can be very precise, with the instrument consistently scanning and producing spectra from samples.
Challenges
One issue of vital importance to producers is variation of results among commercial forage testing laboratories. The National Forage Testing Association (NFTA) carries out a certification program that tests labs’ proficiency in analyzing dried and ground samples using NIRS as well as wet chemistry analysis.
NFTA has published recommended laboratory procedures for dry matter (DM), CP, NDF, ADF, ash, crude fat (CF) and acid-detergent insoluble nitrogen (ADIN). Labs are encouraged to use the methodology as outlined by NFTA, but it is not enforced.
In the mid 2000s, several blind studies were conducted that indicated many commercial testing labs indeed showed variation throughout the sample handling process.
The blind studies were carried out by groups including Nebraska Alfalfa Marketing Association and two separate university groups. Much of the onus was placed back on the commercial testing labs to get their act together and exhibit uniform results.
So what causes variation in feed sample analysis? There will be variation in every step of the sample handling process due to, for example, splitting samples, heterogeneity of samples, variation in the wet chemical techniques used to predict from the NIRS scan or differences in the mathematics and software used to predict samples.
Variation will inevitably occur within a lot of feedstuffs, even before the lab gets the sample.
A proficiency program like NFTA’s helps, but it does not guarantee that methodology is standardized. A further problem is getting the basic facts straight concerning error. The producer, forage broker, nutritionist and other users of lab data need to understand the sources of error and what to expect for statistical variation.
NIRS analysis can only be as good as originating wet chemical analysis. Then there is an additional level of error added to that because chemical data is used as a basis for an NIRS prediction.
Error increases again if an NIRS constituent is based upon a two-stage chemical analysis, for example, with the determination of NDFD (neutral- detergent fiber digestibility, or dNDF – portion of the neutral-detergent fiber digested by animals at a specified level of feed intake, expressed as a percentage of dry matter).
Reducing variation in lab results
The NIRSC is dedicated to countering the sources of error and reducing variation where possible. In the 1990s, much variation in NIRS testing was due to instrument differences, and NIRSC initiated a program to standardize instruments.
Today, the variation comes from differences in instruments where there is no standardization, in sample handling and equation development and use. Any lab with an NIRS instrument can develop a prediction model and all of these can be different. However, if there is a network between labs and monitoring among labs, the variation can be decreased.
NIRSC advocates a standard sample database and standard equations. In addition, the consortium has developed a set of recommendations that will greatly reduce variability.
These include a comprehensive set of standard terminology and units, instrument and equation check testing to ensure equations have not been corrupted or instruments have not drifted and a comprehensive sample handling guide for labs that test using NIRS.
An up-to-date, start-to-finish guide has not previously been available exclusively for NIRS testing. Included in this guide are important directions that connect back to the producer and the very start of the analysis process – taking the sample.
If the lab starts with a proper sample representative of an appropriately sized lot of feedstuff, the NIRS testing lab can then follow standard procedures to ensure the most accurate results and ultimately reduce variability.
The methodologies detailed in the consortium’s guide encompass the most recent standards for techniques and are aimed to overcome the problems as described in previous blind sample studies.
In addition to using these guidelines, NIRS has further advantages because more samples can be taken and run economically that represent subsamples of a lot, the subsample analyses can be averaged, and this reduces the overall analytical error both from sampling and laboratory analysis for the entire lot of feedstuff. PD
Patty Laskowski-Morren
Executive Director
NIRS Consortium
What can you, as a producer, do to minimize forage testing variability?
• Identify single lots of hay (less than 200 tons; don’t mix cuttings, fields or hay types).
• Sample close to feeding or before sale.
• Choose an appropriate sharp sampling device (3/8-inch to 3/4-inch); never use flakes or grab samples.
• Sample at random (do not avoid bales; choosing or avoiding bales introduces bias).
• Take enough cores to represent a lot (20 or more per lot).
• Use proper technique (90º angle, 18-inch to 24-inch depth).
• Handle samples correctly (use plastic, Ziploc-type bags and protect from heat).
• Use appropriately sized samples (half-pound).
• Only split samples to verify lab tests after the complete sample is ground; better yet, ask for your ground sample back to send to another lab, as good labs should be willing to test their performance.
Adapted with revisions from NFTA hay sampling certification exam.
