This is the second article in a series to enhance your understanding of the function and use of NIR technology.

What is a GLOBAL calibration?

As its name implies, a GLOBAL calibration is all encompassing, i.e., a single calibration is developed to analyze a specific nutrient for all samples of a given feed type. For example, the CP calibration for hay may be built using a database comprised of a continuum of reference samples beginning with straw at 2%CP up to 28%CP in immature alfalfa hay. Every new sample of hay submitted for analysis is analyzed using this single hay CP calibration. This is the traditional method of calibrating.

 

What is a LOCAL calibration?

A LOCAL calibration is built specifically for each sample as it's presented for analysis, i.e., each sample is analyzed using its own individual calibration. When a sample is scanned, its spectra is compared to all of the other spectra stored in the calibration database. The software searches the database and identifies the 120 samples that most closely resemble the target sample. These samples are considered the "nearest neighbors", hence the term LOCAL. For example, using the hay CP database of 2% - 28%CP, a calibration for a sample testing 25%CP would most likely be built using samples from 23% - 27%CP. The more the reference population looks like the target sample, the stronger the calibration.

 

In short, what's the difference between GLOBAL and LOCAL?

What is required to make LOCAL work?

A large diverse sample database. The thousands of samples that we've collected over the years has enabled us to make great use of this advanced technology.

 

Aside from the hay example, how have you employed LOCAL to improve predictive ability?

BMR corn silage is an excellent example. Our corn silage database is populated with both conventional and BMR corn silage. In theory, when a sample of BMR is presented for analysis, the NIR software will search the database for other BMR samples. Using LOCAL technology eliminates the need for having separate calibrations for each. In fact, as reported in Newsletter No. 57 , the power of LOCAL software was able to differentiate between BMR1 and BMR3.

 

How about TMRs?

Not surprisingly, TMRs vary widely across the globe. There's even great variation within the US. TMRs from the Northeast and Midwest tend to be corn silage/haylage based with corn and soybean meal making up the primary supplemental energy and protein sources. Florida TMRs tend to be low in forage but high in fibrous byproducts. California TMRs incorporate a lot of vegetable and fruit processing byproducts. The LOCAL program will sift through the 7,000 TMRs in the database to find the samples that most nearly resemble the TMR submitted for analysis.

 

What about the analysis of minerals?

Inorganic molecules (minerals) do not absorb light in the near infrared region. Mineral calibrations are based on an indirect relationship between minerals and organic compounds within the sample matrix. For comparison, Table 1. represents the results from a random sampling of forages analyzed one morning. The samples were analyzed by both wet chemistry and NIR.