Comparing handheld NIR meters on farm

By Dr. Michelle Chang-Der Bedrosian, Vita Plus forage products and dairy technical service
Now is an exciting time to be in agriculture. New technologies are developing everywhere we turn, and forages are no exception. We have the ability to immediately measure the moisture content of common feeds – in the field or silo – simply with the push of a button.

Many people live their lives by the phrase “trust, but verify.” That phrase can also be applied to producers’ relationship with new technologies, including handheld near-infrared (NIR) meters. Here are some tips to verify the accuracy of a handheld NIR meter on your farm.

Start with the low-hanging fruit
At the wavelength spectra that most of the handheld NIR meters read, the biggest peak (and therefore the easiest for the meters to read) is moisture. It is safe to say that, if the handheld NIR meter cannot accurately predict moisture, it certainly will not accurately predict other variables (i.e. crude protein, neutral detergent fiber, fat, etc.).

The logical first step when verifying the accuracy of a portable NIR meter is to determine what you are comparing it to. When comparing a highly flawed instrument to another highly flawed instrument, statistically speaking, the error is magnified, and it is not an accurate verification. Compare the dry matter (DM) readings of an NIR meter to laboratory results (which have the lowest error) to minimize the collective statistical error of the test.

Be clear on the proper protocol for testing
It is essential to compare apples to apples. When testing a sample, send that same exact sample to the laboratory. This may require a small diversion from routine testing protocols. With routine testing protocols, the goal is to get a sample that is representative of the silo/field in question. This is different than the goal of validation (testing one device against another). When validation is your goal, the most important thing is to test the exact same sample on both devices.

For example, when verifying the accuracy of a SCiO Cup, get a good-quality, representative sample, fill the cup, and press “scan.” Then dump the sample into an empty bucket, wipe the dome, and put the same sample back into the cup before pressing “scan” again. Repeat this procedure with the same sample for a third time before pressing “analyze.” Record the DM reading for your records and then send that same exact sample to the laboratory for analysis.

In other words, do not send the laboratory a different sample than what was scanned with the handheld NIR meter because this would artificially increase the error. If you use two different samples for the two different tests, you are actually testing BOTH the equipment AND your ability to take a representative sample, which magnifies human error.

Compare the results
Keep a log of the DM values from both the lab and handheld NIR meter. “Statistical power” refers to how robust a data set is. In a nutshell, the average of 20 samples is stronger than the average of two. Calculate the difference between the lab and handheld NIR meter and keep an average of that. A normal deviation to expect is two percentage points in either direction. The more samples that are in a data set, the closer to this normal deviation you can expect to be.

Ask for help if you need it
If you run into issues with this comparison, don’t be afraid to ask your consultant for assistance.