Home >

Sampling Feedstuffs for Lab Analysis a Necessity


Sampling Feedstuffs for Lab Analysis a Necessity

by Bill Mahanna

Diet formulation is only as good as the feed analysis upon which it is based.

The typical approach is to sample feed ingredients, ship them to a preferred laboratory and load the data into ration balancing software. Submitting the appropriate number of representative samples, along with an understanding of the laboratory analytical error, is pivotal in the decision to either maintain or alter the current dairy diet.

This month’s column will focus on preventing protocol drift when sampling feeds, which is especially important in situations where the dairy is directly submitting samples at the request of the nutritionist.

Representative Sample

Obtaining a representative sample begins with an appreciation of the heterogeneity of different feedstuffs. Sampling 1,200-micron, ground, high-moisture corn almost certainly assures a more representative sample than when sampling heterogeneous feeds like corn silage or a total mixed ration (TMR).

When sampling, the varying nutrient contribution derived from stems, leaves or grain (pieces or fines) represents one of the first challenges to determine the true concentration of feedstuff nutrients.

The variation in the nutrient composition of a diet is a function of the variation in the nutrient composition of the ingredients, the inclusion rate of each ingredient and the number of ingredients in the diet.

On a theoretical basis, the contribution of a feedstuff to the variance of the total diet grows with the square of its inclusion rate. In other words, if you double the inclusion rate of a feed, its impact of diet variation increases by a factor of four.

This is certainly something to consider in diets with high inclusion rates of one feedstuff, like corn silage. Highly variable ingredients can be used successfully if their inclusion rates are kept low (Weiss and St.-Pierre, 2007). This may help explain a personal observation that nutritionists in the West, feeding multiple feedstuffs, tend to be less excited about new laboratory methods than nutritionists in the Midwest and East who tend to feed diets with fewer ingredients at higher inclusion rates.

Sampling, Subsampling

The submission of representative subsamples as well as the replicationof samples and analyses are the main avenues to reduce variation in analytical results. One of the merits of near-infrared spectrometry (NIR)-based analysis is reducing analytical error by testing more samples with the laboratory budget than possible with more expensive wet chemistry methods.

The best and safest approach to sampling forages from bunkers or piles is to take samples directly from the TMR mixer after a loader has put a representative amount of forage into the mixer and the mixer has run for several minutes. Avoid any top or side spoilage that would normally be discarded by the feeder. Several scoop samples can be taken from the discharge chute and composited. If this approach is not feasible, collect 20-scoop samples from a recently exposed bunker or pile face in a 5 gallon bucket.

For either approach, use the quartering technique (Figure 1) to reduce the sample volume to one to two quarts for laboratory submission (Taysom, 2014). It is important to confirm with the laboratory its desired volume of submitted sample. Analyses such as corn silage processing score or grain particle sizing may require larger samples.

Feed analysis - subsample quartering technique.

Feed analysis - maximum average difference in traits that can be detected with a given number of samples.

Hay samples should be taken with a hay probe to reduce leaf loss, which can significantly affect analysis results. The Miner Institute (Cotanch, 2008) reported on a field trial where western alfalfa hay was sampled by hand, based on the knowledge that many producers do not own a hay probe. The hay was sampled either by grabbing loosened hay while trying to capture leaf and stem proportionally or by taking undisturbed, compact clumps of hay. The difference in neutral detergent fiber (NDF) and crude protein between these hand sampling methods was 38% versus 52% and 18% versus 16%, respectively. This clearly shows the value of owning a hay probe.

Field, windrow and baler-type differences lead to bale structure variation requiring that 20 small rectangular or 8-10 large package bales be sampled per lot. For small square bales, it is suggested to take a core sample through the center of either end. For large square bales, take a core sample anywhere on the bale, but insert the corer at a 45-degree angle to the side or a 90-degree angle to the end. For round bales, sample on the curved side with the corer perpendicular to the side (Undersander et al., 2005).

There was a proposal by the National Forage Testing Assn. and the National Hay Assn. for growers and buyers to submit three replicated, distinctive, eight-core samples taken from the same lot of hay (not simply splitting a 20-core sample into three subsamples). The idea has not really taken hold among most producers or laboratories, but the intent was that if a seller and a buyer submitted multiple, independent samples, even to different labs, the average would likely be similar.

In lieu of this scientific approach, hay sellers and buyers should at least agree on a single lab for analyses to avoid differences in analytical methods that commonly occur among labs.

Representative TMR samples can be collected from the mixer discharge chute or by following the feed truck or mixer wagon and collecting about 10 samples as they are discharged in the feed bunk. Insert a latex-gloved hand into the middle of the feed pile about wrist deep, and avoid squeezing or shaking the sample before depositing it into a plastic bucket (Robinson and Meyer, 2010).

Sampling at harvest provides advance knowledge of the quality of grain and forage. Some nutrient fractions will change significantly during storage, such as pH, volatile fatty acids (VFAs), starch digestibility and protein solubility; while other parameters change verylittle (crude protein, acid detergent fiber and NDF) during a reasonably normal fermentation (Dave Taysom, personal communication).

On-farm subsample variation for any type of feed sample (other than hay) can be reduced by employing the quartering technique (Figure 1). The technique begins by thoroughly mixing the material to be sampled (e.g., rolling it back and forth on a piece of plastic), then pouring it into a cone-shaped pile on a clean surface.

Shape the pile into a cone by scooping material from each side up and towards the middle of the sample, and then flatten the cone. Divide the flattened sample into four equal parts (quarters) using a drywall joint knife, trowel or any straight-edged tool. Discard two opposite quarters, combine the two saved quarters into another pile and then quarter again. Be sure to collect fine material at the bottom of the saved sample.

Continue this process until you have a pile that is the amount needed for a laboratory analysis (Taysom, 2014). If the sample exceeds the laboratory’s desired amount for quick drying and analysis, the laboratory will employ the same technique to further reduce the size of the submitted sample.

Some nutritionists incur a slightly higher lab fee by requesting that the laboratory dry and grind their entire sample without any further reduction. Another approach to assure representative samples, especially for TMR analysis, is to monitor mineral or fiber levels to see if they are similar to formulated values before submitting samples for more sophisticated and costly analyses such as Fermentrics.

Finally, personal experience has shown that while the accurate subsampling of the diet may analyze as formulated, TMR mixing or distribution issues can cause performance problems and are the reason successful nutritionists tend to spend as much time observing feeding as they do sitting at the computer.

Sample Handling

The proper approach to sample handling depends on the type of analysis being requested. For normal, wet-chemistry or NIR nutrient analyses, freezing the sample and getting it to the lab before it thaws is an acceptable approach. It is also important to freeze samples if a silage VFA, nitrate or prussic acid analysis is being requested so that spoilage organisms are not active and secondary fermentation does not proceed during sample shipment.

However, if submitting a sample for either microbial/fungi analysis or more sophisticated digestion kinetics, it is best to only refrigerate the sample to avoid freezing, which could cause the rupture of cells and result in misleading counts or digestion rates.

Submitting samples for mycotoxin analysis presents yet another challenge. While most nutrients like protein, fat and carbohydrates are somewhat uniformly distributed in feed, the presence of mycotoxins tends to lack uniform distribution. The second issue is testing for the presence of a toxin at either a parts per million or parts per billion level. When the sample size is too small, the toxins are either missed entirely or detected at much lower levels than are actually present. Research indicates 53% total variance associated with the submission of a 10 lb. sample of corn grain contaminated with 20 ppb of aflatoxin, due to field sampling, laboratory subsampling and lab analysis (Whitaker et al., 1978). A minimum sample size of at least 5 lb. is suggested when analyzing corn for mycotoxins (Romer Labs, 2014). Knowing that dilution is often the best solution, analyzing a representative forage sample gives an indication of the dietary dilution necessary for the contaminated forage source. Some prefer to analyze obviously moldy feed as an indicator of the maximum toxin levels that could be present. However, this can be very misleading because actively growing fungi may not be environmentally challenged to produce toxins, and levels in these locations may actually have lower toxin levels than other areas where fungi in the field experienced environmental stress that caused toxin production before harvest and storage.

How Much? How Often?

Some nutritionists use a rolling average for nutrient values rather than making significant dietary changes based on a single analysis. This makes sense if the variability (standard deviation) of the feedstuff is relatively small and there is no apparent reason why composition should have significantly changed.

Differences between analyses could have been caused by random variation from load to load, by variation within a load (i.e., sampling) or both. In this case, the new number may be no better than the old number, but the average of the two numbers has the lowest probability of being substantially wrong (Weiss and St.-Pierre, 2007).

The goal of feed analysis is for the sample taken on the farm to reflect the actual value of the entire population. The issue becomes defining the population (within an individual truck, field or bunker) and the acceptable level of confidence required for that trait. Figure 2 provides approximate guidelines for the number of samples that should be taken to be 95% confident of the differences that exist for a measured analytical trait. The chart assures a coefficient ofvariation that is within reasonable limits for most laboratory-measured moisture or nutritional traits (Sapienza, 2001).

The acceptable minimum average difference is determined on the Y-axis and then is referenced across the chart until reaching the curve intercept and referencing on the X-axis the approximate number of samples that should be taken. For example, if a 2% unit difference is acceptable for corn silage starch content (e.g., 35%, which can vary from 34% to 36%) and the population is defined as a reasonably uniform field with the same hybrid, then 11 samples are required to be taken from the field.

This can be accomplished by how many truckloads of forage are required to harvest the field and the rate at which trucks are coming in from the field. If 22 trucks are needed to haul chopped forage from the field and 22 trucks per hour are coming to the silo, sampling every other truck will obtain the required 11 samples. However, if the population is defined as one truckload of forage, then the truck needs to be sampled 11 times. If the population is defined as a bunker silo containing one hybrid, then the bunker needs to be sampled 11 times (Sapienza, 2001).

The Bottom Line

Sampling feedstuffs for analysis is a routine and costly necessity to assure properly balanced and economically driven diets. It is critical to the modification of diets that representative samples are obtained and handled in a way to assure that meaningful data are incorporated into ration balancing software.

Implementing sampling protocols and periodically evaluating to prevent protocol drift are critical to the efficiency and productivity of every dairy herd.


Cotanch, K. 2008. Comparison of two hand-sampling techniques of processed alfalfa hay. Miner Farm Report. March.

Robinson, P.H., and D. Meyer. 2010. Total mixed ration sampling protocol. University of California Publication 8413.

Romer Labs. 2014. Sampling procedures for mycotoxin analysis.

Sapienza, D.A. 2001. Number of samples (or locations) required to determine moisture and dry matter of forages. Pioneer Nutrition & Feed Management Today, Vol. 1, No. 3.

Taysom, D. 2014. Sampling tips conventional analysis.

Undersander, D., R. Shaver, J. Linn, P. Hoffman and P. Peterson. 2005. Sampling hay, silage and total mixed rations for analysis. University of Wisconsin Extension Publication A2309.

Weiss, B., and N. St.-Pierre. 2007. Understanding and managing variation in nutrient composition. Proceedings of the Western Dairy Management Conference. Reno, Nev.

Whitaker, T.B., J.W. Dickens and R.J. Monroe. 1978. Variability associated with testing corn for aflatoxin. J. AOCS 56:789- 794.


Originally published in the December 2014 Feedstuffs issue. Reproduced with permission.


The foregoing is provided for informational purposes only. Please consult with your nutritionist or veterinarian for suggestions specific to your operation. Product performance is variable and subject to a variety of environmental, disease, and pest pressures. Individual results may vary.