This article is part of our series of original articles on emerging featured topics. Please check here to see other articles in this series.
Laboratory data from total mixed ration (TMR) samples have potential value when evaluating consistency and accuracy of the diet that was delivered to a pen of cows including:
- Assessing within bunk variation in nutrient delivery. When evaluating consistency of TMR mixing and delivery, samples are taken at various locations across the bunk, analyzed for nutrients or particle size, and then some measure of variation, such as coefficient of variation (CV) or standard deviation (SD), is calculated and compared to a benchmark.
- Assessing day-to-day consistency of TMR delivery. The same basic approach as above except the TMR is sampled over multiple days and then variation is calculated among the daily samples.
- Determining whether the delivered ration matches the formulated one. Because of normal variation in ingredient composition and random or systematic errors associated with the individual doing the feeding and the scales on the mixer wagon, the delivered diet may differ markedly from the formulated diet. To evaluate accuracy (how close the delivered diet matches the formulated diet), samples are taken and results are compared to the specifications of the formulated diet.
Although using TMR composition data to evaluate diets and troubleshoot nutritional problems has potential, to be useful TMR data must meet the following to criteria:
- Sampling variation (e.g., variation among results from samples taken at the same location within a feedbunk on a given day) must be known. Without knowing sampling variation, you might conclude that mixing is poor because you have a high CV across the feedbunk (or across days), but in reality, the high CV might have been caused by poor sampling technique.
- The nutrient composition of the sample must accurately reflect what was delivered to the pen (i.e., sample results must be accurate). If sample results do not match formulated expectations, you might assume ingredients have changed or blame the feeder for not following the recipe, when in reality, it might be the sample (or the sampler) that is to blame.
Should sampling error be a concern for TMR data?
Sampling error (or sampling variation) simply means that if you take multiple samples from the same population, you obtain different values. A TMR is comprised of particles that vary in density, size, shape, and nutrient composition. A stem of hay is light, long, and is generally high in fiber; whereas, a grain of salt is heavy, small, and has no fiber. The extreme heterogeneous nature of TMR makes them extremely difficult to sample accurately, thus sampling error is indeed a major issue with TMR data. In a field study (conducted by The Ohio State University) of commercial dairy farms across the U.S., sampling variation contributed 36 to 70% of the total within farm variation in TMR composition over a 12-month period.
- When assessing day-to-day variation, duplicate samples should be taken and then averaged. Variation among the daily averages should be calculated.
- When evaluating within bunk variation, the process should be replicated and results averaged.
- For example, you could take 5 samples across the bunk, measure particle size on those samples, and calculate the CV; that entire process should be repeated and the 2 CV should be averaged.
Comparing TMR sample results to the formulated diet
An experiment was conducted at Ohio State (see 2016 Tristate Dairy Nutrition Conference Proceedings for full details) to determine whether TMR sample results accurately reflect the TMR that was delivered. The TMR was sampled immediately after it was delivered to the pen using the protocol outlined below. Three different TMR mixes were sampled over 6 days. One TMR mix contained only silages and concentrate; another contained dry hay, silages, and concentrates; and the third contained hay, silages, whole cottonseed, and concentrate. Type of TMR did not have much effect so data from individual TMR will not be discussed. Each day ingredients were sampled and analyzed and amounts of each ingredient put into the mixer was electronically recorded using commercial TMR software. Actual inclusion rates and ingredient composition data were used to calculate the ‘true’ composition of the TMR which was compared to composition determined on TMR samples.
Are TMR samples accurate?
Accuracy has a flexible definition depending on how good is good enough. If you were constructing a nuclear submarine, tolerances might be expressed in nanometers, but if you are digging a hole for a fence post, tolerances may be several inches. For this project, if a sample result was within 5% of the true value, the result was considered accurate. Using that definition, a single TMR sample can accurately reflect the TMR for some nutrients but for other nutrients, even averages of multiple samples are unlikely to be accurate.
Dry Matter (DM) Comparisons:
For DM, a single sample of TMR (using the protocol outlined below) was almost always within 5% of the true value (Table 1) and no sample was more than 10% away from the true value. Although a TMR sample was accurate for DM, the value of knowing the DM concentration of a TMR is questionable because diets are usually not formulated to a specific DM percentage.
Crude Protein (CP) Comparison:
A single TMR sample was usually accurate for CP, but occasionally (approximately 1 out of every 15 samples) sample values were really wrong.
A single TMR was not reliable to evaluate NDF concentration of TMR with approximately 1 out of every 6 samples being more than 10% wrong.
A single TMR sample had no value in estimating the mineral concentrations of the delivered TMR. This finding may have practical implications for phosphorus-based nutrient management plans. If phosphorus intake is calculated by multiplying TMR delivery rates by phosphorus concentration of a TMR sample, estimated phosphorus intake could often be wrong by more than 20%.
Single versus duplicate samples of TMR:
For NDF and all minerals, single TMR samples were not accurate. But, what about means of duplicate samples? Taking 2 TMR samples and averaging them greatly improved the accuracy for NDF. The mean of duplicates were accurate 75% of the time and only about 1 out of 12 times was the mean more than 10% wrong. Averaging TMR samples, however, did not greatly improve accuracy for minerals. Less than 20% of the averages of duplicate samples were accurate for any mineral and between 1 out of every 2 means to 1 out of every 5 means were wrong by more than 10%.
|Nutrient||<5% Deviation||>10% Deviation|
|Dry matter, %||95%||0|
Recommended TMR sampling protocol
Another objective of the projected outlined above was to test different sampling protocols. Details are available in the paper discussed above but based on results from that study the following sampling protocol is recommended.
- As you walk the feedbunk carrying a clean container such as a 5 gal bucket, take a handful of TMR approximately every 10 to 30 ft and place it into the bucket. For shorter bunks sample at 10 ft intervals but for very long bunks sample at 30 ft intervals. You want to have at least 10 handfuls by the time you reach the end of the bunk.
- Alternate samples so that the top, middle and bottom third of the TMR is sampled.
- When taking the handful, ensure that your palm is facing up to avoid dropping small particles.
- After you have walked the entire feedbunk, mix the contents of the bucket and then dump the contents onto a clean floor or large piece of plastic.
- Spread the contents out into a circle, divide the circle into quarters and then using a scoop to ensure you get all the particles, place one of the quarters into a sampling bag and send to the lab. The sample should be larger than a softball but smaller than a volley ball.
Using a simple, yet good sampling technique for obtaining TMR samples was generally accurate for DM and CP; however, using results from a single sample had a high risk of being wrong (>10% different) with respect to NDF and minerals. Taking duplicate samples and averaging NDF values reduced the risk of being wrong to an acceptable level. Sampling TMR did not accurately assess mineral delivery and should not be used.
The Ohio State University