The typical process of formulating a diet for dairy cows goes as follows: (1) sample the forages on the farm, (2) send the samples to a good lab, (3) when the lab results are available then enter the data into a computer, and (4) formulate the diet using a good dairy cow nutrition model. Because forages usually make up more than half the diet dry matter (DM), using incorrect nutrient composition data for the forages could result in an unbalanced diet, which could reduce yields of milk or components, or increase health problems.
We have been conducting a large project evaluating variation in nutrient composition of feeds. Commonly, silages on a farm are sampled about once monthly and the data from that single sample are used to formulate or re-formulate diets. One objective we had was to determine if that approach is in fact adequate. We sampled corn and haycrop (mostly alfalfa but some farms fed mixed grass and alfalfa) silages on several Ohio dairy farms and on a few farms in Vermont each day for 14 consecutive days. Each day, we took 2 independent samples from each silage. Independent means that we took several handfuls of silage, put them in a bucket, mixed that and then took a few handfuls, and put them in a bag to be sent to the lab. We then repeated that process to get the 2 independent samples. All samples were sent to the OARDC dairy nutrition lab and each sample was assayed in duplicate for DM and neutral detergent fiber (NDF). Haycrop silage was also assayed for crude protein (CP), and corn silage was assayed for starch. By taking duplicate samples from multiple farms, over multiple days and then analyzing everything in duplicate, we could partition the variation into that caused by farm, sampling, analytical, and day.
Sources of Variation
The nutrient composition of feeds can vary for a number of reasons. It is important to know what caused the variation when formulating diets.
- Farm variation in nutrient composition of silages reflects different growing conditions on different farms, different hybrids, different harvest times, etc. Because of the numerous factors that differ among farms, this variation is usually very large.
- Analytical variation is usually caused by human error (for example very small differences in weighing), instrument calibrations, reaction conditions, etc. It could also be caused by different labs. In this study, all samples were analyzed in a single lab. So the analytical variation that we observed is less than what would be experienced if samples were sent to different labs.
- Sampling variation can be a difficult concept to understand. If you have a pile of corn silage that will be fed today and you grab 5 handfuls of silage and put each into a separate bag and send each bag to a lab, you will likely get 5 different values for CP, NDF, starch, and DM concentrations. These differences represent sampling variation (sometimes referred to as sampling error). For corn silage, two samples could have different NDF concentrations because one sample had a little more corn cob in it than the other sample. Although one should always try to take representative samples, multiple samples of different feeds will never be identical.
- Day variation can also be called true day-to-day variation. This means that the composition of the feed really did change over time. This change could be caused by differences in harvest time (for example, the sample of alfalfa silage taken on Monday may have been harvested late in the afternoon, but the sample taken on Wednesday was harvested in the morning), field location (e.g., a weedy or dry spot in the field was sampled on a specific day).
Since forages are almost always sampled for each specific farm, farm-to-farm variation is not that important. In this study, farm variation was very large, meaning that silages should be sampled for each farm. However, separating true day-to-day variation from sampling and analytical variation within each farm is important. If a sample of silage is taken this week and it has 40% NDF and another sample is taken next week and it is 45% NDF, if that difference was caused by sampling error (in other words, the silage really did not change) and you reformulate the diet to match the new NDF concentration, the new diet is not going to be properly balanced. On the other hand, if the silage really did change (a true day-to-day change) and the diet is not reformulated, the diet being fed also is not properly balanced.
What We Found
Analytical variation for all nutrients and both types of silages was low, meaning you do not have to pay labs to analyze a given silage sample in
For corn silage NDF and starch and for haycrop NDF and CP, sampling errors were much greater than true day-to-day variation. This means that over a short period (a few weeks), differences between samples in nutrient composition are likely not a real change. The data for the samples should be averaged and the average values should be used in ration formulation.
- True day-to-day variation was the major source of variation for DM concentrations of haycrop silage. This means that when DM concentrations change among samples, the change is likely real and diets should be modified. For corn silage DM, true day-to-day variation was about equal to sampling plus analytical variations. This means you should probably measure DM on duplicate samples and if the averages between 2 sets of samples are different, the silage DM really changed and the diet should be modified.
The nutrient composition of silages is variable. However many times when we think that the silage has changed, it really is simply sampling error. Good sampling techniques should reduce sampling variation, but taking duplicate samples and averaging the results will greatly reduce sampling variation. Be careful when making diet changes based on lab results; make sure the feeds have actually changed.