The Alfalfa Breeder’s experience

As you may remember, at the Consortium annual meeting we discussed the importance of instrument standardization. At that time it was proposed to include a forage standardization set (14 sealed cells) instead of only the characterization set (30 sealed cells) normally used. Before using the new standardization procedure on all the instruments, we had agreed to test it on the 6 alfalfa breeders and make an evaluation comparing the standardization performances with the 2 sets of sealed cells.

Following Mark Matteson’s advice, the 2 standardization sets available to the Consortium were tested for leaks. As a result of those tests, all the cells had to be repaired. The 30 cell characterization set was repacked with the same samples that the cells originally contained. In the 14 cell forage set, six new forage samples were selected from the Alfalfa breeders, including both 1 and 2mm grind, to replace some outdated samples.

The two standardization sets were scanned twice on each breeder’s instrument and samples that did not show good spectra repeatability were scanned again. The same process was carried out on the Consortium Master at US Dairy Forage Research Center before and after each breeder visit. Spectral matching between the Master and each breeder’s instrument was performed using the entire characterization set and also using the 6 cups filled with the breeder’s samples. Comparison of the two standardization techniques were made using breeders equation (99alf2.eqa) to predict the remaining 8 cups of the forage set, which function as an independent set to evaluate instrument differences after standardization. The new standardizations were then compared with the old standardization, already present in the instrument computer, which was developed using only the characterization set.

Summary statistics of the prediction made by the 6 alfalfa breeder instruments using different standardizations are shown in Table 1. The variation among instruments using the standardization already in place was quite large. For example the prediction of the 8 validation samples averaged 21.42 % for CP, but the lowest instrument was predicting 20.78 and the highest 22.23 with a difference of 1.5 %. Large variations were also evident for ADF and NDF.

Running a new standardization with the repaired characterization set reduced the variability as exhibited by the decrease in standard deviation (SD) values. In particular, the new characterization reduced the spread among instruments by improving the performance of the instrument with the oldest standardization file. This instrument with the oldest standardization was consistently extreme in the range of values in the old standardization. This improvement outlines the importance of running the standardization procedure on a regular basis, every one or two years.

The greatest improvement in the network performance, however, was obtained with the forage standardization, which reduced the SD to 0.1 for CP and lower than 0.25 for the fiber fractions. ISI recommends that the check cell predictions that indicate instrument variation should have a maximum SD of 0.30. The SD obtained with the forage standardization set indicates variability among Consortium instruments that is within the variation limits normally accepted for a single instrument.

Further, the 14 cell forage set, repaired and with 6 new samples, has showed consistent scanning in all components. This indicates that leaks are not present and the set is ready to use for standardization on commercial labs.

Moving to the second phase of the standardization process, I would suggest that we use the forage set to standardize the commercial lab instruments. For those labs that also use Consortium non-forage calibrations (i.e. roasted soybean equation), standardization with the characterization set will also be used. The forage standardization is recommended for scanning forage samples, and the characterization standardization is recommended for scanning non-forage products. The appropriate standardization may be chosen; depending on what type of sample is being scanned.

Considering the outcome of this first evaluation, I would suggest the Consortium develop a third standardization for corn silage. This would require the participation of labs in the search for corn silage standards. More details of this CS Standardization will follow, after consensus and input from commercial labs.

Table 1. Mean, Minimum, maximum and standard deviation of prediction of 8 forage standardization set samples by the breeder’s instrument using different standardizations.

	Before STD	STD with Characterization set	STD with Forage set
CP
Mean	21.42	20.92	20.89
Min	20.78	20.17	20.76
Max	22.23	21.16	21.00
SD	0.53	0.36	0.10
ADF
Mean	33.18	33.25	33.44
Min	32.27	32.10	33.13
Max	34.53	34.67	33.76
SD	0.86	0.77	0.23
NDF
Mean	40.37	40.25	40.48
Min	39.08	39.30	40.11
Max	41.24	41.68	40.70
SD	0.84	0.78	0.24