Improved wet Silage analysis through optimized NIR sampling

Introduction: Fast and reliable silage analysis plays an important role in herd improvement programs as well as feeding economy.

Presently silage is sampled at the farm and brought to central labs. Dry analysis involves taking out approx. 200 grams of wet silage. This samples is dried, milled and analyzed with NIR reflectance with good results as a fairly large and homogeneous sample is analyzed.

For wet silage though traditionally a much smaller, therefore less representative and through the structure, much less homogeneous, amount of silage is observed

The theory of sampling sets forth that good sampling is obtained, when replicate sampling variation and replicate scanning of the same sample shows equal fluctuations. This has historically been a challenge for wet silage analysis.

In this study it has been shown that representative and reliable analysis of dry matter, protein and Starch in fresh, wet silage is possible by new improved sampling technique. The presentation will focus on the importance of proper sampling of heterogeneous products and the positive impact on reproducibility.

Materials and Methods: 200 samples of fresh grass and 300 samples of maize from various European farmers have been used for the pilot trial at Sciantec Labs in the UK. All silage samples were sampled properly at the farm, yielding a master sample of 500 gg. Of this 200 gg is packed in Agritubes and scanned with FT-NIR technology (AgriQuant-B8, Spiral sampling Q-Interline). The scanning used Spiral reflection mode with 250 cm2 observed area. All spectral data was collected with 16 cm-1 resolution and 100 sec. observation time in the spectral range 4.000-10.000 cm-1. (1000-2500 nm)
Reference values were obtained with acknowledged chemical reference methods.
PLS-1 calibrations were developed and validated with an independent data set of 19 samples measured in 2 different packings to verify the reproducibility. The same samples were also scanned at the old analyzer setup involving film wrapped sample and standard models.

Results and Discussion: Calibration development was carried out for several components and the secv values in wet maize for dry matter, protein and starch were 1.3, 0.39 and 2.6 respectively. The SEP values for the validations set were found to be 1.2 for dry matter, 0.45 for protein and 2.4 for starch.

The error arising from repacking for the independent test set for dry matter was 0.67 for the spiral sampler and 1.46 for the old analyzer setup leading to more representative sampling when using the spiral sampler.

It is shown that the improved performance can be directly linked to the large observed area and volume of sample in the spiral sampling step.