Effect of long term storage of grains on NIR spectra

ianm's picture

11. Effect of long term storage of grains on NIR-spectra.

From: Hyesun Park

Is there any studies on the effect of long term storage (1-5 years) of whole grain, such as corn, soybeans and wheat, on spectra? I am particularly interested in developing calibration for oil, starch, protein, fiber, moisture and amino acids. To develop robust calibrations for grain samples, it is essential for calibration set to have year variability (mainly due to environmental effects). From our experience, at least 4-5 consecutive years of crops are needed for stable calibrations. Every year, freshly harvested crops are added to the existing calibration pool. A problem rises when new NIR-units arrive. Samples from 1-5 crop years ago are scanned, thus I assume, introducing storage effect into calibration, although samples have been kept in a cooler to minimize any biological changes over time. I really do not want introduce long term storage effect into the models especially when it becomes very influential. But it is hard to quantify to what degree it would affect spectra. Is there any (rough) guide lines for the optimum length of storage of whole grain for NIR-applications?

Hyesun Park

From: Phil Williams [email protected]

Subject: RE: effect of long term storage of grains on NIR-spectra

Grain (at least wheat) does change with time, but most of the changes take place during the first few months. We have studied the functionality of wheat that has been in elevator storage for several years and the physical dough and baking characteristics have remained well-preserved.

One of the difficulties in detecting differences in the spectra over time is whether the instrument has changed over the same period.

More to come, but extremely busy now, during our harvest period.


From: "Gallaher, Ken" [email protected]

Try using analyzers to which calibrations can be directly transfered without having to calibrate each one seperately.

The only laboratory analyzer I know that can do that is the Bomem FTNIR system. The up front capital cost is proably higher than what you are using - but if one includes recalibraion cost and the risk of sample aging - then it is another matter.

Certainly one way to test for aging is just to keep rerunning the same sample set on the same instrument and see what happens to predictions.