Millers and bakers are aware that wheat, being a naturally produced raw material, exhibits variation in its protein content. This is of significant importance for millers who are trying to meet the demands of their bakery customers by delivering them a consistent product with little variation in terms of its overall baking performance. Bakers, also with customers whose expectations are for a consistent product, are looking for a flour that varies little in terms of its overall mixing performance and bake- ability. Protein content is an important indicator of bread quality and can have a considerable impact on a flour’s performance in the bakery.

It is commonly understood that many factors contribute to broad variations in wheat protein content including the geology and soil characteristics of a region, the climate of a particular region (dry vs wet) and year on year seasonal variations (untimely rain at harvest, insufficient snow fall in winter, etc). Even field to field one can find differences based on the production practices employed: the crop rotation, level of fertilizer application, dry-land vs irrigated, etc. The uptake of protein by the wheat plant and it accumulation in the kernels is a particularly complicated process and with many factors contributing to the overall content at harvest.

What, perhaps, is not as well recognized is what could be called ‘micro-variations’ in protein content: What exactly does the protein distribution throughout a single field look like? How about variability between various plots in that field, plant to plant variation, and variation within a single plant, etc.

Nowadays the ‘bulk’ measurement of protein content of a particular wheat sample is performed at a mill or grain receiving location using near-infrared spectroscopy (NIR). In this method the grain to be tested is placed in the instrument and near-infrared radiation between the wavelengths of a 400nm – 1700 nm (you can’t see this ‘light’) is directed at the sample with the instrument measuring the amount of radiation absorbed by the sample. This amount is compared against a calibration and the protein content is indicated on the display on the instrument. It sounds complicated but this method is so easy to perform and remarkably accurate that it is the industry standard for protein content measurement. NIR is also commonly used to measure ash content among other grain quality parameters.

But this is a bulk measurement of protein content on a sample of wheat that has been harvested from many fields in a region (possibly multiple regions) and while it is a good indicator of the protein content of the milled flour that will come from this wheat lot it does little to indicate the ‘micro’ protein variation that was mentioned earlier.

In 2000 a group of experiments were conducted to attempt to map the protein variation as it existed in a group of fields in southwest Kansas, USA (a prime growing region for hard red winter wheat). So, while NIR technology is used for bulk measurements of protein it can also be used on individual kernels using an instrument called the Perten SKCS. With this method a single kernel is placed in the instrument and near-infrared radiation is directed at the kernel and the protein content is determined just as with the bulk samples.

To obtain the kernels for the variation mapping project four varieties of wheat (Jaeger, 2137, Tam 107 and Ike) 47 fields were sampled and each kernel was uniquely identified based on its field, plot within the field, row within the plot, plant within the row, and head within the plant. On the head kernels were removed from locations identified as ‘top’, ‘middle’ and ‘bottom’. In all over 10,000 kernels were sampled, identified and their protein content determined.

The results of this protein measurement was subjected to statistical analysis from which was determined the ‘variance estimate’ of protein variation. A ‘variance estimate’ is basically how much variation one would expect to find overall based on the variation found within a particular sample set. Said another way, based on how much variation we found in our kernels we could expect to find the same variation throughout, perhaps, all of southwest Kansas or even hard red winter wheat grown under similar circumstances throughout all of the US. From these variance estimates a ‘map’ of the levels of protein variation was produced.

The distribution of ‘micro’ protein variation was particularly surprising. If you look at the graph to the right you will see that, as one would expect, a good deal of the variation was at the ‘field to field’ level: different farmers, different production practices would tend to indicate that protein is going to vary.

What came as a real surprise is the increase in protein variation found within a single field. Typically in this type of analysis one finds decreasing variance estimates as they move from the large scale level to the small scale (ie field to field down to kernels within a single plant). The fact that there is actually an increase in protein variability within a single field would indicate that protein uptake by the plants is not uniformly accomplished by all plants at the same rate even within the relatively small area of a single field and that the range of protein content found in a single field can, in fact, be quite wide.

From the peak at the ‘plots within a field’ level we see a sharp drop in the variability of protein at the ‘rows within a plot’ level, and then an increase until we reach a second peak at the ‘heads on a plant’ level. Again, another surprise. This indicates that protein within a single plant is not evenly distributed throughout all the heads on the plant and, in fact, can vary more than protein content between individual plants. A single wheat plant might have 6-8 heads with each head containing many kernels.

So, while the graph is illustrative, what are the actual levels of protein variation. Given these samples the variance estimate for the variety 2137 at that ‘field to field’ level is almost 2.5% points of protein. Therefore one field might have an average of 12% protein whereas another might come it at 14.5% (and another as low as 9.5%). For 2137 at the ‘Plot to Plot’ level the variance is even higher at just over 3% and heads on a plant are estimated to vary as much as 1% point of protein! That’s a lot of variability. If you are using a high quality flour you aren’t seeing that level of variation in the protein content of your flour. But what we all should recognize from this study is that while wheat fields, visually, appear to quite uniform, there is a tremendous amount of protein variability occurring, even within very small areas.

Admittedly, this is an academic exercise from the standpoint of the baker. Given the grain production and marketing systems in place throughout the world wheat is blended many times from harvest until it reaches the mill and is turned into flour. This allows for almost complete ‘smoothing’ of this ‘micro-variability’. On the other hand, it should give some indication that providing a consistent flour in terms of not only protein content but also concerning all the other baking variables is a challenging task. What this research also indicates though is that there may be value in segregating wheat from various locations in order to capture some beneficial characteristics that are occurring at the field (or smaller) level but are being lost once that grain is blended.

If you are interested in reading the full article that this research was published in please use the contact link on the left and send a message indicating your interest. I copy will be emailed shortly.