13 Sep The need for deep understanding of food system data
When working with mathematical modelling, you always need to be on the lookout for discrepancies and gaps. Inaccurate or incomplete data will propagate through to model results. While updating the Sustainable Nutrition Initiative’s DELTA Model®, our team found a potential gap around skim milk. While whole milk, cream and butter were accounted for, it appeared that a proportion of dairy nutrition had disappeared through the processing of skim milk. This was due to the way models like DELTA need to consolidate a wide range of food items to make the data easier to manipulate by the user. Read on to find out how the team tackled and solved this conundrum, giving a better picture of the significant impact skim milk has on global nutrition.
The UN FAO Food Balance Sheets (FBS) are a valuable resource to understand food commodity production and use at both a country and global level, and (when connected with food composition data) the nutrition potentially available to people throughout the world. They provide the basic data that underpins the DELTA Model® and the allocation of primary commodity production for uses as food, feed, or other purposes.
Where food commodities are processed to form another item that also appears in the FBS, this is identified in the Processed element of the balances. Where an item is processed into other food items not included in the balances then the uses of the subsequent items are incorporated in the parent item as primary commodity equivalent amounts allocated to the remaining elements (New FBS methodology (fao.org)).
Overall, this system works well, but it breaks down when the processing of a commodity results in two food products, only one of which is present as an item in the Food Balances. The end uses of the non-FBS item(s) essentially disappear from the FBS. This does not affect the nutritional measures contained within the FBS itself, as the food use of these products is rolled up into the parent item values for protein, fat, and energy supply. However, in models that only use the mass flows from the FBS – such as the DELTA Model® – the contribution to human nutrition of these secondary products is lost.
The biggest challenge of this type occurs in the processing of milk into either Butter/Ghee or Cream (which are FBS Items), which generates a by-product of Skim and/or Buttermilk (which are not in the FBS). Thus, the food and nutritional value of Skim and/or Buttermilk are not accounted for when relying solely on FBS mass flows. A similar challenge may exist with the end use of cakes left over from extracting oils from the various oil-bearing crops, including soybeans.
Accounting for skim milk
To address the skim milk challenge for version 2.1 of the DELTA Model®, we used information in the Supply and Utilisation Accounts (SUA) to generate an extra FBS Item to capture the end uses of the Skim and Buttermilk. In 2020, the FBS reports that approximately 264 million tonnes (Mt) of milk were processed to produce 3.7 Mt of cream and 12.5 Mt of butter. The co-product of this processing is approximately 248 Mt of combined skim milk and buttermilk. Put another way, every 100 tonnes of milk identified as processed in the FBS results in 1.4 tonnes of cream, 4.75 tonnes of butter and ghee, and 93.85 tonnes of skim and buttermilk. The amount of nutrition represented by the skim and buttermilk can be estimated by multiplying by the typical protein content of bovine skim milk (~3.3%), resulting in over 7.5 Mt of protein. This is approximately 4% of total food protein consumption in 2020.
From the more comprehensive data in the SUA, there are four items that represent skim milk and buttermilk:
- 137.7 Mt of Skim milk of cows
- 98.8 Mt of Skim milk of buffalo
- 11.18 Mt of Buttermilk
- 1.2 Mt of Skim sheep milk.
For food system modelling applications – such as the DELTA Model® – we also need to know how these production volumes are used. To do this we need to look at the end use of each of these products, and any other products that result from further processing of these. The processing element is important as a large portion (61%) of skim milk from cows undergoes further processing, which represents over a third of all skim milk produced globally.
The FAO commodity tree (Figure 1) and SUA identifies a range of products that are derived from skim milk, along with the effective yield of this production. By reversing the yield, we can work backwards from the reported end use of these commodities to determine the equivalent amount of skim milk these uses represent and combine this with the direct uses of skim milk to produce an effective allocation table. For example, if the yield of skim milk powder is 0.1 tonnes per tonne of skim, then every tonne of powder used for food is the equivalent of 10 tonnes of skim. This is a similar process to how the whole of the FBS is calculated. Completing this process gives a final skim milk allocation as follows:
- 64.9% is used for food
- 30.5% is used for feed
- 3.9% has industrial uses – mainly industrial use of casein
- 0.7% is further processed.
Global Nutrient Contribution of Skim
So why does this matter? As indicated earlier, the volume of skim produced has the potential to provide up to 4% of global protein consumption. Having considered the end uses of skim and its various derivative products, and created an extra Food Item within the DELTA Model®, we find that skim milk does indeed play a significant role in global nutrition, with key nutrients supplied by skim milk including:
- 9.7% of Calcium (making skim the #3 source of this nutrient in the DELTA Model®)
- 6.3% of Vitamin B12 (#5)
- 5.6% of Riboflavin (#10)
- 2.45% of Protein (#12)
This substantial contribution to human nutrition could easily have been missed without a detailed examination of the data underlying our modelling. This would have resulted in a significant loss from the modelled nutrient flows to populations and could be even more significant in model scenarios designed by our users. Our solution, shared with the FAO team, prevents this issue and makes sure those nutrients are accounted for. At SNi, we endeavour to make food system data more accessible, and that starts with a detailed understanding of the data.
This Thought for Food was written by Dr Andrew Fletcher, Jade Rivers, and Dr Nick Smith, with the support of the SNi team.
