By Dr Gary Barker
In 2005 Lawrence Wein and Yifan Liu described the US milk supply as ‘bow-tie-shaped’. It is becoming increasingly clear to researchers from Livestock, Livelihoods and Health that a similar picture is helpful for modelling the food supply chain in Tanzania.
Wein and Liu’s memorable picture encapsulates the US system for collection of milk, from a large number of dairy cows, followed by sequential elements of pooling as the milk is transferred to tanks, to silos and eventually to a very small number of finished goods manufacturers. Subsequently the milk product is dispersed again – at first to a few distributors, then to larger numbers of retailers and finally to very many consumers to complete the ‘bow-tie’ picture as milk moves from farm-to-fork.
Wein and Liu were specifically interested in the vulnerabilities associated with this structure for the milk supply. But their assessment points clearly to the importance of the supply chain topology in any appreciation of food risks – and particularly in understanding complex dependencies that may be important in the appreciation of separate measured values for the level of contamination. (The Wein and Liu paper proved quite controversial, for many reasons, and developed an argument that progressed all the way to the US government.)
It is difficult to be sure about the topology of the meat supply chain in Tanzania but the large number of small producers, and the small number of slaughter slabs and abattoirs, point to a bow-tie structure with the primary markets close to the bottleneck. Some slabs slaughter animals from several markets and some markets supply animals to several slabs, so traceability is easily lost.
As an added complication, particularly for emerging markets, the structural links that connect a food processor with a supplier, or with a customer, are often transient – here today and gone tomorrow. The intermittent changes in relationships are driven by market forces and the unpredictable nature of these business links precludes precise, system-wide knowledge of a food chain structure at any one moment in time.
When supply chains are not fully integrated, the complexities compromise our ability to track and trace food during its passage from producer to consumer. Once it is difficult to trace the food, tracing food contamination is doubly problematic. Tracing zoonotic pathogens within the fluctuating, and often ‘on-demand’ meat supply in Tanzania is therefore a significant scientific challenge.
Uncertainties necessitate a different approach to food chain modelling and food safety management. To make progress it is necessary to assign probabilities not only to the likelihood of the contamination at one point (i.e. to represent uncertain prevalence), but also to the chances that food travels along different paths that connect producers and consumers within the fluctuating supply network.
This probabilistic view can never produce categorical results which, for example, might be obtained from combining lots of pieces of precise, ‘one up, one down’ trace information. However, based on downstream observations and statistical knowledge of food chain pathways, inference is often sufficient to indicate likely sources of contamination and to act as a precursor for more powerful biotracing techniques such as the matching of genotypes. The statistical picture of food distribution can still support risk assessment and, crucially, help speed up tracking and tracing operations in the event of contamination incidents.
In Africa where individual producers of meat are generally small-scale, and are separated from their desired markets by distributed and often transient transport, slaughter and butchery operations, a system-wide view of food safety can only arise from the statistical appreciation of both the prevalence and the network of processor-supplier relationships.
Following food contamination under these conditions is one of the modelling objectives for the Livestock, Livelihoods and Health research programme. In Tanzania a dedicated group of field scientists and social scientists from LLH are involved in gathering data that support this modelling approach.