by Michelle Ziegler
The era of big data is coming to historic epidemiology. A new study published this month in Scientific Reports took a database of 5559 European outbreak reports (81.9% from UK, France and Germany) between 1347 and 1760 to analyze the role of rivers in the incidence and spread of plague. Their hypothesis was that river trade played a similar role as maritime trade in disseminating the plague but that the correlation would grow weaker over time as movement of goods over land became less expensive. In the 14th century, water transport was approximately ten times cheaper than land transport; the cost ratio diminishes to two to four times as expensive by the 18th century. While it is not surprising that rivers had a role in disseminating the plague, the high correlation Yue, Lee, and Wu (2016) found between not only the proximity of the river but also its size and elevation is striking. Over 95% of the outbreaks occurred within 10 km of a ‘navigable’ river, defined as 5 m or greater in modern width and to differentiate maritime from riverine trade, excluded outbreak sites within 5 km of the maritime coastline. To ensure that rivers were suitable as trade routes, they only included rivers that linked two cities and excluded rivers that flowed into a lake without an outlet.
If we drill down into their results more directly, then we find that 84% of the city centers were less than 1 km from a river with 79.5% of those being on a river at least 20 meters wide. By their calculations, the average river width was 84.6 m. This correlates well with increased traffic and goods following to and through cities on substantial rivers. It is worth nothing here that the specific examples they give in England, Fossdyke, River Great Ouse, and the River Derwent are either canals or fit into an extensive canal system.
Looking at relationships between the outbreak sites and geography also favors high traffic river routes. When they included a “spatial lag in the regression models” they showed that there is a “highly significantly correlation with the spatial lag (p <0.01), indicating that plague outbreaks were spatially dependent upon previous outbreaks in adjacent cities” (Yu, Lee & Wu, p. 2). There was also a negative correlation between elevation and plague incidence, which they attribute to a lack of navigable rivers at higher elevations noting that only 20 incidents were recorded above 1000 m over sea level. They also tested their results with controls for population density and economic status which did not effect their results for the likelihood of plague incidence or the association between outbreaks and river width. This will have to be evaluated by those with more modeling experience than I have.
There are a few caveats. First, such studies are only as good as their database. Yue, Lee and Wu used the digitized database constructed by Büntgen et al (2012) that was itself based on a 35 year old archive published in French. I’ll leave its scrutiny to historians. They also do not address potential biases in all such databases, such as the likelihood that urban sites are recorded at a higher frequency than rural sites or that the political climate can effect the survival of records. Indeed, economic records are likely to note pestilence as a factor effecting commerce. While the environmental destruction of an enduring war could increase plague incidence, the high level of records from the ’30 years war’ needs a historians eye to evaluate. They also note that they are using measurements of modern rivers and canals that may have been significantly different in the past, modified by both natural processes such as silting or flooding and man-made changes such as straightening, dredging, or canal development.
They also assume there were no European reservoirs, which we now know is not true. Ancient DNA studies have indicated that there were at least two strains descended from the Black Death circulating within late medieval Europe (Bos et al, 2016; Spyrou et al, 2016). The European reservoir(s) have not yet been located. However, relatively few of the incidents reported in the database are likely to be actual zoonotic events linked directly to a local sylvatic (wild) reservoir, plus many known reservoirs outside of Europe are found at high elevation (for example in Tibet or Madagascar) and so are unlikely to be in this particular database given the absence of sites at higher elevations. Once a new outbreak emerges from a high elevation reservoir and comes off the mountain so to speak, then its transmissions by rivers is as likely as a strain entering from outside of Europe. On the other hand, if cities or even river networks are the actual reservoir, it would significantly effect their results.
River and canal networks or large river ports could function as reservoirs. River ports are similar to coastal maritime ports in that they have warfs, warehouses and nearby markets that would support large rodent populations. Barge traffic would specialize in transport of grain and other foodstuffs attractive to rodents. Yue, Lee and Wu (2016) state that they did not query their database for the effect of flooding because they could not accurately predict where floods would occur, that flooding is not predictable based solely on river width. Flooding along these river and canal systems is something that needs to be investigated because it would force rodents out of their normal shelter and could be related to human outbreaks (as the plague of 589 in Rome probably was). Floods could also carry infected rodents or fleas downstream on floating debris.
This study is a interesting jumping off point for future work. The database needs to be evaluated by historians and perhaps subdivided into smaller time periods. Division of the database into regional studies would also allow local archaeology and ecology to be more informative on precise outbreaks. I’m looking forward to all of questions big data studies like this one open up!
Bos, K. I., Herbig, A., Sahl, J., Waglechner, N., Fourment, M., Forrest, S. A., et al. (2016). Eighteenth century Yersinia pestis genomes reveal the long-term persistence of an historical plague focus. eLife, 5, 17837. http://doi.org/10.7554/eLife.12994
Spyrou, M. A., Tukhbatova, R. I., Feldman, M., Drath, J., Kacki, S., de Heredia, J. B., et al. (2016). Historical Y. pestis Genomes Reveal the European Black Death as the Source of Ancient and Modern Plague Pandemics. Cell Host and Microbe, 19(6), 874–881. http://doi.org/10.1016/j.chom.2016.05.012