Environmental and economic determinants of temporal dynamics of the ruminant movement network of Senegal

Abstract

Our understanding of the drivers of the temporal dynamics of livestock mobility networks is currently limited, despite their significant implications for the surveillance and control of infectious diseases. We analyzed the effect of time-varying environmental and economic variables-biomass production, rainfall, livestock market prices, and religious calendar on long-distance movements of cattle and small ruminant herds in Senegal in the years 2014 and 2019. We used principal component analysis to explore the variation of the hypothesized explanatory variables in space and time and a generalized additive modelling approach to assess the effect of those variables on the likelihood of herd movement between pairs of administrative units. Contrary to environmental variables, the patterns of variation of market prices show significant differences across locations. The explanatory variables at origin had the highest contribution to the model deviance reduction. Biomass production and rainfall were found to affect the likelihood of herd movement for both species on at least 1 year. Market price at origin had a strong and consistent effect on the departure of small ruminant herds. Our study shows the potential benefits of regular monitoring of market prices for future efforts at forecasting livestock movements and associated sanitary risks.

Figure 1

Results of the principal component analysis performed on the rainfall (left) and biomass production (right) aggregated at month and department level in 2014 and 2019. The top and middle graphs biplots represent the values taken by the departments (arrows) and months (dots numbered from 1 to 12 in chronological order) on the first and second dimensions (the X and Y axis respectively). The bottom graphs represent the value taken by the successive months on the first dimension (which explains most of the data variance) in 2014 (plain line) and 2019 (dashed line).

Figure 2

Results of the principal component analysis performed on the market price of cattle and small ruminants in 2014 and 2019 in a selection of monitored markets. The left hand biplots represent the values taken by the markets (arrows) and months (dots numbered from March to December in chronological order) on the first and second dimensions (the X and Y axis respectively). The color of the arrows corresponds to the cluster the markets were attributed to. The right hand maps represent the geographical location of the markets, with colors corresponding to their attributed cluster. The bottom graphs represent the value taken by the successive months on the first dimension in 2014 (plain line) and 2019 (dashed line) for cattle (left) and small ruminants (right). The plain and dashed vertical red lines correspond to the months of important religious celebrations where cattle and small ruminants are traditionally consumed in 2014 (plain) and 2019 (dashed) (Magal de Touba for cattle, Tabaski for small ruminants). Administrative boundaries were drawn using the GADM database of Global Administrative Areas (www.gadm.org).

Figure 3

Graphical representation of the spatial and temporal distribution of the number of the recorded cattle herd (left) and small ruminant herds (right) movements between departments of Senegal in 2014 (top) and 2019 (bottom). The inter-departments links on which at least one movement was recorded are displayed on the maps with arrows’ width proportional to the logarithm of the number of recorded movements in the year. The evolution of the number of recorded movements through time is displayed in the bar plot, with the number of recorded movements throughout the country and the calendar month in Y and X axis respectively. Important religious celebrations (Magal de Touba for cattle, Tabaski for small ruminants) are indicated with a brown bar. Administrative boundaries were drawn using the GADM database of Global Administrative Areas (www.gadm.org).

Figure 4

Graphical representation of the fitted thin plate spline functions of the effect of the selected variables on the likelihood of bovine herd movement along a defined inter-department link on a defined month in 2014 (A) and 2019 (B). Black lines are spline functions linking the tested variable to the the risk ratio of herd movement, and blue bands correspond to the 95% confidence interval. Blue histograms at the bottom of each graph represent the distribution of the selected variable (the y axis is the count of observations). On the graph displaying the effect of calendar month, the dashed red vertical line corresponds to the Magal de Touba celebration.

Figure 4

Graphical representation of the fitted thin plate spline functions of the effect of the selected variables on the likelihood of bovine herd movement along a defined inter-department link on a defined month in 2014 (A) and 2019 (B). Black lines are spline functions linking the tested variable to the the risk ratio of herd movement, and blue bands correspond to the 95% confidence interval. Blue histograms at the bottom of each graph represent the distribution of the selected variable (the y axis is the count of observations). On the graph displaying the effect of calendar month, the dashed red vertical line corresponds to the Magal de Touba celebration.

Figure 5

Graphical representation of the fitted thin plate spline functions of the effect of the selected variables on the likelihood of small ruminant herd movement along a defined inter-department link on a defined month in 2014 (A) and 2019 (B). Black lines are spline functions linking the variable to the risk ratio of herd movement, and colored bands correspond to the associated 95% confidence interval. Blue histograms at the bottom of each graph represent the distribution of the selected variable (the y axis is the count of observations). On the graph displaying the effect of calendar month, the dashed red vertical line corresponds to the time Tabaski celebration.

Figure 5

Graphical representation of the fitted thin plate spline functions of the effect of the selected variables on the likelihood of small ruminant herd movement along a defined inter-department link on a defined month in 2014 (A) and 2019 (B). Black lines are spline functions linking the variable to the risk ratio of herd movement, and colored bands correspond to the associated 95% confidence interval. Blue histograms at the bottom of each graph represent the distribution of the selected variable (the y axis is the count of observations). On the graph displaying the effect of calendar month, the dashed red vertical line corresponds to the time Tabaski celebration.

Figure 6

Graphical representation of the evolution of the number of registered departures of bovine and small ruminant herds (bars) and average bovine and small ruminant market price (line) in the regions of Tambacounda and Kolda in 2019. The dashed vertical red lines correspond to the months of important religious celebrations where cattle and small ruminants are traditionally consumed (Magal de Touba for cattle, Tabaski for small ruminants).