Mapping livestock movements in Sahelian Africa

Abstract: In the dominant livestock systems of Sahelian countries herds have to move across territories. Their mobility is often a source of conflict with farmers in the areas crossed, and helps spread diseases such as Rift Valley Fever. Knowledge of the routes followed by herds is therefore core to guiding the implementation of preventive and control measures for transboundary animal diseases, land use planning and conflict management. However, the lack of quantitative data on livestock movements, together with the high temporal and spatial variability of herd movements, has so far hampered the production of fine resolution maps of animal movements. This paper proposes a general framework for mapping potential paths for livestock movements and identifying areas of high animal passage potential for those movements. The method consists in combining the information contained in livestock mobility networks with landscape connectivity, based on different mobility conductance layers. We illustrate our approach with a livestock mobility network in Senegal and Mauritania in the 2014 dry and wet seasons.

• The paper introduces a novel framework that uses conductance maps and circuit theory to simulate livestock movement paths.
• It integrates mobility network and landscape connectivity data to identify seasonal transit zones and key geographic hubs.
• The study offers actionable insights for disease control, land use planning, and conflict management in the Sahel region.

The paper "Mapping livestock movements in Sahelian Africa" (1910.10476) presents a framework for mapping potential livestock movement paths and identifying high-traffic areas in the Sahel region, specifically focusing on Senegal and Mauritania. This addresses the challenge posed by the lack of quantitative data and the high spatio-temporal variability of herd movements, which are crucial for livelihoods but also contribute to conflicts and disease spread. The methodology integrates livestock mobility network data with landscape connectivity analysis using circuit theory.

Methodology

The core of the methodology involves constructing a conductance map representing the ease of livestock movement across the landscape and then simulating potential flows between known origins and destinations derived from empirical mobility data.

Data Sources

1. Livestock Mobility Network: This network was constructed using sanitary movement permit data (LPS - "Laissez-Passer Sanitaire") collected by veterinary services in Senegal, Gambia, and Mauritania for cattle movements on foot during 2014.
   • Nodes: Represent origin and destination locations (e.g., markets, administrative units).
   • Links: Directed connections indicating movement between nodes, weighted by the number of cattle (volume) reported on the permits. Links were timestamped monthly.
   • Temporal Aggregation: Data was aggregated monthly and analyzed separately for the wet season (June-October) and the dry season (November-May).
2. Landscape Data: Rasterized environmental and infrastructural data at a 500m spatial resolution were used:
   • Land Use/Land Cover (LULC): Derived from FAO data, aggregated into 14 relevant classes.
   • Road Network: Main roads extracted from OpenStreetMap.
   • Administrative Borders: Senegal-Mauritania border from GADM, with specific official crossing points identified via expert consultation.

Conductance Map Construction

A multi-step process was employed to generate seasonal conductance maps where pixel values range from 0 (high resistance/impassable) to 1 (low resistance/high ease of movement):

1. Walking Layer (W): A base layer assigning conductance values to each LULC type based on expert knowledge regarding suitability for cattle movement. Crucially, these values differed between wet and dry seasons to reflect seasonal changes in land use (e.g., croplands assigned high conductance (1) in the dry season post-harvest, but low conductance (0.125) during the wet growing season). Areas deemed impassable (e.g., large water bodies, irrigated croplands during the wet season) were assigned no value. (Refer to Table 1 in the paper for specific values).
2. Road Layer Incorporation (R): The influence of the road network was incorporated. Pixels containing roads were assigned a maximum conductance value of 1. The conductance of pixels not containing roads ($W_{ij}$) was potentially scaled by a parameter $\delta_W$ (reference value 0.8), representing the relative ease of movement through the landscape matrix compared to roads. The conductance $R_{ij}$ at pixel $(i, j)$ is given by:

   $$R_{ij} = \begin{cases} 1 & \text{if pixel } (i,j) \text{ contains a road} \ \delta_W \times W_{ij} & \text{otherwise} \end{cases}$$

   This formulation implies that even the most suitable off-road terrain (W=1) might be slightly less conductive than a road if $\delta_W < 1$.

3. Border Permeability Adjustment (C): The final conductance map $C_{ij}$ accounts for restricted movement across the international border. Pixels located on the border but not at designated official crossing points had their conductance value from the road-incorporated layer ($R_{ij}$) scaled down by a border permeability parameter $\delta_R$ (reference value 0.1). Pixels at official crossing points retained their $R_{ij}$ value.

   $$C_{ij} = \begin{cases} R_{ij} & \text{if pixel } (i,j) \text{ is an official crossing point} \ \delta_R \times R_{ij} & \text{if pixel } (i,j) \text{ is on the border but not a crossing point} \ R_{ij} & \text{otherwise} \end{cases}$$

   This allows for unofficial crossings but makes them significantly less likely (more resistant) than official ones.

Movement Modeling with Circuitscape

The software Circuitscape was used to model potential movement paths. Circuitscape applies principles from electronic circuit theory, treating the landscape as a conductive surface where movement probabilities are analogous to current flow. High conductance corresponds to low resistance.

1. Pairwise Connectivity: For each individual origin-destination link $(O, D)$ present in the empirical mobility network, Circuitscape calculated a potential connectivity map. This map represents the probability density of movement between $O$ and $D$, given the overall landscape conductance map $C$. Essentially, it computes the expected current flow through each pixel when a current source is applied at $O$ and a sink at $D$.
2. Normalization: The output map for each link was normalized to represent probabilities.
3. Weighting by Flow Volume: Each normalized link-specific probability map was weighted by the proportion of the total animal volume associated with that specific link in the empirical network. This ensures that high-volume routes contribute more significantly to the final map.
4. Aggregation: All volume-weighted probability maps were summed pixel-wise across all links in the network. The resulting aggregated map represents the overall potential paths for livestock movements, integrating landscape permeability with the observed network flows. Higher values in this final map indicate areas with a higher likelihood of animal passage considering both landscape constraints and empirical movement patterns.

Identification of High Potential Areas

To identify regions with high overall potential for livestock passage, the values from the final potential path map were summed within administrative boundaries (regions for Senegal, Gambia, Mauritania). These aggregated regional scores were then normalized to facilitate ranking and comparison, providing a measure of the total potential movement activity traversing each administrative unit.

Results and Analysis

The study yielded insights into both the structure of the mobility network and the spatial distribution of potential movement corridors derived from the connectivity analysis.

Mobility Network Characteristics

• Strong Seasonality: Livestock movement volume, measured by the total number of cattle recorded in the LPS data, was markedly higher during the dry season compared to the wet season. A peak in movement occurred in the late dry season (April-June), preceding the onset of rains. Both the number of active links and the total volume decreased significantly during the wet season.
• Flow Concentration: Movement was highly concentrated along a few major routes. The top 10 links accounted for approximately 66% of the total animal volume during the wet season and about 75% during the dry season.
• Geographic Hubs: Significant activity, including substantial transboundary movements, was observed around the Senegal-Mauritania border, particularly involving locations like Podor, Kaedi, Matam, and Mbout. Another active zone was identified in southeastern Mauritania near the Malian border (e.g., Boustaile). Network analysis metrics, specifically betweenness centrality, highlighted Podor as a crucial transit hub, especially during the dry season.

Potential Path Mapping Outcomes

• Corridor Visualization: The Circuitscape approach generated detailed maps (Fig 7 in the paper) illustrating potential movement corridors at the 500m resolution. These corridors clearly reflected landscape constraints, channeling movement through more conductive areas.
• Seasonal Path Differences: The modeled potential paths showed significant differences between the wet and dry seasons. In the wet season, movement was more constrained, often forced along road networks due to the low conductance assigned to cultivated areas and seasonally flooded zones. In the dry season, potential paths spread more widely, utilizing areas like harvested croplands which become more permissible for passage.
• Identification of Transit Zones: A key finding was the identification of areas with high passage potential that were not major origins or destinations in the original LPS network data. For example, regions along the axis connecting Podor, Kaedi, and Matam showed high potential passage values, highlighting their importance as transit zones, a feature less apparent from analyzing node degrees or strengths alone.
• Comparison with Network-Based Ranking: Ranking administrative regions based on the aggregated potential path map values (connectivity approach) yielded substantially different results compared to ranking them based solely on the network strength (total in-flow + out-flow) of nodes located within those regions. The Kendall's tau correlation coefficient ($\tau$) between the two ranking methods was low (approx. 0.3-0.4), indicating that the connectivity approach provides a more spatially explicit and nuanced understanding of movement intensity across the landscape, rather than just focusing on start/end points.
• Robustness Analysis: Sensitivity analyses indicated that the regional rankings derived from the connectivity model were relatively robust to variations in the expert-defined LULC conductance weights and the border permeability parameter ($\delta_R$). However, the rankings showed some sensitivity to the parameter $\delta_W$, which balances the relative importance of the road network versus the general landscape matrix in determining overall conductance.

Practical Implications

The maps of potential livestock movement paths have significant practical applications for stakeholders in the Sahel.

Disease Control

The generated maps identify corridors and potential congregation areas where animal density is likely to be high. By integrating these potential path maps with spatial data on epidemiological risk factors (e.g., vector habitats for Rift Valley Fever, previous outbreak locations), public health and veterinary authorities can delineate zones with elevated risk of disease transmission and introduction. This enables more efficient resource allocation for targeted surveillance, strategic vaccination campaigns, and focused vector control interventions in critical high-traffic or high-risk transit areas.

Land Use Planning

The identification of key movement corridors is crucial for sustainable land use planning that accommodates pastoral mobility. This information can inform the optimal placement of essential infrastructure such as water points, veterinary posts, grazing reserves, and livestock markets to support pastoral systems. Furthermore, it provides a basis for formally designating and securing official transhumance corridors, helping to ensure continued access to resources for herders and potentially reducing degradation associated with uncontrolled movement.

Conflict Management

By overlaying the potential movement maps, particularly during the wet season when agricultural activity is high, with maps of cultivated land, specific areas of potential conflict between herders and farmers can be pinpointed. These are zones where livestock are most likely to traverse or graze on croplands, leading to crop damage and social tension. This spatially explicit information can guide conflict mediation efforts, facilitate negotiations for passage rights or grazing agreements, inform the demarcation of agricultural versus pastoral areas, and prioritize interventions aimed at securing passage corridors in conflict hotspots.

Limitations and Considerations

The authors acknowledge several limitations inherent in the methodology and data:

• Data Representativeness: The LPS data likely underestimates total movement, as it may not capture all movements (especially smaller, local ones or those avoiding official channels) and relies on reported origins/destinations, potentially missing intermediate stops.
• Spatial Resolution: The 500m resolution may not capture very fine-scale paths or barriers.
• Model Simplifications: The model did not explicitly incorporate dynamic factors like the location of temporary water points or forage availability beyond the LULC classification, which are known drivers of movement decisions.
• Expert Weights: The conductance values assigned to LULC types were based on expert opinion, introducing a degree of subjectivity, although sensitivity analysis suggested robustness for regional rankings regarding these weights.

Conclusion

The methodology presented offers a robust and adaptable framework for mapping potential livestock movement corridors in data-scarce environments like the Sahel. By combining empirical network data with landscape connectivity modeling using circuit theory, it generates spatially explicit outputs that go beyond traditional network analysis. These outputs provide valuable, actionable intelligence for improving animal disease surveillance and control, guiding land use planning to support pastoralism, and mitigating conflicts between herders and farmers.