Network analysis of regional livestock trade in West Africa.

In West Africa, long and complex livestock value chains connect producers mostly in the Sahel with consumption basins in urban areas and the coast. Regional livestock trade is highly informal and, despite recent efforts to understand animal movement patterns in the region, remains largely unrecorded. Using CILSS' database on intraregional livestock trade, we built yearly and overall weighted networks of animal movements between markets. We mapped and characterized the trade networks, identified market communities, key markets and their roles. Additionally, we compared the observed network properties with null-model generated ensembles. Most movements corresponded to cattle, were made by vehicle, and originated in Burkina Faso. We found that live animals in the central and eastern trade basins flow through well-defined, long distance trade corridors where markets tend to trade in a disassortive way with others in their proximity. Modularity-based communities indicated that both national and cross-border trade groups exist. The network's degree and link distributions followed a log-normal or a power-law distribution, and key markets located primarily in urban centers and near borders serve as hubs that give peripheral markets access to the regional network. The null model ensembles could not reproduce the observed higher-level properties, particularly the propinquity and highly negative assortativity, suggesting that other possibly spatial factors shape the structure of regional live animal trade. Our findings support eliminating cross-border impediments and improving the condition of the regional road network, which limit intraregional trade of and contribute to the high prices of food products in West Africa. Although with limitations, our study sheds light on the abstruse structure of regional livestock trade, and the role of trade communities and markets in West Africa.

Introduction

Livestock production involves at least 20 million people across West Africa [1], where long market chains connect producers in the Sahel with consumption basins in urban areas and the coast [2-5]. Livestock production, marketing, and processing generate income for actors along the value chain [1,4] and provide food and nutrition security in the region. Intraregional livestock trade is highly informal; its true magnitude is not captured in official statistics and therefore unknown [6,7]. Livestock (mostly cattle and small ruminants) are traded live and lead the intraregional food trade [8,9].

Despite significant tariff and non-tariff barriers to trade [2,4] and sizeable poultry and beef imports, West African intra-regional livestock flows have increased in the past two decades. The existing gap between production and demand of animal products [3,4,7] is expected to widen in the next decades [10,11] driven by population and income growth, migration and urbanization [3,4], suggesting that there is potential for intraregional livestock trade to satisfy the growing demand for animal products [3]. Moreover, the percentage of intraregional trade in West Africa is considered low when compared to other regions of the world. Because of this, the livestock sector has been recognized as a major agricultural trade opportunity in West Africa, and particularly for the Sahelian countries of Niger, Burkina Faso and Mali [12-14].

The livestock marketing system can be studied as a network of actors or locations connected by animal shipments. Livestock markets are spread throughout the region forming a web with assembling and bifurcating connections through which animals are bought in from rural areas, shipped to wholesalers and then sold in shortage areas, often across borders. Network analysis methods can help gain insights into livestock market structure and functioning, and guide policymakers in designing agricultural development and trade policies. However, network analysis relies on movement data that is often not collected or not available in developing countries.

In West Africa, livestock trade networks appear to be dominated by socially embedded business transactions, generating contrasting opinions on the market system’s ability to respond to future increases in demand for livestock-derived products. On the one hand, the market structure is thought to constrain agricultural growth [2,15]; on the other hand, it confers the system with resilience in an environment where disruptions and shocks to the market system–like climatic events and conflict [16]–are frequent [3]. Few attempts to formally describe the regional trade structure have been made to support either claim [17], partly because comprehensive data are lacking on livestock movements, production (allowing inferences on sales made), sales (allowing inferences on domestic vs. regional sales) and exports.

Walther [17-19] used Social Network Analysis (SNA) to study cross-border trade and policy networks in West Africa, but excluded livestock. Three different livestock-specific initiatives have studied animal movements as networks in the region. Dean et al. (2013) studied the risk of disease transmission through cattle trade in Togo. Motta et al. [20], on the other hand, studied the cattle trade network in Cameroon and discussed the implications of the network structure for regional disease spread. More recently, Apolloni et al. [21] and Nicolas et al. [22] mapped, characterized and attempted to predict livestock movements in Mauritania. All these efforts highlighted the importance of international livestock trade; however, all three used data collected in single countries. Given the high percentage of movements and animals that cross borders according to these studies, an understanding of regional patterns of trade is essential to better prepare and respond to disruptions like climatic events (drought), market closures or disease outbreaks that might trigger or worsen famine [16,23]. Although some understanding of animal mobility in the region exists, animal movements originating in the top three exporters (Niger, Mali and Burkina Faso) have not been analyzed as a network.

Our objective is to advance the formal study of regional trade networks in West Africa. We use network analysis to characterize the network of live animal trade (cattle, sheep, goats and donkeys) originating primarily in Burkina Faso, the third largest exporter of live animals in the region [9] as they are shipped through the central and eastern trade basins. We use survey data compiled by the Permanent Interstate Committee for Drought Control in the Sahel (Comité Inter-Etats de Lutte contre la Sécheresse au Sahel, abbreviated as CILSS). The CILSS database covers significantly more livestock trade than official figures, contains records of multiple years (2013–2017) and includes movements from selected corridors between 10 West African countries. We first provide a (1) brief descriptive summary of the movements by type of movement (national or international), livestock and transport. We then (2) map and characterize the market network using network statistics and (3) identify market communities. Lastly, we (4) identify key markets and their roles in the regional network. Our findings are contextualized, and their policy implications are discussed.

Materials and methods

Study area

Data used in our study were collected in selected markets and border exit points on livestock corridors throughout West Africa mapped in Fig 1. West Africa extends over 7.9 million km2, occupying approximately one-quarter of the African continent [24]. The region is comprised of 16 countries where hundreds of languages are spoken, 35 of which are spoken by more than 1 million inhabitants [25]. West Africa spans over diverse agroecological zones (arid, semi-arid, sub-humid, humid) and production systems [25], and over largely horizontal stripes of increasing rainfall (<100 to >1600mm) and wet season duration (2–10 months) from north to south [24]. Most of the region experiences unimodal rainfall that allows one growing season from April/June to September/October. Both the Sahel and the Guinean Coast have experienced drier conditions since the late 1960s [26,27] which, combined with severe recurring drought periods, have affected agricultural production. The agriculture sector concentrates more than half of each country’s GDP, while the livestock sector contributes 5–44% of the agricultural GDP [4], carrying greater importance in Sahelian countries where most of the region’s 220 million ruminants are located [25,28]. In contrast, there are 362 million humans in West Africa (est. 2016)– 46% residing in cities [29]–the majority of which concentrate in coastal countries [25].

Fig 1

Map of survey, origin and destination markets recorded in the livestock trade database from 2013–2017.

Diamonds indicate livestock markets where data was collected or survey markets; squares highlight reported origin and destination markets. The map was made in R with the sf package [30].

Data source and pre-processing

Data were collected between January 2013 and August 2017 by the CILSS and regional partner organizations to track the direction and magnitude of intraregional trade. The data collection methodology is described in S1 File. Survey locations include markets and border crossings in selected trade corridors connecting Benin, Burkina Faso, Côte d’Ivoire, Ghana, Mali, Niger, Nigeria, Senegal and Togo. The “markets” included in our study, mapped in Fig 1, are defined as geographical locations that were either (i) survey points or (ii) reported as an origins or destinations of animal movements.

Although it does not capture all intraregional livestock trade, Josserand [6] estimated that CILSS’s livestock trade database captured significantly more trade than official figures, especially for animals originating primarily in Burkina Faso and Mali, two of the three biggest livestock exporters in the region (after Niger) [9]. However, the subset of movements used in our analysis primarily captures movements departing Burkina Faso.

Collected information included the date of each movement, the type of livestock (goat, sheep, cattle or donkey), volumes, origin and destination market and country, and type of transport (on the hoof, by truck or train). These variables are defined in Table 1.

Table 1

Variables from CILSS’ database on intraregional livestock trade used in our analysis.

Name	Variablea	Definition	Format
Loading point	from (name)	Origin location of the livestock movement	Code
Unloading point	to (name)	Destination location of the livestock movement	Code
Origin country b c	fromc (country/cid)	Country where the origin market is located	Name/Code
Destination countryb c	toc (country/cid)	Country where the destination market is located	Name/Code
Date	dmonth	Month the movement was recorded	Month
Type	type	Type of livestock (goat, sheep, cattle or donkey)	Category
Heads	weight	Scaled number of animals being transported	Number
Type of transport	transp	Type of transport used (vehicle, on foot or by train)	Category
Type of movementc	intl	Indicates if the movement crossed borders	Category

aThe “variable” column contains the names of the variables provided in S2 and S3 Files. The variable names in the node list are provided in parentheses.

bCountries where the origin/destination markets were located were used. In some cases, these differed from the countries recorded in the database.

cComputed by authors

Data were pre-processed in Microsoft Excel and R Statistical Software [31] to homogenize and code the variable values. In the database, each entry represents a livestock movement from a loading point to an unloading point, passing through a data collection point (or origin, destination and survey markets, respectively). We separated each entry into two movements: one from the origin market to the survey market, and one from the survey market to destination market. To avoid self-loops in the network, entries were only separated if their origin or destination markets differed from the data collection point. Some of the movements were missing either the origin, destination or both markets, while other shipments could not be geolocated. Excluding incomplete and unmapped entries resulted in 42,252 movements.

Because of the geographic scale at which the movements were recorded, the shipment weights have been scaled to protect the identity of individual traders and/or shipments. This scaling does not affect the construction or analysis of the network. Data comprising complete mapped entries are provided in S2 and S3 Files as an edge list and its corresponding node list. The data processing pipeline is shown in Fig 2.

Fig 2

Data pre-processing pipeline.

Processing steps are shown in grey and data in white. Pre-processing included separating movements into two links, excluding self-loops and incomplete and unmapped entries. After pre-processing, a working dataset of 42,252 movements remained (shown in black).

Network construction

The network was constructed by representing markets and shipments between markets as nodes and directed links, respectively. Links were weighted by the (scaled) number of animals involved in each movement, or by the number of movements between the markets. The movements were aggregated over each year (2013–2017), resulting in 5 directed static networks with weighted links. Additionally, we constructed one overall network (2013–2017) to study its degree distribution.

Descriptive summary

We provide a descriptive summary of the data. This serves two purposes: describing the working dataset and allowing for comparisons with previous work. The descriptive summary includes disaggregation of number of movements by type of movement (national or international), livestock (cattle, goat, sheep or donkey), and transport (vehicle or on foot). Additionally, we summarize the shipments by their origin and destination countries.

Network analysis

We characterized the size, connectivity, heterogeneity and centrality of the trade network using the metrics shown in Table 2. Network-level metrics included the number of data collection points, markets, shipments and pairs of trading markets, the network diameter, link density, average link degree, average shipments, transitivity, average path length, the edge, betweenness and closeness centralization, transitivity, and propinquity.

Table 2

SNA metrics used to characterize the livestock trade network.

Type of metric	Network	Subset	Node
Size	Survey markets		Volume (weight)
	Markets
	Shipments
	Pairs of trading markets
	Diameter
Connectivity/Cohesiveness	Link Density	GSCC
	Average Link degree	GWCC
	Average shipments
	Transitivity
	Average Path Length
Heterogeneity	Propinquity	Trade communities
	Degree distribution
	Degree assortativity
Centrality (unweighted)	In degree		Degree
	Out degree		In degree
	Betweenness		Out degree
	Closeness

We expect animals to flow from their origin through local markets of increasing importance before reaching a small number of regional trade hubs and subsequently, consumers. Thus, we hypothesize that the geographic distances between pairs of locations that don’t trade are larger than the distances between those that do. To test our hypothesis, we ran a one-sided Mann-Whitney test between the Haversine distances separating markets that did trade animals and the distances between markets that did not (H0 = The location shift between the distances separating markets that did not trade and those that did equals zero; H1 = The location shift between the distances is greater than zero). The Haversine distances between each pair of markets in the yearly networks were calculated using their geographic coordinates with the geosphere R package [32]. A p-value of the one-sided Mann-Whitney test is reported per year.

It is common practice to compare empirical network to a null model ensemble to distinguish important network properties from trivial ones, and to enable comparisons with other networks [33]. Here, we follow the process described in [34] to normalize the metrics of the observed networks. First, we generated 5 configuration model ensembles (one for each year) without self-loops or multiple edges and with 10l re-wiring iterations. Then, network-level metrics were calculated for the 1000 simulated networks in each ensemble. Using the configuration model implies that some properties will not vary between the observed and simulated networks; thus, the number of survey markets, markets, shipments and links (total and average), the link density, and degree and closeness centralities were not computed, while the minimum ensemble p-values of the one-sided Wilcox test are reported for the propinquity. Finally, we compared how the observed properties deviated from the ensemble by calculating a z score for each observed metric per year for the rest of the statistics.

Additionally, we tested if the degree distribution of the market network aggregated over the study period (2013–2017) follows a power law or a log-normal distribution following Clauset et al. [35] using (a) the number of movements and (b) the number of neighbors as degrees. At the subset level, we calculated the size of the Giant Strongly and Weakly Connected Components (GSCC and GWCC), and detected trade communities with the fast greedy modularity optimization algorithm introduced in [36]. Key nodes were identified with their trade volume and degree (all, in, out) [37]. Market communities and market level metrics were calculated for the most recent year only (2017). The networks were constructed, analyzed, simulated and visualized with the igraph package [38] in R [31]. Additional R packages used to map or analyze the animal movements are specified under each Fig or Table. All the statistics are defined in Table 3.

Table 3

Network metric definitions in the context of livestock trade.

Metric	Definitiona
Survey markets	Number of markets where data collection took place.
Markets	Number of markets (nodes) that were origins or destinations of livestock shipments (n).
Shipments	Number of shipments between markets (m); includes all individual shipments made between all pairs of markets and is different from pairs of trading markets (l).
Links (Pairs of trading markets)	Pairs of markets that traded at least one animal (l); directed link (e.g. shipment from market A→B is different than from B→A).
Diameter	The longest geodesic distance between any pair of livestock markets in the network using the shortest possible walk from one market to another [39]; calculated considering (directed) and neglecting (undirected) link directions.
Link Density	Ratio of links (l) among livestock markets (n) in the network with respect to the maximum possible number of links (2n(n-1)); defined as l/2n(n-1)) [39].
Average link degree	Average number of markets that each market traded with; defined as l/n.
Average shipments	Average number of shipments each market is involved in; defined as m/n.
Transitivity (clustering coefficient)	If we define the neighbors of a specific market as the other markets who are directly linked to it, the clustering coefficient measures the proportion of neighbors of a specific market that are linked to each other (at the node level), or the average of these local clustering coefficients (at the network level) [39,40].
Average path length	The geodesic (shortest path) between two livestock markets averaged over all pairs of livestock markets in the network; defined as 1/n(n-1) Σi≠jd(vi, vj) where d(vi, vj) is the geodesic path between markets i and j.
Propinquity	The tendency of trading markets to be closer than markets that don’t trade. Measured with the p-value of a one-sided Mann-Whitney test between two groups of geographic distances: the distances between pairs of markets that are linked and those between pairs markets that are not linked (or between pairs of markets that traded and those that didn’t trade).
Degree distribution	Probability distribution of the number of neighbors of each market over the whole network and study period.
Degree assortativity	Correlation between the degrees of linked markets, quantifying the tendency of markets to connect with other similar markets in terms of degree centrality (or number of neighbors).
Degree centrality; centralization	Number of markets a specific market is connected to; standardized mean difference between degree centrality of the most central market and the rest of the markets [37] (in- and out-degree refers to the number of markets that ship livestock to a market of interest, and the number of markets that the market of interest sends livestock to, respectively [39]).
Closeness centrality; centralization	Number of markets a specific market is connected to; standardized mean difference between closeness centrality of the most central market and the rest of the markets [37].
Betweenness centrality; centralization	The frequency a market lies in the shortest path between pairs of markets in the network [37]; Standardized mean difference between betweenness centrality of the most central market and the rest of the markets [39]
Giant Strong Connected Component (GSCC)	Maximum connected subset of markets in the network in which all pairs of markets are linked, considering the direction of the links [39].
Giant Weak Connected Component (GWCC)	Maximum connected subset of markets in the network in which all pairs of markets are linked, neglecting the direction of the links [39].
Trade communities	Market community configuration that maximizes the modularity Q or the difference between the links running within communities and those expected by chance [41]. Calculated with the fast greedy algorithm as introduced by [36].
Volume	Livestock volume received or sent by a market.

aMetric sources are cited, definitions adapted to livestock networks were partly drawn from Dubé et al. [42].

Results

Descriptive data summary

Type of movement, livestock and transport

Close to two thirds of the livestock movements were international, while the rest did not cross borders. Table 4 shows that most movements corresponded to cattle, followed by sheep and goats with an insignificant number of donkey shipments. Movement peaks, shown in Fig 3, occurred in the months preceding Tabaski (or Eid al-Adha) some of the years, particularly for sheep. Tabaski, also known as the “festival of the sacrifice” is a worldwide Islamic holiday that affects animal mobility patterns in West Africa because many Muslim families sacrifice sheep to commemorate it. This festival is observed on the lunar calendar and is celebrated on a different Gregorian date (about two weeks earlier) each year. Vehicles (train and truck) were the main form of transport for all livestock types and seasons, as S1 Table and S3 Fig show.

Table 4

Summary of livestock movements for 2013–2017 by type of movement, transport and livestock in absolute and relative quantities.

	Number of movements	Percentage of movements
Total	42,252	100%
International	24,974	58%
National	17,278	42%
On the hoof	1,870	4%
Train	498	1%
Vehicle	39,884	94%
Cattle	31,080	74%
Donkey	1	<1%
Goat	2,546	6%
Sheep	8,625	20%

Fig 3

Number of livestock movements by livestock type and month 2013–2017 for our working dataset.

Sheep movements increase significantly in the months preceding Tabaski, while cattle and other livestock shipments seem to be unchanged by its occurrence. Vertical black dashed lines indicate the occurrence of Tabaski each year. Major ticks (labeled) correspond to the start of each year, whereas minor ticks are quarters (3-month periods).

Origin and destination countries

Fig 4 shows the origin and destination countries of the shipments. Most movements originated in Burkina Faso, Côte d’Ivoire, and Ghana, with smaller proportions departing Benin, Niger and Mali. Close to a quarter of shipments were destined to Ghana, Côte d’Ivoire and Benin each, whereas Nigeria was reported as the destination country for close to a twentieth of all transfers. Less than 5% of the shipments were leaving to Niger, Mali, Guinea and Senegal combined.

Fig 4

Proportion of livestock movements by origin and destination country.

The percentage of all movements that originated and were destined to each country are shown in purple and yellow, respectively.

Network level: Well-defined long-distance corridors

Links for each year are mapped in Fig 5 by type of movement (national or cross-border). A line connecting two markets (squares) indicates that at least one animal movement was recorded between them in that year. There are clear differences between the years, yet some movement patterns persist. For example, in 2014 and 2017 there were more pairs of trading markets than other years, as well as more international connections. In 2017, there were Nigerien shipments towards the Nigerian border that were not captured in other years. On the other hand, a general north-south direction of animal movements persists all years. Repeating national patterns also include movements from North-Western Burkina Faso towards Benin and Togo, shipments from central Côte d’Ivoire towards markets in the south and west, and national flows from Northern to Southern Ghana. International movements that persist were made from Burkina Faso into the Ivorian and Ghanaian coasts.

Fig 5

Livestock shipments by year.

The color of the link indicates if the movement crossed borders (purple) or not (yellow), while black squares mark the origin/destination markets. The maps were made in R by cropping OCHA ROWCA’s administrative level-0 boundaries for West and Central Africa. OCHA ROWCA’s maps are protected under the CC BY license (https://creativecommons.org/licenses/by/4.0/legalcode).

Table 5 shows temporal changes in the data collection during the study period. For instance, the number of survey markets fluctuated reaching their maximum in 2017. The number of shipments peaked in 2014, even when the duration of data collection is considered (full years vs. Jan-Aug for 2017). However, more pairs of trading markets were recorded for 2017. On average, each market traded with 1–2 other markets every 4–10 days each year. Less than 12% of every market’s neighbors were connected to each other, and the shortest distance between any two markets ranged between 1–3 links. The propinquity p-values in Table 5 suggests that markets that traded were significantly closer to each other than those that did not for all years except 2014. On the other hand, a negative correlation between the degrees (disassortativity) of markets that traded was found for all years, indicating that markets tend to trade with markets of dissimilar centrality. The centrality values show that some markets trade with more locations than most markets in the network, or that there are hub markets.

Table 5

Network level metrics for the livestock market network by year.

Year	2013	2014	2015	2016	2017a
Survey markets	25	30	23	33	41
Markets	112	136	108	122	183
Shipments	5997	11872	10098	7914	6371
Pairs of trading markets	154	286	138	146	298
Diameter (directed)	3	4	5	4	8
Diameter (undirected)	7	7	8	7	8
Link Density	1.2%	1.6%	1.2%	1.0%	0.9%
Average Link Degree	1.4	2.1	1.3	1.2	1.6
Average Shipments	53.5	87.3	93.5	64.9	34.8
Transitivity (clustering coefficient)	4.5%	11.6%	2.5%	2.7%	7.4%
Average Path Length	1.6	1.9	1.4	1.7	3.3
Propinquity (p-value)	7.58E-05***	0.17ns	2.94E-05***	4.06E-14***	3.69E-60***
Degree Assortativity	-59.9%	-60.9%	-64.0%	-54.8%	-55.2%
Degree centralization	17.8%	20.4%	21.9%	16.5%	10.7%
In-degree centralization	35.7%	41.4%	43.7%	32.1%	16.1%
Out-degree centralization	13.2%	14.0%	13.8%	14.7%	10.6%
Betweenness centralization	0.1%	0.3%	0.3%	0.4%	0.5%
Closeness centralizationb	0.6%	2.2%	0.4%	0.6%	3.1%

a Data for 2017 include movements from January-August

bNot well-defined for disconnected graphs

*** p-value<0.001

ns Not significant

The z scores of the observed metric values are reported in Table 6. For all years, the degree assortativity and betweenness centrality scores were negative, while the undirected diameter scores were positive. The observed assortativities, however, were 9 or more standard deviations lower than the mean ensemble assortativities. The remaining metrics (directed diameter, transitivity average path length) either had smaller absolute z scores or had both positive and negative scores. Of these, the directed diameter and average path length reached their maximum normalized score in 2017. For all years, the propinquity of the simulated networks was insignificant as the minimum ensemble p-values were all close to 1.

Table 6

Z scores of the observed metric values and propinquity significance of the ensembles.

Observed metrics were normalized using the average and standard deviations of the corresponding ensemble values for each year.

Year	2013	2014	2015	2016	2017
Diameter (directed)	-1.7	-0.8	0.2	-0.8	2.8
Diameter (undirected)	1.3	1.3	2.8	1.3	2.8
Transitivity	-2.4	4.7	-4.4	-4.2	0.6
Average path length	-0.7	-0.1	-1.2	-0.6	2.8
Degree Assortativity	-11.6	-12.1	-13.7	-9.0	-9.2
Betweenness centralization	-1.0	-0.8	-0.8	-0.7	-0.7
Propinquity (min p-value)a	0.99 ns	0.99 ns	1.00 ns	1.00 ns	0.97 ns

aThe minimum ensemble propinquity p-value is reported for each year

ns Not significant

We found that the degree distribution of the network follows either a power-law or a log normal distribution. For the movement distribution shown in Fig 6A, the hypothesis p-value ruled out a log-normal as a plausible fit. However, both a power and a log-normal distribution were plausible for the link (neighbor) distribution shown in Fig 6B. While Vuong’s test results strengthened the case for the power law over the log-normal our sample was too small to significantly rule out either one (R = -0.515, xmin = 5, p = 0.697, see S2 Table for details).

Fig 6

Power-law fit to aggregated network degree distribution.

Power law (solid line) and a log-normal distribution (dashed line) were fit to the degree distribution of the trade network (n = 262) over the whole study period using the sequences for number of movements and aggregated links. (A) A power law distribution provided a plausible fit to the movement sequence (xmin = 84, p = 0.4, α = 1.659). (B) Both a power law (xmin = 5, p = 0.69, α = 2.152) and a log-normal (xmin = 4, p = 0.45, μ = -2.410, σ2 = 0.036) distribution were plausible for the link sequence, and one could not be favored over the other (xmin = 5, R = -0.961, p-value = 0.697). See S2 Table for details.

Subset level: Cross-border trade communities

Few markets were reachable by all the other markets via directed paths, as shown by the size of the Giant Strongly Connected Components in Table 7. When the direction of the links was neglected, all the markets belonged of the Giant Component in the yearly networks except for 2016. Markets were classified into 8 groups or trade communities with the fast, greedy modularity optimization algorithm (Q = 62.7%).

Table 7

Size of the connected components.

	2013	2014	2015	2016	2017
Giant Strongly Connected Component size (markets)	5	4	1	2	12
% of markets in GSCC	4.5%	3.7%	<0.1%	0.2%	9.8%
Giant Weakly Connected Component size (markets)	112	136	108	105	183
% of markets in GWCC	100.0%	100%	100.0%	86.1%	100.0%

Fig 7 shows the 2017 trade network by country and trade community membership. Country membership (Fig 7A and 7C) suggests that many peripheral markets tend to connect to other higher-centrality markets in their country. In turn, these hub markets serve as bridges between different country-communities, and either primarily send or receive animals to/from many other markets. However, different communities emerge when the network is partitioned to maximize the modularity. Some of those communities are essentially single-country groups–like communities 4 and 6 in Burkina Faso and Cote d’Ivoire. Others spread across national boundaries. Cross-border communities include two between Niger and Nigeria (communities 1 and 5), and one extending from Southwestern Niger to Burkina and south towards the coast (2). Some of these groups serve as trade funnels (1) while others are diffusers (6). Both community partitions highlight the relevance of cross-border ties and suggest that markets play distinct structural roles in the network.

Fig 7

Market network by country and trade communities for 2017.

Trade communities were detected using igraph’s implementation of the fast greedy [36] modularity optimization algorithm (Q = 62.7%, n = 8). Node shapes and colors indicate the country (A, C) or community (B, D) the node belongs to.

Market level: Border and urban hubs

Nine key markets, shown in Table 8, were identified with the node level metric values. Key markets were either border markets (within 50 km of an international border), in large urban centers or export markets. Key border markets dominate the in-degrees and include Nadiagou (Burkina Faso, Benin), Dan Barto (Niger, Nigeria) and Dèrassi (Benin, Nigeria). Out-degree leaders are predominantly big urban and export markets, that also serve as hubs but connecting international production to consumer markets in urbanized areas and the coast. Port-Bouët (Côte d’Ivoire), Bouaké (Côte d’Ivoire), Kumasi (Ghana) and Bobo-Dioulasso (Burkina Faso) are located in or near urban settlements. Although close to a border, we also consider Parakou (in Benin, bordering Nigeria) as an urban market because it is in an important urban settlement. The remaining key market is Fada N’Gourma (Burkina Faso) through which livestock mainly destined for export flows.

Table 8

Key nodes in the market trade networks by year and metric.

Year		2013	2014	2015	2016	2017
Volume		Parakou	Port-Bouët	Port-Bouët	Bouaké	Bouaké
Links	Degree	Nadiagou	Nadiagou	Nadiagou	Nadiagou	Dan Barto
	In-deg.	Nadiagou	Nadiagou	Nadiagou	Nadiagou	Dan Barto
	Out-deg.	Bouaké	Derassi	Bouaké	Bouaké	Fada N’Gourma
Shipments	Degree	Parakou	Nadiagou	Nadiagou	Kumasi	Dan Barto
	In-deg.	Parakou	Parakou	Parakou	Kumasi	Port-Bouët
	Out-deg.	Bobo-Dioulasso	Nadiagou	Fada N’Gourma	Bobo-Dioulasso	Bouaké

Markets with the highest total volume, and total, in and out degree in number of links and shipments are reported for each year. Border markets are markets within 50km of an international border. Urban markets are highlighted in italic. Fada N’Gourma, an export market, is in bold.

Limitations

As in Thébaud et al [1], we do not report the traded volumes for various reasons, including the possibility that some animals and/or shipments were accounted for more than once. Two other main limitations of our study pertain the data: its completeness and changes in the data collection. Incomplete and/or imperfect data are pervasive in the developing world, where the mechanisms to thoroughly capture, store, analyze and communicate findings on movements of goods are often not in place. In our case, incomplete entries could not be used in our analysis and thus limited it. Furthermore, survey markets were selected based on a regional assessment of trade and do not constitute a random sample of all markets, so our work does not paint a complete or unbiased picture of trade patterns in the region but is focused on animals originating in Burkina Faso and Mali that cross international borders in the central and eastern basins of trade of West Africa.

There are two main reasons to be cautious when comparing our results with related research. First, observed temporal fluctuations in network structure could reflect the changes in the data collection and incomplete entries, and not real changes in the movement structure. Movement data was collected by the CILSS to track the direction and magnitude of intraregional livestock trade, and not specifically to carry out this analysis. The second reason pertains related work [20-23] having different purpose, geographic coverage and scale, and using different data collection and analysis methods. Because of these differences, it is possible existing research captured transhumance and/or trade-related, formal or informal movements, or different sections of the value chain. Even if identical methods had been used to investigate animal mobility in previous work, networks of different size, density and/or structure should not be directly compared. Recent developments in network similarity and distance measures can be used to compare networks of different characteristics, but such a comparison is out of the scope of our work. To enable comparisons with future animal mobility network research, we normalized the observed statistics using simulated network ensembles for each year.

Discussion

Live animals are the most valued food product traded within West Africa; however, a considerable part of this trade is unofficial [6]. As trade liberalization continues to be pursued in the region, understanding spatiotemporal patterns of regional, live animal movements will become increasingly important to inform policies that support existing and future trade, as well as strategies for disease surveillance and control. We built weighted networks of animal movements originating primarily in Burkina Faso by representing locations (markets) as nodes and shipments as directed links between nodes. We then mapped and characterized the network, identified market communities, key markets and their roles.

Our findings show that in the central and eastern trade basins of West Africa, most animal movements crossed borders, and that those movements were made primarily by vehicle on well-established long-distance routes. The existence of major trade routes is supported by two findings: markets traded with, on average, few other markets multiple times each month, and repeating spatial patterns of trade despite changes in the data collection–like a general north-south movement and defined country roles. Despite these general patterns, we found heterogeneity in the roles that markets play in the network. For instance, markets that traded at least once were significantly closer to each other that those that did not for most years. Combined with the heterogeneous distribution of the number of shipment and trade partners, this indicates that there are hub markets that tend to trade with nearby peripheral ones, and that those hubs connect said periphery to the rest of the network.

The degree distribution followed a power law, which may suggest that the network is more vulnerable to targeted attacks or interventions than to random ones [43]; however, implications of unrecorded and/or incomplete links must be studied further for this specific type of network to assert this. We attempted to reconstruct the observed networks from a null model but local constraints (the degree or link sequences) were not able to reproduce some of the observed network properties. The empirical networks were considerably more disassortative and propinquital than the ensembles, which indicates that other (possibly spatial) processes are necessary to explain higher-order properties of the network. Our findings reflect the fragmentation of the road network inherited from the colonial period and the poor accessibility of many peripheral markets in the region. Cattle trade relies on a handful of paved roads in each country and on a limited number of transnational routes that help connect the Sahel to the main consumption centers [49].

Temporal variation in the network structure from year to year can reflect (not exclusively) changes in the data collection, climatic conditions of historical events, or a combination of them. Untangling their individual contributions is a complex task that we do not attempt here; however, we provide some context that can inform observed changes in the network. This context is provided while considering that our dataset consisted in a subset of all the shipments recorded at a sample of market locations, and that this sample changed during the study period. In 2015, a late start to the rainy season and unfavorable pasture conditions for the second year in a row triggered sale of animals with deteriorated body condition, and uncommon sales in Northern Burkina Faso [44]. It is possible that destocked herds from previous years affected the availability of animals (and therefore, trade) during 2016; however, favorable rainfall during the second half of 2016 and into 2017 increased pasture availability which could explain the increase in shipments for 2017 when compared to the previous years. In contrast, the devaluation of the Naira in 2016 (Nigeria’s official currency) was unfavorable for livestock exports into Nigeria [45]. Unfavorable conditions because of this devaluation yet trade movements towards Nigeria in 2017 suggest that markets between Niger and Nigeria were included in the data collection sample during 2017. In Mauritania, Tabaski affects spatiotemporal animal movement patterns [21]. Because Tabaski was observed at the end of August, Tabaski-related movements had probably started when our study period finished, which could explain why the number of animal shipments for 2017 are comparable to previous years even when the collection interval was shorter.

We found densely connected market communities that shared more internal links than country-specific communities, and more links than expected by chance. Some of these communities corresponded with trade patterns that were persistent through the years, such the community that moves animals from Central to Southern Côte d’Ivoire and the community within Burkina Faso, while most of them spanned over at least one border, like the groups between Benin and Nigeria, and between Niger and Nigeria. Cross-border trade communities support findings from previous country-specific and regional studies: multiple-country approaches are necessary to support intraregional livestock trade [9,20,23].

Hub markets in border areas and urban centers identified previously in the literature [22,46-48] were confirmed with node volumes and degree centralities. Border markets like Nadiagou, and Dan Barto serve as assembly points where livestock is received from numerous markets located closer to production areas (and therefore have large in-degrees). Urban markets such as Bouaké and export markets like Fada N’Gourma, on the other hand, have a high out-degree centrality and receive long-distance traded livestock before it is shipped to consumer markets. A third of the key markets–(Parakou, Dan Barto and Dèrassi)–were located within 50 km of Nigeria, highlighting the importance of the Nigerian market as a destination for regional livestock production that was not apparent from the country-level findings. These key markets also suggest that Benin and Niger may play an intermediary role in shipments ultimately destined to Nigeria.

Previous efforts to study livestock trade in West Africa have concerned national movements [22], transhumant movements [1] or had limited geographic scope [20,21,23]. We analyzed as a network, for the first time, regional animal movements originating mainly in Burkina Faso, one of the top three live animal exporters in the region. Despite its limitations, our study sheds light on the understudied structure of regional livestock trade and the role of markets in West Africa.

Analysis of existing agricultural product mobility data and its communication to policy makers is essential to advance regional trade integration. Our findings advocate for increasing the density and quality of the regional road network, which could help develop livestock and other intraregional trade products that are primarily transported by vehicle. Border areas should be prioritized as their infrastructure is not prepared to withstand increasing regional trade movements and will likely constrain them [48,49]. Additionally, our results substantiate removing border tariffs and delays that contribute to the higher cost of food in West Africa when compared to other regions of the world [50], and that decrease the population base reachable from border towns [49]. One way forward is to operationalize existing One Stop Border Posts (OSBPs) in important cross-border areas that have been identified in this manuscript.

Some future research directions were identified while carrying out this work. Future efforts to study animal mobility in the region (and its implications for disease surveillance, detection and control) should exploit the bilateral nature of trade by streamlining research initiatives and sharing methods, data, findings and lessons learned. Opportunities also exist to synthesize knowledge from different sources and fields such as remote sensing products, infrastructure maps, market information systems and household surveys. Given its importance as a consumption market for the region, an effort should be made to better understand live animal shipment patterns into Nigeria. Unfortunately, the current security situation in the Sahel can complicate collecting data in some areas. Finally, using data-informed simulations, previous work has concluded that the risk of regional disease transmission is high in West Africa [20,23]. The risk of regional disease spread can be re-assessed with recent developments in (a) animal mobility patterns in the region and on the (b) inference of network structure from observed network patterns.

Percentage of livestock movements by type.

(DOCX)

Click here for additional data file.

Fitting the degree distribution.

(DOCX)

Click here for additional data file.

Proportion of movements by type of movement and month 2013–2017.

(DOCX)

Click here for additional data file.

Proportion of movements by type of livestock and month 2013–2017.

(DOCX)

Click here for additional data file.

Proportion of movements by type of transport and month 2013–2017.

(DOCX)

Click here for additional data file.

Data collection methodology.

(DOCX)

Click here for additional data file.

Edge list.

(CSV)

Click here for additional data file.

Node list.

(CSV)

Click here for additional data file.

25 Sep 2019

PONE-D-19-20212

Network analysis of regional livestock trade in West Africa

PLOS ONE

Dear Ms. Valerio,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

We would appreciate receiving your revised manuscript by Nov 09 2019 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Peng Li, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Thank you for stating the following in the Competing Interests section:

I have read the journal's policy and the authors of this manuscript have the following competing interests: RM serves in the Editorial Board of PLOS ONE.

Please confirm that this does not alter your adherence to all PLOS ONE policies on sharing data and materials, by including the following statement: "This does not alter our adherence to  PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests).  If there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include your updated Competing Interests statement in your cover letter; we will change the online submission form on your behalf.

Please know it is PLOS ONE policy for corresponding authors to declare, on behalf of all authors, all potential competing interests for the purposes of transparency. PLOS defines a competing interest as anything that interferes with, or could reasonably be perceived as interfering with, the full and objective presentation, peer review, editorial decision-making, or publication of research or non-research articles submitted to one of the journals. Competing interests can be financial or non-financial, professional, or personal. Competing interests can arise in relationship to an organization or another person. Please follow this link to our website for more details on competing interests: http://journals.plos.org/plosone/s/competing-interests

3.  We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

Reviewer #3: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: General comment

This paper aims to investigate the livestock market and trade network in West Africa. The topic is very interesting and exploratory. A bunch of network analysis metrics is used to discover the pattern of livestock trade origins and destinations. This paper is generally well written. The reviewer recommends minor revision before publishing in this journal. The following comments may help improve the manuscript.

Specific comment:

1. In Line 106-107, the name of the database may need to be explained in English.

2. In Line 265, the p-value is missing.

3. In “Network Analysis Results” and “Discussion and Conclusions” section, please use active voice such as Table 5 shows … instead of referring to the source in parentheses such as …(Table 5).

4. In Line 288, “Large GWCC reinforce” maybe “A large GWCC reinforces”.

5. In Line 111, the authors claim “our findings are then contrasted with existing literature”, please clarify the opinions in the literation and the related conclusion in this paper, and the reason for the difference in the discussion.

Reviewer #2: 1. The significance of the research should be further clarified. Many scholars have studied related content, as listed in the references. In line 35-38, live-animal trade has not been formally studied at a regional scale, you attempt to fill this gap. In fact, it is not a real gap. So where is the innovation of your research?

2. In the paper, one of the conclusion is the trade network has a loose structure (Line 244 and Line 344). What is the criterion for judging whether the structure is loose or tight? Is it described by one indicator or multiple indicators? If multiple indicators are used to describe, an evaluation index system should be given. In addition, compared to what we can say the structure is loose or tight?

3. Why infrastructure construction can affect market structure (Line 351, Line414)? In addition to the infrastructure, what other factors affect the market structure? Please provide evidence.

4. There is too much about the CILSS database (Line107, Line 150, Line330, Line 388). Put the contents about CILSS only in Data source and pre-processing is better.

5. The discussion and conclusions are more likely the results. More specific conclusions need to be reflected. For example, what kind of reginal policies are needed to fulfill the livestock sector’s potential?

Reviewer #3: General comments:

The abstract should be slightly modified to include the weighted property of the studied networks.

Many studies are cited as a basis for the submitted manuscript. An effort has been made at improving the understanding on livestock mobility at regional level, by applying already published methods focusing on the same area, and is appreciated. It will allow consistency in the obtained results and discussion at larger scale. It will also provide support to regional surveillance. However, none explanation nor sufficient details on the used methods are available in the manuscript. The authors should be aware that the reader will have to read all the mentioned published papers to be able to understand the submitted manuscript in a sufficient way... which is not acceptable. Major revision are needed to include few sentences on the "recycle part" of the published studies in order to support the reader understanding.

In addition, non information on the data collection methods and the studied dataset are provided. The authors only refers to a published report supposed to describe both (field data collection method, and related database). Only 2 sentences, provide information on the available data, extracted from the database, and how, the authors cleaned their dataset. Further description have to be provided in the methods section to support the robustness of the used database and the adequacy of the used methods.

Reading the analysis and the shown data, I'm concerned about the 'regional level' appellation... The authors should define the 'region' that they're studying and take into consideration that the geographical region (Fig 1) is not the same that the region concerned by the CILSS database. This database was built by implementing field data collection in 8 country. Niger and Guinea are not part of the CILSS database, but are mentioned as destination and presented on the Fig 1. Benin is part of the CILSS and is not mentioned in the Fig 1 caption. Thanks to the lack of data collection points in Niger and Guinea, the flows could have occurred and being part of the unrecorded data (or recorded in another database). Some discussion on this purpose could be provided.

Moreover, no effort has been made to support a regional understanding on the observed network. General statement are provided as discussion, and deserve further development.

Finally, the writing have to be organised to fit the standard on scientific writing. The current version of the manuscript mixed results and discussion section in multiple part of the manuscript (e.g. L 356, L 361-362).

Other comments:

L 55-56: in the current form, it is unclear about which agricultural products the authors are is talking (species, type of animal products).

L 60-61: "as the projected demand surpasses demographic growth"

If the comment is related to all included at regional scale, the sentence is of poor added value. It'd be valuable to further develop.

L 65: the Livestock trade and livestock products mentioned earlier, are there only related to the trade of meat and live animal in general? Some precisions would be needed to better introduce the study goal. The introduction remain unclear on the purpose: which species, which products, which distinction at regional scale (spatial heterogeneity of the needs and habits related to agriculture...).

L 77-78: the authors should provide precision on what they consider to be source of 'disruptions and shocks'

L 78-81: and related to these livestock data... anthropological studies could be valuable at this scale to better integrate social and human factors and impact on trade habits and agricultural sector. Access to the field is not particularly easy in this area.

L 107 and Material and Methods: The authors should provide sufficient description on the database that they are using in the submitted manuscript (what is CILSS, how the database is organised). The mentioned reference is a 34 pages report which describe a variety of database. Few additional sentences have to be added in order to summarise and support the current study.

Add a brief description on the database, how it was built (compulsory or not, data collection methods).

Is duration of the travel (origin/destination) known?

L 167: The authors have sufficient data to study the market stability at this scale but do not develop. Why the authors decided not to compare the yearly network dataset? Time-series data between 2012-2017 were available but none or a few investigations were done on their structure and evolution over the period. Moreover, even if only 2 trimesters were available in 2017... the recorded data are comparable to those of 2013 and 2016... Do you have information to relate to this point? Could you have part of the explanation?

Discussion on the relation with historical events would be appreciated (2016 = decrease of the national value in Nigeria). And for the regional level? What about quality of the pasture and socio-economical factors (trend in the market, prices... data are part of the CILSS and are from real added value)?

L 207: if ID numbers were collected during the shipment recording... how double counting could have occurred?

L 213: The Tabaski is here mentioned but was not introduced. What is the purpose here? At which scale this religious festivity is practiced? The results are not discussed... The authors should discuss all the results that they are providing, or the results should be withdraw from the submitted manuscript.

Moreover, the Tabaski can hold different name at regional scale and is yearly dependent and occurred between September and October from 2013 to 2017. These specificities could have major impact on the trade habits (related to drought and agricultural/pastoral activities). None discussion of these major aspects was done. An effort should be done to further improve the submitted study and it interest for a regional analysis.

L 332: I'm concerned about the meaning of 'daily data collection'. What is the relation between the recorded date and the effective movement (shipment)? Is this possible that the date of declaration made to the officer do not refers to the date in which the Livestock were traded? In Africa, date can be confusing.

L 352-353: Comparison has to be done carefully. The methods on data collection impact the dataset.

L 393: In the case of Mauritania, the dataset is mostly national with indication on the markets at international scale (regional - origin and destination). The Livestock trade networks were sub-divided into species networks because of the variability on the consumption habits like the Tabaski. At world and regional scale, the authors had highlighted the interest of other socio-cultural habits. Each of the networks were described at both scale, but due to the data collection scheme, it was not possible to implement the predictive study (as well as the finer SNA analysis) at regional scale (end-point markets are not studied, indeed the SNA parameters can't reflect the real centrality of these markets). The methods from Apolloni et al. and Nicolas et al. provided robustness and consistency to the analysis regarding the input dataset (field data collection was scheduled for the purpose of the Livestock trade network analysis). The level of detail is finer, however, only one year was available!

Comparison between the two studies has to be done carefully...

L 406-408: Requirement of a standardised data collection methods at regional scale, would also be expected

S7 Fig. Give the full country name in the figure caption as in the Fig 1.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

17 Feb 2020

Response to reviewers: letter responding to each point raised by the academic editor and reviewer(s)

Dear Dr. Peng Li,

Many thanks for your excellent review of our submission. We have carefully reviewed and responded to all the points raised by yourself and the three reviewers. Please find each point raised and its response below in its corresponding section: Academic Editor, Reviewer #1, Reviewer #2 and Reviewer #3. Specific comments within each section were numbered as E1, E2 (editor section), R1.1, R2.2 (reviewer 1 section), and so on, for simplicity. We have formatted the response to each general and specific comment in red text.

Academic Editor

Journal Requirements:

E1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

RESPONSE:

We appreciate that the Editor called this to our attention. We have made the following changes to meet PLOS ONE’s style requirements and to address reviewer #2 comments (R2.5):

• Modified the style of “Introduction” to heading level 1 (bold, 18pt)

• Corrected the level 3 heading (Network Analysis) to sentence case (changed to Network analysis in bold, 14pt)

• Revised level 2 heading “Network Analysis” to sentence case (“Network analysis”)

• Changed level 1 heading “Discussion and Conclusions” to “Discussion”

• Separated the Supplementary Information file into individual files named according to PLOS ONE’s guidelines (SI_1, SI_2, …).

E2. Thank you for stating the following in the Competing Interests section:

I have read the journal's policy and the authors of this manuscript have the following competing interests: RM serves in the Editorial Board of PLOS ONE.

Please confirm that this does not alter your adherence to all PLOS ONE policies on sharing data and materials, by including the following statement: "This does not alter our adherence to PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests). If there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include your updated Competing Interests statement in your cover letter; we will change the online submission form on your behalf.

Please know it is PLOS ONE policy for corresponding authors to declare, on behalf of all authors, all potential competing interests for the purposes of transparency. PLOS defines a competing interest as anything that interferes with, or could reasonably be perceived as interfering with, the full and objective presentation, peer review, editorial decision-making, or publication of research or non-research articles submitted to one of the journals. Competing interests can be financial or non-financial, professional, or personal. Competing interests can arise in relationship to an organization or another person. Please follow this link to our website for more details on competing interests: http://journals.plos.org/plosone/s/competing-interests

RESPONSE:

We have updated our cover letter, adding a competing interests statement that now reads:

“RM (co-author) serves in the Editorial Board of PLOS ONE. However, this does not alter our adherence to PLOS ONE policies on sharing data and materials.”

We also included a data availability statement that now reads:

“Data are available as a supporting information file of the manuscript (S7-8).”

E3. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

RESPONSE:

We have included the data as a supporting information file of the manuscript. Our cover letter has been modified to reflect this:

“Data are available as a supporting information files in the manuscript (S7-8).”

Reviewer # 1

R1 General comment

This paper aims to investigate the livestock market and trade network in West Africa. The topic is very interesting and exploratory. A bunch of network analysis metrics is used to discover the pattern of livestock trade origins and destinations. This paper is generally well written. The reviewer recommends minor revision before publishing in this journal. The following comments may help improve the manuscript.

RESPONSE:

We thank Reviewer 1 for their time and comments. Below we address each specific point raised, in order.

R1 Specific comments

R1.1. In Line 106-107, the name of the database may need to be explained in English.

RESPONSE:

Lines 100-101 have been modified to include the name of the database in English:

“Permanent Interstate Committee for Drought Control in the Sahel (Comité Inter-Etats pour la Lutte contre la Sécheresse au Sahel, CILSS).”

R1.2. In Line 265, the p-value is missing.

RESPONSE:

All the p-values have been included (now L315-317). The reader is also referred to S5 table for all the details. Thank you.

R1.3. In “Network Analysis Results” and “Discussion and Conclusions” section, please use active voice such as Table 5 shows … instead of referring to the source in parentheses such as …(Table 5).

RESPONSE:

We have modified the Results (L228-358) and Discussion (L386-484) sections to the active voice.

R1.4. In Line 288, “Large GWCC reinforce” maybe “A large GWCC reinforces”.

RESPONSE:

This sentence is no longer included in the manuscript. We thank the reviewer for noticing this error.

R1.5. In Line 111, the authors claim “our findings are then contrasted with existing literature”, please clarify the opinions in the literation and the related conclusion in this paper, and the reason for the difference in the discussion.

RESPONSE: We appreciate the reviewer’s suggestion. To address this comment and reviewers # 2 and # 3 concerns on manuscript organization and clarity, we have done the following:

• Separated and Rewrote the Results and Discussion sections

• Included possible reasons for differences between our findings and existing studies in the Discussion (L372-385)

Reviewer # 2

R2 Specific comments

R2.1. The significance of the research should be further clarified. Many scholars have studied related content, as listed in the references. In line 35-38, live-animal trade has not been formally studied at a regional scale, you attempt to fill this gap. In fact, it is not a real gap. So where is the innovation of your research?

RESPONSE:

We have reframed the importance of our work in the Introduction (L93-99) as follows:

“Although some understanding of animal mobility in the region exists, trade patterns of animals originating in the top exporters (Niger, Mali and Burkina Faso) have not been quantitatively analyzed as a network.

Our objective is to advance the formal study of regional trade networks in West Africa. We use network analysis to characterize the trade network of live animals (cattle, sheep, goats and donkeys) originating primarily in Burkina Faso, the third largest exporter of live animals in the region [9] as they are shipped through the central and eastern trade basins.”

We agree that some scholars have covered related content before in West Africa (Lines 82-95); however, previous work lacks the geographic and time breadth that the data used in our study has. Of the previous 5 formal (quantitative) network analyses of animal mobility in West Africa:

• The study by Thébaud et al [1] concerned transhumant movements only [1], differing from ours because, although some animals can be sold during transhumance, seasonal movements are not primarily fueled by trade. Additionally, this study did not analyze animal movements as a network.

• Apolloni et al. [2] described preliminary findings from a survey on animal movements conducted in Mauritania. Nicolas et al. [3] then attempted to predict these movements within Mauritania only, even though 70% of the animals in their database had crossed borders. Because their work was focused on animals entering and leaving Mauritania, it differs from ours.

• Two studies investigated the spread of diseases regionally through animal movements: Dean et al. [4] and Motta et al. [5]. Dean et al. [4] evaluate the risk of disease spread from Togo to neighboring countries, and from Burkina Faso into Togo. Data was collected in the Savannah region of northern Togo only; therefore, this study differs fundamentally from ours in both its objective (evaluating disease transmission risk) and its geographic extent. It is worth mentioning that CILSS’s database does capture animals entering Togo from Burkina Faso in comparable numbers as the results presented by Dean et al (~40,000 heads in Jan-August 2017) but does not capture animals leaving Togo as well as their study. On the other hand, Motta et al. [5] do not explicitly perform simulations of disease transmission risk, but describe the structure of the Cameroonian animal movement network, discussing the implications for disease spread. This study differs from ours primarily in the geographic coverage of the movement network, although some of the objectives are similar.

Our manuscript focuses on regional movements of animals that originate in Burkina Faso, one of the top three live animal exporters in West Africa (according to the World Bank [6]), and therefore complements the existing literature. Furthermore, findings at three scales of analysis are combined to make sense of regional patterns of trade and identify important locations for the regional network.

Although some studies have covered this subject in the past, we believe that regional trade of live animals in West Africa is still quantitatively understudied. Studying animal trade patterns is important because live animals are the most important food product traded in the region, official statistics do not adequately capture their flows, and because describing trade patterns can inform regional policies on trade, disease control and food security.

R2.2. In the paper, one of the conclusions is the trade network has a loose structure (Line 244 and Line 344). What is the criterion for judging whether the structure is loose or tight? Is it described by one indicator or multiple indicators? If multiple indicators are used to describe, an evaluation index system should be given. In addition, compared to what we can say the structure is loose or tight?

RESPONSE:

We have removed this line from the paragraph. Instead, we compare the observed networks to their corresponding ensemble generated with a null (configuration) model that maintains the observed degree sequence (L 293, 301). With these simulated networks, we calculated a z score for each observed network metric value. Some metrics had to be excluded because they could not be calculated. The density is one of the metrics that could not be benchmarked against the simulated networks, because by construction all the networks in the configuration model ensemble had the same density. We thank the reviewer for this very useful comment.

R2.3. Why infrastructure construction can affect market structure (Line 351, Line414)? In addition to the infrastructure, what other factors affect the market structure? Please provide evidence.

RESPONSE:

Infrastructure can affect the observed market structure because survey points were located on major trade corridors. These trade corridors rely on transnational paved roads that help connect the Sahel to consumption centers [7]. Therefore, it is possible that the location of the survey points affects the observed trade network structure. Most movement being made by vehicle (in contrast to findings in other West African countries) support this claim. We moved those lines to L413-417, that now read:

“Our findings reflect the fragmentation of the road network inherited from the colonial period and the poor accessibility of many peripheral markets in the region. Cattle trade relies on a handful of paved roads in each country and on a limited number of transnational routes that help connect the Sahel to the main consumption centers [49].”

In a broader sense, infrastructure can affect the market structure by incentivizing development and inflow of people and goods towards newly accessible locations. For example, the construction of the Nouakchott-Nouadhibou route in Mauritania in 2014 incentivized fishermen and herdsmen to relocate near the paved road to access new markets [8]. In the case of livestock, historical trade routes inherited from the colonial period have conditioned how and where live animals are traded (in the form of ethnic and social ties, and a fragmented road network).

R2.4. There is too much about the CILSS database (Line107, Line 150, Line330, Line 388). Put the contents about CILSS only in Data source and pre-processing is better.

RESPONSE:

We have moved details on the database to the Data source and pre-processing sub-section. Because no scientific publication exists that describes this database in detail, we included a supplementary information file (S6 File. Data collection methodology) with more information about the data collection and also referred the reader to the official documentation of CILSS.

R2.5. The discussion and conclusions are more likely the results. More specific conclusions need to be reflected. For example, what kind of regional policies are needed to fulfill the livestock sector’s potential?

RESPONSE:

To address this and other reviewer comments, we have modified the sections as follows:

• “Network Analysis Results” has been renamed “Results”. The content has been modified accordingly

• The “Discussion and conclusions” section has been renamed “Discussion”

In the conclusions, we have recommended the following actions concerning regional policies (L463-472):

• Increasing the density and quality of the regional road network to support trade of animals and other food products that are primarily transported by vehicle, prioritizing cross-border areas important for regional trade

• Removing border delays and costs that contribute to the higher cost of food in West Africa when compared to other regions of the world. One specific way forward is to (install) operationalize One Stop Border Posts (OSBPs) in known cross-border areas of importance (some of which have been identified in the manuscript)

• Streamlining regional animal movement and disease surveillance/control research: sharing findings, best practices and lessons learned. This will enable synergies and avoid duplication of efforts led by different countries and/or international organizations

Additionally, the following research priorities have been identified (L473-484):

• Studying spatiotemporal patterns of animal flows into Nigeria (an important regional consumption market)

• Assessing the risk of disease spread at a regional scale using recent developments in (a) regional patterns of animal mobility, (b) inference of network structure from observed trade patterns (see [9])

Reviewer #3

R3 General comment:

The abstract should be slightly modified to include the weighted property of the studied networks.

Many studies are cited as a basis for the submitted manuscript. An effort has been made at improving the understanding on livestock mobility at regional level, by applying already published methods focusing on the same area, and is appreciated. It will allow consistency in the obtained results and discussion at larger scale. It will also provide support to regional surveillance. However, none explanation nor sufficient details on the used methods are available in the manuscript. The authors should be aware that the reader will have to read all the mentioned published papers to be able to understand the submitted manuscript in a sufficient way... which is not acceptable. Major revision are needed to include few sentences on the "recycle part" of the published studies in order to support the reader understanding.

In addition, non information on the data collection methods and the studied dataset are provided. The authors only refers to a published report supposed to describe both (field data collection method, and related database). Only 2 sentences, provide information on the available data, extracted from the database, and how, the authors cleaned their dataset. Further description have to be provided in the methods section to support the robustness of the used database and the adequacy of the used methods.

Reading the analysis and the shown data, I'm concerned about the 'regional level' appellation... The authors should define the 'region' that they're studying and take into consideration that the geographical region (Fig 1) is not the same that the region concerned by the CILSS database. This database was built by implementing field data collection in 8 country. Niger and Guinea are not part of the CILSS database, but are mentioned as destination and presented on the Fig 1. Benin is part of the CILSS and is not mentioned in the Fig 1 caption. Thanks to the lack of data collection points in Niger and Guinea, the flows could have occurred and being part of the unrecorded data (or recorded in another database). Some discussion on this purpose could be provided.

Moreover, no effort has been made to support a regional understanding on the observed network. General statement are provided as discussion, and deserve further development.

Finally, the writing have to be organised to fit the standard on scientific writing. The current version of the manuscript mixed results and discussion section in multiple part of the manuscript (e.g. L 356, L 361-362).

RESPONSE:

We thank Reviewer 3 for their time and comments. Below we address each point raised in the general commentary, in order.

The abstract has been modified to reflect the weighted nature of the networks (L30).

A summary of previous related work and how it is similar or differs from our manuscript has been included in the introduction (L83-96). This paragraph summarizes previous work and how it relates and differs from this manuscript.

We have included more information on the methods. The data collection methodology is described in S6 File in enough detail for the reader to understand how the data was collected, processed and stored. The reader is referred to the website that contains all the official documentation of the data collection (www.agrictrade.net) for more information. At the time this revision was written, this website was being finalized. The data necessary to replicate our analysis are provided in S7-S8 Files. Lines 130-170 detail the steps taken to pre-process the database before analysis, while S7-S8 Files suffice to rebuild the yearly networks. The variables in the provided data are described in Table 1. Concerning the “recycled” part of the paper, the fourth paragraph in the introduction (L82-95) now briefly explains the livestock network literature for West Africa and explains how our study differs from/is similar to it.

The significance and geographic scope of our work has been updated. Although we intended to include more countries in our analysis, many entries in the database were incomplete and therefore could not be included. Our final dataset comprises movements originating primarily in Burkina Faso, one of the top three exporter of animals in the region. The data tracks these animals as they are transported primarily in a North-South direction into Benin, Cote d’Ivoire, Ghana and Nigeria. Thus, our study region comprises the Central trade basin and part of the Eastern trade basin of West Africa (this has been stated clearly in the manuscript, L98, L138-142).

The manuscript has been reorganized for clarity, as suggested by the reviewer. We have modified the content of each section so that:

• The “Results” section only presents our findings

• The “Discussion” section interprets and contextualizes the results

R3 Specific comments

R3.1 L 55-56: in the current form, it is unclear about which agricultural products the authors are is talking (species, type of animal products).

RESPONSE:

L 55-56 now reads: “…therefore unknown. Livestock (mostly cattle and small ruminants) are traded live and constitute the leading product in food trade for the region [8,9]”

R3.2 L 60-61: "as the projected demand surpasses demographic growth"

If the comment is related to all included at regional scale, the sentence is of poor added value. It'd be valuable to further develop.

RESPONSE:

We have removed the sentence and instead added:

“…the existing gap between production and demand of animal products is expected to widen in the next decades driven by population and income growth, migration and urbanization“ (L59-61)

R3.3 L 65: the Livestock trade and livestock products mentioned earlier, are there only related to the trade of meat and live animal in general? Some precisions would be needed to better introduce the study goal. The introduction remain unclear on the purpose: which species, which products, which distinction at regional scale (spatial heterogeneity of the needs and habits related to agriculture...).

RESPONSE:

We have modified this paragraph. We have specified the product and species (live cattle and small ruminants, L56), and summarized it again in the last paragraph of the introduction where the objectives are stated (L98-99).

R3.4 L 77-78: the authors should provide precision on what they consider to be source of 'disruptions and shocks'

RESPONSE:

We have identified two instances of disruptions and/or shocks (climatic and conflict events, L77-78). These types of disruptions can affect both the production of animals and the functioning of markets (and therefore trade).

R3.5 L 78-81: and related to these livestock data... anthropological studies could be valuable at this scale to better integrate social and human factors and impact on trade habits and agricultural sector. Access to the field is not particularly easy in this area.

RESPONSE:

We agree with the reviewer. Much could be done with other types of studies. We believe that spaces exist for synergies between different types of studies to better understand agricultural trade patterns (L478).

R3.6 L 107 and Material and Methods: The authors should provide sufficient description on the database that they are using in the submitted manuscript (what is CILSS, how the database is organised). The mentioned reference is a 34 pages report which describe a variety of database. Few additional sentences have to be added in order to summarise and support the current study.

Add a brief description on the database, how it was built (compulsory or not, data collection methods).

RESPONSE:

We thank the reviewer for highlighting this. We have added “S6 File. Data collection methodology” with more details on the CILSS and describing their data collection purpose and methodology. Additionally, Table 1 describes each variable present in the data provided in S7 File and S8 File. We have removed the report from the references and instead cited more appropriate sources. The data collection was not compulsory for traders, but to our knowledge no problems were reported with trader participation.

R3.7 Is duration of the travel (origin/destination) known?

RESPONSE:

The duration of travel is unknown.

R3.8 L 167: The authors have sufficient data to study the market stability at this scale but do not develop. Why the authors decided not to compare the yearly network dataset? Time-series data between 2012-2017 were available but none or a few investigations were done on their structure and evolution over the period. Moreover, even if only 2 trimesters were available in 2017... the recorded data are comparable to those of 2013 and 2016... Do you have information to relate to this point? Could you have part of the explanation?

RESPONSE:

We briefly present the evolution of the network in Table 5 in the form of network statistics for each year. In addition, Fig 5 presents each yearly network. We did not elaborate on market stability for the following reasons:

• Missing data: As is reported in the manuscript, many movements are missing either the origin or destination market. Only complete, geolocated entries were used in our analysis.

• Changes in data collection: The number of markets surveyed changed between 2013-2017 (L287). Throughout the study period, survey points have been added and removed from the sample.

We chose to not study the evolution of the market structure in detail, because changes can be an artifact of the data collection and not reflect true variation in the trade network.

Regarding the comparability of 2017 (Q1-Q3) volumes with previous years, the following can explain this (L428):

• More markets were surveyed in 2017 than in previous years

• FEWSNET reported above-average cumulative seasonal rainfall for Burkina Faso in July 2017; At the end of 2017 livestock were reported to be in good physical condition, which suggests that enough biomass was available during a number of months prior

• Tabaski is set on the lunar calendar and its date for 2017 was at the end of August; significantly more sheep movements had been reported by August in 2017 than for previous years, when Tabaski was later

• When the duration of the data collection for each year is considered, more movements were made in 2014 than in 2017 (2017 was a “normal” production year according to FEWSNET, https://fews.net/west-africa/burkina-faso/key-message-update/july-2017 )

R3.9 Discussion on the relation with historical events would be appreciated (2016 = decrease of the national value in Nigeria). And for the regional level? What about quality of the pasture and socio-economical factors (trend in the market, prices... data are part of the CILSS and are from real added value)?

RESPONSE:

The naira devaluation and other climatic conditions are now discussed in L417-437. In the discussion, the “Limitations” subsection discusses further why observed changes can be reflecting changes in the data collection.

Regarding market price trends, FEWSNET did not publish livestock prices in their price bulletin series for Burkina Faso. We did not find other official sources of livestock market prices elsewhere, so we could not use them.

R3.10 L 207: if ID numbers were collected during the shipment recording... how double counting could have occurred?

RESPONSE: In this manuscript, we refrain from reporting on total volumes because of possible double counting of animals (L361). CILSS’s objective is to track the magnitude, direction and changes in intraregional trade, and they aim to capture both official and informal movements. Therefore, animal/official custom transaction IDs were not collected during the study period. Under their collection methodology (see S6 File for details), we presume that multiple counting could have happen in three ways:

• An animal is counted early in the trade journey and then counted in another section of the journey. For example, a truck shipment with cattle is registered at Bobo-Dioulasso (collection point in Burkina Faso) as leaving Bobo-Dioulasso towards Daloa (in Cote d’Ivoire). There is a possibility that the same truck shipment is registered at Man (collection point in Cote d’Ivoire) point as going to Abidjan (also in Cote d’Ivoire).

• An animal is counted as part of a specific shipment in a specific data collection pint. Later, this animal leaves the initial shipment it was counted on and becomes part of a new shipment (for example, it is sold to another trader). This new shipment is later reported at another survey point.

• The same shipment is counted by each of the two enumerators at the collection point and reported to the focal point twice. The data collection methodology has measures in place to prevent this type of double counting (enumerators compare and collate their records at the end of each survey day, see S6 File).

If we add all the shipments to report on total volumes, we run the risk of counting some animals more than once and have no reliable method to detect these duplicates.

R3.11 L 213: The Tabaski is here mentioned but was not introduced. What is the purpose here? At which scale this religious festivity is practiced? The results are not discussed... The authors should discuss all the results that they are providing, or the results should be withdraw from the submitted manuscript.

Moreover, the Tabaski can hold different name at regional scale and is yearly dependent and occurred between September and October from 2013 to 2017. These specificities could have major impact on the trade habits (related to drought and agricultural/pastoral activities). None discussion of these major aspects was done. An effort should be done to further improve the submitted study and it interest for a regional analysis.

RESPONSE:

We have now introduced Tabaski to the reader in the second paragraph of the results (L235). We specified that it is not observed on the Gregorian calendar and therefore is celebrated in a different date each year. The importance of the holiday for the region and its consumption patterns is highlighted (), and Alternative names for Tabaski were also introduced to the reader (L234). We have also included a paragraph discussing the possible climatic, historical and data collection conditions that could explain the temporal changes in our networks. (L 417-437)

R3.12 L 332: I'm concerned about the meaning of 'daily data collection'. What is the relation between the recorded date and the effective movement (shipment)? Is this possible that the date of declaration made to the officer do not refers to the date in which the Livestock were traded? In Africa, date can be confusing.

RESPONSE:

We appreciate that the reviewer has brought up this point of concern. CILSS’s data collection is carried out daily or on market days in selected markets and border-crossing points in trade corridors (see S6 File for more details). On data collection days, two enumerators visit each survey point. Movements are recorded by both enumerators on the day that they take place, and not when they are reported to official authorities or when they are collated and transferred to the focal point in charge of each country-commodity. This is reflected in the dates reported in the data (not shown in the manuscript because of the time scale of the analysis). Furthermore, movement data collected in each survey point is transferred to a focal point at the end of each month. Therefore, if the date of the movement is estimated, which is highly unlikely given the collection methodology, collation of data by month to the focal point and subsequent transfer to the CILSS (and a yearly time step in the manuscript) would lessen the impact of day-differences between the actual movement date and estimated dates. Furthermore, the ASSESS project (USAID/West Africa Analytical Support Services and Evaluations for Sustainable Systems in Agriculture, Environment, and Trade project) carried out a Data Quality Assessment (DQA) of CILSS’s intra-regional trade monitoring in 2017. Although the DQA found some improvement opportunities, no issues concerning discrepancies on movement dates were raised. We understand that data collection in the field is far from perfect and appreciate this concern.

R3.13 L 352-353: Comparison has to be done carefully. The methods on data collection impact the dataset.

RESPONSE:

We agree that comparison should be handled carefully. We have all direct comparison between the networks and instead compared the observed network metrics to the ensemble of networks generated with the configuration model (L293). Direct comparisons between the studies are not made with the caveat that the data collection efforts were different (L372). Although it is possible to compare different networks with novel graph distance and other types of algorithms, that is outside of our manuscript scope.

R3.14 L 393: In the case of Mauritania, the dataset is mostly national with indication on the markets at international scale (regional - origin and destination). The Livestock trade networks were sub-divided into species networks because of the variability on the consumption habits like the Tabaski. At world and regional scale, the authors had highlighted the interest of other socio-cultural habits. Each of the networks were described at both scale, but due to the data collection scheme, it was not possible to implement the predictive study (as well as the finer SNA analysis) at regional scale (end-point markets are not studied, indeed the SNA parameters can't reflect the real centrality of these markets). The methods from Apolloni et al. and Nicolas et al. provided robustness and consistency to the analysis regarding the input dataset (field data collection was scheduled for the purpose of the Livestock trade network analysis). The level of detail is finer, however, only one year was available!

Comparison between the two studies has to be done carefully...

RESPONSE:

We agree with the reviewer. We have removed direct comparisons between the network structures and highlight why different network should not be compared in the manuscript (L372).

R3.15 L 406-408: Requirement of a standardised data collection methods at regional scale, would also be expected

RESPONSE:

We thank the reviewer for highlighting the need for more information concerning the data collection. The data collection was coordinated by the CILSS and carried out with the support of 10 regional organizations involved in trade. Standardized methods were used in the data collection (in the form of enumerator training and standardized collection forms), and throughout the data pipeline (from collection, collation, storage, entry, transfer, analysis to reporting). We have included “S6 File. Data collection methodology” in the Supplementary Information to provide more details on the data collection methods.

R3.16 S7 Fig. Give the full country name in the figure caption as in the Fig 1.

RESPONSE: S7 Fig has been replaced by S4 Fig and the ISO codes are no longer necessary.

References

1. Thébaud B. Pastoral and Agropastoral Resilience in the Sahel: Portrait of the 2014-2015 and 2015-2016 Transhumance [Internet]. 2017. Available from: http://www.inter-reseaux.org/IMG/pdf/afl_resilience_study_june2017_abridged_version.pdf

2. Apolloni A, Nicolas G, Coste C, El Mamy AB, Yahya B, El Arbi AS, et al. Towards the description of livestock mobility in Sahelian Africa: Some results from a survey in Mauritania. PLoS One. 2018;13(1):1–24.

3. Nicolas G, Apolloni A, Coste C, Wint GRW, Lancelot R, Gilbert M. Predictive gravity models of livestock mobility in Mauritania: The effects of supply, demand and cultural factors. PLoS One. 2018;13(7):1–21.

4. Dean AS, Fournié G, Kulo AE, Boukaya GA, Schelling E, Bonfoh B. Potential Risk of Regional Disease Spread in West Africa through Cross-Border Cattle Trade. PLoS One. 2013;8(10):1–9.

5. Motta P, Porphyre T, Handel I, Hamman SM, Ngu Ngwa V, Tanya V, et al. Implications of the cattle trade network in Cameroon for regional disease prevention and control. Sci Rep [Internet]. 2017 Apr 7;7(1):43932. Available from: http://www.nature.com/articles/srep43932

6. World Bank. Connecting Food Staples and Input Markets in West Africa - a Regional Trade Agenda for ECOWAS Countries [Internet]. Washington, DC; 2015. Report No.: 97279-AFR. Available from: https://openknowledge.worldbank.org/handle/10986/22276

7. OECD/SWAC. Accessibility and Infrastructure in Border Cities. Paris: OECD Publishing; 2019. (West African Papers).

8. Steck B. West Africa facing the lack of traffic lanes. EchoGéo 20. 2012;

9. Chaters GL, Johnson PCD, Cleaveland S, Crispell J, de Glanville WA, Doherty T, et al. Analysing livestock network data for infectious disease control: an argument for routine data collection in emerging economies. Philos Trans R Soc B Biol Sci [Internet]. Royal Society; 2019 Jul 8;374(1776):20180264. Available from: https://doi.org/10.1098/rstb.2018.0264

Submitted filename: Response to reviewers.docx

Click here for additional data file.

21 Apr 2020

Network analysis of regional livestock trade in West Africa

PONE-D-19-20212R1

Dear Dr. Valerio,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Peng Li, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The reviewer does not have further comments and recommends this manuscript to be published in this journal.

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

29 Apr 2020

PONE-D-19-20212R1

Network analysis of regional livestock trade in West Africa

Dear Dr. Valerio:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Peng Li

Academic Editor

PLOS ONE