Successful artificial reefs depend on getting the context right due to complex socio-bio-economic interactions

Artificial reefs (ARs) are one of the most popular means of supporting marine ecosystem conservation and coastal fisheries, particularly in developing countries. However, ARs generate complex socio-bio-economic interactions that require careful evaluation. This is particularly the case for ARs outside no-take zones, where fish might be subject to enhanced exploitation due to easier catchability. Here, we conducted an interdisciplinary study on how ARs impact fish and fishing yields, combining mathematical and sociological approaches. Both approaches converge to confirm that fishery yields decline when ARs are exploited as if they were open access areas. This situation typically occurs in areas with weak governance and/or high levels of illegal fishing activity, both of which are common in many developing countries. To avoid these adverse effects and their associated ecological consequences, we recommend prioritizing the onset of a long-term surveillance system against illegal fishing activities, and adapting design and location of the ARs based on both and local and academic knowledge, before the deployment of ARs.


b. The complete model description
Let n, and n2 respectively being the fish biomass in the MPA (where the ARs are deployed) and in the fishing area, as well as E1 and E2 the corresponding fishing efforts.
The evolution of the fish biomass and fishing effort is described by the following ordinary differential equations (eq. 2): With £« 1 is a small dimensionless parameter. T is the fast time.
Equations 2 are called the complete model as they take into account processes going on at different time scales. They are composed with two parts, a fast one and a slow one. The fast part relates to fish and fishermen migrations between the MPA and the fishing area while fish growth, landings and fishing effort (investment in fishing means) vary at the slow time t=£T. Slow terms of the complete models are those multiplied by e and correspond to fish growth, fishing mortality and fishing effort. For slow-fast models, we refer to the early work of Tikhonov (1952). Perfect and approximate methods of aggregation of variables were defined in (Iwasa, Andreasen, & Levin, 1987;Iwasa, Levin, & Andreasen, 1989 Where V1* and v2*=1-V1* represent the asymptotic spatial distribution of individuals respectively in zone 1 and 2 due to fast fish migration. They are given by the following expressions (eq. 5): (eq. 5) Similarly, for the fishing effort, we get the next expressions at the fast equilibrium: With (eq. 7) Where y corresponds to the fraction of the total fishing effort that is deployed in the MPA at the fast equilibrium. In other words it represents the share between legal and illegal fishing effort.

d. The aggregated model
The next step is to make an approximation by substituting the fast variables in terms of the fast equilibrium (n.', n2*' E1*, E2*) into the equations of the complete model. Adding fish and boat equations, and using slow time t, we obtain a reduced model, called the "aggregated model", which reads as follows (eq. 8): This "quick derivation method" is valid when the aggregated model is structurally stable, that is the case of the aggregated model used in this study (eq. 8

e. Analysis of the aggregated model
The aggregated model can be rewritten as follows (eq. 9): where Kag is the total carrying capacity of the fish population on the two zones, MPA and fishing, which is defined as follows: The aggregated model is of the same form as the classical Lotka-Volterra predator-prey model with prey logistic growth. Consequently, we know that there exist only three equilibria: (0, 0), (Kag,O) and (n*,E*) where: We know that when equilibrium (n*,E*) is positive, it is globally asymptotically stable in the positive quadrant. Under these conditions, we are going to study the sensitivity of the catch in the fishing area and in the MPA (illegal fishing) at equilibrium with respect to different values of V (AR volume) and y (the proportion of illegal fishing).
In figures S1, S2 and S3, we respectively plot the equilibrium catch in the fishing area (Y z * eq. 13), in the MPA (Yl* eq. 14), and the total catch (Yl* + YD. We compare four cases according to model assumptions on ARs productivity (oK = 0.1 ; 5) and attraction (eq. 14)

-Local knowledge: considering the actors point of view a. Context and study area
Venne and Bargny are two coastal fishing territories in Senegal in which artificial reef was deployed in 2004 in the frame of the Japanese collaboration (JICA). In January 2012, a sociologic survey carry out by the Senegalese fisheries and oceanographic center "CRODT" (Mbaye, 2012) in Venne and Bargny revealed that 95% and 85%, respectively, of the artificial reef area was "lethargic". The fact was that since the end of the JICA project that initiated the reef, monitoring, control and surveillance was no longer performed. The survey suggested that at the community level, the establishment of protected fishing areas and artificial reef areas requires strict monitoring to ensure compliance with management rules. As a result, the lack of monitoring capacity was identified as the main factor that can undermine incentives in MPAs and artificial reef areas.
The lack of cooperation from the fisheries administration was cited as the first factor in non-compliance with management measures. This factor was particularly mentioned in Venne with 75% of fishermen saying that since the end of the project with JICA, they have not received any support from the fisheries administration to continue monitoring as well as control and surveillance of the artificial reef area.
In 2014, we performed a second study, directly designed to collect the fishermen a) low production and attraction effect; b) strong production but low attraction effect; c) low production but strong attraction effect; d) strong production and attraction effects.
The colours correspond to different levels of illegal fishing effort over the ARs. From dark blue to orange respectively 0%, 5%, 10%, 25%, 50%, 75% and 100% of illegal fishing effort. oK: additional carrying capacity per unit of volume. Bo: Strength of the AR attraction effect from the MPA toward the ARs. Images from left to right correspond to an increasing a, and from top to bottom correspond to an increasing q. Blue, red, and yellow curves correspond to total catch, catch in the fishing area, and catch on the AR, respectively. The AR volume, production and attraction parameters were respectively set to 5, 0.5 and 200 rn". The other parameters were set to the value given in table

Data S1. (separate file)
Field report of the survey on artisanal fishermen perception on artificial reef effects on fisheries and ecosystem, including the questionnaire (translated from French).

Data S2. (separate file)
Extraction of the artisanal fishermen answers (converted in modalities) during the survey on artisanal fishermen perception on artificial reef effects on fisheries and ecosystem performed on June 2014 in Venne, Senegal (translated from French).