Page 7 - protected_areas

www.nature.com/scientificreports/

Nevertheless, these commitments can be mitigated in part by implementing SSFs within an MPA framework. If
successful, the benefits can be considerable for local coastal communities via increased revenue, they can achieve
conservation goals and finally, they can maintain profitable exploitation of SSF resources. Products coming from
successfully managed SSFs within MPAs would allow managers and policy-makers to satisfy the growing public
demand for responsible seafood consumption 11,52 .
Although SSF-MPA systems are highly complex, the large range of socio-ecological conditions we exam-
ined make it likely that our key attributes could prove beneficial to SSF in other geographical locations in the
Mediterranean Sea. Therefore, we can suggest that the allocation of some public expenditures from current fish-
56
eries subsidies (globally accounting for more than 30 billion US$/year ) to actions aimed at setting key attributes
in MPAs (e.g. effective patrolling, stakeholders capacity building) will produce substantial ecological, economic
and social benefits to society.
Methods
Database compilation. We use the term MPA sensu lato to define any marine area where human activities
are restricted for conservation and/or management purposes, and that generally embed no-take zones into buffer
zones. Furthermore, we made no distinction based on MPA legal status (e.g. national park, regional MPA, marine
reserve etc.). However because a considerable part of the information gathered for each MPA had to be provided
by MPA managers, we restricted our investigation to Mediterranean MPAs that have a management body. This
criterion excluded a large proportion of MPAs belonging to the Natura 2000 network although many are in the
19
process of establishing a management body. This lead to 153 potential MPAs . From the pool of possible MPAs,
we randomly selected 75, representing approximately half of all possible MPAs. Only half of the MPAs were
selected because of the effort needed to contact each MPA management body (i.e. multiple direct contacts via
e-mail and/or phone calls with MPAs’ managers). Also by selecting only half of the possible MPAs, we were able
to maximise the data gathered for each MPA.
Information about each MPA was obtained through multiple sources. These included: (1) questionnaires
emailed to MPAs managers and scientists, (2) review of international ISI scientific literature, (3) review of studies
published on a national/local level and (4) review of grey literature and unpublished studies (e.g. project report)
(see Supplementary Methods for a detailed description of data gathering procedure). Of the 75 randomly selected
MPAs, 34 MPAs replied to the questionnaire. We therefore retained these 34 in our study while all the others were
discarded due to a lack of critical information specific to the study.
At first we collected information on the largest number of attributes possible in order to thoroughly describe a
range of differing management and social situations. However some attributes were removed because exhaustive
data (e.g. about fishing effort, number of hours of surveillance, MPA funds for surveillance) could not be obtained
or they had low relative discriminating power among the MPAs. For instance, when identifying the “multiple
fishing gears allowed within the MPA”, the same score and/or category was attributable to more than 95% of the
MPAs considered. This procedure resulted in 20 attributes being included in the study (Supplementary Table 2).
Outcomes were considered as three response variable. They were: (1) ecological effectiveness, measured as an
increase in fish density or biomass as a result of the implementation of the MPA or when compared with open
access areas, (2) fishermen incomes, measured as an income increase as a result of the implementation of the MPA
or when compared with open access areas, and (3) fishermen environmental commitment, measured as their
commitment to MPA SSF management practices and participation to research and environmental programmes.
These three outcomes were defined on a binary scale denoting presence or absence (1 or 0 respectively) as done in
34
Gutiérrez et al. . This dichotomous coding scheme was chosen because the studies had differing techniques and
sampling schemes, and a lack of temporal series concerning the number of fines. These problems also prevented
the estimation of a response ratio for each of the three outcomes. However this coding schema is fully suitable to
identify the attributes significantly contributing to successful management of SSF in MPAs that represented the
aim of our study .
34
Ecological effectiveness and fishermen incomes were based on a review of the scientific and grey literature
as well as reports from the management bodies of the MPA. In a few cases the status was determined from our
own unpublished results. Fishermen environmental commitment was determined by using information provided
through questionnaires. For a detailed description of the rationale behind each outcome see the Supplementary
Methods.
Information concerning at least two outcomes was missing from six MPAs and a further three forbid SSFs.
These MPAs were removed from the analysis leading to a total of 25 MPAs being investigated. Based on a sensi-
tivity analyses, reported in Supplementary Methods and Supplementary Table S3, we assessed that our sample
size (i.e. n = the number of marine protected areas included in the present study) represented a replication level
adequate to provide reliable estimation of the relevance of the attributes considered in determining overall success
(see Supplementary Methods and Supplementary Table S3).
Finally, data were compiled for 23 variables (20 attributes and 3 outcomes) for each of the 25 MPAs where SSF
were allowed within their boundaries and for which evidences about at least 2 of the 3 outcomes were available.
To assess the potential effect on our results of misreporting related to outcome and/or oversights of use-
ful literature, we performed a sensitivity analysis. This analysis showed that our data were highly robust
to moderate miscoding of the three outcomes (see Supplementary Methods, Supplementary Table S4 and
Supplementary Figures S7–8).
Statistical analyses. This data matrix contained 575 cells of which 3.6% of the attributes and 6.7% of the
57
outcomes contained missing data. The missing data were compensated via missForest , an iterative imputation
method based on a random forest that can successfully impute missing values. The method uses multicollinearity

Scientific RepoRts | 6:38135 | DOI: 10.1038/srep38135 7