Sei sulla pagina 1di 12

Zoning Marine Protected Areas through Spatial

Multiple-Criteria Analysis: the Case of the Asinara


Island National Marine Reserve of Italy
FERDINANDO VILLA,* LEONARDO TUNESI, AND TUNDI AGARDY
*Institute for Ecological Economics, University of Maryland, P.O. Box 38, Solomons, MD 206880038, U.S.A.,
email villa@cbl.umces.edu
Istituto Centrale per la Ricerca Applicata al Mare, Via di Casalotti 300, Roma, Italy
Conservation International, 2501 M Street NW, Washington, D.C. 20037, U.S.A.

Abstract: As the role of marine protected areas as conservation tools becomes better understood and more
sophisticated, their planning becomes more complicated. Systematic, objective approaches to site selection
and design can help reconcile conflicting interests, represent stakeholders viewpoints fairly and evenly, and
extend the scope of planning studies from single reserves to networks. We illustrate the use of spatial multiplecriteria analysis for determining the suitability of marine areas for different uses and levels of protection.
This technique couples geographic information systems (GIS) for land assessment and evaluation with a formal statement of the design priorities as seen from the different viewpoints of all involved stakeholders. The
planning process, while staying focused on the main purposes of conservation and feasibility, involves all the
main interest groups in the definition of priorities so that conflicts and tensions are kept under control. We
used multiple-criteria analysis to integrate objective data with the contrasting priorities of different stakeholders in the planning of a marine protected area. The results of the analysis can be used to define an optimal spatial arrangement of different protection levels. As a case study, we developed a zoning plan for one of
the first marine protected areas in Italy, the Asinara Island National Marine Reserve.
Zonificacin de reas Marinas Protegidas Mediante el Anlisis de Criterios Espaciales Mltiples: el Caso de la
Reserva Nacional Marina de la Isla Asinara en Italia
Resumen: Puesto que el papel de las reas marinas protegidas est siendo mejor entendido y se est volviendo ms sofisticado, la planeacin para abordar eficientemente estas reas se est volviendo ms complicada para las personas que toman decisiones. Las metodologas sistemticas y objetivas para la seleccin de
sitios y el diseo de reservas pueden ayudar a reconciliar los conflictos de intereses, representar los puntos de
vista de los usuarios de manera equitativa y balanceada y extender la dimensin de los estudios de planeacin para reservas individuales o en redes. Ilustramos el uso de un anlisis de criterios espaciales mltiples para determinar la viabilidad de reas marinas para diferentes usos y niveles de proteccin. Esta tcnica
une sistemas de informacin geogrfica (GIS) para estimacin y evaluacin de suelos con una declaracin
de prioridades de diseo tal y como es percibida por los diferentes usuarios involucrados. El proceso de planeacin, al mismo tiempo que se enfoca en los propsitos principales de la conservacin y en su viabilidad,
involucra a los principales grupos interesados en la definicin de prioridades de tal manera que los conflictos y tensiones pueden ser manejadas. Utilizamos el anlisis de criterios espaciales mltiples para integrar
datos objetivos con las prioridades contrastantes de los diferentes usuarios en la planeacin de un rea marina protegida. Los resultados del anlisis pueden ser usados como una gua para definir arreglos espaciales
ptimos con diferentes niveles de proteccin. Como caso de estudio desarrollamos un plan de desarrollo de
zonificacin para una de las reas marinas protegidas de Italia, la Reserva Nacional Marina de la Isla Asinara.

Paper submitted October 3, 2000; revised manuscript accepted May 16, 2001.

515
Conservation Biology, Pages 515526
Volume 16, No. 2, April 2002

516

Spatial Multiple-Criteria Analysis for MPA Zoning

Introduction
Marine protected areas (MPAs) fulfill three key functions
in modern conservation: conserving marine biodiversity,
maintaining productivity, and contributing to economic
and social welfare ( United Nations Environmental Program 1995; McManus et al. 1998). Increasingly, MPAs are
being used to support other traditional forms of marine
resource management, such as fisheries management,
where these methods have proved ineffective ( Agardy
2000a). In addition, MPAs are used to hedge against management uncertainty and the dynamic conditions of marine ecosystems, providing a buffer against management
error or unforeseen declines in environmental quality and
marine production (Dayton et al. 2000).
As the use of this conservation tool has become more
sophisticated, managers have begun to realize that systematic approaches to MPA site selection and design are
crucial to deriving maximum benefits. But increasing sophistication has also increased the probability that protected-area planning will become too complicated for
decision-makers to address effectively. In such cases,
goals are often unclear and conflicting interests are not
fairly and evenly represented. Systematic and objective
approaches are needed to help overcome such pitfalls.
When MPAs are designed systematically, they confer
not only the benefits listed above, but alsoperhaps
even more importantlyserve as demonstration models
for integrating management priorities with multiple
stakeholder needs (Agardy 1997). Systematic approaches to designing and implementing MPAs on national or even regional scales also foster the formation of
networks that allow large-scale tracking of environmental conditions and conservation of whole ecosystems or
even regions.
Traditionally, planning of MPAs has been based largely
on common sense. The rationale for choosing an area often is more political than scientific, and lack of time, funds,
and data are typical of the process. Scientists are only now
beginning to weigh in with clearly defined criteria for
site selection (e.g., Roberts 2000), but decision-makers are
still largely guided by highly simplistic approaches (e.g.,
Kelleher et al. 1995). A repeatable, rigorous approach
that can be applied systematically to any number of MPAs
in a network is needed to define optimal sites and optimal zoning within those sites. The success of an MPA depends on the level of local involvement, from planning
to management, which in turn depends on local awareness, the development of which must be a priority beginning in the design stage (Kelleher & Recchia 1998).
The case of the Italian MPA system provides a good illustration of all the problems and needs mentioned above.
Despite long-term interest by Italian environmental organizations and the high value and uniqueness of the Mediterranean marine environment, the implementation of
Italian MPAs has been seriously delayed and is only in an

Conservation Biology
Volume 16, No. 2, April 2002

Villa et al.

initial stage. Out of 49 MPAs planned, only 15 have been


created, and the number of functioning MPAs is much
smaller. Because of the delays in instituting MPAs, longterm data and study cases are not available to help plan
new protected areas. Traditionally, enforcement and
control of environmental restrictions have been suboptimal. Current legislation ( Italian Law 1991) stipulates a
number of sites that require protection and creates a
mechanism through which governmental consulting
agencies can design reserves in other environmentally
important areas.
Through the leading marine research agency of the
Italian government, Istituto Centrale per la Ricerca Applicata al Mare (ICRAM ), we were charged by the Italian
Ministry of the Environment to study the natural and socioeconomic environments of two Italian coastal areas
and formulate the zoning proposal for the relative MPAs.
To catalyze the implementation of MPAs in Italy, we focused our activities on designing MPAs with multipleuse zoning and on creating proposals for management of
these MPAs. This effort has resulted in a set of techniques and guidelines to define data-collection standards, priorities, and analytical methods to guide the development of future MPAs. The methods we developed
to define the zoning plan, defined here as the spatial
structuring of different protection levels in the area, are
the focus of this paper and are illustrated by the case
study of the Asinara Island National Marine Reserve.

Italian MPA Design Requirements


Italian MPAs are conventionally implemented according
to three protection levels (Tunesi & Diviacco 1993). The
existing law explicitly suggests the following levels, which
include both open sea and coastal areas without distinction.
(1)

(2)

(3)

Integral reserve: only authorized personnel are allowed access for monitoring, research, and maintenance.
General reserve: low-impact tourism is allowed
(self-guided trails allow educational activities; boating, swimming, snorkeling, fishing, and scuba diving are allowed with restrictions on numbers, size,
and types of boats and on fishing techniques).
Partial reserve: usually a buffer between the exterior of the park and more restrictively protected
zones, usually containing the parks administrative
and educational facilities, and usually allowing restricted recreational navigation and some sporting
and commercial fishing activities.

The structure suggested by the law lacks an important


protection level at the core of the success of many international MPAs, one in which public access is allowed
but nothing in the natural environment can be taken or

Villa et al.

Spatial Multiple-Criteria Analysis for MPA Zoning

altered. Such protection is necessary to allow visitors


the full experience of a marine reserve with a healthy
distribution of species and sizes within populations
(Tunesi 1994 ). This access is fundamental to attracting
public attention through recreational scuba diving and
snorkeling, developing environmental awareness in visitors, and ultimately deciding the success of the MPA. For
these reasons we proposed an additional protection
level, splitting the integral reserve into two categories:
(1) no-entry, no-take zones and (2 ) entry, no-take zones
(Table 1).
The zoning plan for any MPA must take into account
the need for biodiversity conservation and the sustainable use of environmental resources. This is achieved
through an objective balance between comprehensive
resource protection and multiple, compatible uses of resources (Agardy 2000b ). Zoning options should be presented clearly to stakeholders representing different interests in the area (Agardy 1995; Salm & Clark 1999). For
example, a zoning plan that optimizes the potential for
fisheries production will satisfy fisheries interests but
will not necessarily address the needs of tourism or safeguard the broader biodiversity of the area. Articulating
various zoning options with various potential costs and
benefits to all user groups is the most transparent
method of planning a reserve and the one most likely to
receive sustained public support.
In reality, most zoning is arbitrary and often reflects or
favors the interests of a single stakeholder. This may not
be obvious to politicians who make the final decision to
adopt a zoning plan and implement a reserve. To avoid
arbitrariness and potential unfairness, we propose an objective system for creating and evaluating zoning options. We chose to use spatial multiple-criteria analysis

Table 1.

517

(SMCA ) for determining the suitability of various marine


areas for different uses and levels of protection. This
technique couples land assessment based on a geographic
information system ( GIS ) with a formal statement and
analysis of design priorities for each protection level to
create a standardized procedure. The technique ensures
that all-important planning aspects are considered and that
all stakeholders are allowed maximum involvement in the
definition of priorities.

Methods
Multiple-criteria analysis (MCA) encompasses a rich and
diverse set of techniques that are widely used in fields
from economic analysis to environmental impact assessment. Multiple-criteria evaluation methods (Cochrane &
Zeleny 1973; Voogd 1983; Nijkamp et al. 1990) have assisted urban and regional planners, allowing them to
make objectively informed choices and to consider social preferences, development needs, and conservation
requirements. A fundamental technique in MCA is concordance/discordance analysis: a set of observations, described as a set of measured attributes, is ranked according to a concordance (or discordance) score. Scores are
computed by combining objective measures with sets of
priority weights that express the importance of each
attribute within a particular scenario. It is common practice to compare scenarios on the basis of concordance
scores computed from the quantitative description of an
existing (or planned ) situation. One of the strengths of
MCA is the ease with which heterogeneous information
can be combined. Quantitative measures can be used
along with semiquantitative information and ranks;

Activities allowed for each planned level of protection.*

Category

Activity

Research

nondestructive
monitoring
sailing,
motor boating,
swimming
anchoring,
mooring
diving,
guided tours,
recreational
fishing
artisanal,
sport,
scuba, and
commercial
fishing

Sea access
Staying
Recreation

Exploitation

No-entry,
no-take

Entry,
no-take

General
reserve

Partial
reserve

P
P
P
P
P
P
P

R
P
P
P
R
R
R

G
R
G
R
A
A
A

G
R
G
R
G
G
G

P
P
P
P
P

P
P
P
P
P

R
R
P
P
P

G
R
R
P
P

*Abbreviations: A, allowed upon authorization; G, allowed without authorization; P, prohibited; R, subject to specific limitations.

Conservation Biology
Volume 16, No. 2, April 2002

518

Spatial Multiple-Criteria Analysis for MPA Zoning

physical measures and the results of surveys are commonly used together with no need for special data preprocessing.
Conventional MCA has no explicit notion of geographical space. To make it suitable to many environmental
applications, an extension to a spatially explicit context
is necessary. Villa et al. ( 1996 ) developed a spatialized
MCA ( SMCA ) that can be used to assess the concordance of an existing or planned land configuration with
a set of stated objectives. The technique produces maps
of the concordance of an areas characteristics with one
or more land-use scenarios. The latter are expressed as a
numerical ranking of the priority or relative importance
of each classification variable. To apply the method, the
following information is needed: ( 1 ) Quantitative or
semiquantitative (ranks) measures for each variable under consideration. In a spatial context, this usually corresponds to a GIS coverage describing that variable. ( 2 )
Importance or priority values for each variable, in quantitative or semiquantitative form. The magnitude of the
weights expresses the relative importance of each variable, and their sign expresses the negative or positive influence of the variable within the specified scenario.
Herein we use the term scenario interchangeably with
priority weights vector to refer to the formal statement of priorities for each variable under a particular
viewpoint.
Selection and Definition of Characteristic Variables
The unit of analysis in SMCA, or evaluation unit, corresponds to a unique configuration of variable values
found in the source data set. Figure 1 is an overview of
the process that leads to identifying the evaluation units
in a simplified data set. As shown, the information contained in the initial set of variables is processed to identify all areas where a unique combination of variable values exists. These areas are tagged, and the vector of
variable values in each one is recorded. For each set of
priority weights ( scenario), concordance and discordance scores are calculated in each evaluation unit. The
final concordance and discordance maps are produced
for each scenario by mapping the concordance values
obtained on the corresponding units spatial extent.
The choice of variables, of course, is critical to the
success of the study. Ideally, the set of variables should
have as little internal correlation as possible, and the
variables should be suitable for direct comparison with
one another. This usually means defining a uniform conceptual framework to apply to the whole variable set. As
an example, all the variables could express a notion of
value. As exemplified later, the processing capabilities
of GIS software are instrumental in performing the necessary transformations of existing mapped attributes and
in carrying on the necessary data reduction (such as reclassification and ranking).

Conservation Biology
Volume 16, No. 2, April 2002

Villa et al.

There is no particular restriction or statistical constraint


on the kind of variables suitable for MCA. Quantitative
measures can be used along with ranks or semiquantitative appraisal scores that refer to a scale that need not be
linear but must be monotonic. Verbal rankings such as
bad, good, and very good are as acceptable as precise quantitative measures and can be freely mixed with
them.
Definition of Priority Scenarios
The statement of the objectives is the second critical
step in the planning process. The outcome is a set of one
or more priority vectors or scenarios, one for each viewpoint, that assigns relative importance values (weights) to
each variable considered. Numeric weights or importance rankings can be assigned directly through collaborative brainstorming or can be calculated based on pairwise comparison matrices. In the latter case, weights are
extracted from a matrix specifying relative priorities for
each possible pair of variables. This approach, described
as a phase of the analytical hierarchy process ( Saaty
1980), is considered the most useful and informative.
Each element aij of the comparison matrix expresses
the importance of feature i with respect to feature j
within the chosen scenario. The importance value is
usually a real number, with the value 1 indicating that
the two features are equally important. Other values are
assigned according to a chosen rating scale. The matrix
is reciprocally symmetric ( aij  1/aji ). Thus, only half of
the matrix needs to be filled in by the planner. In the
study, we used the scale suggested by Eastman et al.
( 1993 ), in which the value 9 expresses maximum importance for feature j over i, and, conversely, 1/9 expresses complete dominance of i over j.
After the entire matrix has been compiled, the actual
priority weights are computed by extracting its eigenvalues and eigenvectors. The normalized eigenvector associated with the dominant eigenvalue is the required
weight vector. This simple process, described in detail
by Saaty (1980), can be easily automated into an interactive, self-explanatory computer program. Thus, the planner does not have to know the mathematical details. In
addition to the weights, a program can calculate consistency indexes that identify the amount of internal contradiction in the priority structure. Such indexes often
provide important feedback with which to refine the
priorities and better understand the problem.
Compiling the comparison matrix is a useful exercise
because it usually forces the experimenter or decisionmaker into an in-depth investigation of the nature of the
objectives. Expressing priorities as numbers forces the
different stakeholders to speak the same language and at
the same time provides some distance from the core of
the decision process that makes it easier to come to
agreement. For example, when more decision-makers

Villa et al.

Spatial Multiple-Criteria Analysis for MPA Zoning

519

Figure 1. Graphical outline of the


steps involved in computing a concordance map in a simplified case
with two variables ( VAR) (A and B).
Variables A and B can have the values 1 or 2. Evaluation units are
identified as the areas with the four
possible configurations of (A,B)
as follows: a  ( 1,1), b  (2,1), c 
(1,2), and d  (2,2).

are involved, the sets of priority weights obtained from


each involved party can be used to reduce the dimensionality of the decision space through multivariate statistics ( Villa et al. 1996 ). Such analysis can exactly quantify the amount and structure of the disagreement in a
group, identifying the main lines of thought and helping
with the resolution of conflicts and the identification of
(often unsuspected ) common endeavors.
Computation of Concordance Maps
Concordance scores are calculated for each evaluation
unit under the viewpoint expressed by each priority
vector ( Fig. 1 ). The mathematical process that leads
from the variable values and priority vectors to the concordance score relative to each evaluation unit is detailed in the Appendix. The scores, standardized as
needed, are mapped back on the evaluation units to create a concordance map for each scenario, which shows
pictorially the agreement between the priorities specified and the features of the area under consideration. All
these operations are performed by specialized software.
In typical applications, one concordance map is obtained for each scenario. The maps can be aggregated
later and further analyzed with GIS algorithms.

Variable values and priority weights can be expressed


in a quantitative form (real numbers or integers) or in a
semiquantitative form (e.g., by means of verbal scales or
signs [, ]). Semiquantitative notation conveys only
the order of dominance, as opposed to actual measurements, of a variables values, relaxing all assumptions of
linearity in the measure of the characteristic of interest.
This is convenient in many real-world cases, in which
available information is categorical, incomplete, or hard
to quantify. Quantitative and semiquantitative measures
are integrated into a single dominance score according
to the EVAMIX approach developed by Voogd ( 1983 ).
The Appendix details the mathematical procedure for
both the quantitative and semiquantitative cases.

Zoning the Asinara Marine Protected Area


We used SMCA to evaluate the concordance between
the Asinara reserve area and each of the four planned
protection levels. We used concordance maps as a guide
to inform, document, and justify the proposed zoning of
the MPA. In doing so, we tried to maximize the generality of the procedure. One of our goals was to develop a
repeatable procedure to give consistency and some ob-

Conservation Biology
Volume 16, No. 2, April 2002

520

Spatial Multiple-Criteria Analysis for MPA Zoning

jectivity to the planning process of the remaining Italian


MPAs. The goal of generality made it necessary to take
particular care in selecting and organizing the data
needed as a base for the evaluations. The assignment of
priority weights to the variables for the different protection scenarios was done by ICRAM staff, who summarized and translated in pairwise notation the outcome of
a number of planning sessions held with all the involved
stakeholders.
The Study Area
The planned Asinara National Marine Reserve surrounds
Asinara island, a 50.9-km 2 island separated from the
northwestern Sardinian coast by a narrow (1-km) channel bisected by the small island of Piana. Asinara was a
prison island from 1885 to 1997, so the island has had a
quieter history than the other overexploited coastlines
of Sardinia and indeed most of the Mediterranean. Public
access and construction have been forbidden for nearly
a century. This, despite the lack of environmental concern of the small resident population, has allowed the
maintenance of peculiar Mediterranean flora and fauna
and has prevented serious damage to the coastal benthic
and marine assemblages, as opposed to that of the offshore marine environment.
Data Collection and Organization
An SMCA is based on the ranking of variables according
to relative importance. It is thus necessary that the information carried by the variables is clear enough to enable
a decision-maker to unambiguously state a preference.
Raw data, such as direct measurements from surveys or
remote sensing, typically do not meet this need. It
would be confusing to rank, for example, the importance of the potential for scuba diving directly against
the suitability of an area for settling of a rare bird species. Moreover, direct comparison of raw information is
more likely to be interpreted differently by decisionmakers of different backgrounds, creating uncontrollable bias in the final weights. For this reason, raw data are
usually aggregated into variables whose comparison is as
intuitive and unambiguous as possible.
The selection of these variables is a particularly critical
phase. The final set of variables needs to be expressive
enough to capture the most important attributes of the
area as understood by all involved stakeholders, yet also
general (to serve as an example for application to different environments) and simple (to be applied in conditions of data scarcity). Furthermore, it is important that
the data and their aggregation be understandable to a
wide, technically untrained audience.
To meet our objectives of generality and efficiency,
we developed a hierarchically structured knowledge
model in which higher-level variables describe a more

Conservation Biology
Volume 16, No. 2, April 2002

Villa et al.

abstract view of the system, suitable to representing


general MPA characteristics, and the peculiarities of individual sites are incorporated at the lower levels. This
scheme also allows the dimensionality of the system description to be reduced to an optimal size for communicating and understanding the results. The high-level variables need to be (1) expressive and complete enough to
give a full picture of the system; (2) consistent enough
to minimize ambiguity during assignment of priorities;
( 3 ) reflective of the characteristics of a generic system
and not tied to environmental features that are shown
by only some of the systems under evaluation; (4 ) suitable for unambiguous spatial representation; and (5) related clearly and consistently to raw data, maintaining
enough redundancy to allow their evaluation in conditions of data scarcity and to accommodate the environmental peculiarities of different sites.
To meet these goals, we aggregated the available data
(Table 2) into five higher-level variables, whose comparison is made simple by the fact that they all express a
value related to environmental, economic, and social influences on the area. The aggregation procedure resulted in the five spatial coverages described below.
The map of the natural value of the marine environment ( NVC map ) aggregates natural values related to
species diversity and size distribution within populations in the benthic and aquatic communities, the presence of endemic or rare species, and the presence and
status of conservation of habitats that have crucial roles
in maintaining ecosystem function (e.g., nursery areas).
The map was obtained by GIS addition of properly reclassified biocenotic and habitat maps for the most important habitats and species.
The map of the natural value of the coastal environment (NVC map) was obtained by aggregating information for important coastal endemic species on the suitability of habitats for the return or reintroduction of key
species, such as the sea turtle (Caretta caretta) or the
Monk seal (Monachus monachus) and on the ability of
the coastal habitat to support key species that nest on
the mainland, such as Adounis gull ( Larus adouinii ).
We performed the aggregation by applying SMCA to the
raw information, giving separate consideration to natural-value elements directly linked to the sea floor, such
as the distribution of benthic assemblages, and elements
associated with the coastal environment, such as the potential occurrence of the Monk seal. Table 3 summarizes
the variables and the relative weights used to obtain the
value map. We performed the weighting on the basis of
the results of field research, considering the real presence of a species and its status in the area. In the Mediterranean Sea, for instance, the Monk seal population is
more endangered than the gastropod Patella ferruginea,
but the Monk seal is absent along the coasts of Asinara
and the value assigned to this species relates only to the
potential value of a particular stretch of coast as host to

Villa et al.

Spatial Multiple-Criteria Analysis for MPA Zoning

Table 2. Spatial coverages used as raw data for the spatial


multiple-criteria analysis.
Mapped variable
Geology and geomorphology
Benthic assemblages
Diversity of the fish community
(no. of species)
Nursery areas for fish populations
Sites relevant for the biological cycle of
Patella ferruginea
Sites relevant for the biological cycle of
Caretta caretta
Sites relevant in the biological cycle of
marine cetaceans
Potential for transit and settling of
Foca monaca
Hatching sites for marine birds
Interest for archeology based on
presence of archeological sites or
knowledge of previous settlements
Interest for scientific research
and education
Suitability for traditional fishing
techniques based on presence
of installations or existence of
traditional fishing sites
Suitability for commercial fishing
Suitability for aquaculture
Suitability for scuba diving
Suitability for snorkeling
Suitability for whale and dolphin
watching
Suitability for recreational fishing
Suitability for sailing, recreational boating
Suitability for swimming
Density of commercial navigation and
commercial harbors
Industrial installations and power plants
Tourist infrastructure
(e.g., hotels, camping sites)
Tourist harbors
Input of pollutants from urban,
industrial sources, and rivers
Density and severity of acoustic and
other forms of pollution
Areas subject to military control

Source*
DC
DC
DC
DC
DC
DC
DC
DC
DC
DC, S
DC

S
DC, S
DC
DC, S
DC
DC
DC, S
DC, S
DC, S
DC
DC
DC
DC, S
DC
DC
DC

*Abbreviations: DC, data collection in the field; S, information provided by stakeholders.

an errant specimen. In contrast, the gastropod is still living in small populations on the rocky shores of the Asinara Island.
The map of the value of an area for recreational activities (RAV map) was obtained by attributing relative importance values to each variable involved and performing an SMCA to characterize the value as joint
concordance of the area characteristics with the suitability for each feature. We considered suitability for all fundamental recreational and cultural activities ( Table 3,
which also summarizes the weighting used ). The final
value map was obtained from the results of SMCA after

521

weighting with the accessibility of the area. The map of


the value of an area for commercial exploitation of resources (CRV map) considered only the allowed fishing
activities relative to traditional and artisan fishing practices and was prepared by addition of maps identifying
traditional fishing sites and an areas general suitability
for such practices.
The map of the degree of accessibility of area (ease of
access [EAC] map) was used both as a benefit value for
scenarios in which access is allowed and encouraged and
as a cost factor in high-protection scenarios. It was obtained through layering and distance buffering of maps
identifying marine access routes and existing harbors.
Three of the aggregated value maps for the Asinara
MPA are shown in Fig. 1. Three aggregated maps (NVM,
CRV,EAC ) were obtained from the raw data through
simple GISbased aggregation. The other two ( NVC,
RAV ) were obtained by performing an SMCA, ranking
each raw variable in order of its importance in defining
the value, and calculating the actual value as the concordance of the variables configuration with corresponding weights.
Three of the aggregated value maps for the Asinara
MPA are shown in Fig. 2. Three aggregated maps (NVM,
CRV, EAC ) were obtained from the raw data through
simple GIS-based aggregation. The other two ( NVC, RAV )
were obtained by performing a SMCA analysis, ranking
each raw variable in order of its importance in defining
the value, and calculating the actual value as the concordance of the variables configuration with the corresponding weights.
Not all the required raw information ( Table 2 ) was
readily available. Specific campaigns, field activities, and
meetings with stakeholders representing the main socioeconomic categories with interests in the MPA (current
or prospective ) had to be carried out to obtain it. The
aim of this work was also to improve the value of this information and to evaluate the possibility of applying this
kind of approach on a national level for future MPAs in
Italy. We were careful to collect data sets that we considered potentially significant for characterizing the environmental and the socioeconomic components of any
new marine protected area. We also tried to identify a
set of raw data compatible with the temporal and financial constraints that face decision-makers, planners, and
managers of Italian MPAs. Such constraints are not
unique to Italy, of course, so this system can serve as a
reference for other countries attempting to develop strategic systems of MPAs.
It is important to ensure that the raw data collected do
not already contain prejudgment or bias, because they
must form the basis of a multiple-objective evaluation.
For example, variables such as suitability for commercial fishing consider only the size and quality of the harvested fish and ignore the fact that fish populations
might be particularly vulnerable if caught in a particular

Conservation Biology
Volume 16, No. 2, April 2002

522

Spatial Multiple-Criteria Analysis for MPA Zoning

Villa et al.

Table 3. Weights used in aggregating the raw variables through spatial multiple-criteria analysis to obtain maps of natural coastal value (left)
and recreational value (right).
Variable
Abundance of Patella ferruginea
Sightings of coastal cetaceans
Suitability for Monachus monachus
Presence or suitability for key
avian species (Larus adouinii)

Weight

Variable

Weight

420
149
46

Archeological sites
Suitability for recreational diving
Suitability for snorkeling
Potential for dolphin watching
Suitability for recreational navigation
Suitability for swimming

26
191
229
203
104
247

384

area. Unbiased protocols for collection and use of such


information were prepared and incorporated in an extensive set of guidelines to be made available to MPA
planners in future studies. Such guidelines are the basis
of correct ranking and weighting of variables for priority-setting and evaluation.
Selection of Priorities
The definition of priority weights for the value maps
under each protection scenario is the key phase of the
evaluation, where the viewpoints of all involved decision-makers are incorporated. In the Asinara case, local
stakeholders were not directly involved in the matrix
construction but were interviewed personally on several
occasions. We spent a week in 1998 and a week in 1999
meeting with all the parties involved in or concerned
with the establishment and management of the Asinara
reserve. We met with representatives of the municipality on two occasions and conducted several interviews
and field trips with local environmentalists, marine scientists, ornithologists, geologists, and economists. Based
on their opinions and our own judgment, we prepared
pairwise comparison matrices to identify priorities relative to the four protection levels. The matrices were
shown and explained to each group involved, and the
priorities were refined based on their feedback. The
mathematical analysis of the matrix was not done until
the pairwise comparisons were finalized. Because the final outcome of the analysis is not obvious from the pair-

wise comparisons (which was stressed to each group),


it was easier for each involved party to concentrate on
the problem rather than the advantage that any particular outcome could have given their respective causes.
We prepared pairwise comparison matrices between
the five high-level variables and calculated the corresponding weights relative to each protection scenario
(Table 4 ). We also calculated inconsistency indexes for
all matrices. The highest value obtained was 0.06, which
is considered more than adequate (common MCA practice sets a threshold of 0.1 for maximum allowed inconsistency). The concordance maps for each protection
scenario (Fig. 3) were calculated with a module for the
GRASS geographical information system software ( U.S.
Army Construction Engineering Research Laboratory
1994; this program was written and is available free
from F.V.). In all maps, the values of concordance (ranging from 0 to 1) have been distributed over 255 levels.
Development of the Zoning Plan
Several considerations and priorities influenced development of the zoning plan. In preparing the initial zoning,
we followed three fundamental guidelines: (1) Maximum
importance should be given to the outcome of the concordance analysis. In particular, the indications of suitable
sites for no-entry, no-take, and entry, no-take zones should
be followed as closely as possible. (2) Excessive articulation in zoning should be avoided to maximize clarity, efficiency, and ease of management and enforcement. (3)

Figure 2. Aggregated maps of the


natural value of the marine environment ( NVM), the natural value
of the coastal environment ( NVC),
and the recreational value ( RAV )
used in the final concordance analysis. Lighter shades of gray stand
for higher value. The central mainland area is not included in the
analysis.

Conservation Biology
Volume 16, No. 2, April 2002

Villa et al.

Spatial Multiple-Criteria Analysis for MPA Zoning

Table 4. Pairwise matrices comparing variable values under the


viewpoints expressed in each protection scenario.*
NVC CRV RAV EAC Weight NVC CRV RAV EAC Weight
noentry, no-take
NVM
NVC
CRV
RAV
EAC

9
9

9
9
1

9
3
1
1

entry, notake
377
455
51
51
67

general reserve
NVM
NVC
CRV
RAV
EAC

1/2 1/2
1/2 1/2
1/2

1
1
1
1

9
9

5
5
3
5
1/9 1/5
1

399
351
27
126
96

partial reserve
149
171
225
259
196

1/7 1/9 1/5


1/7 1/7 1/5
1
1
1

45
47
309
327
272

*Abbreviations: NVC, natural value of coastal environment; CRV,


value for commercial exploitation; RAV, recreational value; EAC, accessibility and potential disturbance; NVM, natural value of marine
environment.

Views of stakeholders should be considered to reduce


remaining conflicts over use to a minimum.
Guided by these criteria, we developed a preliminary
zoning proposal that was later presented to the stakeholder group in a series of meetings. Based on the feedback received, we devised a final zoning proposal (Fig.
4) for the Asinara Island National Marine Reserve. The
protection areas were articulated as follows:
One no-entry, no-take zone was located in the northern and distant part of the island. Its location was chosen and justified based on the results of the concordance analysis, reflecting the higher priority given to
biological importance and relative remoteness.
Four different entry, no-take zones were chosen to
protect more specific natural values such as endangered species. The priority values elaborated for the
entry, no-take level were similar to those elaborated
for the no-entry, no-take level (Table 4). Selection of
these areas was based mostly on their biological value
and easy access so that these areas could lend themselves to noninvasive, controlled ecotourism.
Two general-reserve zones, one for each side of the island, were devoted to protection of sensitive coastal
benthic assemblages, such as seagrass meadows, that
suffer relatively little effect from the activities allowed
at the general-reserve level.
One partial-reserve zone was designated to serve as a
buffer zone for the western and northern parts of the
MPA. This zone showed good concordance with the
objectives of a partial reserve as shown by the results
of the SMCA. Such areas were arranged to maximize
the compatibility of traditional fishing practices with
conservation needs.

523

The zoning proposal also identified three channels situated to provide maximum access with as little environmental disturbance as possible. In the end, the concordance maps were helpful in highlighting the location of
the main areas of the island and particularly in identifying the areas most suitable to act as no-take reserves that
would be compatible with stakeholders priorities and
needs. Compromises were often reached on the basis of
the SMCA results. For example, the location of entry, notake zones within the general reserve area in the southeast sector was difficult to determine because of the
presence of traditional fishing sites that conflicted with
the need to protect the essential function of seagrass
meadows. The collaborative process, centered around
easily understandable concordance maps, was a great asset in minimizing conflict and in keeping the groups attention focused on the core issues. On the other hand,
the decision analysis highlighted little discrimination between some of the alternatives. For example, the
weights in Table 1 (and, consequently, the concordance
maps in Fig. 3 ) were not different enough to help discriminate between no entry, no-take and entry, no-take
scenarios, and zoning decisions were guided by common sense. Although this might well be due to uncertainties in the decision-makers priorities, such cases
should usually prompt a review of the variables used,
with the aim of maximizing their discriminating power
for the problem under study.

Conclusions and Future Developments


The development of MPA zoning plans has traditionally
been a highly specific process, with little standardization of methods and procedure. This is due in part to the
great variability in the function of MPAs in each specific
geographical and social context and to differences in the
problems planners and managers face at each prospective site. We therefore do not mean to imply that the use
of SMCA is anything more than a general guide to reaching a satisfactory zoning plan. Nevertheless, we also recognize several benefits of identifying a set of procedures
to ensure as much objectivity and standardization as possible. In our experience, these benefits are maximized if
the following general guidelines are adhered to:
Phases of analysis are comprehensive and undertaken
in the correct sequence.
Planning is undertaken to help solve conflicts in opinion among decision-makers. With a formal statement
of priorities, the components of conflict related to language and personality are minimized; our experience
has shown that these components usually play a major
role in most disagreements.
Information is translated into a formal notation to facilitate understanding of the areas environmental sta-

Conservation Biology
Volume 16, No. 2, April 2002

524

Spatial Multiple-Criteria Analysis for MPA Zoning

Villa et al.

Figure 3. Concordance maps for


the four protection scenarios corresponding to protection levels (A1)
no-entry, no-take; (A2) entry, notake; ( B) general reserve; and (C)
partial reserve. Concordance values
range from 0 (black) to 1 (white).
The central mainland area is not
included in the analysis.
tus and priorities. In particular, the pairwise comparison phase forces the involved decision-makers to
identify and resolve doubts, uncertainties, and inconsistencies. The use of inconsistency measures produces precious feedback that ultimately results in a
much clearer understanding of the problem and ones
own priorities.
Objectivity is obvious. An important side-effect of
leaving the final decision on priorities to a mathematical algorithm (the extraction of eigenvalues of the
pairwise comparison matrix) is that malicious statements cannot be justified. It is hard to manipulate pairwise priorities to influence the outcome to suit ones
interests. This benefit is maximized if all involved decision-makers are informed of it diplomatically.
Subjectivity, where it exists, is highlighted and made
transparent in the priorities. Sensitivity analysis can be
directed to the most subjective components of the
study because their identity is known.
Clear, readable documentation provides grounds for
decisions in the form of maps that can be easily understood. Decisions are made by politicians on grounds
provided by scientists. Among the biggest causes of error in such studies is the lack of correspondence between the scientists and the politicians vocabularies,
Conservation Biology
Volume 16, No. 2, April 2002

which can be overcome by a clearly laid out graphical


presentation.
In Italy, where MPAs are in their infancy but a large
number are planned for the near future, a strategic planning method is badly needed. Of course, many important planning issues cannot be satisfactorily solved by
the approach we describe. In particular, the spatial independence of the evaluation units does not allow direct
consideration of features whose definition depends on
spatial organization at multiple scales. It would be difficult, for example, to fully account for aesthetic values,
whose definition is not only subjective but also highly
dependent on the relative arrangement of landscape features. Nevertheless, the methodology we describe allows consistency in both the planning and follow-up
phases, thus ensuring ease of comparison and maximal
feedback. Techniques based on SMCA can be used in
planning as we have illustrated, but they also can be applied in other phases. For example, SMCA can be used
to monitor and evaluate the effectiveness of MPA management to see whether objectives are being met and according to what time frames.
In the particular case of supporting environmental
planning, MPA development, and conservation education in Italy, we have come to appreciate the significance of consistent procedures to provide future MPA

Villa et al.

Figure 4. Final zoning plan for the Asinara Marine


Reserve: (A1) no-entry, no-take; (A2) entry, no-take;
( B) general reserve, and (C) partial reserve.

planners and managers with better instruments to maximize the scientific and environmental return of MPAs.
An SMCA is one such procedure that holds promise as a
user-friendly yet rigorous tool for articulating priorities
and developing planning options, and we intend to continue applying it to Italian reserve planning.

Acknowledgments
We thank G. Notarbartolo di Sciara, president of the Instituto Centrale per la Ricerca Applicata al Mare, for his
lead role in this project and E. Cossu, president of Asinara Natural Park, for his support. T. Di Nora, G. Lauriano, M. E. Piccione, and E. Salvati contributed much excellent fieldwork and important feedback during the
refinement of the theoretical approach. Funding was
provided by the Ispettorato Centrale per la Difesa del
Mare of the Italian Ministry of the Environment. F.V. is
supported by the National Science Foundation (NSF ),
grants NSF 784AT-31057 (PACI Alliance project, Subaward #784 ) and DEB-9714835, and by U.S. Environmental Protection Agency grant R82716901.
Literature Cited
Agardy, T. 1995. Critical area identification and zoning in coastal biosphere reserves: one way to make marine conservation work in Canada. Pages 214219 in N. Shackell and J. H. Martin Willison, editors.
Marine Protected Areas and Sustainable Fisheries, Science and Protected Areas Association, Wolfville, Nova Scotia, Canada.

Spatial Multiple-Criteria Analysis for MPA Zoning

525

Agardy, T. 1997. Marine protected areas and ocean conservation. Landes Press, Austin, Texas.
Agardy, T. 2000a. Effects of fisheries on marine ecosystems: a conservationists perspective. International Council for the Exploration of the
Sea Journal of Marine Sciences 57:761765.
Agardy, T. 2000b. Information needs for marine protected areas: scientific and societal. Bulletin of Marine Sciences 66:875888.
Cochrane, J. L., and M. Zeleny, editors. 1973. Multiple criteria decision
making. University of South Carolina Press, Columbia.
Dayton, P. K., E. Sala, M. Tegner, and S. Thrush. 2000. Marine reserves:
parks, baselines, and fishery enhancement. Bulletin of Marine Sciences 66:617634.
Eastman, J. R., P. A. K. Kyem, J. Toledano, and W. Jin. 1993. GIS and
decision making. Explorations in geographical information systems
technology. Volume 4. United Nations Institute for Training and
Research, Geneva.
Kelleher, G., and C. Recchia. 1998. Editorial: lessons from marine protected areas around the world. Parks 8:14.
Kelleher, G., C. Bleakley, and S. Wells. 1995. A global representative
system of marine protected areas. World Bank, Washington, D.C.
Italian Law. 1991. Legge quadro sulle aree protette n. 394/91. Available at
http://www.parks.it/federparchi/leggi/394.html (accessed February
2000).
McManus, J. W., C. van Zwol, L. R. Garces, and D. Sadacharan, editors.
1998. A framework for future training in marine and coastal protected area management. Conference proceedings 57. ICLARM,
Penang, Malaysia.
Nijkamp P., P. Rietveld, and H. Voogd. 1990. Multicriteria evaluation in
physical planning. Holland Publishers, Amsterdam.
Roberts, C. 2000. Selecting marine reserve locations: optimality versus
opportunism. Bulletin of Marine Sciences 66:581592.
Saaty, T. L. 1980. The analytical hierarchy process. McGraw-Hill, New York.
Salm, R., and J. Clark. 1999. Marine and coastal protected areas: a guide
for planners and managers. 2nd edition. World Conservation Union,
Gland, Switzerland, and Island Press, Washington, D.C.
Tunesi, L. 1994. Aree marine costiere protette e gestione delle risorse
ittiche. Mare Nostrum 1:1215.
Tunesi, L., and G. Diviacco. 1993. Environmental and socio-economic
criteria for the establishment of marine coastal parks. International
Journal of Environmental Studies 43:253259.
U.S. Army Construction Engineering Research Laboratory. 1994.
GRASS 4.1 user manual. Champaign, Illinois.
United Nations Environmental Program (UNEP). 1995. Guidelines for
integrated management of coastal and marine areas. UNEP Regional Seas Reports and Studies 161:180.
Villa, F., M. Ceroni, and A. Mazza. 1996. A GIS-based method for multiobjective evaluation of park vegetation. Landscape and Urban Planning 35:203212.
Voogd, H. 1983. Multicriteria evaluation for urban and regional planning. Pion, Amsterdam.

Appendix
Calculation of Concordance Maps
Each scenario v is defined by a priority vector Wv, containing the weights
for variables 1 . . . j. Each observation (corresponding to an evaluation
unit in the spatial case) is described by a vector of variable values for
each variable considered. The values of all variables for all observations constitute the evaluation matrix E, whose element eij is the value
of variable j for observation i.
The variables are subdivided into two sets: H, composed of all the
quantitative variables, and Z, composed of the semiquantitative variables. Both the priority matrix and the quantitative portion of the evaluation matrix are then standardized in the interval (0,1). For each couple observations ii, a dominance score, miiv, is computed, expressing
Conservation Biology
Volume 16, No. 2, April 2002

526

Spatial Multiple-Criteria Analysis for MPA Zoning

Villa et al.

how much unit i outperforms i under the viewpoint expressed by


scenario v. The computation is done separately for quantitative and
semiquantitative data. The score oiiv for quantitative data is

o ii'v = ( w jv e ji e ji ) ,
jH

(1)

where wjv is the priority weight for variable j in scenario v and  is a


scaling parameter usually set equal to 1. In the semiquantitative case,
the score qiiv is
1
---

q ii'v

= [ w jv sgn ( e ji e ji') ] ,
jZ

(2)

where sgn is the sign function, defined as

1 if e ji < e ji'

sgn ( e ji e ji') = 0 if e ji = e ji' .

1 if e ji > e ji'

Conservation Biology
Volume 16, No. 2, April 2002

(3)

The scores for qualitative and quantitative data are then compounded
into a partial dominance score

m ii'v =

wjv oii'v + wjv qii'.

j H

(4)

j Z

Based on the partial dominance scores, a global dominance score si


for each unit i with respect to all others is calculated. Given the scenario v, the simplest equation for s is

1
s i = --- m ii',
I i

(5)

where I is the total number of evaluation units (for clarity, v is not


shown). More complex aggregation formulas are usually chosen because of their better statistical behavior. The value si, with appropriate
standardization, expresses the degree of concordance between the
variable values in observation i and the scenario, v, chosen. The different values of s, reflecting different combinations of variable values in
the area, constitute the concordance map.

Potrebbero piacerti anche