A Global Evaluation of Coral Reef Management Performance: Are MPAs Producing Conservation and Socio-Economic Improvements?
- First Online:
- Cite this article as:
- Hargreaves-Allen, V., Mourato, S. & Milner-Gulland, E.J. Environmental Management (2011) 47: 684. doi:10.1007/s00267-011-9616-5
- 589 Views
There is a consensus that Marine Protected Area (MPA) performance needs regular evaluation against clear criteria, incorporating counterfactual comparisons of ecological and socio-economic performance. However, these evaluations are scarce at the global level. We compiled self-reports from managers and researchers of 78 coral reef-based MPAs world-wide, on the conservation and welfare improvements that their MPAs provide. We developed a suite of performance measures including fulfilment of design and management criteria, achievement of aims, the cessation of banned or destructive activities, change in threats, and measurable ecological and socio-economic changes in outcomes, which we evaluated with respect to the MPA’s age, geographical location and main aims. The sample was found to be broadly representative of MPAs generally, and suggests that many MPAs do not achieve certain fundamental aims including improvements in coral cover over time (in 25% of MPAs), and conflict reduction (in 25%). However, the large majority demonstrated improvements in terms of slowing coral loss, reducing destructive uses and increasing tourism and local employment, despite many being small, underfunded and facing multiple large scale threats beyond the control of managers. However spatial comparisons suggest that in some regions MPAs are simply mirroring outside changes, with demonstrates the importance of testing for additionality. MPA benefits do not appear to increase linearly over time. In combination with other management efforts and regulations, especially those relating to large scale threat reduction and targeted fisheries and conflict resolution instruments, MPAs are an important tool to achieve coral reef conservation globally. Given greater resources and changes which incorporate best available science, such as larger MPAs and no-take areas, networks and reduced user pressure, it is likely that performance could further be enhanced. Performance evaluation should test for the generation of additional ecological and socio-economic improvements over time and compared to unmanaged areas as part of an adaptive management regime.
KeywordsConservationEffectivenessGlobalEcologicalSocio-economicThreatMarine Protected Area
Global estimates suggest that 34% of coral reefs have been destroyed or are at a critical stage and Marine Protected Areas (MPAs) are seen as the best strategy to conserve coral reef habitats and their biodiversity (Bohnsack 1998; CBD 1996; McClanahan and others 2006; Roberts and Hawkins 2000; Wilkinson 2008). Database entries suggest there are approximately 1000 coral reef MPAs (Tupper and others 2008b). The actual number of MPAs is difficult to assess, as new MPAs are continually being established, many of which are very small (Rodrigues and others 2004) and community managed MPAs are rarely reflected in databases, despite being numerous (Weeks and others 2009). Current numbers are likely to be significantly higher, as there are 700 MPAs in the Caribbean (Guarderas and others 2008) and 985 in the Philippines (Weeks and others 2009). However, the number or area of MPAs is a poor indication of protection (Chape and others 2005; McClanahan 1999; Mora and others 2006a, b), as there is scant evidence that the recent proliferation of MPAs has produced tangible conservation benefits (Jameson and others 2002; Jones 2001; McClanahan 1999; Pollnac and others 2001; White and others 2002; Wilkinson and others 2006).
Conservation requires carefully designed and realistic objectives which can be tested to enable adaptive management and ensure accountability (Hockings 2000; Syms and Carr 2001; White and others 2006). MPAs are expensive and compete with one another for funding, as well as with other conservation measures (Gravestock and others 2008; Pelletier and others 2005). While a consensus has been reached about the need for MPA performance criteria and regular evaluation (Pomeroy and others 2004), the actual indicators and evaluations are less well defined (Alder and others 2002). There is a dearth of comprehensive evaluations of management effectiveness (Day and others 2003; Stem and others 2005) or quantitative assessment on a regional or global scale as to whether MPAs are actually providing conservation and welfare improvements (Vilayleck and Andrefouet 2006; Weeks and others 2009).
The first broad assessment of 383 MPAs found that a third had met their management objectives, one-third had met them partially, and the remaining third had failed to meet their objectives (Kelleher and others 1995; Kelleher 1996). More recent assessments of MPAs in Asia have suggested that 10–25% are effectively managed or successful (Burke and others 2002; Tun and others 2004; White and others 2005a). Mora and others (2006a, b) evaluated 980 coral reef MPAs (18.7% of the global habitat) and found that only 147 were at low risk from external threats such as pollution and less than 0.01% of the world’s corals are within no-take MPAs that have no poaching and a low risk of external threats. In addition, 40% were smaller than 2 km2, rendering them ineffectual for many species. Generally, compliance and enforcement of regulations is disappointing in many MPAs (Byers and Noonburg 2007; Jameson and others 2002; Mora and others 2006a). Other authors have used combinations of criteria to assess MPAs regionally, such as investigating representation of bioregions and/or size and spacing patterns and have found that the extent, distribution and size are inadequate to achieve conservation targets in the Philippines (Weeks and others 2009) and that existing protection is insufficient to preserve either the full range of important habitats or connectivity among populations (Guarderas and others 2008). Additional research has concentrated exclusively on specific ecological outcomes, such as fisheries and coral cover. For example, MPAs have been shown to increase fisheries biomass, size, density and species richness (Lester and others 2009; Pollnac and others 2010). Selig and Bruno (2010) analysed changes in coral cover inside and outside 310 MPAs and found that there was no mean change in coral cover for reefs within MPAs over 38 years in any region. They found that MPAs were effective at slowing coral loss compared to reefs at unprotected sites, although there was a lag-time of 14 years for cover decline to level off. Regional differences emerged, with a linear relationship in the Caribbean and a non-linear one in the Indo-Pacific.
Evaluations of individual or regional groups of MPAs have generally suggested that the great majority of MPAs fail to meet their management objectives and that the reefs inside MPAs remain threatened (Jameson and others 2002; Jones 2001; McClanahan 1999; Mora and others 2006a; Pollnac and others 2001; Selig and Bruno 2010; White and others 2002; Wilkinson and others 2006). However, meaningful comparison is hampered by a lack of standard methodologies, differences in robustness and comprehensiveness of methodologies, as well as inconsistent language and lack of consistency in methods over time (Aronson and others 2002; Gaston and others 2008; Hockings 2003; Hudina 2006; Stem and others 2005). Given the ever increasing number of MPAs and limited conservation funds, establishing performance is critical and case study approaches have limited contribution to understanding general relationships of marine reserve functioning (Ostrom 2007; Pollnac and others 2010).
MPA evaluations are not trivial (Christie 2004) and achievement of objectives alone is insufficient for MPA evaluation, as this would reward MPAs with modest goals (Jones 2001). Terrestrial evaluations, such as those utilising satellite images, are of limited use in the marine environment (Allison and others 1998; Bateman and others 2002; Jones 2001). Similarly, an overview of IUCN category or regulations and secondary data is unlikely to reveal enough information to make sound judgements about the conservation effectiveness of a particular area (Stern 2006) because there is a disconnect between IUCN category and stated goals or level of protection (Guarderas and others 2008) and because management objectives have an unclear link to protected area outcomes (Leroux and others 2010). There is also a tendency to confuse expected outcomes associated with actions with the actions themselves. MPAs typically have several, potentially conflicting objectives, which also make the task of explicitly linking outcome measures to actual performance tricky, especially given heterogeneous and potentially conflicting objectives (Syms and Carr 2001). For example, the degree of local reliance on marine resources is often included as a performance measure, although this has an unclear link to MPA effectiveness (Stern 2006).
One framework which has served as the basis for many protected area assessment methodologies gives equal weighting to six elements; context, planning, inputs, processes, outputs and outcomes (Hockings and others 2000). Subsequently, many evaluation methods have been developed, which vary from rapid qualitative judgements to detailed quantitative monitoring with associated trade-offs in terms of time, cost and data quality (Corbett 1992; Wells 2006). The underlying principles and approaches are often similar between assessments (Stem and others 2005), with common elements including the analysis of status and trends in biodiversity, socio-economic, threat and governance measures (Francis and others 2002; Wells 2006). However, as a result of distinct perspectives on MPA success or effectiveness, assessments differ in their purpose, information used, terminology, fields covered, indicators assessed, the number and type of participants and the financial cost (Hockings 2000). The most popular indicators are often chosen based on feasibility of measurement, rather than because they are better indications of performance. Many assessments focus exclusively on ecological outcomes (Halpern 2003). Biophysical and socio-economic impacts are rarely evaluated, as they are costly and difficult to measure accurately (Day and others 2003; Kay and Alder 1999; Pomeroy and others 2005). Thus evaluations differ in terms of their robustness, credibility, reliability and comprehensiveness (Gaston and others 2008; Hudina 2006).
This paper seeks to add to this body of work by providing a detailed examination of key aspects of MPA performance for coral reef MPAs globally. The MPA literature has identified critical features and actions for management. First, we examine the extent to which these are fulfilled by MPAs. Second we reduce the extensive set of performance measures against which MPA management are typically assessed to a subset of ecological and socio-economic outcome measures related to MPA goals and designed to demonstrate whether the MPA is producing discernable benefits over time or compared to outside the MPA, which has been identified as the most appropriate measure of performance (Kareiva 2006; Mora and others 2006a; Selig and Bruno 2010; Syms and Carr 2001). Third, we evaluate the evidence that MPAs are indeed producing these additional benefits, based on our suite of measures, and finally, we examine differences in outcomes over time and between regions.
This is the first study to assess such a wide range of factors which are indicative of both conservation and welfare benefits of coral reef MPAs globally, using a single methodology. Our performance measures can be assessed using expert knowledge at a low cost, and are able to capture both social and ecological additionality, and hence may be widely applicable for the future evaluation of the success of coral reef MPAs.
Identifying Comprehensive Performance Criteria
Initially, simple performance criteria were developed based on academic literature, which has suggested that there are some critical features which determine MPA performance. These include adequate total MPA size (Lauck and others 1998; Lowry and others 2009; Roberts and Hawkins 2000), a no-take area which is sufficiently large for the species being targeted (Boersma and Parrish 1999; Halpern 2003; Pomeroy and others 2004; Vilayleck and Andrefouet 2006), adequate budgets and staff (Emerton and others 2006; Gravestock and others 2008), with sufficiently strict regulations, often denoted by a low IUCN management category (Dudley 2008) and inclusion in an MPA network and within the context of wider coastal zone management (Allison and others 1998; Cicin-Sain and Belfiore 2005; Murray and others 1999). Research also suggests that MPAs have to be no-take and minimally affected by external risks to provide appropriate protection of coral reefs, hence this was also included in the criteria (Sale and others 2005; Storms and others 2005). Other conditions were a sufficiently low fishing pressure (Davis and Tisdell 1995), low tourism pressure (carrying capacity estimates range from 4000–15,000 dives per site per year, Dixon and others 1993; Hawkins and Roberts 1992; Hawkins and others 1999), local employment benefits (Mascia 2004), user compliance to regulations (Boersma and Parrish 1999) and the cessation of destructive activities (Wilkinson 2006).
In order to produce a comprehensive but feasible set of performance measures, the different evaluations were compared and a long-list of potential variables was generated, based on the principle that MPAs should be assessed both against their stated objectives and against a general set of desirable outcomes (Hockings and others 2000; Syms and Carr 2001). Our list of measures was organised into five categories, representing different facets of effectiveness, which is built on existing frameworks and methodologies, but is a novel approach to evaluating performance remotely:
Perceptions of Overall Success and the Extent of Achievement of Aims in Terms of the MPA’s Primary Stated Goal
An important application of this research is to test the extent to which MPAs meet their management objectives (Halpern and Warner 2003; Hockings and Gilligan 2010; Jameson and others 2002), however assessment in relation to goals requires that they be well-defined, which is not always the case (Kay and Alder 1999; Roberts and Hawkins 2000; Sanchirico 2000). Goals may be ecological, socio-economic or governance related, and fulfilling each may require different actions and regulations (Alder and others 2002; Pomeroy and others 2004; Roberts and Hawkins 2000). Secondary goals typically include increasing tourism, local community economic development, conflict management, education, research, representation of stakeholders and cultural heritage protection (Agardy 2000; Boersma and Parrish 1999; McField and Kramer 2007; Pomeroy and others 2007; Pomeroy and others 2004; Sanchirico and others 2002), but these are rarely evaluated. Hence a number of expected or desirable outcomes needed to be included, as well as information on MPA aims and regulations.
The Cessation of Banned or Destructive Activities
This is included since MPA performance critically depends on the threats that each MPA faces. The cessation of damage to reefs is a pre-requisite of coral reef protection (Lowry and others 2009; Wilkinson 2006), especially as reduced stress is associated with increased resilience to natural and human threats (Tompkins and Adger 2004). Stated compliance is likely to be subject to cultural biases and over-reporting (Pollnac and others 2010). For this reason, we did not ask respondents to report compliance, but instead examined compliance related variables, including percentage illegal activities detected, percentage of those detected punished, the number of banned activities that occur, and changes in destructive activities within and outside the MPA.
Change in Threats Over Time and Compared to Outside
We wished to detect temporal and spatial changes in threats (Guarderas and others 2008; Hockings and others 2004; Pollnac and others 2010). MPAs are unlikely to be effective if they are located in areas that are subject to numerous, and often uncontrollable, external stressors (Jameson and others 2002; Pollnac and others 2010). Similarly, if an improving trend merely reflects trends in the area where the MPA is situated, there is no additionality from the conservation intervention, and therefore it is not likely to be cost-effective. We distinguished between external threats which cannot be effectively addressed by an MPA, and internal threats that can be addressed by managers (Hockings and Gilligan 2010; Mora and others 2006b; Selig and Bruno 2010).
The fundamental precondition for conservation success is a stabilisation or improvement in the status of the conservation target. This is difficult to assess in the absence of rigorous scientific surveys, but we were able to obtain information on change in the percentage of live coral cover, which is at a global scale the most appropriate test of habitat conservation (Mora and others 2006a; Selig and Bruno 2010; Wilkinson 2008). We looked at changes over time and compared to country averages. We also included perceived changes in habitat quality and both fisheries and species conservation, since these reflect distinct ecological outcomes (Pelletier and others 2005) and since fisheries improvements from MPAs remain controversial (Lowry and others 2009; Russ and others 2004). Some ecological outcomes could only be tested indirectly. For example, the extent to which fishers congregate on reserve boundaries is thought to be evidence of spillover of fisheries benefits from the MPA (Murawski and others 2000; Shorthouse 1990).
Social and economic factors were explicitly included in our framework because these are rarely included in performance evaluations of MPAs, despite increasing evidence of their importance for MPA success (Christie 2004; Fiske 1992; Kelleher and Recchia 1998; McClanahan 1999; Roberts and Hawkins 2000). For example, conflict is thought to contribute to the high rate of MPA failure (Christie 2004; Christie and White 2007; Christie and others 2009; Halpern and Warner 2003; Mascia 2003; White and others 2002). Hence the generation of local wealth, employment, economic development and the extent of conflict are important measures of success that would be expected to occur as a result of other benefits such as tourism and fisheries (Russ and others 2004).
A questionnaire survey was chosen as the primary method of data collection, using an expert scoring approach to gather both quantitative and qualitative evidence of performance by describing changes in a number of specific outcomes within and outside the MPA. Expert scoring is increasingly used in global analyses and PA assessments (Balmford and others 2004; Gravestock and others 2008; Mora and others 2006a; Staub and Hatziolos 2003; Wells 2006). Expert scoring is less accurate than long term monitoring or scientific studies, as it is qualitative, reliant on subjective perceptions and may therefore be prone to strategic answers, such as exaggerations, or omission of relevant facts (Hockings 2003). However, it has several advantages. Quantitative monitoring data is rarely available, is often collected using diverse methods that are not easily comparable and can be of varying quality. Scoring, on the other hand, is more easily implemented and cost-effective (Alder and others 2002). Furthermore, responses of MPA managers can be based on years of field experience and may therefore better capture the realities and complexities of MPAs (Hockings 2003).
The questionnaire was designed to gather site level information on many variables related to MPA performance, to enable a holistic understanding of the MPA outcomes and contexts. It included sections covering respondent details, MPA features and management actions, budgets, regulations and uses, threats both inside and outside, and respondent opinions on habitat and welfare benefits and enforcement. The majority of protected area evaluations do not include reference sites outside the MPAs or quantitatively assess effects over time, hence these comparisons were included wherever possible. Since replication over time and randomisation of experimental treatments was not feasible, we included as many samples as possible and modelled potentially confounding factors explicitly (Stone 1993). By applying the same evaluation methodology in all MPAs, we hoped to achieve a more comprehensive picture of the many facets of global coral reef MPA performance than previously.
The survey instrument was developed by adapting previously published indicators, principally the “World Bank GEF MPA project scorecard” (Staub and Hatziolos 2003), “How is your MPA doing?” (Pomeroy and others 2004) and the common reporting framework for marine conservation effectiveness (Stern 2006) and relevant reviews (Stem and others 2005). For perceived changes, pre-defined ordinal categories were used. Questions were designed to produce a quantifiable measure on a scale ranging from low to high effectiveness and open-ended questions were used as little as possible, to aid comparisons at a regional or global level (Stern 2006). All questions related to 2005, unless otherwise stated, to reduce temporal variability. To supplement the questionnaire data, secondary data were obtained from a range of sources. National-level average live coral cover estimates were taken from Wilkinson (2004). Budgetary information was converted into 2005 US dollar equivalents taking into account historical exchange rates and purchasing power parity, which are necessary to normalise budgetary data (OECD 2002).
Population and Sampling
The population of interest was all MPAs containing coral reefs. Management staff themselves were not solely targeted, as many have inadequate time to participate and since they may exaggerate positive MPA impacts (Bhagwat and others 2001), so academics and NGO employees with significant local knowledge were also targeted as respondents. It was assumed that respondent evaluations of specific changes at each MPA were broadly comparable, although respondent affiliation was noted, so that biases could be tested.
The survey was distributed to respondents at an international coral reef management symposium, as well as directly through email correspondence and via a website in Spanish and English which was widely publicised, from October 2006 until November 2007. Contact details were gleaned from conference proceedings, internet sites and using a snowball approach, where respondents were asked to pass on the survey. This approach was necessary given the need for as large a sample size as possible. This sampling procedure suffers from a lack of randomness and self selection bias, which is likely to mean that responses constitute a best case scenario, as those MPAs with better performance and capacity are more likely to contribute, especially in terms of responses from management staff. Seventy-eight responses were received from 33 countries. It is not possible to estimate overall response rate since newsletters and snowballing sampling approaches were employed. For individuals who were directly given or sent the survey, it was approximately 20 and 4% respectively. This low rate was likely to have been due to the time taken to complete each survey and could mean that there is a bias towards MPAs which have performed well or by respondents with a personal interest in evaluation or concerns about management of a given site. Suspected biases were examined using comparisons with other published research. Multiple responses from the same MPA (for 11 MPAs) were found to differ marginally in less than 10% of responses, except for the two very large sites where 11 and 17% of responses differed (the Great Barrier Reef Marine Park and the Komodo National Park). Because responses were similar, one response was randomly chosen for each MPA with multiple responses.
Some independent variables were coded based on open ended responses. Qualification of the suitability of the management action used to address the main threat was coded on a three-point scale. For example, increased funding would be ineffective against outside pollution, potentially effective against poaching and highly effective for understaffing. Other performance measures variables summarised responses, such as the number of threats and unsustainable uses and their comparison to outside MPAs. Data were analysed using bivariate and multivariate statistics using Stata.8 and R version 2.0.0 software.
Sample population statistics (n = 66 unless otherwise stated)
MPA Age (years)
No take area (km−2)
Set-up budget (US$’000)
Overall budget (US$’000) N = 40
Budget km−2 (US$’000) N = 56
Management budget km−2 (US$’000)
Visitor pressure (km−2)
Fisher pressure (km−2)
Proportion of MPAs Fulfilling Key Criteria for Success
Percentage and area of MPAs fulfilling possible evaluation criteria for MPA success
Possible MPA success criteria
% of MPAs
% of MPA Area
MPA size > 20 km2
Have no-take area
No-take area > 2 km2
No-take area > 5 km2
More than 5 years old
IUCN category II or stricter
IUCN category IV or stricter
Part of wider coastal management
Have active fisheries management
One staff member per 10 km2
Had an initial set up budget
Minimum annual budget of US$1000/km2/year
Uses and threats
<10 fishers per km2
<100 tourists per km2
Mangrove extraction never occurs
Commercial fishing never occurs
Coral mining/destructive fishing never occurs
No banned activities occur
Better coral cover than national average
Maintained or improved live coral cover
Evaluation of MPAs Using the Performance Measures
Achievement of Management Aims
Respondents estimated a mean detection rate of illegal activities of 39.8%, although 12% had no detection and 5% claimed almost perfect detection. Of those activities detected, a mean of 48% were punished (ranging from 0 to 100%), so that overall only 19% of the assumed number of infractions were punished. Highly unsustainable uses such as mangrove wood extraction, coral mining, blast fishing and cyanide fishing had been effectively halted in 68%, 82%, 83% and 84% of MPAs respectively. Whilst 20% of the MPAs reported that they had stopped all banned activities occurring, over 40% still had one or two still taking place. Whilst it is difficult to gauge the severity of these activities, a mean of 2.7 banned activities nevertheless occurred within the MPAs. Foreign commercial fishing, which had been banned in 89% of MPAs, nevertheless occurred frequently in 3% and occasionally in 36%. Similarly, blast fishing had been banned in 97% of MPAs, but occurred occasionally in 18% of MPAs, and cyanide fishing had been banned in 98.5% of MPAs, but occurred occasionally in 16.2%. This supports the hypothesis that MPA enforcement is often ineffective (Murray and others 1999).
Reduction in MPA Threats
When asked to identify the most serious threat to the MPA, those threats originating inside the MPA were poaching (28%), tourism related damage (12%), unsustainable use (11%), lack of enforcement (5.5%) and corruption (3.5%). Externally generated threats included coastal development (19%), pollution (12%) and bleaching or climate change (9%). Overall, 60% of the MPAs reported that the greatest threat originated inside the MPAs and 40% outside.
In terms of the occurrence of commonplace large-scale threats inside the MPA over the last 5 years, only 6 MPAs had faced none and the mean number was 3.3 per MPA, where the region was a strong predictor for the type of threats faced. The most frequently cited such threats were coral bleaching (70%), hurricanes (60%) and sedimentation (56%), although different regions faced different levels and combinations of these threats. For example, while African and Pacific MPAs reported a maximum of 4 large-scale threats, 25% of MPAs in the Americas and 33% in Asia had > 5 large scale threats. All of the MPAs in the Americas reported at least one large scale threat, usually cyclones and hurricanes.
The majority of actions the managers could use to ameliorate the threats facing the MPA were either unsuitable (22%) or of limited effectiveness (46%), compared to 32% which were targeted and potentially effective against the threat. The mean reduction in destructive uses (including but not limited to banned activities) achieved since establishment of the MPA was 2.5 (SD = 2.27; Fig. 2b). However, unprotected areas have also seen an average decrease of 1.5 (SD = 1.44) destructive activities in the same period. Thirty nine percent of MPAs are mirroring areas outside their boundaries, 25% have out-performed outside areas by one activity, 36% by more than one activity and 5% four or more. The MPA types that were most successful in reducing destructive activities compared to their surrounding areas were medium-sized (151–1000 km2; χ2 = 28.7, P = 0.052), within less developed countries (f = 6.7, df = 1, n = 60, P = 0.012), multiply managed (z = −2.19, n = 60, P = 0.029); or had no-take areas (z = −2.04, n = 60, P = 0.038).
MPA establishment has not led to improvement in habitat quality in 36% of MPAs and in species conservation in 28%. Fisheries showed the poorest performance of the ecological outcomes assessed, having not improved in 44% of MPAs. This is borne out by only 19% of MPAs reporting regular “fishing the line”, despite 50% having active fisheries management. In the spatial comparisons, 69% thought that the number of fishers had decreased inside the MPA and 28% that it had increased, while outside the MPA 56% thought that fishing pressure had increased. The smallest size class of MPAs (0–25 km2) reported the most improvement in fisheries (χ2 = 11.12, P = 0.085), as did less developed countries (71% compared to 38% of more developed countries; f = 6.7, df = 1, n = 60, P = 0.012).
The respondents took their current coral cover estimates from one-off studies (73%), monitoring data (22%) and expert opinion (5%). Many respondents (24%) were not able to report coral cover at MPA establishment. This was especially true for older MPAs. A comparison of live coral cover inside the MPA to the country average (Fig. 2c) reveals that MPAs contain on average 7.6% more live coral cover than the national average. This is highly variable, however (median = 1.75%, n = 62, SD = 23.2%, range = −22 to +77%).
If we compare respondents’ estimates of coral cover change to their perceptions of habitat change (noting that one pertains to cover and the other to quality, so they are different), we see that while 93% of respondents perceived that their MPAs had maintained or improved habitat, only 66% of MPAs had maintained or improved cover, although these results are not statistically different (χ2 = 1.39, n = 49, P = 0.499), which suggests that perceived habitat changes are similar to those calculated from coral cover estimates.
MPAs with no-take areas had greater improvements in live coral cover (+2.3% compared to −4% for those without; z = −1.82, n = 50, P = 0.075) and had suffered less coral damage from tourists (z = 1.846, n = 66, P = 0.065). Similarly, community managed areas had seen greater coral cover increases (mean +5.2% compared to −5.17% in other MPAs; t = −3.294, n = 50, P = 0.002), despite being in higher (less stringent) IUCN categories.
There had been no improvements since establishment in local economic development in 29% of MPAs and in research in 12%. Most respondents (79%) did not feel cultural erosion was a problem. Conflict was reported to have increased in 50% of MPAs and stayed the same in 26%. This supports the hypothesis that MPAs can are not always social successes (Christie 2004). Conflict had decreased most in multiply managed and community managed areas (χ2 = 13.8, n = 66, P = 0.031) and those which had conflict resolution initiatives (χ2 = 16.5, n = 66, P = 0.000). 85% of MPAs had experienced increased tourist visitation, while 55% reported increased local employment benefits and 45% increased local wealth. The MPAs supported a mean of 69 businesses (SD = 131) and 727 jobs (SD = 1490, max = 6740). Of these jobs, an average of 51% were in the tourism industry (1 per 54 tourists), 46% were fishers and 3% were MPA staff. MPAs support an average of 291 jobs per km2. However this figure is unduly affected by two outliers supporting more than 1000 jobs/ km2. If these are removed, the mean number of jobs supported is 123 per km2 and 23% of MPAs support only 0.3 jobs /km2 or less. Respondents estimated that 82% of jobs were retained by local communities (24% had 100% of jobs locally retained), with 84% of fishing employment retained by locals, compared to 77% of management and 75% of tourism employment.
Examining Temporal and Regional Influences on MPA Performance
Temporal Aspects of Change
The number of years since MPA designation explains 8.5% of the variation in change in live coral cover (f = 4.32, df = 1, n = 50, P = 0.043), and is not a significant predictor of spatial differences. Whereas MPAs up to 12 years old reported increases in coral cover, on average, those older than 12 years reported losses in cover. These losses become greater the older the MPA, with MPAs 30 years or older having experienced a mean of 10% decline in coral cover (Fig. 3). To test a potential impact of the 1998/99 major bleaching event on habitat quality changes, we examined the relationship between date of establishment before this bleaching event and the change in coral cover over time and found that it was not significant (f = 1.22, n = 49, P = 0.275).
Based on the academic literature, we hypothesized that effectiveness would generally increase over time, although habitat or fisheries improvements tend to show a lag and there may be a non-linear relationship between the number of years since designation and improvements (Leverington and others 2008; Syms and Carr 2001). These results suggest that temporal changes are not linear and that for our sample, effectiveness has not increased greatly over time.
Regional Variation in MPAs and Performance Measures
(a) Significant differences between MPAs in more and less developed countries (MDCs, LDCs), as assessed using univariate tests. (b) Variables which differ significantly between regions
MDC (n = 13)
LDC (n = 53)
Age (years since designation)
F = 11.4, df = 1, n = 65, P = 0.001
No take area size (km2)
F = 3.5, df = 1, n = 62, P = 0.067
Median budgets (US$ per km2)
T = 1.86, n = 56, P = 0.068
% funding used directly for management costs
F = 11.5, df = 1, n = 59, P = 0.001
% funding from government
F = 2.51, df = 1, n = 61, P = 0.015
% to achieve fisheries improvements
χ2 = 13.7, n = 64, P = 0.001
Decreased destructive uses over time
F = 6.7, df = 1, n = 60, P = 0.012
Variable (number MPAs)
Africa (n = 10)
Americas (n = 20)
Asia (n = 30)
Pacific (n = 6)
Temporal change in coral cover
F = 2.99, P = 0.040
Spatial coral cover comparison
F = 3.42, P = 0.023
No destructive activities to decrease over time
F = 10.2, P = 0.000
No. destructive activities to decrease compared to outside
F = 8.07, P = 0.045
No. banned activities occurring
F = 5.42, P = 0.003
No large threats inside
χ2 = 31.2, P = 0.006
Main threat inside MPA
χ2 = 7.2, P = 0.066
As expected, we found strong regional influences on performance (Selig and Bruno 2010; Wilkinson 2006; Table 3b). African MPAs tended to protect high quality coral and had maintained it since inception (over an average of 11 years). American MPAs showed the most negative trends in coral cover since establishment (over a mean of 15 years), but these were still better than the trends outside the MPAs. Asian MPAs had increased coral cover by 5.31%, the greatest increase of all the regions (over a mean of 14 years) and were situated in areas with 7.5% better coral cover than the national average. Pacific MPAs contained much greater coral cover than the national average and had seen increases in coral cover, over a mean of 19 years.
MPAs in Africa and the Pacific showed relatively low levels of illegal activities, compared to Asian and to a lesser extent, American MPAs. MPAs in America and the Pacific had reduced relatively fewer destructive uses compared to Asian MPAs. When we compare these changes to areas outside, MPAs in Africa highly outperform outside conditions and the additionality of Asian MPAs is weaker than the temporal change suggests. However MPAs in Asia show the most variation, such that all three MPAs which had decreased six or more destructive activities were in Asia. In addition, the results suggest that while not reducing a large number of activities, MPAs in America were nevertheless outperforming outside conditions, in contrast to Pacific MPAs which were largely mirroring outside events.
MPAs in Africa faced fewer threats per MPA but a larger variety of threats overall (coral bleaching, pollution and war). Eighty eight percent of the most critical threats came from inside and were reported to be poaching, coastal development and unsustainable use. All of the MPAs in the Americas reported at least one large scale threat, especially coral bleaching, sedimentation and hurricanes (a mean of 3.2 large threats per MPA), with 25% having more than five large scale threats. 48% of their main threats were external and therefore beyond the control of management. Asia had the largest number (3.8) of mean large threats per MPA, especially bleaching, sedimentation and pollution, hence the most pressing threat originated inside the MPA in 84% of MPAs. One MPA in Vietnam faced eight large scale threats. Pacific region respondents cited a mean of 1.8 threats, especially coral bleaching and hurricanes, hence 67% reported that their greatest threat originated inside the MPA.
Protected areas are expensive and are often reported as failing to produce tangible conservation and welfare benefits, hence there is an increasing interest in alternative approaches, such as incentive schemes and tradeable permits (Greiner and others 2000; Heinen 1996; Hilborn and others 2005). Given the increasing need for coral reef protection, it is critical to direct funds effectively. The cost of conducting robust quantitative analysis of the conservation and welfare improvements from MPAs on a global scale, with rigorous spatial comparisons, is prohibitive. While site-level studies are invaluable to elucidate complex relationships between MPA context, features, management and outcomes, the large number of effectiveness evaluation methodologies limits the comparability of these studies. Hence there is a need for an approach that provides robust information for global-level comparisons, but which is feasible and cost-effective in the face of limited data and time, and could form the basis for ongoing monitoring at the global scale. This is the gap that this study has aimed to fill.
We have gathered detailed information on a large range of factors at a large number of MPAs, by utilising expert knowledge of both the MPA itself and the ecological and social context in which it operates. We evaluated a range of factors linked to MPA performance, including goal fulfilment, ecological and social measures, and assessing changes through time and in relation to outside areas. We also sought to reduce common limitations of MPA evaluations related to the focus on small sets of ecological variables, the cost and effort of data collection, the lack of comparisons and control sites and the focus on management inputs based on the assumption that these produce the expected outcomes (Holtzman and others 2009; Willis and others 2003).
Our methodology necessitated reliance on potentially coarse perceptions of changes, which has limited the kinds of impacts evaluated. For example, it was not possible to look for changes related to recruitment and spawning, ecosystem resilience, food and employment security, susceptibility to environmental shocks or profitability of fisheries (Pelletier and others 2005). However, such information is rarely gathered and it is not possible to take into account every variable which may play a role in MPA effectiveness (Halls and others 2002). Instead we limited our study to assessing outcomes that respondents were able to judge based on their day-to-day knowledge of an MPA. Experts are increasingly used for global assessments of ecosystem or protected area status or performance (Burke and others 1998; Gravestock and others 2008; Mora and others 2006a; Tupper and others 2008a; Wilkinson 2008; Wood 2007). We acknowledge the limitations of expert opinion, which is likely to be unquantifiably subjective and biased. However the similarity of independent answers from different types of respondents for those MPAs which had duplicate answers, and the range of outcomes reported, suggest that this approach is not fatally flawed.
The MPAs included in this analysis cover 7% of coral reef MPAs contained in the reefbase database and more in terms of area under management. Comparison with other datasets supports the assertion that this sample is biased towards MPAs which are larger and better funded. If paper parks are identified by a lack of staff or any income, 12–23% of the sample MPAs are paper parks, which is MPAs generally (Gravestock and others 2008). Additionally, temporal changes in coral cover are similar to those reported in a larger dataset (Selig and Bruno 2010), which means that the sample may not be biased towards better performing MPAs. Indeed, the MPAs are highly heterogeneous, with a spread across IUCN categories and regions that is not significantly different from the global population, suggesting it is broadly representative of MPAs in general. As a result, this arguably constitutes an adequate dataset to make cautious inferences about MPAs generally (Hughes and others 2002).
The approach of narrowing down criteria for performance measures is very useful when such data are readily available, but the choice of measures used can be arbitrary or misleading. For example, several protected area assessments have excluded small MPAs as a minimum requirement (e.g., Mora and others 2006a; Weeks and others 2009) due to evidence of minimum sizes for fisheries impacts and since small reserves are more vulnerable to periodic disturbances and are harder to enforce (Lowry and others 2009; Roberts and Hawkins 2000). However, the effect of reserve size is complex, with some studies showing a strong effect (Claudet and others 2008; Halpern and Warner 2003) and others showing no effect (Halpern 2003; Lester and others 2009; Pollnac and others 2010; Stallings 2009). This means size alone is not a good indication of likely performance. While MPAs are likely to suffer reduced performance as a result of design that does not follow best available science (Le Quesne 2009), the correct design will depend to some extent on the MPA’s context (Christie and others 2009). Hence the proportion of MPAs fulfilling criteria has limited application as an evaluation tool and changes in several expected outcomes constitute a better measure of performance.
The examination of performance measures suggests that some MPAs are fulfilling objectives related to conservation of habitat quality, improvement of local community welfare and reduction of threats. However this is by no means universal. In contrary to previous observations (Christie 2004), socio-economic benefits are more commonplace than ecological improvements. The principal aim of the MPA has no significant relationship to whether the required outcome is achieved, while targeted management actions (such as fisheries and species conservation or conflict resolution) are generally effective. Therefore simply designating an MPA and specifying an aim is not enough, resources need to be made available to ensure active management effort (Balmford and others 2004; Francis and others 2002; Gravestock and others 2008). Median budgets reported are at the lowest end of minimum budgetary requirements per km2 (Balmford and others 2004), so that budgets are likely to be constraining performance. This hypothesis is further supported by the fact that 60% of the MPAs cited their greatest threat as one which occurred inside its boundaries and since responses suggest that 20% of illegal activities that occur are detected and punished, which is cause for concern, especially given the likelihood of over-reporting. Fortunately, compliance can be improved given increased budgets for enforcement, education and outreach programs.
The finding that only 42% of the MPAs are part of wider coastal management and that few MPAs have adequately sized no-take areas or have been incorporated into a network of MPAs could undermine management efforts, as could the lack of solutions reported to MPAs’ most pressing threats. More emphasis needs to be placed on reducing these threats on a regional basis, as relying on MPAs alone to conserve coral reefs, whilst they face a large number of external threats, is likely to be insufficient for coral reef conservation (Allison and others 1998). This is underscored by the finding that the mean number of large-scale threats inside and outside the MPA did not differ significantly. Importantly, our analyses demonstrate that those sites which had seen reductions in a large number of threats were sometimes simply mirroring the situation outside their MPA. This is important to take into account in MPA evaluations, as donor funds may not be well used even if threats are reducing, if this is part of a wider trend rather than as a result of MPA activities. Our evaluation approach produces the highest performance scores in MPAs which have been able to out-perform outside changes, which can help to identify locations were management can have the most impact.
Coral cover estimates suffer from difficulties in obtaining accurate data and large regional variability. Fortunately, only 5% of the coral cover estimates used here originated from expert opinion, rather than from monitoring data or a published study. Given the caveats, we can tentatively say that our finding that two thirds of the MPAs in the study had maintained or increased coral cover, which constitutes a significant success in the context of the global decline in reef health (Wilkinson 2008), although this represents only 20% of the area under management in our study. Overall, there has been no mean change in coral cover, as was also reported by Selig and Bruno (2010). Our results suggest that the 1998/1999 bleaching event was not a significant predictor of coral cover change over time at these MPAs. However, both local and regional bleaching events are expected to influence coral cover and the effect of management on the resilience to bleaching events should be the subject of future research.
Spatial comparisons were coarse grained but essential to provide local context. MPAs are generally located in areas benefitting from better than average coral cover than the rest of the country, although this is not always the case, which is significant, given that a basis for MPA site selection is usually to protect areas with better coral cover. This supports assertions that in many cases socioeconomic criteria are given greater weight than ecological considerations (Halpern and Warner 2003), which may contribute to poor ecological performance (Roberts and others 2003). Unlike Selig and Bruno (2010), we found that the oldest MPAs had experienced the greatest losses in coral cover, although this was a weak relationship and relatively unsurprising in the context of global coral reef ecosystem declines (Wilkinson 2008). We found fishing impacts difficult to evaluate with a quantitative measure that respondents are able to report on, which is common even in highly funded MPAs (Hockings and Gilligan 2010), but surprising given the emphasis on fisheries benefits as key reasons for establishing MPAs (Alder and others 2002). However our study suggested that fisheries enhancement was not universally attained, although 1 in 5 of the MPAs reported “fishing the line”, which could be evidence of spillover. Species conservation improvements were more widely reported however.
Tourism in our sample has resulted in the creation and support of increased employment and to a lesser extent, local wealth. This has been documented previously in large and renowned MPAs (e.g., Hockings and Gilligan 2010), but was reported here for small MPAs also. Further research in needed to clarify levels of visitation which will not undermine management. Since most employment in fishing, management and tourism is retained by locals, this suggests that MPAs are not encouraging immigration, as may be occurring in terrestrial protected areas (Wittemyer and others 2008). Funding is frequently used for local community projects, including those that are designed to compensate local costs (e.g., alternative livelihood schemes). Nevertheless conflict had increased over time in half of the MPAs, which is contrary to what was expected (Kelleher and others 1995). This is surprising given the generation of local benefits and is of concern as it could undermine local support and compliance (Pomeroy and others 2007).
Trends over time can be hard to discern in the face of temporal variation (Syms and Carr 2001). In our study, as in others, a non-linear pattern emerges in MPA performance over time (Selig and Bruno 2010; Syms and Carr 2001) and the temporal pattern varies by type of outcome. While MPAs often had immediate benefits, these tended to decline and then increased again in MPAs older than 20 years. Thus similar threshold effects to those hypothesized by Syms and Carr (2001) and observed in Indo-Pacific MPAs (Selig and Bruno 2010) were found here. Many of the MPAs in our sample were too young to have achieved their full impact, especially in terms of ecological changes, so these results probably reflect ongoing transient dynamics. In addition, factors other than time are likely to be playing a significant role in terms of changes in coral cover.
There is a need to ensure that MPAs are designed and managed according to best available science, such as relates to areas of higher than average environmental quality, minimum size requirements, zoning or no-take areas, sufficient budgets and staffing, adequate restrictions and enforcement of destructive activities, location within networks of protected areas and wider coastal zone management, which should be achieved through better communication of best available science, which is updated regularly.
MPAs that do not incorporate those aspects that are known to improve performance should be assisted in making necessary changes wherever possible
MPAs need to focus on conflict resolution, to increase long term effectiveness
MPAs whose main aim related to fisheries or species conservation as well as those who face large external threats need additional management actions, beyond MPA designation
There is a need to establish a globally recognized set of indicators of MPA performance which include both biological and socio-economic outcomes and threat reductions. These indicators should be collated for as many MPAs as possible and compared to similar areas without MPAs and in the context of MPA goals
Management and contextual information should be collected for MPAs, which can be used to examine the drivers of different types of outcomes, including those related to MPA design, management outcomes and local ecological and socio-economic conditions
Performance indicators should be incorporated into global datasets and repeated over time, to assess changes in collective MPA effectiveness over time and to help to understand temporal aspects of different outcomes, such as ecological lag times, as well as responses to local or regional shocks or other changes such as increased funding, as a tool for adaptive management
Performance indicators should be used to identify which MPAs require greater financial or technical assistance and/or institutional support.
We thank Dr. Jamie Oliver and Dr. Mark Tupper at the World Fish Centre for their financial support and help in survey design. We thank NERC/ESRC for support to Venetia Hargreaves-Allen, and a Royal Society Wolfson Research Merit Award to EJMG. We thank the many individuals and organisations who completed surveys or contributed contact information or advertised this research, including Conservation International, the World Wildlife Fund, the MPA learning network and MPA news. The authors also wish to thank the three anonymous reviewers who provided valuable comments on the manuscript.