Spatial planning model for optimizing conservation priorities for local community utilization on Arefi Island in the Raja Ampat Marine Protected Area (MPA) Southwest Papua, Indonesia

Study area

This research was conducted in Area III of the MPA in the Dampier Strait, Arefi Island, Raja Ampat Regency, Southwest Papua (Fig. 1). Raja Ampat Regency is group of islands situated between 2°25′N and 4° 25′S latitude and 130°E to 132°55′ E longitude. The regency covers approximately 6,084.5 km² and encompasses around 600 islands of various sizes. The Raja Ampat conservation area is renowned for its ecological richness and is a popular tourist destination. It also holds strategic importance for fisheries due to its well-functioning aquatic ecosystem. Although categorized under “other zones” in the Raja Ampat MPAs, which includes four districts in Southwest Papua, satellite data reveal the significant biophysical potential of Arefi Island (McKenna, Gerald & Suer , 2002; MMAF, 2018). The term “other zone” refers to the zoning classification used by MMAF for the Raja Ampat water conservation area.

Arefi Island, located at 0°47′18.67″S and 130°42′27.72″E, is home to significant marine biodiversity and provides crucial habitats for various species, including corals, fish, and endangered marine mammals (Kovács et al., 2021; Trip et al., 2019). The island’s unique characteristics make it a suitable candidate for various conservation zones, such as core, fisheries, and sustainable utilization areas. In addition to its ecological importance, Arefi Island is home to indigenous cultures that practice the “Sasi” tradition, a customary resource management system deeply rooted in their cultural heritage. The traditional practice of Sasi, implemented by local communities on Arefi Island, plays a key role in ecosystem recovery by regulating resource use, allowing marine habitats to regenerate and preventing depletion due to overfishing. This system involves periodic closures to allow ecosystem recovery and ensure resource sustainability (Sairiltiata, 2023). Under Sasi, indigenous communities impose temporary bans (moratoriums) on the use of marine resources, such as coral reefs and fish, in specific areas for designated periods (Rachma, Mangunjaya & Tobing, 2018). This highlights the need for a comprehensive and nuanced zoning strategy to fully protect and utilize the island’s diverse ecosystems.

Data used

The data used in this research primarily consists of remote sensing data, supplemented by secondary data sources (Table 1). The integration remote sensing data with ground-based observation or secondary data enhances the accurary of the result (Petrou, Manakos & Stathaki, 2011). Specifications of the multispectral bands of the Worldview-3 imagery are detailed in Table 2. The research framework and stages are illustrated in Fig. 2.

The first step involved analyzing satellite imagery to map biophysical parameters, including mangroves, seagrass, and coral reefs. These biophysical parameters serve as breeding grounds for numerous fish species with both commercial and ecological significance (Weeks, 2017; Sutrisno, Sugara & Darmawan, 2021) and are integral to conservation efforts. In the second step, the location of biophysical parameters were used as inputs for determining conservation features. Cost features were calculated based on the current usuage, as presented in Table 3.

Method of biophysical analysis

The biophysical parameters were mapped using high-resolution Worldview-3 satellite images from 2021, provided by the Center for Data and Information, National Innovation and Research Agency (BRIN). The spatial resolution of these images is approximately 0.6 m, allowing detailed analysis and mapping of terrestrial and aquatic ecosystems.

Due to spectral similarities, traditional pixel-based classification methods are limited for biophysical analysis in shallow water areas. To address this, object-based image analysis (OBIA) was employed, which differs from pixel-based methods by using image objects as the basic unit of analysis rather than individual pixels (Hossain & Chen, 2019). OBIA is needed for high resolution or highly variable images, because it is able to group pixels into objects based on spatial and spectral characteristics, thereby increasing classification accuracy (Blaschke, 2010).

OBIA is an iterative process that starts with segmenting satellite images into cohesive and contiguous segments. These image objects are then classified using either supervised or unsupervised approaches (Belgiu & Csillik, 2018). According to Ventura et al. (2018), the OBIA workflow begins with image segmentation, a process based on pixel parameters with similar spectral values. In this study, we used the multi-resolution segmentation (MRS) algorithm to create image objects that minimize average heterogeneity and maximize homogeneity. The three key parameters in the MRS algorithm are shape, compactness, and scale (Darmawan et al., 2022). OBIA analysis was performed using eCognition Developer 64 software.

The segmentation results were then classified using support vector machine (SVM) algorithms, a sophisticated non-parametric classifier widely employed in hyperspectral image classification that operates based on statistical learning theory (Tan et al., 2018). It is designed to seek an optimal decision hyperplane within a high-dimensional space, ensuring optimal separation of classes. SVM consistently performs well in challenging classification scenarios with high-dimensional features, demonstrating its effectiveness even when dealing with a limited number of training samples (Cao et al., 2018). The fundamental concept behind the SVM is to identify a hyperplane that maximizes the margin between distinct classes. This hyperplane is expressed by the following equation (Camps-Valls & Bruzzone, 2009). ${\displaystyle f\left(x\right)=\sum _{i=1}^n{\alpha }_i{y}_{i}K\left(x,x_i\right)+b}$ where $f\left(x\right)$: decision function

α_i: Coefficients obtained during the training process

y_i: class label of training sample x_i

K(x, x_i): kernel function

b: bias term

The predicted class of the input data point x is determined by $f\left(x\right)$. If $f\left(x\right)>0$ then the data point is classified as belonging to one class. If $f\left(x\right)<0$ then the data point is classified as belonging to another class. The classified data and field and secondary observations were then input into the Marxan software.

Method of priority area conservation

Marxan models

To meet conservation feature targets, enhance connectivity between areas, and minimize overall management costs for priority zones, we utilized Marxan v.4.0.6. This software is designed to identify priority conservation areas. The analysis was conducted using the QMarxan Toolbox (2.0.1), a plugin for QGIS 3.18.3. The conservation features of Arefi Island include three critical ecosystems: mangroves, seagrass, and coral reefs. These features were identified using a map generated from object-based image analysis. Arefi Island served as the primary study area, with a buffer zone extending from the coastline to encompass all shallow water habitats within this zone. The study area was divided into hexagonal planning units with a side length of 15 m, resulting in 9,531 planning units (PUs) within the area.

Conservation costs were estimated based on land status of the area or region, modified from Wijayanto, Yulianda & Imran (2021) and Watts et al. (2017), and presented in Table 3. Categories include: Resident Area (3), Land Use (3), Floating Net Cage (1), and Dock Area (1). Land use refers to the land cover on an island that is not identified as part of marine conservation areas.

This study examines the cost attributes of various human spatial utilization activities within the conservation area. Penalty scores are assigned to each cost attribute based on the significance of the activity, following Watts et al. (2017). Higher penalty scores indicate greater difficulty in designating the area as a core conservation zone. For instance, a penalty score of one is assigned to activities such as docks and floating net cages, while higher scores are given to land use and residential areas. These scores reflect the challenge of considering or reclassifying the area as a core zone (Wijayanto, Yulianda & Imran, 2021).

Both conservation features and cost attributes are assigned to each planning unit (PU) without normalization, ensuring that each PU contains values for both conservation features and costs. We then calibrated the Species Penalty Factor (SPF) and the BLM. The SPF was calibrated to appropriately scale the penalty for missing conservation features relative to one another. The BLM was adjusted to identify the optimal value that balances area compactness with cost. As the BLM value increases, the algorithm tends to favor a ‘single large’ design over ‘multiple small’ designs, thereby enhancing connectivity.

Agnew, Fryirs & Leishman (2024) highligted the practicality of using Marxan as an accessible tool to address complex prioritization challenges and to model landscape-scale rehabilitation scenarios over time. Similarly, Chan, Hoshizaki & Klinkenberg (2011) demonstrated that a 50% protection scenario effectively stabilized Marxan solutions for ecosystem services, while Delavenne et al. (2012) found that a 50% conservation target offers stronger ecosystem protection. This threshold is designed to provide optimal protection and ensure ecological sustainability. A comparison of this study’s findings with previous research reveals both advances in conservation planning and the ongoing need for refined spatial analysis in MPAs. For instance, Jones & De Santo (2016) emphasized the importance of integrating both ecological and social data to achieve biodiversity conservation and community benefits when designing the effective MPAs. While Jones & De Santo (2016) focused on balancing ecological and socioeconomic factors, our study concentrated on ecological values and their spatial distribution. This difference highlights the importance of a holistic approach that considers both ecological integrity and human well-being, suggesting that future studies should incorporate more comprehensive socioeconomic analyses to better align conservation efforts with community needs.

Therefore in this research, three EV scenarios were analyzed using QMarxan. Scenario EV I, aligned with Target 3 of the post-2020 Global Biodiversity Framework and IUCN guidelines, aims to protect 30% of identified conservation targets, including coral reefs, seagrass, and mangroves. This scenario is designed to enhance connectivity within the conservation area (IUCN, 2008) while supporting fisheries in the surrounding regions (Firmansyah et al., 2018).

Scenario EV II set a 40% protection target, following the recommendations of Noss et al. (2012) and consistent with Aichi Target 11 (Harris & Holness, 2023) which adopt targets of 10%, 30%, 40%, or 50%, for nature conservation to achieve biodiversity goals.

Aichi Target 11 emphasizes the need for protected areas and other effective conservation measures across geographic regions, including strictly protected zones as well as areas where sustainable use is permitted, as long as species, habitats, and ecosystem functions are adequately protected.

Scenario EV III adopted a 50% conservation target, following the ’Half-Earth’ concept, which advocates protecting 50% of conservation targets. This ambitious scenario aligns with the ecoregional approach proposed by Dinerstein et al. (2017), which seeks to preserve 50% of the terrestrial biosphere for global ecological heritage conservation.

The irreplaceability of each planning unit was measured based on the frequency with which it was selected across 1,000 iterations, with values ranging from 0 to 1,000. Units with higher irreplaceability scores were considered more important for conservation. Planning units scoring between 750 and 1,000 were designated as Priority I, indicating their critical importance for conservation. Units scoring between 500 and 750 were categorized as Priority II, while those with scores between 250 and 500 were labelled as Priority III. Units with scores between 0 and 250 were classified as Priority IV. Any unit with a score of zero was considered a nonpriority zone. Priority I areas were designated as core conservation zones, Priority II areas were allocated for tourism, Priority III areas were identified as fisheries zones, and Priority IV areas were set aside for other uses, such as coastal development. Next, the Priority I areas map (core conservation zones) is overlaid with the biophysical feature areas to identify important habitats.

We validated the outputs of the Marxan model by comparing the conservation areas generated with actual field conditions to ensure that the target species and ecosystems were present in the identified priority areas. This validation was essential to confirm the feasibility of implementing Marxan’s recommendations in the field. Additionally, we refined the model through iterative adjustments, such as modifying the SPF and testing various scenarios. This iterative process allowed us to develop a more robust and optimal conservation strategy.

We simulated three scenarios (30%, 40%, and 50%) with identical costs. Using the output from the Marxan operation, after 1,000 iterations for each scenario, we calculated: (1) the total number of selected planning units, (2) conservation costs, and (3) boundary length (BLM). Based on these results, we calculated conservation cost efficiency, defined as the number of planning units per unit of cost. A higher efficiency value indicates a more suitable scenario (Zhang & Li, 2022).

Conservation cost efficiency is calculated as the ratio of selected planning units to total conservation costs, serving as a metric to evaluate resource utilization in achieving conservation objectives. This method evaluates resource efficiency in achieving conservation goals by running Marxan simulations with multiple iterations (e.g., 1,000) across predefined conservation targets (e.g., 30%, 40%, and 50%). Each iteration generates data on selected planning units, associated conservation costs, and the boundary length modifier. Upon completion, the collected data includes total selected planning units (SPU) and their corresponding conservation costs (CC). Efficiency is calculated using the formula: ${\displaystyle \text{Conservation Cost Efficiency}=\text{Total Selected Planning Units}/\text{Total Conservation Costs}}$ This calculation allows for comparing the number of planning units selected per unit of cost across different scenarios. A higher result indicates a more efficient scenario in terms of cost-effectiveness in achieving conservation objectives. By applying this method, we can effectively evaluate and compare the efficiency of different conservation strategies using Marxan.

Satellite image analysis results

Satellite image analysis using OBIA identified three primary coastal ecosystems on Arefi Island: mangroves, seagrasses, and coral reefs. Additionally, the Marxan model incorporates parameters such as floating nets, residential areas, and docks for zoning analysis. These parameters are crucial for determining conservation priorities within the Marxan framework (Fig. 3). The total area of coastal ecosystems on Arefi Island is approximately 64.78 hectares Coral reefs cover 36.35% of this area, making them a significant component. Mangroves and seagrasses are also present but are distributed unevenly across the island. Mangroves, which cover 11.81 hectares (18.24% of the total area), are primarily concentrated in the southeast, while their presence is relatively sparse in residential areas. Seagrass beds, covering 29.42 hectares and constituting 45.41% of the total area, are dominant in the northern part of Arefi Island. The overall classification has a kappa accuracy value of 0.82. Additionally, the presence of a port and floating fish cages indicates local community activities such as shipping, fishing, and tourism.

Conservation priority area recommendations for Arefi Island

From the analysis of the maps presented in Fig. 4, three important areas—core zone, utilization zone and sustainable fishery zone—were identified with higher selection percentages. These areas are found in the northern, southeastern, and southwestern waters of Arefi Island. Notably, the eastern part of Arefi Island showed a lack of selected areas for conservation.

The spatial zoning arrangements for Arefi Island’s conservation areas under the three EV scenarios revealed significant differences in how space is allocated to optimize conservation priorities, as shown in Table 4.

This study compares three conservation scenarios—Ecological Value I (EV I), Ecological Value II (EV II), and Ecological Value III (EV III)—to evaluate the spatial allocation of core zones and their effectiveness in protecting key habitats, including coral reefs, seagrass beds, and mangroves. The analysis highlights the relationship between the extent of core zones and the level of habitat protection achieved. In EV I, the core zone covers 12.33 hectares, representing only 2.27% of the total area. This scenario provides minimal protection, conserving just 5.82% of coral reefs, 16.25% of seagrass beds, and 29.55% of mangroves (Table 5). A significant portion of the area (92.43%) remains allocated for general use, reflecting limited conservation prioritization.

EV II introduces an expanded core zone of 19.53 hectares, increasing its share to 3.65% of the total area. This expansion results in improved habitat protection, safeguarding 21.74% of coral reefs, 22.16% of seagrass beds, and 29.55% of mangroves (Table 5). Altogether, this scenario ensures the protection of 23.35% of key habitats, indicating a moderate enhancement in conservation efforts compared to EV I.

EV III represents the most ambitious conservation scenario, with the core zone covering 34.37 hectares, equivalent to 6.32% of the total area. This significant expansion enhances habitat protection dramatically, with 40.85% of coral reefs, 40.79% of seagrass beds, and 29.64% of mangroves included in the core zone (Table 5). In total, 38.76% of key habitats are safeguarded under this scenario, illustrating a strong commitment to conservation priorities.

The progression across the scenarios demonstrates a clear trend toward increasing habitat protection through larger core zones, with EV III achieving the most comprehensive conservation outcomes.

As shown in Table 4, increasing the proportion of protected conservation features across the scenarios leads to a corresponding rise in the number of conservation planning units designated as core and utilization zones. Conversely, it results in a reduction in the areas allocated for sustainable fisheries and other zones. Using the 30% conservation scenario (EV I) as a baseline, expanding the protection targets to 40% (EV II) and 50% (EV III) increased the core zone size by 58.39% and 178.75%, respectively.

However, the increase in utilization zone under EV III (34.99%) was less significant comparedto EV II (80.75%) from EV I. Both EV II and EV III led to a reduction in areas designated for sustainable fisheries and other uses. This shift reflects a deliberate reallocation of spatial zones, with EV III showing a substantial reduction in these areas to accommodate an expanded core zone. This reconfiguration highlights the increased prioritization of conservation as the other zones decrease in size, making room for more core conservation areas.

Table 5 presents the impact of different ecological value scenarios on habitat prioritization within the core zone. Under Ecological Value I, the core zone spans 9.64 ha, with mangroves comprising 29.55%, followed by seagrass (16.25%) and coral reefs (5.82%). In Ecological Value II, the core zone expands to 15.13 ha, increasing coral reef (21.74%) and seagrass (22.16%) coverage, while mangroves remain at 29.55%. Ecological Value III further enlarges the core zone to 25.12 ha, with coral reefs (40.85%) and seagrass (40.79%) becoming dominant, while mangrove coverage slightly declines to 29.64%. This trend demonstrates how ecological priorities influence the spatial distribution of key biophysical habitats.

As the scenarios progress from EV I to EV III, the core zone area increases significantly, incorporating larger proportions of coral reefs and seagrass. While mangrove areas remain constant across all scenarios, their percentage within the core zone decreases as the extent of other habitats expands. Scenario III achieves a more balanced representation of coral reefs and seagrass, whereas Scenario I places greater emphasis on mangroves relative to the core zone.

The results for the multitarget scenario are summarized in Table 6. As shown, increasing conservation targets leads to higher conservation costs and longer boundary lengths, although the pattern of conservation efficiency remains irregular.

Table 6 compares conservation scenarios based on selected planning units, costs, boundary length, and efficiency. The results reveal trade-offs among these factors. Scenario 3 selects the most units but incurs higher costs, while Scenario 2 is the most efficient (0.0062), indicating optimal resource allocation. Scenario 1 has the lowest efficiency (0.0045). These findings highlight the importance of balancing cost, spatial coverage, and efficiency in conservation planning.

The analysis of costs, length, and efficiency across the scenarios revealed that efficiency does not consistently correlate with cost. Scenario 2 achieves the highest efficiency despite having moderate unit count and costs, indicating a more effective allocation of resources compared to Scenarios 1 and 3. Conversely, Scenario 3 incurs the highest cost but does not deliver proportionally higher efficiency, suggesting diminishing returns as resource investment increases.