Uncovering Structural Drivers of Coastal Economic Resilience and Mangrove Sustainability: A MICMAC-Based Adaptive Governance Framework

Mangrove forests sit at the interface between terrestrial and marine social-ecological systems (SES) in 123 countries. A hectare can store up to 650 tonnes of carbon and support more than 5,700 dependent species [1]. They buffer coastlines, provide fisheries nurseries, and regulate the hydro-oceanographic processes coastal communities depend on. Global cover has still declined by more than a fifth in four decades: 677,000 hectares were lost between 2000 and 2020 alone, and roughly half of all mangrove ecoregions are now classified as threatened. Most of this loss is anthropogenic. Land-use conversion to aquaculture and agriculture caused 62% of global losses between 2000 and 2016, and 80% of that human-driven loss was concentrated in just six Southeast Asian countries [2]. That concentration points to a governance failure in coupled human-natural systems [3, 4], which is the basis for treating coastal management as an integrated SES problem rather than a purely ecological one.

Indonesia's Ministry of Environment holds the world's largest mangrove area, at 3.46 million hectares in 2026, but it is also among the countries with the steepest absolute losses, along with Saudi Arabia [5]. That scale of loss is why the government set a national target to rehabilitate 600,000 hectares. East Java shows the same trend at a smaller scale. Provincial data put current mangrove cover at 27,221 hectares, A broader East Java forest assessment reports that primary mangrove forest fell from ~15,000 ha in 1990 to ~9,000 ha in 2022, consistent with substantial long-term degradation even if that estimate is narrower than total mangrove cover [6]. Aquaculture-pond conversion and unregulated infrastructure development are the main drivers. The result is a double vulnerability: environmental carrying capacity declines through abrasion, saltwater intrusion, and biodiversity loss, and that decline weakens the economic resilience of small-scale fishers who depend on intact mangrove–fisheries links. This tension is most critical, and institutionally fragmented, within mangrove ecosystems officially designated as production forests. Unlike strictly protected areas, production forests are the primary arenas where the ecological carrying capacity must directly absorb the intensive livelihood demands of local fishing communities.

Prior research on coastal governance and mangrove conservation, in East Java and elsewhere, stays largely within single sectors and small spatial scales. A national review of Drivers, Pressures, State, Impact, and Response (DPSIR) applications across Indonesia's mangrove provinces found that most studies still treat drivers, pressures, and impacts as separate analytical layers, and called for integrated approaches that connect spatial degradation with social outcomes [7]. The same split shows up on the ground: in Banyuwangi's Pangpang Bay, overlapping authority between the Forestry Service and the Fisheries Service over the same coastal space has produced years of institutional friction, with no shared mechanism linking the regional spatial plan to the coastal zoning plan [8]. Even recent methodological advances have not closed this gap; a 2025 study combining DPSIR with participatory GIS in the Sundarbans was, by its own account, the first to give DPSIR a spatial dimension, yet it still works at the scale of individual respondents' perceptions rather than provincial biophysical data [9]. This persistence is not surprising: SES are complex, non-linear, and tightly coupled [3, 4], so a shock to the environmental subsystem feeds quickly into economic resilience, and poverty in turn pushes communities toward more destructive resource use. Macro-level policy usually addresses downstream symptoms, such as cash assistance for poor fishers, without touching the upstream drivers [10], and foresight tools such as DPSIR, however refined, share the same blind spot: indirect, cascading effects between ecological and economic subsystems go unmapped.

Structural analysis responds to this gap directly by shifting the analytical focus from linear cause-and-effect to the architecture of the entire system [11, 12]. By utilizing cross-matrix multiplication, it systematically traces hidden feedback loops and indirect pathways, revealing how a biophysical shock cascades through multiple socioeconomic intermediaries. The Cross-Impact Matrix Multiplication Applied to Classification (MICMAC) method, part of Michel Godet's prospective foresight tradition [13, 14], has mapped driving and dependent variables in strikingly different systems: sugarcane agribusiness strategy in Mexico [15] smart-city eligibility assessment in India [16], gender barriers in green vocational education in Colombia [17], sustainability and food security prospects in the coastal areas of East Java [18] and alternative-livelihood sustainability in Nembrala, East Nusa Tenggara [19]. Across this range, the method consistently separates driving, relay, and dependent variables regardless of sector, which is what makes it transferable to coastal SES in the first place. But every one of these studies, including the only Indonesian coastal case, operates at a single site, institution, or village. None scales up to a provincial diagnostic that combines biophysical degradation data with socioeconomic determinants in one model.

This study applies MICMAC structural analysis to open up that black box of environmental governance in coastal East Java. Therefore, the primary objective of this research is to systematically map the structural relationships between biophysical mangrove degradation and the socio-economic vulnerabilities of small-scale fishers operating within production forest ecosystems. Specifically, it seeks to answer: How do direct and indirect interactions among ecological, economic, and institutional variables dictate the livelihood resilience of small-scale coastal communities? By addressing this question, the contribution is a provincial-scale typological classification of coastal vulnerability that merges biophysical degradation data with microeconomic determinants, used to identify driving, relay, and dependent variables. The output is an adaptive policy zoning map, intended as an operational basis for SDG 1, 8, 13, and 14, and as a diagnostic template other mangrove-dependent SES facing similar governance fragmentation could adapt.

3.1 Direct influence and dependence classification

The direct influence/dependence map generated from the MDI classifies the twenty operationalized variables into four structural categories based on their driving power (Mi) and dependence (Di), following Godet's structural prospective methodology [9, 10]. The resulting distribution, four driving variables (B6, I2, I4, I6), eleven relay variables, four dependent variables (SE1, SE7, I1, I5), and a single excluded variable (SE5) reveals a system in which the relay quadrant is markedly dominant (see Figure 4 and Table 2). Quadrant assignment follows the standard threshold rule relative to the average driving power and average dependence computed across all twenty variables (M_avg = D_avg = 29.70; full scores reported in Table 3).

The MDI classifies twenty operationalized variables into four structural categories by driving power and dependence score. With four driving, eleven relay, four dependent, and one excluded variable, the relay quadrant is by far the most populated zone.

Table 2. Direct influence/dependence scores

tu_pian_4.png

Figure 4. Direct influence-dependence map

The four driving variables aquaculture land conversion (B6), mangrove policy effectiveness (I2), spatial planning compliance (I4), and cross-sectoral coordination (I6) carry high driving power and low dependence.

Eleven variables populate the relay quadrant: mangrove cover (B1), deforestation rate (B2), vegetation health index (B3), shoreline change (B4), blue carbon stock (B5), biodiversity index (B7), fisheries GRDP (SE2), spatial vulnerability (SE3), community participation (SE4), social capital (SE6), and anthropogenic pressure (I3). These variables absorb shocks from the driving tier and retransmit them downward, making the relay core the most structurally volatile zone in the system.

The four dependent variables, coastal poverty rate (SE1), sustainability standard compliance (SE7), fish landing infrastructure (I1), and law enforcement capacity (I5), present low driving power alongside high dependence.

Fisher household livelihood diversification (SE5) is the sole excluded variable, with low scores on both driving power and dependence. Network-level visualization reinforces the quadrant structure: relay variables form a dense central cluster that receives strong inputs from the peripheral institutional and land-use hubs, while law enforcement and livelihood diversification occupy marginal positions consistent with their dependent and excluded status.

3.2 Direct influence graph

To further substantiate the quadrant classification derived from the MDI, the structural data is visualized as a network topology through the direct influence graph (Figure 5). While the influence/dependence map plots the aggregated driving power and dependency of each variable, the direct influence graph explicitly maps the specific relational pathways and the intensity of connections between individual nodes. The visual encoding of the lines where the thickest red edges represent the strongest direct impacts and the solid blue edges indicate relatively strong influences provides an immediate diagnostic view of how systemic shocks are transmitted.

The direct influence graph provides a network-based representation of the structural influence pattern identified in the influence-dependence map. Edge weight and color—red for the strongest direct relationships, blue for moderately strong ones correspond to the MDI scoring matrix, making the graph a visual complement to the MICMAC quadrant classification [35]. The resulting network topology shows a system organized around a densely connected transmission core whose internal coupling is far greater than its periphery.

The highest concentration of crossing red and blue edges falls at the network's center, where deforestation rate (B2), anthropogenic pressure (I3), and mangrove cover (B1) converge. Policy effectiveness (I2) and aquaculture land conversion (B6), by contrast, occupy peripheral positions yet emit a disproportionate share of the strongest outgoing edges. Their structural role is that of primary driving variables: high driving power with low dependence, feeding continuously into the relay core. Law enforcement capacity (I5) and coastal village poverty rate (SE1) present the inverse configuration receiving many incoming edges from across the network while generating few of their own.

tu_pian_5.png

Figure 5. Direct influence graph

The network topology identifies the same small set of upstream variables I2, B6, I4, and I6 that the MICMAC quadrant and displacement analyses flagged as the system's primary control points.

3.3 Indirect influence graph

Building upon the immediate interactions mapped in the direct influence graph, the structural analysis advances to uncover the hidden, multi-step pathways within the system through the indirect influence graph (Figure 6). This network topology is generated mathematically by raising the initial MDI to successive powers, simulating how cascading effects and feedback loops propagate throughout the system over time. While the direct graph captures immediate, observable relationships, the indirect influence graph visualizes the ultimate structural architecture of the coastal SES.

3.4 Displacement analysis

The displacement map generated from the MICMAC analysis visualizes the trajectory of each variable as the computational model shifts from calculating direct influences to uncovering indirect and potential pathways (Figure 7). This step is critical in structural analysis, as complex SES often contain hidden feedback loops where a variable's true impact is mediated through multiple intermediate actors.

tu_pian_6.png

Figure 6. Indirect influence graph

tu_pian_7.png

Figure 7. Displacement map

Table 3. Direct vs. indirect influence and dependence scores

The quantitative shifts between direct and indirect scores for each variable are explicitly reported in Table 3. This table details the initial direct influence (MI) and dependence (DI) scores alongside the computed indirect scores, quantifying the exact displacement (ΔM and ΔD) as multi-step pathways are activated. As visualized in the displacement map (Figure 7) and supported by the quantitative data in Table 3, trajectory lines connecting the direct, indirect, and potential influence coordinates remain short for nearly all variables. No variable migrates across quadrant boundaries.

Among the driving variables (Quadrant I), governance indicators (I2, I4, I6) and aquaculture land conversion (B6) show minimal vertical displacement. Their driving power persists when indirect pathways are incorporated into the iterative matrix algorithm.

Within the relay quadrant (Quadrant II), variables including the vegetation health index (B3), blue carbon stock (B5), and social capital (SE6) shift toward higher dependence or reduced influence when indirect pathways are activated. As these indirect effects propagate through the coastal system, relay variables become more embedded in the feedback structure.

The dependent quadrant (Quadrant III) yields particularly clear results. Law enforcement capacity (I5) and coastal village poverty rate (SE1) show distinct horizontal displacements under indirect analysis, confirming that both are structurally downstream in this system. For poverty, the indirect mapping traces its accumulation through declining fisheries productivity (SE2) and reduced ecological carrying capacity. For law enforcement, the rising indirect dependence on cross-sectoral coordination (I6) and policy effectiveness (I2) is evident.

Fisher household livelihood diversification (SE5) shows no displacement under either direct or indirect analysis. This is a null result worth examining carefully. Even through multi-step hidden pathways, diversification has no measurable structural connection to the broader coastal ecosystem dynamics.

The displacement analysis shows that the hidden pathways in this SES consolidate structural influence rather than redistribute it. Indirect loops increase the effective driving power of governance failures and tighten the dependence of ecological and economic indicators on upstream conditions.

3.5 Adaptive policy zoning map for coastal SES

Building on the quadrant classifications derived from this study, three spatially differentiated management zones are proposed across the five administrative regencies. As visualized in the GIS-based zoning map (Figure 8), these zones are geographically delineated using spatial criteria corresponding to the structural role of the variables concentrated within them. To ensure spatial consistency, these three policy zones are explicitly mapped onto the three ecological clusters defined in the methodology.

The first zone (corresponding to the eastern ecological cluster) covers coastal areas subject to unregulated aquaculture expansion and documented jurisdictional overlap, with Pangpang Bay Banyuwangi being a representative case. The target variables (B6, I2, I4, I6) are Quadrant I drivers whose positional persistence under indirect analysis confirms their function as primary control parameters for the broader SES.

The second zone (corresponding to the northern ecological cluster) targets actively degrading production forests where biophysical decline has measurably reduced local economic resilience notably the heavily abraded estuaries along the northern coast covering Pasuruan and Probolinggo. The relay variables here (B1, B2, B3, B5, B7, I3, SE4, SE6) showed heightened sensitivity to upstream shocks under displacement analysis, and anthropogenic pressure within this zone is partly generated internally through economic incentives.

The third zone (corresponding to the southern ecological cluster) covers marginalized fishing villages in the southern coast of Malang and sections of Situbondo, where the accumulated effects of upstream failures are most visible. Dependent variables (SE1, SE7, I5, I1) and the excluded variable (SE5) define this zone's target portfolio. The displacement analysis shows poverty in this zone accumulates through declining fisheries productivity and reduced ecological carrying capacity; it is a downstream indicator, not an upstream driver.

tu_pian_8.png

Figure 8. GIS-based adaptive policy zoning map across the five coastal regencies of East Java

This study finds that the coastal mangrove system in East Java is governed structurally by four upstream variables—aquaculture land conversion (B6), policy effectiveness (I2), spatial planning compliance (I4), and cross-sectoral coordination (I6)—that precede and condition the ecological and socioeconomic outcomes most visible to policymakers. Displacement analysis confirmed that these variables retain their driving positions under both direct and indirect pathway analysis, indicating that their structural primacy is embedded in the system's architecture rather than a function of first-order expert scoring. The eleven relay variables, including mangrove cover, blue carbon stock, social capital, and anthropogenic pressure, form a densely coupled transmission core that amplifies and retransmits shocks originating from the driving tier. Coastal poverty, enforcement capacity, and infrastructure accumulate at the dependent end of this structure as products of upstream conditions. Livelihood diversification remains structurally isolated throughout, pointing to an institutional gap that current coastal programs have not resolved.

The main implication of these findings is what they contradict. Government programs in coastal East Java have historically concentrated resources on the dependent tier—cash transfers, infrastructure investment, and site-level ecological restoration—while the institutional and land-use conditions generating system-wide degradation have remained structurally unaddressed. The analysis shows that this sequencing cannot produce durable outcomes while the driving variables continue operating undisturbed. The three-zone adaptive policy framework proposed here—upstream institutional reform, relay-zone co-management with ecological rehabilitation, and dependent-zone outcome monitoring—offers a structurally grounded alternative to this pattern.

The relevance of these findings extends beyond the five coastal regencies and three ecological clusters examined in this study. Indonesia's decentralized governance system has produced analogous jurisdictional fragmentation in mangrove-adjacent coastal systems across the archipelago, and the prospective structural approach applied here offers a transferable diagnostic method for identifying governance leverage points in those contexts. More broadly, the finding that institutional and land-use variables occupy the upstream driving position provides empirical support for centering governance reform in coastal management policy, complementing a growing body of evidence that governance capacity constitutes the binding constraint in mangrove conservation outcomes.

Three limitations warrant acknowledgment. The MDI rests on expert elicitation, which introduces scoring subjectivity; livelihood diversification's isolated position may partly reflect conservative assessments of its indirect linkages rather than a genuine structural absence. The analysis also produces a structural snapshot rather than a temporal trajectory, and whether the driving variables' positional dominance persists through active governance reform is an empirical question the MICMAC method cannot resolve without longitudinal validation. Third, because the structural matrix was generated through a consensus-oriented FGD rather than independently scored and retained individual matrices, a formal inter-rater agreement coefficient (e.g., Kendall's W) could not be calculated; agreement was instead addressed procedurally through structured deliberation, which we acknowledge as a methodological limitation relative to designs that preserve individual expert scores.

Future work should move in two directions. Within the study area, participatory re-scoring workshops and field monitoring of relay variables can test whether the structural predictions hold under observed management conditions. Beyond East Java, applying this analytical framework to comparable coastal governance contexts elsewhere in Indonesia would determine whether the institutional dominance identified here reflects a site-specific configuration or a systemic feature of the country's coastal governance landscape.