The king will tell us: traditional authority and the politics of trust in cyclone preparedness Open Access

The intensification of extreme weather events under climate change has placed early warning systems (EWS) at the centre of disaster risk reduction (DRR). Cyclones, floods, and droughts are becoming more frequent and severe, threatening lives, livelihoods, and development gains, particularly in low- and middle-income countries (IPCC, 2023; UNDRR, 2022). Global frameworks such as the Sendai Framework for Disaster Risk Reduction 2015–2030 and the “Early Warnings for All” initiative emphasise the need for inclusive, people-centred systems (UNDRR, 2023). However, despite substantial investments in forecasting technologies and communication infrastructure, failures remain widespread. These breakdowns often lie not in the absence of data, but in the erosion of trust, legitimacy, and authority within the “last mile” of communication (Wisner et al., 2004; Lindell and Perry, 2012).

In this paper, “people-centred” early warning systems refer to approaches that prioritise accessibility, trust, cultural relevance, and local ownership rather than relying solely on technological solutions (UNDRR, 2023). Likewise, the “last mile” refers to the final stage of warning transmission, where messages reach households, often representing the most fragile part of the system due to social, relational, and institutional barriers. These concepts emphasise that early warning effectiveness depends not only on scientific accuracy but also on who communicates the message and how it is interpreted within local governance structures.

In cyclone-prone Madagascar, these limitations are particularly acute. The island experiences three to four cyclones annually, which have a devastating impact on housing, agriculture, and infrastructure (WFP, 2023; FAO, 2022). Despite the establishment of the Bureau National de Gestion des Risques et des Catastrophes (BNGRC), communities in rural districts frequently report warnings that arrive late, inconsistently, or in forms they perceive as untrustworthy (World Bank, 2021; INSTAT, 2021). Past experiences of false alarms, unmet promises, and poor service delivery have heightened scepticism toward formal channels such as SMS alerts and radio broadcasts (Gaillard and Mercer, 2013; Kelman et al., 2009). In contrast, warnings issued by traditional leaders, including fokontany chiefs and kings (mpanjaka), are often acted upon immediately, demonstrating that messenger legitimacy can be as decisive as technical accuracy (Artur and Hilhorst, 2012; Chinsinga, 2011).

This reliance on traditional leaders reflects broader patterns of hybrid governance across Africa and Asia, where state institutions coexist with customary systems of authority (Boege et al., 2009; Meagher, 2012). In rural Madagascar, leaders retain central roles in community life—resolving disputes, organising labour, and mediating external interventions (Englebert, 2002; Logan, 2013). Their legitimacy, rooted in ancestry and daily proximity to communities, contrasts with state officials, often appearing only during crises, reinforcing perceptions of detachment (Marcus and Ratsimbaharison, 2005). Consequently, the phrase often heard in fieldwork, “the king will tell us,” Captures the political reality of warnings filtered through cultural authority. By foregrounding messenger legitimacy and hybrid governance, this paper examines how traditional leaders shape cyclone preparedness in fragile contexts. It argues that disaster governance must formally recognise these actors to design inclusive, trusted, and effective early warning systems (Aldrich and Meyer, 2015; Pelling and Dill, 2010; Sutton and Tierney, 2006).

Madagascar is among the most cyclone-prone countries in the world, facing on average three to four cyclones annually that destroy homes, farmland, and infrastructure (WFP, 2023; FAO, 2022). Its geographic location in the south-west Indian Ocean exposes it to increasing tropical cyclones, devastatingly impacting vulnerable rural communities whose livelihoods depend heavily on agriculture (UNDRR, 2022; IPCC, 2023). These risks are compounded by poverty, limited infrastructure, and inadequate state capacity, heightening disaster vulnerability (World Bank, 2021). The uneven distribution of resources across urban and rural areas further deepens exposure, leaving marginalised rural populations most at risk (INSTAT, 2021; Marcus and Ratsimbaharison, 2005).

The institutional landscape of disaster risk management in Madagascar is dominated by the Bureau National de Gestion des Risques et des Catastrophes (BNGRC), which was established to coordinate risk governance and early warning dissemination. While BNGRC has improved national preparedness, its reach into rural communities remains limited due to infrastructural challenges, weak communication channels, and a lack of public trust in state agencies (UNDRR, 2023; Wisner et al., 2004). State presence is often sporadic and highly visible only during emergencies, undermining its legitimacy and effectiveness in long-term disaster governance (Lindell and Perry, 2012; Pelling and Dill, 2010). Communities frequently perceive official forecasts and warnings as inconsistent or irrelevant to their local realities, and past experiences of false alarms or unmet promises reinforce scepticism (Kelman et al., 2009; Gaillard and Mercer, 2013).

In this context, traditional leaders such as fokontany chiefs and kings (mpanjaka) remain central to community life, mediating disputes, organising collective labour, and legitimising external interventions (Englebert, 2002; Logan, 2013). Their legitimacy derives from ancestry and daily presence within communities, enabling them to act as trusted interpreters of risk in ways that state actors cannot (Boege et al., 2009; Meagher, 2012). This dynamic reflects broader patterns of hybrid governance in Africa and Asia, where state authority coexists with customary power, creating overlapping but complementary systems of legitimacy (Chinsinga, 2011; Artur and Hilhorst, 2012). In Madagascar, warnings endorsed by these leaders are frequently acted upon more swiftly than those issued by formal institutions, underscoring the decisive role of messenger credibility in disaster response (Sutton and Tierney, 2006; Aldrich and Meyer, 2015).

These dynamics illustrate how the politics of trust and legitimacy shape disaster preparedness in fragile contexts. The persistence of hybrid governance challenges assumptions that early warning is only a matter of technical accuracy or infrastructure. Instead, messenger authority and community trust must be recognised as equally vital in last-mile communication. This background highlights the need to examine how traditional authority influences preparedness in cyclone-prone Madagascar, providing a foundation for the subsequent literature review and analysis.

The study of early warning systems (EWS) in disaster risk reduction (DRR) has evolved over the past 3 decades, shifting from purely technical approaches to an emphasis on social and cultural dimensions of communication. A growing body of scholarship indicates that warning effectiveness depends not only on accurate forecasts but also on trust, credibility, and the legitimacy of the messenger (Dynes, 2005; Sutton and Tierney, 2006). In fragile governance contexts such as Madagascar, where formal institutions are weak, traditional authorities play a central role. This review situates the study within four debates: (1) traditional authority in disaster governance, (2) messenger legitimacy, (3) hybrid governance, and (4) relational trust and risk behaviour.

Traditional leaders remain influential across the Global South, rooted in ancestry and cultural legitimacy (Englebert, 2002; Logan, 2013). Their authority influences how communities perceive and manage risk. In Malawi, chiefs disseminate flood warnings that residents often treat as more binding than radio alerts (Chinsinga, 2011). In Mozambique, régulos blend forecasts with indigenous indicators (Artur and Hilhorst, 2012). Similar patterns are observed in Southeast Asia and the Pacific, where barangay captains, church leaders, and village heads serve as trusted intermediaries (Mercer et al., 2009). These authorities often function as “twilight institutions” between state and society (Lund, 2006). In contexts like Madagascar, they fill governance gaps and act as de facto disaster managers.

Additional research underscores the importance of integrating traditional and scientific knowledge. Twigg (2015) argues that last-mile challenges are deeply social, shaped by legitimacy rather than infrastructure. Participatory approaches enhance compliance when community leaders are involved in designing warning processes (Le Dé et al., 2015). The Pacific experience shows that collaboration between customary and state actors can either reinforce or undermine evacuations (Handmer and Choong, 2006; Paton et al., 2008). These insights underscore the value of culturally grounded EWS.

Warning uptake depends heavily on who delivers the message (Burningham et al., 2008). Messenger legitimacy, being perceived as credible and culturally authoritative, is critical (Lindell and Perry, 2012). In fragile contexts, formal institutions lack legitimacy due to political instability and service failures (Marcus and Ratsimbaharison, 2005). Communities thus rely on traditional leaders who are deeply embedded in local life. Farmers in Ghana trust chiefs more than extension officers during droughts (Amanor-Boadu et al., 2019), and during Sierra Leone’s Ebola outbreak, advice gained traction only when endorsed by chiefs (Richards, 2016). Messenger legitimacy is political and relational, explaining why some warnings are heeded and others ignored.

The coexistence of state and customary authority in disaster contexts reflects hybrid governance, where formal and informal institutions negotiate roles in managing public affairs (Boege et al., 2009; Meagher, 2012). Hybrid governance extends the reach of formal systems but also introduces tensions when warnings are reinterpreted, delayed, or contested. In Mozambique, chiefs convene village assemblies to translate meteorological forecasts into culturally meaningful advice (Artur and Hilhorst, 2012), while in Tanzania, mosque announcements and elders’ messages complement SMS alerts to strengthen dissemination (Ndetto and Matzarakis, 2014). However, hybridity can also produce conflict, as seen in Fiji, where chiefs resisted government evacuation orders (Handmer and Choong, 2006). For Madagascar, hybrid governance is not optional but inevitable: state cyclone alerts are filtered through traditional leaders whose legitimacy anchors community responses. Understanding this negotiated, multi-actor landscape is crucial for evaluating the effectiveness of early warning systems.

Trust is not a static attribute but a relational construct built through familiarity, accountability, and shared lived experience (Kasperson et al., 1988; Bankoff, 2004). Communities are more likely to act on warnings from leaders with whom they share daily life and mutual obligations, rather than from distant state officials (Wisner et al., 2012; Tierney, 2014). In Madagascar, reliance on kings and fokontany chiefs illustrates how relational trust transforms warnings into moral obligations, often framed in spiritual or ancestral terms (Gaillard and Texier, 2010). By contrast, SMS alerts and NGO campaigns are often disregarded as impersonal and detached. Relational trust thus explains why messenger legitimacy endures without formal institutional recognition, and why culturally grounded warnings may prompt quicker action than technically precise but socially distant messages.

Although studies highlight the legitimacy of messengers and hybrid governance, gaps remain. Research often assumes uniform trust in traditional leaders, overlooking variations in trust across different demographics, including gender, generation, and socioeconomic status. Tensions arising from conflicting instructions between formal and customary authorities are underexplored. Global DRR frameworks such as EW4All prioritise technology while downplaying legitimacy and trust (UNDRR, 2023). This study addresses these gaps by examining the intersection of messenger legitimacy, hybrid governance, and relational trust in shaping the uptake of warnings in Madagascar. Together, these strands conceptualise EWS as both technical and socio-political systems, embedded in culture and everyday authority structures.

This study employed a qualitative case study design to explore how messenger legitimacy, hybrid governance, and relational trust shape early warning responses in Madagascar. Two cyclone-prone districts, Marovoay and Manakara, were purposively selected to provide contrasting contexts. Marovoay represents an inland floodplain with agricultural livelihoods, and Manakara is a coastal district exposed to storm surges. These sites were chosen based on their high hazard exposure, strong presence of traditional authority structures, and varied geography and livelihoods, offering comparative insights (Seawright and Gerring, 2008; BNGRC, 2022).

Participants for both interviews and focus groups were selected purposively from the settlements most frequently affected by cyclones in Marovoay and Manakara. Local disaster officers and fokontany leaders assisted in identifying individuals with direct experience in issuing or receiving warnings, ensuring variation across gender, age, and livelihoods. Focus-group participants were drawn from specific villages where traditional leaders play an active role in warning dissemination, as these sites allowed the study to capture community-level interpretations of legitimacy and trust. Focus groups were used in conjunction with interviews to facilitate triangulation, as interviews provided individual perspectives—especially from chiefs, kings, and officials and focus groups revealed shared narratives and collective decision-making processes essential for understanding warning uptake.

Data collection, conducted between September and November 2024 under the REPRESA (Resilience and Preparedness to Tropical Cyclones across Southern Africa) project which operates in Madagascar, Mozambique, and Malawi and aims to strengthen multi-hazard early-warning systems, hydrodynamic flood modelling, and community preparedness for tropical cyclones included 24 key informant interviews comprising 14 authority figures (chiefs, kings, disaster officers, and institutional actors) and 10 ordinary residents, along with women’s leaders, teachers, religious figures, and officials from relevant departments; four focus group discussions of 10–12 participants each; and direct observation of community meetings. Interviews and discussions were conducted in local dialects, translated into French and English through back-translation processes to ensure accuracy (Temple and Young, 2004). A purposive sampling strategy was employed to ensure diversity across gender, age, and livelihoods, thereby securing representation of both authority figures and ordinary residents (Patton, 2015).

Thematic analysis guided the interpretation of the data (Braun and Clarke, 2006). Transcripts were coded for recurring themes such as authority, trust, messenger legitimacy, and behavioural responses, which were then synthesised under the study’s theoretical framework. The study was guided by a three-pillar theoretical framework comprising messenger legitimacy, hybrid governance, and relational trust, which together explain how authority, credibility, and social relationships shape community responses to cyclone warnings. Cross-case comparisons highlighted similarities and differences: while both districts relied heavily on traditional leaders, Manakara residents framed warnings spiritually, whereas Marovoay communities emphasised agricultural impacts.

Ethical integrity was maintained by securing informed consent, ensuring anonymity, and separating community members from chiefs during focus groups to reduce coercion risks. The Faculty of Natural and Agricultural Sciences ethics committee at North-West University approved the research and complied with the Belmont Report’s principles of respect, beneficence, and justice (1979). Limitations included potential translation loss, reliance on recollections outside the cyclone season, and the non-generalizable nature of qualitative findings. However, triangulation across interviews, focus groups, and observation, combined with comparative analysis, strengthened the study’s validity (Miles et al., 2014). Overall, the methodology provided a robust foundation for examining the role of messenger legitimacy and hybrid governance in disaster preparedness in Madagascar.

Findings from Marovoay and Manakara indicate that the dissemination of early warnings in rural Madagascar is more influenced by the social legitimacy of messengers than by technical factors. Four themes emerged: (1) trust in traditional leaders over state actors, (2) the symbolic legitimacy of the mpanjaka, (3) parallel systems of warning dissemination, and (4) frictions and opportunities for integration.

A consistent finding across both districts is that community members place greater trust in traditional leaders than government or NGO sources when deciding how to respond to cyclone warnings. Respondents expressed scepticism toward official channels such as SMS alerts or radio announcements, citing inaccuracy, impersonality, and lack of follow-through.

One elder in Lokomby remarked, “We hear the radio, but we wait for the king to confirm. When he speaks, we know it is true.” This statement reflects trust in the mpanjaka and distrust in state institutions. Many participants recounted experiences where official warnings did not materialise into severe events or relief failed to arrive, leading to perceptions of government unreliability. This aligns with Gaillard and Mercer’s (2013) observation that past experiences of false alarms or unmet promises undermine credibility in formal early warning systems.

Participants also expressed limited trust in NGOs, noting that many organisations were visible only during emergency periods and often did not return after distributing relief. This inconsistency created perceptions of weak accountability, with several respondents describing NGOs as “passing actors” who appeared briefly without long-term engagement. Because NGOs were not embedded in everyday community life, their warnings and preparedness messages were often perceived as less credible and reliable than those delivered by customary leaders who maintain daily presence and social accountability.

The embeddedness of traditional leaders in everyday community life enhances their legitimacy. Fokontany Chiefs and kings share daily struggles with residents, are visible in communal rituals, and are accountable through lineage-based legitimacy. In contrast, state officials are often viewed as outsiders, appearing only during crises. This echoes Dynes’ (2005) finding that people rely more on familiar and proximate figures than distant institutions in emergencies.

Comparative evidence supports this dynamic. In Malawi, Chinsinga (2011) found that villagers often disregarded official flood warnings unless confirmed by local chiefs. In Ghana, Amanor-Boadu et al. (2019) reported that farmers placed greater trust in traditional authorities than in extension services when responding to drought forecasts. Similarly, in Sierra Leone’s Ebola crisis, Richards (2016) noted that health advisories gained traction only when endorsed by chiefs. These cases affirm that messenger legitimacy, rather than information accuracy alone, determines the uptake of warnings in fragile governance contexts.

Patterns of messenger legitimacy varied across social groups. Women often relied on women’s leaders and church groups, while men tended to defer to fokontany chiefs or lineage-based elders. Younger residents followed radio and SMS alerts but still sought traditional validation, whereas older residents placed greater trust in traditional authorities, such as kings and senior elders. Livelihood also shaped perceptions: coastal fishers valued the mpanjaka’s spiritual authority, while inland farmers emphasised the chief’s administrative role. Trust likewise differed among traditional leaders themselves, with kings being seen as moral–spiritual authorities and chiefs as procedural–governance figures. These intersections show that messenger legitimacy is not uniform but influenced by gender, age, and livelihoods, with important implications for targeted DRR communication.

Beyond functional trust, the mpanjaka (traditional king) commands symbolic and cultural authority that amplifies the weight of his pronouncements. In Manakara, participants emphasised that evacuation or relocation was unthinkable unless explicitly instructed by the king, even when official warnings had already been issued. This reliance is rooted in both epistemological and cultural dimensions. One informant explained: “The cyclone is not just wind—it is anger. The king tells us when we must move to calm the ancestors.” Such statements highlight that warnings are interpreted not only as technical information but as culturally mediated acts, tied to cosmologies of ancestors, spirits, and moral obligations.

The symbolic role of traditional leaders is well documented in disaster studies. Gaillard and Texier (2010) note that religious and cultural authorities often frame disasters as spiritual or moral phenomena, shaping how communities interpret and act on warnings. Bankoff (2004) similarly emphasises that risk is socially constructed through cultural narratives, which require messengers capable of articulating them. In Madagascar, the king functions as both a spiritual mediator and a political authority, embodying what Lund (2006) describes as a “twilight institution” that straddles the boundaries of state, culture, and society.

The cultural legitimacy of the mpanjaka means that his messages are not merely advisory but obligatory. Failure to heed them risks not only physical danger but also spiritual disobedience. This adds moral weight to warnings, ensuring compliance in ways that technical alerts cannot achieve. Such symbolic authority is less documented in technical EWS literature but is critical for understanding why communities act or fail to act on warnings in culturally rich contexts.

Field data reveal that most communities receive cyclone warnings through formal and informal channels. Official alerts typically arrive via radio broadcasts, SMS, or announcements from local disaster management committees. However, these are rarely acted upon until validated by a trusted traditional authority. Although communities wait for the mpanjaka or fokontany chief to validate warnings, these authority figures themselves rely primarily on the same formal sources: radio broadcasts, SMS alerts, and official bulletins from the BNGRC and district disaster committees to obtain initial information. Chiefs reported that they monitor these channels closely during cyclone season and only convene discussions or community meetings once they have cross-checked the message across multiple sources or confirmed it with other local leaders. This means that authority figures make active judgements about when a warning is credible enough to relay, balancing the urgency of the alert with the need to avoid unnecessary alarm. As a result, the decision to disseminate a warning is both an informational and a social process shaped by verification, perceived severity, and expectations of community trust.

Many communes, especially in Manakara, have a structured process: the fokontany chief receives the formal warning, consults the king, and then convenes a communal meeting to issue the validated warning. Only after this process do households prepare for evacuation or secure their property. This layered dissemination system provides redundancy, ensuring that warnings reach most households, but it also introduces delays that may reduce lead times in fast-moving events. Field accounts suggest that these verification delays can range from several hours to nearly a full day, depending on the availability of the mpanjaka and the time needed to convene a council or community meeting. While this process strengthens trust and ensures that warnings are believed and acted upon, it also compresses the effective time households have to prepare, limiting opportunities to secure property, protect livestock, or evacuate vulnerable members. This illustrates a clear trade-off: communities prioritise social legitimacy over speed, even though doing so can reduce their actionable lead time during rapidly escalating cyclone events.

This dynamic reflects what scholars describe as hybrid governance (Pelling and Dill, 2010; Boege et al., 2009). Official systems provide technical forecasts, while traditional leaders filter, translate, and authorise them. Such arrangements are not unique to Madagascar. In Mozambique, Artur and Hilhorst (2012) found that chiefs often blended meteorological forecasts with indigenous indicators such as bird migrations or river levels before issuing communal warnings. Mercer et al. (2009) documented similar practices in which village heads combined local cosmologies with formal hazard information in Indonesia.

The parallel systems also highlight the persistence of what Twigg (2015) calls “last-mile challenges.” These are not solely logistical (distance, infrastructure) but also socio-political, as warnings must pass through culturally legitimate messengers before being acted upon. While redundancy is beneficial, it requires careful integration to avoid fatal delays. This pattern challenges the assumptions of the information deficit model, which suggests that people fail to take protective action because they lack sufficient or accurate information. The model assumes that providing more technical warnings will automatically improve preparedness. However, both our findings and international evidence show that trust in the messenger, cultural legitimacy, and social validation are far stronger determinants of action than the volume or clarity of information alone.

Traditional leaders are indispensable yet absent from formal governance frameworks, creating friction. Disaster officers noted that communities “wait for the chief” before acting. One official lamented: “We send the alert, but they wait for the chief. We are not always welcome unless we work with them.” Such comments highlight institutional tensions: while state agencies hold legal authority, customary leaders hold practical legitimacy. This reflects institutional tension: state agencies have legal authority, but customary leaders hold practical legitimacy. Examples of constructive collaboration also exist. In Lokomby, traditional leaders joined disaster committees and participated in simulations, thereby boosting public confidence. Comparative evidence shows both conflict (chiefs resisting government orders in Fiji) and cooperation (chiefs and churches enhancing compliance in Vanuatu) (Handmer and Choong, 2006; Paton et al., 2008). For Madagascar, formalising the role of traditional leaders could mitigate friction and leverage their legitimacy. This does not require bureaucratising them but recognising their functional role as intermediaries. As Wisner et al. (2012) argue, resilience requires engaging existing social structures rather than bypassing them.

The findings reinforce and extend the theoretical frameworks guiding this study. Messenger legitimacy was consistently decisive in shaping warning uptake, confirming critiques of the “information deficit” model (Burningham et al., 2008). Hybrid governance was evident in the parallel operation of state and customary systems, highlighting both synergies and frictions (Boege et al., 2009; Pelling and Dill, 2010). Relational trust emerged as the glue connecting communities to their leaders, underscoring the social construction of risk (Bankoff, 2004; Wisner et al., 2012).

In addition, the results resonate strongly with the two-step flow communication model developed in risk communication theory (Katz and Lazarsfeld, 1955). This model proposes that information transmitted through mass media does not directly influence individuals but first shapes opinion leaders’ views. These figures are more knowledgeable, trusted, and engaged within communities. These opinion leaders then interpret, adapt, and disseminate the information to the wider public (Rogers, 2003). In the context of Madagascar, traditional authorities such as kings and fokontany chiefs clearly function as these opinion leaders. Formal cyclone alerts issued through radio or SMS are filtered through them before gaining legitimacy at the community level. Thus, the flow of communication can be conceptualised as: Mass Media → Traditional Leaders (Opinion Leaders) → General Public. Incorporating this model helps to explain why early warnings delivered directly by state agencies often fail. In contrast, when validated and reframed by trusted leaders, the same messages trigger timely protective action.

Importantly, the Madagascar case illustrates that these frameworks must be understood together. Messenger legitimacy cannot be divorced from hybrid governance, as traditional leaders derive their authority from cultural embeddedness within hybrid landscapes. Similarly, relational trust explains why legitimacy endures despite the absence of formal institutional recognition. This integrated approach contributes a richer understanding of how early warnings operate in fragile contexts.

Table 1 below summarises how messenger legitimacy, hybrid governance, and relational trust intersect in Madagascar’s cyclone preparedness, and their implications for designing early warning systems.

The findings from Marovoay and Manakara highlight that messenger legitimacy, hybrid governance, and relational trust are decisive factors in shaping community responses to cyclone early warnings. These insights carry important implications for policy and practice in Madagascar and other disaster-prone contexts where formal institutions are weak and traditional authority structures remain influential. This section outlines five key recommendations.

Current early warning system (EWS) frameworks emphasise infrastructure, speed, and accuracy while neglecting the social terrain through which warnings travel. However, this study demonstrates that who delivers the message can be as important as the content itself. Policymakers should therefore integrate messenger legitimacy as a central design principle in EWS.

This means involving trusted community figures such as fokontany chiefs, mpanjaka, and religious leaders in dissemination, planning, and simulation exercises. Their participation ensures that warnings are culturally resonant and more likely to elicit compliance. Donner and Rodríguez (2008) and Lindell and Perry (2012) emphasise that warnings framed by credible messengers generate higher protective action rates. By institutionalising the involvement of local leaders, governments can move beyond technocratic approaches toward people-centred systems.

Traditional leaders in Madagascar already play a critical but informal role in early warning dissemination. Their authority is acknowledged in practice by communities but remains invisible in national policy. This creates tensions with state actors and limits opportunities for structured collaboration.

Policy frameworks should therefore formally recognise the contributions of traditional leaders within disaster governance. This recognition does not imply bureaucratising customary authority but integrating it functionally into preparedness structures. Possible measures include:

1. Formal agreements between local governments and traditional councils.

2. Representation of chiefs on local disaster risk management committees.

3. Capacity-building programmes that strengthen leaders’ understanding of scientific forecasts while respecting cultural knowledge systems.

Similar approaches have been successful elsewhere. In Vanuatu, chiefs are formally included in disaster committees, resulting in higher compliance rates (Paton et al., 2008). In Mozambique, régulos collaborate with NGOs to translate forecasts into local action (Artur and Hilhorst, 2012). Madagascar could adapt these models, institutionalising hybrid governance rather than leaving it ad hoc.

Disaster preparedness must engage both scientific forecasts and cultural cosmologies. Communities often interpret cyclones as meteorological events and spiritual phenomena tied to ancestral anger or moral imbalance. Ignoring these perspectives risks undermining trust and compliance. EWS protocols should therefore allow space for traditional interpretations alongside technical forecasts. For example, messages might acknowledge both meteorological predictions and the need for spiritual protection rituals. This integration of “dual epistemologies” mirrors recommendations by Mercer et al. (2010), who advocate blending indigenous and scientific knowledge for effective DRR.

Far from diluting scientific authority, such integration enhances community ownership and ensures warnings resonate with cultural realities. Policymakers can design systems that people are more likely to trust and act upon by recognising that risk is socially constructed and physically modelled. Nevertheless, a persistent implementation gap remains between official forecasts and the protective actions taken at the community level. This gap underscores the need to embed people-centred approaches and the integration of dual epistemologies more strongly within national DRR laws and regulations. Without clear legal mandates and institutionalised mechanisms, aligning scientific forecasts with cultural cosmologies risks remaining rhetorical rather than transformative. Ensuring that legislation and policy frameworks explicitly require the inclusion of customary leaders and traditional interpretations would help close this gap and strengthen compliance at the community level.

Simulation exercises (SIMEX) are powerful tools for building preparedness and trust. However, in Madagascar, these exercises are often led by NGOs or state officials without the active participation of cultural leaders. The findings suggest that joint drills involving chiefs, kings, and spiritual leaders alongside formal responders can significantly improve confidence in warnings. Residents in Lokomby, for example, reported greater willingness to evacuate when they had previously seen traditional leaders participate in drills. Simulation exercises thus serve not only to test systems but also to symbolically demonstrate alignment between state and customary authorities. Policymakers should scale up such culturally embedded exercises, using them as opportunities to bridge formal–informal divides (Shaw and Izumi, 2014).

Global DRR discourse often frames the “last mile” as a logistical challenge, ensuring that remote communities physically receive information. However, this study shows it is equally a social challenge: communities may receive warnings but still disregard them if the messenger lacks legitimacy.

Policymakers should reconceptualise last-mile communication as a relational challenge requiring co-design with local actors. This implies:

1. Mapping not only the communication infrastructure but also the trust networks.

2. Identifying key authority figures who act as gatekeepers of risk information.

3. Co-producing warning messages that align with local languages, norms, and cosmologies.

By adopting a social lens, Madagascar can build systems that not only reach communities but also persuade them to act, echoing Twigg's (2015) call for inclusive DRR strategies.

However, relying solely on opinion leaders carries risks. If a leader is misinformed or biased, communities may receive distorted messages, which can reduce their preparedness. Overreliance on a few intermediaries may also marginalise alternative voices. Leadership structures are dynamic and context-specific; individuals trusted in one crisis may not hold the same authority in another, requiring continuous mapping of leadership networks.

While traditional authorities are trusted, participants acknowledged that chiefs may misinterpret or selectively relay warnings or pass on alerts that do not materialise. Such inaccuracies can erode trust, especially when households take costly actions unnecessarily. Messenger legitimacy must therefore be continually reinforced through accuracy and consistency. Traditional leaders would benefit from capacity-building on interpreting scientific warnings, understanding thresholds, and participating in joint preparedness planning. Participatory processes combining scientific and local knowledge can reduce miscommunication and strengthen trust.

To mitigate these challenges, traditional leaders should be more deliberately involved in preparedness planning, training, and simulation exercises. Strengthening their capacity to interpret meteorological information may improve the timeliness and confidence of dissemination. Expanding capacity-building to deputy chiefs, women’s leaders, youth representatives, and religious figures can reduce overreliance on a single individual and enhance redundancy. This ensures continuity even when primary leaders are unavailable. To further reinforce redundancies, NGOs should also be more formally integrated into preparedness planning and simulations. Regular engagement outside emergencies allows NGOs to build relationships, improve trust, and reposition themselves as reliable preparedness partners. Involving NGOs alongside state and traditional authorities can diversify communication pathways and enhance community confidence in early warning processes.

This paper examines the politics of trust and messenger legitimacy in shaping community responses to cyclone early warnings in Madagascar. While national agencies, such as the BNGRC, provide essential forecasts, their messages often fail to resonate with rural communities, reflecting broader challenges of limited state presence, weak service delivery, and a history of unfulfilled promises. In contrast, warnings delivered by traditional leaders, including fokontany chiefs and kings, are widely accepted and acted upon, demonstrating that messenger credibility is as decisive as technical accuracy. The findings highlight the persistence of hybrid governance, where state institutions coexist with customary authority. Rather than treating traditional leaders as peripheral, disaster governance frameworks should recognise their role as trusted risk interpreters. By anchoring warnings in culturally legitimate structures, communication strategies can bridge the “last mile” gap and foster inclusive preparedness. Recognising relational trust as a key dimension of vulnerability reduction and strengthened disaster risk reduction (DRR) outcomes challenges conventional approaches focused primarily on infrastructure and technology.

Policy implications extend beyond Madagascar. In many fragile or resource-constrained settings, similar dynamics shape the effectiveness of early warning systems. Messenger legitimacy is not a substitute for technical accuracy but a complementary requirement for achieving community action. International frameworks such as the Sendai Framework and the “Early Warnings for All” initiative must move beyond a universal communication model to embrace diverse governance contexts.

Future research could build on these findings in several ways. Comparative studies across different cultural and governance settings in the Global South would clarify whether reliance on traditional leaders is unique to Madagascar or part of a wider pattern. Longitudinal work could explore how trust in traditional authority evolves, particularly as climate change intensifies cyclone risks and state institutions expand their presence. In addition, a mixed-methods approach that combines social science insights on trust with advances in climate modelling and communication technology would deepen our understanding of how messenger legitimacy and technical accuracy interact. Such research would contribute to the design of early warning systems that are not only scientifically robust but also socially trusted.