Social–Ecological Dimensions of Wildfire Risk in the Community Forests of Northern Thailand: Leadership Perception, Participation, and Surface Fuel Conditions

Abstract

Community-Based Fire Management (CBFiM) integrates local governance and ecological stewardship, yet the social drivers shaping its effectiveness remain poorly understood. This study examines the relationships among leadership perception, community participation, and surface fuel conditions in two community forests in Lampang Province, northern Thailand: Ban Pong and Ban Rong Ta. Forest floor fuel data were collected through destructive fuel sampling during the 2025 dry season, and social data were gathered through structured questionnaires measuring leadership perception using the Crew Member Perceived Leadership Scale and participation across seven fire management activities. Ban Rong Ta showed lower fuel loads but higher fire occurrence (nine fire detections recorded in 9 of 10 study years), lower leadership perception across all dimensions, and reduced participation in activities. The brigade–community participation gap reflects patterns documented across Southeast Asian community forestry programs, pointing to a structural challenge in fire governance. These findings suggest that awareness and informational participation alone do not reduce wildfire risk, and that integrating social and ecological indicators is essential for designing effective community-based fire governance systems.

1. Introduction

Fire is a common occurrence in forest ecosystems worldwide. For example, tropical deciduous forests are ecologically dependent on low-intensity fires, which stimulate seed germination, resprouting from root stocks, and the flush of early-successional vegetation that follows fire events, thereby maintaining ecological balance [1,2]. However, fire regimes have been altered by climate change and anthropogenic factors, increasing the frequency and severity of wildfires across multiple ecosystems [3,4,5]. Given the escalating risks, it is necessary to introduce and implement fire management strategies that are ecologically and socially sustainable [6,7]. Thailand’s forests are recognized as a biodiversity hotspot. Based on the Thailand land-use map (2000/2001), mixed deciduous forest accounts for approximately 46% of the country’s total forest area, while dry dipterocarp forest accounts for approximately 16% [8]. These forests provide ecosystem services to the local population, such as carbon storage as an ecosystem service which is increasingly recognized within national and international policy frameworks, including REDD+ and Thailand’s nationally determined contributions and soil conservation [9,10]. Despite these benefits, during the dry season, ecological factors and human activities interact to create conditions that increase wildfire frequency [11,12]. In addition to these environmental consequences, human livelihoods remain dependent on forests for subsistence [13,14].

Community forestry (CF) is one of the most common constitutional frameworks worldwide that promotes local participation. About 14% of all forests inside developing countries are governed by this Community Forest Management (CFM) approach. Some successful examples have occurred in South Asia, Africa, and Latin America [15,16,17]. Community Forest Management’s main purpose is to integrate ecological sustainability, equity, and economic efficiency [18,19]. However, as noted in a study by Pagdee et al. [20], measures of success vary, as most case studies focus on one or a few success dimensions rather than achieving their main objectives simultaneously [20,21].

In Thailand, the Royal Forest Department (RFD) has promoted community forestry since the 1980s [22], but legal rights have remained limited until the passage of the Community Forest Act in 2019. This act recognized community forests and established legal assistance between the communities and the state [23]. With this act, community leaders are required to submit a forest management plan and to register it, giving them the rights to use and manage forest resources for activities like ecotourism. This law also established a governance system that includes the Community Forest Policy Committee, the Provincial Community Forest Committee, and the local Community Forest Management Committee. These entities provide policy guidance, approve and monitor management plans, and facilitate joint decision-making between communities and the state [24,25].

Nan Province is an explicit example of community forestry in northern Thailand. This area possesses several community forests that are sustained on account of local conservation actions. In 2018, a project led by local NGOs, professors, and community leaders confirmed forest boundaries, conducted participatory mapping activities, and promoted discussion between stakeholders. These efforts show how local knowledge, combined with technical and institutional support, strengthens forest governance and ecological protection [26]. Furthermore, research studies in Thailand show that household participation in community forests is influenced by sociodemographic factors such as training, knowledge transfer, and the benefits obtained from forest resources [27]. Even though monetary incentives can improve participation, effective leadership and the selection of members in the committee remain important for promoting involvement [21,28].

In the context of fire management, governance and participation are particularly important within fire management. Historically, governments worldwide have addressed wildfire management through suppression. While suppression continues to be an important part of fire management, previous case studies show that excluding local communities from fire governance often results in low effectiveness [3,29]. To incorporate communities not only as fire users but as fire managers in areas such as prevention and response, the concept of Community-Based Fire Management (CBFiM) was introduced. As mentioned by the Food and Agriculture Organization [6], CBFiM is a flexible approach formed by the ecological, social, and political contexts of each region in which it is implemented.

Case studies from different regions show the potential and limitations of Community-Based Fire Management (CBFiM) when local ecological knowledge and governance systems are either promoted or avoided. In Australia, the reintroduction of Indigenous fire practices, which are often known as “cultural burns,” have shown positive results for landscape restoration, fuel reduction, and biodiversity conservation following decades of fire suppression policies [30,31]. In comparison, experiences from parts of Latin America highlight the risks of excluding local communities from fire governance. In Chile, changes towards fire suppression policies and restrictions on traditional burning practices have been accompanied by increases in wildfire area and severity over recent decades [32,33]. While consequences are formed by multiple factors, these cases deepen the importance of local fire management practices. Overall, these examples demonstrate that sustainable fire management needs governance systems that include local knowledge, participation, and ecological context.

In Thailand, Community-Based Forest Management (CBFiM) has come into focus in the past two decades, specifically in relation to deforestation and land ownership—specifically the contested legal status of community-managed forest land under Thailand’s National Forest Reserve Act and the Community Forest Act of 2019, which determine communities’ formal rights to manage, patrol, and enforce fire regulations within their designated areas [34,35]. Nonetheless, community fire management remains understudied, and wildfire risk in community forests of northern Thailand is poorly characterized at the local governance level, as fire management efforts have largely focused on regional integrated approaches rather than community-level governance dynamics [36]. Current studies tend to emphasize operational factors of fire management, while social constructs such as community participation, internal leadership interactions, and governance are often overlooked [6,20]. Fire management naturally places participants under high stress and in life-threatening conditions. Effective leadership plays a central role in the well-being of firefighters who operate in high-stress environments [37,38]. As proposed by Curral et al. [38], strong leadership increases the efficiency and psychological well-being of brigade members. Furthermore, Waldron et al. [39] propose that integrity, competent decision-making, and personal genuineness constitute the main framework of effective leadership in firefighters in the USA. The study shows that leaders perceived as genuine and trustworthy can inspire confidence, increase participation, and facilitate conflict mediation.

To address this knowledge gap, the absence of empirical studies linking internal leadership dynamics and participation patterns to fire management outcomes in community forests of Southeast Asia, this study focuses on two communities in the Ngao Model Forest in Lampang province: Ban Pong and Ban Rong Ta. Both communities were selected due to their shared ecological conditions, including similar forest types and seasonal fire regimes, as well as their participation in Thailand’s Community Forestry structure. Based on the characteristics, this study aims to examine the interaction between social and ecological variables in Community-Based Fire Management. The main objectives are to understand the characteristics of fuel structure (i.e., the composition, load, height, and spatial continuity of combustible surface material on the forest floor, including litter, grass, and undergrowth components) and surface fuel loads in both selected community forests, and how leadership and community participation shape fire management practices inside communities. It is hypothesized that communities with stronger leadership perception and higher participation will demonstrate more effective fire management practices.

Conceptual Framework

This study follows a social–ecological perspective to explore associations within Community-Based Fire Management (CBFiM) systems. The framework is informed by leadership theory [39] and collective action theory [19,40], which highlight the importance of leadership qualities and collective participation in governance consequences. Leadership qualities (integrity, competent decision-making, and personal genuineness) are examined in relation to levels of community participation in fire prevention, preparedness, and response activities. Participation is further associated with the development of trust, legitimacy, and shared norms, which appear to support coordinated action and strengthen social cohesion within local governance systems.

These social processes are related to the implementation of fire management practices, including fuel management, coordination, and monitoring. Implementation, in turn, is linked to ecological outcomes, specifically forest floor fuel load, fuel component composition (litter, grass, and undergrowth), and fire occurrence frequency, as recorded by satellite-based hotspot detections at the local level, particularly fuel load conditions and fire frequency. Rather than assuming direct causal relationships, the framework emphasizes interconnected associations among leadership, participation, governance processes, and ecological patterns. The relationships are understood as dynamic and context-dependent, providing an exploratory basis for identifying patterns that may inform future longitudinal or causal research. In this way, the framework conceptualizes CBFM as an evolving social–ecological system in which governance dynamics and ecological outcomes are closely related, as seen in Figure 1.

2. Methodology

2.1. Study Site

This study focuses on the Ban Pong Community Forest (18°41.8225′ N, 99°56.301′ E), established in 2016, and on the Ban Rong Ta Community Forest (18°41.534′ N, 99°55.949′ E), established in 2007. Ban Pong Community Forest spans 40 hectares, of which 38 hectares are a mixed deciduous forest, and 2 hectares are a teak plantation. In comparison, Ban Rong Ta Community Forest covers 108.5 hectares, of which 6.11 hectares are used for agriculture and 102.40 hectares are classified as mixed deciduous forest [41]. Both forests are classified as deciduous forests, a forest type common in northern Thailand that is characterized by seasonally dry conditions and periodic fire disturbance. Field observations at both sites recorded dominant tree species including teak (Tectona grandis), pradu (Pterocarpus macrocarpus), and teng (Shorea obtusa), with the understory commonly featuring bamboo (Dendrocalamus spp.) and various shrub species typical of seasonally dry forests. These species are consistent with vegetation surveys conducted within the Ngao Model Forest, Lampang Province [42]. Ban Pong Community Forest additionally contains a 2-hectare teak plantation, whereas Ban Rong Ta’s larger mixed deciduous area supports a broader range of native species. Both sites share broadly similar vegetation composition, though Ban Rong Ta’s larger and more fire-affected area exhibits greater structural heterogeneity in the understory layer. The locations of both community forests are shown in Figure 2.

Community forests in northern Thailand are widespread: as of 2022, the Royal Forest Department registered over 11,000 community forest units nationally, with a high concentration in the northern and northeastern regions. The two study forests are representative of a common type in the region: mixed deciduous community forests of small to medium size (40–108 ha), established under the Community Forest Act, located in upland agricultural landscapes, and managed primarily for watershed protection and non-timber forest product use. They differ in age (Ban Pong established 2016; Ban Rong Ta established 2007) and fire history, which makes them useful for comparative analysis while limiting direct generalizability to community forests with substantially different vegetation types, governance histories, or ethnic compositions.

2.2. Data Collection

2.2.1. Forest Floor Fuel Data Collection

Fuel characteristics were assessed in April 2025, during the peak of the dry season in northern Thailand. Sampling was conducted in two mixed dipterocarp community forests. While both communities share similar forest types and a comparable climate zone, site-specific abiotic factors may also influence fuel accumulation and were not systematically assessed; these are acknowledged as potential sources of unexplained variation in the fuel data. Surface fuels were classified into three categories: litter, grass, and undergrowth [2,43]. Forest floor fuels were defined as follows: (1) litter, comprising dead leaves, twigs, bark fragments, and reproductive material deposited on the soil surface; (2) grasses, comprising dead or dry standing and fallen grass biomass, measured separately owing to their distinct combustibility; and (3) undergrowth, comprising dead fine woody debris and low-growing non-grass vegetation including herbaceous dicots and low shrubs below 0.5 m in height. Grass was therefore treated as a distinct fuel component rather than a sub-category of undergrowth, reflecting common practice in tropical dry forest fuel classification. All fuel material was collected in the dry season (February–March), when moisture content is lowest and fuels are at their peak accumulation and flammability. The mixed deciduous forests of northern Thailand exhibit a strongly seasonal fuel accumulation pattern: leaf litter deposition peaks following the dry-season leaf fall (November–January), while grass biomass accumulates through the dry season and senesces by late January–February, coinciding with the peak fire danger period.

A total of 20 circular plots (5 m radius; 78.5 m²) were established using a simple random sampling approach. The number of plots per community (10 plots each) was influenced by recent fire activity in Ban Rong Ta Community Forest, which reduced the availability of measurable surface fuels in portions of the area. Sampling focused on locations where fuel was present and could be assessed. Within each plot, fuel height (cm) and fuel continuity (%) were recorded for each fuel category, with continuity defined as visually estimated percent cover. Fuel load was calculated through destructive sampling in four 1 × 1 m subplots located along the cardinal directions within each plot. Fresh biomass was weighed in the field and corrected to dry mass using oven-dried subsamples to determine moisture content [44]. Final fuel load estimates were expressed in t ha−1Plot level total fuel data are provided in Table S1 (Supplementary Materials).

Forest floor fuel data collection was conducted prior to social data collection, which was completed in June 2025, to capture seasonal fuel conditions while allowing sufficient time for survey and interview implementation.

Fire occurrence data were obtained from NASA FIRMS (VIIRS Suomi-NPP Collection 2) for the period 2015–2025 to provide a recent fire-history context. FIRMS detections show four fire detections in Ban Pong Community Forest and nine fire detections in Ban Rong Ta Community Forest, the latter recorded across 9 of 10 study years [45].

2.2.2. Social Data Collection

Information on fire management activities, participation, and leadership perceptions was gathered using structured questionnaires. Before data collection, research objectives were presented to community leaders, and information letters were distributed to ensure participants were informed about the study. Participants were given opportunities to ask questions, and the researcher obtained feedback. The questionnaire assessed participation in community fire management and respondents’ roles within the village. Leadership perceptions were measured using the Crew Member Perceived Leadership Scale [39], with “leader” defined as the officially recognized head of the community fire brigade. The scale was translated into Thai using a forward–backward procedure [46,47], reviewed for cultural appropriateness, and pilot-tested with 31 participants in Ban Huand Community Forest. Minor wording adjustments were made, and internal consistency was high (Cronbach’s α = 0.910), exceeding accepted reliability thresholds [48,49].

Household survey samples were estimated using Krejcie and Morgan’s [50] formula with a 10% margin of error, resulting in 68 households in Ban Pong and 62 in Ban Rong Ta. A 10% margin of error was chosen to balance reliability with the limitations of fieldwork in remote communities. This margin of error is considered appropriate in exploratory studies that aim to identify general patterns rather than exact population estimates [51]. Demographic data were obtained from community records to avoid duplicate sampling. In addition, all active fire brigade members were surveyed (19 in Ban Pong and 15 in Ban Rong Ta). While some brigade members belonged to the same household, responses were recorded individually, and identifiers were removed prior to analysis to maintain confidentiality.

2.3. Data Analysis

2.3.1. Forest Floor Fuel Data Analysis

Both fresh and oven-dry biomass obtained through destructive sampling were used to estimate fuel load (t ha⁻¹) for each plot and fuel layer. For each plot, fuel load (kg m⁻²) was calculated separately for litter, grass, and undergrowth components, and plot-level values were then averaged across the ten plots per community forest to produce community-level mean fuel loads for each component, reported by community (Ban Pong and Ban Rong Ta) and by fuel type (undergrowth, litter, and grass). Structural fuel characteristics were assessed using fuel height (m) for undergrowth and grass, fuel continuity (%) for all fuel layers, and litter depth (cm) for the litter layer, following standard fuel inventory approaches [52].

To evaluate differences in fuel conditions between communities, plot-level fuel metrics (fuel load, height, continuity, and litter depth) were compared using Wilcoxon rank-sum tests (Mann–Whitney U) due to small sample sizes and non-normal distributions [53]. Where multiple tests were performed within the same variable group, p-values were adjusted using the Benjamini–Hochberg method [54]. Effect sizes (r) were calculated to quantify the magnitude of differences (small, moderate, large).

2.3.2. Social Data Analysis

Data from the Crew Member Perceived Leadership Scale [39] were collected in June 2025 from 68 households in Ban Pong and 62 in Ban Rong Ta, all members of their respective community forests. The questionnaire was administered through face-to-face interviews, with items scored on a five-point Likert scale (1 = never to 5 = always). Internal consistency of the adapted leadership scale was assessed using Cronbach’s alpha for each leadership dimension. Reliability was high across all dimensions—personally genuine (α = 0.91), competent decision-making (α = 0.89), and integrity (α = 0.87)—exceeding commonly accepted thresholds for reliability in social science research [48,49].

Composite scores were calculated by averaging item responses within each dimension, and mean scores were then derived for each respondent group. The use of ANOVA on composite scores derived from Likert items was considered appropriate given the multi-item structure of each dimension and the high internal consistency values reported above; simulation studies and methodological reviews support the robustness of parametric analysis applied to multi-item composites when reliability is high [55]. Prior to analysis, ANOVA assumptions were examined: normality of residuals was assessed using Shapiro–Wilk tests, and homogeneity of variances was evaluated using Levene’s test. Leadership perceptions were compared using one-way ANOVAs across four groups (brigade vs. community members in each village). Separate ANOVAs were run for each dimension, and significant effects were followed by Tukey’s HSD post hoc tests to identify pairwise differences while controlling for multiple comparisons [56,57]. Statistical significance was set at α = 0.05.

To assess participation, respondents were asked about their awareness and participation in seven wildfire management activities (community meetings, awareness campaigns, firebreak construction, patrol, satellite monitoring, fire suppression, and training). Participation frequency was recorded on an ordinal scale (none, rarely, almost always, and always). In order to standardize awareness participation rates were calculated as the proportion of aware respondents who report active engagement in activities.

Given the group sizes, non-parametric tests were employed. Wilcoxon rank-sum tests assessed differences in participation frequency between brigades and communities and between villages, while paired Wilcoxon signed-rank tests compared brigade and community participation within villages for the same activity. A chi-square test of independence on raw participation counts across categories provided an overall assessment of associations between group membership and participation patterns. Analyses were conducted in R version 4.4.3 with significance set at α = 0.05.

3. Results

3.1. Fuel Characteristics

Surface fuel characteristics differed between Ban Pong and Ban Rong Ta Community Forests across fuel layers, reflecting contrasting disturbance histories and local vegetation dynamics (

Table 1. Surface fuel characteristics for Ban Pong and Ban Rong Ta Community Forests.

Fuel Variable	Ban Pong Mean ± SD	Ban Rong Ta Mean ± SD	W	p-Value	r	Effect	Sig.
Fuel Load
Total fuel load (t ha−1)	6.10 ± 1.70	4.85 ± 2.81	60.0	0.4497	0.169	Small	ns
Litter load (t ha−1)	4.79 ± 1.46	4.87 ± 2.24	57.0	0.6310	0.098	Small	ns
Undergrowth load (t ha−1)	0.60 ± 0.32	0.30 ± 0.23	77.0	0.0050	0.622	Large	*
Grass load (t ha−1)	0.27 ± 0.46	0.15 ± 0.14	29.5	0.5304	0.152	Small	ns
Structural
Litter depth (cm)	6.00 ± 1.40	4.00 ± 1.40	88.0	0.0014	0.722	Large	*
Undergrowth height (m)	0.28 ± 0.12	0.23 ± 0.12	50.5	0.3498	0.220	Small	ns
Grass height (m)	0.23 ± 0.25	0.14 ± 0.06	55.0	0.0668	0.445	Medium	ns
Continuity
Litter continuity (%)	85.54 ± 8.80	81.36 ± 23.01	54.0	0.7624	0.068	Small	ns
Undergrowth continuity (%)	14.90 ± 8.69	9.49 ± 7.17	57.5	0.1198	0.455	Medium	ns

W = Wilcoxon rank-sum statistic; r = effect size (<0.3 small, 0.3–0.5 medium, >0.5 large); p-values adjusted using Benjamini–Hochberg correction; * p < 0.05; ns = not significant.

). In both communities, litter constituted the dominant surface fuel component, exhibiting the highest continuity values and contributing the largest share of total fuel load relative to undergrowth and grass.

Total surface fuel load was higher in Ban Pong (6.10 ± 1.70 t ha⁻¹) than in Ban Rong Ta (4.85 ± 2.81 t ha⁻¹), though litter fuel load—the principal contributor in both communities—was comparable between sites (Ban Pong: 4.79 ± 1.46 t ha⁻¹; Ban Rong Ta: 4.87 ± 2.24 t ha⁻¹). The difference in total fuel load between communities was therefore driven primarily by undergrowth and, to a lesser extent, grass components. Undergrowth fuel load was notably higher in Ban Pong (0.60 ± 0.32 t ha⁻¹) than in Ban Rong Ta (0.30 ± 0.23 t ha⁻¹), and this difference was statistically significant and of large magnitude (W = 77, p = 0.005; BH-adjusted p = 0.010; r = 0.622). Grass fuels were low in both communities and did not differ significantly (Ban Pong: 0.27 ± 0.46 t ha⁻¹; Ban Rong Ta: 0.15 ± 0.14 t ha⁻¹).

Litter fuel load did not differ significantly between communities (p = 0.631; r = 0.098), yet litter depth was significantly greater in Ban Pong (0.06 ± 0.014 m) than in Ban Rong Ta (0.04 ± 0.014 m), with a large effect size (W = 88, p = 0.0014; r = 0.722). This dissociation between litter mass and litter depth indicates that the structural configuration of the litter bed—including packing density and aeration—varied between sites independently of total fuel quantity, with potential implications for combustion dynamics and smoldering potential.

Litter continuity exceeded 80% in both communities (Ban Pong: 85.54 ± 8.80%; Ban Rong Ta: 81.36 ± 23.01%). This level of continuity is ecologically relevant, as it suggests that an ignition within the litter layer could propagate across the forest floor under dry-season conditions [7]. In contrast, undergrowth and grass continuity were considerably lower (≤16% in both communities), consistent with a patchy distribution of fine fuels above the litter stratum. Continuity did not differ significantly between communities for most fuel layers; however, a moderate effect size for undergrowth continuity (r = 0.455) suggests potentially meaningful structural differences that the available sample size may have been insufficient to detect.

Fuel height metrics followed a similar pattern. Undergrowth was slightly greater in Ban Pong (0.28 ± 0.12 m) than in Ban Rong Ta (0.23 ± 0.12 m), and grass height was higher in Ban Pong (0.23 ± 0.25 m) than in Ban Rong Ta (0.14 ± 0.06 m). Although neither difference reached statistical significance, a moderate effect size for grass height (r = 0.445) indicates a potentially ecologically relevant difference in near-surface fuel structure between communities. While statistically detectable, the differences in undergrowth and grass height between the two community forests are small in absolute terms (mean difference < 0.05 m for undergrowth) and likely of limited ecological significance in isolation; their relevance is primarily in the context of fuel continuity and structural arrangement rather than as stand-alone metrics of fire risk. Taken together, the most robust community-level distinctions were found in undergrowth fuel load and litter bed depth, both yielding large effect sizes and meeting adjusted significance thresholds, as demonstrated in Figure 3.

3.2. Participation in Community-Based Fire Management Activities

Awareness of Wildfire Management Activities

Awareness of wildfire management activities was mostly high across both villages and sectors, although variation existed for specialized activities. Brigade members showed awareness (n = 19 of 19 in Ban Pong; n = 15 of 15 in Ban Rong Ta) across most activities recorded for community meetings, environmental awareness, firebreak construction, patrolling, suppression, and training. The only exception was satellite monitoring, which demonstrated lower awareness between brigades (n = 14 of 19 in Ban Pong; n = 11 of 15 in Ban Rong Ta). Among community members, awareness was highest for environmental awareness campaigns (n = 68 in Ban Pong; n = 60 in Ban Rong Ta) and lowest for satellite monitoring (n = 44 in Ban Pong; n = 30 in Ban Rong Ta), which registered the weakest awareness across both groups and villages.

3.3. Participation Rates Between Aware Respondents

Among respondents who answered positively, awareness participation rates varied by sector and by activity, as shown in Figure 4. Brigade members belonging to Ban Pong reported near-full participation across most operational activities which means that nearly all registered brigade members who were aware of an activity reported active engagement in it, with rates of 100% for community meetings, environmental awareness, and firebreak construction, and 95% for patrol, suppression, and training. Meanwhile, Ban Rong Ta brigades showed a similar pattern, achieving 100% participation in community meetings, environmental awareness, firebreak, patrol, and suppression. Satellite monitoring was the only activity with a lower participation percentage, as clearly stated in both villages, at 50% in Ban Pong and only 9% in Ban Rong Ta.

In contrast, community participation was consistently lower than brigade participation. In Ban Pong, community engagement was highest in community meetings (66%) and environmental awareness (41%), while operational activities such as firebreak construction (15%), suppression (5%), patrol, and satellite monitoring (2%) showed minimal involvement. Community participation in Ban Rong Ta was comparatively higher for several operational activities, including patrol (37%), firebreak (39%), suppression (29%), and training (43%), though satellite monitoring showed no participation.

The chi-square test of independence revealed a highly significant overall association between group membership and participation frequency across all four categories (χ²(3) = 331.79, p < 0.001), confirming that brigades and community members participated at different rates when considered together. Meanwhile, a weaker association was also found between village and participation frequency (χ²(3) = 17.98, p < 0.001).

Unpaired Wilcoxon rank-sum tests across all activities confirmed that brigade and community members differed significantly in both ‘almost always’ (W = 156, p = 0.008) and ‘always’ (W = 189, p < 0.001) participation categories, while no significant difference was found for ‘rarely’ responses (W = 77.5, p = 0.358). This suggests a gap, driven by high-frequency participation. In comparison, village produced no statistically significant differences for any frequency outcome.

Within villages, brigades reinforced previous findings. Inside Ban Pong, brigade members reported significantly higher participation than community members in the “almost always” (V = 21, p = 0.036) and “always” (V = 28, p = 0.022) categories, with no difference in “rarely” responses. In Ban Rong Ta, a significant difference was observed only in the “always” category (V = 28, p = 0.022). Mean “always” participation rates show the magnitude of this disparity as brigades averaged 56.5% in Ban Pong and 53.3% in Ban Rong Ta, compared with 6.8% and 13.9% among community members, respectively.

3.4. Leadership Perception

The leadership scale showed high internal consistency across dimensions, supporting the use of composite scores for analysis. One-way ANOVA revealed significant differences among the four respondent groups for all three leadership dimensions: personally genuine, competent decision-making, and integrity (p < 0.01 for all dimensions). Post hoc Tukey HSD tests indicated that these differences were primarily driven by contrasts between brigades and community members, as well as between communities. In general, Ban Pong Brigade members reported higher leadership perception scores across all dimensions than Ban Rong Ta Brigade members. Differences between community members across villages were smaller and less consistent. These results suggest that leadership perceptions vary systematically with both organizational role and community context, with higher scores associated with more stable participation patterns and lower recent fire occurrence, as observed in Figure 5.

4. Discussion

Altogether, the ecological and social findings of this study yield a concept: the capacity of community forests to manage wildfire is not determined by a single factor but by the interaction among fuel structure, participation in management activities, and the leadership that coordinates those activities. Ban Pong Community Forests exhibited higher leadership perception scores, more consistent perception regarding fire brigades, and stronger community attendance at management activities, but they also accumulated higher undergrowth fuel loads (0.60 ± 0.32 t ha⁻¹) and higher litter depth accumulation in comparison with Ban Rong Ta Community Forest. These results show that, although current management activities in Ban Pong are attended by a majority and coordinated, they may not be sufficient to reduce the physical fire risk posed by accumulated fuel. Strong participation in meetings and awareness sessions does not translate into efficient fuel reduction in the Community Forest.

In comparison, Ban Rong Ta is the exact opposite. They possess lower leadership perception scores, weaker brigade participation, and higher fire occurrence, as documented by NASA FIRMS records, which show nine fire detections across 9 of 10 study years, compared to four fire detections in Ban Pong. When normalized by forest area, detection density was approximately 0.10 detections per hectare in Ban Pong (40 ha) and 0.083 detections per hectare in Ban Rong Ta (108.5 ha), a difference that partly reflects the smaller extent of Ban Pong. However, the temporal pattern—fire detections recorded in 9 of 10 study years in Ban Rong Ta compared to four total detections in Ban Pong—is not affected by area normalization and remains the more ecologically informative metric for comparing fire frequency between these two sites. The use of absolute counts as a comparative metric is therefore acknowledged as a limitation; area-normalized rates should be used where possible in future multi-community comparisons. These results coincide with lower total fuel loads (4.85 ± 2.81 t ha⁻¹). The high variance in litter continuity at Ban Rong Ta (81.36 ± 23.01%) supports this interpretation, as uneven fuel by repeated fire events would produce this kind of spatial heterogeneity in the litter bed. It must be mentioned that lower fuel loads in Ban Rong Ta should not be taken as evidence of more effective management; they may be a reflection of a landscape that has been burned on numerous occasions, with fuel conditions formed by fire occurrence and not by fire management. It should also be noted that higher fire occurrence in Ban Rong Ta may not be entirely attributable to governance deficits; fire is not uniformly detrimental in deciduous forests. Seasonal burning can stimulate early grass growth, reduce tick and pest populations, facilitate access for non-timber forest product collection, and support the regeneration of certain tree species [2]. Some community members may therefore tolerate or even deliberately set low-intensity fires for these utilitarian benefits, which would confound a purely risk-based interpretation of fire occurrence data. Future research should investigate intentional burning practices and community perceptions of fire benefits alongside governance indicators.

Satellite records show that Ban Rong Ta registers fire detections in 9 of 10 study years. This pattern points towards a lack of participation as a plausible contributing factor. Community engagement in critical operational tasks was consistently lower in Ban Rong Ta, and the brigade received lower leadership perception scores across all three dimensions assessed. The combined effect of weaker leadership and reduced community engagement in preventive activities may have left that area more vulnerable to ignition. Also, the gap in awareness and participation documented in this study is consistent with patterns recorded in Community Forest Management programs across Southeast Asia. Studies from community forestry contexts in northern Thailand have found that participation in formal management activities tends to concentrate among a small portion of members, while most community participation remains in information activities [16,20]. The structure observed in the present study relates to a pattern identified as a sustained challenge in decentralized forest governance throughout the region.

In Indonesia, community fire management programs known as Masyarakat Peduli Api (fire-aware communities) have faced similar challenges: formal brigade structures are established, training is provided, and community awareness is raised, yet participation during actual fire events remains among brigade members, with the main community playing a passive role [58]. This parallel suggests that the brigade–community participation gap observed here is not specific to the Thai context but reflects a wider regional trend in community fire governance.

Direct quantitative comparisons of leadership perception scores with other Thai or Southeast Asian studies are limited by the use of standardized leadership scales in community fire management research in this region. However, the pattern identified is consistent with findings from community forest governance research across Southeast Asia [20,59].

The relationship between fire occurrence and leadership perception in this study is not likely to be unidirectional. Although weaker leadership and lower community participation may contribute to greater fire vulnerability, repeated fire events may, over time, erode community confidence in leadership. In organizational psychology, repeated exposure to high-stakes events perceived as preventable is associated with declining institutional trust [60], and community members who experience recurring fire damage despite the presence of a formal brigade may progressively question brigade competence and commitment. This would appear as lower scores in the competent decision-making and personally genuine leadership dimensions, which coincides with the pattern identified in the Ban Rong Ta Brigade. The brigade’s own lower self-evaluations are consistent with patterns that have been described as organizational fatigue in community conservation contexts across Southeast Asia, where cooperative efforts repeatedly fail to produce visible outcomes [20]. Whether this dynamic is operating in Ban Rong Ta cannot be confirmed from the quantitative survey data alone; substantiating this interpretation would require open-ended interviews tracking brigade members’ perceptions across fire seasons, and this therefore remains a hypothesis for future investigation. The most plausible interpretation is therefore a reinforcing cycle in which weak leadership reduces participation, reduced participation increases fire vulnerability, and repeated fire events erode the collective confidence needed to sustain management efforts.

Several limitations constrain this study. First, the study was limited to two communities, limiting generalizability to the wider northern Thai provinces and Southeast Asia. The two communities may differ in abiotic characteristics that were not accounted for. Future research incorporating more community forests across diverse ecological and social contexts is needed to establish the patterns detected here. Secondly, the cross-sectional design means that causal directionality between social and ecological variables cannot be established with certainty. The relationship between leadership perception, participation, fuel accumulation, and fire occurrence is almost certainly bidirectional and dynamic, with feedback operating across multiple scales. A longitudinal study design tracking the same communities across multiple fire seasons would be required to separate these influences. Third, although fire occurrence was indicated through NASA FIRMS hotspot data, formal correlation analysis between hotspot frequency and social or fuel variables was constrained by the small number of community-level observations. The directional consistency between fire frequency, fuel structure, and leadership perception across the two communities supports the proposed framework, and future studies with larger community samples should incorporate fire occurrence data to enable formal statistical testing.

Fourth, survey responses may have been subject to social bias, particularly for self-reported participation frequencies. In order to address this gap, a mixed-methods approach that incorporates direct observation or participatory assessments would provide a stronger basis for validating self-reported data. Fifth, the Crew Member Perceived Leadership Scale was originally developed for formally trained firefighters in the United States of America and does not entirely capture the leadership processes relevant to Community-Based Fire Management groups in northern Thailand. Indigenous and locally embedded forms of leadership that include reciprocal authority structures, elder-based knowledge systems, and community accountability mechanisms may not map cleanly onto the formal leadership constructs the scale was designed to measure. Adapting leadership assessment tools in collaboration with the communities themselves represents an important line for future methodological development.

5. Conclusions

This study demonstrates that effective community-based wildfire management depends on numerous social factors, such as leadership quality and community participation, as well as ecological outcomes reflected in fuel structure and fire occurrence. The findings suggest that strong leadership and high awareness do not necessarily reduce wildfire risk unless they translate into effective fuel management, and low fuel loads may reflect repeated fire occurrence rather than successful prevention.

Future research incorporating direct measures of forest use intensity alongside fuel sampling and longitudinal governance data would provide a stronger empirical basis for this relationship.

The awareness-to-participation gap documented in this study is consistent with patterns seen across community fire governance in Southeast Asia. This indicates that the challenge lies in institutional design rather than a lack of information. Addressing this issue requires integrating community members into operational management roles, investing in leadership development, and developing frameworks that integrate social and ecological indicators rather than treating them as separate sectors.