Beyond awareness: factors associated with household preparedness in response to early warning information in rural Somaliland

1 Introduction

Climate-related disasters, such as droughts, floods, and erratic rainfall, have intensified in frequency and severity worldwide, with rural communities in developing countries bearing disproportionate losses to their livelihoods and food security because of their dependence on climate-sensitive agriculture and pastoralism (Pienaah et al., 2023; Teku and Tariku, 2026). Early warning systems (EWS) are a cornerstone of disaster risk reduction, as they reduce losses through timely and actionable warnings (Akwango et al., 2017; Beaudet et al., 2025; Macherera and Chimbari, 2016). However, across Africa, weak infrastructure and limited funding constrain EWS delivery, and even where systems function, warning receipt does not reliably translate into household action: response is mediated by education, income, gender, trust in the source, and access to communication channels (Awolala et al., 2022; Liu, 2025; Muema et al., 2018; Raphela and Ekeke, 2025).

Across sub-Saharan Africa, many EWS remain reactive rather than anticipatory, monitoring hazards after their onset instead of enabling proactive preparedness (Nhamo et al., 2019). Their reach is further limited by poor last-mile dissemination, fragmented governance, and the systematic exclusion of marginalized groups, including pastoralists, women, and remote rural households (Josè Moisès et al., 2023). Evidence from Zimbabwe, Nigeria, and Kenya shows that the utilization of climate and warning information remains uneven, shaped by disparities in education, institutional support, and communication access (Mudombi and Nhamo, 2014; Muema et al., 2018). Material and financial resources further determine whether households can convert warning information into concrete preparedness actions. Household preparedness levels remain generally low across multiple vulnerable settings, with most households lacking emergency plans or rapid-response capacity (Esang, 2026; Kitanović, 2020).

This gap between warning receipt and action is most pronounced in last-mile rural communities, where institutional channels for translating hazard forecasts into local actionable guidance are the weakest. In such settings, dissemination is fragmented, sensitization often depends on household heads rather than formal agencies, and insurance and information products rarely reach residents (Ogunwumi and Ihinegbu, 2025). Therefore, preparedness is not a uniform rural problem but a last-mile one, since these households face the highest exposure alongside the lowest institutional reach. Evidence from rural China reinforces this point: households drew on different capital assets at each disaster stage: natural, physical, and financial capital before a disaster; physical and social capital during a disaster; and human, financial, and social capital afterward, with village-level infrastructure and training determining which strategies were accessible (Li et al., 2025). Last-mile status thus shapes both exposure severity and the mechanisms available for preparedness, justifying a dedicated focus on rural, last-mile settings rather than treating “rural” as undifferentiated.

However, the literature on determinants remains theoretically fragmented. Li et al. (2025) explain preparedness through a livelihood's capital lens, in which asset access determines the available strategies at each disaster stage. Borate et al. (2025) found that self-efficacy and response efficacy are the strongest predictors of wildfire mitigation, with housing tenure producing a secondary resource disparity between owners and renters. Ogunwumi and Ihinegbu (2025), studying flood preparedness in Nigeria, found that among sociodemographic variables, only the gender of the household head was significant, while age, occupation, and household size were not. This result sits uneasily against capital- and efficacy-based accounts and suggests that demographic profiling alone weakly predicts preparedness once structural barriers such as poor sensitization and absent insurance are considered. These studies do not converge on a single explanatory framework, and none integrates communication quality or institutional trust factors, which have been repeatedly identified as decisive in the sub-Saharan African EWS literature (Awolala et al., 2022; Muema et al., 2018). This study addresses this theoretical fragmentation rather than a simple scarcity of evidence.

In the Horn of Africa, this fragmentation is compounded by fragility. Recurrent droughts, floods, and displacement have deepened the vulnerability of rural and agro-pastoral communities (Murendo et al., 2025; Omer, 2024; Sharmake et al., 2022), and weak governance and limited trust further undermine EWS effectiveness (Mabumbo et al., 2024). In Ethiopia's Somali Region, accurate flood forecasts failed to prompt action because warnings were not translated into locally actionable guidance (Mabumbo et al., 2024), a communication failure consistent with the last-mile dynamic described above. Somaliland exemplifies these conditions: its rain-fed, pastoral-dependent economy leaves rural households highly exposed to drought and food insecurity (Sharmake et al., 2022), and although EWS institutions exist, dissemination remains weakest among women and less educated populations (Omer et al., 2026a). Related studies confirm awareness of climate risk but low uptake of preparedness measures, constrained by market, extension, and information access barriers (Benito et al., 2025; Omer et al., 2025).

Framing the problem across fragile, pastoral-dependent settings generally rather than relying on Somaliland-specific findings as the evidence base clarifies the broader gap: no study has jointly tested how sociodemographic characteristics, communication quality, institutional trust, household preparedness resource needs shape last-mile preparedness, despite each factor being independently linked to preparedness in fragmented strands of the literature (Awolala et al., 2022; Borate et al., 2025; Li et al., 2025; Ogunwumi and Ihinegbu, 2025). This study addresses this gap using cross-sectional survey data from four rural regions of Somaliland: Maroodijeex, Togdheer, Sool, and Sanaag to jointly test these determinant domains. The findings are intended to inform last-mile disaster risk reduction and EWS design in comparable fragile agro-pastoral settings beyond Somaliland.

2 Methods and materials

2.1 Study area

This study was conducted in rural areas of Somaliland, specifically in villages located within the Maroodijeex, Togdheer, Sool, and Sanaag regions. These regions were selected because they represent diverse rural livelihood systems and varying levels of exposure to climate-related hazards, including droughts, floods, water scarcity, and food insecurity. Somaliland, located in the Horn of Africa, declared independence from Somalia in 1991, although it remains internationally unrecognized (Bryden, 2004). Rural communities in these regions primarily depend on pastoralism and agro-pastoralism, making household livelihoods highly sensitive to environmental shocks and to climate variability. Recurrent climate-related disasters have increased the importance of early warning systems as key mechanisms for disaster risk reduction, preparedness, and anticipatory actions. Therefore, the selected study areas provide an appropriate context for examining the Factors of household preparedness in response to early warning information across diverse rural settings in Somaliland (Figure 1).

2.2 Research design, sampling, and data collection methods

This study employed a cross-sectional quantitative design to examine the factors of household preparedness in response to early warning information among rural households in Somaliland. The analysis draws on the same household survey dataset (n = 390) reported by Omer et al. (2026a) who examined the functionality, community response, and systemic gaps of Somaliland's early warning systems using descriptive and bivariate analyses. The present study extends the use of this dataset by addressing a different research objective through multivariable logistic regression to identify independent factors of household preparedness. Specifically, it jointly examines socio-demographic, communication, institutional trust, and preparedness resource-need factors that were not investigated in the earlier study. Any descriptive results shared between the two studies were transparently disclosed rather than presented as newly collected data.

A multistage sampling strategy was employed to ensure a broad representation across rural, hazard-prone areas of Somaliland. Four regions (Maroodijeex, Togdheer, Sool, and Sanaag) were purposively selected based on their recurrent exposure to climate-related hazards and diverse livelihood characteristics. Maroodijeex includes relatively populous rural and peri-urban communities, whereas Togdheer and Sool are predominantly arid pastoral regions that experience frequent droughts. Sanaag encompasses both mountainous and coastal environments that are exposed to droughts and flash floods. Within each region, districts were randomly selected, followed by the selection of sampling areas based on the population size and exposure to hazard-prone conditions. Households were subsequently selected using systematic random sampling in sampling areas with available household lists; in locations without complete sampling frames, a predefined field-walking procedure supported by community-gathering points was used to identify eligible households.

The sample was proportionally allocated according to the population distribution across the four regions: 36.9% from Maroodijeex, 22.1% from Togdheer, and 20.5% each from Sool and Sanaag. Data collection was conducted at different times of the day to reduce the likelihood of excluding households whose members were absent during the normal working hours. Enumerators were trained in inclusive household engagement to minimize the social selection bias. A total of 390 households participated in this study (Table 1). Because regional selection was purposive and household sampling frames were incomplete in some locations, the sample should not be considered nationally or regionally representative of rural Somaliland. Therefore, the findings should be interpreted as characterizing the surveyed communities specifically, rather than generalized without qualification to the broader rural population.

Table 1. Distribution of study sample across regions and selected sampling areas in Somaliland.

No formal a priori sample size calculation was conducted before data collection. Following data collection, the final sample was assessed for its suitability for the planned multivariable analyses using established methodological guidance. The locations presented in Table 1 represent the principal administrative or geographic reference points used to organize sampling across the four study regions. They include cities, towns, district centers, and smaller settlements and should not be interpreted uniformly as villages. Although some sampling areas are identified by the names of major urban or district centers, such as Hargeisa, Burao, and Aynabo, the study specifically targeted rural households residing in the surrounding rural communities within each respective sampling area, rather than households in the urban centers themselves. Eligible households were subsequently selected from these rural communities according to the sampling procedure described above. Accordingly, the term “Selected Sampling Areas” is used to reflect the geographic hierarchy and sampling design.

Primary data were collected using a structured, predominantly closed-ended questionnaire administered at the household level. The questionnaire covered socio-demographic characteristics, awareness and perceptions of early warning systems, communication quality, institutional trust, preparedness resource needs, and household preparedness. Enumerators received training in the local language, study objectives, interviewing procedures, and ethical considerations before the fieldwork. Prior to data collection, the questionnaire was reviewed by the research team to improve clarity, consistency, and contextual relevance, and revisions were made as necessary. Because the instrument consisted primarily of single categorical and binary items rather than multi-item scales, internal consistency reliability testing was not applicable. Instead, content validity was established through an expert review to ensure that the questionnaire adequately captured the study constructs. Data were collected directly from households in the selected villages to ensure broad representation of communities and livelihood groups.

2.3 Measurement of variables

The dependent variable was household preparedness in response to early warning information, which was derived from a single survey item asking whether the household had taken any measures in response to early warnings. This was coded as 0 (no action) or 1 (took action). Because this measure was based on a single item, it captures whether a preparedness action occurred but does not distinguish the type, intensity, or adequacy of the measures taken.

The explanatory variables comprised socio-demographic, communication, behavioral, and resource-related factors (Table 2). Socio-demographic variables (gender, age group, education, and occupation) reflect characteristics of the individual respondent interviewed per household. The questionnaire did not record the respondent's relationship with the household (e.g., household head status), and these variables should therefore be interpreted as respondent-level rather than verified household-head-level characteristics. Occupation was classified into formal and informal categories. The formal category comprised government employees and NGO workers, whereas the informal category included livestock herders, farmers, and business owners (self-employed traders), consistent with the occupational classification applied in the study dataset. Communication variables captured respondents' perceived timeliness and clarity of early warning information, each measured using a three-category ordinal item. Trust in early warning systems was measured as a binary indicator of whether the respondents reported trusting the warning information provided.

Table 2. Description of dependent and explanatory variables.

Household preparedness-related resource needs were assessed using a composite index derived from four dimensions: emergency response training, financial resources, preparedness information, and participation in preparedness drills. For each dimension, respondents indicated whether the resource was needed, coded 1 if the need was reported, and 0 otherwise. The four indicators were equally weighted and summed to produce a Preparedness Resource Need Index ranging from 1 to 4, with higher scores indicating a greater number of unmet resource needs. No household reported a score of zero, and no missing data were observed for any of the four items (n = 390). For descriptive and regression analyses, the index was categorized as low unmet need (score = 1), medium unmet need (score = 2), or high unmet need (score = 3–4). Although derived from self-reported resource needs rather than a direct measure of household assets, the index is conceptually and empirically distinct from the dependent variable that measures whether a preparedness action was taken.

Ordinal communication variables (timeliness and clarity) were entered into the regression model as categorical (dummy-coded) covariates rather than as continuous linear terms, given the absence of a strong a priori assumption of equal spacing between categories. The regression results showed a monotonic pattern of increasing effect size across the timeliness categories, consistent with an underlying dose-response relationship; however, a formal linear trend test was not conducted, and this is noted as a direction for further analysis.

2.4 Data analysis

The Data were analyzed using Stata version 17.0. Descriptive statistics were used to summarize the respondents' sociodemographic characteristics and key study variables. Bivariate associations between household preparedness and each explanatory variable were examined using Pearson's chi-square tests, providing an unadjusted descriptive screening of the association for each variable independently. A binary logistic regression model was then estimated to identify the independent factors of preparedness while controlling for all covariates simultaneously, as bivariate associations may be confounded by correlations among explanatory variables. Comparing the unadjusted chi-square results against the adjusted logistic regression estimates allows divergence between the two, which is indicative of confounding, to be identified. The results are presented as odds ratios with robust standard errors, p-values, and 95% confidence intervals.

Robust (Huber–White) standard errors were estimated to account for potential heteroskedasticity. However, the multistage sampling structure was not explicitly accommodated through cluster-adjusted variance estimation; therefore, residual intra-cluster correlation may have resulted in underestimated standard errors and narrower confidence intervals for some estimates. This limitation is acknowledged in the Discussion, since accounting for clustering could widen confidence intervals for some estimates. Model diagnostics included the Hosmer–Lemeshow goodness-of-fit test, link test for model specification, and variance inflation factors (VIF) for multicollinearity. Classification accuracy, sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve were reported as in-sample measures of model fit rather than as evidence of out-of-sample predictive performance; no train/test split or cross-validation was performed. The study's objective is explanatory, identifying factors of preparedness rather than predictive, and model-fit statistics are reported as diagnostics of the fitted model rather than as claims of forecasting performance.

3 Results

3.1 Descriptive overview

Respondents were nearly evenly split by gender (53.9% male, 46.1% female), predominantly middle-aged (57.2% aged 35–54), with just over half being literate (51.3%), largely engaged in informal-sector livelihoods (92.3%), and living in large households (54.1% with seven or more members; Table 3). Detailed patterns of EWS awareness by hazard type and information source are reported in the companion study (Omer et al., 2026a); the present analysis focuses specifically on the communication, trust, and preparedness resource-need variables most directly relevant to preparedness factors.

Table 3. Socio-demographic characteristics of the study households.

n = number of households. Percentages are based on the total sample.

Awareness of early warning systems was high but concentrated on weather- and disaster-related alerts, with comparatively little awareness of disease outbreak or food security warnings. Perceptions of communication quality were mixed: 26.2% of respondents rated warnings as very timely and 38.5% as not timely, while 34.6% rated messages as very clear and 23.9% as not clear. Trust in warning systems was high overall (81.3%). Preparedness resource need was distributed relatively evenly across low (32.3%), medium (32.8%), and high (34.9%) unmet-need categories (Table 4).

Table 4. Communication characteristics, trust in early warning information, and preparedness resource needs.

Despite this generally favorable profile of communication quality and trust, actual preparedness behavior was low: 63.3% of households took no preparedness action, and 55.1% lacked a household emergency plan (Figure 2).

Table 5 (below) reports the specific resources respondents identified as necessary to strengthen preparedness: a large majority (81.5%) reported a need for emergency response training, while approximately half identified a need for financial resources (53.1%) or information access (53.1%), and about one in five (19.7%) identified a need for community drills or exercises. This combination of favorable communication and trust indicators alongside low actual preparedness is the central pattern this study investigates in the analyses that follow.

Table 5. Perceived household resource needs for disaster preparedness.

3.2 Bivariate analysis of household preparedness

Table 6 presents the proportion of households taking preparedness action within each category of the study variables. Preparedness was significantly associated with gender: 44.3% of households with male respondents took preparedness action compared with 27.8% of households with female respondents (χ² = 11.37, p = 0.001). Age was also significantly associated with preparedness (χ² = 7.75, p = 0.021): households aged 55 and above showed the highest action rate (55.8%), compared with 35.5% among those aged 18–34 and 33.6% among those aged 35–54. Education showed a significant association in a counterintuitive direction (χ² = 6.72, p = 0.010): 43.2% of illiterate respondents took preparedness action compared with 30.5% of literate respondents. Household size was significantly and inversely associated with preparedness (χ² = 7.40, p = 0.025): smaller households (1–3 members) had the highest action rate (52.1%), which declined to 38.9% among households of 4–6 members and 31.8% among households of 7 or more members. The formal/informal occupation distinction was not significantly associated with preparedness (χ² = 0.62, p = 0.430).

Table 6. Association between household characteristics and household preparedness in response to early warning information (Pearson's χ² test).

Values represent the row percentage of households reporting that they took preparedness action following receipt of early warning information. Associations were assessed using Pearson's chi-square (χ²) test. Statistically significant associations (p < 0.05) are shown by the corresponding p-values.

The strongest associations, which were the most counterintuitive in direction, emerged among the communication and trust variables. Preparedness action was most common among households perceiving warnings as not timely (67.3%), declining sharply to 25.4% among those perceiving warnings as somewhat timely and 6.9% among those perceiving warnings as very timely (χ² = 107.36, p < 0.001). A parallel pattern was observed for message clarity (χ² = 73.71, p < 0.001): 64.5% of households that found warnings unclear took preparedness action, compared with 42.6% among those finding warnings somewhat clear and only 10.4% among those finding warnings very clear. Trust in early warning systems showed the largest effect in the bivariate analysis (χ² = 90.09, p < 0.001): 84.9% of households reporting no trust in official systems took preparedness action, compared with only 25.6% of households reporting trust. Preparedness resource need was also significantly associated with preparedness (χ² = 6.51, p = 0.039), although the pattern was non-monotonic. Households reporting low unmet need had the lowest preparedness-action rate (27.8%), compared with 42.2% among households reporting medium unmet need and 39.7% among those reporting high unmet need. This counterintuitive pattern for communication and trust variables, evident at the bivariate level, motivated the multivariate analysis below to determine whether it held after adjustment.

3.3 Multivariable logistic regression analysis of household preparedness

Table 7 presents the primary multivariable logistic regression model. The model fit the data well [LR χ²(14) = 189.52, p < 0.001; classification accuracy = 81.8%; AUC = 0.878; Hosmer–Lemeshow p = 0.944; pseudo-R² = 0.37], reported as in-sample diagnostics rather than out-of-sample predictive validation. Timeliness of information showed one of the strongest adjusted associations with reported preparedness action: relative to households perceiving warnings as not timely, those describing information as somewhat timely (OR = 0.223, p < 0.001) and very timely (OR = 0.074, p < 0.001) had substantially lower odds of taking preparedness action. Trust in early warning systems similarly showed a strong inverse adjusted association with reported preparedness action: households reporting trust had markedly lower odds of taking preparedness action than those without trust (OR = 0.128, p < 0.001). Preparedness resource need showed a significant, non-linear association with preparedness action: households reporting a medium level of unmet need (two of four resources) had substantially higher odds of taking preparedness action than households reporting only one unmet need (OR = 2.718, p = 0.014), while households reporting high unmet need (three or four resources) did not differ significantly from the low-need reference group (OR = 1.367, p = 0.432). Among the sociodemographic variables, only age remained significant after adjustment: respondents aged 55 and above had significantly higher odds of preparedness action than the youngest age group (OR = 2.588, p = 0.043). Gender, education, occupation, and household size did not retain significance once other covariates were controlled for, suggesting that their bivariate associations were confounded by correlated factors such as preparedness resource needs and age.

Table 7. Multivariable logistic regression analysis of factors of household preparedness (primary specification).

Female, age 18–34 years, literate, formal occupation, household size 1–3 members, not timely, not clear, no trust, and low Unmet need.

4 Discussion

This study examined the factors of household preparedness in response to early warning information among rural households in Somaliland. Rather than confirming the conventional assumption that awareness, timely communication, and institutional trust drive preparedness, the findings reveal an inverse relationship: households receiving less timely and clear information and expressing lower institutional trust were more likely to report taking preparedness action. This pattern persisted after adjustment for sociodemographic, occupational, and resource-related factors, suggesting that the observed associations were not explained by measured confounding variables and are unlikely to reflect a modeling artifact.

Survey data cannot fully explain the mechanisms underlying this pattern; however, Protection Motivation Theory (PMT) provides a useful framework for interpretation. PMT suggests that protective action is shaped not only by perceived threat but also by coping appraisal, including self-efficacy, response efficacy, and perceived costs of action (Borate et al., 2025). In this context, One hypothesis requiring future empirical testing is that households receiving timely warnings and expressing greater trust in official systems may perceive institutional responses as sufficient; however, institutional reliance and perceived response efficacy were not measured in this study. This interpretation should nevertheless be considered cautiously because the preparedness measure captures whether any action was taken but does not distinguish between self-initiated, government-directed, NGO-supported, or community-coordinated responses. Consequently, the observed pattern may reflect reliance on institutional or externally coordinated support rather than reduced preparedness motivation. Households that reported no individual preparedness actions may have perceived the hazard as less severe, acted previously or collectively, relied on external assistance, or retrospectively evaluated warnings differently from households that reported taking action. This interpretation is consistent with evidence from Ethiopia's Somali Region, where access to accurate forecasts did not necessarily translate into community action when institutional intervention was anticipated (Mabumbo et al., 2024), and with broader evidence suggesting that sustained reliance on external assistance can influence proactive household responses (Nhamo et al., 2019). Future research distinguishing the source and nature of preparedness actions will help clarify these mechanisms.

Preparedness resource needs showed a non-linear association with action: households identifying a moderate number of unmet needs were significantly more likely to act than those identifying only one; however, this association did not extend to households reporting the greatest number of unmet needs. One plausible reading, consistent with evidence that recognizing a gap can itself motivate corrective action up to a point (Esang, 2026), is that moderate awareness of specific resource gaps prompts households to act on what they can address, while households facing the broadest set of unmet needs may confront constraints that are financial, logistical, or informational, and too severe to overcome through motivation alone. This reframes the finding: it is not that fewer unmet resource needs do not necessarily enable preparedness in a simple linear manner, but that the number and severity of unmet needs households recognize appear to interact with their ability to respond to them. Age was the only sociodemographic factor that remained significant after adjustment, with older respondents more likely to exhibit a pattern more consistent with an accumulated disaster experience explanation than with the common assumption that older age signals vulnerability and dependence (Kitanović, 2020). Repeated exposure to drought and environmental shocks in rural Somaliland may sharpen risk perception, independent of formal education or resource access (Barre and Osman, 2026; Omer et al., 2025, 2026b; Sharmake et al., 2022).

These findings suggest that improving the timeliness of warning communication and strengthening institutional trust alone may not necessarily translate into greater reported household preparedness. Further research is needed to determine whether reliance on institutional support, perceived response efficacy, previous experience, or other unmeasured mechanisms contribute to the observed inverse associations. Practical priorities include pairing warning dissemination with material support, engaging older community members as sources of experiential knowledge in preparedness programming, and directly addressing the near-universal absence of household emergency plans.

This study had several limitations. First, its cross-sectional design precludes causal inference and limits the assessment of the temporal relationships between perceptions of early warning information and reported household preparedness. Second, preparedness was measured as a binary outcome and therefore does not capture the type, intensity, effectiveness, or source of preparedness actions, including whether the actions were self-initiated or externally supported. Consequently, institutional reliance and other potential mechanisms underlying the observed associations should be regarded as hypotheses for future investigations rather than established explanations. Third, the Preparedness Resource Need Index reflects self-reported unmet needs rather than direct measures of household resource availability, such as assets, savings, or access to credit. Sociodemographic characteristics relate to the interviewed respondents because their roles within the household were not recorded, while self-reported preparedness may also be subject to recall and social desirability biases. In addition, purposive regional selection and incomplete household sampling frames in some locations limit the generalizability of the findings to rural Somaliland. No formal a priori sample size calculation was conducted, although the achieved sample was subsequently assessed as adequate for the planned multivariate analysis.

The multistage sampling design also involved clustering within sampling areas, but cluster identifiers were not available in a form that permitted cluster-adjusted variance estimations. Consequently, the regression standard errors do not account for potential intra-cluster correlation, and the reported confidence intervals and p-values may overstate the statistical precision if substantial within-cluster correlation is present. Therefore, inferential findings should be interpreted cautiously, particularly for associations near the conventional thresholds of statistical significance. Furthermore, region-level fixed effects, interactions across the Preparedness Resource Need Index categories, and formal trend tests for the ordinal communication variables were not examined and warrant consideration in future research. Despite these limitations, this study contributes empirical evidence on household preparedness in an under-researched, resource-constrained setting. It provides a basis for further research on early warning systems in Somaliland and comparable drought-prone contexts.

5 Conclusion

This study provides evidence that household preparedness in response to early warning information in rural Somaliland is associated with factors beyond warning receipt alone. In the adjusted analysis, institutional trust and perceived warning timeliness showed strong inverse associations with reported preparedness action, while preparedness resource needs and respondent age were also associated with the outcome. These findings highlight the importance of considering communication, institutional, resource-related, and sociodemographic factors when examining household responses to early warning information. The mechanisms underlying the inverse associations between institutional trust, warning timeliness, and reported preparedness cannot be established using the present cross-sectional data. In particular, this study did not measure institutional reliance, expectations of external assistance, perceived institutional response efficacy, or the motivations underlying household preparedness actions. Therefore, these mechanisms remain hypotheses for future investigation rather than explanations established by the present findings.

The results suggest that communication-focused approaches to early warning may be insufficient on their own to strengthen household preparedness in resource-constrained settings. Early warning programming may benefit from complementing accessible and timely communication with appropriate preparedness resources and support that enable households to translate warning information into feasible, protective actions. Longitudinal and mechanism-focused studies are needed to clarify the pathways underlying the observed associations and distinguish self-initiated, community-coordinated, and externally supported preparedness actions. Overall, this study contributes context-specific evidence on household preparedness in rural Somaliland and provides a basis for further research and early warning programming in comparable, last-mile, and resource-constrained settings.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Himanshu Grover, Alabama Agricultural and Mechanical University, United States

Reviewed by: Christopher Ihinegbu, University of California, Irvine, CA, United States

Samuel Mores Geddam, Alliance University, India

Pascal Buggs, Alabama Agricultural and Mechanical University, United States