Spatial distribution and determinants of unimproved sources of drinking water among Somali households: spatial and multilevel analysis using nationwide survey data Open Access

SHDS

Somali Health and Demographic Survey

SNBS

Somali National Bureau of Statistics

WASH

Water, sanitation, and hygiene

ICC

Inter-class correlation coefficient

AIC

Akaike information criterion

BIC

Bayesian Akaike information criterion

An unimproved drinking water source is characterized by its lack of protection against external contamination, particularly fecal contamination (Johri et al. 2019). The World Health Organization (WHO) and UNICEF, through their joint monitoring initiative, categorize drinking water sources as either improved or unimproved. Improved sources of drinking water include rainwater collection, tube wells or boreholes, protected springs, protected dug wells, public taps or standpipes, piped water into a home or yard, and protected springs. Conversely, unimproved sources encompass unprotected dug wells, unprotected springs, carts with small tanks or drums, tanker truck-provided water, surface water (such as rivers, dams, lakes, ponds, streams, canals, and irrigation channels), and bottled water, due to potential limitations on the quantity of water available to a household from this source, rather than its quality (Onda et al. 2012; Evans et al. 2013; World Health Organization Fund UNC 2021). Contaminated water from unimproved sources is the leading cause of mortality among children globally (Watkins 2006).

Drinking water quality remains a significant public health concern, particularly in middle- and low-income countries where access to improved sanitation and water supplies is limited (Cassivi et al. 2018; Abegaz & Midekssa 2021). The overall health and well-being of communities are contingent upon access to clean drinking water (Tetteh et al. 2022). It is estimated that approximately 80% of individuals in low- and middle-income countries are affected by inadequate water, sanitation, and hygiene (WASH) services (Prüss-Ustün et al. 2019). In Sub-Saharan Africa, and notably in Somalia, access to improved drinking water sources remains limited (World Health Organization Unicef 2008; Beyene et al. 2015; Ismail et al. 2024).

In 2020, 489 million individuals worldwide lacked access to improved drinking water facilities. This figure includes 122 million people who rely on surface water sources, such as rivers, dams, lakes, ponds, streams, canals, and other bodies of water, which can provide clean water due to their construction and design, as well as irrigation canals (WHO/UNICEF JMP 2017; World Health Organization Fund UNC 2021). Between 2000 and 2020, access to improved sanitation facilities – those designed to hygienically separate human waste from human contact – has increased (WHO/UNICEF JMP 2017; World Health Organization Fund UNC 2021).

Safe drinking water is a fundamental component necessary for sustaining life (Bartram & Cairncross 2010; Meride & Ayenew 2016; Ritchie & Roser 2017; Prüss-Ustün et al. 2019). The Sustainable Development Goal, endorsed by the United Nations in 2015, mandates that by 2030, all individuals should have access to safe and affordable drinking water (Cassivi et al. 2018). Access to improved water and sanitation is associated with reduced morbidity, decreased mortality, and a lower risk of diarrhea in children (Hasanain et al. 2012; Wolf et al. 2014; Wolf et al. 2018; Bain et al. 2020).

Despite limited coverage, water hygiene and sanitation continue to be significant public health concerns in Sub-Saharan Africa (Bain et al. 2014). Low-income households are particularly vulnerable during periods of water scarcity (Majuru et al. 2016). Notable disparities in access to water and sanitation facilities persist between urban and rural communities in Sub-Saharan Africa (Hassan et al. 2024a; Ismail et al. 2024; Ali et al. 2025; Dahir et al. 2025). WASH supply is crucial for the prevention and control of communicable diseases (World Health Organization Fund UNC 2021). WASH practices remain essential for reducing exposure to pathogens that cause waterborne illnesses, particularly among children in low-resource settings (Brown et al. 2013; Cumming et al. 2019; Rogawski McQuade et al. 2020).

The coverage of water supply and sanitation directly impacts population health and well-being (Bartram et al. 2014; WHO/UNICEF Joint Water Supply & Sanitation Monitoring Programme 2014; Hutton & Chase 2018; Hussain et al. 2024; Okesanya et al. 2024). It is well-established that adverse environmental conditions influence the transmission of infectious diseases. It is estimated that the disease burden from WASH accounts for 5.7% of the global disease burden (measured in DALYs) and 4.0% of the total mortality rate (Prüss et al. 2002). Numerous human diseases are present in excreta, and the removal of these pathogens from the environment significantly affects community health (Saunders & Warford 1976; Prüss et al. 2002).

Ensuring access to clean drinking water is imperative for promoting population health and well-being (Tetteh et al. 2022). Contaminated water is a known vector for diseases such as cholera, dysentery, polio, typhoid, and diarrhea, which result from consuming contaminated water from unimproved sources (WHO 2019). It is unsurprising that households sourcing their drinking water from improved sources report greater satisfaction with the availability and quality of the water compared to those relying on less satisfactory sources (Abebaw et al. 2010). The incidence of waterborne illnesses is significantly lower when access to improved water sources is available (Pullan et al. 2014).

Approximately half of all accidental deaths due to diarrhea worldwide are attributed to inadequate drinking water and sanitation (Prüss-Ustün et al. 2019). The utilization of unsafe water sources increases the likelihood of various infectious diseases, including cholera, typhoid, dysentery, schistosomiasis, salmonellosis, and infections affecting the respiratory, skin, and ocular systems (Teka 1977; Blum & Feachem 1983; Mills & Cumming 2016; Saxena et al. 2018; Prüss-Ustün et al. 2019; Aragaw et al. 2023). Furthermore, diseases such as trachoma, scabies, and helminthiasis may arise from insufficient access to water (Freeman et al. 2017; Prüss-Ustün et al. 2019).

Reports indicate that consumption of water from unimproved sources is associated with maternal undernutrition, suggesting that enhancing awareness and education regarding water quality can have broader health implications. It is well-documented that the use of unimproved water sources can adversely affect health. For instance, research indicates that children from households relying on unimproved water sources are more susceptible to morbidity, particularly diarrheal diseases (Manalew & Tennekoon 2019; Afrifa-Anane et al. 2022; Amadu et al. 2023).

In numerous developing countries, the prevalence of gastrointestinal infections linked to inadequate sanitation is a defining characteristic of the disease landscape (Fewtrell et al. 2005). The provision of safe drinking water, adherence to food hygiene practices, and sanitary disposal of excreta are identified as the most cost-effective and beneficial strategies for disease prevention (Rajagopalan & Shiffman 1974; Prüss-Ustün et al. 2014; Chen et al. 2024a, b; Zhang et al. 2025).

Access to clean water, adequate sanitation, and hygienic facilities can mitigate the risk of diarrheal diseases (Montgomery & Elimelech 2007; Edition 2011; Ali et al. 2025). The primary defense against disease involves ensuring a safe and sufficient water supply, proper disposal of human waste and excreta, and the assurance that food, vegetables, and beverages are free from harmful organisms or their byproducts, alongside the control of pests such as flies, lice, and mosquitoes (Teka 1984).

In Somalia, over half of the population lacks access to clean water sources, and the incidence of water-borne diseases is on the rise. According to FAO-AQUASTAT, Somalia's water scarcity is attributed to poor water management, as the total renewable water resources available are significantly less than the total produced (Mourad 2020). This has led to a nationwide water shortage. Consequently, an improved management system could substantially address Somalia's water challenges (El Kharraz et al. 2012).

The destruction of water systems during the civil war has impeded progress toward achieving Sustainable Development Goals (Jama & Mourad 2019; Mourad & Avery 2019). Additionally, upstream innovations in Ethiopia have impacted Somalia's irrigation practices and water availability (Michalscheck et al. 2016). These challenges encompass a broad spectrum of issues, ranging from climate hazards such as droughts and floods to local water management concerns, including livestock access to water intended for human use, exacerbating health vulnerabilities in many rural communities. Addressing these issues requires local governments to seek assistance and international communities to promote national development. However, the nation has been unable to surmount these challenges due to a lack of coordination and cooperation among various stakeholders (Mourad 2020; Balthasar 2022).

A study conducted in Somalia revealed that 51.5% of households utilized hand-dug wells equipped with pumps, while 61.3% of households traveled 3 kilometers to access water. Of these households, 40.5% consumed 120 liters of water daily, whereas 23.9% consumed between 61 and 90 L. More than half of the respondents reported insufficient water supply, and 64.4% did not engage in water treatment practices. Additionally, many households shared latrines, which were found to be unhygienic (Saed et al. 2021).

Approximately 77.7% of the Somali population has access to improved water sources during the rainy season, compared to 74.7% during the dry season. This discrepancy is primarily attributed to the lower usage of improved drinking water sources in rural areas (74.2–69.6%) and nomadic communities (61.6–52.3%) (Somalia 2023). The predominant improved sources of drinking water include boreholes and tubewells (11.7%/11.5% during the rainy/dry season), followed by piped water into homes (44.7%/41.8%). In rural areas, 35.7% of the population has access to piped water, while in urban areas, this figure rises to 54.9%. Tubewells and boreholes are more commonly used in nomadic regions (36.9%). Despite this, non-improved water sources are still utilized, with nomads relying more heavily on unimproved and natural water sources, such as lakes, rivers, dams, and water catchments, particularly during the rainy season (16.9%), followed by rural residents (10.7%/9.5% during the rainy and dry seasons) (Somalia 2023). Significant regional disparities in drinking water sources have been identified in studies conducted in Nepal and other sub-Saharan nations (Pullan et al. 2014; He et al. 2018).

Currently, there is a lack of data on unimproved drinking water sources, particularly regarding their spatial variation at the national level. Some studies have examined water access in specific regions or populations within Somalia (Ismail et al. 2024). A comprehensive national-level analysis of unimproved water sources and their determinants using spatial techniques and modern statistical methods is lacking. Therefore, it is imperative to investigate the spatial hotspots of unimproved drinking water sources in Somalia and inform policymakers about the specific administrative areas that require the most policy attention to achieve the desired goals and address information gaps.

Consequently, this study aims to map and identify the spatial hotspots of unimproved drinking water sources using data from the Somali Health and Demographic Survey 2020, the first of its kind in Somalia. This research contributes to the monitoring of Sustainable Development Goal (SDG) 6, particularly target 6.1, which aspires to achieve universal and equitable access to safe and affordable drinking water for all by 2030.

The study was conducted in Somalia, encompassing both rural and urban areas. The specific regions and geographical sub-divisions within Somalia that are analyzed were dependent on the availability of complete and reliable data from the SHDS 2020 dataset, particularly considering areas excluded due to security concerns during the original survey. Specifically, the Lower Shabelle and Middle Juba regions were not covered in the SHDS 2020 data collection due to security constraints and are therefore excluded from this analysis.

A population-based, cross-sectional study was employed utilizing the SDHS 2020 data, which is nationally representative and provides valuable information on population characteristics, health status, and access to improved sanitation and water sources.

The source of data was SDHS 2020 which is the first ever done in Somalia. Sample size: The SDHS 2020 collected data from 16,360 households, with a successful interview rate of 99.7%. Sampling procedure: The SDHS employed a three-stage stratified cluster sampling design in urban and rural areas. Nomadic populations were sampled using a two-stage stratified cluster design. Sampling probabilities were proportional to the size of primary sampling units (PSUs) and secondary sampling units (SSUs), with systematic sampling of households in the final stage. This ensures the representativeness of the Somali population. After cleaning the sample size of the study was 15,823.

The Somali National Bureau of Statistics (SNBS) implemented stringent data quality control measures throughout the data collection process of the SHDS 2020. These measures encompassed comprehensive interviewer training, standardized data collection protocols, field supervision, and data entry quality checks. Additionally, the SNBS conducted consistency checks to identify and address any inconsistencies or outliers in the data. Furthermore, the data underwent review by technical experts from the SNBS, UNFPA, and other partner organizations to ensure accuracy and compliance with established standards. These validation steps enhance the reliability and validity of the SHDS 2020 data for subsequent analysis and interpretation.

Variables of the study were taken from the previous studies (Bogale 2020; Kassie & Mengistu 2022; Aragaw et al. 2023). The outcome variable of the study was the usage of unimproved drinking water sources. The drinking water source was classified as unimproved if a household gets drinking water from an unprotected dug well, unprotected spring, surface water, and others (Aragaw et al. 2023). The covariates of the study were classified into two, Individual/Household level and Community Level. Individual/Household levels were Age of Household Head, Sex of Household Head, Household size, Educational attainment of the Household Head, Household wealth index, Marital Status of women aged 15–49, and Media exposure (VT and radio). Community Level factors were Region, Place of residence, Community level of education, and Community level poverty.

Community-level variables were derived by aggregating individual household data within each PSU. PSUs are geographic areas that represent the first stage of clustering in the DHS sampling design and serve as proxies for communities in this analysis. ‘Community level of education’ was categorized as ‘High’ or ‘Low’ based on whether the proportion of household heads in the PSU who had ever attended school was above or below the national median proportion across all PSUs. Similarly, ‘Community level poverty’ was categorized as ‘High’ or ‘Low’ based on whether the proportion of households in the PSU belonging to the two lowest wealth quintiles was above or below the national median proportion.

Frequencies and percentages were used to summarize the prevalence of unimproved drinking water sources and to understand the overall patterns in the data. Multicollinearity among the independent variables in the final model was assessed using the variance inflation factor, and values were found to be well below common thresholds for concern (VIF < 2.5; Mean VIF = 1.45), indicating that multicollinearity was not a significant issue (see Supplementary Table S1 within acceptable limits (VIF < 5)), suggesting multicollinearity was not a significant issue. A mixed-effects logistic regression model was also implemented using the software of Stata-16.0 and R-4.3.3 versions to identify factors associated with the use of unimproved drinking water sources. This approach was preferred to account for the multilevel structure of the data, where individuals were nested within communities.

A two-level mixed-effects logistic regression model was implemented. The primary analysis focused on identifying the main effects of the selected individual/household and community-level covariates on the use of unimproved drinking water sources. Interaction effects, such as those between poverty and locality (urban/rural/nomadic), were not explicitly modeled in this study, though their potential importance is acknowledged in the limitations section as an area for future investigation. This approach accounted for the hierarchical structure of the data, with households (Level 1) nested within communities (PSUs, Level 2). The model included fixed effects for the individual/household and community-level covariates listed in Table 2, and a random intercept at the community level to account for unexplained variability between communities and the non-independence of observations within the same community.

A geographic information system application was used for spatial analysis to evaluate regional differences in cases of unimproved drinking water sources across the 2020 SDHS clusters and R Software was used to generate maps and visualize the spatial patterns of unimproved water source use. The Global Moran's ‘I’ statistic was employed to assess whether the patterns of unimproved drinking water sources in the study area were scattered, clustered, or randomly distributed. Moran's ‘I’ is a spatial statistics tool that measures spatial autocorrelation by analyzing the entire dataset and yielding a single output value ranging from −1 to +1. A Moran's ‘I’ value close to −1, +1, and 0 indicates a spread, aggregation, or random distribution of unimproved water sources, respectively. Additionally, a significantly positive spatial autocorrelation suggests the clustering of similar sources in the geographic area, while a significant negative autocorrelation indicates that neighboring water sources are less similar than expected by chance, revealing spatial patterns. A statistically significant Moran's ‘I’ (P < 0.05) led to the rejection of the null hypothesis (which posits a random distribution of unimproved water sources) and confirmed the presence of spatial autocorrelation (Kassie & Mengistu 2022).

The data for this study was accessed from the Somali National Bureau of Statistics (SNBS) SHDS 2020. The SNBS is responsible for protecting the privacy and confidentiality of all survey participants. Informed consent and ethical review were part of the original SDHS 2020 data collection process. This study followed the SNBS's established data privacy and ethical guidelines, ensuring the anonymity of individuals in the analysis. As a secondary data analysis study, ethical clearance was sought from an appropriate ethical review board and Institutional Review Board within Amoud University.

The data illustrated in Table 1 investigates the correlation between various socio-demographic factors and the types of drinking water sources, classified as ‘unimproved’ and ‘improved.’ The analysis utilizes chi-square tests to evaluate the significance of these relationships, with p-values reflecting the strength of the associations. A significant observation is the impact of the gender of the household head on the choice of water source. In households led by males, 29.37% depend on unimproved sources, while 70.63% utilize improved sources. Conversely, in female-headed households, 26.29% rely on unimproved sources, with 73.71% opting for improved sources. The chi-square value of 16.40 and a p-value of 0.000 indicate that the gender of the household head is a critical determinant of the type of drinking water source used. Marital status is another important factor influencing the quality of water sources. The findings indicate that 29.39% of married individuals utilize unimproved sources, whereas the rate is lower at 16.30% for those who are divorced or abandoned. The chi-square statistic of 76.01 and a p-value of 0.000 demonstrate a robust relationship between marital status and the availability of improved drinking water. Furthermore, access to media is significantly associated with the quality of water sources. Households with a television report only 3.15% using unimproved sources, in contrast to 32.26% in households lacking a television. This strong correlation is reinforced by a chi-square value of 767.50 and a p-value of 0.000. Similarly, households with a radio exhibit 20.11% using unimproved sources, while those without a radio show 29.80%. The chi-square value of 93.64 and p-value of 0.000 confirm that media access is related to the use of improved water sources. The level of education is a crucial determinant of water source quality. Individuals who have received formal education demonstrate a marked preference for improved water sources, with 79.60% favoring them, in contrast to only 32.59% of those who have never attended school. The chi-square analysis reveals a significant correlation, underscoring the influence of education on the availability of better drinking water. Economic status, as indicated by the wealth index, presents a distinct disparity in the utilization of water sources. Notably, 54.70% of individuals belonging to the lowest wealth quintile depend on unimproved sources, whereas a mere 2.56% of those in the highest quintile do the same. The chi-square statistic of 3.7e + 03, along with a p-value of 0.000, emphasizes the strong relationship between economic status and the quality of water sources accessible to households. The age of the head of the household also plays a role in determining water source quality. Households led by individuals younger than 26 years show a higher reliance on unimproved water sources, with 34.42% using them, a figure that declines in older age brackets. The significant chi-square value of 34.99 and a p-value of 0.000 indicate that the age of the household head is a vital factor influencing access to improved drinking water. The size of a household significantly influences the quality of water sources utilized. Households with fewer than five members exhibit a higher reliance on unimproved water sources, with 31.54% of them doing so, in contrast to 23.29% of larger households, which consist of five or more members. The chi-square statistic of 126.07, accompanied by a p-value of 0.000, demonstrates a significant relationship between household size and the type of drinking water source employed. The type of residence is another crucial factor affecting access to water sources. Urban households show a marked inclination toward improved water sources, with only 5.94% resorting to unimproved options. In stark contrast, 56.52% of nomadic households depend on unimproved water sources. The chi-square value of 3.5e + 03 and a p-value of 0.000 reveal a strong association between the type of residence and the availability of improved drinking water. Regional disparities also play a significant role in the accessibility of water sources. For instance, 67.52% of households in Bakool utilize unimproved water sources, whereas an impressive 98.26% of households in Banaadir have access to improved sources. The chi-square values across various regions highlight considerable differences, with a p-value of 0.000 confirming the statistical significance of these variations.

Table 2 presents the results from the four nested multilevel logistic regression models. Model I is the null model, Model II includes individual/household factors, Model III includes community-level factors, and Model IV is the final, fully adjusted model including factors from all levels, which serves as the primary basis for interpreting the determinants. The analysis employs four nested models to investigate the effects of individual and community-level factors on the odds of relying on unimproved sources.

Model I, the empty model, reveals a substantial between-community variability, with an Inter class Correlation Coefficient (ICC) of 69%. This indicates that a significant portion of the total variation in unimproved water sources is attributable to differences between communities.

Model II incorporates individual-level variables. The analysis reveals that female-headed households were associated with a 13.6% reduction in the odds of unimproved water source usage compared to male-headed households (AOR = 0.864; 95% CI: 0.767–0.973; p < 0.05). Additionally, households with abandoned household heads had 65% higher odds of using unimproved drinking water sources compared to married household heads (AOR = 1.653; 95% CI: 1.148–2.380; p < 0.05), whereas households with widowed or never-married heads did not demonstrate statistically significant differences. The study also found strong gradients across wealth indexes, with the odds of unimproved water use decreasing significantly as wealth increases, compared to the lowest wealth index. For example, households in the highest wealth quintile had a 94% lower odds of unimproved water use (AOR = 0.057; 95% CI: 0.039–0.084; p < 0.001) compared to the lowest wealth quintile. Other individual-level factors such as the age of the household head, whether or not the household had a television, whether or not the household has ever attended school, and the number of household members were not statistically significantly associated with unimproved drinking water sources. Interestingly, however, having a radio at home was associated with 13% lower odds of using unimproved sources, although this association was not statistically significant in this model (AOR = 0.870; 95% CI: 0.755–1.004; p ≥ 0.05).

Model III introduces community-level variables, finding that geographical region is a significant factor influencing unimproved water use. Compared to Awdal region (the reference region), nearly all regions of Somalia showed significantly different odds of using unimproved drinking water sources, with the strongest associations observed for Banadir (AOR = 6.571; 95% CI: 4.686–9.213; p < 0.001), which had 6.5 times higher odds of using unimproved water sources compared to Awdal. Also, compared to Awdal, households in Mudug (AOR = 3.363; 95% CI: 2.424–4.666; p < 0.001) and Bakool (AOR = 2.664; 95% CI: 1.840–3.855; p < 0.001) also showed significantly elevated odds of using unimproved drinking water sources. However, the central regions of Somalia (Sool, Bari, Nugaal, and Hiraan) had significantly lower odds of using unimproved drinking water sources. The analysis also found that living in urban areas was associated with almost 3 times higher odds of using an unimproved drinking water source, and living in nomadic settings was associated with significantly higher odds of using an unimproved drinking water source (AOR = 31.91; 95% CI: 18.42–55.278; p < 0.001). In addition, community literacy rates were associated with 2.4 times higher odds of unimproved drinking water source use compared to areas with low literacy rates (AOR = 2.443; 95% CI: 1.637–3.644; p < 0.001).

Model IV presents a combined model that includes both individual-level and community-level factors. After controlling for community-level factors, the significant effects of female-headed households were no longer statistically significant. The effect of having a radio also was attenuated. In this full model, the effect of abandoned household heads on unimproved water sources persists. Also, the wealth gradients remain similar to previous models, with higher wealth indices associated with lower odds of unimproved drinking water sources. Among the community-level variables, significant regional effects persist, for example, the odds of unimproved drinking water sources in Banadir are 50% lower, after controlling for individual-level variables (AOR = 3.363; 95% CI: 2.424–4.666 in Model III, compared to AOR = 0.517; 95% CI: 3.747–7.473 in Model IV). The community effects of place of residence (urban, nomadic) and community literacy levels also persist.

The random effects show a reduction in ICC across all models, with Model IV yielding an ICC of 36%. This demonstrates that both individual-level and community-level factors help explain the clustering of unimproved water sources. Finally, the AIC and BIC values progressively decrease across the models, indicating an improved fit with each model addition. Overall, the multilevel analysis underscores the importance of both individual and community-level determinants in influencing access to improved drinking water sources in Somalia, with a strong emphasis on the role of wealth, geographical region, place of residence, community literacy rates, and household head characteristics.

This study offers a comprehensive analysis of the spatial distribution and determinants of unimproved drinking water sources among Somali households, leveraging the first-ever nationally representative Somali Health and Demographic Survey (SHDS) data from 2020. Our findings reveal significant spatial disparities and multilevel factors influencing access to safe drinking water, underscoring the complex challenges of achieving SDG 6 in Somalia. This work stands as a unique contribution given the limited prior national-level research in Somalia, and provides important insights for future public health interventions.

The spatial analysis, employing Moran's I and Getis-Ord Gi* statistics, demonstrates significant spatial autocorrelation and clustering of unimproved drinking water sources across Somalia. The global Moran's I of 0.278 (p < 0.05) confirms a statistically significant pattern of clustering, refuting the idea of a random distribution (6 0). The spatial lag map visually confirms this, highlighting areas with high concentrations of households using unimproved sources, notably in the northern and central regions, and parts of the mid-southern area. Local Moran's I analysis further pinpoints specific hot spots with high rates of unimproved source usage in the northeast, southwest, and middle of Somalia, possibly linked to similar environmental conditions, infrastructure deficits, and localized practices. Getis-Ord Gi* analysis identified hot spots for high unimproved water use in the north and a cold spot for lower use in the southwest. These disparities echo findings from studies in other sub-Saharan African nations, such as Ethiopia and Nepal, which also document spatial inequalities in access to improved water sources (He et al. 2018; Kassie & Mengistu 2022). Pullan et al. (2014) demonstrated geographical inequalities in access to improved water across sub-Saharan Africa (Pullan et al. 2014). Our study's unique contribution is providing this granular, national-level insight into Somalia, where such data has been historically scarce.

Our multilevel analysis identified important factors at both individual/household and community levels that influence the use of unimproved drinking water sources. Household-Level Factors: The descriptive statistics show the gender of the household head to be a statistically significant factor in the type of water source used. The initial multilevel model showed that female-headed households had lower odds of using unimproved water sources compared to male-headed households; however, this association was attenuated in the full model. The nuanced effect of gender highlights the complex social dynamics that affect water source selection (Aragaw et al. 2023). This finding contrasts with some studies that found no gender difference and others that have found higher risks to female-headed households. This nuanced finding is therefore unique to the context of Somalia. Further, households with abandoned heads consistently had higher odds of relying on unimproved sources across all the models, implying the role of household structure and support. After adjusting for other factors, other household factors like the presence of a television, the age of the household head, or the number of household members, were not statistically significantly associated with accessing safe water, suggesting other community factors, poverty, and education play a major role.

Consistent with a multitude of studies (Kassie & Mengistu 2022; Aragaw et al. 2023), household wealth was a strong predictor of water source (Farih et al. 2024; Hassan et al. 2024b; Yousuf et al. 2024). Households in the lowest wealth quintile consistently had higher odds of using unimproved water, while those in the highest wealth quintile showed a significant decrease in their use of unimproved sources, a finding that was consistent across all models. This finding corroborates the understanding that socioeconomic status significantly shapes access to basic services (Hussain et al. 2024). This relationship likely extends beyond just affordability and touches on access to resources that would enable improvements to water storage or treatment in addition to infrastructure.

The study found that households with a radio at home had lower odds of relying on unimproved sources. This underscores the importance of radio as a channel for public health information and promoting safe water use (Ismail et al. 2024), a finding consistent with other studies that have emphasized the role of education in promoting WASH practices. Community-Level Factors: Community-level factors played a substantial role in determining access to safe water. Region: The analysis revealed significant regional variations in unimproved water source use. Households in Banaadir (Mogadishu) had significantly lower odds compared to those in the Awdal region, possibly reflecting differences in urban infrastructure and investments in safe water in the capital, as well as higher infrastructure development and less vulnerable rural areas in Awdal. Conversely, regions like Bakool and Mudug demonstrated a greater reliance on unimproved sources, emphasizing the need for targeted interventions in these underserved regions. This observation of significant regional disparities in water access aligns with previous studies in the literature (Pullan et al. 2014; He et al. 2018). Place of Residence: Nomadic communities exhibited significantly higher odds of using unimproved water, followed by rural communities, which is likely due to their higher mobility, and challenges with water pipelines and other infrastructure. Urban areas had significantly lower odds. These patterns underscore the urban-rural and nomadic disparities, which are consistent with previous research and especially relevant in Somalia (Somalia 2023), where conflict and lack of development have exacerbated inequalities. Community Education: Community literacy rates had significant associations with the use of unimproved drinking water sources. This was a unique finding because, unlike other studies that found individual education of the household head to be a significant predictor, this study found community-level literacy to be more strongly associated with the usage of unimproved water sources. This phenomenon could be attributed to several factors within the Somali context. Higher community literacy may foster a collective understanding and appreciation of health and hygiene principles, facilitating the broader dissemination and adoption of information regarding safe water practices through social networks and community forums. Furthermore, more literate communities might possess greater social capital and collective efficacy, enabling them to better organize, advocate for, and manage communal water resources and infrastructure improvements. In environments with limited individual resources, the collective capacity enhanced by community education could therefore play a more decisive role than individual education alone.

This study makes a unique contribution to the understanding of water access challenges within the Somali context. Firstly, Somalia's history of conflict has severely damaged water infrastructure, caused population displacement, and created instability in rural areas, affecting water availability and distribution (Jama & Mourad 2019; Mourad & Avery 2019). The role of population displacement is particularly critical; prolonged conflict and recurrent environmental shocks like droughts and floods have led to large-scale movements of people, often into urban peripheries or Internally Displaced Persons (IDP) camps. These displaced populations frequently face acute shortages of basic services, including safe water, as settlements are often informal and lack adequate infrastructure. The strain on existing, often meager, water sources in host areas, coupled with the precarious living conditions of the displaced, significantly increases reliance on unimproved and unsafe water sources, thereby perpetuating cycles of vulnerability and ill health. Saed et al. (2021) also emphasized similar issues in IDP camps in Mogadishu (Saed et al. 2021). Secondly, this study emphasizes the unique challenges facing nomadic and rural communities, whose water access often relies on seasonal rains and surface water sources which are particularly prone to contamination (Somalia 2023). Third, while other studies have demonstrated socioeconomic and regional disparities in water access, this study provides unique insights into the national context of Somalia, where such large-scale studies are lacking. It is the first study to map out national-level access to unimproved water sources and their determinants. Finally, this study uniquely identified community education, rather than individual education of the household head, to be a significant factor in utilizing improved water sources, highlighting the importance of promoting literacy and education at the community level.

The results of this study also show common trends, as well as differences when compared to existing literature (Qi et al. 2023; Chen et al. 2024a, c). As found in the literature, socioeconomic status remains a major determinant of using improved water sources (Pullan et al. 2014; Kassie & Mengistu 2022). The role of media in water-related health promotion is also demonstrated in this study (Bain et al. 2020; Ismail et al. 2024). Our findings are consistent with existing literature in Sub-Saharan Africa, including evidence from Ethiopia (Aragaw et al. 2023), which also showed differences between urban, rural, and nomadic communities, and the role of poverty as a key factor (Bogale 2020; Kassie & Mengistu 2022). This study also demonstrated differences in the literature. For example, this study showed that community literacy had a more direct effect than the household head's education and that socioeconomic status was highly dependent on geographical region within Somalia, a nuanced finding that is unique to this study.

Our findings carry several implications for policy and practice. First, interventions should be spatially targeted, focusing on the identified hot spots and regions with high dependance on unimproved water sources in North East, South West and Middle of Somalia. Second, water and sanitation programs must prioritize the unique needs of nomadic and rural populations, which are complicated by their mobility and environmental vulnerabilities (Somalia 2023). Third, interventions should not just focus on infrastructure, but also improve socioeconomic conditions and community education levels. Fourth, the usage of community radio as a key platform for health information and promoting safe water practices should be prioritized. Finally, this study underscores the necessity of a multi-sectoral approach to water security, bringing together health, education, and infrastructure sectors, as well as coordination between governmental agencies, humanitarian bodies, and community-based programs.

This study is not without limitations that should be considered when interpreting the findings. Firstly, the cross-sectional nature of the data inherently limits our ability to infer causal relationships. Data were collected at a single point in time, preventing the establishment of temporal sequences between potential determinants and the use of unimproved water sources. Consequently, we can only report associations and cannot definitively establish causality, acknowledging the potential for residual confounding from unmeasured variables.

Secondly, the data were collected in 2020. While representing the most recent comprehensive national dataset available, Somalia has since experienced significant environmental events, including severe droughts and floods, alongside ongoing localized instability and potential humanitarian responses. These factors may have altered water access patterns and infrastructure conditions in certain regions since the survey was conducted. Therefore, our findings should be interpreted as providing a crucial baseline from 2020, underscoring the need for continuous monitoring and updated data collection to capture the evolving situation.

Thirdly, the study's geographic scope was limited by the necessary exclusion of Lower Shabelle and Middle Juba due to security concerns that prevented access during the SHDS 2020 data collection. This exclusion may introduce selection bias, as these regions could potentially exhibit systematically different patterns of unimproved water source use compared to the included areas. Consequently, while the SHDS 2020 is nationally representative of the accessible areas, our findings may not be fully generalizable to the entirety of Somalia, and the national estimates of unimproved water source use could be affected if these excluded regions differ significantly. Furthermore, although the survey achieved a very high overall response rate (99.7%), the potential for non-response bias, however small, cannot be entirely dismissed if the non-participating households differed significantly regarding water access.

Fourthly, our analysis was constrained by the variables available in the SHDS dataset. For instance, the survey does not include detailed information about the specific physical conditions, functionality, or microbial quality of the water sources used, which could influence household choices and health outcomes. The modeling also focused primarily on main effects to provide a foundational understanding; potential interaction effects between key variables (such as household wealth and type of residence), which could offer more nuanced insights into how determinants operate under different conditions, were not explored in the final models presented here. Future studies could build on these findings by investigating such interactions to refine targeted interventions.

Additionally, the spatial analysis employed exploratory statistics (Moran's I, Getis-Ord Gi*) to identify the presence and location of spatial clustering. While valuable for highlighting geographic patterns and potential hotspots, these methods do not inherently control for underlying covariates like poverty or regional infrastructure levels in the same way spatial regression models would. Although beyond the scope of the current foundational study, future research employing methods such as Geographically Weighted Regression (GWR) or Spatial Autoregressive models, or analyzing the spatial autocorrelation of residuals from the multilevel models, could provide a more robust understanding of spatial dependencies after accounting for known factors. The application of these more advanced spatial regression techniques is planned for subsequent analyses building on the current work.

Despite these limitations, this study utilizes the first-ever nationally representative SHDS data, offering valuable new insights into the complex multilevel and spatial factors influencing access to unimproved drinking water sources in this critical context. The findings provide an important evidence base for targeting interventions and monitoring progress toward SDG 6 in Somalia.

Our findings carry several implications for policy and practice aimed at improving access to safe drinking water in Somalia. First, interventions should be spatially targeted, focusing on identified hot spots and regions with high dependance on unimproved water sources in the North East, South West, and Middle of Somalia, necessitating detailed mapping of water resources and infrastructure to inform resource allocation and intervention strategies. Second, water and sanitation programs must prioritize the unique needs of nomadic and rural populations, which are complicated by their mobility and environmental vulnerabilities. This may involve mobile water treatment units, drought-resistant water storage solutions, and culturally appropriate hygiene education campaigns. Third, interventions should not just focus on infrastructure, but also improve socioeconomic conditions, particularly for the poorest households, by promoting income-generating activities, providing access to microfinance, and addressing food insecurity, all of which can improve the affordability of safe water solutions. Fourth, given the strong association between community literacy and access to improved water sources, investments in basic education and literacy programs are crucial. These programs should be tailored to the specific needs of communities and should emphasize the importance of safe water practices. Fifth, the usage of community radio as a key platform for health information and promoting safe water practices should be prioritized. Radio programs can be used to disseminate information on water treatment methods, hygiene practices, and the benefits of using improved water sources. Sixth, a multi-sectoral approach to water security is essential, bringing together the health, education, infrastructure, and environmental sectors. This requires strong coordination between governmental agencies, humanitarian organizations, and community-based programs to ensure a holistic and sustainable approach. Addressing the underlying issues of conflict, political instability, and weak governance is crucial for creating a stable environment in which water infrastructure can be developed and maintained. This may involve promoting peacebuilding initiatives, strengthening local governance structures, and ensuring the security of water resources. Finally, regular monitoring and evaluation of water access interventions are needed to assess their effectiveness and identify areas for improvement. This requires the collection of reliable data on water source use, water quality, and related health outcomes. By implementing these policy recommendations, Somalia can make significant progress toward achieving SDG and ensuring that all its citizens have access to safe and affordable drinking water.

This study has demonstrated significant spatial variations and multilevel determinants associated with the use of unimproved drinking water sources in Somalia. Our findings highlight the necessity of spatially targeted and context-specific interventions that prioritize vulnerable populations, addressing not just household-level factors, but also community-level factors, and should be done with communities, as well as humanitarian organizations. Achieving SDG 6 in Somalia requires addressing these complex spatial and multilevel challenges through comprehensive and coordinated efforts, with a focus on spatial inequality, poverty, and the promotion of community education.

We gratefully acknowledge the Somali National Bureau of Statistics (SNBS) for providing access to the 2020 Somali Health and Demographic Survey (SHDS) data. My thanks also extended to my wife, Sr. Hayat and my kids for giving me time-space in doing this study.

Conceptualized by M.A.H, and H.A; Developed methodology by M.A.H, Rendered support in data curation by M.A.H, A.M.Y, A.H.M; Analysis by M.A.H, A.M.Y and A.H.M; Wrote and original draft preparation by A.M.Y, H.A; M.A.H; Wrote and intellectual content, reviewed, and edited the article by A.H.M and S.N.; Supervised by A.H.M. and S.N. All authors reviewed and approved the final version of the manuscript.

The authors declare that this study was not supported by any external grants or funding.

All relevant data are available from an online repository or repositories. The dataset was accessed from https://microdata.nbs.gov.so/index.php/catalog/50.

The authors declare there is no conflict.