The study was conducted in five regions of South and Central Somalia: Banadir, Bay, Galgaduud, Gedo, and Lower Shabelle (Figure 1). The regional sample distribution was Banadir (n = 95), Bay (n = 94), Galgaduud (n = 94), Gedo (n = 94), and Lower Shabelle (n = 95), giving a total sample size of 472 mature camels (Figure 1). These regions represent important camel-keeping environments spanning dry inland rangelands, agro-pastoral zones, and comparatively more humid coastal and riverine areas. Therefore, the study area captured a broad ecological gradient within Somalia’s pastoral production landscape while remaining focused on regions where the target ecotypes were accessible under field conditions. These regions are characterised by arid to semi-arid climates, variable rainfall, high temperatures, and extensive livestock mobility, all of which shape camel management and performance.

Because the study covered heterogeneous production environments, region was considered an important contextual factor. Banadir and Lower Shabelle include coastal, peri-urban, riverine, and agro-pastoral settings, whereas Bay, Gedo, and Galgaduud include drier inland pastoral and agro-pastoral landscapes with variable browsing resources, water access, and herd mobility. Exact herd-level coordinates, altitude, rainfall, temperature, vegetation composition, and management variables were not consistently available for all sampled herds; this limited the ability to separate ecotype effects from environmental and regional effects.

A cross-sectional field survey was conducted using structured interviews and direct phenotypic assessments. Herds were identified through local pastoral networks, field contacts, and owner willingness to participate, and animals were selected from mature camels within accessible herds. Therefore, the sampling should be interpreted as field-based sampling of available owner-identified ecotypes rather than a fully probabilistic sample from a complete regional camel census. The sample comprised 472 mature camels, including 188 males and 284 females (Figure 2). Ecotype identity was initially reported by owners using locally recognised names and was checked against the external phenotypic criteria used by field staff, including body conformation, coat characteristics, and qualitative traits. Ecotype assignment was checked by trained field staff using agreed external phenotypic criteria; however, a formal inter-observer agreement statistic was not assessed and should be recognised as a field-classification limitation. Age was estimated using a combined assessment of owner/herder recall and dentition, including tooth eruption pattern, tooth wear, and general dental condition. When owner information and dentition were inconsistent, the dentition-based estimate was prioritised for age-class assignment; animals with highly uncertain age estimates were not included. Camels younger than 8 years were excluded to ensure the analysis focused on animals with stable adult morphology, while animals older than 15 years were excluded to minimise confounding from advanced age, tooth wear, and age-related changes in body condition. The animals were grouped into two age classes: 8–11 years and 12–15 years. This stratification allowed for ecotype comparisons within broadly comparable adult maturity classes. The sample size was based on the availability of mature animals from the four owner-identified ecotypes across the study regions and was considered sufficient for descriptive comparisons within age-sex strata. Accordingly, the study should be regarded as a field-based characterisation survey rather than a controlled genetic or longitudinal production trial.

Qualitative phenotypic observations considered in the study were face profile (facial measurements), ear size and orientation, body colour and colour pattern, hair type and length, hump size and position, and udder and teat characteristics in females, following the FAO framework for phenotypic characterisation of animal genetic resources (FAO, 2008).

The quantitative traits included in the study were 18 linear body measurements recorded in centimetres (cm), namely, face length, distance between the eyes, ear length, neck length, width at shoulders, height at withers, thoracic girth, anterior limb length, height at hump, abdomen circumference, hump length, hump height, hump circumference, body length, posterior limb length, tail length, and foot-pad circumference (Figure 3). These measurements were taken with the animal standing on level ground, using a measuring tape and a measuring stick for height traits. Face length was measured along the face from the frontal/nasal region to the muzzle; distance between the eyes was measured between the medial eye margins; ear length was measured from the ear base to the tip; neck length was measured from the head-neck junction to the shoulder region; height at withers was measured vertically from the ground to the highest point of the withers; thoracic girth was measured around the chest immediately behind the forelimbs; abdomen circumference/barrel girth was measured around the widest part of the abdomen; body length was measured from the shoulder point to the pin-bone region; tail length was measured from the tail base to the tip; and foot-pad circumference was measured around the pad. Limb and hump measurements were recorded using the same field landmarks throughout the survey. The variable coded as width at shoulders (WS) refers to the shoulder-region field measurement used in this survey and should not be interpreted as a simple skeletal transverse shoulder width measurement. The measurement landmarks shown in Figure 2 were checked against the field recording sheet and harmonised with the trait names used in the Methods and tables. For data quality, field data collectors were trained prior to data collection, and a common recording sheet and measurement protocol were used. Measurements were checked in the field by the lead recorder when more than one field worker was present. Values that appeared to be extreme during data analysis were rechecked against the available data files and interpreted cautiously where they could not be fully resolved.

Live body weight was calculated indirectly from shoulder height, thoracic girth, and barrel girth/abdomen circumference using the barymetric formula described by Yagil (1994). The estimated body weight (in kg) was calculated as follows: BW = SH × TG × BG × 50 where BW = estimated live body weight (kg), SH = shoulder height (m), TG = thoracic girth (m), and BG = barrel/abdomen circumference (m).

Shoulder height, thoracic girth, and barrel girth/abdomen circumference were combined after conversion to metres. This approach was used because direct camel weighing facilities were not practically available under field conditions. The equation was retained because no Somali camel ecotype-specific validated alternative was available and because it enabled approximate comparative assessment across animals under field conditions. Because a Somali-ecotype-specific validation or accuracy estimate for this equation was not available, estimated body weight should be interpreted as an approximate field indicator rather than as a direct measurement of live weight. Measurement errors in shoulder height, thoracic girth, or barrel girth/abdomen circumference may propagate into the body-weight estimate. Estimated body weight was reported in kilograms and analysed as a quantitative trait.

Milk production data were collected from lactating females. Daily milk yield was recorded as owner-reported milk offtake rather than direct measurement of total biological production. Herders reported the milk obtained from morning and evening milking using their own familiar local milking containers; these containers were standardised by the study team. Before standardisation by the study team, variation in local container size and fill level may have introduced additional measurement error into reported offtake estimates. Reported quantities were converted to litres during data entry, and the recall period was limited to 1–2 days to reduce recall bias. This short recall period likely reduced, but did not eliminate, recall and rounding error. Because calf suckling, residual milk, and total biological production were not directly measured, reported offtake is likely to underestimate total biological milk production. However, container-based reporting and rounding may cause either over- or under-estimation of human offtake. Key factors affecting milk yield include calf suckling before or after milking, residual milk, calf age, parity, stage of lactation, pregnancy status, health status, diet, watering frequency, and season, although these were not consistently recorded. Accordingly, the milk data should be interpreted as descriptive owner-reported milk-offtake information. Only females that were actively lactating during the 1–2-day recall period were included. A total of 128 lactating cows (45% of the 284 camels sampled) were used for ecotype comparisons of daily yield, with 32 animals in each ecotype.

Data entry was performed in Microsoft Excel, and subsequent statistical analyses were conducted in R (version 4.2.2). Descriptive statistics were calculated for all variables. Differences between means in morphometric traits across ecotypes were assessed using one-way analysis of variance (ANOVA), following inspection of residual patterns and variance homogeneity. For significant ecotype effects (P-value <0.05), Duncan’s multiple range test was employed for post hoc mean separation, and results are presented as means ± standard errors of means (SEM).

Morphometric data were stratified by sex and age class, and milk production variation among ecotypes was evaluated using descriptive statistics and ANOVA. For the multivariate morphometric analysis, to complement the univariate comparisons, exploratory multivariate analysis was conducted using the available individual-level morphometric records. The 18 quantitative traits, including estimated body weight, were standardised before analysis. Principal component analysis (PCA) was used to visualise the overall morphometric structure and ecotype overlap. To evaluate classification accuracy, a linear discriminant analysis (LDA) with stratified cross-validation was performed. Additionally, unsupervised k-means clustering (k = 4) was used to assess whether natural groupings corresponded to the designated pastoral ecotype labels. These methods were used for complementary purposes: PCA provided an unsupervised visual summary of overall morphometric structure and overlap, LDA provided a supervised assessment of how well measured traits classified known ecotypes, and k-means clustering provided an unsupervised test of whether natural groupings aligned with pastoral ecotype labels.

Qualitative assessment revealed consistent morphological distinctions among the four studied ecotypes. While all ecotypes exhibited a straight-to-convex facial profile and erect ears, they varied distinctly in ear size, coat pattern, hair type, and mammary characteristics. The Hoor ecotype was characterised by a diverse colour palette, medium-to-long hair, and relatively large udders and teats. Conversely, Aiden and Gellab individuals displayed varied coat colours alongside generally larger ears. The Sifdacar ecotype was uniquely distinguished by a highly uniform light grey to greenish-grey coat and a consistent colour pattern. These empirical observations align closely with traditional pastoral classifications, confirming that Somali camels comprise recognisable ecotypes with distinct phenotypic appearances (Table 1).

Marked morphometric divergence was observed among younger adult males (Table 2). Sifdacar males exhibited the greatest facial length (57.0 ± 2.26 cm, mean ± SED) and the highest mean height at withers (161.0 ± 7.44 cm). Aiden males exhibited the highest estimated mean body weight (430 ± 30.2 kg), whereas Gellab males displayed the largest thoracic girth (193.0 ± 11.51 cm). Significant ecotypic differences were also detected in neck length, shoulder width, anterior and posterior limb lengths, tail length, foot-pad circumference, and several hump dimensions (Table 2).

Among females within the 8–11 years age cohort, significant ecotypic variation was observed across multiple dimensions (Table 3). Aiden females recorded the longest facial length (58.25 ± 1.91 cm), the largest thoracic girth (200.38 ± 24.60 cm), and the highest estimated mean body weight (363.25 ± 51.13 kg), although not all pairwise body-weight contrasts reached statistical significance. The overall ecotype effect for height at withers was not statistically significant within this cohort (Table 3).

In the older male cohort, phenotypic differentiation among ecotypes remained pronounced (Table 4). Aiden males maintained the highest estimated mean body weight (500.88 ± 38.14 kg), whereas Sifdacar males exhibited the lowest (361.80 ± 48.24 kg). Despite their lower body mass, Sifdacar males retained distinctive skeletal frame features, including the widest interorbital distance (34.00 ± 2.26 cm) and prominent hump-related dimensions. Significant ecotype effects were observed for facial length, interorbital distance, neck length, shoulder width, height at withers, height at hump, hump length, hump circumference, posterior limb length, tail length, foot pad circumference, and body weight (Table 4). These data suggest that conformation differences persist into maturity, though age-related growth trajectories may vary by ecotype.

Older females also displayed distinct phenotypic variation across ecotypes (Table 5). Aiden females again exhibited the highest estimated mean body weight (490.42 ± 11.66 kg), whereas Sifdacar females displayed the greatest shoulder width (232.75 ± 13.80 cm) and among the longest facial measurements (57.38 ± 1.08 cm). Significant ecotypic differences were preserved for facial length, interorbital distance, shoulder width, height at hump, body length, tail length, and foot pad circumference (Table 5). Conversely, height at withers, thoracic girth, abdominal circumference, and body weight did not differ significantly among ecotypes in this mature age class, indicating a partial convergence of structural size traits in aging females (Table 5).

Reported daily milk offtake did not differ significantly among the evaluated ecotypes (P = 0.814; Table 6). The Hoor ecotype exhibited the highest numerical mean (7.70 ± 2.47 L/day), followed by Aiden (7.34 ± 2.46 L/day), Sifdacar (7.25 ± 2.29 L/day), and Gellab (7.24 ± 2.08 L/day). The ranges of observed daily milk offtake overlapped across groups, with Hoor displaying the widest distribution (5–13 L/day) (Table 7). These numerical differences should therefore be interpreted descriptively rather than as evidence of ecotype-level milk-production differences.

Exploratory principal component analysis (PCA) based on standardised morphometric traits showed partial multivariate separation among ecotypes, with substantial overlap at the individual-animal level (Figure 4). The first two principal components accounted for 31.6% of the total standardised variance, with PC1 and PC2 explaining 18.1% and 13.5%, respectively. PC1 was primarily loaded by hump circumference, height at withers, hump height, ear length, and tail length, whereas PC2 was predominantly associated with estimated body weight, thoracic girth, facial length, posterior limb length, and height at hump. Linear discriminant analysis (LDA) with five-fold cross-validation achieved 54.2% classification accuracy. In addition, k-means clustering (k = 4) showed weak alignment with the designated pastoral ecotype labels.

This study provides quantitative evidence that locally recognised Somali dromedary camel ecotypes are associated with measurable phenotypic variation under pastoral field conditions. Across both sexes and multiple age cohorts. The Hoor, Aiden, Gellab, and Sifdacar ecotypes exhibited distinct structural conformations, skeletal frames, and qualitative attributes. These findings indicate that traditional pastoral classification systems reflect genuine bio-physical variation rather than arbitrary nomenclature. This position aligns with broader camel characterisation literature, which consistently demonstrates that regional or pastoral types can be reliably differentiated using linear body measurements and external morphology (FAO, 2008; Yosef et al., 2014; Legesse et al., 2018; Tandoh et al., 2018; Ehsaninia et al., 2020; Yakubu et al., 2022; Akhmetsadykova et al., 2023).

The morphometric profiles identified herein indicate measurable differences in conformation among pastoral types. The Aiden ecotype generally exhibited the highest estimated body mass, particularly in older mature cohorts, which is consistent with pastoral descriptions of a robust, multipurpose animal. Conversely, Sifdacar camels showed a taller, more elongated skeletal frame across several comparisons (notably among younger males), while the Gellab ecotype often showed more pronounced thoracic and shoulder measurements. Although the Hoor ecotype displayed less extreme skeletal dimensions, it was qualitatively associated with larger udder and teat dimensions in this field assessment. These mammary observations should be interpreted as descriptive traits and not as evidence of intrinsic dairy superiority. The ecotypic differences in milk yield were statistically non-significant, and baseline data regarding milk yield was based on owner reports. Parallel intra-country variations have been documented in Sudanese, Ethiopian, Nigerian, and Iranian camel populations, where ecotypes vary not only in absolute size but also in structural proportions. Such morphometric variation may be associated with traits relevant to environmental stress tolerance, foraging range, thermodynamic endurance, consumer market preferences, and production orientation (Ishag et al., 2010; Ishag and Ahmed, 2011; Tandoh et al., 2018; Bekele et al., 2018; Ehsaninia et al., 2020; Alaqeely et al., 2021).

The exploratory multivariate outcomes refined the interpretation of the univariate analyses. The PCA indicated that these ecotypes occupy partially distinct regions of morphometric space, primarily segregated by axes associated with skeletal frame, hump dimensions, and body mass. However, the substantial individual-level overlap confirms that a single multivariate technique cannot justify strict breed-level categorisation. This pattern is consistent with mobile pastoral systems, where herd movement, routine sire exchange, and environmental variation may maintain phenotypic continuity while preserving recognisable local ecotypic profiles. Nonetheless, the cross-sectional field design of this study means that production environment and management confounding cannot be entirely ruled out. Apparent ecotype differences in both body measurements and reported milk offtake should therefore be interpreted cautiously, because unmeasured region, forage availability, water access, herd management, and micro-environmental conditions may have contributed to the observed patterns. Foraging regimes, mobility dynamics, disease burdens, reproductive states, seasonal body condition scores (BCS), and micro-ecological variations may influence linear dimensions and soft-tissue traits.

From a pastoral systems perspective, these morphometric variations may represent biologically and culturally meaningful diversity. Within highly erratic dryland ecosystems, herd diversification constitutes a foundational livelihood strategy; pastoralists select and value specific animals based on prioritised functions such as milk yields, mobility, drought resilience, body fat reserves, reproductive efficiency, or commercial market value. Therefore, the maintenance of distinct ecotypes may reflect a localised strategy of animal management shaped by ecological pressures, selection objectives, and cultural preferences, rather than a mere taxonomic anomaly (Bulliet, 1990; Herrero et al., 2016; Mirkena et al., 2018; Burger et al., 2019; Faye, 2020; Köhler-Rollefson, 2022). For Somalia, where severe climate volatility and rangeland degradation remain primary constraints, documenting this phenotypic diversity is important for resilience-oriented pastoralist community development and animal genetic resource (AnGR) conservation.

Despite clear morphometric divergence, reported daily milk offtake varied minimally among the four ecotypes. The mean daily offtake recorded here (7.2–7.7 L/day) aligns with performance ranges reported for dromedaries managed under extensive pastoral or semi-intensive systems across the Horn of Africa (Bekele et al., 2002; Farah et al., 2007; Abdalla et al., 2015; Gebremichael et al., 2019; Zarrin et al., 2020). However, direct cross-study comparisons should be approached cautiously, as definitions vary regarding extracted volume, saleable yield, and total biological production. Although the Hoor ecotype had the highest numerical mean, this pattern should be interpreted only as a descriptive trend that is directionally consistent with pastoral perception, not as evidence of intrinsic milk-production capacity. The wide overlap among individuals and the non-significant ecotype effect indicate that this dataset cannot establish ecotype-level dairy superiority. Given that these data relied on herder recall, were cross-sectional, and lacked statistical adjustments for parity, lactation stage, calf suckling interventions, seasonality, nutrition, watering intervals, and herd-level management, they must be treated as field estimations of reported offtake rather than definitive evaluations of intrinsic genetic potential.

These results are best interpreted as hypothesis-generating: under extensive conditions, environmental and management variation may obscure subtle ecotype-related differences in reported offtake. Dromedary milk offtake may be associated with seasonal forage availability, hydration intervals, parity, reproductive cycles, calf management, and overall herd health (Ismail, 1987; Al Haj and Al Kanhal, 2010; Nagy and Juhasz, 2016; Ali et al., 2018; Ainani et al., 2022; Nagy et al., 2022). Consequently, any apparent ecotype-related dairy tendency in this cross-sectional dataset should be interpreted cautiously, because immediate environmental factors and husbandry practices may be more influential than ecotypic classification alone. These findings suggest that any future genetic improvement initiative should be coupled with optimised animal husbandry, veterinary interventions, and longitudinal direct milk recording before drawing conclusions about ecotype-specific dairy potential.

The results address an important paradigm in contemporary camel science: distinct phenotypic differentiation does not automatically imply isolated genetic lineages or closed breed boundaries. Genomic evaluations of East and North African dromedaries demonstrate that pastoralist-defined breeds or ecotypes can be morphometrically distinct while maintaining high intra-population genetic diversity and evidence of historical admixture, a direct consequence of historical mobility, herd mixing, and collective grazing boundaries (Bahbahani et al., 2019; Legesse et al., 2018; Piro et al., 2020; Piro, 2021; Alaqeely et al., 2021; Bagiyal et al., 2024). Therefore, our data underscore that Somali camels are phenotypically differentiated pastoral types rather than strictly isolated genetic populations. Integrating baseline morphometrics with molecular markers remains the logical next step. To date, publicly available, Somali-specific genomic datasets capable of resolving these ecotypes remain unexploited, and molecular validation is still required. These findings have direct implications for national breeding and livestock conservation policies: (1) Somali camels should not be managed as a single, undifferentiated population in regional agricultural planning. (2) Ecotype-aware breeding strategies must carefully balance selection for economic traits (e.g., body mass or milk) with the preservation of adaptive fitness and genetic diversity under climate stressors.(3) Genetic improvement initiatives must remain grounded in indigenous pastoral knowledge, as local communities serve as the primary long-term custodians of these specialised genetic resources (Berghof et al., 2019; Abri and Faye, 2019; Köhler-Rollefson, 2022; Yakubu et al., 2022; Bagiyal et al., 2024). In practice, this supports community-based breeding programs (CBBPs). A systematic nation-wide screening program could identify elite sires through progeny testing, particularly by assessing their daughters’ performance, and use them to establish foundation stock for nuclear CBBPs. This will require a large-scale recording of milk production, reproduction and management parameters on participating herds. Future programs should prioritise reproductive soundness and herd-level multi-trait selection indices rather than focusing only on mature body size or presumed dairy ecotypes (Monaco, 2025).

Several limitations should be considered when interpreting these findings. First, the study was cross-sectional and field-based, with accessible herds sampled through pastoral networks rather than through a complete probabilistic regional census. Second, the study covered central and southern Somalia only; therefore, ecotypes and production environments from northern regions, including Puntland and Somaliland, were not captured. Third, exact herd-level coordinates, altitude, rainfall, temperature, vegetation composition, forage availability, water access, and management variables were not consistently available for all sampled herds. These missing spatial, environmental, and management variables limit the ability to separate ecotype effects from regional, ecological, and herd-management effects and should be considered when interpreting both body measurements and reported milk offtake.

Fourth, live body weight was estimated using a barymetric equation rather than direct weighing, and the equation has not been specifically validated for all Somali camel ecotypes. Fifth, age estimation relied on combined owner/herder recall and dentition, which may be affected by nutrition, management, and animal history. Sixth, milk data were based on short-recall owner-reported offtake from local containers rather than independently weighed total biological production. Therefore, milk values may include recall, rounding, and container-related error and should not be interpreted as evidence of intrinsic milk-production capacity.

Finally, parity, lactation stage, pregnancy status, calf-management practices, watering frequency, and detailed health information were not consistently recorded, and mammary assessments were descriptive rather than standardized dairy-conformation scores. Future studies should therefore combine longitudinal, directly measured milk-production records with standardized phenotypic recording, repeated reproductive and management data, and molecular/genomic benchmarking. Digital phenotyping tools, including structured udder and dairy-conformation scorecards, calibrated imaging, photogrammetry, and three-dimensional scanning approaches, could improve repeatability and comparability of morphometric and udder traits (ICAR, 2018; Salau et al., 2017; Strickland et al., 2025). These tools should be implemented alongside genomic characterization to identify the most informative traits, define ecotypic boundaries more rigorously, and create the evidence base for genomic selection in Somali camel breeding programs.