Evidence of subpopulation diversification and traces of introgression within Canarian camel breed zoometric standard: scope and opportunities for selection

Abstract

Extant diversity for phenotypic traits is an essential criterion to be considered when ordering priorities for conservation and improvement of animal genetic resources. Concretely, the characterisation of the distinctive body morphometry of a particular group of animals can aid in the design of selective breeding programs, given the strong correlation between body morphology and productive function. The present research aims to characterise an endangered autochthonous camel breed (Canarian camel), mainly relegated to leisure riding, for its body morphology, with a double objective: to explore the phenotypic diversity and structure of the breed for zoometric traits and assess the zoometric profile of this genetic resource that supports its differentiation from other camel breeds. Overall, the results highlight the existence of a high degree of diversity, which may be linked to genetic factors for zoometric traits in an endangered autochthonous breed with traditional in situ breeding schemes, which predicts the success of the implementation of genetic improvement schemes for such functional characteristics. This phenotypic diversity in body morphology could also be a tool for the evaluation of new functional niches within the efforts of functional valorisation of this camel breed for its sustainable conservation. Body morphology traits in the studied camel breed are significantly influenced by sex, physiological status and coat colour.

Keywords:

• Body morphology
• sexual dimorphism
• epistasis
• phenotypic diversity
• selective breeding

Introduction

Livestock biodiversity arises from the complex process of evolution of domestic animals in human-influenced scenarios with diverse raising and environmental conditions. Directional selection and the natural mechanisms responsible for the dynamics of genetic change in live populations (mutation, genetic drift, gene flow and natural selection) drive biological divergence among farm animal resources (Hoffmann 2011; Eusebi et al. 2019).

A comprehensive understanding of the genetic heritage of a particular population, breed or variety is of primary interest for owners and stakeholders. This knowledge not only sheds light on the impact of the historical background of such animal aggrupation on its current status and functionality but also evaluates possible functional avenues (Ovaska et al. 2021). Achieving this involves contrasting the studied population with similar phylogenetic neighbours (De la Barra et al. 2016) and estimating existing genetic variation to assess adaptive potential to challenging issues (Kardos et al. 2021).

Zoometric characterisation specifically explores the distinctive morphometry of a particular group of animals in reference to their functionality (Brito et al. 2021). Body morphological characteristics, annexed with formal records of functional performance, sex and phaneroptic details (i.e. coat colour and particularities), are used to calculate factor-correlated variability. This variability guides the adapted definition and application of selection criteria (Yusuff and Fayeye 2016; Toalombo Vargas et al. 2019). Additionally, when genealogical records are available, heritability components for traits of interest can be estimated (Poyato‐Bonilla et al. 2021). Zoometrics, as a method, plays a crucial role in decisions related to animal breeding for improvement, utilisation and conservation purposes (Assan 2015).

Among livestock species, camels are notably overlooked in breed characterisation based on morphometric assessment, despite the well-established correlation between body morphology and productive function (Iglesias et al. 2020). The literature on phenotypic and genotypic variability in camel species for morphometric traits is significantly less extensive compared to other livestock (Babelhadj et al. 2017; Alhaddad and Alhajeri 2019; Alhajeri et al. 2019). However, there is a positive trend in camel demographics and geographical distribution (Faye 2022), accompanied by growing socio-economic interest in their production as sustainable species (Iglesias Pastrana et al. 2020).

In this situation, studying existing morphometric and phenotypic variability in camels becomes a crucial prerequisite for assessing variations within and between populations. This is essential for conserving camel biodiversity worldwide through effective selection for specialised traits and the revalorisation of endangered resources, as demonstrated by Iglesias et al. (2020) and Köhler-Rollefson (2022). Moreover, when molecular tools are not affordable, morphological characterisation through discriminant analysis stands out as the most reliable alternative option (Ceccobelli et al. 2016).

Contextually, the Canarian camel (Camelus dromedarius), a local breed primarily found in the Canary Islands (Spain) and the only camel breed in Europe, is currently under threat. Since its arrival in the archipelago around 1405 from the nearest African coast, local dromedaries have undergone anthropogenic selection for rural labour and work-loading activities. Subsequently, these animals transitioned to leisure activities within the tourism industry, serving as the major income source for the breed and ensuring its survival to the present day (Pastrana et al. 2021). Reproductively isolated from other camel populations due to official health-based legislation, this local camel genetic heritage has diverged from that of its North African relatives, as demonstrated by Schulz et al. (2010). These authors performed the first genotyping, phenotypic and ethnographic study of this animal population, which led to the official declaration of the Canarian camel as a singular breed and declared it at risk of extinction in 2012 (Iglesias Pastrana et al. 2020). Distinctive physical features, such as greater muscle development at the fore and hind quarters and chest depth, set the Canarian camel breed apart from North African dromedaries. These characteristics were crucial in proposing the initial breed standard within the official breeding program (Schulz 2008; Fernández de Sierra and Fabelo Marrero 2017). However, there is no morphometric characterisation assessing the distribution of phenotypic variability or subdivision among different populations of the breed. Consequently, the identification of genetic groups acting as reservoirs of genetic variability remains unexplored.

The present research aims to conduct an extended morphological characterisation of Canarian dromedaries. This involves using canonical discriminant analysis to (1) explore the phenotypic diversity and structure of the breed and (2) assess the phenotypic profile of this autochthonous genetic resource, supporting its differentiation from other camel breeds. The findings, including the improved definition of the breed standard conformation traits and understanding its phenotypic variability, will contribute to the adaptation of breeding strategies with a focus on sustainable conservation and genetic improvement.

Materials and methods

Definition of zoometric parameters

In order to establish a comprehensive database of zoometric measurements for camels, we conducted an extensive literature review on the subject throughout the entire month of September 2019. This review was conducted using the Google Scholar search engine (https://scholar.google.com/), accessed on 1 September 2019. This method has been employed in previous studies for efficient data extraction (Iglesias Pastrana, Delgado Bermejo, et al. 2022). From this literature search, we identified six relevant papers related to camel zoometrics, spanning the years 1994–2019 (Iglesias et al. 2020). To enhance the dataset, we included additional variables related to camel functional development, following those outlined by Alhajeri et al. (2021). After compiling this variable list, we identified 30 zoometric measurements to be collected in the field and subsequently extracted from digital images for the same animals. Detailed results and definitions of the zoometric parameters considered in this study can be found in Iglesias Pastrana, Navas González, et al. (2022).

Animal sample

The collection of zoometric data took place between September 2019 and August 2020 and involved 130 individuals of the Canarian camel breed, consisting of 58 females and 72 males. These camels were located in three representative breeding locations within Spain: Matalascañas (Huelva, Doñana National Park, coordinates 36.972330, −6.427498; n = 14; nine males and five females), Almería (coordinates 36.902180, −2.429520; n = 16; 13 males and three females) and Fuerteventura (coordinates 28.186777, −14.158361; n = 100; 50 males and 50 females). To ensure the suitability of the animals for the study, thorough clinical examinations were conducted to assess their well-being. Furthermore, only non-pregnant female camels were included in the study to avoid potential bias in zoometric measurements in the thoracoabdominal region due to pregnancy (Yakubu et al. 2011). Both age and live weight did not follow a normal distribution (p < .05). The live weight was calculated using the formula (Equation (1)(1) $$\text{Live weight}=\text{6}.46\times \text{1}{0}^{-7}{\left(\text{HW}+\text{ChG}+\text{HG}\right)}^{3.17}$$(1) ) proposed by Boujenane (2019): (1) $$\text{Live weight}=\text{6}.46\times \text{1}{0}^{-7}{\left(\text{HW}+\text{ChG}+\text{HG}\right)}^{3.17}$$(1) where HW represents height at the withers, ChG denotes chest girth and HG signifies hump girth. Each of the 30 zoometric measurements collected from each animal was obtained from its left side, following the procedures outlined in Iglesias Pastrana et al. (2021) and Alhajeri et al. (2021). The age of female Canarian dromedary camels in the study spanned from 40 to 423 months, while male Canarian dromedary camels ranged in age from 18 to 385 months. Regarding live weight, female Canarian dromedary camels exhibited weights ranging from 327.03 to 687.13 kg, while their male counterparts weighed between 342.72 and 777.56 kg.

Sampling

To minimise potential bias attributed to hair length and texture, we selected the end of the moulting season, which spans a six-to-eight-week period starting in late spring (Babu 2015), as the sampling moment. A flowchart summarising the research methodology can be found in Iglesias Pastrana, Delgado Bermejo et al. (2022) and Iglesias Pastrana, Navas González, et al. (2022).

On-field zoometrics

For on-field measurements, animals were positioned in a static upright stance with their heads naturally raised and bodies correctly aligned (parallel fore and hind legs perpendicular to the ground, with toes in line). Measurements were taken on a flat and firm ground surface using a non-elastic measuring tape. All operators underwent training, with the first operator responsible for on-field measurements and subsequent digital image-based extraction. The second operator assisted in collecting zoometric measurements, while a third operator recorded the results and held a one-meter measuring bar for reference during digital zoometric extraction, following the procedures outlined in Iglesias Pastrana, Navas González, et al. (2022).

Digital imaging

Three photographs (front, lateral and back views perpendicular to the camera; Sony DSC-RX100 SENSOR CMOS Exmor 1.0 of 20.1 MP, F1.8–4.9, Zoom 20–100, Optical Zoom 3.6×, 3″ LCD Screen Image stabiliser, Minato City, Japan) were captured for each animal just before the on-field zoometric evaluation. The second operator captured these photographs for subsequent digital imaging analysis, while the third operator held a one-meter measuring bar along the same midline as the animal’s body to serve as a reference for distance calibration in the computer measurement software used for digital imaging zoometrics. The obtained images were processed digitally using Kinovea 0.95 (Free Software Foundation, Inc., Boston, MA). Zoometric linear measurements were recorded in pixels by drawing a straight line between two points in the image, automatically converting to centimetres following calibration using the measuring bar as a reference through the software’s Calibrate option (Iglesias et al. 2020). Puig-Diví et al. (2019) previously reported Kinovea software as a valid and reliable tool for accurate measurements at distances of up to 5 m from the object and at angles ranging from 90° to 45°.

Image collection occurred on an open, firm and flat ground surface, with carefully chosen lighting conditions to ensure the animal was not in a shaded area or exposed to light that might distort image capture. The animal’s colour was considered to prevent any distortion or misregistration of measurements due to background colour. The camera, positioned at a standardised height of 1 m on a camera stand, was 4 m away from the camel’s centre of balance. This distance and height allowed for the capture of the entire animal being evaluated. Procedures outlined in Iglesias Pastrana, Navas González, Ciani, Nogales Baena, et al. (2020) and Iglesias Pastrana, Navas González, Ciani, Barba Capote, et al. (2020) were followed to ensure the animal’s correct position, including marking standard lines on the ground before taking photographs to confirm proper alignment.

Observational sample

As suggested by Iglesias Pastrana, Delgado Bermejo, et al. (2022) and Iglesias Pastrana, Navas González, et al. (2022), digital imaging zoometry and zoometry on live animals can be reliably translated and used interchangeably or simultaneously. For this reason, we decided to use both on-live zoometric measurements and digital imaging-based zoometric measurements per animal as independent variables in the statistical analysis developed further in the present study.

Statistical analysis

Following the methodology outlined in González Ariza et al. (2021) and González Ariza et al. (2022), we initially used a discriminant canonical analysis to create a tool for evaluating the most optimal linear combinations of zoometric traits (either on-live animals or digital imaging-based), phaneroptics (eye or coat colour), sex and neutering status within and between the three locations in which Canarian dromedary camels can be mostly found in Spain. Hence, such locations Matalascañas (Huelva, Doñana National Park, coordinates 36.972330, −6.427498), Almería (coordinates 36.902180, −2.429520) and Fuerteventura (coordinates 28.186777, −14.158361) served as categorical dependent and population clustering variables in our analysis.

Our independent variables consisted of zoometric traits (either on-live animals or digital imaging-based), as previously described. Additionally, we considered factors such as phaneroptics (eye or coat colour), sex and neutering status in the analysis due to these variables being reported to condition dromedary camel zoometry in the literature (Pigière and Henrotay 2012; Babu 2015; Tandoh and Gwaza 2017; Alhajeri et al. 2021). These variables, interlinked with genetics and environmental factors, contribute significantly to the variation in zoometric measurements among dromedary camels. Researchers have unveiled intriguing associations between sex, castration status and specific zoometric parameters, highlighting the profound influence of hormonal factors on growth and development. Moreover, the variability existing for qualitative traits such as the colour of the coat and the eyes could also be a reflection of both historical and contemporary preferences in camel breeding programs and evolutionary forces, hence a further potential inference of genetic diversity in these animals. Therefore, understanding these conditioning effects not only contributes to our knowledge of camel biology but also has practical implications for camel management, selective breeding and conservation strategies (Alhaddad and Alhajeri 2019). For these reasons, we decided to consider them independent variables in our discriminant analysis.

For the classification, prediction, interpretation and manipulation of observations related to the aforementioned independent variables, we employed the Chi-squared automatic interaction detection (CHAID) decision tree method. This approach allowed us to discretely analyse the independent variables and their relationship to the three locations where Canarian dromedary camels are bred. To assess the reliability of the CHAID decision tree model, we conducted cross-validation to evaluate its predictive performance when applied to new data samples, comparing it to the training sample. This assessment helped us determine how effectively the model generalised to unseen data. We used 10-fold cross-validation to validate the CHAID decision tree and assess whether the selected predictors effectively explained differences across three locations where Canarian dromedary camels inhabit.

Results

A summary of descriptive statistics for zoometric traits in the Canarian camel breed can be found in Table 1.

Table 1. Descriptive statistics for zoometric traits in Canarian camel breed.

Zoometric trait	Minimum, cm	Maximum, cm	Mean, cm	Standard deviation, cm	Coefficient of variation, %
Head length	36.5	67	53.87	5.78	10.72
Head width	24	36.1	26.3	3.27	12.45
Ear length	4.5	14.3	10.88	1.55	14.29
Ear width	3.7	10.5	7.74	1.21	15.62
Neck length: dorsal line	158	178	122.80	16.50	13.43
Neck length: ventral line	54.6	128.7	105.83	9.89	9.34
Neck girth: cranial third	49.58	103.26	75.67	9.09	12
Neck girth: middle third	46.92	116.43	74.12	10.39	14.02
Neck girth: caudal third	86.44	152.27	119.51	13.32	11.15
Chest width	35.4	98.5	51.31	6.76	13.18
Heart girth	69.84	165.43	137.1	12.20	8.90
Height at withers	158	203.2	184.40	101.06	54.8
Body length	95	193	160.3	15.72	9.80
Hump-to-tail distance	30.9	81.5	49.91	7.37	14.77
Hump length	18.6	113.5	81.31	14.28	17.56
Hump width	12.9	173	80.09	14.02	17.5
Hump height	26.3	92.7	62.17	13.2	21.23
Hump girth	76.42	209.6	141.72	26.18	18.47
Rump length	30.2	57.4	43.07	4.90	11.37
Rump width	26.5	53.1	38.20	5.53	14.48
Tail length	27.6	65.5	52.85	7.65	14.47
Width at the base of the tail	8.1	21.3	12.57	2.1	16.71
Thigh perimeter	47.14	103.35	75.42	9.68	12.83
Hock perimeter	28.01	55.3	41.46	4.92	11.85
Fore cannon bone perimeter	8.12	23.2	17.43	3.50	20.01
Rear cannon bone perimeter	10.37	24.85	19.63	3.45	17.56
Sole length	13	24.6	18.92	2.04	10.78
Length of toe dorsal line	6	44.50	14.3	3.27	22.9
Heel height	4.56	53	7.84	4.14	52.81
Foot perimeter	25.4	54.9	41	5.1	12.45

After conducting three rounds of multicollinearity analyses (to remove any redundant factor(s) from the set of explanatory variables) (Chan et al. 2022), the variables included in the discriminant canonical analysis are those outlined in Table 2.

Table 2. Summary of the value of tolerance and VIF after multicollinearity analysis of zoometric traits (either on-live animals or digital imaging based), phaneroptics (eye colour and coat colour) sex and sexual status in Canarian camel breed.

Statistic	Tolerance (1 – R^2)	VIF (1/tolerance)
Hump width	0.8809	1.1352
Height at withers	0.8658	1.1550
Heel height	0.7222	1.3847
Length of toe dorsal line	0.6812	1.4679
Coat colour–ashed	0.6577	1.5204
Coat colour–black	0.6379	1.5676
Coat colour–white	0.5902	1.6944
Head width	0.5846	1.7105
Hump-to-tail distance	0.5690	1.7576
Eye colour–brownish with blue spots	0.5545	1.8033
Eye colour–brown	0.5342	1.8718
Chest width	0.5109	1.9572
Neck girth: caudal third	0.4778	2.0931
Neck length: dorsal line	0.4715	2.1207
Width at the base of the tail	0.4505	2.2200
Neck girth: cranial third	0.4501	2.2219
Neck girth: middle third	0.4383	2.2816
Sole length	0.4358	2.2945
Neck length: ventral line	0.4333	2.3081
Hump length	0.4313	2.3185
Rump width	0.4270	2.3421
Tail length	0.4115	2.4303
Hock perimeter	0.4064	2.4604
Hump height	0.4064	2.4608
Coat colour–blonde	0.3925	2.5475
Coat colour–bay	0.3617	2.7644
Ear width	0.3516	2.8445
Heart girth	0.3400	2.9414
Coat colour–cinnamon	0.3111	3.2145
Ear length	0.3033	3.2968
Hump girth	0.3012	3.3206
Rear cannon bone perimeter	0.2998	3.3360
Neutered–no	0.2921	3.4240
Rump length	0.2917	3.4285
Coat colour–chestnut	0.2896	3.4525
Thigh perimeter	0.2650	3.7734
Sex–male	0.2649	3.7752
Foot perimeter	0.2530	3.9518
Body length	0.2336	4.2804
Head length	0.2258	4.4287

Interpretation thumb rule: VIF ≥5 (highly correlated); 1 < VIF < 5 (moderately correlated); VIF = 1 (not correlated).

Pillai’s trace criterion reported a significant difference across locations (Pillai’s trace criterion: 1.5648, F (observed value): 19.6869, F (critical value): 1.3075, df1: 800, df2: 438, p value< 0,0001), confirming the validity of the discriminant canonical analysis (Lipovetsky 2015).

Out of the two functions identified through discriminant analysis, both were found significant for their discriminant ability (Table 3). Table 3 also presents the test functions for the present canonical discriminant analysis. These are the functions included in each test, with the null hypothesis that the canonical correlations associated with the functions are all equal to zero. In this example, we have two functions. Thus, the first test presented in this table tests both canonical correlations (‘1 through 2’), and the second test presented tests the second canonical correlation alone.

Table 3. Canonical discriminant analysis efficiency parameters to determine the significance of each canonical discriminant function.

Test functions	Canonical correlations	Eigenvalue	Bartlett’s statistic	Discrimination (%)	p Value
1 through 2	0.9377	7.2867	776.9583	76.9755	.0001
2	0.8279	2.1796	274.7274	23.0245	.0001

Among these, the F1 function exhibited the highest discriminatory power, with an eigenvalue of 7.2867, explaining 76.98% of the variance.

The several variables examined in this study were ranked based on their discriminative capacity. An evaluation of the equality of group means of the independent variables involved in the discriminant canonical analysis is presented in Table 4.

Table 4. Results for the tests of equality of group means to test for difference in the means across sample groups once redundant variables have been removed.

Variable	Wilks’ lambda	F	p Value	Rank
Hump height	0.5988	86.1083	<.0001	1
Tail length	0.6599	66.2208	<.0001	2
Hump length	0.7238	49.0383	<.0001	3
Neck length: dorsal line	0.7343	46.4963	<.0001	4
Body length	0.7852	35.1437	<.0001	5
Head length	0.8024	31.6458	<.0001	6
Thigh perimeter	0.8042	31.2925	<.0001	7
Hump girth	0.8135	29.4504	<.0001	8
Rump length	0.8154	29.0960	<.0001	9
Neck girth: cranial third	0.8282	26.6649	<.0001	10
Ear length	0.8494	22.7841	<.0001	11
Neutered–yes	0.8618	20.6069	<.0001	12
Coat colour–bay	0.8639	20.2422	<.0001	13
Heart girth	0.8671	19.7001	<.0001	14
Rump width	0.8683	19.4942	<.0001	15
Coat colour–ashed	0.8705	19.1103	<.0001	16
Head width	0.8866	16.4413	<.0001	17
Neck girth: middle third	0.8966	14.8817	<.0001	18
Coat colour–chestnut	0.9059	13.3537	<.0001	19
Chest width	0.9075	13.0949	<.0001	20
Sole length	0.9397	8.2511	.0003	21
Length of toe dorsal line	0.9398	8.2364	.0003	22
Heel height	0.9509	6.6388	.0015	23
Hock perimeter	0.9523	6.4418	.0019	24
Sex–female	0.9542	6.1680	.0024	25
Coat colour–blonde	0.9563	5.8769	.0032	26
Coat colour–black	0.9569	5.7816	.0035	27
Neck length: ventral line	0.9577	5.6693	.0039	28
Hump-to-tail distance	0.9681	4.2277	.0156	29
Hump width	0.9756	3.2129	.0419	30
Coat colour–cinnamon	0.9784	2.8337	.0606
Eye colour–brownish with blue spots	0.9856	1.8772	.1551
Eye colour–brown	0.9878	1.5811	.2077
Coat colour–roan	0.9880	1.5672	.2106
Foot perimeter	0.9897	1.3434	.2628
Neck girth: caudal third	0.9906	1.2132	.2989
Rear cannon bone perimeter	0.9926	0.9534	.3868
Width at the base of the tail	0.9960	0.5145	.5984
Height at withers	0.9990	0.1289	.8792
Ear width	0.9997	0.0409	.9600

df1 = 80; df2 = 438.

Higher values of F and lower values of Wilks’ lambda indicate greater discriminating power (El Ouardighi et al. 2007). The analysis revealed that hump height, tail length, hump length, neck length: dorsal line, body length, head length, thigh perimeter, hump girth, rump length, neck girth: cranial third, ear length, neutral status, coat colour, heart girth, rump width, head width, neck girth (middle third), chest width, sole length, length of toe dorsal line, heel height, hock perimeter, sex, neck length, ventral line, hump-to-tail distance and hump width made highly significant contributions (p < .01) to the discriminant functions when locations were the clustering criteria.

Standardised discriminant coefficients were used to assess the relative weight of each independent variable across the two significantly established discriminant functions (Supplementary Table S1).

Press’ Q value of 467.40 (n = 260; n′ = 251; K = 3) was computed for zoometric traits (either on-live animals or digital imaging-based), phaneroptics (eye or coat colour), and sex and sexual status, indicating that predictions can be considered better than chance at a 95% confidence level (Chan 2005).

The calculation of centroids for different locations was conducted. The relative positions of each centroid were determined by substituting the mean values for the observations represented in each of the two detected discriminant functions (F1 and F2).

Canonical relationships with traits were graphically represented to illustrate group distinctions on a territorial map for easier interpretation. To perform variable selection, we employed regularised forward stepwise multinomial logistic regression algorithms. In this process, we regulated the priors based on group sizes calculated using the prior probability function of commercial software (SPSS Version 26.0 for Windows, SPSS, Inc., Chicago, IL). This approach ensured that groups with varying sample sizes did not unduly impact the classification quality, thereby enhancing the reliability of our analysis (Marín Navas et al. 2021).

The utilisation of the same sample size across groups, as employed in this study, is known to provide robust results. In this context, previous research has indicated that a minimum sample size of at least 20 observations per four or five predictors is advisable. Additionally, it is recommended that the maximum number of independent variables not exceed n – 2, where n represents the sample size. This approach helps mitigate potential distortions in the analysis (Poulsen and French 2008; Marín Navas et al. 2021).

Consequently, the present study used a six or seven times higher ratio between observations and independent variables than those described above, which renders discriminant approaches efficient.

The discriminant routine of the Classify package of SPSS version 26.0 software (SPSS, Inc., Chicago, IL) and the canonical discriminant analysis routine of the Analysing Data package of XLSTAT software (Addinsoft Pearson Edition 2014, Addinsoft, Paris, France) were used to perform canonical discriminant analysis.

A territorial map depicting the Canarian dromedary camels in the canonical discriminant analysis is sorted across locations, and the results for the functions at the centroids are presented in Figure 1. This figure depicts the separation among the animals comprising the three locations considered in this study.

Figure 1. Territorial map depicting the Canarian dromedary camels in the canonical discriminant analysis sorted across locations.

The CHAID decision tree is presented in Supplementary Figure 1.

Discussion

Domestic camel populations (dromedaries or one-humped camels and Bactrian or double-humped camels) exhibit relatively low genomic diversity (structural variation) (Burger et al. 2019). Despite this, there is notable phenotypic variability attributed to adaptive plasticity in response to diverse environments, anthropogenic selection pressures and geographical mixing within these livestock species (Faye 2022). As stated by Goswami et al. (2022), diversity in morphological characteristics and the number of species are not closely related in mammals. In this context, accurately estimating levels of phenotypic diversity within and between camel populations becomes crucial for adapting control and selection strategies to prevent and minimise the detrimental loss of such variability (Singh and Verma 2018).

Breed-specific and comparative phenotypic variability for morphological traits: influencing factors

In our study focused on the morphological characterisation of the endangered Canarian camel breed, the coefficient of variation (CV) ranged between 8.90 and 54.80%, indicating substantial population heterogeneity for body morphology traits. This broad variation suggests the potential success of genetic improvement schemes for such characteristics. High phenotypic variability provides opportunities for rapid response through selection (Azimi et al. 2017). Except for Bekele et al. (2018), Chniter et al. (2013) and Bello et al. (2022), who reported CV between 3–13%, 5–11% and 7–30% for morphometric traits in camels from Ethiopia, Tunisia and Nigeria, respectively, any of the other existing morphological characterisations of camels did not specifically calculate this normalised measure of data variability. Then, scientists, stakeholders and animal breeders are ignoring a statistical figure that would allow them to make an accurate comparison between and within different camel populations for body morphology phenotypes. At a global level, this situation may be limiting the opportunities for empirical conservation and sustainable improvement of camel biodiversity.

Several factors, including methodological issues and animal-dependent features, could explain the large general CV values in our dataset. First, sampling an animal population in which a wide range of ages is represented would intrinsically influence the consistency of the data (Melesse et al. 2022). Moreover, sex and neutering status showed significant discriminating power. Particularly, females and neutered animals were those animal groups within which the differences in body morphometrics were more evident. The Canarian camel breed is indeed reported to exhibit modest sexual dimorphism for the general size of the animals and some local body regions (Fernández de Sierra and Fabelo Marrero 2017), a condition that could be affected to a greater extent in a contemporaneous scene given the progressive relegation of females from leisure tourism to milk production (Iglesias Pastrana et al. 2021). The time of castration of camels also impacts body development, as the animals neutered after reaching sexual maturity will develop a wider and more robust physico-anatomical structure and working endurance (Faye 1997; Ucko and Dimbleby 2007; Pigière and Henrotay 2012). Additionally, qualitative attributes like coat colour exhibited significant effects on body measurement traits, similarly to other livestock species (Sánchez‐Guerrero et al. 2019). In particular, the coat phenotypes ‘bay’, ‘blonde’, ‘black,’ 'ashed’ and ‘chestnut’, were the coat colour categories that presented larger differences for body morphology traits between individuals. Other qualitative attributes, such as eye colour, did not explain in a significant manner the differences in morphological characteristics between animals in the studied population. This categorical character is evidenced to impact, however, other criterion of functional interest complementary to zoometric traits in working animals’ breeding, such as leadership behaviour (Iglesias Pastrana et al. 2021).

Aggregating measured variables into qualitative categories, mean CVS were 35.78% for heights, 14.86% for widths, 13.92% for girths and perimeters and 13.70% for lengths. These findings align with the historical and genetic history of the Canarian dromedary camel. Compared with its ancestors, this dromedary is a little shorter but broader than the Western Sahara camel (Schulz 2008). The physical effort required for their working performance in rural labour since their arrival in the Canary Islands, which required them to carry out more traction than large displacements, led to the Canarian camels having developed muscles, firm limbs, and being well planted on the ground. On the contrary, the camels from the Western Sahara are characterised by a light body and long limbs, making them more suitable for long journeys and the development of higher speeds. Additionally, the importation of camels from the North of Africa to the archipelago until the official prohibition of this practice for health reasons at the end of the last century and before the institutional declaration of the local population as a singular camel breed nearly two decades later (Fernández de Sierra and Fabelo Marrero 2017) irremediably left a footprint on the contemporary genetic structure of the Canarian camel breed. Hence, the increased variation in heights and widths in our study sample could be explained by the large variability that it might display depending on the degree of kinship of each of the individuals measured with the respective founder line and inflated by the differential intraherd functional selection across generations and production environments.

Comparing our results to previous studies, both similar and contrasting results are identified. Fatih et al. (2021) have recently reported that the sex and breed fixed factors explain the differences between eight Pakistani working camel breeds, mainly for morphological widths and lengths. Further, Alhajeri et al. (2019) found a main effect of the breed on torso length and shape when exploring the morphometric variation of six common Arabian Peninsula camel breeds. Other authors encountered that wither height, thoracic and abdominal circumferences, length of the head, and length of the neck were the variables that explained in a greater proportion the differences between and within two different Algerian camel’s populations (Belkhir et al. 2013). Similar to this, Meghelli et al. (2020) found a significant effect of sex and breed, mostly for lengths and circular perimeters, in the other two different camel breeds raised in Algeria. On the contrary, Bello et al. (2022) recently encountered significant morphometric variation attributed to the oroclimatic conditions at rearing locations in four different strains of camels from north-western Nigeria, with the heights and widths having the highest CV. In the case of dromedaries from southern Tunisia, Chniter et al. (2013) observed increased CV for lengths and widths, followed by perimeters and heights. This last tendency is identical to that observed for multipurpose camels reared in Ethiopia (Bekele et al. 2018).

The main conclusion derived from the comparison between our results and those obtained in these last cited studies is that the set of morphological characteristics that allow for greater precision in explaining the differences between breeds or populations of camels will be strongly correlated with their main functionality. For example, in the case of working animals (i.e. Canarian and Pakistani camels), the height and positioning of the limbs, as well as the length and width of the head, neck, and anterior and posterior joints, are decisive for excellent functional performance and account for almost the total variation among individuals (Kefena et al. 2011; Vicente et al. 2014). On the other hand, animals destined for meat, milk and other by-products, as is the case of the camel populations evaluated in the Arabian Peninsula, Algeria, Tunisia and Ethiopia, are mainly selected for local and general lengths and depth indexes (Figueiredo Filho et al. 2016; Pourlis 2020). So, the differential inclusion of morphological traits within selection schemes, modulated by the genetic structure and history of each particular breed or population and overall expressed as different productive yields, might explain the relatively high level of heterogeneity for the respective discriminating traits.

Phenotypic diversification between and within Canarian camel subpopulations

In this breed, facing a relative functional emergency and lacking pedigree management, it could be expected to observe low phenotypic variability. Contrarily, the observed phenotypic variability is high, suggesting influences of migration, genetic drift, natural selection and evolutionary changes caused by the exploitation of this breed in different geographical environments with slight peculiarities (Iglesias Pastrana, Navas González, Ciani, Nogales Baena, et al. 2020). Cluster analysis results indicate genetic structuring among separated subpopulations.

The canonical analysis identified two discriminant functions that explained 76.98% and 23.02% of the total variance. Furthermore, the value of Wilks’ lambda in the study sample was 0.0380 (3.85%), which in turn indicates that nearly the total variability encountered in the cluster analysis (96.15%) can be attributed to differences between the subpopulations studied. In fact, the genetic homogeneity within each subpopulation is further supported by the relatively high value of the correct classification of individual camels into their respective source population (96.54%).

The few misclassifications in the discriminant analysis may indicate genetic introgression. This finding can be explained in reference to the gene flow arising from the exportation of some animals, mostly from the eastern Canary Islands, to mainland Spain in a scenario of reintroduction (Iglesias Pastrana, Navas González, Ciani, Nogales Baena, et al. 2020). In this regard, the ongoing exchange of animals between productive environments should be exhaustively controlled so that the magnitude of introgression does not further increase. Thus, the phenotypic variation rates for key morphological and ecological traits within and between subpopulations continue to be high enough to guarantee successful local adaptation to particular climatic and/or orographic conditions (Melesse et al. 2022). Marker-assisted selection, when available, constitutes the best approach for this precision breeding (Eggen 2012).

In the present study, significant differences between subpopulations were found for all morphological traits except a few (‘ear width’, ‘height at withers’, ‘width at the base of the tail’, ‘neck girth: caudal third’, 'rear cannon bone perimeter’ and ‘foot perimeter’). Notably, thigh muscle development, a key distinguishing feature between the Canarian camel and its ancestors, also varies between subpopulations of the Canarian camel breed. The animals from Almeria, followed in descending order by the animals of Fuerteventura and Matalascañas, are those Canarian dromedaries that have the greatest muscle development in this body area. This finding could be explained on the basis of the larger efforts that the animals reared in this location have to deal with higher-inclined, heterogeneous and abrupt grounds during the leisure activities they perform (Cervelló-Royo and Peiró-Signes 2015; Martínez Raya and González-Sánchez 2020). In addition, these animals are mostly imported from the island of Lanzarote, where the animals are used to working on slightly abrupt grounds (Carracedo 2014). The growth of the muscle–skeletal system is also susceptible to being affected by the protocol used for training camels for the performance of leisure work in each productive environment (i.e. time duration of training sessions, number of sessions per week, and maximum permissible load weights, among other factors) (Matsuura et al. 2013).

For the rest of the evaluated morphological characters with discriminating potential, different sets of variables broadly characterise and differentiate each subpopulation. The dromedaries from Fuerteventura have the widest and longest head, ears, rump and tail. By contrast, the camels reared in both subpopulations of mainland Spain have particular characteristics for hump dimensions. The dromedaries reared in Matalascañas have the tallest but globally smaller hump, being this last attribute indirectly inferred from their highest values for the variable ‘hump-to-tail distance’. In Almería, the dromedary camels have the longest hump and the widest anterior torso, having this attribute in close relationship with the abovementioned greater muscle development in the posterior part of the body in this subpopulation for environment-mediated functional reasons.

Globally, it is common to find animals whose height exceeds body length, as well as some quantitative differences between sexes for the proportion of certain local body regions in this camel breed. With a slightly lower height at the withers, males have a larger head, a forehead often divided into two well-defined fat pads, a thicker anterior torso, a larger overall hump size, and thicker and stronger extremities. These differences between sexes for morphological traits would be explained by both hormone-mediated effects and the sex-biased selection within this breed since males are supposed to possess more physical strength for greater working performance (Schulz 2008).

The contemporaneous efforts to promote the milk production potential of Canarian female camels (Pastrana et al. 2021) can be expected to further increase this phenotypic differentiation in the short term. When designing breeding programs, consideration of correlations between morphological characteristics and milk yield is essential to avoid undesired effects on the breed’s genetic structure and viability.

Exploration of future potential niches based on interbreed similarity

The long-term sustainable conservation of the Canarian camel breed entails immediate selective breeding for its primary function (leisure tourism) and exploring alternative niches to diversify products and services. Preliminary evaluation of niche opportunities involves comparing morphological traits with already-characterised dromedary camel populations.

Canarian camels share similarities in neck, body and hump length with pack, beauty and riding camels from Sudan (Ishag, Eisa, et al. 2011; Ishag, Reissmann, et al. 2011), Pakistan (Fatih et al. 2021) and India (Mehta et al. 2007). These resemblances relate to maintaining postural balance and propulsion during riding and loading activities (Harris 2017).

In terms of the general depth of the anterior third (neck circumference and chest circumference), similarities are observed with food production camels from Africa (Belkhir et al. 2013; Chniter et al. 2013; Meghelli et al. 2020), India (Mehta et al. 2007), Saudi Arabia (Al-Hazmi et al. 1994; Abdallah and Faye 2012), Afghanistan and Pakistan (Raziq et al. 2011). These particular body dimensions correlate with productive life and efficiency in livestock species (Sawa et al. 2013; Marković et al. 2019).

The Canarian camel also exhibits comparable morphological characteristics to other camel breeds and populations in various functional destinies and geographic regions. This suggests opportunities for participation in beauty contests, animal-assisted interventions for treating motor disabilities in humans, and food production.

Conclusions

Despite low directional selection and limited pedigree management in Canarian dromedaries, significant phenotypic diversification for zoometric traits was observed between and within subpopulations of this camel breed. Slight genetic introgression between subpopulations, likely influenced by gene flow resulting from the exportation of animals from the Canary Islands to mainland Spain, also existed. Sex and physiological status significantly impacted zoometric traits, with females and neutered animals displaying the most pronounced within-group differences in body morphometrics. Sexual dimorphism for zoometrics in this camel breed can be attributed to both hormone-mediated effects and sex-biased selection favouring greater physical strength in males. Male dromedaries exhibit a larger head size, divided forehead fat pads, a thicker anterior torso, a larger hump size, thicker and stronger extremities, and slightly lower height at the withers compared to their counterparts. Furthermore, gene epistasis can be inferred based on the significant variation in body morphology traits between different coat colour phenotypes. The ongoing exchange of animals between productive environments should be exhaustively controlled to prevent further increases in introgression. Hence, to ensure that the phenotypic variation rates for key morphological and ecological traits within and between subpopulations remain high enough to guarantee successful local adaptation to particular climatic and/or orographic conditions, Additionally, a meta-analysis calculating the CV for body morphological traits evaluated in different camel populations will provide valuable information for researchers and breeders on the magnitude of heterogeneity for such characteristics within and between camel populations. Therefore, it will highlight the opportunities for the conservation and genetic improvement of global camel biodiversity.

Author contributions

Conceptualisation, C.I.P., F.J.N.G. and J.V.D.B.; data curation, C.I.P. and F.J.N.G.; formal analysis, C.I.P., F.J.N.G. and J.V.D.B.; funding acquisition, E.C. and J.V.D.B.; investigation, C.I.P. and F.J.N.G.; methodology, C.I.P. and F.J.N.G.; project administration, E.C. and J.V.D.B.; resources, E.C. and J.V.D.B.; software, C.I.P. and F.J.N.G.; supervision, F.J.N.G., E.C. and J.V.D.B.; validation, F.J.N.G. and J.V.D.B.; visualisation, E.C.; writing–original draft, C.I.P. and F.J.N.G.; writing–review and editing, F.J.N.G., E.C. and J.V.D.B. All authors have read and agreed to the published version of the manuscript.

Ethics statement

All the farms included in this study adhered to specific codes of good practices, ensuring that the animals received humane care in accordance with the national guidelines for the care and use of laboratory and farm animals in research. Written consent from the owners of the animals was obtained for their participation in the study. The research was conducted in compliance with the Declaration of Helsinki. The Spanish Ministry of Economy and Competitiveness, through the Royal Decree Law 53/2013 and its accredited entity, the Ethics Committee of Animal Experimentation from the University of Córdoba, granted permission for the application of the protocols used in this study. These protocols were cited in the 5th section of the 2nd article of the Decree, as the animals assessed were utilised for accredited zootechnical purposes. This national Decree follows the European Union Directive 2010/63/UE, dated 22 September 2010.

Acknowledgements

The authors would also like to thank ‘Aires Africanos’ Eco-tourism Company, Oasis Park Fuerteventura and ‘Camelus’ Camellos de Almería, for their direct technical help and assistance.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data will be made available from corresponding authors upon reasonable request.

Additional information

Funding

The present research was carried out in the financing framework of the international project CA.RA.VA.N–‘Toward a Camel Transnational Value Chain’ (Reference APCIN-2016-00011-00-00) and during the covering period of a predoctoral contract (FPU Fellowship) funded by the Spanish Ministry of Science and Innovation. This work is also supported by Ministerio de Educación, Cultura y Deporte; ARIMNET2; European Union.