VARIABILITY ANALYSIS OF WILD GUATEMALAN AVOCADO GERMPLASM BASED ON AGRO-MORPHOLOGICAL TRAITS † [ANÁLISIS DE VARIABILIDAD DEL GERMOPLASMA DE AGUACATE SILVESTRE GUATEMALTECO BASADO EN CARACTERÍSTICAS AGROMORFOLÓGICAS]

descriptors proved to be useful for grouping trees according to known races and can therefore be used in the characterization of cultivars and wild trees whose ancestry is unknown.


INTRODUCTION
The avocado (Persea americana Mill.) belongs to the Lauraceae family, which includes about 50 genera and is considered one of the oldest angiosperm families (Renner, 1999(Renner, , 2004)).The avocado is a subtropical perennial species; nevertheless, it has been adapted to diverse climates (Galindo-Tovar and Arzate-Fernández, 2010).Therefore, its production has spread practically throughout the American continent, Asia, Africa continents, Middle East, and Europe (Guzmán et al., 2017).It is one of the most economically and culturally important fruit trees in Mesoamerican region its use in the region dates back more than 8,000 years, as shown by archaeological evidence in the same area (Smith, 1966), where the native people have maintained a generalized culture of knowledge, conservation, and use (Galindo-Tovar et al., 2008, 2013).
The current classification of avocado recognizes three horticultural races: Mexican (P.americana var.drymifolia), Guatemalan (P.americana var.guatemalensis) and West Indian (P.americana var.americana) (Bergh, 1992).Each of races maintains specific characteristics in their morphology, phenology, ecology and adaptation, and is part of the diversity of the genus (Storey et al., 1986).Only West Indian race is well suited to genuinely tropical climes, while the Guatemalan and Mexican accessions typically exhibit poor fruit sets (Bergh, 1992).The Guatemalan race is well adapted from about 1,000 to 2,000 m, while Mexican accessions remain in the colder avocado areas above 1,500 until 3,000 m.The three avocado races are compatible, and trees of various races can hybridize when cultivated close to one another and adapt to a wide range of environments (Storey et al., 1986).As one of the three main centers of domestication of the species, the Guatemalan avocado populations exhibit a high diversity with great potential for its use (Galindo-Tovar et al., 2013).
The avocado variability is manifested in various traits, such as the shape, color, size of fruit and leaves, flavor of fruits, and their phenology (Acosta-Díaz et al., 2013).The avocado crop is of high socio-economic importance in Guatemala's farming systems (Ledesma and Carter, 2020).At the same time, Guatemala is a country particularly susceptible to the effects of climate change (Lopez-Lalinde and Maierhofer, 2021).So it is crucial to protect and properly use the native diversity to develop new genotypes able to adapt to actual and coming adverse environmental conditions brought by climate change (Grüter et al., 2022).Nonetheless the wild Guatemalan avocado populations are increasingly displaced and threatened by various factors, such as increase in human population and consequent deforestation (Bullock et al., 2020), but mainly by the increase in cultivated areas with introduced uniform cultivars ( Rincón-Hernández et al., 2011).Thus, developing new strategies for conservation and sustainable use of Guatemalan avocado genetic resources is necessary.
A crucial stage in the description and characterization of germplasm is the evaluation of variability.The morphological markers are suitable because they provide a practical method for determining the degree of variation.These readily visible morphological traits are helpful for conservation and sustainable use strategies (Ranjitha et al., 2021).The assessment of the genetic diversity of avocados using morphological characterization has been reported around the world, such as in Mexico (Rincón-Hernández et al., 2011), Colombia (López-Galé et al., 2022), Tanzania (Juma et al., 2020), Ghana (Abraham et al., 2018), proving its validity and usefulness.In Guatemala, despite being a center of domestication, morphological characterization of populations not subjected to cultivation has not been carried out.These populations present a free genetic exchange unlike populations under cultivation that are usually inbred.Accordingly, the information limitations about Guatemala's avocado can lead to the loss of genes that can be useful for adaptability and resilience to climate change, as well as genes of interest for improving pest and disease resistance and fruit quality.
This study aims to explore and characterize the variability of wild avocado germplasm from Guatemala using 13 qualitative and 7 quantitative morphological traits.The outcomes of this investigation will support future planning for breeding strategies and germplasm management and conservation.

Study site and sampling
Based on information from the Guatemalan atlas of wild relatives of cultivated plants (Azurdia et al., 2011) and with Rafael Landívar University Herbarium staff and local people familiar with wild avocados a reconnaissance of area was carried out to identify wild individuals.A total of 189 unique avocado trees were sampled these were organized into eight geographic populations, covering three different physiographic regions.The populations were in Sacatepéquez, Chimaltenango, Sololá, Totonicapán, Quiché, Huehuetenango, Alta Verapaz, and Baja Verapaz departments; located in central, western, and northern physiographic regions (Figure 1).Table 1 displays the ecological traits of each department examined for this study.The number of individuals per population varied from 8 to 36, due to the accessibility and availability of avocado trees in each population.When an avocado tree was located, the site's latitude, longitude, and elevation were recorded.

Measurement of qualitative and quantitative morphological traits
The International Plant Genetic Resources Institute (IPGRI) field guide for avocado crops was used to characterize the tree trunk, young twigs, leaves of comparable age, flowers, ripe fruits, and seeds for each selected tree through 21 plant descriptors (IPGRI, 1995).Using the methodology described by Juma et al. (2020) each tree was examined for five twigs, three to five leaves, two to four fruits, two to three flowers, and three seeds.These descriptions make differentiating between phenotypes simple and rapid they are often quite observable, highly heritable, and equally exhibited in all environments (IPGRI, 1995).Highly discriminating descriptors available during the study period were used.The 21 descriptors registered and their possible variants are presented in Table 2.
The flesh texture was determined by tasting the ripped fruit, trunk circumference was determined with measuring tape and vernier caliper was used to record leaf length, width, and sepal length.The fruit and seed weight were measured using a portable semi-analytical balance.Bare hands were used to feel the smoothness or roughness of trunk's surface.Samples' leaf, fruit, pedicel, and seed shapes were compared to corresponding pictures in field guide.Cotyledon surface smoothness or roughness was determined by touching the seed with bare hands.Anise odor was identified by crushing and smelling the leaves.Color of a young twig, a mature leaf, petal pubescent, and mature fruit skin color were recorded based on visual observation.

Data analysis
The compareGroups package (Subirana et al., 2014) in R software v.4.2.0 (R Core Team, 2022) was employed to assess statistical significance of the factor location on measured morphological values.We used a one-way analysis of variance (ANOVA) and Tukey's tests at a significance threshold = 0.05 for each quantitative morphological value.The coefficient of variation was estimated to determine the degree of variability in each quantitative attribute across populations.GGally package (Schloerke et al., 2021) was used to estimate a correlation matrix a p-value of < 0.05 was considered statistically significant.Data on qualitative morphological attributes were treated to a cross-tabulation statistical approach to ascertain the frequency distribution of the features amongst populations.The Pearson Chi-square (χ2) test was used to ascertain if the cross-tabulation variables were related.The the ggstatsplot package (Patil, 2021) implemented in R was used to carry out the cross-tabulation and Chi-square tests.The factor analysis of mixed data (FAMD) was performed to explore the morphological relationships of the 189 avocado trees and the association between all quantitative and qualitative variables.The FAMD method may be thought of as combining principal component analysis (PCA) and multiple correspondence analysis (MCA).FAMD is dedicated to analyzing a data set containing quantitative and qualitative variables to balance their influence in the analysis (Pages, 2004).Since variables measured at different scales do not contribute equally to the FAMD analysis, the variables were standardized prior to analysis (Kenkel, 2006), thus making it possible to optimize the variance explained in each of the dimensions.FactoMineR (Lê et al., 2008) and factoextra packages (Kassambara and Mundt, 2020) was used to extract the FAMD results.
Hierarchical Clustering on Principal Components (HCPC) based on FAMD analysis was applied to identify the clustering and population structure of the sampled trees.HCPC is a robust method combining principal component methods, hierarchical clustering, and partitioning clustering, particularly the k-means method, which are the three standard methods used in multivariate data analyses (Husson et al., 2010).HCPC results was visualized through a dendrogram and factor map using the factoextra package.For the purpose of customizing and visualizing the dendrogram was exported in Newick tree format to the Interactive Tree Of Life v6 (iTOL) (Letunic and Bork, 2019).
The test values (Morineau, 1984) were computed and used to rank the most distinctive quantitative and qualitative categories within each cluster found by HCPC.

Quantitative traits
The description of quantitative traits of each of eight geographic avocado populations is displayed in Table 3.About overall tree traits among all populations, highest trunk circumference TC was 118.33 cm from Sac-Chi and Chi population exhibited lowest TC value with 82.59 cm.The coefficient of variation for TC ranged of 10.55 % (Sac-Chi) to 33.4 % (BV) and mean of 22.2 %.The data reported in this study are lower than those reported in the characterization conducted in wild avocados from Ghana, with mean TC values of 133.04 cm (Abraham et al., 2018).
For seed traits, the weight varied from 75.04 g (AV) to 96.02 g (Sac), with an overall mean of 86.56 g and the highest and lowest CV of 22.98% (Chi) to 14.26 % (Sac-Chi).The high variation in each qualitative trait was also reported in the characterization of Mexican avocado germplasm (López- Guzmán et al., 2015;Rincón-Hernández et al., 2011).With the analysis of variance, excluding leaf width and petal length, all quantitative traits exhibited statistically significant (p < 0.05) differences when compared between populations.
Based on Tukey's test, Sac and To-Qui locations had the highest average values for fruit weight and length, while the lowest values were found in AV (Table 3).The quantitative morphological traits assessed revealed significant diversity in all eight populations.There was CV with more than 20 % for 87.5 % of the descriptors taken into consideration.A larger proportion for a property may suggest greater variability (Hidalgo, 2003).
The fruit weight (FW) was the trait that varied the most between the populations.This variability might be accounted for by the different genetic basis of each population as well by the existence of the microenvironments and various agro-ecological circumstances between the populations that directly influence the fruit size and weight (Chen et al., 2007;Henao-Rojas et al., 2019;Cañas-Gutiérrez et al., 2022).

Qualitative traits
The studied trees showed three trunk surface variations, with even, rugged, and very rugged appearing in 54.5, 24.9, and 20.6 % of all samples, respectively.There was a statistically supported association between trunk surface and populations (χ 2 = 29.21,df = 14, p < 0.001) (Figure 2A).Rough and very rough trunk surfaces were predominant, which is in accordance with the study area because most of sampled trees were located in medium or highlands and just a few in lowlands.It has been reported that tree bark Guatemalan and Mexican avocado races, mainly established over 1,500 masl, is less rough.In contrast, those of West Indian race are rougher and typically well-suited to lowlands (Bergh, 1992).Regarding anise aroma in leaves, 48 % of sampled trees did not show this aroma.In comparison, in remaining 52 %, it was present, and statistical analysis did not find an association with populations (p >0.05) (Figure 2B).Anise-scented smell is linked to abundance of estragole, an organic compound detected only in Mexican avocado cultivars ( Pino et al., 2006;Pereira et al., 2013).
Concerning pubescence leaves, it showed sparse, intermediate, and dense with a distribution of 32.80, 37.56 and 29.62 %, respectively.There was not association between leaf pubescence and populations (p > 0.05) (Figure 2D).The presence of leaf pubescence can lessen the photosynthetic light plants absorb.Through growth season, pubescence rises, changing energy balance of leaf and significantly slowing down photosynthetic activity (Ehleringer et al., 1976).Nevertheless, it has been reported that leaf pubescence can increase water use efficiency by promoting condensation (Konrad et al., 2015), which is an essential trait for adaptation to drier environments, especially in areas vulnerable to drastic environmental variations due to climate change.
Nine different leaf shapes were identified in avocado characterization from Guatemala, and there was no statistical association between this trait and geographical populations (p > 0.05) (Figure 2F).The leaf form is a crucial characteristic since it can indicate how much leaf area there is and, therefore, how much light is reflected, which can directly affect plant productivity (Nkansah et al., 2013).Some of these shapes have been previously reported in avocado germplasm characterized from Colombia (López-Galé et al., 2022), Mexico (Acosta-Díaz et al., 2020), Tanzania (Juma et al., 2020) and Ghana (Abraham et al., 2018).The higher number of leaf shapes (9) described in present study is comparable with the reported in Tanzania (Juma et al., 2020).It could be due to a larger sample size because both studies were focused on description of wild populations and not only core collections.
The avocado fruit displayed three and nine pedicel and fruit shapes, respectively (Figure 3A, 3B).Avocado fruit shape is a commercially important morphological trait that appeals to market consumers.Having great diversity in fruit shape and mature skin color offers a chance to diverse interest customers.Most of shapes recorded among sampled tree were previously published by Juma et al. (2020).They also provided details on relationship between various shapes and cultivars developed from three horticultural races of avocados.The existence of these fruit morphologies in the studied trees shows presence of genetic material from all three avocado races in Guatemala.
For fruit flesh texture, classes buttery (34.86%), watery (21.16%), pastose (26.45%), and granular (17.51%) were described (Figure 3E).Since most Guatemalan avocado buyers like it for its flavor, buttery texture has significant economic importance.Many Guatemalan avocado farmers have chosen varieties with a buttery feel.Mexican and certain Guatemalan avocados are typically said to have a buttery flesh texture (Bost et al., 2013).
Similarly, some studies noted that Mexican and Guatemalan avocado cultivars had moderate to high oil contents, which may be related to buttery and pastose flesh textures (Pereira et al., 2013;Espinosa-Alonso et al., 2017).In current study, the presence of buttery and pastose flesh textures suggests the presence of avocados from Mexican and Guatemalan races.In contrast, presence of watery flesh textures indicates presence of avocados from West Indian race.Interestingly, during fruit collection, some producers indicated that some avocado trees produced fruits with buttery flesh in dry seasons and watery texture in rainy seasons.A similar phenomenon was reported by Juma et al. (2020), who suggest that environmental factors may influence changes in texture, so it is essential to consider that fruit quality may be affected in future by climate change, especially in vulnerable regions such, as Guatemala.
Avocado fruits displayed eight seed shapes.Data on distribution of seed shapes encountered are reported in Figure 4A.The Chi-squared test was not significant (p < 0.05) except for BV, which exhibited an association with oblate and cordiform seed shape.For the seed cotyledon surface smooth, rough, and intermediate texture were recorded with a distribution of 36.51,35.97 and 27.51%, respectively, among the eight populations.Statistically, a significant association was not detected between the cotyledon surface and the populations (Figure 4B).
The wide variety of seed forms identified in this study agrees with the report based on the morphological characterization of Tanzanian avocados, where about 17 different seed forms are described (Juma et al., 2020).On the other hand, the Indian and Colombian avocado characterization reports six and three different seed forms, respectively (López-Galé et al., 2022;Ranjitha et al., 2021).The wide difference between this study and the Indian and Colombian reports may be due to the larger sample size and the greater genetic diversity expressed through the various seed shapes.Popenoe (1974) linked the spheroid, obovate and oblong-conic seed shapes with Guatemalan, West Indian, and Mexican avocado races.The current study's observation of these seed morphologies suggests that Guatemala's wild avocado germplasm is derived from all three avocado races.The analyzed avocado plants exhibited a wide variety of the traits investigated.Some of these characteristics, such as fruit form, the color of the ripe fruit's skin, and flesh texture, have economic value and may be utilized by farmers as selection criteria for future production and by breeders to create superior cultivars.According to Barrett et al. (2010) the externally brilliant color, shine, and fruit form entice buyers and cause them to make impulsive purchases.When consumers eat the fruit, the texture, newness, and other flavor characteristics are pivotal to the consumers' enjoyment (Barrett et al., 2010).Consumers utilize visual cues to judge the freshness and flavor quality of food at the moment of sale, albeit these cues can occasionally be deceptive (Shewfelt, 2000;Barrett et al., 2010).

FAMD analysis
The FAMD analysis revealed that dimensions 1, 2, and 3 retained 21.1, 15.6, and 14.9 % of the total variance, respectively.At the same time, 75 % of the cumulative variance was achieved with 5 dimensions.Regarding the correlation between quantitative and qualitative variables, the fruit shape and skin color, seed shape, flesh texture and anise odor in the leaves contributed 13.85, 13.72, 13.56, 12.62 and 5.03 % to the two first dimensions (Figure 5B).These characteristics have been used to discriminate between avocado horticultural races.For example, anise odor in the leaves, fruits with rough surfaces and rough cotyledon surfaces are traits linked to Mexican, Guatemalan and West Indian races, respectively (Bergh, 1992;Janick, 2005;Popenoe, 1935).
We next used coordinates, squared cosine (cos 2 ) and contribution to assess a correlation between variables and the primary two dimensions.The angle between variable point and the axis, or cos 2 , revealed the quality of the factor map's depiction of variables.Additionally, the contribution was calculated using the separation between variable's perpendicularly projected point and associated dimension axis.According to whether a variable has a positive or negative contribution, it is placed in a different quadrant of the two dimensions.Several variables were highlighted in a redder color, such as fruit weight and seed weight in the quantitative plot (Figure S1), as well as FT.buttery, FSS.even, and SS.spheroid in qualitative plot (Figure S2).
The FAMD revealed an association between all quantitative and qualitative variables of the avocado germplasm analyzed and usefulness for profiling among sampled trees (Figure S3).In this study, the FAMD was very useful in demonstrating that the analyzed trees, which compose wild Guatemalan avocado germplasm, exhibit a mixture of different levels of each of analyzed qualitative characteristics.It may be due to the development of hybrids between three horticultural races due to absence of sterility barriers and floral biology, that favors cross-pollination (Alcaraz and Hormaza, 2011;Gross-German and Viruel, 2013).However, the FAMD demonstrated that the features that can be beneficial for distinguishing between individuals are those that most significantly contributed to dimensions 1 and 2, such as fruit shape and skin color, seed shape, flesh texture and anise odor in the leaves.
Separate explorations of connections between quantitative and qualitative factors were conducted.Consequently, the angles in correlation circle for each pair of quantitative variables represented their correlations (Figure S1).The orthogonal angle revealed a weak association, whereas the acute and obtuse angles suggested positive and negative correlations.After the correlation analysis (Table 4) the variables of fruit weight and seed weight demonstrated a positive correlation (r = .49,p < 0.0001).For conducting pre-harvest yield estimation and tracking shifts in avocado production, the relationship between fruit weight and length is ideal (Mokria et al., 2022).The remaining variables were not correlated.

Population structure
Through FAMD analysis, individual data points were visualized in the new feature space created by the first three, and thus the most informative, dimensions (Figure 6A).There was no prominent grouping of samples in the FAMD individuals plotting the avocado's qualitative and quantitative morphological attributes, which showed significant variance among the examined trees at the population level (Figure 6A).It was difficult to see the differences across the populations since there was a high phenotypic divergence within each area, revealing an absence of differentiation among populations.The considerable variety seen may have been influenced by the fact that avocado is a highly heterozygous cross-fertilizing species that produces fruits with monoembryonic seeds, so its progeny can be highly variable (Alberti et al., 2018).In this sense, it is to be expected that the evaluated populations, where repopulation processes happen naturally with little selection interference from producers, will exhibit high morphological variability, particularly in fruit and seed characteristics, which are those with the most significant natural variability (López-Galé et al., 2022).

Cluster analysis
The output of the hierarchical clustering algorithm on principal components (HCPC) resulted in a factor map and dendrogram.The HCPC suggested the presence of three clusters across 189 avocado trees.The visualization of cluster on factor map (Figure 6B) revealed that each one exhibited the presence of trees from the eight populations.Like factor map, the dendrogram from a hierarchical cluster analysis revealed three distinct groupings, each including samples from different populations, suggesting a weak population's genetic structure.The largest group was depicted by green branches, followed by blue branches, and finally, the smallest group was painted red, all containing samples from all populations (Figure 6C).
The clustering of avocado populations from various geographical locations points to a genetic mixing that may have resulted via seed exchanges between these places.Farmers in various locations frequently trade seeds with friends and family.Additionally, avocado fruits from one population can be swapped for fruits from another, which may then be used to sow the seeds.In addition, the avocado's religious, mythical, economic, and therapeutic significance in Guatemala-one of the three domestication centers- (Galindo-Tovar et al., 2008;Landon, 2009) could explain the genetic mingling seen among populations of avocados.This condition still exists now and has taken the lead role in influencing genetic diversity and population structure.
From HCPC analysis all quantitative characters had statistically significant links with the clusters' partition (p < 0.0001) with the largest explained variance for FW, SW and FL (FW: eta 2 = 0.16, p < 0.0001; SW: eta 2 = 0.14, p < 0.0001; TC: eta 2 = 0.11, p < 0.001).Chi-square test for qualitative characters exhibited statistically significant links with the clusters (p < 0.0001).In first cluster 90 and 61.19 % of individuals showed a rugged fruit skin surface and smooth cotyledon surface, respectively.The presence of anise odor in the leaves was not highly represented in this cluster (v.test -3.51).The lowest SW (v.test = -3.30)was found in this cluster (Table 5).
12   13.51 -3.21 > 0.001 TC: trunk circumference, FW: fruit weight, FL: fruit length, LL: leaf length, SW: seed weight.The sign of the v.test indicates if the mean of the cluster is under or over-expressed for the category.FSS: fruit skin surface, MFSC: mature fruit skin color, CS: cotyledon surface, FS: fruit shape, AS: anise smell, CYT: color young twig, PP: petal pubescent, PS: pedicel shape, FT: flesh texture, SS: seed shape, LS: leaf shape, TS: trunk surface, Cla/Mod: proportion (expressed as percentages) of individuals with specific qualitative character category in the cluster; Mod/Cla: proportion (expressed as percentages) of individuals within the cluster with the specific qualitative character category.
These traits fit with description of Guatemalan race, which is characterized by absence of an anise-like fragrance in the leaves, fruits with rough surfaces usually, and seed that is generally not large in proportion to the fruit size (Bergh, 1992;Janick, 2005;Popenoe, 1935).This similarity of morphological characteristics suggests that the genetic pool of the first cluster comes from the Guatemalan race.
In the second cluster 91.67, 68.97 and 60.17 % of the individuals exhibited buttery flesh texture, anise aroma in the leaves and even fruit skin surface, respectively.Agreeing, the individuals showed the lowest FW (v.test -2.98) and significantly higher SW (v.test = 2.58) (Table 5).The qualitative and quantitative characteristics of second cluster are in accordance with the botanical description of the Mexican race, which is characterized by fruits with high oil content, giving a buttery texture, smooth surface and usually under one-half pound in weight, as well as the typical anise-like odor in leaves (Ayala-Silva and Ledesma, 2014;Bergh, 1992).This finding hints that the genetic basis of the individuals in the second group is largely derived from the Mexican race.
In third cluster 77.92, 70.59 and 62.54 % of individuals showed rounded fruit shape, scares petal pubescent and rough cotyledon surface, respectively.The highest values of FW and FL were found in this cluster (Table 5).Both qualitative and quantitative characters seem to be quite close to the description of the West Indian race, which is recognized because the leaves have no anise scent, fruit and seed size ranges from large to very large, the fruits are usually pear-shaped or round and the cotyledons are often rough on the surface (Ayala-Silva and Ledesma, 2014;Popenoe, 1935).The similarities in morphological characteristics point to the third cluster exhibiting a genetic basis from the West Indian race.
The data supplied can also be utilized to identify which individuals own genotypes most similar to avocado races mentioned.This information can be used to create germplasm collection to maintain the natural genetic variability linked to each Guatemalan avocado race, which may then be exploited in further breeding programs.
As demonstrated, the genetic basis of each cluster can be linked to one of the avocado races.However, the clusters also showed traits that are not typical to the related race, which points to a high degree of hybridization, probably due to the lack of sterility barriers between the avocado horticultural races (Alcaraz and Hormaza, 2011;Gross-German and Viruel, 2013).Thus, inter-breeding is possible, resulting in highly diverse genotypes adapted to the country's climatic conditions.The high genetic diversity was also reported in Guatemala using AFLP and SSR molecular markers.Besides, the lack of population structure found in this study is similar to the one revealed by the molecular markers (Ruiz-Chután et al., 2020;Ruiz-Chután et al., 2022).

CONCLUSION
This study represents the first characterization of wild Guatemalan avocado trees based on morphological markers.The FAMD revealed that the fruit and seed weight, seed and fruit shapes, and fruit length are ideal traits for distinguishing between individual trees and clusters.In addition, the clustering methods revealed that the 189 avocado trees could be regrouped into three main clusters.The morphological descriptors proved to be useful for grouping trees according to known races and can therefore be used in the characterization of cultivars and wild trees whose ancestry is unknown.

Figure S3
. Association between all quantitative and qualitative variables from the FAMD analysis.

Figure 1 .
Figure 1.Map of Guatemala, displaying the geographical location of sampled avocado populations.Source: own elaboration.

Figure 2 .
Figure 2. Description of frequency and chi-squared test of the qualitative traits of the overall tree.A panel: trunk surface.B panel: leaf anise smell.C panel: color of the mature leaf.D panel: pubescence of leaf.E panel: color of the young twig.F panel: leaf shape

Figure 3 .
Figure 3. Description of frequency and chi-squared test of the fruit's qualitative traits.A panel: pedicel shape.B panel: fruit shape.C panel: mature fruit skin color.D panel: fruit skin surface.E panel: flesh texture.

Figure 4 .
Figure 4. Description of frequency and chi-squared test of the avocado seed qualitative traits.A panel: seed shape.B panel: cotyledon surface.

Figure 5 .
Figure 5. Scree plot showing the percentage of variance explained by each of the first 8 dimensions (A).Contribution of avocado variables to Dimension 1 and 2 (B).The red dashed line on the graph above indicates the expected average value, if the contributions were uniform.

Figure 6 .
Figure 6.Avocado individuals' visualization-based A) the first three dimensions based on FAMD, B) a factor map constructed from the HCPC algorithm and C) a dendrogram from the hierarchical cluster analysis.In figure C, the same font color is used to indicate samples from the same population.

Table 1 . Ecological characteristics of the studied locations.
N: number of sampled trees, maT: mean annual temperature, arR: annual rainfall range, ar masl: altitudinal range in meters above sea level.

Table 3 . Description of quantitative traits among geographical populations. Mean values, standard deviation (SD) and coefficient of variation (CV) complemented by Tukey post-hoc test.
L: location, A: altitude in meters above sea level, E: estimator, N: number of sampled trees, TC: trunk circumference, LL: leaf length, LW: leaf width, SL: sepal length, FW: fruit weight, FL: fruit length, PL: pedicel length, SW: seed weight.Different letters indicate significant differences (p< 0.05)