1. Introduction
While grapevines feature in Iberian Peninsula prehistory records [
1,
2,
3], initially as wild, Roman era viticulture, documented by Columella in 1548 [
4], grapevine cultivation expanded significantly, emphasizing wine, raisins, and must. Medieval al-Andalus witnessed diverse viticulture with eastern table grapes and “sultana” raisins. Agricultural treatises spanning the tenth to fourteenth centuries CE, including
Kitāb fi Tartīb awqāt al-girāsa wa-l-magrūsāt and
Al-Mugni’ fi l-filāḥa, describe grapevine cultivation techniques and varieties [
5,
6,
7,
8]. Varietal differentiation relied on traits like berry size, shape, color, flowering and ripening periods, cluster type, and sugar content [
7,
8]. Notably,
Almurjardal and
Melar grapes were prized for their flavor and raisin suitability, while the
Jalladi variety was noted for its distinct traits [
7,
8]. Abū l-Jayr transferred rare bush species, and Ibn Baṣṣāl suggested grapevine transportation using seeds, hinting at an interaction in al-Andalus with native and exotic varieties, possibly resulting in hybrids in the post-Muslim conquest era.
The grapevine (
Vitis vinifera L., Vitaceae), represents one of the oldest cultivated fruit plants [
9]. The genus
Vitis encompasses over 70 species, with
Vitis vinifera L. boasting approximately 10,000 cultivars. Historically, some of these cultivars originated from Central Asia and the South Caucasus, spreading across western Europe and the Mediterranean basin [
10].
The South Caucasus region is regarded as the birthplace of viticulture and winemaking [
11,
12,
13,
14,
15]. Archaeological discoveries at sites within the Shulaveri group, such as Dangreuli Gora, Imiri Gora, and Gadachrili Gora, reveal a sophisticated level of grape cultivation and wine production in southern Georgia between the sixth and fourth millennia B.C. [
14]. Furthermore, McGovern et al. [
15] identified Georgia, located in the South Caucasus, as the probable point of origin for European winemaking.
Four theories on grapevine domestication exist. The monophyletic monospecific theory posits that cultivated grapevines (
V. vinifera) originate from local wild populations (
V. sylvestris) [
16]. The monophyletic bispecific theory suggests that cultivated populations stem from a common ancestor, possibly extinct, shared with the wild species, implying present distinct species [
17]. The polyphyletic multispecific theory proposes that regional cultivated grapevine populations derive from independent wild ancestors, each with its own group of wild relatives [
18,
19]. The hybrid theory [
20] suggests cultivated grapevines originated from hybridization between wild species, particularly between dioecious European grapevines and hermaphrodites from Central Asia.
Wild grapevines (
Vitis sylvestris CC Gmelin) and cultivated grapevines (
V. vinifera L.) primarily differ in regard to their reproductive biology, with wild grapevines being dioecious and most cultivated grapevines being hermaphrodites [
11]. Other distinctions arise from the domestication process, including differences in sugar content, berry and bunch dimensions, and blooming and ripening periods [
21].
Vitis vinifera cultivars are grouped into three main types: wine grapes, table grapes, and dried fruits (raisins and sultanas) [
22,
23]. This classification aligns broadly with biogeographical groups of grape cultivars and exhibits a distinct allelic composition and frequencies [
24,
25,
26,
27].
Andrasovzsky [
18] aimed to geographically classify grapevine varieties into five eco-geographic species:
Vitis mediterranea (Mediterranean and Danube),
V. byzantina (Caspian region),
V. alemannica (Rhine Valley),
V. deliciosa (Caucasus), and
V. antiquorum (Central Asia). Negrul [
24] introduced the term “Proles” to categorize major clusters of cultivated varieties, later refined by Gramotenko [
28], Troshin et al. [
26], and Troshin [
29]. Each proles has characteristic distribution and morphology:
Proles Pontica Negrul (V. mediterranea András) spans the Danube Valley, Eastern Mediterranean, east of the Iberian Peninsula, western Caucasus, and Egypt, consisting of wine cultivars with dense clusters and small to medium berries, and comprises around 4000 to 5000 cultivars.
Proles Orientalis Negrul (Caspian region, Eastern Caucasus, Central Asia, western Mediterranean, Romania) encompasses table grapes and some ancient wine cultivars with thick, elongated berries. Subproles Caspica Negrul (V. byzantina András) includes around 300 cultivars, and Subproles Antasiatica Negrul (V. antiquorum András) about 1000.
Proles Occidentalis Negrul (V. alemannica András pp) covers the Rhine Valley, Italy, France, and part of the Iberian Peninsula, consisting of wine cultivars with small leaves and compact clusters, including between 1000 to 2000 cultivars.
Molecular evidence supports an east–west gradient. The chlorotype variation among
V. sylvestris and
V. vinifera suggests two significant origins for the cultivated germplasm: one in the Near East and another in the western Mediterranean. Approximately 70% of Iberian Peninsula cultivars derive from western
V. sylvestris populations [
21]. Among the six haplotypes, half of western European varieties, especially Spanish accessions, share haplotype VI absent in Caucasian and Middle Eastern cultivars. Haplotype I prevalence declines from over 80% in South Caucasus and the Near East to under 20% in Spain, with intermediate values in central Europe [
30]. Molecular analyses by Mercati et al. [
31], confirmed significant east–west gene flow and identified five ancestral populations. The Portuguese germplasm is genetically closer to French varieties than those from Spain. The Maghreb region shows a connection to Spain, except for Tunisia, which ties to the Italian and Tyrrhenian ancestral group.
OIV [
32,
33] and IPGRI/UPOV/OIV [
34] identified 128 morphological, physiological, and agronomic traits to describe grapevine diversity. Ampelographers prioritize lesser-known characters, which are crucial for defining groups, despite being less relevant to viticulturists. These characters remain integral to understanding the origins and characteristics of each group.
Vitis vinifera seeds exhibit a characteristic globose to obovoid shape, yet they display significant polymorphism and remain identifiable even in carbonized, mineralized, or fragmented states. Factors such as grape size, ripening degree, and seed quantity within each grape, contribute to the seed shape variability [
35,
36]. Ampelographic textbooks typically focus on a few seed characteristics for descriptive purposes. These include seed length and weight, which are categorized into distinct classes based on observations on dry seeds from well-developed bunches [
32,
33]. Additionally, the detection of transversal ridges distinguishes between
Vitis vinifera and related species of section
Vitis and those of section
Muscadinia [
32,
33].
Biometric indices have proven to be valuable in the taxonomic study of
Vitis since Stummer’s pioneering work in 1911 [
37]. Discrimination between wild and domesticated grapevine seeds based on morphometrics has been subject to ongoing review in ampelographic and archaeobotanical literature [
38,
39,
40,
41]. Furthermore, recent advancements include image analysis of the seed shape [
42,
43,
44,
45]. The application of distance-based trees for allocating archaeological grape seed samples is well established [
45]. Similarly, Rivera et al. [
46] tentatively assigned archaeological
Phoenix seed samples using a method based on Ward’s minimum variance algorithm.
The main aim of this study is to analyze medieval archaeological grapevine seeds morphometrically, comparing them with samples from other periods and with modern wild and cultivated grapevine populations. Our goal is to determine the identity and level of domestication of these medieval vines from al-Andalus, as well as their relationship with the broad groups of cultivated vines categorized by Negrul as proles. Additionally, we aim to distinguish raisin seed types from other grape types, based on morphology.
2. Materials and Methods
In this paper, we analyze nineteen grapevine seed samples from nine sites in Spain located in the southern Iberian Peninsula (
Figure 1), which have been previously studied by our group [
47,
48,
49,
50,
51,
52,
53,
54,
55,
56,
57]. We compare these samples with a diverse range of published archaeological seeds from the region [
58,
59,
60,
61], as well as with local modern cultivars.
2.1. Description of Sites Samples
The village of Huerta del Inglés (Badajoz) was excavated between 2022 and 2023, due to the installation of a photovoltaic solar plant. This revealed an Andalusi village active during the ninth to the early tenth century CE, possibly linked to a Visigoth farm from previous centuries at the same site. Around sixteen houses and more than eighty negative structures where grapevine seeds were found were identified, indicating a larger original settlement, inferred from associated
maqbara findings [
47]. These monocellular houses exploited the agricultural resources of the area, ideal for livestock and rainfed crops, possibly supplemented by a nearby spring for irrigation.
In 2021, as part of the project “Valorization, Conservation, and Maintenance of the Archaeological Sites of the Alcazaba de Badajoz”, an excavation was conducted in a section located at the northern end of the citadel. This area expanded towards the Guadiana River during the second half of the twelfth century under the Almohads, with no known previous occupation. The excavation revealed urban structures consisting of houses arranged around a central courtyard, along with other simple spaces, such as single or double-room structures with direct street access, likely serving as residences, artisan workshops, or stores. It is known that a blacksmith’s shop was located nearby. Following the conquest of the city in 1230 by Alfonso IX de León, this area was abandoned [
48].
The site of La Graja, located north of Higueruela, in the Cañada de Pajares valley, is identified as an Andalusi alquería spanning approximately 6.8 ha. Archaeological interventions, from 2020 to 2023, aimed to delve into the study of 11th century Islamic societies in Albacete province. These interventions revealed 46 structures on the site, with thirty being houses, consisting mainly of oblong bays where grapevine seeds were found, surrounding a courtyard corral. Additionally, there are 8 small buildings with simple rectangular floor plans, situated in peripheral or central isolated areas, without other associated constructions [
49].
The archaeological campaigns at the fortress of Isso (Hellín, Albacete, Spain) spanned from 2019 to 2022, with the aim of identifying its perimeter, much of which was lost due to the construction of a neighborhood of houses. The objectives also included dating the Andalusi fortress and understanding its various transformations. Through this work, it was revealed that the two large towers visible over the centuries represent only a fraction of a much larger quadrangular fortress, approximately 44 m on each side. Further analysis revealed that the wall and towers underwent at least three distinct phases of construction. Notably, significant chronological data were uncovered within the walls, where excavations reached unmodified rock, providing an excellent stratigraphic record. Analysis of the ceramic materials suggests that the medieval structures cannot predate the second half of the 12th century. If this hypothesis holds true, it implies state or king involvement in the construction, potentially serving military or storage purposes. Supporting evidence includes the discovery of storage structures in the northeastern corner, where grapevine seeds were found [
50].
“Alcázar Menor” of Murcia: written sources and archaeological findings indicate the existence of a significant estate in the northern suburb of Murcia, situated approximately 100 m from the medina walls. This estate served dual purposes as a palatial residence and productive orchard, boasting residential, ceremonial, religious, and recreational structures, including baths, alongside extensive orchards, and gardens where grapevine seeds were found. Archaeological investigations at the site have revealed two superimposed palaces. The older palace, attributed to Amir
Abû ‘Abd Allâh Muhammad ibn Sa’d ibn Mardanîsh (1147–1172), is situated in the basement. It features a large transept garden, consisting of two platforms with a central gutter, crowned by a pavilion (
qubba). The more recent palace, commonly associated with
Ibn Hud al-Mutawakkil (1228–1238), was constructed atop the ruins of its predecessor. Portions of its remains still stand today, repurposed as walls within the present-day monastery [
51].
Archaeological excavations conducted on the roof of the central nave of the 15th century church of Santa María in Alicante yielded numerous ceramic vessels, some containing remnants of plants, including grapevine seeds. The recovered remains reflect a diverse assortment of fruits, particularly dried fruits likely consumed by the workers. The presence of these fruits in the vessels is highly probable, given the distinctive distribution of the samples, all of which consist of monospecific samples [
52].
The ancient settlement of Begastri, located at Cabezo de Roenas near Cehegín in Murcia, spans various historical periods. Initially an Iberian settlement, it later evolved into a Roman municipality and a Visigothic episcopal city. Minor Islamic occupation occurred during the Emirate/Caliphate period. The Roman–Visigoth phase saw the construction of fortified walls around the acropolis using large ashlars and masonry, incorporating ornamental materials like friezes and column bases. The city’s defensive perimeter stretched over 272 m, featuring quadrangular towers and a gate defended by towers and a barbican. Recent excavations in the western sector, where grapevine seeds were found, uncovered artifacts like belt buckles, glass fragments, and the tremis of Recaredo, suggesting an administrative or manorial area within the Visigothic city, affiliated with the kingdom of Toledo, and neighboring the Byzantine domain [
53].
The Roman villa of Los Villaricos, situated in Arreaque, approximately 5 km east of Mula (Murcia), enjoys an ideal environment for agricultural activities. Its proximity to the river Mula and its strategic location along the Carthago Nova–Complutum road axis, via Yéchar towards Archena, highlights its significance. Since 1985, archaeological campaigns have revealed Los Villaricos to be one of the most remarkable Roman villa sites in the Iberian Peninsula. The excavations have unveiled distinct areas, including rustic sections dedicated to agricultural work and storage, notably for olive oil and wine production, evidenced by the discovery of two
torcularia (winepresses). Additionally, the urban area comprises a residential zone with a central courtyard and domestic spaces for the owner’s family, along with a
balneum area featuring baths with hot, warm, and cold water facilities. Carbon-14 analysis of the olive seed samples from the
torcularium places the villa’s peak period of expansion and operation during the 4th century CE. This era coincides with the height of the nearby civitas on La Almagra hill, ancient Mula. The final phase of the villa’s occupation is marked by documented burials, indicating structural abandonment, reuse, and restructuring between the late 5th and early 7th centuries CE [
54,
55].
Located in the Los Baños district of Fortuna, the Roman health resort of Fortuna sits amidst an area known for its hyperthermal waters, shaping infrastructures dedicated to thermal activities across different epochs. The Roman era stands out as the pivotal period, witnessing the inception of the first spa facilities that influenced subsequent medieval and modern phases. The earliest constructions, dating back to the turn of the era, persisted through various reforms, including structural alterations, until the late 4th century CE [
56]. In addition to the principal edifice housing the spring and pool, several ancillary structures complement the complex. These include an outdoor pond receiving surplus water from the spring, a pool, an outdoor
apodyterium adjacent to one in the complex entrance, and a series of masonry rooms situated approximately 80 m away. The grapevine seeds were found in the rooms, which based on their layout and the recovered materials, likely served as essential hospitality facilities within the spa premises [
57].
The territories from which the medieval samples analyzed originate, currently host notable viticultural activity, covered by the following five denominations of origin [
62]:
Extremadura, known for white varieties such as “Alarije”, “Borba”, and “Jaén Blanco”, and red varieties like “Bobal” and “Garnacha Tinta”, grown on siliceous or chalky substrates, with precipitation between 400 and 600 mm, and average temperatures around 16.5 degrees Celsius (sites: Alcazaba de Badajoz and Huerta del Inglés);
Jumilla, cultivating varieties such as “Monastrell”, “Garnacha Tintorera”, and others for reds, and “Airén” for whites, on calcareous substrates at altitudes ranging from 300 to 900 m, with precipitation around 300 mm (site: Fortaleza de Isso);
Alicante, where “Monastrell” predominates among the red varieties, followed by “Garnacha Tintorera”, while “Moscatel de Alejandría” is characteristic among the white varieties. The influence of maritime air, calcareous substrate, and precipitation between 300 and 500 mm are notable (site: Santa María de Alicante);
Bullas, cultivating “Monastrell”, “Garnacha Tinta”, and others for reds, and “Macabeo” for whites, on calcareous substrates at altitudes ranging from 500 to 1500 m, with precipitation around 365 mm, and average temperatures between 14.5 and 16 degrees Celsius (site: Begastri);
Almansa, known for “Garnacha Tintorera”, “Monastrell”, and other red varieties, and “Macabeo” or “Moscatel de Grano Menudo” for whites, grown on calcareous substrates at altitudes between 400 and 700 m, with precipitation around 300 mm, and average temperatures between 16 and 17 degrees Celsius (site: La Graja).
In the area near the city of Murcia, table grapes from cultivars “Aledo”, “Dominga”, and others are cultivated outside of the denominations of origin, in orchard areas irrigated with water from the Segura River or wells (sites: Los Villaricos, Alcazar Menor de Murcia, and Baños de Fortuna).
2.2. Sampling Methods and Criteria for Selection of the Materials
Archaeological samples were obtained during excavation, following specific methodologies outlined in the literature cited. A comprehensive flotation approach was used, and the seeds were retrieved using meshes with a minimum diameter of 0.5 mm. Individual seed isolation from the assemblage was conducted using a binocular microscope, at magnifications ranging from 10 to 40×, and stored in Eppendorf polymer tubes of various sizes.
The primary dataset comprises 4029 rows of analyzed single seeds from 783 seed samples (individuals), with 32 columns of observations. These columns include 13 quantitative variables, 12 allometric indices, and 3 qualitative variables (divided into 11 categories). Additional columns contain information on sample provenance, taxonomic attribution, and other relevant data. To understand the Vitis species and cultivar diversity comprehensively, modern reference materials were utilized, providing a broad representation of cultivars, wild, and feral grapevines. Of the analyzed seeds, 3483 came from modern sources (481 samples), 399 from archaeological contexts (195 samples), primarily preserved in carbonized form, and 147 were fossilized (107 samples).
2.3. Morphometric Studies: Description of Measurements Taken, and Methods Used
An attempt was made to work with samples of ten seeds each for modern varieties and populations of wild vines, which were randomly selected for measurement from larger samples preserved in the seed collection at the University of Murcia. However, when including fossil samples and archaeological materials in the study, a significant number of the samples measured and analyzed consisted of fewer seeds or even just a single seed. Precisely for this reason, we preferred methods that allowed us to work with individual seeds, although the multivariate analysis and cluster creation were based on the parameters of each sample as a whole and not on the isolated values of each seed within it. On average, we worked with 5 seeds per sample, hence the total number of seeds and samples mentioned above. Medieval archaeological seeds underwent individual analysis based on 14 characteristics: 11 quantitative (notably total length, maximum breadth, and seed thickness, along with another 8 characteristics specified in
Supplementary Table S1) and 3 qualitative (contour type, arrangement of fossettes, presence/absence of radial furrows) (
Supplementary Table S2). Allometric variables (
Supplementary Table S3) were reserved for interpreting results [
36].
Digital scaled images were used for the analysis. Up to ten seeds from each sample were placed on a plasticine support with an integrated scale and photographed from dorsal, ventral, and lateral perspectives, using a Samsung A40 camera. Fiji 2.9.0 (14 September 2022) software was employed for image analysis, under consistent zoom conditions. Additionally, the scale images of fossilized and archaeological seeds from specialized literature aided the measurements. The data were recorded in an Excel spreadsheet, and the allometric relationships were automatically calculated.
The SEM analyses were conducted in the University of Murcia’s Scientific and Technical Research Area. Specimens were mounted on aluminum stubs, coated with a 5.0 nm thin layer of platinum using a Leica EM ACE 600, and examined using an FE-SEM device (ApreoS Lovac IML Thermofisher, Waltham, MA, USA), with a selected voltage of 10 kV and a current of 0.20 nA for imaging.
2.4. Domestication Indices: Explanation of Indices Used and Their Significance
2.4.1. Primary Indices
Stummer [
37] proposed an index based on the allometric relationship between the seed width and length, effectively distinguishing extreme forms, with intermediate values common in both wild and cultivated populations (
Supplementary Table S4). Stummer’s index values ranging from 0.44 to 0.53 are typically associated with cultivars, whereas values between 0.76 and 0.83 are specific to Austrian wild vines. However, values falling between 0.53 and 0.76 are observed in both cultivars and wild vines. Levadoux [
38] demonstrated limited validity in regard to this index for distinguishing between wild and cultivated vines.
Facsar, Terpó, Facsar and Jerem, and Perret [
39,
63,
64,
65] introduced a novel index based on the allometric relationship between the length of the beak or column and the total length of the seed, providing effective differentiation between wild and cultivated populations, typically with the boundary between 18 and 19 (
Supplementary Table S5).
Formulas devised by Mangafa and Kotsakis in 1996 [
41] found successful application for local Greek samples, encompassing both modern seeds and archaeological remnants. These formulas (
Supplementary Table S6) rely on relationships and constants, involving variables like seed length (
L), stalk length (
LS), and chalaza position (
PCH).
2.4.2. Derived Indices from Pre-Established Thresholds: Domestication and Wildness-Derived Indices
As the aforementioned indices aim to achieve the same objective of distinguishing between wild and domesticated forms, albeit yielding varying results, their collective utilization may enhance the ability to discriminate between seeds from wild and cultivated grapevines. The combined domestication/wild index is computed for each seed individually using the following formula (1), wherein
NIT denotes indices surpassing, either above or below, the threshold value, and
NI denotes the indices considered [Equation (1)]:
The threshold values for recognizing a seed as wild: Stummer > 75, Perret < 19, Mangafa and Kotsakis F1 < −0.2, Mangafa and Kotsakis F2 < −0.2, Mangafa and Kotsakis F3 < 0, and Mangafa and Kotsakis F4 < −0.9.
The wildness index (
Wi) ranges from 0 to 1, with intermediary values of 0.17, 0.33, 0.5, 0.67, and 0.83. Seeds with wildness index values between (0.67) 0.83 and 1 are unequivocally phenotypically wild, while those between 0 and 0.17 (0.33) deemed domesticated. In this study, a value of 0.5 is considered merely intermediate, according to Rivera et al. [
66], which suggests the presence of a hybrid swarm and introgression phenomena. Notably, the sum of the wildness index and its complement, the domestication index, always equals one.
2.5. Multivariate Analysis: Details of the Methods Employed
2.5.1. Variables
The data matrix comprises 783 samples and 231 columns or variables resulting from the analysis of each sample.
The quantitative and allometric variables are divided into intervals, and for each interval or category, the percentage of seeds from the sample that fall within it is calculated. For each sample, the ordered set of the 231 categories and their relative frequencies based on the observed values in the seeds of that sample constitute a descriptive spectrum, such that two exactly identical samples would present in the same spectrum (probability distributions), with identical frequencies and, therefore, the dissimilarity indices would have a value of zero. The spectrum of variables is structured as follows.
Quantitative continuous characteristics transformed into the following 137 categorical variables:
Length (25 intervals or categories), width (21), thickness (9);
Volume (12);
Beak length (dorsal 9, ventral 9), width (at base 11, at junction 11), thickness (6);
Chalaza length (18) and width (6).
Allometric continuous characteristics transformed into the following 83 categorical variables:
Width/length (29), width/thickness (10);
Beak length/seed length (16), beak width/length (9);
Chalaza width/thickness (9), chalaza apex to seed apex distance (10);
Qualitative categorical characteristics, accounting for 11 variables, as follows:
2.5.2. Data Analyses
This method was previously utilized for date palm seed classification by Rivera et al. [
46]. The chi-square dissimilarity index was calculated following the methodologies outlined by Perrier et al. (2003) [
67] and Perrier and Jacquemoud-Collet [
68]. This index assesses the contribution of a value,
xik, to the overall sum,
xi, across all the variables, serving as a comparison of unit profiles [Equation (2)].
for
j ≠ i.
Where dij is the dissimilarity between units i and j; i, j = 1, 2, …, N (samples, rows), N = 783; k = 1, 2, …, K (variables, columns).
Where dij = 1 means varieties i, j differ in regard to all variables, and dij = 0 means varieties i, j are identical.
Pairwise dissimilarities are mapped onto a multidimensional space for analysis. To achieve a meaningful two-dimensional visualization of these relationships, cluster analysis is employed. Cluster analysis refers to a collection of numerical techniques aimed at categorizing study objects into distinct groups based on their characteristics [
69]. Utilizing the minimum variance clustering, also known as Ward’s method, the analysis focuses on minimizing variation within each cluster, resulting in distinct groupings [
69]. Ward’s method generates a single tree, where the goal is to minimize within-cluster variance.
For graphical representation, we employed Figtree software version 1.4.4. [
70].
2.6. Bayesian Hypothesis Testing: Explanation of the Approach and Its Application
2.6.1. Bayes–Laplace Theorem
For archaeological seed interpretation, we employed a Bayesian approach to determine the conditional probability of an archaeological seed belonging to a specific Vitis taxon Θi. We sought to answer: What is the likelihood that an archaeological seed or sample is assigned to Θi given its domestication index value xj and its cluster yj? Drawing upon a dataset of approximately 600 comparison samples, where the taxonomic identity is known a priori, both from seed morphology and grapevine plant studies, we constructed a discrete joint probability function p(X,Θ). This function assigns a posterior probability value to each combination of a Vitis taxon and a domestication index value or a Vitis taxon and a Ward’s tree cluster.
Bayes’ rule [Equation (3)] was employed to approximate the solution.
where
p(
ϴ│
x) is the posterior probability distribution for the parameter
ϴ given a single observed value of the variable
X =
xj, in our case the degree of domestication, which is represented by the domestication index value, which ranges from 0 (clearly wild) to 1 (cultivar with domesticated traits).
When considering the Bayes rule in terms of individual probabilities, formula [
1] can be read as [Equation (4)]:
For a given value of the data, for instance
X =
x4 and a specific value for the parameter
ϴ (
Vitis taxa), such as,
ϴ =
θ3, we get [Equation (5)].
In [Equation (5)], both the likelihood p(x4|θ3) and marginal likelihood p(x4) can be calculated based on the joint distribution derived from the comparison samples. The prior probability p(θ3) is estimated using data on the regional prevalence of different taxa, from established sources of evidence.
2.6.2. Unveiling Medieval Seeds: Bayesian Analysis in Focus
In a Bayesian context, this framework selection serves as prior knowledge, establishing “a priori” probabilities that influence our results. Bayesian principles allow us to refine probabilities of hypotheses by integrating all available prior evidence, considering temporal and spatial constraints. In this study, we focused on the geographical variation in proportions of different Vitis vinifera “proles”, and the ratio V. vinifera/V. sylvestris. A meticulous formulation of “a priori” probability distributions, based on robust evidence, is as crucial as defining the hypotheses and relevant variables.
The “a priori” probabilities assume that during the 9th-13th centuries CE, 30% of vines were wild, while 70% were cultivated. Altering these proportions would affect the probabilities. Regarding cultivars, we anticipate a proportion similar to the contemporary western Mediterranean, given most variety introductions and translocations had occurred by then (
Table 1).
Table 1.
Alternative hypotheses and their respective priors and likelihoods 1.
Table 1.
Alternative hypotheses and their respective priors and likelihoods 1.
Groups | Prior | L1 | L2 |
---|
Vitis vinifera cultivars | | | |
Proles Orientalis Negrul Subproles Antasiatica | 0.14 | 0.07 | 0.042 |
Proles Orientalis Negrul Subproles Caspica | 0.19 | 0.05 | 0.041 |
Proles Pontica Negrul | 0.24 | 0.18 | 0.035 |
Proles Occidentalis Negrul | 0.19 | 0.13 | 0.056 |
Varieties with intermediate characteristics | 0.04 | 0.06 | 0.077 |
Wild grapevines in natural habitats | | | |
Vitis sylvestris, autochthonous native western wild | 0.21 | 0.39 | 0.104 |
Feral and colonial wild grapevines, descended from cultivated plants | 0.05 | 0.06 | 0.068 |
Direct hybrids of wild Caucasian grapevines with cultivars * | 0.01 | 0.01 | 0.038 |
Purely Caucasian feral * | 0.01 | 0.01 | 0.041 |
Wild autochthonous native Caucasian grapevines * | 0.01 | 0.02 | 0.068 |
Unlikely hypotheses | | | |
American grapevine species * | 0.0008 | 0.001 | 0.074 |
Eastern Asian grapevine species * | 0.01 | 0.02 | 0.101 |
Fossils * | 0.0001 | 0.0001 | 0.255 |
Figure 2.
Ward’s minimum variance tree. Cluster labels: Blue fill, yellow text for medieval seed clusters. Green fill, white text for earlier period clusters mentioned. White fill, black text for others. Color codes (RGB) for sample labels:
Vitis caucasica: 0-153-153;
V. sylvestris: 0-102-0;
V. vinifera: 255-102-0;
V. vinifera ×
V. sylvestris: 204-153-0; other wild
Vitis species: 0-0-204;
Vitis seeds fossil: 102-51-0;
V. vinifera ×
V. caucasica: 153-102-0;
V. vinifera ×
V. amurensis: 102-0-0. A higher resolution version of this graph is available in
Supplementary Figure S2.
Figure 2.
Ward’s minimum variance tree. Cluster labels: Blue fill, yellow text for medieval seed clusters. Green fill, white text for earlier period clusters mentioned. White fill, black text for others. Color codes (RGB) for sample labels:
Vitis caucasica: 0-153-153;
V. sylvestris: 0-102-0;
V. vinifera: 255-102-0;
V. vinifera ×
V. sylvestris: 204-153-0; other wild
Vitis species: 0-0-204;
Vitis seeds fossil: 102-51-0;
V. vinifera ×
V. caucasica: 153-102-0;
V. vinifera ×
V. amurensis: 102-0-0. A higher resolution version of this graph is available in
Supplementary Figure S2.