Introduction

The genus Prunus L. (Rosaceae) includes about 400 species classified into five subgenera including Prunus, Amygdalus, Cerasus, Padus and Laurocerasus, which are mainly distributed in temperate regions of the boreal hemisphere (Krussman 1986; Maynard et al. 1991; Aradhya et al. 2004; Yilmaz et al. 2009). Domesticated Prunus includes European plum (Prunus domestica L.), Japanese plum (Prunus salicina Lindl.), peach [Prunus persica (L.) Batsch], apricot (Prunus armeniaca L.), sweet cherry [Prunus avium (L.) L.], sour cherry (Prunus cerasus L.) and almond [Prunus dulcis (Mill.) D.A. Webb]. With the exception of almonds, where the edible part consists of the seeds, the others are consumed for their fleshy fruits (Janick 2005).

Prunus domestica (plum) is one of the most economically important fruits in temperate regions and represents a major crop in Europe and southwest Asia (Ramming and Cociu 1991; Watkins 1995; Körber-Grohne 1996; Zohary et al. 2012). In 2013, FAOSTAT estimated that the total commercial harvest of plums was 12 million tons, cultivated from 2.5 million ha (FAOSTAT 2013).

The primary centre of plum domestication has been identified in Central Asia, with other secondary centres in East Asia, Europe and North America (Watkins 1995). However, the effective place(s) of origin and domestication of plum are still under investigation.

Crane and Lawrence (1952) and Watkins (1995) suggested that plum might be a polyploid derivative of a cross between the tetraploid Prunus spinosa L. and diploid P. cerasifera Ehrh. P. spinosa (sloe) is a shrub with a distribution range that extends from the western and central parts of Europe to Asia Minor; it is also present in the Caucasus region and North Africa (Hegi 1995). However, as suggested by Zohary et al. (2012), the wild relative of P. domestica is an autopolyploid derived from P. cerasifera that probably also partially contributed to two other wild species, including P. cocomilia Ten. and P. brigantino Vill. Moreover, P. domestica ssp. insititia (L.) Bonnier & Layens (damson) is considered the ancestor of modern plums (Woldring 2000; Zohary et al. 2012).

Recently, genetic studies have shown that P. spinosa, P. domestica ssp. insititia and P. domestica have close genetic relationships (Aradhya et al. 2004; Pollmann et al. 2005; Depypere et al. 2009; Horvath et al. 2011; Xuan et al. 2011; Milošević and Milošević 2012; Athanasiadis et al. 2013). In addition, different authors investigating genetic relatedness between modern Prunus species concluded that the phylogenetic reconstruction is the result of several processes of speciation derived from hybridisation that occurred during a long time span (Bouhadida et al. 2004, 2007; Katayama and Uematsu 2005; Wünsch 2009; Yilmaz et al. 2009; Horvath et al. 2011).

Archaeological evidence of P. spinosa fruitstones has been found in many archaeological sites in the western Mediterranean basin, dating between the Neolithic Age and Bronze Age (Woldring 2000; Zohary et al. 2012). However, in archaeological sites dating between the Bronze Age and early Iron Age, a large number of intermediate forms due to interspecific hybridisation among sloe, damson and plum have also been found (Pollmann et al. 2005). During the Roman period, the domestic plum seems to have appeared and then spread into western Europe (Janick 2005).

The earliest evidence of plum cultivation in Italy was found in a cesspit under the Temple of Fortuna in Pompeii, where a fruitstone of plum, dated to 150 bc, was found (Zech-Matterne et al. 2015). Also from Pompeii in the House of the Orchard, some painted representations of cultivated plums with yellow, blue and purple fruit dating back to 79 bc were found (Table 1; Jashemski and Meyer 2002).

Table 1 In chronological order from the earliest identifications until the 6th century ad, the major records of P. spinosa, P. domestica ssp. insititia and P. domestica found in archaeological contexts in western Europe and Carthage
Full size table

Lastly, written sources provide some descriptions of cultivated plums. For example, Theophrastus mentioned the names ‘Prumnon’ in his Enquiry into Plants (Περὶ Φυτῶν Ιστορίας) and Pliny described several varieties of plums with yellow, red, violet, black, white or bright coloured fruits in his Natural History (cited in Jashemski and Meyer 2002).

Prunus identification at the species level with traditional archaeobotanical methods is difficult due to the morphological range variation within the different taxa (Woldring 2000; Pollmann et al. 2005; Depypere et al. 2007). According to Horvath et al. (2011), the taxonomic classification of Prunus is generally done on the phenotypic characteristics of their flowers and fruits, and it would be better to associate both morphological characteristics and molecular markers, as the phenotypic characteristics are not always reliable due to variation that can occur due to environmental conditions. As argued by Depypere et al. (2007) and Woldring (2000), the fruitstone of Prunus would be the most stable of all diagnostic characters used for their identification at the species level. For this reason, in archaeobotanical studies the characteristics of the fruitstone were successfully used for their classification (Pollmann et al. 2005; Zheng et al. 2014).

During the last two decades, a significant increase in the use of image analysis applications has been highlighted in the plant biology research field, and automatized systems have the potential to replace human visual assessments. Due to the application of new image analysis technologies to plant biology, it is possible to use them on archaeobotanical material to distinguish, in an accurate, reproducible and repeatable way, wild taxa from cultivated ones (Terral et al. 2010; Bouby et al. 2013; Orrù et al. 2013; Pagnoux et al. 2015; Sabato et al. 2015; Ucchesu et al. 2015, 2016).

The recent discovery of several intact Prunus fruitstones recovered from inside various amphorae in the Phoenician and Punic contexts of the lagoon of Santa Giusta (Oristano, Sardinia), dated in a range between the 6th and the 2nd century bc, brings into question the spread of domesticated plums in Italy.

The present work aims to identify and characterise the Prunus remains from the archaeological contexts of Santa Giusta in order to investigate the domestication level of these remains by applying image analysis techniques and to explore the possible relationships among archaeological remains, traditional varieties of plum and wild populations present in Sardinia today.

Archaeological context

The Phoenician and Punic settlement of Santa Giusta is located in the north-central part of the Gulf of Oristano (39°51ʹ57″N, 8°35ʹ21″E) in Sardinia, near the former city of Othoca (Fig. 1). It has an almost circular shape, with a maximum area of 900 ha and a depth ranging from 40 to 150 cm. The site is waterlogged and has been excavated since 2005 under the supervision of the Soprintendenza per i Beni Archeologici per le province di Cagliari e Oristano and the University of Cagliari and is still in progress (Del Vais and Sanna 2009, 2012). The underwater excavation allowed for the recovery of several amphorae dating back to the Phoenician and Punic period in the 6th–2nd centuries bc (ESM 1; Del Vais and Sanna 2009). Various materials were found inside several amphorae and sediments, including animal remains such as Ovis aries, Capra hircus, Bos taurus (Portas et al. 2015) and macro plant remains, which were preserved in excellent condition due to the anaerobic conditions (Del Vais and Sanna 2009).

Fig. 1
figure 1

Location of the Santa Giusta lagoon and distribution of modern P. spinosa populations and P. domestica varieties in Sardinia used in this study

Full size image

Materials and methods

Archaeological samples

A total of 64 waterlogged Prunus fruitstones (code PRU_SG) were analysed in this study. Nine Prunus fruitstones came from four amphorae and 55 fruitstones from the layers R8, R9, R10. The remains were recovered by using the wash-over technique with a fine mesh (0.25 mm) (Kenward et al. 1980). Prunus remains were subsequently kept in distilled water and stored at +5 °C.

Modern samples

Modern samples of P. spinosa were collected from 11 different populations in Sardinia (ESM 2, Fig. 1), and fruitstones of P. domestica, representing 22 traditional varieties, came from different locations in Sardinia, duplicated in the field catalogue of CNR-ISPA (Nuraxinieddu, Oristano, Sardinia) (ESM 2, Fig. 1). Some of these samples were collected and selected from areas closest to the archaeological site to evaluate the potential relationships between the varieties and archaeological remains. In order to ensure the highest number of accessions, the fruit was sampled in three consecutive years, from 2012 to 2014. In addition, two accessions of P. domestica ssp. insititia (AN1 and AN2) preserved in the Sardinian Germplasm Bank (BG-SAR) were added to the study and considered as an outgroup.

Digital image analysis

Digital images of the modern and archaeological fruitstones were acquired using an Epson Perfection V550 flatbed scanner with a digital resolution of 400 dpi for a scanning area not exceeding 1024 × 1024 pixels (Bacchetta et al. 2008). Image acquisition of modern fruitstones was done after cleaning away of the pulp. To minimise shape variations, according to Depypere et al. (2007), image acquisition of the archaeological fruitstones was done on slightly dehydrated samples.

The images were processed and analysed using the software package ImageJ v. 1.49 (http://rsb.info.nih.gov/ij). A plugin, Particles 8 (Landini 2006), freely available on the official website http://www.mecourse.com/landinig/software/software.html, was used to measure 26 endocarp morphometric features (Table 2; Fig. 2). In all, 2,845 Prunus fruitstones were analysed.

Table 2 List of morphometric features measured on the fruitstones
Full size table
Fig. 2
figure 2

Graphical representation of principal morphometric features (see Table 3) measured on each endocarp

Full size image
Table 3 Identification percentages of the archaeological fruitstones of Prunus (PRU_SG) from the Santa Giusta context considered as unknown specimens
Full size table

Statistical analysis

The raw data recorded from the studied fruitstones were statistically analysed by applying the stepwise linear discriminant analysis (LDA) method, using IBM SPSS software package v. 16.0 (SPSS Inc. 2006). This method is commonly used to classify or identify unknown groups characterised by quantitative and qualitative variables (Fisher 1936, 1940; Sugiyama 2007). It allows for finding the combination of predictor variables with the aim of minimising the within-class distance and maximising the between-class distance simultaneously, thus achieving maximum class discrimination (Hastie et al. 2001; Holden et al. 2011; Alvin and William 2012; Kuhn and Johnson 2013).

On the basis of three statistical variables, Tolerance, F-to-enter and F-to-remove, the stepwise procedure selects the best features to use for the discrimination process. The Tolerance value indicates the proportion of a variable’s variance that is not accounted for by other independent variables in the equation. A variable with extremely low Tolerance values provides little information to the model. The F-to-enter and F-to-remove values define the power of each variable in the model and describe what happens if a variable is either inserted or removed from the current model (Grillo et al. 2012). This method starts with a model that does not include any variables. At each step, the variable with the largest F-to-enter value that exceeds the selected entry criteria (F ≥ 3.84) is added to the model. The variables omitted from the analysis at the last step have F-to-enter values smaller than 3.84 and are not added. The process is automatically stopped when no remaining variables are able to increase the discrimination of the method (Lo Bianco et al. 2017). Finally, a cross-validation procedure is applied to verify the performance of the identification system by testing individual unknown cases and classifying them on the basis of all the others. This procedure, also called rotation estimation (Picard and Cook 1984; Kohavi 1995), was applied, both to evaluate the performance and to validate any classifier. The validation procedure used here is the leave-one-out cross-validation (LOOCV) (Grillo et al. 2016). It involves using a single case from the original sample set as the validation dataset and the remaining cases as the training set. Each case is classified into a group according to the classification functions computed from all the data, except the case being classified. The proportion of misclassified cases after removing the effect of each case one at a time is the leave-one-out estimate of misclassification (SPSS 2006).

All the raw data were standardized before starting any statistical calculation. Moreover, in order to evaluate the quality of the discriminant functions achieved for each statistical comparison, the Wilks’ Lambda, the Eigenvalues, the percentage of explained variance, the Chi square and the Standardized Canonical Discriminant Function Coefficients (SCDFCs) were computed.

Results

To test the variability existing in P. domestica, a comparison among the fruitstones of the 22 modern varieties collected in Sardinia was carried out and an overall percentage of correct identification of 86.1% was reached (ESM 3).

A preliminary morphometric comparison was made among the fruitstones belonging to the three taxonomic entities (P. spinosa, P. domestica and P. domestica ssp. insititia) and the 64 waterlogged archaeological fruitstones from Santa Giusta (PRU_SG) were added to the classifier as an unknown group (Table 3).

An overall correct identification percentage of 94% was achieved. The fruitstones of P. spinosa were perfectly identified, while there were a few misidentifications among P. domestica and P. domestica ssp. insititia. Of the archaeological unknown fruitstones from Santa Giusta, 83% (53 fruitstones) were identified as P. spinosa and 17% (11 fruitstones) as P. domestica (Table 3). Table 4 shows the number of Prunus fruitstones identified from the amphorae and the layers R8, R9, R10.

Table 4 Prunus fruitstones identified and other biological remains found inside the amphorae and from layers R8, R9, R10
Full size table

No morphometric difference was observed between Prunus remains of differing ages (data not shown).

Considering these achievements, the 11 archaeological fruitstones identified as being from P. domestica, one more time considered as unknown specimens, were compared with the modern varieties of plum. In this case, the archaeological samples from Santa Giusta showed most similarities with the varieties Sanguigna di Bosa (SBO) in 81.8% of cases and Di Bonarcado (FAR) in 9.1% of cases (Fig. 3). Likewise, the 53 archaeological fruitstones from Santa Giusta, identified as P. spinosa, were considered unknown and compared with the modern wild populations of P. spinosa from Sardinia. These archaeological fruitstones were very similar to those collected at Monte Arci (MRC) in 90.6% of cases (Table 5; Fig. 3).

Fig. 3
figure 3

Representation of the samples analysed. Below are the types of Prunus fruitstones identified from the Santa Giusta contexts and their relationship with the modern accession material

Full size image
Table 5 Correct classification percentages in modern populations and archaeological samples of P. spinosa
Full size table

In the evaluation of the features, the most discriminant five variables, of the 25 selected and used by the stepwise LDA, are reported. The first variable is the area of the endocarp, with a high value of F-to-remove (ESM 4).

Discussion

The domestication process of fruit trees remains unclear, perhaps because fruits have received much less attention than annual crop plants (Goldschmidt 2013). The identification of the place of origin of cultivated species of Prunus is difficult due to their long history of cultivation, to which human dispersion to different places is added (Pollmann et al. 2005). Therefore, Prunus species may have become naturalised, creating difficulties for the distinction between ancestrally wild populations and those which escaped from cultivation (Kole and Abbott 2012). As suggested by Pollmann et al. (2005), attribution of Prunus remains to a specific species is limited due to the imprecise classification of these groups.

From the results obtained through LDA, it was possible to identify correctly the unknown Prunus remains of the Santa Giusta context as cultivated varieties and wild species. From the 64 archaeological remains, 53 of these were classified as P. spinosa, while the other 11 were classified as P. domestica. In particular, none of these fruitstones were attributed to P. domestica ssp. insititia, the wild form at the origin of the domestic plum. Based on these achievements, it can be assumed that the earliest plum cultivation may have occurred in Sardinia at least since the 6th century bc, during the Archaic Period. However, the place of origin of these cultivated fruit trees is still unknown. There are more finds of domesticated fruitstones of plum in waterlogged contexts of the Roman period, suggesting that the Romans contributed to the spread of several varieties of plums into western Europe (Pollmann et al. 2005; Zohary et al. 2012). A further result of this study is that the archaeological remains from Santa Giusta which were identified as P. domestica, are similar to a traditional variety that is cultivated in the territory of Bosa, in northwestern Sardinia.

The close relationship shown by the comparative analysis between the archaeological and modern samples of P. spinosa allow us to hypothesise that the wild fruit found in the Santa Giusta amphorae might have been gathered on the slopes of Monte Arci which is located just 10 km from the Santa Giusta archaeological site. This massif is a volcanic complex rich in obsidian materials which were exploited for millennia by the Neolithic community. In addition, it is probable that during the Phoenician and Punic period this was an important area for the exploitation of natural resources, as shown by the gathered wild fruit found in the amphorae of Santa Giusta.

Possible hypotheses about the use of this wild fruit can be made using ethnobotanical research. The uses of sloes are varied; ethnobotanical literature indicates their use principally for food and medicine (Parada et al. 2009; Tiţă et al. 2009; Łuczaj 2012; Pardo-de-Santayana et al. 2013; Pieroni and Quave 2014). In Sardinia, the consumption of sloes as food and as medicine is well documented, as a liquid extract of flowers or fruit for the treatment of coughs, in addition to their traditional use for dyeing wool (Atzei 2003; Campanini 2009).

Other uses may be related to religious rituals: in some Punic tombs, charcoal remains of sloe could represent firewood for human body cremation or ritual offerings (Gómez Bellard et al. 1990). In addition, in Roman cemeteries, the use of fresh fruit of sloe, damson and plum as ritual offerings is well known (Preiss et al. 2005; Cooremans 2008; Bouby et al. 2011; Rottoli and Castiglioni 2011).

In the case of Santa Giusta, the presence of P. spinosa and P. domestica together with animal remains could represent their use as food or could be linked to food preservation methods. This practice is well known for Phoenician and Punic Sardinia (Del Vais and Sanna 2012).

Conclusions

The discovery of well-preserved waterlogged fruitstones of P. domestica in the Phoenician and Punic contexts of Santa Giusta could be evidence that the introduction of primitive cultivated forms of plums in Sardinia was started by the Phoenician people in the Archaic period. Therefore, these fruitstones represent the first cultivated plum finds in Sardinia and the oldest evidence of cultivated plums in Italy. We hope for future investigations to better understand the history of the beginning of domestication of fruit trees in the Mediterranean basin.