Creating Value for the Montepulciano D’Abruzzo PDO Chain: A Pilot Study of Supply Chain Traceability Using Multi-Elemental and Chemometrics Analysis of Wine and Soil

Abstract

This study aims to enhance the value of the Montepulciano d’Abruzzo PDO supply chain by integrating multi-elemental and isotopic profiling with chemometric analysis. The objective is to establish a pilot study for origin authentication, supporting strategic, managerial, and regulatory decision-making for stakeholders in the wine sector. Wine and soil samples from producers in the Abruzzo region were analyzed for 63 elements and selected isotopic ratios using HR-ICP-MS and MC-ICP-MS. Exploratory data analysis, including PCA and clustering, was employed to investigate intrinsic data structure. Variable selection techniques identified the most discriminant markers, and multiple classification models were tested to assess producer-level differentiation. The combined elemental and isotopic dataset showed strong intrinsic structure. Four variables—Mo, ²⁰⁸Pb/²⁰⁶Pb, P, and ⁸⁷Sr/⁸⁶Sr—emerged as key discriminants. Quadratic Discriminant Analysis and Artificial Neural Networks achieved 100% accuracy in classifying samples by producer. The results demonstrate that integrating multi-elemental and isotopic data with chemometric tools offers a pilot approach to authenticate wine origin and enhance transparency across the PDO supply chain. Beyond scientific innovation, this study provides a pilot decision support model that can strengthen competitive differentiation, regulatory compliance, and sustainable territorial development, highlighting opportunities for digital transformation in PDO management.

1. Introduction

The Protected Designation of Origin (PDO) system represents one of the most effective mechanisms for safeguarding the authenticity, quality, and cultural identity of agri-food products [1]. Within the wine sector, PDO certification serves as a legal and commercial guarantee of provenance, ensuring that production methods, grape varieties, and regional conditions conform to strict regulatory frameworks [2]. Beyond mere certification, PDOs embody the philosophy of terroir—the intricate interplay between soil, climate, human know-how, and local biodiversity—that shapes each wine’s organoleptic profile and typicity. In doing so, PDOs protect traditional viticultural knowledge while providing consumers with a guarantee of authenticity and a sense of connection to place [3,4,5].

Many wine-producing regions have a rich history and cultural heritage associated with winemaking. PDOs help preserve and promote these traditions, ensuring that future generations can continue producing wines deeply rooted in local culture and history [6,7]. PDOs create a strong identity for wines from specific regions, helping them stand out in the market [8,9,10]. This differentiation can lead to increased consumer interest and demand and the potential for higher prices for wines with a recognized and respected PDO designation [11,12].

In economic terms, PDO wines contribute significantly to regional competitiveness and export performance. Italy, one of the world’s largest wine producers, relies heavily on PDO designations as both a quality assurance mechanism and a branding instrument for international markets. The success of PDO systems is closely linked to their ability to ensure traceability, particularly in an era characterized by globalized supply chains, rising counterfeiting risks, and increasingly discerning consumers. Thus, developing scientifically validated tools for origin verification is crucial to maintaining consumer trust, preserving market reputation, and promoting territorial sustainability [13,14,15,16].

Despite its prominence, the Montepulciano d’Abruzzo PDO chain—one of Italy’s most widely consumed and culturally symbolic red wines—has received limited scientific attention regarding traceability at the micro-geographic scale. Montepulciano d’Abruzzo differs from other Italian appellations in its distinctive balance of structural richness and aromatic complexity, reflecting the region’s diverse soils and microclimates. However, traditional analytical approaches have been insufficient to differentiate producers operating within the same PDO boundaries, limiting the capacity to demonstrate authentic terroir expression or detect fraudulent labeling. Furthermore, consumers around the world ultimately enjoy Montepulciano d’Abruzzo wines [17,18]. The wine’s flavors, aromas, and characteristics reflect the Abruzzo region’s terroir and the winemakers’ expertise. Throughout this production chain, various stakeholders, including grape growers, winemakers, regulatory bodies, and distributors, contribute to creating and successfully marketing Montepulciano d’Abruzzo wines [19]. It is essential to recognize that each step in the chain plays a crucial role in maintaining this distinctive Italian wine’s quality, authenticity, and reputation.

This study aims to fill this knowledge gap by proposing an integrated analytical–statistical approach capable of identifying compositional fingerprints at the producer level. Through the combined use of high-resolution and multi-collector inductively coupled plasma mass spectrometry (HR-ICP-MS and MC-ICP-MS), we accurately quantified elemental and isotopic compositions in wine and corresponding vineyard soil samples from five Montepulciano d’Abruzzo producers. Subsequently, we employed a suite of multivariate and machine learning methods—Principal Component Analysis (PCA), Linear and Quadratic Discriminant Analysis (LDA and QDA), Artificial Neural Networks (ANNs), Random Forest (RF), Naïve Bayes (NB), and Support Vector Machines (SVM)s—to build a robust classification model capable of distinguishing producers.

To our knowledge, this is the first study that provides an integrated multi-elemental analysis of wine and soil from the Montepulciano D’Abruzzo PDO chain, coupled to multivariate analyses. Based on the literature review, even if some papers used this approach for other PDOs [20,21,22,23,24,25], no recent papers were found on this topic. Two papers published in 2000 and 2004 have explored possible classification of Montepulciano d’Abruzzo wines [26,27]. In these studies, classification based only on the year of production or the pedoclimatic zone was possible using several analyses, i.e., reducing sugars, total alcohol, volatile acidity, tartaric acid, and trans-resveratrol contents. Another paper provides multivariate clustering of the Montepulciano d’Abruzzo territory based on land use information and a geologic map [28]. Therefore, the present paper represents a step forward in advancing knowledge of the Montepulciano d’Abruzzo chain and it can be used as a benchmark for future publications.

Furthermore, this study contributes to both scientific and managerial domains. Scientifically, it advances the frontier of analytical traceability by coupling isotopic and multi-elemental profiling with data-driven modeling. From a managerial perspective, it provides a practical framework for producers, consortia, and regulatory agencies to implement evidence-based traceability systems that enhance quality assurance, strengthen territorial branding, and improve governance of PDO value chains. In doing so, this paper aligns analytical chemistry with managerial science to promote a more integrated vision of authenticity and competitiveness in the wine industry.

2. Materials and Methods

2.1. Sample Collection and Preparation

Wine (n = 40) and soil (n = 40) samples were collected from five local producers in the Montepulciano D’Abruzzo PDO production area. The relatively small number of producers involved in the sampling is mainly due to structural and territorial factors of the region. First, the Montepulciano d’Abruzzo PDO production area is geographically limited, and the overall vineyard surface is relatively small compared to other Italian wine districts. Moreover, only 200 winemakers produce this wine and a large proportion of them are organized as cooperatives. In such cases, grapes are sourced from a wide range of vineyards without traceability at the single-vineyard or soil level, making it impossible to directly link the resulting wine to a specific soil sample. This structural condition significantly restricted the number of eligible producers for this study. In addition, only five agreed to participate and allow both soil and wine sampling at their estates but they represent the 80% of non-cooperative production. For these reasons, the dataset represents a limited but representative subset of producers that ensure reliable traceability between vineyard soil and the corresponding bottled wine. The geographical location of the producers is reported in Figure S1. The meteorological conditions (Table S1) of the producers were the same, as was the soil parent material (pedological region 35 of Italy). For each wine sample, soil samples were collected at regular intervals at a depth of 30 cm and a distance of 50 cm from the vines that produced that wine. Each wine and soil sample were collected in triplicate (three wine bottles and three soil portions) and then the results were averaged. The samples were collected from two vintages, 2021–2022 and 2022–2023, to mitigate seasonality fluctuation.

2.2. Chemicals, Reagents, and Instrumentation

Nitric acid (HNO₃; Sigma-Aldrich Chemie GmbH, Munich, Germany) of analytical grade was used throughout the study. All water was deionized Milli-Q water (Millipore, Bedford, MA, USA), produced by reverse osmosis followed by ion exchange purification. For the dilution of wine samples and sample digests, the water was further purified by sub-boiling distillation in Teflon stills. Pre-packed 2 mL columns containing Sr-Spec resin (4,4′(5′)-di-tert-butylcyclohexano crown ether in 1-octanol on an inert polymeric support; Eichrom Technologies LLC, Lisle, IL, USA) were used as received. Instrumental mass bias was corrected using NIST SRM 987 (Strontium Carbonate), NIST SRM 951a (Boric Acid), and NIST SRM 981 (Common Lead Isotopic Standard), all obtained from the National Institute of Standards and Technology (Gaithersburg, MD, USA). For quality assurance, certified reference materials (CRMs) were included and processed through the complete analytical procedure, including digestion, separation, and analysis. These comprised SLRS-5 River Water CRM and NASS-6 Seawater CRM for trace metals (National Research Council, Ottawa, ON, Canada), NIST SRM 1547 Peach Leaves (National Institute of Standards and Technology, Gaithersburg, MD, USA), and GBW 07,410 soil (National Research Centre of Geoanalysis, Beijing, China). All measurements were performed as reported by [29,30].

2.3. Sample Preparation and Analysis

Personnel wearing clean room attire prepared the samples in Class 10,000 clean laboratory areas. General precautions detailed by [29] were taken to minimize contamination. Laboratory materials used during sample preparation were soaked in 0.7 M HNO₃ for 24 h at room temperature and rinsed with deionized Milli-Q water before use. Wine sample preparation and analysis were conducted as previously reported in [29]. For multi-element analysis, wine samples were diluted 40-fold with 1.4 M HNO₃. Each batch included method blanks and certified reference materials (CRMs). Concentrations of 63 elements were measured using high-resolution inductively coupled plasma mass spectrometry (HR-ICP-MS). Matrix effects were corrected by internal standardization, with indium added to all solutions at 2.5 µg/L, and quantification was achieved via external calibration using concentration-matched standards. The remaining portion of each original sample was evaporated to dryness in a 25 mL Teflon beaker at 95 °C on a ceramic-top hotplate. The residue was redissolved in 3 mL of 3 M HNO₃ and subjected to Sr and Pb separation using Sr-specific resin columns. Sr was eluted with 0.05 M HNO₃, followed by Pb elution with 0.002% EDTA. A 0.1 mL aliquot of each purified fraction was then diluted 50-fold with 1.4 M HNO₃ and analyzed by HR-ICP-MS. This analysis of the purified fractions served four purposes: (I) determining analyte concentrations for preparing concentration- and acid strength-matched solutions for isotope ratio measurements, (II) assessing analyte recovery, (III) evaluating separation efficiency from matrix elements, and (IV) detecting possible spectral interferences from the matrix or contamination. Separated fractions were diluted to uniform concentrations. Bracketing standards were prepared to match samples in both analyte and internal standard concentrations (with Tl added for Pb isotope ratios) and acid strength. Instrumental mass bias was corrected using a combination of standard-sample bracketing and internal normalization based on tabulated isotope ratios (⁸⁸Sr/⁸⁶Sr, ¹¹B/¹⁰B, or ²⁰⁵Tl/²⁰³Tl). For soil samples, the analyses were carried out as reported in [30]. The sequential extraction procedure (SEP) followed a six-step method. In the first step (F1), the exchangeable fraction was extracted using 40 mL of distilled water. The second step (F2) targeted carbonate-bound species with 40 mL of 0.11 M acetic acid (CH₃COOH). The third step (F3) addressed the reducible phase using 40 mL of 0.1 M hydroxylamine hydrochloride (NH₂OH·HCl). The oxidizable phase (F4) was treated first with 2 × 10 mL of 8.8 M hydrogen peroxide (H₂O₂), followed by extraction with 50 mL of 1 M ammonium acetate (CH₃COONH₄) adjusted to pH 2. The fifth step (F5) targeted residual fraction I, using 20 mL of aqua regia. Finally, the sixth step (F6) addressed residual fraction II with 10 mL of hydrofluoric acid (HF). The studies mentioned above also provide figures of merit for the analytical methods used. Calibration curves are reported in Table S2.

2.4. Statistical Analysis

Analysis of variance (ANOVA) and mean comparison by Tukey’s honest significant difference (HSD) for an unequal number of samples at the 5% level were performed using JMP 17 Pro (SAS Institute, Singapore). Hopkins statistic, Visual Assessment of Cluster Tendency (VAT), and cluster analysis were performed using Rstudio v. 1.3.1093. Outlier detection was performed using three complementary statistical approaches to ensure robustness: Hotelling’s T² statistic (Mahalanobis distance), the interquartile range (IQR) method, and Cauchy distribution-based analysis. Hotelling’s T² was applied to multivariate data to identify observations with an unusually large distance from the multivariate centroid, taking into account correlations among variables. Observations exceeding the critical T² threshold at the chosen confidence level were flagged as potential multivariate outliers. In parallel, the IQR method was used in a univariate context, where values lying outside Q₁−1.5 × IQR or Q₃ + 1.5 × IQR were considered outliers. Additionally, fitting a Cauchy distribution—known for its sensitivity to extreme values—allowed for the identification of observations with disproportionately high influence on distribution tails. Only samples consistently identified as outliers by using these methods were removed.

Chemometric data analyses (PCA, variable selection, LDA, QDA, ANN, RF, NB, SVM) were also performed with JMP 17 Pro. The hyper-parameters were chosen according to Rapa et al. [29]. Before the chemometric assessment, an autoscaling pre-treatment was carried out on the data matrix [29,30]. Validation of classification method was performed by using the Leave-One-Out method [31,32].

3. Results and Discussion

3.1. Wine Analysis Results

A total of sixty-three elements (Al, As, B, Ba, Be, Bi, Br, Ca, Cd, Ce, Co, Cr, Cs, Cu, Dy, Er, Eu, Fe, Ga, Gd, Ge, Hf, Hg, Ho, I, K, La, Li, Lu, Mg, Mn, Mo, Na, Nb, Nd, Ni, P, Pb, Pr, Pt, Rb, Re, S, Sb, Sc, Se, Si, Sm, Sn, Sr, Tb, Te, Th, Ti, Tl, Tm, U, W, V, Y, Yb, Zn, and Zr) were quantified in 20 wine samples from five local Montepulciano d’Abruzzo producers using high-resolution inductively coupled plasma mass spectrometry (HR-ICP-MS). A multi-element screening approach was adopted to identify the most promising elemental markers for differentiating among the producers. The analytical results are summarized in

Table 1. The results of the multi-elemental analysis of Montepulciano d’Abruzzo wines, expressed as µg/L. Rows in the same column not linked by the same letter are statistically different (p-value < 0.01) by Tukey’s HSD.

		A	B	C	D	E
Al	Mean	320 b	240 b	190 b	350 b	770 a
	S.D.	44	150	53	150	81
	Min	280	140	150	150	690
	Median	320	240	170	380	770
	Max	380	340	250	500	850
As	Mean	1.100 b	0.450 b,c	0.270 c	0.460 b,c	2.200 a
	S.D.	0.450	0.014	0.130	0.200	0.058
	Min	0.740	0.440	0.140	0.230	2.200
	Median	0.890	0.450	0.270	0.440	2.200
	Max	1.700	0.460	0.390	0.720	2.300
B	Mean	9000 a	7100 a	7900 a	9200 a	9200 a
	S.D.	1800	140	570	1500	610
	Min	7200	7000	7300	7600	8700
	Median	8600	7100	8200	9300	9100
	Max	12,000	7200	8300	11,000	9900
Ba	Mean	110 b	120 b	73 c	120 b	160 a
	S.D.	18	4.200	8.800	11	8.600
	Min	97	120	66	110	150
	Median	110	120	70	120	160
	Max	140	120	83	130	170
Be	Mean	0.180 a,b	0.066 b	0.270 a,b	0.150 b	0.450 a
	S.D.	0.003	0.025	0.200	0.140	0.080
	Min	0.180	0.049	0.058	0.048	0.400
	Median	0.180	0.066	0.310	0.100	0.420
	Max	0.180	0.083	0.450	0.350	0.550
Bi	Mean	0.016 b	0.014 b	0.017 b	0.005 b	0.050 a
	S.D.	0.011	0.009	0.006	0.003	0.017
	Min	0.005	0.008	0.010	0.002	0.040
	Median	0.015	0.014	0.020	0.005	0.040
	Max	0.030	0.020	0.020	0.008	0.070
Br	Mean	47 b	36 b	52 b	49 b	230 a
	S.D.	4.700	0.710	8.700	4.500	70
	Min	41	35	45	44	150
	Median	48	36	50	49	260
	Max	52	36	62	55	280
Ca	Mean	52,000 c	77,000 a,b	56,000 b,c	63,000 b,c	96,000 a
	S.D.	4300	850	13,000	10,000	5200
	Min	47,000	77,000	41,000	51,000	91,000
	Median	53,000	77,000	61,000	62,000	96,000
	Max	57,000	78,000	67,000	75,000	100,000
Cd	Mean	0.150 b	0.027 c	0.011 c	0.110 b	0.290 a
	S.D.	0.036	0.004	0.006	0.015	0.012
	Min	0.093	0.024	0.006	0.091	0.270
	Median	0.160	0.027	0.011	0.110	0.290
	Max	0.170	0.029	0.018	0.130	0.300
Ce	Mean	0.130 a	0.011 a	0.120 a	0.065 a	0.220 a
	S.D.	0.068	0.003	0.090	0.070	0.150
	Min	0.072	0.009	0.014	0.016	0.076
	Median	0.130	0.011	0.160	0.037	0.210
	Max	0.200	0.014	0.180	0.170	0.370
Co	Mean	1.800 b	2.300 b	0.740 c	2.200 b	4.400 a
	S.D.	0.240	0.160	0.100	0.620	0.530
	Min	1.600	2.200	0.620	1.600	4
	Median	1.700	2.300	0.790	2.300	4.200
	Max	2.200	2.400	0.800	2.800	5
Cr	Mean	9.700 a	6.800 a	5.300 a	8.600 a	60 a
	S.D.	2.700	0.140	1.500	2	60
	Min	6.900	6.700	3.600	5.800	19
	Median	9.400	6.800	6.100	9.100	31
	Max	13	6.900	6.300	11	130
Cs	Mean	3.500 a	2.500 a	3.400 a	3.800 a	3.600 a,b
	S.D.	0.200	0.230	0.410	1.300	0.990
	Min	3.200	2.300	3.100	2.500	3
	Median	3.500	2.500	3.400	3.600	3.100
	Max	3.700	2.600	3.900	5.700	4.800
Cu	Mean	100 a,b	42 a,b	90 a,b	24 b	310 a
	S.D.	41	20	25	4.300	210
	Min	66	28	62	17	71
	Median	98	42	95	25	370
	Max	160	56	110	27	480
Dy	Mean	0.025 a,b	0.004 b	0.015 a,b	0.014 b	0.060 a
	S.D.	0.017	0.001	0.008	0.018	0.026
	Min	0.010	0.003	0.006	0.001	0.040
	Median	0.020	0.004	0.020	0.007	0.050
	Max	0.050	0.004	0.020	0.040	0.090
Er	Mean	0.023 b	0.003 b	0.012 b	0.016 b	0.067 a
	S.D.	0.015	0.001	0.008	0.023	0.021
	Min	0.010	0.002	0.005	0.002	0.050
	Median	0.020	0.003	0.010	0.005	0.060
	Max	0.040	0.003	0.020	0.050	0.090
Eu	Mean	0.008 b	0.008 a,b,c	0.005 c	0.008 b	0.010 a
	S.D.	0.001	0.001	0.001	0.001	0
	Min	0.007	0.007	0.004	0.006	0.010
	Median	0.008	0.008	0.005	0.008	0.010
	Max	0.008	0.008	0.006	0.009	0.010
Fe	Mean	1400 a	1000 a	1300 a	1900 a	1700 a
	S.D.	190	91	200	660	130
	Min	1200	980	1100	1200	1500
	Median	1300	1000	1300	1900	1800
	Max	1600	1100	1500	2500	1800
Ga	Mean	0.070 a,b	0.042 b	0.032 b	0.045 b	0.110 a
	S.D.	0.011	0.023	0.006	0.024	0.036
	Min	0.060	0.026	0.025	0.017	0.079
	Median	0.068	0.042	0.035	0.043	0.100
	Max	0.086	0.058	0.037	0.076	0.150
Gd	Mean	0.017 a,b	0.001 b	0.008 a,b	0.010 a,b	0.033 a
	S.D.	0.007	0	0.006	0.008	0.019
	Min	0.009	0.001	0.002	0.003	0.016
	Median	0.017	0.001	0.008	0.007	0.030
	Max	0.026	0.001	0.014	0.022	0.053
Ge	Mean	n.d. a	0.025 a	0.017 a	0.013 a	0.027 a
	S.D.	\	0.035	0.029	0.025	0.025
	Min	\	n.d.	n.d.	n.d.	n.d.
	Median	\	0.025	\	\	0.030
	Max	\	0.050	0.050	0.050	0.050
Hf	Mean	0.035 a,b	0.004 b	0.028 a,b	0.027 a,b	0.080 a
	S.D.	0.006	0.001	0.023	0.042	0.020
	Min	0.030	0.003	0.004	0.003	0.060
	Median	0.035	0.004	0.030	0.007	0.080
	Max	0.040	0.004	0.050	0.090	0.100
Hg	Mean	0.015 a	0.009 a	0.016 a	0.016 a	0.030 a
	S.D.	0.008	0	0.020	0.011	0.020
	Min	0.004	0.009	0.003	0.008	0.011
	Median	0.017	0.009	0.007	0.013	0.029
	Max	0.023	0.009	0.039	0.032	0.051
Ho	Mean	0.006 a,b	0.001 b	0.003 b	0.004 a.b	0.015 a
	S.D.	0.003	0	0.002	0.005	0.007
	Min	0.003	0.001	0.001	0.001	0.009
	Median	0.006	0.001	0.003	0.002	0.013
	Max	0.010	0.001	0.004	0.012	0.023
I	Mean	9.800 b	7.500 c	10 b	10 b	20 a
	S.D.	0.500	0.710	0	0	0
	Min	9	7	10	10	20
	Median	10	7.500	10	10	20
	Max	10	8	10	10	20
K	Mean	1,300,000 a	1,300,000 a,b	960,000 a,b	1,000,000 b	140,0000 a
	S.D.	64,000	160,000	140,000	120,000	53,000
	Min	1,300,000	1,100,000	800,000	960,000	1,300,000
	Median	1,400,000	1,300,000	1,000,000	970,000	1,300,000
	Max	1,400,000	1,400,000	1,100,000	1,200,000	1,400,000
La	Mean	0.058 a	0.004 a	0.039 a	0.030 a	0.100 a
	S.D.	0.034	0.001	0.031	0.031	0.063
	Min	0.024	0.004	0.004	0.008	0.035
	Median	0.057	0.004	0.054	0.018	0.110
	Max	0.092	0.005	0.060	0.076	0.160
Li	Mean	47 a	7.500 a	41 a	40 a	18 ab
	S.D.	24	2.900	0.440	12	6
	Min	23	5.500	40	24	11
	Median	42	7.500	41	43	21
	Max	79	9.600	41	49	22
Lu	Mean	0.006 b	0.001 b	0.004 b	0.006 b	0.019 a
	S.D.	0.002	0	0.001	0.008	0.007
	Min	0.004	0.001	0.003	0.002	0.013
	Median	0.005	0.001	0.004	0.002	0.018
	Max	0.009	0.002	0.006	0.018	0.026
Mg	Mean	130,000 a	130,000 a	120,000 a	140,000 a	140,000 a
	S.D.	15,000	4900	5700	19,000	5500
	Min	110,000	120,000	110,000	120,000	130,000
	Median	130,000	130,000	120,000	130,000	130,000
	Max	150,000	130,000	120,000	160,000	140,000
Mn	Mean	1100 a,b	1300 a,b	970 b	1200 a,b	1400 a
	S.D.	79	150	35	110	270
	Min	1000	1200	930	1000	1300
	Median	1100	1300	980	1200	1300
	Max	1200	1500	1000	1300	1700
Mo	Mean	2.800 a	2.300 a	2.900 a	2.700 a	3.200 a
	S.D.	0.510	0.480	0.470	0.390	0.230
	Min	2.200	2	2.500	2.400	2.900
	Median	2.700	2.300	2.700	2.500	3.300
	Max	3.400	2.700	3.400	3.300	3.300
Na	Mean	26,000 b	7100 c	13,000 c	12,000 c	63,000 a
	S.D.	6400	230	2700	3000	3500
	Min	19,000	6900	12,000	8800	59,000
	Median	26,000	7100	12,000	12,000	64,000
	Max	34,000	7200	17,000	16,000	66,000
Nb	Mean	0.063 a	0.018 a	0.024 a	0.046 a	0.100 a
	S.D.	0.045	0.007	0.016	0.058	0.034
	Min	0.035	0.013	0.010	0.005	0.081
	Median	0.044	0.018	0.021	0.025	0.083
	Max	0.130	0.023	0.042	0.130	0.140
Nd	Mean	0.070 a	0.015 a	0.056 a	0.043 a	0.120 a
	S.D.	0.040	0.001	0.031	0.038	0.076
	Min	0.036	0.014	0.022	0.018	0.040
	Median	0.062	0.015	0.063	0.027	0.140
	Max	0.120	0.015	0.082	0.098	0.190
Ni	Mean	16 b	11 a,b	9.700 b	20 a,b	43 a
	S.D.	0.950	1.500	0.470	3.100	25
	Min	15	10	9.300	18	24
	Median	16	11	9.700	19	34
	Max	17	12	10	25	72
P	Mean	410,000 a	410,000 a,b	340,000 a,b,c	340,000 b,c	320,000 c
	S.D.	15,000	21,000	40,000	14,000	40,000
	Min	390,000	390,000	300,000	330,000	280,000
	Median	410,000	410,000	350,000	330,000	340,000
	Max	420,000	420,000	380,000	360,000	350,000
Pb	Mean	18 a	1.800 a	2.600 a	13 a	13 a
	S.D.	6.900	0.330	0.260	18	4.200
	Min	8.100	1.600	2.400	1.800	10
	Median	19	1.800	2.600	5.400	12
	Max	25	2	2.900	41	18
Pr	Mean	0.015 a	0.002 a	0.010 a	0.008 a	0.026 a
	S.D.	0.006	0	0.006	0.008	0.016
	Min	0.008	0.002	0.002	0.002	0.009
	Median	0.014	0.002	0.013	0.006	0.029
	Max	0.021	0.002	0.013	0.020	0.040
Pt	Mean	n.d. a	0.001 a	0.001 a	0.002 a	n.d. a
	S.D.	\	0.002	0.002	0.001	\
	Min	\	n.d.	n.d.	0.001	\
	Median	\	0.001	0.001	0.001	\
	Max	\	0.003	0.003	0.003	\
Rb	Mean	2000 a	1700 a,b	1500 b	1700 a,b	1300 b
	S.D.	170	110	40	21	350
	Min	1900	1600	1500	1700	1100
	Median	1900	1700	1500	1700	1200
	Max	2200	1800	1600	1700	1700
Re	Mean	0.009 b	0.003 b	0.004 b	0.010 b	0.040 a
	S.D.	0.002	0.003	0.001	0.001	0.017
	Min	0.007	0.001	0.004	0.009	0.020
	Median	0.009	0.003	0.004	0.010	0.050
	Max	0.010	0.005	0.005	0.010	0.050
S	Mean	230,000 a,b	220,000 a,b	160,000 b	160,000 b	340,000 a
	S.D.	20,000	30,000	48,000	15,000	120,000
	Min	200,000	200,000	120,000	140,000	230,000
	Median	240,000	220,000	150,000	160,000	310,000
	Max	250,000	240,000	220,000	180,000	470,000
Sb	Mean	0.320 b	0.130 b	1 a	0.710 a,b	0.320 b
	S.D.	0.050	0.007	0.058	0.430	0.082
	Min	0.250	0.120	0.980	0.150	0.230
	Median	0.340	0.130	1.100	0.770	0.360
	Max	0.370	0.130	1.100	1.200	0.380
Sc	Mean	0.018 a	0.020 a	0.017 a	0.021 a	0.033 a
	S.D.	0.016	0.028	0.015	0.024	0.006
	Min	0.003	n.d.	n.d.	0.001	0.030
	Median	0.015	0.020	0.020	0.016	0.030
	Max	0.040	0.040	0.030	0.050	0.040
Se	Mean	0.280 b	0.150 b	0.530 b	0.530 b	2.300 a
	S.D.	0.210	0.210	0.250	0.330	0.580
	Min	n.d.	n.d.	0.300	0.300	2
	Median	0.300	0.150	0.500	0.400	2
	Max	0.500	0.300	0.800	1	3
Si	Mean	9200 b,c	12,000 b	8100 b	8900 b,c	18,000 a
	S.D.	600	2300	410	910	1900
	Min	8400	11,000	7600	7900	17,000
	Median	9300	12,000	8300	8800	18,000
	Max	9800	14,000	8300	10,000	21,000
Sm	Mean	0.015 a	0.004 a	0.012 a	0.011 a	0.026 a
	S.D.	0.006	0.001	0.008	0.013	0.016
	Min	0.010	0.003	0.005	0.003	0.009
	Median	0.015	0.004	0.010	0.006	0.030
	Max	0.020	0.004	0.020	0.030	0.040
Sn	Mean	160 a	0.130 b	1.200 b	0.750 b	1.500 b
	S.D.	6.800	0.002	0.830	1	0.340
	Min	160	0.130	0.540	0.190	1.200
	Median	160	0.130	0.830	0.240	1.400
	Max	170	0.130	2.100	2.300	1.900
Sr	Mean	970 a,b	520 b	720 a,b	1200 a	950 a,b
	S.D.	210	190	49	270	48
	Min	670	390	660	840	890
	Median	1000	520	720	1200	980
	Max	1200	650	760	1400	980
Tb	Mean	0.004 a	0.001 a	0.002 a	0.003 a	0.006 a
	S.D.	0.002	0	0.002	0.002	0.004
	Min	0.002	0.001	0.001	0.002	0.003
	Median	0.005	0.001	0.002	0.002	0.006
	Max	0.006	0.001	0.004	0.005	0.010
Te	Mean	0.012 a	0.019 a	0.014 a	0.007 a	0.014 a
	S.D.	0.009	0.004	0.003	0.005	0.006
	Min	0.003	0.016	0.012	0.002	0.007
	Median	0.012	0.019	0.013	0.007	0.015
	Max	0.021	0.021	0.017	0.014	0.019
Th	Mean	0.010 a	0.001 a	0.013 a	0.002 a	0.029 a
	S.D.	0.009	0.001	0.011	0.003	0.023
	Min	0.002	0	0.001	0	0.014
	Median	0.007	0.001	0.015	0.001	0.019
	Max	0.022	0.001	0.023	0.006	0.056
Ti	Mean	2.600 a	0.970 a	0.980 a	3.100 a	2.800 a
	S.D.	0.490	0.330	0.720	3.200	0.990
	Min	2	0.740	0.490	0.550	2.100
	Median	2.700	0.970	0.640	2.200	2.300
	Max	3.100	1.200	1.800	7.600	3.900
Tl	Mean	0.180 b	0.075 c	0.079 c	0.160 b	0.280 a
	S.D.	0.023	0.005	0.011	0.023	0.017
	Min	0.160	0.071	0.070	0.150	0.260
	Median	0.170	0.075	0.077	0.150	0.290
	Max	0.210	0.079	0.092	0.190	0.290
Tm	Mean	0.004 a,b	n.d. b	0.002 b	0.003 a,b	0.013 a
	S.D.	0.002	\	0.001	0.005	0.006
	Min	0.002	\	n.d.	n.d.	0.008
	Median	0.003	\	0.002	0.002	0.010
	Max	0.007	\	0.003	0.010	0.020
U	Mean	0.490 a	0.011 a	0.170 a	0.110 a	0.390 a
	S.D.	0.260	0.001	0.110	0.150	0.200
	Min	0.120	0.011	0.091	0.007	0.270
	Median	0.580	0.011	0.120	0.058	0.280
	Max	0.690	0.012	0.290	0.330	0.620
W	Mean	0.170 a	0.065 a	0.053 a	0.093 a	0.330 a
	S.D.	0.060	0.007	0.006	0.072	0.230
	Min	0.080	0.060	0.050	0.050	0.200
	Median	0.200	0.065	0.050	0.060	0.200
	Max	0.200	0.070	0.060	0.200	0.600
V	Mean	2.200 a	0.490 a	0.750 a	2.800 a	3.500 a
	S.D.	1.500	0.120	0.400	3.500	2.700
	Min	0.830	0.400	0.460	0.220	1.700
	Median	1.900	0.490	0.590	1.400	2.100
	Max	4.300	0.570	1.200	7.900	6.600
Y	Mean	0.190 a,b	0.014 b	0.084 b	0.120 b	0.490 a
	S.D.	0.100	0.002	0.056	0.160	0.220
	Min	0.110	0.012	0.021	0.016	0.290
	Median	0.150	0.014	0.100	0.047	0.460
	Max	0.340	0.015	0.130	0.360	0.720
Yb	Mean	0.033 a	0.004 a	0.020 a	0.031 a	0.120 a
	S.D.	0.019	0.001	0.011	0.046	0.070
	Min	0.020	0.003	0.009	0.007	0.070
	Median	0.025	0.004	0.020	0.008	0.090
	Max	0.060	0.004	0.030	0.100	0.200
Zn	Mean	880 a,b	600 b,c	210 c	950 a,b	1300 a
	S.D.	180	280	33	220	210
	Min	750	400	180	640	1200
	Median	820	600	210	1000	1200
	Max	1100	800	250	1100	1600
Zr	Mean	1.500 a,b	0.200 b	1.300 a,b	0.940 a,b	3.500 a
	S.D.	0.200	0.031	1.100	1.500	1.300
	Min	1.200	0.180	0.210	0.079	2.400
	Median	1.500	0.200	1.300	0.280	3.200
	Max	1.700	0.220	2.400	3.100	4.900

n.d.: not detected.

and are expressed as mean, standard deviation, median, minimum, and maximum values for each producer. Statistical evaluation was performed using Tukey’s honestly significant difference (HSD) test at a 5% significance level, and elements showing significant differences (p < 0.05) are also reported in Table 1.

This paper explored the possibility of building a micro-scale traceability tool for Montepulciano d’Abruzzo selected producers for the first time. Based on the results reported in Table 1, several early differences between the samples can be identified. At first, it is noteworthy that producer E has the statistically significant highest content of sixteen elements, namely Al, As, Ba, Bi, Br, Ca, Cd, Co, Er, I, Lu, Na, Re, Se, Si, and Tl. The presence of these elements in the wine samples can be attributed to several factors, such as soil composition (confirmed by the next paragraph), different agricultural practices, water used for irrigation as well as environmental contamination. Undoubtedly, the wine samples from this producer were richer in these elements, distinguishing it from others. This can create a market opportunity for this producer. Some of those elements have a statistically significantly lower content compared to wine from other producers. Producer B has the lowest I content, while producer C has the lowest content of Ba. Barium is commonly recognized as a soil-derived, essential plant nutrient, with its contribution potentially linked to vine management practices (e.g., fertilization) and soil origin [33].

Furthermore, in the wine sample of producer A, the highest content of Sn was found. The Sn present in those samples is at least 100-fold the Sn present in the samples of the other producers, as well as the Sn content reported from other studies [29,30]. A high Sn content in wine is unusual and typically indicates contamination during the winemaking or storage process. Tin is not a natural component of wine and is not intentionally added. If found in significant amounts, it usually originates from metallic containers or equipment, soldering materials, packaging, or environmental contamination. The high presence of this element can pose health risks (gastrointestinal irritation and neurological issues), but it can also impact taste [34].

Besides the multi-elemental characterization, the isotope ratios of ²⁰⁸Pb/²⁰⁶Pb, ²⁰⁷Pb/²⁰⁶Pb, ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁸Pb/²⁰⁷Pb, and ⁸⁷Sr/⁸⁶Sr were measured. The analytical results and the significant differences according to Tukey HSD are reported in

Table 2. The results of the Pb and Sr isotope analysis of Montepulciano d’Abruzzo wines. Rows in the same column that are not linked by the same letter are statistically different (p-value < 0.01) by Tukey’s HSD.

		A	B	C	D	E
208Pb/206Pb	Mean	2.100 a	2.100 b,c	2.100 a,b	2.100 a,b	2.100 c
	S.D.	0.003	0.006	0.002	0.003	0.003
	Min	2.100	2.100	2.100	2.100	2.100
	Median	2.100	2.100	2.100	2.100	2.100
	Max	2.100	2.100	2.100	2.100	2.100
207Pb/206Pb	Mean	0.860 a	0.850 b	0.860 a,b	0.860 b	0.850 c
	S.D.	0.001	0.004	0.002	0.001	0.002
	Min	0.860	0.850	0.860	0.860	0.850
	Median	0.860	0.850	0.860	0.860	0.850
	Max	0.860	0.860	0.860	0.860	0.850
206Pb/204Pb	Mean	18 a	19 a	18 a	18 a	19 a
	S.D.	0.340	0.140	0.097	0.095	0.210
	Min	18	18	18	18	18
	Median	18	19	18	18	18
	Max	19	19	18	18	19
208Pb/207Pb	Mean	2.400 c	2.400 b	2.400 b,c	2.400 b	2.500 a
	S.D.	0.001	0.004	0.003	0.001	0.002
	Min	2.400	2.400	2.400	2.400	2.500
	Median	2.400	2.400	2.400	2.400	2.500
	Max	2.400	2.500	2.400	2.400	2.500
87Sr/86Sr	Mean	0.710 a	0.710 a	0.710 a	0.710 a	0.710 a
	S.D.	0	0	0	0	0
	Min	0.710	0.710	0.710	0.710	0.710
	Median	0.710	0.710	0.710	0.710	0.710
	Max	0.710	0.710	0.710	0.710	0.710

. No significant differences emerged between the samples from the analysis of isotope ratios. As is well known, the isotope ratios are recognized as powerful geographic tracers. The samples analyzed come from the same region and have the same soil type. This result does not affect the analysis; rather, it confirms that the analyzed wine samples all come from the same area.

3.2. Soil Analysis Results

Soil and wine samples were analyzed to enhance the development of a micro-scale traceability tool for Montepulciano d’Abruzzo. For this matrix, sixty-three elements (Al, As, B, Ba, Be, Bi, Br, Ca, Cd, Ce, Co, Cr, Cs, Cu, Dy, Er, Eu, Fe, Ga, Gd, Ge, Hf, Hg, Ho, I, K, La, Li, Lu, Mg, Mn, Mo, Na, Nb, Nd, Ni, P, Pb, Pr, Pt, Rb, Re, S, Sb, Sc, Se, Si, Sm, Sn, Sr, Tb, Te, Th, Ti, Tl, Tm, U, W, V, Y, Yb, Zn, and Z) were measured by HR-ICP-MS in soil samples collected from different Montepulciano d’Abruzzo producers. The analytical results are reported in Table 1 and expressed as mean, standard deviation, median, minimum, and maximum values for each producer. Tukey’s honestly significant difference test was applied to the data matrix at a 5% significance level, and statistically significant differences (p < 0.05) are reported in

Table 3. The results of the multi-elemental analysis of Montepulciano d’Abruzzo soil, expressed as µg/Kg (dry weight). Rows in the same column that are not linked by the same letter are statistically different (p-value < 0.01) by Tukey’s HSD.

		A	B	C	D	E
Al	Mean	19,150,000 b	34,350,000 b	26,050,000 b	46,820,000 a	19,150,000 b
	S.D.	5,586,143	7,107,038	5,521,503	5,573,299	1,343,502
	Min	15,200,000	28,300,000	19,100,000	39,900,000	18,200,000
	Median	19,150,000	32,650,000	28,250,000	46,250,000	19,150,000
	Max	23,100,000	43,800,000	30,700,000	56,700,000	20,100,000
As	Mean	15,500 a	13,250 a,b	11,280 b	15,830 a	12,500 a,b
	S.D.	707	500	1720	1940	707
	Min	15,000	13,000	8700	13,000	12,000
	Median	15,500	13,000	11,500	16,500	12,500
	Max	16,000	14,000	13,000	18,000	13,000
B	Mean	46,700 b,c	45,475 c	57,100 c	45,000 c	74,950 a
	S.D.	141	1345	6701	2551	3889
	Min	46,600	44,200	47,700	42,100	72,200
	Median	46,700	45,200	57,750	44,100	74,950
	Max	46,800	47,300	64,500	49,300	77,700
Ba	Mean	270,500 c,d	286,750 c	237,833 d	357,667 b	492,000 a
	S.D.	7778	14,268	26,430	28,161	4242
	Min	265,000	272,000	202,000	325,000	489,000
	Median	270,500	286,000	243,000	349,500	492,000
	Max	276,000	303,000	269,000	394,000	495,000
Be	Mean	2790 a,b	3222.5 a	2008.3 b	3701.7 a	1785 b
	S.D.	98	89	162	739	49
	Min	2720	3110	1730	3030	1750
	Median	2790	3240	2100	3415	1785
	Max	2860	3300	2120	4770	1820
Bi	Mean	400 a,b	400 a,b	300 b	467 a	250 b
	S.D.	0	0	0	103	70
	Min	400	400	300	400	200
	Median	400	400	300	400	250
	Max	400	400	300	600	300
Br	Mean	6550 a	8175 a	8417 a	6700 a	6850 a
	S.D.	1484	1381	1528	1515	919
	Min	5500	6600	7100	4600	6200
	Median	6550	8150	7900	6650	6850
	Max	7600	9800	11,000	8800	7500
Ca	Mean	96,850,000 a	62,550,000 b	81,030,000 a	43,650,000 c	104,000,000 a
	S.D.	1,909,188	8,958,236	13,653,082	4,020,820	2,828,427
	Min	95,500,000	49,500,000	66,900,000	36,100,000	102,000,000
	Median	96,850,000	65,400,000	79,200,000	44,300,000	104,000,000
	Max	98,200,000	69,900,000	97,200,000	47,800,000	106,000,000
Cd	Mean	179.5 a,b	267.25 a	189.67 b	226.83 a,b	237.5 a,b
	S.D.	14	23	42	35	9
	Min	169	237	140	186	231
	Median	179.5	268.5	191.5	216	237.5
	Max	190	295	236	278	244
Ce	Mean	36,550 a,b	34,500 b	29,600 b	41,980 a	29,450 b
	S.D.	4030	820	3217	4077	1202
	Min	33,700	33,700	24,900	38,400	28,600
	Median	36,550	34,350	30,850	39,950	29,450
	Max	39,400	35,600	32,800	47,800	30,300
Co	Mean	10,750 c	13,150 b	10,835 c	13,967 b	16,350 a
	S.D.	212	685	728	758	1202
	Min	10,600	12,600	9910	13,100	15,500
	Median	10,750	12,950	11,050	13,800	16,350
	Max	10,900	14,100	11,700	14,900	17,200
Cr	Mean	66,500 b	88,400 a	76,133 b	79,133 a,b	89,300 a
	S.D.	282	6422	5562	4449	2262
	Min	66,300	81,800	70,000	74,300	87,700
	Median	66,500	87,450	74,400	77,600	89,300
	Max	66,700	96,900	85,000	85,300	90,900
Cs	Mean	3995 a,b	4830 a	2062 b	3493 a,b	2435 b
	S.D.	615	619	271	1327	275
	Min	3560	4110	1870	1930	2240
	Median	3995	4815	1975	3430	2435
	Max	4430	5580	2600	5240	2630
Cu	Mean	75,400 b	188,500 a	38,600 b	153,000 a	39,700 b
	S.D.	1697	17,000	9597	34,298	1555
	Min	74,200	172,000	30,400	115,000	38,600
	Median	75,400	188,000	34,450	144,500	39,700
	Max	76,600	206,000	51,700	199,000	40,800
Dy	Mean	2000 a	2000 a	2000 a	2000 a	2000 a
	S.D.	0	0	0	0	0
	Min	2000	2000	2000	2000	2000
	Median	2000	2000	2000	2000	2000
	Max	2000	2000	2000	2000	2000
Er	Mean	1000 a,b	925 b	1000 a	967 a,b	1000 a,b
	S.D.	0	50	0	51	0
	Min	1000	900	1000	900	1000
	Median	1000	900	1000	1000	1000
	Max	1000	1000	1000	1000	1000
Eu	Mean	700 a,b	725 a	650 a,b	683 a,b	600 b
	S.D.	0	50	54.	40	0
	Min	700	700	600	600	600
	Median	700	700	650	700	600
	Max	700	800	700	700	600
Fe	Mean	26,550,000 c	31,425,000 a,b	28,100,000 c	33,370,000 a	29,750,000 b,c
	S.D.	636,396	139,6125	916,515	1,932,528	777,817
	Min	26,100,000	30,000,000	26,900,000	30,300,000	29,200,000
	Median	26,550,000	31,300,000	28,200,000	33,350,000	29,750,000
	Max	27,000,000	33,100,000	29,200,000	35,700,000	30,300,000
Ga	Mean	11,000 c	13,500 a,b	12,500 b,c	14,500 a	11,500 b,c
	S.D.	0	1290	547	836	707
	Min	11,000	12,000	12,000	13,000	11,000
	Median	11,000	13,500	12,500	15,000	11,500
	Max	11,000	15,000	13,000	15,000	12,000
Gd	Mean	3050 a,b	2825 a,b	2683 b	3033 a	2600 a,b
	S.D.	353	125	75	265	141
	Min	2800	2700	2600	2700	2500
	Median	3050	2800	2700	3000	2600
	Max	3300	3000	2800	3500	2700
Ge	Mean	600 a	600 a	617 a	650 a	600 a
	S.D.	141	81	75	54	0
	Min	500	500	500	600	600
	Median	600	600	600	650	600
	Max	700	700	700	700	600
Hf	Mean	850 a,b	975 a,b	567 b	1330 a	550 a,b
	S.D.	212	50	233	516	70
	Min	700	900	300	1000	500
	Median	850	1000	600	1000	550
	Max	1000	1000	900	2000	600
Hg	Mean	25.5 b	45.25 a	23.7 b	42.7 a	17.5 b
	S.D.	0.7	7	6	5.	0.7
	Min	25	36	18	35	17
	Median	25.5	45.5	20	44	17.5
	Max	26	54	34	48	18
Ho	Mean	460 a	340 b	430 a	370 b	460 a
	S.D.	42	18	16	32	28
	Min	430	330	400	330	440
	Median	460	335	435	370	460
	Max	490	370	440	410	480
I	Mean	4000 a,b	4750 a	2830 b	4830 a	3000 b
	S.D.	0	500	752	752	0
	Min	4000	4000	2000	4000	3000
	Median	4000	5000	3000	5000	3000
	Max	4000	5000	4000	6000	3000
K	Mean	15,450,000 a,b	15,975,000 a,b	16,000,000 b	18,033,000 a	13,750,000 b
	S.D.	353,553	1,144,188	1,248,999	956,382	777,817
	Min	15,200,000	15,000,000	14,300,000	17,000,000	13,200,000
	Median	15,450,000	15,700,000	15,950,000	17,900,000	13,750,000
	Max	15,700,000	17,500,000	17,500,000	19,400,000	14,300,000
La	Mean	18,000 a,b	16,000 b	15,000 b	18,000 a	15,000 b
	S.D.	1414	0	1505	1722	0
	Min	17,000	16,000	13,000	17,000	15,000
	Median	18,000	16,000	15,000	18,000	15,000
	Max	19,000	16,000	17,000	21,000	15,000
Li	Mean	42,950 b	51,325 a,b	45,000 b	57,817 a	54,200 a
	S.D.	919	2139	2547	4593	3111
	Min	42,300	48,500	41,600	52,000	52,000
	Median	42,950	51,550	45,550	58,600	54,200
	Max	43,600	53,700	47,900	64,300	56,400
Lu	Mean	89.5 b	132.5 a	74.3 b	120 a	74 b
	S.D.	14	5	10	10	5
	Min	79	130	55	100	70
	Median	89.5	130	78	120	74
	Max	100	140	85	130	78
Mg	Mean	12,200,000 a,b	6,572,500 c	11,640,000 b	5,855,000 c	15,650,000 a
	S.D.	1,697,056	537,672	1,933,297	457,416	1,202,081
	Min	11,000,000	6,060,000	9,470,000	5,010,000	14,800,000
	Median	12,200,000	6,450,000	11,600,000	5,945,000	15,650,000
	Max	13,400,000	7,330,000	14,100,000	6,380,000	16,500,000
Mn	Mean	604,000 a,b	652,250 a,b	585,500 b	570,170 b	800,500 a
	S.D.	14,142	34,422	49,184	109,071	9192
	Min	594,000	615,000	519,000	410,000	794,000
	Median	604,000	655,500	583,500	605,500	800,500
	Max	614,000	683,000	647,000	693,000	807,000
Mo	Mean	1009.5 c	987.5 c	1490 b	1345 b,c	4380 a
	S.D.	85	54	259	153	254
	Min	949	913	1190	1130	4200
	Median	1009.5	998.5	1465	1375	4380
	Max	1070	1040	1820	1570	4560
Na	Mean	4,400,000 b,c	5,990,000 a	6,008,000 a	5,763,000 a,b	2,685,000 c
	S.D.	113,137	535,038	792,575	362,086	7071
	Min	4,320,000	5,420,000	5,400,000	5,280,000	2,680,000
	Median	4,400,000	5,960,000	5,655,000	5,725,000	2,685,000
	Max	4,480,000	6,620,000	7,400,000	6,250,000	2,690,000
Nb	Mean	9400 b	10,225 b	8300 b	13,167 a	9500 b
	S.D.	141	518	787	1940	707
	Min	9300	9900	7300	11,000	9000
	Median	9400	10,000	8500	13,500	9500
	Max	9500	11,000	9100	16,000	10,000
Nd	Mean	13,500 a	12,750 a	12,830 a	15,000 a	12,500 a
	S.D.	707	500	1722	1095	2121
	Min	13,000	12,000	11,000	14,000	11,000
	Median	13,500	13,000	12,500	15,000	12,500
	Max	14,000	13,000	16,000	17,000	14,000
Ni	Mean	38,000 c	50,100 a,b	44,700 b,c	48,100 b	58,100 a
	S.D.	707	2186	3188	3819	3818
	Min	37,500	47,900	41,900	42,800	55,400
	Median	38,000	50,050	43,450	47,600	58,100
	Max	38,500	52,400	49,400	54,500	60,800
P	Mean	939,500 a,b	1,147,500 a	835,830 b	919,670 a,b	658,500 b
	S.D.	10,606	47,871	153,639	142,702	17,677
	Min	932,000	1,080,000	639,000	779,000	646,000
	Median	939,500	1,160,000	916,000	853,000	658,500
	Max	947,000	1,190,000	955,000	1,110,000	671,000
Pb	Mean	21,200 a,b,c.	25,025 a,b	17,600 c	27,417 a	17,800 b,c
	S.D.	989	1567	3059	4853	1555
	Min	20,500	23,100	13,100	23,600	16,700
	Median	21,200	25,100	19,250	24,900	17,800
	Max	21,900	26,800	20,000	34,800	18,900
Pr	Mean	3655 a,b	3437 a,b	3242 b	4018 a	2990 b
	S.D.	346	93	400	323	84
	Min	3410	3300	2740	3680	2930
	Median	3655	3475	3300	3975	2990
	Max	3900	3500	3880	4580	3050
Pt	Mean	4.80 b,c	7.10 a,b	3.18 c	8.88 a	3.95 b,c
	S.D.	1.28	0.83	1.4	1.5	0.3
	Min	3.9	6	0.98	6.6	3.7
	Median	4.8	7.2	3.15	8.75	3.95
	Max	5.7	8	4.9	11	4.2
Rb	Mean	61,450 a,b	80,125 a	50,133 b	70,383 a	43,100 b
	S.D.	7566	5483	5011	12,459	1697
	Min	56,100	74,400	44,000	53,900	41,900
	Median	61,450	79,500	50,750	69,250	43,100
	Max	66,800	87,100	56,300	85,000	44,300
Re	Mean	0.30 a,b	0.30 b	0.67 a	0.32 b	0.55 a,b
	S.D.	0.14	0.21	0.21	0.09	0.21
	Min	0.2	0.1	0.4	0.2	0.4
	Median	0.3	0.25	0.65	0.3	0.55
	Max	0.4	0.6	1	0.5	0.7
S	Mean	303,000 c	451,000 b	229,700 c	239,500 c	636,500 a
	S.D.	5656	57,486	25,144	37,292	40,305
	Min	299,000	398,000	203,000	177,000	608,000
	Median	303,000	441,500	223,000	246,000	636,500
	Max	307,000	523,000	275,000	281,000	665,000
Sb	Mean	722.5 b,c	840 a,b	574 c	952.5 a	748.5 b,c
	S.D.	2	30	45	122	74
	Min	721	796	505	825	696
	Median	722.5	850	593	917.5	748.5
	Max	724	864	612	1160	801
Sc	Mean	4500 a,b	4250 b	3830 b	5830 a	4000 a,b
	S.D.	70	1500	408	408	0
	Min	4000	3000	3000	5000	4000
	Median	4500	4000	4000	6000	4000
	Max	5000	6000	4000	6000	4000
Se	Mean	1500 a	1750 a	2000 a	2000 a	2000 a
	S.D.	707	500	632	0	0
	Min	1000	1000	1000	2000	2000
	Median	1500	2000	2000	2000	2000
	Max	2000	2000	3000	2000	2000
Si	Mean	201,500,000 c	240,750,000 a,b	230,170,000 b	249,830,000 a	205,500,000 c
	S.D.	707,106	6,291,528	12,089,940	10,703,581	707,106
	Min	201,000,000	235,000,000	211,000,000	240,000,000	205,000,000
	Median	201,500,000	240,000,000	236,500,000	247,000,000	205,500,000
	Max	202,000,000	248,000,000	240,000,000	266,000,000	206,000,000
Sm	Mean	4000 a,b	3000 b	3800 a,b	4000 a	3000 a,b
	S.D.	0	0	752	0	0
	Min	4000	3000	3000	4000	3000
	Median	4000	3000	4000	4000	3000
	Max	4000	3000	5000	4000	3000
Sn	Mean	2425 b	3080 a	2533 b	3387 a	2090 b
	S.D.	49	75	129	305	14
	Min	2390	3000	2320	3000	2080
	Median	2425	3075	2545	3345	2090
	Max	2460	3170	2670	3910	2100
Sr	Mean	207,000 b	152,000 c	236,000 b	158,000 c	644,000 a
	S.D.	7071	7615	28,798	12,155	8485
	Min	202,000	146,000	195,000	144,000	638,000
	Median	207,000	149,500	247,000	155,500	644,000
	Max	212,000	163,000	262,000	176,000	650,000
Tb	Mean	500 a	450 a	417 a	433 a	400 a
	S.D.	0	57	40	51	0
	Min	500	400	400	400	400
	Median	500	450	400	400	400
	Max	500	500	500	500	400
Te	Mean	55.5 c	106 a,b	75.5 b,c	112 a	81 a,b,c
	S.D.	6	11	18	14	25
	Min	51	94	49	92	63
	Median	55.5	105	76	115	81
	Max	60	120	98	130	99
Th	Mean	3440 a,b	3237 a,b	3163 b	5302 a	2420 a,b
	S.D.	2446	1119	1221	75	197
	Min	1710	1990	1740	4500	2280
	Median	3440	3265	3320	5210	2420
	Max	5170	4430	4830	6590	2560
Ti	Mean	1,850,000 c	2,250,000 a,b	2,080,000 a	2,430,000 a	1,900,000 b,c
	S.D.	70,710	129,099	160,208	150,554	141,421
	Min	1,800,000	2,100,000	1,900,000	2,200,000	1,800,000
	Median	1,850,000	2,250,000	2,100,000	2,500,000	1,900,000
	Max	1,900,000	2,400,000	2,300,000	2,600,000	2,000,000
Tl	Mean	785 a,bc	824 a,b	504 c	1004 a	549 b,c
	S.D.	48	31	51	222	38
	Min	751	790	438	793	522
	Median	785	823	501.5	926.5	549
	Max	819	861	571	1320	576
Tm	Mean	150 a	125 a	117 a	117 a	100 a
	S.D.	70	50	40	40	0
	Min	100	100	100	100	100
	Median	150	100	100	100	100
	Max	200	200	200	200	100
U	Mean	1765 b	1387 c	1773 b	1663 b	2330 a
	S.D.	21	43	116	158	98
	Min	1750	1350	1580	1500	2260
	Median	1765	1375	1825	1630	2330
	Max	1780	1450	1870	1900	2400
W	Mean	2000 a,b,c	2000 a,c	1000 b	2500 a	1000 b,c
	S.D.	0	0	0	836	0
	Min	2000	2000	1000	2000	1000
	Median	2000	2000	1000	2000	1000
	Max	2000	2000	1000	4000	1000
V	Mean	85,500 b	95,250 a,b	86,830 b	100,830 a	100,000 a,b
	S.D.	707	3774	6177	7600	0
	Min	85,000	91,000	80,000	92,000	100,000
	Median	85,500	95,000	86,000	99,000	100,000
	Max	86,000	100,000	96,000	110,000	100,000
Y	Mean	9450 a,b	6350 b	9330 a,b	10,500 a	10,750 a,b
	S.D.	5020	1588	1670	1676	1767
	Min	5900	4600	7500	8400	9500
	Median	9450	6350	9300	10,700	10,750
	Max	13,000	8100	12,000	12,000	12,000
Yb	Mean	650 b	1000 a	600 b	900 a	650 b
	S.D.	70	0	63	89	70
	Min	600	1000	500	800	600
	Median	650	1000	600	900	650
	Max	700	1000	700	1000	700
Zn	Mean	74,000 b	80,620 a,b	75,430 b	92,070 a	97,500 a
	S.D.	989	10,427	3744	6487	1697
	Min	73,300	65,200	72,000	81,700	96,300
	Median	74,000	84,800	74,100	92,150	97,500
	Max	74,700	87,700	82,400	102,000	98,700
Zr	Mean	39,400 a,b	44,620 a,b	21,330 b	68,230 a	21,950 b
	S.D.	12,445	2783	7961	22,985	1060
	Min	30,600	40,600	10,900	44,200	21,200
	Median	39,400	45,450	24,850	61,800	21,950
	Max	48,200	47,000	30,200	101,000	22,700

.

From the multi-element analysis of soil samples, it is possible to infer some findings. At first, in this case, the soil from producer E resulted in a higher content of certain elements. Indeed, producer E’s samples showed the statistically significant highest content of B, Ba, Co, Mo, and S. It is noteworthy that Ba and Co were also found to have the highest content in the wine samples of producer E. So, the high presence of these two elements in wine is related to their high presence in soil. Among these elements, barium is the only non-essential nutrient for plants; instead, high Ba levels can interfere with plant uptake of essential elements like potassium and calcium. Cobalt is a trace element required in very small amounts by plants, and its high level can be toxic, affecting enzyme function.

Boron and Mo are essential micronutrient for plants, being involved, respectively, in cell wall strength and reproductive development (B) and nitrogen metabolism and enzyme activity (Mo). On the other hand, sulfur is a macronutrient that plants use for protein synthesis and chlorophyll formation. Nevertheless, high S levels may lead to soil acidification over time, affecting overall soil health. Moreover, it emerged that producer C had the lowest content of Ca, a macronutrient for plants. However, the levels of calcium in the analyzed wine were not different, indicating that the plant did not suffer from a shortage of this nutrient. Producer D, instead, presented the highest content of Nb. Niobium is a relatively rare element in soil, and its presence typically originates from natural sources such as weathering of Nb-rich minerals (e.g., pyrochlore, columbite, and euxenite) or anthropogenic activities like mining, industrial waste disposal, and agricultural practices. In this case, the Nb levels in wine samples were comparable, so its high content in the soil of producer D producer do not affect the systems but rather can be an added value for that producer since Nb is a rare earth element.

In the soil samples, isotope ratios of ²⁰⁸Pb/²⁰⁶Pb, ²⁰⁷Pb/²⁰⁶Pb, ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁸Pb/²⁰⁷Pb, and ⁸⁷Sr/⁸⁶Sr were measured. The analytical results, as well as the significant differences by Tukey’s HSD, are reported in

Table 4. The results of the Pb and Sr isotope analysis of Montepulciano d’Abruzzo soil. Rows in the same column not linked by the same letter are statistically different (p-value < 0.01) by Tukey’s HSD.

		A	B	C	D	E
208Pb/206Pb	Mean	2.069 a,b	2.0725 a	2.067 b	2.069 b	2.067 b
	S.D.	0	0.5773	0.002	0.002	0
	Min	2.069	2.072	2.065	2.067	2.067
	Median	2.069	2.0725	2.067	2.0685	2.067
	Max	2.069	2.073	2.069	2.072	2.067
207Pb/206Pb	Mean	0.8338 b	0.8362 a	0.8343 b	0.8342 b	0.8339 b
	S.D.	0	0.1414	0.0010	0.7501	0.212
	Min	0.8338	0.836	0.8334	0.8331	0.8338
	Median	0.8338	0.83625	0.8339	0.8344	0.83395
	Max	0.8338	0.8363	0.8356	0.8348	0.8341
206Pb/204Pb	Mean	18.85 a	18.79 a	18.86 a	18.84 a	18.84 a
	S.D.	0.01	0.01	0.05	0.02	0.03
	Min	18.85	18.78	18.79	18.82	18.82
	Median	18.855	18.79	18.88	18.84	18.84
	Max	18.86	18.8	18.9	18.88	18.86
208Pb/207Pb	Mean	2.481 a	2.479 a,b	2.478 b	2.481 a	2.479 a,b
	S.D.	0	05	0.001	0.983	0
	Min	2.481	2.478	2.476	2.479	2.479
	Median	2.481	2.479	2.478	2.4805	2.479
	Max	2.481	2.479	2.479	2.481	2.479
87Sr/86Sr	Mean	0.7092 b,c	0.7096 a	0.7092 c	0.7094 b	0.7090 c
	S.D.	0	0	0.0752	0.1549	0
	Min	0.7092	0.7096	0.709	0.7092	0.709
	Median	0.7092	0.7096	0.7091	0.7095	0.709
	Max	0.7092	0.7096	0.7092	0.7095	0.709

. In this case, no significant differences emerged between the samples based on the analysis of isotope ratios. This evidence confirms the geographical recognition of samples to a single area. The only exception was found for ²⁰⁷Pb/²⁰⁶Pb of producer B, which has a significantly higher value compared to the others. Despite this statistically difference, the value is not sufficiently different from the others to justify a hypothesis of varied geographical recognition. In conclusion, while soil composition represents an important contextual factor, its direct influence on wine elemental profiles could not be quantitatively established within the scope of the present study.

3.3. Chemometric Results

The analytical results of isotope and elemental determinations have revealed distinct markers for a single Montepulciano Abruzzo producer. Various chemometric tools were employed on the data matrix to enhance characterization and develop classification models for distinguishing each producer’s wine and soil samples.

The first chemometric application utilized cluster analysis. The dataset exhibited a strong intrinsic cluster structure, as confirmed by Hopkins’ statistic (H = 0.700198 for wine samples and H = 0.664036) and VAT (Figures S2 and S3).

For wine samples, K-means and hierarchical clustering identified two clusters (Table S3). In cluster 1, all samples from wineries A, B, C, and D were collocated, while cluster 2 comprised all samples of winery E. These outcomes are in good agreement with the analytical results showing that winery E has the most identifiable chemical pattern.

A step forward was reached by K-means and hierarchical clustering, categorizing soil samples into four clusters (Table S3). Wineries B, D, and E have their own cluster, while wineries A and C are recognized as a unique cluster.

Therefore, the clustering analysis indicates that the dataset, consisting of elemental and isotope analysis results, is well suited for classifying samples.

Principal Component Analysis (PCA) was performed to highlight the natural grouping of samples. For wine data, three outliers (one from producer A and two from producer E) have been identified by the T2 (Mahalanobis distance), quantile range, and Cauchy’s distribution methods and were removed from the dataset. The same procedure was applied to soil data, and one outlier (from producer B) was identified and removed. The scores and loading plots of the unsupervised PCA are reported in Figure 1A for wine and 1B for soil. Autoscaling pre-treatment was carried out on the dataset to exclude the variance related to the different measurement units.

For wine samples, the first two PCs describe 64.6% of the total variance, highlighting early grouping of samples. Producer E’s samples are all located in the right quadrant of the scores plot, well separated from the others. Likewise, for the other producers, partial separation occurs, with the only exception being for producer D whose samples are confused with those of other producers.

The PCA of soil samples gives similar results. The first two PCs explain 58.9% of the total variance. Also, in this case, the samples from producer E appear well separated from the others (upper left quadrant of score plot). In this case, the other producers appear overlapped in pairs (producers D and B; producers A and C).

The exploratory data analysis of the content of sixty-three elements in soil and wine revealed partial grouping of samples based on different producers. This suggests the potential for classifying samples according to their producers. The combined consideration of soil and wine datasets provides contextual information supporting sample differentiation, although no direct quantitative soil–wine transfer relationships were demonstrated, except for Ba and producer E. Consequently, the datasets for soil and wine were combined, and the merged dataset was subjected to multivariate analysis to develop classification models for identifying individual producers.

At first, variable selection was necessary to select the most informative variables for producer recognition. Two methods of variable selection were applied: stepwise and decisional trees. The results of the selection processes are reported in Table S4 for stepwise and Table S5 for decisional trees. To establish a robust selection process, variables were selected by using both methods. Only four variables were included from the sixty-three elements analyzed, i.e., Mo, ²⁰⁸Pb/²⁰⁶Pb, P, and ⁸⁷Sr/⁸⁶Sr, and they were used in the subsequent steps.

Discriminant methods and machine learning methods were applied to build a geographic recognition model based on wine and soil element analyses: linear discriminant analysis (LDA), Quadratic Discriminant Analysis (QDA), Artificial Neural Network (ANN), Random Forest (RF), Naïve Bayes (NB), and Support Vector Machine (SMV).

Table 5. Percentage of samples correctly classified for producers.

	A	B	C	D	E
LDA	67%	67%	89%	60%	60%
QDA	100%	100%	100%	100%	100%
ANN	100%	100%	100%	100%	100%
RF	100%	83%	100%	100%	100%
NB	67%	67%	89%	60%	100%
SVM	100%	75%	83%	50%	67%

reports the correct classification rates of soil–wine samples according to producers using the six methods described. Validation was carried out using the Leave-One-Out approach. Figures of merit of classification methods are reported in Table S6.

From the results, it is immediately evident that producer C was generally classified with high accuracy by all methods, suggesting that the samples belonging to this producer are relatively distinct and easily separable. Producers D and E, however, proved more difficult to classify, particularly for the simpler, linear methods such as LDA and NB, which implies that their data might overlap with other producers or contain greater internal variability. Producers A and B showed moderate performance with the simpler models but achieved perfect classification with more complex methods, reinforcing the idea that nonlinear techniques provide a substantial advantage. Overall, the results indicate that nonlinear and flexible models, such as QDA, ANN, and Random Forest, are far more effective for this dataset than linear methods like LDA or probabilistic ones like Naïve Bayes.

In conclusion, QDA, ANN, and RF emerge as the most reliable methods for classifying producers in this dataset, achieving superior predictive performance and demonstrating strong adaptability to complex data structures. Notably, QDA and ANN attained 100% classification accuracy, which can be attributed to their ability to model nonlinear relationships and class-specific decision boundaries. Unlike LDA, which assumes linear separability and a common covariance structure across classes, QDA relaxes these constraints by allowing each class to have its own covariance matrix, enabling more flexible, nonlinear decision surfaces that better reflect the intrinsic variability in the data.

Similarly, ANN and RF are inherently capable of capturing higher-order interactions and nonlinear patterns among predictors without requiring strong distributional assumptions. This suggests that the relationships between the input variables and producer classes are not strictly linear, but instead involve complex dependencies that linear models such as LDA are unable to represent effectively. While SVM demonstrated promising performance, its results indicate that additional kernel tuning or parameter optimization may be required to fully exploit nonlinear structures present in the dataset. In contrast, the comparatively weaker performance of LDA and Naïve Bayes highlights the limitations of methods that rely on linear boundaries or strong independence assumptions. Overall, these findings reinforce the conclusion that the underlying data structure is predominantly nonlinear, and that classification methods designed to capture such complexity provide the most accurate and robust results.

Moreover, our findings align well with other studies that integrate soil and wine data prior to performing chemometric analyses. For instance, the authors of [35] achieved an 84.82% correct classification of their samples, though they relied solely on linear discriminant analysis (LDA) and focused on samples from a single Italian region. A more recent study by [30], which involved samples from four different regions, reported classification accuracies ranging from 38.5% to 95.85%.

Importantly, employing diverse classification methods enables effective sample differentiation based on their producer. As previously demonstrated [29,36,37], integrating multiple classification techniques with multi-elemental analysis has proven to be a robust approach to traceability and micro-traceability recognition.

3.4. Managerial Implications

The managerial implications of this study are highly relevant to the Montepulciano d’Abruzzo PDO chain, a complex ecosystem involving growers, wineries, cooperatives, and regulatory bodies. This study demonstrates measurable compositional differences among producers and confirms the feasibility of micro-scale traceability, with important consequences for governance and market competitiveness. The dual structure of the chain—combining a few independent estates with a large cooperative sector—creates challenges for vineyard-level traceability, particularly where blending practices dilute geographic specificity. The analytical framework developed here offers consortia and policymakers a scientific tool to strengthen traceability governance, even in complex supply chains. Incorporating multi-elemental and isotopic profiling into control protocols could complement documentary audits, reducing fraud risk while enhancing transparency and consumer confidence in the Montepulciano d’Abruzzo PDO, especially in international markets.

Despite its strong reputation, Montepulciano d’Abruzzo suffers from limited product differentiation. The identification of four key elemental and isotopic markers (Mo, 208Pb/206Pb, P, and 87Sr/86Sr) enables producers to scientifically document individual terroir signatures. This evidence can support marketing and brand storytelling, allowing wineries to highlight geochemical distinctiveness linked to specific sub-areas and to pursue premiumization strategies in higher-value market segments where authenticity is a key driver.

At the policy level, the findings suggest that analytical traceability could support territorial zoning and product segmentation within the PDO framework. Mapping geochemical profiles across sub-regions may enable the identification of micro-terroirs and the development of additional geographical mentions, helping to counteract category homogenization. More broadly, integrating traceability science into regional policy could support rural development, environmental monitoring, and agri-food innovation, positioning the Montepulciano d’Abruzzo PDO as a model of knowledge-based territorial governance.

4. Conclusions

The present study demonstrates that the integration of multi-elemental and isotopic profiling with advanced chemometric analysis provides a powerful and reliable approach to establishing micro-traceability within the Montepulciano d’Abruzzo PDO wine chain. Through comprehensive characterization of both wine and soil matrices, the research successfully identified compositional patterns and discriminant markers capable of distinguishing individual producers within a single PDO territory. The combination of HR-ICP-MS and MC-ICP-MS analyses, together with multivariate classification methods, revealed that Quadratic Discriminant Analysis (QDA) and Artificial Neural Networks (ANNs) achieved complete classification within the available dataset; however, this performance should be interpreted cautiously due to the limited sample size, the use of Leave-One-Out cross-validation, and the exploratory nature of the modeling approach. From a broader perspective, this study contributes to both scientific innovation and managerial practice. Scientifically, it extends the application of elemental and isotopic fingerprinting to an underexplored PDO, offering a methodological template for future research on terroir differentiation. Managerially, it provides actionable insights for strengthening quality control, authenticity verification, and brand reputation. This study underscores the potential of data-driven traceability systems to support evidence-based governance of PDO products, bridging the gap between laboratory science and real-world policy implementation.

By translating analytical data into strategic value, the research reinforces the role of traceability as a managerial instrument—one that supports sustainable competitiveness, consumer trust, and the long-term valorization of regional heritage. Future studies should expand the sample set across different vintages, include temporal variability and climate factors, and test the scalability of the classification model across other Italian PDOs. Such developments could lead to the creation of a national digital traceability infrastructure, connecting producers, regulators, and consumers through transparent, scientifically grounded information systems.

Ultimately, this work illustrates that the intersection of analytical chemistry, data science, and managerial strategy offers not only a tool for verifying authenticity but also a transformative framework for guiding the evolution of the Italian PDO wine sector toward innovation, transparency, and resilience.