Revista de la Facultad de Ciencias
Agrarias. Universidad Nacional de Cuyo. Tomo 57(2). ISSN (en línea) 1853-8665.
Año 2025.
Original article
Data-driven
Method for the Delimitation of Viticultural Zones: Application in the Mendoza
River Oasis, Argentina
Método
basado en datos para la delimitación de zonas vitivinícolas: Aplicación en el
oasis río Mendoza, Argentina
Rosana Vallone3,
Pablo Paccioretti1,
Francisco Corvalán3,
1Universidad Nacional de Córdoba. Facultad de Ciencias
Agropecuarias. Ing Agr. Félix Aldo Marrone 746. Ciudad Universitaria. X5000HUA.
Córdoba. Argentina.
2Unidad de Fitopatología y Modelización Agrícola (UFyMA),
Instituto Nacional de Tecnología Agropecuaria. Consejo Nacional de
Investigaciones Científicas y Técnicas. Av. 11 de Septiembre 4755, X5020ICA. Córdoba.
Argentina.
3Universidad Nacional de Cuyo. Facultad de Ciencias Agrarias.
Almirante Brown 500. Chacras de Coria. M5528AHB. Mendoza. Argentina
*mariano.cordoba@unc.edu.ar
Abstract
In viticulture,
understanding the spatial variability of natural factors influencing vineyard
potential is essential for terroir characterization. In the present study, we
present a data-driven protocol that integrates climate, geomorphology, and soil
data to delineate viticultural zones. The method combines spatial layers with
statistical tools to partition a region into areas with similar
characteristics. The protocol comprises: 1) rescaling multiple spatial data
layers, 2) applying spatial multivariate analysis to group spatial units, and
3) using machine-learning algorithms to identify key zoning drivers. The
approach was applied to the Mendoza River oasis in Argentina. Climate and
geomorphology layers were used first, as they varied at a broader spatial scale
than soil data. Two climatic zones were identified, mainly differentiated by
elevation and thermal indices. Subsequent soil-based zoning within each
climatic zone revealed five distinct edaphoclimatic zones. These zones showed
statistically significant differences in environmental variables and exhibited
spatial coherence aligned with landscape features. Results showed that this
protocol facilitates the integration of diverse data sources and supports a
deeper understanding of the uniqueness of vineyard zones in wine-producing
regions.
Keywords: spatial clustering,
edaphoclimatic zoning, zoning drivers
Resumen
En viticultura,
comprender la variabilidad espacial de los factores naturales que influyen en
el potencial vitivinícola es esencial para caracterizar el terroir. En este
estudio presentamos un protocolo basado en datos de clima, geomorfología y
suelo para delimitar zonas vitivinícolas. El método integra capas de datos
espaciales con herramientas estadísticas para subdividir una región en áreas
con características similares. La metodología comprende: 1) el reescalado de
múltiples capas de datos espaciales, 2) el análisis multivariado espacial para
agrupar unidades espaciales, y 3) la aplicación de algoritmos de aprendizaje
automático para identificar las principales variables determinantes de la
zonificación. El protocolo se aplicó al oasis del río Mendoza, en Argentina.
Primero se utilizaron los datos climáticos y geomorfológicos, que mostraban
variabilidad a una escala espacial mayor, identificándose dos zonas climáticas
diferenciadas principalmente por la altitud y los índices térmicos.
Posteriormente, dentro de cada zona climática, se realizó una zonificación
adicional basada en propiedades de suelo, lo que permitió identificar cinco zonas
edafoclimáticas. Estas zonas presentaron diferencias estadísticamente
significativas en variables ambientales y una alta coherencia espacial en
correspondencia con las características del paisaje. Este enfoque permite
integrar datos diversos y contribuye a lograr una comprensión más profunda de
los ambientes vitícolas en regiones productoras de vino.
Palabras clave: agrupamiento
espacial, zonificación edafoclimática, importancia de variables en zonificación
Originales: Recepción: 17/06/2025- Aceptación: 16/10/2025
Introduction
Viticulture is one
of the most widespread horticultural activities worldwide, with wine grapes
cultivated across diverse climates and landscapes. Sustainable viticulture
requires integrating large volumes of soil and climate data to support vineyard
management and long-term planning (Visconti et al., 2024). In Argentina,
vineyards extend across multiple mesoclimates, landscapes, and soils. National
grape and wine production is led by Mendoza province, in the central west of
the country, accounting for nearly 70% of the total (INV,
2024).
Viticulture in Mendoza province is concentrated in four main oases located
between 33° and 36° S. This arid to semi-arid region depends on irrigation from
Andean meltwater and shows high variability in climate and soils (Puscama
et al., 2025; Straffelini et al., 2023). Recently, the
Argentine Viticultural Corporation (COVIAR) conducted a comprehensive soil and
climate survey of wine-producing regions, fostering a deeper understanding of
natural variability and its influence on productivity and regional differences
in grape production. The concept of terroir -used to explain differences in
wine style and quality- is fundamentally geographical and supports spatial
analysis of edaphoclimatic influences (Van Leeuwen et al., 2010). This concept is
also recognized as a socio-ecological construct, shaped not only by natural
conditions but also by cultural, technical, and economic factors (Vaudour
et al., 2010). Terroir is multifactorial, with climate, topography, and soil
as its main environmental pillars (Van Leeuwen et al., 2010;
Vaudour et al., 2010). Climatic factors like temperature, radiation, precipitation,
and thermal amplitude influence grape phenology and sugar accumulation (Jones et
al., 2005). Topographic or geomorphometric factors -particularly
elevation, slope, and aspect- influence climate, water dynamics, and solar
exposure (Hall
& Jones, 2010; Irimia et al., 2014). Soils regulate water and nutrient
supply, shaping vegetative growth, yield, and grape composition (Morlat
& Bodin, 2006). The interaction among these factors drives spatial
heterogeneity in vineyard performance and wine typicity, underscoring the need
to integrate them into zoning studies. When referring specifically to areas
sharing similar biophysical features that influence vine development and grape
composition, the term “edaphoclimatic zones” is commonly used. Advances in
proximal and remote sensing technologies now allow the acquisition of large
volumes of data, which in turn permits delimiting these edaphoclimatic zones.
Combined with geostatistical and machine-learning methods, these data are used
to develop digital maps of biophysical variables relevant to viticulture (Ferro
& Catania, 2023). Nevertheless, many terroir studies still rely on expert-based
assessments rather than systematic, data-driven approaches (Bramley
et al., 2020). In recent years, data-driven methods have emerged as powerful
tools to analyze terroir, enabling subregional classifications and explaining
within-region variations in grape quality and vineyard performance (Bramley
et al., 2023; Bramley & Gardiner, 2021).
Data-driven
delineation of edaphoclimatic zones requires: 1) statistical methods that
integrate the spatial distribution of multiple variables, and 2) clustering
algorithms that group sites with similar attributes and identify the key
drivers defining each zone. The success of such zoning depends on the spatial
resolution, type, and quality of input data (Van Leeuwen et al., 2010). Although several
zoning approaches exist (Ghilardi et al., 2023), few address the
dual challenge of handling high-dimensional environmental datasets while
accounting for the spatial autocorrelation typical of edaphoclimatic variables.
Many studies have classified viticultural areas by using soil, climate, and
topographic variables (Ferretti, 2020; Ghilardi et al., 2023), yet most rely on
conventional multivariate techniques that do not explicitly incorporate spatial
structure. Ignoring spatial dependence can result in fragmented zones or biased
interpretations of terroir. Integrating spatially explicit methods -such as
spatial principal components or geostatistical models- remains limited but
offers a promising path toward more robust and geographically coherent zoning.
Thus, this study describes a data-driven approach to delineate edaphoclimatic
zones in a viticultural region by integrating multiple environmental variables
across spatial scales. In this approach, spatial principal components are first
computed and used in a hierarchical clustering process to delineate zones.
Because soil variables typically exhibit greater spatial heterogeneity than
climatic and geomorphometric variables, we adopted a nested clustering
strategy: initial zoning was based on climate and geomorphology -variables
occurring at a broader spatial scale- followed by refinement with soil data to
capture finer-scale variability.
Spatial principal
components were applied to ensure that spatial autocorrelation was explicitly
considered in the clustering process (Córdoba et al., 2012). Although the
loadings of input variables on each component provide insights into spatial
correlations, they do not always identify the main drivers of zoning-
particularly when multiple correlated factors shape the results (Jolliffe
& Cadima, 2016). For this reason, we complemented the spatial principal
component analysis with machine-learning-based feature selection. This approach
allowed us to quantify the relative importance of each predictor while handling
high-dimensional, multicollinear datasets. The full analytical workflow was
applied to climatic, geomorphometric, and soil data from the Mendoza River
oasis in Argentina. The resulting maps and descriptions of the edaphoclimatic
zones identified are publicly available through an open-access website (https://caracterizacion-fisico-ambiental-coviar.hub.arcgis.com/), which provides
interactive visualizations and detailed zone characterizations.
Materials
and Methods
Study
area
The study area (figure 1)
is located in the Mendoza River Oasis, Argentina, and includes the departments
of Lavalle, Capital, Las Heras, Guaymallén, Maipú, and Luján de Cuyo.
Figure
1. Study area: Mendoza River oasis, Mendoza province,
Argentina.
Figura 1. Área
de estudio, oasis río Mendoza, provincia de Mendoza, Argentina.
Elevations range from 600 to 1,200 meters above sea level (m a.s.l.).
The region has a warm-temperate arid climate with low annual precipitation
(228.8 mm, minimum 148.8 mm in Lavalle), low humidity, and moderate winds. The
annual mean temperature is 15.8°C, with higher values in the north and urban
areas. The mean diurnal temperature variation is 14.3 °C. Extreme heat (>35
°C) occurs on 15.4 days per year on average, reaching 36 in Lavalle, with 3.9
heatwave events. The area records 1,536.6 annual cold hours, being the highest
in Luján de Cuyo. Frost and hailstorms are major meteorological risks. Frost
occurs on 43.4 days on average, peaking in Perdriel-Agrelo (87 days) and
northern Lavalle (57 days). Rainfall is highest in February (42.1 mm), which
increases the risk of cryptogamic disease before grape maturation.
Spatial
Data Layers
The study area boundaries were defined using digital maps of
geomorphometric and soil variables from vineyard test pits (Vallone
et al., 2023), combined with bioclimatic indices. Bioclimatic maps were derived
from a national survey of Argentina’s wine regions (Cavagnaro
et al., 2023). Weather records were obtained from nine World Meteorological
Organization (WMO)-certified stations with 41 years of data (1980-2020), supplemented
by public and private networks. Climatic maps were generated using kriging
interpolation for variables such as cumulative seasonal precipitation (CSP)
(September-April). Standard viticultural indices included: a) Growing Degree
Days (GDD) (Mullins et al., 1992),
b) Winkler Index (WI) (Amerine & Winkler, 1944),
c) Huglin Heliothermal Index (HI) (Huglin, 1983),
d) Cool Night Index (CNI) (Tonietto &
Carbonneau, 2004), and e) Thermal Integral above 13°C (TIB13). Shuttle Radar
Topography Mission (SRTM) elevation data were processed with SAGA GIS to derive
geomorphometric variables: slope, aspect, curvature, convergence, slope length
factor (L-S), topographic wetness index, multiresolution valley bottom
flatness, and vertical distance to drainage. Additional land suitability maps
and geomorphometric studies from COVIAR were incorporated (Vallone
et al., 2007). A total of 153 soil samples were analyzed for physicochemical
properties. Digital soil maps were generated (McBratney et
al., 2003) and harmonized into 0-50 cm and 50-100 cm horizons (Malone
et al., 2009). Plant available water (PAW), field capacity (FC), permanent
wilting point (PWP), and saturated hydraulic conductivity (Ksat) were estimated
from field data (bulk density, particle size) and pedotransfer functions. A
soil water storage capacity map was produced using geostatistical
interpolation. All spatial layers were resampled to a 4-ha grid (200 × 200 m).
Analytical
Method
Step 1.
Delimitation of Climatic Zones
Climatic zones were
delimited using the KM-sPC algorithm (Córdoba et al., 2012), which combines
fuzzy k-means clustering with spatial principal components (sPCs) to account
for spatial autocorrelation. Input variables included climatic and bioclimatic
indices and elevation, which strongly influence viticultural potential. The
first two sPCs were retained because they explained at least 80% of the total
climatic variance; that is, they captured the main patterns in the data.
Clustering was performed for 2 to 6 classes (i.e., five clustering
runs), with the optimal number determined using the partition coefficient,
classification entropy, and a combined summary index (Albornoz
et al., 2018). The clustering was conducted on a 4-ha grid, resolution value
to which all spatial layers were previously resampled. The analysis was
implemented in R (R
Core Team, 2024) using the “paar” package (Paccioretti et al., 2024).
Step 2.
Soil Zoning within each Climatic Zone
Within each
climatic zone, a finer edaphic partition was performed using KM-sPC with soil
and geomorphometric variables. This nested clustering approach reflects the
natural hierarchy between broader climatic-geomorphological controls and finer
edaphic variability.
Step 3.
Characterization of Delimited Edaphoclimatic Zones
Radar Plots
Radar plots were
used to visualize the multidimensional attributes of each zone. Each axis
represents a variable, with spoke lengths scaled to relative magnitudes. These
plots enabled comparison between individual zones and the overall mean profile.
Radar plots were generated using the “fmsb” package (Nakazawa,
2023).
Random Forest (RF)
RF is an ensemble
method that builds multiple decision trees from bootstrap samples and
aggregates their predictions to improve accuracy and reduce overfitting (Breiman,
2001).
In this study, RF classification was applied to evaluate the relative
importance of each variable in distinguishing individual zones by contrasting
one zone against all others. Variable importance was quantified as the mean
decrease in accuracy after permuting the values of each predictor in the
out-of-bag samples. This analysis identified the predictors that most
contributed to zone distinctiveness. Model tuning involved optimizing the
number of variables randomly selected at each split (mtry) through grid search.
The number of trees and the minimum terminal node size were fixed at 500 and 5,
respectively. Model performance was evaluated using 10-fold cross-validation.
The RF model was implemented in R using the “caret” (Kuhn
& Max, 2008) package.
Identification of
Key values for each Zone
Key characteristics
of each zone were summarized into “zone notes”, describing typical climatic,
geomorphological, and soil attributes. These summaries were developed in
collaboration with domain experts to emphasize the distinctive features of each
viticultural zone.
Step 4.
Zone Validation
The appropriateness of the delineated zones was evaluated by
comparing the means of the most important variables identified by the RF
analysis. A permutation-based statistical test accounting for spatial
correlation was applied to evaluate whether these variables differed
significantly among zones. The analysis was performed using the “ofemeantest”
package (Córdoba et al., 2024)
in R.
Results
and Discussion
Delimitation
of Climatic Zones
Figure 2
shows the first two components of the sPC analysis (sPCA), which explained
96.5% of the total climatic variability. The most influential variables for
differentiating climatic zones were WI, TIB13, and HI. These indices were positively
correlated, as indicated by the small angles between their vectors. In
contrast, these variables were negatively correlated with Digital Elevation
Model (DEM), as shown by the angle close to 180°. GDD and CSP contributed to
spatial variability but were less important, being primarily associated with
the second axis.
CNI:
Cool Night Index, HI: Huglin Index, (°GDD), TIB13: Thermal Integral with Base
13°C (°GDD), WI: Winkler Index (°GDD), CSP: Cumulative Seasonal Precipitation
(mm), GDD: Growing Degree Days (days), DEM: Digital Elevation Model (m a. s.
l.).
CNI: Índice de
Frescor Nocturno, HI: Índice Huglin (°GDD), TIB13: Integral Térmica con Base
13°C (°GDD), WI: Índice Winkler (°GDD), CSP: Precipitación Estacional Acumulada
(mm), GDD: Duración del periodo activo (días), DEM: Modelo Digital de Elevación
(m s. n. m.).
Figure
2. Graphical representation of the first two axes of
the spatial principal components analysis, which together explained 96.5% of
the total variance. The percentage contribution of each variable to the
variance explained by these components is shown.
Figura 2. Representación
gráfica de los dos primeros ejes del análisis de componentes principales
espaciales, los cuales explican el 96,5% de la varianza total. Se muestra el
porcentaje de contribución de cada variable a la varianza explicada por estas
componentes.
Based on the three indices, the climate zoning identifies two
zones (figure
3). Zone 1, located on the left bank of the Mendoza River,
between 600 and 900 m a. s. l., encompasses the transition and low areas of the
river basin, including peri-urban areas of the Capital, Godoy Cruz, and the
departments of Guaymallén, Maipú, Las Heras, and Lavalle (zone 1). This zone
exhibits the highest temperatures and the lowest precipitation levels within
the oasis. Zone 2, covering the river basin and right bank between 900 and
1,100 m a. s. l., is characterized by cooler temperatures and higher rainfall.
Figure
3. Bioclimatic zoning (left) and edaphoclimatic zoning
(right) of the Mendoza River oasis, Argentina.
Figura 3. Zonificación
bioclimática (izquierda) y edafoclimática (derecha) del oasis río Mendoza,
Argentina.
Soil
Zoning within each Climatic Zone
Clustering based on
geomorphometric and edaphic properties showed that topography strongly
influences soil variation in the oasis. The relative importance of variables
for differentiating the finer partitions indicated that DEM was one of the most
important predictors, especially for differentiating areas within climatic zone
1. In zone 2_1, DEM was again influential, together with the CNI. In contrast,
in zones 2_2 and 2_3, the most discriminative variables were CSP and
Sedimentation Volume at 50-100 cm depth (SV_100), respectively. The eigenvector
loadings from the sPCA described the contribution of each variable to overall
variance (Córdoba
et al., 2013). We complemented this with RF classification to evaluate
variable importance in terms of their discriminatory power for clustering. RF
is advantageous for handling complex, non-linear relationships and interactions
among variables, which are common in environmental datasets (Fox et
al., 2020). Combining sPCA and RF enabled a nuanced understanding of
variable influence, both from a structural and predictive perspective.
Characterization
and Validation of Edaphoclimatic Zones
Figure 3 (right) shows the
multivariate zoning of the Mendoza River oasis. The edaphoclimatic zoning
delineated five zones. Figure
4
presents star plots summarizing the main characteristics of each zone. The five
delimited zones showed statistically significant differences for most key soil
and climatic variables (p<0.05).
Variables
are listed clockwise from 12 o’clock. Suffixes _50 and _100 indicate values for
the 0-50 cm and 50-100 cm soil layers, respectively. Variables include: Cool
Night Index (CNI), pH_50, pH_100, Cation Exchange Capacity (CEC_50), Gypsum_50,
Gypsum_100, Sedimentation Volume (SV_50, SV_100), Sodium Adsorption Ratio
(SAR_50, SAR_100), Calcrete depth, Stone depth, Soil depth, Permeability, Total
Nitrogen (TN_50), Organic Matter (OM_50), Saturated Hydraulic Conductivity
(Ksat), Runoff, Erosion, Drainage, Calcium Carbonate (CaCO₃_50, CaCO₃_100),
Electrical Conductivity (EC_50, EC_100), Storage Capacity (SC), Flooding,
Multi-resolution Valley Bottom Flatness (MrVBF), Flow Accumulation (FA),
Vertical Distance to Drainage Network (VDCN), Aspect, Longitudinal Slope (LS),
Slope, Convergence Index (CI), Topographic Wetness Index (TWI), Digital
Elevation Model (DEM), Growing Degree Days (GDD), Cumulative Seasonal
Precipitation (CSP), Winkler Index (WI), Thermal Integral with Base 13°C
(TIB13), Huglin Index (HI).
Las
variables se enumeran en el sentido horario desde las 12 en punto. Los sufijos
_50 y _100 indican valores correspondientes a las profundidades de 0-50 cm y
50-100 cm, respectivamente. Las variables incluyen: Índice de Frescor Nocturno
(CNI), pH_50, pH_100, Capacidad de Intercambio Catiónico (CIC_50), Yeso_50,
Yeso_100, Volumen de Sedimentación (VS_50, VS_100), Relación de Adsorción de
Sodio (RAS_50, RAS_100), Profundidad de tosca, Profundidad de piedra,
Profundidad del suelo, Permeabilidad, Nitrógeno Total (NT_50), Materia Orgánica
(MO_50), Conductividad Hidráulica Saturada (Ksat), Escorrentía, Erosión,
Drenaje, Carbonato de Calcio (CO₃Ca_50, CO₃Ca_100), Conductividad Eléctrica
(CE_50, CE_100), Capacidad de Almacenamiento (CA), Anegamiento, Índice de Multiresolución
del Fondo de Valle (MrVBF), Acumulación de Flujo (AF), Distancia Vertical a la
Red de Drenaje (VDCN), Orientación, Pendiente Longitudinal (PL), Pendiente,
Índice de Convergencia (IC), Índice Topográfico de Humedad (TWI), Modelo
Digital de Elevación (DEM), Grados-Día Acumulados (GDD), Precipitación
Estacional Acumulada (CSP), Índice de Winkler (WI), Integral Térmica con Base
13°C (TIB13), Índice de Huglin (HI).
Figure
4. Star plots illustrating the average profile of each
edaphoclimatic zone in the Mendoza River oasis.
Figura
4. Gráficos de estrellas que ilustran
el perfil promedio de cada zona edafoclimática en el oasis del río Mendoza.
Zone 1_1 is characterized by bioclimatic indices placing it in a
very warm viticultural climate with mild nights. Annual precipitation in this
zone is the lowest in the oasis (140 mm), while the active period lasts more
than 260 days. Morphometric indicators revealed gentle slopes throughout most
of the zone, with terrain aspects oriented mainly northeast and north in the
final stretch of the river in the Lavalle sector. Topography-related factors
(convergence, surface flow accumulation, and slope length) suggest a lower risk
of water erosion in this alluvial plain compared with the lower Mendoza River
basin, where defined channels are absent. Evidence of gully erosion is observed
along the Mendoza River, Leyes Stream, and Tulumaya Stream. Soils in zone 1_1
show high effective rooting depth, with loamy to sandy loam texture. In
addition, they show lower permeability and more restricted natural drainage
than soils of the other zones. Soil salinity reaches the highest levels in the
oasis. Accumulations of limestone and gypsum can be observed at both surface
and subsurface horizons. Mean comparison of key variables indicated
statistically significant differences between zone 1_1 and the other zones
(p<0.05).
Zone 1_2 is
characterized by a very warm climate with mild nights and low precipitation
(170 mm annually). It shows the highest GDD in the oasis. Morphometric indices
suggest a generally lower risk of water erosion than in zone 1_1, except for
the cone area in the Barrancas sector. This zone comprises higher lands
transitioning towards zone 1_1. It includes the areas of Maipú and Guaymallén.
Soils are constrained by shallow groundwater and surface accumulations of
calcium carbonate (locally known as “tosca” in Mendoza). At depth, soils
transition to clayey loam with higher organic matter and total nitrogen,
especially in the green belt area at district of Km 8 and Corralitos. These
soils derive from former lagoons and are classified as intrazonal. Salinity
levels are moderate, lower than in zone 1_1.
Zone 2_1 includes
the highlands of the Mendoza River basin and the right bank of the river
(Chacras de Coria, Las Compuertas, Vistalba, Perdriel, and Lunlunta). The
climate is warm with cold nights. Annual precipitation averages 230 mm. This is
the highest zone and, according to geomorphometric indicators, is at risk of
water erosion. Soils are loamy and sandy loam at depth. They show the lowest
values of Electrical Conductivity (EC) and Sodium Adsorption Ratio (SAR), both
superficially and at depth, with slight salinity. Soil permeability is high,
and drainage is somewhat excessive. Soils are the shallowest in the oasis, averaging
115 cm, due to rocky subsoil.
Zone 2_2 includes
the southern part of Ugarteche and the area of El Carrizal. Bioclimatic indices
classify it as a warm zone with cold nights. Average minimum night air
temperature in March ranges between 12 and 14°C. Seasonal precipitation is the
highest in the oasis, with an annual average of 295 mm.
Finally, zone 2_3
is the coolest zone, with cold nights. The WI classifies it as temperate-warm.
Minimum night air temperatures in March are below 12°C. Average annual seasonal
precipitation is 260 mm. The active period is the shortest in the oasis.
Morphometric indicators revealed a higher risk of water erosion in the proximal
sector of Agrelo, with signs of gully erosion. Soil water storage capacity is
the highest (160 mm on average). Soils are sandy loam with rocky subsoil
limiting effective depth. The predominant slope orientation is southeast. The
edaphoclimatic zone map (figure
3)
closely matches field observations (Vallone et al., 2023).
The methodological
approach used here integrates bioclimatic, soil, and geomorphometric variables
while accounting for their spatial correlation. By combining the original
variables into sPCs and clustering them, the method reduces the effects of
spatial autocorrelation on total variability. As a result, the delineated zones
are more contiguous and geographically coherent, minimizing fragmentation and
better reflecting landscape continuity (Córdoba et al., 2013). This approach
addresses a common limitation of traditional terroir studies that rely on
non-spatial multivariate methods (Ghilardi et al., 2023). The primary link
between wine and soil lies soil regulating water and nutrient availability for
vines. Soil heterogeneity over space and time, and complex soil-climate
interactions are widely recognized as major drivers of terroir differentiation,
especially at regional to sub-regional scales (Lanyon et al., 2004;
Piraino & Roig, 2024). In this study, soil properties emerged as a fundamental
component of zoning, reinforcing their central role in the terroir concept. Our
findings confirm the importance of soils in defining viticultural zones.
However, the direct relationships between soil characteristics, vine
performance, and wine sensory attributes remain a key area for future
validation.
Beyond the
overarching effect of altitude, other key variables identified during the
zoning process included soil depth, depth to stone or hardpan layers, soil
permeability, and thermal indicators. Soil depth influences root development,
water retention, and nutrient uptake, ultimately shaping vine vigor and overall
water status (Morlat
& Bodin, 2006). The presence of stones or calcareous hardpans (“tosca”) can
constrain root penetration, modify drainage, and influence nutrient
availability, often creating moderate water stress conditions that are
beneficial for grape quality (Pracilio et al., 2006). Soil
permeability controls water movement and aeration within the root zone, with
direct consequences for vine vigor and fruit composition (Lanyon et
al., 2004). Elevation, in turn, regulates microclimatic conditions like
temperature regimes and rainfall distribution, both critical for grape
development and ripening (Ferretti, 2020).
Temperature
accumulation, expressed through indices such as the WI and HI, also emerged as
an important climatic factor. These indices are widely used to characterize
grapevine phenology, berry composition, and ripening dynamics, all of which
have direct implications for sugar accumulation, acidity, and aromatic
potential (Jarvis
et al., 2017). Topography -including elevation, slope, and aspect-, although
not explicitly detailed among the most influential variables here, is well
established as a determinant of local microclimatic conditions and wine
characteristics (Biss,
2020).
Other soil properties, although not primary drivers of clustering in this
study, remain essential to vine growth and contribute to intra-zone
variability. These include soil pH and nutrient availability, both influenced
by the underlying geological substrate (Retallack & Burns, 2016). While grapevines
tolerate a relatively wide pH range, deviations from the optimal levels can
hinder nutrient uptake, reduce growth, and affect yield. Soil organic matter
and localized precipitation play pivotal roles in soil fertility and water
availability. Vineyards exposed to high runoff and erosion risk may experience
reduced grape and wine quality due to the loss of topsoil and organic matter (de Sosa
et al., 2023). Furthermore, gypsum-rich soils influence soil structure and
nutrient balance, thereby affecting vine performance and grape composition (Lanyon et
al., 2004).
From a practical
perspective, the successful implementation of this zoning approach depends on
the availability and quality of spatial input data. While some viticultural
regions benefit from long-term climate records and detailed soil surveys,
others lack the resolution or coverage required. Defining a minimum dataset
-including soil depth, topographic indices, and key climate metrics- has been
suggested as a way to enhance the transferability and operational use of zoning
protocols (Bramley
et al., 2023). Moreover, the reproducible workflows applied in this study
facilitate wider adoption and ensure transparency. Importantly, the method
provides more than just classification: by identifying and characterizing zones
based on influential biophysical drivers, it offers a framework for
site-specific vineyard management, land-use planning, and even the development
of appellation criteria. Future work should integrate vine performance metrics
-such as yield, grape composition, and sensory attributes- to validate and
refine the delineated zones in relation to the terroir concept.
Conclusion
This study presents a data-driven approach for delineating
edaphoclimatic zones that ensures spatial coherence and identifies areas
according to the most influential climatic, geomorphometric, and soil variables
driving regional variability. By integrating spatially explicit clustering with
visual and statistical tools -such as star plots and the random forest algorithm-
the method enables the simultaneous assessment of variable importance and the
detailed characterization of each zone. This framework provides a robust basis
for terroir identification and supports a deeper understanding of the
uniqueness of vineyard environments, offering practical guidance for
viticultural zoning, vineyard management, and regional planning in
wine-producing regions.
Albornoz, E. M.,
Kemerer, A. C., Galarza, R., Mastaglia, N., Melchiori, R., & Martínez, C.
E. (2018). Development and evaluation of an automatic software for management
zone delineation. Precision Agriculture, 19(3), 463-476.
https://doi.org/10.1007/s11119-017-9530-9
Amerine, M. A.,
& Winkler, A. J. (1944). Composition and Quality of Musts and Wines of
California Grapes. Hilgardia, 15(6), 493-675.
https://doi.org/10.3733/hilg.v15n06p493
Biss, A. J. (2020).
Impact of vineyard topography on the quality of Chablis wine. Australian
Journal of Grape and Wine Research, 26(3), 247-258. https://doi.org/10.1111/ajgw.12433
Bramley, R.,
Ouzman, J., & Trought, M. C. T. (2020). Making sense of a sense of place:
precision viticulture approaches to the analysis of terroir at different
scales. OENO One, 54(4), 903-917. https:// doi.org/10.20870/oeno-one.2020.54.4.3858
Bramley, R., & Gardiner, P. S. (2021). Underpinning terroir
with data: a quantitative analysis of biophysical variation in the Margaret
River region of Western Australia. Australian Journal of Grape and Wine
Research, 27(4), 420-430. https://doi.org/10.1111/ajgw.12491
Bramley, R.,
Ouzman, J., Sturman, A. P., Grealish, G. J., Ratcliff, C. E. M., & Trought,
M. C. T. (2023). Underpinning Terroir with Data: Integrating Vineyard
Performance Metrics with Soil and Climate Data to Better Understand
Within-Region Variation in Marlborough, New Zealand. Australian Journal of
Grape and Wine Research, 1-23. https://doi. org/10.1155/2023/8811402
Breiman, L. (2001).
Random Forests. Machine Learning, 45(1), 5-32. https://doi.
org/10.1023/A:1010933404324
Cavagnaro, M.,
Pappalardo, C., & Dalmasso, J. (2023). Caracterización climática de
regiones vitivinícolas de Argentina. Provincia Mendoza.
https://caracterizacion-fisico-ambiental-coviar.hub. arcgis.com/
Córdoba, M.,
Balzarini, M., Bruno, C., & Costa, J. L. (2012). Análisis de componentes
principales con datos georreferenciados. Una aplicación en agricultura de
precisión. Revista de la Facultad de Ciencias Agrarias. Universidad Nacional
de Cuyo, 44(1), 27-39.
Córdoba,
M., Bruno, C., Costa, J., & Balzarini, M. (2013). Subfield management class
delineation using cluster analysis from spatial principal components of soil
variables. Computers and Electronics in Agriculture, 97, 6-14.
https://doi.org/10.1016/j.compag.2013.05.009
Córdoba, M.,
Paccioretti, P., & Balzarini, M. (2024). Ofemeantest: On Farm
Experimentation Mean Test. R package version 0.0.900.
https://github.com/PPaccioretti/ofemeantest
de Sosa, L. L.,
Navarro‐Fernández, C. M.,
Panettieri, M., Madejón, P., Pérez‐de‐Mora, A., & Madejón, E. (2023). Application of seaweed and
pruning residue as organic fertilizer to increase soil fertility and vine
productivity. Soil Use and Management, 39(2), 794-804.
https://doi. org/10.1111/sum.12882
Ferretti, C. G.
(2020). A new geographical classification for vineyards tested in the South
Tyrol wine region, northern Italy, on Pinot Noir and Sauvignon Blanc wines. Ecological
Indicators, 108, 105737.
https://doi.org/10.1016/j.ecolind.2019.105737
Ferro, M. V., &
Catania, P. (2023). Technologies and Innovative Methods for Precision
Viticulture: A Comprehensive Review. Horticulturae, 9(3), 399.
https://doi.org/10.3390/ horticulturae9030399
Fox, E. W., Ver
Hoef, J. M., & Olsen, A. R. (2020). Comparing spatial regression to random
forests for large environmental data sets. PLOS ONE, 15(3),
e0229509. https://doi.org/10.1371/ journal.pone.0229509
Ghilardi, F.,
Virano, A., Prandi, M., & Borgogno-Mondino, E. (2023). Zonation of a
Viticultural Territorial Context in Piemonte (NW Italy) to Support Terroir
Identification: The Role of Pedological, Topographical and Climatic Factors. Land,
12(3), 647. https://doi.org/10.3390/ land12030647
Hall, A., &
Jones, G. V. (2010). Spatial analysis of climate in winegrape-growing regions in
Australia. Australian Journal of Grape and Wine Research, 16(3),
389-404. https://doi.org/10.1111/ j.1755-0238.2010.00100.x
Huglin, P. (1983).
Possibilites d’ appreciation objective du milieu viticole. Bulletin de l’OIV,
56, 823-833.
INV. (2024). Informe
anual de cosecha y elaboración 2024.
Irimia, L. M.,
Patriche, C. V., & Quénol, H. (2014). Analysis of viticultural potential
and delineation of homogeneous viticultural zones in a temperate climate region
of Romania. OENO One, 48(3), 145.
https://doi.org/10.20870/oeno-one.2014.48.3.1576
Jarvis, C., Barlow,
E., Darbyshire, R., Eckard, R., & Goodwin, I. (2017). Relationship between
viticultural climatic indices and grape maturity in Australia. International
Journal of Biometeorology, 61(10), 1849-1862.
https://doi.org/10.1007/s00484-017-1370-9
Jolliffe, I. T.,
& Cadima, J. (2016). Principal component analysis: a review and recent
developments. Philosophical Transactions of the Royal Society A:
Mathematical, Physical and Engineering Sciences, 374(2065),
20150202. https://doi.org/10.1098/rsta.2015.0202
Jones, G. V.,
White, M. A., Cooper, O. R., & Storchmann, K. (2005). Climate Change and
Global Wine Quality. Climatic Change, 73(3), 319-343.
https://doi.org/10.1007/s10584-005-4704-2
Kuhn, & Max.
(2008). Building Predictive Models in R Using the caret Package. Journal of
Statistical Software, 28(5), 1-26.
https://doi.org/10.18637/jss.v028.i05
Lanyon, D. M.,
Hansen, D., & Cass, A. (2004). The effect of soil properties on vine
performance. CSIRO Land and Water Technical Report N°. 34/04.
Malone, B. P.,
McBratney, A. B., Minasny, B., & Laslett, G. M. (2009). Mapping continuous
depth functions of soil carbon storage and available water capacity. Geoderma,
154(1-2), 138-152. https:// doi.org/10.1016/j.geoderma.2009.10.007
McBratney, A. B.,
Mendonça Santos, M. L., & Minasny, B. (2003). On digital soil mapping. Geoderma,
117(1-2), 3–52. https://doi.org/10.1016/S0016-7061(03)00223-4
Morlat, R., &
Bodin, F. (2006). Characterization of Viticultural Terroirs using a Simple
Field Model Based on Soil Depth – II. Validation of the Grape Yield and Berry
Quality in the Anjou Vineyard (France). Plant and Soil, 281(1-2),
55-69. https://doi.org/10.1007/s11104-005- 3769-z
Mullins, M. G.,
Bouquet, A., & Willians, L. E. (1992). Biology of the Grapevine.
Cambridge University Press.
Nakazawa, M.
(2023). fmsb: Functions for Medical
Statistics Book with some Demographic Data. https:// minato.sip21c.org/msb/
Paccioretti, P., Córdoba, M., Giannini-Kurina, F., &
Balzarini, M. (2024). paar: Precision
Agriculture Data Analysis. R package version 1.0.1,
https://CRAN.R-project.org/package=paar
Piraino, S.; Roig,
F. A. 2024. Landform heterogeneity drives multi-stemmed Neltuma flexuosa growth
dynamics. Implication for the Central Monte Desert forest managemen. Revista de
la Facultad de Ciencias Agrarias. Universidad Nacional de Cuyo. Mendoza.
Argentina. 56(1): 26-34. DOI: https://doi.org/10.48162/rev.39.120
Pracilio, G., Smettem,
K. R. J., Bennett, D., Harper, R. J., & Adams, M. L. (2006). Site
assessment of a woody crop where a shallow hardpan soil layer constrained plant
growth. Plant and Soil, 288(1-2), 113-125.
https://doi.org/10.1007/s11104-006-9098-z
Puscama, F., Gil,
R., & Berli, F. (2025). Impact of intra-vineyard soil heterogeneity on
Malbec. Vine growth, yield and wine elemental composition and sensory profile. Revista
de la Facultad de Ciencias Agrarias. Universidad Nacional de Cuyo, 57(1),
1-18.
R Core Team. (2024).
R: A Language and Environment for Statistical Computing. R Foundation for
Statistical Computing, Vienna, Austria. R Foundation for Statistical
Computing. https:// www.R-project.org/
Retallack, G. J.,
& Burns, S. F. (2016). The effects of soil on the taste of wine. GSA
Today, 26(5), 4-9. https://doi.org/10.1130/GSATG260A.1
Straffelini, E.,
Carrillo, N., Schilardi, C., Aguilera, R., Estrella Orrego, M. J., &
Tarolli, P. (2023). Viticulture in Argentina under extreme weather scenarios:
Actual challenges, future perspectives. Geography and Sustainability, 4(2),
161-169. https://doi.org/10.1016/j. geosus.2023.03.003
Tonietto, J., &
Carbonneau, A. (2004). A multicriteria climatic classification system for
grape-growing regions worldwide. Agricultural and Forest Meteorology, 124(1-2),
81-97. https://doi. org/10.1016/j.agrformet.2003.06.001
Vallone, R.,
Olmedo, G., Maffei, J., Morábito, J., Mastrantonio, l., Lipinski, V., &
Filippini, M. (2007). Mapa de Aptitud de suelos con fines de Riego y de
riesgo de contaminación edáfica de los Oasis Irrigados de la Provincia de
Mendoza. FCA-DGI-OEI.
Vallone, R.,
Moreiras, S., Cáceres, M., Arzalluz, I., Zuin, J., Martín, T., & Corvalán,
F. (2023). Caracterización geológica, geomorfológica y edafológica de zonas
vitícolas argentinas. Provincia Mendoza. Oasis Norte.
https://caracterizacion-fisico-ambiental-coviar.hub.arcgis.com/
Van Leeuwen, C.,
Roby, J. P., Pernet, D., & Bois, B. (2010). Methodology of soil-based
zoning for viticultural terroirs. Bulletin de l’OIV, 83(947),
0-13.
Vaudour, E., Carey,
V. A., & Gilliot, J. M. (2010). Digital zoning of South African
viticultural terroirs using bootstrapped decision trees on morphometric data
and multitemporal SPOT images. Remote Sensing of Environment, 114(12),
2940-2950. https://doi.org/10.1016/j. rse.2010.08.001
Visconti, F., López, R., & Olego, M. Á. (2024). The Health
of Vineyard Soils: Towards a Sustainable Viticulture. Horticulturae, 10(2),
154. https://doi.org/10.3390/horticulturae10020154