A Multivariate Clustering Approach for Characterization of the Montepulciano d Abruzzo Colline Teramane Area

Transcription

1 A Multivariate Clustering Approach for Characterization of the Montepulciano d Abruzzo Colline Teramane Area Jose C. Herrera Nuñez, 1 Solange Ramazzotti, 2 * Fabio Stagnari, 2 and Michele Pisante 3 Abstract: A multivariate clustering approach combined with geomatics was applied to delineate and define homogeneous zones (terroir units) within the premium Denomination of Guaranteed Origin Colline Teramane winegrowing area, in the Teramo Province, Italy. A geographical information system (GIS) architecture was set up to incorporate the main factors affecting the geographical differentiation of land capability. Vine-related environmental indices and parameters were interpolated and incorporated in the GIS. A principal component analysis followed by a multivariate clustering algorithm was applied to obtain a continuous topography climate map, in which each cluster represented a homogeneous zone or terroir unit. Terroir units were further characterized by overlaying the topography climate map with land-use information and a geologic map of the same area. This territory classification provides a useful framework for further hypotheses testing, in particular vine interaction with environment and new cultivar suitability. Moreover, the approach represents a potential tool for rational territory management and land-use planning. Key words: terroir, geographic information system (GIS), clustering, zoning, Montepulciano d Abruzzo Success in vinegrowing and winemaking is to a large extent determined by natural environmental conditions in the vineyard, referred to as terroir (Seguin 1986, Falcetti 1994, van Leeuwen et al. 2004). Terroir is a unit of land that is characterized by relatively homogenous or dominant patterns of topography, climate, geology, and soil at a given scale (Laville 1990), as well as biology (variety and rootstock) and human actions (social, historical experience, and winemaking). Therefore, a terroir represents the combination of factors that interact to form a unique viticultural unit with an agronomic potential that is reflected in the characteristics of its product. In viticulture, terroir was recently officially defined by the Organisation Internationale de la Vigne et du Vin (OIV 2010): Vitivinicultural terroir is a concept which refers to an area in which collective knowledge of the interactions between the identifiable physical and biological environment and applied vitivinicultural practices develops, providing distinctive characteristics for the products originating from this area (resolution OIV/Viti 333/2010). 1 Associate Scientist, 2 Assistant Professor, 3 Full Professor, Agronomy and Crop Sciences Research and Education Center, Department of Food Science, University of Teramo, via C.R. Lerici 1, Mosciano Sant Angelo (TE), Italy. *Corresponding author ( sramazzotti@unite.it; tel: ; fax: ) Acknowledgments: The research project was cofinanced by the Fondazione della Cassa di Risparmio di Teramo. The authors thank Antonello Moneta, Informatic Engineer and Fellowship at University of Teramo in , the CRA-SCA of Bari, and the Geology Faculty of the University Gabriele D Annunzio of Chieti for their fundamental contribution to this work. Manuscript submitted Feb 2010, revised Aug 2010, Oct 2010, accepted Nov 2010 Copyright 2011 by the American Society for Enology and Viticulture. All rights reserved. doi: /ajev Terroir unit identification and characterization is the goal of viticultural zoning studies aiming to delineate zones that are likely to produce grapes or wines of similar composition and enabling operational decisions at various production stages. In traditional approaches, zoning was performed by assessing the impact of single parameters of terroir on grape quality: climate (Winkler et al. 1974, Huglin 1978), soil and geology (Seguin 1975, van Leeuwen and Seguin 1994, Mackenzie and Christy 2005, Maltman 2008), cultivar (Riou et al. 1994, Huglin and Schneider 1998), and rootstock (May 1994). However, because many factors are involved in terroir expression, its study is multidisciplinary, involving geology, topography, pedology, climatology, and vine physiology (Seguin and van Leeuwen 2006). Several new technologies, in particular computer and communications, have aided in investigating the combined effects of several terroir parameters simultaneously. Such is the case of geoinformatics (or geomatics), which combines geodetic and spatial information processing methods with computing hardware and software technologies. Geomatic applications have made it possible to collect and analyze large amounts of data on crop performance and the attributes of individual production areas at high spatial resolution (Zhou et al. 2003, Vaudour and Shaw 2005). The introduction of these new technologies in agriculture has corroborated the observations that, far from homogeneous, the productivity of agricultural land is inherently variable and that any crop-management intervention should be targeted accordingly (Pierce and Nowak 1999, Bramley and Lamb 2003, Seguin and van Leeuwen 2006). Therefore, zoning studies not only identify territories according to their potential to produce specific wines but also optimize technical vineyard management for sustainable viticulture. Environment variability can be characterized and managed to optimize winegrape production and to define reliable criteria for planting and applying 239

2 240 Herrera Nuñez et al. appropriate techniques (Lamb 2000, Bramley 2001, Bramley and Lamb 2003, van Leeuwen et al. 2004). In this context, spatial and time-related analyses of the physiographic characteristics (topography, geology, climate, etc.) of a growing area and the evaluation of their interactions with vegetation are crucial both in the planning of new planting areas and the management of traditional ones. Identification of relatively homogenous areas of expected crop performance within landscapes also has potential benefits for improved agricultural policy formulation, decision making (Lamb 2000), and resource conservation (Munier et al. 2004, Patel 2004). Our main goal was to apply the new zoning tools to the study of the Montepulciano d Abruzzo premium Denomination of Guaranteed Origin (DOCG) Colline Teramane winegrowing area (Figure 1). The Montepulciano cultivar, with over 35,000 ha of vineyards planted mainly along the Adriatic Coast, is a major Italian red grape variety. A late ripening variety, it requires sustained heat during the summer to mature properly. Montepulciano is a vigorous and generous variety, but overcropped vineyards give disappointing results. The tendency to push production beyond desirable levels is undoubtedly the result of the historic role of Montepulciano in the Abruzzo region, where very little of the wine was bottled and much was shipped elsewhere to give body to cool-climate wines. Since the DOCG recognition in 2003, this is less common in the Colline Teramane area, where the variety is undergoing a gradual transition from bulk-wine production to bottled wines. Nevertheless, the Colline Teramane area lacks an integrated study of its viticulture potential, although there is information on climate, topography, and land use. Thus, this research aims to construct a more useful instrument for decision making. To accomplish this, a multivariate clustering approach combined with geomatics was applied to climate, topography, and geology data to characterize and then delineate and define homogeneous areas within the study area. The results add a specific contribution for a better understanding of a traditional winemaking territory and represent a first step of a zoning study for the region. Materials and Methods Area of interest. The Montepulciano d Abruzzo Colline Teramane premium wine Denomination of Guaranteed Origin (DOCG) is located in the Teramo Province, Abruzzo, Italy. The territory extends from to east longitudes, and from to north latitudes between the the Apennines mountains and the Adriatic Sea (Figure 1). The area of interest (AOI) was defined following the specifications of the DOCG production disciplinary (Decreto Ministeriale del 20 febbraio 2003, G.U. n. 54 del 6 marzo 2003). According to regulations, vineyards cannot be sited over 500 m above sea level (asl) when not exposed to the south, and over 600 m asl when exposed to the south. Thus, the AOI was delimited using the elevation information contained in a 10 x 10 m digital elevation model (DEM). Total AOI surface counts for 1,214 km 2, of which vineyards surface covers about 2,300 ha, or 2% of the total surface area. Acquisition of environmental data set. Environmental data on climate, topography, geology, and land use were incorporated into a GIS architecture using ESRI ArcGIS, version 9.1 (Environmental Systems Research Institute, Redlands, CA). Climate data. Interannual daily data of climatic variables (minimum and maximum air temperature, precipitation, and solar radiation) for the 1996 to 2007 period were acquired from 26 georeferenced meteorological stations, 15 located within the AOI and 11 along boundaries (Figure 2). For each meteorological station, the following bioclimatic indices were calculated according to the multicriteria climatic classification (MCC) system (Tonietto and Carbonneau 2004): Huglin index (HI) (Huglin 1978), where T = mean air temperature ( C), Tx = maximum air temperature ( C), and k = day length coefficient. Cool night index (CI) (Tonietto and Carbonneau 2004), CI = Tn 9 where Tn 9 = average of the daily minimum air temperature ( C) in the month of September. Dryness index (DI) (Tonietto and Carbonneau 2004), DI = Wo + P - Tv - Es Figure 1 Map of the DOCG Montepulciano d Abruzzo Colline Teramane winegrowing area, in the Teramo Province, Italy. where Wo = initial useful water reserve in the soil (mm), P = precipitation (mm), Tv = potential transpiration of the vineyard (mm), and Es = direct evaporation from the soil (mm). In the northern hemisphere, DI is calculated from 1 Apr to 30 Sept. The bioclimatic indices were interpolated using cokriging, as implemented in ArcGIS, with elevation as the secondary variable (covariable). This interpolation method was selected

3 Characterization of the Colline Teramane Winegrowing Area 241 Figure 2 Maps of bioclimatic indices interpolated using data sets from 26 georeferenced meteorological stations located within and along boundaries of the study area: (A) Huglin index (degree days), (B) cool night index ( C), and (C) dryness index. White areas within the area of interest indicate zones above 650 m asl, excluded from cultivation by the specifications of the DOCG Colline Teramane production disciplinary. after comparison to inverse distance weighted and kriging because of its lower root-mean square error. Topographical data. A digital elevation model (DEM) in raster format at 10 x 10 m/pixel resolution (Figure 3) was acquired from the Abruzzo Regional Agency for Development and Services in Agriculture. Maps of slope and incoming solar radiation (ISR) from April to September were derived from the DEM using the ArcGIS Surface and Solar radiation Spatial Analyst Tools, while a topographic wetness index (TWI) was obtained using Saga Terrain analysis modules (SAGA-System for Automated Geoscientific Analyses, version 2.0.3, Institute of Geography, University of Hamburg, Germany). The topographic wetness index can be used to assess the pattern of potential soil moisture on a field and changes in soil texture caused by erosion (Schmidt and Persson 2003). Geological data. Soil parameters (pedology) within the Teramo Province are lacking because of the unavailability of data. Such information is crucial to better define terroir units, especially when aiming at sustainable agronomic practices. However, as soil distribution is related to topography, and in turn related to erodibility of parent rock, a 1: scale geology map of the area (D Alessandro et al. 2007) was used. Land-use data. Spatial data of the boundaries of every vineyard present in the territory was acquired from the Italian Agency for Supplies in Agriculture and from aerial color (RGB) photographs (orthophoto) of the entire Teramo Province at 0.5 m/pixel resolution, taken from September to October Orthophotos and vineyard boundaries were used for photointerpretation of results and further analysis of the spatial distribution of vineyards and their climate, topography, and geology characteristics. Iterative clustering process. Environmental data was used for the identification and characterization of terroir units within the AOI. A terroir unit map was created by performing a principal component analysis (PCA) followed by a multivariate clustering algorithm applied to the following seven layers: Huglin index, cool night index, dryness index, DEM, slope, topographic wetness index, and incoming solar radiation. A similar method was proposed for the definition of agroecozones (Williams et al. 2008). First, variable values of each raster cell were standardized (from 0 to 1). Then the PCA was performed using a forward rotation, so the seven layers formed seven orthogonal axes in the data space into which each cell was plotted. The similarity of cells within the 7-dimensional data space was then coded as Euclidean separation distances. The PCA, as performed in a GIS environment, is a linear transformation that reorganizes the variance of a multiband image into a new set of image bands. Each individual band in the output PCA image receives some contribution from all of the input bands (that is, the contribution of each layer to each resultant principal component). The resultant principal components explaining a cumulative variance of 90 to 95% were then used in the clustering algorithm. Cells were clustered using the nonhierarchical, unsupervised iterative algorithm ISODATA (iterative selforganizing data analysis technique), which calculates class

4 242 Herrera Nuñez et al. means evenly distributed in the data space and then iteratively clusters the remaining pixels using minimum distance techniques. Each iteration recalculates means and reclassifies pixels with respect to the new means. Iterative class splitting, merging, and deleting are done based on input threshold parameters set by the user. All pixels are classified to the nearest class unless a standard deviation or distance threshold is specified, in which case some pixels may be unclassified if they do not meet the selected criteria. This process continues until the number of pixels in each class changes by less than the selected pixel change threshold or the maximum number of iterations is reached. The change threshold was set to 5% and maximum standard deviation to 0.2 for each class. Results and Discussion The spatial analysis of the climatic and topographic data permitted the preliminary characterization of the Colline Teramane winegrowing area. The classification approach produced a map with different homogeneous zones (terroir units) within the AOI. Climate and topography. The 15 meteorological stations within the AOI were located from 5 to 610 m asl and 150 m to 35.8 km from the coast (Figure 2). Growing season (April to September) mean temperature averages 20 C with approximately a 4 C range across the stations According to the season average temperature index (GST) (Jones 2006), these mean temperature values can be classified as warm, hot, and very hot climate (Hall and Jones 2009). Heat accumulation during the growing season was given by the Huglin index (HI), which ranged from 1926 to 2441 (Figure 2A), indicating that the Colline Teramane area could be classified under three of the six HI classes in the multicriteria climatic classification system (Tonietto and Carbonneau 2004). Cool night index (CI) values ranged from 11.4 C to 17.0 C (Figure 2B), placing the AOI in the cool night and temperate night classes. HI and CI showed similar spatial variability, with the lowest values at higher elevations inland. Precipitation ranged from ~600 to 900 mm on an annual basis and from 240 to 630 mm during the growing season; coastal areas had the lowest precipitation. The average precipitation during the hottest months of the year (June, July, and August) was ~145 mm. The dryness index (DI) ranged from 60.5 to mm (Figure 2C), falling into the subhumid (DI-1) and humid (DI- 2) classes of multicriteria climatic classification system. The topography of the AOI was well modeled by the accurate resolution of DEM (Figure 3). The landscape is dominated mainly by rugged hills and river valleys that run from the mountains to the sea. Slopes in the region indicate the inclination grade of the hills, which ranges from 0 to 40. The topographic wetness index (TWI) map showed that the major probability for water logging (higher values) is present near the river courses and downhill where the slope is minimum. The incoming solar radiation (ISR) map depicted the greater sunlight irradiation over the hilly zones exposed to south and across the flatter zones such as those near the rivers, with values ranging from 700 to 1291 kw.h/m 2. Terroir unit delineation: PCA and clustering. The PCA resulted in 81% of the total variance explained by the first principal component (PC), with HI, CI, and ISR as the highest loading parameters (Table 1). The second PC explained Figure 3 Digital elevation model of the study area at 10 x 10 m/pixel resolution. White areas within the area of interest indicate zones above 650 m asl, excluded from cultivation by the specifications of the DOCG Colline Teramane production disciplinary. Table 1 Percentage of explained variance and loading values from PCAs of the bioclimatic and morphologic variables. (Loadings of each layer in the PCs are calculated by squaring the eigenvector s value.) Bold values indicate the most dominant loadings of each parameter in the first three PCs. PC1 PC2 PC3 PC4 PC5 PC6 PC7 Explained variance (%) Layer Eigenvectors Elevation Topographic wetness index (TWI) Slope Incoming solar radiation (ISR) Huglin index (HI) Cool night index (CI) Dryness index (DI)

5 Characterization of the Colline Teramane Winegrowing Area % of the variance, with elevation and slope as the highest loading parameters. The third PC explained 5% of the remaining variance and loads on slope and ISR. The amount that each original input band contributed to each PCA band can be determined by examining the eigenvectors from the PC analysis (Table 1). The multivariate clustering (ISODATA algorithm) was then performed for the first three PCs, resulting in a classification map of the AOI and the identification of four terroir units: zones A, B, C, and D (Figure 4). Zone A is found mainly on flat areas from the coast along the valleys and covers 34% of the AOI. Zone B occupies the inner areas (26% of AOI) with higher elevations, zone C covers the hilly areas near the river valleys (24% of AOI), and zone D (16% of AOI) represents marginal areas with higher elevations and slopes. The mean values of the bioclimatic indices within each terroir unit classify them as relatively similar in the multicriteria Figure 4 Terroir unit map resulting from PCA and multivariate clustering performed using climate and topographic data. The resulting four different clusters corresponding to four terroir units (zones A, B, C, and D) are highlighted in different grays. White areas within the area of interest indicate zones above 650 m asl, excluded from cultivation by the specifications of the DOCG Colline Teramane production disciplinary. climatic classification system: HI +1, CI-1, DI-1, corresponding to temperate warm, subhumid areas with temperate nights during maturation time (Table 2). Zones A and B present the highest energy potential, with the HI slightly higher in zone A because it comprises the flatter zones along the rivers and coast. Zone C has slightly lower, but similar ISR and HI values, than zone A. Zone D has the lowest energy potential as it presents the lowest values of HI, CI, and ISR. The spatial distribution of vineyards within the terroir units was studied using land-use layer information. Zone A accounts for 59% of total vineyards ( ha) within the AOI, zone B accounts for 10% (194.1 ha), zone C accounts for 30% (557.5 ha), and zone D accounts for only 1% (30 ha). The terroir units map was overlaid with a geology map to obtain the percentage distribution of different subsoil or lithotypes within the classified zones. Three lithologic classes were of more viticultural importance: clay deposits accounted for 75% of zone A, 49% of zone B, 54% of zone C, and 39% of zone D; pelitic arenaceous rocks accounted for 23% of zone B, 23% of zone C, and 44% of zone D; fluvial gravel and sand deposits accounted for 11% of zone B, 15% of zone C, and 7.5% of zone D. As mentioned previously, information regarding soil composition is lacking within the Abruzzo region, and therefore we were unable to incorporate this layer into the GIS and the classification algorithm. Nevertheless, the incorporation of geological map information in the study has led to some relevant considerations: clayey lithotypes, mainly in the hilly and coastal areas, are impermeable and therefore characterized by a high water retention capacity. Consequently, the influence of the exposure is strong: when hills are exposed to the south, erosive phenomena could be favored as well as the formation of badlands, with greater soil erosion risk. Otherwise, when hills are exposed to the north, runoff phenomena could be favored. In gravelly-sandy lithotype areas, the high permeability and typical subsurface waterproof layers could be sites of local water tables, even at modest depths. Moreover, when potential evapotranspiration is high (southeast of the AOI), the risk for water stress is increased. Conclusions This research provides a zonation of viticultural potential in the Colline Teramane area by integrating the spatial variability in climate, topography, and geology. The proposed Table 2 Summary statistics of the bioclimatic indices and morphologic parameters used in the analysis. The mean values and standard deviations (SD) of the four zones are reported. Zone A Zone B Zone C Zone D Mean SD Mean SD Mean SD Mean SD Huglin index (HI) Cool night index (CI) Dryness index (DI) Elevation Slope Incoming solar radiation (ISR) Topographic wetness index (TWI)

6 244 Herrera Nuñez et al. method uses PCA and geographic clustering algorithms for a multivariate study of the principal climate indices and topographic parameters important for grapevine growth. The result is a dynamic map where feasible homogeneous zones or terroir units can be outlined according to input parameters and criteria set during data processing. Within the study area, the Huglin index, incoming solar radiation, the cool night index, and elevation were the most influential indices and parameters on spatial variability. Temperature-related climate indices showed higher amounts in the coastal areas and in more exposed areas and lower amounts further inland and in mountain shadow zones. Furthermore, the analysis of vineyard distributions and geologic characteristics showed that ~89% of vineyards were located in two of the four terroir units (zones A and C) and lay primarily on a substrate of clay deposits. The results of the present study allow for recommendations on potential vineyard restructuring operations, site selection for new vineyards, and environmental impact modeling and for more regional agricultural policymaking. The identification of terroir units also provides a valuable framework for further research, particularly for testing hypotheses as vine response to climate change and for evaluating the suitability of new cultivars. Because quantitative validation procedures are currently unavailable, future research is necessary to understand the real impact of environmental variability captured by terroir units on yield and quality of the final product. Literature Cited Bramley, R.G.V Progress in the development of precision viticulture Variation in yield, quality and soil in contrasting Australian vineyards. In Precision Tools for Improving Land Management. L.D. Currie and P. Loganathan (eds.), pp Occasional report no. 14. Fertilizer and Lime Research Centre, Massey University, Palmerson North, New Zealand. Bramley, R., and D. Lamb Making sense of vineyard variability in Australia. In Proceedings of the IX Congreso Latinoamericano de Viticultura y Enologia, pp Pontificia Universidad Católica de Chile, Santiago. D Alessandro, L., L. Del Sordo, M. Buccolini, E. Miccadei, T. Piacentini, and A. Urbani Analisi del dissesto da franca in Abruzzo. In Rapporto sulle frane in Italia. Il Progetto IFFI. Metodologia, risultati e rapporti regionali, pp Rapporti APAT 78/2007. APAT, Rome. Falcetti, M Le terroir. Qu est-ce qu un terroir? Pourquoi l étudier? Pourquoi l enseigner? Bull. OIV 67(2): Hall, A., and G.V. Jones Effect of potential atmospheric warming on temperature based indices describing Australian winegrape growing conditions. Aust. J. Grape Wine Res. 15: Huglin, P Nouveau mode d évaluation des possibilités héliothermiques d un milieu viticole. C.R. Acad. Agric. Fr. 64: Huglin, P., and C. Schneider Biologie et écologie de la vigne. Lavoisier Tec & Doc, Paris. Jones, G.V Climate and terroir: Impacts of climate variability and change on wine. In Fine Wine and Terroir: The Geoscience Perspective. R.W. Macqueen and L.D. Meinert (eds.), pp Geological Association of Canada, St. John s, Newfoundland. Lamb, D.W The use of qualitative airborne multispectral imaging for managing agricultural crops a case study in south-eastern Australia. Aust. J. Exp. Agric. 40: Laville, P Le terroir, un concept indispensable à l élaboration et à la protection des appellations comme à la gestion des vignobles: Le cas de la France. Bull. OIV : Mackenzie, D.E., and A. Christy The role of soil chemistry in wine grape quality and sustainable soil management in vineyards. Water Sci. Technol. 51: Maltman, A The role of vineyard geology in wine typicity. J. Wine Res. 19:1-17. May, P Using Grapevine Rootstocks. The Australian Perspective. Winetitles, Adelaide. Munier, B., K. Birr-Pedersen, and J.S. Schou Combined ecological and economic modelling in agricultural land use scenarios. Ecol. Model. 174:5-18. OIV (Organisation Internationale de la Vigne et du Vin) Summary of Resolutions Adopted by the Eighth General Assembly of the OIV. June Tbilisi, Georgia. Patel, N.R Remote sensing and GIS application in agro-ecozoning. In Proceedings of the Training Workshop (World Meteorology Organization, Geneva, Switzerland). M.V.K. Sivakumar, et al. (eds.), pp Dehra Dun, India. Pierce, F., and P. Nowak Aspects of precision agriculture. Adv. Agron. 67:1-85. Riou, C., P. Pieri, and B. Le Clech Consommation d eau de la vigne en conditions hydriques non limitantes. Formulation simplifiée de la transpiration. Vitis 33: Schmidt, F., and A. Persson Comparison of DEM data capture and topographic wetness indices. Precis. Agric. 4: Seguin, G Alimentation en eau de la vigne et composition chimique des moûts dans les Grands Crus du Médoc. Phénomènes de régulation. Conn. Vigne Vin 9: Seguin, G Terroirs and pedology of vine growing. Experientia 42: Seguin, G., and C. van Leeuwen The concept of terroir in viticulture. J. Wine Res. 17:1-10. Tonietto, J., and A. Carbonneau A multicriteria climatic classification system for grape-growing regions worldwide. Agric. Forest Meteorol. 124: van Leeuwen, C., and G. Seguin Incidences de l alimentation en eau de la vigne, appréciée par l état hydrique du feuillage, sur le développement de l appareil végétatif et la maturation du raisin (Vitis vinifera variété Cabernet franc, Saint-Emilion, 1990). J. Int. Sci. Vigne Vin 28: van Leeuwen, C., P. Friant, X. Choné, O. Tregoat, S. Koundouras, and D. Dubourdieu Influence of climate, soil, and cultivar on terroir. Am. J. Enol. Vitic. 55: Vaudour, E., and A. Shaw A worldwide perspective on viticultural zoning. S. Afr. J. Enol. Vitic. 26: Williams, C., W. Hargrove, M. Liebman, and D. James Agroecoregionalization of Iowa using multivariate geographical clustering. Agric. Ecosys. Environ. 123: Winkler, A., J. Cook, W. Kliewer, and L. Lider General Viticulture. University of California Press, Berkeley. Zhou, Y., S. Narumalani, W.J. Waltman, S.W. Waltman, and M.A. Palecki A GIS-based spatial pattern analysis model for ecoregion mapping and characterization. Int. J. Geophys. Inform. Sci. 17: