Original research articles

Classifying vineyards from satellite images: a case study on Burgundy’s Côte d’Or


Aim: To use Remote Sensing imagery and techniques to differentiate categories of Burgundian vineyards.

Methods and results: A sample of 201 vine plots or “climats” from the Côte d’Or region in Burgundy was selected, consisting of three vineyard categories (28 Grand Cru, 74 Premier Cru, and 99 Communale) and two grape varieties (Pinot Noir and Chardonnay). A mask formed by the polygons of these vine plots was made and projected on four satellite images acquired by the ASTER sensor, covering the Côte d’Or region in years 2002, 2003 (winter image), 2004 and 2006. Mean reflectances were extracted from pixels within each polygon for each of the nine spectral bands (visible and infrared) covered by ASTER. The database had a total of 797 reflectance spectra assembled over the four images. Statistical discriminant analysis of percentage classification accuracy was made separately for Côte de Nuits and Côte de Beaune, and for each year. Results showed that for individual years and Côtes, classification accuracy for vineyard category was as high as 73.7% (Beaune 2002) and as low as 66.7% (Beaune 2003). There were no significant differences in accuracy between spring, summer and winter images. Classification accuracy for grape variety in Côte de Beaune over the four study years was between 73.5% for Pinot Noir climats in 2004 and 91.9% for Chardonnay climats in 2006, including the winter image. Concerning the vegetation index NDVI, there were no significant differences between vineyard categories.

Conclusions: Satellite data is shown to be functional to reveal vineyard quality. Spectral differences between categories of Burgundian vineyards are at least partially due to terroir characteristics, which are transmitted to vine and vine canopy.

Significance and impact of the study: This work indicates that Remote Sensing techniques can be used as an auxiliary tool for the monitoring of vineyard quality in established viticultural regions and for the study of quality potential in new regions.


The observation of Earth from remote platforms like airplanes or satellites has proved to be a powerful resource for land studies, with applications to geology, agriculture, environmental sciences, urban and marine monitoring, and many other fields. Presently, most of these Remote Sensing investigations are performed using digital images collected from satellites, which provide low-cost data, with the advantage of being re-acquired at new over-flights. The typical Remote Sensing process involves cameras and sensors aboard the satellite, which collect sunlight reflected from the Earth’s surface; during reflection by types or classes of surface cover, like soil or vegetation, the solar spectrum undergoes modifications. The resulting reflectance spectra carry characteristic features of the classes present in the imaged surface and identification of these classes is possible. For example, reflectance spectra from plants are characterized by low reflectance in visible light, with a peak at 550 nm due to chlorophyll, which is the reason for the green color of vegetation; at near-infrared (NIR) wavelengths there is an abrupt transition towards stronger reflectances (the Red Edge); and at longer wavelengths (the Short-Wave Infrared, SWIR) reflectance falls, carrying the typical features of absorption by water at 1400 and 1900 nm. The spectral signatures of minerals or water are quite different, and this allows the identification of classes of soil and land cover in Remote Sensing images which have the adequate spectral sensitivity; for comprehensive reviews on the applications of Remote Sensing imagery to land monitoring, see Jensen (2007) or Campbell and Wynne (2011).

This work deals with reflectance spectra of vineyards as a particular class of vegetation. Applications of Remote Sensing techniques to vineyard studies are still in their infancy. Up until now, the majority of studies have focused on precision viticulture management in private properties of limited surface and, for this reason, are based on airborne sensors, either multispectral or hyperspectral (Bramley and Proffitt 1999; Zarco-Tejada et al 2005). Remote Sensing imagery from satellites covers much larger areas and is suitable for regional surveys and monitoring. This field of research is new and much groundbreaking work has to be done. In this aspect, we reported in a series of papers studies performed over several viticultural areas in Europe and South America (Da Silva and Ducati 2009; Blauth and Ducati 2010; Ducati et al 2014). It was demonstrated that Remote Sensing data and techniques allows not only the separation of vineyards from other vegetation, but also, to a certain degree, the identification of grape varieties (Cemin and Ducati 2011). These possibilities were already perceived from laboratory measurements (Lacar et al 2001), but now it becomes clear that satellite images have their own potential in viticultural studies. After using satellite images to study vineyards in France (Bordeaux, Champagne, Loire), Chile and Brazil, we now focus our studies on Burgundy´s Côte d’Or. This choice is justified by three basic factors:

  • a) The hierarchical division of the Burgundian vineyard is historical and emblematical, having been the object of countless studies, but up to the present day few papers, if any, have used observations from space;
  • b) The typical size of vine parcels in Burgundy is of the order of few hectares, being adequately resolved by multispectral images like those from ASTER sensor;
  • c) The Côte d’Or region is generally oriented facing east (Pitiot and Servant 2010; Atkinson 2011), and so most vineyards receive the morning sunlight in fairly equal inclinations of solar rays. This fact is relevant since the ASTER imager gets data in the morning (around 10h30 AM). The illumination of parcels, which in general are on gentle slopes, tends to be homogeneous; this perception was gained during several field trips to the region in the last years by the first author.

In Burgundy, the hierarchy of Grand Cru, Premier Cru, and more generic appellations (Côtes, Villages, Communales, etc.) seems to be linked to soil characteristics, which are at the very root of the terroir concept (Van Leeuwen and Seguin 2006). From a principal component analysis, Wittendal (2004) gave weight to a widespread perception, indicating that most Grand Cru soils have a particular structure that is significantly different from the soils of other categories. Therefore, the objective of this investigation was to verify if these quality categories, which are transmitted from soil to wine, are also transmitted from soil to vine leaves and if they can be detected in the spectral information contained in the images. This is because the observation parameter in digital images, the reflectance, originates mainly from vine leaves reflecting sunlight, if we are using non-winter data. At the high plant density used in Burgundy (up to 10,000 vines/hectare), the soil is almost entirely covered by the plant canopy; besides, at the moment of image acquisition (10h30 AM), there is an important projection of shadow between vine rows and little sunlight is reflected from the shadowed soil. This point will be analysed in greater detail in the Discussion section, but for now, it can be stated, for practical purposes, that the reflected light in images acquired during the vegetative cycle comes essentially from vine leaves.

Materials and methods

1. Image acquisition

Images from the ASTER sensor, which is aboard the Terra satellite, were acquired through the NASA website reverb.echo.nasa.gov/reverb, in the context of a research project submitted by the authors and approved by NASA. Extensive information on the ASTER sensor can be found in Abrams et al (2002). The spatial resolution, or pixel size of the images, is 15 meters at the first three spectral bands in the visible and near-infrared (VNIR) subsystem, and 30 meters at the six bands in the SWIR subsystem. These combined features allow deeper analysis of reflectance compared with those of other orbital sensors like Landsat, CBERS or ALOS. Image dates were September 19, 2002; June 9, 2004; and September 6, 2006, corresponding to late spring or late summer in the Northern hemisphere. To look for pure soil effects, we also analysed a winter image (February 15, 2003), where no grape leaves were covering the soil and hiding it from satellite view. The usual treatments were applied: atmospheric correction (Berk et al 2006) and compensation for the crosstalk effect (Iwasaki and Tonooka 2005). An additional treatment was made concerning pixel size; to run the atmospheric correction algorithm it is necessary that all nine bands be at the same spatial resolution (either 15 meters or 30 meters). When studying vine parcels of a few hectares, the spatial resolution has to be as high as possible, and to take full advantage of the three higher resolution (15 meters) bands, we resampled the six 30-meter bands to match the 15-meter pixels, a procedure already used and discussed elsewhere (Mather 1999; Altaweel 2005).

2. Vineyard selection

The Côte d’Or region has about 1,200 vine parcels which are climats or lieu-dits. This is a large number, and to study them all would be a major constraint to a project whose prime objective is to test the spectral recognition of vineyard categories. Thus, we looked for a sample which would be representative of the three categories (Grand Cru, Premier Cru and Communales), having as selection criteria parcels with adequate areas (not too small, i.e., with more than 40 pixels of 225 m2 each), with adequate geometry (the more square, the better), and evenly distributed over the Côte d’Or region. The maps of Côte de Nuits and Côte de Beaune by Pitiot and Poupon (2009) were used for the selection of vine parcels. The final sample was formed by 201 plots: 28 Grand Cru (10 in Beaune, 18 in Nuits), 74 Premier Cru (51 in Beaune, 23 in Nuits), and 99 Communales (53 in Beaune, 46 in Nuits), this last category corresponding to vineyards which are lieu-dits and which are neither Grand Cru nor Premier Cru. The complete list of these vineyards is given in Appendix A.

A mask containing the polygons of these vineyards was generated based on the maps of the region, which were georeferenced in the same frame of reference as the satellite images. This mask was then superposed on the ASTER images, on which the selected parcels could be identified. For each climat or lieu-dit, the mean reflectance of all pixels inside a polygon was calculated for each one of the nine spectral bands. Great care was taken in doing the mask to ensure that each polygon was entirely inside the climat, avoiding contamination from other spectral classes, like roads, buildings or other vegetation classes. The selected polygons had from 40 to 140 pixels, corresponding to areas within climats from 0.9 to about 3 hectares, since a pixel has 15 m x 15 m or 225 m2.

The final sample had 797 reflectance spectra of 201 climats of the three vineyard categories. The sample was divided in eight files, one file for each year (2002, 2003, 2004, and 2006) and for each Côte. The exact number of each category in each year/image varied slightly because variations in image quality sometimes made it impossible to accurately visualize certain climats. The sample also contained information on grape variety. In Côte de Nuits, only 2 out of the 87 selected parcels were Chardonnay, expressing the massive dominance of Pinot Noir in the northern part of Côte d’Or. On the other hand, in Côte de Beaune many parcels of the Communale category and some of Premier Cru were planted with both grape varieties. In the eight files forming the database, each line contained the climat number and name, the category code, the grape variety, and all nine mean reflectance values. Figure 1 shows how some climats appear in the satellite image.

Figure 1. Examples of polygons superposed on an ASTER satellite image and the corresponding maps. Some selected climats in two areas of Côte de Nuits are shown.

3. Statistical analysis

A statistical discriminant analysis with respect to vineyard category was performed using the nine spectral bands of reflectance spectra for each year and each Côte, since a preliminary analysis using the whole sample revealed that seasonal and spatial differences increased noise level.

An additional analysis using one-way ANOVA with Tukey post-hoc multiple comparisons test was made, looking for which spectral band could be more relevant to separate one vineyard category from the others.

Another discriminant analysis was performed with respect to grape variety for the Côte de Beaune files. Since each image was studied separately, we included the winter data in the analysis, although in winter there are no vine leaves to reflect sunlight. We also performed two separate analyses for grape varieties. The first one was with Pinot Noir (49 parcels) and Chardonnay (37 parcels); all 28 mixed parcels were excluded from this analysis. The second analysis considered all the three grape groups: Pinot Noir, Chardonnay and Pinot/Chardonnay. Preliminary tests showed that the discriminant analysis performed better for the pure groups. This was expected because in parcels with spectral mixture at pixel level, the exact variety proportions are unknown. Therefore, we restricted the analysis to the mono-varietal vineyards.

An additional analysis was performed concerning a vegetation index, in this case the Normalized Difference Vegetation Index (NDVI), defined as (Tucker 1979):

where RIR is the reflectance at the near infrared, which in ASTER subsystems is band 3 (0.760–0.860 µm), and RVIS is the reflectance at red (0.630–0.690 µm), which is band 2. NDVI values were calculated for all 797 spectra and an ANOVA analysis was made comparing the mean values of NDVI with respect to vineyard category.


With respect to vineyard category, the application of discriminant analysis to Côte de Beaune (Table 1) and Côte de Nuits (Table 2) showed that for individual years and Côtes, classification accuracy was as high as 73.7% (Beaune 2002) and as low as 66.7% (Beaune 2003).

Table 1. Discriminant analysis (vineyard category) for Côte de Beaune.

Year Category Predicted Group Membership Total
Communale Grand Cru Premier Cru
2002 Original Count Communale 37 3 13 53
Grand Cru 2 8 0 10
Premier Cru 9 3 39 51
% Communale 69.8 5.7 24.5 100
Grand Cru 20 80 0 100
Premier Cru 17.6 5.9 76.5 100
2003 Original Count Communale 41 2 10 53
Grand Cru 1 7 2 10
Premier Cru 12 11 28 51
% Communale 77.4 3.8 18.9 100
Grand Cru 10 70 20 100
Premier Cru 23.5 21.6 54.9 100
2004 Original Count Communale 37 4 10 51
Grand Cru 0 10 0 10
Premier Cru 13 5 32 50
% Communale 72.5 7.8 19.6 100
Grand Cru 0 100 0 100
Premier Cru 26 10 64 100
2006 Original Count Communale 37 3 11 51
Grand Cru 0 8 2 10
Premier Cru 11 6 34 51
% Communale 72.5 5.9 21.6 100
Grand Cru 0 80 20 100
Premier Cru 21.6 11.8 66.7 100

Table 2. Discriminant analysis (vineyard category) for Côte de Nuits.

Year Category Predicted Group Membership Total
Communale Grand Cru Premier Cru
2002 Original Count Communale 36 3 6 45
Grand Cru 3 12 3 18
Premier Cru 3 8 11 22
% Communale 80.0 6.7 13.3 100
Grand Cru 16.7 66.7 16.7 100
Premier Cru 13.6 36.4 50 100
2003 Original Count Communale 33 6 7 46
Grand Cru 4 10 4 18
Premier Cru 2 4 17 23
% Communale 71.7 13.0 15.2 100
Grand Cru 22.2 55.6 22.2 100
Premier Cru 8.7 17.4 73.9 100
2004 Original Count Communale 32 5 9 46
Grand Cru 2 14 2 18
Premier Cru 2 6 15 23
% Communale 69.6 10.9 19.6 100
Grand Cru 11.1 77.8 11.1 100
Premier Cru 8.7 26.1 65.2 100
2006 Original Count Communale 34 6 6 46
Grand Cru 1 14 3 18
Premier Cru 3 4 16 23
% Communale 73.9 13.0 13.0 100
Grand Cru 5.6 77.8 16.7 100
Premier Cru 13.0 17.4 69.6 100

Results from the ANOVA test which investigated the more relevant spectral bands for vineyard category discrimination were as follows: Grand Cru climats with Pinot Noir grapes had higher mean values of reflectance in bands B2, B4, B5, B6, B7 and B8; that is, in these six bands, Grand Cru plots were different from both Premier Cru and Communale plots. Numbers are less clear for Chardonnay Grand Crus. No systematic differences were observed, and in some bands, Communale was separated from Grand Cru and Premier Cru and in others, Grand Cru was separated, with higher or lower values, from the other classes. It seems that for Chardonnay, the Grand Cru category is separated in a more complex way.

Grape variety results for Côte de Beaune, over the four years, are presented in Table 3. Accuracy was as low as 73.5% for Pinot Noir parcels (2004) and as high as 91.9% for Chardonnay parcels (2006). It is worth of note that the separation of Pinot Noir from Chardonnay vine plots was fairly well done even using 2003 winter data (79.6% for Pinot Noir and 89.2% for Chardonnay), performing better than June 2004 data and with an accuracy similar to late summer data for 2002 and 2006.

Table 3. Discriminant analysis (grape variety) for Côte de Beaune.

Year Grape variety Predicted Group Membership Total
Chardonnay Pinot noir
2002 Original Count Chardonnay 32 5 37
Pinot noir 5 44 49
% Chardonnay 86.5 13.5 100.0
Pinot noir 10.2 89.8 100.0
2003 Original Count Chardonnay 33 4 37
Pinot noir 10 39 49
% Chardonnay 89.2 10.8 100.0
Pinot noir 20.4 79.6 100.0
2004 Original Count Chardonnay 27 9 36
Pinot noir 13 36 49
% Chardonnay 75.0 25.0 100.0
Pinot noir 26.5 73.5 100.0
2006 Original Count Chardonnay 34 3 37
Pinot noir 6 43 49
% Chardonnay 91.9 8.1 100.0
Pinot noir 12.2 87.8 100.0

Results for NDVI are shown in Table 4. Not surprisingly, NDVI values for all vineyard categories were smaller for the 2003 winter image (mean values around 0.25), being typical of classes like soil, which is exposed at this season. The index increased as the vegetative cycle progressed and was 0.46 for the June 2004 image in the two regions. The NDVI index was higher for both late summer images (Sept. 2002 and 2006), with mean values between 0.62 and 0.64. However, there were no significant variations of NDVI comparing vineyard category.

Table 4. Vegetation index (NDVI) for Côte de Beaune and Côte de Nuits.

Côte epoch # vineyards NDVI min NDVI max Mean Std. Dev.
Beaune Sept. 2002 Comm. (53) .53 .70 .64 .040
Gr. cru (10) .53 .74 .61 .068
Pr. cru (51) .48 .74 .62 .052
Total (114) .48 .74 .64 .049
Feb. 2003 Comm. (53) .20 .29 .24 .020
Gr. Cru (10) .20 .30 .24 .029
Pr. Cru (51) .20 .33 .24 .030
Total (114) .19 .33 .24 .025
June 2004 Comm. (51) .39 .54 .48 .033
Gr. Cru (10) .38 .48 .44 .033
Pr. Cru (51) .33 .57 .45 .048
Total (111) .33 .57 .46 .043
Sept. 2006 Comm. (51) .57 .69 .64 .031
Gr. Cru(10) .55 .66 .61 .034
Pr. Cru (51) .53 .68 .63 .040
Total (112) .53 .69 .63 .037
Nuits Sept. 2002 Comm. (45) .51 .72 .63 .043
Gr. Cru (18) .50 .78 .65 .071
Pr. Cru (22) .50 .70 .63 .058
Total (85) .50 .78 .63 .054
Feb. 2003 Comm (46) .22 .32 .27 .022
Gr. Cru (18) .22 .32 .26 .026
Pr. Cru (23) .21 .34 .26 .030
Total (87) .21 .34 .26 .025
June 2004 Comm. (46) .41 .51 .47 .023
Gr. Cru (18) .40 .49 .45 .023
Pr. Cru (23) .40 .51 .45 .033
Total (87) .40 .51 .46 .026
Sept. 2006 Comm. (46) .47 .69 .61 .050
Gr. Cru (18) .55 .74 .66 .055
Pr. Cru(23) .48 .70 .62 .053
Total (87) .47 .74 .62 .054


1. Factors building up reflectance

An initial point to be discussed is if reflectance data from a vineyard, acquired by a satellite, contains information only from vine leaves or if this information is contaminated by other radiant sources, like soil, roads, or buildings. As for pixels located outside the vine parcels, we have already made it clear that great care was taken to precisely extract data well inside the selected parcels. The question is if within-vineyard features like soil, terrain slope, and row orientation play relevant roles in reflectance responses. In a previous paper we studied these influences in a different region, the Loire Valley (Ducati et al 2014). We will now present a short summary of that lengthy study. We argued that the soil contribution to reflectance is strongly dependent on the amount of illuminated soil, which depends on plant density, row orientation, and the hour and epoch of image acquisition. As for plant density, the standard density in Côte d’Or is about 10,000 plants per hectare; such a high density ensures, during the vegetative cycle, a relatively small soil visibility. However, some soil would still be seen from the vantage point of a satellite, and we have to consider its radiance relative to plant leaves, which can be reduced by shadows projected by the vines.

For this geometrical shadow effect, the proportion P of the inter-row surface that is shadowed by a continuous wall of vines is :

where h is the row height, d is the distance between rows, z is the Sun’s zenithal distance or elevation at a given moment, and a is the lateral illumination angle at the same time, that is, a is the angle between the orientation of a given row and the solar azimuth. Here, the angle a modulates the shadow’s length because when a = 0º, the row is oriented towards the Sun and the inter-row soil is entirely illuminated. In equation (2), the proportion P assumes values between 0 and 1, the value 1 corresponding to the extreme situation where the whole area between vine rows is in total shadow.

In Ducati et al (2014) we showed that a shadowed soil has little reflectance compared to fully-illuminated leaves. But it is seen that the amount of illuminated soil is critically dependent on row orientation; if in Loire we perceived that row orientations in vineyards are almost random, in Côte d’Or we saw during our field trips, further supported by visual inspection of high-resolution satellite images available via Internet services, that there is a certain dominance of down-hill rows. The exact geographical direction of these down-hill rows, however, varies depending on the local slope, which can favor an east exposure as well as north-east, south-east and other more extreme exposure. The Sun’s position at the moment of image acquisition (10h30 AM) was east, and the Sun’s zenithal distance in our images never exceeded 65°. Therefore, for the vineyards with down-hill rows, there is a certain amount of sunlight illuminating the soil between rows and its contribution may be considered.

However, for practical purposes, we will not discuss this issue further. This is due to the fact that during our field trips we did not detect any correlation between row orientation and vineyard category. In fact, we even observed that in those Grand Cru vineyards that are not “monopoles” the orientations of vine rows vary considering the plots inside the appellation; this is observed, for example, in the Clos-Vougeot and Échezeaux terrains and, of course, in the two other categories of this study, Premier Cru and Communale. Therefore, even if a contribution to pixel reflectance from the soil exists, this contribution is prone to be common to all three studied categories and would not be a differentiation factor in our study. This conclusion is not surprising, since if row orientation had been relevant to quality discrimination in Burgundy, the categories would have been separated by their respective row orientations a long time ago, which is not the case.

Another consideration, as stated at the beginning of this section, is terrain slope. The less prestigious vineyards in Burgundy are located over the plains on the east side of RN74, but in our study we have selected only a few plots in that zone. By far, the larger part of our sample was composed of vine plots on the hilly part. Here two points must be considered. The first one is that there is no clear slope-based criterion to separate the three categories in our sample; this is simply not observed. The second point is that the reflectance has its origin in vine leaves, and the amount of reflectance depends on the relative orientation of the plant leaf surface with respect to sunlight. However, plant leaves orient themselves to maximize sunlight absorption and do not follow terrain slope; therefore, leaf orientation, and so the part of pixel reflectance that is due to vegetation, tends to be independent of slope.

As a final remark on the possible relevance of slope and solar orientation to vineyard quality in Côte d’Or, we note that these variables were already included in the study by Wittendal (2004), with no positive results.

2. Our results

As presented in the Results section, category discrimination was fairly good over all four images. It could be expected that discrimination accuracy would be poorer for the winter data, since the reflected light comes from the soil, with little contribution from vegetation; however, the difference of winter data with respect to spring or summer data was small. This suggests that spectral differences between categories come primarily from the soil and when the soil is covered by vine leaves these differences persist and even seem to be reinforced by what the soil communicates to plant leaves. Another result that may deserve deeper investigation is the parcels that were placed in a different category in discriminant analysis. Overall, the results indicate that about 30% of all parcels presented spectral features typical of other categories. For example, some climats, which are formally classified as being generic appellations, carried spectral features derived from soil/leaf reflectance that put them in higher categories; the opposite also happened. For each region we had four images, and so we had four spectra treated separately in discriminant analysis. Uncertainties in reflectance determination could eventually change a spectrum to the point of putting it in another category, but some systematic changes did appear. These cases were as follows: one climat of the Communale class was considered to be Grand Cru in all four images; seven Communale climats were classified as Premier Cru four times; the opposite, Premier Cru classified as Communale, happened four times; and two Premier Cru climats were classified as Grand Cru in all four images. Visual inspection of images did not reveal any irregularities, and thus these cases deserve further and deeper investigation, possibly including more detailed soil information and field inspection.

Between 70% and 100% of Grand Cru vineyards were correctly identified in Beaune; these numbers were between 55.6% and 77.8% in Nuits. The question is: in our satellite reflectance data, what differentiates the Grand Cru category from the others? Results showed that a systematically higher reflectance seems to be real for Pinot Noir, and that for Chardonnay the differentiation process seems to be more subtle. It is known that ASTER satellite bands B4 to B8 are sensitive to water, meaning that higher water content in a target will reduce reflectance. Pinot Noir Grand Crus had higher reflectance. Whether this means less water in leaves and soil is not totally clear. Atkinson (2011) suggested that Grand Cru soils have a more stable capacity of water storage, due to sub-soil characteristics; peculiarities in top-soil, the one that is observed from space, seem to be less evident. This question can be better understood in future studies, based on more detailed spectral information and a wider database.

Parcels with Chardonnay were well separated from those with Pinot Noir. Again, the question is whether this is due to vegetation or soil. We already mentioned that during the vegetative season the soil is well covered by the canopy and also that this soil tends to be in shadow at the moment of image acquisition. This perception leads us to believe that the separation is due to spectral differences between the two varieties, expressed by leaf reflectance. In another investigation (Da Silva and Ducati 2009) we demonstrated that red and white grapes could be spectrally separated, the cause being the anthocyanin pigment which is present in leaf cells of red grapes. However, the 2003 winter result provides a deeper insight, since we have only soil reflectance and separation is still well done. This is for Côte de Beaune, where the best terroirs for Chardonnay and Pinot were pinpointed by centuries of study. The extremely variable Burgundian soil was and continues to be determinant for these choices (Fanet 2008; Pitiot and Servant 2010), and in a limestone-dominated environment, subtle differences made Chardonnay to be frequently placed where there is a certain dominance of clay with marl-limestone. Other factors can also play a role, like terrain orientation, with often more south and south-west facing parcels for Chardonnay and a tendency for Pinot Noir to occupy the higher parts of the hilly landscape. All these factors may induce spectral differences in the infrared. On the one hand, a southern exposure would make the top layers of soil drier, possibly increasing reflectance. But on the other hand, more clayed, marl-limestone terrains retain more water, and it is well known that higher soil humidity reduces the reflectance at infrared (Bowers and Hanks 1965); moreover, if Chardonnay is more frequently placed at the lower part of hills, again humidity will be higher, and both factors would reduce reflectance. Therefore, as for vineyard category discrimination, terrain and soil-terroir are the ultimate differentiation factors.


We have shown that satellite data is functional to reveal vineyard quality. This can be seen as surprising. Spectral differences should come essentially from soil features, which are transmitted to the vine and to vine leaves. In this study, Remote Sensing techniques were valuable in the characterization of terroirs, in this case with respect to vineyard quality. It can be noted that the ultimate factor defining quality is the resulting wine, and wine quality results not only from the soil components of the terroir concept but also from viticultural and winemaking practices. These non-geochemical factors can perhaps explain part of the “wrong” identifications commented in the Discussion, keeping the following comment in mind (Thackrey 2001): “I believe that the quality of French wine is due to a French genius for viticulture and winemaking, (...), not to the subsoil”. The fact that the differences between these climat categories are at least partially due to terroir characteristics suggests that terroir, or more precisely, the soil, influences the vine and vine canopy up to the point that detection of vineyard quality by Remote Sensing becomes possible.

Acknowledgements: ASTER L1B data were obtained through the online Data Pool at the NASA Land Processes Distributed Active Archive Center (LP DAAC), USGS/Earth Resources Observation and Science (EROS) Center, Sioux Falls, South Dakota (https://lpdaac.usgs.gov/get_data). JRD is grateful for the hospitality of the staff of the International Vintage Master at the École Supérieure d’Agriculture (ESA) of Angers, France, where this investigation was started during his stay as Visiting Professor in 2011, benefiting from Erasmus Mundus financial support.


  • Abrams M., Hook S., Ramachandran B., 2002. ASTER User Handbook. Available at: http://asterweb.jpl.nasa.gov/content/03_data/04_Documents/aster_user_guide_v2.pdf (accessed on November 12, 2013).
  • Altaweel M., 2005. The use of ASTER satellite imagery in archaeological contexts. Archaeological Prospection, 12(3), 151-166. doi:10.1002/arp.254
  • Atkinson J., 2011. Terroir and the Côte de Nuits. Journal of Wine Research, 22(1), 35-41. doi:10.1080/09571264.2011.550760
  • Berk A., Anderson G.P., Acharya P.K., Bernstein L.S., Muratov L., Lee J., Fox M., Adler-Golden S.M., Chetwynd J.H. Jr., Hoke M.L., Lockwood R.B., Gardner J.A., Cooley T.W., Borel C.C., Lewis P.E., Shettle E.P., 2006. MODTRAN5: 2006 Update. In: Proceedings of SPIE, vol. 6233, 62331F. doi:10.1117/12.665077
  • Blauth D.A., Ducati J.R., 2010. A Web-based system for vineyards management, relating inventory data, vectors and images. Computers and Electronics in Agriculture, 71(2), 182-188. doi:10.1016/j.compag.2010.01.007
  • Bowers S.A., Hanks R.J., 1965. Reflection of radiant energy from soils. Soil Science, 100(2), 130-138. doi:10.1097/00010694-196508000-00009
  • Bramley R.G.V., Proffitt A.P.B., 1999. Managing variability in viticultural production. Australian Grapegrower and Winemaker, 427(July 1999), 11-16.
  • Campbell J.B., Wynne R.H., 2011. Introduction to Remote Sensing. Guilford Press, New York.
  • Cemin G., Ducati J.R., 2011. Spectral discrimination of grape varieties and a search for terroir effects using remote sensing. Journal of Wine Research, 22(1), 57-78. doi:10.1080/09571264.2011.550762
  • Da Silva P., Ducati J.R., 2009. Spectral features of vineyards in south Brazil from ASTER imaging. International Journal of Remote Sensing, 30(23), 6085-6098. doi:10.1080/01431160902810612
  • Ducati J.R., Sarate R.E., Fachel J.M.G., 2014. Application of remote sensing techniques to discriminate between conventional and organic vineyards in the Loire Valley, France. Journal International des Sciences de la Vigne et du Vin, 48, 135-144. doi:10.20870/oeno-one.2014.48.3.1574
  • Fanet J., 2008. Les Terroirs du Vin. Hachette Pratique, Paris.
  • Iwasaki A., Tonooka H., 2005. Validation of crosstalk correction algorithm for ASTER/SWIR. IEEE Transactions on Geoscience and Remote Sensing, 43(12), 2747-2751. doi:10.1109/TGRS.2005.855066
  • Jensen J.R., 2007. Remote Sensing of the Environment: An Earth Resource Perspective. Prentice Hall, New York.
  • Lacar F.M., Lewis M.M., Grierson I.T., 2001. Use of hyperspectral reflectance for discrimination between grape varieties. In: Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (Sydney, Australia), vol. 6, pp. 2878-2880. doi:10.1109/igarss.2001.978192
  • Mather P., 1999. Computer Processing of Remotely-Sensed Images: An Introduction. Wiley, Chichester.
  • Pitiot S., Poupon P., 2009. Atlas des Grands Vignobles de Bourgogne. Lusigny-sur-Ouche, France.
  • Pitiot S., Servant J.-C., 2010. Les Vins de Bourgogne. Beaune, France.
  • Thackrey S., 2001. An Afternoon with Sean Thackrey. Available at: http://www.gangofpour.com/bree/profiles/thackrey/thackrey4.html (accessed on November 23, 2013).
  • Tucker C.J., 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment, 8(2), 127-150. doi:10.1016/0034-4257(79)90013-0
  • Van Leeuwen C., Seguin G., 2006. The concept of terroir in viticulture. Journal of Wine Research, 17(1), 1-10. doi:10.1080/09571260600633135
  • Wittendal F., 2004. Great Burgundy wines: a principal component analysis of “La Côte” vineyards. In: 11th Oenometrics Conference (Dijon, France), 31 p.
  • Zarco-Tejada P.J., Berjón A., López-Lozano R., Miller J.R., Martín Muñoz A.P., Cachorro V., González M.R., de Frutos A., 2005. Assessing vineyard condition with hyperspectral indices: leaf and canopy reflectance simulation in a row-structured discontinuous canopy. Remote Sensing of Environment, 99(3), 271-287. doi:10.1016/j.rse.2005.09.002


Jorge R. Ducati


Affiliation : Centro Estadual de Pesquisas em Sensoriamento Remoto e Meteorologia, Universidade Federal do Rio Grande do Sul, Av. Bento Goncalves 9500, CEP 91501-970, Porto Alegre, Brazil; Departamento de Astronomia, Instituto de Física, Universidade Federal do Rio Grande do Sul, Av. Bento Goncalves 9500, CEP 91501-970, Porto Alegre, Brazil; Visiting professor (2011), École Supérieure d‘Agriculture d’Angers, Groupe ESA, 55 rue Rabelais, 49007 Angers, France

Magno G. Bombassaro

Affiliation : Centro Estadual de Pesquisas em Sensoriamento Remoto e Meteorologia, Universidade Federal do Rio Grande do Sul, Av. Bento Goncalves 9500, CEP 91501-970, Porto Alegre, Brazil

Jandyra M. G. Fachel

Affiliation : Departamento de Estatística, Instituto de Matemática, Universidade Federal do Rio Grande do Sul, Av. Bento Goncalves 9500, CEP 91501-970, Porto Alegre, Brazil


No supporting information for this article

Article statistics

Views: 1525


PDF: 556