Content uploaded by Andrew Reynolds
Author content
All content in this area was uploaded by Andrew Reynolds on Jul 28, 2016
Content may be subject to copyright.
APPLICATION OF REMOTE SENSING BY UNMANNED AERIAL VEHICLES TO MAP
VARIABILITY IN ONTARIO RIESLING AND CABERNET FRANC VINEYARDS
Andrew G. REYNOLDS1*, Ralph BROWN2, Marilyne JOLLINEAU3, Adam SHEMROCK4, Elena KOTSAKI1,
Hyun-Suk LEE1, Wei ZHENG5
1Cool Climate Oenology and Viticulture Institute, Brock University, St. Catharines, Ontario, Canada; 2School of
Engineering, University of Guelph, Guelph, ON, Canada; 3Dept. of Geography, Brock University, St. Catharines, Ontario,
Canada; 4Air-Tech Solutions, Kingston, Ontario, Canada; 5Dept. of Agriculture and Food, University of La Rioja, Logroño,
La Rioja, Spain. *Corresponding author: A.G. Reynolds, Email: areynolds@brocku.ca
Abstract. The objective of this investigation was to verify usefulness of proximal sensing technology and unmanned aerial
vehicles (UAVs) for mapping variables e.g., vine size (potential vigor), soil and vine water status, yield, fruit composition,
and virus incidence in vineyards. Twelve Niagara Peninsula sites (six each of Riesling and Cabernet franc) were chosen in
2015. Data were collected from a grid of vines (≈ 80 per vineyard) geolocated by GPS. Soil moisture and leaf water
potential (ψ) data (three times during the growing season; June to September) and yield components/berry composition were
collected. Ground based GreenSeekerTM data were likewise acquired June to September, while multi-spectral UAV data
were obtained at veraison and processed into geo-referenced high spatial resolution maps of biophysical indices (e.g.,
NDVI). Following harvest, yield/berry composition maps were also prepared. These data layers in conjunction with
growing/dormant season sentinel vine data [e.g. soil moisture, leaf ψ, vine size, winter hardiness (LT50)], were used for map
creation. Vine size, LT50, yield, berry weight, and berry composition data were correlated in several vineyards to NDVI and
other data acquired with the UAV and GreenSeekerTM, while soil and vine water status, and yield components showed direct
relationships with NDVI. Spatial relationships were also apparent from examination of the maps. Principal components
analysis confirmed these relationships. Map analysis to determine spatial relationships was accomplished by calculation of
Moran’s I and k-means clustering. NDVI values were considerably higher in GreenSeeker maps vs. those from UAV
flights. Water status zones, and those of several fruit composition variables, were correlated with UAV-derived NDVI.
Preliminary conclusions suggest that UAVs have significant potential to identify zones of superior fruit composition.
Keywords: Precision viticulture, drones, leaf water potential, soil moisture
1 INTRODUCTION
The Ontario wine industry produces ≈ 65,000 tonnes of grapes and consists of cultivars such as Riesling, Chardonnay, and
Cabernet franc, with lesser quantities of Merlot, Cabernet Sauvignon, and Pinot noir (www.grapegrowersofontario.com).
Soils are characterized as variable as a result of widespread glacial activity over 10,000 years ago, and consequently many
vineyards are situated on several soil series that can range widely in texture, depth of solum, and water-holding capacity [1].
This variability in soil characteristics can impact vine vigor, yield, and water status. A significant growth in the number of
small artisanal wineries has permitted production of wines that are unique to individual vineyard sites and in some cases
unique to specific vineyard blocks. In the past 10-15 years this interest has expanded to include identification of unique
portions of vineyard blocks, some < 1 ha, that might be capable of producing extremely high-value wines based upon yield,
vine size, or water status-based quality levels.
The recent introduction of Greenseeker and other proximal sensing technologies might allow growers to identify unique
zones within vineyards without use of aircraft [2-4]. If unique zones can be identified easily from the ground, it is possible
that different wine products of varying price points could be created from these zones with minimal cost from the producer.
Data validation would be required as with remote sensing to determine relationships between proximally-sensed data and
other variables of agricultural relevance, but the proximally-sensed data are relatively easy to access. Ground-based
(proximal sensing) technologies are relatively recent introductions and their evaluation in viticulture is uncommon. Their
initial use was for continuous compilation of NDVI data from vine canopies [2-4]. Proximal sensing correlated with vine
size and berry color in Merlot vineyards in northern Greece [5]. NDVI sensors explained variation in biomass; its
relationship to vine size was nonlinear and was best described by a quadratic regression. A linear correlation to stable
isotope content in leaves (13C and 15N) provided evidence that canopy reflectance detected plant stresses as a result of water
shortage and limited N fertilizer uptake [6,7]. It was also successful in detection of downy mildew disease levels in Italian
vineyards [4]. Despite these advances research into proximal sensing has mainly been limited to agronomic crops such as
corn [8,9] and thus far no work in Canada has been carried out.
Attempts have been made with limited success to identify unique zones using remote sensing and to thereafter associate
these remotely-sensed regions with variables such as vine water status, soil moisture, vine vigor, yield, and berry
composition. Although less laborious than manual data collection and subsequent production of a multitude of maps, use of
aircraft is costly and remote sensing in agricultural systems is often imprecise [5]. The data that is collected must be
converted to variables, e.g., normalized difference vegetative index (NDVI) through computer software such as ENVI [10-
13]. Moreover, validation of data acquired by remote sensing is still necessary to determine whether ostensibly-unique
zones are relevant from a standpoint of physiology, productivity, and berry composition. One particular challenge involved
masking of cover crop NDVI from all images to assess the vine canopy-specific NDVI [12,13].
381
In viticultural applications, remote sensing has been used in modelling vegetative growth, and to infer grape composition
from those measurements. Johnson et al. [14] used remotely sensed multispectral data to delineate a Chardonnay vineyard
into small-lot production zones. They found that vine size was related to vigor zones identified by airborne images. Vigor
zones were also related to vine water status and grape composition variables. Thus, indirectly, remote sensing was used to
predict vineyard status and grape composition, with direct implications for wine quality [14]. Relationships between
vegetation indices (VIs) and vegetative growth were further explored by Dobrowski et al. [15]. There were strong, positive
correlations between the extracted VIs and pruning weights (vine size) in two years. Additionally, relationships established
in the first season were able to predict the vine size in the second season [15].
The ability of remote sensing to be used to directly predict grape composition variables was explored by Lamb et al. [16].
They found that re-sampling the image to a final pixel size approximately equal to the distance between rows, effectively
combining vine size and density information into a single pixel, resulted in the strongest correlations to color and phenols.
Strongest negative correlations between NDVI and color and phenols occurred around veraison [16]. In Languedoc
(France), Acevedo-Opazo et al. [17] performed a study on remotely sensed VIs, vine water status, and grape composition on
a number of winegrape cultivars in non-irrigated vineyards. Temporally stable relationships occurred between zones
delineated based on the NDVI and vegetative growth, vine water status, and yield. These zones were also consistent with
soil type. They concluded that a combination of remotely sensed data with intimate vineyard knowledge, especially of the
soil, is needed to predict grape composition and ultimately wine quality [17].
Overall, remote sensing has been proven as a useful tool for monitoring vineyard vegetative growth, and for making
inferences about grape composition from multispectral measurements. In Ontario, NDVI data from remote sensing was
associated with numerous variables in Riesling vineyards, including vine water status, yield components, and berry
composition [12,13]. Similar applications were made in Pinot noir vineyards [11]. Remote sensing proved to be a good tool
to determine color and phenolic potential of grapes, in addition to water status, yield and vine size [11]. These studies were
unique by employing remote sensing in cover-cropped vineyards and thereafter using protocols for excluding the spectral
reflectance contributed by the inter-row vegetation.
Remote sensing has likewise been used for making inferences about grape composition from multispectral measurements
[18,19]. However, employment of UAVs for remote sensing in vineyards is a relatively new area of research, heretofore
untested in Canada, and capable of acquiring high resolution spatial data without high cost of aircraft. As with proximal
sensing there has been little published, and most has confirmed their ability to acquire NDVI and related images [20-22].
Zarco-Tejada et al. [23] explored relationships between photosynthesis and chlorophyll fluorescence by hyperspectral
imagery captured via UAVs. Significant relationships were demonstrated between photosynthesis and chlorophyll
fluorescence vs. remote measurements. Other relationships were demonstrated between both chlorophyll a/b and leaf
carotenoids vs. several VIs based on multispectral images acquired by UAVs [24]. UAVs were likewise utilized for
assessment of vineyard water status by correlation of stem water potential with NDVI [25]. Further relationships were
elucidated between several VIs including NDVI vs. leaf water potential and stomatal conductance [26]. Additionally,
photochemical reflectance index (PRI), renormalized difference vegetation index (RDVI), and red edge index were
correlated to water status variables [26]. Other stresses such as those nutritional in nature have been detected by UAVs; e.g.
NDVI was correlated with levels of iron chlorosis, leaf carotenoid pigments, and grape leaf and berry anthocyanins [27].
2 MATERIALS AND METHODS
Sites and cultivars. Six each of Cabernet franc and Riesling vineyards (1-2 ha in area) in different Niagara sub-appellations
were chosen. The sites represented the following sub-appellations: Niagara Lakeshore (Buis), Creek Shores (Lambert), St.
Davids Bench (Chateau des Charmes), Lincoln Lakeshore north (George), Lincoln Lakeshore south (Hughes; Kocsis),
Beamsville Bench (Cave Spring). Soil types [1] varied substantially in these sub-appellations from well-drained coarse-
textured Tavistock and Vineland series (Niagara Lakeshore, Lincoln Lakeshore north), to moderately-well drained
Chinguacousy (Creek Shores, Beamsville Bench), and poorly-drained Jeddo (Lincoln Lakeshore south) and Beverly/Toledo
soils (St. Davids Bench). These soils provided a range of water-holding capacities that impacted vine water status.
Geolocation. Vineyard blocks were GPS-delineated to determine shape using a Trimble Handheld GPS, equipped with
TerraSync software (Trimble Navigation Ltd., Sunnyvale, CA). Sentinel vines (80-100) were identified in a ≈ 8m x 8m grid
within each vineyard and geolocated by the aforementioned GPS system. Post-collection differential correction was
performed using GPS Pathfinder Office (Trimble Navigation Ltd.) to sub-metre accuracy (≈30-50 cm) using the Port
Weller, ON base station correction. Field measurements and berry samples were taken on these vines in all vintages.
Soil moisture, leaf water potential (ψ), and stomatal conductance (Gs). Vineyard soil moisture (SM) was measured by
time domain reflectometry (TDR) using the Field Scout TDR 300 Soil Moisture Meter (Spectrum Technologies, East
Plainfield, IL). The volumetric water content mode was used. A pair of 20-cm stainless steel probes were used for
measurements at all sentinel vines. Measurements took place at berry set, lag phase, and veraison on all sentinel vines. Two
measurements ≈ 10 cm from the vine trunk were taken. Vine water status was measured using midday leaf ψ by pressure
bomb (Soil Moisture Equipment, Santa Barbara, CA). Measurements were made only at designated leaf ψ vines (≈ 15 per
vineyard block), on the same days as SM measurements, from 1000h-1400h (ca. solar noon), under full sun. Leaf Gs was
measured by a hand-held porometer (Decagon Devices, Pullman, WA) on those vines used for leaf ψ.
382
Yield components and vine size. Harvest dates were at the discretion of vineyard managers. Fruit from each sentinel vine
was harvested, cluster number determined, and fruit weighed using a portable field scale. Mean cluster weight was
calculated from these data. Cane prunings were retained and weighed to determine vine size during dormancy.
Berry analysis. Basic composition. A 100-berry sample was taken from each sentinel vine at harvest and frozen at -25°C.
Each sample was weighed to determine mean berry weight, and placed in a beaker in a water bath at 80°C for one hour to
dissolve precipitated tartrates. Samples were homogenized in a commercial juicer. Soluble solids were measured using a
benchtop refractometer (UV Corp., Buffalo, NY). Berry pH was measured using an Accumet pH/ion meter and VWR
Symphony electrode (Fisher, Mississauga, ON). Juice samples (≈35 mL) were clarified by centrifugation at 4500 g for 10
minutes. The remaining juice (≈20 mL) was retained at -25°C for subsequent color analysis (Cabernet franc). Titratable
acidity was measured on centrifuged juice, titrated to an endpoint of pH 8.2 with 0.1N NaOH using a PC-Titrate autotitrator
(Man-Tech Associates, Guelph, ON).
Colour/hue, total anthocyanins, total phenols (Cabernet franc; monoterpenes (Riesling). Re-frozen samples were
centrifuged at 4500 g at 4°C. Absorbances were measured using a spectrophotometric method at 420 nm and 520 nm using
an Ultrospec 2100 pro UV-VIS spectrophotometer (Biochrom Ltd., Cambridge, UK). Color intensity was calculated as
A420+A520, and hue as A420/A520. Total anthocyanins were quantified using the pH shift method [28]. Total phenols
were quantified using the Folin-Ciocalteu method [29] and expressed as mg/L gallic acid. Monoterpenes were analyzed on
250-berry samples taken from leaf ψ vines using a distillation method [30] modified by Reynolds & Wardle [31]. Free
volatile and potentially-volatile terpene concentrations were expressed in mg/kg.
Spatial mapping. GPS coordinates from vineyard blocks and sentinel vines were imported into a GIS environment (ArcGIS
10.3; Environmental Systems Research Institute, Redlands, CA) and linked to all point data collected from sentinel vines.
Spatial interpolation techniques (i.e., kriging) applied to these data were used to estimate the value of vineyard variables at
unsampled locations. This permitted further spatial data analyses; including the integration of these data with the remote-
sensing datasets.
Proximal sensing. A GreenSeeker unit (Trimble Navigation, Englewood, CO) mounted on a four-wheel-drive vehicle was
used to collect Normalized Difference Vegetation Index (NDVI) data on dates close to soil moisture and leaf ψ data
collection. An additional reading was collected mid-September. Data were imported into Farmworks software (Trimble
Navigation, Englewood, CO) and spatial maps created. Shapefiles were thereafter imported into the ArcGIS geodatabase.
GPS coordinates identical to the sentinel vines were identified and NDVI data corresponding to these coordinates were
extracted for statistical analyses.
UAV and sensors. The 2015 UAV flight corresponded to the veraison soil moisture, leaf ψ, and GreenSeeker data
collection. Image acquisition was performed using the ING Robotic Aviation Responder (ING Robotic Aviation, Orléans,
ON), a UAV supplied by Air-Tech Solutions, Kingston, ON. The UAV was powered by an electric motor providing 30 min
of endurance and a 9-kg payload, and equipped with an autopilot system allowing a 1000-m visual range and 5-km radio
line of sight. The UAV was flown at a 122-m elevation and a 60 km/h maximum speed. Two sensors were used for image
acquisition. The first operated in the visible and near-infrared portions of the electromagnetic spectrum (EMS) (Mini-MCA
6; Tetracam, Inc., Chatsworth, CA) utilizing five spectral bands (blue, green, red, red edge, near-infrared) equipped with an
incident light sensor. It ensured acquisition at a resolution of 7.8 megapixels, representing a spatial resolution of 4 cm at an
altitude of 100 m. The second sensor operated in the thermal–infrared (IR) portion of the EMS (A65 thermal imaging
camera; FLIR Systems, Burlington, ON). It ensured acquisition of imagery in the thermal-IR range covering 7500 to 13500
nm at a resolution of 0.3 megapixels, which represents a spatial resolution of 9 cm at a 100-m altitude. Equipment onboard
also consisted of a GPS and an inertial station to correct anomalies in flight attitude (i.e., yaw, pitch, and roll) and ensuring
the verticality and orientation of imaging. This equipment was complemented by a ground receiving station that provided
real-time feedback on the position of the aircraft and its imaging.
Image acquisition, pre-processing, processing. Image acquisition was performed over each vineyard block. Data were
stored onboard and retrieved after the flight mission. Geometric correction was performed to correct the image geometry.
Geometric distortions caused by changes in UAV attitude and altitude were corrected using the information provided by the
inertial station. Radiometric correction was performed to correct the effects of vignetting and bidirectional reflectance. The
series of images acquired during each flight was assembled into mosaics by selecting the overlapping areas near nadir to
limit the viewing angle and the problems of directional effects. Once assembled and corrected NDVI was calculated on
mosaics. Pixel values corresponding to sentinel vines were extracted and compiled into a geodatabase that included all field
based variables (e.g. leaf ψ).
Data analysis. Basic correlations and regressions were performed on the ground-based, proximally-sensed, and UAV data
to determine relationships, particularly those between proximally sensed/UAV data vs. ground-based data using SAS 9.4
(SAS Institute, Cary, NC). PCA was likewise performed by XLStat (Addinsoft, Paris, France). Maps were created for all
variables using ArcGIS. K-means clustering and Moran’s I were used to determine spatial relationships.
383
3 RESULTS AND DISCUSSION
PCA. UAV NDVI and NIR indices were correlated with vine size in all Riesling vineyards (Fig. 1). Other noteworthy
associations included UAV indices and berry weight (five sites), TA (two sites—Buis, George), and proximally-sensed
NDVI (four sites). Inverse correlations of note with UAV data included soil moisture and leaf ψ (five sites). Direct
correlations were also noted for NDVI/NIR and at least one LT50 measurement for four sites: LT50 03 (Buis), LT50 01
(Pondview), LT50 02 (Hughes, Cave Spring). Yield and cluster number were not consistently related to VIs; they were
inversely correlated in three sites (Buis, Hughes, Pondview) but unrelated in two others. UAV NDVI was likewise
correlated with vine size in all Cabernet franc vineyards (data not shown). Other noteworthy associations included UAV
indices and soil moisture (three sites), leaf ψ (two sites), berry weight (five sites), TA (three sites), yield/cluster number
(four sites), and proximally-sensed NDVI (four sites).
Maps. Maps for Buis Riesling vineyard 2015 are depicted in Fig. 2. The UAV NDVI map showed a low NDVI zone on the
west side of the vineyard (Fig. 2A). This corresponded closely with highest regions from the thermal camera (Fig. 2B), and
lowest regions of NDVI by GreenSeeker (Fig. 2C), leaf ψ (Fig. 2D), to a limited degree soil moisture (Fig. 2E), vine size
(Fig. 2F), berry weight (Fig. 2G), and TA (Fig. 2I), and higher LT50 (i.e. less winter hardy; Fig. 2H). Conversely, those
high-NDVI zones identified by the UAV were typically also high in terms of thermal camera data, NDVI by GreenSeeker,
leaf ψ, soil moisture, vine size, berry weight, and TA, and lower in LT50 (i.e. more winter hardy). The Cabernet franc
vineyard (data not shown) likewise showed clustering in the UAV data with low NDVI and high thermal zones in the south
end of the block. These corresponded to low NDVI areas delineated by GreenSeeker, low soil moisture and leaf ψ areas,
low vine size and berry weight, and higher LT50 zones. These showed some spatial correlation with high TA and low Brix
areas, but pH and overall yield were not strongly related spatially.
In most other vineyard blocks the NDVI maps based upon drone flights were comparable to maps created from
GreenSeeker data. There were occasional situations in which maps from thermal data were inversely correlated spatially
with NDVI (data not shown). Most frequent spatial correlations in Riesling with UAV and GreenSeeker NDVI zones were
leaf ψ, Gs, vine size, berry weight, and TA. Noteworthy inverse spatial correlations included soil moisture and yield.
4 LITERATURE CITED
1. KINGSTON, M.S. AND PRESANT, E.W., 1989. The Soils of the Regional Municipality of Niagara. Report 60 of the Ontario
Institute of Pedology, Guelph, ON.
2. DRISSI, R., GOUTOULY, J., FORGET, D. & GAUDILLÈRE, J., 2009. Nondestructive measurement of grapevine leaf area by
ground normalized difference vegetation index. Agronomy J., 101, 226-231.
3. MAZZETTO, F., CALCANTE, A., & MENA, A., 2009. Comparing commercial optical sensors for crop monitoring tasks in
precision viticulture. J. Agric. Eng., 1, 11-18.
4. MAZZETTO, F., CALCANTE, A., MENA, A., & SACCO, P., 2011. Comparing commercial optical sensors for crop monitoring
tasks in precision viticulture. J. Agric. Eng., 2, 1-9.
5. STAMATIADIS, S., TASKOS, D., TSADILAS, C., CHRISTOFIDES, C., TSADILA, E., & SCHEPERS, J.S., 2006. Relation of
ground-sensor canopy reflectance to biomass production and grape color in two Merlot vineyards. Am. J. Enol. Vitic., 57, 415-422.
6. STAMATIADIS, S., TASKOS, D., TSADILA, E., CHRISTOFIDES, C., TSADILAS, C., & SCHEPERS, J.S., 2010. Comparison of
passive and active canopy sensors for the estimation of vine biomass production. Prec. Agric, 11, 306-315.
7. TASKOS, D.G., KOUNDOURAS, S., STAMATIADIS, S., ZIOZIOU, E., NIKOLAOU, N., KARAKIOULAKIS, K., &
THEODOROU, N., 2013. The effects of variable nitrogen and water supply on grapevine canopy reflectance. Proc. Ampelos Conf.,
Trends in Vitiviniculture Development, 30-31 May 2013, Santorini, Greece. http://ampelos2013.conferences.gr/.
8. BARKER, D. & SAWYER, J., 2010. Using active canopy sensors to quantify corn nitrogen stress and nitrogen application rate.
Agronomy J., 102, 964-971.
9. BAUSCH, W.C. & DELGADO, J.A., 2003. Ground-based sensing of plant nitrogen status in irrigated corn to improve nitrogen
management. In T. VAN TOAI et al. (Eds.), Digital Imaging and Spectral Techniques: Applications to Precision Agriculture and Crop
Physiology (pp. 145–157). ASA Special Publication 66. Madison, WI: ASA, CSSA, SSSA.
10.HALL, A., LOUIS, J. & LAMB, D., 2001. A method for extracting detailed information from high resolution multispectral images of
vineyards. In: Proceedings of the 6th International Conference on Geocomputation; University of Queensland, Brisbane.
11.LEDDERHOF, D., BROWN, R., REYNOLDS, A.G. & JOLLINEAU, M., 2015. Using remote sensing to understand Pinot Noir
vineyard variability in Ontario. Can. J. Plant Sci., 96, 89-108.
12.MARCINIAK, M., BROWN, R.B., REYNOLDS, A.G. & JOLLINEAU, M., 2015. Use of GPS, GIS and remote sensing to understand
the terroir of the Niagara Peninsula. Remote sensing applications in a Riesling vineyard. J. Int. Sci. Vigne Vin, 49, 1-26.
13.REYNOLDS, A.G., MARCINIAK, M., BROWN, R.B., TREMBLAY, L., & BAISSAS, L., 2010. Using GPS, GIS and airborne
imaging to understand Niagara terroir. Prog. Agric. et Vitic., 127(12), 259-274.
14.JOHNSON, L., BOSCH, D., WILLIAMS, D., & LOBITZ, B., 2001. Remote sensing of vineyard management zones: implications for
wine quality. Appl. Engin. Agric., 17, 557-560.
384
15.DOBROWSKI, S., USTIN, S. & WOLPERT, J., 2003. Grapevine dormant pruning weight prediction using remotely sensed data.
Austral. J. Grape Wine Res., 9, 177-182.
16.LAMB, D., WEEDON, M. & BRAMLEY, R.G.V., 2004. Using remote sensing to predict grape phenolics at harvest in a Cabernet
Sauvignon vineyard: timing observations against vine phenology and optimizing image resolution. Austral. J. Grape Wine Res., 10, 46-
54.
17.ACEVEDO-OPAZO, C., TISSEYRE, B., GUILLAUME, S. & OJEDA, H., 2008. The potential of high spatial resolution information
to define within-vineyard zones related to vine water status. Prec. Agric., 9, 285-302.
18.GITELSON, A., KAUFMAN, Y. & MERZLYAK, M., 1996. Use of a green channel in remote sensing of global vegetation from
EOS-MODIS. Remote Sensing Environ., 58, 289-298.
19.POST, D., FIMBRES, A., MATTHIAS, A., SANO, E., ACCIOLY, L., BATCHILY, A. & FERREIRA, L., 2000. Predicting soil
albedo from soil color and spectral reflectance data. Soil Sci. Soc. Am. J., 64, 1027-1034.
20.GUILLEN-CLIMENT, M.L., ZARCO-TEJADA, P.J., BERNI, J.A., NORTH, P.R., & VILLALOBOS, F.J., 2012. Mapping radiation
interception in row-structured orchards using 3D simulation and high-resolution airborne imagery acquired from a UAV. Prec. Agric.,
13, 473–500.
21.MATESE, A., PRIMICERIO, J., DI GENNARO, F., FIORILLO, E., VACCARI, F.P., GENESIO, L., 2013. Development and
application of an autonomous and flexible unmanned aerial vehicle for precision viticulture. Acta Hortic. 978: International Workshop on
Vineyard Mechanization and Grape and Wine Quality, pp. 63-69.
22.TURNER, D., LUCIEER, A., & WATSON, C., 2011. Development of an unmanned aerial vehicle (UAV) for hyper resolution
vineyard mapping based on visible, multispectral, and thermal imagery. Proc. 34th International Symposium on Remote Sensing of
Environment, Sydney, Australia, 10-15 April 2011; http://www.isprs.org/proceedings/2011/ISRSE-34/.
23.ZARCO-TEJADA, P.J., CATALINA, A., GONZÁLEZ, M.R., & MARTÍN, P., 2013. Relationships between net photosynthesis and
steady-state chlorophyll fluorescence retrieved from airborne hyperspectral imagery. Remote Sensing Environ., 136, 247–258.
24.ZARCO-TEJADA, P.J., GUILLÉN-CLIMENT, M.L., HERNÁNDEZ-CLEMENTE, R., CATALINA, A., GONZÁLEZ, M.R., &
MARTÍN, P., 2013. Estimating leaf carotenoid content in vineyards using high resolution hyperspectral imagery acquired from an
unmanned aerial vehicle (UAV). Agric. and Forest Meteorol., 171– 172, 281– 294.
25.BALUJA, J., DIAGO, M.P, BALDA, P., ZORER, R., MEGGIO, F., MORALES, F., & TARDAGUILA, J., 2012. Assessment of
vineyard water status variability by thermal and multispectral imagery using an unmanned aerial vehicle (UAV). Irrig. Sci., 30, 511–522.
26.ZARCO-TEJADA, P.J., GONZÁLEZ-DUGO, V., WILLIAMS, L.E., SUÁREZ, L., BERNI, J.A.J., GOLDHAMER, D., & FERERES,
E., 2013. A PRI-based water stress index combining structural and chlorophyll effects: Assessment using diurnal narrow-band airborne
imagery and the CWSI thermal index. Remote Sensing Environ., 138, 38–50.
27.MEGGIO, F., ZARCO-TEJADA, P.J., NÚÑEZ, L.C., SEPULCRE-CANTÓ, G., GONZÁLEZ, M.R., & MARTÍN, P., 2010. Grape
quality assessment in vineyards affected by iron deficiency chlorosis using narrow-band physiological remote sensing indices. Remote
Sensing Environ., 114, 1968–1986.
28.FULEKI, T. & FRANCIS, F.J., 1968. Quantitative methods for anthocyanins. 1. Extraction and determination of total anthocyanin in
cranberries. J. Food Sci. 33, 72-77.
29.SLINKARD, K. & SINGLETON, V., 1977. Total phenol analysis: automation and comparison with manual methods. Am. J. Enol.
Vitic. 28, 49-55.
30.DIMITRIADIS, E. & WILLIAMS, P.J., 1984. The development and use of a rapid analytical technique for estimation of free and
potentially volatile monoterpene flavorants of grapes. Am. J. Enol. Vitic., 35, 66-71.
31.REYNOLDS, A.G. & WARDLE, D.A., 1989. Influence of fruit microclimate on monoterpene levels of Gewurztraminer. Am. J. Enol.
Vitic., 40, 149-154.
385
Figure 1. Principal components analysis of six Ontario Riesling vineyards, 2015.
A-BUIS
B-PONDVIEW
C-CHATEAU DES CHARMES
D-GEORGE
E-HUGHES
F-CAVE SPRING
Soil
Moisture
Leaf ψ
NDVI
NDVI UAV
NIR UAV
LT50 02
LT50 03
Mean
LT50
Clusters
Yield
Vine size
Berry wt
Brix
pH
TA
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1
-1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
F2 (15.97 %)
F1 (19.37 %)
Variables (axes F1 and F2: 35.35 %)
Soil
Moisture
Leaf ψ
NDVI
NDVI UAV
NIR UAV
LT50 01
LT50 02
LT50 03
Mean
LT50
Clusters
Yield
Vine size
Berry wt
Brix
pH
TA
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1
-1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
F2 (14.79 %)
F1 (23.39 %)
Variables (axes F1 and F2: 38.18 %)
Soil
Moisture
Leaf ψ
NDVI
NDVI UAV
NIR UAV
LT50 01
LT50 02
LT50 03
Mean
LT50
Clusters
Yield
Vine size
Berry wt
Brix
pH
TA
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1
-1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
F2 (13.96 %)
F1 (20.12 %)
Variables (axes F1 and F2: 34.09 %)
Soil
Moisture
Leaf ψ
NDVI NDVI UAV
NIR UAV
LT50 01
LT50 02
LT50 03
Mean
LT50
Clusters
Yield
Vine size
Berry wt
Brix
pH
TA
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1
-1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
F2 (16.84 %)
F1 (22.36 %)
Variables (axes F1 and F2: 39.19 %)
Soil
Moisture
Leaf ψ
NDVI
NDVI UAV
NIR UAV
LT50 01
LT50 02
LT50 03
Mean
LT50
Clusters
Yield
Vine size
Berry wt
Brix
pH
TA
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1
-1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
F2 (14.18 %)
F1 (18.34 %)
Variables (axes F1 and F2: 32.52 %)
Soil
Moisture
Leaf ψ
NDVI
NDVI UAV
LT50 01
LT50 02
LT50 03
Mean
LT50
Vine size
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1
-1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
F2 (23.04 %)
F1 (31.03 %)
Variables (axes F1 and F2: 54.06 %)
386
A-NDVI (UAV)
B-UAV THERMAL
C-NDVI (GREENSEEKER)
D-LEAF WATER POTENTIAL
E-SOIL MOISTURE
F-VINE SIZE
G-BERRY WEIGHT
H-LT50
I-TA
Figure 2. Maps of several related variables, Buis Glenlake Riesling vineyard, Niagara-on-the-Lake, ON, 2015. Scale 1:1812 (UAV maps), 1:2500
(others).
387
