GRAPEGROWING WINE EAST Vitiulturl Mpping y UAVs, Prt 2 Applying unmnned eril vehiles in vitiulture By Andrew G. Reynolds, Rlph Brown, Mrilyne Jollineu, Adm Shemrok, Hyun-Suk Lee, Brinn Dorin, Mehdi Shnin, Bozhong Meng Attempts hve een mde, with limited suess, to identify unique zones using remote sensing nd to ssoite regions with vriles suh s vine wter sttus, soil moisture, vine vigor, yield nd erry omposition. Although less lorious thn mnul dt olletion nd prodution of multitude of mps, use of onventionl irrft n e ostly, nd remote sensing in griulturl systems is often impreise. 30 Dt must e onverted to vriles, e.g., normlized differene vegettive index (NDVI) dt through environmentl softwre suh s Environment for Visulizing Imges (ENVI). 9 Moreover, vlidtion of dt quired y remote sensing is still neessry to determine whether ostensily unique zones re relevnt from stndpoint of physiology, yield, nd erry omposition. One prtiulr hllenge involved msking of over rop NDVI from ll imges to ssess the vine nopyspeifi NDVI. 15, 16 In vitiulturl pplitions, remote sensing hs een used in modelling vegettive growth, nd to infer grpe omposition from those mesurements. Remotely sensed multispetrl dt ws used to delinete Chrdonny vineyrd EDITOR S NOTE This is the seond instllment of two-prt series on using unmnned eril vehiles (UAV) to otin dt to help vineyrds implement preision vitiulture. In the first rtile, the uthors looked t the different UAV pltforms, hrdwre, sensors nd imge proessing. Here, the uthors disuss the use of UAVs in vitiulture nd inlude the results of study done in Ontrio vineyrds. Figure 1: The prinipl omponents nlysis for two vineyrds in 2015. Figure 1 (left): the Buis Riesling vineyrd in Nigr-on-the-Lke, Ont. nd Figure 1 (right): the George Cernet Frn vineyrd in Vinelnd, Ont. into smll-lot prodution zones. 12 Vine size (weight of ne prunings; n estimte of vigor) ws relted to vigor zones identified y irorne imges. Vigor zones lso were relted to vine wter sttus nd grpe omposition vriles. Thus, indiretly, remote sensing ws used to predit vineyrd sttus nd grpe omposition, with diret implitions for wine qulity. Reltionships etween vegettion indies (VIs) nd vegettive growth were explored y Dorowski et l. 7 There were strong, positive orreltions etween extrted VIs nd vine size over two yers. Additionlly, reltionships estlished in the first seson were le to predit the vine size in the seond seson. Remote sensing n e used to predit grpe omposition vriles. 14 Re-smpling the imge to finl pixel size pproximtely equl to the distne etween rows, effetively omining vine size nd density informtion into single pixel, resulted in the strongest orreltions to olor nd phenols. The strongest negtive orreltions etween NDVI nd olor nd phenols ourred round verison. In Lnguedo (Frne), Aevedo-Opzo et l. 1 investigted remotely sensed VIs, vine wter sttus, nd grpe omposition on numer of wine grpe vrieties in non-irrigted vineyrds. Temporlly stle reltionships ourred etween zones sed on the NDVI nd vegettive growth, vine wter sttus nd yield. These zones lso were onsistent with soil type. KEY POINTS Remote sensing y unmnned eril vehiles (UAVs) hs een useful in monitoring vineyrd vegettive growth nd to mke inferenes out grpe omposition. A study in Ontrio used UAVs to exmine orreltive reltionships etween vegettion indies nd vineyrd vriles inluding vine size, yield, erry weight, erry omposition nd winter hrdiness. UAVs hve signifint potentil to identify zones of superior fruit omposition nd potentil wine qulity. 60 WINES&VINES July 2018
WINE EAST GRAPEGROWING They onluded tht omintion of remotely sensed dt with intimte vineyrd knowledge, espeilly of the soil, is needed to predit grpe omposition nd ultimtely wine qulity. Remote sensing hs proved to e useful tool for monitoring vineyrd vegettive growth, nd for mking inferenes out grpe omposition from multispetrl mesurements. In Ontrio, NDVI dt from remote sensing ws ssoited with numerous vriles in Riesling vineyrds, inluding vine wter sttus, yield omponents nd erry omposition.16 Similr pplitions were mde in Pinot Noir vineyrds, nd proved to e good tool to determine olor nd phenoli potentil of grpes, in ddition to wter sttus, yield nd vine size.15 These studies were unique in their employment of remote sensing in overropped vineyrds nd therefter using protools for exluding the spetrl refletne ontriuted y inter-row over rops. UAV-sed remote sensing likewise hs een used for mking inferenes out grpe omposition from multispetrl mesurements.23 However, use of UAVs for remote sensing in vineyrds is reltively new re of reserh, thus fr untested in Cnd, nd ple of quiring high-resolution sptil dt without the high ost of onventionl irrft. As with proximl sensing, there hs een little pulished, nd most of tht hs onfirmed n ility to quire NDVI nd relted imges. Reltionships were found etween photosynthesis nd hlorophyll fluoresene y hyperspetrl imgery ptured vi UAVs, s well s etween oth photosynthesis nd hlorophyll fluoresene vs. remote mesurements.35 Other reltionships were demonstrted etween oth hlorophyll / nd lef rotenoids vs. severl VIs sed on multispetrl imges quired y UAVs.38 UAVs were used for ssessment of vineyrd wter sttus y orreltion of stem ψ with NDVI.2 Further reltionships were oserved etween severl Vis, inluding NDVI vs. vine wter sttus [lef wter potentil (ψ) nd stomtl ondutne].38 Addition- d e f Figure 2: Sptil mps of the Cernet Frn vineyrd t Chteu des Chrmes (CDC), St. Dvids, Ont. in 2016. 2: UAVsed NDVI; 2: GreenSeeker-sed NDVI; 2: UAV-sed therml imge; 2d: Soil moisture; 2e: Lef wter potentil; 2f: LT50 (winter hrdiness). Sle 1 =2167. d e f Figure 3: Sptil mps of the Cernet Frn vineyrd t Chteu des Chrmes, St. Dvids, Ont. in 2016. 3: UAVsed NDVI; 3: Yield per vine (kg); 3: Weight of ne prunings per vine (kg); 3d: Berry weight (g): 3e: Berry solule solids (Brix): 3f: Berry ph. Sle 1 = 2167. July 2018 WINES& VINES 61
GRAPEGROWING WINE EAST lly, severl other VIs were orrelted to wter-sttus vriles. 37 Nutritionl defiienies lso hve een deteted y UAVs; e.g., NDVI ws orrelted with levels of iron hlorosis, rotenoid pigments in leves, nd nthoynins in leves nd grpe erries. 21 More reently rop wter-stress index ws lulted sed upon omintion of lef ψ nd stomtl ondutne nd ws ssoited with UAV-derived therml imges. 29 UAVs hve een suessfully linked to wireless sensor networks to ssess orreltions etween VIs nd therml zones in vineyrds. 6 Reently in Ontrio, UAVs were used to exmine orreltive reltionships etween VIs nd severl vineyrd vriles. 25 Vine size, LT 50 (winter hrdiness), yield, erry weight nd erry omposition dt were orrelted in severl vineyrds to NDVI nd other dt quired with the UAV nd GreenSeeker, while soil nd vine wter sttus, nd yield omponents showed diret reltionships with NDVI. Sptil reltionships were pprent from exmintion of the mps. Multivrite sttistil nlysis (e.g., prinipl omponents nlysis; PCA) onfirmed these reltionships. Sptil nlysis lso ws performed (e.g., Morn s I nd k-mens lustering) to verify the existene of tul disrete zones in the study vineyrds. The NDVI vlues were onsiderly higher in GreenSeeker mps vs. those from UAV flights, primrily euse refletne dt were quired from the sides of the nopies with GreenSeeker nd from the tops of the nopies with UAVs. Vine wter sttus nd severl fruit-omposition vriles were orrelted with UAV-derived NDVI. This suggests tht UAVs hve signifint potentil to identify zones of superior fruit omposition nd potentil wine qulity. Cse study in Ontrio We used this tehnology in se study of vineyrds in Ontrio to rete mps illustrting sptil vriility in oth ground-sed vriles nd UAV dt, nd its usefulness in understnding nd improving potentil wine qulity. The Ontrio wine industry produes out 65,000 tons of grpes nd onsists of ultivrs suh s Riesling, Chrdonny, nd Cernet Frn, with lesser quntities of Merlot, Cernet Suvignon, Suvignon Bln, Pinot Gris, nd Pinot Noir (grpegrowersofontrio.om). Soils re hrterized s vrile, result of widespred glil tivity more thn 10,000 yers go, nd onsequently mny vineyrds re situted on severl soil series tht n rnge widely in texture, depth nd wter-holding pity. 13 This vriility n impt vine vigor, yield nd wter sttus. A signifint growth in the numer of smll rtisnl wineries hs permitted prodution of wines tht re unique to individul vineyrd sites nd in some ses unique to speifi vineyrd loks. In the pst 10 to 15 yers this interest hs expnded to inlude identifition of unique portions of vineyrd loks, some less thn 1 hetre, tht might e ple of produing extremely highvlue wines sed upon yield, vine size or wter-sttus-sed qulity levels. Reserhers hose six vineyrds eh of Cernet Frn nd Riesling (1 to 2 h in re) in different Nigr su-ppelltions. The sites inluded the following su-ppelltions: Nigr Lkeshore, Creek Shores, St. Dvids Benh, Linoln Lkeshore north, Linoln Lkeshore south, nd Bemsville Benh. Soil types 13 vried sustntilly in these su-ppelltions from well-drined, orsetextured Tvistok nd Vinelnd series (Nigr Lkeshore, Linoln Lkeshore north), to modertely well-drined Chinguousy (Creek Shores, Bemsville Benh), nd poorly drined Jeddo (Linoln Lkeshore south) nd Beverly/ Toledo soils (St. Dvids Benh). This rry of soil types provided signifint rnge of wterholding pities tht ffeted vine wter sttus. Vineyrd loks were GPS-delineted to determine shpe, nd 80 to 100 sentinel vines were identified within eh vineyrd nd geoloted y GPS. Post-olletion differentil orretion ws performed to su-meter ury (out 30 to 50 entimeters). Field mesurements nd erry smples were tken on these vines over two yers, nd we re ompleting our third yer of dt olletion. Vineyrd soil moisture ws mesured y time domin refletometry. Mesurements took ple t erry set, lg phse nd verison on ll sentinel vines. Vine wter sttus ws me- Figure 4: Mps of two Ontrio Cernet Frn vineyrds in 2016 showing GLRV3 titer [quntifition yle (Cq) vlue] (, ) vs. their respetive orresponding UAV-sed NDVI (, d). Sle (, ): 1 = 2167; (, d): 1 = 1240. d GORDON ROBERT 62 WINES&VINES July 2018
WINE EAST GRAPEGROWING sured using middy lef ψ y pressure om. Mesurements were mde only t designted lef ψ vines (out 15 per vineyrd lok), on the sme dys s SWC mesurements, from 1000h to 1400h (. solr noon), under full sun. Lef gs ws mesured y hnd-held porometer on those vines used for lef ψ. Fruit from eh sentinel vine ws hrvested, luster numer determined, nd fruit weighed. Cne prunings were weighed to determine vine size. A 100-erry smple ws tken from eh sentinel vine, weighed to determine men erry weight, nd therefter solule solids were mesured using refrtometer, ph y ph/ion meter nd titrtle idity y utotitrtor. Monoterpene rom ompounds [free voltile terpenes (FVT) nd potentilly-voltile terpenes (PVT)] were nlyzed on Riesling smples tken from lef ψ vines sed on distilltion method. Cernet Frn juie smple sorne ws mesured using spetrophotometri method t 420 nm nd 520 nm, with olor intensity lulted s A 420 +A 520, nd hue s A 420 /A 520. Totl nthoynins were quntified using the ph shift method, while totl phenols were quntified using the Folin- Ciolteu method. All vines designted for lef ψ nd gs mesurements were smpled Septemer 2016 to determine grpevine lefroll-ssoited virus (GLRV) titer. Totl RNAs were isolted from lef smples using method reently developed in the Meng l. Totl RNAs were used in reverse trnsription using primers speifi to GLRV-2 nd 3, followed y mplifition through PCR using rod-spetrum primers to identify virus presene. To determine virus titer for omprison nd orreltion to the UAV nd GreenSeeker dt, quntittive qpcr ws onduted y Power SYBR Green PCR Mster Mix nd StepOnePlus qpcr (Applied Biosystems). GPS oordintes from vineyrd loks nd sentinel vines were imported into GIS environment (ArGIS) nd linked to ll point dt olleted from sentinel vines. Sptil interpoltion tehniques (i.e., kriging) pplied to these dt were used to estimte the vlue of vineyrd vriles t unsmpled lotions. This permitted further sptil dt nlyses, inluding the integrtion of these dt with the remote-sensing dtsets. A GreenSeeker unit (Trimle Nvigtion) mounted on four-wheel-drive vehile ws used to ollet NDVI dt on dtes lose to soil moisture nd lef ψ dt olletion. An dditionl reding ws olleted in mid-septemer. Dt were imported into Frmworks softwre nd sptil mps reted. Shpefiles were therefter imported into the ArGIS dtse. GPS oordintes identil to the sentinel vines were identified, nd NDVI dt orresponding to these oordintes were extrted for sttistil nlyses. The UAV flights orresponded to the verison soil moisture, lef ψ, nd GreenSeeker dt olletion. Imge quisition ws performed using the SenseFly ebee UAV (see Figure 2A) supplied y Air-Teh Solutions, Inverry, ON. The UAV hd 30 minutes of endurne, 9-kg pylod, nd ws equipped with n utopilot system llowing 1000-m visul rnge nd 5-km rdio line of sight. The UAV ws flown t oth 90 nd 120 m elevtions in seprte flights nd mximum speed of 60 km/h. Two sensors were used for imge quisition. The first operted in the visile nd nerinfrred portions of the eletromgneti spetrum (EMS) utilizing five spetrl nds (lue, green, red, red edge, ner-infrred) equipped with n inident light sensor. The seond sensor operted in the therml infrred (IR) portion of the EMS. It ensured quisition of imgery in the therml-ir rnge overing 750 to 1350 μm. This equipment ws omplemented y ground reeiving sttion tht provided rel-time feedk on the position of the irrft nd its imging. Imge quisition ws performed over eh Reh Thousnds of Winery Deision Mkers The Wines & Vines Online Mrketing System is designed to help your usiness: Inrese sles Generte winery leds Mnge lients nd sles territories Improve mrketing effetiveness Visit winesndvines.om/oms or ll 866.453.9701 to set-up your free demo. ONLINEMARKETINGSYSTEM Wines Vines Anlytis July 2018 WINES&VINES 63
GRAPEGROWING WINE EAST Figure 5: These mps were derived from UAV dt nd re exmples of NDVIsed zones in three Ontrio Cernet Frn vineyrds. 5 is Pondview in Virgil, Ont.; 5 is George in Vinelnd, Ont.; nd 5 is Cve Spring in Bemsville, Ont. Sles re 5: 1 = 1039; 5: 1 = 1240; nd 5: 1 = 1581. vineyrd lok. Dt were stored The series of imges quired onord nd retrieved fter the during eh flight ws ssemled flight mission. Geometri orreinto mosis y seleting the overtion ws performed to orret the lpping res to limit the viewing imge geometry. Georeferening ngle nd the prolems of direws hieved y identifying ontionl effets. One ssemled nd trol points (trgets) on the orreted, NDVI ws lulted on ground. Geometri distortions mosis. The NDVI pixel vlues used y hnges in UAV ttitude orresponding to the field points nd ltitude were orreted using were extrted nd ompiled into the informtion provided y the geodtse thtpge inluded ll EsternWineL_Mr09.qxp 1/22/09 9:47 AM 1 inertil sttion. field-sed vriles (e.g., lef ψ). EASTERN WINE LABS Serving the Anlytil needs of Est Cost Wineries WWW.EASTERNWINELABS.COM Ph 609-859-4302 Cell 609-668-2854 hemist@esternwinels.om AOAC Memer 64 WINES& VINES July 2018 An exmple of UAV orreltions with other vriles is shown in Figure 1 (pge 60). It should e noted tht mthemtil reltionships etween NDVI nd other vriles of importne never will e extly the sme for every vineyrd. There is sustntil vriility etween vineyrds, vrieties nd yers. Tht is the nture of mking iologil mesurements nd ttempting to orrelte those to physil mesurement suh s lef refletne. The UAV NDVI nd therml indies were orrelted with vine size in ll Riesling vineyrds in 2015, inluding tht of the Buis vineyrd in Nigr-on-the-Lke depited in Figure 1. The pplition of PCA provides pitoril method of portrying orreltions etween multiple vriles in lrge dt set. The vriles represented y lines (eigenvetors) re onsidered orrelted if the eigenvetors re prllel to eh other, nd inversely orrelted if they re roughly 180 from eh other. In Figure 1, NDVI quired y UAV ws orrelted with GreenSeeker NDVI, soil moisture, lef ψ, gs, nd erry weight, nd inversely orrelted to therml dt, FVT nd PVT in Riesling. This suggests tht zones of low NDVI might e linked to fruit with higher terpenes, nd y inferene higher-qulity wine. Noteworthy ssoitions in other Riesling vineyrds inluded UAV indies nd erry weight (five sites), TA (two sites Buis, George), FVT/PVT (three sites), nd GreenSeeker NDVI (four sites). Inverse orreltions of note with UAV dt inluded soil moisture nd lef ψ (five sites). Diret orreltions lso were noted for NDVI nd t lest one LT50 mesurement for four sites: LT50 mesured in Mrh (Buis), Jnury LT 50 (Pondview), Ferury LT50 (Hughes, Cve Spring). This suggests tht high NDVI ws ssoited with low winter hrdiness, sine LT50 is negtive vlue. Yield ws not onsistently relted to NDVI; yield nd NDVI were inversely orrelted in three sites (Buis, Hughes, Pondview) ut unrelted in two others. For the most prt we found similr ptterns in ll Riesling vineyrds in 2016. In Cernet Frn, UAV NDVI ws ssoited with soil moisture, lef ψ, gs, yield, nd erry weight, nd inversely orrelted to olor, nthoynins nd phenols in the George vineyrd in Vinelnd in 2015 (Figure 1). UAV NDVI ws orrelted with vine size in ll other Cernet Frn vineyrds. Other noteworthy ssoitions inluded UAV indies nd soil moisture (three sites), lef ψ (two sites), erry weight (five sites), TA (three sites), yield/luster numer (four sites), nd GreenSeeker NDVI (four sites). NDVI ws in-
WINE EAST GRAPEGROWING versely orrelted with nthoynins, A520 nd totl phenols in four of five Cernet Frn vineyrds in 2015. This suggests tht zones of low NDVI might e ssoited with grpes of higher olor nd phenols, nd possily higher wine qulity thn high-ndvi zones. Applition of PCA of 2016 dt reveled similr reltionships. This projet hs produed literlly hundreds of mps! With 12 vineyrds under study, three seons of dt olletion, nd minimum of 14 vriles, tht mens more thn 500 mps. Mps for Chteu des Chrmes Cernet Frn vineyrd 2016 re depited in Figure 2 (pge 61). The UAV NDVI mp showed low NDVI zone in the north end of the vineyrd (Figure 2). This orresponded losely with mps produed using GreenSeeker (Figure 2), highest regions from the therml mer (Figure 2) nd soil moisture (Figure 2d), nd lowest regions of lef ψ (Figure 2e), nd higher LT50 (i.e. less winter hrdy; Figure 2f). Zones of low UAV NDVI (Figure 3) (pge 61) were ssoited with low yield (Figure 3), vine size (Figure 3), erry weight (Figure 3d), higher Brix (Figure 3e) nd higher ph (Figure 3f), one gin suggesting enhned fruit mturity in low NDVI zones. Conversely, those high-ndvi zones identified y the UAV were typilly low in therml mer dt nd high in GreenSeeker NDVI, lef ψ, soil moisture, vine size, erry weight nd TA, nd lower in LT50 (i.e., more winter hrdy). Other Cernet Frn vineyrds showed ssoitions in the UAV dt etween low NDVI nd high therml zones. These orresponded to low GreenSeeker NDVI, soil moisture, lef ψ, vine size nd erry weight, nd higher LT50 zones. There ws some sptil orreltion with high-ta nd low-brix res, ut ph nd overll yield usully were not strongly relted sptilly. In most vineyrd loks the UAV NDVI mps were omprle to GreenSeeker NDVI mps. In generl, there were good diret sptil orreltions etween UAV nd GreenSeeker NDVI vs. lef ψ, lef gs, soil moisture, vine size, LT50 nd TA, nd inverse ones with Brix nd ph. There were lso mny situtions in whih mps from therml dt were inversely orrelted sptilly with NDVI. Most frequent sptil orreltions in Riesling with UAV nd GreenSeeker NDVI zones were lef ψ, gs, soil moisture, vine size, erry weight, yield nd TA. Noteworthy inverse sptil orreltions inluded Brix, ph, FVT nd PVT. The 2016 ptterns in lef ψ, soil moisture, yield omponents nd erry omposition were for the most prt onsistent with those oserved in To see referenes for this rtile, go to winesndvines.om nd serh under Mgzine Fetures July 2018. 2015 for oth Riesling nd Cernet Frn. GVLRV3 titer ws determined in 2016 in ll vineyrds for ll vines used for lef ψ. Zones of low GLRV titer orresponded sptilly with zones of low NDVI in two Cernet Frn vineyrds (Figure 4) (pge 62). The sme trend ws pprent for three Riesling vineyrds. However, low NDVI zones lso orresponded in some ses with low lef ψ, low soil moisture nd high onentrtions of nthoynins nd phenols. This is mjor hllenge, euse we need to identify unique spetrl signtures tht designte virus-ffeted res tht do not orrelte sptilly with other vriles suh s lef ψ; otherwise there is strong possiility of misdignosis. Consequently, more reserh is needed to identify other vegettion indies tht re unique nd speifi to zones of vrying virus sttus. This work is ongoing. Mps lso identify mngement zones within vineyrds tht might orrespond to fruit of different potentil wine qulity. This is, in essene, the rison d être of preision vitiulture. Figure 5 depits three Ontrio vineyrds tht show ler zones of high nd low NDVI determined y UAV. During hrvest, it ws ler tht the high NDVI zones hd lrger lusters, lrger erries, generlly higher yields per vine nd, in the se of Riesling, more unh rot thn low NDVI zones. Conlusions Unmnned Aeril Vehiles (UAVs) re vlule tool to quire high-resolution eril imges of vineyrds. The dt within the high-resolution eril imges n e used to determine the sptil distriution of vriety of nopy vriles within vineyrd lok nd etween different vineyrd loks. In this wy UAVs n e used to mesure the sptil distriution of vigor, wter stress, nutrient sttus, disese, yield omponents nd erry omposition. One of the mjor hllenges is the mssive mount of dt tht re olleted, the time required for postflight dt proessing nd the need to derive useful vegettion indies tht re speifi to vriles of interest. Andrew G. Reynolds is professor of iologil sienes/ vitiulture t the Cool Climte Oenology nd Vitiulture Institute, Brok University, St. Cthrines, Ontrio, Cnd; Hyun-Suk Lee nd Brinn Dorin re grdute students t CCOVI, Brok University. Mrilyne Jollineu is ssoite professor in the deprtment of geogrphy t Brok University. Rlph Brown is professor of engineering t the Shool of Engineering, University of Guelph, Guelph, Ontrio. Mehdi Shnin is ssistnt professor nd Bozhong Meng is ssoite professor t the University of Guelph. Adm Shemrok is with Air-Teh Solutions, Inverry, Ontrio. Qulity Rugged 12 guge, Custom & Stok Stinless Tnks Ctwlk Systems Pump Over Crts Custom Hoppers, Chutes & Bins 251 Gmee Rod, Genev, NY 14456 Phone: 315-759-2118 Toll free: 888-234-6752 Fx: 315-789-1848 Emil: tnks@vnemetl.om Wesite: www.vnemetl.om July 2018 WINES& VINES 65