Using RPAS Multi-Spectral Imagery to Characterise Vigour, Leaf Development, Yield Components and Berry Composition Variability within a Vineyard

Remo e Sens. 2015, 7, 14458-14481; doi:10.3390/ s71114458 emo e sensing ISSN 2072-4292 www.mdpi.com/jou nal/ emo esensing A icle Using RPAS Mul i-Spec al Image y o Cha ac e ise Vigou , Lea De elopmen , Yield Componen s and Be y Composi ion Va iabili y wi hin a Vineya d Cla a Rey-Ca amés 1, Ma ía P. Diago 1, M. Pila Ma ín 2, Agus ín Lobo 3 and Ja ie Ta daguila 1,* 1 Ins i u o de Ciencias de la Vid y del Vino, Uni e si y o La Rioja, Finca La G aje a, Ca e e a de Bu gos Km 6, Log oño 26007, Spain; E-Mails: [email p o ec ed] (C.R.-C.); [email p o ec ed] (M.P.D.) 2 Ins i u o de Economía, Geog a ía y Demog a ía. Cen o de Ciencias Humanas y Sociales, CSIC, Albasanz 26–28, Mad id 28037, Spain; E-Mail: [email p o ec ed] 3 Ins i u e o Ea h Sciences Jaume Alme a, ICTJA-CSIC, Lluis Sole Saba is s/n, Ba celona 08028, Spain; E-Mail: [email p o ec ed] * Au ho o whom co espondence should be add essed; E-Mail: [email p o ec ed]; Tel.: +34-941-894-980. Academic Edi o s: Mu lu Ozdogan, Yoshio Inoue and P asad S. Thenkabail Recei ed: 1 Sep embe 2015 / Accep ed: 26 Oc obe 2015 / Published: 30 Oc obe 2015 Abs ac : Implemen a ion o p ecision i icul u e echniques equi es he use o eme ging sensing echnologies o assess he ineya d spa ial a iabili y. This wo k shows he capabili y o mul ispec al image y acqui ed om a emo ely pilo ed ae ial sys em (RPAS), and he de i ed spec al indices o assess he ege a i e, p oduc i e, and be y composi ion spa ial a iabili y wi hin a ineya d (Vi is ini e a L.). Mul i-spec al image y o 17 cm spa ial esolu ion was acqui ed using a RPAS. Classical ege a ion spec al indices and wo newly de ined no malised indices, NVI1 = (R802 − R531)/(R802 + R531) and NVI2 = (R802 − R570)/(R802 + R570), we e compu ed. Thei spa ial dis ibu ion and ela ionships wi h g ape ine ege a i e, yield, and be y composi ion pa ame e s we e s udied. Mos o he spec al indices and ield da a a ied spa ially wi hin he ineya d, as showed h ough he a iog am pa ame e s. While he co ela ions we e signi ican bu mode a e among he spec al indices and he ield a iables, he kappa index showed ha he spa ial pa e n o he spec al indices ag eed wi h ha o he ege a i e a iables (0.38–0.70) and mean clus e weigh (0.40). These esul s p o ed he u ili y o he OPEN ACCESS Remo e Sens. 2015, 7 14459 mul i-spec al image y acqui ed om a RPAS o delinea e homogeneous zones wi hin he ineya d, allowing he g apeg owe o ca y ou a speci ic managemen o each suba ea. Keywo ds: p ecision i icul u e; emo e sensing; emo ely pilo ed ae ial sys em; spec al indices; kappa index 1. In oduc ion Mode n and sus ainable i icul u e equi es objec i e and con inuous moni o ing o key pa ame e s o a ional and di e en ia ed ag onomic managemen o ineya ds based on he spa io- empo al a iabili y o g ow h, yield, and g ape composi ion wi hin he plo . The di e en ial managemen wi hin he ineya d and he concep o being spa ially a iable a e co e ed by a discipline known as p ecision i icul u e. This new app oach allows he g ape-g owe o op imize c op p oduc ion and p o i abili y by educing inpu s such as machine y, labou , chemicals, wa e , ene gy, e c. I also con ibu es o diminishing he en i onmen al impac o pes icides, nu ien s leaching, and ossil uels, among o he s [1,2]. One o he ools ha has demons a ed i s capabili y o help in c op managemen , and mo e speci ically, in s udying he wi hin- ield a iabili y o assess spa ial and empo al changes in soil mois u e, canopy g ow h, and plan wa e s a us, is emo e sensing. Se e al wo ks ha e shown he sui abili y o emo e sensing o p ecision i icul u e pu poses du ing he las wo decades [3]. Hence, his echnique has been used o app aise he ineya d spa ial a iabili y o wa e s a us [4,5], chlo ophyll and ca o enoid con en [6,7], ineya d canopy s uc u e [8,9], g ape colou and phenols con en [10,11], as well as g ape quali y [6,12]. Remo e sensing in ol es he acquisi ion o spec al da a om se e al pla o ms, such as sa elli es, ai c a s, and emo ely pilo ed ae ial sys ems (RPAS). Vineya ds a e no a con inuous c op, unlike ce eals. Thus, spec al mixing be ween indi idual ines and ow soil backg ound is one o he main d awbacks o ope a ional applica ion o emo e sensing da a o ineya d moni o ing [13]. E en o e y high- esolu ion space bo ne senso s such as Pleiades Cons ella ion (h p://smsc.cnes. /PLEIADES/), he highes spa ial esolu ion a ailable is 50 cm/pixel, which could s ill be oo coa se o moni o some common ine aining sys ems whe e c op canopies a e e y na ow (i.e., 40–60 cm in he case o e ical shoo posi ioned ineya ds), and a mixed signal o lea es, shoo s, shadow, and soil is commonly ound in mos pixels. Senso s on boa d manned ai bo ne ehicles allow hese limi a ions o be o e come by acqui ing image y o inc eased spec al, spa ial, and empo al esolu ion compa ed o hose moun ed on sa elli es, al hough hei ope a ional cos s a e e y high. RPAS is an in e es ing al e na i e in p ecision i icul u e o o he pla o ms such as helicop e s, ai bo ne, and sa elli e, ega ding mainly he spa ial esolu ion and he cos o acquisi ion [14]. RPAS allow o he acquisi ion o image y o e y high spa ial esolu ion, o en a a sub-decime e esolu ion, a much lowe cos han adi ional ai bo ne emo e sensing. Fu he mo e, i has been p o en ha esul s ob ained om a RPAS o ag icul u al applica ion may yield simila es ima ions in e ms o accu acy and p ecision o hose de i ed om adi ional ai bo ne pla o ms [15]. These ac s, oge he wi h hei lowe cos and highe empo al lexibili y, explain he inc eased use o RPAS o ag icul u al pu poses [16]. As a ma e o ac , a e y ecen epo on he economic impac o unmanned ai c a s in he Uni ed S a es (US) [17] e ealed ha he use o RPAS in ag icul u e in 2015 in he US would yield $2096.5 and an Remo e Sens. 2015, 7 14460 expec ed c ea ion o 21,565 jobs. Among he applica ions o RPAS in i icul u e, he assessmen o wi hin- ineya d a iabili y s ands ou [18], as he spa ial and empo al esolu ions a e key a ibu es when he goal o he s udy is moni o ing and managing he ineya d [3]. Mos o he emo e sensing s udies in ag icul u e a e based on he use o spec al indices. These a e a i hme ical combina ions o he o iginal spec al bands cap u ed by he senso , which educe he complexi y o he da ase [3] and acili a e he analysis o a ious ege a ion pa ame e s ha a e di ec ly ela ed o hei spec al esponse a gi en wa eleng hs. Vege a ion indices, like he no malised di e ence ege a ion index (NDVI), ha e been ex ensi ely used o moni o c op s a us by ela ing isible and nea in a ed spec al da a wi h bio-geophysical p ope ies o he ines such as igou and lea a ea [18]. O he spec al indices like he pho ochemical e lec ance index (PRI) combine in o ma ion om wo na ow bands in he isible egion o moni o ege a ion s ess [19]. The p esen wo k aims o show he capabili y o mul ispec al image y acqui ed om a RPAS, and he de i ed spec al indices, o assess and map he spa ial a iabili y o ege a i e and yield componen s, as well as be y composi ion in a ineya d (Vi is ini e a L.) o p ecision i icul u e managemen . 2. Ma e ials and Me hods 2.1. S udy A ea The s udy was conduc ed in a 3.0 ha Temp anillo (Vi is ini e a L.) ineya d loca ed in Na a a, Spain (Figu e 1). G ape ines we e plan ed in 2004 in a sandy-clay soil (5% slope owa ds No h), using a 41B oo s ock a 2.4 m × 1.6 m (in e - and in a- ow) wi h no h-sou h ow o ien a ion. Vines we e ained o a e ically shoo -posi ioned (VSP), spu -p uned co don e aining 16 nodes pe linea me e / ine. Vines we e uni o mly i iga ed wice ac oss he season. Figu e 1. Expe imen al Temp anillo ineya d plo wi h he 72 sampling poin s (a). Schema ic desc ibing he posi ion o he h ee adjacen ines comp ising a sampling poin (b). Remo e Sens. 2015, 7 14461 2.2. Field Da a Acquisi ion o Vege a i e and Yield Componen s. Be y Composi ion Analysis. A egula g id o 72 sampling poin s a 20 m in e als was gene a ed wi hin he ineya d using a Leica Zeno 10 GPS (Hee b ug, S . Gallen, Swi ze land), wi h eal- ime kinema ic co ec ion, wo king a <0.4 m p ecision. Each sampling poin consis ed o h ee adjacen ines (Figu e 1b). Vege a i e and yield- ela ed da a we e measu ed in he ield om all sampling poin s. As indica o s o ine igou , main and seconda y shoo leng h and p uning weigh we e measu ed. Rega ding igou - ela ed a iables, main and seconda y lea a ea pe shoo we e de e mined. Seconda y shoo leng h and seconda y lea a ea pe shoo we e measu ed o he h ee ines included in each sampling poin one week p io o he RPAS ligh , on 13 Sep embe 2011. Fo each ine, wo shoo s we e andomly chosen and hei main shoo leng h, as well as hose o la e als wi h mo e han h ee lea es, we e measu ed using a me e ape [20]. Main and seconda y lea a ea pe shoo we e es ima ed by de olia ing each shoo and weighing he lea es wi h a po able scale o p ecision ±2 g (Ke n and Sohn GmbH, Balingen-F omme n, Ge many), using he disc-lea me hod by Sma and Robinson [21]. The p uning weigh o each ine was manually de e mined in he ield using a hanging scale wo mon hs a e he ligh , on 20 No embe 2011, p io o ine p uning. Addi ionally, non-des uc i e measu emen o pho osyn he ic pigmen s in plan s we e ob ained on 13 Sep embe 2011, using a Mul iplex™ (Mul iplex 3.0, Fo ce-A, O say, F ance), a handheld luo escence-based p oximal senso ha allows he compu a ion o luo escence-based indices ela ed o he chlo ophyll con en in lea es (SFR_RAD) and he ni ogen balance index (NBI_GAD). The Mul iplex™ senso used o his s udy was a modi ied e sion o he de ice ecen ly desc ibed by Ben Ghozlen e al. [22]. Mul iplex™ measu emen s we e pe o med on h ee main lea es (adaxial side) pe sampled ine. F om hese measu emen s, he luo escence indices we e calcula ed as ollows: 𝑆𝐹𝑅_𝑅𝐴𝐷 = 𝐹𝑅𝐹_𝑅𝐴𝐷 𝑅𝐹_𝑅𝐴𝐷 (1) 𝑁𝐵𝐼_𝐺𝐴𝐷 = 𝐹𝑅𝐹_𝑈𝑉𝐴𝐷 𝑅𝐹_𝐺𝐴𝐷 (2) whe e FRF_R and FRF_UV e e o he a ed luo escence emission exci ed by ed and UV ligh , espec i ely, and RF_R and RF_G e e o ed luo escence exci ed by ed ligh and g een ligh , espec i ely [23]. The subsc ip AD s ands o he adaxial side o he lea . SFR_RAD has been epo ed as a p ecise indica o o he chlo ophyll con en in kiwi lea es [24] and in g ape ine lea es [25], and NBI_GAD has also been shown as a eliable indica o o he ni ogen s a us o he g ape ine [25]. A ha es ime (17 Oc obe 2011), main yield componen s, such as clus e numbe pe ine, yield pe ine, clus e weigh , and be y weigh , we e de e mined o each sampled ine. Each plan was manually ha es ed, hei clus e s coun ed and weighed in he ineya d using a hanging scale, and he a e age clus e weigh calcula ed. Two clus e s pe ine we e hen labelled, kep apa and aken o he labo a o y o he Uni e si y o La Rioja in po able e ige a o s whe e hey we e manually des emmed. Be ies om each clus e we e weighed (TE3102S, Sa o ius, Goe ingen, Ge many) and coun ed, and he a e age be y weigh calcula ed. To de e mine be y composi ion, om he be ies co esponding o each ine, a ep esen a i e subsample o 50 be ies was aken and weigh ed. These be y subsamples we e hen s o ed ozen a −20 °C un il colou and phenolic analyses. The emaining be ies o each ine we e p essed and he ob ained mus was analysed o o al soluble solids concen a ion (TSS) and acidi y pa ame e s. TSS (exp essed in B ix) Remo e Sens. 2015, 7 14462 was de e mined using a empe a u e-compensa ing digi al e ac ome e (A ago, Tokyo, Japan), while i a able acidi y and pH we e de e mined ollowing he OIV me hods [26]. An hocyanins and o al phenols we e analysed o each be y sub-sample applying he me hod o Iland e al. [27]. An hocyanin concen a ions we e exp essed as mg/g o esh be y mass, whe eas o al phenols we e exp essed as abso bance uni s (AU) a 280 nm/g o esh be y mass. Fo all da a collec ed in he ield and ela i e o be y composi ion, he alues o he h ee ines pe sampling poin we e a e aged and a mean alue was hen assigned o each sampling loca ion. The ege a i e a iables, p uning weigh , main lea a ea and seconda y lea a ea, and he p oduc i e a iable, yield, we e ans o med o su ace uni s (pe m2) by di iding hem by he ine spacing [28]. 2.3. RPAS Mul i-Spec al Images Mul i-spec al image y was acqui ed wi h a mul i-spec al came a, a mini mul iple came a a ay (MCA 6, Te acam Inc, Cha swo h, CA, USA) moun ed on a RPAS Md4-1000 (Mic od ones GmbH, Siegen, Ge many) on 20 Sep embe 2011 (Figu e 2). The RPAS Md4-1000 is a quad o o able o ca y 1.2 kg o payload mass and able o ly up o 88 minu es, wi h e ical ake-o and landing. The came a consis ed o six independen image senso s and op ics wi h use -con igu able il e s. Image esolu ion was 1280 × 1024 pixels wi h 10-bi adiome ic esolu ion and op ics ocal leng h o 9.6 mm. Fo his s udy, he came a was equipped wi h six 25 mm-diame e bandpass il e s o 10 nm ull-wid h a hal -maximum (FWHM) (Ando e Co po a ion, NH, U.S.), wi h cen e wa eleng hs a 531, 551, 570, 672, 701, and 802 nm, and bandwid h o ±9.31, 10.13, 9.29, 9.82, 9.47, and 10.11 nm, espec i ely. The wa ebands we e selec ed in o de o ob ain he spec al bands equi ed o he calcula ion o he spec al indices commonly used in ag icul u e, and, speci ically, i icul u e- ela ed li e a u e. Figu e 2. Mul ispec al came a moun ed on he RPAS Md4-1000 and de ail o he mul iple came a a ay. The RPAS lew a 250 m heigh and allowed cap u ing images wi h a spa ial esolu ion o 17 cm. Images we e aken a noon, be ween 11:15 a.m. and 12:15 p.m., unde s able wea he condi ions and clea sky (mean wind eloci y o 1.39 m/s). A his ime, g ape be ies we e a he hal -way poin o hei ipening p ocess. A Remo e Sens. 2015, 7 14463 o al o 6 scenes we e acqui ed (wi h an o e lapping a ea o 60%) in o de o moni o he whole ineya d plo leng hwise. Each image co e ed a g ound su ace o app oxima ely 200 × 165 m. I was able o co e he whole plo wi h one ligh due o he educed ex ension o he ineya d, he lying heigh , and he came a weigh (700 g), which allowed he RPAS o ha e enough ligh ope a ional ime. 2.4. Image P ocessing F om he 6 o e lapping images ob ained wi h he mul i-spec al came a, wo scenes ha co e ed he whole ineya d plo we e selec ed o he p esen s udy. An ini ial p e-p ocessing o he images acqui ed wi h he Mini MCA came a was ca ied ou using i s companion applica ion Pixel W ench 2 (PW2), w i en and copy igh ed by Te acam Inc., o p oduce mul i-page Tagged Image Fo ma (TIF) iles. As he mul i-spec al came a uses six di e en lenses (one o each spec al band), i is necessa y o con i m ha he bands a e aligned among hem. Band alignmen is impo an o ensu e consis en spec al in o ma ion, bu when his co ec ion is ca ied ou by PW2, i u ns ou o be insu icien acco ding o Lalibe e e al. [29]. In ag eemen wi h hese indings, he esul s o he in e -band alignmen pe o med wi h PW2 showed subs an ial e o s in his s udy and, he e o e, i was decided o skip he alignmen p ocessed included in PW2 by se ing a null aligning ma ix and ca y ou a speci ic co ec ion. Band alignmen was pe o med by applying he Fine Regis a ion algo i hm o he O eo Toolbox [30] by using a e e ence band (band 2). This p ocess calcula es he local shi s in he x and y di ec ions ha esul in he bes local co ela ion be ween he band o be p ocessed and he e e ence band, and c ea es a de o ma ion ield ha is subsequen ly applied o he band o be p ocessed. A e a se o es s wi h a ange o adii, he local windows o explo e and calcula e co ela ions we e se o 5 × 5 pixels o bands 1, 3 and 4, and 7 × 7 pixels o bands 5 and 6. In addi ion, as he displacemen wi h espec o he e e ence band was la ge o bands 5 and 6, a igid ansla ion was applied in bo h x and y di ec ions p io o he ine egis a ion p ocessing o hese wo bands. Once he band alignmen co ec ions we e comple ed, he images we e geo e e enced o allow o he co ela ion o image da a wi h ield measu emen s. This was done by using ca og aphic co-o dina es om eigh e e ence a ge s (whi e discs o 30 cm o diame e ) loca ed a he bounda ies o he ineya d plo . Those a ge s had been geo e e enced in he ield du ing image acquisi ion using a Leica Zeno 10 Global Posi ioning Sys em (GPS) (Hee b ug, S . Gallen, Swi ze land). An o hopho o o he a ea o s udy p o ided by he Spanish Na ional P og am o Ae ial O opho og aphy (PNOA) wi h 25 cm spa ial esolu ion was complemen a ily used o iden i y 45 g ound con ol poin s in each scene. These poin s we e e enly dis ibu ed o e he image a ea including he bo de s, whe e displacemen e o s we e highe because o he la ge dis ance om he image nadi . A second deg ee polynomial unc ion was applied using a bilinea esampling me hod o ende he geome ically co ec ed image. Image geo e e encing was ca ied ou wi h an accu acy o 0.62 and 0.35 m ( oo mean squa ed e o ). In e ac i e loca ion o g ound con ol poin s, polynomial i , and in e pola ion we e pe o med using ENVI 4.8 (Exelis, McLean, VA, USA). Radiome ic calib a ion was pe o med using an image o a whi e calib a ion panel ha was acqui ed om he g ound wi h he mini MCA (a a dis ance abou 1.5 m) p io o he ligh . A e lec ance spec um o he same panel was measu ed a he En i onmen al Remo e Sensing and Spec oscopy Labo a o y (SpecLab) using an ASDFieldSpec 3 spec o adiome e (ASD Inc., Boulde , CO, USA). Image DN (Digi al Numbe s) Remo e Sens. 2015, 7 14464 om he calib a ion panel measu ed in he ield we e a e aged o each band o calcula e he obse ed e e ence DN spec um. The e lec ance spec um measu ed in he labo a o y was in eg a ed o calcula e he heo e ical e lec ance alues o he panel o each mini MCA band using he ansmission cu es o he il e s (www.ando e co p.com). A ec o o ac o s was compu ed o ans o m he obse ed DN spec um o he image in o he modelled e lec ance spec um. These ac o s we e hen applied o each band o he image in o de o ob ain calib a ed e lec ance o each image pixel. Images we e calib a ed o appa en e lec ance, i.e., e lec ance as i he came a had been ope a ed a 1.5 m. A mosphe ic co ec ion was no pe o med, bu as he a ea was small and he a mosphe ic condi ions e y clea , he uni o m a mosphe ic e ec e en ually p esen in he indices was aken in o accoun by he linea models be ween indices and ield da a. The adiome ic calib a ion ope a ions we e conduc ed by means o speci ic unc ions w i en in R Co e Team [31] and using packages gdal [32] and as e [33]. The nex s ep was he spa ial assembly o he images o build he inal mosaic, so he whole ineya d plo i wi hin a single image ile. The mosaicking p ocess was pe o med using ENVI (Exelis Visual In o ma ion Solu ions, Inc, Boulde , CO, USA) so wa e by combining he wo scenes ha included he whole ineya d. We assigned, as base image, he scene ha co e ed mos o he plo , so he colou balance in he second image was ca ied ou o ma ch he base image by using his og am ma ching. A dis ance o 40 pixels was se o blend he wo images. Using he co ec ed image mosaic, a o al o 11 spec al indices, selec ed om he li e a u e as being he mos commonly used o cha ac e ise ege a ion s a us, we e calcula ed (Table 1 [34–42]). These indices ha e been p oposed o es ima e a wide a ie y o ege a ion p ope ies including pigmen con en s, lea a ea index, plan heal h s a us, nu ien s ess, e c., and some o hem ha e been widely used in p ecision i icul u e. Some indices specially designed o minimize he e ec s o soil backg ound on he ege a ion signal as he OSAVI and MSAVI we e also included in he analysis. In addi ion o hese indices, a se o no malised indices we e calcula ed as all possible-combina ions be ween e e y image band, ollowing he equa ion: 𝑁𝑜𝑟𝑚𝑎𝑙𝑖𝑠𝑒𝑑 𝑖𝑛𝑑𝑒𝑥= 𝐵𝑎𝑛𝑑1−𝐵𝑎𝑛𝑑2 𝐵𝑎𝑛𝑑1+𝐵𝑎𝑛𝑑2 (3) 2.5. S a is ical Analysis Wi h he aim o es ablishing an empi ical ela ionship be ween image da a and ield measu emen s, image pixels co esponding o he 72 sampling poin s we e iden i ied. Fo each sampling poin , he wo mos cen ed pixels (6 pixels pe poin ) we e selec ed. Image alues we e ex ac ed om hose pixels, he spec al indices we e compu ed and a e aged pe sampling poin o each band. P io o he s a is ical analysis, he da a we e p e-p ocessed in o de o de ec po en ial ou lie s. Desc ip i e s a is ics we e compu ed o all a iables (mean, minimum, maximum, s anda d de ia ion and coe icien o a ia ion), as well as he sp ead (Equa ion (4)), exp essed in pe cen age, as an indica o o he a iabili y in he sample [43]. 𝑆𝑝𝑟𝑒𝑎𝑑= (𝑚𝑎𝑥𝑖𝑚𝑢𝑚−𝑚𝑖𝑛𝑖𝑚𝑢𝑚) 𝑚𝑒𝑑𝑖𝑎𝑛 (4) Remo e Sens. 2015, 7 14465 To cha ac e ise he spa ial a iabili y o he a ious pa ame e s wi hin he ineya d, a iog ams o he a iables we e calcula ed by he R package “gs a ” [44]. The pa ame e s o he a iog ams we e also used o in e pola e all he a iables, using k iging echniques. Finally, o quan i y he ag eemen be ween he maps ob ained om he spec al indices and he ones de i ed om ield a iables, all he maps we e classi ied in h ee zones co esponding o low, medium, and high alues applying an iso-clus e unsupe ised classi ica ion. C oss abula ion o he esul ed classi ied maps be ween he spec al indices and he in- ield a iables was pe o med o measu e he s abili y in he spa ial pa e ns using he Kappa index ollowing he equa ion p oposed by Hudson and Ramm [45]. S a is ical and spa ial analysis we e ca ied ou using Mic oso O ice Excel 2013 (Mic oso Co po a ion, Washing on, USA), S a is ica 9.0 (S a So , Inc., Tulsa, OK, USA), and A cGIS Desk op 10.3 (ESRI, Redlands, CA, USA). Table 1. Spec al ege a ion indices commonly used in i icul u e s udies. Spec al Index Re e ence No malized Di e ence Vege a ion Index (NDVI) (Rouse e al. 1974) [34] 𝑁𝐷𝑉𝐼= 𝑅𝑁𝐼𝑅 −𝑅𝑅𝐸𝐷 𝑅𝑁𝐼𝑅 +𝑅𝑅𝐸𝐷 Modi ied Simple Ra io (MSR) (Chen 1996) [35] 𝑀𝑆𝑅= (𝑅𝑁𝐼𝑅 𝑅𝑅𝐸𝐷−1) [(𝑅𝑁𝐼𝑅 𝑅𝑅𝐸𝐷)0.5+1] Modi ied T iangula Vege a ion Index (MTVI1) (Haboudane e al. 2004) [36] 𝑀𝑇𝑉𝐼1=1.2×[1.2×(𝑅800−𝑅550)−2.5×(𝑅670−𝑅550)] Reno malized Di e ence Vege a ion Index (RDVI) RDVI= RNIR-RVIS √RNIR+RVIS (Roujean and B eon 1995) [37] G eenness Index (G) - 𝐺= 𝑅𝐺𝑅𝐸𝐸𝑁 𝑅𝑅𝐸𝐷 Modi ied SAVI (MSAVI) (Qi e al. 1994) [38] 𝑀𝑆𝐴𝑉𝐼= 1 2 [2× 𝑅800+1−√(2×𝑅800+1)2−8×(𝑅800+𝑅670)] Op imized Soil Adjus ed Vege a ion Index (OSAVI) (Rondeaux e al. 1996) [39] 𝑂𝑆𝐴𝑉𝐼 = (1+0.16)(𝑅𝑁𝐼𝑅 −𝑅𝑅𝐸𝐷) (𝑅𝑁𝐼𝑅 +𝑅𝑅𝐸𝐷 +0.16) Modi ied Cab Abso p ion in Re lec ance Index (MCARI) (Daugh y e al. 2000) [40] 𝑀𝐶𝐴𝑅𝐼 =[(𝑅700+𝑅670)−0.2×(𝑅700−𝑅550)]×(𝑅700 𝑅670) T ans o med CARI (TCARI) (Haboudane e al. 2002) [41] 𝑇𝐶𝐴𝑅𝐼 = 3 × [(𝑅700–𝑅670)–0.2×(𝑅700−𝑅550)×(𝑅700 𝑅670)] TCARI/OSAVI (Haboudane e al. 2002) [41] 𝑇𝐶𝐴𝑅𝐼/𝑂𝑆𝐴𝑉𝐼 = 3×[(𝑅700–𝑅670)–0.2×(𝑅700−𝑅550)×(𝑅700 𝑅670 ⁄ )] (1+0.16)×𝑅800−𝑅670 𝑅800+𝑅670+0.16 Plan Cell Densi y Index (PCD) o Ra io Vege a ion Index (RVI) (Jo dan 1969) [42] 𝑃𝐶𝐷=𝑅𝑉𝐼= 𝑅𝑁𝐼𝑅 𝑅𝑅𝐸𝐷 Remo e Sens. 2015, 7 14466 3. Resul s and Discussion 3.1. Spa ial Va iabili y o Spec al Indices, Field Va iables, and Be y Composi ion Spec al indices selec ed om li e a u e (Table 1) and he no malised indices (Equa ion (3)) we e co ela ed wi h ield (g ound u h) da a in o de o exclude om u he analysis hose indices ha exhibi ed weak o no signi ican co ela ions. As a esul , co ela ions wi h a Pea son coe icien , “ ”, smalle han 0.35 in absolu e alue we e disca ded. Fu he mo e, he spec al indices wi h a high co ela ion coe icien we e also disca ded o a oid edundan in o ma ion. The spec al indices ha showed he lowes co ela ion we e MTV1 o RDVI, which yielded mean co ela ion coe icien s wi h in- ield a iables o a ound 0.2. The indices inally selec ed o u he analysis we e NDVI and G. Two indices om he g oup o no malised indices we e also selec ed, which we e named as No malised Vege a ion Index 1, NVI1 (R802 − R531/R802 + R531), and No malised Vege a ion Index 2, NVI2 (R802 − R570/R802 + R570), bo h esul ing om he combina ion be ween NIR and g een e lec ance alues, simila o he G een NDVI p oposed by Gi elson e al. [46]. Desc ip i e s a is ics o he spec al indices and ield a iables a e epo ed in Table 2. As i can be obse ed, mos o hem exhibi ed conside able a iabili y wi hin he ineya d. On he basis o coe icien o a ia ion (CV) and Sp ead pa ame e s, seconda y shoo leng h and seconda y lea a ea we e, by a , he ield a iables showing he la ges a iabili y (CV = 80%, Sp ead > 370%), ollowed by yield pe ine (CV = 57%, Sp ead = 280%), and numbe o clus e s pe ine (CV = 51%, Sp ead = 250%). These alues we e simila o hose obse ed by Baluja e al. [47] in Temp anillo in La Rioja (Spain) and by B amley and Lamb [48] and B amley and Hamil on [49] in Aus alia, in o he a ie ies. The luo escence index indica i e o lea chlo ophyll con en (SFR_RAD), he no malised indices NVI1 and NVI2, be y weigh , pH, TSS, and i a able acidi y showed a CV smalle han 10%. NBI has been desc ibed as a good indica o o he ni ogen s a us o g ape ine lea es [24,50,51]. I s calcula ion depends on he concen a ion o la onoids and chlo ophyll in he lea , so i s a iabili y mus be highe han ha o chlo ophyll and la onoids alone [51]. Fu he mo e, he p esence o wa e de ici due o e y wa m empe a u es ( he a e age o he mon h mean empe a u es du ing he s udy yea exceeded be ween 1.5 °C and 3.2 °C he alue o he a e age o he his o ical se ies mean empe a u es) and d yness ( o al ain all om July o Sep embe was 30.9% o a e age his o ical ain all in his pe iod) as epo ed by he Na a a Go e nmen ’s me eo ological s a ion o Es ella, may ha e induced a weak mine aliza ion o soil ni ogen, which in u n led o ni ogen de iciency on he lea es o he ines in some a eas o he ineya d, hence con ibu ing o he la ge a iabili y obse ed o he NBI_GAD index. All o he expe imen al a iog ams o he a iables s udied we e i ed o a Sphe ical o Gaussian model, wi h he excep ion o yield pe ine and he an hocyanin concen a ion, which showed an absence o spa ial s uc u e as also ound by Ce o ic e al. [52]. The ange o he a iog am is he dis ance a which he sill (maximum semi a iance) is achie ed and indica es whe he he samples a e spa ially dependen (samples sepa a ed by dis ances close han he ange) o spa ially independen (samples sepa a ed by dis ances la ge han he ange) [53]. The ange alue was a iable o all pa ame e s s udied (Table 2). Main lea a ea, main shoo leng h, pH, and i a able acidi y showed a ange a ound 125 m. The spec al index, G, was he a iable wi h he highes ange alue (202 m). The lowes alues o he ange ( om 38 o 65 m), ha is he lowes dis ance o spa ial dependence, we e ound o he Remo e Sens. 2015, 7 14473 Figu e 6. Maps ob ained o g ape quali y pa ame e s (pH, o al soluble solids, Ti a able acidi y, an hocyanins, and o al phenols) in a Temp anillo (Vi is ini e a L.) ineya d. Maps we e ep esen ed by e ciles. AU: abso bance uni s. Rega ding be y composi ion, pH and TSS ha e shown he highes alues on he Eas side o he plo and he lowes alues on he Wes side (Figu e 6). Consequen ly, i a able acidi y showed he opposi e beha iou , wi h highe alues on he Wes side and lowe on he Eas side, as he spec al index G (Figu e 3). The an hocyanin concen a ion has showed a andom pa e n ac oss he plo , while o al phenols yielded he lowes alues a he cen e o he plo (Figu e 6), inc easing owa d he sides, simila o he spa ial pa e n o NVI1 and NVI2 (Figu e 3). The isual ma ch o he spa ial pa e ns be ween he spec al indices and he in- ield a iables and be y composi ion was quan i a i ely assessed by ca ying ou a c oss- abula ion and he compu a ion o Remo e Sens. 2015, 7 14474 he kappa index om each a iable 3-clus e -map. Table 4 shows he kappa index alue o he c oss- abula ion be ween he spec al indices and he in- ield and be y composi ion a iables. The kappa index yielded alues anging om –0.27 o 0.70, ha is, om poo o subs an ial ag eemen , ollowing he classi ica ion p oposed by Landis and Koch [62]. Table 4. C oss- abula ion ou pu s and kappa index ob ained o he classi ied maps o spec al indices (NDVI, G, NVI1, and NVI2), ege a i e pa ame e s (p uning weigh , main shoo leng h, seconda y shoo leng h, main lea a ea pe shoo , and seconda y lea a ea pe shoo ), luo escence-based indices o ni ogen balance index (NBI_GAD) and chlo ophyll con en in lea es (SFR_RAD) o yield pa ame e s (clus e numbe pe ine, clus e weigh , yield pe ine, and be y weigh ) and o g ape composi ion a iables (pH, o al soluble solids, i a able acidi y, an hocyanins, and o al phenols) in a Temp anillo (Vi is ini e a L.) ineya d. NDVI G NVI1 NVI2 Main shoo leng h 0.41 0.67 0.18 0.17 Seconda y shoo leng h 0.47 0.59 0.45 0.44 Main lea a ea 0.43 0.70 0.19 0.16 Seconda y lea a ea 0.47 0.57 0.46 0.45 P uning weigh 0.32 0.08 0.39 0.35 SFR_R 0.11 0.10 0.39 0.30 NBI_G 0.32 0.36 0.38 0.35 Clus e numbe −0.26 −0.27 −0.17 −0.15 Mean clus e weigh 0.35 0.35 0.37 0.40 Yield 0.01 −0.14 0.01 −0.01 Mean be y weigh 0.12 0.11 0.13 0.12 pH 0.14 0.35 −0.05 −0.05 To al soluble solids 0.11 0.25 −0.06 −0.04 Ti a able acidi y −0.16 −0.15 −0.24 −0.23 An hocyanins −0.01 0.00 −0.04 −0.07 To al phenols 0.00 −0.02 −0.09 −0.08 These esul s show ha , spa ially speaking, he classical index G was he one showing a spa ial pa e n ha bes ag eed wi h he spa ial pa e n o he ege a i e a iables, especially hose ega ding shoo leng h and lea a ea, wi h alues anging om 0.59 o seconda y shoo leng h (mode a e ag eemen ) o 0.70 o main lea a ea (subs an ial ag eemen ). P uning weigh , lea chlo ophyll con en , and ni ogen s a us be e concu ed wi h he no malized index NVI1, yielding a ai ag eemen (0.40). On he o he hand, ega ding he p oduc i e and be y composi ion a iables, mean clus e weigh and pH showed he highes kappa index alues, especially wi h he no malized index NVI2 and G, espec i ely (0.40 and 0.35, ai ag eemen ), while clus e numbe , yield, mean be y weigh , pH, o al soluble solids, i a able acidi y, an hocyanins, and o al phenols showed he lowes kappa coe icien s wi h alues indica ing poo o sligh ag eemen wi h he spec al indices, con i ming he poo co ela ions obse ed. The ma ch o he spa ial pa e ns is showing ha , in gene al, he spec al indices de i ed om he mul ispec al images aken om a RPAS a e use ul ools o he g ape-g owe o manage he spa ial a iabili y o he ineya ds. These indices seem o be mo e eliable o he managemen o he ege a i e Remo e Sens. 2015, 7 14475 a iables han o he p oduc i e and g ape composi ion pa ame e s, which is in ag eemen wi h he ac ha he mul ispec al came a is e ie ing he da a om an o e head poin o iew, ha is, om he canopy, no “seeing” he p oduc i e pa o he g ape ines loca ed below he canopy. The e o e, while he mul ispec al senso is di ec ly sensing he ege a i e componen o he ineya d, he ela ion wi h he p oduc i e componen o he ineya d emains indi ec . Mo eo e , depending on he me eo ological condi ions in he season, mo e igo ous suba eas o a ineya d plo may exhibi a highe wa e demand han less igo ous plan s in a d y yea , expe iencing wa e s ess ha may jeopa dize bo h g ape yield and composi ion, while in coole and less d y seasons, highe yield alues a e expec ed in igo ous plan s. In summa y, he spec al indices ob ained om he mul ispec al images cap u ed om a RPAS ha e been shown o signi ican ly co ela e wi h he ege a i e a iables and some o he yield a iables. Two newly de ined spec al indices ha e also been p o ided ha ha e shown he bes co ela ions wi h impo an ege a i e a iables such as p uning weigh , ni ogen s a us, o lea chlo ophyll con en . The classical me hods o he measu emen o ege a i e and yield pa ame e s a e labo ious and gene ally des uc i e. The e o e, he spec al indices could be a use ul ool o es ima e hese a iables o in e es in a as and non-des uc i e way, a a ime ea lie han hey could be assessed by he classical me hods. Fu he mo e, mul ispec al images and hei de i ed spec al indices p o ide he end-use wi h a con inuous su ace o da a om he ineya d wi h high spa ial esolu ion; ha is, a la ge da abase ha makes i possible o ake in o accoun he spa ial a iabili y o he ineya d s a us. This ac makes i ope a ionally easible o ca y ou a segmen a ion o he ineya d in 2 o 3 homogeneous managemen zones. In addi ion, compa ed o o he emo e sensing pla o ms, RPAS p o ide he g ape-g owe no only wi h highe spa ial esolu ion, bu also wi h highe lexibili y o moni o hei ineya ds, especially he small ineya ds p e ailing in Eu ope (a e age ineya d size is less han 5 ha). Ano he impo an ac o o he g ape-g owe ha has no been p o usely add essed in li e a u e is he cos o applying emo e sensing echnologies. A e y in e es ing pape has ecen ly been published by Ma ese e al. [14] in which he au ho s ca y ou a cos compa ison among h ee di e en emo e sensing pla o ms: sa elli e, ai c a , and RPAS. Thei esul s ha e shown ha , o small su aces (5 ha), as in he p esen wo k, he use o RPAS is he mos cos -e ec i e solu ion due o he lowes acquisi ion cos s [14]. The si ua ion changes wi h a highe su ace o s udy (50 ha), as mo e images a e needed o co e i , inc easing he cos s o acquisi ion, p ep ocessing, and analysis o hese images. In his case, he sa elli e would be a mo e cos -e ec i e choice, as i is able o co e a high su ace wi h one image, dec easing he subsequen cos s. To summa ize, as shown by Ma ese e al. [14], each echnology has echnological, ope a ional, and economic ad an ages and disad an ages and mus be selec ed acco ding o he wo k planned. In his s udy, some incon eniences ega ding he ope a ional ac o s o he ligh we e e ealed. Despi e g ea e independence and lexibili y o RPAS sys ems, one o hei disad an ages is ha hey a e signi ican ly a ec ed by he wind, e en by ligh wind, gene a ing mo emen s o he pla o m ha we e e lec ed in he images as geome ic dis o ions. In addi ion, a lowe lying heigh would esul in a highe spa ial esolu ion o he images, which should educe he p oblem o he spec al mix u e, p o iding pu e ine pixels. These ac s may lead o imp o ed co ela ions be ween spec al indices and in- ield a iables. Ou esul s ha e shown ha he h ee sub-zones delinea ed in he ineya d based on he spec al indices ag eed wi h key ege a i e pa ame e s such as p uning weigh , main lea a ea, o main shoo Remo e Sens. 2015, 7 14476 leng h, and e en wi h yield pa ame e s such as mean clus e weigh . I has been explo ed he e by he applica ion o kappa index, a common index usually used o assess land co e changes [63] ha has p o en e y use ul in assessing he ag eemen be ween he suba eas delinea ed wi h spec al indices and he in- ield a iables. These a iables ha e been p o en o be empo ally s able [49,64], so i could be use ul o ca y ou ege a i e managemen ac ions such as e ilisa ion, soil a ming (i.e., in e - ow co e c opping o limi he plan igou in e y igo ous suba eas) and win e p uning in he ollowing yea s. Fu he mo e, he possibili y o cus omizing he il e selec ion o he mul ispec al came a may allow o collec ion o da a in he sho wa e in a ed egion o he spec um i il e s wi hin his egion a e chosen, which could p o ide ele an in o ma ion on he plan wa e s a us and could also p o ide in e es ing da a ega ding yield [65]. In his wo k, hese new echnologies made i possible o delinea e h ee subzones o homogenous ege a i e s a us wi hin he ineya d, whe e he g ape-g owe could de ine i s equi emen s and apply he app op ia e inpu s (wa e , ni ogen, e c.) o managemen p ac ices o each subzone— ha is, he implemen a ion o p ecision i icul u e echniques in he ineya d. 4. Conclusions The esul s p esen ed in his s udy con i m he po en ial o mul i-senso RPAS sys ems o he assessmen o he spa ial a iabili y o he ineya d ege a i e and p oduc i e s a us in p ecision i icul u e. The wo de eloped ege a ion spec al indices, NVI1 and NVI2, we e hose bes co ela ed wi h ield ege a i e pa ame e s measu ed in sampling plo s. Mo eo e , he kappa index has shown ha he spa ial pa e n o he spec al indices ag eed wi h ha o he ege a i e and some o he p oduc i e a iables measu ed in he ineya d. In his case, he classical spec al index G was he one showing he highes spa ial ag eemen wi h he ege a i e a iables. The spec al indices enabled he delinea ion o h ee sub-zones o homogeneous managemen , wi hin he ame o p ecision i icul u e, o op imize he managemen o inpu s (wa e , ni ogen, e c.), soil managemen , and win e p uning. These esul s a e a e y impo an ou come because hey p o ed ha he use o a mul ispec al senso moun ed on a RPAS could become a eliable ool o he g ape-g owing indus y o manage he spa ial a iabili y wi hin he ineya d. The possibili y o acqui ing “on-demand” low-cos images wi h unp eceden ed high spa ial esolu ion will ce ainly p omo e i s ope a ional applica ion o ineya d managemen in o de o imp o e c op p oduc i i y and sus ainabili y. Acknowledgmen s This s udy was ca ied ou as pa o he Tele i is p ojec (Re . ADER-2008-00187) unded by he Agencia de Desa ollo Tecnológico de La Rioja (La Rioja, Spain). Special g a i ude goes o Ai es udio Geoin o ma ion Technologies S. Cop. o he acquisi ion o images wi h he RPAS. Addi ionally, we hank Ja ie Baluja and Vic o Sicilia o collec ing ield da a, and In i Luna o his wo k on he image co ec ions. Remo e Sens. 2015, 7 14477 Au ho Con ibu ions MPD and JT designed he s udy and implemen ed he expe imen s. CRC, MPM and AL analysed he da a. CRC and MPD w o e he manusc ip d a . All au ho s e iewed and app o ed he inal manusc ip . Con lic s o In e es The au ho s decla e ha his wo k was pa ially unded by Fo ce-A by he lending o he senso s used. Re e ences 1. Tey, Y.S.; B indal, M. Fac o s in luencing he adop ion o p ecision ag icul u al echnologies: A e iew o policy implica ions. P ecis. Ag ic. 2012, 13, 713–730. 2. Schie e , J.; Dillon, C. The economic and en i onmen al impac s o p ecision ag icul u e and in e ac ions wi h ag o-en i onmen al policy. P ecis. Ag ic. 2015, 16, 46–61. 3. Hall, A.; Lamb, D.W.; Holzap el, B.; Louis, J. Op ical emo e sensing applica ions in i icul u e-a e iew. Aus . J. G ape Wine Res. 2002, 8, 36–47. 4. Ace edo-Opazo, C.; Tissey e, B.; Guillaume, S.; Ojeda, H. The po en ial o high spa ial esolu ion in o ma ion o de ine wi hin- ineya d zones ela ed o ine wa e s a us. P ecis. Ag ic. 2008, 9, 285–302. 5. Baluja, J.; Diago, M.P.; Balda, P.; Zo e , R.; Meggio, F.; Mo ales, F.; Ta daguila, J. Assessmen o ineya d wa e s a us a iabili y by he mal and mul ispec al image y using an unmanned ae ial ehicle (UAV). I ig. Sci. 2012, 30, 511–522. 6. Ma in, P.; Za co-Tejada, P.J.; Gonzalez, M.R.; Be jón, A. Using hype spec al emo e sensing o map g ape quali y in “ emp anillo” ineya ds a ec ed by i on de iciency chlo osis. VITIS 2007, 46, 7–14. 7. Za co-Tejada, P.J.; Guillén-Climen , M.L.; He nández-Clemen e, R.; Ca alina, A.; González, M.R.; Ma ín, P. Es ima ing lea ca o enoid con en in ineya ds using high esolu ion hype spec al image y acqui ed om an unmanned ae ial ehicle (UAV). Ag ic. Fo . Me eo ol. 2013, 171–172, 281–294. 8. Hall, A.; Louis, J.P.; Lamb, D.W. Low- esolu ion emo ely sensed images o wineg ape ineya ds map spa ial a iabili y in planime ic canopy a ea ins ead o lea a ea index. Aus . J. G ape Wine Res. 2008, 14, 9–17. 9. Ma hews, A.; Jensen, J. Visualizing and quan i ying ineya d canopy LAI using an unmanned ae ial ehicle (UAV) collec ed high densi y s uc u e om mo ion poin cloud. Remo e Sens. 2013, 5, 2164–2183. 10. Lamb, D.W.; Weedon, M.M.; B amley, R.G.V. Using emo e sensing o p edic g ape phenolics and colou a ha es in a cabe ne sau ignon ineya d: Timing obse a ions agains ine phenology and op imising image esolu ion. Aus . J. G ape Wine Res. 2004, 10, 46–54. Remo e Sens. 2015, 7 14478 11. Ma ese, A.; Cap a o, F.; P imice io, J.; Guala o, G.; Genna o, S.F.D.; Aga i, G. Mapping o ine igo by UAV and an hocyanin con en by a non-des uc i e luo escence echnique. In P ecision Ag icul u e’13; S a o d, J., Ed.; Wageningen Academic Publishe s: Wageningen, The Ne he lands, 2013; pp. 201–208. 12. Se ano Po a, L.; Za co Tejada, P.; González Flo , C.; Pons Valls, J.M.; Go chs Al a iba, G. Assessmen o be y quali y using ai bo ne de i ed NDVI and PRI in ain ed ineya ds. In P oceedings o he 9 h Eu opean Con e ence on P ecision Ag icul u e, Lleida, Spain, 7–11 July 2013. 13. Tu ne , D; Luciee , A.; Wa son, C. De elopmen o an unmanned ae ial ehicle (UAV) o hype esolu ion ineya d mapping based on isible, mul ispec al, and he mal image y. In P oceedings o he 34 h In e na ional Symposium on Remo e Sensing o En i onmen , Sydney, NSW, Aus alia, 10–15 Ap il 2011; p. 4. 14. Ma ese, A.; Toscano, P.; Di Genna o, S.; Genesio, L.; Vacca i, F.; P imice io, J.; Belli, C.; Zaldei, A.; Bianconi, R.; Gioli, B. In e compa ison o ua , ai c a and sa elli e emo e sensing pla o ms o p ecision i icul u e. Remo e Sens. 2015, 7, 2971–2990. 15. Be ni, J.; Za co-Tejada, P.J.; Suá ez, L.; Fe e es, E. The mal and na owband mul ispec al emo e sensing o ege a ion moni o ing om an unmanned ae ial ehicle. IEEE T ans. Geosci. Remo e Sens. 2009, 47, 722–738. 16. Zhang, C.; Ko acs, J. The applica ion o small unmanned ae ial sys ems o p ecision ag icul u e: A e iew. P ecis. Ag ic. 2012, 13, 693–712. 17. Jenkins, D.; Vasigh, B. The Economic Impac o Unmanned Ai c a Sys ems In eg a ion in he Uni ed S a es. A ailable online: h p://www.au si.o g/econ epo (accessed on 10 Augus 2015). 18. Sepulc e-Can ó, G.; Diago, M.P.; Balda, P.; Ma ínez de Toda, F.; Mo ales, F.; Ta dáguila, J. Moni o ing ineya d spa ial a iabili y o ege a i e g ow h and physiological s a us using an unmanned ae ial ehicle (UAV). In P oceedings o he 16 h In e na ional Symposium o GiESCO, Da is, CA, USA, 16–20 July 2009. 19. Gamon, J.A.; Peņuelas, J.; Field, C.B. A na ow-wa eband spec al index ha acks diu nal changes in pho osyn he ic e iciency. Remo e Sens. En i on. 1992, 41, 35–44. 20. Ma ínez de Toda, F.; Ta dáguila, J.; Sancha, J.C. Es ima ion o g ape quali y in ineya ds using a new i icul u al index. VITIS—J. G ape ine Res. 2007, 46, 168–173. 21. Sma , R.; Robinson, M. Sunligh in o Wine: A Handbook o Wineg ape Canopy Managemen ; Wine i les: Adelaida, SA, Aus alia, 1991. 22. Ben Ghozlen, N.; Moise, N.; La ouche, G.; Ma inon, V.; Me cie , L.; Besancon, E.; Ce o ic, Z. Assessmen o g ape ine ma u i y using a new po able senso : Non des uc i e quan i ica ion o an hocyanins. J. In . Sci. Vigne Vin 2010, 44, 1–8. 23. Ben Ghozlen, N.; Ce o ic, Z.G.; Ge main, C.; Tou ain, S.; La ouche, G. Non-des uc i e op ical moni o ing o g ape ma u a ion by p oximal sensing. Senso s 2010, 10, 10040–10068. 24. T emblay, N.; Wang, Z.; Ce o ic, Z. Sensing c op ni ogen s a us wi h luo escence indica o s. A e iew. Ag on. Sus ain. De . 2012, 32, 451–464. 25. Rey-Ca amés, C. Assessmen o he Spa ial Va iabili y o Vege a i e S a us in Vineya ds Using Non-Des uc i e senso s. Applica ion o Remo e and P oximal Sensing Technologies in P ecision Vi icul u e. Ph.D. Thesis, Uni e si y o La Rioja, Log oño, Spain, 10 July 2015. Remo e Sens. 2015, 7 14479 26. OIV. Compendium o In e na ional Me hods o Wine and Mus Analysis; In e na ional O ganisa ion o Vine and Wine: Pa is, F ance, 2012. 27. Iland, P.; B ue , N.; Edwa ds, G.; Weeks, S.; Wilkes, E. Chemical Analysis o G apes and Wine: Techniques and Concep s; Pa ick Iland Wine P omo ions: Campbell own, SA, Aus alia, 2004. 28. Sánchez-de-Miguel, P.; Baeza, P.; Junque a, P.; Lissa ague, J. Vege a i e de elopmen : To al lea a ea and su ace a ea indexes. In Me hodologies and Resul s in G ape ine Resea ch; Del o , S., Med ano, H., O , E., Ba a esco, L., G ando, S., Eds.; Sp inge Ne he lands: Do d ech , The Ne he lands 2010; pp. 31–44. 29. Lalibe e, A.S.; Go o h, M.A.; S eele, C.M.; Rango, A. Mul ispec al emo e sensing om unmanned ai c a : Image p ocessing wo k lows and applica ions o angeland en i onmen s. Remo e Sens. 2011, 3, 2529–2551. 30. Inglada, J.; Ch is ophe, E. The o eo oolbox emo e sensing image p ocessing so wa e. In P oceedings o he IEEE In e na ional Geoscience and Remo e Sensing Symposium (IGARSS’09), Cape Town, Sou h A ica, 12–17 July 2009; pp. 733–736. 31. R Co e Team. R: A Language and En i onmen o S a is ical Compu ing; R Founda ion o S a is ical Compu ing: Vienna, Aus ia, 2012. 32. Kei , T.H.; Bi and, R.; Pebesma, E.; Rowlingson, B. Rgdal: Bindings o he Geospa ial Da a Abs ac ion Lib a y. A ailable online: h p://CRAN.R-p ojec .o g/package= gdal (accessed on 10 Augus 2015). 33. Hijmans, R.J.; an E en, J. Ras e : Geog aphic Analysis and Modeling wi h Ras e Da a. A ailable online: h p://CRAN.R-p ojec .o g/package= as e (accessed on 10 Augus 2015). 34. Rouse, J.W.; Haas, R.H.; Schell, J.A.; Dee ing, D.W.; Ha lan, J.C. Moni o ing he Ve nal Ad ancemen s and Re og ada ion o Na u al Vege a ion; NASA/GSFC Final Repo ; NASA: G eenbel , MD, USA, 1974; pp. 1–371. 35. Chen, J.M. E alua ion o ege a ion indices and a modi ied simple a io o bo eal applica ions. Can. J. Remo e Sens. 1996, 22, 229–242. 36. Haboudane, D.; Mille , J.R.; Pa ey, E.; Za co-Tejada, P.J.; S achan, I.B. Hype spec al ege a ion indices and no el algo i hms o p edic ing g een LAI o c op canopies: Modeling and alida ion in he con ex o p ecision ag icul u e. Remo e Sens. En i on. 2004, 90, 337–352. 37. Roujean, J.L.; B eon, F.M. Es ima ing pa abso bed by ege a ion om bidi ec ional e lec ance measu emen s. Remo e Sens. En i on. 1995, 51, 375–384. 38. Qi, J.; Chehbouni, A.; Hue e, A.; Ke , Y.; So ooshian, S. A modi ied soil adjus ed ege a ion index. Remo e Sens. En i on. 1994, 48, 119–126. 39. Rondeaux, G.; S e en, M.; Ba e , F. Op imiza ion o soil-adjus ed ege a ion indices. Remo e Sens. En i on. 1996, 55, 95–107. 40. Daugh y, C.S.T.; Wal hall, C.L.; Kim, M.S.; De Cols oun, E.B.; McMu ey, J.E. Es ima ing co n lea chlo ophyll concen a ion om lea and canopy e lec ance. Remo e Sens. En i on. 2000, 74, 229–239. 41. Haboudane, D.; Mille , J.R.; T emblay, N.; Za co-Tejada, P.J.; Dex aze, L. In eg a ed na ow-band ege a ion indices o p edic ion o c op chlo ophyll con en o applica ion o p ecision ag icul u e. Remo e Sens. En i on. 2002, 81, 416–426. Remo e Sens. 2015, 7 14480 42. Jo dan, C.F. De i a ion o lea -a ea index om quali y o ligh on he o es loo . Ecology 1969, 50, 663–666. 43. B amley, R.G.V. Unde s anding a iabili y in wineg ape p oduc ion sys ems 2. Wi hin ineya d a ia ion in quali y o e se e al in ages. Aus . J. G ape Wine Res. 2005, 11, 33–42. 44. Pebesma, E.J. Mul i a iable geos a is ics in S: The gs a package. Compu . Geosci. 2004, 30, 683–691. 45. Hudson, W.D.; Ramm, C.V. Co ec o mula ion o he kappa coe icien o ag eemen . Pho og amm. Eng. Remo e Sens. 1987, 53, 421–422. 46. Gi elson, A.A.; Kau man, Y.J.; Me zlyak, M.N. Use o a g een channel in emo e sensing o global ege a ion om EOS-MODIS. Remo e Sens. En i on. 1996, 58, 289–298. 47. Baluja, J.; Diago, M.P.; Goo ae s, P.; Ta daguila, J. Assessmen o he spa ial a iabili y o an hocyanins in g apes using a luo escence senso : Rela ionships wi h ine igou and yield. P ecis. Ag ic. 2012, 13, 457–472. 48. B amley, R.; Lamb, D. Making sense o ineya d a iabili y in Aus alia. In P oceedings o he In e na ional Symposium on P ecision Vi icul u e, Nin h La in Ame ican Cong ess on Vi icul u e and Oenology, San iago, Chile, 24–28 No embe 2003; pp. 35–54. 49. B amley, R.G.V.; Hamil on, R.P. Unde s anding a iabili y in wineg ape p oduc ion sys ems 1. Wi hin ineya d a ia ion in yield o e se e al in ages. Aus . J. G ape Wine Res. 2004, 10, 32–45. 50. Aga i, G.; Foschi, L.; G ossi, N.; Guglielmine i, L.; Ce o ic, Z.G.; Vol e ani, M. Fluo escence-based e sus e lec ance p oximal sensing o ni ogen con en in Paspalum agina um and Zoysia ma ella u g asses. Eu . J. Ag on. 2013, 45, 39–51. 51. Ca ela , A.; Ce o ic, Z.G.; Goulas, Y.; Meye , S.; Lela ge, C.; P ioul, J.L.; Ba bo in, A.; Jeu oy, M.H.; Ga e, P.; Aga i, G.; e al. Op ically assessed con en s o lea polyphenolics and chlo ophyll as indica o s o ni ogen de iciency in whea (T i icum aes i um L.). Field C ops Res. 2005, 91, 35–49. 52. Ce o ic, Z.G.; Gou ouly, J.-P.; Hilbe , G.; Des ac-I ine, A.; Ma inon, V.; Moise, N. Mapping wineg ape quali y a ibu es using po able luo escence-based senso s. F u ic 2009, 9, 301–310. 53. Oli e , M.A. Geos a is ical Applica ions o P ecision Ag icul u e; Sp inge : London, UK, 2010. 54. Taylo , J.A.; Ba es, T.R. Sampling and es ima ing a e age p uning weigh s in conco d g apes. Am. J. Enol. Vi ic. 2012, 63, 559–563. 55. Camba della, C.A.; Moo man, T.B.; Pa kin, T.B.; Ka len, D.L.; No ak, J.M.; Tu co, R.F.; Konopka, A.E. Field-scale a iabili y o soil p ope ies in cen al Iowa soils. Soil Sci. Soc. Am. J. 1994, 58, 1501–1511. 56. Bonilla, I.; Ma inez de Toda, F.; Ma ínez-Casasno as, J.A. Vine igo , yield and g ape quali y assessmen by ai bo ne emo e sensing o e h ee yea s: Analysis o unexpec ed ela ionships in c . Temp anillo. Span. J. Ag ic. Res. 2015, 13, 1–8. 57. Dob owski, S.Z.; Us in, S.L.; Wolpe , J.A. G ape ine do man p uning weigh p edic ion using emo ely sensed da a. Aus . J. G ape Wine Res. 2003, 9, 177–182. 58. Hall, A.; Lamb, D.W.; Holzap el, B.P.; Louis, J.P. Wi hin-season empo al a ia ion in co ela ions be ween ineya d canopy and wineg ape composi ion and yield. P ecis. Ag ic. 2011, 12, 103–117. 59. P o i , T.; Malcolm, A. Implemen ing zonal ineya d managemen h ough ai bo ne emo e sensing. Aus . N. Z. G apeg ow. Winemak. 2005, 502, 22–27. Remo e Sens. 2015, 7 14481 60. Kelle , M.; Ta a a, J.M.; Mills, L.J. Sp ing empe a u es al e ep oduc i e de elopmen in g ape ines. Aus . J. G ape Wine Res. 2010, 16, 445–454. 61. Ma ínez de Toda, F. Cla es de la Vi icul u a de Calidad, Nue as Técnicas de Es imación y Con ol de la Calidad de la U a en el Viñedo; Mundi-P ensa: Mad id, Spain, 2011. 62. Landis, J.R.; Koch, G.G. The measu emen o obse e ag eemen o ca ego ical da a. Biome ics 1977, 33, 159–174. 63. Congal on, R.G.; G een, K. Assessing he Accu acy o Remo ely Sensed Da a; CRC P ess: Boca Ra on, FL, USA, 2008. 64. Tissey e, B.; Mazzoni, C.; Fon a, H. Whi hin- ield empo al s abili y o some pa ame e s in i icul u e: Po en ial owa d a si e speci ic managemen . J. In . Sci. Vigne Vin. 2008, 42, 27. 65. Se ano, L.; González-Flo , C.; Go chs, G. Assessmen o g ape yield and composi ion using he e lec ance based wa e index in Medi e anean ain ed ineya ds. Remo e Sens. En i on. 2012, 118, 249–258. © 2015 by he au ho s; licensee MDPI, Basel, Swi ze land. This a icle is an open access a icle dis ibu ed unde he e ms and condi ions o he C ea i e Commons A ibu ion license (h p://c ea i ecommons.o g/licenses/by/4.0/).