Modeling Impcts of Viticulturl nd Environmentl Fctors on 3-Isobutyl-2-Methoxypyrzine in Cbernet frnc Grpes Justin J. Scheiner, 1 Justine E. Vnden Heuvel, 2 Bruce Pn, 3 nd Gvin L. Scks 4 * Abstrct: A study ws conducted to determine the key environmentl nd viticulturl vribles ffecting the concentrtion of 3-isobutyl-2-methoxypyrzine (IBMP) in Cbernet frnc grpes. Berries were smpled from individul vines t 30 dys fter nthesis (DAA), 50 DAA, nd hrvest from 10 nd 8 commercil New York Stte vineyrds in 2008 nd 2009, respectively. IBMP concentrtions t 50 DAA were significntly higher in the wrmer 2008 growing seson (2008, 103 to 239 pg/g; 2009, 12 to 87 pg/g). However, in the cooler 2009 growing seson smller percent decrese in IBMP from 50 DAA to hrvest ws observed, so tht IBMP t hrvest ws not significntly different between yers (2008, 1 to 13 pg/g; 2009, 5 to 14 pg/g). IBMP ccumultion up to 50 DAA nd log-fold decrese of IBMP from 50 DAA to hrvest ws modeled s function of >120 viticulturl nd environmentl vribles (122 in 2008 nd 140 in 2009). Importnt vribles identified for modeling IBMP t 50 DAA were those ssocited with vine vigor, which ws positively correlted with IBMP ccumultion. Cluster light exposure did not explin differences in IBMP ccumultion cross sites, but it ws importnt for modeling smller differences within some sites. IBMP decrese could not be stisfctorily modeled cross multiple sites, but within sites the decrese ws most consistently correlted with clssic fruit mturity indices (totl soluble solids [TSS], TSS*pH 2 ). The intensity of herbceous roms in wines produced from ech site ws not correlted with IBMP concentrtion, but multivrite models indicted tht lower vine wter sttus ws the best predictor of incresed herbceousness. Key words: IBMP, prtil lest squres regression, vigor, methoxypyrzines, modeling The 3-lkyl-2-methoxypyrzines (MPs) re clss of odornts ssocited with herbceous roms of some Bordeux winegrpe (Vitis vinifer L.) cultivrs (e.g., Cbernet frnc, Merlot, Cbernet Suvignon, Crménère, Suvignon blnc). In grpes nd wine, the MPs most often studied re 3-isobutyl-2-methoxypyrzine (IBMP), 3-isopropyl-2-methoxypyrzine (IPMP), nd 3-sec-butyl-2-methoxypyrzine (sbmp). Quntittively, IBMP is predominnt nd is typiclly present in concentrtions n order of mgnitude higher thn IPMP nd sbmp (Alberts et l. 2009). In red wine, the sensory detection threshold of IBMP is reportedly 15 pg/ml (Roujou de Boubée et l. 2000). IBMP msks fruity roms (Hein et l. 2009) nd the positive correltion of IBMP with bell pepper rom intensity is widely 1 Grdute Student nd 2 Assistnt Professor of Viticulture, Cornell University, Deprtment of Horticulture, NYSAES, Genev, NY 14456, nd Plnt Sciences Building, Ithc, NY 14853, 3 former Postdoctorl Scientist, E&J Gllo Winery 600 Yosemite Ave Modesto, CA, 95353, nd 4 Assistnt Professor of Enology, Cornell University, Deprtment of Food Science, NYSAES, Genev, NY 14456. *Corresponding uthor (emil: gls9@cornell.edu) Acknowledgments: The uthors thnk Krl Siebert nd John Brnrd for sttisticl consulttions, Imeld Ryon for technicl ssistnce, nd Anthony Rod Wine Compny, Anyel s Vineyrds, Bedell Cellrs, Fox Run Vineyrds, Hzlitt 1852 Vineyrds, Mike Jordn, Rphel Winery, Shlestone Vineyrds, nd Vetter Vineyrds for coopertion on this project. Supplementl dt is freely vilble with the online version of this rticle t www.jevonline.org. Mnuscript submitted Jn 2011, revised Jul 2011, ccepted Sept 2011. Publiction costs of this rticle defryed in prt by pge fees. Copyright 2012 by the Americn Society for Enology nd Viticulture. All rights reserved. doi: 10.5344/jev.2011.11002 reported (Allen et l. 1991, Roujou de Boubée et l. 2000), lthough the reltionship is less cler for concentrtions round threshold (Preston et l. 2008). Becuse the herbceous roms ssocited with MPs re generlly undesirble in red wine, there hs been interest in developing mngement strtegies to control MP levels. In mture grpe berries, most (>95%) IBMP is locted in the grpe skins t hrvest (Roujou de Boubée et l. 2002) nd 67% ± 13% is extrcted with conventionl red winemking prctices (Ryon et l. 2009). Thus, the concentrtion in finished red wine is lrgely dependent on the concentrtion in grpes t hrvest. Since current remedition techniques to remove MPs from musts or wine re either ineffective or result in other nonselective chnges (Pickering et l. 2006), MPs re most effectively controlled with viticulturl prctices tht reduce their content in grpes (Bogrt nd Bisson 2006). IBMP ccumultes from fruit set until ~0 to 14 dys prior to verison (Roujou de Boubée et l. 2002, Ryon et l. 2008). During fruit mturtion, MPs rpidly degrde to concentrtions <10% of their preverison pek vlues. Preverison IBMP concentrtions correlte strongly to IBMP concentrtions t hrvest within the sme growing region (Ryon et l. 2008). Severl viticulturl nd environmentl prmeters re reported to correlte with MPs. In generl, grpes nd wines produced in cool regions nd in cooler yers re reported to hve higher MP concentrtions (Allen et l. 1994, Flcão et l. 2007). Higher tempertures during the ripening period re thought to enhnce MP degrdtion, leding to lower concentrtions t hrvest (Lcey et l. 1991). Severl groups hve reported tht cluster light exposure cn reduce MPs t hrvest (Mris et l. 1999, Ryon et l. 2008). Exposed 94
Modeling Impcts on IBMP 95 clusters ccumulte less IBMP thn shded clusters, nd proportionl differences persist until hrvest. In contrst, cluster exposure does not influence the rte of postverison degrdtion (Ryon et l. 2008). Consequently, mngement prctices tht improve cluster exposure (e.g., lef removl) cn reduce IBMP when imposed preverison, but postverison tretments re less effective (Scheiner et l. 2010). Conditions tht stimulte vine vigor, such s high wter vilbility nd low bud numbers, re ssocited with high MP concentrtions (Chpmn et l. 2004, Sl et l. 2005), lthough the mechnism is uncler. Higher preverison IBMP concentrtions hve been observed t more vigorous sites in Cliforni (Noble et l. 1995), nd vine growth during the ripening period, induced by high rinfll, ws reported to result in higher IBMP concentrtions in Bordeux (Roujou de Boubée et l. 2002). In Cbernet frnc grown in cool climte, higher IBMP concentrtions were observed in vigorous vines with similr cluster exposures to vines of lower vigor, suggesting tht vine vigor nd cluster light exposure my independently influence MPs (Ryon et l. 2008). In summry, cluster light exposure, temperture, fruit mturity, nd conditions ssocited with vine vigor re linked to MPs, but it is uncler if the observed effects occur independently or indirectly. Studies hve quntified MPs in fruit only t hrvest or in wine, so it is not possible to determine if the observed differences were function of chnges in MP ccumultion or of degrdtion. The objective of this study ws to conduct multivrite nlysis to evlute the correltion mong viticulturl vribles, vine physiology, nd mesond microclimte nd IBMP concentrtions in Cbernet frnc grpe berries t preverison nd hrvest. Mterils nd Methods Experimentl design. Ten nd eight commercil Cbernet frnc vineyrds in New York Stte were used for this study in 2008 nd 2009, respectively (Tble 1). At ech site, two 5-vine pnels were selected for dt collection nd smpling bsed on uniform vine ge within ech vine pnel. Pnels were selected with the intent of cpturing wide rnge of soil types, clones, nd other vribles in order to mximize the rnge in MPs mong ll sites. Mesurement nd smpling were performed on individul vines within ech pnel. Vine mngement ws performed by the cooperting growers ccording to the regionlly pproprite viticulturl prctices for V. vinifer in ech respective re. Vine/cnopy chrcteriztion. The number of count nd noncount shoots ws recorded t nthesis nd 50 dys fter nthesis (50 DAA), nd shoot density ws determined by dividing the number of totl shoots per vine by in-row vine spcing. On divided cnopy systems (i.e., Scott Henry), shoot density ws determined by dividing the number of totl shoots per vine by twice the in-row spcing. At nthesis, shoot length nd number of nodes per shoot were mesured from node 1 to the shoot tip on 20 rndomly selected shoots per vine. Averge internode length ws determined by dividing the verge shoot length by the number of nodes. Enhnced point qudrt nlysis (EPQA) ws conducted t 10 cm intervls in the fruiting zone t nthesis, 30 DAA, 50 DAA, nd hrvest (Meyers nd Heuvel 2008). On trining systems with multiple fruiting zones (i.e., Scott Henry, two-tier fltbow), insertions were mde long ech tier. In 2009, EPQA ws lso performed t 30 cm bove the fruiting zone (midcnopy) t 50 DAA nd hrvest. Photosyntheticlly Tble 1 Loction of Cbernet frnc vineyrds (sites) nd chrcteristics for 5-vine pnels used in the multivrite study. Site Pnel Am. viticulturl re Clone Rootstock Spcing (m/row x Vineyrd m/vine) Trining system ge b Soil series c 1 1 Finger Lkes unknown 3309C 2.7 x 1.6 Two-tier fltbow VSP 7 Auror 1 2 Finger Lkes unknown 3309C 2.7 x 1.6 Two-tier fltbow VSP 7 Auror 2 1 Finger Lkes, Senec Lke 1 3309C 2.7 x 1.8 Two-tier fltbow VSP 13 Cyug 2 2 Finger Lkes, Senec Lke 1 3309C 2.7 x 1.8 Two-tier fltbow VSP 11 Honeoye 3 1 Finger Lkes, Senec Lke 1 SO4 2.7 x 1.7 Two-tier fltbow VSP >15 Cyug 3 2 Finger Lkes, Senec Lke 1 SO4 2.7 x 1.7 Two-tier fltbow VSP >15 Cyug 4 1 Finger Lkes, Senec Lke 1 SO4 3.0 x 2.2 Cordon spur/flt cne SH VSP 6 Howrd 4 2 Finger Lkes, Senec Lke 1 SO4 3.0 x 2.2 Cordon spur/flt cne SH VSP 6 Howrd 5 1 Finger Lkes, Senec Lke 214 SO4 3.0 x 2.2 Cordon spur/flt cne SH VSP 9 Howrd 5 2 Finger Lkes, Senec Lke 214 SO4 3.0 x 2.2 Cordon spur/flt cne SH VSP 9 Howrd 6 1 Finger Lkes, Senec Lke unknown 3309C 2.7 x 1.7 Cordon spur SH VSP 22 Honeoye 6 2 Finger Lkes, Senec Lke unknown 3309C 2.7 x 1.7 Cordon spur VSP 15 Auror 7 1 Long Islnd, North Fork 1 unknown 2.7 x 1.8 Cordon spur/flt cne VSP 12 Hven 7 2 Long Islnd, North Fork 332 unknown 2.7 x 1.5 Cordon spur/flt cne VSP 12 Hven 8 1 Long Islnd, North Fork 1 3309C 2.7 x 1.9 Cordon spur VSP 7 Hven 8 2 Long Islnd, North Fork 1 SO4 2.7 x 1.9 Cordon spur VSP 7 Hven 9 1 Lke Erie unknown 3309C 2.7 x 1.4 Two-tier fltbow VSP 5 Chenngo 9 2 Lke Erie unknown 3309C 2.7 x 1.4 Two-tier fltbow VSP 5 Chenngo 10 1 Lke Erie 327 3309C 2.7 x 1.9 Cordon spur 8 Hornell 10 2 Lke Erie 327 3309C 2.7 x 1.9 Cordon spur 8 Hornell VSP: verticl shoot-positioning; SH: Scott Henry. b Age determined s yers from plnting in 2008; site 3 ge unknown. c Soil Survey Stff, USDA [http://soils.usd.gov/technicl/clssifiction/osd/index.html].
96 Scheiner et l. ctive rdition (PAR, 400 700 nm) ws mesured in the fruiting zone with n AccuPAR LP-80 ceptometer (Decgon Devices, Cmbridge, UK) on cloudless dys between 10:30 nd 15:00 hr. The probe ws inserted prllel to the row in the interior of the cnopy t the fruiting zone nd midcnopy, nd the verge of four redings ws recorded. At 50 DAA nd hrvest, shoot dimeters were mesured midwy between nodes 1 nd 2 on 20 rndomly selected shoots per vine with Storm 3C301 Electronic Digitl Cliper (Centrl Tools, Crnston, RI). At hrvest, the number of nodes of ripe periderm ws counted on 20 rndomly selected shoots per vine. Climtic mesurements. PAR, rinfll, nd ir temperture were monitored from 1 My through hrvest with Hobo Micro Sttion Dt Loggers (Onset Computer Corp., Bourne, MA) plced in ech vineyrd within 50 meters of n experimentl vine pnel. PAR, rinfll, nd temperture were mesured ech minute, nd 5-min verges were logged. Growing degree dys (GDD) were determined s GDD = [(mximum dily temperture + minimum dily temperture) / 2] 10 from 1 My to hrvest, where tempertures were expressed in Celsius. PAR ws not mesured in My nd June in 2008. Smpling nd hrvest prmeters. Fifty-berry smples t 30 nd 50 DAA nd 200-berry smples t hrvest were collected t rndom from ech vine for chemicl nlyses, including IBMP quntifiction (smpling dtes listed in Supplementl Tble 1). Sites in regions outside the Finger Lkes were not visited s frequently due to trvel distnce; therefore slight discrepncies my exist between ctul nd reported phenologicl stges. The berry smples were plced in storge bgs, immeditely frozen with liquid N 2, nd stored t -23 C until nlyses were performed. At ech site, hrvest dte ws determined by the respective winery, nd smple hrvest occurred within three dys of commercil hrvest. Yield per vine ws mesured with hnging scle ccurte to 0.01 kg (model SA3N340; Slter Brecknell, Firmont, MN) nd cluster counts were recorded. Averge cluster weight ws clculted s yield divided by cluster count. Averge fresh berry weight ws determined by weighing the 200-berry hrvest smples with Setr SI410S blnce (Setr Systems Inc., Boxborough, MA). In 2008, percent berry dry weight ws clculted t hrvest on 50-berry subsmple by grinding t 1600 strokes/min for 2 min using 2000 Geno/Grinder (SPEX Certiprep, Metuchen, NJ), nd drying 20 g subsmple in drying oven t 60 C for 48 hr. Percent dry berry weight ws clculted by dividing the initil subsmple wet weight by the finl dried weight. The reminder of the 200-berry smple ws used for crbon isotope composition (δ 13 C) nlysis nd IBMP quntifiction. In 2009, verge fresh berry weight ws dditionlly determined t 30 nd 50 dys fter nthesis using the 50-berry smples. During the winter, vines were pruned ccording to grower specifictions nd dormnt cne pruning weight ws recorded. Crop lod ws clculted s yield divided by pruning weight, nd verge cne weight ws determined s totl pruning weight divided by the number of cnes. Berry nlysis for Brix, titrtble cidity, nd ph. A subsmple of 150 frozen mture berries ws plced in 250 ml beker nd heted to 65 C for 1 hr in wter bth to redissolve bitrtrte crystls, pressed through cheesecloth with pestle, nd the juice ws collected for nlyses. Soluble solids (Brix) were mesured using digitl refrctometer (model 300017; SPER Scientific, Scottsdle, AZ) with temperture correction. Titrtble cidity (TA) nd ph were mesured with n utomtic titrter (Titrino model 798, Metrohm, Riverview, FL), nd TA ws mesured with 5.0 ml liquot of juice by titrtion ginst 0.1 N NOH to ph 8.2 nd expressed s trtric cid equivlents. Berry nd lef nlysis for crbon isotope composition. Crbon isotope composition (δ 13 C) nlysis ws performed on 10 g subsmples of berries collected t 50 DAA nd hrvest nd on three sun-exposed leves collected t 50 DAA nd hrvest from nodes 15, 16, nd 17 from three rndomly selected count shoots. Lef nd berry smples were dried in drying oven t 60 C over 48-hr period nd ground with coffee grinder into fine homogenous powder. δ 13 C nlysis ws performed using Finnign MAT Delt Plus (Bremen, Germny) isotope rtio mss spectrometer interfced to Crlo Erb NC2500 elementl nlyzer nd expressed s δ 13 C = [(R s R pdb ) / R pdb ] x 1000, where R s = 13 C/ 12 C rtio of the smple nd R pdb = 13 C/ 12 C rtio of the Pee Dee Belemnite) stndrd. Berry nlysis of IBMP. 3-Isobutyl-2-methoxypyrzine nlysis ws conducted on 50-berry smples collected t 30 DAA, 50 DAA, nd hrvest. The extrction method ws hedspce solid-phse microextrction (HS-SPME) nd quntifiction ws performed by comprehensive two-dimensionl gs chromtogrphy time-of-flight mss spectrometry (GCxGC- TOF-MS), described in detil elsewhere (Ryon et l. 2009). Fermenttions. Fruit from ech experimentl plot (i.e., 5-vine pnel) ws pooled for fermenttion. Grpes were destemmed, crushed, nd seprted into 7.6 L fermenttion vessels, with two fermenttion replictes used per vine pnel. Musts lower thn 22 Brix were chptlized to 22 Brix to simulte stndrd regionl industry prctices. SO 2 (50 mg/l) ws dded s potssium metbisulfite nd musts were inoculted with Llvin ICV GRE yest (Lllemnd, Snt Ros CA). Yest nutrients were dded s follows: GoFerm (0.15 g/l), Fermid K (0.1 g/l), nd dimmonium hydrogen phosphte (1 g/l DAP, 210 mg/l s N). YAN ws not mesured in individul vine pnels, but N.Y. Stte Cbernet frnc is generlly YAN deficient: the verge in 2010 ws 63 ± 30 mg/l s N (http://grpesndwine.cls.cornell.edu/cls/grpesndwine/ verison-to-hrvest/uplod/verison-to-hrvest-2010-7.pdf), which necessitted the high DAP dditions. Fermenttions were crried out in temperture controlled room t 20 C. The musts were punched down twice dy until the end of lcoholic fermenttion (four to six dys). After the completion of lcoholic fermenttion, n extended mcertion ws crried out for 5 dys. Following extended mcertion, wines were pressed through cheesecloth into 3.78 L crboys nd inoculted for mlolctic fermenttion with Enoferm Alph (Lllemnd). At the end of mlolctic fermenttion, 60 mg/l SO 2 ws dded to finished wines in the form of potssium metbisulfite, followed by cold stbiliztion t 2 C for 60 dys until bottling.
Modeling Impcts on IBMP 97 Wine nlysis for IBMP. At bottling, wine smples were collected for IBMP quntifiction by SPME extrction followed by GCxGC-TOF-MS. Five ml of wine ws trnsferred into 20 ml SPME vil nd diluted with 5 ml Milli-Q wter. NCl (3 g) ws dded to the vil followed by the [ 2 H 2 ]-IBMP internl stndrd t 10 pg/ml. Prmeters for SPME nd GCxGC-TOF-MS were similr to those for whole berries, but with SPME extrction temperture of 40 C nd desorption temperture of 250 C. Sensory evlution. Ten experimentl wines from ech vintge, selected to represent the widest rnge in IBMP concentrtion, were evluted for intensity of herbceous nd fruity roms. A 10-member sensory pnel consisting of eight femles nd two mles, ges 24 to 48, nlyzed the 2008 wines, nd nine-member sensory pnel (sme pnelists s 2008) consisting of eight femles nd one mle, ges 25 to 49, nlyzed the 2009 wines. Sensory pnelists were selected bsed on interest nd vilbility. Prior to the first session, sensory pnelists were fmilirized with the evlution protocols nd scorecrd (nchored 9-point line scle). Arom stndrds were provided to fmilirize sensory pnelists with rnge of herbceous nd fruity roms (Tble 2). Multiple sensory stndrds were used to illustrte herbceousness nd fruitiness, since these terms refer to generl ctegories of roms nd the intent ws to void the ssignment of n herbceous perception to fruity becuse n pproprite stndrd ws not presented. Pnelists were not sked to determine the intensity of ech ttribute bsed on the individul sensory stndrds, s the pnel ws not trined to rte intensity becuse of time nd cost constrints. Following bottling, the 2008 vintge ws stored t the N.Y. Stte Agriculturl Experiment Sttion (NYSAES) wine cellr for eight months (Mrch to October) t 12 C prior to testing, nd sensory nlysis ws performed one yer fter hrvest. For the 2009 vintge, sensory nlysis needed to tke plce erlier becuse of time constrints, nd wines were stored for 5 months following bottling (My to September) t wrmer temperture (21 C) to compenste for the shorter storge time. Sensory evlution ws performed on 2009 wines 9 months fter hrvest. Twenty-four hours prior to testing, wines were moved to the testing re to equilibrte their temperture. Fifteen minutes prior to serving, 40 ml wine ws poured Tble 2 Attributes nd reference stndrds for 2008 nd 2009 sensory pnels for wines mde from individul pnels from sites 2 8 nd 10 (2008) nd sites 1 3 nd 5 8 (2009). Arom ttribute Herbceous Herbceous Herbceous Herbceous Herbceous Fruity Fruity Fruity Reference stndrd 2 g bell pepper 5 ml juice from Gret Vlue cnned sprgus (Bentonville, AR) 5 ml juice from Gret Vlue cnned pes 15 µg/l IBMP 30 µg/l IBMP 2 g Smuckers rspberry jm (Orrville, OH) 2 g Smuckers blckberry jm 2 g Smuckers strwberry jm Stndrds prepred in 40 ml Pinot noir bse wine. into cler, 250 ml tulip-shped glsses. Plstic covers were plced over the glsses to retin roms. Sensory evlution took plce t the NYSAES sensory evlution room under red lighting. Wines were evluted for intensity of herbceous nd fruity roms in triplicte using rndomized complete block design with order of presenttion rndomized within session. Five wine sets were presented, nd sensory pnelists were instructed to evlute ech wine seprtely from left to right. Sensory pnelists wited 30 min nd were then presented nother set of five wines. Ech pnelist completed three sessions over 3-week period. Sttisticl nlysis. Prtil lest squres regression (PLSR) ws conducted with Minitb 15.0 sttisticl softwre (Minitb, Reding, MA) to model IBMP concentrtions in grpes hrvested from the individul vines. For model building, ll x vribles (122 in 2008 nd 140 in 2009; Tble 3) were used to crete n initil model of IBMP in grpes from individul vines (100 vine smples in 2008; 80 vine smples in 2009). Dt were normlized nd the number of ltent vribles in ech model ws determined by the lowest predicted residul sum of squres (PRESS). Leve one out crossvlidtion ws used to clculte p vlues nd cross-vlidted regression coefficients, Q 2 (Brereton 2007). X vribles tht did not contribute to the model were removed mnully by the forwrd selection process (Andersen nd Bro 2010). The x vrible with the lowest regression coefficient ws removed nd the model regenerted. The process ws repeted until no further improvement in PRESS ws observed. PLSR ws implemented in similr mnner to model the intensity of herbceous nd fruit wine rom. In this cse, vine pnel verges were used for viticulturl nd climtic mesurements. One-wy nlysis of vrince (ANOVA) nd Welch s t-test were conducted using SPSS 19.0 sttisticl softwre (SPSS Inc., Chicgo, IL). Mens were seprted using Gmes Howell test t the 5% significnce level. Wine sensory dt were subjected to the mixed models procedure in SAS (SAS Institute Inc., Cry, NC), with judge treted s rndom effect, nd mens were seprted using the Tukey Krmer procedure t the 5% significnce level. Liner regression nlysis ws conducted using SAS (PROC REG). Results Bsic juice chemistry nd climctic prmeters. Rinfll, men verge mximum temperture, nd ccumulted GDDs differed mong sites over ech seson, with greter vrition in climte generlly existing mong regions rther thn mong sites within region (Supplementl Tble 2). The two yers differed significntly in GDD, with 2008 (men GDD, 1458) wrmer thn 2009 (men GDD, 1342) t the sites mesured in both yers (p < 0.05 by pired t-test). Men vlues for bsic juice chemistry (Brix, ph, TA) re reported in Supplementl Tble 3. IBMP concentrtions in Cbernet frnc berries. There were significnt differences ( p < 0.001) in IBMP concentrtion cross sites t ll phenologicl stges (30 DAA, 50 DAA, nd hrvest) in 2008 nd 2009 (Tble 4). In 2008 nd 2009, IBMP concentrtions incresed t nine of 10 sites nd
98 Scheiner et l. Tble 3 X vribles (mesurements) included in initil 2008 nd 2009 PLSR models to predict IBMP concentrtion t 50 DAA nd log percent decrese in IBMP from 50 DAA to hrvest. Vine Crop c Climte Metric Phenology b Metric Metric d Phenology Shoots/vine AN, 50 DAA Yield Rinfll My, Jun, Jul, Aug, Sep, 1 Oct-HAR, 1 My-50 DAA, AN-50 DAA, 50 DAA-HAR, 65 DAA-HAR, 1 My-HAR Shoots/meter AN, 50 DAA Cluster no. Temp My, Jun, Jul, Aug, Sep, 1-15 Oct, 1 My-50 DAA, AN-50 DAA, 50 DAA-15 Oct, 65 DAA-15 Oct, 1 My- 15 Oct Shoot length AN Clusters/shoot GDD My, Jun, Jul, Aug, Sep, 1 Oct-HAR, 1 My-50 DAA, AN-50 DAA, 50 DAA-HAR, 65 DAA-HAR, 1 My-HAR Nodes/shoot AN Cluster wt PAR My, Jun, Jul, Aug, Sep, 1 Oct-HAR, 1 My-50 DAA, AN-50 DAA, 50 DAA-HAR, 65 DAA-HAR, 1 My-HAR Internode length AN Berry fresh wt Shoot dim. 50 DAA, HAR Percent berry dry wt δ 13 C berries 50 DAA, HAR Pruning wt δ 13 C leves 50 DAA, HAR Averge cne wt LLN AN, 30 DAA, 50 DAA, HAR Crop lod PIL AN, 30 DAA, 50 DAA, HAR Yield/shoot PIC AN, 30 DAA, 50 DAA, HAR Soluble solids (TSS) PG AN, 30 DAA, 50 DAA, HAR Titrtble cidity OLN AN, 30 DAA, 50 DAA, HAR ph CEL AN, 30 DAA, 50 DAA, HAR TSS/titrtble cidity LEL AN, 30 DAA, 50 DAA, HAR TSS*pH 2 EP1 AN, 30 DAA, 50 DAA, HAR CEFA AN, 30 DAA, 50 DAA, HAR CEFA* AN, 30 DAA, 50 DAA, HAR LEFA AN, 30 DAA, 50 DAA, HAR LEFA* AN, 30 DAA, 50 DAA, HAR Periderm HAR Nodes per shoot nd internode length mesured in 2008 only. LLN: lef lyer number; PIL: percent interior leves; PIC: percent interior clusters; PG: percent gps; OLN: occlusion lyer number; CEL: cluster exposure lyer; LEL: lef exposure lyer; EP1: cnopy clibrtion coefficient; CEFA: cluster exposure flux vilbility; CEFA*: CEFA computed using dynmic clibrtion model; LEFA: lef exposure flux vilbility; LEFA*: LEFA computed using dynmic clibrtion model; periderm: nodes of ripe periderm. LLN, PIL, PG, OLN, LEL, EP1, LEFA, nd LEFA* mesured t 30 cm bove the fruiting zone in 2009 t 50 DAA nd hrvest. b Mesurement t ech phenologicl stge entered s seprte independent vrible in models; AN: nthesis; DAA: dys fter nthesis; HAR: hrvest. c Crop mesurements tken t hrvest; pruning conducted during winter dormncy; berry weight collected t 30 nd 50 DAA nd hrvest in 2009. d GDD: growing degree dys; PAR: photosyntheticlly ctive rdition. PAR expressed s n ccumultion of dily verge; ws not mesured in My nd June 2008. t seven of eight sites from 30 DAA to 50 DAA by n verge of 63 nd 82%, respectively. Oddly, IBMP decresed from 30 DAA to 50 DAA in 2009 t one vineyrd (site 8) by n verge of 72%. From 50 DAA to hrvest, IBMP concentrtions decresed t ll sites by n verge of 93 nd 72% in 2008 nd 2009, respectively. Although IBMP ws significntly lower t 50 DAA t ll sites in 2009 compred to 2008, t hrvest only sites 5 nd 8 hd significntly lower concentrtions, nd only site 1 hd significntly higher concentrtions. Across sites, IBMP concentrtions in 2008 did not correlte with those in 2009 t ny phenologicl stge (30 DAA, R 2 = 0.09; 50 DAA, R 2 = 0.01; hrvest, R 2 = 0.26). Additionlly, no correltion ws observed within ech yer between 50 DAA nd hrvest IBMP (R 2 < 0.1 for both yers) in contrst to previous report (Ryon et l. 2008). Multisite PLSR models for IBMP t 50 DAA. Sttisticlly significnt PLSR models (p < 0.05) were constructed for IBMP in both 2008 (R 2 = 0.71, R 2 vlidtion = 0.68) nd 2009 (R 2 = 0.45, R 2 vlidtion = 0.37). IBMP concentrtion in berries t 50 DAA ws best predicted by vribles ssocited with vine vigor nd wter vilbility such s shoot dimeter, pruning weight, verge cne weight, shoot length, nd δ 13 C of berries (Tble 5). Notbly, vribles ssocited with vine vigor hd positive correltion coefficients. In 2008, verge temperture from nthesis to 50 DAA hd significnt nd positive regression coefficient, while in 2009, crop lod hd significnt nd positive regression coefficient. Regression plots of predicted IBMP versus observed IBMP in 2008 nd 2009 re shown (Figure 1). While the ll-sites PLS model is pproprite for predicting IBMP within vines for some sites, there re other sites where the ll-sites PLS model poorly explins differences within sites. Using 2008 dt, when the predicted IBMP versus observed IBMP from the ll-sites model is plotted for the 10 vines t ech site individully (dt not shown), we observed significnt correltion t hlf of the sites (sites 1, 2, 3, 7, nd 9; p < 0.05) nd nonsignificnt correltion t the other sites. In 2009, there were no significnt correltions between predicted IBMP versus observed IBMP t individul sites using dt from the ll-sites model.
Modeling Impcts on IBMP 99 Tble 4 Men IBMP concentrtion (picogrms per grm fresh fruit) in Cbernet frnc berries t 30 nd 50 dys fter nthesis (DAA) nd hrvest (HAR) for ech site in 2008 nd 2009. Men nd SD were clculted for the 10 vines smpled from ech site, i.e. represent vine pnel verges. 2008 2009 p vlue (yer) Site 30 DAA SD 50 DAA SD HAR SD 30 DAA SD 50 DAA SD HAR SD 30 DAA 50 DAA HAR 1 61.5cd 19.7 102.5bc 24.7 4.5d 1.0 9.0d 2.1 29.2c 6.1 13.8 6.7 <0.001 <0.001 0.002 2 112.1 20.8 167.6 37.2 7.6bd 3.0 22.4bc 5.5 35.4c 8.2 6.4b 1.7 <0.001 <0.001 0.293 3 49.7cd 31.8 103.2bc 31.8 11.8d 7.9 6.5d 3.0 17.4d 4.4 8.3b 1.8 0.002 <0.001 0.202 4 86.6c 26.4 107.6b 16.3 9.5c 2.9 14.6cd 6.7 32.8c 4.7 7.9b 2.1 <0.001 <0.001 0.097 5 100.4b 19.3 104.3bc 24.2 13.3b 5.3 17.8c 4.0 51.1b 7.6 8.8b 4.3 <0.001 <0.001 0.050 6 49.0d 15.6 107.4bc 23.3 11.5b 3.1 32.7b 8.7 60.1b 5.8 11.2 3.9 0.012 <0.001 0.816 7 101.0b 16.3 191.5 48.1 5.6cd 4.4 33.8 3.1 87.0 5.9 10.2b 6.0 <0.001 <0.001 0.078 8 86.6c 26.4 238.5 48.6 12.1 2.1 41.6 6.1 11.8d 6.4 4.6b 2.7 <0.001 <0.001 <0.001 9 74.8bd 20.0 108.4bc 25.0 1.3e 0.7 10 74.0c 20.9 6.0cd 1.9 p vlue (site) <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 Within column, mens followed by different letter re significntly different (p < 0.05, Gmes-Howell). indictes not mesured. Tble 5 Prtil lest squres regression models: bsic sttistics of the models nd regression coefficients of x vribles for the best model for IBMP concentrtion t 50 DAA (y vrible) in Cbernet frnc berries t ll sites in 2008 (100 vines) nd 2009 (80 vines). Model Vrible 2008 2009 Temp nthesis 50 DAA 0.26 Shoot dim 50 DAA 0.23 δ 13 C berry 50 DAA -0.16-0.62 δ 13 C lef 50 DAA b 0.58 Shoot length 0.24 0.39 Pruning wt 0.08 Averge cne wt 0.11 Crop lod 0.44 NLV 2 4 RMSE 32.0 16.2 R 2 (clibrtion) 0.71 0.46 RMSECV 33.5 17.4 R 2 (vlidtion) 0.68 0.37 p vlue <0.001 <0.001 DAA: dys fter nthesis; NLV: number of ltent vribles; RMSECV: root men squre error of cross-vlidtion. b indictes vrible not in the model. Vrition mong sites ws greter thn vrition within sites (Tble 4), which wekens the bility of our ll-sites model to consider the more subtle within-site vrition. The lodings on the ltent vribles used in the regression models re shown (Tble 6) nd plotted (Supplementl Figure 1). The first ltent vrible in ll of the models ws chrcterized by vribles ssocited with vigor such s shoot dimeter, shoot length, pruning weight, verge cne weight, nd rinfll in July; the second ltent vrible ws chrcterized by δ 13 C of either the leves or berries. Crop lod nd verge berry weight loded to the sme ltent vribles s the previously mentioned vribles ssocited with vine vigor. Single-site PLSR models for IBMP t 50 DAA. The ll-sites model did not lwys effectively model differences within sites. Constructing PLSR models for ech individul Figure 1 Plot of predicted vs. observed IBMP from the (A) 2008 (R 2 = 0.71) nd (B) 2009 (R 2 = 0.45) ll-sites models described in Tble 5. Ech point represents n individul vine, nd different symbols represent different vineyrd sites. Sites 9 nd 10 were not used in 2009. site eliminted vribles ssocited with locl climte nd llowed for n in-depth evlution of vribility in berry IBMP concentrtion within individul sites. In 2008 nd 2009, single-site models were constructed for ll 10 nd 8 sites, respectively. Although the PLSR models were significnt for ll single sites (p < 0.05), the predictive qulity of
100 Scheiner et l. the 2008 models ws poor (R 2 vlidtion rnge = 0.11 to 0.64, verge R 2 vlidtion = 0.356). In contrst, the 2009 singlesite models constructed for IBMP concentrtion t 50 DAA hd higher predictive power thn even the multisite models. The R 2 vlidtion of the 2009 single-site models rnged from 0.29 to 0.93 nd verged 0.66. For ech model, the signs of the regression coefficients of vribles tht contributed to the model were tbulted for ech vrible ctegory: vigor included mesurements of vine growth nd δ 13 C vlues; mturity included ph, TSS, TA, nd relted metrics; nd light exposure included cluster exposure flux vilbility (CEFA) nd other pproprite EPQA metrics (Figure 2). Similr to the multisite models, vribles ssocited with vine vigor were included in 17 of 18 single-site models. At 15 of these sites, there were positive correltions between IBMP nd vigor, versus two with negtive correltions, sttisticlly significnt difference (p < 0.05, chi-squred test). δ 13 C in the berries Tble 6 Prtil lest squres regression models: lodings of the x vribles for the first two components for the IBMP concentrtion t 50 DAA models in 2008 nd 2009. 2008 b 2009 b X vrible LV1 LV2 LV1 LV2 Temp nthesis 50 DAA 0.44 0.35 Shoot dim 50 DAA 0.46 0.05 δ 13 C berry 50 DAA -0.13 0.55 0.30-1.17 δ 13 C lef 50 DAA c 0.72-0.21 Shoot length 0.46 0.23 0.73 0.10 Pruning wt 0.40-0.55 Averge cne wt 0.45-0.48 Rinfll July Berry wt 50 DAA Crop lod -0.19-0.06 Explined vrince (%) 68.9 2.0 27.1 16.2 DAA: dys fter nthesis. b LV: ltent vrible. c indictes vrible not in the model. ws negtively correlted with IBMP ccumultion t nine of these models, versus two with negtive coefficients (p < 0.05). Although CEFA ws not predictor in the multisite models, it ws negtively correlted with IBMP ccumultion in five of the single-site models in 2008, nd no models showed positive correltion between cluster exposure nd IBMP ccumultion (p < 0.05). However, CEFA loded on the sme ltent vribles (dt not shown) s the vribles ssocited with vigor, indicting correltion mong the vribles. In ll 18 single-site models, the first ltent vrible ws chrcterized by vribles ssocited with vine vigor (dt not shown). Multisite PLSR models for log-fold decrese in IBMP, 50 DAA to hrvest. Sttisticlly significnt models could be generted for the log-fold decrese in IBMP from 50 DAA to hrvest, but the predictive power ws wek (Supplementl Tble 4). The R 2 vlidtion of the models ws 0.29 nd 0.10 in 2008 nd 2009, respectively, considerbly weker thn our models of IBMP ccumultion. Shoot length loded hevily on the first ltent vribles of both yers (Supplementl Tble 5) nd ws positively correlted with IBMP decrese (loding, 0.50 in 2008; 0.65 in 2009). Postverison temperture lso loded hevily in 2008 (0.58), nd shoot dimeter t hrvest loded hevily in 2009 (0.51). Single-site PLSR models for log-fold decrese in IBMP, 50 DAA to hrvest. In comprison with the multisite models, the single-site models for log-fold decrese in IBMP concentrtion from 50 DAA to hrvest hd higher predictive power. The rnge in R 2 vlidtion for the single-site models ws 0.75 to 0.95 nd 0.02 to 0.86 in 2008 nd 2009, respectively, nd the verge R 2 vlidtion ws 0.62 nd 0.46 in 2008 nd 2009, respectively. It ws not possible to construct stisfctory model to predict log-fold decrese in IBMP t sites 6 nd 8 (R 2 vlidtion < 0.10). In ll 16 of the stisfctory models, t lest one mesurement ssocited with vine vigor ws included in the model (Figure 3). Specificlly, in 12 of the models, δ 13 C of mture berries nd crop to vine size were Figure 2 Sign of regression coefficients from single-site models for IBMP concentrtion t 50 dys fter nthesis in 2008 nd 2009. Vrible ctegories described in the Results text. * indictes tht the distribution of signs for vrible ctegory ws significntly different from wht ws predicted by chnce (p < 0.05, chi-squred). Figure 3 Sign of regression coefficients from single-site models for logfold decrese of IBMP from 50 dys fter nthesis to hrvest in 2008 nd 2009. Vrible ctegories re described in Results text. * indictes tht the distribution of signs for vrible ctegory ws significntly different from wht ws predicted by chnce (p < 0.05, chi-squred).
Modeling Impcts on IBMP 101 included. However, the reltionship between log-fold decrese in IBMP nd mesurements ssocited with vine vigor, crop to vine size, nd δ 13 C were site dependent nd hd both positive nd negtive regression coefficients. Mesurements of fruit mturity were vribles in six of the models nd in ll cses hd positive regression coefficients. Sensory evlution of wines. IBMP concentrtions in wines rnged from undetectble to 17 pg/ml (2008) nd undetectble to 13 pg/ml (2009) (Supplementl Tble 3). Becuse the men reported vlues for IBMP in berries reported re not weighted bsed on the yield from ech vine (Tble 4), nd becuse yield vried mong sites, it ws possible for wine IBMP to be greter thn the men grpe IBMP cross vines for given pnel. There ws no correltion (R 2 = 0.03 nd 0.02) between IBMP concentrtion nd intensity of herbceous rom of wines in either yer (Figure 4). Significnt differences were observed mong the 2008 wines in the herbceous rom intensity nd in the 2009 herbceous nd fruit rom intensity (Supplementl Tble 6). There ws significnt sensory pnelist x wine interction for intensity of fruit rom in the 2009 wines, indicting some inconsistency in pnelist use of this descriptor. No other interction terms were significnt. PLSR models for rom intensity. PLSR models were constructed to predict the intensity of herbceous rom in the 2008 nd 2009 wines nd intensity of fruit rom in the 2009 wines (Tble 7). The R 2 vlidtion of the 2008 herbceous, 2009 herbceous, nd 2009 fruit models were 0.76, 0.73, nd 0.32, respectively (p < 0.05). It ws not possible to construct stisfctory model to predict the intensity of fruit rom in the 2008 wines. In the 2008 herbceous model, δ 13 C of mture berries ws the most importnt vrible (regression coefficient = 0.76), followed by shoot length (-0.57), Brix*pH 2 (-0.53), nd ph (-0.29). In the 2009 herbceous model, δ 13 C of mture berries nd leves t hrvest hd the highest regression coefficients (1.77 nd -1.44, respectively), nd verge cne weight hd negtive correltion coefficient (-1.02). A plot of δ 13 C versus herbceousness is shown (Figure 5). Discussion Grpes nd wine produced in cool regions nd cool yers re reported to hve higher IBMP concentrtions (Allen et l. 1991, Flcão et l. 2007, Lcey et l. 1991), but we did not observe this phenomenon. In 2008 nd 2009, the verge growing degree ccumultion ( C) from 1 My to 50 DAA cross sites ws 894 nd 799, respectively, nd IBMP concentrtions t 50 DAA were significntly higher (69%) t ll sites in 2008. In concordnce with other studies (Hshizume nd Umed 1996, Lcey et l. 1991), the decrese in IBMP from 50 DAA to hrvest ws lower in the cooler yer (2009), but concentrtions t hrvest were still significntly lower in 2009 thn 2008 t two sites nd not significntly different t five sites s result of the reduced ccumultion. We constructed Tble 7 Prtil lest squres regression models: bsic sttistics of the models nd regression coefficients of x vribles for the best model for intensity of herbceous nd fruity rom (y vrible) for 2008 nd 2009 wines. Vrible 2008 herbceous Model 2009 herbceous 2009 fruit δ 13 C berry HAR 0.76 1.77 0.67 δ 13 C lef HAR b -1.44 Shoot length -0.57 0.16 LLN nthesis 0.66 Periderm -0.80 Avg cne wt -1.02 Brix*pH 2-0.53 ph -0.29 NLV 2 3 2 RMSE 0.10 0.10 0.24 R 2 (clibrtion) 0.92 0.88 0.72 RMSECV 0.32 0.20 0.60 R 2 (vlidtion) 0.76 0.73 0.32 p vlue 0.002 0.026 0.044 HAR: hrvest; LLN: lef lyer number; NLV: number of ltent vribles; RMSECV: root men squre error of cross-vlidtion. b indictes vrible not in the model. Figure 4 Liner regression of IBMP concentrtion nd intensity of herbceous rom for 2008 nd 2009 Cbernet frnc wines. Coefficient of determintion = 0.03 (2008) nd 0.02 (2009). No significnt correltion ws observed (p > 0.05). Figure 5 Correltion between δ 13 C in berries t hrvest nd intensity of herbceous rom for 2008 nd 2009 Cbernet frnc wines. Coefficient of determintion = 0.35 (2008) nd 0.32 (2009). Regression eqution = -0.99x + 31.34 (2008) nd -0.61x + 21.25 (2009). The correltion ws significnt in both yers (p < 0.05).
102 Scheiner et l. PLSR models, combining ll sites nd yers to predict IBMP t 50 DAA nd log-fold decrese in IBMP from 50 DAA to hrvest (dt not shown), nd temperture ws n importnt predictor in both models. Averge temperture from nthesis to 50 DAA positively correlted with IBMP concentrtion t 50 DAA, nd verge temperture from 50 DAA to hrvest positively correlted with the log-fold decrese in IBMP, suggesting tht both pre- nd postverison tempertures re importnt determinnts of finl IBMP concentrtions. Erlier studies cited bove hve focused on temperture during the ripening period nd quntified IBMP fter the onset of degrdtion, but did not ddress the impct of temperture on IBMP ccumultion. There ws significnt vrition within nd mong sites in IBMP concentrtions t both preverison time points nd hrvest, but the site-to-site vrition ws greter (Tble 3). The PLSR models indicted tht IBMP ccumultion ws predicted by vribles ssocited with vine vigor. In generl, vines with higher vigor reflected in greter shoot length nd dimeter, cne weight, nd other ssocited vribles ccumulted more IBMP. The most frequently used vrible in 50 DAA models ws the berry crbon isotope rtio. The crbon isotope rtio (δ 13 C) of grpe leves nd berries correltes with predwn lef wter potentil nd is n indictor of vine wter sttus (de Souz et l. 2003). In response to stress (e.g., drought), the stomtl perture decreses, resulting in depletion of 12 C nd less discrimintion ginst the hevier crbon isotope, 13 C, resulting in n increse in δ 13 C. Berry δ 13 C ws better predictor of IBMP concentrtion t 50 DAA thn lef δ 13 C. Interestingly, lef δ 13 C hd n opposite regression coefficient of berry δ 13 C in severl of the models. Becuse we only smpled sun-exposed leves from nodes 15, 16, nd 17, it is likely tht berry δ 13 C ws better representtion of the whole cnopy over berry development. In ll but two of the models for IBMP t 50 DAA, berry δ 13 C ws negtively correlted with IBMP. Thus, vines with higher discrimintion (i.e., less wter stress) tended to be more vigorous nd hve fruit with higher IBMP concentrtions. However, in the multisite models, δ 13 C loded hevily on different ltent vrible thn the vribles ssocited with vine vigor. This discrepncy could potentilly be explined by other vribles tht influence vigor, including bud number, nutrient vilbility, rootstock, nd crop level. It is not entirely cler why MP ccumultion should be linked to temperture or to wter vilbility. Severl metbolic processes in grpes, such s photosynthesis, re known to increse with incresing temperture to point before decresing (Keller 2010). However, only single enzyme fmily ssocited with MP biosynthesis hs been identified: two O-methyltrnsferses (VvOMT1/2) responsible for methyltion of hydroxypyrzine (HP) intermedites (Dunlevy et l. 2010). To our knowledge, the temperture dependence of VvOMT1/2 ctivity hs not been chrcterized. A recent report hs chrcterized nother V. vinifer OMT responsible for methyltion of flvonoids (FAOMT), which showed four-fold greter ctivity t 37 C thn t 25 C nd mximum ctivity t 50 C (Lücker et l. 2010). While this could potentilly explin greter MP ccumultion t high tempertures, it is possible tht the other unchrcterized steps in hydroxypyrzine synthesis re lso temperture sensitive. Although vribles ssocited with vine vigor explined the mjority of vrince in IBMP concentrtion t 50 DAA cross nd within sites in ech yer, vigor did not pper to correlte well with explining differences in IBMP between yers. In 2009, vines were significntly more vigorous, p < 0.05, (verge cne weight cross sites, 65.7 ± 26 g) thn the previous yer (verge cne weight cross sites, = 46.1 ± 22 g). The most noticeble difference between the two yers ws tht 2008 hd higher preverison tempertures (p < 0.05, Supplementl Tble 2). In field studies, preverison cluster light exposure cn reduce IBMP ccumultion (Mris et l. 1999, Ryon et l. 2008). One exception, reduction in IBMP ccumultion in the bsence of light, hs been reported (Hshizume nd Smut 1999), but this experiment ws performed on hrvested unripe berries, nd extrpolting these results to clusters still on the vine my not be pproprite. Cluster light exposure ws not predictor in ny of the multisite models for IBMP concentrtion t 50 DAA. Becuse cnopy mngement prctices (e.g., lef removl, shoot thinning, shoot positioning, hedging) were imposed by growers t different phenologicl stges nd differing intensities cross sites, it is possible tht we did not cpture the true dynmics of exposure during the preverison period. Even with these cvets, the lck of predictive bility of cluster light exposure in the multisite study is not surprising; the differences in IBMP between fully shded nd fully exposed fruit within vineyrd do not usully exceed fctor of two (Mris et l. 1999, Ryon et l. 2008, Scheiner et l. 2010). However, the vrition in IBMP mong nd within regions, including in our current study, is reported to exceed n order of mgnitude (Allen et l. 1994, Ryon et l. 2008). Furthermore, in five of the single-site models for IBMP t 50 DAA, cluster light exposure, reported s cluster exposure flux vilbility (CEFA), ws negtively correlted with IBMP but loded to the sme ltent vribles s the x vribles ssocited with vigor. Thus, in ll but one instnce where CEFA contributed with negtive loding to single-site model of IBMP ccumultion, it ws coincident with positive loding from vigor metrics. By comprison, in 9 of the 15 single-site models where vigor metrics hd positive lodings, CEFA hd no contribution to the model. As previously observed, in contrst to vigor, CEFA ws not useful in modeling IBMP ccumultion in the multisite models. Tht my indicte tht vine vigor independent of cluster exposure cn hve greter influence thn cluster exposure on IBMP ccumultion, but better controlled experiments would be necessry to be conclusive. In generl, IBMP concentrtions in Cbernet frnc followed the expected pttern over the growing seson of preverison ccumultion followed by postverison degrdtion. However, in 2009, there ws 72% decrese in IBMP from 30 DAA to 50 DAA t one site (site 8), in contrst to severl reports tht IBMP peks in concentrtion ~0 to 14 dys before verison (Roujou de Boubée et l. 2002, Ryon et l. 2008)
Modeling Impcts on IBMP 103 s well s ll other sites in our current studies. Becuse the results were highly nomlous, we confirmed the numbers by rerunning the smples, strting with berries, nother two times. We lso verified tht pek shpe nd qulifier ion rtios were unffected, indicting tht interferences were unlikely. At site 8, the 50 DAA smples were collected ~10 dys before verison, indicting tht IBMP did decrese erlier thn previously reported. The vines t site 8 were highly vigorous nd thinned to less thn one cluster per shoot (verge yield/ pruning weight, 1.18, verge clusters/shoot, 0.6) t ~20 DAA, but we cnnot estblish tht this prctice led to n erly decrese in IBMP. At hrvest, Brix nd TA were 21.6 nd 7.3 g/l, vlues not indictive of dvnced mturity. In contrst to previous report (Ryon et l. 2008), no correltion ws observed between preverison nd hrvest IBMP concentrtions. In this previous work, the study ws performed in single region (Finger Lkes, NY), nd soluble solids ccumultion nd the time elpsed between verison nd hrvest were comprble mong sites. It my be tht preverison IBMP is not useful predictor of hrvest IBMP cross regions which differ in their rte of mturtion. In comprison to IBMP ccumultion, our ttempts to model IBMP decrese cross sites were less successful, s severl x vribles importnt to modeling IBMP decrese within given site did not behve consistently from site to site. For exmple, vribles ssocited with vine vigor were lmost eqully likely to hve negtive or positive correltions in our single-site models (Figure 3). Vine vigor mesurements were tken before verison, thus we cnnot estblish if they were n ccurte representtion of vine growth fter verison. However, ll growers in our study reported using similr prctices postverison: no lef removl ws performed fter verison, but shoot tips were hedged t ll sites. Therefore, prctices should not hve resulted in ny systemtic bis. We did observe tht vines t some sites were still ctively growing t hrvest, nd vine vigor hs been reported to dely fruit mturity (Crbonneu 1997). Thus, the opposing reltionship between these mesurements nd decrese in IBMP from site-to-site my hve resulted from differences in fruit mturtion s ffected by vigor. In ccordnce, nother group (Roujou de Boubée et l. 2000) tenttively correlted decresed rte in IBMP degrdtion in Cbernet Suvignon to lte-seson vine growth induced by high rinfll. It is lso well estblished tht overcropping cn slow fruit mturtion (Jckson nd Lombrd 1993). In 11 of the single-site models, mesurements of crop to vine size (crop lod, yield per shoot, clusters per shoot) were importnt predictors of IBMP decrese, but both positive nd negtive correltions were observed depending on the site. At three of the four sites where mesurements of crop to vine size (crop lod, yield per shoot, clusters per shoot) negtively correlted with log-fold decrese in IBMP, fruit mturity ws positively correlted with log-fold decrese in IBMP, indicting tht vines with higher crop to vine size hd less mture fruit. Severl groups hve reported correltions between IBMP concentrtions nd fruit mturity indices such s Brix nd mlic cid (Chpmn et l. 2004, Hshizume nd Umed 1996, Roujou de Boubée et l. 2002). Although neither IBMP t hrvest nor the IBMP decrese could be predicted by Brix, ph, nd TA cross multiple sites, similr to previous observtions (Ryon et l. 2008), within site the log-fold decrese in IBMP positively correlted with these simple mturity indices in six of the single-site models (Figure 3), significnt result ( p < 0.05, chi-squred test). In the wines vinified from the 5-vine pnels t ech site (20 in 2008 nd 15 in 2009), IBMP concentrtions were t or bove the reported sensory detection threshold, 15 pg/ml (Roujou de Boubée et l. 2000) in only single wine (Figure 4). We hve seen similr results in other recent studies; for exmple, in 2007, >15 pg/ml IBMP in only two of 13 sites in the Finger Lkes (Ryon et l. 2008). The percentge of wines with IBMP exceeding 15 pg/ml is comprble to the percentge reported for wrmer climtes (Allen et l. 1994, Chpmn et l. 2004), even though MP concentrtions re reportedly higher in cool regions. The concentrtion of IBMP in grpes ws correlted with the IBMP concentrtion in wines in the individul vine pnel vinifictions (R 2 = 0.76, p < 0.05, one-wy ANOVA). IBMP concentrtions in some wines were higher thn the grpes from their corresponding pnels (Supplementl Tble 3) becuse the wines represent weighted verges of multiple vines. Tht lso likely resulted in weker correltion coefficient thn is reported elsewhere (Ryon et l. 2009). The intensity of herbceous rom of the Cbernet frnc wines ws not linerly correlted with IBMP (p > 0.05 in both yers) (Figure 4). While correltion between IBMP nd herbceousness hs been reported in wines from New Zelnd nd Austrli (Allen et l. 1991) nd Frnce (Roujou de Boubée et l. 2000), these studies included some wines with >30 pg/ml IBMP, t lest two-fold over its reported sensory threshold of 15 pg/ml (Roujou de Boubée et l. 2000). In our work, IBMP rnged from well below to just bove the sensory detection threshold in red wine, suggesting tht IBMP is not useful proxy for herbceousness in N.Y. Cbernet frnc in the sites under study. Similr observtions hve been mde for Cliforni Cbernet Suvignon with comprble IBMP concentrtions (Preston et l. 2008). Fruity roms cn msk the perception of herbceous roms, nd vice vers (Hein et l. 2009), nd we observed similr inverse correltion between fruity nd vegetl roms in our work (Supplementl Figure 2). Thus, the difference in vegetl rom intensity my relte to n bsence of fruity roms or the presence of other herbceous odornts (e.g., C 6 lcohols), s wines produced from grpes not known to produce significnt levels of IBMP cn lso hve herbceous roms (Guinrd nd Cliff 1987). As cvet, the wines from the two studies were stored under different conditions following bottling (9 months t 12 C for the 2008 vintge vs. 5 months t 21 C for the 2009 vintge), which could ffect the reltive rtes of formtion or loss of other odornts such s esters cross the two yers. The intensity of herbceous rom ws modeled using viticulturl nd climtic mesurements, nd the single most importnt predictor ws the δ 13 C of mture berries. Relted