Reltionship Between Microclimtic Dt, Arom omponent oncentrtions nd Wine Qulity Prmeters in the Prediction of Suvignon blnc Wine Qulity J. Mris, F. litz 2 nd P. D. Hsbroek3 1) AR lnfruiiec-nietvoorbij, Privte Bg X526, 7599 Stellenbosch, South Afric 2) AR Biometry nit, Privte Bg X5() 13, 7599 Stellenbosch, South Afric 3) AR Agromet, Privte Bg X513, 7599 Stellenbosch, South Afric Submitted for publiction: September 2 Accepted for publiction: April 21 Key words: Microclimte, temperture, rdition, Suvignon blnc qulity, prediction model Suvignon blnc grpe chemicl nd wine sensory dt, s well s meteorologicl dt (temperture nd visible light rdition), collected in three climticlly different wine regions in South Afric over three sesons nd from two different cnopy tretments were sttisticlly nlysed. A model for the prediction nd/or definition of Suvignon blnc wine qulity ws developed. The model utilises bove- nd within-cnopy rdition nd cn explin 68.8% of the vrition in the cultivr-typicl vegettive/sprgus/green pepper intensity of Suvignon blnc wine. Other significnt correltions, e.g. between temperture nd monoterpene concentrtions, were lso obtined. Further reserch is necessry to test nd refine this model for ppliction under different environmentl conditions. Over mny yers reserch hs been imed t developing simple model or recipe tht could define nd predict wine qulity. Exmples re sugr/cid rtio ( B/TTA) (Du Plessis & Vn Rooyen, 1982), the glycosyl-glucose (G-G) ssy (Frncis etl., 1998) nd the red wine colour index (Holgte, 2. Such model should lso be useful s method to compenste the grpe producer ccording to grpe qulity. The im further is tht the model should incorporte prmeters tht re esily mesurble under field conditions, insted of prmeters such s rom impct voltiles, the quntifiction of which is dependent on complicted nd expensive gs chromtogrphic nd mss spectrometric nlyses. Due to the complexity of fctors ffecting grpe nd wine composition nd qulity, ttempts to dte hve chieved vrying degrees of success nd foolproof, simple model still hs to he developed. Nevertheless, erlier ttempts were vluble nd ll contribute something to the eventul solution of this problem. Significnt correltions between microclimtic prmeters nd wine sensory chrcteristics were reported erlier (Noble, Elliot- Fisk & Allen, 1995). Models using light rdition in the prediction of photosynthesis under field conditions, were lso developed (Blckburn & Proctor, 1983; Goudrin, 1986; Spitters, 1986; Hsbroek, Myburgh & Hunter, 1997). Photosynthesis is dependent on light rdition nd temperture nd is indirectly n importnt fctor in grpe composition development. The pproch in this study ws to investigte the possible use of microclimtic prmeters s qulity predictor, which cn he 1111.d W. -qig Th. lm, ul enzymtic nd chemicl rections in grpes, in terms of development nd/or degrdtion of components, is dependent minly on temperture nd light rdition. Therefore these two climtic prmeters were used s vribles. nopy density, which is function of different climtic nd viticulturl fctors, lrgely determines the effects of temperture nd light rdition. The purpose of this study ws to correlte microclimtic dt (temperture nd rdition) with rom component concentrtions nd wine qulity prmeters of Suvignon blnc nd to develop model for predicting wine qulity. MATERIALS AND METHODS Grpes, regions nd sesons Vitis vinifer L. cv. Suvignon blnc grpes, grown in three climticlly different regions, nmely Robertson (Region IV, 1945-2222 degree dys), Stellenbosch (Region III, 1668 1944 degree dys ) nd Elgin (Region II, 1389-1667 degree dys) (Le Roux, 1974), were used. A single cnopy mnipultion tretment, which ltered microclimte in nturl wy to obtin higher degree of shding, ws pplied in ech region (Archer & Struss, 1989; Mris, Hunter & Hsbroek, 1999). ontrol vines were not mnipulted. Three replictions, consisting of 15 vines per replic, were used. The generl viticulturl prctices followed were described by Mris ci l. (1999). Grpes were obtined over three sesons, nmely 1997, 1998 nd 1999. Results of the 1997 nd 1998 sesons were published by Mris et l. (1999) nd trils during the 1999 seson were repetition of the first two sesons. Grpe smples (whole hunches, 2kg per smple) were collected weekly t rndom from ech tretment nd replicte over three weeks between pproximtely 16 B (close to vérison) nd 2l B (ripeness). Grpes were hrvested t ripeness (fourth ripening stge) for wine production, following stndrd Nietvoorhij prctices for smll-scle white wine production (Mris ci l., 1999). Mco- nd mieroeljmtjc mesurements MS dt loggers, which continully mesured temperture nd visible light rdition (3-I 2nm) over the whole ripening period on n hourly bsis, were instlled within selected vine cnopies in the vicinity of the clusters, s well s bove the cnopies. Ech rdition sensor consisted of 1 individul sensors, fitted in prllel onto metre-long rod, thus giving more S. Aft. j. Enol. Vitic., Vol. 22, No. 1, 21
A Model for Predicting Suvignon blnc Wine Qulity 23 relible verge reding within the cnopy (Mris et l., 1996; 1999). These mesurements were done in ech experimentl unit. Arom component nlyses Grpe smples from ech tretment, region, seson nd replicte were nlysed for free nd bound monoterpenes nd bound 13 - norisoprenoids by gs chromtogrphy, nd for 2-methoxy-3- isobutylpyrzine (ibmp) by gs chromtogrphy / mss spectrometry (Mris et l., 1996; 1999). Sensory nlysis Wines of ech tretment, region, seson nd replicte were sensorilly evluted for fruitiness nd vegettive/sprgus/green pepper intensities by pnel of six experienced judges. A line method ws used, i.e. evluting the intensity of ech chrcteristic by mking mrk on n unstructured, stright 1 min line (Mris etl., 1999). Sttisticl procedures Seventy-two independent dt sets consisted of three sesons, three loclities, two cnopy tretments nd four ripening stges. Only 7 dt sets were used, due to two missing plots. A 15-vine plot ws considered s n experimentl unit. For ech experimentl unit the microclimtic mesurements, such s the men mximum, men minimum nd verge tempertures during ripening week, were clculted nd served s independent vribles. The verge light rditions bove nd within the cnopies were lso recorded s independent vribles. The grpe nd wine mesurements (verges of three replictes), such s rom voltiles (monoterpenes, norisoprenoids nd ibmp) nd sensory dt (fruitiness intensity nd vegettive/sprgus/green pepper intensity) were recorded s dependent vribles. Person's correltion coefficients were clculted between the bove-mentioned independent nd dependent vribles nd sctter plots were exmined for trends. After exmining the sctter plots, it ws cler tht the rdition vrible showed non-liner trend. Therefore three dditionl terms were dded (i.e. the squres of the bove- nd within-cnopy light rdition nd the interction between them) before the dt were subjected to the stepwise regression procedures. For ech dependent vrible stepwise regression ws performed with specified significnce level of t lest 5% for inclusion in the models, using SAS (Version 6.12 ) sttisticl softwre (SAS, 199. For the bove-mentioned nlyses, the dt were hndled in two wys. Firstly, ll the dt were sttisticlly processed. Secondly, the dt obtined t ripeness (fourth ripening stge) only were sttisticlly processed, since wines were produced from grpes t this stge. RESLTS AND DISSSION The finl models produced by the stepwise regression procedures re given in Tble 1 nd Figures 1 to 3 for the complete dt set, nd in Tble 2 nd Figures 4 to 6 for the dt of hrvest week four only. These coefficients correspond to significnt correltions between dependent nd independent vribles (dt not shown). The figures presented were selected with the min purpose of this study in mind, i.e. to predict wine qulity from esily-mesurble microclimtic prmeters. Although ibmp ws significntly described by the interction of bove- nd within-cnopy rdition, only 43.4% of the vrition could be explined by the model (Tble 1, Fig. 1). For fruitiness intensity 33.3% of the vrition ws explined by mximum temperture nd the bove-cnopy rdition (Tble 1, Fig. 2). The vegettive/sprgus/green pepper intensity (complete dt set) showed the best coefficient of determintion, i.e. R 2 = 68.8% (Tble 1, Fig. 3). The bove-cnopy rdition showed greter effect thn the within-cnopy rdi- TABLE 1 oefficients of independent vribles selected for model nd stndrd errors. Models fitted to ll the dt (N=7 Vrible Intercept T mm. T mx. T ver. A W AxW A2 W2 R2 Monot Noris ibmp 65.4387-2.381 37.2% ±8.75 - ±.3752 - - - si<.1 37.8475 - -.5634 6.8% ±5.9782 - - ±.2537 - - sl=.3 91.8313-2.8195-14.8243.592 43.4% ±17.5359 - - - ±.7455 ±3.936 ±.1616 si<.1 21.955-1.191 2.1351 33.3% Fl - - - - -- ±1.4858 ±.4318 ±.3767 sl<.1 v gp 265.119 15.6791 8.3895. 43.555 68.8% ±29.648 - - ±2.6458 2.9499 - ±.159 ±.2492 sk.1 Monot = Totl free monoterpenes (reltive concentrtion) Noris = Totl bound mterpenes nd 13-norisoprenoids (reltive concentrtion) ibmp = 2-Methoxy-3-isobutylpyrzine concentrtion (ng/l) Fi = Fruitiness intensity (%) Vgp = Vegettive/sprgus/green pepper intensity (%) A = Visible light rdition bove the cnopy (MJ) W = Visible light rdition within the cnopy (MJ) T min. = Minimum temperture ( ) T mx. = Mximum temperture ( ) T ver. = Averge temperture () - = Not selected
24 A Model for Predicting Suvignon blnc Wine Qulity TABLE 2 oefficients of independent vribles selected for model nd stndrd errors. Models fitted to the dt from hrvest week four only (N=I 8) Vrible Intercept T mm. T mx. T ver. A W AxW A2 W2 R2 Monot 66.5851-1.725 - - - - - - 7.5% ±8.3672 ±.2792 sl<.1 Noris - - - - - - - - - - ibmp Fl Vgp 146.672 - - - -11.951 - -.2516-4.8% ±55.455 ±5.1776 ±.118 sl=.2 6.779 - - - 1.5152 - - - - 22.7% ±16.2588 ±.6995 sl=.46 313.451 - - - -22.3616 - -.4499-65.% ±82.2622 ±7.686 ±.175 sl<.i Monot = Totl free monoterpenes (reltive concentrtion) Noris = Totl bound monoterpenes nd 13-norisoprenoids (reltive concentrtion) ibmp = 2-Methoxy-3-isobutylpyrzine concentrtion (ngil) Fi = Fruitiness intensity (%) Vgp Vegettive/sprgus/green pepper intensity (%) A = Visible light rdition bove the cnopy (MJ) W = Visible light rdition within the cnopy (Mi) T mm. Minimum temperture ( ) T mx. = Mximum temperture ( ) T ver. = Averge temperture ( ) - = Not selected L Y=91.2819A.14824W+592A Y = 21.955-1.19T mx. + 2.1351A R 2 = 33.3% SI <.1 I 12 4 flopy rdition (MJ) bove.cnopy rdition (Mi; FIGRE 1 Effect of bove-cnopy (A) nd within-cnopy (W) rdition (between vérison nd ripeness) on 2-methoxy-3-isobutylpyrzine concentrtion in Suvignon blnc grpes. FIGRE 2 Effect of bove-cnopy (A) rdition nd verge mximum temperture (T mx.) (between vérison nd ripeness) on the fruitiness intensity of Suvignon blnc wine. tion. The higher the verge bove-cnopy rdition, the lower the vegettive/sprgus/green pepper intensity of the wine (Fig. 3). This is in greement with dt presented by Mris et l. (1 Althoiieh nturl chndin nlvc n extremel y importnt role in Suvignon blnc qulity development (Mris et l., 1999), models developed seprtely for the shded nd control tretment dt were more or less the sme s for the full dt set nd re therefore not reported. Models rising from the dt of ripening week four (hrvest time) only re presented in Tble 2. As much s 7.5% of the vntion in monoterpene levels ws explined by the verge mximum temperture for the week in which the grpe smples were tken. Monoterpenes my contribute to the flower-like nunces of Suvignon blnc wine, but re probbly less importnt thn the cultivr-typicl methoxypyrzines. The negtive coefficient indicted tht the higher the mximum temperture, the lower the monoterpene concentrtions in the grpes, which is in greement with results reported by Mris et l. (1999). There ws not enough evidence, i.e. no vribles were selected, to fit model on the norisoprenoids in the grpes. 2-Methoxy-3-isobutylpyrzine,
A Model Ji,r Predicting Suvignon blnc Wine Qulity 25 p 1 Y 265.119-15.6791A +.343A2-8.3895W +.555W2 35 3- Y = 146.672-11.951A +.2516A2 = 4.8% SI =.2. L. 91 71 = 68.8% si <.1. 2-1 to-j. 4) > Abovecnopy rdit;o (MJ) c s-i -1 I : Above-cnopy rdition (MJ) FIGRE 3 Effect of bove-cnopy (A) nd within-cnopy (W) rdition (between vérison nd ripeness) on the vegettive/sprgus/ green pepper intensity of Suvignon blnc wine. FIGRE 4 Effect of bove-cnopy (A) rdition t hrvest (week 4) on 2-methoxy-3-isobutylpyrzine concentrtion in Suvignon blnc grpes. 7 6 OR >, 5, 4 I.. 1 8 6 Y = 313.451-22.3616A +.4499A2 65.% Si <.1 u 3 2 1 Y = 6.779 + 1.5152A R 2 = 22.7% Si =.46 (6 L 4 5. 6) 5 ) 2 Above-cnopy rdition (MJ) FIGRE 5 Effect of bove-cnopy (A) rdition t hrvest (week 4) on fruitiness intensity of Suvignon blnc wine. Above-cnopy rdition (MJ) FIGRE 6 Effect of bove-cnopy (A) rdition t hrvest (week 4) on vegettive/sprgus/green pepper intensity of Suvignon blnc wine. fruitiness intensity nd vegettive/sprgus/green pepper intensity showed significnt response to the bove-cnopy rdition only (Figs. 4, 5 nd 6, respectively). Agin the vegettive/sprgus/green pepper intensity vrition ws best explined by qudrtic response of bove-cnopy rdition with 65.%, nd minimum turning point of 24.9% for the bove-cnopy rdition. In spite of the fct tht only 18 dt points were pplied for ripening week four, results obtined for vegettive/sprgus/green pepper intensity (65.%, Tble 2) were close to those of the full dt set (68.8%, Tble 1). ONLSIONS There is significnt reltionship between gs chromtogrphiclly-nlysed grpe rom components, sensoril ly-evluted wine qulity prmeters nd microclimtic dt. Models re presented, one of which cn predict Suvignon blnc wine qulity (cu ltivr-typicl vegettive/sprgus/green pepper intensity) from bove- nd within-cnopy rdition dt, collected over grpe ripening, with firly high degree of ccurcy. It is therefore possible to predict wine qulity from esily-mesurble microclimtic dt. The results of this investigtion were
26 A Model fin- Predicting Sui'igno,i blnc Wine Quit/itt obtined under specific South Africn conditions nd it is not climed tht they re vlid under ll conditions. Nevertheless, they cn be considered s contribution towrds wine qulity prediction. Future reserch is necessry to test nd refine this model for ppliction under different environmentl conditions. LITERATRE ITED Archer, E. & Struss, H.., 1989. Effect of shding on the performnce of Vitis vmtf'er L. cv. bernet Suvignon. S. Aft. J. Eno]. Vitic. 1, 74-76. Blckburn, W.J. & Proctor, J.T.A., 1983. Technicl note. Estimting photosyntheticlly ctive rdition from mesured solr irrdince. Solr Energy 31, 233-234. Du Plessis,.S. & Vn Rooyen, P.., 1982. Grpe mturity nd wine qulity. S. Aft. J. Enol. Vitic. 3.41-45. Frncis, IL., Hnd, RG., ynkr, W.., Kwitkowski, M., Willims, P.J., Armstrong, H., Botting, D.G., Gwel, R. & Ryn.., 1998. Assessing wine qulity with the G-G ssy. In: Blir, Ri., Ss, AN., Hyes, RF. & Hoj, PB., (eds,). Proc. lo Aust. Wine Ind. Tech. onf., 2-5 August 1998, Sydney, Austrli. pp. 4-18. Goudrin, J., 1986. A simple nd fst numericl method for the computtion of dily totls of crop photosynthesis. Agric. For. Meteorol. 38, 249-254. Hsbroek, PD., Myburgh, J. & Hunter, ii., l997. Modelling sunlight interception nd photosynthetic ctivity - correltion with field dt. Poster. 21 Annul SASEV ongress, 27-28 November 1997, pe Town, South Afric. l-iolgtc, A., 2. Berry colour index ssy - ppliction nd interprettion of the ssy s n objective mesure of red wine grpe qulity in commercil vineyrd. Proc. 5th mt. Symp. ool limte Vitic. & Eno]., 16-2 Jnury 2, Melbourne, Austrli. In press. Le Roux, E.G., 1974. ' ii Klimtsindeling vn die Suidwes-Kp]ndse wynhougehiede. M.Sc. Thesis, niversity of Stellenhoseh, Stellenbosch, South Afric, Mris, J., Hunter, J.J. & Hshroek. PD., 1999. Effect of cnopy microclimte, seson nd region on Suvignon blnc grpe composition nd wine qulity. S. Aft. J. Enol. Vitic. 2, 19-3. Mris, J., Hunter, ii.. Hsbroek, P.D. & Augustyn, O.P.I-I., 1996. Effect of cnopy microclimte on Suvignon blnc grpe composition. In: Stockley,..S., Ss, AN., Johnstone, R.S. & Lee, T.H. (eds.). Proc. 9th Aust. Wine Ind. Tech. onf., 16-19 July 1995, Adelide, Austrli. pp. 72-77. Noble, A., Elliot-Fisk, DL., & Allen, MS., 1995. Vegettive flvor nd methoxypyrzines in bernet Suvignon. In: Roussel', R.L. & Lehy, M.M. (eds.). Fruit flvors, Biogenesis, hrcteriztion nd Authentiction. AS Symposium. Series 596, 22-27 August 1993, hicgo. Illinois. SA. pp. 226-234. SAS. 199. SAS/Stt sers Guide. Version, 4 Edition, Vol. 2, SAS Institute Inc.. SAS mpus Drive, ry, N 27513. Spitters,.J.T., 1986. Seprting the diffuse nd direct component of globl rdition nd Its implictions for modelling cnopy photosynthesis. Prt II. lcultion of cnopy photosynthesis. Agric. For. Mcteorol. 38, 23 1-242. S. Mr. J. Enol. Vitic., Vol. 22, No. I, 21