‘5: v. . .. .“ .oOWo.‘Oolo«Iu9AXu vv‘fi..fitvla I‘nlov‘ a $33.? .. ll. . I uuvduu...‘ .u. I . . ..‘wu.u.u.. .wawws.~“-.vdfln. '1 .‘ I n ‘ JO It: toc“:i. €Q$ - nucsvmii..l llVOv r0- I~V 10w own—r a"! w J. tL 0‘... v in ‘19.- '0‘ ‘vt .1 0C a '0 t}: n . ‘0 wnloL 0 ~‘. 51.0...“ (v x =1 ai V.“ 1 .v . v: 1L.II:I livu 1Ivtr\.t‘l 0.0%.... .' §¢§iviv ‘04! o1|.a1b§t¢$!1||.>‘b \Ic . Ittfiunl 0 ll\ 0 ‘I4 fins! 0"! ) ‘1’ fix. ' d‘vo. . '3'. ct .. .nuoha u‘v¢ 0(1va 1‘ :1 .1130 v“ \I v.0. '1 1?: In '0‘. '4-"'cmxmem mg» mcwm: mooe mmcm “a smug-aupou co_qu:E:uuu acu-owcmwo 2.5.. uh.._. N 3000 A9010N3Hd on 25 Reproductive Growth (Refer t Table ll Figures 4,5) Previous Stage: Green Tip Stage: Pre-Cluster Leaf Not all reproductive buds express the 'pre-cluster leaf' stage. Both inter and intra-varietal variation occurs. This variability in development occurs as the buds grow thru the late 'green tip' stage into the 'tight cluster' stage. When the 'pre-cluster leaf' stage is present, one or two small leaves subtending the cluster of flower buds will unfold before the cluster completely outgrows the bud (Figure 46). If the 'pre-cluster leaf' stage is not present, then the cluster of flower buds starts to become visible, but remains sheathed by the subtending leaves, which do not unfurl. Three different variations for development exist: 1) within year - same variety, 2) between years - same variety, and 3) between varieties. 'Golden Delicious' seems prone to both within year and between year variation. 1) In 1983, almost all flower buds passed directly from the 'green tip' stage into the 'early tight cluster' stage without clearly exhibiting the 'pre-cluster leaf' stage. Some flower buds did display the 'pre-cluster leaf' stage. It is probable that this represents variances in growth rates, with those clusters showing pre-cluster leaves having slower rates. This could be attributed to bud to bud 26 variation in chilling hour accumulation (8,13,21,26,35) or carbohydrate reserve status (1). Landsberg discussed a study conducted by Abbott showing that when winter chilling was delayed buds continued to grow (swell in size) but did not break until after chilling, opening directly to the 'green cluster' stage (23). 2) In 1984, flower buds exhibited the 'pre-cluster leaf' stage with greater frequency. 3) In contrast to 'Golden Delicious', varieties such as 'Mutsu', 'McIntosh', and 'Northern Spy' all tend to exhibit the 'pre-cluster leaf' phase in all years. Adjustments to the phenocode were made based on the variations in development observed in the latter green tip stages through the early tight cluster stage (eg. both 1.5+cm green tip and pre-cluster leaf = 10). Stage: Early Tight Cluster The ‘early tight cluster' stage is reached when the cluster of flower buds is clearly visible, but the number of individual buds is not yet descernible. ,The leaves subtending the cluster will have begun to unfurl by this time. (Figure 4H). Stage: Tight Cluster The stage 'tight cluster' is reached when the cluster leaves begin to unfold and the individual buds start to become recognizable. The buds are still tightly appressed (Figure 41). 27 Stage: Early Bud Expansion As the flower cluster continues to grow, the individual flower buds begin to swell and separate. The 'early bud expansion' stage is reached when it is clear that one or two buds are beginning to show swelling and separation (Figure 4J). A number of cluster leaves generally will have unfolded and begun expansion during this phase. Stage: Bud Expansion As the flower buds swell and separate, the pedicels also begin to grow up and out. The flower buds are all clearly decernible at this stage, the pedicels are extended, and many cluster leaves are now out. The flower buds have not yet cracked to show pink (except possibly the king bloom in the latter occurence of this stage ). Stage: Early Pink 1 Full Pink After ' the clustered flower buds have separated (although some clusters may occasionally remain tightly bunched) they continue to swell, revealing petal tissue when the buds crack open at their apex. The 'pink' stage is subclassed into three categories, ranked according to the amount of pink petal tissue visible. The subclasses are 'Early Pink' (Figure 4L), 'Pink'(Figure 4M), and 'Full Pink' (Figure 4N). Flower buds are considered to be in the 'early pink' stage when only a few buds are initially showing pink, 28 or all buds are showing very little pink. A cluster with a full compliment of flowers has up to 5 - 6 flowers / cluster. Normally, the king bloom (center blossom) is more advanced than the other blossoms in the cluster, and frequently 1 or 2 buds lag behind in development. This variation makes averaging of the stages of all the buds within a cluster and subjective judgement important in determining bud/bloom class in the flowering stages. It is during the 'pink' stages that the variation of growth among flower buds within a cluster first becomes apparent. As the flower buds expand, more pink petal tissue is exposed and the buds are classified into the 'pink' and 'full pink' stages, respectively. Stage: Blossom Classes Following the 'pink' stages the flower buds continue to expand until they fully open. They open in order of relative maturity within the cluster, with the king bloom usually opening first. There are three bloom classes, with each ranking dependent upon the number of blossoms which have opened. A blossom is classed as being open if all of the petals have unfurled. The bloom classes are '1 blossom'(Figure 40); '2-3 blossoms'(Figure 5A); '4+ blossoms'(Figure 58). Stage: Full Bloom 'Full bloom' is a frequently used development marker. 29 It is really an averaged classification that occurs when 70 - 80% ‘of the blossoms on a tree are open (F.G. Dennis, personal communication). Hence, there can be some overlap with the '4+ blossom' stage and the 'early petal fall' class. Once apple blossoms open they may remain on the tree for a number of days before dropping their petals. When clusters are first observed to have 4 or more blossoms open then the '4+ blossom' stage is the most appropriate. If all the flowers in a cluster are open or the blossoms appear aged (the pollen darkens from a bright yellow to dark yellow/tan) the use of the 'full bloom' class would be more appropriate than the '4+ blossom' class. It is not unusual to observe the king blossom beginning to lose petals while the other four blossoms remain intact. The 'full bloom' class is most useful when referring to averaged values (see monitoring segment) (eg. if average observations indicate 70 - 80% of the flowers open, use the 'full bloom class'.) Stage: Petal Fall After bloom, if the flower is fertilized, fruit may be set and begin to grow. If fertilization does not occur (or is incomplete) the flower will abcise. As these processes are initiated, the petals will begin to drop. The 'petal fall' classes are grouped by the number of petals lost. To classify a cluster, the average number of petals lost/cluster is used: 'Early petal fall' = <25% petal fall (Figure 50); 'Mid petal fall' = 25 - 65% petal fall (Figure 30 5E); 'Late petal fall' = i<65 - 100% petal fall (Figure 5F). This particular stage is directly affected by wind and rain, which can physically knock petals off blossoms. ‘Stage: Fruit Set Determination of fruit set is very subjective, based on visual expansion of the recepticular tissue of the apple flower. Immediately following 'late petal fall', little, if any swelling of the fruit is visible (Figure 56). As fruit growth begins, a slight rounding is evident (Figure 5H). During 'early fruit set' 1 fruit will show swelling (usually the king blossom ); 'mid fruit set', 2-3 fruitlets swollen; 'late fruit set', 4+ fruitlets swollen'. The fruit set classification is based on 4-5 fruitlets/cluster. With fewer fruitlets use an approximate percentage (eg. 2-3 fruit set = 40 -60 % set.) To be classed as set the fruit must appear to have initiated swelling (Figure 5H). Stage: Fruit Growth During growth fruits are classed by diameter (0.5 cm incremental increases in diameter) (Figure SF-SM). Fruit diameter measurements are made at the largest width of the fruit, where length is the stem/calyx axis, and width is the axis perpendicular to the stem/calyx axis (Figure 5N). 31 Figure 2: Comparative distribution of phenophase (stages of development) as a function of degree-day accumulation (base 40) and date, for reproductive spur growth in apple: "early silver“ thru early "fruit growth" stages:(cv) Red Delicious (HRC, 1983). 32 .mmmH .um: nouwm neococo .fi P_La< u=o_uc~=s=oue autummcmmu Low mace ugmum .Avv vogume :_Em can m_—_>cmxmcm one m:_m= meow wmam um umuopsupcu :owumpze=uom xou-mmcmmo N Uh‘fl wzas ><2 ><2 ...:t< a _ a m a «a. ... ...... ... ... E... :13... ... ...wL.” m ON .fl>cmxmom .H _mga< “:o_pu~:s:ooo amcnmmcmmu gem aunt “Loam mzu m:_m: meow omen an uwuw_=upou :o_unP:E:uuc xou-mmcmmo N .8 .3553 .25.; rang—k0 2.2. :3 ...:< e— ” 4 e a a a a a a a a a a a a a fi _ ~m>_h<._.u0u> I: ....t Lina I 15.53. 2:35 ...... ...» ...-s .....w l :in "...-.5 ...... "E3020 m>=03825m¢ 37 Figure 4. A.- 0. Photographs documenting morphological stages of development for reproductive phenophases (dormant-early bloom) 39 Figure 5. A.- M. Photographs documenting morphological stages of development for reproductive phenophases (bloom - harvest) N. Measuring fruit growth 0. Bourse shoot 41 Figure 6. A.- L. Photographs documenting morphological stages of development for vegetative phenophases (silver tip - terminal bud set) M. Watersprouts N. Suckers 0. Leaf emergence 43 Figure 7. A.- H. Photographs documenting morphological stages of development of vegetative regrowth 45 References Cited 1. Abbott, D. C. 1977. Fruit bud formation in 'Cox's Orange Pippin'. Rep. Long Ashton Sta. for 1976: 167-176. 2. Anstey, T.H. 1965. Prediction of full bloom date for Apple, Pear, Cherry, Peach, and Apricot from air temperature data. J. Amer. Soc. Hort. Sci. 88: 57- . 3. Ballard, J.K., E.L. Proebsting, and R.B. Tukey. 1982. Apples. Wash. State Univ. Coop. Ext. Bull. 913. 4. Baskerville, G.L. and P. Emin. 1969. Rapid estimation of heat accumulation from maximum and minimum temperatures. Ecology 50:514-517. 5. Cain, J.C. 1973. Foliage Canopy development in 'McIntosh' apple hedgerows in relation to mechanical pruning, the interception of solar radiation, and fruiting. J. Amer. Soc. Hort. Sci. 98: 357-360. 6. Chapman, P.J. 1966. Standard names for key apple bud 149. stages. Proc. New York State Hort. Soc. 111: 146- 7. Chapman, P.J. and G.A. Catlin. 1976. Growth stages in fruit trees - from dormant to fruit set. New York's Food and Life Sci. Bull. No. 58. 8. Couvillion, G.A. and A. Erez. 1985. Influence of prolonged exposure to chilling temperatures on bud break and heat requirement for bloom of several fruit species. J. Amer. Soc. Hort. Sci. 110(1): 47-50. 9. Eisensmith, S.P., A.L. Jones, and J.A. Flore. 1980. Predicting leaf emergence of 'Montmorency' sour cherry from degree-day accumulations. J. Amer. Soc. Hort. Sci. 105(1): 75‘780 10. Eisensmith, S.P., A.L. Jones, E.D. Goodman, and J.A. Flore. 1982. Predicting leaf expansion of 'Montmorency' sour cherry from degree-day accumulations. J. Amer. Soc. Hort. Sci. 107(5): 717-722. 11. Edson, C.E. and J.A. Flore. 1986. 'PHENOTE': A prediction program for apple. (In preparation). 12. Edson, C.E. and J.A. Flore. 1986. Modeling shoot development, flowering, and fruit growth in apple. (In preparation). 46 13. Gilreath, P.R. and 0.W. Buchanan. 1981. Rest prediction model for low-chilling 'Sungold' nectarine. J. Amer. Soc. Hort. Sci. 106(4): 426-429. 14. Harding, P.H., J. Cochrane, and L.P. Smith. 1976. Forecasting the flowering stage of apple varieties in Kent, England, by the use of meterological data. Agricultural Meteorology 17: 49-54. 15. Haun, J.R. 1973. Determination of wheat growth- environment relationships. Agron. J. 65: 813-816. 16. Haun, J.R. 1973. Visual quantification of wheat development. Agron. J. 65: 116-120. 17. Haun, J.R. 1982. Early prediction of corn yields from daily weather data and single predetermined seasonal constants. Agricultural Meteorology 27: 191-207. 18. Haun, J.R. and 0.0. Costen. 1983. Relationship of daily growth and development of peach leaves and fruit to environmental factors. J. Amer. Soc. Hort. Sci. 108(4): 666-671. 19. Higgins, J.J. 1952. Instructions for making phenological observations of garden peas. John Hopkins Univ. Lab. of Climatol. 5: 1-11. 20. Higgins, J.J., J.R. Haun, and E.J. Koch. 1964. Leaf development: Index of plant response to environmental factors. Agron. J. 56: 489-492. 21. Jonkers, H. 1979. Bud dormancy of apple and pear in relation to the temperature during the growth period. Scientifia Horticulturae. 10: 149-154. 22. Kronenberg, H. G. 1983. Relationships between temperatures and blooming dates of apple trees. Neth. J. 23. Landsberg, J. J. 1974. Apple fruit bud development and rowth: Analysis and an empirical model. Ann. Bot. 38: 013-1024. 24. Landsberg, J. J. 1975. The mechanisms of apple bud morhogenesis: Analysis and a model. Ann. Bot. 39: 689- 699. 47 25. Landsberg, J. J. 1975 Effects of weather on plant development: Apple bud morphogenesis. In (J. J. Landsberg and C. V. Cutting (eds.) "Environmental effects on crop physiology“. Academic Press. London. 26. Landsberg, J. J. 1977. Studies on the effect of weather on the growth and production cycle of apple trees. J. Roy. Agric Soc. Eng. 138: 116-133. 27. Lewis, A. J. and J. R. Haun. 1971. Detection and evaluation of plant growth responses to environmental conditions. Amer. J. Bot. 58(5): 394-400. 28. Michigan State University Spray Calendar. 1985. M. S. U. Cooperative Extension Publication. Bull E454. 29. Phenology and plant species adaptation to climates of the western U. S. 1978. Oregon State Univ. Ag. Expt. Sta. Bull. 632. 30. Richardson, E. A., et al. 1973. A model can help save Utah's fruit. Utah Science. Dec: 111-112. 31. Seem, R. C. and M. Szkolnik. 1978. Phenological develgpment of apple trees. Vermont Agri. Exp. Station Bull. 684. 6-20. 32. Seem, R. C. and V. T. Cullinan. 1982. A phenology model for aggle tree development. MS. Thesis. Cornell University, N.Y. pp. 33. Shaltout, A. D. and C. R. Unrath. 1983. Rest completion prediction model for 'Starkrimson Delicious' apples. J. Amer. Soc. Hort. Sci. 108(6): 957-961. 34. Schwartz, H. J. and L. E. Powell, Jr. 1981. The effect of long chilling requirement on time of bud break in apple. Acta Horticultuae. 120: 173-178. 35. Thompson, W. K.,D. L. Jones, and D. G. Nichols. 1973. Effects of dormancy factors on the growth of vegetative buds of young apple trees. Part II. Aust. J. Agric. Res. 24: 813- 820. 36. Tukey, H. B. 1942. Time interval between full bloom and fruit maturity for several varieties of apples, pears, peaches, and cherries. Proc. Amer. Soc. Hort. Sci. 40: 133- 140. 48 Section 11 Modeling Shoot Growth, Flower Development and Fruit Growth in Apple. 49 Abstract: An empirical predictive model of apple tree growth was developed for Red Delicious, Golden Delicious, and spur type Golden Delicious. Observations of morphological development (Phenophase) and selected environmental parameters were made at 1 to 7 day intervals. A model that would predict apple shoot growth by phenophase was desired. For monitoring and predictive purposes, shoot growth was partitioned by type as follows: vegetative spur, reproductive spur, vegetative terminal, reproductive terminal. Linear regression analysis using several environmental parameters as independent variables was performed. Non-linear equations using degree-day accumulation as the independent variable were also fitted. The best statistics of fit and predictions were generated by the asymptotic non-linear Logistic equation: Y=B(1)/(1.+ EXP(B(2)-B(3)*X+B(4)*X2-B(5)*X3)) for all shoot types and varieties. Predictions of stage within 0 to 2 phenophases up to 4 months in advance are possible. The concept of Biofix, and a method of initializing the model for use in different locations and years, using Biofix values, is discussed. 50 Introduction The Role g: the Model Every orchardist strives to produce quality fruit at the lowest cost. Timely cultural management and effective pest control cannot be overstated. One of the major and increasingly expensive production costs is the application of chemical sprays for pest control or growth management. Concerns other than cost for pesticide use also exist. There is an increased awareness of the environmental hazards associated with pesticide use, and the efficacy of newly developed pesticides has been reduced as target insects and diseases develop resistance to these chemicals. A Michigan program for Integrated Pest Management (IPM) , largely designed and administered by scientists at Michigan State University, addresses some of the problems associated with pesticide use by providing information designed to improve the timing of pesticide applications with the goal of reducing the amount of pesticide required. This has been accomplished by correlating weather data - with pest phenological development (in the form of field population counts) to generate predictions based on pest life cycle models (26, 69). The ecology of the apple production system is one of complex interactions (13,14,21). Subsystems (eg. insect life cycles, disease epidemiology and physiological mechanisms of growth) have been studied, but to date a field 51 applicable, holistic model of the apple production system has not been developed (21). Insect and disease development, cultural practices, and the environment are tied closely to the growth cycle of the apple tree (7,13,49,69). Croft et. al. (13) states "the trend in tree fruit pest control research is toward increased study of pest-plant interaction..." . A greater understanding of the growth and development of the tree and how it responds to its environment should contribute to more efficient insect and disease control by improved timing of management strategies. Use of tree phenology to develop a model to predict shoot growth would contribute to an understanding of plant development and could later be coupled with insect and disease development predictions resulting in more efficient orchard management. The Utility 9: the Model Many growers still rely on published spray schedules where timing of pesticide application is based on the occurrence of a particular stage of growth (eg. pink, bloom)(49). Timing of apple thinning is also tied to developmental stage (days after bloom or size). Many growers thin their fruit with chemical sprays that require precise timing (if the fruit are too large thinning will not occur). Timing of other growth regulators (Alar, NAA, Ethyphon) has been based on some particular stage of development. The ability to accurately predict the 52 developmental growth stages of the apple tree would make decisions concerning pesticides and growth regulators more efficient. Between 1982 and 1985 an apple phenology study was conducted at Michigan State University. The primary goal of this study was to provide updated and precise growth information necessary to model the vegetative and reproductive growth of the apple pest-tree-environmental system. A phenology model capable of predicting key stages of growth (eg. green tip, pink, full bloom, etc.) was developed. Based on predictive ability the main uses of this model would be: 1) improved timing and more efficient scheduling of pesticide and growth regulator applications, 2) to function as a structural base for more complex models that might later include the physiological components of growth, or interface with models of insect life cycles, disease epidemiology, orchard floor management (weed control), etc. Types g: Models Several apple-crop system models have been designed (13, 21, 45, 58). "A systems model may be made up of a series of either theoretical or empirical equations“ (64). Theoretical models describe the system in terms of the modeler's perception (theory) of how the system functions (64,66,71). Theoretical models help identify interactions, potential parameters for model expansion, and subsystems for 53 empirical modeling (eg. organ systems, physiological mechanisms, etc.) (71) (see Russo and Seem (58), and Evans (23) for a review of the hierarchy of plant subsystems). Empirical models are based on factual data, and have the ability to 'summarize' data (as mathematical equations) (36), and to assess (rather than theorize) behavior under a variety of conditions (71). In reality, most models contain elements from both the theoretical and empirical approaches (64). Mechanistic models view the response of the system in terms of the structure (underlying mechanisms) of the system (66). Both theoretical and empirical models can be mechanistic in nature, and while the empiricist describes the structure in terms of observable output, it is the theorist who is more likly to describe the underlying mechanisms of the system (46,64,66). Mechanistic models have been developed for system components of some fruit crops: apple physiology and growth (42,45), photosynthesis and stomatal conductance in apple (41,42,46), and carbohydrate partitioning in cherry (37). An empirical phenology model was designed to function as a basic structural component of an apple-crop system model. The model was not designed to be a mechanistic model (does not describe the mechanisms of growth), but was designed to describe shoot growth based on observations of the morphological development. Theoretical component models (eg. physiological) could be used to help refine and expand 54 the utility of the empirical model. Model Variables Environmental variables have been used to predict growth (19,30,45,62,68). Physiological events in apple have been related to various environmental parameters (41, 42,43,44,45,46), but are difficut to quantify non- destructively. Haun and others have used non-destructive methods to quantify growth based on systems of measuring changes in growth correlated to changes in the environment (31,32,35,45,47,59). Each morphological stage of development (termed 'phenophase' by Seem & Szkolnik (59)) is assigned a numerical indexing code. The relationship between growth and the environment can be determined by regressing the numerical indexing code (dependent variable) with the environmental parameters (17, 19,20,32,33,60). Cumulative or daily growth relationships can be determined by regression analysis. Several environmental parameters have been used as independent variables. Lombard, et al. (48) concluded that air temperature was the major environmental parameter related to the phenological development of fruit trees. Degree day accumulation (or degree hour accumulation) has been correlated with growth (33,55,56,61,68). Eisensmith, et al.(19,20) developed leaf emergence (leaf number) and expansion (leaf area) predictions by regressing growth with degree-days accumulated in sour cherry Prunus cerasus (cv. 'Montmorency'). 55 Environmental parameters should be related to growth based on: 1) the biological significance of the independent variable being modeled; 2) increased mathematical correlation between dependent and independent variables (16, 29,36); 3) availability of environmental data ; and 4) considerations of model complexity (will the model be difficult to use?). The Functional Approach The functional approach is used to develop equations predicting developmental stages of growth (9, 10, 36). Mathematical functions are used to describe (or fit to) the data (frequently using regression analysis). The functional approach is especially useful to the empiricist since instantaneous values and errors are derived from equations fitted to data (eg. stage = f(time)) (36). Fitting mathematical functions tends to smooth out slight irregularities in the data, helping to show the overall trend (10, 36, 66). With the aid of the computer a number of mathematical functions ranging from simple linear models with one parameter to complex models estimating multiple parameters can be fitted to the data set. Types of functions available, and the rational behind the use of these functions are discussed in detail by Hunt (36), Causton (9), Causton and Venus (10), Gold (29), Thornley (66) and Spain (63). Statistical estimation of errors and estimation of 56 derivatives becomes difficult when using complex functions (36). Complex functions may closely describe the idiosyncrasies of the data fitted, but fail to show the overall trend. (36). Occasionally, complex functions will produce a curve that tends to waver about the data points, yielding high correlation statistics but never truly 'fitting' the data (18). The assignment of biological significance to model parameters is difficult, if not impossible with complex functions(36,54). Erickson (22) and Richards (54) discuss the importance of matching model parameters to biological events. Selection of the simplest model that will achieve the goals of the modeler is preferred (9,36,38). A more complex function that predicts well might be selected if accuracy of prediction was more important than relating the biology of the system to function parameters. Concept 9: Biofix The determination of the onset of growth is a major problem in modeling perennial plant growth. Apple enters a state of rest each fall and must be exposed to approximately 1200 hours of chilling temperatures ( 1 9.50 C to 13° C) before growth can resume (total hours varies with author and method of calculation)(42,61). Temperatures below 1.6° C are thought to contribute little to chilling hour accumulation and temperatures above 13° C to 19° C can negate chilling hours previously accumulated, with optimum 57 chilling occurring at 6.10 c to 7.20 c (42,61). Growth will resume in the spring only when rest has been completed (sufficient chilling hours have accumulated) and favorable temperatures for growth occur. Two approaches have been used. One method predicts the end of rest, based on accumulated chilling hours (3, 11,55,57,61). Models predicting spring bud development through full bloom have been developed for apple which use chilling unit accumulation to determine the onset of spring growth (Utah (42) and North Carolina (61)). The Utah model has not resulted in accurate predictions in regions outside of Utah (2,52,61), and neither model is accurate in Michigan (J. A. Flore, personal communication). Two major problems are encountered when making predictions based on chill unit accumulation. 1) It is difficult to accurately determine the onset of dormancy (one of the chilling model requirements). 2) The growth response of apples varies based on the temperature regime at which chilling hours are accumulated (11.24.65). There is disagreement among authors as to the optimum chilling temperature (41,55,65). Additionally, in Michigan accurate hourly weather data, nor data before April 1 are available using the PMEX system (50). A second approach eliminates the necessity of determining the occurrence of a physiological event (eg. onset of growth or dormancy). A verifiable biological event (biological reference marker or Biofix) is chosen 58 arbitrarily to represent the onset of growth. Biofix markers have been used in models predicting insect development (7,26). Seem and Szkolnik (59) have adapted this concept, using green tip as a reference marker from which to calculate degree-day accumulation. Eisensmith et al. (19,20) have successfully used degree-days accumulated to a selected date (April 15) as the Biofix point in their predictive models of cherry leaf emergence and expansion Degree-Day Accumulation: Base Temperature Different base temperatures have been used to calculate degree-day accumulation for apple (21,48,59,60,61). Richardson, et. al. (55) concluded that any temperature greater than 4.4°C would result in growth. Kronenberg (40) gives a good review, reporting base temperature values of 6.10 C and 6.6°C if day maximum temperatures are used , concluding that using average day temperatures gives better results than maximum day temperatures, and base temperatures may vary with variety (base temperature = 0°C to 8°C). Different phases of growth may respond to different threshold temperatures (40,42,67,68). Early spring bud growth may respond to a lower threshold temperature (lower base temperature) than fruit growth during the summer months (42,67,68). Since this has not been well documented, a single base temperature was utilized in this study. 59 Materials and Methods Plant Material Phenological observations were made at three different orchard sites at the Michigan State Horticultural Research Center (HRC) at East Lansing, Michigan, between 1982 to 1985. Trees in each orchard were spur type Golden Delicious/MM 111, Golden Delicious/seedling, and Red Delicious/seedling 9,16, and 16 years old in 1982, respectively. Five trees/orchard were selected and four observations/tree/shoot type were recorded on each sampling date. Growth was separated into four categories: vegetative or reproductive on spurs or terminal extension shoots. Observations were made on Golden Delicious/MMIII, Red Delicious/MMIII, Ida Red/MM111, and Jonathon/MMIII at the Graham Research Station in Grand Rapids, Michigan in 1983, and on Golden Delicious/Std, Golden Delicious/EM7, Golden Delicious/MM106, Empire/EM7, Red Delicious/EM7, and Red Delicious/MM106 at two commercial orchard sites in Leelanau County, Michigan in 1985. Data collected at the H.R.C. in 1983 was used to generate the predictive models, while data from other observation sites and other years (HRC) were used to test prediction equations. Observational procedure A numerical code (phenocode) was developed based on precisely defined morphological stages of growth (16) in 60 which each developmental stage was assigned a numerical value (Table 1,2). Each stage is mutually exclusive, with development through consecutive stages. The mean phenocode was used to calculate a growth indice (GI) for that particular sampling date, where: ZXJ GItJ- t = time in days X = phenocode at time tj for j = 1...k N = number observations of a particular type of shoot at time tjfor j = 1...k This generated a series of values: GItl, GIt2,...GItk, which could be used as the dependent variable for correlation to selected environmental (independent) variables. Rate observations The differences between vegetative and reproductive development during the early stages of growth: dormant through silver tip are not readily apparent (1, 16). Errors in bud classification may occur during early growth. To decrease bud classification error, daily observations were 61 Table 1: Reproductive Phenocodes: Used to numerically code morphological development of reproductive shoot growth in apple. STAGE NAME CODE Dormant 1 Early Silver Tip 3 Silver Tip 5 Green Tip: 0 - 0.4 cm 7 : 0.5 - 0.9 cm 8 1.0 - 1.4 cm 9 1.5+ cm 10 Pre-cluster Leaf 10 Early Tight Cluster 11 Tight Cluster 12 Early Bud Expansion 13 Bud Expansion 14 Early Pink 15 Pink 16 Full Pink 17 1 Blossom 18 2-3 Blossoms 19 4+ Blossoms 19.5 Full Bloom 20.5 Early Petal Fall 22 Mid Petal Fall 23 Late Petal Fall 24 Early Fruit Set 25 Mid (2-3) Fruit Set 26 Late (4+) Fruit Set 27 Fruit Diameter (cm): <0.5 cm 28 0.5 - 0.9 cm 29 1.0 - 1.4 cm 30 1.5 - 1.9 cm 31 2.0 - 2.4 cm 32 2.5 - 2.9 cm 33 3.0 - 3.4 cm 34 3.5 - 3.9 cm 35 4.0 - 4.4 cm 36 4.5 - 4.9 cm 37 5.0 - 5.4 cm 38 5.5 - 5.9 cm 39 6.0 - 6.4 cm 40 6.5 - 6.9 cm 41 7.0 - 7.4 cm 42 7.5 - 7.9 cm 43 8.0 - 8.4 cm 44 62 Table 2: Vegetative Phenocodes: Used to numerically code morphological development of vegetative shoot growth in apple. STAGE NAME CODE Dormant 1 Early Silver Tip 3 Silver Tip 5 Green Tip: 0 - 0.4 cm 7 : 0.5 - 0.9 cm 8 1.0 - 1.4 cm 9 1.5+ cm 10 Shoot Growth: 1 Leaf 11 2 Leaves 12 3 Leaves 13 4 Leaves 14 5 Leaves 15 6 Leaves 16 7 Leaves 17 8 Leaves 18 9 Leaves 19 10 Leaves 20 11 Leaves 21 12 Leaves 22 13 Leaves 23 14 Leaves 24 15+ Leaves CODE = leaf number + 10 Terminal Bud Set: Dwarfing Rootstocks Spur Types: Spur 25 Terminal 30 Standard Rootstocks: Spur 30 Terminal 32 63 made on 40 spur and 40 terminal buds (tagged), and the stage of growth recorded for each bud. As growth progressed each bud was classified as being either vegetative or reproductive. Mean phenocodes for each sampling date and shoot type were calculated. A weighted growth indice (WGI) was calculated as follows: where ti = t1 < t2 < tk for j = 1...k 80’ + X) t = time in days P = phenocode (from tagged shoots) at time tj for j = 1...k PM = number of observations of tagged shoots/shoot type at time tj for j = 1...k X = phenocode (from random shoots) at time tj for j = 1...k N = number of random observations of a particular shoot type at time tj for j = 1...k WGI = weighted growth indice. Observations were continued throughout the growing season, and the weighted growth index was used as the dependent variable from dormant through tight cluster. Beyond tight cluster there was virtually no difference 64 between the unweighted growth indice and the weighted indice, so only the unweighted indice was used. Weather Variables Daily maximum and minimum temperatures, rainfall, pan evaporation, humidity, soil temperature, and wind speed were recorded beginning April 1 for each of the monitored orchard sites and compiled on the M.S.U. PMEX network. Degree-day accumulation was calculated using the Baskerville and Emin method (6). Degree hour information was not available at each site, therefore it was not used even though this method is considered more accurate for prediction (48). Weather forecasts on the PMEX network could be a good source of inputs for the predictive model (50). In most years in East Lansing, measurable growth does not begin until April 1 (eg. biofix is not reached). Weather data are not recorded on the PMEX system prior to April 1, and therefore are unavailable to growers and were excluded from the development of the model. Adjusting the Phenocode The vegetative phenocode for leaf number was set at: leaf code = leaf number + 10. The Terminal Bud Set (TBS) code for spurs and terminal extension shoots was set at 43 to be consistent for all shoots and to allow an interval for coding extension shoots with large ( £25 ) leaf numbers. This caused large standard deviations for calculated growth 65 indices on sampling dates during early bud set (eg. some shoots were still growing and some were set) (Table 3). Larger standard deviations were observed for spur shoots, since they are shorter and develop fewer leaves. Ideally, a 'floating' bud set code could be used where: TBSt = leaf codet+1. This would result in the lowest variance if the growth indice was calculated for each individual shoot. Since consistent coding/shoot type is necessary to maintain morphological significance, (eg. a leaf code of 30 should not = 20 leaves in one case and TBS code in another ), the floating bud set code was not used. Average leaves/shoot were calculated for three seasons. Based on 1983 data (HRC) the numeric value for TBS code was adjusted to equal average leaves/shoot + 5. The adjusted TBS code allowed an interval to code most long shoots without overlapping the TBS value, and reduced sampling variances during the early terminal bud set phase (Table 3). Determination g: Degree-Day Base Temperature A series of linear regressions (using growth indice from 1983 HRC data as the dependent variable) were calculated testing base temperatures between 40 °F (4.4°C) to 50 °F (10°C) (temperature selection based on a review of the literature). Degree-day accumulations used for model development were calculated using the base temperature having the highest coefficient of determination. To verify 66 Table 3: The influence of Terminal Bud Set (TBS) phenocode standard deviation and growth 0" index using two phenology coding systems: 1)Terminal Bud Set code = 43, or . 2) revised Terminal Bud Set code = 30, related percent terminal bud set (cv. Red Delicious, vegetative spurs. 1983, HRC). GI usingz GI using2 Date ud TBSs43 Std. Dev. TBS=30 Std. Dev. 5/27 0 18.9 1.12 18.9 1.12 6/8 50 32.1 11.26 25.6 4.62 6/10 65 35.5 10.59 27.0 4.27 6/14 80 39.3 7.71 28.9 2.41 6/17 85 39.5 7.19 29.1 1.86 6/24 95 42.5 2.46 30.1 0.45 2 Growth Index (GI) calculated TBS=43 or revised with TBS=3O to (based on observed leaves/shoot). using coding system with standard deviation 67 base temperature selection the prediction equation selected for Red Delicious (reproductive spur) was tested using degree-day accumulations calculated at base temperatures between 40°F (4.4°C) and 50°F (10°C). The base temperature with the highest coefficient of determination and lowest sums of squares was selected (Table 5). Model Development: Phase I A series of linear and non-linear regressions using growth indice (GItj or WGItJ) as the dependent variable, and degree-day accumulation (base 40°F (4.4°C), at time=tj) as the independent variable, were calculated to develop the most accurate prediction equation. Since the simplest model was preferred, linear models were tried first. Linear and multiple regression analysis were calculated using Genstat (28). Non-linear regression analysis was calculated using PLOTIT (18). The variables age, minimum and maximum temperature, precipitation, pan evaporation, and wind speed were considered as independent variables in the multiple linear models. Non-linear functions ranged in complexity from a two parameter exponential equation, to a five parameter logistic saddle (sigmoidal) function. The first phase of model selection was based on the residual sums of squares, the coefficient of determination, and the visual fit of the regression line to the observed data. 68 Adaptation g: the Biofix Concept for Prediction Methods to.adapt the predictive equations for use in other locations and years than those for which they were developed were examined. A stage of development was arbitrarily chosen as a biofix marker to represent the onset of growth. where: Prediction equation: Y=B(1)/(1+EXP(B(2)-B(3)*(DD)+B(4)*(DD)2 -s*(oo)3>) DD= degree-days accumulated from April 1. If silver tip was selected as the Biofix the prediction equation would be adjusted as follows: Y=B(1)/(1+EXP(B(2)-B(3)*(DD-BIOFIX)+B(4)*(DD-BIOFIX)2 -B(S)*(DD-BI0FIX)3)) BIOFIX = degree-days accumulated from April 1 to silver tip A second method utilizing a biofix marker adjusted the input values for the year (or location) to be predicted (eg. 1985) to reference the equivalent biofix marker from the original prediction curve (fit to the original data set 1983,HRC). The equations were adjusted by calculating the difference between the biofix value derived from the original prediction curve ( based on 1983 data) and the biofix value 69 observed in the year to be predicted (1985). The Biofix Difference equation follows: 80 = (BIOFIX B) - (BIOFIX A) where: BO = Biofix Difference BIOFIX A = degree-days accumulated from April 1, 1983 to biofix stage (eg. silver tip), derived from the prediction CUY‘VE BIOFIX B = degree-days accumulated from (eg.) April 1, 1986 biofix stage = silver tip Therefore, the prediction equation would be adjusted as follows: Y=B(1)/(1+EXP(B(2)-B(3)*(DD-BD)+B(4)*(DD-BD)2 -B(5)*(oD-Bo)3)) Model development: Final Selection and Testing Predicted phenocode values were calculated (using Genstat) for a range of sampling dates (representing the whole season) using the four models selected as having the best statistical and visual 'fits'. Testing and final model selection were combined by calculating predictions for each case (variety and shoot type) using data sets collected at the HRC in 1982 and 1984 or at the Graham Station in 1983. Predictions were made using both the biofix and the biofix t0 70 difference values to modify the degree-day accumulation input value. The predicted values calculated using each model were compared to the observed values for each date. The final model selected predicted values closest to the observed values through the observed range of dates, over years and/or locations. Developing 5 Partial Season Sub-model Whole season growth was subdivided and these portions were modeled separately, with the objective of developing a model with fewer parameters and more accurate predictions for all portions of the growth curve. Linear and non-linear regression used the same variables as the whole season modeling. Vegetative curves were sectioned: a) dormant through green tip and b) first leaf through TBS. Reproductive curves were divided into stages: a) dormant through tight cluster b) tight cluster through fruit set and c) fruit growth. A sub-model describing flower development from tight cluster through fruit set was developed using a refined phenocode. Phenological observations during flower development were recorded on 4 reproductive spurs/tree and 4 reproductive terminals/tree, in two orchards (Red Delicious/Std and spur type Golden Delicious/MM106), at the HRC, in 1983. A detailed phenocode dividing flowering into 20 stages of flower bud expansion was developed (Table 4, 71 Figure 1). Individual flowers were rated, and an average phenocode/cluster was calculated. A Flower Development Index (FDI) was calculated daily by averaging the phenocode values for all clusters/shoot type/variety (100-110 observations). Linear and non-linear regressions using FDI as the dependent variable and degree-day accumulation (base 40°F) as the independent variable were calculated. Model selection procedure and criteria were the same as that used to develop the whole season prediction equations. 72 Table 4: Expanded Flowering Phenocodes: Used to numerically code flower development in apple. ~——"‘v Phenophase Code Figure 1 Early Bud Expansion 0 A Bud Expansion 1 n.a. Early Pink 2 8 Pink 3 B Expanded Pink 1 4 B Expanded Pink 2 5 B Expanded Pink 3 6 8 Full Pink 7 B Pre-bloom 1 8 B,C Pre-bloom 2 9 C Pre-bloom 3 10 C Open Petal 1 11 C Open Petal 2 12 C Full Open Blossom 1 13 C,D,E Full Open Blossom 2 (aged pollen) 14 D,E Petal Fall 1 (1 petal off) 15 E Petal Fall 2 (2 petals off) 16 E Petal Fall 3 (3 petals off) 17 E Petal Fall 4 (4 petals off) 18 E Early Fruit Set 19 E Fruit Set 20 F 73 Figure 1. A-F. Photographs documenting the phenophases used to develop the flowering sub-model referenced by in Table 3. 75 Results Degree-Day Accumulation: Base Temperature Determination Linear regression using growth indice (GI) (cv. Red Delicious, reproductive spurs ,1983, HRC) as the dependent variable and degree-day accumulation as the independent variable, resulted in the highest coefficient of determination when degree-days were calculated at a base temperature of 40°F (4.4°C) (data not shown). Non-linear regression using growth indice (GI) (cv. Red Delicious, reproductive spurs, 1983, HRC) as the dependent variable and degree-day accumulation as the independent variable calculated with the model : Y=B1/(1.+EXP(BZ +B3*X + B4*X*X + 85*X**3 )) also resulted in the highest coefficient of determination and lowest residual sums of squares when degree-days were calculated at a base temperature of 40°F (4.4°C) (Table 5). Model Development: Whole Season Model Non-linear regression using degree-day accumulation base 40°F (4.4°C) as the independent variable yielded higher coefficients of determination, lower sums of squares, and better visual fits for all shoots and varieties than either linear or multiple regression equations (Table 6,7). Multiple regression with models containing some or all of the independent variables degree-day accumulation, age, 76 «min mafia nommindmsa ow 833.3ng 25 1835‘ mgm 3. 351mm 31 comm $253233 8.8. «21.1 :5 331833 8969 am mm 35. ea 9.9.3: 338 BC 3 x8 omindocm 383,38 31 :5 38m: < >ooczcr>joz w>mm LAO 101 Figure 7.A: Observed flower development by stage and flowering predicted by the Logistic equation Golden (Gold) Delicious sgur growth as a function day accumulation, base 40 F. Figure 7.8: Observed flower development by stage and flowering predicted by Heibull's function Golden (Gold) Delicious sgur growth as a function day accumulation, base 40 F. (symbols) (--) for of degree- (symbols) (--) for of degree- SPUR FLOWERING UNITS = PHENOCODE 102 Y-20.405/(1 .O+809.04¢EXP(—(.Ol 7*X+.649E—06*X4-*2))) 2°“ 35 = 4.624 .79/ -I R2 - .995 / d O) L I 1 1 \ \ y 1 / / A 4.- . /'// : /.. —- PREDICTED 0 I GOLD DEUCIOUS t l f r V I r l 200 300 400 500 600 Y-20.31*(1.0-EXP(-(-.097+.00264*Xu4.255))) PHENOCODE SPUR FLOWERING UNITS 2°] 55 = 3.546 ../...-—----- . Ra - .996 ,r’ 1 /) 16'I // .1 / . //- /I 121 / .1 /. GI /( Bi V/ d _ // .1 //./ B 4- . / r// 1,; —- PREDICTED 4 :1 an GOLD DELICIOUS r T ' I 200 . 350 '4Oo 550 600 DEGREE DAY ACCUMULATION BASE 4O Figure 7.C: Observed flower development 103 and flowering predicted by the Logistic Golden (Gold Delicious degree-day accumulation, terminal growth base 40°F. Figure 7.0: Observed flower development and flowering predicted Golden (Gold) Delicious 5 day accumulation, base 40 by Neibull's Bur growth as a by stage (symbols) equation (--) for as a function of by stage (symbols) function (--) for function of degree- PH ENOCODE TERMINAL FLOWERING UNITS = PHENOCODE TERMINAL FLOWERING UNITS = 104 Y=22.28/(1.O+1433.94EXP(-(.0185*X+.4813E—OStXau-2)) 2°) ss= 1.7195 .. - R2 a .998 ,2” .. / 16- /’ d /A . /‘ 12" // - t ‘ / / d /l‘ 8- x. .1 /l/ 4. "/ C -' / -I ///./ . (r - —- PREDICTED 0‘ 4”“ 1. GOLD DEUCIOUS . I . I ' I . I 200 300 400 500 600 Ya20.068¢(1.O—EXP(—(.165+.001814XMS.935))) 20'] s’s = 1.68913 ,. / - R’ - .998 ,2" 16- ,I/ - A ‘/ . / ' 12- / - I ‘ / . // 8~ /A a /' . /./ D 4- / a / // - . (r - —- PREDICTED o« ”/4 1. GOLD DELICIOUS 1 I ' l ' I 200 360 460 500 600- DEGREE DAY ACCUMULATION BASE 4O 105 Figure 8.A: Observed flower development by stage (symbols) and flowering predicted by the Logistic equation (--) for Red Delicious spur growth as a function of degree-day accumulation, base 40°F. Figure 8.8: Observed flower development by stage (symbols) and flowering predicted by Heibull's function (--) for Red Delicious spur growth as a function of degree-day accumulation, base 40°F. SPUR FLOWERING UNITS = PHENOCODE PHENOCODE SPUR FLOWERING UNITS 106 Y-21.16/(1.0+1 10.944EXP(-(.0095-x+.7516E-054xa-42)) 2°), 95 = 5.962 [.y’r’i I R2 - .994 '7" ‘I // 16- // .1 // . / ' . /- 12" //I 1 2 4 I/ ' ‘ / y 8- 7 .1 (l j v/ A 4.1 );// .67)”- —- PREDICTED O I a RED DEUCIOUS r ' T fl I I ' 200 350 400 500 600 Y-20.164-(1 .o—Exp(-(.5997+.9786E-034x»1 1.096») 2°“ 53 = 4.962 -y"""" u y/ . R2 =- .995 '/ ‘ / 16- / -1 // .9 ///I / I 121‘ //. 8" /7/ J ‘/ 1 -/'/ B 4~ . 1 // fix” - —- PREDICTED O I RED DEUCIOUS r fi fi If ' I r I 200 300 400 500 600 DEGREE DAY ACCUMULATION BASE 4O 107 Figure 8.C: Observed flower development by stage (symbols) and flowering predicted by the Logistic equation (--) for Red Delicious terminal growth as a function of degree-day accumulation, base 40°F. Figure 8.0: Observed flower development by stage (symbols) and flowering predicted by Neibull's function (--) for Red Delicious spur growth as a function of degree-day accumulation, base 40°F. TERMINAL FLOWERING PHENOCODE UNITS = PHENOCODE TERMINAL FLOWERING UNITS = 108 Y-21.13/(1.O+142.474EXP(-(.0105*X+.6614E-054-XMZ)) 53 = 5.18754 / 2 R’ - .995 /" / c 2 ./ - PREDICTED II RED DEIJCIOUS ' T ' T T I ' I 200 300 400 500 600 Y-20.18*(1 .O-EXP(-(.495+.OOI 23tXuB.847))) $5 = 4.12012 I' ' - Ra - .996 ' / /l / /I / D . v" ‘f/ - —- PREDICTED ‘ a. RED DEUGIous . I ._ I I I . I 200 300 400 500 600 DEGREE DAY ACCUMULATION BASE 4O 109 Hw Table 9: Regression statistics for selected models correlating Flower Development Index (FDI) and temperature: reproductive spurs (1983,HRC). H‘A‘__ RED DELICIOUS Coefficient of Sums of 8(1) Function Determination Squares Parameter Gompertzz .996 3.142 ---- Logisticy .994 5.962 21.16 Neibull'sx .995 4.694 20.16 GOLD DELICIOUS Gompertz .992 7.928 ---- Logistic .995 4.698 20.41 Heibull's .996 3.525 20.31 2 Y=B(1) * EXP(-B(2) * EXP(-B(3) * x)) Y v=3(1)/(1.o + 3(2) * EXP(-(B(3) * X)) x Y=B(1) * (1.0 - EXP(-(B(2) + 3(3) * X) ** B(4)))) 110 Table 10: Regression statistics for selected models correlating Flower Development Index (FDI) and temperature: reproductive terminals. RED DELICIOUS Coefficient of Sums of 8(1) Function Determination Sguares Parameter Gompertzz .992 8.6064 ..... Logisticy .995 5.1875 21.13 Neibull'sx .996 4.1201 20.18 GOLD DELICIOUS Gompertz .997 3.0291 ---- Logistic .998 1.7195 22.28 Heibull's .998 1.6891 20.07 2 Y=B(1) * EXP(-B(2) * EXP(-B(3) * X)) Y Y=B(1)/(1.0 + 8(2) * EXP(-(B(3) * X)) x Y=B(1) * (1.0 - EXP(-(B(2) + 8(3) * X) ** B(4)))) 111 and Red Delicious (Figure 8), but Weibull's function gave a better estimate of the 81 equation parameter (should = 20) for all shoots, than the Logistic equation (Figure 7,8). Heibull's function was selected as the model to represent flowering : Red Delicious: spurs: Y=20.16*(1.0 - EXP(-(.5997 + .9786E-03*X**11.096))) terminals: Y=20.18*(1.0-EXP(-(.495+.00123*X**8.847))) Golden Delicious: spurs: Y=20.31*(1.0-EXP(-(-.O97+.00264*X**4.255))) terminals: Y=20.07*(1.0-EXP(-(.165+.0018*X**5.935))) Adaptation of the Biofix Concept for Prediction Use of the plant biofix adjusted the prediction equation by eliminating the need for determining the onset of growth. However, predictions using degree-days accumulated to biofix as the modifying factor were not necessarily an improvement over those simply using degree- days as an input value (Table 11), where: Prediction equation: Y=B(1)/(1+EXP(B(2)-B(3)*(DD)+B(4)*(DD)2 -B(S)*(DD)3)) DD= degree-days accumulated from April 1. If silver tip was selected as the Biofix, the prediction equation would be modified as follows: Y=B(1)/(1+EXP(B(2)-B(3)*(DD-BIOFIX)+B(4)*(DD-BIOFIX)2 -B(5)*(DD-BIOFIX)3)) BIOFIX = degree-days accumulated from April 1 to silver tip 112 Use of the biofix difference factor to modify the prediction equations resulted in the best predictions, especially when predictions were made at alternate locations (Table 11), where: the Biofix Difference Equation 80 = (BIOFIX B) - (BIOFIX A) and 80 = Biofix Difference BIOFIX A = degree-days accumulated from April 1, 1983 to biofix stage (eg. silver tip), derived from the prediction. CUI‘VB BIOFIX B = degree-days accumulated from April 1, 1986 to silver tip The prediction equation would be adjusted as follows: Y=B(1)/(1+EXP(B(2)-B(3)*(DD-BD)+B(4)*(DD-BD)2 ‘ ’ -B(5)*(oo-Bo)3)) Model Testing: Predictive Capabilities The similarity between seasonal growth curves across years and locations is striking, but demonstrates graphically the reliability of the selected equations (Figure 3-5). Plots of the vegetative growth for spur type Golden Delicious show the greatest difference between years, with final leaves/shoot lower in 1985 (HRC)(Figure 5C,D). 113 qmcdm HH” nostmwdmoz om camadnama mag ocmm1r m>rr Hwnv wwm.~ ~A.m .m -.o ~.o mw.m .m mrch Hszszm Ho 33 Hmm.mv moo.m wo.o o No.0 .m mo.m o mxcHa oH>z.u~.OIN.o as mev Hoom.~ w~.o o wH.m .m um.o 0 mmch oH>3.uw.me.o as Humv HmoV.H uw.m H.m wn.o H.o wa.o H.o mxcHa o~>3.um.oIm.a as mev mmmw.o um.m .m w~.m .m wm.o .m ~ 28 aoqdwdnmnfloam no azm aoqmd" x u ommxmm cpxm « 308m. soafimfimq" x «Homoxmm swam I adomfixv Hmmm «mxav x 282 89.22: x "A8848 88.6 .. 9.9.; 329.903 38 8x: 114 The predictive model was tested using data from the HRC(1984‘ or 1985 data), or the Graham Station (1983 data). Predictions were calculated using the biofix difference equation to modify degree-days input. Models for reproductive growth predicted best, with an accuracy of O - 1.5 stage throughout the growing season (Table 12,13,14,15). Predictions for spur type Golden Delicious were most accurate, predicting full bloom and fruit thinning (10mm) accurately (0 stage difference between predicted and observed values), by 18 and 34 days in advance, respectively (Table 12). Accurate predictions for fruit size (1 1.0 stage) using Biofix=Silver Tip (selected up to 3 months in advance) were made for all varieties tested (Table 12,14). The model underpredicted final harvest size if fruit load was low to medium (data not presented). Vegetative curves generally predicted with less accuracy than the reproductive curves, with differences between observed and predicted values of O - 2.0 stage (Table 13,15). Models for terminal extension shoot growth predict best, usually predicting to within 1.0 stage (Table 15). Models predicting spur shoot growth are the least reliable, frequently generating prediction errors of 1.0 - 2.0 stages (Table 13). 115 43.6 .N. 8581.83 8.. 9.8.291 a... 3328 15:81.88 31 8.83.. .8329; a... «31m 52. 8 3:38 :5 39.8.. <82.)H.+mx22~.+2u.»x+2£»x»x+234x38.. ..9. 13109823 2.5. 2.8." 9.93.. a: 513 31.3.3. m3! 992.: #38825 152.35% 5.: 31m 093.. 35293 2.8 58. 9.9.8.. 852993 33.3.. mg; me 3228.0. 311.8. 93.8. 229.38 c.2338 c.2338 8813.. 5.8.35 3.8.29. E E 3.8.2.... Hfiaaoai Em 3.8.39. .Ezfioai BB. .... ... 35% I... .m a... 3... c 8.. a.“ .m :2: 8.3% :2 ~28 8:1. ..o ~2.~ 8.... ..m .88 Ba .... 1:... 2.8.. 8.8 3... 8. c 29o -.m ..m gob -b .... 13? 2F m... mum; Sh .m 8gb 8... a... 2...... ~58 .u 2...: 2.22:... 8 a: 39... 8o... R; o 89u 8.... ..n. «~93 8.... .m 2...: Ea..n~.o-~.a 0.. «N. 88h 3... c 58... 8... ..o ..Rh 8... .... 2...: c.§.uu.m-u.o 0.. mm ESL 2... ..o ~89: um... c =8.~ 2% .m 3:: 22.1999. 0.. 8 33... 8... 9 BE... Sh .m ESL 8:1. .... ~ 113.22. 5.5“ 19.2.9. 3 :5 3213" Pa 38? a 5313 9.5 2.2.39. 3. 26 39812.8 ...... m3... .852. 3 an 33 8. x #33 8.233.. 5.3 33.... 5.2388 3.823 $8 «an» . an 2.81 n... “Sam. 116 :55 5. 8.313. 3 33.nnmau ...... 8329. 1.9333 3.. 3.83.. .8929; ...... .233 3.... S 3.3%.. 2... .88.. 72:2p.$§.w.~.$8.£&2.££$.$.51..... «.... 3332.3 35. 2.8.. 99.»: .... 923 31.2.3. m3! 99:: .4... Ba. 8.8. 8:28” E .E 8.8. 5.2.33... 83...... m2... E 8:28. 32...... 3.23» ...r mzooqm mac: mzooqm 28; we. we. .8. E. we. 8 u1d. Hm.~mmu mo.o oc.o mo.c mo.o ma.c .mo.o Baum—x... 9:. cmdaocmxgoa momnmo.a cmd‘n.o=m\mna .6 £8 8:22mxm3 "WE’VWE'L'TMTIL'Wfl'fl'flmflflr