MSU LIBRARIES m \— RETURNING MATERIALS: PIace in book drop to remove this checkout from your record. FINES wil] be charged if book is returned after the date stamped below. ‘ car! 0 7 1999 STUD- OP THE PABASITES OF THE JACK PINE BUDWORM: LIFE HISTORY STUDIES; HYPERPARASITISM; AND THE APPLICATION OF TIME SERIES ANALYSIS TO THE STUDY OF ADULT ACTIVITY PATTERNS By Norman Elliott A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Entomology 1985 ABSTRACT STUD. OF THE PARASITES OF THE JACK PINE BUDWORM: LIFE HISTORY STUDIES; HYPERPARASITISM; AND THE APPLICATION OF TIME SERIES ANALYSIS TO THE STUDY OF ADULT ACTIVITY PATTERNS BY Norman Elliott Aspects of the life history of the parasites Glypta fumiferanae Viereck, Ayanteles fumiferanae (Viereck) and Apanteles M Mason were studied in populations in Michigan's lower peninsula. Host-parasite synchrony, spatial and temporal activity patterns of adult parasites, impact on jack pine budworm populations, and adult food relationships were studied. The results indicated that the activity of populations of adult parasites varied temporally and spatially in relation to budworm phenology, time of day, temperature, and vertical stratum within tree crowns. The three species exhibited consistent levels of parasitism in the cohorts studied. Honey was found to increase the longevity of _G_. fumiferanae and _A_. fumiferanae, but honeydew produced by an undetermined aphid species that was abundant in the study sites did not increase the longevity of adults over those not fed. Mortality of Apanteles spp. pupae was studied in the field. Hyperparasites, predators, and unknown causes accounted for high levels of mortality in the cohorts studied. Mortality caused by one hyperparasitic species was found to be positively spatially dependent on the density of Apanteles spp. pupae. A A univariate ARIMA model was fitted to three years of daily trap catch records for adult female Glypta fumiferanae. The properties and problems associated with the use of time series models fitted to population data of the type employed here were discussed. Multivariate (transfer function) models incorporating average daily temperature and total daily rainfall as input processes were fitted. The resulting models showed little improvement in predictability over the univariate model. To Janice, David, Robbie, and Christina; and to Tabbetha too. ii ACKNOWLEDGEMENTS I wish to express my gratitude to Gary Simmons for serving as my major advisor. I am particularly grateful for the intellectual stimulation, encouragement, and confidence he provided me during my program. Thanks also to Lal Tummala for serving as co-major advisor and providing insight on quantitative methodology. To my committee, Dean Haynes, Stuart Gage, and Roland Fischer I extend my appreciation for their guidance during my graduate training. A number of people assisted in collecting data for this thesis. In particular, I thank Charley Chilcote, Ray Drapek, and Frank Sapio, each of whom took time out from their busy schedules to help me at critical times during my research. Finally, I thank Susan Battenfield and Kelly Barden for their help in preparing this manuscript. TABLE OF CONTENTS General Thesis Introduction .................................. 1 Literature Cited ........................................... 5 Literature Review ............................................ 6 Life history and impact of the jack pine budworm .......... 6 Population dynamics of the jack pine budworm .............. 7 Parasites of the jack pine budworm ........................ 12 Hyperparasites of jack pine budworm parasites ............. 14 Literature Cited .......................................... l7 Studies of the Natural History of Glypta fumiferanae and Apanteles spp. ........................................ 2l Introduction .............................................. 22 Materials and Methods ..................................... 24 Description of study areas ............................. 24 Meteorological information ............................. 26 Determining the relative abundance of parasites of early instar JPB larvae ........................... 26 Pupation and emergence rates of parasite species ....... 28 Temporal and spatial activity patterns of adult parasites ............................................ 28 Food sources of adult parasites ........................ 29 Results ................................................... 3l Relative abundance of parasites of early instar jack pine budworm larvae ............................. 3l The phenologies of parasites of early instar jack pine budworm .................................... 33 Spatial and temporal activity patterns of adult parasites ............................................ 37 Food sources of adult parasites ........................ 50 Discussion ................................................ 57 Literature Cited .......................................... 67 The Aplication of Time Series Analysis for Predicting Activity Patterns of Adult Female Glypta fumiferanae Populations ................................................ 72 Introduction .............................................. 73 Construction of Time Series Models ........................ 76 Field techniques ....................................... 76 Environmental data ..................................... 77 Statistical methodology ................................ 79 Model development ...................................... 83 iv Discussion ................................................ 97 Literature Cited .......................................... 108 Mortality of Apanteles spp. Pupae Due to Hyperparasitism, Predation, and Unknown Causes .............................. llZ Introduction .............................................. 113 Materials and Methods ..................................... ll4 Temporal distribution of hyperparasitism ............... ll4 Spatial distribution on mortality ...................... llS Results and Discussion .................................... ll7 Literature Cited .......................................... 126 Appendices .................................................... 128 Appendix I ................................................ l28 Literature Cited ......... .............................. l35 Appendix II ............................................... l36 Literature Cited ....................................... 139 Appendix III .............................................. l40 TABLE 10. LIST OF TABLES Parasitism of third and fourth instar C, pinus in samples from four cohorts ........................... 32 Results of three stage analyses of variation among sites, strata (upper crown, lower crown, and supp- ressed trees), and branches ............ . ............... 34 ADV tables for regression of Malaise trap catch of female parasites (separate regressions done for A, fumiferanae and G. fumiferanae using the independent variabTes trap height TTHT'and’average temperature over the two hour interval between observations ............. 49 Comparison of numbers of G, fumiferanae visiting and feeding on three food sources .......................... 55 Mean longevity (f SE) of Glypta fumiferanae and Apanteles fumiferanae fed various foods ................ S6 Four models for forecasting transformed trap catch of female GLypta fumiferanae ........................... 88 Estimated variance of forecasting error for transformed trap catch ................................. 96 Percent mortality of Apanteles spp. pupae due to hyperparasitism, predat on, and unknown causes at the GR site during 1982 and 1983 ....................... 118 Percent parasitism of Apanteles spp. pupae in diff- erent microhabitats in our study sites in 1984 ........ 121 Results of linear regression analysis of mortality versus host density for the hypreparasites E, phycidis and fl, percussor ....................................... 123 vi FIGURE 1. 10. 11. 12. LIST OF FIGURES Map of locations of study plots in Michigans Lower Peninsula ....................................... 25 Frequency distribution of pupation of A. morrisi and A, fumiferanae reared in an outdooF'insectaEy in relation to accumulated degree days base 8.9 C ..... 35 Frequency distribution of emergence of A, morrisi and A. fumiferanae reared in an outdoor insectaBy, in rélation to accumulated degree days base 8.9 C ..... 36 Cumulative emergence of A anteles spp. and Glypta fumiferanae in the GR site in 1982 and 1983 ........... 38 Malaise trap catch of adult g, fumiferanae at the GR site in 1982 and 1983 in relation to accumulated degree days base 8.9 C ................................ 39 Malaise trap catch of adult female A, fumiferanae in relation to accumulated degree days base 8.9 at the GR site in 1982 and 1983 ....................... 41 Malaise trap catch of adult female A, morrisi in relation to accumulated degree days base 8.90C at the GR site in 1982 and 1983 .......................... 42 Average proportion of total daily Malaise trap catch of adult female 5, famiferanae and G. fumiferanae for seven time intervals within a day ................. 44 The relationship between Malaise trap of adult female G. fumiferanae and temperature for various time int- ervals within the day ................................. 47 The relationship between Malaise trap catch of adult female A. fumiferanae and temperature for various time intervaTs within the day .............................. 48 The relationship between Malaise trap catch of female A. fumiferanae and trap height in the crownns of Eack pine trees ....................................... 51 The relationship between Malaise trap catch and the height of traps in the crowns of jack pine trees ...... 52 vii FIGURE 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. The relationship between Malaise trap catch of adult male G. fumiferanae and the height of traps in the crowns oijack pine trees ............................ 53 Daily Malaise trap catch of adult femaleG Glypta fumiferanae over a three year period ................. 78 Daily average standardized temperature over a three year period corresponding to the period of adult female Glypta fumiferanae activity ................... 80 Daily cm of precipitation over a three year period corresponding to the period of adult femaleG 1ypta fumiferanae activity ................................. 81 Transformed daily Malaise trap catch of adult female GLypta fumiferanae over a three year period .......... 85 The sample autocorrelation and partial autocorrel- ation functions of the transformed and differenced adult female Glypta fumiferanae daily Malaise trap catch series ......................................... 86 The autocorelation function of the residual series generated from the fitted ARIMA (1 1 1)x(0 1 1)34 model ................................................ 87 One-step ahead predictions for the Grand Traverse Co. 1982 and Grand Traverse Co. 1983 trap catch data based on an ARIMA (1 1 1)x(0 1 1)34 model ............ 90 One-step ahead predictions for the Delta Co. 1981 and Nexford Co. 1983 trap catch data based on an ARIMA (1 1 1)x(0 1 1134 model ........................ 91 'Four-step ahead predictions for the Nexford Co.1983 trap catch data based on an ARIMA (1 1 1)x(0 1 1)34 model ................................................ 93 Predictions for the Hexford Co. 1983 Malaise trap catch data based on an ARIMA (1 1 1)x(0 1 l)34 model without the use of an updating recurrsion ............ 94 The likelihood surface as a function of the parameters of the Box-Cox transformation ........................ 102 viii FIGURE 25. 26. Percent parasitism of Apanteles spp. pupae by four species of hyperparasites 1n successive samples of pupae collected from a study site in Grand Traverse Co. Michigan during 1982 and 1983 ........... 120 A flow chart of the proceedure for estimating the prediction error variance that minimizes the like- lihood function for a Gaussian process ............... 134 ix GENERAL THESE INTRODUCTION A major objective of present day insect pest control is to insure, via environmentally compatible and economically sound methods, that damage stays below an acceptable level. The use of parasitic insects in biological control has proven to be an environmentally compatible and economically sound form of pest control. Attempts at biological control have resulted in a number of successes but have also resulted in numerous failures. Since projects to initiate biological control are initially expensive, failures to achieve effective control are highly undesirable. Thus, basic research on the theory and practice of biological control is necessary in order to increase the probabilities of success in attempts to establish this form of pest control. Methods of achieving or improving biological control using parasitic insects fall into three general categories: importation of exotic parasites, augmentation of established exotic or endemic parasites, and management of existing parasites (Debach 1973). Management entails any action taken to improve the effectiveness of a parasite in reducing the population levels of its host. When management can be successfully implemented, it will likely be more cost effective than importation or augmentation. Before we can successfully implement management strategies for parasites a set of "first principles" (or relations) must be deveIOped which relate factors operating on the parasites life system to sets of management options. Once this is accomplished, we have a basis for evaluating specific parasites-host systems in terms of alternative management strategies. While each system will clearly be unique, this set of relations will narrow the scape of potentially important interactions, thereby reducing the expenditure of time and resources necessary to permit the identification of available management options. Because of their stability and the diversity of ecological interactions, many forest ecosystems provide excellent experimental environments for conducting investigations aimed at determining important factors impacting on host-parasite systems. The factors which determine the suitability of particular parasite species in biological control have not been well studied. These factors can however be thought of as belonging to one of the following general categories: 1. Intrinsic factors, i.e., factors operating from within the biological system of the parasite. Examples of intrinsic factors are sex ratio, fecundity, host searching capacity, longevity, host range, temporal synchrony with host, etc. 2. Extrinsic factors, or factors extraneous to the biological system of the parasite. Examples of extrinsic factors are environmental variability, mortality factors operating on the parasite population, abundance of food sources for free living stages of the parasite, abundance of alternate hosts of the parasite, etc. Hyperparasitism, spatial incoincidence, and temporal asynchrony between the host and parasite populations are thought to be important factors influencing the dynamics of host-parasite systems (Huffacker and Messenger 1976, Hassel 1978). However, few investigations have been undertaken to illucidate the importance of these factors in actual ecological systems. Two factors may account for the rarity of such investigations. First, parasites have highly specialized habits and sampling their papulations requires basic knowledge of their biology; as a result, sampling techniques often need to be developed for each situation (Roach et a1. 1979). In many instances, techniques appropriate for such studies are not available. Most studies of host- parasite systems involve the collection of hosts from which parasites are reared. The parasites are then identified and an expression of relative abundance and percentage parasitism is computed to measure their impact on the hosts papulation (Huffacker 1971). Estimates derived by this method are frequently subject to considerable bias (Simmonds 1948, Van Driesch 1983) and furthermore, they yield no information about the dynamics of the adult stage of the parasite, a life stage for which the potential for increased efficacy via management practices may be high (Van Den Bosch and Telford 1964, Weseloh 1976). Second, a theoretical basis for interpreting the results of such studies is not well developed. The research reported in this thesis was conducted in the jack pine (Pinus banksiana Lamb.) ecosystem because of the diversity and relative stability of ecological interactions in such a forested system. Specifically, a system consisting of the jack pine budworm (Choristoneura pinus Freeman), and certain parasites and hyperparasites which operate within this system was studied. Relative techniques were used for monitoring the adult population densities of the parasites and hyperparasites and these techniques were used to relate adult density to spatial and temporal patterns of environmental variability and to parasitism rates. Field and laboratory studies of the ecology and behavior of the adults of certain parasitic species were done in order to determine the importance of various factors on the dynamics of their populations. Lastly, the influence of environmental variability on the activity of adult papulations of the parasite Glypta fumiferanae Viereck was studied by the construction of a mathematical model. The utility of the model for predicting variation in activity levels as a function of previous levels of activity and environmental factors that influence activity was discussed. LITERATURE CITED Debach, P. 197 3. Biological Control By Natural Enemies. Cambridge University Press, London. Hassell, M. P. 1978. Arthropod Predator-Prey Systems. Princeton University Press, Princeton, NJ. Huffacker, C. B. 1971. Biological Control. Plenum Press, New York. Huffacker, C. B. and P. S. Messenger. 1976. The Theory and Practice of Biological Control. Academic Press, New York. Roach, S. H., J. W. Smith, S. B. Vinson, H. M. Graham, and J. A. Harding. 1979. Sampling predators and parasites of Heliothis species on crops and native host plants. In: W. L. Sterling (ed.) Economic Thresholds and Sampling of Heliothis Species on Cotton, Corn, Soybeans, and Other Host Plants. Southern Coop. Ser. Bull. 23. Simmonds, F. L. 1948. Some difficulties in determining by means of field samples the true value of parasitic control. Bull. Ent. Res. 39:435-40. Van Den Bosch, R. and A. D. Telford. 1964. Environmental modification and biological control. In: P. DeBach (ed.) Biological Control of Insects, Pests, and Weeds. Reinhold Publ. Corp., New York. Van Driesche, R. G. 1983. Meaning of "percent parasitism" in studies of insect parasitoids. Env. Ent. 12:1611-22. Weseloh, R. M. 1976. Behavior of forest insect parasitoids. In: J. F. Anderson and H. K. Kaya (eds.) Perspectives in Forest Entomology. Academic Press, New York. LITERATURE REVIEW In this section the relevant literature on the ecology of the jack pine budworm, its parasites, and hyperparasites is reviewed. Life history and impact of the jack pine budworm Depending on the locality and weather conditions, adult jack pine budworm emerge from late June to early August, with mid-July being about average for Michigan (Graham 1935). Males emerge slightly earlier than females and are active fliers. Gravid females are relatively sedentary, but spent females fly actively and can sometimes disperse great distances. Flight activity is greatest at night. The females lay egg masses consisting of approximately 50 eggs on old growth needles throughout the crowns of jack pine trees. Eggs are laid from late June to early August depending on weather and locality; they hatch seven to ten days after they are deposited. The first instar larvae wander and disperse on the wind for a few days and then spin silken hibernaculae in sheltered locations on the trees. First instar larvae do not feed. The small overwintered larvae break diapause in May at about the same time that male staminate flowers are opening. They undergo another period of dispersal before settling in to begin feeding. Development of the new shoots is well advanced before the larvae leave the male flowers and begin to feed on new foliage (Dixon 1961). Feeding on foliage is wasteful and many partially eaten needles are incorporated into the feeding shelters constructed along the axis of the shoots. The larvae are fully deve10ped by mid-June to mid-July and pupate within their feeding shelters. In about a week adults emerge to form the next generation (Kulman and Hodson 1961b). Jack pine budworm defoliation may cause mortality, top-killing, reduced growth, and reduced pollen production. Up to 33 percent mortality of merchantable trees and 90 percent mortality of intermediate and suppressed trees has been reported as a result of severe outbreaks of the jack pine budworm (Batzer and Millers 1970, Benjamin 1965). Understocked, overmature, or stands growing on poor sites tend to suffer more damage than well stocked stands growing on good sites (Batzer and Miller 1970, Benjamin 1965). Tree mortality is more likely to occur when budworm outbreaks coincide with or are preceded by drought or outbreaks of other jack pine defoliators (MacAloney 1944, Batzer and Millers 1970). Population dynamics of the jack pine budworm Numerous studies have been undertaken to investigate the biology of the jack pine budworm, its natural enemies, the influence of pollen abundance and stand characteristics, and the influence of environmental factors on the population dynamics of the jack pine budworm (Allen et a1. 1969, Allen et al. 1970, Batzer and Jennings 1980, Benjamin 1965, Benjamin and Drooz 1954, Dixon 1961, Dixon and Benjamin 1963, Drooz and Benjamin 1956, Foltz et a1. 1972, Jennings 1971, Kulman and Hodson 1961a, Lejeune 1950, Mattson et a1. 1968, Simmons and Sloan 1974, and others). However only Foltz et a1. (1972) and Batzer and Jennings (1980) utilized quantitative methods for studying the population dynamics of the budworm. These studies utilized life tables similar to those developed for the spruce budworm (Morris and Miller 1954). They employed age interval models for the description of within generation survival, and a components of variance technique described by Mott (1966) was used to elucidate the influence of survival of various age classes on generation survival. Foltz et al. (1972) analyzed 29 life tables based on sampling ten areas in Michigan from 1965-1968. Survival over the interval from eg to third instar had the greatest impact on population trends. Survival of late instars (fifth and sixth instars) and realized fecundity were also found to influence population trends. Of factors influencing survival from egg to third instar, fall and Spring dispersal were thought to be the most important, with egg parasitism by Trichggramma minutum (Riley) exerting only a minor effect. Predation by birds was thought to be an important component of large larval survival, however a major portion of large larval survival could not be accounted for by any known factor. Parasitism generally accounted for 10 percent or less of total large larval mortality. Five species of Diptera were responsible for most of the parasitism of this stage. The major mortality factors affecting small larval survival (third and fourth instars) was parasitism by HymenOpterous species; a complex of species of the genus Apanteles and Glypta fumiferanae (Viereck) were the primary parasites involved. In a supplementary study of three cohorts of jack pine budworm it was found that apparent combined parasitism by these parasites was 51.3, 26.5, and 36.2 percent of small larvae. Small larval survival was not considered to be a significant factor in determining population levels in the succeeding generation. Parasitism was considered to be the most important factor influencing pupal survival. Apparent parasitism ranged from 7 to 25 percent in their study. Itoglectis conguisitor (Say) was the major pupal parasite recovered from reared pupae. I. conguisitor is a highly ubiquitous pupal parasite (Weseloh, 1976); it is highly polyphagous and multivoltine and as such cannot be expected to exhibit a significant numerical response to increases in jack pine budworm density. A major proportion of pupal mortality was attributed to unknown causes. Pupal survival contributed only a minor amount to pOpulation fluctuations. Realized fecundity accounted for the second largest portion of the variance in influencing p0pulation numbers in the next generation. This term was made up of a number of component sources which were not individually measured, such as the proportion of adults that were female, mating success, survival and dispersal of adults, and potential fecundity; as such the major factors influencing realized fecundity were not determined. Batzer and Jennings (1980) based their study on 48 life tables from sampling 20 plots in Minnesota from 1965-1968. They found that large larval survival was most closely associated with population trends at high budworm densities. However, at low budworm densities, small larval survival was more important. There was no age interval that was consistently correlated with the size of the next generation for all data groups when the data were grouped according to year or jack pine budworm population density. Mortality during the egg stage was consistently low, though there was a reduction in average egg mass size as the outbreak aged. Mortality of first instar larvae was low and attributed primarily to fall dispersal losses. Spring dispersal losses averaged 32-40 percent of third instars. However, undetermined causes, perhaps overwintering mortality, were as important as dispersal in determining total mortality of this stage. Large larval (third through sixth instars) mortality was decomposed into that due to parasitism by Apanteles spp. and Q. fumiferanae, and that due to unknown causes. Combined apparent mortality due to parasites averaged 12-33 percent, with mortality due to unknown causes accounting for most of the variation in large larval mortality. 10 Pupal mortality was decomposed into that due to parasites and that due to unknown causes. Pupal parasites accounted for an average of 34 percent apparent mortality but this factor was highly variable. Survival of adults was highly correlated with papulation size in the next generation, however little was known of the factors affecting adult survival. Numerous factors could have been acting, including mortality due to predators, moth dispersal, mating failure, and factors determining characteristics such as body size and egg mass size. Natural enemies, weather, food, habitat, and genetic attributes of pOpulations have been implicated as factors influencing population fluctuations of the jack pine budworm. It appears that no single factor can be identified that can account for a high proportion of observed fluctuations in budworm numbers under all conditions. Natural enemies are thought to be important at low budworm densities. Parasitism by Apanteles spp. was found to be more important in dense stands than in poorly stocked stands (Batzer and Jennings 1980). Parasitism by g. fumiferanae was found to be greater in the interior of stands than at the periphery (Kulman and Hodson 1961a). It is thought that parasites and predators exert some control over p0pulations that are at endemic levels, but that once an outbreak is triggered by favorable environmental conditions natural enemies have little or no effect on damping the release phase of the outbreak; they may, however speed the rate of population collapse once the outbreak has begun to decline. It appears that parasites and predators show little numerical response or functional response to budworm population fluctuations. In the case of parasites responses may be limited because most budworm parasites have alternate hosts. However, other than rearing hosts to 11 obtain estimates of parasitism rates, little biological or ecological work has been done on budworm parasites. The possible effect of weather on the release and decline of outbreaks has been described. Batzer and Jennings (1980) believed that the collapse of an outbreak in Minnesota may have been associated with increased humidity and rainfall during the late larval stage as the infestation aged. They observed that large larval survival was greater in low density stands, and that the drier warmer conditions in the low density stands might favor survival because disease levels might be lower than in cool moist environments. Temperature and humidity were thought to be important factors influencing pupal survival, with warm dry conditions favoring pupal survival (Foltz et al. 1972, Graham 1935). Foltz et al. (1972) hypothesized that variation in dispersal losses to early instar larvae and fecund adults were the principal factors responsible for p0pulation fluctuations. They theorized that reduced mortality due to sunny, nonwindy weather could cause populations to increase above endemic levels. The subsequent defoliation caused by the high populations caused changes in the quality of the vegetation and perhaps also in the quality of the jack pine budworm pOpulation, which increased dispersal losses. This led to the collapse of the outbreak. It is believed that pollen, the preferred food source of larval budworm may provide a nutritional advantage over foliage, and feeding in the male cones may also provide a microhabitat that is favorable to larval survival (Lejeune and Black 1950, Lejeune 1950). Fluctuations in male cone production from year to year may affect population trends because of fluctuations in the quality of the food source, and because of changes in dispersal behavior associated with male 12 cone abundance. When male cones are abundant, larvae settle in and feed with little wandering and thus, less dispersal loss (Batzer and Millers 1970, Heron 1956, Kulman et al. 1963, Foltz et al. 1972). Tree and stand characteristics play an important role in determining the abundance of jack pine budworm (Graham 1935, Hodson and Zehngraff 1946, Benjamin 1954, Heron and Prentice 1957, Rudolf 1958, Dixon 1961, Rose 1973, Batzer and Jennings 1980). Stands which tend to be most susceptible to heavy defoliation by the budworm are typically understocked, overmature, growing on poor sites, and composed mostly of orchard type trees, or trees of poor vigor. Young, well stocked stands in good growing condition are generally resistant to heavy budworm attack. Trees in such stands generally produce few male flowers. In contrast, trees in poorly managed stands, such as those described above, produce large numbers of male flowers. Thus, the differing susceptibility of jack pine stands to jack pine budworm feeding damage may be related to differences in the resistance of trees to attack (Batzer and Jennings 1980). Parasites of the jack pine budworm Sixty-three species of primary parasites and hyperparasites have been reared from the jack pine budworm (Walley 1953, Benjamin and Drooz 1954, Drooz and Benjamin 1956, Kulman and Hodson 1961a, Dixon and Benjamin 1963, Allen et a1. 1969). The vast majority of species are of sporadic and infrequent occurrence. Only ten or so parasite species appear to be consistently associated with jack pine budworm populations, and of these, parasites which attack the early larval and pupal stages predominate. Apanteles fumiferanae Viereck and (Lypta fumiferanae are the only parasite species essentially limited to the 13 budworm. All other species are polyphagous, and some, such as Itoplectis conguisitor have very broad host spectra. A complex of three or more species of AEteles attack early instars of the budworm; but whether these species attack first and second instars in the summer or attack overwintered larvae the following spring is unknown. Previous reports suggest that first and second instars are attacked in the summer (Dixon and Benjamin 1963, Allen et al. 1969). However, Miller and Renault (1976) noted that Apanteles species, other than A. fumiferanae associated with spruce budworm p0pu1ations, probably attack the budworm the following spring. The Agnteles parasites emerge primarily from fourth instars. If the parasites must pass through an alternate host, their abundance would be primarily determined by the abundance of their alternate hosts. All existing studies of jack pine budworm parasites were done in outbreak and declining populations. Thus, the relationships among parasite species observed in those studies may not have been indicative of those in endemic p0pu1ations. Miller and Renault (1976) observed major differences in the abundance and diversity of species in the parasite complex attacking endemic and outbreak phase budworm p0pu1ations. No information exists concerning the functional or numerical responses of jack pine budworm parasites to increases in budworm density. In the spruce budworm system A. fumiferanae and g. fumiferanae both show a limited numerical response to increases in host density. However, the magnitude of the numerical response is insufficient to dampen the release phase of spruce budworm population outbreaks (Miller 1959, Miller and Renault 1976). In the spruce budworm system, A. fumiferanae shows a decrease in host searching efficiency with increasing host density (Miller 1959). Miller 14 felt that changes in fecundity, adult food supply, parasite mortality of parasite larvae, mortality of parasite adults, mortality of parasite cocoons due to hyperparasitism, or sex ratio could have accounted for this phenomenon. A general conclusion from studies of the population dynamics of the jack pine budworm is that parasites and other natural enemies have limited effectiveness in maintaining budworm populations at endemic levels. Natural enemies are thought to be able to maintain budworm populations at low levels for a period of time, but when environmental conditions favor population increases, natural enemies cannot dampen the population explosion. Once the outbreak has run its course, parasites and other natural enemies may speed the rate of collapse. Studies directed at determining factors other than host density which limit parasite populations have not been undertaken. Hyperparasites of jack pine budworm parasites The role of hyperparasites in the dynamics of jack pine budworm populations has not been studied. Studies of other host-parasite-hyperparasite systems are relatively uncommon in the literature, and studies directed at determining the effects of hyperparasitism on host-parasite system dynamics are even less common. Notable exceptions include the work of Gutierrez (1970a, l970b, 1970c, 1970d), Gutierrez and Van Den Bosch (1970), Vansickle and Weseloh (1974), Weseloh (1978, 1979), Beddington and Hammond (1977), Ehler (1979), Stamp (1981), Morris (1976). The following important results have emerged from these studies: 1. Hyperparasites are important extrinsic factors limiting the effectiveness of parasites in biological control. 6. 15 Parasites are frequently attacked by complexes containing numerous species of polyphagous hyperparasites. Hyperparasite species within this complex exhibit variation in both their temporal and spatial patterns of attack. Host switching may be a common phenomenon for many polyphagous hyperparasite species (Hassel 1978). Host switching implies that the proportion of a host type attacked by a polyphagous hyperparasite changes from less than expected to more than expected. as the proportion of the host type available increases relative to other host types. Host switching can be viewed as an aggregative response whereby the hyperparasites allocate a greater fraction of their searching time to whichever microhabitats are most profitable (in terms of hosts attacked per unit time). Regardless of the behavioral mechanism which causes this phenomenon, the essential result is that host switching can be considered as merely an extension of the aggregative response where the hyperparasite spends more time searching in patches of high host density than in low density patches. The importance of hyperparasites varies greatly among different parasite complexes. The importance of hyperparasites varies among species of primary parasites within a given parasite complex. The effect of hyperparasites on the efficiency with which the parasite complex exploits the host population may depend on the precise nature of the interaction. For example, the presence of a facultative hyperparasite which exploits both the host and the primary parasite may increase the effectiveness of the parasite complex (Ehler 1979). 16 Hyperparasites can potentially be considered objects of control when developing strategies for parasite management. In order to evaluate this management option, principles must be developed which allow us to relate management actions to their probable effects on the dynamics of the host- parasite system under consideration. 17 LITERATURE CITED Allen, D. C., F. B. Knight, J. L. Foltz and W. S. Mattson. 1969. Influence of parasites on two populations of Choristoneura pinus (Lepidoptera: Tortricidae) in Michigan. Ann. Ent. Soc. Amer. 62:1469—75. Allen, D. C., F. B. Knight, and J. L. Foltz. 1970. Invertebrate predators of the jack pine budworm Choristoneura pinus in Michigan. Ann. Ent. Soc. Amer. 63:59-64. Batzer, H. O. and D. T. Jennings. 1980. Numerical analysis of a jack pine budworm outbreak in dense jack pine. Envir. Ent. 9:514-24. Batzer, H. O. and I. Millers. 1970. Jack pine USDA Forest Service Forest pest leaflet 7. Beddington, J. R. and P. S. Hammond. 1977. On the dynamics of host-parasite- hyperparasite interactions. J. Anim. Ecol. 46:811-21. Benjamin, D. M. 1954. The jack pine budworm in Michigan, 1953.USDA Forest Service, Lake States Forest Exp. Station, Tech. Note 414. Benjamin, D. M. 1965. Evaluation of outbreak populations of the jack pine budworm, Choristoneura m Freeman (Lepidoptera). Proc. XII Int. Cong. Ent., London, 1964:697. Benjamin, D. M. and A. T. Drooz. 1954. Parasites affecting the jack pine budworm in Michigan. J. Econ. Ent. 47:588-91. Dixon, J. C. 1961. The biology and ecology of the jack pine budworm in Wisconsin with special reference to insect parasites. Univ. of Wise.- Madison. Unpubl. Ph. D. Thesis. Dixon, J. C. and D. M. Benjamin. 1963. Natural control factors associated with the jack pine budworm, Choristoneura pinus. J. Econ. Ent. 56:266-70. 18 Drooz, A. T. and D. M. Benjamin. 1956. Parasites from two jack pine budworm outbreacks in the upper peninsula of Michigan. J. Econ. Ent. 49:412-13. Ehler, L. E. 1979. Utility of facultative secondary parasites in biological control. Env.Ent. 8:829-32. Foltz, J. L., F. B. Knight, and D. C. Allen. 1972. Numerical analysis of p0pu1ation fluctuations of the jack pine budworm. Ann. Ent. Soc. Amer. 65:82-89. Graham, S. A. 1935. The spruce budworm on Michigan pine. Univ. Mich. School For. and Cons. Bull. 6. Gutierrez, A. P. 1970a. Studies on host selection and host specificity of the aphid hyperparasite Charips vitrex (Hymenoptera: Cynipidae). 3. Host suitibility studies. Ann Ent. Soc. Amer. 63:1485-91. Gutierrez, A. P. 1970b. Studies on host selection and host specificity of the aphid hyperparasite M _v_it_r§ (HymenOptera: Cynipidae). 4. The effect of host on host selection. Ann. Ent. Soc. Amer. 63:1491-94. Gutierrez, A. P. 1970c. Studies on host selection and host specificity of the aphid hyperparasite Charips vitrex (Hymenoptera: Cynipidae). 5. Host selection. Ann. Ent. Soc. Amer. 63:1495-98. Gutierrez, A. P. l970d. Studies on host selection and host specificity of the aphid hyperparasite M 1i_t_r_e_x_ (Hymenoptera: Cynipidae). 6. Description of sensory structures and a synopsis of host selection and specificity. Ann. Ent. Soc. Amer. 63:1705-09. Gutierrez, A. P. and R. Van Den Bosch. 1970. Studies on host selection and host specificity of the aphid hyperparasite m _vjtre_x (Hymenoptera: Cynipidae). 1. Review of hyperparasitism and the field ecology of M m. Ann. Ent. Soc. Amer. 63:1345-54. 19 Hassel, M. P. 1978. The Dynamics of Athropod Predator-Prey Systems. Princton Univ. Press, Princeton, NJ. 237 pp. Heron, R. J. 1956. Jack pine staminate flower production. Bi-mon. prog, Rpt. Div. For. Biol. Sci. Serv. Can. Dept. Agr. 12:2. Heron, R. J. and R. M. Prentice. 1957. Jack pine budworm in pine plantations in the spruce woods reserve, Manitoba, in 1956. Bi-mon. Prog. Rpt. Div. For. Biol. Sci. Serv. Can. Dept. Agr. 13:2-3. Jennings, D. T. 1971. Ants preying on dislodged jack pine budworm larvae. Ann. Ent. Soc. Amer. 64:384-85. Kulman, H. M. and A. C. Hodson. 1961a. Parasites of the jackpine budworm, Choristoneura pinus, with special reference to parasitism at particular stand locations. J. Econ. Ent. 54:221-24. Kulman, H. M. and A. C. Hodson. 1961b. Feeding and oviposition habits of the jackpine budworm. J. Econ. Ent. 54:1138-40. Kulman, H. M., A. C. Hodson, and D. P. Duncan. 1963. Distribution and effects of jack pine budworm defoliation. Forest Sci. 9:146-57. Lejeune, R. R. 1950. The effect of jack pine staminate flowers on the size of larvae of the jack pine budworm, Choristoneura sp. Can. Ent. 82:32-43. Lejeune, R. R. and W. F. Black. 1950. POpulations of the jack pine budworm. For. Chron. 26:152-56. MacAloney, H. J. 1944. Relation of root condition, weather, and insects to the management of jack pine. J. Forestry. 42:124-29. Mattson, W. J., F. B. Knight, D. C. Allen, and J. L. Foltz. 1968. Vertebrate predation on the jack pine budworm in Michigan. J. Econ. Ent. 61:229-34. Miller, C. A. and T. R. Renault. 1976. Incidence of parasitoids attacking endemic spruce budworm (Lepidoptera: Tortricidae) p0pu1ations in New Brunswick. Can. Ent. 108:1045-52. 20 Miller, C. A. 1959. The interaction of the spruce budworm, Choristoneura fumiferanae (Clem.), and the parasite Apanteles fumiferanae Vier. Can. Ent. 61:457-77. Morris, R. F. 1976. Hyperparasitism in populations of Hyphantria cunea. Can. Ent. 108:685-87. Morris, R. F. and C. A. Miller. 1954. The development of life tables for the spruce budworm. Can. J. 2001. 32:283-301. Mott, D. G. 1966. The analysis of determination of population systems. In. Systems Analysis and Ecology. K. E. F. Watt (ed.). Academic, New York. Rose, D. W. 1973. Simulation of jack pine budworm attacks. J. Env. Man. 1:259-76. Rudolf, P. O. 1958. Silvicultural characteristics of jack pine (Pinus banksiana) USDA Forest Serv. Lakes States Forest Exp. Sta. Paper 61. Stamp, N. E. 1981. Behavior of parasitized aposematic catterpillars: advantageous to the parasitoid or the host? Amer. Nat. 118:715-24. Simmons G. A. and N. F. Sloan. 1974. Consumption of jack pine budworm, Choristoneura pinus, by the eastern chipping sparrow, Spizella passerina. Can. J. 2001. 52:817-21. Van Sickle, D. and R. M. Weseloh. 1974. Habitat variables influence the attack by hyperparasites of Apanteles melanoscelus cocoons. J. N. Y. Ent. Soc. 82:2-5. Walley, G. S. 1953. Hymenopterous parasites of Choristoneura pinus Free. (Lepidoptera: Tortricidae) in Canada. Can. Ent. 85:152. Weseloh, R. M. 1976. Behavior of forest insect parasitoids. pp. 99-110. In: Perspectives in Forest Entomology. Academic Press, New York. 428 pp. 21 STUDIES OF THE NATURAL HISTORY OF GLYPTA FUMIFERANAE AND APANTELES SPP. 22 INTRODUCTION Numerous studies have been done to determine the factors responsible for fluctuations of populations of the jack pine budworm (Choristoneura pinus Freeman) (hereafter referred to as JPB). Foltz (1972) and Batzer and Jennings (1980) constructed life tables and used components of variance analysis (Mott 1966) to elucidate the importance of survival of various age classes to total generation survival. In addition they determined the contribution of some specific mortality factors to stage specific mortality. Their studies could isolate no factor consistently accounting for a high proportion of the observed fluctuations in JPB p0pu1ations. Predators and parasites were thought to be important at low J PB population densities and in populations declining at the end of an outbreak. However, natural enemies were thought to be of limited importance in damping the release of populations to outbreak. The importance of parasites and predators has been established under certain ecological conditions. Thus, studies of the life systems of these organisms are warranted. Such studies could yield insight into those factors limiting natural enemy populations. Field studies on parasites and predators which include detailed analyses are rare (Hughes et al. 1984). This reflects the difficulties involved in carrying out such studies rather than their lack of importance. The first step in studying the life system of an organism is determining its life history characteristics and phenology. Then, the trophic relationships between it and other organisms in the community need to be determined as do the relationships to environmental factors (Hughes et al. 1984). 23 This represents the initiation of life system studies of parasites of JPB. Parasites that attack early instar larvae of JPB were chosen because theoretical work in biological control suggests that certain of these parasites possess biological characteristics often found in species that make successful biological control organisms (Beddington et al. 1978, Hassell 1977) and because of their apparent importance in the life system of JPB. The parasite species of primary interest in this study were Apanteles fumiferanae Viereck and Qtha fumiferanae (Viereck). Both species are univoltine, monOphagous, and are apparently well adapted to their host. The parasites emerge from 4-6th instars and pupate on jack pine foliage. Adults emerge and attack lSt and 2"d instars (Brown 1946a, 1946b, Miller 1959,1960). In this chapter the results of studies of host-parasite synchrony, adult food relationships, spatial and temporal activity patterns, impact on the hosts populations, and relationships with other parasite species are presented. 24 MATERIALS AND HETHODS Description of Study Areas The field studies were done at a number of locations in Michigan's Lower Peninsula during the summers of 1982, 1983, and 1984. In 1982 all studies were done at a single site (sec. 25, T26N, RIOW, Grand Traverse Co. (GR)). In 1983 studies were done at the GR site and an additional site (sec. 15, T24N, R9W, Wexford Co. (WE)). In 1984 the following four sites were used: see. 32, T24N, R4W, Crawford Co. (CR); sec. 21, T32N, R2E, Montmorency Co. (MO); sec. 11, T32N, RlW, Otsego Co. (OTl); sec. 20, T29N, RlW, Otsego Co. (0T2) (Figure 1). Study plots were constructed in the following manner: at each location a random point was selected along a stretch of road passing through a mature jack pine (Pinus banksiana Lamb.) stand; from the random point an azimuth was chosen randomly and followed a random distance (between 60 and 160 m); the point reached by this method was established as the plot center; a plot of radius ca. 25 m was established around the center point. The composition of the plant communities at each of the sites was similar but varied somewhat in the composition of minor species. All sites were predominantly jack pine with other tree species comprising less than 5 percent of dominant, codominant, and intermediate trees. Jack pine trees averaged between 9-12 m in all of the stands, and the trees were of mixed age classes. The composition of the understory vegetation varied from mostly conifers such as white pine (Pinus strobus L.) and red pine (Pinus resinosa Ait.) to predominantly oaks (Quercus sp.) and cherry (Prunus spp.). Ground vegetation at most sites was predominantly sweetfern (Comptonia peregrina (L.)), blueberry 25 , a, .1- .1-1.-.._- 1- 7-1-1.. _ ”In I— . . . m ..l- . .111 m . . _ 1.1.. 1.1.11... .111. Ten-iii... .w Iii _ m . _ Tun—”1111f . T.I1-.«1|1|. 1--- ir-L l a .1 . 1, m w.-.- WB J. .- _ ,1 .. m _ .111 I . a - . .-I-1 I-I _, --.--......---..H _. T-m- El- 1 -1- . 71-111- . x G .W "-11.11...- J. .1111. . I. .1 1.11110'5- . _ WI. _ _ .1 A _ .Tfill-l. _ -l..-I..J . .FIIIL illlIMII— . . m . w _ I. n _ Map of the locations of the study plots in Michigan's Lower Penninsula. Figure l. 26 (Vaccinium sp.), and reindeer lichen (Cladonia spp.) and various grasses; wild flowers such as sweet goldenrod (Solid_a_go odora L.), indian blanket (Gaillardia pulchella L.), harebell (Campamfla l‘Otlmdifolia L.), and rattlesnake weed (Hieracium venosum L.) occurred sporadically in Open areas within the stands. Meteorological Information Meteorological data used in the studies done in 1981 and 1982 were obtained from two sources. Daily rainfall, relative humidity, and average daily temperature was calculated from records compiled at the Cherry Capital Airport, Traverse City, MI. The airport was located ca. 15 km northwest of the GR site. Degree day accumulations were obtained from records maintained by the Cooperative Crop Monitoring System (CCMS) for the Lake City, MI weather station. Degree day accumulations for emergence of reared specimens were determined from minimum and maximum daily temperatures measured at a weather instrument shelter located adjacent to the Open air insectary. Determining the Relative Abundance of Parasites of Early Instar J PB Larvae Two methods were used to assess the relative abundance of parasites of early instar larvae of JPB: first, foliage sampling was done to obtain host material from which parasites were reared; second, parasite pupae were collected directly from foliage. In both methods samples were obtained by removing the terminal portion of branches at various heights in randomly selected trees within the study plots. Parasites that emerged were identified to species by comparison with known specimens. 27 During 1982 parasite pupae were sampled at intervals of 2-3 days from the date the first pupae were observed on foliage until all parasites had emerged as adults. Samples consisted of the terminal 46 cm of a single branch cut from the lower crown and mid crown of each of 15 randomly selected trees within the GR site. All parasite pupae were removed and held individually in l dram glass vials for emergence of adults. During 1983 host material was obtained from branches pruned from trees within the study plots. Host collections were made at time intervals which corresponded as nearly as possible with the peak occurrence of the 3rd, 4th, and 6th instars of JPB. At each sampling interval, larvae were obtained by removing the terminal 46 cm of a single branch taken from the lower crown and mid crown of each of 15 trees. All larvae on each branch were removed and placed in plastic cottage cheese containers (10-15 larvae per container) with 0.5 cm holes cut in the top; the holes were covered with fine cloth which permitted enough air flow to prevent excessive moisture from building-up within the containers. Freshly cut foliage was placed in each container daily; parasites that had emerged from hosts were removed, placed in l dram glass vials, and reared for emergence of adults. Parasite pupae were sampled as described above for the GR site in 1982. In 1984 a host collection was made at the peak of the 4th instar in the CR site by the same method used in 1983, and at the peak of the 5th instar at an additional site in Oscoda Co. (Parasite pupal collections were made in the CR, 0T1, 0T2, and MO sites by the following method: the terminal 60 cm of 3 branches were taken from both mid cron and lower crown from each of 10 dominant or codominant jack pine, from 10 suppressed jack pine, and from 10 28 suppressed white pine. Thus, each sample consisted of 120 branches with the exception of the 0T2 in which jack pine and white pine were not present in the understory and the MO site in Which White pine was not present in the understory. Parasites were reared as described above. Pupation and Emergence rates of parasite species Pupation and emergence rates of A. fumiferanae, A. morrisi, and g. fumiferanae were determined for parasites reared from a large sample of predominantly 4th instars during 1983 at the GR site. Accumulated degree days (base 8.900) were calculated from minimum and maximum temperatures recorded at a weather instrument station located adjacent to the insectary. Degree day accumulations (Baskervile and Emin 1969) for on-site data were begun on June 24 and added to the accumulated Degree days through June 23 recorded at the Lake City CCMS weather station. Teleral and Spatial Activity Patterns of Adult Parasites The seasonal, diurnal, and spatial activity patterns were studied during 1982 using Malaise traps (Nyrop 1982) to measure the relative densities of adult parasites over time and in various strata within jack pine stands. In 1982, 48 Malaise traps were positioned in trees in two plots within the GR stand; 24 traps were stationed in each of the two plots. One of the plots (plot A) was the same plot in which larval sampling was done. The second plot (plot B) was located ca. 300 m from plot A in a portion of the stand with a more diverse composition of tree species. In this area trees such as red oak (Quercus macrocarpa Michx.), white pine, red pine, cherry, and aspen (Pepulus tremuloides Michx.) made up 29 more than 10 percent of the total dominant, codominant, and intermediate trees. In each plot, a single trap was positioned in each of 24 randomly selected dominant or codominant jack pine trees; 12 of the traps were placed in the lower half of the crowns and 12 traps were placed in the upper half of the crowns. Traps were monitored daily at 0800 hours from June 23 through August 13. All parasites of early instar JPB larvae were removed and stored in 5 oz plastic vials; they were then returned to the laboratory and identified to species. Because it was not possible to make unambiguous determinations of male Aggteles, only female populations of the two species of this genus were studied. To determine the diurnal activity patterns of the adult parasites traps were checked bihourly (from 0800 through 2000 hours) for a period of 7 days during the period of peak seasonal activity of the parasites. In 1983 traps were placed in both the GR (32 traps in plot A) and WE (15 traps) sites. In the GR site, a single trap was placed in the upper half of the crown of each of 25 randomly selected trees. Seven traps were placed on the forest floor at distances of 10 m apart. In the WE site a single trap was placed in the upper half of the crown of each of 10 trees and 5 traps were placed on the forest floor. Traps were checked at 0800 hour from June 11 through August 22. Food Sources of Adult Parasites To obtain information on the food sources of _G_. fumiferanae and A. fumiferanae two tests were performed. In the first, parasites were exposed in a cylindrical 14.5 x 12 cm clear plastic container to 3 potential food sources. The food sources were the following: jack pine sprigs soaked with diluted honey, aphid honeydew on jack pine sprigs collected in the field, and bouquets composed 30 of three flower species (A. venosum, _S_. m, and Q. rotundifolia) which were collected in the field. These three flower species were the most common species occurring in the jack pine stand from which they and the parasites were obtained (site CR). The food sources were held at 40-5020 until the tests were done. All tests were conducted between 1000 and 1400 hours using naive parasites. The food material was used in tests within 48 hours from when it was collected in the field. In each repetition of the test, 6-10 parasites of a single sex and species were introduced into the container and observed at 15 minute intervals. At each interval the number of parasites observed on each food source and the number of parasites apparently feeding on a food source were recorded. Chi-square tests were used to determine whether the number of parasites visiting and feeding on particular food sources differed from what would be expected if there were no preferences for any of the foods. A second test was performed to determine the quality of honey and aphid honeydew as food sources for _G_. fumiferanae and A. fumiferanae. From 6-8 parasites of a particular sex and species were maintained in glass canning jars covered with fine mesh saran screen. Water was provided on moist cotton and jack pine needles coated with either aphid honeydew, honey, or distilled water were introduced to the containers daily. The number of parasites that died was recorded each day until all parasites had succumbed. Analysis of variance was used to determine if differences existed in average number of days parasites survived on the different treatments. 31 REUL’IS Relative Abundance of Parasites of Early Instar Jack Pine Budworm Larvae In Table 1 percent parasitism of 3-4 instar JPB in 4 cohorts is presented. Van Driesche (1983) cautioned against the use of estimates of parasitism based on a single sample as estimates of parasite impact per host generation. However, results to be presented later indicate that the estimates for the GR, WE, and CR cohorts are probably reasonably good estimates of impact of the respective parasite species since parasite oviposition was complete and neither hosts nor parasites had begun to exit the life stages sampled. For these three cohorts parasitism averaged 14.5, 8.9, and 3.6 percent for A. fumiferanae, A. M, and g. fumiferanae respectively. The estimates of impact for the other cohort was not reliable. Apanteles M was not reared from samples of 2-3 instar JPB collected when the larvae were in the flower mining stage. However, the species was present in later samples from the same cohort (Table 1). This suggests that the species parasitizes predominantly 3-4 instar JPB in the spring of the year unlike A. fumiferanae and _G_. fumiferanae which parasitize 1-2 instar larvae in the summer of the- preceding year. Reexamination of Agnteles reared from Q. m by Allen et al. (1969) indicated that A. M was absent from samples taken when JBP were in the flower mining stage (2-3 instars), but the species was present in collections of large larvae. Analysis of variance was used to determine if there were differences in the numbers of cocoons per dm2 of foliage surface area among the upper crown, lower crown, and suppressed trees for data collected at the CR, OT], and MO sites in 1984. There was no evidence that differences existed in the numbers of Table l. 32 Parasitism of third and fourth instar £- Pigus in samples from four cohorts. “mp." 9“" 3154 3174 3173 4166 4176 (Julian) Locouon on on V}: on 032040 0. Degree Days a .92; 194 475 455 625 659 (From April 1) Total number no,“ 176 244 141 1 10 55 Ami“ 20.6 23.0 13.9 22.2 10.6 22.6 19.1 23.7 30.9 26.2 Ansel-.212 mist 0 6.1 21.5 4.3 21.7 16.4 .35 14.5 24.7 91 p 9 3‘9 2.8 21.2 3.3 21.1 2.1212 5.5 22.2 9.1231 1111121121192: 33 parasites per dm2 surface area of foliage among the three strata for A. fumiferanae, A. morrisi, or _(_1. fumiferanae (Table 2). This does not mean that there were no differences in percent parasitism among strata since the host density in each stratum was unknown. The Phenologies of Parasites of Early Instar Jack Pine Budworm The base temperatures required for deve10pment of the parasite species are not known; however since degree day accumulations for similar base temperatures are highly correlated (Hughes et al. 1984), a base temperature of 8.9°C was used in determining developmental rates for the species, since 8.9°C is probably close to the actual base temperatures of the parasite species. The rate of pupation was estimated for A. fumiferanae and A. M from reared specimens collected at the GR study plot (Figure 2). Insufficient numbers of _G_. fumiferanae were present in the sample to provide a meaningful estimate. The average number of degree days necessary for pupation (and hence emergence from JPB larvae) for A. morrisi was 568 (se=10.0) and 593 (se=l2.l) for males and females, respectively; with an average of 582 (se=8.6) for males and females combined. For A. fumiferanae the corresponding values were 575 (se=l3.2) for males and 642 (se=l4.2) for females, with an average of 614 (se=ll.2) for both sexes combined. Adult emergence was estimated from the same sample (Figure 3). The mean adult emergence for A. _m_o_rri_si was 718 (se=7.4) for males and 715 (se=12.8) for females and averaged 716 (se=7.7) for the combined sexes. For A. fumiferanae adult emergence averaged 725 (se=9.8) for males and 806 (se=15.3) for females; for the combined sexes average adult emergence was 773 (se=12.0)- Thus, the peak emergence of adults of A. morrisi occurred before that for A. 34 Table 2. Results of three stage nested analyses of variance among sites, trees, strata (upper crown, lower crown, and suppressed trees), and branches. The dependent variables for each analysis were the number of parasites per dm2. foliage surface area. F-tests were used to determine treatment differences. $195.11! $9.911! ILL 55 E Am 5119 2 0.977 0.12 Tree (5110) 27 0.622 0.76 Stratum (Site,Tree) 60 2.776 0.34 Branch (Si te,Tree,Stratum) 160 7.967 - 9.1111111126111112 Site 2 0.238 0.06 Tree (5118) 27 0.900 0.23 Stratum (Site,Tree) 60 2.331 0.60 Branch (Site,Tree,Stratum) 100 3.91 1 - A-JDQEEISI Site 2 0.026 0.01 Tree (3110) 27 0.333 0.1 1 Stratum (Site,Tree) 60 1.126 0.36 Branch(31te,Tree,Stratum) 100 2.960 - 35 790 V\\\\\\.\\\ “was“ § 4.0.0.0...O-Od .4646464646464. m 4.. .6. 6666666666.. 7/ .464646464646464646464646464. h _ q u . u o w v n N .— vouoaaa g. . .1 3.3952 o p .. m .. n w _ o 665%.... g .V .9 LBEaz Accumulated Degree Days (base 8.9‘0) Frequency distribution of pupation of A. morrisi Figure 2. in relation 1’ and A. fumiferanae reared in an outdoor insectar t0 accumulated degree days base 8.90 C. 36 V «- md \ - male/ r ~ s . s. N 53 ..- .0. .O .0 .. .0 Number of A. W’ Emerged 4 a I O O ... ‘ a a" D‘O cod female ‘0..- male % 1323?: W '///////'//////I. Number of A. fumiferanae Emerged ’l/l. 'l/l. Will/Ill. N — 99219! 9! a V 6’6 6’6 6’6 6’6 6’6 \ \ \ start-asst: '96 [A 20.: 20; 20; :0; 20; L\ L\ O 660 700 740 780 820 860 900 940 Accumulated Degree Days (base 8.9‘0) Figure 3. Frequency distribution of emergence of A. morrisi and A. fumiferanae reared in an outdoor insectary. in relation to accumulated degree days base 8.9° C. 37 fumiferanae. Estimated emergence dates for Apgteles spp. based on percent of eclosed cocoons agreed well with the data for reared specimens (Figure 4); however, it was not possible to determine emergence for each species individually from the field data because cocoons of the two species could not be differentiated. Emergence of _G_.. fumiferanae was later than for Awteles spp. in field samples (Figure 4). Spatial and Temporal Activity Patterns of Adult Parasites Seasonal ActivitLPatterns of Adult Parasites of J PB Both A. fumiferanae and g. fumiferanae exhibited seasonal activity patterns that corresponded well with the eclosion of JPB eggs. During 1982 the females of Q. fumiferanae were active in the GR study plot for about 600 DD (950-1550 DD) and exhibited a peak in activity at about 1200 DD (Figure 5); the female parasites were active for a total of 36 calendar days. During 1983 the females of this species were active for about 675 DD (900-1675 DD) and exhibited a peak in activity at about 1220 DD (Figure 5); during this year the parasites were active for 34 calendar days. Peak activity of male _G_. fumiferanae preceded that of females. The peak in seasonal activity for females corresponded closely to the peak in the rate of eclosion of J PB eggs in both years suggesting that the parasite females are synchronized with the peak activity of lst instar JPB. It has been suggested that Q. fumiferanae is adapted for parasitizing 2nd instars in hibernaculae (Miller 1959, Lewis 1960). Thus, the species may not be well synchronized with the life stage which it is best adapted for parasitizing. It should be mentioned here that the estimates of activity were based on daily Malaise trap catch totals and numerous factors may influence 38 .0 l 1 982 0 Apanteles spp. e GL,waniferaruae 0.8 l 1 0.6 1 Proportion Emerged 0.4 1 0.2 J. .0 _L c, ‘ ‘ " ' 1 ° 1 1 1 575 675 775 875 975 1076 1176 Accumulated Degree Days (base 8.9°C) I l 983 0 Apanteles spp. 0 6L,fbwniferawuze 0-8 0:8 .4 Proportion Emerged 0-1 N O o :54 ' ’ 1 ' 1 1 1 1 575 675 775 875 97s 1075 1175 Accumulated Degree Days (base 8.9°C) Figure 4. Cumulative emergence of Apanteles spp. and G. fumiferanae in the GR site in 1982 and 1983. 39 O nnah: A female 82 4m .5 lglnwfle ~t::"“‘“HL .‘d 4 female ‘34 fl :5 '5 «o 2 1‘ E . ' s 4‘ '6 g o - 'I 5'1 ‘ i'r 1 T 1 ‘ “7 660 660 1060 1260 1460 1660 Degree Days (base 8.9‘0) '9 a“ 1983 ¢::fig::1 2.0 1-6 Catch/Trap/Day of G. Muifme 1.0 l '9 ‘3- O o. 1* I I f V 700 900 1100 1300 1600 1700 Degree Days (base 8.9‘C) Figure 5. Malaise trap catch of adult 6. fumiferanae at the GR site in 1982 and 1983 in relation to accumulated degree days base 8.9° C. 40 catch; thus it cannot be definitely concluded that the patterns observed in the daily trap catch totals were associated primarily with host searching and may in fact have been related to other activities such as mating and feeding. However, it is assumed in what follows that host searching parasites made up a major portion of those captured, but further studies would be needed to support this assumption. Seasonal activity patterns of A. fumiferanae were similar to those for g. fumiferanae but are slightly longer and begins somewhat earlier. In 1982 A. fumiferanae was active for about 700 DD (850-1550 DD) and exhibited a peak in activity at about 1240 DD (Figure 6). In 1983 the parasite was active for 750 DD (850-1600 DD) and exhibited a peak in activity at about 1175 DD. In both years the peak in activity coincided with the peak in the rate of eclosion of JPB eggs. This species is thought to be most efficient at parasitizing lst instars (Lewis 1960, Simmons personal communication) and the data reported here suggest that adult activity is well synchronized with the presence of lst instar J PB. Apanteles wig was infrequently caught in Malaise traps in 1982 in spite of the fact that more than half of the Aflteles emerging from field collected cocoons that year were A. M. The few that were caught appeared in the traps between 750-1100 DD (Figure 7). In 1983 traps were monitored from an earlier date than in 1982, and two peaks in activity were exhibited by females of A. g1_9r_ri_s_i, one at about 350 DD and another at about 800 DD (Figure 7). Previous workers have suggested that A. _1_n_<_>EI_s_i_ attacks the spruce budworm (Choristoneura fumiferanae) in the spring and probably overwinters in an alternate host (Mason 1975, Miller and Renault (1976). In conjunction with the estimates of percent parasitism presented above, the observed bimodal seasonal 41 ID -- 6 1°. ream D . «j 1932 "3 N- P 4 ed at N- '8 '9‘ fi- 0 g d ‘\ ‘2- 3... K e‘ 6 O- s -1 0 9 9‘1 1 ' 1 ' I ' ' 1 ~ 650 050 1050 1250 1450 1850 Degree Days (base 8.9’0) '9 c E .‘ ram 0 k of 1983 ‘ “3 Nd 4' =2: a N ‘6 “a: E c,- 2 ' K '2' '5 O- 3 =2 D V VI V I V 700 900 1100 1300 1500 1700 Degree Days (base 8-9'0) Figure 6. Malaise trap catch of adult female A, fumiferanae in relation to accumulated degree days base 8.9°C at the GR site in 1982 and 1983. 42 J 0.15 I 1982 10 0. 0.05 l Catch/Trap/Day of 9 A. morrisi 00 . ' -( °660 760 060 960 1060 1160 Degree Days (base 8.9‘C) 0230 ' 1983 0.20 0-10 C"1911/7 r99/ Day of 9 A. morrisi 0.00 V '— — I ‘- : 1 350 450 550 650 750 850 Degree Days (base 8.9‘C) Figure 7. Malaise trap catch of adult female A, morrisi in relation to accumulated degree days base 8.9°C at the GR site in 1982 and 1983. 43 activity pattern exhibited by the Species supports their suggestion; thus it appears that A. 1192111 has multiple generations and apparently one, or more, alternate hosts in which it overwinters, but JPB is apparently not, or only an infrequent host, for the overwintering stage of the parasite. The alternate host(s) of A. M151 are not known though the species has been reared from at least four different Lepidopteran species (Mason 1975). Mason lists Zelleria haimbachi Busck, which feeds primarily on jack and Ponderosa pines (Wilson 1977) as a possible host of A. M; during June and July of 1983 a total 92 larvae of A. haimbachi were collected from the GR and WE sites and reared for parasites. No A. M were among the parasites reared from the collected specimens. Daily Activity Patterns of Adult Parasites Females of A. fumiferanae were active throughout the daylight hours and were most active from 1200-1800 hours (Figure 8); however 11 percent of the total daily activity occurred between 2000 and 0800 hours indicating that the parasites were active after sunset and probably in the early morning, before 0800 h. Traps were emptied at dusk (2130 hours) and again at dawn (0600 hours) on two consecutive days and female A. fumiferanae were captured between 2130 and 0600 hours and a few were caught between 0600 and 0800 hours confirming that the parasites were active at night and at a low level in the early-morning hours. A similar pattern of diurnal activity was observed for g. fumiferanae, however even a greater proportion (28 percent) of the total daily activity occurred between 2000 and 0800 hours (Figure 8). Females of this species were also found to be active after dusk and again between 0600 and 0800 hours on the 44 .4 Female 0. fumiferanae 0.3 0.2 .1 .4 Male 0. filmiferanae 0.3 0.2 .1 .4 PROPORTION OF TOTAL DAILY CATCH Female A. fismiferame 0.3 0 .1 1‘ 0800 1000 1200 1400 1600 1800 2000 0800 TIME (hours) Figure 8. Average proportion of total daily Malaise trap catch of adult A, fumiferanae and Q, fumiferanae for seven time intervals within a day. The sample period was six days. The data was collected at the GR site in 1982. 45 two days traps were checked at those times. The daily activity pattern of male g. fumiferanae was similar to that for females but did not exhibit a high level of activity between 2000 and 0800 hours (Figure 8). Relationship of Temperature to Adult Parasite Activity Two factors likely account for the observed diurnal patterns of adult parasite activity: temperature and climatological variables closely correlated to temperature and behavioral mechanisms that produce cyclical diurnal patterns of activity that are relatively independent of temperature over a fairly broad range. Linear statistical models were fitted to the bihourly trap catch data and corresponding bihourly average temperature data to determine if temperature (and related variables) played a significant role in determining diurnal activity patterns over the limited range of temperatures (IQ-30°C) encountered in the field study. Models were fitted to the activity data for females of A. fumiferanae and g. fumiferanae assuming linear relationships between trap catch and temperature and assuming different slopes and intercepts for each of the lines relating trap catch to temperature for each two hour period from 0800 to 2000 (6 time periods). The models thus accounted for fixed differences in the overall level of activity (different intercepts), which would result from cyclical variation in behavior and average temperature effects, and for the response of the parasites to variation in temperature within each two hour time interval (different slopes). The regressions for the model described above were significant and accounted for 62 percent and 60 percent of the variation in trap catch for _G. fumiferanae and A. fumiferanae, respectively. By comparing the residual sum of squares of reduced models with the residual sum of squares of 46 the full models, as described by Neter and Wasserman (1974), it was possible to obtain further insight of the influence of temperature on daily activity patterns of the parasites. For _G_. fumiferanae a model with unique slopes and intercepts for the time intervals 0800-1000 hours and 1000-1200 hours, and a common slope and intercept for observations taken at time intervals between 1200-2000 hours was found to provide as good a fit to the data as the full model (Figure 9). The slopes for the 0800-1000 and 1000-1200 time intervals were greater than for the data from the afternoon for which the line is almost horizontal. This suggests that variation in temperature has a greater affect on the activity levels of the parasites in the early morning hours when the parasites need to increase body temperatures to levels necessary for flight. Similar results were evident for A. fumiferanae for which a unique slope and intercept was fitted for the 0800-1000 h time interval and data from all other time intervals were adequately described by a common slope and intercept (Figure 10). Relationship of Trap Height to Parasite Activity Apanteles fumiferanae and _G_. fumiferanae were trapped more frequently in traps placed in the upper crowns of trees than in traps placed in the lower crown and no individuals of either species was ever caught in traps stationed on the ground within the study plots. Regression was used to explore the relationship of trap catch of female A. fumiferanae and female _G_. fumiferanae to temperature and trap height. Both of the variables were related to trap catch, but there was no evidence of interaction among the variables in their effects on trap catch (Table 3). This suggests that the parasites stratified within 1— 47 44.0 + 0800—1000 h 4 1001—1200 h + x 1201—2000 h ’90? 4 .- I” feranae 1'- 3.0 I ’9. ll .0 01 CD 2-(1 1 l g I .Trap Catch of Female G. fum- 0 1192.1 1.0 T I' U i V I U I V I I’ I V 18.0 20.0 22.0 24-0 28.0 28.0 30-0 Average Temperature (°C) 0. Figure 9. The relationship between Malaise trap catch of adult female 6. fumiferanae and temperature for various time intervals within the day. Data collected at the GR site in 1982. 48 ‘2 90 § , + 0800—1000 n g 3- x 1001—2000 11 c: '7 x. -- ~ 6°) ‘1 $ E «’3 E ‘ ‘9 . c: *1: ,5- .9. (J 4' 5 , c: 14. a3“ .4... o d .C: «9 =2- (3 .4 O- 0.1:- 0:? L. ’— ‘3 c: ' l ' I ' 1 r r ' l ' '— 18.0 20.0 22.0 24.0 26.0 28.0 30.0 Average Temperature (’C) Figure 10. The relationship between Malaise trap catch of adult female A, fumiferanae and temperature for various time intervals within the day. Data collected at the GR site in 1982. Table 3. 49 ADV Tables for regression of Malaise trap catch of female parasites (separate regressions done for WM! and fl. fumiferanae) using the independent variables trap height (TH) and average temperature (AT) over the two hour interval between observations. F-tests were used to assess the significance of the regressions. 1" indicates a test of the full versus reduced model. 5111619.! W Sum Regression TH TH 11 TH AT TH 11 AT Residual Regression TH TH it TH AT Residual Total Regression TH TH 11 TH AT TH 11 AT Residual Regression TH TH 11 TH AT Residual Total 11 4 1 1 I 1 5.3 5.85 l 68.83 5.83 1 68.85 1 74.69 4.93 157.9 1 4.87 I 57.97 1 62.84 E 1 7.42“ 23. 17* 1 5.70” 20.67” I: 0.024 0.76 50 the tree crown and that the parasites did not migrate vertically within the crown in response to variation in temperature, but rather tended to be consistently more active within particular strata. To explore the relationship between trap catch and trap height in more detail, regression was used to relate the total catch per trap over a 29 day period from 15 July to 13 August, 1982 for the 48 Malaise traps located in the two study plots in the GR site. The overall level of activity did not differ among plots for females of A. fumiferanae or g. fumiferanae but did differ for male 9. fumiferanae. Thus, a single regression was used to relate trap catch to temperature for females, but the two sites were considered independently when fitting regression models for male 9. fumiferanae activity. For females of A. fumiferanae and g. fumiferanae total trap catch was found to increase with increasing trap height up to a height of about 8.5 m and then decrease with increasing trap height above this level. The trap catch versus trap height relationships were adequately described by quadratic equations for females of both species (Figures 11,12). For male 9. fumiferanae trap catch was approximated well by a linear relationship with trap catch increasing with increasing trap height; there was no decline in trap catch for traps placed high in tree crowns (Figure 13). The average height of jack pine in the two plots was 9.53 m (se=0.26), thus the activity of females of both species was greatest in the upper crowns of trees. Food Sources of Adult Parasites In preference tests females of g. fumiferanae preferred to visit and feed on jack pine shoots coated with honey over branches coated with the honeydew 51 CD 8 3.- c» "' =-1o.41+:5.:513x-o.20x2 5'; 2 g R =o.41 . 2: Q o + ++ :9 + + 2 (3 E a) d 11. s... O O _C N (J 4.; C) o . 05 C) _1 2, . 2 08.0 8.0 710 I 910 I 11.0 Trap Height (M) Figure 11. The relationship between Malaise trap catch of female A, fumiferanae and trap height in the crowns of jack pine trees. Data transformed to natural logarithms before analysis. Data coll- ected at the GR site in 1982. 52 Q) 6 {,3 Y=-5.07+2.03X-0.12X’ &~ 9) 2 9.. R =O.46 it?" + 4. + 1 1 Log, Catch of Female 0. fu 8.0 1.0 ' I ' 1 ' 1 1 1 5.0 740 8.0 12.0 Trap Height (M) Figure 12. The relationship between Malaise trap catch and the height of traps in the crowns of jack pine trees. Data transformed to natural logarithms before analysis. Data collected at the GR site in 1982. 53 J m plot A =24 + plot B " Rz=0.46 a 1 8-0 L, 4.0 8-0 1 l l L Sqrt. Catch of male 0. fumiferanae 0 Cd N J 4' ‘0‘ c, I . ‘ l V ' U I I °3.o 6.0 7.0 9.0 11.0 Trap Height (M) Figure l3. The relationship between Malaise trap catch of adult male g. fumiferanae and the height of traps in the crowns of jack pine trees. Data transformed to square roots before analysis. Data collected at the GR site in 1982. 54 of aphids which feed on jack pine, and only infrequently were observed visiting bouquets of wild flowers; they were never observed feeding on flowers (Table 4). The results of similar tests on A. fumiferanae were inconclusive because the parasites were lethargic and did not move about the arena. Only six of the parasites were observed feeding during the tests, but none of these was observed feeding on flowers though some sat motionless on the bouquets for long periods of time. The individuals of this species used in the tests had eclosed 6-10 days before the tests were conducted and may have been senescent, possibly accounting for their lack of response. Since _q. fumiferanae apparently does not utilize the forest floor it has probably not adapted to to use flowers as a food source; this would explain the lack of response observed in the tests. Females of Q. fumiferanae that were fed honey throughout their adult lifetime lived 8-18 days while males lived 7-14 days. Females of A. fumiferanae lived from 7-19 days. Parasites of both species and sexes that were maintained on honey lived longer than those fed honeydew, but those maintained on honeydew lived no longer than those which were not fed (Table 5). The honeydew used in the tests was produced by an unidentified species of aphid that occurred commonly on jack pine in the plot where the tested parasites were obtained. Honeydew from this aphid species is apparently of little or no nutritional value to the parasites. 55 Table 4. Comparison of numbers of 6. fgmifgranae visiting and feeding on three food sources. It Chi-square test was used to determine if the frequency of observed visitations and feedings differed from enpectations if the parasites showed no preference for any of the three foods. Sex repetition number of finite Chi-square flowers honeydew honey Female 1 5 33 43 28.7* 2 2 9 49 643* male 1 3 8 25 22 .2* 2 6 30 ‘2 258* Sex repetition number of feeding vieite Chi-square flower. honeydew honey Female 1 O l l 16 149* 2 0 2 22 337* male 1 O l 9 14.5* 2 0 5 20 260* 56 Table 5. Mean longevity (25E) of mgpta fumiferanae and Apantglgs fygifergngg fed various foods. Analysis of variance was used to determine treatment differences. Multiple comparisons were made using Scheffe's Method. Means with different superscripts differ at the 95% level. Species Sex Food Source n Mean Longevity (23E) 919mm Male None 6 ‘32 20.75 Honeydew 6 ‘28 20.41 Honey 6 b10.3. 22.3 Female None 7 '40 20.82 Honeydew 7 ‘44 20.53 Honey 7 I’12.? 22.87 on M lo 1 miter n Female None 8 a30 20.53 Honeydew 8 ‘3.1 20.64 Honey 8 t’13s. 24.31 57 DISCUSSION The role of parasites and predators in the life system of JPB has only been studied at a superficial level and firm conclusions regarding the influence of natural enemies on JPB p0pu1ation dynamics are difficult to make with the information presently available. The general consensus is that parasites and other natural enemies are important at low p0pu1ation densities and exert some control over endemic populations and perhaps speed the rate of decline of outbreak p0pu1ations already in the process of declining as the result of other factors such as insufficient food supply. Once an outbreak is triggered by an appropriate combination of environmental and biological factors, natural enemies exert little effect on damping or halting the release of populations to outbreak (Foltz et a1. 1972, Batzer and Jennings 1980, and others). However, there are many exceptions to this generality; populations in some geographic areas are remarkably resistant to J PB outbreaks. The factor or factors accounting for such phenomena are not known, though natural enemies may play a significant role in this resistance. Thus, more must be known about the biology and ecology of parasites and predators of J PB before firm conclusions can be formulated concerning the conditions under which natural enemies will be effective in controlling J PB populations, if in fact such conditions exist. Mortality of 3-6 instar JPB larvae is believed to be an important factor accounting for a large portion of total generation survival of J PB populations (Batzer and Jennings 1980). The most important source of this mortality is apparently associated with the 5-6 instars, with parasitism of 1-3 instars being of lesser importance (Foltz et al. 1972); but the factors accounting for mortality of 58 these life stages are poorly known. However, even the influence of parasitism of early instars, has not been established with certainty. Theoretical studies suggest that the parasite é. fumiferanae possesses many of the biological attributes associated with parasite species that are successful in biological control (Beddington et al. 1978, Hassell 1977), including monophagy and a high intrinsic rate of increase. In addition, the interactions of natural enemies with JPB, each other, and other components in their life systems are generally unknown. The results of this study indicate that the complex of parasites that parasitize instars 1-4 of JPB in outbreak and declining populations in Michigan was consistent, and composed of three species that were primarily responsible for the observed levels of for parasitism of these stages. Other parasites known to utilize 3-4 instar JPB were uncommon and of sporadic occurrence. Apanteles wig was not reared from 2-3 instar JPB sampled early in late May or early June when the larvae are mining flowers and new growth. However, the species was consistently reared from 4-5 instar JPB sampled later in the season. The bimodal pattern of adult female g. m activity suggests that the species parasitizes JPB in the spring and that parasite oviposition is complete by about 450-500 DD (base 8.9°C). At the on site in 1983 the JPB population was near the peak of the fourth instar at about 450-500 DD. To assess the role of a parasite as a mortality factor the percentage of susceptible hosts parasitized must be determined (Simmonds 1948, Van Driesche 1983). The life stages of JPB parasitized by Q. M are not known, though it is clear that the parasite must utilize 3rd instars, 4th instars, or both. If there is no differential mortality between JPB larvae parasitized by A. morrisi and those not parasitized by this 59 parasite, sampling to determine impact should be done at about 500 DD to obtain reliable estimates. If sampling is done earlier some parasites will still be ovipositing and if sampling is done later some parasites will have emerged from hosts; in both instances biased estimates of parasitism will result. The results of some studies suggest that spruce budworm larvae parasitized by A. fumiferanae and Q. fumiferanae suffer proportionally fewer losses to dispersal than unparasitized larvae because of a reduced propensity to disperse caused by the presence of the parasite or chemical substances injected into the host by the adult parasite during oviposition (Lewis 1960). However, other studies indicate that there are no differential losses of unparasitized and larvae parasitized by the two species (Miller 1960, Mcleod 1977). In any event it appears that A. _mo_rri_s_i parasitizes JPB after the major periods of larval dispersal, fall and spring, and differential mortality due to dispersal is not major a factor. Additional work needs to be done to determine if JPB parasitized by A. M are subject to differential mortality due to other causes. The results of this and other work indicate that reliable estimates of parasitism by A. fumiferanae and g. fumiferanae can probably be obtained at any point up to about 550 DD if parasitized and unparasitized larvae have equal probabilities of mortality to causes such as dispersal and predation. However, until more is known about the existence of differential mortality of unparasitized and parasitized larvae the dissection of overwintering larvae might be the most reliable method for estimating the impact of A. fumiferanae and g. fumiferanae (Kemp and Simmons 1976). The significance of parasitism by A. ELOLiSi. as a mortality factor operating on JPB p0pu1ations in Michigan is not known. Life table studies suggest that 60 unidentified sources of mortality Operating on 3-6 instars are significantly correlated with total generation survival (Foltz et al. 1972, Batzer and Jennings 1980). Parasitism by A. M made up at least a portion of this unidentified mortality since in both studies the sampling intervals were constructed in such a way that impact due to this parasite would go largely undetected. Miller and Renault (1976) studied parasitism in endemic p0pu1ations of the spruce budworm and found that parasitism by Apanteles species other than A. fumiferanae was generally low and sporadic, probably because most species have alternate and/or alternative hosts the abundance of which, in part, determines the abundance of these parasite species. The ubiquity of A. 9.9.21.5} in the populations surveyed in this study suggests that parasitism by this species may be an important source of mortality in outbreak and declining p0pu1ations of J PB. Further studies would be necessary to determine its influence in low density populations. The vertical patterns of adult female parasite activity observed in this study suggest that parasitism rates should be higher in the mid- and upper crowns of trees than in the lower crown or on suppressed trees provided that parasite activity levels measured by Malaise trap catch are related to host searching and other activities in equal proportions at all heights (e.g., the parasites do not use some strata for purposes of feeding, mating, etc. and others for host searching). Kaya and Anderson (1974) found that the catch of the parasite Ooencyrtus ennomgh_agu_s (Yoshimoto) on sticky panels in various microhabitats was consistent with observed patterns of parasitism by the species. Weseloh (1972) found that the distribution of the catch of Awteles melanoscelus (Ratzeburg) on sticky panels agreed with the distribution of parasitism, but that parasitism and trap catch of _q. kuwanai (Howard) gave contradictory results. Results of 61 previous studies have found parasitism by A. fumiferanae and g. fumiferanae to be higher in the mid-and upper crowns of trees (Jaynes 1954, Miller 1959, Kulman and Hodson 1961, Allen et al. 1969). These observations support the contention that much of the activity measured by Malaise trap catch was related to host searching. The observation that cocoons of A. fumiferanae and g. fumiferanae were were equally distributed among the upper crown, lower crown, and suppressed trees suggests that considerable redistribution of parasitized larvae occurs during larval dispersal periods in the fall and spring since little adult parasite activity occurred in the lower crown and virtually none occurred within 1 m of ground level. Apanteles morrisi cocoons were also equally distributed among strata. This suggests that this species parasitizes JPB larvae in all strata since adults of this species apparently oviposit after the majority of larval dispersal has occurred. The peak in activity of female populations of A. fumiferanae and Q. fumiferanae is well synchronized with the peak eclosion rate of JPB eggs. Whether or not the observed peak in trap catch coincided with the peak in oviposition activity of the parasite p0pu1ations is not known, but if so it suggests peak oviposition by A. fumiferanae is well synchronized with the peak in abundance of 1St instar JPB, the life stage it is apparently best adapted to exploit (Lewis 1960). Mason (1975) suggested the possibility that A. fumiferanae may oviposit in eggs of JPB. While this phenomenon has not been documented, the activity patterns of females of this species suggest that considerable activity occurs at a time when few lSt instars are available for oviposition. Utilization of eggs for oviposition could increase the reproductive efficiency of the parasite. 62 Glypta fumiferanae is most efficient at parasitizing 2nd instars in hibernaculae (Miller 1960, Lewis 1960). In laboratory studies this parasite begins to oviposit immediately upon emergence (Stairs 1983). If the observed peak in trap _catch of females of this species corresponds to the peak oviposition, primarily lSt instar JPB would be present in the population. In this case the parasite would be poorly synchronized with the life stage it is most efficient at parasitizing. Oviposition experiments with g. fumiferanae and A. fumiferanae indicate reproductive potentials of 100-150 eggs per female for both species (Miller 1959, 1960, Stairs 1983). The consistently low parasitism rates by Q. fumiferanae, as compared with A. fumiferanae, observed in field studies may be attributed in part to poor temporal synchrony with 2nd instar of JPB. However, other factors could be responsible for the low parasitism rates such as lower searching efficiency and longer handling time (Hassel 1977). There are many opposing pressures operating on p0pu1ations. The apparent lack of synchrony between E. fumiferanae and its host could result from the costs associated with increased mortality. Predation and hyperparasitism would play a role if more time were spent in the pupal stage. Increased losses could occur due to intrinsic competition with A. fumiferanae. In cases of multiparasitism, the ability of A. fumiferanae to inhibit embryonation of g. fumiferanae eggs increases sharply with the number of days that Q. fumiferanae precedes that by A. fumiferanae (Lewis 1960). At low host densities competition for hosts may be intense. The nd lack of synchrony of Q. fumiferanae with the presence of 2 instar JBP probably results from the necessity of optimizing a number of variables within its life system. 63 AMteles morrisi pupates and emerges slightly earlier than A. fumiferanae; both parasites spin silken cocoons and pupate on foliage within the feeding shelter of the host, or nearby. The species are morphologically similar and apparently closely related (Mason 1975). Because of their phylogenetic and ecological similarity they are probably subject to many of the same mortality factors such as disease, predation, and hyperparasitism. The presence of A. M could influence the probabilities of survival of A. fumiferanae. The earlier pupation and emergence of A. M could negatively influence the survival of A. fumiferanae from polyphagous hyperparasites which attack the parasites in the pupal stage; these hyperparasites might increase their host searching activities in the microhabitat where the parasites pupate in response to the increase in pupal density of A. mg, particularly when population densities of the parasite are high. The mechanism proposed to account for this behavior by the hyperparasites is host-switching Murdoch (1969). Hassel (1977) has discussed this phenomenon in reference to parasitic insects and suggests that it might be a common phenomenon. The increase in hyperparasite activity might be somewhat delayed as has been observed in for hyperparasites of the gypsy moth (Lymantria m L.) parasite Apanteles melanoscelus (Weseloh 1979). In this instance the host searching population might increase in response to A. _m_or_ris_i and host searching might be at a high level when A. fumiferanae begin to enter the pupal stage, resulting in increased losses to hyperparasitism. The adults of A. fumiferanae and Q. fumiferanae are active primarily during the daylight hours. Most parasites of forest insects are diurnal (Weseloh 1976). There is some indication that female Q. fumiferanae have two periods of increased activity, in the early morning and again at dusk. Juillet (1960) found 64 that ichneumonids were generally most active in the early morning and evening hours. A portion of the total daily activity of both species occurs during the night, but whether this activity is related to host searching or some other function is unknown. Much of the variation in daily activity is related to temperature, the effects of which vary and are superimposed on an apparent inherent rhythmicity. During the early morning hours parasite activity levels are very sensitive to temperature with slight increases in temperature resulting in marked increases in activity. Later in the day, activity levels are relatively insensitive to temperature over the range of temperatures observed in this study (19-3000). These results suggest that the relationship of parasite activity to temperature is nonlinear, as might be expected. Temperature exerts only a minor influence on parasite activity over a fairly wide range of temperatures, but at lower or higher temperatures activity may be rapidly reduced. When temperatures are high early in the morning total daily activity levels may be high because peak activity levels are reached earlier in the day. The insensitivity to variation in ambient temperature during mid-day may be due to the effects of radiant heat. The days during which this study was performed were all cloudless and calm. On sunny days the parasites may be capable of regulating their internal temperature via behavioral mechanisms (Matthews and Matthews 1978). Thus, the relationship of flight activity to temperature is not independent of other variation in the environment. The activity of Q. fumiferanae and A. fumiferanae was greater in the mid- and upper crowns of jack pine trees than in the lower crown, but was reduced in traps placed in the upper crowns of trees that were above the average height of the campy. Responses to physical factors (light intensity, temperature, 65 humidity, etc.), tree form, or host abundance may be responsible for the vertical distribution of parasite activity. The distribution of lSt instar JPB would be difficult to determine, but the population intensity of JPB egg masses is generally greater in the lower half of the crowns of trees in mature jack pine stands (Drapek personal communication). However, since the larvae are positively phototrOphic (Miller 1959,1960, Lewis 1960) the density of small larvae may, in fact, be greater in the upper half of crowns. It is clear that the spatial and temporal relationships of these parasite species are complex. In order to measure the activity of adult parasite p0pu1ations and relate this activity to parasitism in a meaningful way, additional information is needed on the relationship of reproductive condition of the parasite p0pu1ation, perhaps measured as ovariol content, to temporal patterns of activity of the populations. The feeding ecology of the parasite species should be studied in detail to determine if the availability of food is a limiting factor in determining rates of parasitism and under what ecological circumstances food is limiting. Results of this study indicate that the longevity of A. fumiferanae and Q. fumiferanae is increased by provision of adequate food sources. However, even though the parasites were observed to feed on aphid honeydew, this food had no apparent nutritive value to the parasites. Leius (1961a) found that the honeydew of two aphid species was detrimental to Scambts buolianae (Htg.), reducing both the longevity and fecundity of the parasite. Neither of the parasite species appear to be attracted to flowers and did not feed on them in tests. Furthermore, neither species utilizes the forest floor since no individuals of either species was ever caught in this stratum. However, it is possible that the parasites might make use of food sources occurring at ground level in 66 situations such as plantations where trees are small and the both the hosts and parasites occur in close proximity to the ground. Nyrop (1982) estimated that Q. fumiferanae lived an average of 10 days in the spruce-fir stands in which he studied the parasite. Q. fumiferanae appeared to live as long or longer in the populations studied here, and A. fumiferanae lived slightly longer than Q. fumiferanae. Thus, the parasites apparently do derive food from their environment. Flowers occurred only sporadically on the forest floor in jack pine stands studied, and to utilize nectar or pollen from them, the parasites would need to spend a large portion of their time foraging in a microhabitat spatially quite removed from that where their host occurs. The subsequent expenditure of energy, and other associated costs, such as predation risk and the risk of being lost to the system because of failure to successfully orient back to the tree crown, suggests that flower feeding might be an inefficient method for these parasites to obtain energy. It is probably not correct to assume that the parasites feed on all available food sources or that adequate nutrition can be obtained from any available food source. Many adult parasites have rather specific nutrient requirements and food preferences (Leius 1960, 1961a, 1961b, 1967, Syme 1974, Shahjahan 1974). The food sources utilized by A. fumiferanae and Q. fumiferanae were not determined during this study, however it was evident that they probably obtain food in the same microhabitat in which they search for hosts and that their food requirements may be rather specific. 67 LITERATURE CITED Allen, D. C., F. B. Knight, J. L. Foltz and W. S. Mattson. 1969. Influence of parasites on two populations of Choristoneura EM (Lepidoptera: Tortricidae) in Michigan. Ann. Ent. Soc. Amer. 62:1469-75. Baskerville, G. L. and P. Emin. 1969. Rapid estimation of heat accumulation from maximum and minimum temperatures. Ecology 502514-17. Batzer, H. O. and D. T. Jennings. 1980. Numerical analysis of a jack pine budworm outbreak in dense jack pine. Envir. Ent. 92514-24. Beddington, J. R., C. A. Free and J. H. Lawton. 1978. Modeling biological control: on the characteristics of successful natural enemies. Nature 273:513-19. Brown, N. R. 1946a. Studies of the parasites of the spruce budworm, Agghigs fumiferana (Clem.). 1. Life history of Agnteles fumiferanae (Viereck) (Hymenoptera: Braconidae). Can. Ent. 782121-29. Brown, N. R. 1946b. Studies of the parasites of the spruce budworm, A_rgh_ip_s_ fumiferana (Clem.). 11. Life history of Q_lyp£a_ fumiferanae (Viereck) (Hymenoptera: lchneumonidae). Can. Ent. 782138-47. Foltz, J. L., F. B. Knight, and D. C. Allen. 1972. Numerical analysis of population fluctuations of the jack pine budworm. Ann. Ent. Soc. Amer. 65282-89. Hassel, M. P. 1978. The dynamics of arthropod predator-prey systems. Princeton Univ. Press, Princeton, NJ. 237 pp. Hughes, R. D., R. E. Jones, and A. P. Gutierrez. 1984. Short-term patterns of population change. In: C. B. Huffacker and R. L. Rabb (eds.). Ecological Entomology. Wiley, New York. 844 pp. 68 Jaynes, H. A. 1954. Parasitization of spruce budworm larvae at different crown heights by Apanteles and Q1293; J. Econ. Ent. 472355-56. Juillet, J. A. 1960. Some factors influencing the flight activity of hymenopterous parasites. Can. Jour. Zool. 3821057-61. Kaya, H. K. and J. F. Anderson. 1974. Flight and ovipositional activity of the elm spanworm egg parasitoid, Ooencyrtus sp. Env. Ent. 321028-29. Kemp, W. P. and G. A. Simmons. 1976. Field instructions for dissection and rearing of spruce budworm larvae and pupae for parasite identification and determination of parasitism rates. Univ. Maine Agr. Exp. Stat. Miss. Rep. 180. Kulman, H. M. and A. C. Hodson. 1961. Parasites of the jack pine budworm, Choristoneura H, with special reference to parasitism at particular stand locations. J. Econ. Ent. 542221-24. Leius, K. 1960. Attractiveness of different foods and flowers to the adults of some hymenOpterous parasites. Can. Ent. 922369-76. Leius, K. 1961a. Influence of various foods on the fecundity and longevity of adults of Scambus buolianae (Htg.) (Hymenoptera: lchneumonidae). Can. Ent. 9321079-84. Leius, K. 1961b. Influence of various foods on the fecundity and longevity of adults of Itoplectis cormisitor (Say) (Hymenoptera: lchneumonidae). Can. Ent. 932771-80. Leius, K. 1967. Food sources and preferences of adults of the parasite, Scambus boulianae (HymenOptera2 lchneumonidae). and their consequences. Can. Ent. 992865-71. 69 Lewis, F. B. 1960. Factors affecting Assessment of parasitism by Apanteles fumiferanae Vier. and Glypta fumiferanae (Vier.) on spruce budworm larvae. Can. Ent. 922888-91. Mason, W. . M. 1974. The Apanteles species (Hymenoptera: Braconidae) attacking lepidoptera in the micro-habitat of the spruce budworm (Lepidoptera: Tortricidae). Can. Ent. 10621087-1102. Matthews, R. W. and J. R. Matthews. 1978. Insect Behavior. Wiley, New York. 507 pp. ' Mcleod, J. M. 1977. Distribution of ovipositional attacks by parasitoids on overwintering larvae of the spruce budworm, Choristoneura fumiferana (Lepidoptera: Tortricidae). Can. Ent. 1092789-96. Miller, C. A. 1959. The interaction of the spruce budworm, Choristoneura fumiferanae (Clem.), and the parasite Apanteles fumiferanae Vier. Can. Ent. 612457-77. Miller, C. A. 1960. The interaction of the spruce budworm, Choristoneura fumiferana (Clem.), and the parasite Qlyptg fumiferanae (Vier.). Can. Ent. 922839-50. Miller, C. A. and T. R. Renault. 1976. Incidence of parasitoids attacking endemic spruce budworm (Lepidoptera: Tortricidae) pOpulations in New Brunswick. Can. Ent. 10821045-52. Mott, D. G. 1966. The analysis of determination of population systems. In: Systems Analysis and Ecology, K. E. F. Watt (ed.). Academic, New York. Murdoch W. W. 1969. Switching by general predators: experiments on predator specificity and stability of prey populations. Ecol. Monogr. 392335-54. Neter, J. and W. Wasserman. 1974. Applied Linear Statistical Models. Irwin Press, Homewood, IL. 842 pp. 70 Nyrop, J. P. 1982. Studies related to the concept of pest-crop system design: 1) Adult parasitoid activity and its relation to biological control and 2) forest harvesting and the spruce budworm. Ph.D. Thesis, Michigan State Univ., East Lansing. Simmonds, F. J. 1948. Some difficulties in determining, by means of field samples, the true value of parasite control. Bull. Ent. Res. 39:435-440. Shahjahan, M. 1974. Erigeron flowers as a food and attractive odor source for eristenus pseudopallipes, a braconid parasitoid of the tarnished plant bug. Ent. 3:69-72. Stairs, G. R. 1983. Oviposition by Glypta fumiferanae (Hymenoptera: Ichneumonidae) in Choristoneura fumiferana (Lepidoptera: Tortricidae) larvae and the effect of honey on their fecundity and survival. Can. Ent. 1152689-92. Syme, P. D. 1975. The effects of flowers on the longevity and fecundity of two native parasites of the European pine shoot moth in Ontario. Env. Ent. 42337-46. Van Driesche, R. G. 1983. Meaning of "percent parasitism" in studies of insect parasitoids. Env. Ent. 12:1611-22. Weseloh, R. M. 1972. Spatial distribution of the gypsy moth (Lepidoptera: Lymantridae) and some of its parasitoids within a forest environment. EntomOphaga 172339-51. Weseloh, R. M. 1976. Behavior of forest insect parasitoids. In: Perspectives in Forest Entomology. J. Anderson and H. Kaya (eds.). Academic, New York. 428 pp. 7l Weseloh, R. M. 1979. Competition among gypsy moth hyperparasites attacking Apanteles melanoscelus, and influence of temperature on their field activity. Env. Ent. 8:86-90. Wilson, L. F. 1977. A Guide to Insect Injury to Conifers in the Great Lake States. USDA Agric. Hnbk. No. 501. 218 pp. 72 THE APPLICATION OF TIME SERIES ANALYSIS FOR PREDICTIN G ACTIVITY PATTERNS OF ADULT FEMALE Glypta fumiferanae POPULATIONS 73 INTRODUCTION Box-Jenkins modeling and forecasting techniques (Box and Jenkins 1976) have been in existence for many years, however only in recent years have the techniques been employed for modeling and predicting ecological processes. The number of published examples of the use of the methodology in the ecological and entomological literature are few (Poole 1972, Hacker et al. 1973a, 1973b, 1975, Poole 1972, Garsd and Howard 1982, Roubik 1983). Statistical forecasting techniques have been used extensively in economics, a field that is burdened with many problems similar to those in ecology. In economics, it has been found that so-called "naive" models of the Box-Jenkins type usually outperform the usual econometrics models in terms of accurately forecasting future trends (Cooper 1972, Makridakis 1976, Granger and Newbold 1975). It has been suggested that this class of models might also be more efficient than simulation models for predicting trends in ecological systems (Getz and Gutierrez 1982). The potential for the application of discrete time stochastic difference equation models of the Box-Jenkins type for investigating population processes has been discussed by several authors (Royama 1977, 1981, Pielou 1981, Tong 1983). Stochastic difference equation modeling techniques can be used to determine aspects of population dynamics in instances where the biological information available about a species is insufficient to warrant the development of more complex models. The resulting models, while simple in structure, can clarify the important features of population processes that are present in an observed time series. For example, they can provide insight into the existence 74 of density dependence or the strength of interactions among species in a community (Royama 1981, Poole 1976). All that is required in order to develop a Box-Jenkins model is a time sequence of measurements of a p0pu1ation process taken at regular intervals of time. For example, estimates of the annual density of a population of a univoltine insect taken over many years would provide the necessary data from which to build a time series model; if two or more processes were measured simultaneously, then the resulting time series would provide the data base for constructing a bivariate time series model. Furthermore, the models are relatively efficient in terms of computer and deve10pment time. While Box-Jenkins models have many desirable features they also have limitations. Perhaps the most important limitation is the requirement that the stochastic process being modeled be wide sense stationary. Wide sense stationarity means that all the first and second order moments of the process are independent of time. If a stationary stochastic process is the input to a linear time invariant system then the resulting output will be a stationary stochastic process. Most ecological processes are not linear and hence Box-Jenkins models can only serve as approximations of system dynamics. Thus, the models may not be well suited for data exhibiting sudden bursts of large amplitude such as the double equilibria processes studied by May (1977). However, linear models often provide robust approximations of nonlinear phenomena (Roughgarden 1975, Nisbet et al. 1977, Poole 1977, Patten 1975). Furthermore, many types of nonstationarity can be removed from the time series data to produce a process satisfying the stationarity criteria (Box and Jenkins 1976). Once a model is estimated for the stationary series the nonstationarity is reintroduced. For 75 example, nonstationarity in the mean and variance can sometimes be removed by the application of appropriate transformations. Thus, models for nonstationary processes can often be deve10ped using the Box-Jenkins methodology. The purpose of this study was to determine the usefulness of Box-Jenkins modeling techniques for a particular ecological time series; the series consisted of a sequence of data collected over a three year period on the daily Malaise trap catch totals of the Ichneumonid parasite, 9.121% fumiferanae (Viereck), which parasitizes the jack pine budworm (Choristoneura piAug Freeman) and the spruce budworm (Q. fumiferana (Clemens)). This series may be typical of many series obtained in studies of ecological systems in that it has a large variance and is nonstationary. Univariate models, and multivariate models incorporating meteorological variables known to affect parasite activity levels were developed. In the paper a detailed account of the analyses is given and the practical problems associated with the application of the modeling techniques is provided. The major features of the fitted models are described and the utility of the models for predicting future population trends is assessed. 76 CONSTRUCTION OF THE TIME SERIES MODELS Field techniques The data from which the parasite activity models were constructed and validated was generated at three study sites in Northern Michigan. The primary study site was located in the Fife Lake State Forest, Grand Traverse Co., 2 in a mature, Michigan. This site consisted of two study plots each ca. 2000 m naturally seeded jack pine stand. The plots were separated by a distance of ca. 200 m. A more detailed description of the site is given in Elliott (1985). A single Malaise trap (Nyr0p 1982) was secured in the mid- or upper-crown of each of 24 randomly selected dominant or codominant trees within the site (12 trees in each plot). The traps were monitored each day at 0800 hours throughout the adult life span of the parasite species, and the number of Glypta fumiferanae females caught per trap per day was recorded. The resulting data provided a time series of the relative activity of adult Q. fumiferanae over the two year period. In order to obtain a larger series of data for analysis, data from another p0pu1ation of Q. fumiferanae was obtained from Nyrop (1982). Nerp used Malaise traps of the same design as those used in 1982 and 1983, and operated them under similar conditions in two spruce-fir stancb in Delta Co., MI. Because of the striking similarity in the seasonal activity patterns of Q. fumiferanae in the two different locations and because the objective of this study was to explore the properties of stochastic difference equation models fitted to ecological data of this type (rather than constructing forecasting models for a particular population), it was felt that pooling data from the two sites for the 77 purpose of modeling was justified. The time series was constructed from 34 days of trap catch data for each year starting from the date of the first occurrence of non-zero trap catch for each year and continuing for 34 successive days. Thus a total of 102 observations on the relative daily activity of female Q. fumiferanae were available for analysis (Figure 14). A third study site was established in a mature jack pine stand in Wexford Co., MI during the summer of 1983. Ten malaise traps were secured in the mid- crown of 10 randomly selected jack pine trees within this stand. The trap catch data obtained from this study site were used to validate the fitted time series models. Environmental Data Meteorlogical data was obtained from the National Climatic Center, Asheville, North Carolina; hourly temperature measurements and rainfall measurements taken at six hour intervals were extracted from the daily records submitted by the Cherry Capital Airport, Traverse City, Michigan, located ca. 15 km Northwest of the Grand Traverse Co. study site; and by Sawyer Airforce Base, located ca. 30 km from the Delta Co. study sites. Hourly temperatures from 0700 to 2200 hours were averaged for each day for which parasite trap catch data were available. This was done because most of the daily activity of adult Q. fumiferanae occurred during this time interval (Elliott 1985). The temperatures recorded at Sawyer Airforce Base were generally lower than those from Cherry Capital Airport. The parasite populations in each of the two locations are probably adapted to the range of temperatures encountered at their particular geographic location. In order to avoid misrepresentation of the 78 TRAP CATCH o I _ _ v D Grd. Trvs. Co. 1982 Delta Co. 1981 Grd. Trvs. Co. 1983 Figure 14. Daily Malaise trap catch of adult female Glypta fumiferanae over a three year period. 79 effects of temperature on parasite activity in the transfer function models, temperatures needed to be standardized in some manner. Thus, the temperature data for each year were transformed by subtracting the mean of each year's data from each observation from that year and dividing by the standard deviation; the result was a time series of "standardized" daily temperatures (Figure 15). The number of cm of rainfall from 0700 to 2200 hours was also determined from the meteorological records (Figure 16). Statistical Methodology Box-Jenkins time series models (ARIMA models) were fitted to the daily parasite activity (trap catch) series. For the purposes of model identification the properties of the mean, variance, and covariance of the series were used to indicate stationarity (i.e., wide sense stationarity) rather than the strict definition (Granger and Newbold 1977), since the strict stationarity is not a useful definition in practice, when attempting to determine an appropriate model for an observed series. A two parameter power transformation was employed to instantaneously transform the observed series (Box and Cox 1964). A modification of a method described by Granger and Newbold (1977) was used to determine the power transformation parameters such that the likelihood function of the transformed process was maximized, assuming the process to be stationary and Gaussian. An algorithm was developed for this purpose and is described in Appendix I. A test described by Granger and Newbold (1977) was used to test for normality of the marginal distribution of the instantaneously transformed time series. 80 Standardized Temperature 0 I l I . . . 1982 o n c .1981 cm Tm CO Grd. Trvs. Co. 1983 e ° 0 Figure l5. Daily average standardized temperature over a three year period corresponding to the period of adult female Glypta fumiferanae activity. 81 a? 11 i v E '3 8, 1 c i .9 "d: I it": a .. .a 3 '6 ; 8 N: a a -3 ' Grd. Trvs. Co. 1982 Delta Co. 1981 Grd. Trvs. Co. 1983 Figure l6. Daily cm of precipitation over a three year period corresponding to the period of adult female Glypta fumiferanae activity. 82 Model identification, estimation, and checking were done as described by Box and Jenkins (1976) and Abraham and Ledolter (1983) using computer programs written by Dr. D. J. Pack of Ohio State University. The sample autocorrelations and partial autocorrelations were estimated from the transformed data for lags ranging from 1 to 36. Differencing operators were used to produce stationarity in the mean (Box and Jenkins 1976). The sample autocorrelations and partial autocorrelations were used to determine the orders of the autoregressive and moving average polynomials for the appropriately transformed and diff erenced time series. The sample autocorrelation function of the model residuals and statistical tests (Q-tests) based on the sample autocorrelation function of the residual series were used to check the adequacy of the fitted model (Ljung and Box 1978). Once an adequate model was identified and estimated more complex models were estimated to determine if they provided significant improvements. Once an appropriate ARIMA model was identified, estimated forecasts were produced. Both median point and mean point forecasts were produced. Median point forecasts were derived by backtransforming the forecast to the original scale, mean point forecasts were derived by a method described by Granger and Newbold (1976). The derivation of mean point forecasts for the univariate ARIMA model is outlined in Appendix II. The properties of the fitted model and the median and mean point forecasts were investigated by computing forecasts for various lead times for the data used in model estimation and for the independent data set (Wexford Co., 1983). Transfer function models (Box and Jenkins 1976) were fitted to the Q. fumiferanae activity series using the standardized temperature series and the 83 total precipitation series as input processes. Model identification, estimation, and checking were done by the methods described by Granger and Newbold (1977). The statistical properties of the fitted transfer function models were compared with those of the univariate ARIMA model. Model Development The time series consisted of 102 data points taken over a three year period (Figure 14). Inspection of the series suggested that it was nonstationary in both the mean and variance. The variance of the series was dependent on the mean as indicated by the significant correlation between the observations x(t) and the absolute value of x(t+1)-x(t) r2=0.381 (p< .05). The presence of a non- stationary mean was evident from the cyclical nature of the observed series. Nonstationarity in the mean of a discrete stochastic process can be removed by the application of an appropriate differencing transformation (Box and Jenkins 1976), however nonstationarity in variance cannot be corrected by differencing (Granger and Newbold 1977) and an instantaneous two parameter power transformation (Box and Cox 1964) was employed to remove that source of nonstationarity (Granger and Newbold 1977). The transformation parameters were chosen to maximize the likelihood function of the conditional distribution of the differenced process, X'(t), where X'(t)=(l-B)(l-Bs4)X(t) was assumed to have constant mean (see Appendix I). The resulting transformed series was given by, 0.3333 Y(t)= X'(t) /0.3333 84 (Figure 17). The correlation of y(t) and the absolute value of y(t+l)-y(t) was r2:- 0.155 (.10, 0,02) = (211087“ IMnli exp{-S/232} where, o, 9, , 0 , are polynomials in B of orders p, q, P, and Q respectively, s =t§ [E(e(t)| wu,t); 0,6, <2, 012 (2) and, 02 M‘;=az M’;(¢1,0,,0) is the variance-covariance matrix Of W01 ,t). Assume that the first d+sD Observations in the time series are fixed. Our 131 objective is to determine the joint probability density function of the It most recent Observations, x(n), x(n-l), . . . , x(l), Of the process conditional on x(0), x(- l), . . . , x(-c-sD+l). Given these assumptions the jacobian Of the transformation W(il.,t) = Vd V2 Y(2,t) = Vd VET(2,X(t)) is, BY(9’9t) I1 _ (e -1) 1 I 3x(t) "1.31 MM” 2 Thus, the conditional joint probability density function of the series X(t), t=1, . . . , n is h {x I y(O), . . . , y(-d-sD+1); 0, 0, 0, 0, 02, it} = (21“;th IMn I exp{-S/202})J Thus the likelihood function of X(t) is L = constant - inlno2 + iln IMn I - S/Zo2 + an (3) Assuming n to be moderately large, the term iln anl in equation (3) is small compared to S/Za2 (Box and Jenkins 1976); ignoring this term yields L' = constant - inlnaz - S/Za2 + an (4) which is the approximation of the likekihood function, given by equation (3), we will seek to maximize. Note that 62 = S/n is the solution Of the normal equation thus Obtained. In order to simplify equation (4) an additional transformation suggested by Box and Cox (1964) was applied to W(iL,t) to produce a "standardized" series 202,1) = W(9.,t)/J1/n Assuming x(l), x(2), . . . , x(n) to be a realization of the process X(t) we can write 0(9)0(BS)E{W(9.,1) lw; o, e, 0, 0, IL} = 0(B)0(BS)E{e(t) |w; 12,9, 0, o, 9.} l/ and dividing by J 11 yields 132 00312 (shat Z(9~,t)| z; 0, 6, 9, e, 9.} = 9(9)e(9'°‘)E{e(t)l z; 0, e, 2, e, 9.} where E[z(9~,t)| z; 11>, 6, 9, 9, £1 = E[W(9~,t)| 2,0, 6, 4, 9, 2] J"1/n E[e(t)l z; i, e, 9, 9, 21 = E[e(t)lW,¢,9, 9, 9.9.1 J'll" Thus, it is clear that the coefficients of the ARMA model for W(92,t) are identical to those obtained for Z(Ro,t). The sum of squares, S, of equation (2) is s = J"2/"sz where s =9 [B(eit)lw,¢,e,0,e,nn’ z t=l Thus, equation (3) becomes " = constant - 1lnln(J2/nsz) + InJ = constant - InlnSz (5) from which it is clear that the problem of maximizing the likelihood function, (3), is reduced to minimizing the sum of squares, Sz of the transformed time series Z(E,t). Estimation Of the Error Variance Assume that X(t) is a stationary, discrete, time series. Then the error variance of the best linear predictor, in the sense Of minimum squared error, is given by 0’ = expt-é—fi-l: 111(27 20))901 where “0) is the spectral density function Of X(t) (Kolmogorov 1941). Hannah and Nichols (1977) derived a family Of estimates of 022 )l ‘1 -i(m)} (a) A 2 l W’l 0 = max { — In W . m p M jzo (I (DJmi'k where M = [(n-l)/2m], ‘l’(m) = d In I‘(m)/dm 133 is the digamma function, and _ n . . w0j1= n *1331 x(tie‘"’.wj = 211/11. 0