NH \ \ 1 n W W. 1.le 1 t \ W! i \ 1 um FHESlF‘ This is to certify that the thesis entitled THE EFFECTS OF SPRING CLIMATE, SPAWNER ABUNDANCE, AND CANNIBALISM ON THE ABUNDANCE 0F RAINBOW SMELT (OSMERUS MORDAX) AT TWO SITES IN THE WEE-WEN WIRES presented by Robert Daniel Sluka has been accepted towards fulfillment of the requirements for Master of Fisheries and Science—degree in Midfie— 56%de Major professor Date _Augusi‘._9_._129L_ 0-7639 MS U is an Affirmative Action/Equal Opportunity Institution LIBRARY Michigan State l University PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES retmn on or before due due. DATE DUE DATE DUE DATE DU {N {19 7GB NUV 11142153“l ‘34;ng 0‘3 ' H l l l MSU Ie An Affirmative Action/Equal Opportunity Institution omens-9.1 THE EFFECTS OF SPRING CLIMATE, SPANNER ABUNDANCE, AND CANNIBALISM ON THE ABUNDANCE 0F RAINBOW SMELT (OSMERUS MORDAX) AT TWO SITES IN THE UPPER GREAT LAKES. By Robert Daniel Sluka A THESIS Submitted to Michigan State University . in partial fulfillment for the degree MASTER OF SCIENCE Department of Fisheries and Wildlife 1991 ABSTRACT THE EFFECTS OF SPRING CLIMATE, SPAWNER ABUNDANCE, AND CANNIBALISM ON THE ABUNDANCE OF RAINBOW SMELT (OSMERUS MORDAX) AT TWO SITES IN THE UPPER GREAT LAKES. By Robert Daniel Sluka Rainbow smelt are an integral part of the Great Lakes fish community. They are netted as a commercial fish in Green Bay and as a sport fish throughout the Great Lakes. Recent fluctuations in abundance have raised concerns about the stability of rainbow smelt population- levels. Data from the commercial catch in Green Bay from 1966 - 1989 were . analyzed for relationships between abundance, spawner abundance, and five biologically relevant spring climatic factors: air temperature, water temperature, solar intensity, wind speed, and wind direction. Young-of- the-year recruitment data from the United States Fish and Wildlife Service trawl survey at Alpena, Michigan 1975 - 1986 were also analyzed for relationships between these variables. However, the Alpena data are age structured so the effects of different age classes on recruitment could be analyzed. Stepwise multiple regression is used to quantify and model these relationships. The abundance of rainbow smelt in Green Bay was related to spawner abundance, the mean solar intensity in April and May one year previous, and the mean air temperature in April one year previous (R? - 0.73). The catch per unit effort of young-of—the-year at Alpena was related to spawner abundance, cannibalism by adults, and spring temperature (R2 == 0.82). Spawner abundance positively affected the production of eggs and larvae while the cannibalism negatively affected age 1 abundance. A warm, sunny early spring and a cool, overcast late spring affected the abundance of rainbow smelt positively. These models are useful for better management of the commercial, sport, and forage fisheries in the upper Great Lakes. ACKNOWLEDGMENTS This research was funded by the Michigan Sea Grant program, project number R/GLF 35. This research could not have been completed without the help of several researchers. Ray Argyle, Ed Brown, Guy Fleisher, and Ralph Steadman provided data from the United States Fish and Wildlife trawl surveys. Eva Moore provided commercial catch data as well as a friendly face or voice in Ann Arbor. Scott Nelson provided most of the above data on computer disks (for which I am very grateful). Fred Nurnberger at MSU provided climatic data as well as good advice on climate related issues. Ken Kunkel at the Midwest Climate Center provided large climatic data sets in a timely fashion. Water temperature data were provided by Bernie Vigue (Green Bay Water Utility), Jerry Plume (Alpena Wastewater Treatment Plant), Anne Stafford (Detour Village Water Department), Ron Marshman (Sheboygan Water Utility), Clark Creguer (Harbor Beach Water Utility), Bill Ward (East Tawas Water Utility), Jerry Kolaski (Ludington Water Utility). A special thanks goes to the members of my advisory committee. For Dr. Ramm, I have appreciated your commitment to verbal accuracy. Its always kept me on my toes in our conversations. To Dr. Taylor, thank you for you concern for my welfare both academically and personally. Our conversations about life have been much appreciated, even if I still don’t think that there are many paths to Heaven. Thank you for letting me be a part of your lab and the use of the equipment. It is much appreciated. For my major professor Dr. Winterstein, I thank you for your patience and your attitude. I’ve appreciated the times spent talking over the project and the good ideas and words of wisdom. I am indebted to my fellow graduate students for breaking me into the world of science. I’ve enjoyed the time spent in the lab with you all, even if the lab is too cold. Thanks to Andre for teaching me more about cross cultural relationships, to John, Missy, and Michaela for the good talks, and to Steve for the great passes underneath the hoop. Special thanks to Sue for putting up with me for two years in the office. I’ve loved the times spent talking about the world. Also to Paola for the special friendship that has developed. I look forward to seeing you down in the Caribbean! The extra special good guy and how could I ever have graduated with out him award goes to Russ Brown. Thanks so much for your desire to see me do things right, put up with my endless questions, and the junk food out in the field. Sorry I missed that last alley-00p. I’ve enjoyed our times. Don't get squeezed into the world’s mold. Remember to take time for the things that are most important: God, family and friends! The graduate and undergraduate InterVarsity chapters have been an invaluable source of the three F’s: Food, Fun, and Fellowship. I’ve appreciated the commitment of the grad chapter to integrating our disciplines and our faith. Thanks to Karen for her leadership and friendship. This fish-face loves you much! Thanks to my buds Rob and Carlos. Our times of prayer have been a highlight each week. Also to my roomie tree—Bob, its been excellent. A guy couldn’t ask for a better roommate ( who else wouldn’t care if you didn’t clean for a whole iv year?). To everyone in IV from South complex, its been fun. The last year has been great to get to know you all, especially Dave and Tolee my hoop buddys, and Laura (your an awesome v-nector betty). Special thanks to Geetha and Rose whose room became my haunt. Geetha you've always challenged me to think, even if we discuss loudly too much. Rose, what else can be said. You made the last 5 months of grad school wonderful! We’ll see you in the Caribbean! There’s plenty of other IV folk who deserve mention, but this acknowledgment would go on for ever. Thanks Rob for being the concert of prayer dude. Thanks to the friends at home who have pursued our friendships. Especially Shadd the fireman dude, Randy Bonser, and Mike Whitekus. Thanks to you all for coming up to visit and for having the time when I came home! Thanks also to Jamie the larval fish guru. You know how I feel. ' The best seems to always be last. One quick question, how can MSU be both an equal opportunity employer and affirmative action? Typical bureaucratic rhetoric that means nothing. Thanks to my family, especially Mom and Dad for their love and support. Without your sacrifices I couldn’t be here today. Thanks to Jon my bro for sticking with me all the time and his friendship. Thank you to my grandparents for modeling Godly lives to imitate. Biggest thanks of all goes to my Lord and Savior Jesus Christ who loved me, died for me, redeemed me, and will come back for me. Amen, come quickly Lord Jesus! TABLE OF CONTENTS Page LIST OF TABLES ....................................................... vii LIST OF FIGURES ..................................................... viii INTRODUCTION ............................................................. 1 STUDY SITES .............................................................. 4 METHODS .................................................................. 8 Commercial Production Data ....................................... 8 USFWS Trawl Survey Data ........................................ 13 Climatic Data ..................................................... 13 Model Development - Green Bay, Lake Michigan ...................... 18 Model Development - Alpena, Lake Huron ............................ 24 RESULTS ................................................................. 29 Green Bay, Lake Michigan ......................................... 291-. Al pena, Lake Huron ............................................... 34 DISCUSSION .......................................... i .................... 43 LITERATURE CITED ........................................................ 51 APPENDIX ............................................................... 56 vi LIST OF TABLES Page Table 1: Base period unit of effort and CPUE for each gear type used in the abundance index for Green Bay' .................... 11 Table 2: Climatic variables used for modeling, their biological importance to rainbow'smelt, their sources, and references ........ 15 Table 3: Data for’model calculations at Green Bay ....................... 31 Table 4: Multiple regression report for Green Bay model ................. 32 Table 5: Percent age composition of rainbow smelt CPUE at Alpena ........ 35 Table 6: CPUE of rainbow smelt in each age class at Alpena .............. 35 Table 7: Data for model calculations at Alpena .......................... 39 Table 8: Multiple regression report for Alpena .......................... 41 Table 9: Rainbow smelt abundance for each of 6 base periods at Green Bay ......................................................... 57 Table 10: Mean abundance index (x), standard deviation (s.d.), coefficient of variation (CV), and year for base period simulation data ....... 58 LIST OF FIGURES Page Figure 1: Location of Green Bay within Lake Michigan ..................... 5 Figure 2: Location of'Alpena site on Lake Huron .......................... 7 Figure 3: Location of statistical districts WM-l, HM-Z, and MM-l within Green Bay .................... 9 Figure 4: Rainbow smelt abundance index from Green Bay 1939 to 1989 ............................................. 12 Figure 5: Catch per unit effort of young-of-the-year rainbow smelt from Al pena 1975 — 1986 ..................................... 14 Figure 6: Flow chart for Green Bay model development ................... 22 Figure 7: Flow chart for model development at Alpena .................. 27 Figure 8: The detrended abundance index from Green Bay 1966 to 1989 ....................................... 30 Figure 9: Observed and predicted detrended abundance index for Green Bay, 1966 - 1989 ........................ 33 Figure 10: CPUE of age 1 fish one year previous - versus YOY CPUE for Al pena ........................................ 36 -. Figure 11: CPUE of age 1 rainbow smelt the year previous versus CPUE of age 1 rainbow smelt ............................... 37 Figure 12: CPUE of Age 1 rainbow smelt at Alpena 1975 - 1986 ............ 38 Figure 13: Detrended time series of young-of—the-year rainbow smelt for Al pena 1975 - 1986 ...................................... 40 Figure 14: Observed and predicted detrended YOY abundance for Alpena ........................................................ 42 Figure 15: Mean value of rainbow smelt abundance index for allbaseperiodscombined ......................................... 59 Figure 16: Rainbow smelt abundance index versus time for each base period used in the simulation ........................... 6O viii INTRODUCTION Rainbow smelt (Osmerus mordax) were introduced into the Great Lakes on April 6, 1912 in Crystal Lake, Benzie County, Michigan as forage fish for lake trout (Van Oosten 1937). Rainbow smelt were first noted on the eastern shore of Lake Michigan in 1922 (Creaser 1925) then spread rapidly to form detectable populations in Lake Huron (1925), Lake Superior (1929), and Lake Erie (1935) (Van Oosten 1937). Rainbow smelt are anadromous and spawn in the upper Great Lakes in late-April. The fish migrate up coastal streams or tributaries at night and return to the lake at daylight (Langlois 1935). The smelt shed adhesive eggs which attach to the stream bottom in shallow areas (Rothschild 1961). The larvae hatch out in late-May and begin to feed on rotifers. As the fish grow they begin to eat cyclopoid and calanoid copepods (McCullough and Stanley 1981). Rainbow smelt are netted commercially and are a sport fish dipped - during their spring spawning runs (Raab and Steinnes 1979). Rainbow smelt are also an important part of the forage base that supports a valuable salmonid fishery in the upper Great Lakes. Population levels of rainbow smelt in the Great Lakes have fluctuated greatly over the past 24 years with periods of large increases followed by major dieoffs (Baldwin et al. 1979). The fluctuations in the abundance of rainbow smelt have raised concerns about the stability of this forage base (Stewart et al. 1981). Rainbow smelt are a highly fecund fish subject to population fluctuations. In such fish, small changes in the survival of the early life stages could result in large fluctuations in the observed 2 recruitment (Sissenwine et al. 1988; Francis et al. 1989). These changes in the number of fish surviving can be caused by environmental factors (Van Winkle et al. 1979; Evans and Loftus 1987) which have their greatest impact on highly fecund fish (Larkin 1973). Spring environmental variables should be most important in affecting rainbow smelt abundance because of the deposition of eggs and hatching of larvae during this time. To understand the way that specific environmental factors affect fluctuations in fish populations we must examine the mechanisms underlying the relationship (Tautz et al. 1969). Dementieva (1973) suggested that the main factor governing survival of fish may be the direct influence of abiotic environmental factors. Environmental factors could affect the density of prey after larval hatching, an important factor for larval fish survival (McCullough and Stanley 1981).- Environmental factors also affect growth rates and thus time spent ; vulnerable to predation during a certain life stage (Evans and Loftus 1987). The number of fish recruited into the population may also be related to intraspecific biotic factors. Henderson and Nepszy (1989) showed that the abundance of yearlings in any given year was influenced by the abundance of smelt present during the previous year. This inverse relationship is most likely due to predation of young-of-the- year by one year old smelt, as yearling smelt did eat smelt fry (Henderson and Nepszy 1989). A similar inverse correlation was reported by Evans and Waring (1987) for rainbow smelt in Ontario. Additionally the number of recruits could be related to the number 3 of spawning adults. Walter and Hoagman (1975) found that the year- class-strength of rainbow smelt in Green Bay was positively related to spawning stock size for the years 1931 - 1967. The relationship between the adult stock and recruitment is usually related to the density of the eggs or larvae produced (Ricker 1975). The objective of this study was to evaluate the relationship between spring climatic variables, spawner abundance, cannibalism, and the abundance of rainbow smelt in Green Bay 1966 to 1989 and at Alpena 1975 - I986. Descriptive models were developed for commercial catch data from Green Bay and for United States Fish and Wildlife Service (USFWS) trawl data from Alpena. This will allow a better understanding of the mechanisms underlying recruitment of rainbow smelt. Thus managers of the commercial and sport fisheries as well as those managing the forage base for salmonids will have more information to make management decisions. STUDY SITES Rainbow smelt abundance data were available from two different sources, the commercial catch records and the United States Fish and Wildlife Service (USFWS) trawl surveys. Rainbow smelt have been an important part of the commercial fishery in Green Bay since the 1930’s (Schneberger 1937). The majority of rainbow smelt caught commercially in Lake Michigan are taken from Green Bay (Baldwin et al. 1979). This long history of commercial activity has yielded a series of catch and effort data from 1939 to 1989. The National Weather Service Office at Green Bay has been collecting meteorological data since 1948. The combination of detailed catch and weather data as well as the importance of rainbow smelt made Green Bay a good site to begin studying fluctuations in rainbow smelt abundance. Green Bay is located in the Northeast corner of Lake Michigan (Figure 1). The Bay itself is 118 miles long with a mean width of 23 miles and a mean depth of 65 feet (Ditton and Goodale 1972). The USFWS conducts trawl surveys throughout the Great Lakes. The USFWS data collected in Lake Huron at Alpena (Figure 2) are collected systematically and are age structured, allowing a more refined model to be constructed. The trawl survey data are available from 1975 to 1986. Climatic data are available from the National Weather Service station located at Alpena. The combination of detailed climatic observations and age structured biological data make Alpena a good site to model rainbow smelt abundance. Figure 1: Location of Green Bay within Lake Michigan. d u 1 H u c g g c o o o. O- 06 05 04 s 2 4 4 4 4 u u u c . . . . ................. Ow . I. lllllllllllllll L lllllllllllllll 1 lllll II I‘II II I II IIII. . a c u o - q o o . o . c o o n o c o o o u o . .1 O I I ---------r I I I I I I I I I I I I I I I I .C------‘--- O'Cocdqocouo -o--.‘----------- DoooocccT Oucooooobfloooocuoo -- --------P-- nuccooc y <50 . 0 w a Lak. .9 e Huron , Alpena Michigan Lake Michigan Figure 2: Location of Alpena Site on Lake Huron. METHODS Commercial Production Data The Great Lakes Fishery Commission (GLFC) has divided the Great Lakes into statistical district to facilitate easier and more ecologically meaningful data collection and analysis (Hile 1962). Green Bay is comprised of three statistical districts: WM-l and WM-2 in Wisconsin waters and MM-l in Michigan waters (Figure 3). Records for these three districts are available from 1939 to 1989 from the USFWS. The catch and effort are given for each gear type. I tested the null hypothesis of no difference among the three districts using parametric and nonparametric analysis-of—variance (ANOVA). Homogeneity of variance was tested using Bartlett’s test (Steel and Torrie 1980). There was no significant difference in the catch per unit effort (CPUE) variances for pound nets (1949 - 1971) between statistical districts WM-l and MM-l (X2 - 3.720; d.f. - 1; P > 0.05). Rainbow smelt are not fished using pound - nets in statistical district WM-Z. Differences in CPUE variances for 2 ; inch gill nets (1960 - 1971) among all three districts were not significant once data from the year 1967 were removed (X2 = 0.003; d.f. - 2; P > 0.05). The CPUE for 1967 was up to 100 times higher than the rest of the years. There were no significant differences in catch per effort among the three districts (ANOVA; F = 0.99; d.f. = 2,33; P = 0.3811). Since inclusion of the 1967 data indicated significant differences in the variances (X2 = 125.347; d.f. = 2; P < 0.05), a Kruskal-Wallis test also was used to test for differences among the CPUE for each district. There were no significant differences among CPUE (T = 2.601; P > 0.25). AS neither test indicated any differences among Figure 3: Location of statistical districts WM~1, WM-2, and MM-I within Green Bay. 10 Michigan Eeeenebe . MM-1 Leke Michigan Menominee. WM-1 .Green Bey 11 sites, the data were combined from the three districts into one set for Green Bay. A number of different gears were employed by the commercial fisherman. This necessitated the use of an abundance index to combine the different measures of effort. The GLFC standard abundance index was used (Hile 1962). The gears used in calculating the abundance index were 1 and 2 inch gill nets, pound nets, and otter trawls (Table I). A base-period mean catch per unit effort (CPUE) is calculated and then multiplied by the current effort to calculate an expected value of the catch. The ratio of the actual catch to the expected catch summed over all the gears is the abundance index. Simulation analysis indicated that the choice of the base period is not crucial as long as the time frame used to calculate the abundance index is the same for each gear (Appendix). The years 1975-1987 were designated as the base period because otter trawls were not used in the fishery until this time. Figure 4 displays the abundance index for Green Bay 1939 - 1987. Table 1: Base period unit of effort and CPUE for each gear type used in the abundance index for Green Bay. Gear type Unit of effort Base period CPUE ======================================================================= one inch gill net 1000 feet 246 pounds two inch gill net 1000 feet 25 pounds pound net one lift 9227 pounds otter trawl one hour 4314 pounds 12 140 120'- fl .4 O O I on O T 03 O I Abundance Index .h D N O 0 [[111 1939 1951 1963 1975 1987 1945 1957 1969 1981 Year Figure 4: Rainbow smelt abundance index from Green Bay 1939 to 1989. 13 USFWS Trawl Survex,Data The USFWS has been conducting fall trawl surveys of forage fish in Lake Huron since 1973. These surveys provide information on key forage species in the lakes (Argyle 1990). Size and condition of individual fish of the forage species are examined to assess the current status of the populations. The trawl data are collected systematically, with the standard unit of effort being a 10 minute tow of a 39 foot bottom trawl. Data are collected at a number of different depths and combined to give a total catch per unit effort (CPUE). Fish are separated into adult and young-of-the-year (YOY) categories based on their length (Brown 1989). All rainbow smelt under 100 mm are designated as YOY (pers. comm. Ed Brown USFWS National Fisheries Research Center - Great Lakes Ann Arbor Michigan). The CPUE of YOY is displayed in Figure 5. QM Climatic variables were chosen for analysiswhich have been proposed by prior researchers to affect the survival of rainbow smelt. The weather variables selected were hypothesized to affect rainbow smelt survival in the early life history. Thus for rainbow smelt these weather variables could be important during the spring months. Six weather variables from the months of March, April, May, and June were analyzed: wind speed, wind direction, solar intensity, mean cloud transmission, air temperature, and water temperature (Table 2). Air and water temperature can affect the timing of the development of larval fish relative to the availability of food resources (Ware 1975; Tin and Jude 1983). Also, the deposition of eggs by rainbow smelt in Maine was 14 .2500 12000 - L500 - CPUE L000 - 500 - ‘3 l f 1 l l I I, 1975 1977 1979 1981 1983 1985 Year Figure 5: Catch per unit effort of young-of-the-year rainbow smelt from Alpena 1975 - 1986. co_uuo.om ou_m mc_c:oam muuocwo xar .m «so. agenda: co_mm_5mcocu co_>agun nee >u_>_uoa mc_uoo. uuovwa >az .~ uao.u cane coucoo oume_.u umo:b_z ooo— ouuouceaed uco beacon m0uc30mog too. .0 co_uu:nOLQ on» uuowyo xoz .p >u_mcou:_ La.om muao benzene «coca cc .a>_>c:u ogu uneven >uz .~ co_uuoc_n uc_3 coucoo came_.u umozu_z moo. can: Loan: Lagoon soc» auco_cu:c ego! nou_>oca gu_:3 oc_..oza: umaau >ox .P pecan uc_3 ace.a acuiuoocp Loaoxuumoz ecoa.< maao we co_u_wooou ecu uuocco >ex .n asap ace: >u_._u: Laue: >ac ceeco «005:0moc $003 *0 co_uu:ooca ogu wooden >0: .~ noo— «now can c_» ocnuacoasou Loan: «ammo.0uu§_.u «unencuec pace *0 >u___no._a>a cu o>_ua.oc euaum cou_zu_x «hop acon.:: ga_w .e>cad we ace-no.e>on we uc_e_u on» uuoevo >ez .p ocauecoQEOu L_e oucaom eueo moucocoyoc u.elw sonc_e¢ cu ouceucoa§_ deu_ao.o_m odne_ca> .moucocoeoc ace .aooc:Oa c_o:u .udoae zonc_ec cu ouceacoas_ .eu_oo.o_n c_ocu .oc_.ou0I Lew new: mo_ne_ce> o_»es_.u “N e.nep mp 16 affected by water temperature (Hulbert 1974). Wind speed and direction can be related to upwelling effects in large bodies of water such as Lakes Michigan and Huron. Upwelling would provide more nutrients from the bottom of the lake and could produce a better food resource for the larval fish. Wind effects have been implicated in the mortality of shore spawned rainbow smelt eggs (Rupp 1965). Solar intensity and percent cloud cover are two measures of sunlight availability. Light intensity has apparently affected spawning site selection of rainbow smelt (Hulbert 1974) as well as the feeding activity and behavior of larval rainbow smelt (Bedard and Lalancette 1989). Primary production is related to solar intensity which should affect the availability of larval food resources. Climatic data for this study were collected by the National Weather Service office. Monthly average wind speed is recorded in miles ' per hour and wind direction in degrees. The wind direction data were transformed for linear regression because of their circular rather than linear nature (Batschelet 1981). The transformation formula used was: wdt . M + A*cos (wd - p) where X . A cos (p) Y = A sin (p) M = mean level or mesor (X,Y, and M are fit using simultaneous equations) A (x2 + Y2)l/2 p arctan (Y/X) if X>0 180 + arctan (Y/X) if X 0.88). Mean cloud transmission was not used in the final analysis because of this high correlation. Monthly average air temperature values in degrees Fahrenheit were obtained from the state Climatologist Fred Nurnberger located at Michigan State University. The air temperature values for 1988 - 1990 were obtained from the Midwestern Climate Center. Monthly average water temperature data were obtained from two different sources (Table 2). All posible pairwise plots of climatic variables and rainbow smelt abundance were examined. The pairwise plots supported the assumption of a linear relationship between climatic factors and abundance. 18 Model development - Green Bay. Lake Michiqan A unit change in an environmental factor may affect a fish population by a constant quantity (additive effect) or by a constant fraction or multiple (multiplicative effect). The effects of environmental variables on fish populations are generally thought to be multiplicative rather than additive (Ricker 1975). In a favorable environment all fish have a chance of benefitting, whereas in an unfavorable environment a certain fraction (not a fixed number) will be adversely affected (Ricker 1975). Thus the rainbow smelt abundance data (1948 - 1989) were examined using the following multiplicative model: sabun = mean x trend x season x cycle + error. Where sabun . the abundance index mean - the mean abundance index trend - the linear trend season = the differential affect on the abundance index of seasonal effects cycle a the cyclic component of the abundance index. An additive model was also examined but was not used as the multiplicative model proved to be more accurate. Data may exhibit trends that occur over the collection period. The effects of these trends should be removed before examining year-to- year variations in the data due to environmental effects (Ricker 1975). 19 The mean and trend of the time series were determined by regressing the abundance index against time, coded as 1 to 22. The regression residuals, which are now corrected for the mean and trend, were held for further analysis. These data were yearly, thus season a 1 and this component was dropped from the model. In the midst of this data analysis it became apparent that two different processes were occurring during the time period spanned by the data. The first 27 years of the rainbow smelt abundance time series were relatively flat, whereas the last 24 years exhibited large fluctuations (Figure 4). The data set was broken into two parts at 1966 because the fluctuations start about this time. Only the time series from 1966 to 1989 was examined further as my interest was in the variability in the abundance. Walter and Hoagman (1975) found that prior to 1967 the year class strength of rainbow smelt in Green Bay was . positively related to spawning abundance. Cursory investigations of the; data prior to 1966 confirmed this relationship. For this study, the detrended abundance index the previous year accounted for 85% of the variation in the detrended abundance index from 1948 to 1965 (F = 57.94; d.f. - 1,16; P < 0.001). Once the abundance index is divided by the mean and trend the information left should be cyclic and error: sabun/(mean*trend*l) = cycle + error/mean*trend*l One means of examining these cycles is through spectral analysis. A spectrum represents the range of possible frequencies occurring in the time series arranged by their relative importance (Platt and Denman 1975). Spectral analysis is used to determine periodicities in time 20 series data (Rao et al. 1984). It is "... a form of analysis of variance of a time series in which the variance of the series of numbers about their mean is partitioned into contributions at frequencies that are harmonics of the length of the data set (Platt and Denman 1975 p.191).' Then using, harmonic regression, a sine and cosine model can be fit to the data using the cycles determined in the spectral analysis (Reckahn 1986). To validate cycles in the data a time series needs to be at least 10 times the length of the cycle (pers. comm. Ray Assel Great Lakes Environmental Research Lab, Ann Arbor MI). The length of the most significant cycle calculated in this analysis (13 years) indicated that a longer time series would be needed to validate conclusions. Thus spectral analysis and harmonic regression were not used in the final analysis. Analysis of the cyclical properties of the rainbow smelt abundance‘ was continued using Box-Jenkins ARIMA methodology (Box and Jenkins 1976). This has been used in a number of studies to examine properties of fish populations (Saila et al. 1980 ; Jensen 1985; Quinn and Marshall 1989). Autoregressive and moving average models were examined for their utility. A moving average model seemed to be the most useful and theoretically sound, but the sample autocorrelation and partial autocorrelation plots did not give any conclusive evidence for a particular model. Time series analysis was not used in further in this investigation. The effect of previous smelt abundance (essentially what a moving average model would explain) could be analyzed in conjunction with regression analysis of the climatic data. This would allow analysis of 21 spring climatic factors and spawning abundance using one methodology. Multiple regression is one of the most robust techniques employed in the study of fish populations. This technique is not merely searching for correlations between the recruitment and environmental variables, but is a process whereby abiotic (or biotic) factors are selected based on predetermined criterion and then systematically tested for their significance. The final methodology was to analyze the detrended abundance index using multiple stepwise regression (Figure 6). The simplest model that explains the most variability in the data was selected. The addition of model parameters was stopped after a probability level of inclusion greater than 0.10 and/or none added more than ten percent to the R2 value. Events occurring previous to the year in which rainbow smelt were caught would have affected the fishes’ early life history. Rainbow smelt enter the commercial catch at age 1 (pers. comm. Kenneth Gepler University of Wisconsin-Stevens Point). Data collected by the USFWS during their fall trawl surveys in Lake Huron 1975 -1986 indicate that the CPUE of age 1 fish range from 28% to 82%. In five of the years age 1 is the dominant age class and on the average age 1 smelt make up the majority (52%) of the catch. Thus events one year previous could be expected to influence abundance. Data were lagged up to four years to thoroughly investigate the nature of possible time lags in the data. 22 Figure 6: Flow chart for Green Bay model development. 23 detrended abundance Index time series analysis sample and partial autocorrelatlo functions do not indicate a particular ARIMA model l ' multiple \ regression / ——_‘> Independen- variables + spectral analysis > V length of abundance time series too short to conclude about significance of cycles choose simplest model that explains the most variability using biologically relevant data MODEL 24 Model Development - Alpena. Lake Huron The development of the Green Bay model provided a framework to begin the analysis of the data from Alpena. The data were detrended by regressing the abundance index against time. Stepwise multiple regression was used to examine the relationships among variables. The same criterion and format were used to develop and select multiple regression models as described for the Green Bay analysis. Data were analyzed from 1975 to 1986 as there are age composition data for these years. The methodology is essentially the same as described for the commercial production data for Green Bay. However, since the data are apportioned to different age classes the spawning abundance variable can be defined by those age classes that are the majority of spawners. The dependant variable itself was refined by using YOY abundance as opposed - to adult abundance. The data are divided into age classes as YOY, age 1, age 2, ..., age 6. YOY is used as the index of recruitment. It has been shown that age 1 rainbow smelt are predators on YOY (Henderson and Nepszy 1989). Thus an intra-specific predation variable was designated as age 1 fish. Rainbow smelt reach sexual maturity by age 2 (pers. comm. Ralph Steadman USFWS National Fisheries Laboratory - Great Lakes Ann Arbor, MI). The effects of rainbow smelt ages 2 - 6 were examined as spawner abundance. The relationships between biotic variables were assumed to be linear. However, stock - recruitment relationships were examined for their utility in describing rainbow smelt recruitment. Both a Ricker and a Beverton-Holt stock - recruitment relationship were fit using adults as stock and YOY as recruits. Both of these 25 relationships are curvilinear. The Ricker model of stock and recruitment is outlined in Ricker (1975). The mathematical relationship is: R «elseBS where R = the number of recruits S = size of parental stock a = a dimensionless parameter 8 = a parameter of dimensions 1/S This is a dome shaped curve which assumes that at high parental stock size the level of recruitment will decrease. This type of curve is most appropriate if there is a severe mortality event during the early life history (Ricker 1975). The Beverton-Holt model of stock and recruitment assumes that at high parental stock densities there is an asymptotic relationship with recruitment. This tends to be imposed by a fixed level of available food or habitat or when predators adjust predation immediately and continuously to the abundance of prey. A full discussion of this model is presented by Ricker (1975). The mathematical relationship is: R = 1 / (a + B/S) where R and S are as given above a and B are new parameters Both the Ricker and the Beverton-Holt relationships are parameterized 26 using linear regression. They are put into a regressable form and then the regression coefficient (B) and intercept (a) are calculated. The stock - recruitment relationships were compared for their descriptive value using the highest R2 as a criterion. Recruitment was estimated from stock levels using the best stock - recruitment relationship and used in a multiple stepwise regression analysis with the climatic data (Figure 7). Time lags were not used in these models. The recruitment data are YOY and weather factors could only affect the recruits in the year they hatch. 27 Figure 7: Flow chart for model development at Alpena. Climatic Variables Blolo ical Aauu \ //vov ‘\ > Varia les CPUE ' i cpue L \l/ L i i Ricker Beverton - Holt ' l l t l I Choose Stepwise Best by + Multiple R2 Regression V Choose the simplest MODEL +—— model that explains the most variability RESULTS Green Bay, Lake Michiqan The mean of the abundance index was 37.745 and the trend 0.416 + 0.049 * Time, where Time = l to 22. Data for 1988 and 1989 were not used in calculations so that the utility of the model could be examined. The abundance data were divided by the product of the mean, the trend, and the season for each time period from 1966 to 1987. The linear time trend was significantly correlated to the abundance index having an F- ratio of 3.44 (d.f. = 1,20; P < 0.10; R2 = 0.15). Therefore: detrended abundance index = abundance index / (37.745 * (0.416 + 0.049 * Time ) * 1 ) = cycle + error The detrended abundance index was used for subsequent analysis (Figure 8, Table 3). The best model developed included three variables listed in order of inclusion into the model: detrended rainbow smelt abundance the previous year, solar intensity in April the previous year, solar intensity in May the previous year and the air temperature in April the previous year. The regression was significant (F),17 . 11.26 P < 0.001) with an R2 of 0.73 (Table 4). .The observed versus predicted values of the detrended abundance index are displayed in Figure 9. The observed abundance index decreased in both 1988 and 1989. In 1988 the difference between observed and predicted values was 153 percent. The difference between the model value in 1989 and the observed value was 33 percent. 29 30 3.5 i a— 'o E 025— /( o c B 2 r: :3 .0 <1.5— u o '0 :1- o h an! 80.5— olllllllllllLIILllllllllil 1966 1970 1974 1978 1982 1986 1990 Year Figure 8: The detrended abundance index for Green Bay 1966 to 1989. 00 m.0~ 0.0. 0.m.o 000.0 o.- 000. 00 o..~ ~.o. 0~0.o «.m.o o.on 000. 00 ..- 0... 0N..o 0~0.o «.mm .00. 00 o.- 0.x. mom.o 0~N.Q 0.0m 000. m0 ~.o~ n.m. .0m.~ noe.o 0.~0 «no. .0 0.0. o.m. coo.~ .0m.~ n..~. 000. .0 ~.o. o... 0...o woo.~ o.mo. mwo. 00 n.- 0.0. .0..o 0...o m.m ~00. m0 m..~ 0.m. .0m.o .0..o n.. .00. N0 ..o~ o.m. -0.o .0m.o 0.0~ 000. 00 o..~ 0.0. ....o -0.o o.m~ oko. 00 n.0m 0.9. m0m.o .~..o ..om 0.0. 00 ~.- 0.0. ~0~.. m0m.o ~.- sue. on ~.o~ 0.5. 00M.. ~0~.. o.~0 0.0. 00 n.0. u.0. 000.. 00m.. o.m0 on. 00 0... 0.0. 0mm.~ 000.. o.mm «so. 00 m.- n... m~0.. 0mm.~ 0.0. muo. 00 0..~ ~.o. 0mm.. m~0.. 0.00 mno. .0 ~.o. m... .00.o 0mm.. o.mm .so. 00 0..~ m.~. 00n.o .00.o o... 0.0. .0 0.0. 0.m. nm0.o 00m.o m.0 000. 00 n..~ o.0. 0.0.0 mn0.o ~.o 000. N0 o.- o... -~.c 0.0.0 o.m. N00. .0 0.0. m.0. nqu.o -~.o o.m 00o. coo> coo> cau> can» mao_>oca on. w:o_>oca as. mno_>oca on. m:o_>oca oz. xouc_ ._LQ< c. >0: c_ ..LQ< c. none. oucaocana xouc. coo» ocauocoaeo. >u_acouc_ >._mco.c_ oucoucsne poococuoo uoccvcan< 0.0 endow endow voucocuoo >00 c0000 .0 mco_uo.:u.nu .0005 Lo. 0000 "m 0.00. 32 Table 4: Multiple regression report for Green Bay model. Source df Sums of Mean Square F-ratio Prob. Squares level constant 1 22.047 22.047 model 4 8.102 2.026 11.26 < 0.001 error 17 3.058 0.280 total 21 11.161 0.531 R2 - 0.726 adjusted R2 = 0.662 Independent Parameter Probability Simple Cumulative Variable Estimate Level R R2 Detrended abundance 0.426 < 0.01 0.233 0.233 index one year previous Solar intensity in 0.400 < 0 001 0.106 0.400 April one year previous Solar intensity in -0.164 < 0.05 0.153 0.576 May one year previous Air temperature in -0.093 < 0.01 0.199 0.726 April one year previous 33 Observed Predicted — 5 2 5 1 axon... 00:00:19. 00.053006. 5 1970 1974 1978 1982 1986 1966 1976 1980 1984 1988 Year 1972 1968 Figure 9: Observed and predicted detrended abundance index for Green Bay, 1966 - 1989. 34 Alpena, Lake Huron The CPUE data from Alpena is age structured so the biological variables could be refined to estimate spawner abundance and intra- specific predation. As stated earlier rainbow smelt generally reach sexual maturity at age 2 and the majority of the catch are age 1 and 2. The spawning abundance data were divided into age classes based on the percent composition of ages in the catch (Table 5). The total catch of adults (age 1 - 6) were then multiplied by the percent in each age class to give total numbers of fish in each age class (Table 6). The relationship between total spawners (ages 2 - 6) and YOY is very weak (r = 0.050; n - 12; P > 0.85). The total spawning population (ages 2 - 6) was then used to calculate stock - recruitment relationships so that a better relationship could be developed. The Ricker stock - recruitment relationship had an R2 of 0.15 (a . 9.688; B . 0.005) while the Beverton- - Holt relationship had an R2 < 0.01 (a - 0.003; B = 0.083). The Ricker: relationship was included in subsequent analyses. There is a strong positive linear relationship between age 1 fish the previous year and YOY in the current year (Figure 10; r = 0.787; n = 12; P < 0.005). In those years that had high YOY (e.g., 1979, 1980, 1982, and 1986; Table 5) the percentage of age 1 the previous year dominate the catch. Predation of rainbow smelt by the year class the year before is evidenced by the negative correlation between age 1 fish and age 1 fish the previous year (Figure 11; r = -0.32; n = 12; P = 0.31). The effects of predation on YOY were examined by adding age 1 fish into the variable pool for stepwise multiple regression. The alternating effect of age 1 fish on the next year is shown in Figure 12. 35 Table 5: Percent age composition of rainbow smelt CPUE at Alpena. Year A e 1 A e 2 A e'3 Age 4 A e 5 A e 6 1975 28.1 49.7 19.1 2.7 0.3 0 1976 39.7 46.5 11.7 1.7 0.4 O 1977 35.9 40.4 18.8 4 0.8 0.2 1978 57.6 28 11.8 2.1 0.5 O 1979 77.7 16.7 5 0.6 0 0 1980 39 48.2 10.6 2.2 0 O 1981 59.6 30 9 1.2 0 0.2 1982 46.4 47.7 5.7 0 O 0.1 1983 67.3 17.4 14.3 1.1 0 O 1984 20.2 64.1 15.3 0.4 O O 1985 82.1 12.5 4.4 0.8 0.2 0 1986 29 65.9 4.4 0.7 0 0 Table 6: CPUE of rainbow smelt in each age class at Alpena. Year YOY Age 1 Age 2 Age 3 Age 4 Age 5 Age 6 Total Adults , 1975 600 107 190 73 10 l 0 381 1976 565 87 102 26 4 1 0 220 1977 166 159 179 83 18 4 1 442 1978 873 77 38 16 3 1 0 134 1979 1120 379 82 24 3 0 0 488 1980 1973 144 I78 39 8 0 0 369 1981 981 403 203 61 8 0 1 676 1982 1780 109 113 13 0 0 0 236 1983 1009 178 46 38 3 0 0 265 1984 339 39 125 30 I O 0 195 1985 404 229 35 12 2 1 0 279 1986 1568 101 230 15 2 0 0 349 36 2,500 2,000 - . .1 ll '5' 1,500 - O. 0 S e >_ 1,000 - 0 ,. l. 5 500 - 1. Ci Ci 0 l l l l 0 100 200 300 400 500 CPUE Age 1 One Year Previous Figure 10: CPUE of age 1 fish one year previous versus YOY CPUE for Alpena. 37 500 400 - 0 Ci I.l.l :3 300 — O. O 'F- i. ‘0’: 200 — '<: i. l. l. 100 ~ ’0 .' Ci 1. o l l l l 0 100 200 300 400 500 Age 1 CPUE the year previous Figure 11: CPUE of age 1 rainbow smelt the year previous versus CPUE of age 1 rainbow smelt. CPUE age 1 rainbow smelt 500 400 300 200 100 0 1975 1977 38 1 1 1 1 1 1 1 1 1 1 1 1979 1981 1983 1985 1976 1978 1980 1982 1984 1986 Year Figure 12: CPUE of Age 1 rainbow smelt at Alpena 1975 - 1986. 39 The effects of climate on recruitment were also examined prior to stepwise regression analysis. Mean daily air temperatures were summed for one week periods and correlated with YOY abundance using multiple regression. It was found that warmer temperatures prior to spawning were positively correlated to YOY abundance. However, about the second week in April the correlation became negative. This negative trend continues until mid-May. The YOY abundance time series had a mean value of 948.167 and a trend of 0.642 + 0.047 * Time, where Time - 1 to 12. Even though the linear time trend was not significantly related to the YOY CPUE (F = 0.74; d.f. - 1,10; P > 0.40; R2 - 0.07). The YOY time series was detrended to be consistent with the previous analysis of Green Bay (Table 7;Figure 13). All data were included in analyses as there were only 12 years of data available. Table 7: Data for model calculations at Alpena. YOY Detrended Age 1 Age 1 air temp. Year Abundance YOY abundance Tag 1 in April 1975 600 0.918 102 107 32 1976 565 0.810 107 87 40 1977 166 0.224 87 159 43 1978 873 1.110 159 77 36 1979 1120 1.348 77 379 35 1980 1973 2.254 379 144 38 1981 981 1.067 144 403 47 1982 1780 1.847 403 109 35 1983 1009 1.000 109 178 41 1984 339 0.322 178 39 40 1985 404 0.368 39 229 37 1986 1568 1.374 229 101 37 4O 2.5 2 .— LL! 3 0. 0 >. 1.5 — O >. 1: 0 '0 1 _ I: 0 h std 0 D 0.5 — o L l l l l l 1975 1977 1979 1981 1983 1985 Year Figure 13: Detrended time series of young-of-the-year rainbow smelt for Alpena 1975 - 1986. 41 The multiple stepwise regression analysis yielded a model with the spawner abundance (age 1 one year previous), predation (age 1 abundance) and the air temperature the second week in April significantly affecting YOY abundance (F - 12.13; P < 0.005; R2 - 0.82). The multiple regression report is given in Table 8. The observed value of detrended YOY abundance was plotted against the predicted values to determine goodness of fit (Figure 14). Table 8: Multiple regression report for Alpena. Source df Sums of Mean Square F-ratio Prob. Squares level constant 1 13.318 13.318 model 3 3.310 1.103 12.13 0.002 error 8 0.728 0.009 total 11 4.037 0.367 R2 . 0.820 adjusted R2 = 0.752 Independent Parameter Probability Simple Cumulative Variable Estimate Level R R2 CPUE of age 1 fish 0.005 < 0.001 0.619 0.619 one year previous CPUE of age 1 fish 0.002 < 0.10 0.004 0.731 Average air temperature the -0.049 < 0.10 0.125 0.820 second week of April 42 2.5 Observed .1 :1. Predicted 2 " ‘ 1 1 ..... 1. ' ' ,u. : : |.l.l 1' i :' '1 => " ‘. : : D. '1' “ : I - ; : O .‘ >- x : r U ” . .I 0’ 0) f 1 , .5 " 1 c: 1 _ , 1 0 ,’ 1 r i- , 1 r H p I I 0) ', ,’ o ..- ‘‘‘‘‘ 1' 0.5 - 0 1 1 1 1 1 1 1975 1977 1979 1981 1983 1985 Year Figure 14: Observed and predicted detrended YOY abundance for Alpena. DISCUSSION The results of analyses show that three different factors are important in affecting the abundance of rainbow smelt: spawner abundance, cannibalism, and spring climate. In the Green Bay model it was shown that spawner abundance and spring solar intensity were important. The analysis of the Alpena data showed that spawner abundance, predation, and spring temperatures were important. In both models the spawner abundance was positively related to abundance. This confirms prior evidence that the abundance of rainbow smelt in Green Bay is related to spawner abundance (Walter and Hoagman 1975). In the Green Bay model the simple R2 for this relationship was 0.23 and in Alpena it was much higher at 0.62. These relationships appear to be linear as the stock - recruitment relationships, which are curvilinear, were not as significant as the linear relationship. However, more data are needed to confirm the linearity of the relationship. The increase in the R2 value at Alpena is most likely due to the refining of the spawner abundance variable. As mentioned previously, rainbow smelt are sexually mature at age 2 and the majority of fish in the USFWS trawls were age 1 or 2. The commercial data for Green Bay are a combination of all age classes into one value. Thus at the Alpena site the value for spawner abundance is a better estimate than for the Green Bay site. The spawner abundance variable for Green Bay had a lag of one year. This is important as the age of entry into the fishery is at age 1 and the majority of the catch is age 1. Thus spawning events occurring one year previous most likely influence the early life history of most fish caught that year. Spawner abundance is generally thought 43 44 to influence the number of eggs deposited (Ricker 1975). Lett et al. (1975) found that stock biomass was important in determining egg abundance levels. Assuming a constant survival rate, if more eggs are spawned then more will survive. The number of age 1 fish one year previous was used as the spawner abundance at the Alpena site. It is not obvious why this might be a good choice, but it is highly correlated. One might expect that the number of fish age 2 - 6 would make a better predictor of spawner abundance. However, the correlation between this and the young-of—the- year is very low (r . 0.14). Schaefer et al. (1981) observed significant post spawning mortality of rainbow smelt in Lake Superior. Since smelt spawn in the spring, and the trawl survey data are from the fall, if there is significant post-spawning mortality of Lake Huron smelt the fall index of spawner abundance that year would be unreliable.- There is little evidence that there is a large over-wintering mortality. This suggests that the number of age 1 fish in the fall the year previous to spawning should be a good index of spawner abundance as it is a better representation of the number of fish spawning in the spring. The percent composition data (Table 5) support the use of age 1 fish the year previous as spawner abundance. In years of high YOY abundance the majority of the catch the year previous was made up of age 1 fish. Assuming that over-wintering survival is higher than post-spawning survival, the years with high age 1 composition would have higher spawning numbers the next spring. The number of age 2 fish would not be a good representation of abundance the next fall because of the post- spawning mortality. 45 The effects of older fish on YOY are not limited to increasing production of eggs. Cannibalism of YOY by adults has been shown on a number of occasions (Scott and Crossman 1973; Evans and Waring 1987; Henderson and Nepszy 1989). There is evidence from my study that cannibalism is affecting the abundance of the rainbow smelt. The data from Green Bay have all ages classes greater than one lumped together. This did not allow examination of the effects of predation on the abundance in Green Bay. At Alpena the effect of predation is seen on the age 1 fish. The graph of age 1 fish versus age 1 fish the year previous shows a negative relationship (Figure 11). There are four points in the graph which are separated from the rest of the data. The two points that reflect the highest age 1 CPUE (1979 and 1981 in Figure 12) have low predation and high spawner CPUE. While the two points that reflect the highest age 1 CPUE the previous year (1978 and 1980 in Figure 12) have high predation and low spawner CPUE. There seems to be a gradient from (1) high spawner CPUE and low predation which yields high age 1 CPUE to (2) low spawner CPUE and high predation which yields low age 1 CPUE. These two conditions cause extremes in the CPUE of age 1 rainbow smelt, whereas moderate values of spawner CPUE and predation yield moderate CPUE of age 1 fish. Predation has been documented to cause alternations in the abundance of rainbow smelt (Evans and Waring 1987; Henderson and Nepszy 1989). My results suggest an alternating year class phenomena for the age 1 data at Alpena (Figure 12). The analysis of the climatic data Showed that early warm temperatures or high solar intensity were favorable for increased abundance. However as the spring progresses towards May this 46 relationship becomes negative. A warmer, sunnier early spring could affect the early life history of rainbow smelt through an increase in the food available for the larvae. The amount of food available to larval fish can be a major determinant of survival (Freeberg et al. 1990). Larval rainbow smelt initially primarily feed on six kinds of rotifers, then begin feeding on cyclopoid and calanoid copepods as the larvae’s size increases (McCullough and Stanley 1981). The density of these prey could be important for larval survival (McCullough and Stanley 1981). With an early warm spring the density of these prey items could build so that survival of larvae are enhanced. One would most likely expect the positive trend to continue, with warmer, sunnier weather having a positive affect on smelt abundance. However, the data show the opposite effect. Solar intensity in May could have negative effects on abundance either directly from the light - intensity or from events occurring as a result of behavioral changes induced by the light. Dementieva (1973) suggested that in some cases the main factor governing survival of fish is the direct influence of abiotic environmental factors. It is possible then that, with high solar radiation during the first weeks after hatching, the larvae cannot survive the increase in water temperature or the amount of direct exposure to the sun’s rays. High temperatures have been indicated as causing mortality of rainbow smelt (Shaefer et al. 1981; Buckley 1989). Bedard and Lalancette (1989) found that larval rainbow smelt are positively phototactic. At high solar intensity the larvae could be drawn to the surface where they might experience the direct effects of solar intensity or temperature. 47 The air temperature could also be affecting the growth and timing of the hatch of the larvae relative to food resources (O’Gorman 1983). The yolk sac of larval smelt is absorbed in a few days and so the transition to exogenous feeding must be made fairly soon thereafter (Blaxter 1965; O’Connell and Raymond 1970; Blaxter 1971; Akielaszek et al. 1985). The time the larvae is dependant on the yolk sac can also be dependant on temperature (Braum 1967). This type of effect could be interacting with the effect of air temperature earlier in the year to change the length of time eggs are incubating and thus the size of larvae produced. Braum (1967) found that eggs incubated at higher temperatures produced larvae smaller than those at lower temperatures. The negative relationship between late spring temperatures or solar intensity and YOY could be related to this growth effect on larvae. Larvae which hatch at a larger size have a better competitive and predatory avoidance advantage over smaller larvae. Further research is necessary to conclusively explain the unexpected change in the sign of the regression coefficient between April and May solar intensity at Green Bay. However, the possibility that the difference is an artifact of the analysis (e.g. reflecting multicolinearity) can be tentatively rejected. The correlation between monthly solar intensity values were low. In particular the correlation between solar intensity in April and May was only 38%. Both of the models presented are significant, describing 73% and 82% of the variability in the abundance of rainbow smelt in Green Bay and Alpena respectively. However in each model there are years in which the predicted values do not match the observed values very well. In 48 Green Bay the years 1973, 1983, 1984, and 1988 are not accurately described by the model (Figure 9). In 1973 there is the largest value of the abundance index over the time period examined. However, there was a warm late Spring in 1972 which would be expected to have negative affects on the abundance. In 1983 and 1984 the model underestimates the values of the abundance index. This failure of the model to fit the large peak in 1983 is due to low spawner abundance in 1983. A cold early Spring in 1983 should have caused a decline in abundance in 1984, but the abundance increased. In 1988 the model overestimates the abundance index. The previous year there was a very warm early spring indicating a good year for survival. However, the actual abundance index continued to decrease. Factors such as predation by salmonids, and competition with other forage fish could account for the discrepancies in the model. However, these factors were not examined 50- all of the variance in the rainbow smelt abundance cannot be described. - The Alpena model is better at fitting the peaks and troughs of the YOY abundance (Figure 14). In 1984 the model overestimated the abundance. This was due to a higher than average spawner abundance in 1984 which would indicate an increase in YOY CPUE when there was actually a decrease. In 1984 there was also a cooler than average late spring which would indicate an increase in the YOY CPUE in 1985. However, the CPUE stayed lower than the model. These discrepancies in the model could also be accounted for with factors not included in the model such as competition or predation. In addition to the factors mentioned above there are other sources of variance that could not be examined. At Green Bay the commercial 49 catch data have been collected for a longer period, but there could be errors associated with the reporting procedures. The commercial fisherman may not have carefully recorded effort data resulting in less accurate values of the abundance index. The Green Bay data are also not age structured. It was shown in the analysis of data from Alpena that this is important for understanding more fully the effects of adult fish on recruits. The effects of other fish competing or preying on rainbow smelt were not analyzed. The inclusion of salmonid predators and competitors such as alewives into future models would increase our understanding of the dynamics of rainbow smelt throughout their lifespan. This study focused on the early life history. Future research should seek to include dynamics from all life stages into the modeling process. These models are useful for those managing the commercial, sport, and salmonid fisheries of the upper Great Lakes. The information gained‘ from knowledge of the factors affecting abundance will increase our understanding of the early life history of rainbow smelt. This will help managers of the smelt commercial fishery to set regulations. The salmonid fishery can also benefit from this information by adjusting the stocking rate of salmonids based on future predictions of the forage base. The salmonid harvest could also be regulated so that in years when the forage base is low more could be harvested so that those remaining would have sufficient food. This information is also useful for the USFWS personnel conducting their trawl surveys. These models are also useful for guiding future explorations into the factors affecting the abundance of rainbow smelt. Clearly the 50 relationship of spawner abundance to recruitment is important and should be more thoroughly investigated. The analysis of data from Alpena indicated that predation affects the abundance of rainbow smelt. The timing and nature of this predation should be examined in more detail. The different effects of over-winter mortality and post-spawning mortality should be examined to determine what percent of the annual mortality can be attributed to each. The relationship of spring climate to rainbow smelt populations should be investigated further. This study showed that there is an interesting dynamic occurring between early and late spring climatic events. The nature of this relationship should be pursued further. It is possible that the rate of warming is affecting the rainbow smelt populations. Characterization of this relationship will yield further insight into the early life history of rainbow smelt. The relationship between food production and climate should be examined ' further. The collection of zooplankton samples along with abundance data would allow a better analysis of the feeding characteristics of the larval smelt. This could be examined in relation to the spring climatic events producing the zooplankton levels. Future field studies should be able to better quantify the relationships between rainbow smelt abundance, spawner abundance, predation and climatic factors. These studies will give better understanding of the early life history of rainbow smelt and those fish which are affected by its presence in the Great Lakes. LITERATURE CITED LITERATURE CITED Akielaszek, J.J., J.R. Moring, S.R. Chapman, and J.H. Dearborn. 1985. Experimental culture of young rainbow smelt Osmerus mordax. Trans. Am. Fish. Soc. 114:596-603. Argyle, R. 1990. Status of forage fish stocks in Lake Huron 1989. Great Lakes Fishery Commission, Lake Huron Committee Meeting, Ann Arbor, Michigan. 6 pp. Baldwin, N.S, R.W. Saalfield, M.A. Ross, and H.J. Buettner. 1979. Commercial fish production in the Great Lakes 1867-1979. Technical report no. 3, Great Lakes Fishery Commission, Ann Arbor, Michigan. Batschelet, E. 1981. Circular Statistics in Biology. Academic Press, New York. 371 pp. Bedard, D. and L.M. Lalancette. 1989. Comportement des larves d’eperlan, Osmerus mordax, en fonction de l’intensite lumineuse, du courant d’eau et du type de nourriture. Canadian Field Naturalist 103:75-79. Blaxter, J.H.S. 1965. The feeding of herring larvae and their ecology in relation to feeding. California Cooperative Oceanic Fisheries Investigations Report 10:79-88. Blaxter, J.H.S. 1971. Feeding and condition of Clyde herring larvae. Rapports et Proces-Verbaux des Reunions, Conseil pour L’Exploration de la Mer 160:128-136. Box, G.E.P. and G.M. Jenkins. 1976. Time Series Analysis: forecasting and control. 2nd ed. Holden-Day, San Francisco, CA. 575p. Braum, E. 1967. The survival of fish larvae with reference to their feeding behavior and the food supply. In: The Biological Basis of Freshwater Fish Production. Ed. S.B. Gerking, Blackwell. pp. 113-29. Brown, E.H. Jr. 1989. Status of bloaters, alewives, rainbow smelt, slimy sculpins, deepwater sculpins, and yellow perch in Lake Michigan. Great Lakes Fishery Commission, Lake Michigan Committee Meeting, Sault Ste. Marie, Ontario. 7 pp. Buckley, J. 1989. Species profiles: Life histories and environmental requirements of coastal fishes and invertebrates (North Atlantic)- -rainbow smelt. U.S. Fish Wildl. Serv. Biol. Rep. 82(11.106). U.S. Army Corps of Engineers, TR EL-82-4. 11 pp. 51 52 Creaser, C.W. 1925. The establishment of the atlantic smelt in the upper waters of the Great Lakes. Pap. Mich. Acad. Sci., Arts, and Letters. 5:405-423. Dementieva, T.F. 1973. Methods of investigation and factors affecting the survival of fish during early stages of their development as a basis feature of the dynamics of recruitment to commercial fish stocks. Rapp. et Proces Verbaux Des Reunions 164:255—260. Ditton, R. and T. Goodale. 1972. Marine recreational uses of Green Bay: a survey of human behavior and attitude patterns. Technical Report #17, University of Wisconsin Sea Grant Program. WIS-SG-72- 217. 228 pp. Evans, 0.0. and D.H. Loftus. 1987. Colonization of inland lakes in the Great Lakes region by rainbow smelt, Osmerus mordax: their freshwater niche and effects on indigenous fishes. Can. J. Fish. Aquat. Sci. 44(Suppl. 2):249-266. Evans, 0.0. and P. Waring. 1987. Changes in the multispecies, winter angling fishery of Lake Simcoe, Ontario, 1961 - 1983: Invasion by rainbow smelt, Osmerus mordax, and the roles of intra- and inter- specific interactions. Can. J. Fish. Aquat. Sci.44(suppl. 2):182- 197. Francis, R.C., S.A. Alderstein, and A. Hollowed. 1989. Importance of environmental fluctuations in the management of Pacific hake (Merluccius productus), p. 51-56. In R.J. Beamish and G.A. ; McFarlane [ed.] Effects of ocean variability on recruitment and an evaluation of parameters used in stock assessment models. Can. Spec. Publ. Fish. Aquat. Sci. 108. Freeberg, M.H., W.W. Taylor, and R.W. Brown. 1990. Effect of egg and larval survival on year-class strength of lake whitefish in Grand Traverse Bay, Lake Michigan. Trans. Am. Fish. Soc. 119:92-100. Henderson, B.A. and S.J. Nepszy. 1989. Factors affecting recruitment and mortality rates of Rainbow smelt (Osmerus mordax) in Lake Erie. J. Great Lakes Res. 15(2):357-366. Hile, R. 1962. Collection and analysis of commercial fishery statistics in the Great Lakes. Great Lakes Fishery Commission Technical Report No. 5. 31 pp. Hulbert, P.J. 1974. Factors affecting Spawning site selection and hatching success in anadromous rainbow smelt (Osmerus mordax. Mitchill). M.S. Thesis University of Maine at Orono. 43 pp. Jensen, A.L. 1985. Time series analysis and the forecasting of menhaden catch and CPUE. N. Amer. J. Fish. Manage. 5:78-85. 53 Langlois, T.H. 1935. Notes on the Spawning habits of the atlantic smelt. Copeia 3:141-142. Larkin, P.A. 1973. Some observations on models of the stock and recruitment relationship. Rapp. et Proces Verbaux Des Reunions 164:316-324. Lett, P.F., A.C. Kohler, and D.N. Fitzgerald. 1975. Role of stock biomass and temperature in recruitment of souther Gulf of St. Lawrence cod, Gadus murhua. J. Fish. Res. Board Can. 32:1613- 1627. McCullough, R.D. and J.G. Stanley. 1981. Feeding niche dimensions in larval rainbow smelt (Osmerus mordax). Rapp. P.-v. Reun. Cons. Int. Explor. Mer 178:352-354. Meyers, T.P. and R.F. Dale. 1983. Predicting daily insolation with hourly cloud height and coverage. Journal of Climate and Applied Meteorology 22:537-545. O’Connell, C.P. and L.P. Raymond. 1970. The effect of food density on survival and growth of early post yolk-sac larvae of northern anchovy (Enqaulis mordax Girard) in the laboratory. Journal of the Exploration of Marine Biology and Ecology 5:187-197. O’Gorman, R. 1983. Distribution and abundance of larval fish in the nearshore waters of Western Lake Huron. J. Great Lakes Res. 9:14- - 22. . Platt, T., and K.L. Denman. 1975. Spectral analysis in ecology, p. 189-210. In R.F. Johnston et al. [ed.]. Annual review of ecology and systematics. Annual Reviews Inc., Palo Alto, Calif. Quinn, T.J.,II, and R.P. Marshall. 1989. Time series analysis: quantifying variability and correlation in SE Alaska salmon catches and environmental data, p. 67-80. In R.J. Beamish and G.A. McFarlane [ed.] Effects of ocean variability on recruitment and an evaluation of parameters used in stock assessment models. Can. Spec. Publ. Fish. Aquat. Sci. 108. Raab, R.L. and D.N. Steinnes. 1979. The economics of smelting as a recreational activity. Minnesota Sea Grant Publication 04-8-M01- 26, final report, Duluth, Minnesota. Rao, A.R., G. Padmanabhan, and R.L. Kashyap. 1984. A comparative analysis of recently developed methods of spectral analysis. In: "Frontiers in Hydrology." Water Resources Publications, Fort Collins, CO. pp 127-149. Reckahn, J.A. 1986. Long-term cyclical trends in growth of Lake Whitefish in South Bay, Lake Huron. Trans. Am. Fish. Soc. 115:787-804. 54 Ricker, W.E. 1975. Computation and interpretation of biological statistics of fish populations. Fish. Res. Board Can. Bull. 191. Rothschild, B.A. 1961. Production and survival of eggs of the american smelt, Osmerus mordax (Mitchell), in Maine. Trans. Am.Fish. Soc. 90:42-48. Rupp, R.S. 1965. Shore spawning and survival of eggs of the american smelt. Trans. Am. Fish. Soc. 94:160-168. Saila, 5.8., M. Wigbout, and R.J. Lermit. 1980. Comparison of some time series models for the analysis of fisheries data. J. Cons. int. Explor. Mer 39:44-52. Schaefer, W.F., R.A. Heckman, and W.A. Swenson. 1981. Postspawning mortality of rainbow smelt in Western Lake Superior. J. Great Lakes Res. 7:37-41. Schneberger, E. 1937. The biological and economic importance of the smelt in Green Bay. Trans. Am. Fish. Soc. 66:139-142. Scott, W.B. and E.J. Crossman. 1973. Freshwater fishes of Canada. Bulletin 184 Fisheries Research Board of Canada, Ottawa. Sissenwine, M.P., H.J. Forgarty, and W.J. Overholtz. 1988. Some fisheries management implications of recruitment variability. In Fish Pooulation Dynamics (second edition). Ed. J.A. Gulland. John Wiley and Sons Ltd, New York. Steel, R.G.D. and J.H. Torrie. 1980 Principles and Procedures of Statistics: A Biometrical Approach (second edition). McGraw-Hill Book Company, New York. 633 pp. Stewart, D.J., J.F. Kitchell, and C.B. Crowder. 1981. Forage fishes and their salmonid predators in Lake Michigan. Trans. Am. Fish. Soc. 110(6):751-763. Tautz, A., P.A. Larkin, and W.E. Ricker. 1969. Some effects of simulated long-term environmental fluctuations on MSY. J. Fish. Res. Board Can. 26:2715-2726. Tin, H.T. and D.J. Jude. 1983. Distribution and growth of larval rainbow smelt in eastern Lake Michigan, 1978-1981. Trans. Am. Fish. Soc. 112:517-524. Van Oosten, J. 1937. The dispersal of smelt, Osmerus mordax (Mitchill), in the Great Lakes region. Trans. Am. Fish. Soc. 66:160-171. Van Winkle, W., B.L. Kirk, and B.W. Rust. 1979. Periodicities in Atlantic coast striped bass (Morone sexatilis) commercial fisheries data. J. Fish. Res. Board Can. 36:54-62. 55 Walter, G. and W.J. Hoagman. 1975. A method for estimating year class strength from abundance data with application to the fishery of Green Bay, Lake Michigan. Trans. Am. Fish. Soc. 104:255-255 Ware, D.M. 1975. The relation between egg size, growth, and natural mortality of larval fishes. J. Fish. Res. Board Can. 32:2503- 2512. APPENDIX 56 Hile’s (1962) method of calculating abundance was examined to determine if the time period used to calculate the abundance index significantly effects the values of the abundance index. The abundance index (Table 9) was calculated using six different 13 year time intervals as the base period: 1968 - 1980, 1970 - 1982, 1972 - 1984, 1973 - 1985, 1974 - 1986, and 1975 - 1987. The mean, standard deviation, and coefficient of variation are calculated for each base period (Table 10). The graph of the mean abundance over time (Figure 15) showed the same shape as the abundance index used for the Green Bay study. The scale of the curves changed based on the years chosen for the base period, but not the general shape. The graph of abundance for each site (Figure 16) on the same graph showed the overlap of the graphs. It was concluded from this analysis that as long as the base period is the same for all gears the abundance index will retain the same general shape. 57 Table 9: Rainbow smelt abundance for each of 6 base periods at Green Bay. 1968 1980 l 1 Hid Hid O O D I O 0 O O O O O H P’ O O O O O O O O O hUIQQmUOWGQQQQNOSOHOO‘mmU-biu muonoeomHmmoxoxoeHHsoooomiouuouHHo \OUTLJ .c-u 53.6 46.1 64.4 21.6 29.1 21.6 22.5 5.3 2.7 15.8 40.9 55.5 51.6 47.3 44.4 38.4 1970- 1982 Hid Hid O O O O O O O O O I O H 010QOuhOO‘i-‘mUICDI-‘OIOQOOONNIOUHOO 1.1 O O O O O uhUIUIQmUOWOiQQQQNOiOI-‘OGUIUIU-HN \OU‘IU ubU 53.5 45.9 63.9 21.4 28.9 21.5 22.4 5.2 2.6 115.6 139.9 55.4 51.6 47.2 44.4 38.4 1972 1984 Base Period 1.1 H e e e e e e e e e 1.1 Hmmmbuiooomqmo‘sJHooooommmc-NuH U A O\ NNN00000101~011009UH01101041~osu00mm00 \I N 48.6 38.8 54.6 19.1 26.0 19.0 20.4 4.8 2.5 102.9 97.4 43.5 41.7 36.6 31.7 28.6 1973 1985 1 1.1 H 0 O O I I O O O O O O O O 0010053001109\IO‘OHUUHU‘OGO‘UGIF‘OOQIO 1.1 1.1 U'IQQQUIUO'GO‘QGOS\iH'OGOO'vWUIbUUI-I. U :5an ”GUT 39.7 5".2 21.2 28.1 18.4 20.9 6.0 2.7 02.3 93.1 41.2 39.6 34.7 29.0 26.4 1974 1986 1.1 NOUI‘OOtatmiOUIUIQQ-hUINI-IQUIHOUIfl-bm 1.1 U‘QQQ-bUIOQWOIWMQOmQIOQUIUI-hNUI-I 1975- 1987 1.1 H O10OUINQ'AOIJIO‘GONNOUIOQUIUIQUIOOQW H H P‘ chuimtuc:min01m 40 - c m a: E 20 - 0 l l l l l l I l l l 1948 . 1956 1964 1972 1980 1988 1952 1960 1968 1976 _ 1984 Year Figure 15: Mean value of rainbow smelt abundance index for all base periods combined. 60 160 140 - 120 - 100 - 81) F' 60 - Abundance Index 40— ‘V 20 P l l l l l l l l L L 0 1948 1956 1964 1972 1980 1988 1952 1960 1968 1976 1984 Year 1968-80 1970-32 1972-34 1973-85 1974~86 1975-87 --.--.-u-u o-o-c-uo Figure 16: Rainbow smelt abundance index versus time for each base period used in the simulation.