Energy Density and Size of Pelagic Prey Fishes in ...

2 downloads 0 Views 8MB Size Report
Wedge, Dan Bishop, and Larry Skinner of the NY-. DEC for providing us with unpublished data on. Lake Ontario salmonines. This is contribution 837.
Transactions of the American Fisheries Society 123:519-534, 1994 American Fisheries Society

Energy Density and Size of Pelagic Prey Fishes in Lake Ontario, 1978-1990: Implications for Salmonine Energetics PETER S. RAND AND BRIAN F. LANTRY College of Environmental Science and Forestry, State University of New York 1 Forestry Drive, Syracuse, New York 13210, USA

ROBERT O'GORMAN AND RANDALL W. OWENS National Biological Survey, Great Lakes Center, Oswego Biological Station 17 Lake Street, Oswego, New York 13126, USA

DONALD J. STEWART College of Environmental Science and Forestry, State University of New York, and Research Center, State University of New York at Oswego 319 Piez Hall, Oswego, New York 13126, USA Abstract.—We describe dynamics of energy density and size of Lake Ontario alewife Alosa pseudoharengus and rainbow smelt Osmerus mordax, and we use a bioenergetics model of a common pelagic piscivore, chinook salmon Oncorhynchus tshawytscha, to demonstrate the effect of these factors on piscivore daily ration during 1978-1990. The energy density of alewives varied more than twofold between peaks in September (age 1) or October-November (age £2) and the lows in May (age 1) or July-September (age a 2). The previously described seasonal pattern of energy density of Lake Michigan alewives was similar except that energy density of older alewives (age >3) was markedly higher in Lake Michigan. During 1978-1990, the spring energy density of Lake Ontario alewives peaked in 1979 (6,259 J/g wet weight), declined irregularly until 1985, and then remained stable through 1990 (at approximately 4,600 J/g). The initial decline may have been a density-dependent response to a burgeoning alewife population, but the lack of an increase in alewife condition in the late 1980s, when alewife biomass fell, suggests a decline in lake productivity. Energy density of rainbow smelt increased with age in Lake Ontario and condition was invariant during 1978-1990 despite a threefold change in rainbow smelt biomass. Rainbow smelt energy density was lower and fluctuated less seasonally in Lake Ontario than in Lake Michigan. Mean weight of alewives aged 2 and older dropped from 41 g in 1978 to 19 g in 1989 in Lake Ontario. Rainbow smelt aged 2 and older showed a drop in mean weight from 13-17 g in 19781982 to 8 g in 1990. This downward trend in mean size of alewives was correlated with the sizes of alewives consumed by Lake Ontario chinook salmon during 1983-1987. For adult chinook salmon to maintain a constant growth rate during 1978-1990, mean individual daily ration during June-October had to increase from a low of 2.2% body weight/d (or 1.5 prey fish/d) in 1979 to 3.1% body weight/d (3.7 prey fish/d) in 1988. This increase in forage demand may have caused the observed declines in individual condition of salmonines over this period.

,

Alewives Alosa pseudoharengus and rainbow smelt Osmerus mordax compose the majority of the pelagic planktivore biomass in Lake Ontario (Christie and Thomas 1981) and, along with the native bloater Coregonus hoyi, are the dominant planktivores in Lake Michigan (Eck and Wells 1987; Brandt et al. 1991). Alewives and rainbow smelt also form the bulk of the adult salmonine diets in these lakes (Brandt 1986; Jude et al. 1987; Stewart and Ibarra 1991). Because the salmonine populations are maintained by annual releases of hatchery fish and because the salmonines support economically important sport fisheries, there is considerable interest in the ability of the plank-

tivore stocks to sustain populations of piscivores. Recent studies have documented year-to-year changes in the numbers of alewives and rainbow smelt (O'Gorman and Schneider 1986; O'Gorman et al. 1987; Eck and Wells 1987). Growth of hatchery-reared brown trout Salmo trutta and coho salmon Oncorhynchus kisutch during their first year in Lake Ontario was positively correlated with numerical abundance of young alewives (O'Gorman et al. 1987). The growth of adult salmonines is affected not only by changes in prey numbers but also by changes in prey condition. Condition factor (AT) of adult alewives in Lake Ontario declined during

519

520

RAND ET AL.

1977-1985 and varied inversely with catch per unit effort in trawl surveys. This decline in alewife condition may have influenced the growth of age-2 coho salmon through a reduction in the nutritional quality of their preferred prey fish (O'Gorman et al. 1987). Condition factor has been used to measure the nutritional value of prey fish to predators but its use is often restricted or erroneous because of the number of assumptions that have to be met (Cone 1989; Springer et al. 1990). These assumptions are often violated, as revealed by the poor relationship between condition indices and more direct measures of energy content (e.g., Rath and Diana 1985; Strange and Pelton 1987). Length-weight parameters are a more accurate and useful measure of fish condition (Cone 1989; Springer et al. 1990), but we argue that more work is needed to test whether length-weight parameters are related to whole-body energy content, which provides one of the most direct measures of fish condition and allows us to better quantify the energy flux between planktivores and their predators (Stewart et al. 1983; Stewart and Binkowski 1986). Seasonal and annual cycles of energy density in Great Lakes planktivores and the mechanisms underpinning them are poorly understood. The seasonal cycles of energy density of alewives and rainbow smelt in Lake Michigan were measured in previous studies (Foltz and Norden 1977a; Flath and Diana 1985), but little is known about these processes in Lake Ontario prey fishes. In previous modeling analyses of prey fish consumption by salmonines in Lake Michigan, one or two years of prey fish energy density data were extrapolated over a much longer simulation period (e.g., Stewart and Ibarra 1991). This may be invalid when prey fish condition varies dynamically in response to changes in community biomass, food density, or climatic conditions. Mean size of available prey may also change over time, which has implications for predatorprey relations in aquatic ecosystems (Brooks and Dodson 1965; Peters 1983). By applying a bioenergetics model to estimate daily ration of piscivores in terms of both wet weight and number of prey fish consumed, one can integrate changes in energy content of prey with mean size of prey fish. Daily ration expressed in terms of number of consumed fish can serve as an index of the effort expended in foraging by pelagic predators and can help indicate the presence of foraging constraints. We describe the dynamics of length-weight parameters, mean size, energy density, and abun-

dance of alewives and rainbow smelt in Lake Ontario during 1978-1990 and compare values of energy density to published data and additional data that we collected during 1986-1987 in Lake Michigan. We relate mean size of trawl-caught alewives to mean size of alewives consumed by chinook salmon Oncorhynchus tshawytscha in Lake Ontario during 1983-1987. We use these data in a bioenergetics model to determine the effect of changing energy density and mean size of prey fish on daily ration of chinook salmon in Lake Ontario during 1978-1990. Methods Monthly sampling in Lakes Ontario and Michigan.— Alewives and rainbow smelt were caught in Lake Ontario with bottom trawls and frozen in water immediately after capture. Collections were made during the first third of each month, May through November 1989, and in mid-March 1990. The fish were caught mostly at depths of 25-75 m in southeastern Lake Ontario, 10-25 km east of Oswego, New York, but in September fish were caught near Rochester, New York, about 90 km west of Oswego. Alewives and rainbow smelt were collected in Lake Michigan with bottom trawls in January, February, and March 1986 near Two Rivers, Wisconsin. To supplement these data, a sample of alewives was collected from Lake Michigan during May and September 1987. These fish were captured in bottom trawls near Ludington. Michigan, Two Rivers and Sturgeon Bay, Wisconsin, and Waukegan, Illinois. For details on the 1987 sampling methodology, see Brandt et al. (1991). All fish in the Lake Michigan collections were preserved in ice as described above. Energy density determinations. — Wet and dry weights (g) offish from Lakes Ontario and Michigan were measured in the laboratory to track seasonal and ontogenetic dynamics of tissue water content. Five to 10 fish per size-group (approximately 10-mm size-groups) per month were thawed under running water. Power analysis of sample size (Neter et al. 1985) indicated that five or more fish were needed to get a stable variance for mean dry weight in each size-group. Before wet weights were measured, the stomachs of fish were opened and cleared of food. Total length (to the nearest millimeter) and wet weight (to the nearest 0.001 g) were recorded. The fish were placed on tared aluminum pans and oven-dried to a constant weight at 60-65°C. Final dry weights

ENERGY DENSITIES OF LAKE ONTARIO PREY FISHES

were recorded after the same weights (±0.01 g) were obtained on three separate days. We developed species-specific regression models to predict energy density from percentage dry weight (weight of dry matter expressed as percentage of wet weight). Energy densities of fish were determined with a Phillipson Oxygen Microbomb Calorimeter.1 Composites of two to five fish from selected size-classes were chosen as stratified random subsamples from monthly samples representing the greatest ranges of percentage dry weight to derive a robust regression model over observed dry weights. The coefficient of variation of percentage dry weight among the individuals in these composite samples did not exceed 5%. Each composite sample of two to five fish was ground with dry ice to a uniform particle size in a blender fitted with a stainless steel cup. Dry ice made the fish brittle and permitted grinding to a fine powder without tissue adhering to the walls of the blender cup. Each sample was then redried to constant weight at 65-70°C. Samples of dried tissue were formed into pellets and ignited inside a bomb calorimeter. The number of joules liberated was determined according to the user's manual for the bomb calorimeter and by following the methods of Phillipson (1964). Three to five determinations were made from each composite sample. The resulting values were averaged and used to represent the mean of the composite sample. These means were then regressed against the corresponding measures of percentage dry weight of the composite samples. Nine composite samples of Lake Ontario alewife and 15 of Lake Ontario rainbow smelt were processed. Five composite samples of Lake Michigan alewife (1987 collections) were processed. Energy values in this paper are in joules per gram wet weight. To evaluate the seasonal energy density cycle of the two fish species in both lakes, we developed species-specific regression models to estimate energy density (J/g wet weight) from percentage dry weight from our monthly samples in Lakes Michigan and Ontario. To track ontogenetic changes in the monthly samples, we partitioned the samples by years of life, which we assumed to begin on July 1 for alewife and June 1 for rainbow smelt, the approximate day of first feeding (Hewett and Stewart 1989; Lantry and Stewart 1993). Ages were incremented on these dates. Ages were estimated

1 Use of trade names does not constitute government endorsement of commercial products.

521

from length-frequency distributions in each month of collection. Annual spring sampling in Lake Ontario. — Biomass and size distributions of alewives and rainbow smelt were determined annually during 19781990 by the National Biological Survey (NBS) and the New York Department of Environmental Conservation (NYDEC). The survey methods were described in detail by O'Gorman and Schneider (1986) and O'Gorman et al. (1987) and are reviewed only briefly here. Two surveys with bottom trawls were conducted each spring during 1978-1990 in U.S. waters, one in late April-early May for alewives (average, 103 tows; range, 49122) and the other in late May-early June for rainbow smelt (average, 86 tows; range, 68-104). The 12-m-headrope trawls had cod ends of 9-mm stretch-measure mesh made of knotless nylon. The trawls were fished at 10-m depth intervals through the range of depths occupied by the target species along 12 transects (usually only 11 of the 12 transects were fished in late May-early June) spaced roughly 25 km apart from Olcott, New York (29 km east of the Niagara River mouth), east to near the head of the St. Lawrence River. Standard tows lasted 10 min and were made along a contour at a speed of 4.6 km/h. The biomass of prey fish was measured by stratified catch per standard trawl tow (CPUE). It was calculated by first dividing U.S. waters into strata, seven for alewife and six for rainbow smelt (O'Gorman et al. 1987). Stratification was by depth for alewife and by depth and lake region for rainbow smelt. Stratification was appropriate because of large, consistent differences between catches at different depths for both species and in different regions for rainbow smelt. We recorded total lengths (mm) at six (alewife) or all (rainbow smelt) transects. We measured all fish in small catches (generally 100 fish or less) and a random sample of about 100-200 individuals from large catches. The random sample was used to calculate the number offish in each 5-mm length group in that trawl haul. Lengths from all trawl hauls and transects were summed to obtain a length-frequency distribution. To calculate the CPUE in kilograms of fish in the length range used in our analysis, we first multiplied the numbers offish in each 5-mm length-group in the entire length-frequency distribution by the body weight of a fish at the midpoint of that 5-mm length-group (body weight was calculated with the year-specific length-wet weight relation) and then summed the products. The ratio of the weight of individuals in each of the length

522

RAND ET AL.

ranges to the total weight of the catch was multiplied by the total stratified CPUE in kilograms to apportion CPUE among the observed length ranges. Length-wet weight relationships of alewives and rainbow smelt in each year during 1978-1990 were calculated from fish collected during these annual surveys. The fish were caught in southern Lake Ontario near Olcott, Rochester, and Oswego, placed on ice, weighed (g) and measured (total length, TL, mm) within 30 h of capture. Alewives were collected from a wide range of depths (8-150 m) during the 13-year period because the depth at which they concentrated in April-May differed widely from year to year (O'Gorman and Schneider 1986; O'Gorman et al. 1991). Because the depth distribution of rainbow smelt in May-June was similar among years, the rainbow smelt lengthweight samples were mostly spread across the same depth range (8-75 m). Length-weight statistical analysis.—The total lengths and wet weights were logio-transformed before the length-weight relationship in each month or year was calculated. The lengths and numbers of fish in the tails of length distributions varied widely among years, so we restricted the lengths used to calculate length-weight relationships to a range in which the distribution of observations was similar (alewife: 120-199 mm TL, rainbow smelt: 100-169 mm TL). We used the same length restrictions for our analysis of the monthly samples. Because the monthly samples contained substantial numbers of smaller fish (consisting of young of year and some yearlings), we conducted separate analyses on these smaller fish (alewife: TL < 120 mm, rainbow smelt: TL < 100 mm). We used the homogeneity-of-slopes model of the general linear model program (SAS Institute 1988) to perform an analysis of covariance (ANCOVA) and estimate slopes and intercepts in each month and year. When the F-value indicated a significant difference (P < 0.05) among slopes, we used a series of Mests to determine which slopes differed. Two slopes were considered different only when P was less than 0.05/N (N being the number of paired comparisons in the data set). When the F-value indicated that the slopes were not significantly different (P > 0.05), we used the ANCOVA model of the general linear model program (SAS Institute 1988) to estimate a pooled slope, a least-squares mean (LSM) weight in each year or month, and the intercepts in each year or month. We generated a table of all probability values (P)

for the hypothesis HQ : LSM(,} = LSMQ) (pdiff command: SAS Institute 1988) and considered two means different when P < 0.05/N. If significant differences were found in lengthweight regression parameters in the monthly samples, we regressed the parameters against measured energy density. This served as a test of the hypothesis that length-weight parameters can serve as a reliable index of condition for these species. When there were interaction effects between slopes and intercepts across the monthly samples, we regressed energy density against weight at the mean length (calculated for all fish in each of the small and large size categories). The adequacy of this length selection was evaluated by recomputing the regression equation with lengths from the entire length range in each size-group. Chinook salmon diet survey.—We analyzed stomach contents of chinook salmon caught in the sport fishery in Lake Ontario to describe wet weight diet composition of prey fishes and to monitor mean lengths of consumed alewives during 19831988. Ports along the New York shoreline were staffed with volunteers throughout the fishing season during the years of the study. Data recorded included month of capture, total length, and wet weight of predator and species and mean length of ingested prey (when identifiable). To summarize data on diet composition, we converted individual prey fish total lengths (when necessary, standard lengths were multiplied by 1.12 to convert them to total lengths) to wet weight with species-specific total length-wet weight regression equations. We included only data on nonempty stomachs with identifiable prey items. For modeling purposes, we extracted diet data for Chinook salmon in the 2,000—6,600-g weight range, which brackets individuals in the age-2 class. This age designation was based on weight at age measured at the Salmon River hatchery in Altmar, New York (mean offish returning in 1986-1987, L. Wedge, NYDEC, unpublished data). The diet data were further stratified by season (April, May, June-August, September). Because of small seasonal sample sizes, we pooled across years. Prey weights were summed across predators during each season and wet weight composition (as a percentage of total weight of stomach contents) was computed for each diet category (small alewives, 8 g; rainbow smelt; other) as described by Stewart and Ibarra (1991). We also computed a mean total length of consumed large alewives for each chinook salmon (age, >2; wet weight, >2,000 g) caught from June

ENERGY DENSITIES OF LAKE ONTARIO PREY FISHES

through October 1983-1987 (sample size was inadequate in 1988). We computed a mean alewife total length across all Chinook salmon in each year of the survey. Small sample sizes for rainbow smelt precluded us from performing a similar size analysis of this species. Chinook salmon energetics.—'We used the bioenergetics model of Stewart and Ibarra (1991) and the software of Hewett and Johnson (1992) to evaluate the compensatory response of chinook salmon prey consumption to the observed changes in energy density and mean size of prey fishes during 1978-1990. Although data are not sufficient to accurately estimate changes in consumption patterns by chinook salmon over this period, our objective was to evaluate the importance of prey size and energy content on consumption while holding all other model parameters constant. Our modeling objective was to measure the amount of prey consumed (in g • g~' • d ~' and number of prey • predator"' • d ~ ] ) that would be necessary to achieve a stable growth rate during 1978-1990. We chose to model mature chinook salmon because their diet is composed almost entirely offish in Lake Ontario (Brandt 1986) and the forage demand of this species is recognized as the highest of all stocked salmonines (Stewart et al. 1981). We modeled daily energetics of an average individual over the period of fastest growth (June-October) during the third year of lake residence (age 2). The population of age-2 chinook salmon can be partitioned into two life history forms based on the timing of spawning runs (Stewart etal. 1981). Fish of one form migrate to streams for spawning at the end of their third year; fish of the other form spend an additional year in the lake before spawning. These two life history forms have different growth rates (Stewart et al. 1981). Weights input to the model were based on mean weight-at-age data recorded for the Salmon River population monitored during the fall migrations of 1986-1987 (age 2, 6.603 g; age 3, 9,584 g; Wedge unpublished data). Simulations were run from May 1 to November 30. The initial and final weights input to the model (age 2: 1,978 g on May 1, 6,603 g on November 30; age 3: 1,428 g on May 1, 5,441 g on November 30) were calculated based on the ratio of spring weights to fall spawning weights estimated for Lake Michigan chinook salmon (Stewart et al. 1981). The data on wet weight diet composition derived above were used in the model. We used water temperature profiles from Rand et al. (1993) and the temperature preferendum of chinook salmon from Stewart and Ibarra (1991).

523

To estimate seasonal patterns of alewife energy density in each year during 1978-1988 (for which we had no data) for use in the bioenergetic models, we used the energy densities predicted for the spring (see results below) to scale the monthly pattern observed in 1989-1990. Energy density of chinook salmon was based on a weight-dependent model developed by Stewart and Ibarra (1991) for Lake Michigan chinook salmon. Separate simulations were run on both life history forms of chinook salmon to estimate daily rations necessary to achieve the observed growth input into the model. The total prey consumption for the two diet categories designated in the model (rainbow smelt and alewives larger than 8 g) was cumulated during June 1-October 31. This value was then divided by 152 (days) to arrive at an estimate of mean daily ration during this period. Mean weight of chinook salmon over this period (used to calculate mass-specific consumption) was estimated by averaging the wet weights predicted at the end of each month for each life history form during June-October. The estimates of daily ration for each life history form were then averaged and used to represent the rate of prey consumption during the period of rapid growth in each year of the simulation. Daily ration, expressed as number of prey fish consumed per day, was calculated by dividing the model-estimated total weight of large alewives and rainbow smelt consumed daily by the mean weight of the respective prey fish in a given year. We calculated the ratio of the mean weight of consumed alewives during June-October to the mean weight of alewives caught in the spring trawl survey for each year of the diet survey (1983-1987; sample size was insufficient in 1988). To estimate mean size of consumed alewives for the other years of the model period (1978-1982, 1988-1990), we multiplied the mean ratio (1.1) for 1983-1987 by the mean size of alewives measured in the spring trawl survey. We approximated the size of rainbow smelt consumed by chinook salmon during June-October by multiplying mean size of rainbow smelt aged 2 or more in the spring trawl catch by 1.2, the ratio of the mean weight of rainbow smelt (age, >2) captured in August 1989 to the mean weight from catches in June 1989 (based on our monthly trawling survey, see above). Results Alewife condition, as measured by changes in slopes of the length-weight relationships, differed significantly across months (during 1989-1990)

524

RAND ET AL.

TABLE 1. —Parameters of regressions of total length (XL, mm) on wet weight (g) for alewives collected monthly in Lake Ontario in 1989 and 1990. The length-weight equation takes the form logio(weight) = j3-logio(length) + a. Energy density values (J/g wet weight) were derived from the observed linear relationship between percentage dry weight and energy density. Tests for homogeneity of slopes were significant for both size-classes (P < 0.002). Slope values followed by the same letter (within each alewife size-group) were not significantly different (small alewives: P > 0.05/42; large alewives: P > 0.05/48). Entries are in order of decreasing /3. Regression parameter 0 (slope)

Energy density

a (intercept)

Month

Value

SE

Mar May Oct Nov Sep Jul Jun

3.999 z 3.406 z 3.352 z 3.084 z 3.071 z 2.144x 2.099 x

0.180 0.085 0.052 0.044 0.038 0.137 0.114

Small alewives ( 0.98, P < 0.001; Figure 1). We compared the regressions for alewife developed in this study to an earlier relationship developed for Lake

525

ENERGY DENSITtES OF LAKE ONTARIO PREY FISHES

TABLE 2.—Parameters of regressions of total length (TL, mm) on wet weight (g) for rainbow smelt collected monthly in Lake Ontario in 1989 and 1990. The length-weight equation takes the form logio(weight) = 0-logio(length) + a. Energy density values (J/g wet weight) were derived from the observed linear relationship between percentage dry weight and energy density. Least-squares mean (LSM) wet weights (g) were calculated with analysis of covariance (ANCOVA). Tests for homogeneity of slopes were significant for small rainbow smelt (P = 0.002) but not for large rainbow smelt (P = 0.319). Slopes for small rainbow trout followed by the same letter are not significantly different (P > 0.05/42). Entries are in order of decreasing /} or LSM. Regression or ANCOVA parameter 0 (slope) Month

Value

a (intercept) SE

Value

Energy density SE

N

Sep Nov Aug Mar Jul May Jun

2.409 z 2.366 zy 2.326 z 2.315 z 2.234 zy 2.067 zy 2.002 y

0.076 0.223 0.092 0.122 0.056 0.058 0.055

Small rainbow smelt ( 2) and was lowest in May (age 1) and July-September (age >2). The seasonal pattern of Lake Michigan alewives in 1979-1981 (Flath and Diana 1985) was similar to that in Lake Ontario in 1989-1990, but the older alewives in Lake Ontario (age > 3) exhibited lower energy densities than similar-aged Lake Michigan

LSM

.152 .151 .145 .136 .134 .122 .114

Value

SD

3,430 3,567 3,430 4,266 3,473 3,197 3,478

388.8 274.1 1,225.4 597.2 435.7 631.5 332.3

4,942 4,457 5,771 4,599 4,842 4,814 4,631

582.1 801.9 517.7 816.1 603.5 696.8 606.0

alewives (Figure 2). Our estimates of energy density of Lake Michigan alewives in 1986-1987 were generally lower than those reported for 19 7 9-19 81 by Flath and Diana (1985; Figure 2). Energy density of Lake Ontario rainbow smelt increased gradually with age but without any marked seasonal fluctuations (Figure 2). Energy density of rainbow smelt was generally lower and less variable than that of alewives, especially during late fall when energy density of alewives increased sharply. Seasonal energy densities for the three age-groups of rainbow smelt in Lake Ontario (Figure 2) were also lower and less variable than those for age-2 and age-3 fish from Lake Michigan (Foltz and Norden 1977a), which dropped at spawning (late April) and then increased through the summer and early fall. The data of Foltz and Norden (1977a), along with values for energy density reported for Lake Michigan rainbow smelt (6,654 J/g wet weight) by Rottiers and Tucker (1982) and our estimates of rainbow smelt energy density in Lake Michigan, are all higher than those measured for Lake Ontario rainbow smelt (Figure 2). All three age-classes of rainbow smelt from Lake Ontario were lower in energy density in late autumn (November) than in spring (March and May), which was the reverse of what we observed for sympatric Lake Ontario alewives.

526

RAND ET AL.

10.0

Alewife Ontario J-g'1= 360-(%dry) - 2414 Michigan: J-g'1= 387-(%dry) - 4102

8.0

.4 •S>

40

! |>

Lake Ontario 1989-90 (this study) Lake Michigan 1987 (this study)

2.0

Lake Michigan (Stewart & Binkowski 1986, based on data of Rath & Diana 1985)

D)

3 o.o 10.0