Competitive exclusion in Cladocera through elevated mortality of adults

The population dynamics of two cladocerans, Ceriodaphnia pulchella and Diaphanosoma brachyurum competing under laboratory conditions in lake water was analysed using crosscorrelations. Both mixed and isolated populations of the two cladocerans showed delayed densitydependence in the death rates of juveniles and adults as well as in fecundity rate. The regressions for each of the three rates on total density of competitors were compared between the two species. There were no significant differences in the slopes of regressions for fecundity rates and the death rates of juveniles. However, in the inferior competitor (Diaphanosoma) which went extinct in all treatments, the death rate of adults increased with total density much more quickly than in the superior competitor (Ceriodaphnia). The intraspecific comparisons indicated that while Ceriodaphnia adults survived better than juveniles under conditions of crowding, in Diaphanosoma, juveniles were better survivors than adults. These data suggest that the contention of higher vulnerability of cladoceran juveniles than adults to starvation and crowding may prove to be not a universal phenomenon. Introduction In Cladocera, juveniles are usually considered more sensitive to crowding and starvation than adults, and this contention has been used to predict the outcome of competition. For example, Neill (1975) has found that adults of Ceriodaphnia outcompeted Daphnia in laboratory cultures through suppression of its juveniles. Lynch (1978) gave evidence of higher sensitivity to starvation of juvenile Daphnia during co-exploitative interactions with other cladocerans in the lake. Tessier el al. (1983) have found in laboratory experiments that starvation time in Daphnia is generally shorter in juveniles than adults; however, it varied strongly with the amount of lipid reserves accumulated in the body of animals. The contention of higher sensitivity of juveniles to starvation was finally incorporated in a model, predicting the outcome of competition among Cladocera as a function of food supply (Romanovsky and Feniova, 1985). However, Gliwicz (1990) tried to predict the outcome of competition among zooplankton by measuring Daphnia threshold food requirements in young individuals only (2-6 days old). As his approach aimed to predict events at a population level, his measurements implicitly assumed similar sensitivities to starvation of juveniles and adults. Earlier, Threlkeld (1976) proposed a model of survival time of zooplankton as a function of weight-specific respiration rate and the fraction of prestarvation body weight under conditions of starvation. Testing of that model suggested a possibility that aged adults may be very sensitive to starvation as well. Assuming a strong correlation between starvation and increased population density via reduced food concentration, we can summarize the different views by © Oxford University Press 1083 V.Matveev and W.Gabriel proposing three types of dependencies to predict the responses of mortality to an increase in population density (or to a decline in per capita food supply): (i) the mortality of adults remains constant or increases at a slower rate than that of juveniles (Romanovsky and Feniova, 1985), (ii) the mortalities of both juveniles and adults increase at the same rate (a view implicit in Gliwicz, 1990), (iii) the mortality of adults increases while the mortality of juveniles remains constant or increases at a slower rate. These three alternatives would probably cover major conceivable cases. Below, we re-analyse an experiment on competition between Ceriodaphnia pulchella Sars and Diaphanosoma brachyurum (Lievin) described earlier by Matveev (1987a). In this experiment Diaphanosoma was invariably outcompeted under conditions of epilimnion water of meso-eutrophic Lake Glubokoe, Moscow Region. Our objective here was to elucidate the demographic mechanism of that competitive exclusion. We analysed densitydependencies of three finite population rates: fecundity rate, as well as juvenile and adult death rates. Firstly, we tried to find out if these rates were correlated to population densities. As birth rates and death rates are normally delayed functions of density (Matveev, 1985,1987b) we used time lag analysis. Secondly, by comparing the slopes of regressions of the three rates we determined which of them were more sensitive to changes in density and identified which age group is more vulnerable in competition. Method The experiment was conducted in 0.6 1 stoppered flasks with epilimnial lake water void of zooplankton (95-u.m filtration). The filtration provided a natural spectrum of phytoplankton as food for cladocerans in the range of cell sizes of 240 u,m; there were no larger cells. The algae were represented by two categories of naked flagellates and three categories of Chlorococcales, the densities of which were monitored throughout the experiment (Matveev, 1987a). The estimated overlap in feeding niches of Ceriodaphnia and Diaphanosoma was the highest possible and symmetrical according to MacArthur-Levins' index of overlap (Matveev, 1987a). There were five control flasks for Ceriodaphnia and five for Diaphanosoma, which were inoculated with 1-2-instar juveniles. These flasks had isolated populations with starting densities of 2, 4, 6, 8 and 10 animals flask" for each species. In five experimental flasks, where mixed populations were created, the starting densities of the two competitors varied yielding the ratios of 2:10, 4:8, 6:6, 8:4 and 10:2 animals flask" (one replicate per treatment). Thus, for a given species, the starting density was the same in the control flask and the corresponding experimental flask. Such a design implies that under the null hypothesis of no competition, there will be no difference between the mean size of an isolated population for the growing period and of a population in the presence of an alien species. The effect of competition is inferred when the difference is significant or/and one of the populations becomes extinct. The experiment was started on 9 June 1983, and was run for 77 days at a temperature of 20 ± 2°C. Every 3-4 days animals and eggs were counted, the 1084 Exclusion in Cladocera water changed and the corpses removed, counted and measured to distinguish between juveniles and adults. By the end of a 4-day interval between censuses, when animal densities were at their maximum, oxygen in the flasks was not depleted, remaining above 9.4 mg I". For more details of the experiment see Matveev (1987a). Finite rates of mortality and fecundity were estimated using the formulae: d; = D,/(N;M), da = Da/(NaA/),/= F/(Na.T) where d, = finite death rate of juveniles, da = finite death rate of adults, / = finite rate of fecundity, D, = number of juvenile corpses found by the end of the interval between consecutive censuses (A/ = 3 or 4 days), Nj — number of live juveniles (at the beginning of At), Da = number of adult corpses by the end of Ar, Na = number of live adults, F = total number of eggs, T = egg development time, determined from the dependence on the temperature (Bottrell etal., 1976). To determine density-dependencies in both mixed and isolated cultures, we regressed dt, da and / versus an estimate of population density introducing time lags (T). For mixed cultures, total biomass of two competitors was used as an estimate of population density. The biomasses were calculated according to assumed constant mean weights of juveniles and adults of each of the two cladocerans. The weights were assessed on the basis of body length measurements and using length-weight relationships (Bottrell et al., 1976). In singlespecies cultures, the calculation of biomasses by multiplying population numbers by constant weights did not make density estimates more accurate. Therefore, density estimates for isolated populations were total numbers only. Prior to calculating cross-correlations, the numbers of eggs, live and dead adults and juveniles were smoothed using a three point moving average to reduce noise in the data sets. Because time intervals between measurements were not constant, we weighted the points inversely according to the corresponding time intervals. The values of density estimates between days of censuses, necessary for cross-correlations, were obtained by cubic spline interpolation. We used Spearman rank correlation coefficients, but also always computed parametric product-moment correlation coefficients. Both gave very similar results. This might mean that the assumptions underlying parametric tests were not severely violated. After a preliminary scanning for time lags that maximize the absolute value of correlation using 0.5-day time step, we determined population delays with a numerical accuracy of 0.005 days so that the error in time delay resulting from the computations is negligible compared to the error resulting from experimental noise. The range of time delays considered was restricted to biologically meaningful intervals (<2 weeks, Matveev, 1983). By this means we avoided pseudocorrelations which could arise (e.g. between phases of population increase and decrease). Because all finite rates contained an estimate of density in the denominator and these rates were compared with population densities, there was a possibility of obtaining misleading autocorrelations rather than true biological density1085 V.Matveev and W.Gabriel dependencies. To overcome this problem, we compared a cross-correlation curve for a given rate with a cross-correlation curve where the numerator of the rate formula was substituted by 1 ('placebo'). For example, if D/Nfit) was compared with N(t — T) for a density-dependence at a time lag T, a comparison was also made between l/(NjAr) and N(t T). In the computed placebo crosscorrelation a constant value of 1 had no effect on the variation of the rate itself. In the normal cross-correlation the numerator was a variable which affected the variation of a given rate. By comparing placebo and normal cross-correlations one could infer the influence of the normal numerator on the rate and its variation. We accepted a normal cross-correlation as valid, reflecting biological density dependence, if it differed significantly from placebo cross-correlation (e.g. Figure 1). If the two resulting curves had not differed, the detection of a density-dependent relationship for that rate was considered to be impossible. A time lag corresponding to a maximum (by the absolute value) of a crosscorrelation curve was taken to construct a density-dependent regression. The estimates of the slope of the regression remained virtually the same if 1-2 day shifts in the time lag were allowed. Even when replicated, each experimental population of Cladocera has a unique history of development (Slobodkin, 1954). This can make impossible the analysis of density-dependence if it is based on the pooled data from several populations. Therefore, like in previous studies (Matveev, 1985, 1987b), we analysed each species population separately. For 77 days of the experiment, the maximal number of data points per experimental unit could reach 22. However,