Sensitivity Analysis of Population Growth Rates Estimated from Cladoceran Chronic Toxicity Tests

Four variables mean brood size, day of first reproduction, longevity and per capita rate of increase ( r ) -were compared in four 70-d chronic toxicity tests in which Daphnia pulex were exposed to cadmium or copper in continuous and pulsed exposures. With these data we asked the question: Is the population-level variable ( r ) more sensitive than the three organism-level variables (mean brood size, day of first reproduction and longevity) as an indicator of toxicant stress? We define two variables to be equally sensitive if, for both variables, differences between treatments and the control have statistical significance levels in the same probability range. In these four tests, none of the four variables was consistently the most sensitive. We also evaluated the sensitivity of r by simulating shorter test durations, delays in reproduction and less frequent observation schedules. Simulated test durations of less than 21 d produced biased underestimates of r and increased coefficients of variation of the estimates of r relative to the 70-d values; simulated 1and 2-d delays in reproduction also produced biased underestimates of r. However, estimates of r computed for a simulated Monday-Wednesday-Friday observation schedule did not differ significantly from estimates of r computed for a daily observation schedule. We conclude that although the estimator of per capita rate of increase is not always the most sensitive statistic that can be computed from cladoceran chronic toxicity test data, it can be useful for evaluating apparently conflicting effects of pollutants on survival and reproduction, as occurred in the copper continuous-exposure toxicity test.