Detectability of fifteen aquatic micro/mesocosms

Abstract

Zooplankton abundance and species richness in 15 untreated 12,000 L outdoor microcosms (n = 15) were monitored over the course of 1 year to document the inherent variability and statistical detectability between replicates. Statistical power analysis were applied to derive the statistically minimal detectable difference (MDD) between replicates with default values set at; alpha = 0.1 and beta = 0.2. Copepod abundance and species richness generally demonstrated the best detectability at 0.31 and 0.16, respectively, (n = 15); 0.59 and 0.33 (n = 3). Total zooplankton abundance and species richness had the lowest detectabilities at 0.19 and 0.14, respectively, (n = 15); 0.35 and 0.3 (n = 3). Rotifers, due to their opportunistic and rapid life traits, had the lowest single-species abundance detectabilities at 0.54 (n = 15); 0.8 (n = 3), whereas macroinvertebrate species richness had the lowest detectability at 0.43 (n = 15); 0.7 (n = 3) over 1 year. We recommend a priori calibration of the study design relative to relevant MDDs. Moreover, it is suggested to consider alternatives to statistical null hypothesis testing.

Traits and stress: keys to identify community effects of low levels of toxicants in test systems

“…Secondly, biotic interactions may have been different between replicates as a result of differences in environmental parameters and stochastic processes, for example, those that occur within the process of colonisation. Hence, communities within each mesocosm were unique to a certain degree, similar to natural communities, which resulted in considerable inter-replicate variation (Sanderson et al 2009). Multivariate analyses lose power in the identification of statistical links when variation is high.…”

Pyrethroid effects on freshwater invertebrates: A meta-analysis of pulse exposures

“…Often the Coefficients of Variation (CV: stdev/mean) on various population density measures of control treatments are in the range of 30e50% (Sanderson, 2002). This means that the lowest detectable difference between control and treatment when using 3e4 replicates per treatment is often more than 50% (Sanderson et al, 2009). In other words, if it is a protection goal for a given EPA to tolerate maximum a 25% reduction in the population density of a sensitive species (after X weeks recovery), the calculated NOEC (t-test between control and treated using the required 4 replicates, a CV of 30% of both controls and treated populations and an a ¼ 0.05) only states with 27% certainty (power of the test) that there is actually no difference between control and treatment.…”

Rebuttal related to “Traits and Stress: Keys to identify community effects of low levels of toxicants in test systems” by Liess and Beketov (2011)

“…There is often much variability between mesocosms under the same treatments, the different control mesocosms can have quite different communities (Sanderson et al 2009), as can replicates of the various treatments. Under multivariate methods this adds noise and effects of the treatment will only be detected if they are greater than this noise.…”

“…They state that “Such analyses would no doubt also highlight the sensitive responses at the population level as are presented by the SPEAR method.” But we showed previously that Thiacloprid effects at low concentrations (0.1 μg/l) could only be observed for one species at some time points and long-lasting effects—as shown by SPEAR mesocosm —were not identified (Beketov et al 2008). The cause for the difficulty to identify low-level effects for single taxa lies in the low numbers of replicates available for all mesocosm investigations and the high variance generally present in such complex systems (Sanderson et al 2009). As stated repeatedly, only the grouping of taxa—as done in SPEAR mesocosm —allows this problem to be tackled.…”

“…Mesocosm investigations, which are one valuable tool to identify such long-term community effects, face the problem that results are obscured by confounding factors and high variability between replicates (Sanderson et al 2009). This problem is even greater when assessing long-term effects of toxicants because inter-replicate variation increases with time.…”