No abstract

Current water-quality criteria for metals typically are derived from toxicity tests with the metal dissolved in clean laboratory water. Estimating the toxicity of iron from such tests, however, is extremely difficult because of the complex solubility and toxicity characteristics of the ferrous and ferric forms of the metal in freshwater. Consequently, a criterion for dissolved iron in freshwater derived from standard laboratory bioassays may not accurately describe the actual bioavailability and toxicity of this metal. A new approach is necessary to adequately protect aquatic life from the direct (toxic) and indirect (physical) negative effects of iron. We present a novel methodology to derive bioassessment-based benchmarks for total iron. This approach involves the use of quantile regression to model the decline in maximum abundance of taxa along a gradient of increasing iron concentrations. The limiting function (e.g., 90th quantile) is used to project the iron concentration associated with a selected reduction in maximum number of organisms (e.g., 20%). The projected declines in abundance of aquatic organisms are interpreted within the larger context of biological responses to increasing levels of stress (i.e., a biological condition gradient). Projections of iron concentration associated with multiple levels of reduction are selected to establish acceptable levels of change in the various tiers of a biological community. The bioassessment-based benchmarks that we establish for total iron (0.21 and 1.74 mg/L) are based on the assumption that if ecological effects-based criteria for total iron are derived and applied, the structure and function of the aquatic community will be protected.

Summary1 It is widely recognized that negative effects of anaerobic stress on growth and survival of flooded plants influence the distribution of numerous species. Less explored is the possibility that heterogeneity in abundance of plants between habitats with distinct flooding regimes may also result from variation in rates of herbivory or in the ability of plants to tolerate losses to herbivores.2 Flooding and herbivores were tested as factors underlying variation in abundance of two tropical forest palms, Socratea exorrhiza (which is associated with lowlands adjacent to streams) and Oenocarpus bacaba (which is more abundant on plateaux and upper edges of slopes). 3 In a bench experiment, seeds of both palms were either completely immersed in water for a period of 101 days or not subject to inundation. Flooding inhibited germination of both species but, as expected, the adverse effects were much stronger on Oenocarpus. 4 In a field experiment, seeds of both palms were planted with increasing levels of protection against herbivores on plateaux and in lowlands. Seeds were either not protected or placed within poultry-netting exclosures, half of which were sprayed with insecticide. 5 After 17 months, only Oenocarpus had experienced differential mortality between habitats, and this was clearly associated with the negative effects of flooding on seed germination in lowlands. In contrast, growth differed between habitats only for Socratea seedlings, where average above-ground biomass was greater in lowlands. 6 Although protection with exclosures and insecticide increased survivorship of both species, herbivores caused similar proportions of mortality on plateaux and in lowlands, and had no significant effect on seedling growth. Therefore at this site, herbivores do not appear to influence variation in abundance of species between habitats.

Association of species abundance with a continuous environmental variable is frequently tested with regression or correlation analyses. However, because these methods ignore the range and frequency distribution of levels of the variable occurring in the study area, they may generate misleading results. We give examples to illustrate the argument. A better approach to test the association between species abundance and a continuous variable should compare levels of the variable in the study area to levels of the variable occurring in sites occupied by the species. If a particular species abundance is not associated with a given continuous variable, then the frequency distribution of levels of this variable measured where individuals of the species occur should mirror the frequency distribution of levels of the variable measured over the study area. We explain how to use the one-and two-sample KolmogorovSmirnov statistics to compare the cumulative relative frequencies of levels of the variable where individuals are present with points in the study area. We discuss the statistics, assumptions, limitations, and advantages of these tests.