Toxicity of proton–metal mixtures in the field: Linking stream macroinvertebrate species diversity to chemical speciation and bioavailability

Abstract

Understanding metal and proton toxicity under field conditions requires consideration of the complex nature of chemicals in mixtures. Here, we demonstrate a novel method that relates streamwater concentrations of cationic metallic species and protons to a field ecological index of biodiversity. The model WHAM-F_TOX postulates that cation binding sites of aquatic macroinvertebrates can be represented by the functional groups of natural organic matter (humic acid), as described by the Windermere Humic Aqueous Model (WHAM6), and supporting field evidence is presented. We define a toxicity function (F_TOX) by summing the products: (amount of invertebrate-bound cation) × (cation-specific toxicity coefficient, α_i). Species richness data for Ephemeroptera, Plecoptera and Trichoptera (EPT), are then described with a lower threshold of F_TOX, below which all organisms are present and toxic effects are absent, and an upper threshold above which organisms are absent. Between the thresholds the number of species declines linearly with F_TOX. We parameterised the model with chemistry and EPT data for low-order streamwaters affected by acid deposition and/or abandoned mines, representing a total of 412 sites across three continents. The fitting made use of quantile regression, to take into account reduced species richness caused by (unknown) factors other than cation toxicity. Parameters were derived for the four most common or abundant cations, with values of α_i following the sequence (increasing toxicity) H⁺ < Al < Zn < Cu. For waters affected mainly by H⁺ and Al, F_TOX shows a steady decline with increasing pH, crossing the lower threshold near to pH 7. Competition effects among cations mean that toxicity due to Cu and Zn is rare at lower pH values, and occurs mostly between pH 6 and 8.

Introduction

Assessment of the toxic effects of metals and protons in the environment would benefit from an ability to deal with mixtures. Despite this, recent treatises on metals contamination of aquatic systems give scant treatment to the topic of mixture toxicity (Adams and Chapman, 2007, Luoma and Rainbow, 2008). To interpret the results of laboratory single-metal toxicity studies, the concept of multi-substance Potentially Affected Fractions has been developed, in which results from experiments with single toxicants are used to describe mixtures, assuming either concentration or response additivity (De Zwart and Posthuma, 2005, De Zwart et al., 2006). The Biotic Ligand Model has been applied in a meta-analysis study of metal mixtures, by assessing competition for binding at the biotic ligand between toxic metals (Zn, Cu, Cd), calcium and other major cations (Kamo and Nagai, 2008). A ‘saturation’ model, where toxicity is assumed to be a function of the amount of metal bound to a specific binding site on the organism, has also been tested for mixtures (e.g. Borgmann et al., 2004, Norwood et al., 2007). Expanding these approaches to create predictive models for the full range of conditions, mixtures and organisms in environmental systems would be a formidable task as a large quantity of new data would be required. A pragmatic means of combining single-metal Environmental Quality Standards (EQS) for the field assessment of mixtures is through the use of Cumulative Criterion Units (Clements et al., 2008), calculated as the sum of the ratios of the concentration of each metal to its individual EQS. However, this only predicts an acceptable limit, not a concentration–response relationship, and as applied to date does not take account of chemical speciation and bioavailability.

A second issue in toxicity assessment is the need to demonstrate that estimates based on laboratory data apply in the field (Luoma and Rainbow, 2008). While this has been researched quite extensively for protons and aluminium in relation to acidification (Gensemer and Playle, 1999), results for “heavy metals” are sparse. This is clearly important to establish the credibility of EQS, to develop remediation measures, and more generally to understand the relationship between exposure, toxicity and ecological damage.

In the work presented here, we took a field-based approach to address both issues by developing a model, WHAM-F_TOX, that describes the speciation, bioavailability and toxicity of metal mixtures, and is fitted to field ecological observations. We used the Windermere Humic Aqueous Model (WHAM; Tipping, 1994, Tipping, 1998) to calculate both aqueous chemical speciation and to estimate the accumulation of protons and metals by freshwater macroinvertebrates. These model predictions were compared to actual measurements of metal levels in field collected invertebrates. As an ecological response variable we employed macroinvertebrate species richness, specifically of Ephemeroptera, Plecoptera and Trichoptera. The EPT index is routinely employed as a measure of biological integrity and is sensitive to a wide range of stressors (Plafkin et al., 1989). It was appropriate for the present work in view of data availability, established sensitivity, widespread distribution, functional importance in stream ecosystems and sufficient variability to indicate graded responses to metals. We applied the model to published field chemistry and EPT data from over 400 individual streams, studied with respect to either acidification or the impacts of metals from abandoned mines.

Although analysis of field data is attractive because it addresses the problem directly, a drawback is that EPT species richness may be reduced by environmental factors other than cation toxicity. Such factors may be biotic (competition, predation, dietary exposure), physical (water flow, suspended sediment), or chemical (nutrients, anionic toxicants such as arsenic, organic pollutants). Furthermore, benthic macroinvertebrates may be in contact with hyporheic groundwater, with a different chemical composition to that of the main streamflow, and may experience variations in chemistry during hydrological events. However, such factors are not widely quantified, and for our analysis are therefore “unknown”. To take unknown factors into account in assessing metal and proton toxicity, we performed quantile regression modelling (Cade and Noon, 2003), which has previously been applied to assessing organism responses to metals in freshwaters, but without including chemical speciation (Pacheco et al., 2005, Crane et al., 2007, Linton et al., 2007). Fig. 1 is included to show how the different data sources, modelling and statistics are combined into the model.