Assessment methods and limitations
There are a number of assessment methods and approaches that more effectively characterize exposure, effects, and identify stressor importance. These include field-based species sensitivity distributions for both chemical contaminated sediments, noncontaminated TSS, temperature, salinity, and sediment burial 207–213, improved exposure models of clean and contaminated soils and sediments 214, 215, stressor toxicity identification and interactions 9, 184, 216, 217, in situ experimental exposures including biomimetic monitoring 4, 218, biological trait-based analysis 219, and data analyses methods for discerning reference versus impairment and stressor-effect relationships 16, 220.
It is easy to highlight the limitations of any assessment method, whether they are from ecological or methodological perspectives 1, 116. Nevertheless, all of the popular assessment methods have been used effectively to assess ecosystem quality, when they are used judiciously. While the science has improved, the judicious use of the various methods with suitable acknowledgement of their limitations is not altogether common. In addition, a growing body of recent literature is showing the importance of new chemical stressors, their accumulation in sediments, and their interactions showing greater than additive effects 221–224. The often common and significant stressors are frequently not characterized from an exposure-and-effects (risk) perspective. So, if the goal is restoring ecosystem quality, beneficial uses, or ecosystem services, how can that occur if the dominant stressors are not identified?
Few studies have actually evaluated the accuracy of assessment methods; however, previous studies usually document large error rates. In Ohio (USA), evaluation of indigenous biota showed 36% of the impaired stream segments could not be detected using water chemical criteria alone 225. Evaluations of sediment quality guidelines have been better, with prediction rates of 70 to 75% 226, 227. However, error rates of 25 to 30% (and higher for metals) seem unacceptable given the ecological and economic implications of some of the resulting risk management decisions. A comparison of multiple assessment methods (lines-of-evidence) at several sites with metal and organic contaminants, showed accuracy rates ranged from 50 to 60% 40. This suggests that no one method is adequate and points toward the use of multiple methods to best characterize whether or not sediments are significantly contaminated and the aquatic ecosystem is impaired 1, 8.
Tools for experimental designs
Recently, Downes 228 noted that sorting out the effects of multiple stressors would be best accomplished by using the basic components of good experimental design. Experimental designs are available to determine the effects of suspended and deposited clean and contaminated sediments under controlled laboratory conditions 180. Such studies help identify sensitive species and the causal stressor(s). The experimental designs should include exposing multiple species to different levels and types of suspended or depositional solids alone and in combination with other factors (such as contamination, salinity, and temperature) to determine direct effects of clean sediments versus other stressors. Ideally the frequency and duration of sediment exposures is also investigated 57. In addition, it is important to consider the connectivity between organism traits (see below) and their exposure, because this is essential when establishing actual exposures and casuality.
Mesocosms can be used in the laboratory or field to determine effects on populations and communities. Stream-side mesocosms and artificial streams with multiple interaction stressor treatments provide another way to discern which stressors are dominating. Culp et al. 229 constructed a portable mesocosm system that contains 16 circular streams. Here, indigenous benthic biota were introduced to the streams, and natural stream water was combined with pulp mill effluent to simulate discharge dilution effects and separate nutrient-related stress from other factors. Many simple mesocosm designs have been used to assess the role of bioturbation on sediment resuspension and contaminant flux to filter-feeding organisms 119. It is more difficult to create sustained sediment resuspensions within mesocosms without substantially affecting relevant physico-chemical parameters such as dissolved oxygen.
Sublethal effects or subtle interactions that occur in the field are not easily measured in standardized sediment toxicity tests 230, 231. Laboratory toxicity testing can result in artifacts that may alter exposures found in situ, such as changes in: oxidation, redox, pH, sorption and complexation, microbial activity and their by-products (e.g., ammonia), nonequilibrium conditions, organic material via sieving, predation, food availability, ultraviolet light, flow, and suspended solids 5, 9, 40, 232–235. For example, seasonal complexation of Cu to DOC was affected by solar irradiation, causing dissolved organic matter to photochemically degrade and release Cu2+129, 236.
Many have called for more population- and community-based approaches looking at food chain-based effects, integrating key ecological factors beyond laboratory-based, single-species testing 6, 237, 238. This includes a greater focus on ecologically relevant endpoints not traditionally used in ecotoxicology, such as predator–prey interactions, drift, preference-avoidance, and other behavioral measures. These endpoints have been shown to be more sensitive than mortality, growth, or reproduction measures of sediment toxicity and habitat degradation due to uncontaminated fine sediments 3, 6, 181, 239–242. In food webs with organisms interacting strongly through predatory or competitive relationships, indirect effects of contamination will arise 6, 243. Manipulative experiments conducted in the field can be used to test for effects of metal-contaminated sediments on multiple components of an ecosystem simultaneously. When sediment fauna are negatively affected by metal contamination, the recruitment of invertebrates to patches of hard substrate directly above the sediments may increase (N.A. Hill, personal communication). This is likely due to a reduction in infaunal predators such as annelids and crustaceans.
Hence, another useful approach for discerning sediment, habitat, and water quality stressors and factors that control bioavailability is by using caging, colonization, and transplant experiments 9, 181, 244–246. Because sediments are usually altered chemically, physically, and biologically when they are sampled, there is a growing science that uses in situ observations, which can accurately document exposure and effects. The exposure of caged organisms has been recommended for fish and benthic macroinvertebrates to better characterize site and source exposures, thereby allowing for improved risk predictions or measures of remediation effectiveness 4, 15. Field-based tissue residues have been one of the oldest and most common assessment methods 247. For sediment quality assessments, both fish and benthic macroinvertebrates, particularly bivalves, have been commonly collected and tissue levels compared to sediment concentrations. This technique was used in a before–after, control–impact experimental design to establish that dredging increased the bioavailability of sediment contaminants 248. By separating sediment and water exposure compartments and documenting upwelling versus downwelling conditions using mini-piezometers, exposure sources and related toxicity can also be established 9, 10.
Some excellent papers on methods for determining free metal ion concentration, labile species fraction, metal complexation capacity in waters, and sediment flux have been published 218, 249–262. A wide range of biomimetic approaches exists 263. These include organic/plastic fibers, tubes, bags and sheets, gel probes, micro-ion selective electrodes, voltammetric electrodes, and optodes. These methods have provided new insights into controls on organic matter mineralization, benthic fluxes, impacts of redox alterations on metal fate, and geochemical reaction pathways 263. As nanotechnology improves, electrodes will become smaller and more sensitive, and provide a way to determine chemical speciation of metal (organic and inorganic metal complexes and free metal) in situ by encasing microelectrodes into diffusive gradients in thin film- or diffusive equilibrium in thin film-type membranes 263.
In contaminated sediment scenarios, new stressor toxicity identification methods include whole sediment manipulations or in situ exposures with various stressor partitioning methods and substrates that may reduce the likelihood of artifacts 9, 245, 264. The phase I type toxicity identification evaluations use similar resins to those used in the U.S. EPA methods, however organisms and benthic communities are exposed directly to pore or surface waters for 24 to 96 h directly in the field. These methods have been shown to be more sensitive than side-by-side laboratory-based toxicity identification evaluation tests, suggesting manipulation artifacts are causing a loss of toxicity. They effectively identified which stressors dominated at contaminated field sites by separating nonpolar organic, metal, and ammonia fractions. In addition, when combined with modified cages to restrict suspended solids or remove solar ultraviolet radiation, adverse effects from turbidity and photo-induced toxicity are possible 9. Recent lab-based toxicity identification evaluations have effectively documented the importance of sediment-associated PAHs and pyrethroids as dominant sediment stressors in urban watersheds, and chlorpyrifos dominated in agricultural watersheds 205, 265, 266. The importance of sediment metals in urban lakes is declining, as evidenced by a nationwide survey of cores from 1970 to 2001 267. Median changes ranged from −3% (Hg, Zn) to −46% (Pb), but remained elevated over undeveloped watersheds. These declines suggest that insecticides, such as the pyrethroids and PAHs will increasingly be the toxicants of concern in urban sediments.
A promising new assessment tool, particularly for benthic invertebrate-based ecosystem quality assessments, is the use of biological traits (morphological or functional) to develop geographically broad (continental) lotic ecosystem assessments of dominant stressors 196, 219, 268–271. This approach may assist in discerning effects that are due to the physical presence of solids, as opposed to solids-associated chemicals. Proponents of this approach caution against the blind use of excessive traits for the indication of too may stressors and recommend a focus on mechanistic a priori predictions. In addition, one stressor may affect many traits, thus confounding other stressor-specific relationships 271. For this field to progress, larger databases and poorly studied taxonomic groups are needed 270, 271.
Perhaps the most important aspect of an experimental design aimed at assessing when sediments are stressors, is establishing the appropriate reference conditions. In a regulatory sense, this may be straightforward, whereby impairment is simply based on exceedance of environmental quality benchmarks. However, determining what constitutes ecologically significant impairments from a high-quality state is not simple. Hawkins et al. 16 noted that researchers have become increasingly more sophisticated in their approaches for determining reference conditions using site-specific modeling. These approaches have been based on ecological, thermal, hydrologic geomorphic, and chemical benchmarks. These advances have better linked the spatial and temporal dynamics of biota and their natural environment, and are inextricably linked to how well their environments are characterized in terms of accuracy and precision. Useful approaches that have emerged include variations of the River Invertebrate Prediction and Classification System 272–274, ecoregions 275, landscape and typological classifications 276, 277, the Benthic Assessment of Sediment 278, 279, and others 16. Similar approaches in marine and estuarine systems are not as advanced, particularly where taxonomic clarity is not well established. However, extensive efforts are being made around the world to develop integrative tools and methods to assess the ecological health of estuaries, and substantial advances have been made by U.S. and European researchers 280, 281.
A large number of landscape approaches to studying aquatic ecosystems have established causal linkages between landscape variables and biota 62, 282. These have ranged from microhabitat patches to regional in scale, with high to low resolution, respectively. As noted by Allan 283 in a review of land use effects on streams, there has only been moderate success in quantifying the underlying mechanisms due to covariation of natural and anthropogenic factors over multiple scales, and due to legacy and nonlinear responses. Hawkins et al. 16 suggest that these schemes tend to produce overly coarse estimates that are lacking in accuracy and precision, and should be replaced by predictive modeling approaches. These will be based on a better understanding of natural variability of ecological, physical, and chemical characteristics, which then allows one to discern significant effects (impairment) from sampling and prediction error.
By using data-rich models that are statistically based, the decision of reference versus impaired areas along with stressor rankings is possible and can enhance the decision-making process. Two such ecoepidemiological approaches were independently applied to the same environmental monitoring dataset of biological, physical, and chemical variables for the State of Ohio. The methods are the effect-and-probable-cause pie diagram method and the WoE-weighted logistic regression method 173, 220. Both methods yield predictions of local impacts and their probable causes, which provided a statistical ranking of the dominant stressors.
Cross-validation of these models demonstrated that the methods yield significantly similar results in the identification of stressors impacting local fish communities and their relative influence 173. However, key differences were also observed between the methods that reflected the variance in objectives and sensitivities of each. The findings show that scientific interpretation of eco-epidemiological analysis output requires understanding of method distinctiveness, and suggest the potential value of utilizing multiple methods as lines of evidence in an environmental assessment 173. Marine studies of sediment contamination using the sediment quality triad found that salinity and grain size, not contamination, predicted benthic community responses best 284. Obviously greater amounts of colocated physical, chemical, and biological data will provide stronger spatial and temporal characterizations, thus a greater certainty of which parameters have the most significant stressor–response relationships.
Another useful eco-epidemiological model is SPEARorganic (average community sensitivity to organic toxicants) that has been demonstrated to distinguish between the effects of natural longitudinal lotic and organic toxicants using principal components analysis 285. Benthic richness and diversity responses were linked to petrochemical and synthetic surfactant exposures and separated from natural factors (altitude, velocity, temperature width, nitrate, macrophytes, periphyton cover, substrate size, and habitat heterogeneity). Another study using SPEAR found that runoff potential, stream width, and the presence of clay and dead wood in sediments predicted benthic assemblages 286. This approach provides another useful tool for separating natural and anthropogenic stressor responses.
Role of sediment and biota patchiness
A recent review of the benthic community patchiness literature stated that the theories have outpaced experimentation 62, which suggests that while we know patchiness is important, we do not know how to measure its significance. It has been posited that if disturbed community patches are smaller than one meter and benthic invertebrate generation times are less than a year, then recovery from disturbance will be rapid 295. However, as generation times and patch size disturbance increases, recovery times lengthen. Nevertheless, patchy sediment (or biofilm) contamination at a small scale may be ecologically significant for some communities and driven by habitat matrix retention 296. Roberts et al. 297, 298 found that storm water impacts were highly ephemeral, whereas bioaccumulation of stormwater contaminants by macroalgae had ongoing chronic toxic effects on herbivorous epifauna. Moreover, different macroalgal species accumulated contaminants to different degrees within the same site, creating a patchwork of varying contamination loads that organisms could choose from. We will advance our ability to predict the effects of heterogeneously distributed contaminants as we develop the field of landscape ecotoxicology 299, and this may be possible with existing data sets.
The U.S. EPA performed a national sediment quality survey that reviewed data for 19,398 stations in 5,695 river reaches. Sampling bias tended to occur towards stations that likely had sediment contamination 300. They identified areas of probable concern for sediment contamination where the exposure of benthic and fish communities would occur frequently. These watersheds had 10 or more sampling stations classified as having probable adverse effects. We reanalyzed this dataset using a much higher spatial resolution (1:100,000) compared with reach file version 1 (RF3) used in the U.S. EPA report (RF3 at 1:250,000 to 1:500,000) to better delineate local spatial variation. The patchiness of Cu- and Zn-contaminated sediments exceeding sediment quality guidelines in the United States was conducted for river reaches (16-digit reach level RF3, National Hydrology Dataset) on which four or more samples occurred (Figs. 2 and 3). River reaches that had one or more samples with SQG exceedances for Cu and Zn, were primarily within the lowest sample exceedance ranges (1–40%), indicative of more localized, patchy, sediment contamination. An exception was observed at the highest range (80–100% sample exceedance), in which approximately 10% of river reaches had high rates of threshold effect levels exceedances for both Cu and Zn, and 1 and 3% of river reaches had high rates of probable effect levels exceedances for Cu and Zn, respectively. These river reaches are examples of waterbodies potentially dominated by sediment contamination, with a more widespread spatial extent of contaminated sediment as compared to other river reaches. While localized sediment contamination may dominate ecological status on a highly local level, other environmental stressors acting on a wider spatial extent, including habitat quality, flow regime, water chemistry, and point and nonpoint sources of other contaminants, can play a major role in ecological status at the watershed (reach) level. Identification of contaminated sediment as a major (dominant) watershed stressor is, therefore, more convincing for river reaches with high rates (widespread) of SQG exceedances compared with river reaches having low rates (localized) of SQG exceedances. Where contaminated sediments are patchy in their distribution, this will lower their potential to be a dominant stressor in aquatic systems.
Figure 2. Distribution of copper sediment quality guidelines (SQG) exceedance rate (per river reach) for all reaches evaluated. Low percentages of samples exceeding an SQG indicate localized (or absence of) contaminated sediment within a given river reach; high percentages indicate more widespread contamination within a given river reach. TEL = threshold effect level; PEL = probable effect level.
Download figure to PowerPoint
Figure 3. Distribution of zinc sediment quality guidelines (SQG) exceedance rate (per river reach) for all reaches evaluated. Low percentages of samples exceeding an SQG indicate localized (or absence of) contaminated sediment within a given river reach; high percentages indicate more widespread contamination within a given river reach. TEL = threshold effect level; PEL = probable effect level.
Download figure to PowerPoint