Large impoundments remove substantial amounts of sediment and nutrients from rivers and often limit production by downstream primary producers and secondary consumers. Nutrient levels and macroinvertebrate and fish abundance in the lower Kootenai River (7^th order, mean annual discharge = 454 m³/s) in Idaho and Montana declined dramatically after Libby Dam was built in 1972. A subsequent study implicated ultraoligotrophic conditions (total dissolved P [TDP] ≤ 2 µg/L TDP) as a principal causative agent and prompted an on-going experimental nutrient-addition program for the Kootenai River downstream from Libby Dam, with dosing at the Idaho—Montana border. Pre-treatment monitoring began in 2003 and liquid ammonium polyphosphate fertilizer (10-34-0) was added each year during the growing season from 2006 through 2010 with a target TDP concentration of 3 µg/L and TN:TP near 20:1. We studied benthic macroinvertebrate responses to the experimental addition and hypothesized moderate increases in invertebrate richness, abundance, and biomass with little change in assemblage structure. We used a before—after control—impact BACI design with macroinvertebrate samples collected pre- and post-treatment from July to early November 2003–2010 from fertilized and unfertilized reaches. After treatment, mean modified (Oligochaeta and Chironomidae subtaxa excluded) total abundance increased 72%, mean total abundance increased 69%, and mean biomass increased 48%. Abundance of Ephemeroptera, the principal insect order in the study area increased 66%. Filter-feeder abundance also increased, indicating increased suspended organic matter in addition to the attached forms consumed by other benthic macroinvertebrates. The first 5 y of experimental treatment resulted in increased food resources for resident native fishes with no major alteration of macroinvertebrate community structure or trophic pathways.

During the past century, the Kootenai River, Idaho (USA), has experienced cultural oligotrophication following extensive levee construction, channelization, wetland drainage, and impoundment. A multiyear, whole-river nutrient-addition experiment was undertaken to mitigate these effects. The river was dosed with liquid agricultural-grade ammonium polyphosphate fertilizer (10-34-0) from June through September 2006–2010 to achieve an in-river total dissolved P (TDP) concentration of 3.0 µg/L. A fine-scale monitoring program included 8 sites over a 20-km reach (2 upstream control sites, one injection site, and 5 downstream treatment sites). Nutrient addition did not significantly increase N and P concentrations in the water column, but it significandy increased chlorophyll accrual rates and densities of edible green algae and diatoms. Nutrient addition significantly reduced NO₃^_ NO₂^_ concentrations, atomic TN:TP ratios, and densities of inedible cyanophytes. Mean NO₃^_ NO₂^_ values decreased along a downstream gradient below the nutrient-addition site, whereas chlorophyll accrual rate typically peaked immediately downstream from the nutrient addition site then decreased progressively down-stream. Our study showed that nutrient addition is a useful river restoration technique for the Kootenai River.

Storm-flow disturbances are frequent during the wet season of Australian tropical savannas. We examined benthic algal resistance and resilience in open-canopy streams in the Daly River watershed. Storm flows occurred every 2 to 3 d at 1^st- to 4^th-order sites, with sharp rises and falls and relatively long periods of shallow, low-turbidity base flow. At a 5^th-order site, storm-flow duration was longer and base flows were deeper and more turbid. We hypothesized that: 1) storm flow would dislodge benthic algal biomass, 2) baseflow biomass would be low, 3) taxon richness would be low, and 4) algal composition would be dominated by resistant algae with prostrate or erect growth forms or by fast growing colonizing algae. Hypothesis 1 was supported. Storm flows dislodged ~93% of epilithic biomass. Support for hypothesis 2 was equivocal. At the 5^th-order site, sand mobility prevented establishment of benthic algae until seasonal flows receded. At the other sites, epilithon net growth rates were much greater than rates in some temperate streams. Benthic biomass was typical of temperate oligotrophic streams, but maximum biomass was typical of mesotrophic streams. We attributed the relatively rapid growth and high biomass to warm water temperatures (mean = 29°C, maximum = 36°C), high incident light, rapid algal nutrient uptake, loss of grazing invertebrates caused by storm flows, and physical impediments to fish access. Hypothesis 3 was not supported. Mean taxon richness was high because of the occurrence of rare taxa. Hypothesis 4 was not supported. Epilithic algal biomass was dominated by resistant filamentous chlorophytes. Epilithic algal resistance was similar to resistance in higher latitude streams, but resilience was greater. Epilithic algae potentially could supply autochthonous C to the Daly River and other tropical aquatic food webs.

Channelized streams are common in North American agricultural regions, where they minimize water residence time and biological nutrient processing. Floodplain restoration done via the 2-stage-ditch management strategy can improve channel stability and nutrient retention during storms. We examined the influence of floodplain restoration on whole-stream metabolism by measuring gross primary production (GPP) and ecosystem respiration (ER) for 1 y before and 4 y after restoration of an upstream, unaltered control reach and a downstream, restored reach. Both reaches were biologically active and dynamic. GPP ranged from 0.1 to 22.1 g O₂ m^-2 d^-1, and ecosystem respiration (ER) rates ranged from -0.1 to -38.7 g O₂ m^-2 d^-1. We used time-series analysis and found that GPP increased postrestoration during floodplain inundation when expressed per unit length, but not per unit area, of stream. GPP was more resilient post- than prerestoration and returned to prestorm levels more quickly after than before floodplain construction. In contrast, the floodplain restoration had no effect on ER or on any metric of metabolism during base flow. Overall, we showed that floodplain—stream linkages can be important regulators of metabolism in restored agricultural streams.

Sewage effluents are recognized as one of the most common sources of river degradation. However, very few investigators have tried to evaluate ecosystem recovery after cessation of wastewater treatment plant (WWTP) discharge. Our goals were to: 1) analyze invertebrate community responses to chemical water quality and habitat changes after wastewater treatment improvement, and 2) evaluate the abilities of taxonomy-based and trait-based approaches to detect and explain community recovery processes in a multistressor context. We studied the benthic macroinvertebrate community of 3 sites of a heavily impaired stream (the Vistre River, France) during a 4-y period that included the decommissioning of a deficient WWTP (WWTP-A) and the commissioning of a new one (WWTP-B). We assessed the recovery of the benthic community by comparing observed taxonomic and trait-based (i.e., functional) metrics at the study site to values estimated with information from least-impaired reaches of the same river type. Most taxonomy-based characteristics of benthic communities subjected to WWTP-A recovered in 3 mo, but the recovery time of several trait-based characteristics was ∼2 y. No change was observed in taxonomic and trait-based diversity during the 4-y study. Taxonomy-based metrics detected the first signs of river reach recovery rapidly, but combinations of trait-based metrics and taxonomic abundance-based metrics are more likely to identify functional recovery of invertebrate assemblages faced with water-quality improvement, even when multiple stressors impaired upstream reaches.

Aquatic macrophytes suffer from widespread and profound habitat deterioration that has led to dramatic changes in species distributional patterns. We explored linkages between species functional traits and abundance patterns by analyzing mean community trait values in 1083 Danish lowland streams. We expected that widespread macrophytes would share common traits and that species with smaller ranges would be negatively associated with these traits. We selected 11 traits to characterize the ecological features of 52 amphibious and submerged species. We examined the relationships between species abundance and species traits with multivariate ordination and coinertia analysis (COIA), and identified groups of species with similar distributional patterns and ecological traits by hierarchical cluster analysis. Local abundance and geographical range size of submerged and amphibious species were significantly positively related. Species abundance and species traits tables were significantly correlated, and species were separated into 5 groups based on life form, productivity, meristem characteristics, ability to fragment, and the size of the root-rhizome system. Widespread plant species in Danish lowland streams shared common traits, such as meristem, dispersal, and productivity characteristics, and we infer that these traits are likely to have adaptive value in eutrophic and hydromorphologically disturbed habitats. On the other hand, species with narrower range sizes were characterized by a different set of trait characteristics. We conclude that species abundance and distribution patterns are closely correlated with ecological trait characteristics and that traits associated with resilience to anthropogenic disturbance confer success in Danish streams.

Selective feeding and differential nutrient excretion by aquatic invertebrates plays a substantial role in nutrient recycling. Grazing larvae of the caddisfly Glossosoma intermedium construct a portable case for protection that also serves as a good substrate for periphyton colonization because it is constantly fertilized by larval excreta. We tested whether case periphyton was nutrient enriched compared to streambed periphyton and whether selective feeding by caddisfly larvae on case periphyton facilitates P remineralization. We measured total N, total P, and N:P in stream water and streambed cobble periphyton in 3 western Wisconsin streams. We collected larvae and measured N and P content and N:P of case periphyton, G. intermedium body, and G. intermedium excretion products. Cobble periphyton N:P at 2 streams suggested P limitation (368 ± 109 and 50 ± 10), but case periphyton N:P at those streams did not (11 ± 3 and 11 ± 0.8, respectively). Neither cobble nor case periphyton N:Ps suggested P limitation at the 3^rd stream. Measured excretion N:P was more similar to the excretion N:P predicted for caddisfly grazing on case periphyton than for caddisfly grazing on cobble periphyton. Our results suggest that larval excretion alleviated P limitation of case periphyton and that case periphyton may serve as an important dietary resource for the grazing caddisfly larvae. Feeding on this P-rich case periphyton promotes P remineralization in P-limited, lotic ecosystems.

One common stoichiometric approach to predicting patterns of nutrient release (excretion egestion) by animals in aquatic ecosystems is to base predictions on elemental mass-balance constrained by homeostatic maintenance. An easily measured resource composite (i.e., seston, epilithon, or leaf litter) often is used to represent ingested stoichiometry, but whether such a composite is a good indicator of food actually ingested is a relatively unexplored assumption. We examined the application of a stoichiometric model to the diets of 4 generalist stream invertebrates. We fed 3 trichopteran and 1 amphipod taxa rations consisting of cultured algae, stream epilithon, and several species of conditioned leaf litter. The rations ranged widely in C:N from 10 to 69 (molar) and in C: P from 165 to 3500. After a 2-d feeding period, we measured NH₄ and PO₄^3- excretion, and C, N, and P egestion rates. The relationships observed between the stoichiometries of release and ration were unexpected. Total N: P release rates conformed to stoichiometric predictions for only 1 taxon. Excretion and egestion rates and ratios were generally similar across diets and rarely varied with ration stoichiometry. These patterns were the result of smaller-than-expected responses to leaf-litter rations, which were the most imbalanced relative to body stoichiometry. Analysis of the C:N stoichiometry of foregut material for 2 taxa showed selective ingestion of an N-rich fraction of leaf litter, in 1 case reducing an apparent 8.4:1 C:N imbalance between diet and body composition to 1.5:1. Our results show that selective feeding can reduce potential stoichiometric imbalances, altering patterns of nutrient release relative to expectations based on bulk-diet stoichiometry. Assuming that stream invertebrates consume materials stoichiometrically similar to a resource composite can obscure understanding of stoichiometric imbalances and the role of invertebrates in nutrient cycles.

Permafrost disturbance (shoreline retrogressive thaw slumping [SRTS]) causes solute-rich terrestrial inputs to Arctic tundra lakes. Eight upland tundra lakes (3 undisturbed [U], 5 disturbed [D]) in the Inuvik region of the Northwest Territories, Canada, were sampled to assess the effects of SRTS on benthic invertebrate community structure, abundance, and whether localized SRTS effects in D lakes could be discriminated by comparing samples taken adjacent to (Da) or further from (Do) the disturbance. Community composition and abundance differed between U and D lakes. Macroinvertebrates were more abundant in areas adjacent to SRTS (Da) than in areas away from SRTS (Do), but community composition did not differ between Da and Do areas. Abundance was >2× greater in D than in U lakes. Ostracoda abundance was 4× greater in D than U lakes and 2× higher in Da than Do areas. Nematoda abundance was ∼10× higher in D than U lakes, whereas Chironomidae abundance was lower in D than U lakes. Bivalvia and Oligochaeta had similar abundance in D and U lakes, and both groups were more abundant in Da than Do areas. Differences in abundance and community composition were related primarily to higher concentrations of Ca² in D than U sediments and to higher organic C and N in U than D sediments. Sediment organic C and Mg concentrations, macrophyte biomass, and dissolved organic C concentrations best explained differences in community composition among lake types and areas. Our study adds to our understanding of cascading changes in the foodweb structure and ecological states of freshwater ecosystems caused by climate change in the Arctic.

Rare species are difficult to study or conserve. Beloneuria jamesae Stark and Szczytko (Cheaha Stone) (Plecoptera:Perlidae) is endemic to the Talladega Mountains, Alabama, USA, rarely collected, and considered imperiled. We collected stonefly larvae at 181 locations and found Cheaha Stone at 33% of these. Species distribution modeling by classification tree analysis identified elevation, stream size, and permanence as key habitat variables. Cheaha Stone was most prevalent in small permanent streams at higher elevation but extended into larger streams with sources at high elevation. Populations occurred infrequently in small streams at lower elevations. The known distribution of Cheaha Stone is protected within public lands, but climate warming and possible increases in drought frequency and severity could restrict the species to fragmented populations in small streams draining the highest summits of this low mountain range. Our georeferenced data and explicit classification tree are foundations for monitoring and management and are guidelines for seeking Cheaha Stone outside its known geographic range.

Growth and longevity of freshwater mussels (Unionida) are important for defining life-history strategies and assessing vulnerability to human impacts. We used mark—recapture and analysis of shell rings to investigate age and growth of the hyriid, Westralunio carteri, at 5 sites in southwestern Australia. We tested the utility of the in situ marker calcein for validating the assumption of annulus formation in adults. Calcein was incorporated into the shells of all recovered individuals, but it provided an interpretable reference mark in only 4 of 16 individuals. These 4 individuals produced 1 shell ring subsequent to the mark, supporting the assumption of annulus production during the austral winter. Maximum age ranged among populations from 36 to 52 y and maximum size ranged from 72.9 to 82.8 mm. Mean age and length did not differ between sexes, and growth trajectories differed between sexes at only 1 site. Estimates of growth measured by the von Bertalanffy growth constant, K, ranged from 0.021 to 0.336 among sites. Estimates from mark—recapture experiments were 20 to 52% lower than values from shell annuli at all sites except 1 where K from shell annuli was ∼½ that estimated from mark—recapture. Both methods showed a positive relationship between K and mean water temperature among sites, suggesting a role of riparian shading in regulating stream temperature, and hence, indirectly influencing mussel growth. Mussel growth and mean N or P concentrations were not related among sites, but total N at the site with highest mussel growth was >2× higher than at any other site. Westralunio carteri is a long-lived, slow-growing bivalve. Maximum age, K, and probable age at maturity (4–6 y) are similar to other slow-growing freshwater bivalve groups. This suite of life-history traits is considered an adaptation for stable aquatic habitats. Therefore, W. carteri can be expected to adapt poorly to human impacts, such as riparian clearing and water extraction, which increase the temporal variability of environmental conditions in streams.

Freshwater ecosystems are highly vulnerable to warming climates. However, macroecological studies of climate-change effects on aquatic biodiversity are rare because of a lack of standardized large-scale surveys, e.g., along large latitudinal gradients. Our goal was to assess the overall richness pattern and projected differences in present and future patterns of the stream insect orders Ephemeroptera (E), Plecoptera (P), Trichoptera (T), and combined EPT along an extensive latitudinal gradient across North America (30 to 70° N). We applied Bioclimatic Envelope Models (BEMs) to project present-day and future climatically suitable areas for EPTs on a spatial resolution of 10 arc-min (20 km × 20 km) across North America. To overcome issues related to spatially biased sampling, we assessed climatically suitable areas (CSAs) for each genus and modeled potential generic richness/grid cell, rather than assessing observed generic richness patterns directly. Projected present-day generic richness was greatest between 40 and 48°N latitude, with peaks at 44, 45, and 47°N for the E, P, and T orders, respectively. Our models projected that CSAs would shift an average 4.2 to 5.2°, 4.4 to 5.3° and 3.4 to 4.1°N latitude by 2080 for E, P, and T genera, respectively. Overall, the present-day projected generic EPT richness is highest in the warm and cool temperate zones and shows a unimodal richness pattern that is projected to shift northward under climate-change conditions. A similar northward shift of richness patterns might also apply to other aquatic insects with relatively narrow thermal sensitivity and terrestrial, winged adults, e.g., freshwater Diptera or many aquatic Coleoptera. This large-scale application of genus-by-genus models gives a first approximation of the likely consequences of climate-change effects on freshwater biota across North America.

Disentangling the influences of multiple environmental factors on ecosystem integrity is not straight-forward because environmental factors may interact and biotic responses may be nonlinear. We aimed to understand better the relationships between freshwater invertebrate assemblages and multiple, interacting environmental factors. We analyzed stream monitoring data for 689 sampling sites in the state of Ohio (USA) with Boosted Regression Trees (BRTs). We used 16 environmental predictors covering geography, water chemistry, physical-habitat quality, and toxic pressure. We represented freshwater invertebrate assemblages by the Invertebrate Community Index (ICI) and its 10 component metrics. The ICI was mainly related to physical-habitat quality, nutrient concentrations (P and N), and pH. Responses of the ICI component metrics to physical-habitat quality and water-chemistry variables were similar and were associated with amplified importance of these predictors for the ICI, whereas heterogeneous responses of the component metrics to geographic variables appeared to cancel each other out at the level of the ICI. Models including predictor interactions explained 22 to 54% of the deviation in the biotic endpoints, whereas the no-interactions models explained 14 to 47%. The gain in predictive power was largest between the no- and the pairwise interaction models and decreased rapidly for each additional interaction level. We conclude that a focus on pairwise interactions is a good compromise between higher predictive power and interpretability of the results.

The nearshore zones of the Great Lakes provide essential habitat for biota and are perhaps the region of the lakes most susceptible to human impacts. The objective of our study was to develop a fish habitat classification for the nearshore zone of Lake Ontario based on physical characteristics of that zone, land cover in the surrounding watershed, and fish community patterns. Nearly 80% of the spatial variation in fish community data was described by 2 physical variables (average fetch and bathymetric slope of the nearshore zone) and 2 land-cover variables (urban/industrial development and mixed forest cover) in adjacent watersheds. These variables are likely to be surrogates for other conditions in the nearshore, such as wave action, circulation, vegetation, and water quality. A 12-group fish habitat classification was developed from those variables. Validation and significance tests identified similarities and differences among the fish communities in the classes and indicated that the number of classes should be collapsed to 3: exposed, sheltered, and developed/urbanized. In general, the western basin of the lake was developed, the central region was exposed, and the eastern region of the lake was a mix of exposed and sheltered classes. These results highlight that even in lakes as large as Lake Ontario, the nearshore fish community is influenced by watershed land cover, and emphasize that management or restoration of the nearshore ecosystem in lakes will require integration of aquatic, watershed, and land-cover management.

Accurate knowledge of the distribution of rare, indicator, or invasive species is required for conservation and management decisions. However, species monitoring done with conventional methods may have limitations, such as being laborious in terms of cost and time, and often requires invasive sampling of specimens. Environmental DNA (eDNA) has been identified as a molecular tool that could overcome these limitations, particularly in aquatic systems. Detection of rare and invasive amphibians and fish in lake and river systems has been effective, but few studies have targeted macroinvertebrates in aquatic systems. We expanded eDNA techniques to a broad taxonomic array of macroinvertebrate species in river and lake systems. We were able to detect 5 of 6 species (Ancylus fluviatilis, Asellus aquaticus, Baetis buceratus, Crangonyx pseudogracilis, and Gammarus pulex) with an eDNA method in parallel to the conventional kicknet-sampling method commonly applied in aquatic habitats. Our eDNA method showed medium to very high consistency with the data from kicknet-sampling and was able to detect both indicator and nonnative macroinvertebrates. Furthermore, our primers detected target DNA in concentrations down to 10^-5 ng/µL of total extracted tissue DNA in the absence of background eDNA in the reaction. We demonstrate that an eDNA surveillance method based on standard PCR can deliver biomonitoring data across a wide taxonomic range of macroinvertebrate species (Gastropoda, Isopoda, Ephemeroptera, and Amphipoda) in riverine habitats and may offer the possibility to deliver data on a more refined time scale than conventional methods when focusing on single or few target species. Such information based on nondestructive sampling may allow rapid management decisions and actions.

Accurate and precise estimates of relationships between stressors and environmental responses can inform management decisions most usefully when models can be easily interpreted. Here, we describe an approach for classifying lakes and reservoirs that can improve estimates of the relationships between total P (TP) and chlorophyll a (chl a) concentration, while preserving a model that can be readily interpreted by environmental managers and stakeholders. We selected classification variables statistically with a classification and regression tree in which relationships between TP and chl a were the terminal nodes of the tree. We developed a set of classification trees from bootstrapped replicates of the calibration data to explore a broader range of possible trees. We chose a final tree based on its predictive performance with a validation data set. The total N:TP mass ratio was the classification variable selected most frequently from a broad array of biological, chemical, and physical candidate classification variables. Relationships between TP and chl a in the resulting lake classes provided predictions that were substantially more accurate than predictions computed using nutrient ecoregions based on aggregations of Omernik Level III ecoregions, but predictions from a random forest model that averaged an ensemble of trees were even more accurate. Thus, the classification approach presented here sacrifices a small amount of predictive accuracy to retain a tree structure that is readily interpretable.

Biogeochemical studies done to elucidate sediment—water transfer of solutes and benthic reaction rates in limnic and marine ecosystems often rely on the study of porewater distribution and temporal dynamics of target solutes and subsequent diagenetic modeling. Hydrophilic poly-ether-sulfone (PES) membranes are used increasingly often to sample soil and sediment porewater because they are versatile and easily adapted to field studies in benthic biogeochemistry. Nevertheless, possible interference with components of filtered water by membrane surface properties impedes accurate measurement of concentrations of solutes in filtrates. We identified and described NH₄ sorption onto PES membranes. We quantified and constrained sorption characteristics of these samplers because of the potential influence of PES membranes on NH₄ filtrate composition. Maximum adsorption capacity was associated with low operational temperatures and reduced filtrate salinity. NH₄ partition coefficients (K>^*) > 8 × 10^-2 cm were measured at 5°C and 0‰ salinity. However, K>^* followed an exponential decay curve with salinity and a negative exponential decay with temperature. NH₄ desorption from the membranes was rapid, but affected the 1^st mL of filtered sample after removal of internal volume. We put forward recommendations for improved practice to minimize the influence of membrane adsorption capacity on measurement of NH₄ composition of porewater samples when using PES membranes.

Although used in many jurisdictions around the world, analytical approaches of the Reference Condition Approach (RCA) to bioassessment of freshwater ecosystems have evolved quite slowly over the past 2 decades. For this special series of papers in Freshwater Science, researchers analyzed 3 data sets that included both benthic macroinvertebrate and environmental data from a number of reference sites. Australian Capital Territory (ACT) reference sites (n_total = 107) were wadeable streams in the upper Murrumbidgee River catchment, Australian Capital Territory, Australia. Yukon Territory (YT) reference sites were wadeable streams (n_total = 158) in the Yukon Territory, Canada, part of the Yukon River basin. Great Lakes (GL) sites (n_total = 164) were all nearshore (<20 m) lentic sites in the North American Great Lakes. For each data set, sites were divided into model-building (training) and model-testing (validation) groups. Each validation site was further subjected to 3 levels of simulated degradation based on the sensitivity of the biota to eutrophication. The analytical approaches ranged from standard or slight modifications of methods used in national programs (Australian River Assessment [AUSRIVAS], Canadian Aquatic Biomonitoring Network [CABIN]), to improved matching of sites to be assessed and appropriate reference sites, and Bayesian and machine-learning modeling. In comparing Type 1 error rates (proportion of validation sites deemed not in reference condition) and power (proportion of simulated impairment sites deemed not in reference condition), we found no obvious pattern among the 3 data sets or approaches. Approaches commonly used in RCA programs would benefit from incorporating newer methods that better match reference and test-site environments and build better predictive models.

Confidence in any bioassessment method is related to its ability to detect ecological improvement or impairment. We evaluated Australian River Assessment (AUSRIVAS)-style predictive models built using referencesite data sets from the Australian Capital Territory (ACT), the Yukon Territory (YT; Canada), and the Laurentian Great Lakes (GL; North America) area. We evaluated model performance as ability to correctly assign reference condition with independent reference-site data. Evaluating model ability to detect human disturbance is generally more problematic because the actual condition of test sites is usually unknown. Independent reference-site data underwent simulated impairment by varying the proportions of sensitive, intermediate, and tolerant taxa to simulate degrees of eutrophication. Model performance was related to differences in data sets, such as number and distribution of invertebrate taxa. Sensitive taxa tended to have lower expected probabilities of occurrence than more-tolerant taxa, but the distribution of taxa grouped by tolerance categories also differed by data set. Thus, the models differed in ability to detect the simulated impairment. The ACT model performed best with respect to Type 1 error rates (0%) and the GL model the worst (38%). The YT model performed best (10% error) for detecting moderate impairment, and the ACT model detected all severely impaired sites. AUSRIVAS did not assign most mildly impaired sites to below-reference condition, but a reduction in observed/expected values for some of the mildly impaired sites was observed. Models did not detect mild impairment that simply changed taxon abundances because presence—absence data were used for models. However, in comparison with other models described in this special issue (that did use abundance data), the AUSRIVAS model performance was comparable or better for detecting the simulated moderate and severe impairments.

The objective of this study was to evaluate the performance of 3 bioassessment models for reference data sets collected from the Australia Capital Territory (ACT), the Yukon River Basin (YT), and the Laurentian Great Lakes (GL) built following the standard Canadian Aquatic Biomonitoring Network (CABIN) method. To evaluate the models, we used validation reference-site data, which were artificially impaired to simulate 3 levels of eutrophication by varying the proportions of sensitive, intermediate, and tolerant taxa. Models correctly classified 56 to 62% of reference sites. Type 1 errors (assessing reference sites as degraded) were high for all data sets and ranged from 30 to 75%, in part because the biological communities of the validation sites extended to or beyond the range of the reference-site data used to build the models. Capturing the full range of ecological variation with adequate sample size is critical for reference-condition approach (RCA)-type models. Type 2 errors (assessing degraded sites as in reference condition) varied greatly among data sets and for each reference group within each data set. Resource managers must carefully consider the risks associated with making errors. Thus, standard methods for quality assurance of assessment models should include simulated data so that error rates and adjusted assessment thresholds can be reported to ensure that degradation can be detected and that undisturbed sites are not mistakenly subjected to unnecessary management action.

Reference Condition Approach (RCA) predictive models are used to assess a test site against reference sites probabilistically matched based on habitat. These models are the basis of several major national stream bioassessment programs in the UK, Australia, and Canada. In the usual approach to developing predictive models, discriminant function analysis (DFA) is used to assign a test site to a group of matched reference sites. These groups typically are established by classification of a macroinvertebrate assemblage and matched to the habitat attributes in a single-step DFA model. We examined an alternative to standard DFA in which a series of tiered models are used. This tiered method constructs a model for the 1^st division in a hierarchical classification, and then develops models for each further step in the hierarchical classification. We examined the method with 3 training and validation data sets. Validation data consisted of data from reference sites and those same sites after they underwent simulated impairment. We compared the tiered approach to the standard approach based on prediction accuracy and Type 1 and Type 2 error rates for each data set. The tiered DFA models were similar to or slightly better than the standard single-step DFA models in correctly matching validation sites to reference groups, but this improvement in accuracy did not necessarily translate into improved bioassessment error rates.

Reference Condition Approach bioassessment programs have been in place in the northern and Muskoka regions of Ontario, Canada, for many years. Assessments are carried out regularly to evaluate and monitor the effects of a variety of activities, including mining, forestry, and cottage development. These programs are run by the Co-operative Freshwater Ecology Unit (CFEU) at Laurentian University in Sudbury, Canada, and the Dorset Environmental Science Centre (DESC) in Muskoka, Canada. We applied 2 bioassessment methods used at the CFEU and DESC to 3 data sets that were subjected to simulated impact by nutrient enrichments to compare their performance with a number of other bioassessment methods. We used Assessment by Nearest Neighbour Analysis (ANNA) and a Redundancy Analysis (RDA) variation of ANNA with Test Site Analysis (TSA) to identify subsets of reference sites to compare with a given simulated impact test site based on habitat matching. We compared the benthic macroinvertebrate (BMI) communities and evaluated the differences between the validation and training sites to assess the degree of impairment. After assessing all impacted sites, we calculated Type 1 and Type 2 error rates. ANNA and RDA separated sites with different levels of simulated impact in an Australian data set of diverse benthic macroinvertebrate communities distributed along a habitat gradient. In contrast, our assessments did not perform well with 2 data sets for which the simulation did not behave as expected, perhaps because of impoverished communities.

The status of biological assemblages is often inferred by comparing observational assemblage data with reference data generated by a predictive model. Limiting environmental difference analysis (LEDA) applies the ecological concepts of distance decay of similarity and limiting factor theory to model reference data by extrapolation from samples collected at reference sites. I applied LEDA modeling to data sets from the Australian Capital Territory (ACT), Laurentian Great Lakes (GL), and Canadian Yukon Territory (YT) to test its accuracy and precision in predicting the composition of benthic macroinvertebrate samples from reference sites, and its specificity and sensitivity in detecting mathematically simulated human impacts on sample composition. LEDA models were significantly and substantially more accurate than null models for all 3 data sets, but not significantly more precise. They mostly assessed unimpacted samples correctly but seldom detected mild simulated impacts. However, they often detected moderate and usually detected severe impacts. Model performance for the test data sets probably was constrained by limitations of the environmental predictor variables and low taxonomic richness in many samples. The simulated impacts provided insight into model behavior but were limited in their scope and realism. I suggest that future simulations should separately test detection of different kinds of assemblage change, not only different intensities, and should address the consequences of human alteration of predictor variables.

Despite the existence of many approaches to reference-condition modeling, Bayesian statistical methods have not been used. We assessed whether a hybrid approach that combined features of existing reference-condition approaches with Bayesian model fitting and assessment of test sites could provide superior results to existing established methods. We used 4 Bayesian models of increasing complexity to develop and test reference-condition models for 5 biotic endpoints across 3 data sets. Our best models were comparable or superior to standard approaches (Benthic Assessment of Sediment, Australian River Assessment System) using the same data. Those of our models with the simplest endpoint (species richness) performed best. On average, those models with the simplest model structures also performed best, but differences in performance among models of different complexity were small. All models performed poorly at detecting the lower levels of simulated impact in the test data. However, these impacts were small relative to the variation among validation sites and consequent predictive uncertainty of the models. The Bayesian approach to reference-condition modeling shows promise as an alternative to existing methods. It also has advantages in terms of the ease of interpretation of model outputs. However, for the approach to be relevant, further development work should be driven by a perceived need to revise standard methods used by management agencies.

We developed a bioassessment tool (HYDRA) to predict the taxa present at a site based on the best performing machine learning tool (Support Vector Machines [SVM], Multi-Layer Perceptron [MLP], K-Nearest Neighbour [KNN]) for each taxon. HYDRA differs from standard models based on discriminant function analysis (DFA) in 2 main ways: 1) HYDRA predicts taxa directly without a priori reference-site classification, and 2) all environmental variables provided for model building contribute to predictions instead of only those that best explain differences among groups. Probabilities of taxon occurrence were used to calculate the Observed/Expected index (O/E50), based on taxa predicted with >50% probability of occurrence. We tested the hypothesis that a combination of models (HYDRA) would perform better than each model alone (SVM, MLP, KNN). We measured performance as: 1) taxon prediction accuracy, 2) precision given by the O/E50 standard deviation (SD O/E50), 3) accuracy of the validation O vs E linear regression, and 4) sensitivity to impairment. We used 3 data sets covering a wide range of environmental conditions (Yukon Territory, Great Lakes, Australian Capital Territory) and calculated O/E50 values for reference, validation, and sites with 3 known levels of simulated impairment. We created 4 quality classes (Good—Severe) and used the 10^th percentile O/E50 values of training sites to define the boundary between Good (= reference) and Moderate classes. HYDRA was the best solution for all data sets and was able to distinguish levels of impairment. Taxon prediction accuracy was not related to taxonomic group. Models (SVM, MLP, KNN) varied in accuracy among data sets, and accuracy seemed to depend on the distribution of the taxa across training sites. SVM provided good models, but showed poor sensitivity with 1 data set, which indicated inability to deal with low-richness communities.