Machine learning (ML) is increasingly considered the solution to environmental problems where limited or no physico‐chemical process understanding exists. But in supporting high‐stakes decisions, where the ability to explain possible solutions is key to their acceptability and legitimacy, ML can fall short. Here, we develop a method, rooted in formal sensitivity analysis , to uncover the primary drivers behind ML predictions. Unlike many methods for explainable artificial intelligence (XAI), this method (a) accounts for complex multi‐variate distributional properties of data, common in environmental systems, (b) offers a global assessment of the input‐output response surface formed by ML, rather than focusing solely on local regions around existing data points, and (c) is scalable and data‐size independent, ensuring computational efficiency with large data sets. We apply this method to a suite of ML models predicting various water quality variables in a pilot‐scale experimental pit lake. A critical finding is that subtle alterations in the design of some ML models (such as variations in random seed, functional class, hyperparameters, or data splitting) can lead to different interpretations of how outputs depend on inputs. Further, models from different ML families (decision trees, connectionists, or kernels) may focus on different aspects of the information provided by data, despite displaying similar predictive power. Overall, our results underscore the need to assess the explanatory robustness of ML models and advocate for using model ensembles to gain deeper insights into system drivers and improve prediction reliability.

The Alberta oil sands (AOS), Canada, represent a vital energy resource; however, the extraction of oil from these deposits poses significant environmental challenges. In particular, the eventual reclamation of significant volumes of waste materials is required. As a potential solution to incorporate and remediate AOS mine wastes in-situ, a pilot-scale pit lake, Lake Miwasin (LM), was constructed in 2017-18. It used a combination of coagulated and flocculated fluid fine tailings (bottom substrate) capped with a blend of oil-sands process affected water and runoff water from the surrounding catchment. The objective of the study described here is to assess the resultant surface water quality of this artificial lake and its regional comparability with natural systems. Furthermore, we explored the physicochemical characteristics, the mechanisms regulating lake water chemistry, and calculated various water quality indices for both LM and surrounding natural water bodies. Based on water quality indices, our study results showed that the overall water quality of LM's surface water was not yet similar to surrounding water bodies in the AOS region. Surface water from LM is slightly alkaline with elevated total dissolved solids (TDS) and differs chemically from surrounding water bodies. Predominant ions in LM are Na+, K+ (pore water from tailings), whereas surrounding water bodies exhibit Ca2+, CO3-, and HCO3- (gypsum, dolomite, and calcite dissolution). Therefore, water chemistry of LM is more typical of a Na-Cl water type, while surrounding bodies show a Ca-HCO3 water type. Evaluations based on different water quality indices highlighted the importance of monitoring specific variables in LM such as conductivity, TDS, alkalinity, and NH3. The analysis and assessment of LM surface water is imperative for gaining insight into the potential character and trajectories of water quality in future pit lakes and informing design considerations and application of this novel reclamation approach in the AOS region.

Trace elements play diverse roles in animal physiology ranging from essential micronutrients to potent toxicants. Despite animals accumulating many trace elements through their diets, relationships between trophic positions and biological concentrations of most trace elements remain poorly described. We report trophic transfer rates of Al, As, Ba, Cd, Co, Cu, Fe, Hg, Mn, Ni, Pb, Se, Sr, Ti, Tl, U, V, and Zn from 31 freshwaters located in distinct biogeographic regions. Elemental concentrations and stable nitrogen isotope ratios (proxies for trophic position) were determined in zooplankton, molluscs, insects, and fishes from all sites. Trophic magnification factors (TMFs) were calculated as the mean fraction of each element that transferred from prey to predators at each site. TMFs >1 indicate biomagnification and TMFs <1 indicate biodilution. Mercury was the only biomagnifying element (median TMF = 3.77), and selenium neither biomagnified nor biodiluted (median TMF = 1.01). All 16 remaining elements biodiluted, with median TMFs ranging from 0.07 (uranium) to 0.60 (thallium). We used a model selection procedure to determine whether intrinsic physical and chemical elemental properties explained differences in TMFs among elements. Elements with high covalent bonding indices (Q) had marginally greater TMFs than elements with low Q values. Based on their high Q values, we recommend investigation into the trophic transfer rates of ten additional trace elements, some of which may biomagnify through some aquatic food webs. The high variability in TMFs within elements suggests that ecological factors are likely more important than intrinsic elemental properties at determining elemental TMFs.

Substantial quantities of fine tailings and oil sands process affected water (OSPW) require reclamation in the Athabasca oil sands (AOS) region, Canada. Towards this end, Lake Miwasin was created as a pilot-scale pit lake containing treated fluid tailings (bottom sediment) capped with a blend of OSPW and surface water. This is a recent approach to waste reclamation and long-term monitoring is ongoing to determine the trajectory of water quality in this test lake. The current study characterized spatial and temporal changes in surface water chemistry using a wireless sensor network (WSN), particularly to identify contaminant release from the consolidating tailings and potential periods of sediment resuspension. The WSN technology was deployed during the open water season from September 2020 to October 2022 to remotely measure water quality parameters at different depths of the water column. Field measurements and manual water sampling were conducted periodically to validate sensor measurements and to analyse additional variables requiring more complex analysis. During the study, increased electrical conductivity (EC) near the sediment-water interface during water column stratification indicated expression of pore water with elevated salt content, as the bottom tailings progressively consolidated. A decreasing trend in EC towards the end of the monitoring season suggested water input from the surrounding catchment and possibly a decline in porewater expression. A Tier 1 screening-level risk assessment using hazard quotients (HQs) was conducted for individual metals using Canadian water quality benchmarks and published toxicity data. Total osmolarity was used to identify risks associated with major ions. Cumulative HQs for trace elements were >1 but decreased over time. The risk due to major ions was <1 for all years and seasons except 2019 and winter 2020. Modelling results indicated that the predicted 90th percentile HQs for Se and As were 1.6 and 0.44, respectively, suggesting that Se represents a potential ecotoxicological risk and should be further investigated. Overall, water quality monitoring and modelling insights gained from this study have the potential to inform AOS pit lake design and prediction of reclamation trajectories.

Machine learning (ML) is increasingly considered the solution to environmental problems where only limited or no physico-chemical process understanding is available. But when there is a need to provide support for high-stake decisions, where the ability to explain possible solutions is key to their acceptability and legitimacy, ML can come short. Here, we develop a method, rooted in formal sensitivity analysis (SA), that can detect the primary controls on the outputs of ML models. Unlike many common methods for explainable artificial intelligence (XAI), this method can account for complex multi-variate distributional properties of the input-output data, commonly observed with environmental systems. We apply this approach to a suite of ML models that are developed to predict various water quality variables in a pilot-scale experimental pit lake. A critical finding is that subtle alterations in the design of an ML model (such as variations in random seed for initialization, functional class, hyperparameters, or data splitting) can lead to entirely different representational interpretations of the dependence of the outputs on explanatory inputs. Further, models based on different ML families (decision trees, connectionists, or kernels) seem to focus on different aspects of the information provided by data, although displaying similar levels of predictive power. Overall, this underscores the importance of employing ensembles of ML models when explanatory power is sought. Not doing so may compromise the ability of the analysis to deliver robust and reliable predictions, especially when generalizing to conditions beyond the training data.

Pit lakes are currently being investigated as a way to store and reclaim waste materials in the Alberta Oil Sands (AOS) region, Canada. Lake Miwasin (LM) is a pilot-scale pit lake consisting of treated fine tailings overlayed with oil sands process-affected water (OSPW) blended with fresh surface water. In October 2021, the surface water contained a mean concentration of 1.33 ± 0.04 µg/L dissolved selenium (Se), slightly above the Canadian Council of Ministers of Environment water quality guideline for long-term protection of aquatic life (1 µg Se/L). This study assessed the bioaccumulation of Se by the cladoceran Daphnia pulex under laboratory conditions through both aqueous and dietary exposure routes for comparison to field-collected specimens. In 12-day semi-static tests, lab-cultured D.pulex were exposed to water, and algae grown in media spiked with selenate. Results showed that Se bioaccumulation by lab-cultured D. pulex increased in all exposure treatments from days 5 to 12, with maximum Se concentrations of 3.08–3.47 µg/g dry weight (dw) observed within the exposure range tested. Interestingly, lower Se bioaccumulation concentrations (1.26–1.58 µg/g dw) were observed in the highest dissolved Se and dietary Se treatments, suggesting potential internal regulatory mechanisms. In addition, native D. pulex (LM) collected from Lake Miwasin and cultured in-house were exposed in 8-day semi-static tests to Lake Miwasin surface water and algae cultured in Lake Miwasin surface water. Selenium bioaccumulation in native D. pulex (LM) ranged from 2.00 to 2.04 µg/g dw at day 8 and was not significantly different (p > 0.05) compared to Se concentrations in D. pulex collected from Lake Miwasin (2.15 ± 0.28 µg/g) in summer 2022.

Significant amounts of tailings and oil sands process-affected water (OSPW) are generated by bitumen extraction in the Alberta Oil Sands region. These by-products are potentially toxic to aquatic organisms and require remediation. The study site was Lake Miwasin, a pilot-scale pit lake integrated into broader reclamation efforts. It consists of treated tailings overlaid with blended OSPW and freshwater, exhibiting meromictic conditions and harboring aquatic communities. This study assessed the potential toxicity of Lake Miwasin surface water (LMW) and pore water (LMP) using saline-acclimated Cladocera, including lab strains of Daphnia magna and Daphnia pulex and native Daphnia species collected in brackish Humboldt Lake (HL) and Lake Miwasin (LM). The pore water evaluation was used to represent a worst-case water quality scenario during pond stratification. Additionally, the inclusion of native organisms incorporated site-specific adaptations and regional sensitivity into the toxicity evaluation. Our results showed that LMW did not display acute or chronic toxicity to lab species and native Daphnia sp. (HL). Conversely, LMP was acutely toxic to both lab species and native D. pulex (LM). In chronic tests (12 days exposure), LMP negatively affected reproduction in D. pulex (lab), with reductions in the number of offspring. Limited ability to acclimated organisms to the high salinity levels of LMP resulted in a shortened exposure duration for the chronic toxicity test. In addition to salinity being identified as a stressor in LMP, toxicity identification evaluation (TIE) phase I findings demonstrated that the observed toxicity for D. magna (lab) and D. pulex (LM, native) might be attributed to ammonia and metals in LMP. Further investigations are required to confirm the contributions of these stressors to LMP toxicity.

Many large inland rivers are contaminated by historic industrial and agricultural activities along their shorelines. Redistribution of contaminated river sediments via dredging or flooding can reintroduce previously stored toxic persistent organic pollutants into the aquatic environment. We used multiple lines of evidence, including chemical measurements using gas chromatography-high-resolution mass spectrometry and cell-based bioassays using the fish liver cell line RTL-W1, to determine the risk to aquatic biota associated with sediment in three reservoirs and one delta lake in the Saskatchewan River, a large Canadian prairie river. Concentrations of dioxins and furans (maximum 30 ng kg−1 dry weight) and polycyclic aromatic hydrocarbons (maximum < 300 μg kg−1 dry weight) were generally low in sediments, though hotspots were identified in parts of reservoirs rich in clay and organic matter wherein these chemicals accumulated. Concentration measurements agreed well with bioassay results, as the samples with the highest chemical concentrations induced the greatest EROD activity (biological equivalent concentration = 4010 ng kg−1). With one exception, all fishes tested were below the tolerable intake level for dioxins, and PAH metabolites were only rarely detected in bile, suggesting limited risk to human consumers. Based on our findings, the overall effects of dioxins and dioxin-like chemicals on biota in this system are likely limited. Nonetheless, disturbance and intentional movement of sediments in this river basin for potential sediment restoration in the future should be done cautiously, paying attention to sediment characteristics and local hotspots.

Selenium (Se) is both an essential micronutrient and a contaminant of concern that is of particular interest in mining-influenced waterbodies in Canada. The objective of this research was to characterize the trophic dynamics of selenium along a gradient of exposure concentrations in a Canadian boreal lake ecosystem. From June 20 to August 22, 2018, six limnocorrals (littoral, ∼3000 L enclosures) were spiked with mean measured concentrations of 0.4, 0.8, 1.6, 3.4, 5.6 and 7.9 μg Se/L as selenite, and three limnocorrals served as untreated controls (background aqueous Se = 0.08-0.09 μg/L). Total Se (TSe) concentrations in water, periphyton, phytoplankton, sediment, benthic macroinvertebrates, zooplankton and female finescale dace (Phoxinus neogaeus; added on day 21 of the experiment) were measured throughout and at the end of the experiment. Total Se bioaccumulation by organisms was generally non-linear. Greater uptake by phytoplankton than periphyton was observed. Taxonomic differences in accumulation of TSe by invertebrates (Heptageniidae = Chironomidae > zooplankton) were observed as well. Fish muscle and ovary tissue TSe bioaccumulation was more variable than that at lower trophic levels and uptake patterns indicated that fish did not reach steady state concentrations. This research provides field-derived models for the uptake of Se by algae and invertebrates, and contributes to a better understanding of the dynamics of TSe bioaccumulation over a gradient of exposure concentrations in cold-water lentic systems.

Selenium (Se) enrichment has been demonstrated to vary by several orders of magnitude among species of planktonic algae. This is a substantial source of uncertainty when modelling Se biodynamics in aquatic systems. In addition, Se bioconcentration data are largely lacking for periphytic species of algae, and for multi-species periphyton biofilms, adding to the challenge of modelling Se transfer in periphyton-based food webs. To better predict Se dynamics in periphyton dominated, freshwater ecosystems, the goal of this study was to assess the relative influence of periphyton community composition on the uptake of waterborne Se oxyanions. Naturally grown freshwater periphyton communities, sampled from five different water bodies, were exposed to environmentally relevant concentrations of selenite [Se(IV)] or selenate [Se(VI)] (nominal concentrations of 5 and 25 mu g Se L-1) under similar, controlled laboratory conditions for a period of 8 days. Unique periphyton assemblages were derived from the five different field sites, as confirmed by light microscopy and targeted DNA sequencing of the plastid 23S rRNA gene in algae. Selenium accumulation demonstrated a maximum of 23.6-fold difference for Se(IV) enrichment and 2.1-fold difference for Se(VI) enrichment across the periphyton/biofilm assemblages tested. The assemblage from one field site demonstrated both high accumulation of Se(IV) and iron, and was subjected to additional experimentation to elucidate the mechanism(s) of Se accumulation. Selenite accumulation (at nominal concentrations of 5 and 25 mu g Se L-1 and mean pH of 7.5 across all treatment replicates) was assessed in both unaltered and heat-killed periphyton, and in periphyton from the same site grown without light to exclude phototrophic organisms. Following an exposure length of 8 days, all periphyton treatments showed similar levels of Se accumulation, indicating that much of the apparent uptake of Se(IV) was due to non-biological processes (i.e., surface adsorption). The results of this study will help reduce uncertainty in the prediction of Se dynamics and food-chain transfer in freshwater environments. Further exploration of the ecological consequences of extracellular adsorption of Se(IV) to periphyton, rather than intracellular absorption, is recommended to further refine predictions related to Se biodynamics in freshwater food webs.

Human activities have increased the release of selenium (Se) to aquatic environments, but information about the trophic transfer dynamics of Se in Canadian boreal lake systems is limited. In the present study, Se was added as selenite to limnocorrals (2‐m‐diameter, 3000‐L in situ enclosures) in a boreal lake in northwestern Ontario to reach nominal concentrations of 1 and 10 μg Se/L in triplicate each for 77 d, and 3 additional limnocorrals were controls with no Se added. Total Se concentrations were determined in water, sediment, periphyton, benthic macroinvertebrates, zooplankton, and reproductively mature female fathead minnows (Pimephales promelas; added on day 33) collected throughout (and at the end of) the exposure period. Mean measured water Se concentrations in the control, 1‐, and 10‐μg/L treatments were 0.12, 1.0, and 8.9 μg/L. At the end of exposure (day 77), enrichment functions ranged from 7772 L/kg dry mass in the 8.9‐μg/L treatment to 23 495 L/kg dry mass in the 0.12‐μg/L treatment, and trophic transfer factors for benthic macroinvertebrates ranged from 0.49 for Gammaridae to 2.3 for Chironomidae. Selenium accumulated in fathead minnow ovaries to concentrations near or above the current US Environmental Protection Agency criterion (15.1 μg/g dry mass for fish ovary/egg) in the 1.0‐ and 8.9‐μg/L treatments, suggesting that, depending on aqueous Se speciation, such exposures have the potential to cause Se accumulation in fish to levels of concern in cold‐water, boreal lake systems. Environ Toxicol Chem 2019;38:1954–1966. © 2019 SETAC

Selenium (Se) is a contaminant of concern in many aquatic ecosystems due to its narrow range between essentiality and toxicity in oviparous (yolk-bearing) vertebrates. The objective of the present study was to determine the effects of Se, experimentally added to in situ limnocorrals as selenite, on invertebrate communities and fathead minnow (Pimephales promelas) at environmentally realistic Se concentrations. Nine limnocorrals were deployed in a mesotrophic lake at the International Institute for Sustainable Development - Experimental Lakes Area in Ontario, Canada in May 2017. From June 1 to August 17, 2017, selenite was added to six enclosures to attain mean measured aqueous Se concentrations of 1.0 +/- 0.10 or 8.9 +/- 2.7 mu g/L Se (in triplicate) and three limnocorrals were untreated controls (background mean aqueous Se = 0.12 +/- 0.03 mu g/L). Benthic macroinvertebrates were collected throughout and at the end of the exposure period using artificial substrates to determine density, dry biomass, diversity, and taxa richness at the family level. Reproductively mature female fathead minnows (added on d 33 of the study) were collected throughout and at the end of the exposure period. After 77 d, Chironomidae and Gammaridae densities and biomass were significantly lower in the 8.9 mu g/L Se treatment relative to the 1.0 mu g/L Se treatment and the control. Invertebrate diversity (measured as Shannon's and Simpson's indices) significantly declined in the 1.0 mu g/L and 8.9 mu g/L Se treatments relative to the control (0.12 mu g/L Se group). Fulton's condition factor for fathead minnow was significantly less in the 8.9 mu g/L treatment compared to 0.12 and 1.0 mu g/L Se experimental groups. The results of this study indicated that exposure to relatively low aqueous selenite concentrations can negatively affect invertebrate density and biomass, as well as fish condition. More research is necessary to characterize the risk of selenite exposure to aquatic invertebrates under realistic field conditions, and future risk assessments may need to consider reduced food availability as a factor that may impair the health of higher trophic level organisms in areas with elevated selenite.

Trophic transfer of contaminants dictates concentrations and potential toxic effects in top predators, yet biomagnification behaviour of many trace elements is poorly understood. We examined concentrations of vanadium and thallium, two globally-distributed and anthropogenically-enriched elements, in a food web of the Slave River, Northwest Territories, Canada. We found that tissue concentrations of both elements declined with increasing trophic position as measured by δ15N. Slopes of log [element] versus δ15N regressions were both negative, with a steeper slope for V (-0.369) compared with Tl (-0.099). These slopes correspond to declines of 94% with each step in the food chain for V and 54% with each step in the food chain for Tl. This biodilution behaviour for both elements meant that concentrations in fish were well below values considered to be of concern for the health of fish-eating consumers. Further study of these elements in food webs is needed to allow a fuller understanding of biomagnification patterns across a range of species and systems.

Freshwaters worldwide are under increasing pressure from anthropogenic activities and changing climate. Unfortunately, many inland waters lack sufficient long-term monitoring to assess environmental trends. Analysis of sedimentary ancient DNA ( sedaDNA) is emerging as a means to reconstruct the past occurrence of microbial communities of inland waters. The purpose of this study was to assess a combination of high-throughput sequencing (16S rRNA) of sedaDNA and traditional paleolimnological analyses to explore multidecadal relationships among cyanobacterial community composition, the potential for cyanotoxin production, and paleoenvironmental proxies. DNA was extracted from two sediment cores collected from a northern Canadian Great Plains reservoir. Diversity indices illustrated significant community-level changes since reservoir formation. Furthermore, higher relative abundances in more recent years were observed for potentially toxic cyanobacterial genera including Dolichospermum. Correlation-based network analysis revealed this trend significantly and positively correlated to abundances of the microcystin synthetase gene ( mcyA) and other paleoproxies (nutrients, pigments, stanols, sterols, and certain diatom species), demonstrating synchrony between molecular and more standard proxies. These findings demonstrate a novel approach to infer long-term dynamics of cyanobacterial diversity in inland waters and highlight the power of high-throughput sequencing to reconstruct trends in environmental quality and inform lake and reservoir management and monitoring program design.

Cumulative environmental impacts driven by anthropogenic stressors lead to disproportionate effects on indigenous communities that are reliant on land and water resources. Understanding and counteracting these effects requires knowledge from multiple sources. Yet the combined use of Traditional Knowledge (TK) and Scientific Knowledge (SK) has both technical and philosophical hurdles to overcome, and suffers from inherently imbalanced power dynamics that can disfavour the very communities it intends to benefit. In this article, we present a ‘two-eyed seeing’ approach for co-producing and blending knowledge about ecosystem health by using an adapted Bayesian Belief Network for the Slave River and Delta region in Canada's Northwest Territories. We highlight how bridging TK and SK with a combination of field data, interview transcripts, existing models, and expert judgement can address key questions about ecosystem health when considerable uncertainty exists. SK indicators (e.g., bird counts, mercury in fish, water depth) were graded as moderate, whereas TK indicators (e.g., bird usage, fish aesthetics, changes to water flow) were graded as being poor in comparison to the past. SK indicators were predominantly spatial (i.e., comparing to other locations) while the TK indicators were predominantly temporal (i.e., comparing across time). After being populated by 16 experts (local harvesters, Elders, governmental representatives, and scientists) using both TK and SK, the model output reported low probabilities that the social-ecological system is healthy as it used to be. We argue that it is novel and important to bridge TK and SK to address the challenges of environmental change such as the cumulative impacts of multiple stressors on ecosystems and the services they provide. This study presents a critical social-ecological tool for widening the evidence-base to a more holistic understanding of the system dynamics of multiple environmental stressors in ecosystems and for developing more effective knowledge-inclusive partnerships between indigenous communities, researchers and policy decision-makers. This represents new transformational empirical insights into how wider knowledge discourses can contribute to more effective adaptive co-management governance practices and solutions for the resilience and sustainability of ecosystems in Northern Canada and other parts of the world with strong indigenous land tenure.

A major source of uncertainty in predicting selenium (Se) distribution in aquatic food webs lies in the enrichment factor (EF), the ratio of Se bioconcentration in primary producers and microorganisms relative to the concentration of Se in the surrounding water. It has been well demonstrated that EFs can vary dramatically among individual algal taxa, but data are lacking regarding the influence of periphyton community composition on EFs for a given geochemical form of Se. Therefore, the goals of this study were first to assess whether different periphyton communities could be established in aquaria with the same starting inoculum using different light and nutrient regimes, and second, to determine if the periphyton assemblage composition influences the uptake of waterborne Se (as selenite) and subsequent Se transfer to a model macroinvertebrate primary consumer. Periphyton biofilms were grown in aquaria containing filtered pond water (from Saskatoon, SK) spiked with approximately 20 μg Se/L (mean measured concentration 21.0 ± 1.2 μg Se/L), added as selenite. Five different light and nutrient regimes were applied to the aquaria (three replicates per treatment) to influence biofilm community development. After 6 weeks of biofilm maturation, 40 to 80 immature cultured snails (Stagnicola elodes) were added to each aquarium. The bacterial and algal members of the periphyton community were characterized by targeted metagenomic analyses before and after addition of snails to ensure the snails themselves did not significantly alter the biofilm community. Samples were collected for Se analysis of water, periphyton, and whole-body snail. The nutrient and light treatments resulted in substantially different compositions of the periphytic biofilms, with each being relatively consistent across replicates and throughout the study. Although the aqueous concentration of dissolved Se administered to treatments was constant, uptake by the different periphytic biofilms differed significantly. Both the low-light (61.8 ± 12.1 μg Se/g d.w.) and high-light (30.5 ± 4.7 μg Se/g d.w.) biofilms, which were found to have high proportions of cyanobacteria, contained statistically higher concentrations of Se relative to the other treatments. Furthermore, the concentration of Se in bulk periphyton was predictive of Se bioaccumulation in grazing snails but as an inverse relationship, opposite to expectations. The trophic transfer factor was inversely correlated with periphyton enrichment factor (r = -0.841). A number of different bacterial and algal taxa were correlated (either positively or negatively) with Se accumulation in periphyton biofilm and snails. Recent advancements in genetic methods make it possible to conduct detailed characterization of periphyton assemblages and begin to understand the influence that periphyton composition has on Se biodynamics in aquatic systems.

Across the circumpolar world, intensive anthropogenic activities in the southern reaches of many large, northward-flowing rivers can cause sediment contamination in the downstream depositional environment. The influence of ice cover on concentrations of inorganic contaminants in bed sediment (i.e., sediment quality) is unknown in these rivers, where winter is the dominant season. A geomorphic response unit approach was used to select hydraulically diverse sampling sites across a northern test-case system, the Slave River and delta (Northwest Territories, Canada). Surface sediment samples (top 1cm) were collected from 6 predefined geomorphic response units (12 sites) to assess the relationships between bed sediment physicochemistry (particle size distribution and total organic carbon content) and trace element content (mercury and 18 other trace elements) during open-water conditions. A subset of sites was resampled under-ice to assess the influence of season on these relationships and on total trace element content. Concentrations of the majority of trace elements were strongly correlated with percent fines and proxies for grain size (aluminum and iron), with similar trace element grain size/grain size proxy relationships between seasons. However, finer materials were deposited under ice with associated increases in sediment total organic carbon content and the concentrations of most trace elements investigated. The geomorphic response unit approach was effective at identifying diverse hydrological environments for sampling prior to field operations. Our data demonstrate the need for under-ice sampling to confirm year-round consistency in trace element-geochemical relationships in fluvial systems and to define the upper extremes of these relationships. Whether contaminated or not, under-ice bed sediment can represent a worst-case scenario in terms of trace element concentrations and exposure for sediment-associated organisms in northern fluvial systems. Environ Toxicol Chem 2017;36:2916-2924. (c) 2017 SETAC

Indigenous community members along the Slave River in Canada have voiced their concerns for the health of ecosystems under pressure from resource extraction, hydroelectric development and global climate change. We present a test case of traditional knowledge and scientific results about the spawning and migration patterns of fish in the Slave River and Delta. This dual knowledge system approach elucidates the broader connectivity of local study regions and can improve monitoring programmes by extending beyond the usual context/confines of the present or recent past, increasing the spatial and temporal range of system information.

Aside from a companion investigation to this study, there are currently no peer-reviewed phosphorus (P) release rate data for northern North American (i.e., Canadian) reservoirs. Using Lake Diefenbaker, Saskatchewan, Canada as a case study, we tested the effect of variation in overlying water DO conditions on the P release rates from sediment cores. Sediment cores from four down-reservoir locations in Lake Diefenbaker were incubated under high (>8 mg l−1), low (2–3 mg l−1), or anoxic (<1 mg l−1) DO concentrations. Sediment cores were then analyzed for total P (TP) and three geochemical P fractions to assess how the DO regime influenced sediment P inventory. Maximum P release rates were highest under anoxic conditions and similar among sites (15.0–20.3 mg m−2 day−1), with the low-DO rates intermediate to the high-DO and anoxic P fluxes. Predictive internal P loading models considering only hypolimnetic anoxia may therefore oversimplify and thus underestimate P mobilization in situ. Non-apatite inorganic P (54 ± 10% across sites) from the top 1 cm of the sediment profile was the main source of P released during incubations, indicating that sampling on a coarser scale of resolution could obscure the relationship between sediment geochemistry and short-term P flux.

Tracking the uptake and transfer of toxic chemicals, such as mercury (Hg), in aquatic systems is challenging when many top predators are highly mobile and may therefore be exposed to chemicals in areas other than their location of capture, confounding interpretation of bioaccumulation trends. Here we show how the application of a less commonly used ecological tracer, stable sulfur isotope ratios (34S/32S, or δ34S), in a large river-delta-lake complex in northern Canada allows differentiation of resident from migrant fishes, beyond what was possible with more conventional 13C/12C and 15N/14N measurements. Though all large fishes (n = 105) were captured in the river, the majority (76%) had δ34S values that were indicative of the fish having been reared in the lake. These migrant fishes were connected to a food chain with greater Hg trophic magnification relative to the resident fish of the river and delta. Yet, despite a shallower overall trophic magnification slope, large river-resident fish had higher Hg concentrations owing to a greater biomagnification of Hg between small and large fishes. These findings reveal how S isotopes can trace fish feeding habitats in large freshwater systems and better account for fish movement in complex landscapes with differential exposure pathways and conditions.