Increased size and relative abundance of migratory fishes observed near the Athabasca oil sands

Environmental concerns pervade oil sands activities in northern Alberta (Schindler 2010). Much of the focus of studies to evaluate the concerns has targeted chemical stressors accumulating in various environmental media (Kelly et al. 2009; Kurek et al. 2013). Among chemical stressors, oil sands constituents, including polyaromatic compounds (PACs) and metals, have been found in snow close to mines (Kelly et al. 2009, 2010) and may migrate into local waterbodies where effects in aquatic organisms may occur. Recent work suggests loads of total metals are higher in tributaries during freshet in years with low pH snowmelt (<6.5; Alexander et al. 2017). Loading of PACs in the sediments of local and regional lakes (Kurek et al. 2013) and depositional areas of some local rivers (Evans et al. 2016) has also increased over time suggesting the influence of local oil sands activity. Although the accumulation of oil sands constituents has been observed in various environmental media, the association with biological effects in the local aquatic environment is less clear. Acute toxicity and changes in the physiology of early life stages of fish after exposure to oil sands constituents have been observed in laboratory experiments (Colavecchia et al. 2006). Acute toxicity to early life stages of fish may be a relevant indicator, but documenting changes using this measurement can be challenging in field settings. Changes in fish associated with chemical stressors have also been observed during field studies, but the detected differences are largely limited to the induction of biochemical markers of exposure, occurrence of PACs in fish muscle, PAC biotransformation products, and changes in reproductive physiology (Tetreault et al. 2003; Ohiozebau et al. 2016, 2017; Arens et al. 2017).

Although biota in the oil sands region show biochemical and physiological indications of exposure, the full extent of effects in aquatic organisms in response to anthropogenic loading of metals or PACs is not currently known. Documenting change in field studies using reproductive physiology in adults may, in many cases, be simpler than acute toxicity to early life stages of fish, but the consequences of either may emerge over time as reduced abundance of fish captured during field surveys (Gibbons and Munkittrick 1994). Although documenting declines in relative abundance may be rare in the oil sands region, some work has shown this response in some species (RAMP 2010; Schwalb et al. 2015). Declines in abundance of three migratory species, mountain whitefish (Prosopium williamsoni), longnose sucker (Catostomus catostomus), and Arctic grayling (Thymallus arcticus) were documented using data from a series of fish fences in the Muskeg River (RAMP 2010; Schwalb et al. 2015). The declines first appeared between 1977 and 1995, but also persisted until 2009, the most recently reported data from a fish fence in the Muskeg River (Schwalb et al. 2015). In contrast to chemical exposure, these changes were attributed to altered hydrology and increased landscape disturbance (Schwalb et al. 2015), another concern in the oil sands region. Although the declines in these three species were clear when compared with the data from 1976/1977, until 2006 they were also accompanied by reduced abundance of white sucker (Catostomus commersoni). In contrast to mountain whitefish, longnose sucker, and Arctic grayling, the declines in white sucker reversed in 2009. In that year, the largest recorded upstream migration of white sucker into the Muskeg River was observed (RAMP 2010; Schwalb et al. 2015). These findings suggest heterogenous and complex dynamics among fish populations utilizing the Athabasca River and its tributaries near the oil sands mines. The contrasting results further suggest additional work is needed to identify or clarify the status of aquatic organisms residing in the oil sands region.

Studies addressing the ongoing and growing need for information on the status of biota in the oil sands region are underway (AEMERA and ECCC 2016). Despite the intense effort across multiple media and receptors similar to other monitoring programs, studies in the oil sands are affected by persistent and widespread challenges. A primary difficulty in ecological monitoring is the lack of information describing the undisturbed baseline conditions. This information is necessary for a widely discussed and ideal design, before-after-control-impact (BACI; Smokorowski and Randall 2017). Where data to fulfill the BACI design or its variants are not available, other study types such as paleolimnology may be employed (Kurek et al. 2013). If paleolimnology or other techniques based on environmental archives are not appropriate for the focus of a study, such as riverine fishes, other approaches are needed. A common alternative design uses reference and exposure sites sampled only during the “after” period, but identifying effects with these designs can be challenging (Munkittrick et al. 2002). In the absence of data to fulfill a BACI design, a desire to examine certain ecological receptors, and uncertainty in the realtionships between reference and exposure locations, additional approaches are required to rigorously examine data.

Recently, the general concept of adaptive monitoring has been proposed to accommodate and overcome several challenges commonly affecting ecological monitoring, including cases with uncertain reference and missing baseline data (Arciszewski et al. 2017). The analyses in adaptive monitoring program often focus first on changes at a given location over time reducing the primary reliance of analyses on reference sites (Arciszewski and Munkittrick 2015). In place of a robust estimate of an undisturbed baseline, an initial estimate can be derived, often from the first three years of data at a site. The initial estimate is made without assuming the existing conditions are acceptable, but does assume organisms can respond in measurable ways to new, modified, or progressively influential stressors (Arciszewski and Munkittrick 2015; Arciszewski et al. 2017). This initial estimate of existing data, commonly set at 95%, is used to define a range of values predicted to occur in the future, assuming conditions influencing the biological attributes have remained stable over time. Statistically unexpected observations suggest that the predicted range is erroneous (i.e., the observed difference originates from uncaptured normal variation), the observed difference is attributable to the built-in error rate (α), or the drivers affecting the measured endpoint are changing. If a statistically unexpected difference is identified, confirmation of changes is sought to reduce inference errors and provide confidence in the conclusions (Hodson et al. 1996). In addition to confirmation, supporting evidence from parallel studies or more detailed, specific, and future studies may also be initiated to explore any hypothesized causes of the changes observed. In contrast, if no likely differences are observed, a new estimate of an expected range of observations is derived, including the new information (Arciszewski et al. 2017).

Although adaptive monitoring is intended for prospective studies, the process and tools can also be applied retrospectively. An historical data set that has not been thoroughly examined following an adaptive approach and that may contain valuable information on the state of fishes in the Athabasca River is known as the “fish inventory”. The inventory was a boat electrofishing program initiated by Syncrude in the spring of 1987 (Hatfield Consultants Partnership et al. 2016). Fall and summer collections were added in 1989 and 1990, respectively (Hatfield Consultants Partnership et al. 2016). In 1997, the fish inventories were embedded in the Regional Aquatics Monitoring Program (RAMP; Hatfield Consultants Partnership et al. 2016). Collections for the fish inventory continued until 2014. During recent reviews, the relevance and capabilities of the program were criticized (Main 2010), and like many historical data sets, the fish inventory is necessarily constrained. Consequently, any analyses using these data have important limitations, including the questions that can be addressed, the strength of the conclusions drawn, and the additional sources of information that can be accessed. Despite this, the fish inventory data have some beneficial attributes. Data records from the inventory are extensive. Records include effort, individuals and species captured, and length and weight for many individuals. Subsets of fish were also aged and some were sexed. Although collections started 20 years after the first commercial development, oil sands production in the first year of the fish inventory (1987) was less than 14% of production in the final year of sampling (2014; CAPP 2016). The increase in production from 1987 to 2014 included both the expansion of existing developments and opening of new mines and in situ facilities (Fig. 1). Sites used in the inventory also span the length of the Athabasca River in the minable area of the oil sands. The duration, period of coverage, volume of data, and spatial coverage of the fish inventory data may outweigh the absence of detailed baseline information collected prior to any development.

The intention of this paper is to determine if any evidence of changes in fishes residing in the Athabasca River near oil sands facilities is present in the fish inventory data. Based on the initial analyses on white sucker and walleye, hypotheses to explain the observations, including reduced fishing pressure and eutrophication, were formed. To scrutinize and explore these hypotheses, additional data including other species, relative abundance, commercial fishing records, nutrient concentrations, and discharge were sought.

Successfully detecting environmental change with monitoring is affected by many factors, including the availability of data from an undisturbed baseline period and the study designs selected (Smokorowski and Randall 2017). Where a baseline data set is missing or deficient, the capability to detect change is compromised and alternative approaches are required. Here, techniques to evaluate environmental change used statistically defined and adaptive thresholds (Arciszewski et al. 2017). Based on an adaptive baseline range of data, white sucker and walleye in the oil sands area of the Athabasca River have been longer than expected since at least 2007. The earliest confirmed evidence was observed in 2009. Given the suggestion from the bootstrapped thresholds for CPUE that the relative abundances of several species may also be also increasing, the differences we detected further suggest that some changes potentially relevant for resident and transient fish populations have occurred in the lower Athabasca watershed in the last decade.

In contrast to effects suggesting exposure to oil sands constituents or watershed development, a more general hypothesis arose to explain a majority of our results: the overwintering location of fish species. For instance, the largest changes in the relative abundances of walleye and white sucker appear to be spatially and temporally non-random; they occur in the spawning season near respective spawning areas (Bond 1980; Bond and Berry 1980a, 1980b). Lake whitefish also have a higher relative abundance in the fall near suspected spawning areas (Bond and Berry 1980a, 1980b). Many species showing changes in the fish inventory data, including white sucker, walleye, lake whitefish, and goldeye have suspected upstream spawning migrations from Lake Athabasca (Bond 1980; Bond and Berry 1980a, 1980b). Telemetry studies done using walleye and lake whitefish during RAMP support this contention (RAMP 1999).

Similarly, overwintering location may explain the differing conclusions between this study and analyses from the fish fences (RAMP 2010; Schwalb et al. 2015). Among the fishes showing persistent declines in 2009 compared with the earliest fences, Arctic grayling likely overwinter in the Athabasca River (Bond 1980). Mountain whitefish may overwinter in both the Athabasca and Clearwater rivers upstream of Fort McMurray (Bond 1980), but are also rare in the inventory (Tables S3 and S4). The final species showing a persistent decline in fish fence data, longnose sucker, also shows a pattern of changes in length in the fish inventories, albeit muted and delayed when compared with white sucker (Fig. S9). The occurrence of attributes of overwintering in the Athabasca River and Lake Athabasca suggest a life cycle in longnose sucker inhabiting the Athabasca basin that includes both behaviours. This contention is supported by radio-telemetry studies done during RAMP, which suggest the longnose sucker migrating from Lake Athabasca spawn in the mainstem, whereas those spawning in tributaries overwinter in the Athabasca River (RAMP 2002). Furthermore, until 2006 white sucker showed similar declines in the fish fence data when compared with Arctic grayling, mountain whitefish, and longnose sucker. Patterns in the relative abundance of white sucker compared with the others diverged in 2009. This suggests that some alterations in some of the factors affecting the relative abundances of fishes in the lower Athabasca basin occurred within the last decade, and indicate the importance of considering ecology in monitoring studies of migratory species (Arens et al. 2017). More specifically, changes occur most frequently in species that migrate from Lake Athabasca into the oil sands area to spawn.

In the fish inventory data, clear changes were observed in several migratory species (white sucker, walleye, and lake whitefish) and emerged in a specific form (increased size and (or) relative abundance). Among the interpretive frameworks used to evaluate the results of fish health studies, increased size and relative abundance can be associated with more food per individual (Gibbons and Munkittrick 1994). If demographic factors, such as changing sex ratios, are not important, more energy available per individual fish may be attributed to eutrophication. Evidence from other studies suggests a possible role of eutrophication. Increases in N and P over time have been observed in the Athabasca River in earlier work (Dubé and Wilson 2013). Results from RAMP have also suggested intermittent nutrient enrichment in benthic invertebrates in the Peace–Athabasca delta and some tributaries, including the Ells, Muskeg, MacKay, lower Clearwater, and Firebag rivers (RAMP 2010, 2011, 2012). Furthermore, white sucker has shown responses associated with changes in nutrient loading elsewhere (Arciszewski et al. 2015) and other work also suggests that walleye populations in Alberta may be sensitive to eutrophication (Mogensen et al. 2014) suggesting the plausibility of this effect.

If eutrophication is occurring in the Athabasca River, the nutrients could originate from multiple sources. Nutrients may arise from the Athabasca watershed upstream of Fort McMurray, but nutrients may also enter the Athabasca River within the study area. The spatial relationships in TN and TP support both routes. Although the upstream nutrient load is not likely related to the oil sands, nutrient inputs between the locations of the water sampling may derive from atmospheric deposition of nitrogenous compounds emitted primarily from stacks and mine fleets (Fenn et al. 2015; Lynam et al. 2015; McLinden et al. 2016). Total N deposition in the oil sands region estimated for current development is 165 eq/ha per year (Cho et al. 2017) and may be associated with increased productivity in local waterbodies (Hazewinkel et al. 2008). Summers et al. (2016) suggested, however, that warming, rather than nutrient enrichment, is driving increased productivity in local lakes, which may have a role in explaining some of our results. Inconsistencies in the deposition of nitrogenous compounds on the landscape and transportation to watercourses over time may also explain some noise in TN among stations compared with TP.

The discharge of sewage may also be affecting productivity in the Athabasca River. The mean daily loading of TP from the Fort McMurray wastewater treatment plant rose from 10 to 15 kg/d from 2011 to 2014 (Regional Municipality of Wood Buffalo 2012a, 2015). Other possible sources of nutrients from wastewater are industrial facilities, communities, and work camps. Although detailed information on the treatment types and discharge volume or chemical characterization from these additional sources was not sought, a surrogate measurement, such as the shadow population, may be informative. The shadow population, including those living in work camps increased from 3500 to 40 000 between 2000 and 2012 (Regional Municipality of Wood Buffalo 2010, 2012b), including a specific one-year increase from 10 442 to 18 572 from 2006 to 2007 (Regional Municipality of Wood Buffalo 2010). This increase in population indirectly suggests an increase in wastewater production and may further explain some noise in the relationship of TN among the water chemistry locations.

Although an expected response of fishes to eutrophication does explain some observations, others are not consistent with this explanation. No differences are apparent in length, weight, and condition of walleye or white sucker collected in the fall (Figs. S10, S11) and there is no substantive evidence of increases in growth, estimated as length-at-age, in these species (Fig. 3). Signals of eutrophication do not appear in all species and are missing from small-bodied species residing throughout the study area year-round, including trout-perch, lake chub (Couesius plumbeus), and flathead chub (Platygobio gracilis; Fig. S3). Although detecting change in the fish inventory data set using small-bodied species is challenging, the clearest patterns are focused in fishes that overwinter in Lake Athabasca. There may be some role of eutrophication, but nutrients are not likely the key driver in explaining the changes we observed in walleye and white sucker. Other factors, either alone or in combination with nutrient enrichment, are likely responsible for the changes observed here.

An effect of mortality may also explain some of the observations presented here, but may occur to specific life stages. First, emergence of more food per individual may be related to the increased mortality of juveniles (Gibbons and Munkittrick 1994). A predicted effect of exposure to oil sands constituents is acute toxicity, especially in early life stages (Colavecchia et al. 2006). Regionally, juvenile white sucker are also likely predated by walleye (Arens et al. 2015). Unfortunately, the changes in the relative abundances of juvenile fishes, including white sucker, cannot be estimated from the inventory data; collections before and after 2003 did not equally emphasize small fish. However, an effect of mortality on juveniles would likely eventually emerge as declines in the abundance of adults. Furthermore, an expected pattern of responses and subsequent observations, if juvenile mortality occurred in the oil sands region, is increased growth rates of survivors along with reduced abundance (Gibbons and Munkittrick 1994). In the fish inventory data, walleye and white sucker showed no substantive evidence of a change in growth rate, and the relative abundances of both species has increased over time. An increase in relative abundance suggests a mechanism other than increased juvenile mortality.

An increase in the relative abundances of some species during spawning runs may instead be related to the reduced mortality of adults. The reduction of adult mortality, especially in potamodromous species, may be associated with a decrease in downstream fishing pressure in Lake Athabasca. Reduced effort in commercial fisheries explains changes observed in goldeye, lake whitefish, and walleye. The increased relative abundance of white sucker is more difficult to explain. The non-commercial catch of white sucker may have declined (Balagus et al. 1993), but the higher relative abundance of this species may also be related to a possible collateral effect of the decreased harvesting of walleye. Gillnets used for walleye in Lake Athabasca are 10.2 cm (Paul 2013). Gillnets with 10.2 cm mesh sizes were also used to collect fishes, including northern pike, walleye, and white sucker from Lake Athabasca and the Peace–Athabasca delta for tissue analyses during the Northern River Basins Study (Balagus et al. 1993). Although by-catch data were not found for Lake Athabasca, 10.2 cm gillnets have also efficiently captured walleye, white sucker, longnose sucker, and northern pike where those species co-occur (Arciszewski et al. 2015). Furthermore, a reduction in fishing effort may have coincided with a slight effect of nutrient enrichment.

If reduced mortality of adult fishes migrating from Lake Athabasca is responsible for the upstream changes, the reasons for a change in commercial harvest are not clear. The notable increases in the relative abundances suggest that the stocks were likely healthy during the fish inventory and when fishing pressure decreased. This suggests the reported reductions in catch rates were not related to the health of the fishery but could instead be linked to sociological factors, including rising fears and uncertainties of consuming fish caught locally (Tolton et al. 2012; Ohiozebau et al. 2017), changing demographics and vocations of members of local communities (Balagus et al. 1993; Parlee 2015), some other reason, or a combination of factors. However, all commercial fisheries in Alberta lakes were closed in August 2014 (AEP 2014). Any relationship between the closure decision and the data presented here is not clear.

Finally, there is another possibility to explain the observations presented here. The patterns we observed could be attributed to a change in sampling methodology. The inventory protocols expanded in 2003, but these later collections biased the data against the patterns we observed. The change in fishing protocol after 2002 increased the emphasis on collecting small fishes, including both small-bodied species and juveniles of large-bodied species. In relative numbers, larger white sucker and walleye have become more common over time, suggesting that the change in emphasis after 2002 did not affect the sampling of these species. Furthermore, if the changes we observed in the Athabasca mainstem were related to sampling methods, we would also expect similar changes in the Clearwater River over the same sampling period; this pattern, however, was not observed (Fig. S4).

Although the evidence for an effect of overwintering location and spawning migrations seems clear, not all observations suggest a direct and simple connection. Mysteriously, possible increases in CPUE were observed in some species in non-spawning seasons. Although these changes in the relative abundances during the non-spawning seasons may be important, the influence of low relative abundances observed during many sampling campaigns between 1987 and 2006 on the calculation of the bootstrapped thresholds affects their clear interpretation. Corresponding to the difficulty of identifying real change in monitoring data (Arciszewski et al. 2017), some changes in the relative abundances of some species in some seasons at some locations may also be statistical anomalies or may be due to the minor influence of juveniles of these species.

Although decisions to pool fish length data from consistently sampled reaches overcame the low sample sizes in the initial years, the detection of changes via the development of triggers was also clearly beneficial. However, understanding the performance of tools in various scenarios is necessary. For instance, normal ranges may not behave as expected in some cases (Kilgour and Somers 2017). In three of six endpoints examined using tolerance intervals (white sucker length and weight and walleye length) the tools and their application performed well. In the remaining endpoints, walleye weight and condition and white sucker condition, tolerance intervals, and more likely, the processes used for their application, including adapting triggers, did not perform well. Tolerance intervals configured as adaptive triggers may drift where observations are close to, but do not exceed a trigger (e.g., walleye weight) or where measurements alternate inside and outside triggers (e.g., white sucker condition). More generally, a lack of variability in annual means may also generate triggers that are likely too sensitive. This likely occurred in walleye condition. In the instances where tolerance intervals may not be appropriate, such as when fewer than three years of data are available, data are non-normal, or little variation in means is apparent in initial years of collection, an alternative approach, such as weighted means, nonparametric bootstrapping (Barrett et al. 2015), or alterations in the number of observations expected above thresholds (Arciszewski and Munkittrick 2015) may be useful to detect unexpected observations and further account for unequal sample sizes each year (Fig. S12). An important challenge in the use of normal ranges is when early data reflect a specific (and potentially uncommon, but also unknown) exposure scenario. This can be especially problematic during the initial years of a monitoring program. In these instances, relying on OTLs may be one of the few options available for understanding the occurrence of change. However, ecological considerations are also likely important in the development and application of thresholds. In this work, the exact timing of the fish collections relative to spawning within each year may also explain the unusual behavior of annual means of the condition of white sucker relative to the estimated normal ranges.

Monitoring programs are often limited by the absence of information on the relevance of the differences inevitably encountered (Munkittrick et al. 2009). Here, thresholds calculated for fish size using tolerance intervals and for relative abundance using parametric bootstrapping identified several unusual observations after 2006. Our work suggests that some potamodromous species likely overwintering in Lake Athabasca are thriving. The form of changes observed suggests a role of either eutrophication or reduced fishing pressure in Lake Athabasca. Although some influence of increased nutrients on fish may be occurring, an effect of reduced fishing pressure in Lake Athabasca is likely a predominant driver in the changes detected. An effect of reduced fishing pressure has not been previously considered in the oil sands region. Although the origins may not be explicitly known, the patterns we detected are difficult to ignore. If this approach was in place during the original collections, focused studies examining the increasing size would have been initiated in 2009.

No external funding was provided for this work. We would like to thank the COSIA Monitoring Priority Area Steering Committee and other industry representatives for valuable comments provided during the preparation of this manuscript. Thank you to Rod Hazewinkel at Alberta Environment and Parks for providing the wastewater reports for the Fort McMurray wastewater facility. We greatly appreciate the insights provided by two anonymous and kind reviewers for improving the interpretation of the data.