Are we accurately estimating the potential role of pollution in the decline of species at risk in Canada?

Pollution of the air, soil, and water is a significant threat to many species (Novacek and Cleland 2001; Dudgeon et al. 2006; Hernández et al. 2016). In some cases, pollution can cause sudden and immediate mortality, such as an oil spill leading to massive kills of birds (e.g., Haney et al. 2014) or wildlife deaths after contact with cyanide-bearing wastes from gold mining (Donato et al. 2007). Similarly, nutrient pollution can lead to rapid mortality of fish or waterfowl as a result of the production of potent toxins in algal blooms (e.g., Papadimitriou et al. 2018) or upwelling of anoxic bottom waters (e.g., Rao et al. 2014). However, many pollutants cause sublethal effects on wildlife that are less obvious and more challenging to assess (Nilsen et al. 2019). Nonetheless, sublethal effects of environmental pollutants on individuals can reduce their fitness, leading to population-level impacts (e.g., Kidd et al. 2007; Incardona et al. 2015). Notably, pollutants not only have direct toxic effects on organisms, but may also potentially cause harm through indirect pathways (e.g., trophic cascades; Kidd et al. 2014; Sasaki et al. 2016). Pollution can also compound the effects of other drivers of decline, weakening populations already affected by loss of habitat, overexploitation, or other threats (Brook et al. 2008; Darling and Côté 2008).

The task of assessing the effects of pollution is daunting: tens of thousands of chemicals exist in commerce today and the size of the global chemical industry is projected to double by 2030 (UNEP 2019). Contaminants, such as certain flame retardants, undergo transformations into more toxic breakdown products in the environment that contribute to heightened environmental effects (Liggio et al. 2016). For many pollutants there are no analytical methods to detect their presence in either the environment or wildlife, or detection methods may be costly and invasive (Choy et al. 2010). Most chemical substances available in commerce are not monitored in the air, soil, or water (Muir and Howard 2006).

Numerous sources of pollution are of concern to Canada’s biodiversity. On an annual basis, about 5 million tonnes of pollutants, including over 320 substances, are released to the environment from more than 7000 facilities across Canada (Government of Canada 2018). These include pollutants released into the air, land, and water from oil and gas extraction, manufacturing, mining, quarrying, electricity, and other sectors, including municipal wastewater treatment plants (Government of Canada 2018). Over 2000 wastewater facilities across the country contribute to the over 150 billion litres of untreated and undertreated sewage that is discharged into Canadian waterways every year (Government of Canada 2017a). In addition to these intentional discharges, pipeline and other transportation incidences result in the accidental release of pollutants to the environment. For example, close to 700 pipeline spills over the last decade resulted in the release of natural gas, crude oil, and other substances to the Canadian environment (National Energy Board 2019). Runoff from urban, agricultural, and industrial landscapes contaminates Canada’s groundwater and downstream aquatic ecosystems (Government of Canada 2017b). Finally, over 23 000 known or suspected contaminated sites have been identified and classified in urban, rural, and remote areas across Canada, many of which are contaminated with petroleum hydrocarbons, metals, and (or) persistent organic pollutants, such as polychlorinated biphenyls (Government of Canada 2017c).

To standardize the consideration of threats during species assessments, the International Union for the Conservation of Nature (IUCN) developed a threat classification system. Categories include anthropogenic threats such as Residential and Commercial Development, Agriculture and Aquaculture, Energy Production and Mining, and Pollution (Salafsky et al. 2008; IUCN 2019). In Canada, the IUCN’s threat classification system has been adopted by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC) in an effort to standardize federal species assessments. COSEWIC is an independent committee of experts charged with determining which species are at risk of extinction in Canada and re-assessing the status of each species at risk once every 10 years (Government of Canada 2016a). The scope and severity of each threat are considered independently; both are critical to accurately assess the impact of a threat on the target population. Scope estimates the proportion of the species’ population that can reasonably be expected to be affected by the threat within the next 10 years (NatureServe 2019a; Table 1). Severity represents the likely impact of each threat on the size of the population within 10 years, or three generations of the species, whichever is longer (NatureServe 2019a). Thus, a threat with high scope and high severity affects most individuals of the species and is likely to cause substantial declines. A threat with high scope and low severity affects most individuals of a species but is unlikely to cause the population to decline.

In Canada, species assessments are conducted by panels of experts in the biology of particular taxa. The members of these panels work together to complete a “Threats Calculator”, which quantifies the scope and severity of each category of threat. The accuracy of a species assessment relies on the available data, but complete “gold standard” data sets for most species and threat categories are lacking (Joppa et al. 2016). Species experts are well qualified to assess the ecological needs or population trends of their study species. However, these experts may not necessarily have expert knowledge of all threat categories, or of the ways in which each threat is most likely to affect their study species. In the experience of some of us who have served on COSEWIC committees, the lack of data available for assessing the scope of pollution in particular became painfully clear, which led us to wonder: how accurately is the scope of pollution as a threat to Canadian species at risk currently being assessed and, furthermore, could accuracy be improved by a rigorous, quantitative method?

In this study, we quantified the scope of pollution threat as the extent of geographic overlap between point sources of pollution and the distributions of Canadian species at risk. We then compared our estimate of scope with the scope of pollution listed in the Threats Calculators completed using expert opinion through the COSEWIC process. Our goals were to determine: (i) what proportion of the observed distribution of each species at risk coincides with conservatively mapped sources of pollution, (ii) the geographic or taxonomic patterns in pollution and distribution co-occurrence, and (iii) the extent to which Threats Calculator estimates of scope correspond to the geographical co-occurrence of pollution sources and species. We do not address the severity of pollution as a threat to Canadian species at risk, as this would require data that, in most cases, are simply not available. However, our results provide a tool to estimate the minimum scope of exposure to pollutants for species assessed or re-assessed in the future.

Our final database included 588 terrestrial and freshwater species at risk, comprised of 24 amphibians, 64 arthropods, 83 birds, 71 fish, 19 lichens, 37 mammals, 31 molluscs, 17 mosses, 44 reptiles, and 198 vascular plants (data available in Supplementary Material 1). Of these species, 555 had a published status report, and 488 had mapped occurrence records. A Threats Calculator was available for 109 species and the combination of a Threats Calculator and mapping data was available for 80 of these species.

We observed a strong geographic bias in species at risk and pollution (Fig. 1). The richness of species at risk overlapped strongly with areas of greatest urbanization and landscape modification, such as southern Ontario, the Prairies, and the Lower Mainland of British Columbia (Fig. 1A). The number of categories of mapped pollution sources within grid squares occupied by one or more species at risk also tended to be higher in southern regions (Fig. 1B; see Supplementary Material 2 for overlap between occurrences of species at risk and each pollution category separately).

The proportion of mapped species occurrences that coincided with one or more categories of pollution (considering all pollution categories) ranged widely across the 488 species with mapping data (Fig. 2). On average, over half (57%) of a species’ range overlapped with at least one pollution source (all categories) and almost one-quarter (22%) of all species (n = 108) had at least one pollution source in all of the 100 km² grid cells they are known to occupy (100% range overlap). Just over 12% of all species (n = 59) did not overlap with a single pollution source. Significant differences in scope_GEO existed between taxonomic groups; on average, fish and molluscs exhibited a significantly higher proportion of overlap compared with lichens and mammals, whereas vascular plants exhibited a significantly higher proportion of overlap compared with mammals (Fig. 2; Supplementary Material 3). Between taxonomic groups, differences were also present when considering each pollution category separately (Supplementary Material 3). For example, fish had significantly greater average overlap with Household Sewage and Urban Waste Water (category 9.1) than birds or mammals (respectively, mean percentage overlap = 50%, 30%, 20%). Mosses had significantly greater average overlap with Air-borne Pollutants (category 9.5) than birds, lichens, or mammals (respectively, mean percentage overlap = 49%, 20%, 8%, 16%; Supplementary Material 3).

We also observed high inter-specific variation among scope_TC scores (Fig. 3). However, most species with Threats Calculators received either the highest scope (“pervasive”) or zero; fewer species received intermediate scores. Freshwater fish and birds had significantly higher scope_TC than vascular plants. Pairwise differences between taxonomic groups were otherwise not significant (Table S4b, Supplementary Material 4). There were also some significant differences when considering each pollution category separately (Supplementary Material 4). For example, fish and molluscs had significantly higher scope for Agricultural Effluents (category 9.3) than mammals and vascular plants.

For the 80 species with both mapping data and Threats Calculators, the relationship between scope_GEO and scope_TC was weak. Although scope_TC was positively related to the proportion of geographic overlap with pollution sources, there was substantial unexplained variation (Fig. 4). A cumulative logit model for scope_TC based on scope_GEO was significantly better than an intercept-only model (p(χ²) = 0.029; ΔAIC = 2.79); however, the model explained only 1.77% of the residual deviance. The significant relationship is driven by the tendency for species with scope_TC pervasive to have higher proportion overlap, and species with scope_TC zero to have lower proportion overlap—there was little or no correspondence for the scope levels in between (Fig. 4; Supplementary Material 5).

Scope_GEO and scope_TC correspond strongly for some taxonomic groups, but poorly for others (Fig. 5). For example, mollusc species generally had a higher scope_TC with higher scope_GEO. The Yellow Lampmussel (Lampsilis cariosa) and Round Hickorynut (Obovaria subrotunda) both received a Threats Calculator scope of pollution of pervasive, corresponding to geographic overlap of 66.7% and 100%, respectively, whereas the Haida Gwaii snail with scope_TC of 0 correspondingly had a scope_GEO of 0%. However, some taxonomic groups, including molluscs, include species that scored 0 for scope_TC despite a large proportion of their known occurrences coinciding with one or more pollution sources. Incongruous scoring for scope was especially common in Threats Calculators for vascular plants and terrestrial mammals (Fig. 5).

Results comparing scope_TC and scope_GEO for each pollution category separately varied by pollution category. Species with a higher Threats Calculator scope for Household Sewage and Urban Waste Water (category 9.1) and Agricultural Effluents (category 9.3) also tended to have a higher proportion overlap, but there was a great deal of variation (Supplementary Material 5). Importantly, scope_GEO was not a significant predictor of scope_TC for any of the other pollution categories, and scope_TC was heavily zero-inflated (Supplementary Material 5).

When considering all pollution categories together, we found little correspondence between the proportion of geographic overlap (scope_GEO) and assessed scope (scope_TC). Many species at risk that were assessed geographically as having high proportions of their known occurrences co-occurring with pollution sources received a score of zero or “negligible” for pollution in Threats Calculators. Mismatches of this type are concerning and suggest that the decision to score the scope of pollution as null or negligible may be made in the absence of supporting evidence. For example, the American hart’s-tongue fern (Asplenium scolopendrium var. americanum) occurs in 55 grid cells in southern Ontario, 45 of which also contain one or more pollution sources. Yet the Threats Calculator scores pollution scope as zero. The Lake Ontario population of the channel darter (Percina copelandi) occurs in only 10 grid cells, six of which co-occur with one or more point sources of pollution, but the Threats Calculator ranks the scope as negligible (<1% of the population affected). The gypsy cuckoo bee (Bombus bohemicus) and the spotted turtle (Clemmys guttata) have 86% and 67%, respectively, of their occupied grid cells co-occurring with at least one pollution source. However, the Threats Calculator for each of these species ranks the scope of pollution as “small” (1%–10% of the population). Other studies have suggested that exposure to pollution may be underestimated in conservation assessments and recovery planning. For example, an analysis of the impacts of oil development in Saskatchewan’s grasslands on species at risk concluded that the potential impacts of water pollution from oil or fracking were not mentioned in recovery documents, even though evidence supports that oil development has an impact on grassland birds and other species (Olive 2018).

There were also cases of mismatches in the opposite direction, whereby species with low geographic overlap with pollution sources nonetheless were scored a “large” or “pervasive” scope by experts participating in Threats Calculator panels. For example, many lichen species have received a scope of large for Air-Borne Pollutants, even though they rarely co-occur with point sources of this pollution category. Experts in these cases may have factored in the well-known ability of air pollutants to disperse over immense distances, and the documented accumulation of atmospheric contaminants in lichens, in assessing the scope of this pollution category (e.g., Carignan et al. 2002; Graney et al. 2017).

Our analysis shows that species at risk and pollution sources co-occur at a high rate in Canada. In general, the highest densities of pollution sources and species at risk are concentrated in the south, where the human population density is also highest. This observation corroborates a well-known pattern in Canada, in which species are disproportionately threatened in southern portions of the country, corresponding with regions of high human habitation (Coristine and Kerr 2011). There is no question that species at risk in Canada are threatened by habitat destruction, natural resource extraction, hunting and gathering, and disturbance via human recreation and work (McCune et al. 2013). Nonetheless, the scope of pollution as a threat is also high—especially for southern species—and should not be overlooked despite the dominant and sometimes more obvious presence of other threats. The taxonomic differences that we observed in proportion of geographic overlap between species’ ranges and pollution sources largely conforms with the distribution patterns of taxonomic groups. For all pollution categories combined, mammals and lichens have the lowest proportional overlap because they tend to have more northerly distributions, whereas freshwater fishes, molluscs, and vascular plants are dominated by species with southerly distributions.

We assume that if a species is present in a 10 km × 10 km grid cell that also contains one or more pollution sources, there is potential for exposure to the pollutant. In practice pollution effects will vary with the category of pollution, type of species, and proximity of the species to the source of pollution. For example, if a vascular plant species is present in the same grid cell as a municipal garbage dump, but is still several kilometres away, there may be a very low chance that the dump will threaten the plant species. In the case of other threat categories, the potential for a pollution source to threaten a species that co-occurs in the same grid cell is not easy to estimate. For example, nearly one-third of mapped vascular plant species at risk have 50% or more of their occurrences in grid cells that also include intensive agricultural land use. There are few studies considering the impact of herbicide or pesticide drift, or other agriculture-related pollution, on vascular plants growing nearby or at what distances such effects could manifest (but see Marrs et al. 1989; Boutin and Jobin 1998). Based on the definition of scope, it is reasonable to expect that those plants could be affected by agricultural effluents, yet only one vascular plant with a Threats Calculator has a scope >0 for agricultural pollution. In general, especially when little is known about contaminant concentrations, the distances they can move, or susceptibility of the species, it is prudent to assume that populations within a 10 km × 10 km grid cell that also contains a pollution source are more likely to be exposed to that pollution category than those in grid squares that do not contain a potential source of pollution. Therefore, species that co-occur with a pollution category should receive a non-zero value for the scope of that pollution category.

The scope of pollution as assessed by COSEWIC expert panels in Threats Calculators (scope_TC) was significantly lower for vascular plants than for freshwater fishes and birds. Taxonomic differences in scope_TC may reflect the expertise or conventions of the groups conducting assessments. COSEWIC relies heavily on the expertise of Species Specialist Sub-Committees (SSCs) that are responsible for particular taxonomic groups and whose members participate in completing Threats Calculators. Different SSCs may have slightly different tendencies in their scoring of pollution as a threat. For example, members of Threats Calculator groups for vascular plants may tend to record the scope of pollution as zero when the severity of pollution is deemed minimal or not a factor. We believe it is important to keep scope and severity separate, especially given the rapid evolution of new sources of pollution and changing scientific understanding of the effects of pollution on species. Data on geographic overlap provides a standardized and more rigorous method to quantify minimum scope. The species experts can then use these geographic data to inform more nuanced, species-specific discussions.

Our geographically explicit determination of overlap contributes to an evidence-based, systematic determination of minimum scope, which can then be modified based on evidence if and when evidence is available. Additional evidence could include substantiated information about: the area of pollutant influence, environmental persistence, and susceptibility of a particular species or taxonomic group. For example, excess energy such as light pollution has a pronounced effect on bird species’ migration routes, foraging times, and reproduction (Longcore and Rich 2004), yet the impacts to plants may be less direct (Bennie et al. 2015).

Pollution has been ranked below other threats, including habitat loss, overexploitation, and invasive species, in literature reviews of the dominant threats to species at risk (e.g., Venter et al. 2006; Prugh et al. 2010; McCune et al. 2013). A recent global analysis revealed that, even though the production and diversification of synthetic chemicals outpaces other agents of global change, mainstream ecological journals, meetings, and funding devote only a tiny effort (<2%) to the study of synthetic chemicals—which, unfortunately, has contributed to an ignorance among ecologists about how synthetic chemicals affect biodiversity and ecosystem function (Bernhardt et al. 2017). Our study, together with others, shows how the paucity of information about contaminants translates into pollution being underestimated as a threat to species at risk in Canada.

This project was made possible through funding from the Liber Ero Fellowship to all authors, an NSERC postdoctoral fellowship to JLM, and a Banting postdoctoral fellowship to DMO. We gratefully acknowledge NatureServe, the Nature Conservancy of Canada, and the Atlantic Canada Conservation Data Centre for sharing the species at risk occurrence data.