Federal environmental quality guidelines - aluminium
Official title: Canadian Environmental Protection Act, 1999 - Federal environmental quality guidelines - aluminium
Environment and Climate Change Canada
Federal Environmental Quality Guidelines (FEQGs) provide thresholds of acceptable quality in the ambient environment. They are based solely on the toxicological effects or hazards of specific substances or groups of substances. FEQGs serve three functions: first they can be an aid to prevent pollution by providing targets for acceptable environmental quality; second, they can assist in evaluating the significance of concentrations of chemical substances currently found in the environment (monitoring of water, sediment and biological tissue); and third, they can serve as performance measures to assess the effectiveness of risk management actions for the chemical substance. The use of FEQGs is voluntary unless prescribed in permits or other regulatory tools. Thus FEQGs, which apply to the ambient environment, are not effluent limits or “never-to-be-exceeded” values but may be used to derive effluent limits. The development of FEQGs is the responsibility of the Minister of Environment under the Canadian Environmental Protection Act, 1999 (CEPA 1999) (Government of Canada (GC) 1999). The intent is to develop FEQGs as an adjunct to risk assessment/risk management of priority chemicals identified in the Chemicals Management Plan (CMP) or other federal initiatives. Where data permit, FEQGs are derived following Canadian Council of Ministers of the Environment (CCME) protocols. FEQGs are developed where there is a federal need for a guideline (e.g., to support the Screening Assessment of Certain Aluminium-containing Substances under the CMP) but where the CCME guidelines for the substance have not yet been developed or are not reasonably expected to be updated in the near future.
This factsheet describes the Federal Water Quality Guideline (FWQG) for the protection of aquatic life from adverse effects of aluminium (Al) in freshwaters and is based on total aluminium. A multiple linear regression (MLR) approach was used to incorporate toxicity modifying factors (TMFs) into the draft guideline. The FWQG for aluminium follows CCME methods and meets CCME minimum data requirements for a Type A statistical approach (CCME 2007). There is no pre-existing FWQG for aluminium however there is an outdated CCME guideline that was published in 1987 expressed as total Al; 5 µg/L at pH <6.5 and 100µg/L ≥ pH 6.5 (CCREM 1987). The CCME guideline is contingent only on pH and was not derived using the preferred Type A approach. The derivation of this FWQG is based on the collection and evaluation of aquatic toxicity data published up to June 2019. No FEQGs have been developed for biological tissue compartments, sediment, or marine water at this time.
|FWQG Equation||FWQG (μg/L) = exp([0.645 × ln(DOC)] + [2.255 × ln(hardness)] + [1.995 × pH] + [-0.284 × (ln(hardness) × pH)] -9.898)|
|Example FWQGb||170 μg/L|
a The FWQG is expressed as an equation in order to calculate a site-specific FWQG. The FWQG is for total aluminium in fresh waters and is calculated using the equation above or using the FWQG calculator (Appendix B). The FWQG equation is valid between hardness 10 and 430 mg/L, pH 6 and 8.7, and dissolved organic carbon (DOC) 0.08 and 12.3 mg/L.
b As an example, the FWQG at a hardness of 50 mg/L, pH 7.5, and DOC concentration of 0.5 mg/L is 170 µg/L.
Aluminium (Al; CAS # 7429-90-5; molar mass 26.98 g/mol) is the third most abundant element and the most common metal in the Earth’s crust (USEPA 2018). Aluminium is often found combined with other elements, typically complexed with oxygen as oxides and silica as silicates, but rarely in the elemental state (ATSDR 2008; USEPA 2018).
Aluminium is commonly found in rocks, particularly in aluminosilicate minerals where it is considered toxicologically irrelevant (i.e. essentially inert, not bioavailable). When these minerals weather they slowly release potentially toxic forms of aluminium to the environment (i.e. Al3+, Al hydroxides etc.) (GC 2010; USEPA 2018). The most common ore for aluminium metal is the mineral bauxite (ATSDR 2008). Aluminium metal is light-weight, ductile, and silvery-white in appearance. It is considered a non-essential element because it plays no important biological function and offers no beneficial properties to life. The speciation and solubility of aluminium in surface waters are greatly affected by various water quality parameters, most importantly pH (Cardwell et al. 2018). In the water column, aluminium may be present as dissolved complexes (both organic and inorganic), as a free ion (Al3+), in association with particles, as colloids, or as solids precipitating to the sediment (GC 2010). Aluminium is commonly found in aquatic systems as a result of both natural and anthropogenic inputs. Elevated levels in surface waters can cause toxic effects to aquatic organisms.
Sources and uses
Aluminium metal and aluminium compounds are used in a variety of applications in Canada as well as worldwide. Aluminium sulphate and chloride salts are primarily used in municipal drinking water and wastewater treatment as a flocculating agent to help remove suspended particles and bacteria from the water (ATSDR 2008; GC 2010). They are also used as an additive in the pulp and paper industry for paper sizing (GC 2010). Consumer products containing aluminium include: antacids, astringents, buffered aspirin, food additives, antiperspirants, natural health products, cosmetics, beverage cans, pots, pans, and foil (ATSDR 2008; GC 2010). As a light-weight conductive metal, it is widely used in the construction, transportation, and electronic and electrical industries for products ranging from airplanes to power lines (ATSDR 2008; NRCan 2018).
Bauxite, the primary aluminium ore, must be chemically refined into alumina, and then smelted to form pure aluminium metal. Bauxite is not mined in Canada; however there is one alumina refinery (located in Quebec) and ten smelters; nine located also in Quebec and one in British Columbia (NRCan 2018). Canada is the world's fourth largest primary aluminium producer after China, Russia, and India, producing an estimated 2.9 million tonnes in 2018 (NRCan 2018). Some aluminium compounds are manufactured in Canada, notably aluminium chloride and aluminium sulphate, primarily for use within Canada as opposed to exportation (GC 2010). Anthropogenic sources of aluminium include effluent from water treatment plants where aluminium compounds are added as clarifying agents (industrial water, drinking water or wastewater), fossil fuel combustion, and emissions from the processing of aluminium ore and aluminium production (USEPA 2018; GC 2010; ATDSR 2008).
National Long-term Water Quality Monitoring Data (ECCC 2018a) were queried for total aluminium concentrations in surface waters (2000-2018) and then organized by province/territory in Table 2. An ECCC database, referred to as GENIE (ECCC 2018b) was also queried for aluminium concentrations in the Great Lakes (2012-2017) and data are included in Table 2. Total aluminium concentrations for Canadian jurisdictions varied from detection limit (0.04 µg/L) for Nova Scotia to 58,500 µg/L for Alberta. The median (50th percentile) was less variable, ranging from 6 µg/L for Ontario to 500 µg/L for Manitoba. Canadian monitoring data for other parameters (e.g., DOC, hardness, pH) are presented in Appendix A.
|British Columbia||13620||< 0.2||11||23||70||256||877||25600|
|New Brunswick||301||< 4.0||12||18||68||156||264||1782|
|Newfoundland and Labrador||4571||< 0.5||34||59||88||137||244||9680|
|Northwest Territories||1145||< 0.1||4||22||99||702||1986||13600|
|Nova Scotia||2151||< 0.04||60||98||168||255||355||2900|
|Ontario/Great Lakes||245||< 0.5||1||2||6||25||99||1410|
|Prince Edward Island||40||12.4||29||54||82||171||466||3420|
Mode of toxic action
Aluminium has no known biological function and is therefore considered a non-essential element. The toxic mode of action of aluminium for fish has been widely investigated, however information is less available for invertebrates and is especially limited for aquatic plants and algae. Aluminium elicits toxic effects on fish by two main modes of action; disturbance of ionoregulatory processes and respiratory disruption (Exley et al. 1991; Gensemer and Playle 1999; GC 2010; Gensemer et al. 2018; Cardwell et al. 2018). Gills are the primary biological ligand to which aluminium binds to fish (Exley et al. 1991; Teien et al. 2006; USEPA 2018). Aluminium binding to the gill surface disturbs ionoregulation, leading to reduced ion uptake, loss of plasma ions, and changes in blood parameters (GC 2010; USEPA 2018). Damage to ionoregulation, respiration, or a combination of the two may ultimately lead to death. The chemical impact on ionoregulatory processes, such as a decrease in plasma Na+ and Cl¯ ions, are more common under acidic conditions where dissolved monomeric aluminium species (Al3+) are dominant (Gensemer and Playle 1999; GC 2010; Gensemer et al. 2018). Physical effects are more common at circumneutral pH values (6-8), where aluminium hydroxide precipitates at the gill surface causing the clogging of the interlamellar spaces with mucous which can eventually lead to hypoxia (Gensemer and Playle 1999; GC 2010; Gensemer et al. 2018).
Aluminium accumulates on mostly respiratory or ionoregulatory surfaces of invertebrates but can accumulate over the whole body (Gensemer and Playle 1999). Ionoregulatory effects are the most documented responses to aluminium exposure for invertebrates, while respiratory effects are reported much less frequently in invertebrates than in fish (Gensemer and Playle 1999; GC 2010; USEPA 2018). Respiratory effects occur when aluminium binds to or precipitates onto the bodies of invertebrates, forming a physical barrier that obstructs respiration (GC 2010).
The mode of toxic action of aluminium to aquatic plants and algae is not well understood. Aluminium can bind to polyphosphates; forming non-bioavailable complexes and thus making phosphorus unavailable for growth (Gensemer and Playle 1999; GC 2010; Petterson et al. 1988; USEPA 2018). This can occur intracellularly as well in the surrounding water. Aluminium is also adsorbed into the cell wall when cyanobacteria are exposed to high concentrations of phosphate (Petterson et al. 1985).
Fate, behaviour and partitioning in the environment
Aluminium chemistry in surface waters is complex. Aluminium may be present as dissolved complexes (with both organic and inorganic ligands), as a free ion (Al3+), in polynuclear aluminium species, in association with particles, as colloids, or as solids precipitating to the sediment (GC 2010). There are many factors that influence the fate, behaviour, and bioavailability of aluminium including temperature, the presence of complexing ions or ligands, and, most importantly, pH. Aluminium is amphoteric, which means it can act as either an acid or base. Aluminium is relatively insoluble at more neutral pH levels (6-8) (USEPA 2018; Gensemer and Playle 1999; GC 2010). Aluminium solubility is also dependent on dissolved organic carbon (DOC) and temperature (Wilson 2012; USEPA 2018; Rodriguez et al. 2019). DOC is an important ligand with which aluminium forms complexes, reducing concentrations of monomeric aluminium in the water column. Aluminium is a strongly hydrolysing metal, and, unlike some metals (e.g. iron and manganese), aluminium speciation does not depend on redox conditions (Gensemer and Playle 1999; GOC 2010).
At low pH values (<6), dissolved aluminium is present mainly in the free ion form (Al3+). As pH rises, hydrolysis occurs forming hydroxide complexes (e.g., Al(OH)2+, Al(OH)2+). Solubility reaches a minimum at circumneutral pH (6-8). Solubility starts to rise again at high pH values (>8) due to the formation of the anion Al(OH)4– (Driscoll and Schecher 1990; GC 2010). Figure 1 depicts the solubility of aluminium species in relation to pH.
Solubility of aluminium species (and total aluminium) in relation to pH in a system in equilibrium with microcrystalline gibbsite. At low pH values (<6), dissolved aluminium is present mainly in the free ion form (Al3+). As pH rises, hydrolysis occurs forming hydroxide complexes (e.g., Al(OH)2+, Al(OH)2+). Solubility reaches a minimum at circumneutral pH (6-8). Solubility starts to rise again at high pH values (>8) due to the formation of the anion Al(OH)4–
Under circumneutral pH conditions, aluminium changes from dissolved monomeric forms to insoluble polymers, which precipitate out of solution. Transient forms of polymeric aluminium (colloidal and amorphous) exist for a short-time (minutes to hours) during this transformation. Larger polymers and minerals in crystalline forms take several days to weeks to fully form. Aluminium toxicity to aquatic species under these conditions may be of a lesser concern since the transient forms do not exist long enough to cause harm. However, an exception to this generalization occurs when there is a continual input of an acidic solution containing aluminium. For example, aluminium toxicity is a particular concern where episodic acidic pulses occur and in mixing zones, where aluminium-rich acidic waters meet more neutral water (Rodriguez et al. 2019). Episodic acidic pulses, for example winter snowmelt or acid rain events, may mobilize aluminium from soil and sediment, increasing bioavailability and potential for toxicity to aquatic organisms (Gensemer and Playle 1999; Wilson 2012; USEPA 2018). Acid rain was the focus of a lot of research during the late 1970s to early 1990s due to observed toxic effects in both terrestrial and aquatic environments. It was observed that not only were organisms affected by the decline in pH but also by the mobilization of metals. Aluminium like most metals increases in solubility at low pH and the combination was subsequently found to be a major factor in the decline of the affected ecosystems (Wilson 2012).
Most aluminium from waterborne exposure rapidly adsorbs to external gill and body surfaces of fish and invertebrates. Internalization from cellular uptake also can occur but takes place more slowly, accumulating in internal organs like muscle, kidney, and liver over time (Wilson 2012; USEPA 2018). Uptake and bioaccumulation of aluminium via diet is considered unlikely and there is no evidence of biomagnification through the food chain (Wilson 2012; USEPA 2018).
Aluminium in air is transported as windblown particulate matter and can be deposited onto land and water through deposition (USEPA 2018). Aluminium concentrations in the atmosphere are considered to be negligible compared to the majority of aluminium entering surface water from the weathering of rocks or soil (GC 2010). Aluminium is ubiquitous in rocks and soil (silt and clay) in the form of aluminosilicate minerals. Gibbsite (Al(OH)3) is generally considered to be the most important mineral in modelling the geochemistry and transport of aluminium in aqueous systems (Driscoll and Postek 1996; Gensemer and Playle 1999; Wilson 2012). As these rocks and minerals weather and factors such as pH fluctuate, aluminium from soil can be transported into the aquatic environment. Aluminium in sediment is generally considered non-bioavailable when it is bound with DOC or in the form of silt or clay. Therefore, sediment can act as a sink for aluminium. However, as conditions change, such as a decrease in pH, aluminium in the sediment can mobilize back into the water column.
Aquatic toxicity data
Data compiled by the USEPA for the Aquatic Life Ambient Water Quality Criteria (AWQC) for Aluminum (USEPA 2018) formed the foundation of aquatic toxicity data considered for development of the aluminium FWQG. A detailed review of studies from this source was performed by ECCC following the CCME (2007) guidance for data quality. Determinants of test acceptability included, but were not limited to, exposure duration, analytical determination of aluminium exposure concentrations and other water quality parameters, documentation of the control response, the use of suitable biological endpoints and the inclusion of appropriate statistical analyses of the data collected in the study. Aluminium nitrate, sulfate, and chloride salts were the aluminium compounds used in the toxicity tests considered for the derivation of the guideline. A total of 733 chronic toxicity endpoints for 24 species from 26 studies were identified as acceptable. Many acceptable studies reported results for multiple effects (e.g., reproduction, growth, mortality) with multiple endpoints (e.g., NOEC, LOEC, ECx). The full toxicity dataset is presented in Appendix A.
It is often not possible to measure the absolute total aluminium concentration in water because of the limitations in routine sampling and analytical methods. The total recoverable aluminium is often used to represent the total aluminium concentration. Within the total recoverable fraction, herein referred to as total, both particulate (bound to or incorporated into suspended matter and minerals) and dissolved aluminium fractions are included. The FWQG for aluminium is based on measurements of total aluminium. Aluminium toxicity studies were only considered if total concentrations were reported in the toxicity test.
Often metal toxicity is best characterized by the dissolved fraction of a metal (operationally defined as the concentration recovered after being passed through a 0.45μm filter), as it is often shown to correlate with toxicity better than total concentrations (e.g. zinc, copper). However, aluminium behaves differently because of chemical speciation and solubility characteristics at different pH values. Multiple studies available in the scientific literature demonstrate the dissolved fraction alone does not correspond with aluminium toxicity. Gensemer et al. (2018) conducted both acute and chronic tests using Pimephales promelas, Ceriodaphnia dubia, and Pseudokirchneriella subcapitata at circumneutral pHs (6–8), finding that toxicity was either reduced or removed by filtration and that dissolved concentrations did not correlate with toxicity. This finding is consistent with results of Cardwell et al. (2018) where similar tests were conducted on several other freshwater species. These two studies also showed that concentrations of dissolved aluminium remained relatively constant regardless of the initial added aluminium, suggesting that concentrations of dissolved aluminium are limited by the solubility of the aluminium test compounds (Cardwell et al. 2018; Gensemer et al. 2018). Colloidal and precipitated forms of aluminium, which are removed by a filter in dissolved measurements, were found to cause toxicity to aquatic organism under circumneutral pH conditions (Cardwell et al. 2018; Gensemer et al. 2018).
Since a FWQG based on dissolved aluminium would underestimate toxicity, dissolved measurements were not used. The FWQG is instead based on total aluminium measured in laboratory water in order to reflect all forms of aluminium that result in toxicity. This decision is consistent with the AWQC for Aluminum from the USEPA (USEPA 2018). All aluminium concentrations are expressed as total aluminium herein unless otherwise specified.
Toxicity modifying factors
Toxicity modifying factors (TMFs), such as pH, DOC, and water hardness as CaCO3 (herein referred to as hardness) can alter the bioavailability of aluminium and hence the toxicity to aquatic organisms. Therefore, it is important in guideline derivation to incorporate TMFs when the data are available. TMFs are often incorporated into water quality guidelines by either a multiple (or single) linear regression (MLR) approach or a biotic ligand model (BLM). MLRs (DeForest et al. 2018) and a BLM (Santore et al. 2018) for total aluminium were published in 2018. Both approaches were investigated for potential use for the development of the FWQG for aluminium.
Biotic ligand model
A Biotic Ligand Model (BLM) was developed by Santore et al. (2018) and then customized by Windward under contract with ECCC in 2019 to follow ECCC methodology, dataset, and format. Since the source of aluminium toxicity can be caused by both dissolved and precipitated forms depending on the chemical conditions, the BLM models the toxicity of both forms of aluminium, attributing the toxic effect to the dissolved portion of aluminium until it reaches the solubility limit, then attributing the rest of the toxic effect to precipitated aluminium. The effects caused by each form of aluminium are modeled as a concentration-response relationship. The slopes of the response curves were calibrated for three species: P. promelas, C. dubia, and P. subcapitata. These three species are used as representatives for fish, invertebrate, and plant/algae species, respectively, for which specific parameter files have not yet been calibrated. Following CCME (2007) methodology, the aluminium BLM software also creates species sensitivity distributions (SSDs) and produces hazard concentration values for the fifth percentile (herein referred to as HC5 values). Please refer to Santore et al. (2018) for more information on the approach.
The ECCC version of the aluminium BLM was investigated as a method to incorporate bioavailability into the FWQG. While investigating the BLM, it became apparent that it produces considerably low HC5 values. A seemingly disproportionately large effect of temperature in the BLM is assumed to be one of the reasons for the low values, having a larger effect than even hardness and DOC at some water chemistry combinations. There are limited experimental data showing the effect of temperature on aluminium toxicity, while holding other water chemistry parameters constant. When BLM-based HC5 values were compared to Canadian monitoring data very high numbers of exceedances resulted, regardless of the type of site (i.e. whether the site is considered reference condition, or is potentially exposed to anthropogenic inputs). The BLM produces very low guideline values, which have not been validated fully by experimental data, making the method difficult to use in practice (e.g., a potentially very high false-positive rate).
Since the peer review of the draft Aluminium FWQG (January 2020), a new version of the BLM, has been developed by Windward Consulting (unpublished) and was provided to ECCC to review. The updated BLM produces much higher HC5 values and different trends in HC5 values compared to the previous version. In addition, the effect of temperature, which has a significant effect on the results using the previous version, had no clear effect in the updated version. The differences between the two BLM versions was unexpected and is not fully understood.
In addition, the BLM generally is a relatively complex model making it more difficult for users to fully understand, requires software to implement which can be a source of user frustration, and ideally uses a complete characterization of water quality, which users may not have. Because of the current uncertainties surrounding the aluminium BLM (e.g. unexplained differences between the versions) and for the other reasons noted above, the BLM method was not used for the development of the FWQG.
Multiple linear regression
A multiple linear regression (MLR) approach was used to incorporate TMFs into the draft FWQG for aluminium Chronic MLRs were developed by DeForest et al. (2018) for the three main trophic levels within a freshwater environment, represented by the fathead minnow (P. promelas), the water flea (C. dubia) and an alga (P. subcapitata). Most data used to create the MLR relationships were published by Gensemer et al. (2018). Nine additional C. dubia and P. promelas toxicity tests were conducted by Oregon State University (OSU) in order to expand the ranges of water chemistry conditions for model development (DeForest et al. 2020; OSU 2018a,b,c). The MLRs were therefore updated by the authors and made available to ECCC for use in this assessment. Three-day EC10s (growth) for P. subcapitata (n= 27), 7-d EC10s (reproduction) for C. dubia (n=32), and 7-d EC10s (biomass) for P. promelas (n=31) were used to create the MLR relationships (DeForest et al. 2020). One 33-d EC10 (survival) for P. promelas was also included. The inclusion of this endpoint was justified by the authors because the 7-d survival and growth test had a similar sensitivity as the 33-d survival and growth test. A pooled MLR model was also derived, combining C. dubia and P. promelas aluminium toxicity datasets (DeForest et al. 2020).
MLR models were developed for a variety of terms including the independent variables of DOC, pH, and hardness. A pH2 term and the following interaction terms were also considered based on the knowledge of aluminium speciation and bioavailability: DOC × pH; DOC × hardness; and hardness × pH. The pH2 term takes into account that aluminium bioavailability decreases from pH 6 to 7 and then increases from pH 7 to 8 (DeForest et al. 2018). A negative DOC × pH term characterizes the tendency for a decrease in the mitigating effect of DOC as pH increases; a negative DOC × hardness term would reflect the tendency of a decrease in the mitigating effect of DOC as hardness increases; and a negative hardness × pH term would reflect the tendency of a decrease in the mitigating effect of hardness as pH increases (DeForest et al 2018). A summary of the results for the best fit MLR models are presented in Table 3. All three MLRs for the different taxa retained DOC, hardness, and pH but different interactive terms. For more detailed information on the MLR analyses see DeForest et al. (2018; 2020). The DeForest et al. (2018; 2020) MLRs do not include temperature as a TMF and there are currently not enough data to do so.
Ninety-one percent of predicted C. dubia EC10 values (29 of 32), 94% of predicted P. promelas EC10 values (29 of 31), and 100% of predicted P. subcapitata EC10 values (27/27) were within a factor of two of observed EC10 values from the dataset used to create the individual species MLR relationships (DeForest et al.2018; 2020). Using the pooled MLR model, predictability of P. promelas endpoints decreased slightly from 94% to 90% and predictability of C. dubia endpoints remained the same at 91%.
|Species||n||Adj. R2||Intercept||DOC||Hardness||pH||pH2||DOC ×|
|DOC × Hardness||Hardness×|
|Pooled (C. dubia + P. promelas)||63||0.88||- 8.618|
An approach was investigated which used the C. dubia MLR to normalize all invertebrate endpoints, the P. promelas MLR to normalize all fish endpoints, and the P. subcapitata MLR to normalize all aquatic plant endpoints before plotting SSDs. Since this approach involves multiple MLRs with different slopes, a final guideline equation could not be calculated. The CCME (2007) protocol requires the use of SSD software to create fitted SSD curves. Therefore, one y-intercept for use in the guideline equation cannot be derived when using multiple MLRs. Instead, look up tables of HC5 values derived from different SSDs normalized to various water chemistry combinations were used, requiring rounding when user inputs fall between the pre-calculated SSDs. In addition, because all three individual MLRs differ in slope, including interaction term slopes, combining them into SSDs caused trends in HC5 values that may not be supported by the science, and some of which were believed to be statistical artifacts of the SSD. Following this approach, P. subcapitata was often an outlier in SSDs normalized to high pH values ( pH>8). This caused particularly poor fit of the SSD at this pH range. The BLM does not show the same sensitivity of algae at high pH that is suggested by the P. subcapitata MLR. The approach was not used to develop the FWQG due to the above reasons.
A pooled MLR (C. dubia and P. promelas) was also investigated. The pooled MLR incorporates 68 toxicity data points from 2 different species and taxonomic groups, has a high R2 value of 0.88, and has a similar level of accuracy in predicted EC10s compared to the individual species models. Algae data were not incorporated into the pooled MLR since the data showed significantly different slopes compared to fish and invertebrate data. The lack of algae data in the pooled MLR is recognized as an uncertainty, however the protectiveness assessment concluded plants/algae are protected by the FWQG (see Protectiveness Assessment). The pooled MLR approach allows for a guideline equation to be derived, results in a SSD with good fit, is considered protective and predictive, and is transparent and easy to use. The pooled (invertebrate and fish) EC10 MLR model was therefore chosen to be used in the guideline derivation for aluminium.
This approach is generally aligned with the USEPA AWQC (USEPA 2018). The USEPA also applied the DeForest et al. (2018 a,b) MLR approach however chose to use the separate fish and invertebrate MLRs instead of the pooled MLR. In addition, the two jurisdictions differ in general guideline derivation methods which includes the USEPA preference for EC20 values compared to EC10s preferred following CCME (2007) protocol.
Federal water quality guideline derivation
Federal Water Quality Guidelines (FWQGs) are preferably developed using the CCME (2007) protocol. In the case of aluminium, there were sufficient acceptable chronic toxicity data to meet the minimum data requirements for the preferred CCME Type A approach. A Type A guideline is a statistical approach that uses SSDs comprised of primarily “no effect” data to calculate HC5 values, which in turn become the final guideline value (CCME 2007).
Only data that fell within the acceptable ranges of the MLR (Table 6) were used in guideline derivation in order to avoid extrapolations beyond the MLR relationship. EC10 values were calculated using the USEPA Toxicity Relationship Analysis Program (TRAP v. 1.3) (USEPA 2015) where needed and the necessary underlying data were available. Reported DOC values of less than a detection limit (i.e. <1 or <0.5 mg/L) were changed to half the detection limit for use in equations based on USEPA recommendations (USEPA 2007; 2018). Reported DOC values of 0 mg/L were changed to 0.3 mg/L representing near zero values for use in equations. Seven endpoints used in the SSD dataset did not have reported DOC concentrations and therefore were estimated following USEPA recommendations (USEPA 2007; 2018). All SSD endpoints had reported hardness and pH values. Refer to Appendix A for the full list of toxicity endpoints, experimental conditions, water chemistry, etc.
The pooled MLR model and slopes (Table 3) were used to normalize all acceptable toxicity data points to a common water chemistry (DOC 0.5 mg/L, pH 7.5, and hardness 50 mg/L) using the equation:
ECx (at DOC 0.5 mg/L, pH 7.5, and hardness 50 mg/L) =EXP[(ln (original ECx))-0.645*(ln( original DOC)-ln(0.5))-2.225*(ln(original hardness)-ln(50))-1.995*(original pH-7.5)+0.284*((ln(original hardness)*original pH)-(ln(50)*7.5))]
A geometric mean was calculated where multiple comparable endpoints were available for the same species, effect, life stage and exposure duration. The most sensitive and preferred endpoint (or geometric mean) was then selected for each species following CCME (2007). A total of 54 endpoints for 14 species (3 fish, 8 invertebrates, 2 aquatic plants/algae, and 1 amphibian) were included in the SSD dataset and are summarized in Table 4. Salvelinus fontinalis (fish) was the most sensitive species in the dataset with a normalized effect concentration of 171 µg/L. Lemna minor (plant) was the least sensitive species in the dataset with a normalized effect concentration of 14,607 µg/L.
|Species scientific name||Species common name||Group||Endpoint||Effect concentration (µg/L)||Normalized effect concentrationa (µg/L)||Reference|
|Salvelinus fontinalis||Brook trout||Fish||60-d EC10 (Weight)||103.24||170.65||Cleveland et al. 1989|
|Pimephales promelas||Fathead minnow||Fish||7-d EC10 (Mean dry weight)||Geomean (n=2)||271.52||ENSR 1992a|
|Hyalella azteca||Amphipod||Invertebrate||28-d EC10 (Biomass)||142.6||307.46||Cardwell et al. 2018|
|Lampsilis siliquoidea||Fatmucket||Invertebrate||28-d EC10 (Dry weight)||109||312.73||Wang et al. 2018|
|Pseudokirchneriella subcapitata||Green algae||Plant/algae||72-h EC10 (Biomass)||Geomean (n=30)||358.77||Gensemer et al. 2018|
|Danio rerio||Zebrafish||Fish||33-d EC10 (Biomass)||98.2||397.42||Cardwell et al. 2018|
|Ceriodaphnia dubia||Water flea||Invertebrate||6-d EC10 (Reproduction)||Geomean (n=3)||435.88||ENSR 1992b|
|Bufo bufo||Common toad||Amphibian||7-d >NOEC||Geomean (n=2)||421.44||Gardner et al. 2002|
|Daphnia magna||Water flea||Invertebrate||21-d EC10 (Reproduction)||709.4||535.04||Gensemer et al. 2018|
|Lymnaea stagnalis||Great pond snail||Invertebrate||30-d EC10 (Dry weight)||Geomean (n=3)||870.38||OSU 2018d|
|Brachionus calyciflorus||Rotifer||Invertebrate||48-h EC10 (Reproduction)||Geomean (n=6)||1506.69||OSU 2018e, Cardwell et al. 2018|
|Chironomus riparius||Midge||Invertebrate||10-d EC10 (Growth)||971.6||1722.97||Cardwell et al. 2018|
|Aeolosoma sp.||Oligochaete||Invertebrate||17-d EC10 (Reproduction)||987.9||5942.63||Cardwell et al. 2018|
|Lemna minor||Duckweed||Plant/algae||7-d EC10 (Weight)||2175||14607.41||Cardwell et al. 2018|
a Effect concentrations normalized using the Pooled MLR model to a common water chemistry.
The R package (R version 4.03) ‘ssdtools’ (ssdtools version 0.3.2) as well as the corresponding user friendly “Shiny App” were used to create SSDs from the dataset (Dalgarno 2018, Thorley and Schwarz 2018). The package fit several cumulative distribution functions (CDFs) (log-normal, log-logistic, and log-gumbel) to the data using maximum likelihood estimation (MLE) as the regression method. Akaike information criterion (AIC), which is a measure of the relative quality of fit to the data set, was calculated for each distribution (Burnham and Anderson 2002). Using AICc, which is AIC corrected for small sample size, a model averaged HC5 can be established. The smaller the AICc the better the distribution fits the data set. Each model was then weighted; models with high value weight fit the data well compared to the others. See Schwarz and Tillmanns 2019 for more information on the approach.
The SSD and accompanying summary statistics at water hardness 50 mg/L, pH 7.5, and DOC 0.5 mg/L are presented in Figure 2 and Table 5, respectively. The full R script is available in Appendix A.
The 5th percentile value of the plot is 170 µg/L. This value is the site-specific federal water quality guideline for the site water that has the DOC concentration of 0.5 mg/L, hardness of 50 mg/L, and pH of 7.5. The guideline value represents the concentration below which one would expect either no, or only a low likelihood of, adverse effects on aquatic life.
|Distribution||AICc||Predicted HC5 (µg/L)||95% LCL (µg/L)||95% UCL (µg/L)||Weight||Weighted HC5 (µg/L)||Weighted 95% LCL (µg/L)||Weighted 95% UCL (µg/L)|
a Final guideline values are rounded to two significant figures. For example, 165 µg/L is used in the guideline equation derivation however 170 µg/L is presented as a final guideline value.
Because pH, DOC and hardness were identified as significant toxicity modifying factors as well as the interaction between hardness and pH, the FWQG is expressed as an equation in order to calculate a site-specific FWQG. The equation is based on the pooled MLR model slopes of 1.995 (pH), 0.645 (DOC), 2.255 (hardness), and -0.284 (hardness x pH), and the 5th percentile value of 165 μg/L derived from the SSD at pH 7.5, DOC of 0.5 mg/L and hardness of 50 mg/L.
Based on the pooled MLR model and the HC5 from the SSD, the y-intercept can be derived using the following equation:
y-intercept = ln(5th percentile) – [DOC slope × ln(DOC)] – [hardness slope × ln(hardness)] – [pH slope × pH] – [hardness*pH slope × (ln(hardness) × pH)]
= ln(165) – [0.645 × ln(0.5] – [2.255 × ln(50)] – [1.995 × 7.5] – [-0.284 × (ln(50) x 7.5)]
The FWQG equation for total aluminium is therefore:
FWQG (μg/L) = exp([0.645 × ln(DOC)] + [2.255 × ln(hardness)] + [1.995 × pH] + [-0.284 × (ln(hardness) × pH)] -9.898)
where the FWQG is in μg/L total aluminium, hardness is measured as CaCO3 equivalents in mg/L, pH is in standard units, and DOC is in mg/L.
The FWQG is for total aluminium and is found using the FWQG equation above, which has also been incorporated into the FWQG calculator in Excel (Appendix B). The FWQG equation is valid between hardness 10 and 430 mg/L, pH 6 and 8.7, and DOC 0.08 and 12.3 mg/L, which are the ranges of data used to derive the MLR slopes (DeForest et al. 2018; 2020) (Table 6). Only values within these ranges should be entered into the guideline equation to ensure the equation is accurate and the FWQG is protective. Any user inputs into the FWQG Calculator that are outside of these ranges are automatically rounded to the upper or lower bounds. If site-specific water hardness, pH and/or dissolved organic carbon (DOC) is not known, use the corresponding lower limits from Table 6 (the calculator will do this automatically).
It is recognized that some water bodies in Canada may have water chemistry measurements (See Appendix A) outside the valid range of the FWQG (Table 6). The FWQG Calculator was designed to only work within the domain of the MLR model and therefore, if users wish to calculate a more stringent guideline, they must use the guideline equation separately. Users may extrapolate only to more stringent guideline values. Users may extrapolate below pH 6 but not above pH 8.7. The CCME guideline for the protection of aquatic life for pH is in the range pH 6.5-9.0 (CCREM 1987) and should be considered if extrapolating below the pH limit. Due to the complexity of the hardness and pH relationship, it is suggested to not extrapolate outside the hardness range (10-430 mg/L), since less stringent guideline values may result. Users should not extrapolate beyond DOC 12.3 mg/L. Guideline values generated outside valid ranges are more uncertain and should be used with caution. Sites with parameters consistently outside the valid ranges may warrant consideration for the derivation of Site-Specific Water Quality Objectives (CCME 2003).
|Variable||pH||DOC (mg/L)||Hardness (mg/L)|
There is some uncertainty with comparing total aluminium measured in exposure water from laboratory based toxicity studies (as well as guideline values based on those measurements) with total aluminium in field-collected water. When total metal samples are measured in the field-collected water, all forms of aluminium are captured including potentially high amounts of non-bioavailable crystalline aluminium forms (e.g. minerals and large polymers). Total aluminium measured in water from lab-based toxicity studies lacks this crystalline form (Santore et al. 2018). The MLR and BLM are based on total aluminium measured in laboratory studies and the BLM specifically does not consider mineral aluminium in the model since it is considered toxicologically irrelevant (i.e. essentially inert). Therefore, it is being suggested by some in the scientific community that it is not ideal to compare guideline values to field-sampled total aluminium (Ryan et al. 2019). An alternative method, a pH 4 extraction, has been introduced by Rodriguez et al. (2019) with the expectation that the method will better estimate the bioavailable fraction of aluminium in natural waters, avoiding most of the mineral phases in the measurement. Additional validation tests are currently ongoing. If users experience exceedances while comparing the guideline to total aluminium measurements and if there is a reason to suspect a false-positive, it is suggested to consider other methods, such as the pH 4 extraction method, in place of the total aluminium measurements.
In addition, since aluminium is ubiquitous in the natural environment, it is also suggested to consider natural background concentrations at sites with guideline exceedances. In some cases, natural background concentrations of a substance may exceed the guideline without any apparent effect on biota (i.e., if the substance is not present in a bioavailable form). Under these circumstances, it might be necessary to modify WQGs to account for conditions that occur at the site. CCME (2003) provides guidance on two methods to establish site-specific water quality objectives, which can be: 1) slightly above the natural background level, or 2) at the upper limit of natural background concentrations. To define natural background levels, it is recommended that research is conducted into historical records of elevated aluminium concentrations with historical land uses (i.e., before and after human activity, and analysis on aluminium concentration trends). An extensive dataset of water parameters over several consecutive years for each site is required to estimate natural background levels.
A protectiveness assessment was conducted to determine if the protection clause of the CCME (2007) protocol should be invoked. Note that only laboratory derived data was used in this assessment. Assessing protectiveness using data from natural ecosystems, such as species diversity, is beyond the scope of this document. To determine whether the guideline is sufficiently protective, FWQGs were calculated for each of the acceptable endpoints in the toxicity dataset within the valid water chemistry ranges of the MLR. The FWQGs were then compared to measured toxicity values at their tested water chemistry. Ratios (measured concentration:FWQG) >1 indicate that the FWQG is protective of the toxicity value in that particular test, while ratios <1 indicate that the FWQG is higher than the observed toxicity, and hence may require further evaluation (Figure 3). This protectiveness assessment resulted in 98% (668/680) of acceptable toxicity data points being above the site-specific guideline. To ensure protectiveness, each of the 12 endpoints with ratios <1 were further examined to ensure none of them triggered the protection clause (CCME 2007). Endpoints plotting below the site specific FWQGs are for C. dubia (n=2; NOEC and LOEC (reproduction)), H. azteca (n=1; NOEC (biomass)), S. fontinalis (n=1; NOEC (growth), and P. subcapitata (n=8, seven EC10s (biomass) and one EC50 (biomass)). The geometric mean of all species ratios were above 1. For example, the species geometric mean ratio for P. subcapitata was 5, meaning on average the reported measured toxicity values were approximately 5 times higher than the site-specific FWQG. None of the endpoints below the guideline were for a species at risk, or for lethal effects equal to or above a level of 15% (CCME 2007). Overall examination of the available data suggests that the MLR-based aluminium FWQG is protective.
Ratios of effect concentrations to federal water quality guidelines of 680 acceptable endpoints are plotted against the measured effect concentrations. Data points that plot above the dashed line (1 to 1 line) indicate that the federal water quality guideline is protective of the toxicity value in that particular test, while values below the line indicate that the guideline is higher than the measured toxicity, and hence may require further evaluation. The figure shows that the water quality guideline for aluminium is protective.
[ATSDR] Agency for Toxic Substances and Disease Registry. 2008. Toxicological profile for aluminum. United States Department of Health and Human Services, Public Health Service, Atlanta, GA. (accessed 2018-11-28)
Cardwell, A.S., W.J. Adams, R.W. Gensemer, E. Nordheim, R.C. Santore, A.C. Ryan and W.A. Stubblefield. 2018. Chronic toxicity of aluminum, at a pH of 6, to freshwater organisms: Empirical data for the development of international regulatory standards/criteria. Environ. Toxicol. Chem. 37: 36–48.
[CCREM] (Canadian Council of Resource and Environment Ministers).1987. Canadian water quality guidelines. Prepared by the Task Force on Water Quality Guidelines [CCME] Canadian Council of Ministers of the Environment. 2007. A protocol for the derivation of water quality guidelines for the protection of aquatic life 2007. In: Canadian environmental quality guidelines, 1999, Canadian Council of Ministers of the Environment, 1999, Winnipeg.
[CCME] Canadian Council of Resource and Environment Ministers. 2003. Guidance on the Site-Specific Application of Water Quality Guidelines in Canada: Procedures for Deriving Numerical Water Quality Objectives. Prepared by the CCME Water Quality Guidelines Task Force.
Dalgarno, S. 2018. ssdtools: A shiny web app to analyse species sensitivity distributions. Prepared by Poisson Consulting for the Ministry of the Environment, British Columbia. (accessed 2020-03-3)
DeForest, D.K., K.V. Brix, L.M. Tear and W.J. Adams. 2018. Multiple linear regression models for predicting chronic aluminum toxicity to freshwater aquatic organisms and developing water quality guidelines. Environ. Toxicol. Chem. 37: 80-90.
DeForest, D.K., K.V. Brix, L. Tear, A.S. Cardwell, E.N. Stubblefield and W.J. Adams. 2020. Updated aluminum multiple linear regression (MLR) models for predicting chronic aluminum toxicity to freshwater aquatic organisms and developing water quality guidelines. Environ. Toxicol. Chem. 39: 1724-1736.
Driscoll C.T. and W.D. Schecher. 1990. The chemistry of aluminum in the environment. Environ. Geochem. Health 12: 28–49.
Driscoll C.T. and K.M. Postek. 1996. The chemistry of aluminum in surface waters. In: G. Sposito (Ed.). The environmental chemistry of aluminum. 2nd edition. Boca Raton (FL): CRC Press. p. 363–418.
[ECCC] Environment and Climate Change Canada. 2018a. National Long-term Water Quality Monitoring Data. Open Data. (accessed 02/01/2019).
ECCC. 2018b. National Long-term Water Quality Monitoring Data. Genie. Unpublished. (accessed 2019-01-02).
Exley, C., J.S. Chappell and J.D. Birchall. 1991. A mechanism for acute aluminium toxicity in fish. J. Theor. Biol. 151: 417-428.
Gardner, M.J., E. Dixon, I. Sims and P. Whitehouse. 2002. Importance of speciation in aquatic toxicity tests with aluminum. Bul. Environ. Cont. Toxicol. 68: 195-200.
Gensemer, R.W. and R.C. Playle.1999. The bioavailability and toxicity of aluminum in aquatic environments, Crit. Rev. Environ. Sci. Technol. 29: 315-450.
Gensemer, R., Gondek, J., Rodriquez, J. Arbildua, W. Stubblefield, A. Cardwell, R. Santore, A. Ryan, W. Adams and E. Nordheim. 2018. Evaluating the effects of pH, hardness, and dissolved organic carbon on the toxicity of aluminum to aquatic organisms under circumneutral conditions. Environ. Toxicol. Chem. 37: 49–60.
[GC] Government of Canada. 1999. Canadian Environmental Protection Act, 1999. S.C., 1999, c. 33, Canada Gazette. Part III, vol. 22, no. 3. Published by the Minister of Justice, Government of Canada.
GC. 2010. Priority substance list assessment report for aluminum salts. Environment Canada and Health Canada. (accessed 2018-11-28).
[NRCan] Natural Resources Canada. 2018. Aluminum facts. Mining/Materials. Natural Resources Canada, Ottawa. Site last modified: 21/08/18. Accessed 2020-06-22
[OSU] Oregon State University Aquatic Toxicology Laboratory. 2018a. Chronic Toxicity of Aluminum to the cladoceran, Ceriodaphnia dubia: Expansion of the empirical database for bioavailability modeling. Prepared by Oregon State University Aquatic Toxicology Laboratory, Corvallis, OR, USA. Prepared for Aluminum Reach Consortium, Brussels, Belgium.
OSU. 2018b. Short-term chronic toxicity of Aluminum to the fathead minnow, Pimephales promelas: Expansion of the empirical database for bioavailability modeling. Prepared by Oregon State University Aquatic Toxicology Laboratory, Corvallis, OR, USA. Prepared for Aluminum Reach Consortium, Brussels, Belgium.
OSU. 2018c. Short-term Chronic Toxicity of Aluminum to the Fathead minnow, Pimephales promelas: Validation of Aluminum Bioavailability Models. Prepared by Oregon State University Aquatic Toxicology Laboratory, Corvallis, OR, USA. Prepared for Aluminum Reach Consortium, Brussels, Belgium.
OSU. 2018d. Chronic toxicity of aluminum to the great pond snail, Lymnaea stagnalis: Validation of aluminum bioavailability models. Prepared by Oregon State University Aquatic Toxicology Laboratory, Corvallis, OR, USA. Prepared for Aluminum Reach Consortium, Brussels, Belgium.
OSU. 2018e. Chronic toxicity of aluminum to the rotifer, Brachionus calyciflorus: Validation of aluminum bioavailability models. Prepared by Oregon State University Aquatic Toxicology Laboratory, Corvallis, OR, USA. Prepared for Aluminum Reach Consortium, Brussels, Belgium.
Pettersson, A., L. Kunst, B. Bergman and G.M. Roomans. 1985. Accumulation of aluminium by Anabena cylindrica into polyphosphate granules and cell walls: an X-ray energy-dispersive microanalysis study. J. Gen. Microbiol. 131: 2545-2548.
Pettersson, A., L. Hallbom and B. Bergman. 1988. Aluminum effects on uptake and metabolism of phosphorus by the cyanobacterium Anabaena cylindrica. Plant Physiol. 86: 112-116.
Ryan, A.C., R.C. Santore, S. Tobiason, G. WoldeGabriel and A.R. Groffman. 2019. Total recoverable aluminum: not totally relevant for water quality standards. Integr. Environ. Asses. 15: 974-987.
Rodriguez, P. H., J. Arbildua, G. Villavicencio, P. Urrestarazu, M. Opazo, S.A. Cardwell, W. Stubblefield, E. Nordheim and W. Adams. 2019. Determination of Bioavailable Aluminum in Natural Waters in the Presence of Suspended Solids. Environ. Toxicol. Chem. 38: 1668-1681.
Santore, R.C., A.C. Ryan, F. Kroglund, P. Rodriguez, W.A. Stubblefield, A.S. Cardwell, W.J. Adams and E. Nordheim. 2018. Development and application of a biotic ligand model for predicting the chronic toxicity of dissolved and precipitated aluminum to aquatic organisms. Environ. Toxicol. Chem. 37: 70–79.
Schwarz, C.J. and A.R. Tillmanns. 2019. Improving statistical methods to derive species sensitivity distributions. Water Science Series, WSS2019-07, Province of British Columbia, Victoria.
Teien, H.C., W.J.F. Standring and B. Salbu. 2006. Mobilization of river transported colloidal aluminium upon mixing with seawater and subsequent deposition in fish gills. Sci. Total. Environ. 364: 149-164.
Thorley, J. and C. Schwarz. 2018. ssdtools: Species Sensitivity Distributions. R package version 0.0.1.
[USEPA] United States Environmental Protection Agency. 2007. Aquatic life ambient freshwater quality criteria - copper, 2007 Revision. EPA-822-R-07-001. Office of Water, Office of Science and Technology, Washington, DC. February 2007.
USEPA. 2015. Toxicity Relationship Analysis Program (TRAP) Version 1.30a. Mid-Continent Ecology Division, 6201 Congdon Blvd., Duluth, MN 55804.
USEPA. 2018. Final Aquatic Life Criteria for Aluminum in Freshwater. Accessed 16/01/19
Wilson, R.W. 2012. Chapter 2: Aluminum. In: C.M Wood, A.P. Farrell and C.J. Brauner (Eds.). Homeostasis and toxicology of non-essential metals: Volume 31B. Elsevier Inc.
List of acronyms and abbreviations
Akaike information criterion
Agency for Toxic Substances and Disease Registry
biotic ligand model
Canadian Council of Ministers of the Environment
Canadian Environmental Protection Act
Chemicals Management Plan
Canadian Council of Resources and Environment Ministers
dissolved organic carbon
Environment and Climate Change Canada
federal environmental quality guideline
federal water quality guideline
Government of Canada
hazard concentration of the fifth percentile
lower confidence limit
lowest observed effect concentration
maximum acceptable toxicant concentration
multiple linear regression
Natural Resources Canada
species sensitivity distribution
toxicity modifying factor
toxicity relationship analysis program
upper confidence limit
United States Environmental Protection Agency
Appendix A. Aluminium toxicity dataset
Appendix B. Federal water quality guidelines calculator - aluminium
Report a problem or mistake on this page
- Date modified: