1 Introduction

This study proposes a new method to calculate freshwater ecotoxicity fate (FF) and characterization factors (FF) for metals emitted to soil. This method accounts for metal speciation and uses an adapted version of USEtox. USEtox is a consensus-based life cycle impact assessment (LCIA) model developed within the UNEP-SETAC Life Cycle Initiative. It allows to calculate CF (i.e., the quantity of environmental impact per quantity of substance emitted) for human toxicity and ecotoxicity (Rosenbaum et al. 2008). According to Clearwater consensus (Diamond et al. 2010), metal ecotoxicity was not considered as appropriately modeled by USEtox. The model’s net sedimentation did not account for diffusive release from sediment to water. The particle and dissolved distribution coefficients were considered same regardless of the different water properties in freshwater systems (Diamond et al. 2010).

The inclusion of speciation to obtain characterization factors (CF) in order to improve the ecotoxicological impact assessment of metals in LCA was identified as one of the key priorities by a group of experts in Clearwater consensus (Diamond et al. 2010). The bioavailability factor (BF) was introduced in the definition of CF. It is the ratio of metals’ truly dissolved to total dissolved concentration. It was recommended by Diamond et al. (2010) to use a geochemical speciation model (WHAM6) to calculate the freshwater BF’s. In addition, WHAM6 was suggested to use in calculating the freshwater “distribution coefficients” (Kd) to replace the default Kd’s inside USEtox (Diamond et al. 2010).

Kd is also a critical parameter determining the fate of metals for terrestrial ecotoxicity in predicting metals’ mobility and (bio)availability in soils (Degryse et al. 2009). According to Owsianiak et al. (2013), metal distribution between the solid phase and solution phase (i.e., distribution/partition coefficient, Kd) and speciation in soil pore water control the availability of toxic forms of metals in both solid and dissolved phases (Owsianiak et al. 2013). Soil Kd is a ratio of metal bound in the solid phase (mg/kg) by total metal concentration (mg/L) in the dissolved phase (pore water) assuming that the dissolved metals are mobile and could possibly be taken up by adjacent roots and are important for various soil biological organisms (Sauvé et al. 2000).

Recent developments were made to consider metal speciation in freshwater and terrestrial ecotoxicity. Gandhi et al. (2010) have used WHAM6 speciation model to include metal speciation in aquatic environment to determine the freshwater ecotoxicological CFs for Cu, Ni, Co, Cd, and Pb for seven European freshwater archetypes. The resulting CFs using this method were 1 to 4 orders of magnitude lower than the CF’s which did not account for speciation (Diamond et al. 2010; Gandhi et al. 2010, 2011a, b). The results highlighted the fact that only certain portion of the total dissolved metal (truly dissolved) is responsible for the detrimental ecotoxicological effects.

Dong et al. (2014) generalized the same methodology and applied it to all the other metals available in WHAM for speciation calculation. The resulting site-specific CF’s were similar or slightly higher than the generic CF computed by USEtox for Cd, Co, Ni, Pb, and Zn, but 1–2 orders of magnitude higher for Cu. Interestingly, according to Dong et al. (2014) site chemistry did not affect much for Cd, Mn, Ni, and Zn results; their CF’s varied only 0.7–0.9 order of magnitude across archetypes. These metals have shown the highest CFs where freshwater had the lowest DOC concentrations, allowing to conclude that their affinity toward organic ligands were responsible in decreasing their bioavailability to the biota/environment (Dong et al. 2014).

To stay coherent with the calculations of freshwater ecotoxicity, BF was also introduced in the definition of terrestrial ecotoxicological CF by Plouffe et al. (2016). Plouffe et al. (2016) and Plouffe et al. (2015) validated and calculated CFs for terrestrial ecotoxicity using WHAM6 for all soil units of the world from the harmonized world soil database (HWSD). The generic USEtox CF results were compared with CF’s using soil labile Zn and soluble Zn concentrations. According to Plouffe et al. (2015), the soil labile Zn was chosen because it was the closest to the definition of “true solution Zn” as provided by Diamond et al. (2010) for freshwater. Plouffe et al. (2016) also selected soil soluble Zn for the determination of BF and Kd since this fraction included most of the Zn bioavailable fraction. In addition, the true solution (labile) Zn did not allow any robust validation of the WHAM model parameterization in soil due to lack of field data in the literature (Diamond et al. 2010; Plouffe et al. 2015). Using soil soluble Zn to calculate the bioavailable concentrations allowed a better validation of the WHAM speciation model as there were enough worldwide data in the literature to compare and validate with, provided that, it was less in agreement with the Clearwater consensus recommendations (Diamond et al. 2010; Plouffe et al. 2015). Resulting true solution CFs varied over 14 orders of magnitude all over the world and the global aggregated CF was 50 times lower than the generic USEtox CF. The global variability of the CF’s using soluble Zn was reduced to 1.76 orders of magnitude (Plouffe et al. 2015, 2016).

In a similar study, Owsianiak et al. (2013) used linear regression models instead of geochemical speciation model to assess the Cu and Ni speciation in soil and calculated terrestrial ecotoxicological CF’s. A new terminology, “accessibility factor (ACF)” was introduced for soil with the suggested Diamond et al. (2010) CF calculation framework. ACF was defined as the reactive fraction of total metal in soil (kg reactive/kg total). The calculated CF’s using this method showed global spatial variabilities of 3 and 3.5 orders of magnitude for Ni and Cu respectively (Owsianiak et al. 2013). In another study, Owsianiak et al. (2015) introduced aging models for metals in soils for Cd, Co, Cu, Ni, Pb, and Zn to improve the current LCI and LCIA methodology. A metal’s aging time in soil represents the time the metal has been in soil prior to the measurement of its reactivity (Owsianiak et al. 2015). This study also showed how metals’ reactivity (as defined in Owsianiak et al. (2013)) affected their CF’s. The study concluded that the emission source did not have significant or consistent effects on the reactivity of metals. The reactive fraction for metals for spiked soils aged more than 0.5–2.5 years and from the anthropogenic sources were similar, but remained statistically significant for organic-related Cu and airborne Zn. According to Owsianiak et al. (2015), the spiked soils did not undergo any aging process since the short terms did not allow the occurrence of long-term aging mechanisms like precipitation of salts or formation of non-reactive minerals to lower their reactivity. They acknowledged that, when considering the time scales of decades to centuries, the influence of metal aging on its reactivity for anthropogenic sources is difficult to capture and is statistically uncertain. The study, therefore, did not draw definitive conclusions about the influences of metal aging and emission sources on its reactivity (Owsianiak et al. 2015).

After the Clearwater consensus (Diamond et al. 2010), it was recommended to include metal speciation also in terms of EF (effect factor). To be reasonable with the calculation of BF and FF, the use of free metal ion activity concentration was suggested (Diamond et al. 2010). It was also recommended to use consistent parameter values (i.e., same TSS value in USEtox and WHAM) throughout different steps of the calculations (Diamond et al. 2010).

The free ion activity model (FIAM) and biotic ligand model (BLM) allow predicting the effect concentration of metal in aquatic environment accounting for speciation. BLM uses a mechanistic approach that is based on the hypothesis that the metal–biotic ligand interaction can be represented like any other chemical reaction of a metal species with an organic/inorganic ligand. The toxicity of metals to organisms is assumed to occur as the free metal ion is reacting with the organisms’ binding sites (Di Toro et al. 2001; Niyogi and Wood 2004). BLM calculates metal complexation with the biotic ligand and the dissolved organic carbon content (DOC) while also considering competition within the surrounding cations present in the water (Ca2+, Mg2+, K+, H+). Currently, acute and chronic BLM metal stability constants (LogKMe) are available for Cu (daphnia, fish), Ni (algae, daphnia, fish), Zn (daphnia, fish and algae), Cd, Pb, and Ag (De Schamphelaere et al. 2004; Cheng et al. 2005; Cheng and Allen 2006). Gandhi et al. (2010) applied BLM to calculate effect factors (EF) for Cu, Ni and Zn and showed that speciation of metals in aquatic environment lowered the toxicity by 5–50 times (Cu, Ni, Zn) depending on metal and selected water chemistry.

From all these studies (Gandhi et al. 2010; Gandhi 2011; Gandhi et al. 2011a, b; Owsianiak et al. 2013; Dong et al. 2014; Plouffe et al. 2015, 2016), it is evident that speciation plays a critical role in metal fate and characterization in natural aquatic and terrestrial ecosystems emphasizing that the consideration of total metal concentration should be avoided in fate and characterization calculations. These studies and their results highlight that there is a need to consider metal speciation both in soil and in water in ecotoxicological impact assessment. Hence, the calculations of metals’ soil and water toxicity should be based on regional properties for soil and freshwater.

In the current version of USEtox, the soil compartment is considered as a sink for the contaminants: the fraction of contaminant that reaches the deeper soil layers “disappears” and is never transferred to the surface water. This may be an appropriate assumption in most cases for organic chemicals, which might degrade before its resurgence to groundwater. However, this assumption represents a bias for metals, since they are not biodegradable and may travel from soil to groundwater through the deeper soil layers and ultimately reach to freshwater. The metal fate should therefore be properly addressed within USEtox.

Most of the aquifers and groundwater tables are interconnected with the freshwater bodies and the changes in their quality and quantity affect the freshwater (Fleckenstein et al. 2010). Water movement between groundwater and surface water is a major pathway for chemical transfer between terrestrial and aquatic systems (Winter 1998; Menció and Mas-Pla 2008, de Souza Machado et al. 2016). After precipitations, a fraction of the rainwater infiltrates through the land surface and moves vertically downward to the water table. The infiltration capacity of a soil determines whether and how much of the water can seep into the deeper soil layer. The transfer of contaminants in the soil depends on physicochemical properties of the soil and the substance. Generally, preferential (macropore) flow and sorption (matrix flow) are two main types of transfer processes in soil. The fraction of macropore flow in soils can be high reaching up to more than 90% of the total flow. It is often difficult to estimate the fractions of water that are subjected to macropore flow, interrupted macropore flow. According to Iqbal (1999), rain events and macropore channel geometry can contribute to changes in their already non-homogeneous flow pattern.

Adsorption, complexation, and precipitation also influence the speed of metal transport in soils. These processes depend on speciation of the metals and available surface sites (Hellweg et al. 2005). The preferential adsorption of heavy metals strongly depends on the metal concentration and the type of interaction with the solid surface. Some heavy metal cations like Cu2+ and Pb2+ may form complexes with dissolved organic acids in the soil. Whether the subsoil can retain these organic complexes depends on the content of organic matter. Dissolved organic complexes have been observed to directly reach the groundwater. In soils, where there was a higher percentage of calcite, (pH of the soil is neutral to basic), the anions were also observed to directly penetrate to groundwater. The mobility of organic metal complexes was found generally limited in soils (Hellweg et al. 2005). In case of higher pH value of soils, the specific adsorption of the metals increased as the mobility of metals was reduced. This is true for Cd2+, Ni2+, Cu2+, and Pb2+ since they form hydroxo complexes, which are adsorbed to sesqui-oxides. The presence of other heavy metals might influence the mobility of metals in soil. Hellweg et al. (2005) claimed that Cd2+ mobility could be enhanced by high concentration of Pb2+ (Hellweg et al. 2005).

By not considering the fate of metals through groundwater, the fate factors of metals to the soil are overestimated (since soil is considered the ultimate compartment where most of the metal emitted to soil “disappears”) and the fate factor of metal to the surface water is underestimated (as metals leaching from the soil compartment never potentially reaches surface water in the model).

The ground water moves slowly both vertically and laterally with a three-dimensional flow, which moves along flow paths of varying lengths from areas of recharge to areas of discharge. This movement is governed by established hydraulic principles. Flow through the aquifers can be expressed by hydraulic conductivity or Darcy’s law.

A watershed (analogous with “drainage basin” or “catchment area”) is defined as an area of land that drains down the precipitation until it reaches a specific water body (a river, a lake, or the ocean). Watershed boundaries are based on soil topography, watercourse, and stream locations. The watershed can be considered as the geographical resolution level at which any raindrop that falls on the soil will reach the same water body (and so do all the contaminants transported by this water flow if not degraded before).

Estimation of the groundwater behavior requires modeling of the interaction between all of the important processes in the hydrologic cycle, such as land cover, soil profile, infiltration, surface runoff, evapotranspiration, snowmelt, and variations in groundwater. The quantitative description of the hydrologic processes may become very complicated due to the high uncertainty and complexity in the underlying physical parameters (Jyrkama and Sykes 2007). However, in LCIA, toxic impact assessment is generally conducted using simplified steady-state models such as USEtox. One of the important principles of USEtox is to be parsimonious and to include only the most relevant environmental mechanisms. Hence, the integration of the transfer of contaminant through groundwater in LCIA should also be done parsimoniously in an adapted version of USEtox, only allowing to quantify the mass of contaminant transferred from soil to surface water at steady state through groundwater, without details about the pathway of the contaminant in the subsoil/groundwater compartment and about the hydrologic process kinetics.

The goal of the present study is to develop a modified version of USEtox to calculate regionalized freshwater ecotoxicity characterization factors (CF) for soil metal emissions. This modified version of USEtox established the missing link between soil, groundwater, and freshwater and considered speciation in all the passing environmental compartments (soil, subsoil and groundwater, freshwater). World’s watersheds have been chosen as native resolution scale for the results. CF’s were aggregated at the native resolution scale to regional, country, and continental levels. Their spatial variabilities were determined. The newly obtained fate and characterization factors were compared with the current methods.

We have applied the new approach in the case of zinc as a proof of concept. Zn was chosen for this study for several reasons. Despite of being a trace metal in earth’s crust, Zn can lead to toxic effects when it is found in the environment in excess. At lower pH, the solubility and toxicity of free Zn changes. Divalent Zn trace cations become more mobile and it is mostly responsible for phytotoxic effects after free Al and Mn (Kabata-Pendias 2010; Plouffe et al. 2015). Zn is used in industries as protective coating against corrosion, as a catalyst in various chemical processes (e.g., rubber, pigments, plastics, lubricants, and pesticides), in batteries, automotive equipment, etc. (Kabata-Pendias 2010). Thus, it is important to better evaluate its ecotoxicological impacts. Zn is also a major contributor to the average Canadian’s ecological footprint primarily in relation to terrestrial ecotoxicity (Lautier et al. 2010). In addition, according to Pizzol et al. (2011), the calculation for determining the ecotoxicological contribution of Zn is dependent on the chosen LCA methods (Pizzol et al. 2011; Plouffe 2015). Further study on Zn therefore becomes crucial in determining its actual contribution.

2 Methodology

Figure 1 summarizes all the methodological steps that are detailed in the next subsections.

Fig. 1
figure 1

General methodological steps followed for this study

(Step 1) Initially, Zn soil-water partitioning coefficients (Kd) in soil and subsoil were obtained using the WHAM7 speciation software for all different soil and subsoil units from the harmonized world soil database (HWSD v.1.2). The HWSD-database (2014) has listed the properties of topsoil (depth 0–30 cm) and subsoil (depth 30–100 cm). We have calculated partition coefficients for topsoil and subsoil following the same methodology as Plouffe et al. (2016). We made the assumption that the entire soluble Zn fraction is mobile and travels with water through the soil column and reaches the groundwater and ultimately the surface water. (Step 2) In the second step, the labile (true solution) fraction of Zn in freshwater is calculated using WHAM 7 as being the bioavailable fraction for freshwater ecosystems as recommended by the Clearwater consensus (and as already operationalized by Gandhi et al. (2010) and Dong et al. (2014)) using the freshwater properties around the world available from different databases (GEMStat, EuroGeoSurveys, etc.). This true solution fraction Zn was used to determine watershed specific suspended solid-water partitioning coefficient (Kpss), organic carbon-water partitioning coefficients (KDOC), bioavailability factors (BF), and effect factors (EF). EF’s were determined using biotic ligand model (BLM). (Step 3) The third step was to intersect world’s watershed (basins and subbasins) map with the soil mapping units from the HWSD database using geographical information system (GIS) to generate the native resolution geographical cells that had the soil and subsoil Kd’s from the soil unit (determined at the first step) and the freshwater Kpss, KDOC, BF, and EF from the receiving watersheds. (Step 4) In step four, a modified version of USEtox was created in which the subsoil and groundwater compartment was connected to the topsoil and to the surface water compartments. For each native resolution cell Zn fate and characterization factor was calculated considering that Zn had different properties in each cell (i.e., the soil and subsoil specific Kd’s in soil and subsoil and groundwater compartments and the receiving freshwater specific Kpss, KDOC, BF, and EF). (Step 5) The results obtained at the native resolution scale in step 5 were aggregated at more operational regionalized scales: watershed, country, continent, and world levels. The corresponding spatial variability due to aggregation at a coarser scale is determined. (Step 6) Ultimately, the results were compared with the current version of USEtox results.

2.1 General consideration

The UNEP-SETAC recommendation in Clearwater consensus (Diamond et al. 2010) was followed in calculating the characterization factors (PAF m3 day-kg−1) from soil to water (Eq. (1)).

$$ \mathrm{CF}=\mathrm{FF}\cdotp \mathrm{BF}\cdotp \mathrm{EF} $$
(1)

Here, FF stands for fate factors (days), BF stands for bioavailability factor, and EF stands for effect factor (PAF m3 kg−1).

2.2 Fate factor determination

Fate factors (days) were obtained with USEtox. The fate factor (FF) represents the residence time of a substance in an environmental compartment and depends on the properties of the substance and the properties of the emitting and receiving compartments.

To calculate FF for soil along with the metal’s molecular weight, substance-specific input parameters, and properties of the soil compartment, USEtox also needs partition coefficient values between soil particles and water (KPSI) (Plouffe et al. 2016). For water, two types of partition coefficients are required (Kpss-suspended solids-water partition coefficients and KDOC-dissolved organic carbon-water partition coefficients). We have calculated the water-specific and soil-specific partition coefficients following the methods by Gandhi et al. (2010) and Plouffe et al. (2016) to calculate soil to water fate factors using USEtox (see Section 2.3 for more information).

As defined in Gandhi et al. (2010), fate factor is the change in steady-state total dissolved amount of any contaminant/substance in an environmental compartment due incremental change in its emission.

FFi,s for freshwater compartment is defined as (Eq. (2)) (Gandhi et al. 2010):

$$ {\mathrm{FF}}_{\mathrm{i},\mathrm{s}}=\frac{{\Delta C}_{d,s}\cdotp V}{{\Delta m}_{i,s}} $$
(2)

Here, ΔCd,s is the incremental change in the steady-state concentration of the total dissolved substance s (kg/m3), V is the volume of freshwater compartment (m3), Δmi,s is the incremental change in the emission of total substance s (total dissolved and particulate phases) to compartment i (kg/day), and d refers to the total dissolved fraction of that substance (Gandhi et al. 2010).

2.3 Topsoil and subsoil regionalized partitioning coefficients (Kd) calculation

The WHAM6 model was recommended for metal speciation assessment in LCIA in the Clearwater consensus (Diamond et al. 2010) due to its sophisticated treatment of metal binding to humic and fulvic acids (common for soil and water) in particulate and total dissolved phases. Various researchers (Gandhi et al. 2010; Dong et al. 2014; Plouffe et al. 2016) have already used the WHAM model in LCIA to determine metal speciation both in soil and in water. We used the speciation model WHAM7 (Lofts 2012) since it is the most recent version of the WHAM model.

WHAM is an equilibrium-based model and it contains submodels that represent ion binding on humic substances (humic ion-binding model VI) (Tipping 1998) and mineral solids (SCAMP submodel) (Lofts and Tipping 1998). The latter submodel consists a surface complexation model for four types of surfaces: silica, iron, manganese, and aluminum oxides and a cation exchange model for clays (Lofts and Tipping 1998; Tipping 1998).WHAM can also estimate metal adsorption to Fe and Mn oxides.

For soils, the model was used with default parameters considering soil was oxic water containing high level of particulate phases (particulate oxides, silica, quartz, clay, and organic matter) following the method by Plouffe et al. (2015).

The speciation calculations were performed considering the major soil attributes available in the HWSD v1.2 database (pH, CEC, DOC, Na+, Ca2+, Mg2+, K+, SO42−, Cl) (HWSD-database 2014). Since metals are naturally occurring, we added background Zn concentrations to all soil samples based on the classification provided for five major types of soil texture groups in Kabata-Pendias (2010) (see ESM 1).

We made the assumption that the transportation of Zn in soil and in the aquifer occurs only in the aqueous phase and all soluble Zn species are mobile with water in the soil and the subsoil. Our calculated Kd’s for soils and subsoils are based on the soluble fraction of Zn as performed in Plouffe et al. (2016) and presented in Eq. (3).

$$ {K}_{\mathrm{d}}=\frac{{\left[\mathrm{Zn}\right]}_{\mathrm{soil}}}{{\left[\mathrm{Zn}\right]}_{\mathrm{d}\mathrm{issolved}}} $$
(3)

The soluble fraction comprises free ion, true solution complexes, and species bounded to colloidal phases (Lofts 2012), all the remaining Zn species being considered as the fraction of Zn in soil ([Zn]soil).

For groundwater, we made a simplifying assumption that the speciation in groundwater is same as in the subsoil. In reality, there are different types of groundwater and aquifer with varying depth, geologic structure, properties, and locations which will not be same as the subsoil compartment. If we consider only the shallow unconfined aquifers with impermeable or low-permeability bedrock layer underneath which serves more as a source to the freshwater than the sink, we can therefore make an approximation that they can be presented with the properties of the subsoil compartment, as found in Springs Creek basins in Fulton et al. (2005). We are aware that this is far from an accurate proxy, but the groundwater properties around the world are too poorly documented to have a better assumption of the subsurface environment.

2.4 Freshwater regionalized partitioning coefficients (Kd) and bioavailability factor calculation

The ratio between suspended particulate matter (SPM)-bound/particulate ([Zn]particulate) and truly dissolved metal concentration is expressed as Kpss (unit L/kg). Similarly, KDOC (unit L/kg) represents the ratio between dissolved organic carbon (DOC)-bound/colloidal ([Zn]colloidal) and truly dissolved metal concentration (Eqs. (4) and (5)).

$$ {K}_{\mathrm{pSS}}=\frac{{\left[\mathrm{Zn}\right]}_{\mathrm{particulate}}}{{\left[\mathrm{Zn}\right]}_{\mathrm{true}\ \mathrm{solution}}} $$
(4)
$$ {K}_{\mathrm{DOC}}=\frac{{\left[\mathrm{Zn}\right]}_{\mathrm{colloidal}}}{{\left[\mathrm{Zn}\right]}_{\mathrm{true}\ \mathrm{solution}}} $$
(5)

Bioavailability factors (BF) were calculated based on the recommended methodology in the Clearwater workshop (Diamond et al. 2010) and the method followed by Gandhi et al. (2010) as ratio of the true solution Zn to total Zn concentration. The true solution concentrations were calculated from WHAM7 results ([Zn]true solution) and comprise the free ion and the true solution complexes (Eq. (6)).

$$ \mathrm{BF}=\frac{{\left[\mathrm{Zn}\right]}_{\mathrm{true}\ \mathrm{solution}}}{{\left[\mathrm{Zn}\right]}_{\mathrm{total}}} $$
(6)

To be consistent with the already performed studies in soils and subsoils, we used the speciation model WHAM7 to determine the speciation in freshwater. We followed the methodology from Gandhi et al. (2010) considering the major water properties (pH, DOC, alkalinity/hardness/CaCO3, Na+, Mg2+, Al3+, Ca2+, K+, SO42−, Cl). Data was collected from different sources covering different regions of the world with different levels of completeness and geographical resolution, as there was not a single database that contained worldwide data. For Europe, the Forum of European Geological Surveys (FOREGS, now EuroGeoSurveys) database (BGS_NERC 2017) was used. For Canada, the Canadian aquatic data based on the ecozones was collected from Wiken (1996) and Gandhi et al. (2011a, b) were used. For the other regions of the world, we used available data from the GEMStat portal (GEMStatPortal 2017) database, which allowed covering Central and South America, the Russian Federation, and parts of Africa. Some data gaps were also completed by extrapolating data from neighboring watershed data (see details in the Electronic Supplementary Material, section ESM 1).

When available, we have used the listed dissolved Zn concentration from the databases and the literature for Zn speciation calculation. In some cases where dissolved Zn concentration was unavailable, a background value of 2 μg/L was assumed. This value is an average estimation of the dissolved Zn found worldwide in sites where no anthropogenic contaminations were detected (Shiller and Boyle 1985; ATSDR 2005; CCME 2016; BGS_NERC 2017; GEMStatPortal 2017). For more information on the background Zn concentration in freshwater, please refer to the Electronic Supplementary Material, section (ESM 1).

2.4.1 Calculation of the effect factors

Following the methodology from Gandhi et al. (2010), we used the BLM (Di Toro et al. 2001) model to calculate Zn effect factors (EF, in PAF-m3/kg) considering the free metal ion concentration was the contributor to the toxicity.

BLM parameters are available for daphnia. For fish and algae, we have extrapolated the BLM parameters (conditional stability constants-LogKMe) from De Schamphelaere et al. (2004) assuming that the LogKMe for cations and biotic ligand, mechanism of binding and modes of action were similar across the organism class (Gandhi et al. 2010). The resulting BLM models for daphnia, fish, and algae were used to determine watershed specific EC50’s for all the three species. These EC50’s were then used to calculate the regionalized EF’s using Eq. (7) (Haye et al. 2007; Rosenbaum et al. 2008).

$$ \mathrm{EF}=\frac{0.5}{\mathrm{HC}{50}_{\mathrm{EC}50}} $$
(7)

2.5 Intersection of the soil units of the HWSD and of the watersheds to define the “native resolution cells”

Depending on the publications and sources, the watershed geographical limits may vary slightly. In regionalized LCIA, some watershed maps are already used for other impact categories than toxicity/ecotoxicity (freshwater eutrophication, water use impacts) (Boulay et al. 2011; Helmes et al. 2012). The AWARE consensual model for water scarcity impact assessment in LCA used the watershed definition from the WaterGap project (Döll et al. 2003; Boulay et al. 2011). For coherence purpose across impact categories in LCIA, we decided to adopt the same map.

The drainage basins or watersheds map from WaterGAP (courtesy of Döll et al. (2003)) was generated from world’s 34 biggest river basins (downstream) and later subdivided into > 11,000 smaller subwatersheds. We have intersected the soil map with the (sub)watershed map. In some instances, the information on parent watersheds (which are comprised with more than one subwatershed) was taken from UN-Global (2016) database.

The HWSD-database (2014) soil database is composed of a GIS raster image file which links to a soil attribute database all over the world. The underlying map is in 30-arc-second grid cells, corresponding to 16,112 soil mapping units (SMU or MU_Global). They were converted as vector layers in QGIS for calculation.

The native resolution cells are obtained by intersecting the HWSD soil map and the WaterGAP watershed map using QGIS (Quantum GIS 2.18).

Each (sub)watershed was paired with the aquatic properties data (described in Section 2.4) of the world in another shape layer in QGIS. The variabilities of the documented properties across the samples available in the (sub)watershed were averaged since the reported data were spread for different seasons over few years (GEMStatPortal 2017).

The native resolution cells resulting from the intersection of the soil map, watershed map, and water monitoring stations are considered to have two types of information: soil properties from the soil map and the aquatic properties from the water monitoring stations from different databases.

2.6 Creation of a modified version of USEtox in which the groundwater compartment is connected to the soil and to the surface water compartments

In USEtox, intermittent rain is the primary carrier of chemical deposition from air onto soil. After that, the runoff event carries some of the deposited chemicals from the soil surface to the surface water. There is also a fraction of rain that is considered to be leaching from topsoil surface.

According to the model structure, half of net precipitation onto soils is evaporated, with the remaining half being split equally between surface water runoff (25%) and water infiltration (25%) through soil (see Eq. (8)); representation of transfer from soil to water in k-matrix) (Henderson et al. 2011). Considering this fraction of runoff from soil to water is reasonable, since it is being received by another compartment, by contrast, another 25% of the rain which is the fraction infiltrating (leaching) via soil is lost from the system (see Eq. (9)): how current USEtox calculated removal rate of chemical from soil compartments). This fraction has not been accounted as reaching to any other compartments.

$$ \mathrm{Transfer}\ \mathrm{soil}\ \mathrm{to}\ \mathrm{water}\ \left(\mathrm{USETox}\right)=\frac{\frac{\mathrm{Rain}\ \mathrm{rate}\times \mathrm{RunoffFraction}}{\mathrm{Dimensionless}\ \mathrm{soil}-\mathrm{water}\ \mathrm{part}.\mathrm{co}-\mathrm{eff}}+\mathrm{Erosion}}{{\mathrm{Depth}}_{\mathrm{soil}}} $$
(8)
$$ \mathrm{Removal}\ \mathrm{from}\ \mathrm{soil}\ \mathrm{compartment}\ \left(\mathrm{USETox}\right)=\frac{\mathrm{Rain}\ \mathrm{rate}\times \mathrm{Fraction}\ \mathrm{infiltrating}\ \mathrm{through}\ \mathrm{soil}}{\mathrm{Dimensionless}\ \mathrm{soil}-\mathrm{water}\ \mathrm{part}.\mathrm{co}-\mathrm{eff}\times {\mathrm{Depth}}_{\mathrm{soil}}} $$
(9)

To remedy this, we connected the “lost” portion of metal with the freshwater compartment via subsoil and groundwater (Fig. 2). We proposed Eq. (10) to be the new relationship between the soil-subsoil-freshwater link in obtaining the metal fate factors which considered the topsoil and subsoil speciation.

Fig. 2
figure 2

Modification of the fate model in USEtox

This modification was added with the current version of USEtox that linked the topsoil, subsoil, and freshwater compartment together. As described in Section 2.3, the subsoil compartment is considered as a proxy for groundwater compartment in the modified USEtox. The subsoil and groundwater compartment have the same surface area as the soil compartment in USEtox at the continental and global scale with a depth of 100 cm. This is the reported depth of subsoil in the HWSD database (HWSD-database 2014). Total volume of groundwater at the global scale was collected (10.7E15 m3) from Margat (2008). The porosity of the subsoil and groundwater compartment was set to 48% which was calculated from soil particle density and average bulk density found in HWSD database (see supporting info ESM 1) (HWSD-database 2014). With the consideration of the inclusion of the groundwater compartment, the transfer coefficients (Kd-subsoil) were added within the calculation (described in Section 2.3). We have calculated Kd’s based on the dissolved Zn fraction.

$$ \mathrm{Transfer}\ \mathrm{Soil}\ \mathrm{to}\ \mathrm{water}\left(\mathrm{modified}\ \mathrm{USETox}\right)=\left(\frac{\frac{\mathrm{Rainrate}\times \mathrm{Runoff}\ \mathrm{fraction}}{\mathrm{Dim}.\mathrm{soil}-\mathrm{water}\ \mathrm{part}.\mathrm{co}-\mathrm{eff}}}{{\mathrm{Depth}}_{\mathrm{soil}}}+\mathrm{Erosion}\right)+\frac{\left(\mathrm{Removal}\ \mathrm{from}\ \mathrm{soil}\ \mathrm{comp}.\left(\mathrm{leaching}\right)\times \mathrm{volume}\ \mathrm{of}\ \mathrm{water}\ \mathrm{in}\ \mathrm{the}\ \mathrm{groundwater}\ \mathrm{comp}\mathrm{artment}\right)}{\left(\mathrm{Volume}\ \mathrm{water}\ \mathrm{in}\ \mathrm{the}\ \mathrm{groundwater}\ \mathrm{comp}\mathrm{artment}+{\mathrm{K}}_{\mathrm{d}}\mathrm{subsoil}\times \mathrm{Volume}\ \mathrm{soil}\ \mathrm{in}\ \mathrm{the}\ \mathrm{groundwater}\ \mathrm{comp}\mathrm{artment}\right)} $$
(10)

In the matrix algebra within USEtox (Rosenbaum et al. 2007), the fate matrix is the result of the inversion of k-matrix. This matrix is also the removal rate matrix which consists of all the rate of removals from different compartments. The removal rates in k-matrix are calculated among other environmental mechanisms, based on the pollutants’ standard chemistry included in USEtox’s substance database, substance’s partitioning coefficients in air, soil, subsoil and water. The new proposed Eq. 10) which been obtained from mass balance in the subsoil and groundwater compartment was added in the matrix. This way the law of mass conservation was not violated as this was simply addition of a mass fraction to another recipient with the consideration that part of the metal was sorbed onto the soil surface and part of it were carried along with the water flow into the freshwater system.

2.7 Aggregation of the CF’s at the watershed, country, continental, and global levels

CF’s at the native resolution scale are hardly operational in LCA context; hence, there is a need to calculate aggregated CF’s at broader geographical scales. As the Zn emission to soil in the different native resolution scale cells is unknown, we considered as a proxy that this was proportional to the cell surface. In our calculation, it was assumed that each contributing soil cells were equally important and any emission in soil was uniformly distributed over the entire soil or watershed unit before passing to the next compartment. Hence, the characterization factor at the watershed level, country level, continental scale, and the global scale CFagg were calculated as:

$$ {\mathrm{CF}}_{\mathrm{agg}}=\frac{\sum_i{S}_i\times {\mathrm{CF}}_{\mathrm{native}\ i}}{\sum_i{S}_i,\mathrm{total}\ } $$
(11)

where CFnative i is the CF for each native resolution cells and ΣiSi,total is the total number of cells within the considered aggregation level (country, region, continent etc.). For each CFagg, the spatial variability within the broader geographical cell was documented using with their minimum, maximum, average, median value, frequency of occurrence, and percentiles.

2.8 Comparison of the CF’s obtained at step 5 with the default CF values for Zn from USEtox

The CF values obtained at the different geographical scales were compared with the USEtox generated generic value of soil to water CF’s for Zn emitted to soil in order to see the influence of the modifications done to the model and to determine if the speciation and the transfer through groundwater are influent enough to be worth being integrated within USEtox.

3 Results and discussions

3.1 Topsoil and subsoil regionalized transfer coefficients (Kd)

The obtained Kd’s for each soil cell of the HWSD can be found in the Electronic Supplementary Material section for topsoil and subsoil (ESM 2). Topsoil Kd’s ranged from 5.58E+01 L/kg to 9.29E+16 L/kg varying over 15 orders of magnitude with a median of 3.12E+04 L/kg. Subsoil Kd’s also showed a wide variability (16 orders of magnitude) from 4.64E+01 to 1.05E+17 L/kg with a median of 1.66E+04 L/kg. Both the topsoil and subsoil medians were far from their average values which points out to their outliers on both ends (topsoil average Kd = 1.36E+13 L/kg; subsoil Kd = 1.59E+13 L/kg) signifying wide variability in world soil characteristics. The lower Kd values were found for soils with high organic matter content with a moderate clay and quartz concentration. However, few extreme Kd values (on higher side) in Africa (larger than 1E+07 L/kg) were possibly due to the sandy soils in some specific soil units and therefore in watersheds (Orange and Vaal river watersheds). In those zones, contributing soils have more than 60% sand (Arenosols, Leptosols, and Cambisols) which exhibit high quartz content with very poor organic matter content and higher pH value (≈ 8.3) which led to higher partition coefficients (Hassan et al. 1996; HWSD-database 2014). According to Sauvé et al. (2003), sandy soils have less CEC (cationic exchange capacity) as opposed to clayey soils. CEC acts as buffer against soil acidification. These soil units have higher pH along with higher Kd’s, which can mean that acidification increased the available metal pool in the soil (Sauvé et al. 2003).

However, 90% of the Kd values are between 1.01E+03 and 9.7E+6 L/kg (5th and the 95th percentiles) which means we can accept the median value of Kd for topsoil and subsoil. Eighty-eight percent of the total number of soils for topsoil and 98% for the subsoil Kd values occur between 3rd and 4th orders of magnitude (Fig. 3). Similar range of results for soil Kd’s were also found in Plouffe et al. (2016). However, caution should be taken if using the generic value of Kd since it might be far enough to represent the world or a chosen specific soil unit or watershed.

Fig. 3
figure 3

Violin plot showing the variability of the soil-water partitioning coefficient in top soils (Kd), subsoil, and groundwater (Kd) compartments. The violin representing the probability density and the box plots representing the values with 95% confidence interval, median values, and the interquartile ranges

3.2 Freshwater regionalized Kpss, KDOC, BF, and EF

The regionalized freshwater Kpss at the native resolution scale varied over 7 orders of magnitude across the globe (1.96 to 2.05E+07 L/kg), with a median value of 1.34E+04 L/kg. The regionalized freshwater KDOC values varied over (1.02E+01 to 2.37E+06 L/kg) 5 orders of magnitude all over the world. The Zn Kpss and KDOC values found in Gandhi et al. (2010) and Dong et al. (2014) varied over 1–2 orders of magnitude for Europe only. Given that our results span over bigger geographical region (the world), the variability in the results are to be expected.

The regionalized BF varied over 2 orders of magnitude across the globe (3.10E−3 to 9.82E−1) with a median value of 3.85E−01 and an average value of 4.07E−01 (Fig. 4a). When we compared with the BF values found in Gandhi et al. (2010) (1.32E−1 to 9.79E−2) and Dong et al. (2014) (1.9E−1 to 7.4E−1), less variations are expected since their studies were limited for Europe. Our median and average values of BF are in the same order of magnitude for the entire world; however, it was not because there were little variations in worlds’ water properties rather lack of data available on freshwater properties.

Fig. 4
figure 4

Plots showing the variability and distribution in (a) bioavailability factors (BF) and  (b) effect factors (EF) for all water properties around the world

The regionalized EF varied over 3 orders of magnitude across the globe (8.35E+01 PAF m3/kg to 9.75E+04 PAF m3/kg), with a median value of 4.69E+03 PAF m3/kg and an average value of 5.79E+03 PAF m3/kg (Fig. 4b). Our calculated value is in the same order of magnitude with the USEtox generic EF value (2.84E+03 PAF m3/kg). Average ecotoxicity (HC50EC50) for Zn is 2.46E-04 kg/m3. For Europe, our effect factor varied over 1 order of magnitude (1.85E+02 to 9.82E+03 PAF m3/kg) as also found by Gandhi et al. (2010) and Dong et al. (2014) with a median value of 2.41E+03 PAF m3/kg.

3.3 Intersection of the soil units of the HWSD and of the watersheds to define the “native resolution cells”

The correspondence between the native resolution (subwatershed) and the HSWD soil units resulting from the overlap/intersection of the HWSD soil units and the watershed maps is available in the Electronic Supplementary Material in excel format (ESM 3).

This overlapping of the two maps resulted in 5327 native resolution cells containing both topsoil and subsoil units to perform necessary calculations.

3.4 Results from modified version of USEtox in which the groundwater compartment is connected to the soil and to the surface water compartments

The spatial variability in FFsw around the world ranged from 8.55 to 1.27E+02 days with a median value of 4.23E+01 days. The generic USEtox value for FFsw is 4.78E+01 days, which is amazingly close to our global median calculated by default USEtox. Although all the regions showed larger spatial variability in native resolution (watershed) Kd’s, which were calculated from topsoil and subsoil Kd’s (see Section 2.5 for more information on how it was calculated), it did not reflect on the soil to water metal fate factors (FFsw). They resulted in a variability of 2 orders of magnitude. According to Humbert et al. (2014), this is an acceptable range in the uncertainty in LCA results (Humbert et al. 2014). Interestingly, 80% of the all native resolution FFsw’s varied within the same order of magnitude (1.71E+01 to 8.2E+01 days) which means a single value from this range can represent global soil to water fate factor. Figure 5a) shows the distribution of the native resolution scale FFsw’s at the continental and global scales. Spatial variability in FFsw (min-max values) for Africa, Asia, Australia, Canada, Europe, Mexico and Central America, Russia, South America, and the USA are as follows: 2.85E+01 to 4.27E+01 days for Africa, 5.25E+01 to 7.88E+01 days for Asia, 1.88E+01 to 2.81E+01 days for Australia, 1.14E+01 to 1.33E+02 days for Canada, 3.06E+02 to 4.57E+05 days for Europe, 2.61E+01 to 3.91E+01 days for Mexico and Central America, 8.68 to 2.93E+01 days for the Russian Federation, 1.52E+01 to 7.83E+01 days for South America, and 8.21 to 1.28E+02 days for the USA.

Fig. 5
figure 5

Variabilities at the native resolution scale regionalized values of (a) the soil to water fate factors (FFsw) and (b) the freshwater ecotoxicity characterization factors (CFsw) for an emission to soil

The native resolution soil to water characterization factors (CFsw) varied over 3 orders of magnitude (2.44E+02 to 3.50E+05 PAF day m3/kg) at the global scale. Figure 5b shows the frequency and distribution of the average world CFsw with the most probable value for (sub) watershed CFsw being 9.55E+04 PAF day m3/kg. The USEtox default soil to water CFsw (1.36E+05 PAF day m3/kg) was slightly higher than our calculated value but it fell within the acceptable uncertainty range in LCA (Humbert et al. 2014).

When we compared between the continent/region specific aggregated CFsw’s, we observed greater spatial variabilities in regions where freshwater properties data were available (Europe, Canada, South America). European (with UK) CFsw results varied over 3 orders of magnitude (3.06E+02 to 4.45E+05 PAF day m3/kg) with the median value of 4.10E+04 PAF day m3/kg (Fig. 6).

Fig. 6
figure 6

Soil to water characterization factors (CFsw) for watersheds for different regions of the world

The CFsw results showed a spatial variability of 2 orders of magnitude (3.69E+03 to 1.18E+05 PAF day m3/kg) within Canada. We assumed that the freshwater properties data remained consistent over one ecoregion (for detailed information, see ESM 1, Wiken (1996), and Gandhi et al. (2010)). The results are presented for all ecoregions in Canada in the Electronic Supplementary Material (ESM 2).

Larger watersheds dominated the ranges of the characterization factors for South America. The Amazon River and the Parana River watersheds are the biggest watersheds in South America and cover almost 50% of the entire continent. For the Amazon River watershed, CFsw’s varied over 2 orders of magnitude (2.16E+4 to 2.84E+05 PAF day m3/kg), whereas for the Parana River watershed, the range of CFsw was on same order or magnitude (1.39E+04 to 7.16E+04 PAF day m3/kg). The CFsw for the entire continent stayed within the range of these two large watersheds’ CFsw values (1.39E+04 to 2.84E+05 PAF day m3/kg).

For the Russian Federation, freshwater properties were available for few major watersheds (data collected from the GEMStatPortal (2017) database). CFsw’s ranged over 1 order of magnitude (1.41E+03 to 1.66E+04 PAF day m3/kg). For Africa, Asia, and the USA, CFsw’s ranged from 3.77E+04 to 7.86E+04 PAF day m3/kg, 2.21E+05 to 3.32E+05 PAF day m3/kg, and 1.4E+04 to 2.09E+04 PAF day m3/kg respectively. CFsw’s stayed over the same order of magnitude for Australia.

3.5 Aggregation of the CF’s at the watershed, country, continental, and global levels

The values of the aggregated CFsw’s at the watershed and regional (country/continental/global) levels with their corresponding spatial variabilities are available in the Electronic Supplementary Material (ESM 2).

USEtox calculated value for soil to water characterization factor is 1.36E+05 PAF day m3/kg which is close to our calculated aggregated world CFsw value (9.18E+04 PAF day m3/kg) after the inclusion of groundwater compartment and metal speciation (Fig. 7).

Fig. 7
figure 7

Spread and frequency of the soil to water characterization factors (CFsw) with the aggregated CFsw value and the default USEtox value

4 Conclusions

The objective of this work was to obtain new regionalized CFs for freshwater ecotoxicity for Zn emitted to soils. This is an important step in regionalized assessment of metals in LCIA with the inclusion of watersheds as native resolution. We observed a large spatial variability the soil and water properties across different soils and watersheds of the world. Worldwide soil to water characterization factors (CFsw) for Zn varied over 2 orders of magnitude and the aggregated world CFsw is 1.5 times higher than the USEtox calculated value. It highlights the fact that a specific fraction of Zn is contributing to the higher characterization factor for freshwater.

The results also showed that the limited availability of freshwater properties data impacted in the FF and CF results. The regions where we had freshwater properties data from different water monitoring stations exhibited larger spatial variability in both fate and characterization factors. For example, for Europe, we observed a variability of 3 orders of magnitude in FFsw and CFsw’s. It emphasizes the fact that we cannot generalize or assign a single value (generic FF or CF) for fate or characterization factor over Europe.

If soil data are available, LCA users will be able to match the data from the HWSD database then define the watershed and can find the soil and water Kd’s and their corresponding FF’s and CF’s. If the soil data is unavailable but the location is known, it can be tracked with the watershed using the HWSD map and the information provided in ESM 2 regarding the relationship between soil units and watersheds and eventually the corresponding values of FF and CF would be obtained.

Since the inclusion of the missing soil-subsoil and groundwater-freshwater link with speciation yield to considerably higher values in LCA results, these values should be tested in case studies to determine how they are altering the LCA case study results.

Groundwater movement is unpredictable and invariably changing. Our assumption of saturated aquifer that is in equilibrium with subsoil should be explored further. Also, we made the assumption of a single porosity value that might change since this is a highly specific scenario. The groundwater position, soil characteristics, and aquifer’s type and structure all contribute to different porosities for the groundwater.

Soil to water fate and characterization factors and their spread around the globe can be applied in risk assessment. They can be categorized with an archetypical approach. The highest and the lowest emitters, potential risk of contaminating other compartment (based on wind flow, flux, rain etc.), etc., could be identified and remediation or preventative measured can be taken to avoid future metal contaminations.