Ireland has a relatively mild climate due to the influence of the Atlantic Ocean, with inland daytime temperatures generally ranging from 8°C in the winter to around 20°C in the summer (Met Éireann 2020). The prevailing wind direction is south-west, leading to greater annual rainfall in the west, which receives 1000–1400 mm, compared to 750–1000 mm in the east. In the mountainous regions of Ireland, rainfall tends to be greater, with some areas receiving 2000 mm or more per year (Met Éireann 2020).
During April–July 2017 and 2018, 31 upland acid-sensitive lakes in Ireland were sampled to determine the concentrations of THg in lake water, lake sediment, and catchment topsoil (Fig. 1). Irish uplands are generally undeveloped areas at elevations greater than 150 m above sea level (masl) along the coast margins, comprising habitats such as heaths, semi-natural grasslands, and bogs, and predominately used for rough grazing (Perrin et al. 2014). Soils within the upland catchments are shallow and tend to be highly organic with greater exposed rock in the southern catchments (Supplementary Information Fig. SI-1). The dominant soil type is peat or peaty podzols (Teagasc and EPA 2015), which is generally underlain by slowly weathering base-poor geologies, such as granite, quartzite, schist, and gneiss (Bowman 1991; Aherne et al. 2002).
Ireland is home to more than 12 000 lakes; in the upland areas these lakes tend to be small precipitation-fed headwaters. Upland lake catchments are relatively un-impacted by development or local pollution owing to the rough landscape. The 31 study lakes were selected from a set of lakes sampled in 1997 to assess the impacts of long-range transboundary air pollution (Aherne et al. 2002) and re-sampled in 2007 and 2008 (Burton and Aherne 2012; Scott and Aherne 2013); the dominant inputs of mercury pollution to the study lakes are from long-range atmospheric transport. The lakes in the 1997 study were selected using stratified random sampling, with greater weighting on high elevation lakes in acid sensitive regions (Aherne et al. 2002). In the present study, the lakes were selected to ensure spatial coverage of all sampled regions, with preference for higher elevation sites (median elevation 476 masl), to limit the effects of local disturbance (Nelson and Aherne 2020). The study lakes are small (median: 1.6 ha) in small catchments (median: 11 ha) and were assumed to be representative of the ‘average lake’, as 88% of all lakes in Ireland are less than five hectares (EPA 2006).
Lake water samples were collected from the shore, approximately 15 cm below the surface in an area free from emergent vegetation. The lakes are small, shallow, and well-mixed by wind; therefore, shore samples were assumed to be representative of the entire lake. Prior to collection, the sample container was rinsed three times with lake water, and rinse water was poured onto the shore to avoid disturbance of sediment. At each lake, two 30 mL water samples were collected in 40 mL glass I-Chem vials and acidified with 0.5% HCl for THg analysis. Precautions were taken to limit potential contamination during sampling; vials were individually double bagged, and gloves were worn when handling the vial and inner bag during sampling. Field blanks were used to test for contamination during sample collection or from the vials themselves. A 250 mL unfiltered bulk water sample was also collected for all remaining chemical tests (e.g., DOC, pH, conductivity).
Lake sediment was also sampled from shore, using a stainless-steel trowel. The top 5–10 cm of sediment was collected at three randomly selected locations along the lake edge, mixed, and stored in a 250 mL glass jar as a composite of the three locations. Sampling of sediment was attempted at all 31 sites, however rocky lake bottoms prevented sediment collection at three of the lakes.
Soil was collected at three randomly selected locations around the perimeter of the lake, within 10 m of the shore. Vegetation was removed, and the top 5–10 cm of soil was collected. A composite of the three locations was stored in a 250 mL glass jar, consistent with the sediment. At each of the three locations a bulk density core (radius 5 cm, height 5 cm) was also collected and stored in individual Ziploc bags.
Samples were kept cool and transported to Trent University for analysis. Prior to sampling, vials for THg (Glass I-Chem with Teflon-lined caps) were cleaned in 10% BrCl/HCl bath (v/v) for 48 hours, triple rinsed with reverse osmosis water, once with b-pure water, and oven dried (> 100°C). All other sample containers were soaked in a 10% HCl bath for 24 hours, triple rinsed with reverse osmosis water, once with b-pure water, and then air dried.
The water samples were analyzed for THg on a Tekran 2600 mercury analyzer, using US EPA method 1631. The samples for DOC were filtered using 0.45 µm disposable nylon syringe filters and measured on a Shimadzu TOC Analyzer. Conductivity and pH were determined on unfiltered samples using a Mantech PC-titrate.
The composite sediment and topsoil samples were oven dried for 72 hours at 40–50°C (to avoid volatilization of mercury). The dried samples were ground and sieved (< 2 mm). The fine material was used to determine THg, pH, percent organic matter (%OM), and particle size analysis (PSA); THg was determined using a DMA 80 (Direct Mercury Analyzer). Percent organic matter was calculated by loss-on-ignition (LOI) using a muffle furnace at 400°C for 10 hours. The residual ignited sample was used for PSA on a Horiba Particle Size Analyzer. Prior to PSA, samples were soaked in Calgon solution (30 g/L) for 12 hours to promote separation of particles during analysis. Bulk density samples were oven dried at 105°C for 24 hours and the dry mass was used for bulk density calculations.
In this study mercury was quantified as THg concentration (ng/L for water, ng/g for sediment and soil), THg normalized by organic carbon (THg/OC mg/kg), and THg pool (µg/m2) in the catchment soil (see Supplementary Information A for calculations). The use of THg/OC allowed comparison of THg concentrations between the three environmental media (water, soil, and sediment), and the examination of interlinkages that were hidden by organic carbon. Total Hg and THg/OC were used as dependent variables in statistical tests.
The number of times each lake was sampled during April–July 2017 and 2018 varied; most lakes were sampled once or twice, though a small number had additional observations. To allow for comparison between lakes the data were averaged to one value per variable per lake to represent the 2017–2018 data. All variables were tested for normality using the Shapiro-Wilk test (p < 0.05); due to the non-normal distribution of the data, non-parametric tests were used. All statistical tests were performed using the statistical software PAST version 4.03 (Hammer et al. 2001).
Variability between lakes was calculated as normalized median absolute deviation (NMAD) to best represent the non-normally distributed data (see Supplementary Information A). To assess regional variation across the north–south and east–west gradients, all variables were tested for significant Spearman Rank Correlations (p < 0.05) with easting and northing as an indicator of spatial autocorrelation. Significant differences (p < 0.05) between THg/OC in water, sediment, and soil, and between THg and organic matter in sediment and soil were assessed using Wilcoxon signed rank tests.
The associations between water, and sediment and soil were explored using Spearman rank correlations (p < 0.05). Spurious correlations were excluded from this analysis, e.g., for THg/OC in lake water, correlation with THg and DOC were omitted; similarly, for sediment and soil, associations between THg/OC, %OM, and THg, or between THg in soil and THg soil pool, were excluded.
The influence of catchment soil and lake sediment on THg and THg/OC in lake water was assessed using Redundancy Analysis (RDA). Further, a second RDA was run with sediment THg and THg/OC as the dependent variables to determine the influence of soil on sediment. The biplots for RDA were scaled to allow easier visual interpretation, using scaling type 2 as determined by Legendre and Legendre (1998). The correlation strength of the individual axes (THg and THg/OC) with their dominant drivers is indicated as an r value within-text. The strength of the overall RDA (both THg and THg/OC axes) is noted in the RDA biplot figure caption as R2 and R2adj.