Comparing impacts of metal contamination on macroinvertebrate and fish assemblages in a northern Japanese river

Study site

Field sampling of macroinvertebrates, fishes, and physicochemical characteristics was performed at nine sites in the Tokushibetsu River system on Hokkaido Island, northern Japan (Fig. 1) from 26 to 28 June 2018. Except for the fish survey, the methods adopted were described for a previous study in June 2017 by Iwasaki et al. (2020b). The basin area of the Tokushibetsu River system is approximately 300 km², and the predominant land uses are forest (88%) and agriculture (<10%; mainly meadow), with little urban use (<1%) according to Iwasaki et al. (2020a). The geology of the studied catchment is characterized mainly by igneous rocks, followed by sedimentary rocks (Geological Survey of Japan, 2015).

Five of the nine sites (sites S1a–S4) were in the Ofuntarumanai River, a metal-contaminated stream receiving treated mine discharge, and four reference sites (R1–R4) were in the main stream of the Tokushibetsu River. The reference sites, with similar physicochemical characteristics other than metal contamination (e.g., channel width), were established at similar elevations as the contaminated sites to avoid problems caused by natural longitudinal changes in community structure (Clements, 1994; Tokeshi, 2009; Iwasaki et al., 2012; Morita, Sahashi & Tsuboi, 2016). Study sites with the same numbers had similar elevation levels, for example S1 (a and b) and R1 (Table 1). These study sites were on third- or fourth-order rivers. Sites S1a and S1b were upstream and downstream of the inflow of treated mine discharge, respectively (Fig. 1). Because the treated discharge was not the sole source of metal contamination in the Ofuntarumanai River (Iwasaki et al., 2020b) and there was a contaminated tributary upstream of S1a (Takayasu et al., 1964), we regarded S1a as a contaminated site (see below for more details). Permits for field sampling in the river were obtained from the local municipal office and Hokkaido government.

Similar field sampling was performed at the same nine sites in the Tokushibetsu River system on 26 and 27 September 2018. This field sampling was carried out by the Japanese Ministry of Economy, Trade and Industry, and the results are publicly available (see Supplemental materials for more details). The field sampling method for benthic macroinvertebrates in this survey was different from that in the present study and no replicates were taken at seven of the nine sites; thus, it was difficult to simultaneously analyze the field data collected in both June and September 2018. Instead, we note the results of the additional field study in the Discussion section to evaluate our findings based on the field sampling conducted in June 2018.

Water-quality parameters

During field sampling, three water samples (50 ml) were filtered from each study site for dissolved metals analysis (0.45 µm pore-size) and refrigerated in the field. Ultrapure nitric acid was added to those water samples on the day of sampling so that the pH was less than 2. Concentrations of dissolved Cu, Zn, Cd, and Pb were measured by using an inductively coupled plasma mass spectrometer (Element XR, Thermo Fisher Scientific, Tokyo, Japan) according to method 200.8 of the U.S. Environmental Protection Agency (U.S. EPA, 1994). The limits of quantification were 0.001 µg/L for Cu, 0.06 µg/L for Zn, and 0.005 µg/L for both Cd and Pb.

Water temperature, dissolved oxygen, pH, and electrical conductivity were measured by using multi-parameter portable meters (Multi 3630IDS, Xylem Analytics Germany, Weilheim, Germany). Filtered water samples were also collected for measuring concentrations of dissolved organic carbon (DOC) and major ions (Na⁺, K⁺, Ca²⁺, Mg²⁺, Cl⁻, and SO₄²⁻). DOC was measured with a total organic carbon analyzer (TOC-L CPH, Shimadzu, Kyoto, Japan). Concentrations of major ions were measured with an ion chromatograph (Dionex ICS-1100/2100, Thermo Fisher Scientific). We calculated water hardness as 2. 497 × [Ca²⁺] + 4. 118 × [Mg²⁺].

As an index of contamination by multiple metals, we calculated the cumulative criterion unit (CCU; Clements et al., 2000) as the sum of the ratios of measured concentrations of four metals to the US EPA hardness-adjusted water-quality criteria for aquatic life (WQC; U.S. EPA, 2002) because the water quality standards for aquatic life are available only for Zn in Japan:

where m_i is the concentration of dissolved metal i and c_i is the corresponding WQC. Hardness-adjusted WQC for Cu, Zn, Cd, and Pb were calculated at a water hardness of 10 mg/L based on the observed range of water hardness in this study (Table 2) and a previous study of the same river (Iwasaki et al., 2020b). Note that, because the hardness of 10 mg/L is below the lower end of the hardness range of toxicity data used in the WQC development (20 mg/L; U.S. EPA, 2002), caution is required for the interpretation of the calculated CCU values. Also, we did not consider water quality variables other than water hardness (e.g., pH and DOC) in this calculation (Iwasaki et al., 2020b). This is because these variables varied little among study sites (Table 2), and US EPA WQC based on biotic ligand models that can consider the influence of water chemistry on metal toxicity were available only for Cu (U.S. EPA , 2007).

Physical parameters

Average channel width (surface-water width measured at run) and riffle width were measured at each study site. Riffle width was averaged if benthic macroinvertebrates were collected at multiple riffles within individual sites. The catchment area of each site was quantified using a digital elevation model (50-m grid; Geographical Survey Institute of Japan, http://www.gsi.go.jp/ENGLISH/index.html) and a geographic information system (ArcGIS 10.2 for Desktop, Esri Japan, Tokyo, Japan). Maximum water velocity and depth were evaluated on the basis of measurements at multiple places in riffles from which macroinvertebrates were collected at each study site. Current velocity was measured at 60% of water depth using an electromagnetic velocity meter (VR-301; Kenek, Tokyo, Japan).

Macroinvertebrates

At cobble-dominated riffles at each site, we collected macroinvertebrates from five randomly chosen stones (maximum diameter, 14–27 cm) using a Surber net (mesh size, 0.355 mm). Samples were preserved in the field in 99.5% ethanol and washed through a 0.5-mm sieve in the laboratory. Macroinvertebrates remaining on the sieve were preserved in 70% ethanol and identified generally to genus or species level. For each stone from which macroinvertebrates were collected, water depth and current velocity (at 60% depth) were measured above its upper surface before collecting macroinvertebrates. The relative surface area of each stone was estimated as the product of its maximum diameter and maximum boundary length. Although the mesh size of the sieve used in this study (0.5 mm) is commonly adopted in nationwide samplings of benthic macroinvertebrates in Japan, the smaller mesh size (i.e., 0.35 mm) can collect smaller individuals, which are more sensitive to metal contamination than larger ones (Cadmus et al., 2020). However, because our sampling was performed in late spring, when most insect individuals are expected to be at late larval stages and larger than 0.5 mm, the use of the smaller mesh size should not have materially changed our results.

We analyzed eight community metrics for abundance (the number of individuals per stone) and richness (the number of taxa per stone): total abundance, total taxon richness, and the abundance and richness of three major aquatic insect orders in the benthic samples collected: Ephemeroptera (mayflies), Trichoptera (caddisflies), and Diptera (true flies). We also determined the abundance of the dominant families (i.e., Ephemerellidae, Baetidae, Heptageniidae, Hydropsychidae, Chironomidae, and Simuliidae) of the three major aquatic groups, which were defined as those families that accounted for more than 5% of the total abundance at each sampled stone and that were collected at more than 30% of the sampled stones (i.e., more than 14 stones of a total of 45 stones collected). For all macroinvertebrate metrics, the means and standard errors (as indicators for the uncertainty in site mean) of five stones at each site were calculated and used for further analyses. Macroinvertebrate abundances were log₁₀-transformed (x + 1) before calculation of the site means to satisfy the assumptions of further analyses.

Fishes

At each site, we established five fish-sampling areas of approximately 5 m ×10 m to cover all of the habitats available (e.g., run, riffle, pool, and backwater) as much as possible. The distance between sampling areas was set to be >20 m. Fishes were collected from the downstream to the upstream end of each sampling area by using a backpack electrofishing unit (200–300 VDC; LR-20B, Smith-Root, Inc., Vancouver, WA, USA) and by throwing a cast-net. After one pass electrofishing, we used a cast-net four or five times within each sampling area to catch fishes in places where the pool was too deep for electrofishing to work. For the fish collection, we did not use block nets because it was impractical to effectively put them in place without disturbance. However, by expending a certain amount of effort, our sampling methods should be acceptable for comparing the abundances of fish communities. The captured fishes were anesthetized with phenoxyethanol and identified to species level if possible. The fork length was measured to the nearest one mm and body weight was measured to the nearest 0.1 g onsite.

A total of five fish species were collected: Oncorhynchus masou (masu salmon; Salmonidae), Salvelinus leucomaenis (white-spotted char; Salmonidae), Barbatula oreas (stone loach; Nemacheilidae), Lethenteron spp. (lamprey; Petromyzontidae), and Tribolodon spp. (Cyprinidae). We excluded Tribolodon spp. from the analyses because of their very limited abundance in our samples (only two individuals collected at R4) and determined the abundance (the number of individuals per sampling area) and condition factor of the other four species. The abundances of fishes were log₁₀-transformed (x + 1), and the means and standard errors of the five replicate samplings at each site were used for later analyses. Also, the condition factor (CF) was calculated as an indicator representing the health status of individual fish by using the following equation: (2)${\displaystyle \mathrm{CF}=\mathrm{body}\mathrm{weight}\left(\mathrm{g}\right)∕{\left[\mathrm{fork}\mathrm{length}\left(\mathrm{cm}\right)\right]}^{\text{3}}\times 1000.}$

The condition factor is relatively easy to measure in the field and is a sensitive measure for detecting population-level consequences of metal contamination (Munkittrick & Dixon, 1989a; Environment and Climate Change Canada, 2015). Condition factor data were pooled at individual sites and used in later analyses.

Approximately 128,000 individual hatchery-reared masu salmon fry (O. masou; mean fork length: 5.6 cm) were released at a location between S1b and S2 on the contaminated river on 6 June 2018. Masu salmon were also released at three other locations including a tributary between R1 and R2 in the Tokushibetsu River basin in April and June 2018. All released fry had thermally induced otolith marks (Volk, Schroder & Grimm, 1999). To estimate the proportion of wild (natural-origin) and hatchery fish at each site, we sampled and checked the otolith marks of 20–27 masu salmon captured from each site in the laboratory. We then tested whether the inclusion of hatchery fish affected the results of our analyses.

Data analysis

All statistical tests were performed using R version 3.6.1 (R Core Team, 2019). A significance level (α) of 0.05 was used. All the data used are available in the Supplementary File. In order to evaluate any effects at the five contaminated sites in the river receiving the mine discharge (i.e., S1a–S4), we first evaluated whether the site mean for each biological metric was within the 90% prediction interval calculated by fitting an intercept-only linear regression model to the reference site means. We refer to the 90% prediction intervals as “reference ranges” that are assumed as likely observed ranges of means at reference sites. We then examined whether there were statistically significant differences in biological metrics between each contaminated site and the corresponding reference site with a similar elevation (R1 vs. S1a, R1 vs. S1b, R2 vs. S2, R3 vs. S3, R4 vs. S4) by using a multiple comparison test (the single-step P-value adjustment; Bretz, Hothorn & Westfall, 2010) after analysis of variance.

We used the results of these two analyses to operationally interpret the findings in three ways. If the mean of a given biological metric at a contaminated site was lower or higher than the corresponding reference range and was significantly lower or higher than that of the corresponding reference site by the multiple comparison test, we report that as an “adverse effect”. If either one of these two results was observed, we report that as “some effect of concern” and if neither was observed, we conclude that there was “no effect of concern.”

Physicochemical parameters

Concentrations of the four trace metals (Cu, Zn, Cd, and Pb) at the contaminated sites (S1a–S4) were approximately 2 to 190 times higher than the concentrations at the corresponding reference sites at similar elevations, except for the concentration of Zn (25 µg/L) at reference site R1, which was similar to the concentrations at S1a and S1b (Table 2). Concentrations of the metals excluding Cu at many contaminated sites were higher than the values of the US EPA hardness-adjusted WQC for aquatic life, with higher concentrations and CCU values at the upstream sites. As previously observed (Iwasaki et al., 2020b), there was little difference in metal concentrations between the site just upstream (S1a) and just downstream (S1b) of the inflow of treated discharge. This is most likely due to the high concentrations of metals in an upstream tributary draining the mining area (Iwasaki et al., personal observations, 2019; this is beyond the scope of the present study). CCU values were greater than 1 at all of the contaminated sites except for S4, indicating potential ecological risks based solely on the concentrations of the trace metals measured.

There were marginally lower values of pH, DOC, and water hardness at the metal-contaminated sites compared with reference sites (Table 2), all of which generally increase the bioavailability of metals (Adams et al., 2020). The estimated catchment areas of the metal-contaminated sites were generally larger than those of the corresponding reference sites with similar elevations (particularly between S2 and R2, and S3 and R3; Table 1), but other physical parameters were similar at those sites.

Macroinvertebrates and fishes

All eight community metrics for macroinvertebrates at S3 and S4 were within the reference ranges and were not significantly different from those at the corresponding reference sites (Fig. 2), indicating that there were no effects of concern at those contaminated sites. On the other hand, there were adverse effects or some effects of concern for several of the community metrics at the upstream contaminated sites (S1a, S1b, and S2). For example, the mayfly abundance at S1b (58% lower than at R1) and the caddisfly abundance at S1b (83% lower that at R1) were lower than the reference ranges and significantly lower than at the corresponding reference sites.

As with the metrics for the macroinvertebrate community, there were no effects of concern for the abundances of any of the six dominant macroinvertebrate families at S3 and S4 (Fig. 3). Although the variations within individual sites (i.e., the 90% confidence intervals of site means) were relatively large, the abundance of heptageniid mayflies at S1b (68% lower than R1) and the abundance of hydropsychid caddisflies at S1a (84% lower than R1) were lower than the reference ranges and significantly lower than at the corresponding reference sites, indicating adverse effects. Furthermore, there were some effects of concern for the abundances of Simuliidae and Chironomidae at some of the upstream contaminated sites (S1a, S1b, and S2).

No adverse effects were detected for the abundances or condition factors of the four fish species sampled, except for the abundance of masu salmon at S3. Although there were some occasional effects of concern (e.g., the abundances of white-spotted char at S1a and S1b and B. oreas at S2 and S3; Fig. 4), the sites where significant differences were observed or the mean value was higher or lower than the reference range varied depending on species. Lamprey (Lethenteron spp.) were not collected at most of the contaminated sites (i.e., S1a–S3), but the numbers of lamprey collected were also limited at the reference sites (R2–R4; a total of 5–15 individuals per site) and their variations were relatively large (see Fig. 4). An adverse effect was detected for the abundance of masu salmon at S3, whereas there were no effects of concern for this metric at the other contaminated sites. The estimated proportions of released hatchery masu salmon at three of the reference sites (R1, R3, R4) and two of the contaminated sites (S1a, S1b) were 0%, whereas at R2, S2, S3, and S4 the proportions were 9% (2 of 23), 48% (13 of 27), 5% (1 of 21), and 18% (4 of 22), respectively. We estimated the abundances of wild masu salmon at each site using these proportions and reran the two analyses. The reanalysis did not change the conclusions on the effects of mine contamination on the abundance of masu salmon at contaminated sites.