Recovery of Fish Communities in the Finniss River, Northern Australia

The Finniss River in the wet-dry tropics of northern Australia has received acid rock drainage (ARD) contaminants from the Rum Jungle uranium/copper m...
0 downloads 0 Views 211KB Size
Environ. Sci. Technol. 2001, 35, 2932-2941

Recovery of Fish Communities in the Finniss River, Northern Australia, Following Remediation of the Rum Jungle Uranium/Copper Mine Site ROSS A. JEFFREE,* JOHN R. TWINING, AND JEREMY THOMSON Environment Division, ANSTO, PMB 1, Menai, 2234, Australia

The Finniss River in the wet-dry tropics of northern Australia has received acid rock drainage (ARD) contaminants from the Rum Jungle uranium/copper mine site over more than four decades. Annual-cycle loads of Cu, Zn, Mn, and sulfate, calculated from daily water and flow measurements, have been determined both prior to and following mine-site remediation, that began in the early 1980s. The effects of varying contaminant loads on the relative abundances of seven fish species, sampled by enmeshing nets during dry seasons, were determined by nonmetric multidimensional scaling (nMDS), in combination with clusteranalysis and other nonparametric statistical techniques. These analyses showed that (i) prior to remediation, the impacted region of the Finniss River in 1974 had significantly dissimilar (P < 0.001) and more heterogeneous fish communities, generally characterized by reduced diversity and abundance, compared to sites unexposed to elevated contaminant water concentrations and (ii) postremediation, recovery in fish communities from the impacted region was indicated because they were not significantly dissimilar from those sampled at contemporary (P ) 0.16) unimpacted sites, that were also similar to preremedial unimpacted sites. Even though considerable contaminant loads are still being delivered to the impacted region of the Finniss River over the annual cycle, the recovery in fish diversity and abundances is consistent with (a) reductions of in situ contaminant water concentrations at the time of fish sampling, (b) reductions in annual-cycle contaminant loads of sulfate, Cu, Zn, and Mn by factors of 3-7, (c) greatly reduced frequencies of occurrence and magnitude of elevated contaminant water concentrations over the annual cycle, that was most pronounced for Cu, and (d) the absence of extensive fish-kills during the first-flushes of contaminants into the Finniss river proper at the beginning of the wet season, that were observed prior to remediation. As such, the results indicate that there has been ecological benefit to the Finniss River attributable to remedial works undertaken at the Rum Jungle mine site. Recovery in abundances of these fishes may also be due to their time-dependent evolution of tolerance to minewaste contaminants over their long period of exposure.

* Corresponding author phone: 61(2) 9 717 3584; fax: 61 (2) 9717 9260; e-mail: [email protected]. 2932

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 14, 2001

FIGURE 1. Map of the Finniss River in the vicinity of the mined area of Rum Jungle, showing the sites from which fish were sampled with enmeshing nets and the gauging station GS 8250097 (f) (modified from ref 3). Sites marked (O) are adjacent sites sampled in 1992 (2a, 2b, and 3a).

Introduction The Rum Jungle uranium/copper mine site (RJ) in tropical northern Australia has been a source of acid rock drainage (ARD) contaminants since the 1950s, which have had adverse impacts on the receiving waters of the Finniss River (FR). During the late 1960s/early 1970s unabated contaminant loads from RJ were quantified, and the geographical scale and intensity of detriment to the aquatic biota of the FR were investigated, particularly with respect to fish diversity and abundance (1-3). This investigation defined the impacted and unimpacted regions of the river. Mine site remediation began in 1982 followed by monitoring of water quality and flow, based on daily measurements within the FR system. These data have been used to determine both annual-cycle contaminant loads and frequency distributions of contaminant water concentrations and their changes following remediation (4). Investigations have also been undertaken to measure aspects of the postremedial ecological status of the FR system (5, 6). In this study we compare fish diversity and abundances in the impacted region of the FR, with sites unexposed to elevated water concentrations of contaminants, both before and after remediation. Our results show both the preremedial impact of unabated contaminants from RJ on fishes and their postremedial recovery. Their recovery is explained as primarily due to measured reductions in annual contaminant loads and, more particularly, the reduced occurrence of their elevated water concentrations in the FR, that diminishes the likelihood of fish kills at the beginning of the wet season.

Materials and Methods Study Site. Figure 1 shows RJ and the receiving waters of the upper reaches of the FR. Its ephemeral East Branch (EB) receives contaminant loads from RJ, calculated from daily water sampling and flow measurements at Gauging Station (GS) 8150097 on the EB, 5.6 km downstream RJ. This sampling station is reasonably assumed to measure all effluent from RJ that is received by the FR. Sampling, measurement, and 10.1021/es001719+ CCC: $20.00

 2001 American Chemical Society Published on Web 06/06/2001

analytical methods are described by Lawton (7). Since 19911992, real-time flow weighted composite samples have been used to generate the load estimates with the use of programmed data-loggers and auto-samplers. Remedial measures at RJ were initiated during 19811982 and aimed to reduce annual-cycle loads of Cu, Zn, and Mn by 70, 70, and 56%, respectively, and to modify the patterns of pollutant transport to the EB to reduce exposure of the biota to short-term elevated water concentrations (6, 8-10). Previous studies of the impacts of RJ on the FR, undertaken in the early 1970s, had quantified pollution inputs into the FR, identified the most probable factors affecting the pollutant input, and described the cyclical nature of the process (1, 11). In brief, pollution generation is dominated by ARD processes, mainly bacterially assisted sulfidic oxidation of metal-containing minerals. This is an ongoing process that is not overly influenced by seasonality. In contrast, contaminant transport is dominated by the climatic regime, particularly the highly seasonal rainfall patterns, characterised by the wet (summer, December to April) and the dry (winter and spring, May to December) seasons (3). With pollution generation being practically continuous, seasonality in the transport process leads to build-up of contaminants within or adjacent to RJ during periods of little or no flow through the system. Consequently, variable pollutant concentrations occur, with the higher contaminant loads being delivered to the FR particularly at the onset of rain (3). Mine waste contaminants build up in the bed of the EB over the dry season. This is a result of precipitation and evaporative concentration of solutes in the recessional flow from the mine-site, predominantly comprising seepage from the waste rock heaps. This pollutant load is resuspended by the first flush down the East Branch to be delivered to the FR at the beginning of the wet season. The observed critical toxicological events for the biota in the FR, particularly at the onset of the wet season, are also affected by the variability of rainfall/runoff between catchments due to the sporadic nature of storm events. This variability can lead to flow in the polluted EB being relatively high compared with that in the main river. Prior to remediation this circumstance was observed twice and in both cases was associated with extensive fish kills in the main river (2). These events become less common as the wet season progresses, and runoff from all areas of the catchment become more consistent in flow. The RJ uranium mining project was initiated in 1953 at a time when its environmental consequences received minor attention. Accordingly, no premining ecological studies were undertaken in the FR. The first ecological investigations in 1973 (2), prompted by previous anecdotal reports of these fish-kills, evaluated their likely importance in determining any patterns of detriment to the fish communities during the subsequent dry season, when flow from the EB was relatively diluted, and when there was vehicle access to lower reaches of the FR. Field investigations were undertaken in the FR during both the pre- and postremedial periods during the dry season and start of the wet season to (i) characterize water chemistry in billabongs (deeper regions of low flow in the main channel) during the periods of fish sampling and of first flushes of contaminants, (ii) observe the occurrence and nature of fish-kills and their geographic extent, and (iii) investigate pollution effects on fish diversity and abundance. Measurement of Fish Diversity and Abundance. Between the pre- and postremedial studies the following changes were made to taxonomic nomenclature for species reported in this study: (i) Fluvialosa erebi to Nematalosa erebi; (ii) Nematocentrus sp. to Melanotaenia splendida inornata; (iii) Hephaestus sp. to three teraponid species, viz., H. fuliginosus, Syncomistes butleri, and Syncomistes (?) n. sp.; (iv) Neosilurus

sp. to N. ater and N. hyrtlii; and (v) Madigania unicolor to Leiopotherapon unicolor. In the preremedial study (2) “yellow-tailed” and “black-tailed” forms of eel-tailed catfish were distinguished but pooled for analysis; these forms have subsequently been classified as two species that have again been pooled in this study for comparative purposes. Similarly, with Hephaestus sp. their great variation in form was originally noted; subsequently, three species have been identified but here are grouped as “black bream”. Generic names of individual fish species are mostly used in the following text (12). During preremedial investigations a range of sampling methodologies, including enmeshing nets, was used to determine the presence and abundances of fish species in the FR, at six major sites both upstream and downstream of the point of inflow of contaminants from RJ, via the EB (Figure 1). This region of the FR typically consists of a series of long pools of ca. 3 m in depth with flat bottoms and riparian vegetation consisting of shrubs, trees, and bamboo. Their physical characteristics are described in more detail in Jeffree and Williams (2). The relative amount of shoreline occupied by Pandanus was lowest, preremediation, just downstream of the junction with the EB (2, 3). This effect may have been caused by pollution loads from the EB, where severe detriment was caused to the riparian vegetation. Two sampling sites were located upstream of the EB junction (6 and 5) and four downstream (4, 3, 2, and 1). The site furthest downstream (1) was also below the confluence with another major perennial tributary, Florence Creek (Figure 1). Samplings were undertaken at three periods prior to remediation, during the dry season of 1974 (May-June, August-September, and November) following a pilot study in November-December 1973. Once during the dry seasons of 1992 (July-August) and 1995 (July-August) the setting of enmeshing nets was used to replicate a major component of the preremedial studies at the original sites or at those in close proximity, i.e. sites 2a, 2b, and 3a (Figure 1). Ten nets ranging in bar mesh size from 12 to 75 mm were set once at each of the six sites within a 3 week period, for the full night during 1974 (with clearing at midnight), to enhance the catch at those three sites immediately downstream of the junction with the EB, where the catch rate was very low. During the two postremedial samplings this group of nets, that was set at each of six sites within a three week period, was lifted at midnight because catches were adequate at all sites. All half-night catches were accordingly adjusted up to equivalent effort for a full night before their subsequent statistical analysis. The likely effect of this change in sampling approach was assessed at site 6 in 1992, where catch data for both morning and evening were determined for the majority of nets. These data showed that the seven species used for subsequent statistical analysis maintained very similar rankings of abundance between half and full night samplings. These taxa showed a 36% increase in collective abundance in the morning sampling, suggesting that our adjustment procedure (doubling) was a conservative underestimate of the postremedial abundances. A previous study on the taxonomic dietary composition of a small eleotrid, Mogurnda mogurnda, from shallow embayments in the FR (3), showed habitat parameters were significant (P < 0.01) predictors of the abundances of some foods. These included physical and biological habitat characteristics such as structural heterogeneity, coarseness of substrate and abundance of other fish species likely to alter the abundances of prey items through their consumption. In this study we have not attempted to correct for any potential effect of habitat parameters on the abundances of individual fish species/groups, prior to multivariate analysis, for the following reasons: (i) Information on some of these parameter types for the deeper pools sampled was simply VOL. 35, NO. 14, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2933

not observable by us, viz; structural complexity, substrate composition, quantitative abundance of fish-consuming crocodiles (Crocodylus porosus, C. johnstoni). Moreover, both pool size and bank-side vegetation may be adversely affected by the deposition of sand removed from the EB and preremedial pollution load, respectively, so any pollution impact could be a combined effect of water chemistry and structural habitat changes, as indicated for the EB. Most of the sites were sampled repeatedly during the pre- and postremedial studies or were in close proximity to the original sites. (ii) Observations of fish-kills during the preremedial period clearly identified adverse impacts on fish diversity and abundance in the FR immediately downstream of its junction with the EB. These appreciable impacts in this region were also apparent from simple plots of fish species versus distance downstream (2). Lack of assessment of the possible effects of habitat parameters may a priori reduce the capacity of the multivariate analyses to discern pollution effects, or their absence, on the fish communities exposed to elevated contaminant water concentrations. Physicochemical Measurements. These were measured at each site at the time of sampling. Preremediation measures included depth profiles of pH and dissolved oxygen (3). During the postremedial studies the following parameters were measured at three locations within each site, where nets were subsequently set: depth profiles of temperature, dissolved oxygen, pH, turbidity, and conductivity. The water chemistry results for samples taken during 1974 were reported in Jeffree and Williams (2). During postremedial studies duplicate surface water samples were taken at the upstream end of each site in acid-washed and river-rinsed 250 mL polyethylene bottles. The water samples were acidified with 2 mL of concentrated HCl, then sealed, frozen, and returned to the Australian Nuclear Science & Technology Organization (ANSTO) for analysis of total Ca, Mg, Cu, Zn, and Mn by ICP-AES and total Co and Ni by ICP-MS, as recommended in ANZECC (13). Statistical Methods. Preremedial data on the abundances of the seven most commonly occurring species caught by enmeshing nets were reported previously (2). These data and the abundances of these same species (or species groups), determined during the postremedial field sampling, formed the basis of our analyses. Because the preremedial data on numbers of each of seven fish species in each sample contained many zeros, parametric multivariate-normality could not be achieved. Therefore the data were analyzed with nonparametric multivariate analytical techniques in the PRIMER program (14). Nonmetric Multi-Dimensional Scaling (nMDS) was used to construct a two-dimensional map of the rank-order similarities among samples from sites, that were a priori designated impacted (sites 4, 3, and 2) or unimpacted (sites 6, 5, and 1), based on the findings of the preremedial study (2, 3). A Bray-Curtis similarity measure was chosen because of its ability to generally give a good representation of community structure. A fourth root transformation of the raw data was used to reduce any undue influence or distortion caused by species with large numbers, e.g. Nematalosa in postremedial samples. Cluster analysis (hierarchical dendrogram, average linkage) grouping was also performed and superimposed onto the two-dimensional nMDS plot to test for consistency between the two analytical methods. A one-way ANOSIM test (analysis of similarity) was applied to the rank similarity matrix. The test statistic R was used to reflect differences between samples from the impacted and unimpacted zones, in contrast to differences within zones. It was used in combination with a general randomization procedure (Monte Carlo test) to generate significance levels, so as to compare the following subsets of samples: (a) preremedial impacted versus unimpacted 2934

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 14, 2001

TABLE 1. Average Physicochemical Measures of Sampling Sites within the Finniss River in (a) 1992 and (b) 1995 site

dist,a km

depth, m

6 5 3a 2b 2a 1

-18 -0.1 3 13 15 30

1.7 1.4 2.1 1.4 2.8 3.6

6 5 4 3 2 1

-18 -0.1 1 3 15 30

2.2 1.5 1.7 2.4 2.3 3.9

turb, NTU

temp, °C

D.O. % sat.b

(a) 1992 7.1 0.11 6.9 0.11 6.7 0.15 7.5 0.15 6.9 0.14 5.2 0.01

4.2 3.0 3.3 3.7 3.0 0.9

21.4 18.9 20.8 24.2 24.0 23.5

50 20 22 66 62 59

(b) 1995 7.5 0.33 7.3 0.38 7.5 0.37 7.2 0.37 7.4 0.27 5.8 0.02

1.7 2.5 2.5 2.7 4.5 1.8

22.2 21.4 21.9 21.2 23.2 22.9

44 19 43 34 53 58

pH

cond, mS/cm

a Distance (dist) refers to kilometers downstream of the EB confluence. b Dissolved oxygen values consistently declined from the surface. Refer to the text.

sites and (b) postremedial impacted versus unimpacted sites. SIMPER (similarity percentages) analyses were performed to determine those fish species that were most important in determining the differences between specified groupings of samples.

Results and Discussion Physicochemistry. The 1973/1974 study, encompassing a wider range of sampling times across the dry season, reported pH values from 5.5 to 7.5, temperatures ranging from minima of 16 °C to maxima of 30 °C and dissolved oxygen at near saturation levels. These conditions were nearly uniform throughout the water column (3). In both 1992 and 1995 there were only moderate differences in physicochemical properties (except dissolved oxygen) through the water column at each site, thus their profileaveraged values are shown in Table 1. Sites were generally shallow, with maximum depths from ca. 2 to 4 m. Temperatures varied between 19° to 24 °C over both postremedial sampling periods, a range encompassed by the preremedial values. Turbidity was usually low, in the range 0.05) difference can be detected between them. These statistical comparisons between fish communities have shown a clear and interpretable pattern of pollution impact and recovery. The effect of habitat differences on the community structure of these fishes is suggested in Figure 5, by the clustering lower in the dendrogram (i.e. greater similarity) of samples taken from the same site, but at different periods, e.g. sites 2, 3, 4, and 5 during May and August 1974, site 1 during May 1974 and 1995, site 6 during 1992 and 1995sthe clustering indicating that site-specific characteristics determine similar fish community structure over time. However, there is still a clear separation of all samples from the impacted sites, prior to remediation, at lower levels of similarity (