Uptake of selenium and productivity in waterfowl in the Elk River Valley, British Columbia

Transcription

1 Uptake of selenium and productivity in waterfowl in the Elk River Valley, British Columbia Final Report November 19, Sumpter Dr., Coquitlam, B.C., Canada V3J 6Y3 ph: / fax: /

2 November 19, 2004 WATERFOWL i EXECUTIVE SUMMARY A previous study of the effects of selenium on waterbirds in lotic (rapidly flowing) streams of the Elk River Valley in 2002 showed that: uptake of aqueous selenium from the water to invertebrate prey of spotted sandpipers and American dippers was less than expected, based on uptake models in the scientific literature, uptake of selenium was occurring in spotted sandpipers from selenium-exposed areas, but not in American dippers, hatchability of spotted sandpiper eggs was depressed in the areas where increased uptake was occurring. The authors of that study recommended that uptake of selenium be studied in birds of lentic habitats (such as marshes and bogs with little flow), where biotransformation processes are expected to be more active. This study of waterfowl (ducks and geese) was in response to that recommendation, and complemented a concurrent study of uptake and effects in redwinged blackbirds. Preliminary sampling of waterfowl eggs in 2003 showed that waterfowl nests in these upland marsh and river oxbow habitats are extremely difficult to find, but that among the 3 eggs that were found, 2 (both from selenium-exposed areas) had high selenium concentrations. In 2004, more intensive effort was made over a longer period of waterfowl courtship and nesting. Objectives were to: 1. Obtain a representative sample of waterfowl eggs for selenium analysis from nests in selenium-exposed and reference area marshes in the study area. 2. Observe 2 measures of productivity: clutch size (the number of eggs laid), and brood size. Brood sizes were based on counts of all broods seen in each wetland. 3. Compare selenium concentrations, measures of productivity (clutch size and brood size), and prevalence of teratogenic effects in reference and seleniumexposed areas. 4. Determine whether there was any association of selenium concentrations with the above measures of productivity or with prevalence of teratogenic effects. In 2004, we collected 26 waterfowl eggs from 19 active nests (7 were salvaged after the nests were predated) and counted 50 broods of 8 species. Selenium concentrations in waterfowl eggs ranged from 1.32 mg/kg dry weight in a Canada goose at the lower Elk River Oxbow to 34.7 mg/kg dry weight in an American coot at Clode Pond. Overall, there was no statistically significant relationship between selenium concentrations in eggs and in the water of the marshes where the eggs were collected. In Goddard Marsh, however, the aqueous selenium was very high (78 µg/l) and the egg selenium concentration was moderate (4.4 mg/kg dry wt.). With this outlier removed, a weak, positive relationship emerged of increasing selenium in eggs with selenium in water. Only Canada goose and mallard eggs were collected at > 2 sites. These 2 species showed a pattern consistent with the above relationship: lower selenium concentrations at reference

3 November 19, 2004 WATERFOWL ii sites and higher at exposed sites. In mallards, these between-site differences in selenium concentrations were not statistically significant owing to the small sample sizes, despite a considerable spread from the low at the lower Elk River Oxbow (3.23 mg/kg dry wt.) to the high at Otto Creek (15.7 mg/kg dry wt). In Canada geese, between-group differences were statistically significant, with eggs from Clode Pond (17.1 mg/kg dry wt.) and Goddard Marsh (4.4 mg/kg dry wt.) having much higher egg selenium than those from Otto Creek and the upper and lower Elk River Oxbows (2.26, 1.69 and 1.38 mg/kg dry wt., respectively). When all species were combined and grouped by selenium exposure categories (high, low and reference areas), the selenium concentrations were significantly different between the reference and high exposure sites. This confirmed that significant selenium uptake was occurring in the more highly exposed sites. Based on published selenium toxicity thresholds of 12 to 15 mg/kg dry weight, the eggs and young of Canada geese (17.1 mg/kg Se dry wt., N=1), and American coots (29.6 mg/kg Se dry wt., N=2) at Clode Pond, mallards at the Fording River Oxbow (12.0 mg/kg Se dry wt., N=2) and mallards at Otto Creek (15.7 mg/kg Se dry wt., N = 1) could be at risk to selenium toxicosis. Only Canada geese were represented at enough sites to compare clutch sizes (number of eggs laid) by species. Mean Canada goose clutch sizes were nearly identical at all 7 sites. To compare clutch size of all waterfowl species, which normally have different clutch sizes, the counts were normalized to express deviation from the provincial average. There was no difference of normalized clutch size among exposure categories. Clutch sizes for all species were within the normal range for British Columbia. This showed that, within the limitations of the modest sample sizes, selenium had no discernable effect on the number of eggs laid. There were also no significant differences of normalized brood sizes (number of young that hatched and survived long enough to be counted by observers) among exposure categories. As with clutch sizes, brood sizes of all species were within the normal range for British Columbia. However, there was more variation in brood sizes in the high exposure sites than in the low exposure and references, suggesting that significantly greater sample sizes would be required to detect effects of selenium on brood size. Of 34 broods that we were able to monitor for > 7 days, 30 had 100% survival after an average of 20 days (range = 7 to 41 days). Those that had less than 100% survival were a brood of 4 coots at Graves Marsh that had 1 surviving after 18 days; a Barrow's goldeneye brood of 12 at the Fording River Oxbow that had 10 surviving after 9 days; a brood of 4 buffleheads at the Fording River Wetland with 3 surviving after 24 days; and a brood of 5 Canada geese at the Elk River Oxbow-Lower that had 4 surviving after 33 days. Among the longest surviving broods were 4 goslings out of a brood of 5 at the Elk River Oxbow-Lower that survived at least 33 days; a brood of 7 hooded mesrgansers, also at the Elk River Oxbow-Lower, that survived at least 41 days; and a brood of 4 goslings at Clode Pond that survived at least 31 days. These results show that if selenium is affecting overall brood size or brood survival, it is at such a low level as to be undetectable with a sample size of 50 broods. However, this generalization may not apply to the 3 high exposure sites (Clode Pond, Line Creek Marsh and Goddard Marsh) which were poorly represented in the brood counts. No unusual mortality was observed except for 4 dead adult ducks at Clode Pond. While suspicious in view of the high selenium concentrations in the water, there are many natural and

4 November 19, 2004 WATERFOWL iii unnatural hazards that could account for them. Moreover, adult ducks can survive with far higher selenium concentrations that those that affect reproduction. It is surprising, however, that no ducks nested at Clode Pond, despite the regular occurrence and courtship throughout the nesting period, of mallards, buffleheads, and Barrow's goldeneyes, with occasional pairs of ring-necked ducks, American widgeons, cinnamon teal, lesser scaups, and common mergansers. Two pairs of coots did nest in Clode Pond, one of which hatched out 7 coot chicks, (the other nest was destroyed by a predator after a storm exposed it to view) but the chicks were never seen despite intensive searching. The apparent absence of the coot chicks, too, is suspicious, but neither the 4 dead ducks nor the apparently missing coot chicks are evidence of contaminant effects. This study resulted in the following conclusions: 1. Waterfowl eggs contained higher concentrations of selenium at the more highly contaminated sites than at the reference sites. The selenium concentrations in the eggs of 3 species (mallards, American coots and Canada geese) at 3 sites (Otto Creek, Fording River Oxbow, and Clode Pond) were above reported toxic thresholds. 2. Goddard Marsh seems to be an outlier in terms of the modest selenium concentration in a single goose egg, relative to concentrations in the water. It may simply be because the parents fed in uncontaminated or less contaminated locations prior to egg-laying, but it may alternatively relate to the unique physico-chemical properties of the marsh. 3. There were no overall differences in clutch size or brood size at selenium-exposed sites compared to reference sites. Clutch and brood sizes were within or above the normal ranges for the province at all sites. Among individual broods monitored, survival was as high in the exposed sites as in reference sites; however, few broods were counted at the most highly contaminated sites (1 at Line Creek Marsh and 2 at Clode Pond and none at Goddard Marsh). 4. There was no direct evidence of selenium toxicosis in waterfowl at any of these study sites. The apparent absence of coot chicks that hatched at Clode Pond, as well as the anecdotal observations of 4 dead ducks could be due to selenium toxicosis, but could also be due to other factors such as predation or disease. 5. The limitations of this study were in the modest sample sizes (19 clutches and 50 broods counted) and in our inability to definitely link the individual broods observed to specific nests from which eggs were counted and analysed for selenium concentration. These limitations, which were anticipated in the approved study design, resulted from the low numbers of waterfowl nesting at each marsh, the difficulty in finding their nests, and the mobility of broods immediately after hatching. Nevertheless, sample sizes were adequate to determine that significant uptake is occurring at the more highly contaminated sites, and that, overall, productivity is within provincial norms for each species. Recommendations included continued monitoring of waterfowl productivity, better linking of broods to nests (such as by radio-tagging hens) and assessment of the physicochemical properties of Goddard Marsh, Clode Pond, and Line Creek Marsh that may influence conversion of selenium to biologically available forms and consequent uptake through the food chain.

5 November 19, 2004 WATERFOWL iv DISCLAIMER This report is the property of the Elk Valley Mines Environmental Management Committee, comprised of Elk Valley Coal Corporation s 5 operating mines (Greenhills Operations, Coal Mountain Operations, Line Creek Operations, Elkview Operations, and Fording River Operations). It may be provided in confidence to government agencies on the Elk Valley Selenium Task Force. The report contains proprietary and business information and may not be shared with people or agencies other than those listed above. It is a work in progress and disclosure of the information contained in the report, beyond the Task Force, could potentially harm the owners of the information. Canada goose

6 November 19, 2004 WATERFOWL v TABLE OF CONTENTS Executive Summary i Table of Contents v List of Tables vi List of Figures vi Introduction 1 Introduction 1 Methods 3 Study Sites... 3 Permits... 7 Sample Collections and Brood Observations... 7 Chemical Analysis... 8 Statistical Analysis... 9 Results 10 Weather Selenium Concentrations Productivity Survival Mortality Discussion 21 Conclusions 24 Recommendations 25 References 26

7 November 19, 2004 WATERFOWL vi LIST OF TABLES Table 1. Study Sites Table 2. Bonferonni comparison of log-transformed Se (dry wt.) among high and low exposure areas and reference areas Table 3. Selenium concentrations (mg/kg) in waterfowl feathers, Table 4. Results of analysis by ICP/MS for a suite of elements (total, mg/kg dry wt.) in waterfowl eggs, Table 5. Waterfowl egg statistics...33 Table 6. Waterfowl clutch and brood statistics LIST OF FIGURES Figure 1. Exposed and reference area sampling sites... 1 Figure 2. Selenium concentrations (µg/l) in water from exposed and reference areas (Lower Elk River is at EVS Stn. 11 below Sparwood)... 7 Figure 3. Minimum temperatures at Sparwood, May July 2004, from the Canadian Climate Data Archive. Values missing from the archive are blank. Horizontal lines are long-term normals Figure 4. Selenium concentrations (mg/kg) in all species by site. Error bars are given for N > 1 samples Figure 5. Selenium in waterfowl eggs and water (Goddard Marsh omitted) Figure 6. Mean selenium (mg/kg) concentrations in mallard eggs Figure 7. Mean selenium concentrations (mg/kg) in Canada goose eggs Figure 8. Selenium concentrations (mg/kg) in all species by exposure category. Error bars are given for N > 1 samples Figure 9. Mean ± 2SE (Standard Error) clutch size, all taxa by site. Error bars are given for N > 1 samples Figure 10. Clutch size by exposure category (mean deviation from average ± 2SE (Standard Error)) Figure 11. Mean ± 2SE (Standard Error) brood size, all taxa by site. Error bars are given for N>1 samples Figure 12. Brood counts by exposure category (mean deviation from average ± 2SE (Standard Error)

8 November 19, 2004 WATERFOWL 1 INTRODUCTION Elevated concentrations of selenium have recently been found in water, sediments, and aquatic biota in portions of the Elk River and some of its tributaries, particularly those downstream of several coal mines (Kennedy et al. 2000; McDonald and Strosher 1998, 2000). Selenium can be toxic to birds, affecting both developing embryos and newly hatched young. Selenium concentrations as high as 107 ± SD 14 µg/l have been recorded in tributary streams below active mining areas (EVS Environment Consultants 2003b). In the Elk River itself above the coal mines, and in tributaries above mining areas (Figure 1), selenium concentrations are generally < 1.0 µg/l (rarely up to about 5.0 µg/l in creeks that pass through coalbearing strata), while those below the coal mines have been as high as 7.4 µg/l (EVS Environment Consultants 2003a; Kennedy et al. 2000; McDonald and Strosher 1998, 2000). Waterborne selenium concentrations below 1.0 µg/l have been reported to pose no risk to waterbirds, while those above about 5.0 µg/l pose a risk to birds (Adams et al. 1998; Lemly 1993, 1996). The British Columbia water quality guideline to protect aquatic life is 2.0 µg/l (Nagpal and Howell 2001) and the Canadian "no effect" guideline is 1.0 µg/l (Environment Canada/Health Figure 1. Exposed and reference area sampling sites. Canada 1995). Background water concentrations from reference areas in selenium exposure and ecotoxicology studies have ranged from 0.1 to < 3.0 µg/l (Adams et al. 1998; Casey and Siwik 2000; Ohlendorf et al. 1990; Seiler 2003; U.S. Department of the Interior 1998).

9 November 19, 2004 WATERFOWL 2 A government-industry task force, the Elk Valley Selenium Task Force, is directing studies to determine whether adverse effects of selenium may be occurring in areas surrounding mine activity. During , the Elk Valley Mines Environmental Management Committee (Greenhills Operations, Coal Mountain Operations, Fording River Operations, Elkview Operations, and Line Creek Operations), retained SciWrite Environmental Sciences Ltd. to determine if selenium was affecting waterbirds in lotic (flowing streams) environments. The study area included portions of the Elk and Fording Rivers and their tributaries, where elevated levels of selenium have been measured downstream from coal mines, and reference areas representing the natural background of the region. Preliminary analysis of data in water and benthic macroinvertebrates in selenium-exposed areas were high enough, based on the scientific literature, that some effects on waterbirds might have been expected ( 2001). That study complemented others of fish and invertebrates in lentic environments (EVS Environment Consultants 2000, 2002; Minnow Environmental Inc 2002). The lotic system avian study results (Harding and Paton 2003) showed that, although selenium concentrations in spotted sandpiper eggs were higher in exposed areas than in reference areas, they were lower in both American dippers and spotted sandpipers than would have been expected, based on the concentrations in the water, sediments, and benthic macroinvertebrates. There were no effects on overall productivity of either American dippers or spotted sandpipers, although hatchability was slightly depressed in spotted sandpipers in the seleniumexposed areas (Harding et al. 2004). There were no teratogenic effects. The authors speculated that the Clode Pond lotic environment short residence time, low sediment volumes, few rooted plants, and undeveloped detritus-based community was the most likely explanation for the low bioaccumulation of selenium from the water phase through the macroinvertebrates to the waterbirds, and for the lack of severe effects on the waterbirds. They recommended further studies of selenium uptake in marsh birds of lentic systems within the study area. The present study of waterfowl in lentic systems was initiated in response to that recommendation, and complements a study of red-winged blackbirds undertaken at the same time. Preliminary sampling was undertaken at exposed and reference sites (Figure 1) during May-June 2003 to determine whether uptake was occurring in waterfowl and to develop methods of finding nests and making productivity observations (Harding 2003). This study reports the results from 2003 and 2004.

10 November 19, 2004 WATERFOWL 3 METHODS Study Sites We sampled from 6 selenium-exposed areas (i.e., exposed to high selenium from mine runoff, relative to background) (Table 1) and 3 reference areas. Exposed areas were Clode Pond (CP), Goddard Marsh (GM), Line Creek Marsh (LC), Fording River Oxbow (FRO), Elk River Oxbow Lower (EROL), and Otto Creek (OC). Otto Creek is near, but not connected to, Goddard Marsh and drains the area around the Elkview Operations coal processing plant. These areas all showed moderate to high water and sediment selenium concentrations in previous sampling. Also, cutthroat trout were found at Clode Pond and Fording River Oxbow and contained relatively high tissue concentrations of selenium (Minnow Environmental Inc 2002). The reference areas were Elk River Oxbow Upper (EROU), Fording River Wetlands (FRW), and Otto Ditch (OD) (Table 1). EROU is upstream of any direct mine inputs. FRW is an isolated marsh upstream of lower Fording River and has no mine runoff in its drainage. Otto Ditch was added during the study to take advantage of a mallard brood observed there; although near both Goddard Marsh and Otto Creek on the Elkview Operations coal mine property, it is fed by groundwater seepages and receives no runoff from the coal mine. Table 1. Study Sites. Location Reference Areas Fording River Wetland Otto Ditch Elk River Oxbow Upper Exposure Areas High Exposure Clode Pond Line Creek Marsh Goddard Marsh Low Exposure Fording River Oxbow Otto Creek Elk River Oxbow Lower Code FRW OD EROU CP LC GM FRO OC EROL

11 November 19, 2004 WATERFOWL 4 Goddard Marsh Otto Creek (background) and Otto Ditch (foreground)

12 November 19, 2004 WATERFOWL 5 We supplemented previous aqueous selenium concentration data (Harding 2003) with new samples from Otto Ditch (OD), where selenium concentrations averaged 0.65 µg/l (N=3). Based on these results and a review of 2002 and 2003 concentrations in water and sediment (EVS Environment Consultants 2003a, b; Minnow Environmental Inc 2002), we classified exposed and reference areas as follows: High selenium exposure >10 ppb Se: Goddard Marsh (GM), Line Creek Marsh (LC), and Clode Pond (CP). Low selenium exposure 1.0 to 10 ppb Se: Fording River Oxbow (FRO), Otto Creek (OC) and Elk River Oxbow-Lower (EROL). Reference <1.0 ppb Se: Elk River Oxbow-Upper (EROU), Otto Ditch (OD), and Fording River Wetlands (FRW). The lower Elk River Oxbow site is, however, somewhat ambiguous. Minnow Environmental Inc. (2002) considered this site to be "exposed" to selenium on the basis that the Elk River floods into it during freshet, notwithstanding that they measured <0.5 µg/l in the water and 2.0 mg/kg in the sediment (almost identical to EROU, < 0.5 µg/l and 2.1 mg/kg Se, respectively). The lower Elk River ranges from 4.6 below Sparwood (EVS station 11) to 3.0 µg/l Se at Highway 93 (EVS Environment Consultants 2003b). EVS Environment Consultants (2002) referred to Station 11 as a "far-field" site, i.e., low exposure. For this report, we retained the classification of the lower Elk River Oxbow as a low exposure site because waterfowl feed in other wetlands besides their immediate nesting environment and would probably be exposed to selenium in the river and other connecting water bodies, even if they received little exposure from EROL itself. Waterfowl broods seen on the river adjacent to the lower Elk River Oxbow were included in the EROL site.

13 November 19, 2004 WATERFOWL 6 Elk River Oxbow-Lower Average aqueous selenium concentrations are shown in Figure 2. Data for Otto Ditch are from this study (mean of 3 in 2004), those for Line Creek Marsh are from Line Creek Operations (mean of 3 in 2004), those for Otto Creek are from Elkview Operations (2003 average of the creek's inlet and outlet), and those for other stations are means for 2002 as previously reported (EVS Environment Consultants 2003a, b; Minnow Environmental Inc 2002).

14 November 19, 2004 WATERFOWL 7 Aqueous selenium concentrations (µg/l) Goddard Marsh Clode Pond Line Creek Marsh Fording River Oxbow Lower Elk River Stn 11 Otto Creek Otto Ditch Elk River Oxbow-Lower Elk River Oxbow-Upper Fording River Wetland Figure 2. Selenium concentrations (µg/l) in water from exposed and reference areas (Lower Elk River is at EVS Stn. 11 below Sparwood). Permits In 2003, we collected waterfowl eggs under Environment Canada Scientific Permit Numbers and , issued May 20, 2003 and June 6, 2003, respectively. In 2004, we operated under Environment Canada Scientific Permit Number , issued March 8, Sample Collections and Brood Observations In 2003, we searched for nests from May 30 through June 2, and June 12 through June 16. We used a canoe or chest waders in open water areas. In shallow and upland areas, we searched by dragging a chain across open meadows and by beating brushy areas with willow switches. The chain was 3 m in length, to which 2 m lengths of 3 mm steel cable were added at both ends, so Fording River Oxbow that 2 investigators could cover a 7 m wide swath of marsh. The objective was to prompt the hens to flush so that nests could be found. The chain-dragging technique is a

15 November 19, 2004 WATERFOWL 8 standard method that works well in prairie marshes, where investigators normally deploy longer, heavier chains behind a pair of all-terrain-vehicles. In this study, where the topography and trees prevented the use of ATVs, 2 investigators dragged the chain stretched between them while walking. This technique proved ineffective and no active nests were found, although a few eggs were salvaged from predated nests. In 2003 we also collected a mix of breast and flight feathers from inactive nests, kill sites, preening sites, and dead waterfowl that we found. In March 2004, we began monitoring waterfowl migration, courtship, and nesting. When the first nesting behaviour was seen in geese, which mate for life and remain together as pairs throughout nesting and clutch-rearing, we began searching for nests by walking marsh edges and by observing mated pairs to locate nest sites. For ducks, which Collecting a goose egg generally do not mate for life and segregate by gender following mating, after observing courtship behaviour to determine when and where nesting was likely, we searched probable nesting locations. At several locations we tried using trained bird dogs and experienced houndsmen to find nests, but this also proved ineffective. When we found nests, we collected an egg from each and continued monitoring the nests to document their status and outcome. The status of active nests was defined as either laying or incubating and the number of eggs was determined. Nesting outcomes were either unsuccessful (i.e., destroyed or predated) or successful. If successful, we determined the number of hatched eggs by the number of eggs present when last seen, the number of egg shells with membranes indicative of hatching, or the presence of egg shell fragments that had been predated. After nesting, we counted all broods seen in each study site. Ducklings unaccompanied by parents were identified with reference to Harrison (1978). Since waterfowl leave their nests immediately upon hatching, it was not possible to associate each brood with a given nest. Therefore, the productivity data are based on mean clutch sizes and brood sizes per site. We examined all ducklings and goslings with 10x binoculars or a 20x spotting scope to identify any deformities that might have indicated teratogenic effects of selenium. We weighed (Ohaus balance beam scale) and measured (Fowler electronic calipers) collected eggs and examined them for viability and health status of embryos, including any physical abnormalities that might indicate teratogenic effects of selenium. Chemical Analysis Egg contents and feather samples were submitted to a private laboratory (ALS Environmental of Vancouver, British Columbia, certified by the Canadian Association for Environmental Analytical Laboratories Inc. and accredited by the Standards Council of

16 November 19, 2004 WATERFOWL 9 Canada). Analysis followed Environment Canada and U.S. Environmental Protection Agency protocols for the Puget Sound-Georgia Strait region (Puget Sound Water Quality Action Team 1995). The samples were homogenized and digested with nitric acid followed by repeated additions of hydrogen peroxide. Instrument analysis for arsenic, mercury and selenium was by HVAAS (Atomic Absorption Spectrophotometry; EPA Method 7000 series). Other metals were analysed by ICP/MS (Inductively coupled plasma mass spectrometry; EPA Method 6020). A series of internal standards covered the mass ranges of Li6, Sc45, Y89, In115, Re187, and Th232. Results were reported in both wet and dry weight. In 2003, 4 waterfowl eggs that were salvaged from predated nests were analysed for 25 elements (metals and metalloids) by HVAAS and ICP/MS as described above, with results reported in wet weight and dry weight. Waterfowl feather samples were analysed for selenium only, with results reported in wet and dry weight. In 2004, the samples were analysed in 3 batches. In the first, the laboratory's Data Quality A brood of goslings Objectives (accuracy 70% 120% recovery, precision 45% relative difference) were met for all samples. A pair of replicate samples produced identical results at 2 significant digits. Two laboratory method blanks were 0.18 and 0.16 mg/kg, respectively, compared to a target of < 0.10 mg/kg. Recoveries of selenium in 2 standard reference materials (NRC Dogfish Liver, DOLT-3 and NRC Lobster Hepatopancreas, TORT-2) were -4.0% and - 8.3%, respectively. In the second batch, the laboratory's Data Quality Objectives (accuracy 70% 120% recovery, precision 45% relative difference) were met for all samples. Five laboratory method blanks were all less than the detection limit (<1.0 mg/kg). Recoveries of selenium in 4 standard reference materials (1 NRC Dogfish Liver, DOLT-3, 3 NRC Lobster Hepatopancreas, and TORT-2) averaged -0.5% of the certified value. A pair of replicates produced an 11% difference. The third batch of 3 eggs was submitted in September 2004 for 25 elements (metals and metalloids) by HVAAS and ICP/MS, as described above, to check for the presence of elements that could interfere with selenium uptake or toxicity. Results reported in wet weight and dry weight. Three laboratory method blanks were below detection limits for all measurements. Certified reference materials NRC Dogfish Liver DOLT-3, NRC Lobster Hepatopancreas TORT- 2, and NRC Dogfish Liver DOLT-2 had recoveries 100%, 100% (mean of 3), and 87.3%. Statistical Analysis Statistical analysis was performed with SPSS. Some graphs were made with Excel. Comparisons of selenium concentrations and clutch and brood sizes were by ANOVA with

17 November 19, 2004 WATERFOWL 10 Bonferroni post-hoc analysis for multiple comparisons. Selenium concentrations were logtransformed prior to comparative analysis. Different waterfowl species have characteristic clutch and brood sizes. Geese, for example, normally have 5 6 eggs, while ring-necked ducks have 9 10 (Appendix, Table 6). Therefore, to compare clutch and brood sizes of all species across all sites, these data were normalized to express the value as the percent deviation from the provincial average using the formula: % deviation = (brood count/provincial average) 1 Common names and species codes in the graphs and tables conform to British Columbia standard nomenclature (Resources Information Standards Committee 2003). Species codes used in this report are: AMCO=American coot, HOME=hooded merganser, MALL=mallard, BWTE=blue-winged teal, GWTE=green-winged teal, RNDU=ring-necked duck, CAGO=Canada goose, BAGO=Barrow s goldeneye, BUFF=bufflehead. RESULTS Weather Although weather during the study was generally within the normal range, in June, when most broods were hatching, there were 3 minimum temperature excursions to well below normal (Figure 3; data from Environment Canada, Canadian Climate Data Archive, downloaded August 28, 2004). Minimum temperature ( C) at Sparwood May June July Figure 3. Minimum temperatures at Sparwood, May July 2004, from the Canadian Climate Data Archive. Values missing from the archive are blank. Horizontal lines are long-term normals. Selenium Concentrations We collected 3 waterfowl eggs in 2003 (not counting a common snipe egg), all salvaged from inactive nests or predators' caches (see Appendix, Table 4 for locations). We also collected

18 November 19, 2004 WATERFOWL samples of feathers (mix of breast, back, and flight feathers) from nest sites, preening sites, kills and a duck found dead. Analytical results are given in the Appendix, Table 3 for feathers, and Table 4 for eggs. In 2004, we collected 26 waterfowl eggs from 19 active nests (7 were salvaged after the nests were predated). There being no basis for separating the egg results by year, we combined results for both years for statistical analysis. Selenium concentrations and egg measurements are given in the Appendix, Table 5. Three additional waterfowl eggs (nestmates of eggs previously analysed for selenium that were salvaged from destroyed nests) were submitted for ICP/MS scan of 24 elements; results are included in the Appendix, Table 4. Selenium concentrations in waterfowl eggs ranged from 1.32 mg/kg dry weight in a Canada goose at the lower Elk River Oxbow to 34.7 mg/kg dry weight in an American coot at Clode Pond. Mean egg selenium (MES) concentrations by species by site are shown in Figure 4. MES (mg/kg) by Site Mean +- 2 SE Se dry wt Species AMCO BAGO CAGO HOME MALL RNDU 0.00 CP EROL EROU FRO FRW GM OC Site Figure 4. Selenium concentrations (mg/kg) in all species by site. Error bars are given for N > 1 samples. CP and GM are high exposure sites; EROL, FRO, and OC are low exposure sites; EROU and FRW are reference sites. Only Canada goose and mallard eggs were collected at > 2 sites. With such a small data set, statistical comparisons among species or sites are tenuous at best. Nevertheless, some basic trends are apparent. Overall, there was no clear relationship between selenium concentrations in eggs and in the water of the marshes where the eggs were collected. In Goddard Marsh,

19 November 19, 2004 WATERFOWL 12 however, the aqueous selenium was very high (78 µg/l) and the egg selenium concentration was moderate (4.4 mg/kg dry wt.). With this outlier removed (see Discussion), a weak, positive relationship emerges of increasing selenium in eggs with selenium in water (Figure 5). All taxa, all sites except GM Selenium in water (ppb) R 2 = Selenium in eggs (ppm) Figure 5. Selenium in waterfowl eggs and water (Goddard Marsh omitted). The 2 species for which eggs were collected at more than 2 sites (Canada geese and mallards) showed a pattern consistent with this relationship: lower selenium concentrations at reference sites and higher at exposed sites. In mallards (Figure 6), these between-site differences for log-transformed selenium concentrations were not significant (ANOVA: p = 0.15) owing to the small sample sizes, despite a considerable spread from the low at the lower Elk River Oxbow (3.23 mg/kg dry wt.) to the high at Otto Creek (15.7 mg/kg dry wt). In Canada geese (Figure 7), between-group differences were significant (ANOVA: p = 0.008), with eggs from Clode Pond (17.1 mg/kg dry wt.) and Goddard Marsh (4.4 mg/kg dry wt.) having much higher egg selenium than those from Otto Creek and the upper and lower Elk River Oxbows (2.26, 1.69 and 1.38 mg/kg dry wt., respectively). Measuring a mallard egg

20 November 19, 2004 WATERFOWL 13 Figure 6. Mean selenium (mg/kg) concentrations in mallard eggs. Figure 7. Mean selenium concentrations (mg/kg) in Canada goose eggs.

21 November 19, 2004 WATERFOWL 14 When grouped by species and site, the tendency toward higher concentrations at exposed sites was suggestive (Figure 8), although the differences were not significant due to the poor representation of any species across all sites. When all species were combined and grouped by exposure category, however, the log-transformed selenium concentrations were significantly different between the reference and high exposure sites (Table 2). MES (mg/kg) by Exposure Category Mean +- 2 SE Se dry wt Species AMCO BAGO CAGO HOME MALL RNDU Reference Low High Exposure Figure 8. Selenium concentrations (mg/kg) in all species by exposure category. Error bars are given for N > 1 samples.

22 November 19, 2004 WATERFOWL 15 Table 2. Bonferonni comparison of log-transformed Se (dry wt.) among high and low exposure areas and reference areas. (I) Exposure (J) Exposure Mean 95% Confidence Interval Difference (I-J) Std. Error Sig. Lower Bound Upper Bound Reference Low High (*) Low Reference High High Reference (*) Low * The mean difference is significant at the.05 level. Productivity We counted eggs in 19 waterfowl clutches of 5 species and 50 broods of 8 species. Data are given in the Appendix, Table 6. Clutch Size Figure 9 shows the clutch sizes normalized to percent deviation from normal so that they can be represented at the same scale. Only Canada geese were represented at enough sites to compare clutch sizes by species. Mean Canada goose clutch sizes were nearly identical (ANOVA: p > 0.05) at all 7 sites. Ring-necked duck nest

23 November 19, 2004 WATERFOWL 16 Mean +- 2 SE Clutch Normalized Species AMCO CAGO HOME MALL RNDU CP FRO FRW EROU GM OC EROL Site Figure 9. Mean ± 2SE (Standard Error) clutch size, all taxa by site. Error bars are given for N > 1 samples. CP and GM are high exposure sites; EROL, FRO, and OC are low exposure sites; EROU and FRW are reference sites. American coot nest

24 November 19, 2004 WATERFOWL 17 To compare clutch size of all waterfowl species, which normally have different clutch sizes, the counts were normalized to express deviation from the provincial average. There was no difference of normalized clutch size among exposure categories (ANOVA: p > 0.05; Figure 10). Normalized Clutch Size 0.20 Mean +- 2 SE Clutch Normalized Reference Low High Exposure Figure 10. Clutch size by exposure category (mean deviation from average ± 2SE (Standard Error)). A brood of Barrow's goldeneye ducklings

25 November 19, 2004 WATERFOWL 18 Brood Size Brood sizes also showed no significant differences among sites for most species (ANOVA: p > 0.05). However, brood counts were poorly represented at the high-exposure areas: 2 at Clode Pond (1 of which, a brood of American coots, was inferred because all 7 eggs hatched out, but the hatchlings were never seen), 1 at Line Creek Marsh and none at Goddard Marsh. Line Creek Marsh is poor habitat and Goddard Marsh, although seemingly excellent habitat, is so congested with cattails and willows that observing broods was extremely difficult. Figure 11 shows the brood sizes normalized to percent deviation from normal so that they can be represented at the same scale. For mallards, however, the low mean at Otto Creek (3.5 ducklings, N = 2) and the larger brood at Otto Ditch (8 ducklings, N=1) resulted in a barely statistically significant difference (p = ). Mean +- 2 SE Brood Normalized Species AMCO BAGO BUFF BWTE CAGO GWTE HOME MALL CP FRO FRW EROU OC OD EROL GMU LC Site Figure 11. Mean ± 2SE (Standard Error) brood size, all taxa by site. Error bars are given for N>1 samples. CP and GM are high exposure sites; EROL, FRO, and OC are low exposure sites; EROU, OD, GMU, and FRW are reference sites.

26 November 19, 2004 WATERFOWL 19 To compare brood sizes of all waterfowl species, which normally have different brood sizes, the counts were normalized to express deviation from the provincial average. There was no difference of normalized brood size among exposure categories (ANOVA: p > 0.05; Figure 12). The wide variance in highly exposed sites was notable, however, relative to the other 2 exposure categories. Normalized Brood Count 1.00 Mean +- 2 SE Brood Normalized Reference Low High Exposure Figure 12. Brood counts by exposure category (mean deviation from average ± 2SE (Standard Error).

27 November 19, 2004 WATERFOWL 20 Egg Health Significant between-site differences (ANOVA: p = 0.039) were found in Canada goose eggs, those from the upper Elk River Oxbow having smaller eggs than other sites (Appendix, Table 5). In mallards, 2 smaller-thanaverage size (ANOVA: p=0.018) and weight (ANOVA: p = 0.003) eggs at the lower Elk River Oxbow contributed to the difference. Of 6 embryos that were large enough that abnormalities might have been detected (> ~ 40 mm length), none were seen. Similarly, of 258 ducklings and goslings observed, no physical abnormalities were observed. Survival Examining a duck egg Of 34 broods that we were able to monitor for > 7 days, 30 had 100% survival after an average of 20 days (range = 7 to 41 days). Those that had less than 100% survival were a brood of 4 coots at GMU that had 1 surviving after 18 days; a Barrow's goldeneye brood of 12 at FRO that had 10 surviving after 9 days; a brood of 4 buffleheads at FRW with 3 surviving after 24 days; and a brood of 5 Canada geese at EROL that had 4 surviving after 33 days. Among the longest surviving broods were 4 goslings out of a brood of 5 at the lower Elk River Oxbow that survived at least 33 days; a Goose egg contents showing vein structure and brood of 7 hooded mergansers, also at embryo (lower right) EROL, that survived at least 41 days; and a brood of 4 goslings at Clode Pond that survived at least 31 days.

28 November 19, 2004 WATERFOWL 21 Mortality Of the 19 nests that we found, predators took some or all of the eggs from 8 (42%). These nests contained 121 eggs, of which 27 (21%) were taken by predators. Predators or their sign seen in marsh areas included: Red-tailed hawk, common crow, and raven at all sites Great-horned owl, grizzly bear, and river otter at Elk River Oxbow-Upper Great blue heron and American kestrel at Clode Pond Cooper's hawk a Elk River Oxbow-Lower Great blue heron, coyote, grizzly bear and black bear at Goddard Marsh Coyote and river otter at Otto Creek Four dead adult ducks were found that had not been eaten or scavenged by predators and that showed no obvious signs of trauma or injury. In 2003, we found a dead mallard drake and a dead green-winged teal female at Clode Pond. In 2004, we found a dead mallard drake and a dead hooded merganser, also at Clode Pond. Samples of feathers from the mallard and teal found on June 16, 2003 had 2.51 mg/kg and 3.41 mg/kg selenium dry weight, respectively (Appendix, Table 3). This was within the range (1.45 mg/kg to 5.49 mg/kg) of other waterfowl feathers from a kill, a preening site, and a nest at the same site (Appendix, Table 3). The mallard was freshly dead and floating in the water in the east side of the pond, while the teal was a dried-up carcass found on the spit that bisects the west side of the pond. The 2 dead ducks found in 2004 were too decomposed to sample, having died some weeks previously. DISCUSSION The limitations of this study were in the modest sample sizes (19 clutches and 50 broods counted) and in our inability to definitely link broods observed to specific nests from which eggs were counted and analysed for selenium concentration. These limitations, which were anticipated in the approved study design, resulted from the low numbers of waterfowl nesting at each marsh, the difficulty in finding their nests, and the mobility of broods immediately after hatching. Since we found only 19 active waterfowl nests but counted 50 broods and some broods were undoubtedly missed, or hatched after our field work ended it is apparent that we found only approximately one-third of the nests. Considering the amount of field time spent searching (170 person-days during March July), a more effective nest-finding method would be needed to increase the sample size. Given a potential breeding population at these 9 sites of approximately 6 7 successful pairs per site, virtually all of them would have to be found to generate adequate sample sizes for a robust comparison of productivity. Otto Ditch was added as a reference area, despite its proximity to Otto Creek, because during our concurrent blackbird study we observed a mallard brood there and did not want to miss the opportunity to add to our sample size. However, since mallards feed widely around their nests sites before and after nesting, this site is ambiguous as to its exposure classification and should not be sampled in future studies.

29 November 19, 2004 WATERFOWL 22 The concentrations of selenium in birds' eggs in this study were generally lower than reported toxic thresholds, but some were higher. Skorupa and Ohlendorf (1991), in a synthesis of toxicity data, suggested a toxicity threshold in birds eggs of about 8.0 µg/g dry weight. Other studies also showed effects at the 3 to 8 µg/g dry weight level (Fairbrother et al. 1994; Heinz et al. 1990; Lemly 1996, 1997; Skorupa 1998; Stanley et al. 1994; U.S. Department of the Interior 1998; U.S. Environmental Protection Agency 1998). More recently, higher effects thresholds of 12 to 15 µg/g dry weight MES for embryotoxicity and nestling mortality have been recommended (Adams et al. 2003; Fairbrother et al. 1999; Fairbrother et al. 2000; Ohlendorf 2003). Based on these thresholds, the eggs and young of Canada geese (17.1 mg/kg Se dry wt., N=1), American coots (29.6 mg/kg Se dry wt., N=2) at Clode Pond, and mallards at the Fording River Oxbow (12.0 mg/kg Se dry wt., N=2) and mallards at Otto Creek (15.7 mg/kg Se dry wt., N = 1) could be at risk to selenium toxicosis. The above thresholds were based mainly on laboratory toxicity studies of mallards and black-necked stilts. Diving ducks that eat more benthic invertebrates and fish-eating waterbirds often have higher selenium concentrations than other waterfowl (Henny et al. 1991; Ohlendorf et al. 1986c), although toxic effects in these species have not been reported (Ohlendorf 1993). In this study, diving ducks included ring-necked duck and Barrow's goldeneye. About 75% of Barrow's goldeneye food is animal, including molluscs and other benthic invertebrates, crustaceans, and fish, while pondweeds and other vegetation make up the balance (Kortright 1942). The ring-necked duck, although a diving duck, occupies habitats and has a diet more like those of the mallard, consuming about 80% vegetation. The hooded merganser eats small fish, frogs, tadpoles, and invertebrates such as beetles and caddis fly larvae (Kortright 1942). At Kesterson Reservoir in California, a site highly contaminated with selenium, eared grebes, which have a diet similar to that of hooded mergansers, had up to 67.7 mg/kg mean egg selenium (Ohlendorf et al. 1986b). Canada geese often feed on early greening meadow grasses in the spring, but also eat sedges and other aquatic plants, insects, crustaceans, and small molluscs, as do mallards (Kortright 1942). Therefore, the species studied here are representative in habitat and dietary habits of others reported in the literature. Several studies have shown that the American coot, an herbivore, accumulates higher concentrations of selenium than dabbling ducks that feed more on aquatic insects (DuBowy 1989; Ohlendorf et al. 1986a). This is consistent with our finding of 2 coot clutches at Clode Pond that averaged 29.6 mg/kg Se dry weight, which was higher that a Canada goose egg from the same site with 17.1 mg/kg Se dry weight. Coots may also be more sensitive than mallards (Fairbrother et al. 1999; Ohlendorf et al. 1989; Williams et al. 1989). At Kesterson Reservoir, coots with a mean egg selenium concentration of 30.9 mg/kg dry weight had a 64.4% frequency of nests with embryotoxicity including dead or deformed chicks (Ohlendorf et al. 1986b). Those coots also had only a 57.9% nest success (percent of nests that produced at least one juvenile) (Ohlendorf et al. 1989). Although 7 coot chicks hatched at Clode Pond on July 9, they were never seen despite 3 days' unsuccessful searches for the brood (July 10, 17, and August 27). Conversely, 4 Canada goose goslings hatched at the same site were among the longest survivors: 31 days. Although the apparent absence of the coot chicks is suspicious, it is not, by itself evidence of selenium toxicosis. The lack of high selenium accumulation in a Canada goose egg from the middle of Goddard Marsh is surprising in view of the high selenium concentrations in the water. This site

30 November 19, 2004 WATERFOWL 23 seems to be an outlier in terms of selenium uptake in waterfowl. The most likely explanation is that, prior to egg-laying, the geese fed in less-contaminated marshes or uplands. However, other elements such as boron, arsenic, and mercury, can act antagonistically with selenium in waterfowl, ameliorating adverse effects (Adams et al. 2003; Hoffman and Heinz 1998; Stanley et al. 1994; Stanley et al. 1996). Although these elements have not been measured in birds' eggs or in invertebrate tissues from Goddard Marsh, mercury was present in duck eggs from 2 other sites (EROU and FRO) in trace amounts (Appendix, Table 4). Sulphate can also interfere with both the uptake and the toxicity of selenium (Ogle and Knight 1996). The sediments at Goddard Marsh are anaerobic and reducing, while those at Clode Pond are aerobic and oxygenated (Minnow Environmental Inc. 2004). The dissolved sulphate in the water at Goddard Marsh (302 mg/l) was much higher than at Clode Pond (165 mg/l and 192 mg/l in 2 samples) (Minnow Environmental Inc. 2004). The higher amount of sulphate in Goddard Marsh might have reduced the uptake of selenium there. Alternatively, the Goddard Marsh pair of geese may have fed primarily at uncontaminated sites nearby. The waterfowl feather selenium concentration data presented here may be useful in long term studies or regional comparisons, but provide little information on uptake and possible toxicity at these sites. Feathers in waterfowl can be useful in metal and metalloid pollution monitoring (Burger 1994; Furness et al. 1986; Goede and DeBruin 1984). Feather data must, however, be interpreted with caution because they represent uptake at the time that the feathers were formed, rather than when collected; moreover, feathers may accumulate additional contaminants directly from the environment after formation, or they may lose certain contaminants through leaching (Ohlendorf 1993). In this situation, feathers collected in the spring of 1993 were formed after the previous summer's moult, and would therefore reflect the selenium concentrations of their post-moulting environment, rather than the spring feeding and courtship habitats where they were collected. Apart from the relatively high incidence of nest predation, productivity in terms of both clutch size and brood size was at least as high as the provincial averages, and no different at exposed sites compared to reference sites. For the few broods that could be monitored for more than 1 week, survival was high. Most selenium toxicity studies of duckling survival found that selenium-induced mortalities occurred within 1 to 2 weeks after hatching, although Mallard drake mortality can occur as late as 6 weeks post-hatch (Heinz et al. 1988). We followed broods with 100% for more than 4 weeks even in the most contaminated sites. This shows, at least, that if selenium is affecting clutch size, brood size or brood survival, it is at such a low level as to be undetectable with a sample size of 50 broods.

31 November 19, 2004 WATERFOWL 24 Moreover, no unusual mortality was observed except for the 4 dead adult ducks at Clode Pond. While these anecdotal observations are suspicious in view of the high selenium concentrations in the water, there are many natural and unnatural hazards that could account for them. These include a variety of viral, bacterial, and mycological infections, as well as anthropogenic sources including fish hooks and lead fishing sinkers, which fish-eating waterfowl such as mergansers sometimes ingest, and lead shot that dabbling waterfowl often ingest in heavily hunted marshes. Furthermore, adult waterfowl may show no ill effects at dietary selenium concentrations that affect survival of eggs and young (Ohlendorf 1993). It is surprising, however, that no ducks nested at Clode Pond, despite the regular occurrence and courtship throughout the nesting period, of mallards, buffleheads, and Barrow's goldeneyes, with occasional occurrence of pairs of ring-necked ducks, American widgeons, cinnamon teal, lesser scaups, and common mergansers. In 2002, Anatum Ecological Consulting (2002) also noted the lack of waterfowl broods: "It was surprising that Clode Pond did not have at least one nesting pair of ducks on the water body. It has a rich fish food supply and likely invertebrates for nesting waterfowl. There is sufficient nesting cover, comparable to other sites surveyed in the valley, which had pairs of ducks and were successful in raising broods. Possible reasons for the lack of nesting waterfowl could be vehicle or people disturbance from adjacent roads, unusually wet/cold spring weather, or some other influencing effect. Numerous waterfowl do feed or stop over in the area during the spring" The apparent paucity of nesting waterfowl in what appears to be good habitat is also a cause for concern about this site. The small data set reported here on waterfowl clutch and brood size and egg health in selenium-exposed and reference sites does not reveal whether selenium is affecting productivity. Waterfowl move around widely while courting and selecting nests sites. Those that made nests in selenium-exposed marshes may have fed primarily at unexposed sites, and vice versa. The few significant differences found were more likely due to random effects or habitat quality differences. As shown in the Appendix, Table 6, mean clutch and brood sizes of species except Canada geese were within or above the normal ranges for British Columbia. Therefore, despite the possibility that the selenium at Clode Pond may be toxic to individual birds or their progeny, there is no direct evidence of toxicity to waterfowl, and no evidence of depressed waterfowl productivity in selenium-exposed sites. The results of this preliminary study should be interpreted with caution, in view of the small sample sizes that, for the most part, preclude statistical comparisons. CONCLUSIONS 1. Waterfowl eggs contained higher concentrations of selenium at the more highly contaminated sites than at the reference sites. The selenium concentrations in the eggs of 3 species (mallards, American coots and Canada geese) at 3 sites (Otto Creek, Fording River Oxbow, and Clode Pond) were above reported toxic thresholds. 2. Goddard Marsh seems to be an outlier in terms of the modest selenium concentration in a single goose egg, relative to concentrations in the water. It may simply be because the