Introduction

Differential effects of pesticides

Fish kills are episodes of massive fish mortality over a short period of time. Agricultural pesticide runoff has caused numerous fish kills in recent years (USEPA, 2001; Mutch et al., 2002). It is clear that different species of fish respond differently to chemical exposure (e.g., Post and Leasure, 1974; Arnold and Braunbeck, 1994) and that such interspecies differences in behavioural responses to chemical exposure may contribute to changes in community structure. Hansen et al. (1999), for example, suggested that differential sensitivity to metal contamination was responsible for a change in the relative proportions of rainbow trout (Oncorhynchus mykiss) and brown trout (Salma trutta) in different sections of a Montana river.

Within a species, different life stages may experience differential sensitivity to many contaminants (McKim, 1977; Van Leeuwen et al., 1985). For example, when DDT was aerially applied over a forest in New Brunswick in the 1960s to control spruce budworm, Atlantic salmon parr (2+ fish) numbers decreased by 40%, but young of the year (0+) salmon numbers decreased by 98% (Elson, 1967). Such age-related mortality after exposure to anthropogenic chemicals is likely to have long-term effects on population structure and needs to be considered more closely in management planning.

Fish kills and azinphos-methyl

Since 1962, at least 44 documented fish kills that were either proved or suspected to have been caused by pesticides have occurred on Prince Edward Island (PEI) (Mutch et al., 2002); 22 of these kills occurred between 1994 and 2002. Sixteen of the fish kills on PEI since 1994 have had specific pesticides attributed to them; of these 16, 10 involved azinphos-methyl. As of 2000, the pesticide azinphos-methyl was determined to be responsible for 143 fish kill incidents in the United States, which is 21% of all reported pesticide-associated fish kills in that country (USEPA, 2001). Azinphos-methyl is an organophosphate insecticide used on a number of food crops, including potatoes. It has a half-life of 26 d at 30 °C and pH 7. Its metabolites (dimethylphosphorothioic and dimethylphosphoric acids, desmethyl azinphosmethyl, and azinphos-methyloxon) are thought to be more toxic than the parent compound (WHO, 2002). Azinphos-methyl is highly soluble in water (25.10 mg/l at 25 °C), and does not readily leach through soil. It is therefore likely to reach watercourses as runoff associated with rainfall (USEPA, 1998). This pesticide is an acetylcholinesterase inhibitor and, as such, is thought to affect various fish behaviours, such as schooling, temperature selection, respiration, feeding, and rheotropism (Smith, 1984). Disruption of some of these behaviours (e.g., respiration) could cause acute mortality; disruption of other behaviours (e.g., feeding, rheotropism) could lead to mortality over a longer period of time. Ninety-six hours LC50 values of azinphos methyl of the three salmonid species found on PEI are 5.3 μg/l for rainbow trout, Oncorhynchus mykiss (number of studies = 5), 2.9 μg/l for Atlantic salmon, Salmo salar (n = 7), and 1.2 μg/l for brook trout, Salvelinus fontinalis (n = 1) (USEPA, 1998).

The Wilmot River

The Wilmot River is one of PEI’s longest rivers and has a watershed area of approximately 166 km2 (Fig. 1). In 2000, over 77% of the land in the watershed was devoted to agriculture; of this, 36% was in potato production (determined using MapInfo Professional Version 6.5 from 2000 Corporate Resource Inventory GIS layer obtained from the PEI Dept. of Agriculture and Forestry). Fish communities in the river had been sampled in 2001 as part of a study concerning the relationship between land use and salmonid populations. Brook trout and rainbow trout were found at high densities while only three Atlantic salmon were captured during that year. Three-spined stickleback (Gasterosteus aculeatus) were found seasonally (June); no other fish species were captured at the study sites, although nine-spined stickleback (Pungitius pungitius) and American eel (Anguilla rostrata) have been found in the river further downstream (S. Hill, Watershed Coordinator, PEI Dept. of Fisheries, Aquaculture, and Environment, pers. comm.). In July 2002, heavy rainfalls were responsible for two separate runoff events of soil, pesticides, and fertilizers into the Wilmot River from adjacent potato fields. Analysis of water samples by Environment Canada after both events for pesticides known to have been recently used on fields in the watershed revealed that azinphos-methyl (O,O-dimethyl S-[(4-oxo-1,2,3-benzotriazin-3(4H)-yl)methyl] phosphorodithioate) was present in high concentrations (B. Birch, Environment Canada, pers. comm.). Two days after the first runoff event (10 July 2002), water samples were taken from the river and adjacent land for pesticide analysis by Environment Canada. Azinphos-methyl was found in standing water adjacent to the river at concentrations up to 1225 μg/l, and was found at 0.8 μg/l in the river. There were 4500 dead salmonids (4494 brook trout and 6 rainbow trout) recovered after this pesticide runoff event; other species (three-spined stickleback and nine-spined stickleback) were observed but not recorded (S. Hill, Watershed Coordinator, PEI Dept. of Fisheries, Aquaculture, and Environment, pers. comm.). Water samples taken from the river by Environment Canada one day after the second runoff event (19 July 2002) were found to have azinphos-methyl concentrations ranging from 0.39 to 0.8 μg/l (B. Birch, Environment Canada, pers. comm.). There were 6500 dead salmonids (6412 brook trout and 88 rainbow trout) recovered after this pesticide runoff event (S. Hill, pers. comm). Azinphos-methyl is highly toxic to fish (USEPA, 1998) and was likely responsible for the observed fish mortality (B. Birch, Environment Canada, pers. comm.). Previous pesticide monitoring of streams in agricultural areas on PEI has found detectable levels of azinphos methyl in approximately 25% of samples; the median concentration in these samples was 0.073 μg/l (Mutch et al., 2002) and higher concentrations were associated with heavy rainfall events.

Figure 1.
figure 1

The Wilmot River watershed, Prince Edward Island, indicating sampling sites (arrows; 1–3) and entry points of the pesticide runoff (stars; black: 10 July and gray: 19 July).

Full size image

These runoff events provided a unique opportunity to study the effects of pesticide runoff on salmonid population and community structure. The objectives of this study were to (1) describe temporal and spatial patterns of fish survival after two pesticide runoff events, and (2) examine differences in community and population structure of salmonids after pesticide runoff events.

Materials and methods

Three sites on the Wilmot River (Fig. 1) were sampled for salmonids in July and August of 2001, 2002, and 2003. Site 1 was located at least 60 m upstream of all pesticide runoff events. Site 2 was downstream of the first runoff event (10 July 2002) and upstream of the second runoff event (19 July 2002). Site 3 was approximately 7 km downstream of the first runoff event and 2 km downstream from the second pesticide runoff entry point. Sampling dates for each site are given in Table 1.

Table 1. Fish sampling dates for 2001, 2002, and 2003
Full size table

An area of approximately 100 m2 was sampled at each site with barrier nets placed at the upstream and downstream boundaries. The fish were sampled by electrofishing with a Smith-Root backpack electrofishing unit, Model 12 POW. Three consecutive sampling runs were performed to permit the estimation of population sizes, probability of capture (usually between 80% and 99%) and 90% confidence limits using the program POPDN3, version 1.3 (1985). Captured fish were sedated with clove oil (1.25×10−2 ppt), identified to species, measured (fork length; ±1 mm) and weighed (±0.1 g). To estimate age categories (0+, 1+, 2+), length histograms of fish at each site were examined, and fish were visually separated into groups. Fish were placed in 50 L tubs of clear water after processing and released into the stream when fully recovered.

Water chemistry analyses were done to ensure

the water met requirements for salmonids and did not contribute to the observed fish mortality. Data loggers that recorded the temperature every hour were present from June to November of each year. Dissolved oxygen (mg/l; determined with a YSI Model 55 dissolved oxygen meter), conductivity (μS; determined with a YSI Model 33 Salinity-Conductivity-Temperature metre) and pH (determined with a Fisher Scientific Accumet® portable AP61 pH meter) readings were obtained during each fish sampling period.

Results

Changes in community structure

The salmonid community at the upstream, unaffected site (Site 1) was composed primarily of brook trout in all 3 years (Fig. 2a). Population density decreased throughout the summer in all 3 years in a consistent manner.

Figure 2.
figure 2

Population density (number/100 m2) of brook trout (BT) and rainbow trout (RT) during each sampling period (JL = July; AU = August) at (a) Site 1, (b) Site 2, and (c) Site 3.

Full size image

In 2001, the salmonid community at Site 2 (approximately 3.5–4.5 km downstream of first runoff event entry points) was composed primarily of brook trout (Fig. 2b). On 13 July 2002, 3 days after the pesticide runoff event, the total salmonid population had decreased by 40% from the previous year, and rainbow trout were the dominant species. The density of both species decreased throughout the summer. Salmonid densities 1 year later (10 July 2003) were similar to those in 2002; however, brook trout were once again the dominant species.

The most noticeable change in salmonid communities occurred at Site 3 (7 and 2 km downstream of first and second runoff events, respectively). Brook trout were the dominant species at this site in 2001 and 13 July 2002 (Fig. 2c). After the pesticide runoff event on 19 July 2002, overall salmonid density decreased by 90% from 7 days previous, with a 98% decrease in brook trout density and a 66% decrease in rainbow trout density. Salmonid density remained low for August. In 2003, salmonid numbers had increased from the previous year, but were still well below numbers occurring before the pesticide runoff events. However, rainbow trout outnumbered brook trout in all samples after the pesticide runoff event except July 2003, in which the two species were captured in approximately equal proportions.

Changes in population structure

The brook trout population at Site 1 was composed primarily of 0+ fish in all 3 years (Fig. 3a), but there were more 1+ fish captured in 2002 than in the other 2 years. Four of the five rainbow trout captured in 2001 were 0+; all of the rainbow trout captured in 2002 and 2003 were 0+ (Fig. 4a).

Figure 3.
figure 3

Population density (number/100 m2) of brook trout age classes (0+, 1+, and 2+) during each sampling period (JL = July; AU = August) at (a) Site 1, (b) Site 2, and (c) Site 3.

Full size image
Figure 4.
figure 4

Population density (number/100 m2) of rainbow trout age classes (0+, 1+, and 2+) during each sampling period (JL = July; AU = August) at (a) Site 1, (b) Site 2, and (c) Site 3. Note the different scales in each graph.

Full size image

Brook trout populations at Site 2 were dominated by 0+ fish (97%) in 2001 (Fig. 3b). This percentage declined during 2002 (after the pesticide runoff event) to 71%, 68%, and 47% for 13 July, 20 July, and 11 August 2002 samples, respectively, but by July 2003 had returned to ≥96%. Rainbow trout populations were comprised of mostly 0+ fish (≥80%) during all sampling times (Fig. 4b).

The age class structure at Site 3 before the pesticide runoff event was also dominated by 0+ fish (Figs. 3c and 4c). Before the closest pesticide runoff event (19 July 2002), 0+ brook trout comprised 95–96% of the brook trout sampled (Fig. 3C). One day after the closest pesticide runoff event, none of the three brook trout captured were young-of-the-year. There was only one brook trout captured in August (0+). In 2003, the majority (89–96%) of brook trout captured were again 0+ fish. The low density of 1+ brook trout in 2003 (none in July; one in August) was likely due to the decimation of this age class in 2002. However, naturally low densities of 1+ fish at this site make the interpretation of data difficult. Similarly, high proportions of the rainbow trout captured in 2001 (93–96%) and 2002 before the runoff event (89%) were 0+ fish (Fig. 4c); one day after the pesticide runoff event, 0+ rainbow trout accounted for only 54% of the remaining population. By August, this percentage had increased to 87%, and in 2003 the majority of rainbow trout captured (≥90%) were again 0+ fish.

Water chemistry results

All water chemistry values were within normal ranges for salmonids. Specifically, in 2002 the mean water temperatures for each site from 1 July to 1 September were 11.6, 16.9, and 15.0 °C, while the highest water temperature at each site was 15.1, 18.6, and 16.5 °C for sites 1, 2, and 3, respectively. The dissolved oxygen levels ranged from 9.8 to 10.3 mg/l over all 3 years.

Discussion

The fish kill events of 2002 had a significant short term impact on the structure of the salmonid communities of the Wilmot River. The population structure of both brook trout and rainbow trout was altered due to the apparent greater sensitivity of younger fish to the pesticide runoff. Similarly, the species composition of the study sites was altered through the differences in the response of brook trout and rainbow trout to the runoff. One year after the runoff events responsible for the two fish kills, noticeable differences were still apparent in both population and community structure of salmonids in affected sections of the river; however, fish populations do appear to be recovering.

Changes in community structure

After the pesticide runoff events of July 2002 on the Wilmot River, brook trout populations just downstream of the entry points decreased substantially; rainbow trout populations also decreased (site 3 only), although not to the same extent as brook trout. In affected sites, the fish community structure changed from brook trout to rainbow trout dominated communities. This change in the proportion of each species carried over into 2003 at Site 3.

In a study examining the salmonid population and community structure of 37 sites on 15 different rivers across PEI in 2001 and 2002, the only river (in addition to the site on the Wilmot River) where rainbow trout outnumbered brook trout was the Tryon River (Gormley, 2003). Interestingly, the section of the river sampled was documented to have a fish kill related to pesticide runoff in July 1999. The pesticide deemed responsible for the fish kill was, again, azinphos-methyl (Mutch et al., 2002). While data on the fish community before the fish kill are not available for this site, sampling two years after the event (in 2001) revealed an extremely low density of salmonids (36/100 m2) dominated by rainbow trout (89% of individuals captured). One year later, salmonid population density at the site increased to 110 salmonids/100 m2 with a higher proportion of brook trout (38%) than had previously been found (Gormley, 2003). In an earlier study, Johnston and Cheverie (1980), found that the entry of the pesticide endosulfan into a tributary of the Dunk River, PEI, was followed by a 67% decrease in brook trout population density, but only a 27% decrease in rainbow trout population density. These observations suggest that differential effects of pesticide runoff on salmonid populations may contribute to current species composition in various streams across the province.

The higher proportion of rainbow trout found in areas of rivers affected by pesticide runoff suggests a difference in the response of the two species to pesticide exposure. There are two possible mechanisms that may account for this difference: behavioural and physiological. A weaker or slower response to pesticide exposure by brook trout would result in higher mortality. Avoidance of chemicals is well-documented, however, and generally occurs at concentrations lower than that required to cause mortality (e.g., Hansen et al., 1972; Atchison et al., 1987). Brook trout on PEI have been previously observed to swim downstream, away from a pesticide runoff event on the Mill River (Saunders, 1969). On this river, salmonid (Atlantic salmon and brook trout) movement was monitored by a two-way counting fence; this fence demonstrated the mass movement of fish downstream immediately after the pesticide runoff event. Fish may also modify their chances of being exposed to the runoff by their position in the water column. For example, fish that remain mostly in high flow areas would receive a higher dose of the contaminant then fish which remained in groundwater upwellings. Brook trout are known to seek the lower temperatures supplied by groundwater upwellings if the temperature in the river increases (Power, 1980). However, the response of either species to contaminant exposure is currently unknown.

It is more likely that brook trout and rainbow trout exhibit differential physiological sensitivity to certain chemicals. For example, different salmonid species exhibit different levels of sensitivity to azinphos-methyl, the pesticide deemed responsible for the fish kills on the Wilmot River. Indeed, the 96-h LC50 value of azinphos-methyl is five times lower for brook trout (1.2 μg/l) than rainbow trout (5.3 μg/l) (USEPA, 1998), indicating the brook trout’s greater sensitivity to this compound.

Changes in population structure

For both species, young of the year (0+) fish suffered a larger decline in numbers than did older age classes. This observation was more evident at Site 3, as population data from 1 week prior to the runoff event were available. Conclusions at Site 2 are more difficult to make as the fish population was not surveyed that year until after the massive fish mortality episode.

Smaller bodied fish are generally found to be more susceptible than larger fish to toxicants (McKim, 1977; Van Leeuwen et al., 1985; Murty, 1986). The near-elimination of the 0+ fish in July 2002 was reflected in an extremely small cohort of 1+ fish in 2003. The appearance of a strong cohort of 0+ fish in 2003 is encouraging for the reproduction of the population; however, the impacts of the decimation of the 0+ age class of 2002 will not be felt until this cohort reaches reproductive age.

Conservation implications

The long-term impacts of pesticide runoff events are not well-understood, as data on fish community and population structure before the insult are often not readily available. In our study, short-term (1 year) impacts included changes in relative species abundance as well as population age structure. Atlantic salmon are declining through much of their range (Parrish et al., 1998). While the specific cause of this decline is unclear, the input of pesticides into historical salmon rivers is likely a contributing factor (Parrish et al., 1998). If rainbow trout, an exotic species, are less sensitive to stream changes associated with pesticide runoff events than are native species such as brook trout (as suggested by this study) and Atlantic salmon, pesticide runoff may also select for a non-native species.

The massive fish mortality events observed on the Wilmot River were two of at least eight events that occurred on different rivers across PEI in 2002. These events are unpredictable, and their occurrence is related to unforecasted, major rainfall episodes in areas recently sprayed with pesticides. In recent years, there have been summers with no fish kills (e.g., 2001, 2003), few fish kills (e.g., one in 2004, one in 1998), and numerous fish kills (e.g., eight in 1999, eight in 2002). Summers that experience less major rainfall events also experience fewer fish kills.

Although there have been numerous fish kills on PEI in the past decade, the Wilmot River sites were the only sites to have salmonid population data that were collected immediately prior to these events. This provided a unique situation to assess the magnitude of the effect of these events on salmonid populations. Although the populations appear to be recovering, the near-elimination of an age class will have impacts on the future reproduction of the populations. The fate of the salmonid populations in this river will largely depend on the presence or absence of future pesticide runoff events as well as restocking programs by local community groups.