Robert A. Ronconi, Department of Biology, University of Victoria, PO Box 3020, Stn CSC, Victoria, BC, Canada V8W 3 N5 (fax +250 721 7120; e-mail firstname.lastname@example.org).
1Oil sands mining is one of several industrial activities that produces effluent that is dangerous to waterfowl. Such industries require effective systems to deter birds, but current deterrents are not always successful, presumably because wildlife ignore or habituate to them.
2We tested a new radar-activated on-demand system of deterrence in the oil sands region of Alberta, Canada, by comparing the proportion of birds that landed on a tailings pond while it was activated with the proportion that landed during two other treatments: a continuous, randomly activated, deterrent system, and control periods with no deterrents. We also assessed the efficacy of different stimuli types within the on-demand system.
3Across several bird guilds, only the on-demand deterrent system significantly reduced the probability of birds landing in comparison with the control treatment. In addition to treatment effects, birds were more likely to land earlier in the spring and when they flew at lower altitudes, and shorebirds were more likely to land than ducks, geese and gulls.
4The comparison of stimuli revealed that cannons elicited significantly more response by birds in flight than mechanized peregrine falcon effigies with speakers broadcasting peregrine sounds.
5Synthesis and applications. Our results promote the use of on-demand systems for waterfowl deterrence at tailings ponds and recommend cannons over effigies as stimuli. We suggest that oil sands deterrence efforts should (i) be operational in the early spring, when tailings ponds appear to be most attractive to migrating waterfowl, (ii) target low-flying waterfowl and shorebirds and (iii) be effective during both day and night. These results and recommendations have potential application for problems of bird deterrence at several other industrial sites.
The problem of effective deterrence is particularly acute for mining industries (Allen 1990; MEM 1996; Read 1999), which often produce toxic tailings ponds. Waterfowl and shorebirds especially are attracted to freshwater ponds for foraging, roosting and nesting, and as stopover sites during migration. Spring migration is a particular problem in north-eastern Alberta, Canada, when warm-water effluent to tailings ponds from oil sands mines creates limited open water ponds while natural waterbodies are still frozen. When waterfowl land in these ponds, they may ingest oil and their plumage may become oiled with waste bitumen, potentially preventing birds from flying or leading to lost insulating capability and death from hypothermia (Golder Associates 2000). Shorebirds may also land along shorelines where oil collects, after perceiving these areas as mudflats. Several hundred birds appear to be oiled in a typical year at individual ponds (Gulley 1980; Van Meer & Arner 1985) and there are more than 10 tailings ponds operating in the north-eastern Alberta region (Golder Associates 2000; R. A. Ronconi, personal observation). The problem of oiled birds is already large and is likely to grow for two reasons. First, Alberta's oil sands are situated along a major migratory flyway (Bellrose 1976; Hennan & Munson 1979) for waterfowl travelling to the Peace-Athabasca Delta, an internationally significant staging area (Hennan 1972; Pollard et al. 2000). Secondly, oil sands development is certain to grow in the 21st century in Alberta, which may contain the second largest oil reserves in the world (CAPP 2003). Existing tailings ponds range in size from 150 ha (Boag & Lewin 1980) to roughly 3000 ha (Yonge 1981) and already provide the largest water body in this part of the migratory flyway (Golder Associates 2000).
Early efforts to deter waterfowl from landing in oil sands tailings ponds used human effigies (Boag & Lewin 1980; Gulley 1980). In subsequent years, propane cannons were introduced as additional deterrents. Habituation to these visual and auditory deterrents is likely to occur because migratory birds encounter ponds (and deterrents) along migration routes repeatedly within and among seasons. Recent technological advances offer promise to reduce habituation whilst maintaining the efficacy of deterrent systems. These advances pair automated detection of incoming birds via radar with a demand-based deployment schedule for aversive stimuli, thus the frequency with which birds experience the stimulus is reduced and habituation is less likely. Such a system has been successful at deterring migratory waterfowl from landing on contaminated ponds elsewhere (Stevens et al. 2000; C. Johansson, personal communication). Our objectives in this study were to compare (i) the efficacy of two deterrent systems, including a current industry-standard and a new radar-activated system; and (ii) the efficacy of the stimulus type used within these two systems.
The study area (Fig. 1) was situated at the Muskeg River Mine (MRM) site, operated by Albian Sands Energy Ltd, 75 km north of Fort McMurray, Alberta, Canada. The study was undertaken from 10 April to 30 May 2003, when the mine was in late developmental and early bitumen production phases. The tailings ponds complex (Fig. 1) consisted of three ponds: the two smallest ponds had combined dimensions of approximately 180 ha (1 × 1·5 km) and the main pond was roughly 350 ha (or 1·6 × 2·3 km) during the study period. During this early production stage, oil extraction processes were not running at anticipated capacity, thus reducing bitumen effluent and creating safer conditions for use of the control (no deterrents) treatment for deterrence experiments.
monitoring and deterrence equipment
Bird activity on the ponds and migration patterns were monitored by combining visual observations via binoculars with radar images that were interpreted simultaneously. The radar was located at the tailings ponds and radar images were observed between 3 and 29 May. During this period, we recorded the numbers of bird groups travelling over the ponds, maximum distances at which birds were detected, and direction of bird travel over a 1·85-km radius. Visual observations quantified flock sizes, identified species or guilds, and determined the number and approximate locations of bird landings on tailings ponds for as many of the radar targets as possible.
Two categories of bird deterrent systems were compared with control periods when no deterrents were used. The first standard type consisted of human effigies and propane cannons (Zon Mark III Plus, Margo Supplies, Alberta, Canada) set to fire continuously on various time intervals to resemble random firing. The second type consisted of an on-demand radar-activated system (BirdAvert™, Peregrine Systems, Salt Lake City, Utah, USA) comprising a marine radar (Furuno 1942 Mark 2, 1·2 m antenna, 4 kW output, 9·410 GHz, X-band) linked to a computer station. Incoming birds detected by the radar were automatically interpreted by the computer, which then activated deterrents via radio signals. The associated software included options to customize deterrence strategies for a variety of specific conditions, such as the direction and size of target stimuli and the schedule and chronology of deterrent deployment (for customized software see http://www.birdavert.com).
In our experiment, deterrent stimuli of the on-demand system included six floating and two shore-based platforms (Fig. 1) each equipped with solar panels, batteries, a peregrine falcon effigy with flapping wings, speakers broadcasting peregrine calls, a high-intensity strobe light and a propane cannon (Ronconi et al. 2004). Although distress and alarm calls of target species are frequently used as audio-deterrents (e.g. gull calls; Gosler, Kenward & Horton 1995), predator calls were used in this study because peregrines are potential predators of several target guilds under investigation (e.g. ducks and shorebirds). Because the tailings ponds were large, we confined our equipment to a small section (c. 50 ha) of the pond (Fig. 1) to (i) maintain accuracy of visual observations quantifying bird landings and reactions to deterrents; (ii) ensure that equipment was placed at high-enough densities to elicit bird responses; and (iii) approximate industry standard spacing (Golder Associates 2000) of current oil sands deterrents. Identical spacing was used for both deterrent systems, with floating and shore-based deterrent platforms spaced, on average, 297·3 m apart (SD = 74·8, range 194–420 m, or one per 8 ha). This spacing was a higher density, and thus more conservative, than the 500-m (one device per 13 ha) industry-standard spacing used for floating deterrents. Shore-based industry-standard spacing (one cannon and human effigy per 200 m) was achieved with the addition of an extra cannon and effigy used during industry-standard treatment periods.
Experiment 1: efficacy of deterrent systems
Throughout peak migration in May, deterrent efficacy was tested under a temporally stratified schedule. Testing was divided into a regular rotation of 72-h research blocks each containing three 24-h treatment blocks. Each 24-h block was randomly assigned to (i) an on-demand deterrent system, (ii) an industry-standard deterrent system or (iii) no deterrent system (control). The efficacy of these deterrents was assessed by comparing among treatments the proportions of birds flying over the study area that also landed within the study area; landings were considered as negative responses (no deterrence) and birds flying over the study area but not landing as positive responses. Because birds tended to fly in groups, we used groups of birds within a guild as the unit of statistical replication. We did not include in our analyses observations of birds for which flights began and ended within the tailings ponds because our study was intended to assess the response of migrating birds to deterrents and resident birds are generally harder to deter than migrants (Sema 1976).
Observation periods typically lasted 4·5 h beginning either 0·5 h before sunrise or 4 h before sunset, and treatment types were switched at midday. Observation periods were restricted to daylight hours because we needed to determine visually whether birds landed or not to assess the efficacy of our treatments. This daytime protocol did not provide a good test of lights as deterrents during the day or of any deterrents at night. Nevertheless, we believe it provided a relevant test of deterrent needs at tailings ponds because long-distance migratory flights in this region peak at night (Blokpoel 1973; Berthold 1993) and typically end in the early morning (Richardson & Gunn 1971) when birds are most likely to land.
During on-demand treatments, deterrent equipment was activated remotely by an observer who simulated automatic detection in the radar booth while a second observer recorded bird observations. By simulating an automated on-demand system in this way, observers had sufficient time to spot birds before activating deterrents in order to assess pre- and post-deterrence reactions. During industry-standard treatments, we covered the peregrine effigies of the on-demand system with camouflage, erected human effigies on the adjacent shore and on each floating platform, and programmed the cannons to fire continuously at various intervals, comparable with the protocols of active mines in the area. During control periods, deterrents were inactive and platforms were camouflaged.
Experiment 2: efficacy of deterrent stimuli types
In our first experiment, the on-demand system combined three types of stimuli: cannons, peregrine models with calls and strobe lights. In late May, using the manually activated on-demand (but simulated) approach (above), we compared (i) on-demand cannons and (ii) on-demand peregrine effigies with calls in alternating day-long treatment periods. For this experiment, we used a dependent variable of in-flight bird reactions (below), which made it impossible to observe reactions to the stationary, soundless, deterrent provided by human effigies. All groups were watched before, during and after deterrence, and responses were categorized as a binary variable of 0 (no reaction) or 1 (reaction). The following types of reactions were recorded as positive responses to deterrents by birds in flight: startle response (change in flight pattern or break in group formation), sudden increase or decrease in altitude, sudden increase in speed, and abrupt change in direction.
environmental and ecological variables
As part of both experiments, several environmental and ecological variables were measured. Because weather conditions can affect migratory bird movements (Richardson 1978; Elkins 1983), we collected standard weather data (temperature, wind speed and direction, precipitation) from an on-site weather station at the mine. We also collected information on bird group size, species, guild, flight direction, flight altitude (< 100 m or > 100 m), time of day and date. We noticed a profound change in migratory volume part way through our experiment, and so we coded date as a categorical variable for two periods, 3–6 May and 7–29 May. Use of tailings ponds by ducks and other waterfowl peaked on 6 May, after which pond use remained below 17% of those peak numbers. Circular direction data for wind and flight paths were coded on a linear scale from 1 to 5 that represented five classes along the north–south axis to reflect that fact that northerly headwinds were least favourable (coded 5) for north-bound spring migrants. Flight direction was also coded in five classes along the north-west–sout-east axis, because north-west was the predominant flight direction and the most likely direction of migratory travel.
Logistic regression was used to analyse the binary response variables of both experiments. Analyses were restricted to bird guilds potentially vulnerable to oiling (Anatidae, ducks, geese/swans; Charadriidae, shorebirds; Laridae, gulls) and with sample sizes > 30, thus excluding terns (Laridae, n= 7), cranes (Gruidae, n= 14), herons (Ardeidae, n= 5), cormorants (Phalacrocoracidae, n= 2), loons (Gaviidae, n= 4) and grebes (Podicipedidae, n= 1).
We built each of the two logistic regression models with several steps (after Hosmer & Lemeshow 1989) using SPSS 11·5 (2002). In brief, these steps included (i) identifying potential main effects with liberal, univariate tests, (ii) building a main-effects model using likelihood-ratio tests, (iii) testing the significance of biologically plausible two–way interactions by adding them one at a time to the main-effects model, and (iv) fitting a final reduced model. Prior to building the main effects model, we tested for confounding effects by examining changes in beta coefficients (retaining variables that changed the beta coefficients of other variables by more than 20%) and compared the performance of linear (x) and quadratic (x2 + x) terms using a likelihood-ratio test (Hosmer & Lemeshow 1989). We assessed model significance with the deviation chi-square statistic (-2LL) and model-fit with Nagelkerke's r2 statistic and the Hosmer & Lemeshow (1989) test. Both predicted probabilities and odds-ratios were used to interpret the logistic regressions.
bird activity and radar performance
We observed more than 16 000 waterfowl flying over the tailings ponds during spring migration (Table 1), with 7914 individuals (773 groups) observed during the experimental periods. Although higher proportions of landings were observed in April, more individuals landed in May, when the bird deterrent experiments were conducted. The highest numbers and highest proportions of landings were for ducks (536 individuals and 50–52% landing) and shorebirds (444 and 18–63%) compared with other guilds (42 and 0–18%). Overall, mean daily duck numbers on the pond were 21·0 ± 6·3 SE and ducks were prevalent on the tailings ponds for 2 weeks before 7 May, while numbers remained consistently low after.
Table 1. Numbers of waterfowl observed at the Muskeg River Mine tailings ponds, prior to and during deterrent experiments in 2003
Numbers observed during survey periods
18–27 April: 37·0 survey h prior to deterrent experiments
3–29 May: 97·4 survey h during deterrent experiments
10 April−30 May
Loons, grebes, cranes, herons, cormorants and coots.
Between 3 and 29 May, the radar detected 3309 bird groups, while observers were only able to detect 615 (19%) of those groups. Of those groups identified, 72·8% were large waterfowl (terns and larger), 19·2% were shorebirds, 4·4% were passerines and 3·6% were large non-waterfowl (e.g. ravens). The mean distance of maximum radar detection was 1·19 km (± 0·36 SD, n= 38, range 0·5–1·9 km). A logarithmic regression showed detection distance increased with group size (F = 11·34, d.f. = 40, P= 0·002, R2 = 0·22). The mean maximum detection distances were 1·15 km for ducks (n = 20), 1·47 km for geese (n = 3) and 1·20 km for gulls (n = 15), but these values did not differ significantly (F = 0·204, d.f. = 1, P= 0·654).
experiment 1: efficacy of deterrent systems
On average 1·27 bird groups per hour (SD = 1·49, range 0–7) landed in the study area during the first experiment. Initial univariate analyses revealed the following variables as producing significant differences (at P≤ 0·25) in the proportion of birds landing: treatment type, observation period (pre- vs. post-peak migration), wind speed, wind direction, bird guild, flight direction and flight altitude. The final model, based on 372 observations of bird groups (Table 2; −2LL = 125·61, model χ2 = 46·72, d.f. = 8, P < 0·001), showed that bird landings were significantly less likely (P = 0·003) with the on-demand treatment than the control (Fig. 2). The industry-standard treatment produced an intermediate response (Fig. 2) that did not significantly differ from the control (P = 0·278) or (measured by changing the reference category) the on-demand system (Table 2; Wald = 1·408, d.f. = 1, P= 0·235). Although the industry-standard appeared to be effective in reducing the probability of duck and goose landings relative to the control (Fig. 2), the interaction between treatment and guild was not significant (Wald = 0·081, d.f. = 6, P= 1·00). Together, these results suggested that the radar deterrents worked well for most guilds but the industry-standard deterrents were not effective for shorebirds and showed inconclusive (i.e. intermediate) effects for ducks.
Table 2. Final results of logistic–regression models for two bird deterrence experiments at the Muskeg River Mine tailings ponds. Significant variables are indicated in bold
Reference category for two-way comparisons were as follows: *control, **industry-standard, ***ducks ,****cannons.
Period defined by peak (date of highest waterfowl count on tailings pond): pre-peak (3–6 May), post-peak (7–29 May).
Observation period (pre- vs. post-peak in pond use)†
Interaction (observation period by flight direction)
ON-DEMAND DETERRENT STIMULI EXPERIMENT
Deterrent type (cannons vs. peregrine effigies)****
Four other effects were liberally significant (P ≤ 0·1) in the final model. First, birds flying at altitudes greater than 100 m were between 4·6% and 12·8% (depending on treatment type) less likely to land (Table 2; P= 0·083) than birds flying at below 100 m. Secondly, shorebirds were more likely to land than ducks (Table 2; P= 0·014), but ducks, geese and gulls did not differ in their propensity to land (Table 2). Thirdly, observation period (defined above) affected the probability of landing (Table 2; P= 0·004) whereby ducks, geese and shorebirds (Fig. 3) were more likely to land before 7 May than after. Finally, there was a significant interaction between observation period and flight direction, which showed that landing probability generally increased with non-migratory flight directions (e.g. south-westerly flights) in the early spring.
The final model describing the efficacy of deterrent systems provided a very strong fit to the data (Nagelkerke's r2 = 0·319, Hosmer and Lemeshow goodness-of-fit test = 2·754, d.f. = 8, P= 0·95). Because small numbers of gull and goose observations might obscure industry-standard treatment effects (Fig. 2), the analysis was rerun without these guilds. This slightly increased the significance of the on-demand treatment (P = 0·002) but the industry-standard treatment remained non-significant (P = 0·131).
To facilitate interpretation of our final model, we expressed our model-derived probabilities (Fig. 2) as odds ratios (Hosmer & Lemeshow 1989). These showed that, for ducks, the odds of landing during the on-demand treatment were 53% of those for the control, whereas the odds for shorebirds landing during the on-demand treatment were only 20% of those of the control. Comparable odds ratios for the industry-standard treatment relative to the control were 38% and 69% for ducks and shorebirds, respectively. These values suggested that the preferable performance of the on-demand system was caused by the substantially reduced odds for shorebirds landing. They also suggested that the industry standard was markedly more effective for ducks than shorebirds.
experiment 2: efficacy of deterrent stimuli types
Deterrent type did affect bird responses (Wald = 5·741, d.f. = 1, P= 0·017) and no confounding variables or significant two-way interactions were identified. The final model (Table 2; –2LL = 59·449, model χ2 = 6·858, d.f. = 1, P= 0·009, Nagelkerke's r2 = 0·164) showed that the odds of birds changing flight pattern was 5·6 times greater when cannons were used as a deterrent stimulus instead of peregrines. As for the first experiment, the model was qualitatively unchanged by the exclusion of gulls, which seldom landed on the pond (Table 1).
More information about stimulus efficacy was provided by comparing specific bird responses. Among the 28 peregrine activations used in the analysis, only three bird responses were observed (11%): one flushing event by a group that had already landed and two startle reactions by groups in flight. In contrast, cannons elicited reactions during 12 of their 30 activations (40%) and often produced more than one reaction type per event: two flushes by landed groups and in-flight reactions including four startles, two increases in speed, six changes in altitude and five direction changes.
patterns of bird activity and limitations of radar technology
Ducks and shorebirds were most susceptible to oiling because they landed in tailings ponds more often than other guilds. Our data also showed greater proportions of landings in early spring (April), but in May higher absolute volumes of migratory birds occurred and more birds landed on the ponds, comparable with previous studies in this region (Schick & Ambrock 1974; Hennan & Munson 1979). Together, these broad patterns suggest an early spring dependence on tailings ponds, perhaps because many natural waterbodies in the region are still partially frozen at that time (Gulley 1980). However, the increase in absolute numbers later in the spring suggests that birds are still vulnerable to being oiled then.
Assessment of radar data with visual sightings showed both advantages and limitations of radar use as part of an on-demand deterrent system. Radar was able to detect four times as many birds as visual sightings and was able to detect birds at night, which is particularly critical for bird deterrence because shorebirds, ducks and geese are nocturnal as well as diurnal migrants (Berthold 1993). Unfortunately, none of our experiments tested deterrent efficacy or bird behaviour at night. Another advantage of the radar was that it adequately detected desired target bird groups for deterrence. Visual sightings confirmed that more than 90% of radar detections were of target species (shorebirds and waterfowl). However, radar detection distances (mean maximum of 1·19 km, this study) could not encompass all pond diameters in the region. The tailings ponds in this study were more than 3 km in combined length yet smaller than other tailings ponds in the area. Despite this limitation, radar offers much greater detection ranges than visual observation (this study), infra-red motion detection (Jacobson et al. 1997) and thermal infra-red detection (Focardi et al. 2001).
experiment 1: efficacy of deterrent systems
Our analysis showed an overall and significant effect of on-demand deterrents in reducing bird landings on tailings ponds. We detected no similar decline for the industry-standard treatment as a main or interaction effect, despite the apparent reduction in landing frequency, predicted probabilities and odds ratios for ducks relative to other guilds (Fig. 2). Stevens et al. (2000) likewise found on-demand deterrents successful relative to unguarded ponds. Our failure to detect an effect of the industry-standard, or an interaction between guild and treatment, may have been a function of inadequate samples size or the equitability of sample size between our dichotomous response categories (Hosmer & Lemeshow 1989). Nevertheless, the overall model robustly supports the effectiveness of the on-demand system across guilds (Fig. 2; ducks, geese and shorebirds) and our interpretation of the industry-standard treatment is that it appears to reduce landing probability for ducks but not shorebirds (Fig. 2; ducks only).
In addition to treatment and guild effects, birds were significantly more likely to land earlier in the season and when they were flying at lower altitudes, and both of these variables have potential management implications. The observation periods (pre- vs. post-peak in pond use) showed that landing probability was higher in early spring, making it imperative to have bird deterrents fully functional then. The interaction between observation period and flight direction suggests that it will be more important to deter southward-flying birds in the early spring, as these birds are more likely to land. These birds are more likely to be residents and hence the system may be less effective at deterring them.
Our final model also showed that birds flying higher than 100 m were less likely to land. Future deterrent efforts may wish to target low-flying birds, thus limiting on-demand deterrent activation to the individuals that are most likely to land in ponds. Thus detection systems would benefit from increased horizontal, but not vertical, range, and may additionally benefit from deterrent stimuli with long-distance horizontal ranges (e.g. rotating spotlights; Read 1999) or placement in susceptible areas (Somers & Morris 2002). Weather did not affect landing propensity in this study. Because weather is known to affect migration behaviour, including altitude of migration (Richardson 1978; Elkins 1983), it may merit further study in the context of deterrents.
experiment 2: efficacy of deterrent stimuli types
Consistent with earlier work by Boag & Lewin (1980), cannons were more effective than peregrine effigies with calls. The lack of response to peregrine calls may, however, not be surprising because peregrines only call during the breeding season (Ratcliffe 1993). Therefore ducks and shorebirds may not have associated the peregrine calls with the threat of predation and other calls (e.g. distress calls of ducks) may present better auditory stimuli.
In other experiments, stimuli types are frequently confounded. For example, Stevens et al. (2000) tested multiple stimuli in an on-demand system (pyrotechnics, distress calls and aerosol repellents) but component effects were not separated. Similarly, we cannot speculate about the effectiveness of human effigies in our study because they were not tested in experiment 2 and they were confounded with cannon stimuli in experiment 1. Earlier research at tailings ponds (Boag & Lewin 1980) concluded that human effigies were effective, but studies elsewhere showed that they are prone to habituation effects (Stickley, Mott & King 1995; Andelt, Woolley & Hopper 1997). Nevertheless, it can be inferred from our results that human effigies in association with randomly firing cannons do not significantly decrease landing probabilities. Equally important to the type of stimuli is the spacing with which they are deployed (Ward 1978; Sharp 1987). While it was not feasible to manipulate spacing in our experiments, the responses to the on-demand treatments suggest that the 300-m spacing used was effective.
The issue of habituation may explain why numerous other studies have shown reduced effectiveness over time for both cannons and other auditory deterrents (Spanier 1980; Sharp 1987; Bomford & O’Brien 1990; Koski, Kevan & Richardson 1993; Stickley, Mott & King 1995; Andelt & Hopper 1996; Andel, Woolley & Hopper 1997). Bomford & O’Brien (1990) suggested that auditory deterrents were most effective when activated at random intervals, and when sounds were supported with visual deterrents, although our results showed that cannons do not necessarily need associated visual deterrents to elicit bird reactions. Rather, activating them in response to bird approaches and synchronizing cannon firing to create more intense sound stimuli may have contributed to the effectiveness of the on-demand treatment. This approach minimized pre-exposure to cannon sounds as birds approached the ponds and made it possible for the birds to associate their activity with the deterrents (sensuShettleworth 1998). It may have also increased the apparent threat of the deterrents, which is also likely to have increased the responses of the birds to them (Beale & Monaghan 2004).
Our study shows promise for radar-activated on-demand deterrents over existing industry standards in the oil sands, particularly for shorebirds, and demonstrates that cannons are more effective deterrent stimuli than human effigies at this site. Some recommendations and limitations of our study and its interpretation are apparent. First, industries operating in northern latitudes should ensure that deterrents are operational in early spring when the surrounding natural waterbodies are frozen and early migrants may be most vulnerable. Secondly, although radar accurately detects target bird groups (ducks, geese and shorebirds), maximum detection ranges may be too low for very large tailings ponds.
Profitable lines of future research would include effective spacing for floating deterrents, the efficacy of deterrents at night, the range and sensitivity limits of radar detection, and the percentage of flocks evaded by radar detection. In particular it may be most important to determine how landing probabilities vary temporally among guilds and with flight altitude, so that deterrence strategies can be customized to target those birds that are most likely to land. Successful bird deterrence from ponds may also be contingent on the availability of nearby alternative waterbodies (Gosler, Kenward & Horton 1995; Stevens et al. 2000). Tracking individual birds with radio-telemetry (Béchet, Giroux & Gauthier 2004) may provide a better means of assessing the effects of disturbance in the oil sands region.
Although we have demonstrated the efficacy of a deterrent system, bird deterrence is not the long-term solution. In addition to deterrence, the oil sands industry is committed to the reclamation of mines and tailings ponds post-production (10–20 years at this site). Moreover, the industry is developing ‘dry-tailings’ processes that will negate the need for hazardous tailings ponds, although such technology is likely to be at least 10 years away. In the meantime, on-demand cannon deterrent systems offer the potential of better avian deterrence at industrial sites. As the oil sands industry continues to expand and develop, such systems could considerably reduce the long-term cumulative effects on waterfowl migrating through the region. More broadly, the on-demand approach to avian deterrence has potential applicability for deterrence in other settings, including airports, oil spills, aquatic and terrestrial farms, wind farms and other mining and industrial facilities (Ronconi et al. 2004).
Logistical support and funding was provided by Albian Sands Energy Ltd, Canada Foundation for Innovation, Syncrude Canada Ltd and the University of Alberta. We thank Darrell Martindale for thoughtful support throughout. We are grateful for field assistance from Andrew Forrest and Sarah Wong, for technical support from Will Neagle and Peregrine Systems Inc., and for assistance from Albian employees C. Theriault, M. Kay, B. McKenzie, T. Einarson, B. Kean and Tailings and Site Services Teams. We thank L. Foote, J. Gulley, C. Johansson, C. McParland, C. Paskowski, J. Thompson and C. White for advice and/or proposal reviews, and three anonymous referees. Scientific permit no. CWS03-A002 was issued by the Canadian Wildlife Service.