2013 NORTHEASTERN NATURALIST 20(1):1–18 Assessing Efficacy of Non-Lethal Harassment of Double-Crested Cormorants to Improve Atlantic Salmon Smolt Survival James P. Hawkes1,*, Rory Saunders1, Adam D. Vashon2, and Michael S. Cooperman3 Abstract - Salmo salar (Atlantic Salmon) smolts are exposed to predation pressure as they migrate from freshwater into the estuary and near-shore marine environment. In particular, Phalacrocorax auritus (Double-crested Cormorants) are a predator of Atlantic salmon smolts during their estuary and near-shore migration. National Oceanic and Atmospheric Administration’s (NOAA) National Marine Fisheries Services’ (NMFS) telemetry data collected prior to this study (1997–2003), suggest that smolts are being removed from the Narraguagus River on their downstream out-migration. This removal may be the result of Cormorant predation. We investigated whether smolt survival could be improved by disrupting normal Cormorant foraging activity by integrating passive smolt tracking and active harassment techniques. Smolt movement and usage of various portions of the estuary according to light condition and tidal stage were explored along with concurrent avian harassment. Although harassment only occurred in approximately 33% of available daylight hours during this study, the impacts were easily recognized. Non-lethal harassment effectively displaced Cormorants from feeding locations and reduced loss of emigrating smolts. In 2004, 83.3% (15 of 18) of all smolt mortalities occurred on days of non-harassment, compared to only 16.7% (3 of 18) on days when harassment occurred. Similarly in 2005, 87.5% (7 of 8) of all smolt mortalities occurred on days of non-harassment, compared to only 12.5% (1 of 8) on days when harassment occurred. Non-lethal harassment appeared to be an effective means to reduce loss of emigrating smolts in the Narraguagus River estuary. Introduction Salmo salar L. (Atlantic Salmon) populations in the United States are near historic low abundance (USASAC 2011). The Gulf of Maine Distinct Population Segment (GoM-DPS) represents the last remnant native stock in the US and is listed as endangered under the US Endangered Species Act (Fay et al. 2006; 74 Federal Register 29344, 19 June 2009). The freshwater habitats of Maine’s salmon rivers have been degraded by a variety of land-use practices, chemical pollution, and introduction of exotic species; and these changes have contributed to the decline of Atlantic Salmon (NRC 2004). Despite years of intensive efforts and freshwater habitat restoration coupled with population enhancement via conservation hatcheries, populations within the GoM-DPS remain low (USASAC 1NOAA, National Marine Fisheries Service, Maine Field Station, 17 Godfrey Drive, Orono, ME 04473. 2United States Department of Agriculture, Animal and Plant Health Inspection Service Wildlife Services, 79 Leighton Road, Suite 12, Augusta, ME 04330. 3Conservation International, 2011 Crystal Dr. Suite 500, Arlington, VA 22202. *Corresponding author - James.Hawkes@noaa.gov. 2 Northeastern Naturalist Vol. 20, No. 1 2011). This lack of recovery of the GoM-DPS is not surprising given the continued reduction in marine survival of Northwest Atlantic Salmon stocks described by Chaput et al. (2005). The seaward migration of Atlantic Salmon smolts can be a time of high mortality owing to interactions amongst legacy effects as a result of exposure to acid rain and aluminum during rearing (McCormick et al. 2009), unique physiological challenges associated with the transition from freshwater to saltwater, and exposure to abundant and novel predators (Beland et al. 2001, Blackwell and Juanes 1998, Dieperink et al. 2002, Järvi 1989, Martin et al. 2009). Identifying and potentially managing sources of smolt mortality within estuaries are high priority research and conservation needs that could improve the effectiveness of recovery measures through adaptive management practices. Knowledge of the ecology and migratory behavior of Atlantic Salmon smolts within estuaries is limited but increasing. Typically, estuarine migration has been described as ebb-tide transport, whereby smolts move seaward on receding tides and hold their location on flood tides (Lacroix 2008, McCleave 1978). Recent studies have demonstrated that appreciable numbers of smolts make pronounced upstream “reversal” movements during flood tides (Kocik et al. 2009, Martin et al. 2009). Unlike in freshwater, where smolts typically move downstream at night presumably to avoid visual predators (Martin et al. 2009, Moore et al. 1995, Ruggles 1980), smolts within estuaries have been observed moving at all hours of the day (Kocik et al. 2009; LaCroix et al. 2004; Martin et al. 2009; Moore et al. 1995, 1998). It is unclear if different estuarine behaviors, such as migration during day or night, on flood or ebb tides, or rapid transit of the estuary versus prolonged estuary residency, result in differential survival. The Narraguagus River, ME supports a naturally reproducing Atlantic Salmon population, which is supplemented by a river-specific, multi-life stage juvenile stocking program, designed to recover the natural population. This population is severely diminished compared to historic levels (USASAC 2011). Estimates of the number of smolts emigrating from this river over the past decade have ranged from fewer than 1000 to 2500 annually (USASAC 2011). Prior migration studies have demonstrated high smolt mortality during the transition from the river to the estuaries. Specifically, Kocik et al. (2009) reported almost 60% of 581 telemetry-tracked smolts died as they transited though the lower river, estuary, and inner bay of the Narraguagus system from 1997 to 2004. They observed only 10% mortality in the outer portions of Narraguagus Bay, despite smolts having comparable residency times in the two areas. Specific sources of mortality were not identified. Our post-study review of smolt telemetry-tracking data in Kocik et al. (2009) revealed strong spatial overlap between locations of presumed smolt losses and our observation of areas occupied by Phalacrocorax auritus Lesson (Double-crested Cormorant; hereinafter “Cormorants”). We hypothesized the primary cause of smolt mortality was predation removal by avian sources, particularly Cormorants. Cormorants are considered generalists, foraging on a variety of seasonally abundant fresh and saltwater fish and invertebrates from benthic and pelagic 2013 J.P. Hawkes, R. Saunders, A.D. Vashon, and M.S. Cooperman 3 habitats (Blackwell and Krohn 1997, Blackwell et al. 1995). Atlantic Salmon smolts are a significant prey item; in some cases, smolt losses to Cormorants can be exceptionally high. For example, Meister and Gramlich (1967) recovered from Cormorants rookeries 3724 tags applied to hatchery stocked smolts on the Machias River in 1966 and 1967, representing 4.9 and 8.1%, respectively, of fish released during these two years. These authors estimated that Cormorants consumed between 16 to 50% of stocked smolts during this period. Similarly, Krohn and Blackwell (1996) reported an estimated 7.5 to 9.2% (43,318 to 78,675) of hatchery-stocked smolts in the Penobscot River were preyed upon by Cormorants between the years of 1992 to 1994. Our goal was to test the efficacy of non-lethal harassment and d escribe smolt survival during harassment and non-harassment periods. Specific objectives were to describe 1) the distribution of Cormorants within the lowest portion of the Narraguagus River, 2) the relationship between Cormorant habitat use and Atlantic Salmon smolt behavior, and 3) the effectiveness of a pilot Cormorant harassment study. We discuss our findings in the context of management recommendations and future research needed for limiting smolt mortality in the estuary. Site Description The Narraguagus River is 70 km long with a drainage area of approximately 600 km2. The river flows southeast from its headwaters at 124 m above sea level to the town of Cherryfield, ME, where the estuary begins (Baum and Jordan 1982). Between Site A and the start of the estuary (above Site B; Fig. 1), the river is dynamic in composition. The upper third is a slow-moving pool separated by a dam; traveling further downstream, extreme shifts in gradient and habitat result in riffle and chutes with coarser substrate consisting of cobble, boulders, and ledge transitioning to the tidal portion of the watershed. The estuary is 6 km long, ranges from 50 to 120 m wide, has a mean depth at low water of 3 m and a mean tidal fluctuation of 3.4 m (Baum and Jordan 1982, Kocik et al. 2009). The tidal freshwater zone is relatively small, 260 ha, compared to the estuary-mixing zone totaling 1300 ha (Strategic Assessments Branch 1985). Much of the estuary exhibits brackish water characteristics, with salinity ranges observed between <1‰ in the upper reaches to as much as 26‰ before entry into the inner bay area (Baird Software 2004). Our visual surveys of the estuary found that the makeup is primarily fine sediments and sand in the upper reaches transitioning to constricting bedrock ledges, boulders, and moving sand bars in the lower portion. There are four islands in Narraguagus Bay, within 15 km of the mouth of the Narraguagus River, which supported ≈200 nesting pairs of Cormorants at the time of the study (Rookery survey–NOAA, Orono, ME, unpubl. data). Methods Smolt telemetry The telemetry methods used in our study were identical to those described in Kocik et al. (2009). Acoustic receivers (VR2 Telemetry Receivers; Vemco Ltd.) 4 Northeastern Naturalist Vol. 20, No. 1 were moored at sites throughout freshwater and the estuary in the Narraguagus system (Fig. 1). Owing to distinct differences in the habitat structure, salinities, Figure 1. Map of the study area and telemetry network on the lower Narraguagus River and Narraguagus estuary, ME. Receivers (gray circles) were deployed at the start of the study area at Site A. Traveling downstream, the first tidally influenced estuary unit which is above the salinity gradient is found at Site B. Continuing seaward, the salinity influence begins at Site C and increases in salinity along the remaining of the study array (Site J/K) before entering Narraguagus Bay. The upper estuary consists of sites B–D, and the lower estuary includes sites E–J/K. 2013 J.P. Hawkes, R. Saunders, A.D. Vashon, and M.S. Cooperman 5 and tidal currents in the estuary, we partitioned the estuary into upper (Sites B–D) and lower (Sites E–J/K) portions (Fig. 1). Additional telemetry arrays were moored outside our study area in Narraguagus Bay to determine the efficiency of receiver sites and the fate of tagged smolts. We used two 1.52-m rotary screw traps to collect emigrating smolts 7.65 km above head of tide on the Narraguagus River (Fig. 1). We trapped smolts from early April through early June in 2004 and 2005, effectively covering the entire date range of the spring smolt out-migration. We used smolt run timing and abundance records to deploy telemetry transmitters (Vemco Ltd. 69 kHz serial coded acoustic transmitters; 2004 = V9-6L, 20 x 9 mm, 3.3 g in air, 20–60 s random delay, 68 d lifespan, 2005 = V7-2L, 18.5 x 1.6 mm, 1.6 g in air, 20–50 s random delay, 74 d lifespan) in proportion to the anticipated daily abundance of the emigrating population (Kocik et al. 2009). Smolts were anesthetized using buffered tricaine methanesulfonate (MS-222), surgically implanted with a transmitter and released within a half-hour of surgery. We tagged smolts of 145 mm fork length and larger to minimize tagging affects. All smolts were released at the trapping site at river km 7.65. To minimize surgery effects, only smolts making it to Site A (River km 2.11) were included in this study. Detection of tagged smolts at Site A marked entry into the study area. The fate of each smolt was determined by detection at successive receivers. Successful smolt movement was defined as whenever the last detection at one VR2 location was followed by a subsequent detection, either upstream or downstream. Time of the movement was assigned as the time of arrival to the latter site. An unsuccessful movement (mortality), was defined as whenever a last detection at one site, excluding site J and K (the most downstream site in our network), was not followed by a subsequent detection at another site. When assigning mortalities, our assumptions were 1) transmitters do not malfunction, 2) transmitters no longer detected (removed from the system) or which were stationary in the river had been depredated, and 3) last detection of a smolt deemed “unsuccessful” was the time which mortality occurred. Each individual smolt could have numerous successful movements, but only one unsuccessful movement. We classified each smolt movement as either day or night based on the timing of civil twilight (i.e., limit of terrestrial objects to be clearly distinguished) as reported by the US Naval Observatory (aa.usno.navy.mil) for Milbridge, ME and as flood or ebb based on tide data obtained from mobile graphics (mobilegeographics.com) for Milbridge, ME. Cormorant distribution and harassment During the smolt migration of 2004 and 2005, we documented the presence of Cormorants in the Narraguagus estuary and conducted non-lethal harassment activities. From 21 April through 25 May 2004, enumeration and harassment occurred on weekdays, and involved one person. From 9 May through 27 May 2005, enumeration and harassment occurred during alternating four-day periods and involved two people, one each in the upper and lower estuary. In the freshwater and upper estuary portion of the study area, Cormorant enumeration and harassment occurred eight hours per day, between 6 Northeastern Naturalist Vol. 20, No. 1 the dam and just below the head of tide (at approximately Site C; Fig. 1). In the lower estuary, harassment effort was variable. In 2004, hazardous water levels and difficult access limited Cormorant observation and harassment in the lower estuary to a three-hour period around the daytime high tides. We conducted Cormorant harassment for nearly eight hours each day in 2005. Observation and harassment involved researchers patrolling the study area on foot or by motorboat. A consistent daily patrol pattern was employed. Activities began at the top of the study area, with movement seaward, while the remainder of the day consisted of up and downstream movements for the duration of harassment. During each event in which Cormorants were observed, time, location, and number of Cormorants present were recorded. We were unable to identify individuals; therefore, these data represent Cormorant movement patterns, with individuals likely sighted several (unknown) times. The specific method used for each harassment event was selected on a case-by-case basis: approaching by foot or boat, throwing rocks, or using slingshot, lasers (Avian Dissuader®, SeaTech, Albuquerque, NM) or pyrotechnics (Bird Bombs® and Bird Whistlers®, Sutton Agricultural Enterprises, Inc., Salinas, CA; and Shellcrackers®, Stoneco Inc., Trinidad, CO). We did not use pyrotechnics earlier than 0900 hours or in areas where Haliaeetus leucocephalus L. (Bald Eagle) were present. Harassment activities for each event took less than 30 seconds, and Cormorants were observed leaving the site at the completion of harassment. We considered the different methods to be equally effective and, for analysis purposes, we grouped them together as “harassment”. The location and frequency of harassment events were determined by Cormorant presence at the time of harassment patrol. Hence, the number of harassment events conducted each day and at each location varied. We complemented Cormorant data collected via point counts with time-lapse photography to document diel behavior and daily use patterns during treatment and control periods. In 2004, we deployed one camera at an isolated location with a broad vantage point and frequent Cormorant use (Fig. 1). The camera was operated 13 May to 24 May (12 days) and took one image every hour during daylight hours, yielding 14 pictures daily. In 2005, we attempted to increase spatial coverage by adding an estuary site and temporal resolution of the images, but were not successful due to weather-related equipment failures. Smolt behavior and survival We determined the number of successful and unsuccessful smolt movements which occurred during each of the four possible light level and tide direction combinations (day–ebb, day–flood, night–ebb, night–flood) and plotted each movement as a function of these variables. For successful and unsuccessful movements, we summarized proportions of movement during light condition (day or night) and tide direction (ebb or flood) on the frequency of movements. Overlap of Cormorants and smolts We explored the overlap in the timing of smolt movements and Cormorant presence by plotting the mean (±1 SD) number of first smolt detections (i.e., 2013 J.P. Hawkes, R. Saunders, A.D. Vashon, and M.S. Cooperman 7 indicative of arrival to a site) that occurred at each hour of the 24-hour daily cycle. For example, for the 0900 values during the 2004 effort, if smolts moved through the upper estuary over a 30-day period and there were five sites within the upper estuary, we determined how many first detections occurred at any of the five receivers between 0830 and 0930 hours on each of the 30 days and calculated summary values for the 0900-hour range. We then overlaid a plot of the mean (±1 SD) number of Cormorants present in photographs on days of no harassment efforts at each hour of the daily cycle. Effectiveness of Cormorant harassment We used one-way ANOVA to test for differences in daily count of Cormorants present in 2004 photographs on days with and without harassment. We also used the hourly Cormorant count data from the 2004 photographs to plot the mean (±1 SD) number of Cormorants present at each hour of the day on both harassment and non-harassment days. To evaluate the potential for either additive or declining effectiveness of repeated harassment events within a day with a count of Cormorants, we fit an inverse first order polynomial regression of the form y = yø + (a / x) + (b / x2), where y is the number of Cormorants present within a 1-km section of the river located between Sites A and B, yø is the number of Cormorants present at the time of the first harassment event of the day, and x is the number of the harassment events within a day. For example, on the first day of Cormorant harassment in 2005, we conducted 13 unique harassment events between Sites A and B, and for each of these 13 events we counted the number of Cormorants present before each harassment event started. On the second day, we conducted 27 unique harassment events in this area, and so on for each day of the year’s field effort. We then calculated the mean (±1 SD) number of Cormorants present prior to the first harassment event, prior to the second harassment event, etc., up to the largest number of harassments (i.e., 46 events within a single day), and we then fit a curve to the resulting plot. In an effort to determine if harassment of Cormorants affected smolt mortality, we plotted the date and time of each unsuccessful movement event and overlaid a plot of the timing of harassment activities and calculated the percentage of unsuccessful movements occurring during harassment and non-harassment periods. Results Smolt telemetry In 2004, the mean (±1 SD) fork length (mm) and weight (g) of the 66 smolts in the study was 174.9 mm (14.1) and 53.0 g (13.6), respectively. In 2005, the mean fork length and weight of the 61 smolts in the study was 169.0 mm (11.8) and 47.7 g (11.0), respectively. As a comparison, smolts collected by rotary screw traps for abundance estimates during this period in the Narraguagus had a mean fork length (n = 569) of 170.7 mm (14.4) and a mean weight (n = 569) of 49.1 g (13.5). Therefore, the mean length and weight of the smolts in the telemetry study was representative of the smolt population in 2004 and 2005. For the two years combined, transmitter weight (wet weight; g) averaged 6.2% (V9-6L tags) or 8 Northeastern Naturalist Vol. 20, No. 1 3.4% (V7-2L tags) of mean smolt weight, and smolt survival was not related to the ratio of smolt to transmitter weight. Movement within the upper (Sites B–D) and lower (Sites E–J/K) segments of the estuary were different. In the upper estuary, approximately 80% of all successful smolt movements occurred between 2100 and 0400 hours (Fig. 2A, C). In 2004, smolts successfully moved at night regardless of tidal stage. Fifty percent (n = 144) of the successful movements occurred during ebb-night conditions, 31% (n = 89) during flood-night, 12% (n = 35) during ebb-day, and 7% (n = 20) during flood-day conditions. We observed a similar night-use pattern in 2005, but smolt movements were more common during ebb tides than flood tides. Forty-seven percent (n = 109) of the successful movements occurred during ebb-night conditions, 32% (n = 74) during flood-night, 14% (n = 33) during flood-day, and 7% (n = 16) during ebbday conditions. In contrast to the upper estuary, most movements in the lower estuary were associated with tidal stage rather than the light-dark cycle. In both years of the study, >72% of successful smolt movements occurred during ebb tides (Fig. 2B, D). In 2004, there was no indication of an effect of the light-dark cycle on smolt movements. Forty-two percent (n = 198) of the successful movements occurred during the ebb-night conditions, 30% (n = 140) during ebb-day, 16% (n = 76) during flood-night, and 12% (n = 56) during flood-day conditions. In 2005, smolt movements were influenced by ebb tide, but were also associated with darkness. Fifty-seven percent (n = 199) of successful movements occurred during ebb-night conditions, 21% (n = 72) during ebb-day, 14% (n = 50) during flood-day, and 9% (n = 30) during flood-night conditions. There were 20 unsuccessful smolt movements (mortalities) in 2004. Two of these mortalities occurred in freshwater at night. The remaining 18 smolt mortalities occurred in the estuary during daylight hours, with five mortalities in the upper estuary (ebb-day, n = 3; flood-day, n = 2) and 13 mortalities in the lower estuary (ebb-day, n = 10; flood–day, n = 3). Twelve of the 18 estuarine smolt mortalities (66.7%) occurred at or before 1000 hours. The mortality rate for smolts in 2004 was 30.3%. In 2005, there were 10 unsuccessful smolt movements. Two of these mortalities occurred in freshwater at night. All eight mortalities in the estuary occurred during daylight hours, with five in the upper estuary (ebb-day, n = 5) and three in the lower estuary (ebb-day, n = 2; flood-day, n = 1). Six of the eight smolt mortalities (75.0%) occurred at or before 1000 hours. The mortality rate for tagged smolts in 2005 was 16.4% within the study area. Overlap of Cormorants and smolts There was notable separation between the timing of peak smolt movements and peak Cormorant presence within the upper estuary. Most smolt movements occurred from 2100 to 0400 hours, while Cormorant presence peaked between 0600 and 1000 hours (Fig. 3). 2013 J.P. Hawkes, R. Saunders, A.D. Vashon, and M.S. Cooperman 9 Figure 2. The timing of both successful (circles) and unsuccessful (stars) smolt movements in (A) upper estuary 2004, (B) lower estuary 2004, (C) upper estuary 2005, and (D) lower estuary 2005, as a function of both the light-dark cycle and tidal conditions. Diagonal lines represent high (solid line) and low (dashed line) tide conditions and shaded areas are hours of low light levels or darkness. 10 Northeastern Naturalist Vol. 20, No. 1 Effectiveness of Cormorant harassment Harassment affected Cormorant behavior and use in the estuary. In 2004, there were a total of 344 daylight hours during this study period, in which 108 hours of harassment (approximately 160 person hours) were conducted, which represents approximately 32% of daylight hours. 679 harassment events displaced 8132 Cormorants (likely including individuals displaced multiple times) from the study area. In 2005, there were a total of 274 daylight hours during this study period, in which 104 hours of harassment (approximately 319 person hours) were conducted, which represents approximately 38% of daylight hours. 1054 harassment events displaced 9303 Cormorants. Activity patterns of Cormorants derived from the 2004 time-lapse photography indicate that fewer Cormorants were present on days of harassment compared to non-harassment days (one-way ANOVA of Cormorant counts: F[1, 9] = 9.37, P = 0.014) (Fig. 4 A). On non-harassment days at the fixed camera location, Cormorant activity peaked between 0600 and 1000 hours, with the hourly mean number of Cormorants (± 1 SD) observed always greater than 3.7 Cormorants, and with a peak of 9.5 (±7.3) observed at 0700 hours. On days with harassment efforts, subtle peaks in the number of Cormorants occurred at 0600 (mean: 2.6 ± 4.5) and Figure 3. Mean smolt arrival detections to receiver sites used to compare day-night patterns of successful smolt movement during a 24-hour period in the upper estuary, as well as the overlap of Cormorants during this period. Circles are mean number of smolt detections (± 1 SD). Squares are mean hourly Cormorant count (± 1 SD). Data are offset for easy viewing. 2013 J.P. Hawkes, R. Saunders, A.D. Vashon, and M.S. Cooperman 11 1100 hours (mean: 2.6 ± 4.8). Regardless of harassment intensity, the number of Cormorants observed after 1100 hours was consistently lower than earlier in the day, with a mean of 1.25 Cormorants observed per hour (Fig. 4 B). The number of Cormorants in the study area is correlated with cumulative harassment events (Fig. 5A, B) in 2004 (inverse first order polynomial regression: r2 adj. = 50.2%, P less than 0.0001) and 2005 (r2 adj. = 75.0%, P less than 0.0001). Smolts benefited from harassment during both years of this study. In 2004, there were a total of 251 successful and 18 unsuccessful within-stream movements of smolts during daylight hours. During periods of non-harassment, 9.0% (15 of 167) of all smolt movements were unsuccessful. During periods of harassment, only 2.9% (3 of 102) of all smolt movements were unsuccessful. Further, of all unsuccessful movements in 2004, 83.3% (15 of 18) of all smolt mortalities occurred on days of non-harassment, compared to only 16.7% (3 of 18) on days when harassment occurred. In 2005, there were a total of 171 successful and 8 unsuccessful withinstream movements of smolts during daylight hours. During the periods of non-harassment, 8.2% (7 of 85) of all smolt movements were unsuccessful. During the periods of harassment, only 1.1% (1 of 94) of all smolt movements were Figure 4. Data from photographs (n = 168) collected using a fixed-position automated time-lapse digital camera. A) Daily number of Cormorants observed during harassment (solid bar) and non-harassment (open bar) days by hour from 13 May–24 May 2004. B) Mean (± 1 SD) count of Cormorants on days of harassment (solid square) and non harassment (open square). Data are offset for easy viewing. 12 Northeastern Naturalist Vol. 20, No. 1 unsuccessful. Further, of all unsuccessful movements in 2005, 87.5% (7 of 8) of all smolt mortalities occurred on days of non-harassment, compared to only 12.5% (1 of 8) on days when harassment occurred (Fig. 6 A, B). Discussion Atlantic Salmon smolts in the upper and lower Narraguagus River estuary responded differently to light levels and tidal conditions. These behaviors are consistent with smolt movements in other systems (Martin et al. 2009, McCleave 1978, Moore et al. 1995). In the upper estuary, where tidal influence begins, smolts moved mostly at night and were not strongly influenced by the tidal cycle, effectively a continuation of freshwater behavior. In contrast, in the lower portions of the estuary, where tidal influence is significant, movements were primarily associated with ebb tides, but were still more common during darkness than daylight. During our study, there were 30 unsuccessful smolt movements that we classified as mortalities. The vast majority of mortalities occurred during daylight (particularly during early morning hours) and ebb tide conditions. We believe that the four mortalities classified as occurring during nighttime might have occurred during daylight hours before the next detection occurred because of the area where they occurred (heavy use by Cormorants) and when they occurred. This pattern of mortality is consistent with predation by visual predators, which Figure 5. Mean (± 1 SD) Cormorant number by successive harassment event for A) 2004 (r2 adj. = 50.2%) and B) 2005 (r2 adj. = 75.0%). 2013 J.P. Hawkes, R. Saunders, A.D. Vashon, and M.S. Cooperman 13 is a leading hypothesis as to why smolts avoid moving during daylight (Mc- Cormick et al. 1998, Ruggles 1980). We observed a strong temporal separation between smolt movements and Cormorant “on-water” counts, which suggests that innate smolt behaviors (e.g., a reduction in movement) likely improve survival when visual predators would be actively foraging or present. Cormorants were at least partially responsible for losses during this study. We searched for transmitters at the four largest nearby Cormorant rookeries in Narraguagus Bay and found one of the 30 missing telemetry transmitters. The extent for which Cormorants are responsible for the remainder of the missing transmitters is unclear, but mortality of smolts in early hours of daylight on ebb tide is consistent with the foraging behavior of Cormorants (Anderson et al. 2004, Dunn 1975). Predation by other piciverous birds species is also possible. We regularly observed several species, including Pandion haliaetus L. (Osprey), Bald Eagle, Larus argentatus Pontoppidan (Herring Gull), Larus delawarensis Ord (Ringbilled Gull), Ardea Herodias L. (Great Blue Heron), Lophodytes cucullatus L. (Hooded Merganser), Mergus merganser L. (Common Merganser), and Mergus serrator L. (Red-Breasted Merganser). However, the number of Cormorants encountered during field activities indicates that Cormorants are the most abundant picivorous bird in the study area. We observed Cormorants throughout the entire study area, but they most commonly congregated at constriction points within the estuary (ledge, mud Figure 6. Harassment period and last detection of unsuccessful smolts (circle) by day in A) 2004 and B) 2005. Open boxes represent periods of harassment (start and stop times) and shaded area (top and bottom of figure) represents low-light level or dark conditions. 14 Northeastern Naturalist Vol. 20, No. 1 or sand banks, etc.) at low tide conditions. Presumably, Cormorants used these constriction points to efficiently forage for smolts and other prey. Harassment successfully displaced Cormorants away from the areas targeted in the Narraguagus estuary, as evidenced by fewer Cormorants present on the water on days with harassment and a reduction in Cormorant count with cumulative harassment events. Most smolt mortality occurred during daylight hours outside of harassment time and location. The proportion of mortality observed during the day was significantly less when harassment activities were taking place. Mortality during harassment periods was as low as 12.5% during harassment, compared to 87.5% during non-harassment periods. In 2004, harassment effort only consisted of one field staff, who had limited access to the lower estuary where two (of three) mortalities occurred during harassment. In 2005, there were fewer smolts lost overall. Increased harassment effort (two people vs. one person; 319 vs 160 man hours) likely resulted in fewer birds in feeding areas (Fig. 5) and lower smolt losses. Survivorship in 2005 may also have been enhanced as a result of greater river discharge in that year. Several rain events between 15 April and 1 June 2005 resulted in mean discharge of 43.4 m3/s, twice the discharge for the same time period in 2004 (16.5 m3/s; USGS 2004, 2005). Hvidsten and Hansen (1988) and Milton et al. (1995) suggest high discharge events, coinciding with smolt outmigration and stocking, likely result in an accelerated migration time, potentially aiding in survival of out-migrating smolts. It is possible that these pulses of high discharge resulted in an improvement in survival in 2005. We believe that harassment improved survival further because the shallow nature of the estuary would still make smolts vulnerable under ebb and low tide conditions. Once harassment efforts began, photographs at our fixed-point location showed fewer Cormorants observed on the first day of harassment during each harassment period with a subsequent increase and variability in Cormorant counts during the remainder of each harassment period. We considered the possibility of Cormorants becoming conditioned to our treatment, as observed by Stickley et al. (1995), who used fixed-position human-effigy units (i.e., inflatable scarecrows) to harass Cormorants at catfish aquaculture ponds. Unlike Stickley et al. (1995), we did not have spatial or temporal consistency between events, which they found to be necessary for conditioning to occur; therefore, conditioning in our study is less likely. During the egg-laying and chick-rearing periods, Cormorants often forage early in the morning and then later in the day (Coleman and Richmond 2007). In Cormorant colonies near Milbridge, ME, breeding commences in late May or early June (B. Allen, Maine Inland Fisheries and Wildlife, Bangor, ME, pers. comm.). Therefore, activity on the Narraguagus system at the time of the year of this study appears to be either pre-breeding or a different foraging pattern entirely. Photographs, harassment data, and data on unsuccessful smolt movements, suggest Cormorant foraging in this area occurs primarily early in the day (0400–1000 hours) followed by sporadic events during the remainder of daylight hours. Since Cormorant breeding occurs just after the peak of the smolt run, timing of this study may explain why we did not document a pulse of Cormorant foraging activity on the ebb tide in the evening. 2013 J.P. Hawkes, R. Saunders, A.D. Vashon, and M.S. Cooperman 15 Acoustic telemetry allowed us to passively track smolt success on an individual basis through the estuary. Combining passive smolt tracking and active Cormorant harassment activities revealed that smolts actively moved and may have benefited from Cormorant disturbance during daylight hours. Cormorants were successfully displaced, which effectively altered their normal behavior. Cormorant harassment in the estuary successfully provided protection to smolts during daylight movements and improved smolt survival. The efforts during the two years of this study only carried out harassment activities approximately 33% of total daylight hours. We can realistically expect to observe a population effect if we expand this study to a larger scale. While population mortality reductions were not the objective of this study, we could expect to see improvements of survival similar to those observed by Kocik et al. (2009) from the river to the inner bay. It would be reasonable to expect an increase of smolts entering the near-shore environment and GoM, with mortality trending near the 12.5% observed during the second year of this study. This result may require additional harassment effort, both temporally and spatially. If used as a management tool, harassment should occur throughout the emigration window (mid-April through early June) and focus the bulk of activities in the early morning (0400 to 1000 hours) with a reduced effort later in the day. Maintaining elevated activities beyond this period may exhibit diminishing returns and overextend resources. Integrating a Cormorant-count program with smolt monitoring could further increase effectiveness by allowing low-cost monitoring until bird counts or abundance and smolt emigration were at target levels, which would trigger harassment. Management activities such as this could present new options to improve, or at least maintain, remnant Atlantic Salmon populations. Harassment is a moderate intervention targeted on minimizing predation and managing ecosystem structure and function. Such manipulations are an especially attractive management approach when it is suspected that a predator is at abnormally high abundance and prey at low abundance. During the mid- to late 1900s, we believe the ecosystem may have become out of balance because Cormorant abundance was at record high levels (Krohn et al. 1995, Wires et al. 2001) while Atlantic Salmon and other diadromous species were at historic low abundance (Saunders et al. 2006, USASAC 2011). With recent Cormorant declines being attributed to populations exceeding their carrying capacity and the resurgence of the Bald Eagle, a known predator of juveniles (Wires et al. 2001), there is a possibility that the ecosystem is trending towards a lower Cormorant predation rate on smolts. Ongoing restoration activities (dam removal, habitat alterations, and restoration hatchery practices) should result in increased abundance of multiple diadromous species. With decreased predator abundance and an increased prey field for piscivores, the combined effects should result in a higher percentage of smolts successfully entering the GoM. It is, however, unclear whether this would ultimately increase adult returns because it is not known whether early marine mortality is compensatory or additive. Given the relatively large spatial scale of the Northwest Atlantic and relatively small number of Atlantic Salmon in that area, the assumption of additive mortality (i.e., mortality that is not constrained 16 Northeastern Naturalist Vol. 20, No. 1 by density dependence) seems reasonable. Thus, our results suggest that predator manipulation and management can be a reasonable and important management tool to temporarily protect endangered populations until co-occurring prey species (and the concomitant ecological functions they perform) can recover. Acknowledgments We gratefully acknowledge the many individuals who worked on the Narraguagus Cormorant harassment project. Particular thanks are extended to T. Trinko Lake for GIS assistance, and D. Schick of the US Fish and Wildlife service for rookery site visits on coastal islands. We also thank M. Renkawitz, P. Music, J. Kocik, D. Bean, and two anonymous reviewers for helpful comments on earlier versions of this manuscript. We further acknowledge the understanding people in the towns of Cherryfield and Milbridge, and the Crane family for their tolerance of a sometimes noisy management activity. Literature Cited Anderson, C.D., D.D. Roby, and K. Collis. 2004. Foraging patterns of male and female Double-crested Cormorants nesting in the Columbia River estuary. Canadian Journal of Zoology 82:541–554. Baird Software. 2004. Spatial data analyzer (SDA): Bio-tracking user manual. Baird and Associates, Ottawa, ON, Canada. Baum, E., and R. Jordan. 1982. The Narraguagus River: An Atlantic Salmon river management report. Maine Atlantic Sea Run Salmon Commission, Augusta, ME. 69 pp. Beland, K.F., J.F. Kocik, J. van de Sande, and T.F. Sheehan. 2001. Striped Bass predation upon Atlantic Salmon smolts in Maine. Northeastern Naturalist 8(3):267–274. Blackwell, B.F., and F. Juanes. 1998. Predation on Atlantic Salmon smolts by Striped Bass after dam passage. North American Journal of Fisheries Management 18:936–939. Blackwell, B.F., and W.B. Krohn. 1997. Spring foraging distribution and habitat selection by Double-crested Cormorants on the Penobscot River, Maine. Colonial Waterbirds 20:66–76. Blackwell, B.F., W.B. Krohn, and R.B. Allen. 1995. Foods of nestling Double-crested Cormorants in Penobscot Bay, Maine, USA: Temporal and spatial comparisons. Colonial Waterbirds 18:199–208. Chaput, G., C.M. Legault, D.G. Reddin, F. Caron, and P.G. Amiro. 2005. Provision of catch advice taking account of non-stationarity in productivity of Atlantic Salmon (Salmo salar L.) in the northwest Atlantic. ICES Journal of Marine Science 62:131–143. Coleman, J.T.H., and M.E. Richmond. 2007. Daily foraging patterns of adult Doublecrested Cormorants during the breeding season. Waterbirds 30(2):189–198. Dieperink, C., B.D. Bak, L.F. Pedersen, M.I. Pedersen, and S. Pedersen. 2002. Predation on Atlantic Salmon and Sea Trout during their first days as postsmolts. Journal of Fish Biology 61:848–852. Dunn E.H. 1975. Caloric intake of nestling Double-crested Cormorants. The Auk 92(3):553–565. Fay, C., M. Bartron, S. Craig, A. Hecht, J. Pruden, R. Saunders, T. Sheehan, and J. Trial. 2006. Status review of anadromous Atlantic Salmon (Salmo salar) in the United States. Report to the National Marine Fisheries Service and US Fish and Wildlife Service. 294 pp. Available online at http://www.nmfs.noaa.gov/pr/species/statusreviews. htm. Accessed 15 September 2010. 2013 J.P. Hawkes, R. Saunders, A.D. Vashon, and M.S. Cooperman 17 Hvidsten, N.A., and L.P. Hansen. 1988. Increased recapture rate of adult Atlantic Salmon (Salmo salar L.) stocked as smolts at high water discharge. Journal of Fisheries Biology 32:153–154. Järvi, T. 1989. Synergistic effect on mortality in Atlantic Salmon, Salmo salar, smolt caused by osmotic stress and presence of predators. Environmental Biology of Fishes 26:149–152. Kocik, J.F., J.P. Hawkes, T.F. Sheehan, P.A. Music, and K.F. Beland. 2009. Assessing estuarine and coastal migration and survival of wild Atlantic Salmon smolts from the Narraguagus River, Maine using ultrasonic telemetry. Pp. 293–310, In A.J. Haro, K.L. Smith, R.A. Rulifson, C.M. Moffitt, R.J. Klauda, M.J. Dadswell, R.A. Cunjak, J.E. Cooper, K.L. Beal, and T.S. Avery (Eds.). Challenges for diadromous fishes in a dynamic global environment, American Fisheries Society Symposium 69. Bethesda, MD. Krohn, W.B., and B.F. Blackwell. 1996. Part I: Concerning a study to determine whether or not this controversial Maine nester is a major predator of Atlantic Salmon smolts in the Penobscot River. Maine Fish and Wildlife 38(2)9–12. Krohn, W.B., R.B. Allen, J.R. Moring, and A.E. Hutchinson. 1995. Double-crested Cormorants in New England: Population management histories. Colonial Waterbirds 18(Special Publication):99–109. Lacroix, G.L. 2008. Influence of origin on migration and survival of Atlantic Salmon (Salmo salar) in the Bay of Fundy, Canada. Canadian Journal of Fish and Aquatic Sciences 65:2063–2079. Lacroix, G.L., P. McCurdy, and D. Knox. 2004. Migration of Atlantic Salmon postsmolts in relation to habitat use in a coastal system. Transactions of the American Fisheries Society 133:1455–1471. McCleave, J.D. 1978. Rhythmic aspects of estuarine migration of hatchery-reared Atlantic Salmon (Salmo salar) smolts. Journal of Fish Biology 12:559–570. Martin, F., R.D. Hedger, J.J. Dodson, L. Fernandes, D. Hatin, F. Caron, and F.G. Whoriskey. 2009. Behavioural transition during the estuarine migration of wild Atlantic Salmon (Salmo salar L.) smolt. Ecology of Freshwater Fish 18:406–417. McCormick, S.D., L.P. Hansen, T.P. Quinn, and R.L. Saunders. 1998. Movement, migration, and smolting of Atlantic Salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 55:77–92. McCormick, S.D., D.T. Lerner, M.Y. Monette, K. Nieves-Puigdoller, J.T. Kelly, and B.T. Björnsson. 2009. Taking it with you when you go: How perturbations to the freshwater environment, including temperature, dams, and contaminants, affect marine survival of salmon. American Fisheries Society Symposium 69:195–214. Meister, A.L., and F.J. Gramlich. 1967. Cormorant predation on tagged Atlantic Salmon smolts. Final report of the 1966–67 cormorant-salmon smolt study. Atlantic Sea Run Salmon Commission. Orono, ME. Milton, G.R., P.J. Austin-Smith, and G.J. Farmer. 1995. Shouting at shags: A case study of cormorant management in Nova Scotia. Colonial Waterbirds 18(Special Publication 1):91–98. Moore, A., E.C.E. Potter, N.J. Milner, and S. Bamber. 1995. The migratory behaviour of wild Atlantic Salmon (Salmo salar) smolts in the estuary of the River Conwy North Wales. Canadian Journal of Fisheries and Aquatic Sciences 52:1923–1935. Moore, A., S. Ives, T.A. Mead, and L. Talks. 1998. The migratory behavior of wild Atlantic Salmon (Salmo Salar L.) smolts in the River Test and South Hampton Water, southern England. Hydrobiologia 371/372:295–304. 18 Northeastern Naturalist Vol. 20, No. 1 National Research Council (NRC). 2004. Atlantic Salmon in Maine. National Academy Press, Washington, DC. Ruggles, C.P. 1980. A review of the downstream migration of Atlantic Salmon. Canadian Technical Report of Fisheries and Aquatic Sciences 952. 39 pp. Saunders, R., M.A. Hachey, and C.W. Fay. 2006. Maine’s diadromous fish community: Past, present, and implications for Atlantic Salmon recovery. Fisheries 31(11):537–547. Stickley, A., D.F. Mott, and J.O. King. 1995. Short-term effects of an inflatable effigy on Cormorants at catfish farms. Wildlife Society Bulletin 23(1):73–77. Strategic Assessments Branch. 1985. National estuarine inventory data atlas. National Oceanic and Atmospheric Administration (NOAA), National Ocean Service (NOS), Ocean Assessments Division, Rockville, MD. US Atlantic Salmon Assessment Committee (USASAC). 2011. Annual report of the US Atlantic Salmon Assessment Committee 23:2010 activities. Portland, ME. US Geological Survey (USGS). 2004. USGS gauge 01022500: Narraguagus River at Cherryfield, Maine. Available online at http://waterdata.usgs.gov/me/nwis/rt. Accessed 27 August 2004. USGS. 2005. USGS gauge 01022500: Narraguagus River at Cherryfield, Maine. Available online at http://waterdata.usgs.gov/me/nwis/rt. Accessed 30 August 2005. Wires, L.R., F.J. Cuthbert, D.R. Trexel, and A.R. Joshi. 2001. Status of the Double-crested Cormorant (Phalarocorax aruitus) in North America. Final Report to the United States Fish and Wildlife Service. University of Minnesota, St. Paul, MN.