Using dive behavior and active acoustics to assess prey use and partitioning by fin and humpback whales near Kodiak Island, Alaska

Transcription

1 MARINE MAMMAL SCIENCE, **(*): *** *** (*** 2014) 2014 Society for Marine Mammalogy DOI: /mms Using dive behavior and active acoustics to assess prey use and partitioning by fin and humpback whales near Kodiak Island, Alaska BRIANA H. WITTEVEEN, 1 School of Fisheries and Ocean Sciences, Alaska Sea Grant Marine Advisory Program, University of Alaska Fairbanks, 118 Trident Way, Kodiak, Alaska 99615, U.S.A.; ALEX DE ROBERTIS, Resource Assessment & Conservation Engineering Division, Alaska Fisheries Science Center, National Marine Fisheries Service, NOAA, 7600 Sand Point Way, Seattle, Washington 98115, U.S.A.; LEI GUO and KATE M. WYNNE, School of Fisheries and Ocean Sciences, Alaska Sea Grant Marine Advisory Program, University of Alaska Fairbanks, 118 Trident Way, Kodiak, Alaska 99615, U.S.A. Abstract Near the Kodiak Archipelago, fin (Balaenoptera physalus) and humpback (Megaptera novaeangliae) whales frequently overlap spatially and temporally. The Gulf Apex Predator-prey study (GAP) investigated the prey use and potential prey partitioning between these sympatric species by combining concurrent analysis of vertical whale distribution with acoustic assessment of pelagic prey. Acoustic backscatter was classified as consistent with either fish or zooplankton. Whale dive depths were determined through suction cup tags. Tagged humpback whales (n = 10) were most often associated with distribution of fish, except when zooplankton density was very high. Associations between the dive depths of tagged fin whales (n = 4) and the vertical distribution of either prey type were less conclusive. However, prey assessment methods did not adequately describe the distribution of copepods, a potentially significant resource for fin whales. Mean dive parameters showed no significant difference between species when compared across all surveys. However, fin whales spent a greater proportion of dive time in the foraging phase than humpbacks, suggesting a possible difference in foraging efficiency between the two. These results suggest that humpback and fin whales may target different prey, with the greatest potential for diet overlap occurring when the density of zooplankton is very high. Key words: pelagic backscatter, Balaenoptera physalus, Megaptera novaeangliae, suction-cup tag, multi-frequency differencing, prey partitioning, humpback whale, fin whale, Gulf of Alaska. As long-lived and large-bodied mammals with high energetic demands, fin (Balaenoptera physalus) and humpback (Megaptera novaeangliae) whales are apex predators and consume substantial amounts of prey (Katona and Whitehead 1988, Kenney et al. 1997, Witteveen et al. 2006), likely influencing the structure of marine ecosystems (Trites et al. 1997, Croll et al. 1998, Witteveen et al. 2006). These two rorqual (Family Balaenopteridae) whale species are often sympatric and may exploit common 1 Corresponding author ( bree.witteveen@alaska.edu). 1

2 2 MARINE MAMMAL SCIENCE, VOL. **, NO. **, 2014 prey resources; both are known to target pelagic schooling fish and zooplankton (Thompson 1940, Overholtz and Nicolas 1979, Whitehead and Carlson 1988, Piatt et al. 1989, Payne et al. 1990). Because these species often consume similar prey in shared habitat, some degree of partitioning is required to avoid resource competition where they co-occur (Gowans and Whitehead 1995). The degree to which habitat or prey partitioning is present determines whether species are occupying different ecological niches and whether they compete for limited resources. Understanding the extent of resource partitioning is critical when investigating species roles in regional ecosystems. Such investigation is arguably of increased importance for baleen whales given their extreme population fluctuations over the past several decades; near extirpation resulting from intense commercial harvest has been followed by significant recovery of some populations (Baker and Clapham 2004, Zerbini et al. 2006, Barlow et al. 2011). However, climate change and potential alterations to lower trophic level production and assemblage structure may threaten future recovery and growth. While the mechanisms are not fully understood, it is likely that the removal and replacement of large whales has had ecosystem level impacts (Springer et al. 2003, Baker and Clapham 2004, Ballance et al. 2006). The waters surrounding the Kodiak Archipelago in the central Gulf of Alaska support feeding populations of both fin and humpback whales (Wynne and Witteveen 2005, Zerbini et al. 2006, Witteveen et al. 2007). Both species occur year-round, with the highest abundances occurring between May and October, and are showing marked population growth (Zerbini et al. 2006, Barlow et al. 2011, Allen and Angliss 2013). The Gulf Apex Predator-prey (GAP) program based out of the University of Alaska Fairbanks, between 1999 and 2013, has sought to address trophic level questions of biological and economic concern in the Gulf of Alaska triggered by declines in the western stock of Steller sea lions. The goals of the program have been to document trophic relationships between sea lions, their prey, competitors, and predators in the Kodiak area ecosystem. GAP projects ( map/gap) have been designed to broadly assess the degree of spatial and temporal variability and dietary overlap among the region s apex predators from an ecosystem perspective and have involved dozens of researchers. In recent years, a specific component of GAP has been to increase our understanding of habitat and prey partitioning that may be occurring between sympatric fin and humpback whales to better describe the ecological niches of these whales. Baraff (2006) analyzed fin and humpback whale distribution off northeast Kodiak Island relative to physical and biological habitat characteristics to investigate potential habitat partitioning. As in other regions, spatial segregation of these species was found to be related to habitat differences, with fin whales favoring more offshore, shelf edge habitats and humpback whales using more onshore shelf or bank regions (Nemoto 1959, Reeves et al. 1985, Brueggeman et al. 1988, Baraff 2006). Baraff (2006) also showed that humpback and fin whale distribution overlapped in areas with high densities of euphausiids, but that fin whales were generally associated with the distribution of large-bodied copepods (Neocalanus spp.), suggesting the potential for prey partitioning in nearshore Kodiak waters. To further explore this potential prey partitioning, we analyzed whale foraging behavior while concurrently assessing pelagic prey fields in coastal waters of the Kodiak Archipelago. Whale foraging behavior was assessed using dive data obtained from acoustic time depth transmitters (ATDTs) temporarily attached to free-swimming whales. Past studies have demonstrated the utility of combining whale tagging with prey assessment to determine species being targeted by foraging whales (e.g.,

3 WITTEVEEN ET AL.: PREY USE OF FIN AND HUMPBACK WHALES 3 Baumgartner et al. 2008, Witteveen et al. 2008, Friedlaender et al. 2009, Hazen et al. 2009). The ATDT tags provided real-time depth data that were compared to concurrent prey assessment to elucidate targeted prey fields. Using a dual-frequency differencing technique of active acoustic backscatter to discern fish from zooplankton prey fields (e.g., Madureira et al. 1993, Kang et al. 2002, De Robertis et al. 2010), we inferred the prey targeted by foraging fin and humpback whales. Materials and Methods Survey Effort Whale and prey data were collected during six surveys conducted in and around Kodiak Island, Alaska between 2004 and 2011 (Fig. 1, Table 1). All surveys were conducted in August with the exception of a winter survey in December of 2007 (2007Winter). Each survey used two vessels; one dedicated to acoustic assessment of prey and one dedicated to whale tagging, tracking and documenting the number of individual whales (Table 1). Whale Tagging and Dive Analysis Free-swimming fin and humpback whales were tagged with suction-cup tags containing an ATDT (V-16 or V22, Vemco, Bedford, Nova Scotia, Canada) and a VHF transmitter (MOD-125, Telonics, Mesa, AZ). Tags were deployed from either a cross bow (for tag bodies made of syntactic foam (2004 and 2005) or pneumatic launcher (for tag bodies made of PVC [2007 and 2009] or aluminum pipe [2011]). In addition, a pole-deployed 3-dimensional archival tag (Acousonde 3A, Acoustimetrics, Santa Barbara, CA) was attached to a single humpback whale in Depths from the ATDTs were received through a vessel-mounted directional hydrophone in realtime (VH100, Vemco). As such, it was necessary to closely follow the tagged whale for the duration of the tag attachment (focal follows). All received dive data were time-linked to the position (latitude and longitude) of the tracking vessel that then, due to the close proximity of the whale (<1 km), was considered to represent the track of the tagged whale. A number of dive parameters, including maximum depth, mean foraging depth, duration of dive, and surface interval, were derived from the depth-time series from the tagged whales (Table 2). The dive profiles were visually inspected and categorized as either foraging or nonforaging dives (see Witteveen et al. 2008). Briefly, foraging dives were characterized by dive shape or presence of lunges at depth (Fig. 2). Dive categorization allowed for foraging and nonforaging behavior to be distinguished and enabled the mean depth of foraging dives to be calculated for each whale. Only parameters for dives classified as foraging dives are reported here. Means for each dive parameter were calculated for each tagged whale and then averaged across whales by year, species, or prey type as needed for comparisons. Prey Surveys and Analysis of Pelagic Backscatter Acoustic surveys from a separate vessel were used to assess the vertical distribution of potential prey in the vicinity of tagged whales. Acoustic backscatter was measured

4 4 MARINE MAMMAL SCIENCE, VOL. **, NO. **, W 152 W PERENOSA BAY Kilometers 24 TAGGED WHALE TRACKS HB04-01 FW07-04 HB04-02 HB07-05 HB04-03 FW09-01 HB04-04 FW Summer HB05-01 FW Winter HB07-01 HB HB07-02 HB11-02 ACOUSTIC TRACKS N IS LA N D KO D IA K N MARMOT BAY UGANIK BAY 59 N ALASKA 2011 Figure 1. Kodiak Island study area. Whale tagging and prey surveys were conducted in Marmot Bay, Perenosa Bay, and Uganik Bay between 2004 and The solid lines represent tracks of tagged whales, and the dashed lines represent the acoustic survey tracks. with a 38 and 120 khz split-beam echo sounder in all cases except 2007Winter when a 38 and 200 khz split-beam echo sounder was used (Table 1). Acoustic surveys followed either predetermined transects designed to assess prey broadly in areas with whale aggregations, or fine-scale transects in the immediate vicinity of tagged whales (Fig. 1; Witteveen et al. 2008).

5 WITTEVEEN ET AL.: PREY USE OF FIN AND HUMPBACK WHALES 5 Table 1. Summary acoustic sampling effort between 2004 and Fish were sampled with a midwater trawl (MW) and zooplankton with a tucker trawl (TT) in some surveys. Also shown are the number of individual humpback(hb) and fin whales(fw) sighted and tagged during each survey. For acoustic gear HM = hull mounted and T = towed. Survey Dates Location Acoustic gear Frequencies (khz) Sampling Tags deployed No. HB No. FW August Marmot Bay Simrad EK60 (HM) 38 & 120 MW 4 HB August Marmot Bay Simrad EK60 (HM) 38 & 120 MW, TT 1 HB Summer August Marmot Bay Simrad EK60 (HM) 38 & 120 MW, TT 2 HB Winter 27 November 6 December Uganik Bay BioSonics DXT-70 (T) 38 & 200 MW, TT 1 HB 1FW August Marmot Bay Simrad EK-60 (T) 38 & 120 None 3 FW August Perenosa Bay Simrad EK-60 (T) 38 & 120 None 2 HB 42 1

6 6 MARINE MAMMAL SCIENCE, VOL. **, NO. **, 2014 Table 2. Description and units of dive parameters calculated for each tagged whale. Parameter Unit Description MaxDepth m Maximum depth reached AvgDepth m Average depth reached MeanBotDepth m Mean depth during bottom phase Duration min Total duration of the dive Surface min Surface interval Botttime min Duration spent at the bottom phase BottVertDist m Sum of vertical distance traveled during the bottom phase BottRange m Range of depth transisted during the bottom time Efficiency Botttime/(Duration + Surface) Figure 2. Examples of foraging and nonforaging whale dives; foraging dives are most often identified by the presence of lunges at depth, as indicated by the black arrows. In the first three surveys, personnel and equipment availability allowed for net sampling of prey fields to determine species composition and size distribution. Pelagic schooling fish were assessed by targeted trawling with a mid-water trawl (headrope/footrope length: 39 m; vertical opening: 20 m; 2.22 cm codend liner mesh) and zooplankton were sampled using a triple-net Tucker Trawl (1 m 2 opening, 300 lm). Because prey sampling did not occur in all years, these results were used as qualitative indicators of the species likely dominating the acoustic backscatter at these locations. Acoustic assessment of sound-scattering organisms is an efficient means of documenting the abundance and distribution of potential prey fields (MacLennan and Simmonds 1992, Simmonds and Maclennan 2005). However, unambiguous identification of prey species in these fields requires additional sampling such as net or optical sampling that can be logistically difficult and expensive, especially when coupled with whale tagging efforts. As an alternative, we analyzed dual-frequency acoustic data to discern whether acoustic backscatter of prey fields near whales was consistent with either fish or zooplankton. Acoustic backscatter was processed in Echoview 4.9 (Myriax Pty. Ltd., Hobart, Tasmania, Australia). Backscatter was partitioned into signals consistent with fish or zooplankton based on their relative frequency response, using a two-frequency variant of the method described in De Robertis et al. (2010). Acoustic records were averaged into 5 ping by 5 m depth cells, and the frequency response (S V120kHz S V38kHz ;S V = backscatter strength) in each cell was computed.

7 WITTEVEEN ET AL.: PREY USE OF FIN AND HUMPBACK WHALES 7 We applied the method of De Robertis and Higginbottom (2007) to correct for background noise and exclude low signal-to-noise areas. Analysis cells with a frequency response in the range of > 16 to <8 db were assigned to the fish category and those in the range of 8and<30 db were assigned to the zooplankton category (cf.fig.2 in De Robertis et al. 2010). In some cases, a persistent near-surface scattering layer that was unlikely to be from forage fish or zooplankton (e.g., Woillez et al. 2007) was manually excluded prior to applying the frequency difference criteria. The fish (38 khz, S V threshold = 70 db re 1 m 1 ) and zooplankton categories (120 or 200 khz, S V threshold = 80 db re m 1 ) were echo-integrated into the Nautical Area Scattering Coefficient (s A,m 2 nmi 2 ; see MacLennan et al. 2002) in 5 m depth layers at an along-track spatial resolution of m intervals. This was accomplished by integrating all of the analysis cells with the appropriate frequency response (e.g., in the case of fish, if the cell has a frequency response consistent with fish then S vfish = S V38kHz,elseS Vfish = 999) in the m along-track integration interval. In the summers of 2004, 2005, and 2007, the echo sounders were calibrated using the standard sphere method (Foote et al. 1987), but no calibrations were conducted in the 2007Winter, 2009, or 2011 surveys (Table 1). For the surveys in which no calibrations were conducted, we implemented a correction procedure to account for potential biases. First, we identified regions of pelagic fish schools suspected to be walleye pollock (Theragra chalcogramma), capelin (Mallotus villosus) or Pacific herring (Clupea pallasii) and as regions of backscatter consistent with euphausiids (for a total of n = 76, 17, and 10 regions in 2007Winter, 2009, and 2011, respectively) based on their appearance. We then computed the discrepancy between the observed frequency response from the expected mean value as reported in De Robertis et al. (2007) for pollock, capelin, and euphausiids and in Gauthier and Horne (2004) for herring. Next, we adjusted the high-frequency echo sounder gains such that the mean frequency response of all aggregations matched the previously published mean values and then reprocessed the data as described above. This correction forces the ratio of backscatter at the high and low frequencies to match the expected values. In essence, this will maintain the expected frequency response between zooplankton and fish, allowing for improved taxonomic discrimination with uncalibrated systems. In other words, the ratio of backscatter across frequencies will be approximately correct (as it has been forced to match the expected frequency response of the observed species), but the absolute levels have been scaled by an unknown factor and cannot be compared across data sets. This method minimizes the impacts of calibration uncertainty on the dual frequency classification method, does not impact estimates of the vertical distribution of prey, and allows for the ratio of fish to zooplankton backscatter to be compared across surveys. In each survey, water column s A from both prey categories was averaged to calculate the ratio of fish to zooplankton backscatter. As described above, because acoustic equipment was not calibrated or consistent among all surveys, only direct comparisons of average s A among the 2004, 2005, and 2007Summer surveys are valid. However, the ratios of fish to zooplankton s A could be compared as an index of the relative availability of fish and zooplankton prey among all surveys. This ratio will depend on the relative abundance of fish and zooplankton as well as their sizes, species composition, and the frequencies used (recall that a 200 khz echo sounder was used in 2007Winter). For example, although there is considerable uncertainty in measurements of acoustic reflectivity, published estimates suggest that acoustic scattering from a pollock is expected to be 1.3 times higher than that of a herring of similar size and 2.7 times higher than that of a capelin of similar size (Traynor 1996, Ona 2003,

8 8 MARINE MAMMAL SCIENCE, VOL. **, NO. **, 2014 Guttormsen and Wilson 2009). Given that the species and size composition of the scatterers was not known in all cases, the fish to zooplankton s A ratio should be considered a proxy for prey availability that can be used to identify large shifts in the composition of potential prey available to the fin and humpback whales. In addition, fish and zooplankton s A was computed in 5 m vertical depth layers for comparison to the distribution of whale foraging dives for each survey. Spatial and Statistical Analysis To ensure a meaningful temporal and spatial comparison, s A data were limited to include only daytime data that were collected within 1,500 m and 60 h of each tagged whale s track. While efforts were made to match tagging and prey sampling efforts as closely as possible, there were occasions when circumstances (e.g., inclement weather, equipment failure) prevented this from occurring. A number of alternative scales were explored and, ultimately, the temporal and spatial sampling selected, while broad, maximized the use of all available tag and backscatter data. To explore vertical relationships between whale dives and prey, the distribution of whale foraging dive depths was compared to the vertical distribution of fish and zooplankton s A using box plots (Fig. 3). The degree of overlap between whale foraging dives and fish and zooplankton backscatter was quantified by calculating the Global Index of Collocation (GIC; Woillez et al. 2007). The GIC estimates how spatially related two groups are by comparing the distance between their mean locations (as the center of gravity) and the variance of these locations (as inertia). Center of gravity (CG) and inertia (I) were adapted from Wollilez et al. (2007) and calculated as: Figure 3. Examples of dual-frequency backscatter collected during each survey: (A) 2004, (B) 2005, (C) 2007Summer, (D) 2007Winter, (E) 2009, and (F) The left side of each panel represents the lower frequency (38 khz) and the right represents the higher frequency (120 khz in all surveys except for 200 khz in 2007Winter). Zooplankton aggregations are characterized by higher backscatter at the higher frequency while fish aggregations exhibit similar backscatter at both frequencies. The solid black line indicates the mean depth of foraging for all tagged whales for the survey.

9 WITTEVEEN ET AL.: PREY USE OF FIN AND HUMPBACK WHALES 9 P N i¼1 CG ¼ x iz i P N i¼1 Z ; ð1þ i and I ¼ P N i¼1 ðx i CGÞ 2 Z i P N i¼1 Z ; ð2þ i where i represents depth in 5 m bins, x is a specific depth bin and z is the frequency distribution of whale dives, fish backscatter or zooplankton backscatter. The GIC between each whale dive and the prey distribution was calculated as: GIC ¼ ðcg 1 CG 2 Þ 2 ðcg 1 CG 2 Þ 2 þ I 1 þ I 2 ; ð3þ where CG 1 CG 2 is the separation between the centers of gravity of the depths of whale foraging dives and either fish backscatter or zooplankton backscatter and I 1 and I 2 are their respective inertias. The GIC values range from 1 (complete collocation) to 0 (complete separation). The GIC values were then used to quantify whether a given foraging dive profile more closely matched the vertical distribution of fish s A or zooplankton s A on a per dive basis. Each dive was assigned as either a fish dive or a zooplankton dive based on which GIC value was higher. A binomial sign test was then used to determine if dive assignments for a given tagged whale deviated from random (i.e., 50% fish and 50% zooplankton). The GIC values were also averaged across all dives to give each whale an overall fish or zooplankton designation, based again on whether the fish or zooplankton GIC value was higher. For whales in which GIC did not result in a significant relationship to either prey type, the whale dive depths and prey vertical distributions were compared using a Kolmogorov-Smirnov (KS) statistic (D) to assign presumed prey, with the lower D value indicating a more similar distribution between whale foraging dives and a given prey type. Finally, all backscatter and whale dive data were combined to calculate a GIC value and assign dives and presumed prey on a per survey basis. Dive parameters were examined through analysis of variance (ANOVA) to explore the influence of whale species and targeted prey type on dive behavior. All analyses were conducted in R (R Core Development Team) or PASW 18.0 (IBM SPSS). Results Across all years of study, 10 humpback and 4 fin whales were successfully tagged. Analysis of the dive profiles showed that humpback whales dove deeper for shorter durations on average than fin whales (Table 1, 3). More humpback whales were seen in all survey years, with the exception of 2009 when whale densities were lower (Table 1). Sightings of both whale species were highest in 2007Summer (Table 1). Average s A varied substantially among the surveys in which calibration allowed direct comparison (2004, 2005, and 2007Summer; Table 4). The water column zoo-

10 10 MARINE MAMMAL SCIENCE, VOL. **, NO. **, 2014 Table 3. Summary of humpback and fin whales tagged during GAP surveys between 2004 and The tag attachment time, number of dives and mean foraging depth are given. HB = humpback whales, FW = fin whale. Tag_ID codes represent the species, year and sequential tag number. Tag_ID Date Species Attachment duration (h) No. of dives Mean foraging dive depth (m) SD HB August 2004 HB HB August 2004 HB HB August 2004 HB HB August 2004 HB HB August 2005 HB HB August 2007 HB HB August 2007 HB FW December 2007 FW HB December 2007 HB FW August 2009 FW FW August 2009 FW FW August 2009 FW HB August 2011 HB HB August 2011 HB plankton backscatter was particularly variable: zooplankton backscatter was higher in 2007Summer than in 2004 and The ratio of fish to zooplankton backscatter, an index that reflects the relative backscatter of fish and zooplankton (and is comparable among all surveys), was also highly variable among surveys, reaching a maximum in 2011 and a minimum in 2007Summer (Table 4). Fish backscatter in most years was attributed primarily to capelin and, to a lesser extent, age-0 and juvenile (age 1 2) pollock (Table 4). The exception was in 2007Winter, when most pelagic fish backscatter was attributed to adult Pacific herring (Table 4). Zooplankton backscatter was sampled with nets only in the two 2007 surveys. In 2007Summer, a persistent zooplankton scattering layer that extended from 90 to 160 m deep in some areas was dominated by euphausiids. The zooplankton assemblage in 2007Winter was comprised of a diverse mix of copepods, amphipods, and euphausiids (Wynne and Witteveen 2009). Though qualitative, visual assessment of echograms from each frequency revealed relative differences in available prey and obvious relationships between tagged whales and available prey in all surveys (Fig. 3). Forage fish dominated prey fields in 2004, 2005, and 2011 (Fig. 3A, B, F), while a dense band of euphausiids was evident in 2007Summer (Fig. 3C). The moderately dense mixed zooplankton layer observed during the 2007 winter survey (Fig. 3D) correlates with the depth of whale dives, although schools of Pacific herring were evident as well. In 2009, fish and zooplankton backscatter appeared to be low relative to other surveys (Fig. 3E): the pelagic schools of forage fish seen in 2004, 2005, and 2011 were absent and zooplankton backscatter was reduced. Relationships between humpback whale foraging dive depths and prey vertical distribution were evident during each survey, with boxplots and GIC indices indicating that the tagged whales were targeting either fish or zooplankton (Table 5, Fig. 4, 5). Applying the temporal (within 60 h) and spatial (within 1,500 m) restriction to the

11 WITTEVEEN ET AL.: PREY USE OF FIN AND HUMPBACK WHALES 11 Table 4. Acoustic backscatter apportioned to fish and zooplankton using the dual-frequency technique for each survey. Asterisks indicate the year between which echo sounders were calibrated and backscatter can be directly compared. The dominant prey species as determined by sampling from midwater trawls (MW) and Tucker Trawls (TT) are also shown. No net sampling occurred in 2009 or Survey Average fish sa (m 2 nmi 2 ) s A Average zooplankton s A (m 2 nmi 2 ) Fish:zooplankton ratio 2004* Capelin Juvenile walleye pollock Eulachon 2005* 1, Capelin Age-0 walleye pollock 2007 Summer* 914 2, Capelin Juvenile walleye pollock MW TT N/A N/A Euphausiids Copepods 2007 Winter 3.4 Pacific herring Copepods N/A N/A N/A N/A Amphipods Euphausiids

12 12 MARINE MAMMAL SCIENCE, VOL. **, NO. **, 2014 Table 5. Summary of mean Global Index of Collocation (GIC) comparing the vertical distribution of tagged whale foraging dives and fish and zooplankton backscatter (s A ). Backscatter data were truncated to include only data collected within 1,500 m and 60 h of the tagged whale tracks. The mean GIC values for all whales and all s A combined for each survey are also shown. A presumed targeted prey type is listed for each tagged whale based on either dive assignment tests or mean GIC values. Asterisks denote sign tests showing dive assignments were significantly different than random. Prey for nonsignificant tests were determined based on similarity of box plots using Kolmogorov-Smirnov comparison of distributions. N/A indicates that no data within the temporal or spatial restrictions matched to the tagged whale. Average GIC Whale Fish Zooplankton Presumed prey 2004 HB04-01 N/A HB Fish* HB Fish HB Fish* All Fish* 2005 HB Fish* All Fish 2007 Summer HB Zooplankton HB Zooplankton* All Zooplankton* 2007 Winter FW Zooplankton HB Zooplankton* All Zooplankton* 2009 FW09-01 N/A FW09-02 N/A FW Zooplankton All Zooplankton 2011 HB11-01 N/A HB Fish* All Fish* acoustic and dive data eliminated four tags (HB04-01, FW09-01, FW09-02 and HB11-01) from individual comparisons, although these data were used in surveywide analyses (Table 5). Humpback whales tagged during the 2004, 2005, and 2011 surveys were most strongly affiliated with the vertical distribution of fish backscatter. Significantly more dives in these years were assigned to fish than zooplankton, and mean GIC values showed stronger collocation with fish as well (Table 5, Fig. 4, 5). Sign tests conducted on dives by whale HB04-03 did not show a significant association with either fish or zooplankton, though more of its dives were characterized as targeting fish and the difference between dive and fish depth distributions was less (Kolomogorov-Smirnov D = 0.38) than for dive and zooplankton depths (D = 0.46) (Fig. 4, 5). In the 2007Summer survey whale HB07-02 showed a clear relationship with zooplankton distribution. The dives depths of HB07-01 exhibited a more variable pattern (Table 5), but more closely matched zooplankton (D = 0.29) than fish (D = 0.47) (Fig 4). Acoustic data were available in the vicinity for only two of the four tagged fin whales during the 2007Winter and 2009 surveys. Associations between fin whale dives and prey distribution were less definitive than for humpback whales (Table 5, Fig. 3, 4, 5). While the fin whale, FW07-04, tagged in 2007Winter showed a higher

13 WITTEVEEN ET AL.: PREY USE OF FIN AND HUMPBACK WHALES 13 Figure 4. Box plots showing the depth distributions of whale foraging dives and fish and zooplankton backscatter in the vicinity of the whale for whale tags (A) HB04-01, (B) HB04-02, (C) HB04-03, (D) HB05-01, (E) HB07-01, (F) HB07-02, (G) FW07-04, (H) HB07-05, (I) FW09-03, and (J) HB number of dives associated with zooplankton backscatter, the assignment of dives was not significantly different than random. However, box plots suggest that dive depths were more strongly associated with zooplankton (D = 0.19) than fish vertical distribution (D = 0.26) (Fig. 4G). In 2009, FW09-03 dive depths showed a closer association with zooplankton (D = 0.39) than fish (D = 0.52) vertical distribution, although assignments of individual dives to fish and zooplankton were nearly equal (Fig. 4, 5). Dive parameters were compared for three scenarios: FWs vs. HBs with no consideration of targeted prey species, HBs vs. FWs during 2007Winter only (since this was the only survey during which both species were tagged and likely both targeting zooplankton), and HBs only with respect to targeted prey species (Table 6). Dive parameters for FWs were not compared across prey types, as their targeted prey was less certain. When species (FW vs. HB) was the only factor considered, there was no significant difference between the mean values of any dive parameter (Table 7). However, when mean values of parameters of the single humpback and fin whales tagged in the 2007Winter survey were compared, significant differences were seen in all parameters except BottVertDist and BottRange (Table 6, 7). Most notably, the tagged humpback whale dove deeper and longer, but the tagged fin whale had a greater efficiency parameter (Table 6). The characteristics of humpback whale dives were compared across targeted prey type as their dives were shown to significantly correlate to both fish and zooplankton prey types. The mean values of MaxDepth, AvgDepth and MeanBotDepth for dives

14 14 MARINE MAMMAL SCIENCE, VOL. **, NO. **, * * * * * * 0.75 Proportion of Dives 0.50 Prey F Z HB04-02 HB04-03 HB04-04 HB05-01 HB07-01 HB07-02 FW07-04 HB07-05 FW09-03 HB11-02 Whale Figure 5. Proportion of foraging dives of tagged fin and humpback whales assigned as targeting either fish (F) or zooplankton (Z) as determined by global index of collocation (GIC). Asterisks indicate that the distribution of dive assignments is significantly different from random. associated with these prey types were significantly different; humpbacks dove deeper when targeting zooplankton than when targeting fish (Table 6, 7). Discussion The combined whale tagging and prey assessment surveys revealed that humpback and fin whales do exhibit prey partitioning near Kodiak Island. Evidence for this was found in differences in the targeted prey types and mean dive parameters between these two species, as well as variability in their relative abundance across surveys. Humpback Whales The foraging dives of tagged humpback whales were associated with either fish or zooplankton backscatter depending on the survey, supporting the broad perspective of the humpback whale as an opportunistic generalist predator. In each of the summer surveys, tagged humpback whales targeted the prey type with the highest relative backscatter. Foraging dive distributions were more similar to fish vertical distributions (i.e., exhibited higher GIC values) when the backscatter ratio was high and dive depths were more similar to zooplankton vertical distributions when the ratio

15 WITTEVEEN ET AL.: PREY USE OF FIN AND HUMPBACK WHALES 15 Table 6. Mean values (SD) of dive parameters for foraging dives recorded from tagged humpback and fin whales summarized by prey type or survey. Means for each parameter were calculated for each tagged whale and then averaged across whales by year, species, or prey type as needed for comparisons. Parameters are defined in Table 2. HB FW All Fish Zooplankton 2007Winter All Zooplankton 2007Winter N MaxDepth AvgDepth MeanBotDepth Duration Surface Botttime BottVertDist BottRange Efficiency

16 16 MARINE MAMMAL SCIENCE, VOL. **, NO. **, 2014 Table 7. Results of ANOVA for each parameter by species, by prey type without considering species, and by species for the 2007Winter survey only. Asterisks indicates the level of significance with * indicating P < 0.05, ** indicating P < 0.01, *** indicating P < and ns meaning P The parameters are defined in Table 2. ANOVA results HB vs.fw HBvs.FW Zooplanktonvs.fish Parameter All 2007Winter only HB only MaxDepth ns *** * AvgDepth ns * * MeanBotDepth ns *** * Duration ns *** ns Surface ns *** ns Botttime ns * ns BottVertDist ns ns ns BottRange ns ns ns Efficiency ns * ns was low. During the 2009 survey, the fish to zooplankton ratio was relatively high and only a moderate zooplankton layer was observed, but few humpback whales were sighted and none were tagged (Table 1). Although no net sampling was conducted in 2009, visual assessment of backscatter suggested the prey field was dominated by fish occurring at low density (likely adult pollock), rather than the schooling forage fish observed in the other years (Fig. 3). Thus, the fish observed in 2009 may not have been suitable prey for humpback whales. The relative paucity of humpback whales in this year, the fewest observed among all surveys, was therefore likely the result of the absence of aggregations of forage fish or a dense euphausiid layer. Calibration of the echo sounders for the 2004, 2005, and 2007Summer surveys allowed for direct interannual comparisons of humpback whale foraging behavior in Marmot Bay. In these years, humpback whale dive depths changed in a manner consistent with a shift from targeting capelin in the summers of 2004 and 2005 to targeting euphausiids in the summer of This prey switch coincided with a relative increase in the amount of zooplankton backscatter in 2007Summer as well as a decrease in the fish to zooplankton s A ratio. The zooplankton scattering layer observed during the Summer2007 survey, which was determined to be primarily euphausiids, was dense and extensive (Fig. 3C). Humpback whales have been shown to rely more heavily on fish resources than zooplankton in other foraging areas (Overholtz and Nicolas 1979, Friedlaender et al. 2009, Hazen et al. 2009, Laidre et al. 2010). Significant differences in the mean values of MaxDepth, AveDepth, and MeanBotDepth show that when foraging on zooplankton, humpback whales dove to deeper depths than when targeting fish. Many fish prey also show seasonal peaks in energy densities that are higher than crustacean prey, making shallower fish a more attractive prey choice than deeper layers of zooplankton in some cases (Davis et al. 1998, Boldt and Haldorson 2002, Weitkamp and Sturdevant 2008). Dolphin (1988) concluded that the energy cost of foraging was determined by depth of dive and dive duration. Therefore, an increase in foraging costs and reduced prey energy density may make deeply distributed zooplankton, even though it may take less energy to capture, too energetically expensive to target unless densities are high (Goldbogen et al. 2008, 2012).

17 WITTEVEEN ET AL.: PREY USE OF FIN AND HUMPBACK WHALES 17 In contrast to the summer surveys, the humpback whale tagged during 2007Winter was not shown to target the prey layer that showed the highest relative backscatter. The ratio of fish to zooplankton backscatter was 3.4 during this survey, which would suggest that the whales would target fish yet foraging dives were clearly correlated to the zooplankton layer (Figs. 3D, 4H & 5). A likely explanation for this discrepancy is that fish backscatter during this survey was attributed primarily to large Pacific herring. The herring sampled during this survey were age 5 or older and measured greater than 28 cm (Wynne and Witteveen 2009), which may have made them more difficult to catch than smaller and younger age classes. While humpback whales are known to target herring in many regions, herring foraging may be a specialized or learned behavior developed in regions where herring is a more dominant forage fish species than in the Kodiak archipelago, such as Southeast Alaska or Prince William Sound (Jurasz and Jurasz 1979, Sharpe 2001, Heintz et al. 2010, Teerlink 2011). Individual specialization on specific prey types, in fact, may also explain some of the observed differences in targeted prey across surveys for humpback whales. Specialization of foraging behaviors and in preferred prey has been observed in marine mammals and often arises as the result of intraspecific competition (Estes et al. 2003). Humpback whales have experienced tremendous population growth since the cessation of commercial whaling and may be at or near preexploitation levels (Calambokidis et al. 2008, Barlow et al. 2011). Such marked population growth could increase intraspecific competition and facilitate specialized foraging. Regardless of the mechanism, whether it is prey availability or preference or cost of foraging, results from all surveys suggest that humpback whales more frequently targeted forage fish, specifically capelin, but that they can forage on zooplankton when found in sufficient densities in the absence of forage fish. Fin Whales Fin whales were tagged during two of the six surveys, 2007Winter and Fin whales were present in substantial numbers, but not tagged during 2007Summer, when a dense euphausiid layer was observed (Table 1, Fig. 3C). In 2004, 2005, and 2011, fin whales were not present in high enough densities in the study area to attempt tag attachments (Table 1). Fin whales were not shown to have significant relationships to either fish or zooplankton backscatter during years in which they were tagged. However, the fin whale tagged in 2009 showed high, though not significant, GIC values to both fish and zooplankton (Table 5, Fig. 5). As described above, much of the fish backscatter in that year appeared to be from low-density aggregations of individual fish such as adult walleye pollock rather than pelagic schooling fish, such as capelin or juvenile pollock (Fig. 3E). The diet of adult pollock in the Gulf of Alaska is known to include euphausiids, copepods, and amphipods (Adams et al. 2007, Urban 2012). Therefore, the relatively weak association of fin whales to either fish or zooplankton backscatter may be the result of the co-occurrence of fish and whales that were targeting the same zooplankton layer. Generally, the lack of strong correlation between fin whale foraging dives to either fish or zooplankton may be the result of multiple factors. Fin whales may target both prey types equally or are less directed by the relative densities of available prey. Alternatively, the frequencies used to document prey distribution in this study may not have adequately detected the presence or distribution of the fin whales targeted prey. Copepods are known to be a significant prey resource for fin whales (Thompson 1940,

18 18 MARINE MAMMAL SCIENCE, VOL. **, NO. **, 2014 Nemoto 1957, Nemoto and Kasuya 1965), but they are unlikely to dominate the acoustic zooplankton index computed in this study as the backscatter from copepods is weak at the frequencies used during these surveys (Matsukura et al. 2009). The contribution of zooplankton depends on the material properties and size, and at the frequencies used, the backscatter is expected to scale with the fourth power of length (Greenlaw 1979), which indicates that the acoustic zooplankton index used in this study was likely dominated by the contributions from larger organisms such as euphausiids and amphipods. Comparison of Fin and Humpback Whales Evidence of resource partitioning between fin and humpback whales may be gleaned, in part, by comparing their relative abundance across surveys. For surveys during which humpback whale abundance was relatively high, forage fish were abundant relative to zooplankton (i.e., fish to zooplankton backscatter ratio >1; Table 3). In contrast, fin whales were all but absent during these surveys (Table 1). The only survey that contradicted this pattern was 2009, which was characterized by an absence of forage fish and relatively low of zooplankton backscatter. In years such as this, fin whales may have been targeting copepods, which are not considered primary prey of humpback whales (Nemoto and Kasuya 1965, Gaskin 1982). Directed sampling of copepods is required to investigate this possibility. Both species were present in substantial numbers during 2007Summer, suggesting that fin and humpback whales may overlap more when euphausiid density is high. Ryan et al. (2014) come to a similar conclusion for fin and humpback whales in the Celtic Sea, reporting that stable isotopes indicate humpback whales forage to a lesser extent on krill than do fin whales and that any resource overlap between the two species may be short-lived. Numerous other studies have shown that resource partitioning decreases and dietary overlap increases among sympatric predators when high-quality prey are abundant (e.g.,croxall et al. 1999, Forero et al. 2004, Barger and Kitaysky 2012, Durso et al. 2013). The apparent differences in prey utilization between whale species did not translate to significant differences in foraging behaviors, as measured by various dive parameters. Fin whales had a higher mean efficiency parameter than humpback whales (0.49 vs. 0.44, respectively). Efficiency can be increased by either increasing the amount of time spent in the bottom (or foraging) phase of the dive or decreasing dive duration and/or surface interval. Examination of mean dive parameters showed that humpback whales might spend more time in the descent and/or ascent phases of their dives than fin whales. However, humpback whales and fin whales were not always targeting the same prey type confounding direct comparison of parameters. When mean dive parameters were compared for tagged fin and humpback whales from the 2007Winter survey, which represented the only survey during which both species were tagged (although it was only a single individual of each species), significant differences were evident in a number of parameters. Despite both species showing a relationship between foraging dive depths and zooplankton backscatter, fin whale dives were shallower and shorter with a higher efficiency parameter. It has been shown in other studies that despite what is predicted by theoretical aerobic dive limit (TADL), the larger rorquals (i.e., fin and blue (Balaenoptera musculus) whales) often have shorter dive times (Croll et al. 2001). Therefore, differences in dive time and depth between fin and humpback whales in this case may be reflective of simple size differences. However, Doniol-Valcroze et al. (2011) cautions that while allometric relationships have the potential to explain observed differences, these relationships often fail to consider

19 WITTEVEEN ET AL.: PREY USE OF FIN AND HUMPBACK WHALES 19 specific ecological considerations. The comparison here was made between only one animal of each species and, therefore, additional data from both species tagged in the same region is needed to further explore the observed differences. Differences in targeted prey between fin and humpback whales may be the result of multiple factors. Humpback whales have been classified as fast maneuverers with a body form beneficial for catching elusive prey (e.g., schooling fish; Woodward et al. 2006) and has having baleen specialized for larger prey (Werth 2013). Fin whales were not explicitly categorized by Woodward et al. (2006), but blue whales, the closest relative to the fin whale, were classified as fast cruisers with a body form designed for efficient travel to widely distributed patches of zooplankton. Fin whales are sometimes characterized to be among the fastest of the rorquals (Mizroch et al. 2009). In addition, fin whales have finer baleen plates than humpback whales, making them more adept at trapping smaller zooplankton prey (i.e., copepods; Gaskin 1982). Large body size and limited maneuverability has been cited as the reason for blue whales being obligate krill-feeders (Potvin et al. 2012). It has been hypothesized that as a result of their finer baleen, faster swim speed, and horizontal lunging behavior fin whales have a foraging threshold that requires less dense patches of prey than humpback whales (Baraff 2006). This study supports this hypothesis, but is in seeming disagreement with the prediction that the foraging efficiency of rorqual whales decreases with an increase in body size, implying that fin whales should require higher densities of prey than the smaller humpback whales (Goldbogen et al. 2012, 2013). However, these assumptions about foraging efficiency and prey densities were based only on krill as the targeted prey type and did not address the impacts of alternative prey sources, such as forage fish or copepods. Although the exact mechanisms determining prey preference and foraging behavior in these two species remains to be determined, species-specific morphological and physiological differences, in addition to the size, density, and accessibility of prey, are likely contributing to prey partitioning. Ecosystem Implications This study suggests that humpback whales likely have a more diverse diet than sympatric fin whales. This difference could have implications in the marine food webs in which they play predatory roles. A diet comprised of a wider variety of prey species may, in part, facilitate the substantial recovery from commercial harvest exhibited by humpback whales. In the North Pacific, humpback whale population growth rates have been estimated between 6% and 10% per year (Mizroch et al. 2004, Calambokidis et al. 2008) and humpback whales have been sighted in the Arctic Ocean, an area not previously known as humpback habitat (Hashagen et al. 2009). Fin whales are also recovering and have exhibited population growth at annual rates of 4.8%, but fin whales are failing to return to some of their historical habitats (Zerbini et al. 2006). For example, fin whales were one of the dominant species in southeastern Alaska prior to and during commercial whaling, but sightings of fin whales in this area today are quite rare (Mizroch et al. 2009, J. Straley 2 ). The diet flexibility of humpback whales may contribute to their successful recovery and the less flexible diet of fin whales may inhibit their recovery in certain parts of their range. 2 Personal communication from Jan Straley, University of Alaska Southeast, Sitka, AK 99835, January 2014.

20 20 MARINE MAMMAL SCIENCE, VOL. **, NO. **, 2014 While humpback whales were shown to be more generalist predators than fin whales, this study and others have also provided evidence that their diets are more likely to be comprised of fish than zooplankton (Overholtz and Nicolas 1979, Friedlaender et al. 2009, Hazen et al. 2009, Laidre et al. 2010, Ryan et al. 2014). Given their higher consumption of fish relative to fin whales, humpback whales may have greater potential to impact prey populations important to other higher trophic level consumers such as sea lions, commercial fish species, and human harvesters. Conclusions By combining dive data from foraging whales with concurrent data on the distribution of prey, we provide evidence for prey partitioning between fin and humpback whales within the Kodiak Archipelago. This supports previous findings that fin and humpback whales in Kodiak are spatially segregated except in areas where high euphausiid densities promote a high degree of spatial overlap When both fin and humpback whales were present in similar densities both species were targeting zooplankton prey (2007Summer and Winter). Clearly, the small sample sizes presented here require that the conclusions related to dive parameters should be interpreted with caution. Additional tagging of whales with concurrent estimates of prey abundance determined from multiple acoustic frequencies is likely to provide further insights into differences in foraging behavior and prey preferences between these two whale species, which will contribute to the understanding of fin and humpback whales as consumers in changing ecosystems. Acknowledgments The authors are grateful to Casey Clark, Aaren Ellsworth, Brian Ellsworth, Jane McKenzie, Sophie Piersolowski, Beth Pingree, Natura Richardson, Mike Trussell, and Jordy Thomson for their assistance with tagging efforts in the field. We thank the captain and crews of the F/V Alaska and F/V Mythos. We thank the Alaska Department of Fish and Game, including Matthew Foster, Mark Witteveen, and the captain and crew of the R/V Resolution. Special thanks to Dr. Robert Foy (NMFS, Alaska Fisheries Science Center) for his assistance with prey surveys and assessment. We thank the three anonymous reviewers who provided valuable comments to improve this manuscript. The findings and conclusions in this paper are those of the authors and do not necessarily represent the views of the National Marine Fisheries Service. Reference to trade names does not imply endorsement by the National Marine Fisheries Service, NOAA. Funding was provided by NOAA Grants NA04NMF , NA07NMF , NA09NMF439039, NA10NMF to the Gulf Apex Predator prey project at the University of Alaska Fairbanks. All whale research was conducted under NMFS Federal Research Permits # and All data collection was covered under University of Alaska Fairbanks IACUC protocols and Literature Cited Adams, C. F., A. I. Pinchuk and K. O. Coyle Seasonal changes in the diet composition and prey selection of walleye pollock (Theragra chalcogramma) in the northern Gulf of Alaska. Fisheries Research 84: Allen, B. M., and R. P. Angliss Alaska marine mammal stock assessments, U.S. Department of Commerce, NOAA Technical Memorandum NMFS-AFSC pp.