Chapter 12 : Summary of Boat and Aerial Data

Transcription

1 Chapter 12 : Summary of Boat and Aerial Data Emily E. Connelly, Sarah M. Johnson, Iain J. Stenhouse, Kathryn A. Williams Boat and high-definition digital video aerial surveys were used to collect information on the distributions and abundance of animals in the mid-atlantic offshore region over a two-year time period ( ). A focused comparison study of the two methods was conducted in 2013 (see Chapter 11); however, a more general examination of the full datasets can also provide valuable information about the relative utility of the two methodologies. In many ways these two datasets are complementary, and together can provide a more complete understanding of the ecology of the mid-atlantic. Maps of persistent hotspots of wildlife abundance, as well as bar charts illustrating the timing and relative abundance of species presence in the study area throughout the year, were developed as illustrative methods to convey variation in wildlife abundance through space and time. While they are still under development, these are the first methods used in this study to combine information from both the boat-based and digital aerial survey datasets. Comparison of Boat and Aerial Survey Results Observation Rates Digital aerial survey data were easily effort-corrected to present observations per square kilometer, as aerial transects had a consistent strip width. Boat surveys were designed to have a strip width of at least 300 m, but the effective strip width varied by taxon. Detection of objects in boat surveys is known to vary with distance from the observer (Thomas et al. 2010), and thus species that were readily detected large distances away from the boat had a larger surveyed area, or effective strip width, than species that were generally only detectable near the boat. We calculated effective strip half width for the four avian taxa where data was sufficient to parameterize a null distance model in package unmarked in the R Statistical Computing Environment. These groups included Sulidae (gannets), Laridae (gulls and terns), Gavidae (loons), and Anatidae (scoters, ducks, and geese) (Figure 12-2). For most species, there were insufficient boat data to model detection, and the median observation distance was used in these cases. Effective strip width was calculated in unmarked by applying distance-based detection functions (halfnormal distributions) to species groups during distance modeling, and integrating the area underneath the distance curve. Because of specific properties of distance detection curves, this number is equal to the distance at which there is a 50% chance of the detecting an object (Royle, Dawson, and Bates 2004). Because we surveyed on both sides of the ship, this effective strip half-width was multiplied by two. The full effective strip width for each species group, multiplied by the total linear distance of the survey, was used as the effective boat survey area for that species group. Preliminary assessment of the boat-based and aerial survey data indicated that the two methods differ in their abilities to detect and identify certain taxa. Boat surveys detected larger numbers of more Part IV: 12-1

2 species of birds than the aerial surveys, while the digital aerial surveys appeared to be better at detecting certain aquatic animals (Figure 12-1, Table 12-1). Some of the discrepancies in observations point towards potential differences in detectability between the two survey types; for example, Northern Gannets (Morus bassanus) and larger gull species are visible at great distances from the boat survey, as observers can look from the vessel all the way to the horizon. Reviewers of aerial survey data, in contrast, can only see animals present in the narrow strip of the transect onscreen, and aerial survey speed is roughly 13.5 times that of the boat, potentially limiting onscreen appearances by highly mobile animals (Chapter 11). Differences in identification abilities between survey methods may also play a role in explaining the lower detections (see Chapters 2 and 5 for more details on survey methods). Digital aerial surveys were highly effective at detecting several unexpected groups of animals: sharks, fish, and rays (Figure 12-1). While some of these animals were also observed in the boat survey, the aerial surveys provided an excellent platform for detecting and identifying animals within the upper reaches of the water column. In particular, higher counts and species diversity of sea turtles and mammals were detected on the aerial surveys (Figure 12-1, Chapter 4) than from the boat. A similar efficiency in detecting and identifying sea turtles and marine mammals from high definition digital aerial platforms (as compared to visual aerial or boat surveys) has also been observed elsewhere (Normandeau Associates Inc. 2013). Several migratory movements were captured through the use of digital aerial surveys. Major migrations of Cownose Rays (Rhinoptera bonasus) were observed in the aerial study, but went undetected in the boat surveys; almost 48,000 rays were observed in aerial surveys (Chapter 4). Many schools of baitfish were observed in the aerial data as well, some spanning hundreds of meters. Both surveys detected bats, specifically Eastern Red Bats (Lasiurus borealis), though more bats were detected in the aerial survey, and the bats observed from the aerial platform were flying higher than could be detected from the boat (Chapter 13). Identification Rates There appeared to be differences in observers ability to identify animals between the aerial and boatbased surveys in some cases (Figure 12-3). This was partially due to the much more rigorous species identification and audit processes used for the aerial video (see Chapter 11 for a more detailed discussion). More than twice as many bird species were definitively detected in the boat surveys than from the air, with many more aerial observations limited to the family or genus level of identifications (Table 12-1). Gulls and terns (Laridae), loons (Gaviidae), and auks (Alcidae) all had much higher identification rates to the species level from the boat surveys than the aerial (Figure 12-3). Low rates of aerial species identification were not altogether surprising for auks and terns, given their small size and subtle differences between species, but based on results from European studies, higher identification rates had been expected for loons. Aerial video reviewers faced difficulties in differentiating the two loon species that use the mid-atlantic during the non-breeding season, however, due to the high degree of suspected size overlap (particularly for birds sitting at the water s surface) in this time period and region of the U.S. (Gray et al. 2014). Project partners are hoping to use loon identifications and habitat Part IV: 12-2

3 relationships from the boat survey modeling efforts (Chapter 8) to calculate likelihoods of aerial observations belonging to either species; if successful, these results will be presented in future reports. Aerial observers were better than boat observers at identifying scoters, ducks, and geese (Anatidae), which is likely due to boat observers having difficulty differentiating large flocks of Black Scoters (Melanitta nigra) and Surf Scoters (M. perspicillata) at a distance (see Chapter 11 for a more detailed discussion). Observers from both survey types had similarly high identification rates of shearwaters (Procellariidae). Identification rates of toothed whales (Odontoceti) and baleen whales (Mysticeti) were similar between survey platforms (Figure 12-4), and each method observed a few species that were missed by the other (Chapters 4 and 6). Turtles were identified to the species level at much higher rates from the boat survey platform, but only two species (Loggerhead Turtle, Caretta caretta; and Leatherback Turtle, Dermochelys coriacea) were observed; despite difficulties with differentiating subsurface small turtles in the aerial footage, video observers detected three additional species of turtles (Kemp s Ridley Turtle, Lepidochelys kempii; Hawksbill Turtle, Eretmochelys imbricate; and Green Turtle, Chelonia mydas). As fish were not a focal taxon for research in this study, neither platform identified fish to species, aside from Ocean Sunfish (Mola mola); the aerial observers detected 168 sunfish, while the boat observers detected three. While we saw more large whales in the boat surveys (n =35) than in the aerial surveys (n =15), we saw eight North Atlantic Right Whales (Eubalaena glacialis) in the aerial surveys compared to the one in the boat study (Chapters 4 and 6). Large whales were observed throughout the study area (Figure 12-5). North Atlantic Right Whales (Eubalaena glacialis) were seen primarily on aerial surveys, with several sightings within the Virginia WEA and one further offshore on a boat survey. Two further sighting occurred on the saw-tooth transects between Maryland and Virginia s WEAs. All sightings were reported to NOAA and the New England Aquarium. Other whales, including Humpback Whales (Megaptera novaeangliae), Fin Whales (Balaenoptera physalus), and Minke Whales (Balaenoptera acutorostrata), were observed throughout the study area (Figure 12-5). Disturbance Boat surveys are known to affect animal behavior and, in some cases, to influence detectability as a result (see Chapter 11, Schwemmer et al. 2014). Gulls are often attracted to boats as potential sources of food, while scoters are sensitive to disturbance by boats (Buckland et al. 2012, Chapter 11). Marine mammals are also known to be attracted to or disturbed by boats (Mattson, Thomas, and St. Aubin 2005). Integrating Data from Boat and Aerial Surveys Boat and digital aerial surveys appear to produce different results for some taxa, which can present a challenge when comparing data obtained from the two survey methodologies. However, these differences also present an opportunity, in that the two surveys can provide complementary data that may be used, in tandem, to provide a better overall view of wildlife distributions and relative abundance in the mid-atlantic study area. Two such integrative efforts have been developed to date, including the Part IV: 12-3

4 identification of geographic hotspots of persistent abundance, as well as the identification of temporal patterns of persistence and relative abundance of animals within the study area. Persistent hotspot maps highlight locations where individuals within a species or species group have been consistently observed in numbers greater than a standardized baseline anomaly, to identify spatial patterns of species abundance that persist over time and may indicate the locations of important habitat areas (Santora and Veit 2013). Effort-corrected data from boat and aerial surveys were analyzed independently to determine singular occurrences of hotspots within surveys, and then combined to show where hotspots consistently occurred across survey methodologies. This approach allowed us to examine spatial patterns of persistent abundance for a wide range of taxa, including seabirds, marine mammals, and turtles, that may be present within the study area for varying amounts of time depending on each group s life history traits. Bar charts summarize the temporal patterns of species and species groups within the study area. Relative abundance by two-month time period was calculated independently for boat and aerial data, and the two values are presented adjacent to one another to allow for comparison between survey methods for each species and species group. By pairing persistent hotspot maps with temporal bar charts for each species or taxon of interest, we are able to develop a comprehensive picture of geographic and temporal patterns of wildlife within the study area. Persistent hotspot analysis Aerial and boat survey transects and species observations were binned into 4.8 x 4.8 km (23.0 km 2 ) grid cells in ArcGIS version (ESRI, Redlands, California). Grid cells represented Bureau of Ocean Energy Management (BOEM) Outer Continental Shelf (OCS) lease blocks for development on the Atlantic coast of the U.S.; this grid was extended west of the Submerged Lands Act boundary to include the entire study area. This approach allowed us to independently analyze data collected via the two survey methods, and subsequently combine the resulting hotspot estimations from each survey methodology into a unified hotspot persistence map summarized by lease block. A mean (± SD) of 145 ± 10.6 blocks per survey were sampled by boat, covering a mean (± SD) distance of 4.06 ± 1.41 km per block. Aerial surveys sampled a mean (± SD) of 379 ± 86.9 blocks per survey, covering a mean area of 2.80 ± 2.81 km 2 per block. Methods developed by Santora and Veit (2013) were adapted to quantify the variance and anomaly persistence of counts for a single species or species group within each lease block. For the subsequent steps described, boat and aerial survey data were analyzed independently. The first step in determining variance of counts was to calculate an effort-corrected count for the species of interest per lease block for each survey. For boat surveys, this was done by dividing the number of individuals observed by the transect length (km) within each block per survey. For aerial surveys, the number of individuals observed of a given species or group was divided by the total survey area (km 2 ) within each block per survey. Z- scores for each block per survey were then calculated by subtracting the survey mean (of effortcorrected counts) from the effort-corrected count for each lease block, and dividing by the survey s Part IV: 12-4

5 standard deviation (of effort-corrected counts) to normalize each survey s data around the survey mean (after Santora & Veit 2013): Z i,j = [(N i,j /E i,j ) (N j /E j )]/S j sd where N is the number of individuals observed (in block i, for survey j), E is the transect length (for boat surveys) or survey area (for aerial surveys) per block per survey, N j is the mean number of individuals observed per survey, E j is the mean survey length/area per survey, and S j sd is the standard deviation of effort corrected counts for all blocks within a survey. Each lease block within a survey was then examined to determine whether or not the block represented a hotspot, or an area where species abundance was higher than a standardized baseline anomaly. To achieve this, the grand sample mean and standard deviation was calculated for all lease blocks and surveys for a given species (or species group) and survey method. Each block was tabulated as an abundance hotspot for a survey if its z-score was greater than one standard deviation above this standardized baseline. After determining which survey lease blocks represented a hotspot for a given species and survey, the two survey datasets were combined in order to index the percentage of time each lease block represented a hotspot across survey methods. Blocks that had been surveyed fewer than 8 times within a survey method were excluded from further analyses (e.g., if a block was surveyed < 8 times by boat but 8 times by aerial video, only the boat survey data for the block was excluded; if a block was surveyed < 8 times by each survey method, all data for the block was excluded). Thus, while hotspot calculations used all available data for a species, data are not presented for blocks surveyed in less than half of all surveys. Using these criteria, 446 lease blocks were included in subsequent analyses. For each lease block, the number of times a block represented a hotspot was divided by the number of times the block was surveyed to calculate the percentage of surveys in which each block represented a hotspot for the species of interest. Cells with higher percentages represented areas where the species more often showed abundance levels above the standardized baseline anomaly. Persistence classes (e.g., the percentage of time each block was a hotspot) were defined for mapping purposes by percentile (<80 th, th, th, and >95 th percentile) to allow for comparison between species with different life histories that were present in the study area for varying amounts of time throughout the year. Examples of resulting maps are shown below in Figure 12-6 and Figure Refinement of this approach for identifying persistent hotspots, as well as the development of additional hotspot persistence maps for other taxa, are ongoing and will be discussed in future reports. Temporal bar charts To represent the relative abundance of animals in the study area through the seasons, boat and aerial survey data were analyzed independently, as detection and geographic coverage was known to vary between survey methods. The total count of individuals observed was summed for each species and species group by two-month time period. Thus, each time period included data from two to four surveys Part IV: 12-5

6 over the two year study. The two-month length of these time periods was found to best serve for data visualization purposes, as it allowed for variation in the data while also controlling somewhat for variation in effort between periods (see Chapters 4 and 6). Counts were standardized for survey effort (km 2 for aerial survey; linear km for the boat survey) over each two-month time period. Bar charts were built for all species and species groups that were observed more than 10 times (within a survey methodology) over the course of the study. Percentiles were then calculated for all effort-corrected survey data for individual species (Table 12-2) and species groups (Table 12-4). Seven symbol categories were used to represent the percentiles of the data; boat and aerial percentile values were calculated separately, and are presented adjacent to one another to allow for comparison between the two study methods. Preliminary Results The maps and bar charts presented below contain preliminary aerial data, and do not represent final analyses for the species presented. These figures represent examples of products that will be refined for the final report. Northern Gannets are an example of an individual species that was commonly observed and easily identifiable by both boat and aerial survey methods. Northern Gannets were observed in 81% of all surveys (13 out of 16 boat surveys and 12 out of 15 aerial surveys). Across all surveys, Northern Gannets displayed persistently higher abundance close to shore, across all latitudes within the study area (Figure 12-6). Northern Gannets were consistently observed in high numbers from September to April for both boat and aerial surveys, with numbers decreasing in the spring and reaching a low in July and August (Table 12-3). The two survey methods show very similar temporal variance for Northern Gannets, indicating that detection rates for this species were similar between survey methods (Table 12-3). Sea turtles are an example of a taxon that was analyzed at the group (e.g., multi-species) level, and that displayed apparent differences in detectability between survey methods. Sea turtle species were observed in 68% of all surveys (9 out of 16 boat surveys and 12 out of 15 aerial surveys). Sea turtles were more consistently located in large numbers in the southern half of the study area, and generally further from shore (Figure 12-7). Turtles were most commonly observed during summer months, with fewer observations made in the spring and fall, and very few observations in the winter (Table 12-5). Sea turtles consistently represented a greater proportion of aerial observations compared to boat survey observations, although persistent hotspots for this taxon were generally located in parts of the study area with relatively little aerial survey coverage. This suggests the possibility that sea turtles are more easily detected by aerial video surveys than by traditional boat surveys (Table 12-5), as has been seen elsewhere (Normandeau 2012). It should be noted that these approaches to summarizing and analyzing our two-year datasets do not explicitly measure or account for inter-annual variation. Just as there was some level of variation in observations between survey platforms, and the patchy nature of seabird distributions (Fauchald 2009) ensured some level of random variation in our observations, there was also substantial variation in the Part IV: 12-6

7 community composition, distribution, and abundance of species observed between each year of surveys. This type of inter-annual variation in the marine environment is thought to be driven primarily by variation in abiotic variables such as sea surface temperature and currents (Ballance, Pitman, and Fiedler 2006). The Bureau of Ocean Energy Management (BOEM) recommends a minimum of two full annual cycles for offshore surveys for wind energy development for this reason (Rein et al. 2013). For example, though digital aerial surveys for this study were conducted in June and September of 2012 and July and September of 2013, only in 2013 surveys were large numbers of cownose rays observed; some variation in water temperatures, ray populations, or other factors meant that very few rays were seen in Similarly, scoters were observed in high numbers each winter surveyed on the boat survey, but more than twice as many scoters were seen in January of 2013 than in January of Generally speaking, however, we did find roughly similar patterns between years, and conducting surveys over a longer time frame would characterize this interannual variability still further. Comparison and integration of survey data collected using different approaches is essential for characterizing habitats and identifying key habitat use areas. The analyses presented here will be further developed and refined for the final project report in Literature Cited Ballance, Lisa T., Robert L. Pitman, and Paul C. Fiedler Oceanographic Influences on Seabirds and Cetaceans of the Eastern Tropical Pacific: A Review. Progress in Oceanography 69: doi: /j.pocean Buckland, Stephen T., M. Louise Burt, Eric A. Rexstad, Matt Mellor, Adrian E. Williams, and Rebecca Woodward Aerial Surveys of Seabirds: The Advent of Digital Methods. Journal of Applied Ecology 49 (4): doi: /j x. Fauchald, P Spatial Interaction between Seabirds and Prey: Review and Synthesis. Marine Ecology Progress Series 391 (September): Gray, Carrie E., James D. Paruk, Christopher R. DeSorbo, Lucas J. Savoy, David E. Yates, Michael D. Chickering, Rick B. Gray, et al Body Mass in Common Loons ( Gavia Immer ) Strongly Associated with Migration Distance. Waterbirds 37 (sp1): doi: / sp109. Mattson, Megan Cope, Jeanette a. Thomas, and David St. Aubin Effects of Boat Activity on the Behavior of Bottlenose Dolphins (<I>Tursiops truncatus</i>) in Waters Surrounding Hilton Head Island, South Carolina. Aquatic Mammals 31 (1): doi: /am Normandeau High-Resolution Aerial Imaging Surveys of Marine Birds, Mammals, and Turtles on the US Atlantic Outer Continental Shelf Utility Assessment, Methodology Recommendations, and Implementation Tools. Report Prepared under BOEM Contract M10PC U.S. Department of the Interior Bureau of Ocean Energy Management. Part IV: 12-7

8 Rein, Christopher G., A. Scott Lundin, Stephanie J.K. Wilson, and Elana Kimbrell Offshore Wind Energy Development Site Assessment and Characterization: Evaluation of the Current Status and European Experience. Herndon, VA. Royle, J Andrew, Deanna K Dawson, and Scott Bates Modeling Abundance Effects in Distance Sampling 85 (6): Santora, Jarrod A., and Richard R. Veit Spatio-Temporal Persistence of Top Predator Hotspots near the Antarctic Peninsula. Marine Ecology Progress Series 487: doi: /meps Schwemmer, Philipp, Bettina Mendel, Nicole Sonntag, Volker Dierschke, and Stefan Garthe Effects of Ship Traffic on Seabirds in Offshore Waters: Implications for Marine Conservation and Spatial Planning. Ecological Applications : A Publication of the Ecological Society of America 21 (5): Thomas, Len, Stephen T. Buckland, Eric a. Rexstad, Jeff L. Laake, Samantha Strindberg, Sharon L. Hedley, Jon R B Bishop, Tiago a. Marques, and Kenneth P. Burnham Distance Software: Design and Analysis of Distance Sampling Surveys for Estimating Population Size. Journal of Applied Ecology 47 (1): doi: /j x. Part IV: 12-8

9 Figures and Tables Observations per km Gannets (Sulidae) Gulls and Terns (Laridae) Loons (Gaviidae) Scoters, Ducks, Geese (Anatidae) Toothed Whales (Odontoceti) Fish and Sharks Rays (Batoidea) Turtles (Testudines) Aerial Boat Figure Comparison of boat and aerial survey densities by taxon for all surveys. Densities are calculated by the total number of counts divided by the total survey area. Aerial data have transect strip widths of 200 meters. Effective boat transect strip widths were calculated for each group based on the effective half strip width (birds, see Figure 12-2) or were based on the median distance of observations from the boat, in meters (non-avian animals: Odontoceti, 300; Fish/Sharks, 50; Batoidea, 7.5; and Testudines, 100 meters). Aerial observations of groups that were not individually counted or identified (e.g., bait balls, ray schools) are excluded from this figure (see Chapter 3 for more information). Part IV: 12-9

10 A B C D Figure Distance functions for a) Sulidae, b) Laridae, c) Gavidae, and D) Anatidae from the boat survey data. Effective strip half-widths, or the distance from the boat at which there is average detection probability, are indicated by the vertical line in each chart. They are 330m for Sulidae, 236m for Laridae, 272m for Gavidae, and 271m for Anatidae. Part IV: 12-10

11 Identification Rate DE-EE % 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Aerial Boat Aerial Boat Aerial Boat Aerial Boat Aerial Boat Aerial Boat Anatidae Sulidae Laridae Gaviidae Alcidae Procellariidae Figure Identification rates for common bird taxa observed in boat and high-definition video aerial surveys in order of abundance. Dark colors indicate animals identified to species, and light colors indicate animals identified to higher taxonomic levels. Details on species sighted within each taxonomic group can be found in Chapters 4 and 6. The most common avian families observed in surveys were scoters, ducks, and geese (Anatidae); gannets (Sulidae); gulls and terns (Laridae); loons (Gaviidae); auks (Alcidae); and fulmars and shearwaters (Procellariidae). Part IV: 12-11

12 Identification Rate DE-EE % 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Aerial Boat Aerial Boat Aerial Boat Odontoceti Testudines Mysticeti Figure Identification rates of mammals and turtles observed on the boat and high-definition video aerial surveys. Dark colors indicate animals identified to species, and light colors indicate animals identified to higher taxonomic levels. Details on species within each taxonomic group can be found in Chapters 4 and 6. Groups are toothed whales (Odontoceti); sea turtles (Testudines); and baleen whales (Mysticeti). Part IV: 12-12

13 Figure Large whale observations (Mysteceti) from boat and aerial surveys (March 2012 May 2014). Part IV: 12-13

14 Figure Classified persistence abundance hotspots for Northern Gannets (Morus bassanus) observed in boat and aerial surveys March 2012 May For each percentile category shown in the legend, the corresponding percentage of time a block was a hotspot is shown parenthetically; each block is labeled by the number of times the block was considered a hotspot for the given species (values vary due to differences in survey effort between lease blocks). Part IV: 12-14

15 Figure Classified persistence abundance hotspots for sea turtles observed in boat and aerial surveys March 2012 May For each percentile category shown in the legend, the corresponding percentage of time a block was a hotspot is shown parenthetically; each block is labeled by the number of times the block was considered a hotspot for the given species (values vary due to differences in survey effort between lease blocks). Part IV: 12-15

16 Table Total number of animals and species observed in the boat and high-definition video aerial surveys. Counts are not corrected for survey effort or detection. Avian Animals Non-Avian Animals Survey Type Number observed Species Number observed Species Boat 59, , Aerial 46, , Table Legend for individual species temporal persistence and relative abundance charts (see Table 12-3 below). Darker and larger bars show time periods when a species or group was more commonly observed in surveys. Percentile 0 50% 50-60% 60-70% 70-80% 80-90% 90-95% % Boat Aerial E E E E E Symbol Table Temporal persistence and relative abundance of Northern Gannets in boat (B) and aerial (A) studies during the same time period as Figure Common Name Sep-Oct Nov-Dec Jan-Feb Mar-Apr May-Jun Jul-Aug B A B A B A B A B A B A Northern Gannet Part IV: 12-16

17 Table Figure legend for grouped species temporal persistence and relative abundance charts (see Table 12-5 below). Darker and larger bars show the time periods when a species or a group was more commonly observed in surveys. Percentile 0 50% 50-60% 60-70% 70-80% 80-90% 90-95% % Boat Aerial E E Symbol Table Temporal persistence and relative abundance of sea turtle species in boat (B) and aerial (A) surveys during the same time period as Figure Leatherback, Loggerhead, Kemp s Ridley, Hawksbill, and Green turtle species are included. Species Group Sept-Oct Nov-Dec Jan-Feb Mar-Apr May-June July-Aug B A B A B A B A B A B A Turtles Part IV: 12-17