Observer effort for wildlife sightings data. Quantifying observer effort for opportunistically-collected wildlife sightings

Transcription

1 1 Observer effort for wildlife sightings data 2 3 Quantifying observer effort for opportunistically-collected wildlife sightings ERIN U. RECHSTEINER 1, CAITLIN F. C. BIRDSALL 1, DOUG SANDILANDS 1, IAIN U. SMITH 1, ALANA. V. PHILLIPS 1, LANCE G. BARRETT-LENNARD *, Cetacean Research Lab, Vancouver Aquarium, PO Box 3232, 845 Avison Way, Vancouver, BC, V6B 3X8, Canada 10 * Lance.Barrett-Lennard@vanaqua.org 1

2 Abstract. Data contributed to opportunistic sightings networks can provide important information on the presence of species when systematic surveys are unaffordable or impractical. However, without records of when and where observers traveled, it is impossible to determine whether geographic variation in sightings reflects variation in observer effort or variation in the distribution or abundance of the species of interest. Here, we describe a Geographic Information Systems (GIS) model that we used to reconstruct the distribution of volunteer observer effort for observers reporting whale, dolphin and sea turtle sightings to the British Columbia Cetacean Sightings Network (BCCSN). Observers were grouped into seven categories: residents of population centers, crew of large marine vessels, park users, lighthouse keepers, ecotourism operators, coastal workers, and frequent observers. Effort for each observer group was estimated using distribution and travel patterns that we determined were typical of that group, including trip distances, proximity to home port, standard travel routes and maximum sighting distances. We then estimated the relative effectiveness of each category at sighting, identifying, and reporting cetaceans, and used this estimate to weight the effort layer maps for each observer category. The layer maps were summed to give a spatially explicit overall estimate of observer effort, and the model was tested for sensitivities to model inputs. We applied the effort model to our sightings database to calculate an index of sightings density per unit effort (SDUE) for humpback whales (Megaptera novaengliae) and killer whales (Orcinus orca) in British Columbia waters. Areas with high SDUE corresponded well with critical habitats of humpback and killer whales previously designated by Fisheries and Oceans Canada (DFO). Our method of estimating 2

3 opportunistic observer effort provides a practical approach applicable globally to wildlife studies relying on opportunistically-collected data. Key words: GIS, spatial distribution, observer effort, opportunistic wildlife sightings, incidental sightings, citizen science, cetaceans, sea turtles, killer whales, humpback whales, hotspots. INTRODUCTION Conservation planning and environmental impact analyses require reliable information about the status of wildlife populations and their habitat requirements. For species that are wide-ranging, highly mobile or sparsely distributed, systematic studies of distribution and abundance are challenging and costly. In the case of cetacean populations, abundance and distribution estimates are typically derived from systematic line-transect surveys using dedicated platforms (vessels or aircraft), or by using platforms of opportunity such as ferries, cruise ships or ecotourism vessels (e.g., Hauser et al. 2006, Williams et al. 2006, Thomas et al. 2007, Leeney et al. 2012). Shore-based counts have also been shown to be reasonably accurate for some species that inhabit or migrate through inshore or coastal waters (Rugh et al. 2008), and capture-recapture methods using photo-identification have proven effective for more widely ranging species (e.g., Bigg et al. 1990, Ford and Ellis 1999, Ford et al. 2000, Calambokidis and Barlow 2004). When broad spatial or temporal (e.g. year-round) surveys of target cetaceans or other wildlife populations in marine and terrestrial environments are not practical or affordable, opportunistic data contributed by citizen science programs can be valuable in providing data on the occurrence and distribution of species at both local and global scales (Bonney et al. 2009, Dickinson et al. 2010, Hochachka et al. 2011). In the marine 3

4 environment, sightings collected opportunistically by networks of volunteer observers have provided reliable data on occurrence and distribution for conspicuous marine megafauna such as cetaceans, sharks, sea turtles, sea horses, and sunfish (family Molidae) (Evans and Hammond 2004, Goffredo et al. 2004, James et al. 2006, Fulling et al. 2007, Bradshaw et al. 2008, McPherson and Myers 2009, Witt et al. 2012). Many of these species are also targeted by ecotourism operations, including whale-watching and recreational diving, providing additional opportunities for citizens to contribute to sightings databases (Theberge and Deardon 2006, Ward-Paige and Lotze 2011). Although opportunistic observing programs have an established ability to provide meaningful wildlife sightings data, these programs are prone to limitations regarding sampling bias including variation in detectability of species, reliability of species identification, observer expertise, and spatial and temporal heterogeneity in sampling effort (Evans and Hammond 2004, Granek et al. 2008, Dickinson et al. 2010, Nagy et al. 2012). In particular, measures of observer effort are often lacking from opportunistically collected data. Estimates of observer effort add an important dimension to species occurrence data by allowing additional inferences to be made about species relative abundance and absence (Evans and Hammond 2004). Appropriate methods must be applied to detect and correct for such biases, in order to make the data useful for purposes such as measuring abundance, assigning conservation status and defining critical habitat. Estimating the observer effort used in collections of wildlife sightings makes it possible to distinguish hotspots of wildlife activity from areas where oberservers are concentrated. The issue of quantifying observer effort in wildlife monitoring has been addressed using several approaches. Some sightings networks for bird species (e.g., 4

5 Christmas Bird Count, North American Breeding Bird Survey) have formalized the recording of effort by conducting systematic volunteer surveys in which the timing and location of effort is organized in advance. Survey data collected in these programs are typically subjected to statistical corrections, in order to use the data for estimating population trends (Sauer et al. 1994, Peterson 1995, Link and Sauer 1999, Dunn et al. 2004). Other programs have used systematic surveys to determine baselines against which opportunistic sightings over broader areas or longer time scales may be calibrated (e.g., Witt et al. 2012). Ecological modeling approaches have also been developed to predict distribution and habitat use of a species from presence-only data such as platform of opportunity surveys, opportunistic sightings, museum collections, and other long-term datasets with temporal and/or spatial heterogeneity (Elith et al. 2006, Firestone et al. 2008, Fink et al. 2010, Hassall and Thompson 2010). We created a model using Geographic Information Systems (GIS) to construct a plausible distribution of the effort of volunteer observers reporting to the British Columbia Cetacean Sightings Network (BCCSN), which solicits opportunisticallycollected reports of cetaceans and sea turtles along the Pacific coast of Canada. Even though it was not possible to retroactively sighting effort associated with each sighting record in the BCCSN database, it was possible to group the network s observers into categories with predictable distributions. Our model develops those distributions for each observer category and then sums them to provide a density map of observer effort. When the model was completed we conducted a sensitivity analysis to determine which groups of observers were most important to model outputs and to provide direction for future research. We applied the effort model to our sightings database to calculate an 5

6 index of sightings density per unit effort (SDUE) for two cetacean species and compared them to critical habitat previously designated by fisheries managers. Our methodology can be applied to any opportunistically collected sightings data to increase the applicability of these data to conservation initiatives globally, by providing spatially explicit relative abundance and distribution of wildlife. METHODS Sightings network overview The BCCSN, founded in 2000, is a collaborative effort between the Vancouver Aquarium and Fisheries and Oceans Canada (DFO). The BCCSN collects sightings of whales, dolphins, porpoises and sea turtles from the waters adjacent to British Columbia (BC; Appendix A). Participants in the BCCSN observer network include whale watching operators, marine researchers, commercial and sport fishers, lighthouse keepers, park wardens and rangers, officers and crew on coastal ferries and coast guard vessels, fisheries observers, recreational boaters, and waterfront residents. Observers are recruited via targeted presentations to coastal communities and groups such as ecotourism naturalists, sport fishing guides, fisheries associations and yacht clubs. All presentations contain information on species identification and include a request for participation in the BCCSN. Observers report sightings via a toll-free phone number, address, online webform, or logbook. BCCSN staff respond personally to every report to verify species identification and sightings details, and to reinforce participation. 6

7 The BCCSN collects information on date and time, location, weather, species identification, group composition and animal behaviour. Observers self-report their confidence in their species identification (certain, probable, possible, and uncertain). In addition, BCCSN staff-assign a level of confidence in the observer s species identification ability (expert, reliable and unknown), and to each sighting reported by the observer (certain, probable, possible, and uncertain). After review by BCCSN staff, sightings are entered into a database. The sightings database is regularly analyzed to identify errors of data entry and reporting, in both species identification and location. Because each sighting report may include an individual animal or groups ranging up to hundreds of animals, we use the terms sighting and report to refer to a unique sighting event; the number of sightings cannot be used to infer the number of animals seen. The BCCSN database currently contains over 70,000 reports of 23 species of whales, dolphins and porpoise, as well as three species of sea turtles (Appendix B). For our analyses all sightings which were not labeled as either certain or probable by either the BCCSN or the observer were removed. Sightings housed in the BCCSN database that were initially collected by the DFO Cetacean Research Program were removed because these sightings are already corrected for effort. Sightings collected by OrcaNetwork, a sightings network based in Washington State, USA, were also excluded because they are generated primarily from US observers for whom we have no distribution of effort. In addition, sightings that lacked enough information to be assigned to an observer category were removed. Thus, the number of reliable sightings included in 7

8 this analysis was ~47,000 sightings out of the total of ~71,000 sightings in the BCCSN database. Estimating effort of observer categories Observer distribution was represented by seven different categories: 1) large vessel crew, 2) ecotourism operators, 3) residents of population centers, 4) lighthouse keepers, 5) parks users, 6) coastal workers, and 7) frequent observers. Two separate models were constructed (summer: May 01 Sept 30 and winter: Oct 01 April 30) to reflect the seasonal variation in the distribution of each of our major observer groups. Separate layer maps were built for each observer category in each season (summer and winter). The spatial distribution of each of the observer groups was determined on a 5km x 5km raster, and each grid cell in the raster was assigned a value of effort (= effort index). A raster consists of a matrix of cells (or pixels) organized into rows and columns (or a grid) where each cell contains a value representing information, such as temperature (ESRI). The 5km x 5km resolution was chosen as the maximum sighting distance from most observer platforms is less than 5km and it would be unlikely for an observer to accidently describe a cetacean in a neighboring grid cell. Effort estimates were tailored for each layer map for example, effort was measured for observers belonging to the lighthouse keepers layer map in the number of hours spent observing each day, whereas effort of large vessel crew was measured in the number of visits of the ship to each grid cell in each season. Values in each grid cell were smoothed using the filter tool in the Spatial Analyst extension in ArcGIS, which reduced potential errors or anomalies resulting from very high or very low effort index values, and prevented neighboring cells from having 8

9 drastically different effort index values. The rasters were then mathematically normalized by dividing each grid cell value by the maximum grid cell value in either the summer or winter layer map. The normalization provided a unitless effort index value, which allowed us to quantify effort output from each of our observer categories appropriately (i.e. number of hours on the water vs. number of visits to a particular grid cell) and to compare effort between seasons. This formed an effort index from 0 1 for each observer group in each season, representing the relative level of observer effort in each grid cell (Figs. 1 and 2; details below). For categories of observers who had travel patterns that typically radiated outward from a central location, we used cost-distance analysis using ArcGIS version 9.3 (ESRI) to map the distribution of effort. The BC coastline is a complex network of inlets, passages and islands. Therefore, a straight-line distance is not an accurate representation of vessel travel. We assumed that the effort expended by observers in these groups would be concentrated closest to the starting point and decrease the further an observer traveled away from that point. A cost-distance analysis requires the user to define a specific starting point, a cost raster (to establish the cost of traveling through each cell) and a maximum distance parameter. Starting points and maximum travel distances varied between observer groups, but the cost raster was constant. We assigned raster cells that occurred entirely on land (representing a very difficult path for boats to pass) a very high cost (1000) and ocean cells a very low cost (1). We used a Python script (see Supplement 1) to run each feature in the feature layers (i.e. in the park users observer category (see below) each individual park) one at a time through the cost-distance tool (ESRI; and see Supplement 9

10 ) before outputting each feature from the feature layer as an individual raster. Costdistance analyses raster output layers were then resampled to a 5km grid cell resolution and the cost-distance values were reclassified in equal intervals from 5 to 1 using the resample and reclassify tools in Spatial Analyst (ESRI). The final values from each feature were summed using the cell statistics tool (ESRI). Large Vessel Crew. Marine Communications and Traffic Services (MCTS), operated by the Canadian Coast Guard (CCG), monitors ship traffic in Canadian waters via a combination of radio contact, radar detection and satellite tracking. Ship location, direction and speed are recorded and stored in the MCTS database. Participation is required for most vessels longer than 20 m. We used MCTS data from 2003 to estimate the distribution of sightings effort from observers on large vessels. BCCSN observers in the large vessel crew category included marine pilots and tug boat operators, ferry captains and crew, offshore commercial fishing vessels, Canadian Coast Guard (CCG) personnel, and Canadian naval vessels. Although some of our observers that use large vessels are not recorded by the MCTS (e.g. CCG vessels, Canadian Navy), we assumed these large vessels would approach and leave the same ports and would be similarly constrained by shipping lanes and marine topography as ships reporting to the MCTS, and would therefore have travel patterns comparable to those reporting to the MCTS (Figs. 1a, 2a). The MCTS database was provided to the Canadian Wildlife Service to study ship movement as an indication of potential oil-spill distribution (O'Hara and Morgan 2006). The continuous stream of shipping data was sub-sampled so that each transiting of a ship through a grid cell was counted only once (O'Hara and Morgan 2006). Shipping intensity 10

11 was calculated by summarizing the total number of unique ship observations in each season and in each grid cell. The data were smoothed and normalized using the cell statistics tool and the methods described above. Residents of population centers. The population center residents category was created to capture the effort expended by observers who are coastal residents and recreational boaters. Waterfront resident observers tend to be located near population centers and spend a few hours daily scanning the water for passing wildlife. Recreational boaters typically travel from coastal towns and cities, spend a few hours on the water, and return to port the same day. We used cost-distance analyses to map distribution of residents of city centers, with a starting point of the population center and a maximum distance of 25km. We included all towns and cities along the BC coast with residents that report to the BCCSN (87 population centers). After completing the cost-distance analysis for each population center individually, it was weighted according to population size using the times tool in Spatial Analyst (ArcGIS version 9.3; ESRI). Rasters from each population center were then added together, smoothed, and normalized using the cell statistics tool (Figs. 1b, 2b). Lighthouse Keepers. There are 27 staffed light-stations along the BC coast, operated by Canadian Coast Guard personnel. Lighthouse keepers record meteorological and oceanographic data, and many (22) voluntarily participate in the BCCSN by scanning for marine mammals and sea turtles from their light-station. Correspondence with lighthouse keepers revealed that they can view up to 10 km from their high land-based vantage points, and many use boats to travel in a limited range (typical max distance of 10 km) near their station. 11

12 We mapped the 19 light-stations manned by people who report to the BCCSN (three other lighthouse keepers report but were put in the frequent observers category, see below). We then used a cost-distance analysis (maximum distance of 10 km) to map observer distribution. Each light-station was then weighted by the average daylight hours it experiences in summer or winter using the times tool in Spatial Analyst (ESRI). Rasters from each individual lighthouse keeper layer map were then summed using the cell statistics tool, smoothed, and normalized (Figs. 1c, 2c). Park Users. Over 22% of BC s coastline is designated as a Park, Ecological Reserve or Marine Protected Area by BC Parks or Parks Canada. Many BCCSN observers use these areas for conservation or recreation, including park wardens, interpreters and rangers, as well as residents of population centers. Based on conversations with Parks Canada Wardens and BC Parks Rangers, we found that park visitors were typically constrained to within 15 km of the park s shoreline while boating or paddling, and that the highest density of park users were concentrated along the shorelines of the coastal parks. We mapped all coastal parks in BC and clipped the park boundaries to the coastline using the clip tool (ESRI). We then mapped the shoreline of each park and ran the cost-distance routine on each park (starting point: shoreline; max distance 15km). We used monthly visitor statistics supplied by BC Parks and Parks Canada to determine the average number of visitors using each park daily in summer and winter. Not all parks had visitation statistics, so we determined the number of visitors per day per 25km 2 grid cell in parks that had that visitor statistics, and used the Inverse Distance Weighted (IDW; ESRI) tool to interpolate visitor use in parks that were data deficient from neighbouring 12

13 parks (within 150 km of deficient parks; measured from park polygon centroid to park polygon centroid). This method required an assumption that nearby parks receive similar visitation. To make this assumption more robust, we included a barrier polygon in the IDW analysis, which prevented parks in the lower mainland of BC, the central BC coast, Haida Gwaii, and the west coast of Vancouver Island from influencing each other (ESRI). This prevented visitation rates in parks with exceptionally high use (i.e. in the lower mainland of BC) and parks that were within 150km but geographically very isolated (i.e. no road access) from influencing one another. After assigning each park a season-specific visitation rate (number of visitors per day per grid cell in winter and summer) we used the times tool to multiply visitation rates by the length of the park shoreline, and then weighted each cost-distance park raster by this value. The resulting raster layer maps for each park were summed in each season using the cell statistics tool, and then smoothed and normalized (Fig. 1d, 2d). Ecotourism Operators. BCCSN observers in the ecotourism category included commercial whale watch companies and kayak or charter boat companies focused on whale watching. We limited this layer map to companies who have reported 50 sightings, to eliminate bias from companies who reported infrequently despite expending effort and that would indicate a false lack of cetacean presence in the areas they traveled. Eighteen ecotourism companies had reported over 50 cetacean sightings to the BCCSN, however 8 of those companies were assigned to the frequent observers category (see below), leaving 10 companies that were included in the ecotourism operators observer category of our model. Companies were divided into groups that a) start and end trips at a 13

14 specific point i.e. commercial whale watch daytrips, and b) companies that run boat charter or kayak-based trips that are typically multi-day. We distributed surveys to each commercial whale watching company in our ecotourism operators category requesting their home port (where they start and end most trips), the number of boats on the water daily in summer vs. winter, and the average trip length in hours in summer vs. winter. Correspondence with this user group revealed that the typical trip length for whale watch operators was specific to their home-port (Campbell River: 100km; Vancouver: 75km; Victoria: 35km; Telegraph Cove/Port McNeil: 25km; Ucluelet/Tofino: 15 km). We mapped each whale watch operator s homeport and used a cost-distance analysis with their home-port associated maximum distance to map effort distribution of each whale watching operator. We multiplied the number of boats on the water daily with the number of hours per trip and used this value to weight the cost-distance raster using the times tool (ESRI). The distribution of whale-focused boat charter and kayaking trips was better represented by tracklines of typical routes taken by each company. A cost-distance analysis with a 5km maximum was used to represent declining effort the further one strayed from the trackline. We multiplied the number of boats each company had on the water daily in each season with the typical daily hours on the water for each boat, and divided this product by the number of grid cells crossed by the trackline. The output was used to weight the cost-distance raster and a map layer was built for each boat-charter or kayak focused ecotour operator. Individual Ecotourism operators layer maps were summed using the cell statistics tool, then smoothed and normalized (Fig 1e, 2e). 14

15 Coastal Workers. Many BCCSN observers spend time working on the BC coast, including marine mammal scientists, fisheries observers, employees of sport fishing lodges, employees of aquaculture operations, sea plane pilots, and commercial fishers. The four marine mammal scientists who regularly report to the BCCSN were designated as frequent observers, as was the boat-based fisheries observer organization in BC (see below), however employees of resorts and lodges, fish farms, coastal airlines and fisheries observers using aircraft to monitor commercial fisheries were included in the coastal workers layer map. We mapped the location of each of the ten resorts or lodges that report to the BCCSN. We used a cost-distance analysis to represent distribution of guides and guests of fishing lodges. Correspondence with fishing guides indicated that the typical maximum distance traveled daily was 35 km from the lodge, so we used this as the maximum distance in the cost-distance analyses. The cost-distance rasters were weighted by average daylight hours in summer. None of the fishing lodges that report to the BCCSN operate in winter, so the winter lodge and resorts raster was given a constant value of zero. We mapped all fish farms that were operated by companies who report sightings to the BCCSN (two companies; 70 fish farms). We ran a cost-distance analysis on clusters of fish farms rather than on each individual farm, because the same crew generally tends to all farms in the vicinity. This was achieved by running the costdistance on all farms at once while in one raster layer map, rather than separating out each farm to its own raster and then summing the raster layer maps (as done with all previous observer categories). The maximum distance used in the cost-distance analysis 15

16 was 15km because fish farmers typically tend to farms within a 15km radius of the center of the farm cluster. The aquaculture layer map was weighted by summer or winter daylight hours. The aquaculture and fishing lodge layer maps were the only raster layer maps in the coastal workers category that were given effort values based on cost-distance analyses, so they were added together using the cell statistics tool, then smoothed, and normalized. Two sea plane companies report sightings to the BCCSN. These companies were contacted and provided the number of flights and flight routes covered in winter and summer. Their observation areas were mapped and weighed by the number of planes flown over the area per day divided by the number of grid cells covered by the observation area. The same method was employed for inshore and offshore fisheries however we were unable to get distribution data from fishers themselves, and instead used distribution of the fisheries enforcement officers. We contacted fisheries enforcement staff and found their aerial surveys were conducted within 60 nautical miles of the BC shoreline. Thus, we created a 60 km buffer of the political border of BC Area within this buffer was weighed by the number of fisheries observers flying per day per season, divided by the number of grid cells covered by their observation area. The sea plane pilots and fisheries categories were the only raster layer maps in the coastal workers category that were given effort index values based on visits to grid cell, so they were added together using the cell statistics tool, then smoothed and normalized. The aquaculture + fishing lodge layer map was summed with the sea plane + fisheries layer map and normalized to finalize the coastal workers layer map (Figs. 1f, 2f). 16

17 Frequent Observer category. Preliminary iterations of this model using only the six categories described above revealed that certain areas with known high observer effort were not being captured. For example, we have observers in Powell River and Barkley Sound that spend a large portion of their time observing the water and are very consistent in reporting what they see. The frequent observers category was designed to capture the effort of these and other prolific observers in our network. We analyzed our sightings to determine what constitutes a frequent observer. We found that a relatively small number of observers (30) had contributed 450 sightings to the program (Appendix C). We set the threshold at 450 sightings or more to define our frequent observers. We were able to get distribution information from 20 of the 30 selected frequent observers, and used these 20 observers to build our frequent observers layer map. Our 20 frequent observers included 11 ecotourism operators, 1 crew of large vessel traffic, 3 lighthouse keepers and 2 residents of population centers). The distribution of frequent observers that could also belong to one of the categories above was mapped the same way as the categories above in terms of either running a cost-distance analyses on their home port (i.e. a frequent observer who is a commercial whale watch operator or a lighthouse keeper) or determining the number of visits to a grid cell (i.e. a frequent observer who is also a seaplane pilot or a ferry captain). This approach will facilitate future iterations of the model where different observers may be designated as frequent observers and our current frequent observers may move back to their alternative categories. These frequent observers were asked how many days they spend on the water in summer vs. winter and the typical number of hours per day they spend observing. The number of days on the water was multiplied by the 17

18 number of hours spent observing, and then divided by the season length to determine daily effort by which to weight their layer maps. Frequent observers who could not be associated with the 6 categories (n=3) for which effort was defined were asked to draw a map of their spatial distribution. These frequent observers were also asked how many days they spend on the water in summer vs. winter and the typical length of day they spend observing. We digitized the handdrawn maps from these frequent observers and weighted each 5km x 5km grid cell within their spatial distribution. Grid cells were weighted with the product of the number of days spent on or observing the water times the number of hours spent observing per day, and divided by both the number of days in the season (summer vs. winter) and the number of grid cells covered by their distribution. Frequent observers whose distribution was modeled using cost-distance analyses methods were summed, smoothed, and normalized separately from frequent observers whose distribution was better represented by the number of visits to each grid cell. The two frequent observers layer maps where then summed and re-normalized following the same methods as those used for the coastal workers category (Figs. 1g, 2g). Sightings Effectiveness A sightings effectiveness (SE) value was created to compensate for differences in the effectiveness of observer categories in making and reporting sightings. To calculate SE, we identified sample regions in each of the winter and summer models where the effort distribution of 6 observer categories overlapped (Appendix D). In each sample region, the summed value of each of the observer effort index categories was divided by the area of the study region to obtain an observer density value (OD) in each region and 18

19 in each season. In each sample region, the proportion of sightings (PS) belonging to each observer category was calculated by dividing the number of sightings obtained from each category by the total number of all cetacean sightings in that specific sample region. The proportion of sightings and observer density values for each observer category were summed separately in each region and season (Appendix E). Sightings effectiveness in each region (SE reg ) was calculated for each of the observer categories using (1) where, SE reg is the sightings effectiveness for a specific observer category in a specific region, PS reg is the proportion of sightings made by that observer category in that region, and OD reg is the observer density of that observer category in that region based on the effort index values in that region (Appendix E). The total sightings effectiveness for each observer category was determined as the sum of the regional sightings effectiveness values using (2) where, SE is the total sightings effectiveness and SE reg is the regionally specific sightings effectiveness (Table 1). To obtain a coefficient that could be used to weight each of the observer categories, the relative sightings effectiveness (RSE) was calculated by dividing each observer category s SE value by the sum of all the observer categories SE values in each season, using (3) 19

20 where, RSE is the relative sightings effectiveness of each observer group in each season, SE is the total sightings effectiveness of each group in a season, and SE max is the sum of the sightings effectiveness values of all groups in each season (Table 1). Final Effort Model Once all seven raster layer maps were created, normalized and weighted by the corresponding relative sightings effectiveness value for either summer or winter, we combined them to form a single raster layer map for each season using (4) where, FE is the final effort value in each grid cell specific to each season, LV is the large vessel crew category (raster of effort index), SE is sightings effectiveness for a specific observer category, RC is the residents of population centers category, LK is the lighthouse keepers category, PU is the park users category, ET is the ecotourism operators category, CW is the coastal workers category and FO is the frequent observers category. The final effort model layer map in each season, (resulting from Equation 4) was smoothed, and then mathematically normalized using the methods described above (models for both seasons were divided by maximum effort value of either season, i.e. of summer). Sensitivity Analysis We conducted a sensitivity analysis to explore how changes in observer effort index and relative sightings effectiveness changed our model results, using R version 2.15 (Supplement 2). We used a Monte Carlo analysis to vary one input parameter (EI raster or 20

21 SE raster ) at a time by ±25% of its usual input value (i.e. multiplied an entire observer category s effort index raster by ±25% of its value in each grid cell, or multiplied its relative sightings effectiveness by ±25%) while leaving all other values constant. We used the Monte Carlo to run 1000 model iterations each time we varied a parameter, assuming a uniform distribution of values between -25% and +25%. Input parameters that, when varied, resulted in the largest change to mean effort per grid cell were deemed the most sensitive. Sightings Density per Unit Effort (SDUE) We analyzed the sightings records of humpback whales (Megaptera novaengliae) and killer whales (Orcinus orca) in both summer and winter. We used these two species to test our model because critical habitat has previously been assigned to them (DFO 2010, 2011) and can provide a measure of qualitative ground-truthing to our SDUE predictions. For each species, we calculated the Sightings Density (SD) using (5) where, SD is the sightings density per grid cell, S is the number of sightings reported in that grid cell, and A is the area of the grid cell in km 2 For each grid cell, we divided the sightings density by the corresponding Effort Index (EI) to derive a Sightings Density per Unit Effort (SDUE) using (6) where, SDUE is the sightings density per unit effort, SD is the sightings density, and EI is the effort index. For the SDUE calculation, all EI values below were discarded to prevent inflated SDUE values in areas of exceptionally low effort. 21

22 All figures where mapped in ArcGIS version 9.3 (ESRI) using geographic interval symbology. This symbolic classification system was designed for continuous data that is not normally distributed (ESRI). RESULTS Effort Model Summer and winter effort models reveal hotspots in observer effort particularly around southern Vancouver Island (Figs. 3a and b). Frequent observers and ecotourism operators were the most effective sightings groups in summer and the influence of these layer maps can be seen in the final effort model. For example, the ferry routes up the central coast of BC and across to Haida Gwaii are hotspots of effort in summer due to an exceptionally well-reporting ferry crew in our frequent observers category, as are waters surrounding the home-port of Vancouver where several frequent observers and ecotourism operators are based (Fig. 3a). Large vessel crew, frequent observers, and ecotourism operators had the highest sightings effectiveness in winter, and their influence can be seen in the winter effort model - most evident in the shipping traffic lanes which appear as lines radiating north west from Vancouver and Vancouver Island (Fig. 3b). Sensitivity Analysis Sensitivity analyses revealed that summer effort model outputs were most sensitive to changes in effort index (EI) and relative sightings effectiveness (RSE) values of the ecotourism and frequent observers category (Fig. 4). Changing large vessel crew, and coastal workers parameters resulted in small changes to model outputs, and changes to other observer categories resulted in negligible changes to model outputs. Winter effort model outputs were most sensitive to changes in ecotourism operators and large vessel 22

23 crew categories (Fig 4; note secondary Y axis for winter model). Changing coastal workers parameters resulted in small changes to model outputs, and changes to other observer categories resulted in negligible changes to model outputs. Because of the structure of Equation 4, multiplying either the RSE or the EI values by a constant between + and 25% of their respective mean values (see methods) resulted in the same amount of change to the mean model output. Therefore, one figure (Fig. 4) is shown to depict change to either parameter (RSE or EI) within an observer category. Sightings Density per Unit Effort (SDUE) Mean (±SD) humpback whale SDUE was 7.09 (± 37.71) sightings per grid cell (1 grid cell = 25 km 2 ) per unit of effort in summer and 1.22 ± sightings per grid cell per unit of effort in winter. Humpback whale SDUE in summer was highest off the west coast of Vancouver Island, Queen Charlotte Sound, southeastern Haida Gwaii, central Hecate Strait, and in Caamano Sound. In winter, humpback whale SDUE was greatest off the west coast of Vancouver Island, and in Dixon Entrance, and Caamano Sound (Fig. 5 a and b). Mean (±SD) killer whale SDUE was 2.65 (± 16.96) sightings per grid cell per unit of effort in summer and 0.57 ± 5.77 sightings per grid cell per unit of effort in winter. In summer, the SDUE of killer whales was highest off southeastern Vancouver Island - mainly in Canada s Southern Gulf Islands and USA s San Juan Islands, and in Queen Charlotte and Johnstone Straits, Caamano Sound, Dixon Entrance, and along the southeastern coast of Haida Gwaii (Fig. 6 a). In winter, the SDUE of killer whales was 23

24 highest in the Southern Gulf Islands, the San Juan Islands and the northern Strait of Georgia (Fig. 6 b). DISCUSSION Our effort model allowed us to use over 35,000 opportunistically collected cetacean and sea turtle sightings, at times covering geographic areas not often surveyed by other survey methods, to inform us of cetacean use of the BC coast. In particular, by adjusting data for effort, we were able to deduce areas of particular importance to frequently observed wildlife (humpback whales and killer whales) in BC, some of which had not been previously identified. Our method provides a cost-effective and geographically wide-ranging complement to formally collected scientific data in BC, and provides a methodology applicable to other opportunistically collected wildlife sightings data globally. Observer Effort Model The distribution of observer effort in our model is a plausible representation of where participants in the BCCSN spend their time observing the water. The model revealed high observer effort throughout the Strait of Georgia, particularly around the port of Vancouver in both seasons, and moderate to high effort through the Inside Passage on the central BC coast and around Prince Rupert. In winter, observer effort was less pronounced on the northern and central coasts of BC than it was in summer. Low observer effort occurred in most offshore waters in BC and along the northwest coast of Vancouver Island in both seasons. Effort was also low on the coastline around Haida Gwaii in both seasons. Similar spatio-temporal distribution patterns were observed in public sightings of basking sharks (Cetorhinus maximus) around the UK; there, observer 24

25 effort was concentrated in summer months, in straits and passages between islands, on islands with prominent views of the sea, and in near-shore regions with deep-water access (Witt et al. 2012). The distribution of mariners, including ferry crew, commercial and recreational fishers, recreational boaters, and shore-based nature enthusiasts, is necessarily constrained by the geography of each region, so our model can be applied to any region where movements of ocean users can be predicted (Leeney et al. 2012). Sensitivity analyses revealed that ecotourism operators, frequent observers, coastal workers, and large vessel crew are the most important contributors to the effort model in both seasons. Some ferry crews in our region were particularly motivated to provide consistent reports of cetacean sightings. In other regions of the world, ferry vessels have been used by researchers as platforms of opportunity for surveying wildlife (e.g., Leeney et al. 2012), but engaging ferry crews to directly report cetacean sightings, as in BC, may provide quality sightings at minimal cost to researchers. Similarly, ecotourism operators are a major contributor to sightings of cetaceans and sea turtles in BC, as in areas worldwide. Large marine species such as whales, dolphins, sea turtles and sharks are increasingly valued in terms of their economic draw for ecotourism, and some sightings networks have successfully engaged ecotourism operators and participants in collecting sightings data of them (Goffredo et al. 2004, Theberge and Deardon 2006, Higby et al. 2010, Ward-Paige and Lotze 2011). Recreational fishers can also be valuable allies in species monitoring (Granek et al 2008). Retention of observers in these categories should be a priority to the BCCSN as should documenting their effort accurately. Collecting long-term data on the effort of observers 25

26 in these groups could contribute to longitudinal studies documenting changes in cetacean hotspots over time. Sightings density per unit effort hotspots Our model revealed hotspots in the summer distribution of humpback whales along the shelf waters from Vancouver Island up to Haida Gwaii, along the edge of Dogfish Bank in Hecate Strait and over deep water canyons in Queen Charlotte Sound (Fig. 7a). In winter, Dixon Entrance and the west coast of Vancouver Island provided high SDUE for humpback whales; however, to a lesser extent than seen in the same areas in summer. Humpback whales use the BC coast primarily for foraging (DFO 2010), and migrate southward to warm waters in winter months where they reproduce. The focused effort by humpback whales on foraging in summer months (Gregr et al. 2000) may explain a possible affinity for shelf waters where prey is likely plentiful. Our model predictions of lower SDUE of humpback whales in winter are supported by seasonal migration patterns of this species typically moving to tropical waters for calving and breeding in winter (DFO 2010; and references therein). The spatial and temporal (seasonal) distribution of humpback whale hotspots identified by our model were similar to distributions observed using line transect surveys in recent years (Ford et al. 2010). Our model revealed hotspots in killer whale distributions in Dixon Entrance, southeastern Haida Gwaii, Caamano Sound, Queen Charlotte, Johnstone, and northern Georgia Straits, and the Southern Gulf Islands, San Juan Islands, and southern Vancouver Island in summer. In winter months, high SDUE values occurred in Queen Charlotte, Johnstone, and northern Georgia Straits, and the Southern Gulf Islands, San Juan Islands, and southern Vancouver Island - but to a lesser extent than in summer. SDUEs were 26

27 exceptionally strong throughout southeastern Haida Gwaii, the Southern Gulf Islands and San Juan Islands, and in the Johnstone Strait area. The latter two of these areas are wellknown globally as tourist destinations for resident killer whale viewing. Johnstone Strait and the Southern Gulf Islands and San Juan Islands have previously been identified as important habitat for resident killer whales (Ford 2006, DFO 2011). Occurrences of Bigg s (transient) killer whales in the Strait of Georgia have increased year round over the last few decades and Bigg s (transient) sightings are now common in this area (L.G. Barrett-Lennard, pers. obs.). The hot-spots identified in southeastern Haida Gwaii indicate an area that should receive more research attention to verify its importance as habitat for killer whales. Line-transect ship-based surveys of BC have revealed similar patterns of killer whale hotspots (Ford et al. 2010); which supports our model s predictions of where areas of high SDUE occur, and serves as a qualitative validation of the effort model. Comparison of SDUE hotspots and federally designated Critical Habitat To ground-truth model predictions of SDUE hotspots we compared known killer whale and humpback high-use areas (listed as critical habitat) to model-predictions of high SDUE areas of each species. The Humpback Whale Recovery Strategy (DFO 2010) used hotspots of humpback whale sightings identified via surveys conducted from 1984 to 2009 to identify four areas of proposed Critical Habitat. When overlaid with our SDUE results, we found these proposed areas contained relatively high SDUE values in our summer effort model (Figure 7a and b). Of the ~12,000 km 2 identified as Critical Habitat (DFO 2010), ~

28 km 2, or about 50%, had above average SDUE in summer, and ~3500 km 2, or about 30%, had above average SDUE in winter. We also compared the SDUE of killer whales with the Canadian Critical Habitat identified in the Recovery Strategy for Northern and Southern Resident Killer Whales (DFO 2011), with Critical Habitat in US waters, as identified under the Endangered Species Act (ESA), and with the potential Critical Habitat in Canadian waters as identified by Ford (2006). Of the approximate 3,500 km 2 area identified as Critical Habitat in Canadian waters, above average SDUEs occurred in ~2,000 km 2, or ~60% of the Critical Habitat in both summer and winter. Of the approximate 7,000 km 2 of area identified as Critical Habitat in US waters, above average SDUEs occurred in ~1300 km 2, or about 20% of the critical habitat in summer and in winter. Of the 5700km 2 area identified as potential critical habitat by DFO, above average summer SDUEs occurred in ~1300 km 2, or ~20% of the area. In winter, above average SDUEs occurred in ~900 km 2, or ~15% of the area (Fig. 8a and b). Any area with above average SDUE is likely an important area for a species, however not all areas with elevated SDUEs may designated as Critical according the Species at Risk Act criteria (Species at Risk Act, 2013). Our model identified many areas on the BC coast as having higher than average SDUEs, which are not currently encompassed in Critical Habitat designations. Both recovery strategies noted that the identified Critical Habitat is not exhaustive (DFO 2010, 2011). Given the overlap of above average SDUE values with areas currently identified as Critical Habitat or potential Critical Habitat for killer whales and humpback whales, our model is a useful tool to aid in the identification of further Critical Habitat. 28

29 Method limitations Our model can be used to determine SDUE of any cetacean species in BC waters. One drawback is that we were unable to assign a quantitative, true value of effort of each of our observer categories and were thus required to report effort as a relative index. Although effort can be directly compared between seasons and areas, the lack of a unit of effort is problematic because SDUE values are not comparable with other methods of collecting SDUE or catch per unit effort of other species. There may be potential to scale our SDUE predictions against areas with known sightings per unit effort and use that ratio to inform the rest of the BC coast with discrete units. Another limitation to our model is the identification of some high SDUE values in some areas where effort was exceptionally low i.e. where high SDUE may be an artifact of low effort. For example, very high SDUE values emerged in some offshore areas offshore with effort values of < We minimized this occurrence by removing all areas with an effort index below units from our calculations of SDUEs, however other areas of low effort combined with a few sightings may still create high SDUE values. Of course, this provides direction for future research to determine if these exceptionally high SDUE values are an artifact of the model, or true hotspots. Conversely, some of our observers are concentrated in hotspots and so SDUEs may be under-represented where effort is exceptionally high. This is more likely in Johnstone Strait and southern Vancouver Island (particularly off the port of Vancouver) where high effort exerted by our frequent observer category may dampen the strength of SDUE outputs (i.e. SDUE may be higher than presented in our analysis). To this end, it would be useful to focus future research on determining true SDUE values in areas where 29

30 model-predictions for SDUE were very high and effort was very low, and in areas where effort was very high. From a regional management perspective, a limitation to our model is the lack of genetic, ecotype, or individual identification of cetaceans within the sightings database. This is a particular problem for determining what habitats are important to killer whales in BC. Three ecotypes of killer whales occur in BC: residents, Bigg s (transients), and offshores (Ford et al. 2000). Different ecotypes of killer whales have different conservation status for example, the southern residents are listed as endangered by the Canadian Species at Risk Act (SARA) and by the US Endangered Species Act, whereas Bigg s (transient) and offshore killer whales are listed by SARA as threatened and therefore, habitat planning objectives should differ between these populations. Different ecotypes of killer whales also have unique prey requirements. For example, resident killer whales are fish-eaters whereas Bigg s killer whales eat other marine mammals (Ford and Ellis 1999, Ford et al. 2000). Therefore, each ecotype is likely to have different habitat requirements based in part on the geographic distribution of their prey. However, with the data we currently receive, splitting the killer whale sightings up by ecotype would result in high loss of data as most observers are not able to differentiate between ecotypes (i.e. of 22,647 killer whale sightings 9,780 have no ecotype assigned, 2,999 are assigned transient or possible transient, and 9,790 are assigned resident or possible resident ). Our data regarding killer whale SDUE should thus be used as a hypothesis of areas on the BC coast important to all killer whale ecotypes, which could be assessed by scientists for occurrences of each ecotype of killer whale within each high SDUE area identified in our model. 30

31 Our model provides testable hypotheses of where observer effort and cetacean hotspots exist on the BC coast. Predictions of cetacean hotspots identified by this model could be ground-truthed by comparing the findings to those of field studies where effort was recorded (i.e. DFO cetacean surveys; Ford et al. 2010). Our model provides an important step in the assessment of habitat requirements of cetaceans in BC waters and will be an asset to future research in conservation management of marine wildlife in BC. Finally, our model adds to the growing range of methods that aim to optimize the usefulness of opportunistically collected data (Caruana et al. 2006, Fink et al. 2010, Hochachka et al. 2011, Catlin-Groves 2012). Citizen science programs are increasingly recognized as a useful source of data for determininge critical habitat, measuringe population dynamicsand detecting ecosystem changes their impacts on biodiversity (McPherson and Myers 2009, Schmeller et al. 2009, Devictor et al. 2010, Magurran et al. 2010). The novel methodology we applied to our sightings data is widely applicable to any opportunistically collected or citizen science dataset globally, and can enhance the application of citizen science data in determining areas of importance to wildlife. ACKNOWLEDGMENTS The BC Cetacean Sightings Network is a collaboration between the Vancouver Aquarium Marine Science Center, Fisheries and Oceans Canada and the Government of Canada Habitat Stewardship Program for Species at Risk. We thank the 3,600 observers who have reported sightings and the Vancouver Aquarium and the volunteers who have entered our verified sightings into the database. J.K.B. Ford of Fisheries and Oceans Canada has provided guidance and advice since the BCCSN s inception in The Cetacean Research Program at the Pacific Biological Station provided shapefiles showing 31

32 critical habitat. G.M. Ellis, and B. Gisbourne provided advice on methods for modeling the distribution of some of our observers. We thank P. O Hara of the Canadian Wildlife Service and G. McGowan for providing the relevant data from the MCTS database. B. Lindsay and N. Koshure contributed to earlier drafts of the manuscript. G. Allen provided valuable assistance with methodologies in ArcGIS. M. Foster provided the observer effort layer for the Parks Users category and programmed Python and Microsoft Access scripts. R. Joy provided statistical advice. We thank C. Nordstrom and two anonymous reviewers for comments which greatly improved this manuscript. The BC Cetacean Sightings Network received additional support for this project from the National Fish and Wildlife Foundation (Grant # ), the Royal Caribbean Cruise Line Ocean Fund, North Gulf Oceanic Society, and British Columbia Institute of Technology (BCIT). LITERATURE CITED Bigg, M. A., P. F. Olesiuk, G. M. Ellis, J. K. B. Ford, and K. C. Balcomb Social organization and genealogy of resident killer whales (Orcinus orca) in the coastal waters of British Columbia and Washingon State. Report of the International Whaling Commission Special Issue 12: Bonney, R., C. B. Cooper, J. Dickinson, S. Kelling, T. Phillips, K. V. Rosenberg, and J. Shirk Citizen science: A developing tool for expanding science knowledge and scientific literacy. BioScience 59: Bradshaw, C. J. A., B. M. Fitzpatrick, C. C. Steinberg, B. W. Brook, and M. G. Meekan Decline in whale shark size abundance at Ningaloo Reef over the past 32

33 decade: The world's largest fish is getting smaller. Biological Conservation 141: Calambokidis, J., and J. Barlow Abundance of blue and humpback whales in the eastern North Pacific estimated by capture-recapture and line-transect methods. Marine Mammal Science 21: Caruana, R., M. Elhawary, A. Munson, M. Riedewald, D. Sorokina, D. Fink, W. M. Hochachka, and S. Kelling Mining citizen science data to predict prevalence of wild bird species. Pages in Proceedings of the 12th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (Ungar, L., Craven, M., Gunopulos, D., Eliassi-Rad, T., Eds.). ACM Press, New York, New York. DFO (Fisheries and Oceans Canada) Recovery strategy for the North Pacific humpback whale (Megaptera novaeangliae) in Canada. Fisheries and Oceans Canada, Ottawa. ix +61p. DFO (Fisheries and Oceans Canada) Recovery strategy for the northern and southern resident killer whales (Orcinus orca) in Canada. Fisheries and Oceans Canada, Ottawa. ix + 80p. Dunn, E. H., C. M. Francis, P. J. Blancher, S. Roney Drennan, M. A. Howe, D. Lepage, C. S. Robbins, K. V. Rosenberg, J. R. Sauer, and K. G. Smith Enhancing the scientific value of the Christmas Bird Cound. The Auk 122: Elith, J., C. H. Graham, R. P. Anderson, M. Dudik, S. Ferrier, A. Guisan, R. J. Hijmans, F. Huettmann, J. R. Leathwick, A. Lehmann, J. Li, L. G. Lohmann, B. A. Loiselle, G. Manion, C. Moritz, M. Nakamura, Y. Nakazawa, J. M. Overton, A. 33

34 T. Peterson, S. J. Phillips, K. S. Richardson, R. Scachetti-Pereira, R. E. Schapire, J. Soberon, S. Williams, M. S. Wisz, and N. E. Zimmermann Novel methods to improve predition of species' distributions from occurrence data. Ecography 29: ESRI. [Environmental Systems Research Institute] ArcGis 9.3. Environmental Systems Research Institute, Redlands, California, USA. Evans, P. G. H., and P. S. Hammond Monitoring cetaceans in European waters. Mammal Review 34: Fink, D., W. H. Hochachka, B. Zuckerberg, D. W. Winkler, B. Shaby, M. A. Munson, G. Hooker, M. Riedewald, D. Sheldon, and S. Kelling Spatiotemporal exploratory models for broad-scale survey data. Ecological Applications 20: Firestone, J., S. B. Lyons, C. Wang, and J. J. Corbett Statistical modeling of North Atlantic right whale migration along the mid-atlantic region of the eastern seaboard of the United States. Biological Conservation 141: Ford, J. K. B An assessment of critical habitats of resident killer whales in waters off the Pacific coast of Canada. DFO Can. Sci. Advis. Sec. Doc. 2006/072. Ford, J. K. B., R. M. Abernethy, A. V. Phillips, J. Calambokidis, G. M. Ellis, and L. M. Nichol Distribution and relative abundance of cetaceans in western Canadian waters from ship surveys, Canadian Technical Report of Fisheries and Aquatic Sciences No Ford, J. K. B., and G. M. Ellis Transients: Mammal hunting killer whales. UBC Press, Vancouver, British Columbia. 34

35 Ford, J. K. B., G. M. Ellis, and K. C. Balcomb Killer whales: the natural history and geneology of Orcinus orca in British Columbia and Washington. UBC Press, Vancouver, British Columbia. Fulling, G. L., D. Fertl, K. Knight, and W. Hoggard Distribution of Molidae in the northern Gulf of Mexico. Gulf and Caribbean Research 19: Goffredo, S., C. Piccinetti, and F. Zaccanti Volunteers in marine conservation monitoring: a study of the distribution of seahorses carried out in collaboration with recreational scuba divers. Conservation Biology 18: Granek, E. F., E. M. P. Madin, M. A. Brown, W. Figueira, D. S. Cameron, Z. Hogan, G. Kristianson, P. de Villiers, J. E. Williams, J. Posts, S. Zahn, and R. Arlinghaus Engaging recreational fishers in management and conservation: Global case studies. Conservation Biology 22: Gregr, E. J., L. Nichol, J. K. B. Ford, G. M. Ellis, and A. W. Trites Migration and population structure of northeastern Pacific whales off coastal British Columbia: An analysis of commercial whaling records from Marine Mammal Science 16: Hassall, C. and D. J. Thompson Accounting for recorder effort in the detection of range shifts from historical data. Methods in Ecology and Evolution 1: Hauser, D. D. W., G. R. Van Blaricom, E. E. Homes, and R. W. Osborne Evaluating the use of whalewatch data in determining killer whale (Orcinus orca) distribution patterns. Journal of Cetacean Research and Management 8:

36 Higby, L. K., R. Stafford, and C. G. Bertulli An evaluation of ad hoc presenceonly data in explaining patterns of distribution: Cetacean sightings from whalewatching vessels. International Journal of Zoology Article ID pp. Hochachka, W. M., D. Fink, R. A. Hutchinson, D. Sheldon, W.-K. Wong, and S. Kelling. Data-intensive science applied to broad-scale citizen science. Trends in Ecology and Evolution 27: James, M. C., S. A. Sherrill-Mix, K. Martin, and R. A. Myers Canadian waters provide critical foraging habitat for leatherback turtles. Biological Conservation 133: Kremer, H., L. Koch, B. Adloff, B. Baschek, O. Schneider, and W. Fischer Results from a pilot study on sightings of harbour porpoises, Phocoena phocoena, at the North-Sea coast of Schleswig-Holstein, F.R.G. Archive of Fishery and Marine Research 42: Leeney, R. H., M. J. Witt, A. C. Broderick, J. Buchanan, D. S. Jarvis, P. B. Richardson, and B. J. Godley Marine megavertebrates of Cornwall and the Isles of Scilly: relative abundance and distribution. Journal of the Marine Biological Association of the United Kingdom 92: Link, W. A., and J. R. Sauer Controlling for varying effort in count surveys - an analysis of Christmas Bird Count data. Journal of Agricultural, Biological, and Environmental Statistics 4: Magurran, A. E., S. R. Baillie, S. T. Buckland, J. McP. Dick, D. A. Elston, E. M. Scott, R. I. Smith, P. J. Somerfield, and A. D. Watt Trends in Ecology and Evolution 25:

37 McPherson, J. A. and R. A. Myers How to infer population trends in sparse data: examples with opportunistic sighting records for great white sharks. Diversity and Distributions 15: Nagy, C., K. Bardwell, R. F. Rockwell, R. Christie, and M. Weckel Validation of a citizen science-based model of site occupancy for Eastern Screech Owls with systematic data in suburban New York and Connecticut. Northeastern Naturalist 19(6): O'Hara, P. D., and K. H. Morgan Do low rates of oiled carcass recovery in beached bird surveys indicate low rates of ship-source oil spills? Marine Ornithology 34: Peterson, A. P Erroneous party-hour data and a proposed method of correcting observer effort in Christmas Bird Counts. Journal of Field Ornithology 66: R Core Development Team R: a language and environment for statistical computing. Vienna, Austria. ISBN , URL Reid, J. B., P. G. H. Evans, and S. P. Northridge Atlas of cetacean distribution in north-west European waters. Joint Nature Conservation Committee, Peterborough. Rugh, D. J., M. M. Muto, R. C. Hobbs, and J. A. Lerczak An assessment of shorebased counts of gray whales. Marine Mammal Science 24: Sauer, J. R., B. G. Peterjohn, and W. A. Link Observer differences in the North American Breeding Bird Survey. The Auk 111:

38 Species at Risk Act Government of Canada, S.C. 2002, c.29. Accessed April 2013: Theberge, M., and P. Dearden Detecting a decline in whale shark (Rhincodon typus) sightings in the Andaman Sea, Thailand, using ecotourist operatorcollected data. Oryx 40: Thomas, L., R. Williams, and D. Sandilands Designing line transect surveys for complex survey regions. Journal of Cetacean Research and Management 9:1-13. van der Meij, S. E. T., and C. J. Camphuysen The distribution and diversity of whales and dolphins (Cetacea) in the southern North Sea. Lutra 49:3-28. Ward-Paige, C. A., and H. K. Lotze Assessing the Value of Recreational Divers for Censusing Elasmobranchs. PLoS ONE 6(10): e doi: /journal.pone Williams, R., S. L. Hedley, and P. S. Hammond Modeling distribution and abundance of Antarctic baleen whales using ships of opportunity. Ecology and Society 11:[online] URL: Witt, M. J., T. Hardy, L. Johnson, C. M. McClellan, S. K. Pikesley, S. Ranger, P. B. Richardson, J.-L. Solandt, C. Speedie, R. Williams, and B. J. Godley Basking sharks in the northeast Atlantic: spatio-temporal trends from sightings in UK waters. Marine Ecological Progress Series 459:

39 TABLE 1. Sightings Effectiveness (SE) and Relative Sightings Effectiveness (RSE) in a) summer and b) winter. Observer effort category Sightings effectiveness Relative sightings effectiveness Large vessel crew Residents of population centers Lighthouse keepers Park users Ecotourism operators Coastal workers Frequent observers Sum Observer effort category Sightings effectiveness Relative sightings effectiveness Large vessel crew Residents of population centers Lighthouse keepers Park users Ecotourism operators Coastal workers Frequent observers Sum

40 FIG. 1. Observer effort in summer for a) large vessel crew b) residents of population centers c) lighthouse keepers d) park users e) ecotourism operators f) coastal workers and g) frequent observers categories. FIG. 2. Observer effort in winter for a) large vessel crew b) residents of population centers c) lighthouse keepers d) park users e) ecotourism operators f) coastal workers and g) frequent observers categories FIG. 3. Observer effort in BC waters in a) summer and b) winter as predicted by an observer effort model. Relative effort <.0008 was considered data deficient and removed from the model in later analyses. FIG. 4. The effect of changes to model parameters on final effort outputs. Parameters where changed one at a time while all other parameters remained fixed at mean input values. Bold lines in center of the boxes represent median effort, box edges represent 25 th and 75 th percentiles of the data distribution, and whiskers represent the range. The summer model demonstrated the most variation in model outputs when variation was introduced to the ecotourism and frequent observer categories through our sensitivity analysis. The winter model was most sensitive to changes in large vessel crew and ecotourism observer categories. Changes to coastal workers categories resulted in similar effects to outputs of both models. FIG. 5. Sightings density per unit effort (SDUE) of humpback whales in BC waters in a) summer and b) winter. FIG. 6. Sightings density per unit effort (SDUE) of killer whales in BC waters in a) summer and b) winter. 40

41 FIG. 7. Above average sightings density per unit effort (SDUE) for humpback whales in a) summer (SDUE > 7.10) and b) winter (SDUE > 2.73), overlaid with critical habitat (DFO 2010) and ocean depth. FIG. 8. Above average sightings density per unit effort (SDUE) for killer whales in a) summer (SDUE > 2.65) and b) winter, (SDUE > 1.37) overlaid with Canadian southern and northern resident killer whale critical habitat (DFO 2011) and potential critical habitat (Ford 2006) and approximate southern resident critical habitat in US waters (DFO 2011). 41

42 879 FIG

43 881 FIG

44 883 FIG

45 886 FIG

46 888 FIG

47 890 FIG

48 892 FIG