2.1 Data collection 48!
2.1.5 Sources of covariate data 52!
Data to be used as model covariates were obtained from a range of external sources (Table 2.1). A number of studies have identified links between cetacean distribution and environmental and oceanographic covariates (like serving as proxies for prey distribution or other biologically significant phenomena)(Acevedo 1998; Baumgartner 1997; Cañadas et al. 2003; Davis et al. 1998; Fiedler and Reilly 1994; Gaskin 1984; Gowans and Whitehead 1995; Gregr and Trites 2001; Karczmarski et al. 2000; Loughlin et al. 1999; Marubini et al. 2009; Moore 2000; Mullin et al. 1994; Naud et al. 2003; Panigada et al. 2008; Rowe et al. 2003; Smith et al. 1986; Tynan et al. 2005; Wilson et al. 1997; Wimmer and Whitehead 2005). Previous studies have
investigated whether cetaceans make diurnal shifts in behaviour, distribution or habitat
preferences (Dietz and Heide-Jørgensen 1995). To assess whether harbour porpoise exhibit such shifts ‘Time of Day’ was included in the models - as continuous indexes between 0 and 1 to incorporate spatial and temporal variation in the data. ‘Time of Day’ was included in models as a ratio; calculated by dividing the time elapsed since sunrise by the time between sunrise and sunset for the survey day. The time of sunrise and sunset was determined from POLTIPS (Version 3.0, Proudman Oceanographic Laboratory) for Tobermory, the start and end point of the majority of surveys.
prevalent or more detectable during certain phases of tide (e.g. slack, flood, ebb). For harbour porpoises in particular, a range of studies of their distribution have identified site-specific patterns associated with tidal activity (Calderan 2003; Embling et al. 2010; Johnston et al. 2005; Pierpoint 2008; Skov and Thomsen 2008). Here, ‘Position Relative to Tidal Range’ and
‘Position in Daily Tidal Cycle’ were used to determine whether the same patterns were occurring west of Scotland. For tidal variables it was necessary to determine the nearest tidal port from which to source tidal data in POLTIPS. Distances to 15 sea-ports for which tidal predictions were available were calculated using a custom routine in Manifold (Version 8.00. 32- bit, Manifold® Systems). Maximum spring tidal range for the closest tidal port and tidal range for the tide cycle that the data point fell within were determined. Position Relative to Tidal Range provides an indication of variations in the lunar tidal cycle (i.e. the spring-neap tidal cycle). This was calculated by first determining the tidal range for the time and location for each data point and subtracting from it the minimum tidal height for the same location. This value was then divided by the maximum spring tidal range for the nearest tidal port for the current lunar cycle, to generate providing a value between 0 and 1. Values close to 0 represent times close to neap tides and values close to 1 were indicative of periods close to full spring tides. Position in the daily tidal cycle was calculated by dividing the time from the nearest low water to the data point by the time elapsed between successive low waters for that day. This ratio
provided values between 0 and 1, where values 0.0 – 0.1 and 0.9 – 1.0 represented the low water slack tide; 0.1 – 0.3 represented the flood tide; 0.3 – 0.6 was the slack high water tide and 0.6 – 0.9 represented the ebb tide (from Embling 2007). Current speed data were obtained from the POLCOMS CS20 model – resolution: 1.8 km); which predicted the maximum current speed at the time and location of each data point. While this model provides excellent coverage of the west coast of Scotland, it does not cover the northern Sound of Jura, northeast region of the Firth of Lorn and the Sound of Mull. Tidal current models developed for these regions by Andrew Dale (at the Scottish Association of Marine Science - SAMS) were used. The Sound of Mull model had a resolution of 200 metres and the Firth of Lorn/Sound of Jura model at 100 m.
A number of studies have used ‘Distance to nearest land’ as a covariate. This may function as a proxy for other oceanographic factors, e.g. salinity, (Mann and Lazier 2006) or as a reflection of a species remaining close to land for shelter, or navigational cues – e.g. the harbour porpoise echolocation clicks do not travel > 300m so it may be difficult to resolve information from deeper, offshore regions (Able 1995; Alerstam 2006; Mouritsen 2001). Distance to nearest land for each data point was calculated using a script written by Clint Blight (SMRU) in Manifold. The minimum recorded distance to land was 10 m. Sediment data were obtained primarily from
United Kingdom Hydrographic Office (UKHO) and for regions not covered by these data, the Marine European Seabed Habitats (MESH) EUNIS model was used. These datasets were only available as categorical data (RSDB codes describing the different sediment types from the UKHO), so they were converted into percentages of gravel, sand and mud in the sediment using the Folk Classification (Folk 1980). Both depth and slope have been found to be important in explaining cetacean distribution in many regions (Acevedo 1998; Azzellino et al. 2008; Bailey and Thompson 2009; Baumgaertner and Mate 2005; Brager et al. 2003; De Segura et al. 2008; Forcada et al. 1996; Panigada et al. 2008; Wilson et al. 1997). Depth may be made important by prey species occurring certain depth ranges, focusing the distribution of predators. It has been suggested that slope may help drive productivity at fine scales by functioning as anchor points for eddies and currents (Mann and Lazier 2006). Bathymetry data (average seabed depth and average seabed slope), were sourced from EDINA as they provided the best coverage of the study region and the highest resolution available (EDINA averages depth and slope data over a 200 x 200 m grid). Average slope is the change in depth over the resolution of the grid. Slopes of over 20° exist on the west coast of Scotland, though these are extremes. The majority of slope measurements found there are between 0 - 6° (by comparison the slope of the continental shelf is typically around 3-6° and rarely exceeds 10° (Pinet 2009)).
Table 2.1 – Details of covariates used in models showing details of sources, units and
temporal/spatial resolution of data used. Acronyms: UKHO – United Kingdom Hydrographic Office; MESH – Mapping European Seabed Habitats; NEODAAS - NERC Earth Observation Data Acquisition and Analysis Service; POL – Proudman Oceanographic Laboratory; SAMS – Scottish Association for Marine Science.
Covariate Information Resolution Unit Source
Date/Time Recorded in situ from vessel GPS every 10 seconds
(! 30 m) -- In situ
Boat Speed Recorded in situ from vessel GPS every 10 seconds
(! 30 m) Knots In situ
Sea State Recorded by Observers every 30 minutes
(! 5.2 km)
Beaufort Sea
State In situ
Engine Status Recorded by Observers -- On / Off In situ
Time of Day Ratio: Time from Sunrise/Time
between sunrise and sunset for day
at every GPS
location -- POLTIPS
Position Relative to Tidal Range
Ratio: (Tidal Range at location on day – The minimum tidal height at location on day)/ Maximum Spring
Tidal Range for location
at every GPS
location -- POLTIPS
Position in Daily Tidal Cycle
Ratio: Time since Low water for nearest tidal port / Time between
successive low waters for nearest tidal port
at every GPS
location -- POLTIPS
Max. Spring Tidal Range
Maximum Spring Tidal Range for nearest tidal port
at every GPS
location m POLTIPS
Distance
from Land Calculated in Manifold
at every GPS
location m Manifold
Percentage
Gravel Calculated from RSDB codes Variable %
UKHO / MESH EUNIS Percentage
Sand Calculated from RSDB codes Variable %
UKHO / MESH EUNIS Percentage
Mud Calculated from RSDB codes Variable %
UKHO / MESH EUNIS
Depth Depth of seabed 0.2 km m EDINA
Slope Slope of seabed 0.2 km ° EDINA
Current
Speed Maximum current speed
POL: 1.8 km / SAMS: 0.1 or 0.2
km
m / s POLCOMS /
SAMS
Chlorophyll Average chlorophyll concentration
for midpoint of each 2 km segment 2 km mg m-3 NEODAAS
Temperature Average temperature for the
midpoint of each 2 km segment 2 km °C
NEODAAS / in situ Noise: 100 –
150 kHz Calculated from in situ recordings
every 2 minutes
(!600 m) dB re 1 !Pa In situ recordings Noise: 50 - 75
kHz Calculated from in situ recordings
every 2 minutes
Chlorophyll-a concentration and sea surface temperature (SST) were sourced from Natural Environmental Research Council Earth Observation Data Acquisition and Analysis Service (NEODAAS) satellite data. Due to the presence of cloud cover over the study area, weekly (or if unavailable, fortnightly) composites of satellite-derived values were used. Even with the use of composites, it was only possible to obtain useable chlorophyll-a and SST data for approximately 60% of the total dataset. In 2008, sea temperature data were also collected in situ using an autonomous temperature probe (Vemco 8-bit Mini Data Logger) deployed off the stern of the survey vessel at an approximate depth of 1.5 metres.
To investigate the impact of noise on the detection of harbour porpoises, ambient noise measurements were made (sampling at 500 kHz) using the towed hydrophone array during the 2007 & 2008 seasons. To enable this, recordings were made every 2 minutes for between 2 and 5 seconds whilst the hydrophone was deployed. Sound levels were calculated (in dB re 1 µPa) averaged over two frequency bands: the ‘porpoise band’ between 100 – 150 kHz; the frequency range over which harbour porpoises vocalise and a ‘control band’; between 50 – 75 kHz. The ‘control band’ frequency range was chosen as it represents the measure used in Rainbow Click as a control, against which noise in the porpoise band is compared in determining porpoise
detections. ‘DC Noise’ levels, a measure of background noise measured in Porpoise Detector were also included in some of the yearly models (2004 & 2005) to see if it impacted acoustic detection rates.