Marine mammals: methodology for combining data

Introduction

Within the marine renewables industry, regulators need to make decisions regarding the consenting of proposed offshore wind developments. As part of Environmental Impact Assessments (EIA), Habitat Regulation Appraisals (HRA) and Strategic Environmental Assessments (SEA) there is a requirement to assess the potential impacts to marine mammals from the development of marine renewable sites. Marine mammal abundance and distribution data form an important part of assessments to determine the potential effect of such activities. Therefore, there is a need to ensure that abundance and distribution data are collected and analysed to ensure robust estimates to inform the planning, consenting and licensing processes.

There are several sources of information that can contribute to estimating the abundance and distribution of marine mammals. The current standard within the UK is to use density and abundance estimates for cetaceans from aerial and ship-based visual surveys conducted under the SCANS programme (Gilles et al., 2023; Hammond et al. 2002; 2013; 2021) to define reference populations, against which the likely effects at development sites can be gauged. Such surveys are conducted infrequently over large spatial scales, and so abundance and distribution data at a finer scale are not available for the development sites (Hague et al. 2020). As part of EIA processes, there is a need to both characterise the development site and provide baseline density and abundance estimates for eventual pre- to post-impact monitoring. Digital aerial surveys (DAS) have become the offshore industry standard primarily for offshore ornithology studies. However, the method is not taxon-specific and data on marine megafauna, including cetaceans, are also collected. Static passive acoustic monitoring (PAM) is an alternative method of data collection that detects the presence of vocalising animals. While there are other types of acoustic surveying platforms, such as ship-based towed arrays or glider-mounted hydrophones, these mobile acoustic monitoring platforms are not the focus of this study. Both digital aerial and static PAM surveys can collect data on fine temporal and spatial scales, though they have their strengths and limitations. Aerial surveys can typically provide better spatial survey coverage than static acoustic recorders, while acoustic recorders generally provide improved temporal coverage owing to their extensive deployment durations and ability to collect data during hours of darkness and poor weather. Static PAM surveys can be, however, spatially constrained when compared to dynamic surveys such as DAS.

Absolute animal abundance and density can be estimated from both aerial and static passive acoustic data (e.g., Buckland et al., 2015; Marques et al., 2013). Abundance and density estimation methods for both survey modes share many of the same survey design and analysis attributes but differ in some key aspects, as reviewed below.

All absolute animal abundance methods such as distance sampling (e.g., Buckland et al., 2015) and spatial capture-recapture (e.g., Borchers, 2012) require various inputs (both known constants and parameters that need to be estimated). These inputs form an estimator, an equation designed to convert detections of the target species into an estimate of absolute abundance. A general estimator may take the form (Eqn 1.)

where N is the estimated abundance, n is the number of detections, P is the estimated probability of detecting the target species and m is a general term for other multiplying terms that are needed to estimate abundance e.g., group size if n is the number of detected groups or cue production rate if n is the number of detected acoustic cues (see below for more detail).

Further, if the monitored survey area is quantified, then animal density, D, can be estimated (Eqn. 2):

where A is the size of the surveyed area. If the monitored area was selected at random within a wider survey area, abundance estimates can be obtained for the wider survey area by using the random properties of the design. Such estimates are known as design-based estimates. If not, model-based approaches, where density is modelled over space as a function of spatially explicit covariates, might be useful (e.g., Miller et al. 2013).

The probability of detection, P, is a key parameter for abundance/density estimation methods, correcting for objects of interest (i.e., individual animals, animal groups, or cues the animals produce such as sounds) that were available to be detected but were missed during a survey. In other words, the detection probability corrects for perception bias, which would, if ignored, cause density/abundance to be underestimated. Estimating detection probability is an essential step for absolute abundance estimation. Without P, or other required parameters (such as probability of detection on the transect line or call production rate) an estimator might be interpreted as a relative index of abundance. Relative indices of abundance rely on the major assumption that the missing parameter(s) remain constant so that temporal or spatial changes in the relative measure are due to real changes in the abundance or density, and not changes in the parameter(s) missing from the estimator. Therefore, absolute measures of density or abundance should be estimated when possible (e.g., Anderson 2001).

In addition to detection probability, both aerial and acoustic density estimators require additional parameters that are challenging to estimate. Aerial survey data analyses require an estimated availability parameter, which in the marine context accounts for diving animals that are missed on the survey trackline because they are unavailable to be detected (e.g., Borchers et al., 2013). By estimating the probability of detecting an animal on the trackline, often described using the notation g(0) in the distance sampling literature, potential availability bias is corrected for, which would otherwise cause density/abundance to be underestimated. Passive acoustic data analyses require a parameter to account for animals’ vocal behaviour e.g., the number of calls produced per minute, or the proportion of time that an animal is acoustically active, discussed in Marques et al. (2013).

Often, data do not exist to directly estimate detection probability, the availability parameter or the call production parameter. In these cases, to avoid a relative abundance index, data from two survey modes may be combined. This was the motivation for this project, which had the following goals:

1. Produce a modelling framework integrating DAS data and PAM data, including the ability to incorporate seasonal and diurnal uncertainty.

2. Produce a test case study on harbour porpoise to validate the methods, producing density maps for a specified site in Scotland.

3. Provide recommendations on standards for static PAM and DAS data collection.

These goals were achieved by completing the following tasks over the project’s 1-year timeline:

Task 1: A technical meeting was held to discuss available methods for data integration (Section 1).

Task 2: A second technical meeting was held to assess how existing software tools developed within CREEM could be adjusted for combined data types (Section 2).

Task 3: Based on discussions in Task 2, a comprehensive roadmap of how the available software tools could be extended was produced (Section 2).

Task 4: A dataset was selected and prepared for the case study (Section 3).

Task 5: The case study analysis was completed to produce density maps of harbour porpoise for the selected Scottish study site (Section 3).

Task 6: The R-based dssd/dsims survey design tools were used for survey planning recommendations. Tracklines for digital aerial surveys and the placement of acoustic instruments for an integrated survey were designed, using the same location as the case study (Section 4).

Task 7: Finally, this report summarises all objectives and deliverables across the project.