General Data Requirements for the PCoD Approach
Harwood & King (2012) describes ten critical sets of information that are required to implement a full PCOD approach for offshore renewable energy developments (Box 1). Here we describe how these sets of information can be obtained, and the assumptions that must be made when there is insufficient, or no, empirical data to estimate the relevant parameters.
There are a number of computer models that can be used to predict the way in which noise produced during the construction of a particular development propagates through the marine environment (item 1 in Box 1). Under ideal conditions, the predictions of these models are likely to be accurate. However those predictions are sensitive to variations in the temperature and salinity across the water column, bottom topography and bottom sediment characteristics and it is important that these variations are accounted for. Southall et al. (2007) provide suggestions for the sound exposure levels ( SEL) that are likely to result in PTS for different marine mammal groups (items 2. and 3. in Box 1). They also provide weighting functions that can be used to account for way in which the hearing sensitivity of marine mammals varies with sound frequency. Nedwell et al. (2007) have suggested a different set of weighting functions that are based explicitly on audiograms. The experimental data on which Southall et al. (2007) based their conclusions can be used to construct dose-response relationships that relate the probability of developing PTS to the SEL experienced by an individual, as was done by Thompson et al. (2013) for harbour seals and the SAFESIMM computer package (Donovan et al., 2012), which includes dose-response relationships for all of the marine mammal groups identified by Southall et al..
Box 1. Information Required to Implement a PCoD Approach
1. The sound field produced during construction and operation of a particular development (with associated uncertainty).
2. The sound levels that are likely to cause Permanent Threshold Shift ( PTS), preferably in the form of a dose-response relationship, with associated uncertainty, for each priority species.
3. The sound levels that are likely to result in a behavioural response that may impair an individual's ability to survive, breed, reproduce, or raise young, or that may result in that individual being displaced from an area for a longer period than normal. We refer to this as a 'significant' behavioural response. This information should, preferably, be in the form of a dose-response relationship, with associated uncertainty, for each priority species.
4. Estimates of the number of animals of each species that may be exposed to sound levels that could result in PTS, and of the number that may show a 'significant' behavioural response during one day of construction of an offshore wind farm. Estimates of the number of animals of each species that may collide with or become entrapped in a marine renewables device, or be exposed to sound levels that could result in a 'significant' behavioural response, during one day of construction or operation of this device.
5. The number of animals that are likely to be exposed to sound levels likely to result in PTS or a 'significant' behavioural response over the entire course of construction of an offshore wind farm. The number of animals that might be killed or injured during one year of operation for an operational wave or tidal project (given that the probability of collision or entanglement per day is likely to be low), or a range of possible values, if it is impossible to provide an actual estimate.
6. The potential effect of experiencing PTS at a specified frequency on the vital rates for an individual of each species, by age/stage class ( e.g. adult males, adult females, calves, juveniles), with associated uncertainty;
7. A mathematical function linking the number of days on which an individual experiences a 'significant' behavioural response and its vital rates, with associated uncertainty, for the different age classes of each priority species.
8. The current population size and population history for each Management Unit ( MU) of the five priority species, with associated uncertainty.
9. Estimates of the key demographic parameters (adult survival, calf survival, juvenile survival, annual probability of pupping/calving, age at first pupping/calving, longevity) for each species, in each MU (if parameters are likely to vary between MUs) with an indication of likely levels of variation between years.
Thresholds and weighting functions for the onset of a 'significant' behavioural response (item 4. in Box 1) are more difficult to define, but Southall et al. (2007) include some suggestions for these, while acknowledging that thresholds are likely to vary with the context in which disturbance occurs. Thompson et al. (2013) were able to develop a relationship between the probability of a 'significant' behavioural response (in this case, a prolonged change in vocal behaviour) by harbour porpoises and exposure to noise associated with piling operations, which they used to predict the response of harbour seals to the same kind of noise.
The number of animals that may experience PTS, or exhibit a 'significant' behavioural response (item 5. in Box 1) can be calculated by combining estimates from items 1.-4. with information on the anticipated density of each species in the immediate vicinity of the noise source. These density estimates may come from developers' own surveys or meta-analyses of data from many surveys, such as those in Paxton et al. (2012). It may also be appropriate to take account of the way in which animals respond to noise exposure. Individuals that flee directly away from the source of the sound are likely to be exposed to lower SELs than those that follow a more circuitous path; this will reduce the probability that they experience PTS. The interim PCOD approach does not make any adjustments for the different potential noise assessment approaches from which such values are derived.
As noted above, published methods are available for estimating all of the quantities specified in items 1.-4. Therefore, we have not attempted to estimate them as part of the PCoD approach. As a result, the approach can only be implemented if the ESs and HRAs provided for UK offshore renewable energy developments include the information described in item 4. of Box 1. That is, the ES and HRA for a wind farm should include estimates of the number of animals that may be exposed to sound levels that could result in PTS or a 'significant' behavioural response during one day of construction. The equivalent documents for a tidal or wave energy array should include an estimate of the number of animals that might be killed or injured during one year of operation (given that the probability of collision or entanglement per day is likely to be low), or a range of possible values, if it is impossible to provide an actual estimate.
As noted above, there are a number of different ways in which these estimates can be calculated and there is considerable uncertainty associated with some of the values, such as the estimated density of different marine mammal species in the vicinity of a development, that must be used in the estimation process. Ideally, the estimates provided in ESs and HRAs should attempt to quantify these uncertainties. However, we recognise this may not be practicable, and we have therefore assumed a pre-specified level of uncertainty in these estimates, which can be replaced with specific estimates if these are available.
The interim PCoD approach uses highly simplified models of the way in which individual animals within a MU may be exposed to the noise associated with the construction of these developments over the course of a year to estimate the information specified in item 6. of Box 1. The results of the expert elicitation have provided the information required for item 7. of Box 1.
It should be recognised that all of these estimates are based either on strong assumptions or on the opinions of the experts we consulted. They are not based on empirical data, and there is clearly an urgent need to collect the information that can be used to provide more realistic estimates of the parameters that define these relationships.
The report of the IAMMWG on Management Units for Marine Mammals in UK Waters (Anon. 2013) provides estimates of the current size of the populations of the priority species in each MU (item 8. in Box 1). Harwood and King (in prep.), which will be made available together with the software for implementing the interim approach, provide preliminary estimates of the demographic parameters specified in item 9. in Box 1 for each MU of the five priority species. These values should be refined as further evidence becomes available.
The Effects of a 'Significant' Behavioural Response and PTS on Survival and Fertility
As noted above, we asked the experts who participated in our elicitation process for their best estimate of the likely impact of experiencing PTS on survival (for pups/calves, and juvenile animals) or the probability of giving birth (for breeding age adults) for each priority species. We also asked them for the likely range for each estimate, which was used to calculate uncertainty levels. The results were then summarised in a series of probability distributions. We assumed that animals which experience PTS incur the additional mortality or reduction in fertility, suggested by these distributions for the remainder of their lives. However, if regulators and their scientific advisors consider this assumption is inappropriate, the computer code can easily be modified to incorporate Thompson et al.'s (2013) assumption that the effects of PTS on vital rates are only evident in the year in which PTS is first experienced.
We asked the same experts for their best estimates of the number of days of disturbance that an individual calf or juvenile animal of each species could tolerate before it would have any effect on its probability of survival, and that an individual mature female could tolerate before it had an effect on the probability of giving birth (labelled as B in Fig. 4), and on the likely range for these estimates. We also asked what they thought the maximum effect of disturbance on these probabilities might be (A in Fig. 4), and how many days of disturbance would be required to have this effect (vertical line C in Fig. 4). We assumed that the maximum effect of disturbance on the probability of giving birth would be to reduce it to zero. These estimates were used to define a relationship between days of disturbance and the effect on vital rates that took account of the uncertainty each expert associated with his or her judgements, and the variation among experts.
Figure 4. Hypothetical relationship between the number of days of disturbance experienced by an individual marine mammal and its survival used as the basis for the expert elicitation process. B is the number of days of disturbance an individual can tolerate before its probability of survival is affected, A is the maximum effect of disturbance on this probability, and C is the number of days of disturbance required to cause this maximum effect.
When the computer software that implements the interim approach is run, any simulated animal that experiences less than B days of disturbance is categorised as undisturbed. Those that experience between B and C days of disturbance are categorised as experiencing 'moderate' disturbance, and their survival or fertility is reduced by (1 + A)/2 ( i.e. the mid-point between 1 and A). For example, if A was 0.5, the undisturbed survival rate was multiplied by (1 + 0.5)/2 = 0.75. Individuals that experience more than C days of disturbance are categorised as experiencing 'high' levels of disturbance and their survival rate is multiplied by the value for A.
Defining Vulnerable Sub-Populations
The way in which animals use the space within an MU will almost certainly vary between individuals and, as a result, their risk of injury or disturbance from a particular development will also vary. At present, we do not know enough about this variation to model it explicitly, although existing telemetry data could be analysed to provide this information for grey seals and harbour seals. We have therefore adopted a broad modelling approach to characterise the range of population consequences of this variation. At one extreme, users may choose to specify that all members of the population within an MU are equally vulnerable to the effects of a particular development. This is most likely to be the case where the geographical extent of the MU is relatively small and/or the development is at a key location for the species concerned. At the other extreme, users can specify that only a proportion of the population within an MU is likely to spend time in the region around a particular development where sound exposure levels are sufficiently high that they will cause a behavioural response or injury. We refer to these animals as being members of a vulnerable sub-population (see Glossary). For example, wind farm construction in the northern North Sea may only affect a small proportion of the harbour porpoise population in the North Sea MU, and these animals may be unaffected by wind farm construction in the southern North Sea. Similarly, harbour porpoises that spend most of their time in the southern North Sea may be unaffected by the construction of wind farms in the northern North Sea. Figure 5 shows boundaries that could be used to define two vulnerable sub-populations of harbour porpoises within this MU in relation to planned wind farm developments. Paxton et al. (2013) provide estimates of the proportion of the North Sea harbour porpoise population in a number of "areas of interest for offshore development". The proportion of the MU population in the northern vulnerable sub-population shown in Figure 5 could be calculated from the total proportion of harbour porpoises that are estimated to occur in the Moray Firth and Firth of Forth "areas of interest" (or a simple multiple of this proportion). The proportion of the MU population in the southern vulnerable sub-population could be calculated from the total proportion (or a simple multiple) of harbour porpoises that Paxton et al. estimate to occur in the Norfolk Bank, southern Dogger Bank and Dogger Bank "areas of interest". The results of simulations assuming the entire population within the MU is vulnerable to disturbance and those which assume only a proportion of that population is vulnerable can then be compared to determine which assumption results in the worst case scenario for this harbour porpoise population.
Figure 5. Potential boundaries for two vulnerable sub-populations (see Glossary for a definition) of harbour porpoises within the North Sea MU (as defined by Anon. 2013) that could be used in assessing the effects of disturbance associated with UK wind farm developments in the southern and northern North Sea. These are overlayed on a map showing the location of offshore wind farms that are in development taken from p10 of The Crown Estate's UK Offshore Wind Report 2012 . Numbered sites indicate wind farms (or demonstrators) that are in operation or under construction, unnumbered sites are those that are under development. The blue line indicated the limit of UK territorial waters, and the grey line the limit of the UK continental shelf. The approximate boundaries of the North Sea MU for harbour porpoises is shown in red.
Modelling Disturbance and PTS Within a Year
Modelling the cumulative impact of an offshore renewable energy development over the entire period of construction requires a series of assumptions about the way in which marine mammals respond to the disturbance associated with the development. There is considerable evidence (Brandt et al., 2011; Teilmann & Carstensen, 2012) that harbour porpoises which have been disturbed by piling noise do not return to the area where piling occurred until some time after piling ceases (which may vary from hours to years). We have therefore assumed that a marine mammal which is disturbed by a single piling event will vacate the area around that event for the remainder of the day on which piling occurs. This allows us to use one day as the smallest time interval that is modelled. However, the available evidence for harbour porpoises suggests that animals may not return to an area where disturbance occurred for several days after that event, and this 'residual' exclusion from an area may also have a negative effect on their vital rates. We have therefore allowed users to specify the number of 'residual' days of disturbance that may be associated with each day of actual disturbance. We assume that individual marine mammals exhibiting 'residual' disturbance are not vulnerable to PTS or any direct disturbance associated with construction during the time they are experiencing this effect because they will not be exposed to additional noise during this time.
The basic model assumes that animals are at risk of PTS every time they enter the region around a development where they may be disturbed by construction noise. However, the area where they are at risk of PTS is likely to be relatively small. It is therefore possible that they will avoid the immediate vicinity of operations after they have been disturbed once. We have therefore included a capability to model a scenario in which animals are only at risk of experiencing PTS on the first day they experience a sound exposure level sufficient to cause disturbance, although we recognise that this probably underestimates the risk of experiencing PTS.
The probability of experiencing PTS can be modelled in many ways; we have implemented only a few of these and we are not proposing that any one of them is definitive. As we will see later in the report, the choice of model for the risk of PTS has a substantial effect on predictions of the population consequences of construction activities. More research is clearly required to clarify the way in which marine mammals respond to repeated exposure to noise levels that might cause PTS.
Modelling the Effects of Renewable Energy Devices other than Wind Turbines
The interim PCoD approach has been designed to assess the potential effects of a range of offshore renewable energy developments on marine mammal populations. However, its focus is on assessing the potential effects of the construction of large offshore wind farms, and the cumulative effects of multiple renewable energy developments in the same geographical area. Although it can be used to assess the potential population effects of any mortality that may be associated with the operation of an individual tidal or wave energy development, it is better suited to assessing the cumulative effects of such devices if they are operating at the same time as wind farm construction. We have not, therefore, developed a separate protocol for assessing the potential population effects of a particular tidal or wave energy development in isolation.
To the best of our knowledge, the only empirical information on the potential effects of renewable energy devices other than wind turbines on marine mammals comes from studies of the movements of harbour seals and harbour porpoises around the tidal turbine installed in the Narrows of Strangford Lough in Northern Ireland (although, to date, there have been no collisions observed or recorded at this site). We assume that developers of other similar devices will attempt to provide a range of values for the potential number of individuals from a particular MU that might collide with them during the course of a year's operation as part of their ES. The interim PCoD approach can accept these values and can be used to assess their effects in combination with those of other offshore renewable energy developments, as long as some value for the probability of death following a collision is available. In the absence of empirical information about the probability of death following a collision, a precautionary approach would be for the model to assume that all collisions are fatal. In the current implementation we assume that adult and juvenile animals are equally likely to be involved in collisions, and that a collision always results in death. However, this assumption can easily be modified if evidence emerges that certain age classes are more vulnerable than others. The effects of any disturbance that may be associated with the operation of these devices can be handled in the same way as for other renewable energy devices.
The interim PCoD protocol attempts to model many of the major sources of uncertainty involved in the calculation of the potential effects of an offshore renewable energy development on a population of marine mammals. These are shown in Box 2.
Box 2 - Sources of Uncertainty in the Interim PCoD Approach
1. Uncertainty about the size of the population in a particular MU;
2. Uncertainty about what proportion of that population will be affected a particular development;
3. Uncertainty in the predictions of the number of animals that will experience disturbance and PTS as a result of one day of construction or operation;
4. Uncertainty about predictions of the total number of days of disturbance an individual animal will experience during the course of construction of a development and of the total number of animals that will experience PTS;
5. Uncertainty about the effects of disturbance and PTS on vital rates;
6. The effects of demographic stochasticity and environmental variation.
The way in which these uncertainties are modelled is described in detail in Appendix 2. However, we note that accounting for demographic stochasticity, the fact that, even if survival and fertility rates are constant, the number of animals in a population that die and give birth will vary from year to year because of chance events, can produce predictions that appear counter-intuitive. For example, two otherwise identical populations that experience exactly the same sequence of environmental conditions will follow slightly different trajectories over time. As a result, it is possible for the size of a 'lucky' population that experiences the effects of disturbance and PTS associated with an offshore renewable energy development to increase over time, whereas an identical undisturbed but 'unlucky' population may decrease.
The concept of density dependence is central to an understanding of the way in which animal and plant populations respond to a reduction in their size. The standard assumption is that resources (such as food, or suitable places to breed or escape from predators) will become more abundant as population density declines, and this should result in an increase in fertility and survival among at least some members of the remaining population. This assumption underpins the Potential Biological Removals ( PBR) formula that is currently used by the Sea Mammal Research Unit ( SMRU) to estimate the number of 'unnatural' deaths that different Scottish seal populations can sustain ( e.g. SMRU, 2012). These values are then used by the Scottish Government to decide the number of licences that may be granted. However, it should be recognised that Wade (1998) developed the PBR formula specifically for use in the context of the US Marine Mammal Protection Act. Wade (1998) estimated that, if annual, human-induced mortality is restricted to a level below the PBR limit, there was a very high probability (>95%) that populations which were at their maximum net productivity level ( MNPL), which is approximately 60% of carrying capacity - see Glossary, would "stay there or above after 20 years". Populations that were at 30% of carrying capacity would have an equally high probability of recovering to their MNPL after 100 years. It is not immediately obvious how these performance measures relate to the requirement that the Habitats Directive places on member states to maintain favourable conservation status for certain species. Lonergan (2011) provides an interesting critique of the use of PBR for management purposes. PBR does provide a valuable benchmark for assessing the potential population consequences of any mortality that may be associated with the operation of an offshore energy development. However, it is of limited value for assessing the population consequences of the short-term effects associated with the construction of such a development.
With the exception of grey seals, there is no published evidence for density dependence in UK populations of marine mammals. It is therefore difficult to know how to incorporate density dependence into the interim PCoD protocol in a scientifically-justifiable way. For example, density dependence in grey seals seems to be the result of a relationship between pup survival and the density of adult females on individual breeding colonies (Harwood & Prime, 1978). However, Russell et al. (2013) used telemetry data to show that female grey seals tagged in the East Coast and Northeast England MUs are more likely to breed at colonies in other MUs than they are to breed at colonies in their own MU. As a result, changes in the size of the populations in those MUs will have a negligible effect on the density of adult females on individual breeding colonies, and therefore on the survival of pups born to females from those MUs.
Ultimately, agent-based population models (also known as individual-based models, Grimm et al. 2007) that track changes in the energy expenditure and energy intake of individual animals over time, can provide an insight into the way in which survival and fertility are likely to change with population size. Nabe-Nielsen et al. (2011) have developed such a model for harbour porpoises in Danish waters. However, such models are not yet available for any UK marine mammal populations.
For these reasons, we have not included density dependence for any species in the current implementation of the interim PCoD protocol. The implications of this are discussed later.
The interim PCoD protocol can provide a large, and probably unmanageable, amount of information about the changes in population size and structure that are forecast to occur in response to a particular development scenario. We have chosen to focus on the information which we believe is most relevant to national implementation of the Habitats Directive. For each iteration ( i.e. each occasion on which the statistical distributions that are used to capture uncertainty are resampled) of each scenario, we have also simulated the dynamics of an identical undisturbed population which experiences exactly the same history of environmental variation as the disturbed population. We then compare the sizes of the two populations at regular intervals to determine the effects of disturbance, and then summarise the results across all iterations. We also provide a summary figure that indicates the probability that the simulated disturbed population will have declined by at least 1%, 2% or 5% from its initial size immediately after construction work ends, and the decline experienced by 50% of the simulated populations (roughly equivalent to the mean decline) over this period.