The terms abundance and population size are sometimes used interchangeably. The common usage of the English word “abundance” indicates a large quantity of something but, in the context of estimating species abundance, it simply means the number of animals, whether large or small. Abundance may refer to the number of animals in any defined area or it may refer to a “population,” which begs the question – what is a population?

As a precursor to his discussion of comparative demography, Cole (1957) defined a population as: “a biological unit at the level of ecological integration where it is meaningful to speak of a birth rate, a death rate, a sex ratio, and an age structure in describing the properties of the unit.” This resonates with the commonly considered definition that a population comprises individuals of a species that live and interbreed in the same space.

However, such a definition is challenging to implement for many species, perhaps particularly so for marine mammals. Krebs (1972, p139) proposed a more pragmatic definition: “a group of organisms of the same species occupying a particular space at a particular time … [its] boundaries … both in space and time are vague and in practice are usually fixed by the investigator arbitrarily.” Indeed, studies of abundance using at-sea transect surveys (see below) need to be defined in space and time. Estimates of abundance from such surveys therefore relate to the “population” in an area, not to a wider biological population. In studies to estimate abundance using mark-recapture methods (see below), the population estimated comprises those individuals that use the study area, which may be different from the number of animals present in the area during a transect survey. Calambokidis and Barlow (2004) discuss this difference with respect to humpback and blue (Balaenoptera musculus) whales.

A population may also be defined for the purposes of assessment and/or management; such definitions may be general or specific. For example, the IUCN defines a population as the “set of individuals from the same wild species that share the same habitat” (IUCN, 2021). The United States Marine Mammal Protection Act defines a “population stock” as a group of animals “of the same species or smaller [taxon] in a common spatial arrangement that interbreed when mature” (MMPA Section 3, 16 United States Code 1362). The International Whaling Commission (IWC) manages whaling on the basis of defined management stocks, for which abundance estimates are required to be incorporated within a management procedure [e.g., bowhead whales (Balaena mysticetus) in the Bering-Chukchi-Beaufort Seas⁵ ]. The International Council for the Exploration of the Sea has defined “Assessment Units” for some species of small cetacean in the European Atlantic [e.g., harbor porpoise (Phocoena phocoena); ICES, 2014].

Whatever the context, for any study reporting abundance, it is critical to define clearly the “population” to which abundance refers.

The simplest measure of the number of animals in a population is a count. On rare occasions (e.g., “southern resident” killer whales, which are all uniquely identifiable and do not disperse from their natal units), all individuals in the population can be counted (Olesiuk et al., 1990). However, a count is highly unlikely to represent true population size for most marine mammals, especially cetaceans, because of problems of access (individuals spend all or most of their time at sea and mostly underwater), behavior (individuals are highly mobile), and scale (most species occur over large areas). Nevertheless, a count is a minimum measure of abundance and this may be of value in the absence of an estimate of abundance. However, use of a count, in the large majority of cases, would substantially underestimate the capacity of the population to sustain human impacts, such as bycatch mortality, and would therefore lead to much lower bycatch limits than are necessary to achieve management objectives.

To obtain a realistic measure of the number of animals in a marine mammal population, abundance must typically be estimated. Estimation involves collecting a sample of data that can be extrapolated to the entire population using one or more statistical models. The principles of extrapolating sample data to estimate abundance using models can be illustrated through a simple example.

Consider the need to obtain a measure of the number of animals in a defined study area of size A = 450 km². It is not logistically feasible to count all the animals in the area. Instead, a smaller subarea of size a = 50 km² is defined, within which it is believed possible to count all the animals, and in which the density of animals is assumed to be representative of the larger area A. The count in the subarea (n = 35 animals) forms the sample. Extrapolating to obtain an estimate of abundance is achieved by (i) dividing the count n by the size of the subarea a to provide an estimate of animal density, n/a = 35/50 = 0.7 animals per km², and (ii) multiplying estimated animal density by the size of the whole study area, A = 450 km² to provide the estimate of abundance, 315 animals.

The statistical model in this case is simply: N^=na×A. The “hat” over the N indicates that it is an estimated quantity.

The sizes of the areas, A and a, are known quantities because they can be measured exactly, but the number of animals in the subarea is unknown and the count, n, is thus a random variable with associated statistical error. Estimating this error allows us to provide a measure of precision of the estimate of abundance (see below).

All models, whether simple or complex, make assumptions about the accuracy and representativeness of the sample data. The resulting estimates of abundance can be biased if these assumptions are not met. A biased estimate is not an accurate measure of true abundance.

In the example in the previous section, the fundamental assumption made is that it is valid to extrapolate the estimate of animal density in the subarea, n/a, to the whole study area. This assumption has two elements: first, that the sample count, n, accurately represents the number of animals in the subarea a; and second, that animal density in the subarea is representative of density in the study area as a whole.

The first element, accuracy, can be thought of as the capacity of the data collection protocol to provide accurate data. If it over- or under-counts the animals, animal density will similarly be over- or underestimated and extrapolation to the whole study area will result in an over- or underestimate of abundance. For example, the count will be incomplete and abundance will be underestimated if some animals are unavailable for detection at the time of the count because they are hidden from view, or if they react to the presence of the observer and move away before they are counted. These two situations can occur in surveys for marine mammals and the solution is to collect additional data so that a correction factor can be estimated to avoid under- or over-estimation of abundance.

The second element, representativeness, is determined by the sampling design. In this example, a single subarea has been selected for estimating animal density. This could be appropriate if animals were distributed randomly in space, but this is rarely the case and cannot be assumed. Counting within the subarea would lead to overestimating animal density and, therefore, abundance, if, e.g., there were a gradient in density across the study area and the small area fell within an area of high density. The solution is to modify the sampling design, for example by splitting the single subarea into several smaller subareas and distributing them randomly or systematically over the entire study area. This should ensure that estimated animal density is representative of overall density.

If assumptions about the data and methodology are met, the resulting estimate of abundance should be an unbiased measure of the true abundance. The concept here is that if multiple sets of sample data are collected, the resulting repeat estimates of abundance should be centered about the true abundance. In practice, because true abundance is unknown, the only way to determine that an estimate is unbiased is to ensure that the assumptions about the data and the estimation method are met.

Obtaining an unbiased estimate is ideally the aim of all studies of abundance. It is therefore important (a) to use a method of data collection that maximizes the accuracy of the data, and (b) to implement a sampling design that ensures the sample data are representative.

The point was made in the previous section that estimating statistical error enables an estimate of abundance or population size to be presented with a measure of precision. The measure of precision expresses the level of uncertainty we have in the estimate. Statistically, it indicates how much the estimate would vary if it were calculated from repeated sets of sample data. Measures of precision are primarily driven by sample size; they include the Standard Error (SE), Coefficient of Variation (CV, the SE divided by the estimate, expressed as a proportion or a percentage), and a 95% Confidence Interval (CI). An estimate with poor precision indicates that its value was driven substantially by chance, whereas an estimate with higher precision indicates that its value would be similar regardless of the sample of data collected. We should thus be less certain about an imprecise estimate, and be more cautious about its use, than a precise estimate. As a general rule, an estimate of abundance should always be accompanied by a measure of precision.

It is important to be clear about the difference between bias and precision in this context. An estimate may be based on few data and thus be imprecise but nevertheless be unbiased if assumptions about the data and estimation methods are met. In contrast, a precise estimate based on a large sample may be biased because of violation of sampling or analytical method assumptions. For informing conservation and management, the worst-case scenario is arguably a precise but biased estimate because it conveys a false sense of confidence in an inaccurate measure of abundance. For example, consider a population of 10,000 animals, for which there is an estimate of abundance of 15,000 with a CV of 0.1. The estimate is inflated because the sampled subarea contained a higher density of animals than the entire study area. The CV is small because a lot of data were collected in the sampled subarea. In this case, the 95% confidence interval for the estimate (12,000–18,000) would not include the true value. A mortality limit reference point (see Moore et al., In review, Frontiers in Marine Science) (see text footnote 4) calculated on the basis of this estimate would be considerably inflated, potentially leading to bycatch levels that threaten the population. This reinforces the importance of understanding and meeting assumptions about the data and the estimation methods.

Abundance estimates provided to management authorities should be accompanied by enough information to allow assessment of their validity. As well as the estimate (for a particular location and time) and measure(s) of its precision, it is also good practice to provide:

An example of this is the specification of information that should accompany any abundance estimate intended for use in IWC management applications (e.g., the Revised Management Procedure or an Aboriginal Whaling Management Procedure; IWC, 2012, 2020).

The primary basis for estimating the abundance of pinniped populations (seals, sea lions, fur seals, walruses) is information on counts of animals on land or ice. Cetaceans (whales, dolphins and porpoises) and sirenians (dugongs and manatees) are never similarly accessible for such counts. Counts may be of pups (or walrus calves) born during a pupping season or of non-pups hauled out at that time or at other times of the year.

Pinnipeds that haul out on land typically occur in breeding colonies (where pups can be counted) or in other types of aggregation outside the pupping season (e.g., for molting), thus facilitating the counting of animals. These counts can be corrected, if necessary, and then extrapolated to provide an estimate of population size. Ice-associated pinnipeds are typically distributed over wide areas of ice requiring information on counts to be obtained from aerial surveys (see “Transect sampling from ship or aerial surveys”).

More generally, estimates using count data may be combined with those obtained using other methods such as surveys (see “Transect sampling from ship or aerial surveys” below) and mark-recapture (see “Mark-recapture” below), for example for Hawaiian monk seals (Neomonachus schauinslandi) (Baker et al., 2016).

Most of the information on the abundance of cetaceans comes from surveys of defined areas conducted at sea on ships (or smaller boats), or from the air using sampling along define transect lines. Pinnipeds can also be detected during at-sea surveys, although it is unusual for these data to be used to estimate abundance (but see Brediñana-Romano et al., 2014; Williams et al., 2017). However, an aerial survey is the only option for obtaining information on abundance for some ice-associated pinnipeds. The abundance of sirenians is typically estimated from aerial surveys.

Unlike extrapolating counts or mark-recapture (see below), transect methods sample animal density within defined areas, not individual animals. Typically, a team of observers searches on either side of a set of transect lines placed across a survey area and counts the number of animals detected. Most cetaceans occur clustered in groups, so observers typically count groups of animals and determine (or estimate – see below) the size of each group. Use of camera systems instead of observers to capture data on aerial survey is discussed below.

Surveys may be conducted in so-called “passing mode” or “closing mode.” In passing mode, the ship, small boat or aircraft does not divert from the transect line to investigate detected groups of animals. In closing mode, when a group of animals is detected, searching effort is suspended and the group is approached by the ship or small boat or circled overhead by the aircraft. The additional time spent observing the animals, especially in close proximity from ships or small boats, may allow more accurate species identification and group size estimation, including taking photographs for this purpose. Surveys conducted in passing mode maximize searching time, and are necessary for double observer team surveys (see below), but they may be subject to greater error in species identification and group size estimation than surveys conducted in closing mode.

Abundance is estimated by first estimating the density of animals per unit area and then extrapolating density to the whole study area. Abundance is therefore defined as the estimated number of animals in a specified area during the period of time that the survey took place. It is thus conceptually similar to the example described in “Counting and estimation of abundance from samples,” but the area sampled is the area around the set of transect lines placed across the survey area (see Figure 1), not a subarea.

Proper placement of the transect lines is critical for unbiased estimation of abundance. To guarantee that estimated animal density is representative of the entire study area, the set of transect lines must be placed according to a design that ensures that every point in the study area has the same probability of being sampled. This is known as an equal coverage probability design. Clearly any one set of transect lines can only sample a small proportion of the available area. However, incorporating a random starting point into an appropriate design ensures that many repeated, randomly chosen sets of lines would sample everywhere in the study area. Estimates of abundance from surveys that achieve equal coverage probability are known as design-based estimates.

Stratification of survey areas into a number of blocks is commonly done, often for logistical reasons. Survey design should then ensure equal coverage probability within each block. If density differs among blocks, for example because they cover different habitats, stratification may improve the precision of abundance estimates. Transect line placement can be challenging in coastal, inland, and riverine areas; Dawson et al. (2008) provide guidance for such areas. In some circumstances, data from surveys that do not achieve equal coverage probability can be used to estimate abundance in a model-based approach (see “Model-based estimation of abundance”).

Transect surveys for cetaceans typically collect data visually. However, for deep-diving species, abundance estimated from visual data is likely to be negatively biased because of the long dive times of these species (see availability bias below). For these species, using passive acoustic data (i.e., recordings of sounds made regularly, such as for echolocation, by the animals) to estimate abundance may be possible. For example, sperm whale abundance has been estimated from a combined acoustic and visual survey (Barlow and Taylor, 2005) and solely from acoustic data collected from hydrophones towed behind a survey ship (Lewis et al., 2007). Taylor et al. (2017) combined a line transect survey with acoustic data from static recorders to estimate the abundance of vaquitas (Phocoena sinus). Data from static acoustic recorders have also been used to estimate the abundance of beaked whales (Marques et al., 2009) using point-transect distance sampling, a variant of line transect sampling in which the area around a series of points is searched, instead of the area on either side of a series of transect lines (Buckland et al., 2001). These methods are under development; recent work also includes the use of drifting passive acoustic recorders to estimate the density of deep-diving cetacean species (e.g., Keating et al., 2018; Barlow et al., 2021a,b,c).

Some populations of baleen whales, namely gray, humpback and bowhead whales, migrate close to land or ice and can be counted as they pass a suitable observation point. However, even if the whole population migrates past such a point, not all animals will be observed and counted, so the counts need to be corrected and extrapolated to obtain unbiased estimates. Whales can be missed because they migrated past the observation point before or after the observation period, at distances beyond visual range, or during periods when counting could not take place, such as at night or during poor weather. These whales are thus unavailable to be counted, leading to availability bias described above, which needs to be corrected for with correction factors derived from additional data and/or models.

Whales may also be missed because they were available to be seen but were not detected (equivalent to perception bias described above). A correction factor for this bias can be estimated from data collected at two observation points in the form of a mark-recapture experiment, equivalent to mark-recapture distance sampling. In addition, in studies where whales are typically detected in groups, the number of whales in a group may be underestimated and need to be corrected for using data collected additional to the main study.

Studies to estimate abundance of migrating baleen whales that include correction factors such as those described above include: gray whales (Rugh et al., 2005; Laake et al., 2012), humpback whales (Noad et al., 2011) and bowhead whales (George et al., 2004; Givens et al., 2016).

The number of migrating whales can also be estimated using aerial survey, rather than land- or ice-based counts. Salgado Kent et al. (2012) employed this method to estimate the abundance of humpback whales off western Australia on both north- and southbound migration. In this study, challenges in allocating temporarily milling whales to the north- or southbound component of migration was identified as a factor that could influence estimates of abundance.

Section “Correcting and extrapolating counts of a defined portion of a population” and Table 3 describe the basis and requirements for studies to estimate the abundance of pinniped populations from counts. When considering such studies, it may be required to decide whether to count pups or animals of all ages. The practical aspects of ensuring that counts are as accurate as possible are similar, but the choice may be informed based on how available the animals are to count.

In most species, pups are available at a defined time of the year and, in many species, pups are ashore for many days or even weeks, facilitating both planning and the actual counting. If there is a period of peak pupping, pup counts can be focused at this time both to maximize the count but also to ensure consistency across years. Care may be needed, however, if the pupping period shifts over time. Counts of mature animals can also be made at pupping/breeding colonies but immature animals are unlikely to be included.

Counts of pups may not be straightforward to obtain for some species; for example, harbor seal (Phoca vitulina) pups can enter the water soon after birth. For such species, or where counting pups is logistically challenging, counting animals of all ages may be required or desirable. In these cases, it is important to determine an appropriate time to make counts, ideally when the largest proportion of the population is hauled out. For phocid seals, the annual molt may be a good time.

The difference in how pup counts and all-age counts are extrapolated to the whole population may also be an important factor in choice of method. For pups, life history information is needed to calculate a “pup multiplier” or as input to a population model. For all-age animals, information on haul-out behavior, preferably from telemetry, is needed to estimate the proportion of animals ashore when counts are made.

If individuals in a population of cetaceans possess natural markings suitable for photo-identification, e.g., humpback or blue whales, or bottlenose dolphins (Tursiops truncatus), viable alternative methods could be mark-recapture analysis of photo-id data or a line transect sampling using a ship, small boat or aerial survey.

For mark-recapture estimation, the distribution and/or size of the population and the behavior of the animals can make it challenging to achieve capture probabilities that are sufficiently high for abundance to be estimated with the desired level of accuracy and/or precision. In such cases, a line transect survey would be more effective. Multiple case-specific factors mean that it is not possible to generalize on how high capture probabilities should be, but the studies in Table 2 provide examples for guidance.

As mentioned above in relation to definition of the population being estimated, surveys using line transect methods sample density within a defined area and thus estimate abundance within that area at the time of the survey. Mark-recapture methods sample individuals and therefore estimate the size of the population using the study area during a study period. Estimates of these two quantities may not be the same even for the same study/survey area. If the mark-recapture study period is longer than the line transect survey and if animals move in and out of the study area during that period, the population sampled will comprise more animals than are present in the area at one time. This may be important in terms of the objectives for the abundance estimate. If the objective of the study is to estimate abundance in an area at a particular time, a line transect survey may be most appropriate, but if it is to estimate the number of animals using the area over a longer period, mark-recapture could be more appropriate. Abundance has been estimated and compared using these two methods for humpback and blue whales off the west coast of the United States and Mexico (Calambokidis and Barlow, 2004), and for killer whales in British Columbia (Williams and Thomas, 2009).

Two studies that consider some of these issues for small cetaceans are Minton et al. (2013) and Kuit et al. (2021). The first study used small boat line transect surveys to estimate the abundance of Irrawaddy dolphin (Orcaella brevirostris) and Indo-Pacific finless porpoise (Neophocaena phocaenoides) in coastal waters of Sarawak, Borneo, and also used mark-recapture analysis of photo-id data to estimate the number of Irrawaddy dolphins using the study area. The estimates of Irrawaddy dolphin from the two methods were broadly comparable but those from mark-recapture were around 50% larger, a result interpreted by Minton et al. (2013) as reflecting the mark-recapture analysis representing a larger population beyond the study area.

The second study used small boat line transect surveys to estimate the abundance of Irrawaddy dolphin and Indo-Pacific finless porpoise in coastal waters of Matang, Malaysia, and mark-recapture analysis of photo-id data to estimate the number of Indo-Pacific humpback dolphins (Sousa chinensis) using the study area. Kuit et al. (2021) did not use mark-recapture for Irrawaddy dolphins because the animals’ behavior led to poor-quality photographs and adversely affected individual identification. Minton et al. (2013) also noted that obtaining photo-id pictures of Irrawaddy dolphins was challenging.

There may be practical implications in terms of time to conduct the sampling and/or process the data. A line transect survey can be conducted relatively quickly and only one survey is needed to estimate abundance. Mark-recapture studies require at least two and ideally several capture occasions over a period of time that allows the population to mix between occasions. It is likely to take longer to process photo-id data than line transect survey data prior to analysis. Implementing multiple sampling occasions and processing photo-id data have cost implications that may exceed those of conducting a line transect survey (see Table 3).

There may also be practical implications in terms of observation platform. Photo-id can be done from ships but is best done from small boats. In riverine or enclosed marine habitats where navigation is impractical for larger vessels and where survey conditions are good, small boats may be the best platform for line transect surveys. However, small boats may have limitations, depending on the survey area and the method used. The observer is typically not very high above water level thus limiting the area that can be searched. It is unlikely that animals missed on the transect line as a result of perception bias will be able to be corrected for. Any movement of animals in response to the approaching boat prior to detection will also be impossible to account for; this may be important for some coastal species. Aerial surveys, on the other hand, are immune to responsive movement, and it is feasible to collect two-team data to estimate perception bias with an aircraft of sufficient size. However, aerial surveys are likely to be much more expensive and the logistics (surrounding terrain, location of suitable airports, etc.) and safety issues are more challenging (see below). Practical advice for designing and conducting line transect surveys for cetaceans in coastal and riverine habitats can be found in Dawson et al. (2008).

Some cetacean species are not typically confined to coastal waters but nevertheless may not occur a long way offshore. For example, harbor porpoises are largely found on the continental shelf in European Atlantic waters (Hammond et al., 2013). Even if the distribution of the species of interest extends beyond this, the objective may nevertheless be to estimate abundance within a limited range, for example a country’s 200 nm exclusive economic zone. The choice of whether to survey using a ship or an aircraft may therefore arise.

A crucial overarching consideration is safety. Accidents can happen at sea and in the air but the consequences are inherently far more serious in aerial surveys; fatal accidents have occurred during aerial surveys of cetaceans and seals. The risks can be reduced by careful consideration of the type of aircraft, the environmental conditions, and the logistics. Twin-engine aircraft should always be used when at all possible, especially in waters away from the coast. Experienced pilots are essential. Important logistical issues include the location of suitable airports and the availability of fuel. Aerial surveys should only be undertaken by personnel who recognize the risks and where those risks have been mitigated to the extent possible. Regarding choice of equipment to maximize the effectiveness of data collection, high-wing aircraft with bubble windows optimize visibility on and around the transect line. A clinometer is needed for measuring perpendicular distance; the recently developed “geometer” (Hansen et al., 2020) facilitates this and minimizes error.

Camera systems have been used on aerial surveys for some time to count pinnipeds on land or ice (e.g., Speckman et al., 2011; Russell et al., 2019; Thompson et al., 2019; Stenson et al., 2020) and also some cetaceans (e.g., beluga and narwhal – Gosselin et al., 2017; Doniol-Valcroze et al., 2020). They have also been used on aerial surveys for seabirds (e.g., Buckland et al., 2012) and are increasingly being used or considered for cetaceans (e.g., Williamson et al., 2016; Lennert-Cody et al., 2018). There remain challenges regarding data processing, which is time consuming without automatic image analysis, and in data analysis for unbiased abundance estimation, but there are recent developments. In particular, work on analytical methods to use data from two camera systems on one aircraft is promising (Stevenson et al., 2018; Borchers et al., 2020).

Advances in technology have also led to the development of various unoccupied aircraft systems (UAS) [also known as unmanned aerial vehicles – UAV] to collect photographic data on marine mammals with the aim of assessing their effectiveness for estimating abundance. Results show promise for pinnipeds (Goebel et al., 2015; Moreland et al., 2015), sirenians (Jones et al., 2006; Hodgson et al., 2013) and cetaceans (Aniceto et al., 2018); see also reviews in Marine Mammal Commission (2016) and Johnston (2019). UAS may also be useful to confirm species identity and estimate group size of cetaceans on ship surveys.

Unoccupied aircraft systems eliminate the human risk of flying but there are currently considerable challenges to use them to conduct surveys to estimate abundance. In a comparison of surveys of Arctic cetaceans conducted using UAS and piloted aircraft, Ferguson et al. (2018) found that estimates of density from UAS were smaller, less precise and an order of magnitude more expensive to obtain than estimates from conventional visual survey methods. UAS powered by jet fuel can survey for several hours (e.g., Angliss et al., 2018; Ferguson et al., 2018) but the endurance of more widely available battery powered UAS is usually too short (e.g., 20 min; Aniceto et al., 2018) for most surveys.

Ships typically survey at speeds that are an order of magnitude slower than aircraft, and thus cover transects much more slowly than aircraft per unit time. Even though the effective strip width on aerial surveys is typically narrower than on ship surveys, the effective search area is still much smaller per unit time on ship surveys than on aerial surveys. For example, on the SCANS-II survey of European Atlantic waters, the effective search areas per unit time for harbor porpoises, white-beaked dolphins (Lagenorhynchus albirostris) and bottlenose dolphins were 5–9 times larger for an aerial survey than a ship survey (Hammond et al., 2013). Aerial surveys are thus able to “cover a lot more ground.”

However, because of the difference in survey speed, the probability of detecting animals is much smaller on aerial surveys than on ship surveys, although aircraft survey speed does mean that responsive movement does not need to be considered. On aerial surveys, animals are available to be seen for a shorter period of time and it is easier to miss them by chance. In other words, both availability bias and perception bias are larger on aerial surveys than on ship surveys. This also means that sample sizes of sightings may be smaller on aerial surveys, which may make it more challenging to estimate detection probability.

There are also differences between ship and aerial surveys regarding ease of species identification and determining group size. On an aerial survey, groups of animals are seen for only a very short period of time (a few seconds) but on a ship there is much more time available to identify the species composition and size of groups, especially if conducted in closing mode.

One advantage of ship surveys is that they allow different data types to be collected. These may include photo-id data (which can also be collected on aerial surveys for some species, e.g., right and bowhead whales), biopsy sampling for genetic studies, environmental data for studies of habitat use and potentially for model-based abundance estimation, and acoustic data for deep-diving cetacean species such as sperm or beaked whales. However, photo-id and biopsy sampling take time away from surveying and it is important to consider whether the objectives of the study would be compromised by collection of such additional data. Ships can also accommodate more people, which may be useful in the context of training and outreach activities.

In terms of cost, chartering and using ships and aircraft is expensive and the cost varies depending on the equipment. The cost of larger ships is typically charged for a block of time on a per-day basis. Time at sea not surveying due to poor weather is part of this cost, which means that the expected proportion of time thus lost should be factored into survey planning. For aerial surveys, equipment charter costs are typically primarily dependent on flying hours, so time not surveying may not incur much financial penalty. However, in remote areas, aircraft need to be chartered for blocks of time in the same way as a ship and unfavorable weather also needs to be considered in planning.

Table 3 does not consider analysis of the data to estimate abundance. To estimate abundance from pup or all-age counts of pinnipeds, or land-based counts of migrating whales, analysis can follow methods described in studies referred to in the associated sections of this paper.

For line transect (distance) sampling, the software DISTANCE (¹² Thomas et al., 2010) is a commonly used analysis engine for Windows platforms. Several packages within the R analysis software (R Core Team, 2020) are also available for line transect analysis (Miller et al., 2019; Laake et al., 2020). Mark-recapture models to estimate population size are available in the widely used software MARK (see text footnote 8; Cooch and White, 2019) for Windows platforms. Efficient implementation of analyses in MARK can also be achieved through the R package RMark (Laake, 2013). DISTANCE and MARK can be used to implement standard methods and also more complex analyses that relax some of the assumptions of the more basic methods, as described above. The R package multimark (McClintock, 2015) fits mark-recapture models to data from more than one mark type, e.g., left- and right- side photo-id images, or photo-id and genetic marks.

However, a quick scan of the studies referred to above indicates that some use statistically complex methods that cannot be implemented in standard software and will be beyond the expertise of most biologists. If such methods are pursued, experienced analysts will be required.