Occupancy estimation of marine species: dealing with imperfect detectability

Abstract

Underwater visual surveys are frequently used in monitoring programmes of marine populations. Species occupancy, defined as the probability of presence in a sampling unit, is a commonly used state variable. Imperfect detectability is a serious issue in such studies and, if ignored, may lead to incorrect inferences and erroneous management decisions. In this paper, we propose a methodology and field protocol for underwater visual surveys implemented by multiple observers. This approach can be applied for an unbiased occupancy estimation of marine species by explicitly incorporating imperfect detection into the modelling process. Based on a case study carried out in a Greek coastal area (Saronikos Gulf), the benefits of the proposed approach were demonstrated. Using a sufficient number of observers, the probability of recording false absences (i.e. the probability that the target species was present in a site but not detected) was minimized and occupancy estimation was greatly improved. For the whelk Stramonita haemastoma in the case study area, single-observer occupancy estimates were negatively biased and varied significantly (between 0.64 and 0.89) depending on the observer, while with the proposed methodology, using 5 observers, the obtained occupancy estimate had the value of 0.93. The probability of false absence was high in the single-observer case (between 0.10 and 0.30), and rather low with any combination of 3 observers (< 0.025), while it dropped to practically 0 with 5 observers. As demonstrated in the case of the alien green alga Codium fragile fragile, occupancy models provide a flexible framework for relating occupancy to spatial and environmental covariates, testing ecological hypotheses and producing predictive distributional maps. Overall, the presented methodology and its potential extensions could prove extremely useful in a variety of applications in the marine environment. KEY WORDS: Occupancy · Underwater visual survey · Detectability · Monitoring Resale or republication not permitted without written consent of the publisherMar Ecol Prog Ser 453: [95][96][97][98][99][100][101][102][103][104][105][106] 2012 state variable, appropriately replicated in space and time. The most commonly used state variables are population density, abundance and percent cover. However, such estimations are often costly and require substantial effort (e.g. Lancia et al. 1994, Pollock et al. 2002, MacKenzie et al. 2006 or may be unfeasible for various reasons, e.g. in the case of rare or elusive species (Thompson 2004). Alternatively, species occupancy, defined as the proportion of area, patches or sampling units occupied (or as the probability of presence in a sampling unit) may be seen as a low-cost surrogate (MacKenzie et al. 2006). Moreover, in some cases occupancy is the appropriate state variable and would be chosen in the first place, e.g. in studies of distribution and range (Scott et al. 2002), alien invasions (Hanspach et al. 2008), metapopulation studies (Moilanen 2002), community studies (M...

Estimating relative abundance and species richness from video surveys of reef fishes

“…First, occupancy or N-mixture modeling approaches can estimate detection or capture probabilities separately from the underlying distribution or abundance of a species (MacKenzie et al, 2002;Royle, 2004), but multiple site visits may be necessary each year (Issaris et al, 2012) unless spatial autocorrelation is modeled (Johnson et al, 2013). Second, if the emphasis is on es-timation of species richness over an entire study area, species accumulation (i.e., rarefaction) curves may be a useful approach (e.g., Nichols et al, 1998;Thompson et al, 2003).…”

Chasing fish and catching data: recreational spearfishing videos as a tool for assessing the structure of fish assemblages on shallow rocky reefs

“…Sampling inconsistencies can be managed using this procedure based on maximum likelihood (MacKenzie et al 2002). Here, we report briefly on modeling procedures, as these have been described in full detail by previous papers (MacKenzie et al 2002, Katsanevakis et al 2011, Issaris et al 2012, Salomidi et al 2013. Species occupancy Ψ was modelled jointly with detection probability p; a site can be scored as occupied, with probability Ψ, or as unoccupied, with probability 1 − Ψ, by the target species.…”

“…In the general formulation by MacKenzie et al (2002), the presence/ absence of a species at a given number of s sites is recorded K times by the same or independent observers. In the approach proposed by Katsanevakis et al (2011), and more recently applied by Issaris et al (2012) and Salomidi et al (2013), the replication of sampling through time is replaced by repeated independent observations by different observers. Detections by each single observers are considered as surveys sensu MacKenzie et al (2006), generating, for each study site, a detection history consisting of a string of 1 (detection) and 0 (nondetection).…”

“…site covariates), detection (i.e. survey covariates) or both (MacKenzie et al 2002, Katsanevakis et al 2011, Issaris et al 2012. Covariates can be incorporated using the logistic model…”

“…In addition, in order to show the potential use of spearfishing videos for testing broad hypotheses, we investigated how estimates of species occupancy (i.e. probability of presence in a sampling unit) can be improved by taking into account imperfect detection (the probability of correctly recording the presence of a species when present in a sampling unit; MacKenzie et al 2002, Katsanevakis et al 2011, Issaris et al 2012. In order to provide a worked example, we modelled the occupancy of a species (the wrasse Symphodus rostratus) at increasing distances from a commercial port, explicitly incorporating imperfect detectability as a function of covariates extracted from videos (video length and habitat characteristics).…”