Spatial and temporal assessments of overlap are becoming increasingly popular as indicators of encounter risk (i.e., relative risk of interaction) between protected species and human activities. Bedriñana-Romano et al. (2021) identified priority areas for blue whale conservation by exploring overlap between the whales’ critical habitat and vessel traffic. Best and Halpin (2019) quantified tradeoffs in site selection and timing of offshore wind energy development by exploring overlap between the industry’s profitability and sensitivities of several seabird and cetacean species. Numerous studies have explored the overlap between protected species’ distributions and commercial fishing to help inform discussions on bycatch mitigation (Cronin et al., 2016; Hatch et al., 2016; Stepanuk et al., 2018; Baird et al., 2021; Murray et al., 2021) and to prevent seasonal area closures (Howell et al., 2008; Howell et al., 2015; Eguchi et al., 2017; Hazen et al., 2018; Myers and Moore, 2020).

Fisheries bycatch is often cited as a major source of mortality for protected species in the marine environment (Read et al., 2006; Lewison et al., 2009; Wallace et al., 2010) because many protected species exhibit life-history traits (e.g., long-lived, delayed sexual maturity) that make them particularly vulnerable (Crouse et al., 1987; Heppell et al., 1999; Heppell et al., 2000; Fujiwara and Caswell, 2001; Heppell et al., 2005, Niel and Lebreton, 2005, Gray and Kennelly, 2018; Hatch et al., 2019). Minimizing fisheries-related mortality and morbidity, then, can better position threatened populations to recover (Lewison et al., 2004; Warden et al., 2015). Interactions between fisheries and protected species can also pose an economic cost to fishers, possibly leading to depredation of catch (Tixier et al., 2021), damage to fishing gear (Panagopoulou et al., 2017), and loss of fishing opportunities through time-area closures or reductions in fishing effort if bycatch limits are exceeded (Hazen et al., 2018; Myers and Moore, 2020). Here, interaction refers to any direct or close contact between protected species and fishing gear, whereas bycatch refers to a subset of interactions that result in serious injury or mortality.

Mitigating bycatch of protected species often revolves around issues of when and where animals occur and how they interact with commercial fishing gear (Senko et al., 2014). This where, when, and how approach is often associated with implementing gear modifications and time-area closures as preferred management strategies to reduce fisheries bycatch. The U.S. Atlantic sea scallop fishery is an important commercial fishery (NMFS, 2021) that primarily extends from Georges Bank to the Mid-Atlantic and overlaps with the distribution of threatened loggerhead sea turtles requiring vessels to use gear modifications aimed at minimizing sea turtle bycatch. Those gear modifications have had an apparent conservation benefit (Smolowitz et al., 2010; Smolowitz et al., 2012), with reductions in serious injury and mortality rates of sea turtles since their implementation (Murray, 2011; Murray, 2015). However, those same gear modifications complicate bycatch monitoring and increase interest in alternative monitoring approaches such as using overlap metrics as a proxy for encounter risk.

Assessing susceptibility of protected species to commercial fishing can be complicated, as most threatened populations are data-limited. Attributes that define a species susceptibility to fishing are often classified into four ordinal categories (Hobday et al., 2011): overlap (in geographic space [hereafter referred to as space] and time), encounterability (depth usage, given overlap), selectivity (caught and retained by the fishing gear, given overlap and encounterability), and post-interaction mortality (death, given overlap, encounterability, and selectivity). Ideally, each of these parameters should be quantified before defining a species’ risk to fishing activities (i.e., bycatch). But quantifying each of those attributes is difficult when available data are sparse, as is the case for most threatened populations. Using simple measures of spatial-temporal overlap may be useful (Howell et al., 2008; Howell et al., 2015; Cronin et al., 2016; Hatch et al., 2016; Eguchi et al., 2017; Hazen et al., 2018; Stepanuk et al., 2018; Baird et al., 2021; Murray et al., 2021), especially if the other attributes of susceptibility can be simplified or assumed known. For example, assuming high encounterability may be reasonable if turtles are known to be predominately benthic in a certain time and area that overlaps with bottom fishing.

Here, we explore the spatial and temporal overlap between the U.S. Atlantic sea scallop fishery and loggerhead turtles (Caretta caretta) using 2 overlap metrics (i.e., Bhattacharyya’s coefficient and the Morisita-Horn index), and we assessed the ability of one of those metrics to track estimated interactions over time. We characterized the spatial and temporal distributions of loggerhead turtles using satellite tag data (Winton et al., 2018) and the U.S. Atlantic sea scallop fishery using data from the Vessel Monitoring System (VMS). The characterized distributions were combined to develop a time series of overlap measures that spanned roughly 18 years (i.e., 2003 - 2020). We then extracted estimates of fishery interactions from published reports over the same time frame and examined the relationship between overlap and interactions. We decided to focus on loggerhead turtles because they broadly overlap with the U.S. Atlantic sea scallop fishery where fisheries bycatch of marine turtles has historically been a conservation issue.

While higher degrees of overlap were associated with more estimated interactions between loggerhead turtles and the U.S. Atlantic sea scallop fishery, we would not characterize that relationship as strong. The overlap between distributions of loggerhead turtles and the U.S. sea scallop fishery poorly reflected the estimated number of interactions (R² = 0.296), yet the times and areas of higher risk aligned well with patterns in observed bycatch ( Figures 3 , 4 , C1 ). This suggests that while overlap does not encode enough information to be indicative of the magnitude of interactions, it may still be a good indicator of when and where interactions are likely to occur. With overlap in space and time being a relatively limited measure of encounter risk and probability of interaction (Murua et al., 2021, as well as this case), further refinements would be necessary to use overlap as a measure of absolute risk between turtles and U.S. commercial fisheries (Hobday et al., 2011).

We might expect that overlap metrics and interaction estimates would correlate given their direct linkages to commercial fishing effort. Interaction estimates are produced by expanding observed bycatch rates to the entire fleet using fishing effort, most of which is not observed. The overlap metrics cited in this paper, by definition, include the relative intensity of fishing effort over space and time. While the data sources used to define commercial fishing effort differ between the two (VMS for overlap and VTR for interaction estimates), they should conceivably be tracking similar patterns. So it is not surprising that overlap and interactions would associate, at least to some degree. But the low R² value suggests poor predictive performance in the ability of overlap to predict interactions, at least in our case, again reinforcing the complexity of the fine-scale dynamics that produce bycatch and the inability of simple overlap measures to capture that complexity (Hobday et al., 2011; Murua et al., 2021), as well as potential data issues that could influence the comparison between the two (discussed in more detail below).

Fisheries bycatch is the culmination of a complicated process that starts with overlap in space and time and ends with incidental mortality (Hobday et al., 2011). The progression through this complicated process may ultimately determine the strength of association between overlap and interaction. For example, scallop vessels can dredge around the clock and turtles are known to have diel patterns in behavior (Ogden et al., 1983; James et al., 2006; Casey et al., 2010; Christiansen et al., 2017). This context-dependent behavior may influence encounterability and hence the potential for a dredge to catch a turtle during daytime vs. nighttime operations. Overlap and gear selectivity, then, may be high in this example, but encounterability would vary depending on the time of day resulting in a continuum of encounter risk. This shows the multifaceted nature of interactions and the processes leading up to them, and it illustrates a need to better understand what variables could influence risk (e.g., animal behavior and abundance, fishery characteristics, spatial and temporal scale and extent, etc.) using a more integrated approach (Gilman et al., 2019).

The veracity, quality, and comprehensiveness of data are other sources of uncertainty in the comparison between overlap and interactions. First, both of the derived data products (commercial fishing effort and loggerhead turtle distribution) used in the overlap calculations have their limitations. The VMS data does not capture the type (net or dredge) or number of gear (1 or 2) used by a vessel and may not be compatible with the measure of commercial fishing effort used to estimate interactions derived from self-reported VTRs (Murray, 2011; Murray, 2015; Murray, 2021). Similarly, a simple speed-rule approach was used to define fishing activity from the VMS polls for convenience, but there are other methods (Muench et al., 2018). In addition, the relative distribution of loggerhead turtles is based on a limited number of satellite-tagged animals compared to their population size (Winton et al., 2018) and assumed to be static over the 18-year study period; yet climate change may affect turtle distribution (Patel et al., 2021), especially in what appears to be a period of slow (or no) recovery for this threatened population (Ceriani et al., 2019). Second, the limited number of observed interactions has led to high uncertainty in the estimates and has prompted the adoption of different analytical methods for different time periods, possibly resulting in an inconsistent time series (Murray, 2011; Murray, 2015; Murray, 2021).

The ability of overlap to capture some of the observed patterns in bycatch is promising for certain aspects of mitigation. Simple measures of overlap may be useful in determining where and when more research is needed (Hatch et al., 2016; Murray et al., 2021) or the targeting of observer coverage when resources are limited (Baird et al., 2021). Examples exist where simple approaches of quantifying bycatch risk using overlap have been successful at providing near real-time information to fishers (Howell et al., 2008; Howell et al., 2015). Although, as pointed out by Eguchi et al. (2017), the success of overlap as an indicator of bycatch risk might ultimately be context-specific and depend on the complexity of the ecosystem and the amount of data available to capture that complexity through statistical modeling.

In summary, we found that overlap can be loosely thought of as an indicator of relative but not absolute risk. Simple measures of spatial and temporal overlap can provide a way to identify when and where bycatch is likely to occur relative to other times and areas. However, those measures may be limited in their ability to determine the probability of interaction and subsequent interaction estimates needed for fisheries management. Simple overlap studies, then, might be best suited for informing prioritization of resources in a relative sense, like focusing research in certain times and areas, providing targeted observer coverage, or initiating management discussions to delineate where and when bycatch mitigation approaches should be implemented. While overlap may contain some information about the actual risk of an interaction, we would caution against using it as a meaningful predictor of absolute risk unless there is evidence to suggest otherwise.

The datasets presented in this article are not readily available because they are confidential. Requests to access the datasets should be directed to https://www.fisheries.noaa.gov/new-england-mid-atlantic/enforcement/vessel-monitoring-contacts.

JH led the conceptualization, analysis, discussion, and writing with input from all authors. All authors contributed to the article and approved the submitted version.