Tidal turbines are a potential Marine Renewable Energy (MRE) source that can be deployed in high flow current regions (Pelc and Fujita, 2002). Tidal energy technology has been deployed but widespread adoption is hampered, in part, by concerns of aquatic animal impacts (Copping and Hemery, 2020). Primary concerns include collisions with stationary turbine components and/or strikes from rotating blades that could inhibit growth or affect survival of fish, seabirds, or marine mammals (Hemery et al., 2021). Knowledge gaps and inadequate empirical data on animal-device interactions necessitate obtaining, quantifying, and interpreting physical and biological data that can be used to discern the influence of animal-turbine encounters (e.g., Copping and Hemery, 2020), collisions (e.g., Müller et al., 2023), and blade strikes (e.g., Castro-Santos and Haro, 2015; Courtney et al., 2022). Unverified perceptions of mortality from animal-device interactions can impede development of monitoring regulations for tidal turbine sites. Comprehensive baseline and post-installation monitoring data on animal-tidal device interactions are not available, resulting in uncertainty among regulators who are cautious when permitting full-scale MRE sites (Copping et al., 2020a).

Encounter and collision rates between aquatic animals and tidal turbines are not well quantified due to limited opportunities and appropriate technologies to observe, measure, and characterize interactions (Fox et al., 2018; Courtney et al., 2022; Bender et al., 2023). Worldwide, there have been few (Copping et al., 2021) acoustic and optical technologies deployed to monitor tidal energy sites (e.g., Williamson et al., 2017; Staines et al., 2022). Even though stationary acoustic multibeam and multi-frequency echosounders are available, their deployment is often limited due to operational constraints including limited detection of weaker targets (Williamson et al., 2017). An approach that supplements empirical measures when animal behavior and hydrodynamic data are limited is the use of models to estimate the probability of animal-device interactions (Buenau et al., 2022). These studies include fish swimming trajectories during approach (Shen et al., 2016) or interaction (e.g., Viehman and Zydlewski, 2015; Bevelhimer et al., 2017) with tidal turbines. There remains a need for a comprehensive model that quantifies probabilities as fish approach and potentially interact with a hydrokinetic turbine.

To fully estimate potential encounter and interaction risks that influence MRE monitoring requirements and operational regulations, additional risk factors should be incorporated into a conditional, encounter-impact probability model. Current empirical observations and many encounter models lack active and passive avoidance behaviors of fish as they approach and interact with a device. Collision with stationary components of a device is another factor that is not commonly separated from blade strikes in published models. Collisions with stationary structures (Müller et al., 2023) could disorient fish (Courtney et al., 2022) and potentially lead to a blade strike.

This study develops a conditional probabilistic model that quantifies encounter and interactions of fish with tidal turbines. The encounter-impact model estimates probabilities of approach, encounter, collision with stationary components, blade strike by rotating blades, and sequential collision and blade strike using acoustic data from Admiralty Inlet, Washington, United States, and literature values when empirical data are lacking. Existing data gaps are identified along with appropriate next steps for model application. This encounter-impact model is designed to be generic and can be applied to any potential tidal energy project site.

Probabilities of occurrence for each component of the encounter-impact model are influenced by turbine type, time of day, and avoidance. Probabilities of occurrence for the zone of influence range between 0.0636 to 0.0649 for both turbine types (Tables S1A–D, Supplementary Material). Entrainment probabilities range between 0.00245 to 0.0408 for an axial-flow turbine (Tables S1A, B, Supplementary Material) and 0.0118 to 0.0408 for a cross-flow turbine (Tables S1C, D, Supplementary Material). Probabilities of impact depend on occurrences of collision, blade strike, or sequential collision and blade strike. Collision probabilities between fish and tidal devices span three orders of magnitude from 0.000364 to 0.324 for both turbine types, with similar probabilities of blade strike ranging between 0.000261 to 0.40 for both turbine types (Tables S1A–D, Supplementary Material). As expected, probabilities of collision and blade strike are lower than either single event impact, ranging between 0.0000242 to 0.0678 for both turbine types (Tables S1A–D, Supplementary Material). Overall impact probabilities, a summation of subcomponents, for the two turbine types are nearly identical ranging between 0.00110 to 0.666 for an axial-flow turbine and 0.00110 to 0.689 for a cross-flow turbine (Table 2).

When comparing occurrence probabilities for each model component, estimates are higher at night than during the day for both turbine types averaging between 0.00194 to 0.100 (Tables S1A–D, Supplementary Material). Blade strike occurrence probabilities are the only model component that have higher average probabilities during the day than at night, with an average 0.00238 difference for the axial-flow turbine (Tables S1A, B, Supplementary Material), and an average 0.00237 difference for the cross-flow turbine (Tables S1C, D, Supplementary Material). When comparing overall impact probabilities in light regimes, probabilities are higher at night than during the day for both turbine types, with probability variations ranging over three orders of magnitude (Table 2). Turbine design influences impact probabilities, with an axial-flow turbine exhibiting the lowest risk of impact across factors and avoidance scenarios (Table 2).

As expected, probabilities of occurrence for each model component are higher when no avoidance is included and model components are not conditioned on preceding events in model calculations (Tables S1A–D, Supplementary Material). Probabilities of occurrence are lowest when Admiralty Inlet avoidance rates are applied, reflecting the inclusion of conditional probabilities in model calculations. Probabilities of occurrence for the zone of influence are similar across all avoidance scenarios for both turbine types ranging between 0.0636 and 0.0649 (Tables S1A–D, Supplementary Material). Probabilities of occurrence for entrainment are higher when Shen et al.'s 2016 avoidance rate (0.0408) is applied to the model for both turbine types (Tables S1B, D, Supplementary Material). Probabilities of impact are highest by two to three orders of magnitude when no avoidance is included for a cross-flow turbine (Table 2). Collision probabilities (0.324), blade strike probabilities (0.40), and sequential collision and blade strike probabilities (0.0678) are all highest for both turbine types when subcomponents are modeled with no avoidance (Tables S1A–D, Supplementary Material). Minimum and maximum probability values are similar between subcomponents and overall impact estimates, with larger values occurring when no avoidance is applied and lowest when avoidance rates from Admiralty Inlet were used in model calculations (Table 2).

Conditional probability estimates from this study are both lower and higher than other published values (Table 3). Shen et al. (2016) and Bangley et al. (2022) observed order of magnitude higher probabilities of fish approach and encounter with a tidal turbine than average approach estimates in this study. Similarly, Viehman and Zydlewski (2015) report order of magnitude higher average probabilities of entrainment at night with a 0.290 probability estimate difference between day and night calculations. Band et al. (2016) observed order of magnitude higher probabilities of collision for Harbor seals with turbine rotors when compared to results of this study. In contrast, Wilson et al.'s (2006) non-conditional encounter probabilities for Pacific herring were two orders of magnitude lower than those estimated in this study.

Regardless of the combination of factors, probabilities of fish-turbine encounters and impact range from a minimum of 0.0000242 to a maximum of 0.324. Overall impacts ranged from a minimum of 0.00110 to a maximum of 0.689. Probability values are particularly low when conditioned on fish occurring within a turbine’s zone of influence, where subsequent entrainment may lead to an impact. All highest probability values occur at night with no avoidance in calculations for a cross-flow turbine.

Occurrence probability estimates of fish-turbine interactions are influenced by model component, device surface area, and turbine design. The concept of a zone of influence, demonstrated by Shen et al. (2016) and Grippo et al. (2020), represents a large range in which fish can respond to turbine presence. Larger turbine designs, exemplified by the 30 m by 10 m silhouette of the ORPC TidGen cross-flow turbine, also present a large surface area for potential collisions and interactions with a device. The cross-flow turbine is approximately six times longer than the Verdant Power KHPS axial-flow turbine. Greater cross-flow impact probability estimates and congruent empirical blade strike estimates from Courtney et al. (2022) demonstrate high probabilities of entrainment and collision associated with cross-flow turbines, attributable to the large size of the device. Design characteristics of turbines also affect the entrainment model component, as entrainment dimensions scale with turbine size.

Light and dark cycles have limited influence on empirical data-based variations in occurrence and impact probabilities. A slight increase in probability values is observed for model estimates based on night empirical data compared to those sampled during the day. Fish behavior in light and dark conditions provides insight on fish-turbine detection distances where field (Viehman and Zydlewski, 2015; Viehman et al., 2015; Williamson et al., 2019) and experimental (Yoshida et al., 2021) studies found that light intensity affects fish distribution in the presence of MRE devices. Williamson et al. (2019) observed a 2.63 times greater increase in fish aggregation rates around turbine structures at night compared to day, supporting previous studies that show greater probabilities of turbine entry for fish at night (Viehman and Zydlewski, 2015). Viehman et al. (2015) reported that fish are more evenly distributed at night, even at dynamic tidal turbine sites, demonstrating the persistence of fish in dark conditions where turbines are present.

Variations in model component and impact probability estimates are contingent on turbine type, light conditions, avoidance scenarios, and blade strike probabilities. Occurrence probabilities for the zone of influence are unaffected by avoidance scenarios, with slight elevation in nighttime estimates, as probabilities are directly extracted from the Admiralty Inlet empirical data that had no turbine present at the site. Probabilities of entrainment modeled with no avoidance are similar or an order of magnitude lower for the smaller axial-flow turbine compared to the cross-flow turbine. Analogous encounter rate studies (Bangley et al., 2022; Sanderson et al., 2023) using acoustic telemetry tags found encounter rate probabilities an order of magnitude higher compared to the conditional encounter probability estimates. Impact probabilities in model scenarios with no avoidance (e.g., Wilson et al., 2006) result in higher values by one to two orders of magnitude compared to impact subcomponents that include avoidance. It is worth noting that fish-turbine encounter probabilities reported by Wilson et al. (2006) are based on Pacific herring but are not conditional and do not incorporate active or passive avoidance. Comparing blade strike probabilities derived from literature values, Yoshida et al.'s (2021) probabilities result in the lowest overall impact probability estimates when combined with an avoidance scenario. These lower probability values are attributed to a lower turbine blade rotational speed to fish swimming speed ratio, resulting in greater avoidance and lower blade strike rates. In contrast, impact probabilities are highest when using blade strike probabilities from the Romero-Gomez and Richmond (2014) blade strike model that does not include fish avoidance, despite model probabilities decreasing as flow speeds increase. In combination, our probability estimates demonstrate that avoidance is an important factor influencing impact probability estimates, both as a scenario within the conditional model and experimentally with fish and a turbine present.

Additional empirical data are needed to quantify collision rates with stationary structures and blade strikes. Data that track individual fish through turbine encounters supports probability estimates of avoidance, collision with turbine structures, and the combination of collision followed by blade strike. Long-range fish trajectory data can be used to quantify active and passive turbine avoidance behaviors through each step of a sequential encounter model and be used in conditional probability of occurrence calculations. The lack of collision and sequential collision and blade strike data or suitable published values necessitated modification of blade strike rates for model subcomponent calculations. Parameter values for these impact subcomponents are derived by multiplying blade strike probabilities from the literature using Shen et al.'s (2016) avoidance rate of 0.372. The use of published blade strike probabilities in calculation of collision probability estimates may have increased collision probabilities. To illustrate by example, Courtney et al. (2022) observed greater blade strike occurrences compared to other studies that found no blade strikes in natural environments (e.g., Hammar et al., 2013; Viehman and Zydlewski, 2015; Shen et al., 2016; Bevelhimer et al., 2017 Romero-Gomez and Richmond's, 2014) blade-strike model, parameterized for Pacific herring in Admiralty Inlet, does not include avoidance, which increased blade strike estimates compared to field (Hammar et al., 2013; Courtney et al., 2022; Bender et al., 2023) and experimental (Zhang et al., 2017; Berry et al., 2019; Yoshida et al., 2020; Yoshida et al., 2021; Müller et al., 2023) studies. Probabilities calculated using blade strike rates from Courtney et al. (2022) and Romero-Gomez and Richmond (2014) resulted in overall impact probability values that range one to two orders of magnitude higher than estimates derived from Yoshida et al.'s (2021) blade strike estimates. The lack of data or published probability values for collisions and blade strikes, that also include avoidance rates, illustrate a current knowledge gap. Availability of animal-turbine interaction datasets facilitates validation of probability estimates and will allow resource managers to quantify injury and mortality of aquatic animals including species of special status such as the threatened Puget Sound Chinook salmon (Oncorhynchus tshawytscha) (Hall et al., 2018).

Empirical data from a demonstration tidal turbine site are used in this study but sample conditions may affect probability estimates. For example, fish densities derived from the Admiralty Inlet acoustic dataset are categorized as Pacific herring. Representing a mixed fish community by a single species in the conversion of acoustic backscatter measurements to density and abundance estimates is potentially biased, but any biases in the data are assumed constant and low. Pacific herring are used to represent pelagic, schooling fish that are common constituents of any fish community at a MRE site. The encounter-impact model does not explicitly include diel vertical migration, a factor that may affect probability estimates (e.g., Rossington and Benson, 2020), but did include day and night periods to reflect differences in vertical distributions. We did not include crepuscular changes in vertical distributions, which represent brief periods (<3% total) during a 24-hour cycle. Williamson et al. (2019) examined day/night fish aggregation vertical distributions in proximity to turbine structures and observed lower school heights at night compared to those during the day, suggesting that vertical migration can potentially influence probability estimates for specific times during a day, but also emphasizing our limited understanding of fish behavior in proximity of tidal devices. At the time of data collection there were no hydrokinetic devices deployed in Admiralty Inlet. Fish density data used in probability calculations lack information on fish-turbine interactions, necessitating the use of published avoidance, collision, and blade strike values. The absence of a turbine at the data collection site also precludes the ability to measure indirect or delayed impacts of animal-device interactions. Indirect or delayed impact examples include hydraulic shear stress (Hammar et al., 2015) and/or barotrauma (Copping et al., 2020b) that may lead to additional fish injury and/or mortality. Conditional probability values are calculated using empirical acoustic transect data along sequential steps in the encounter-impact model. The data serve as a series of spatiotemporal snapshots of fish distributions but do not explicitly include individual fish trajectories as they pass through a model domain.

The encounter-impact model provides a robust framework for estimating impact probabilities for both individual fish and specific turbines. The model is not structured to estimate encounter or impact probabilities for fish populations or for arrays of tidal turbines. Models such as Hammar et al.'s (2015) population collision risk model and the Exposure Time Population Model (ETPM) from Grant et al. (2014) estimate impacts of animal-tidal turbine interactions on populations encountering and interacting with a device. Hammar et al.'s (2015) model incorporates a component for tidal turbine array passage and co-occurrence, depicting fish approaching and being at the same depth as rotor blades during turbine operation. This component is used to estimate probabilities of a fish population passing through a tidal turbine array and potentially encountering rotor-swept areas of any turbine within the site. The ETPM estimates collision risk for diving birds interacting with MRE devices by evaluating mortality rates that would lead to population-level impacts, which could be adapted to estimate fish and aquatic mammal interaction and mortality rates. The challenge remains to scale high resolution, individual animal-device interactions to populations or species in a single model that accurately estimates impacts of MRE arrays on aquatic communities.

Our model combines analyses of empirical data from Admiralty Inlet with literature values to estimate probabilities of device encounters and impacts on fish. Numeric models can also be used to estimate values for unknown variables such as fish approach and turbine interaction that identify data gaps in the MRE research portfolio (Buenau et al., 2022). An alternate approach to address challenges associated with incomplete empirical data is agent-based modeling. Agent-based models represent populations using individual agents (i.e., organisms) containing unique traits and interact with other agents and their environment (DeAngelis and Mooij, 2005). Within a specified domain, agent-based models are initialized with parameter values and behavioral rules (e.g., aggregation, avoidance) formulated using field observations, existing datasets, or published values (Murphy et al., 2020). When applied to MRE, agent-based models can include individual and aggregative behaviors in response to hydrokinetic devices (e.g., Goodwin et al., 2004) that are then used to quantify impact probabilities for individuals (e.g., Rossington and Benson, 2020) or populations (Hammar et al., 2015). The combination of empirical data with numerical models is a formidable tool to assess fish interactions with MRE devices and is essential for informed regulatory decision-making, conservation strategies, and sustainable development of the MRE blue economy.

MRE is an applied science and engineering field that requires foundational and operational understanding through increased research and environmental monitoring. One area of uncertainty is characterizing fish avoidance behaviors, including reaction distances to MRE devices. This knowledge gap hinders permitting/consenting and subsequent development of MRE projects worldwide. To facilitate progress from demonstration projects to commercial-scale sites, it is essential to implement effective risk management strategies, comprehensive environmental monitoring, and regulatory frameworks that provide clear standards for operation of all MRE sectors including tidal energy (Inger et al., 2009). The encounter-impact empirical model in this study estimates probabilities of occurrence for sequential stages of fish interactions with tidal turbines and can be adapted for any species, location, and device. The comprehensive yet flexible structure of this model serves as a starting point to quantify encounter and impact risks and to further discussion on impact uncertainty.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Only remotely sensed, acoustic data were collected in the original study by the University of Washington. Species composition and length data from net catches were originally collected by and under the authority of the Washington State Department of Fish and Wildlife (WDFW). We received data that were used to choose the type species and calculate an average length for conversion of acoustic backscatter (i.e., reflected energy) measurements to fish density and abundance estimates.

JP: Conceptualization, Data curation, Formal Analysis, Investigation, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing. JH: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing – review & editing.