Recreational vessels are an important pathway for spreading marine non-indigenous species (NIS) around coastal environments globally. However, most vessels are not tracked, limiting our ability to map their movements and identify locations at greater risk of NIS introductions. Using New Zealand as a case study, we quantified spread and risk patterns of recreational vessel movements, using a web-based survey allowing the more than 1,800 respondents to map significant trips with over 12,000 visits. These vessel routes were used to build a network model representing nationwide recreational vessel movements. Two proxies were used to quantify the risk of introduction of marine NIS: (i) incoming hull length and (ii) cumulative residency periods at sites. There was significant variation in the distances travelled, the destinations they visited, and the duration of their stays. New Zealand’s recreational boating network contained 317 destinations with over 4,000 unique connections, concentrated within two distinct areas of the country. Network-based metrics and risk proxies quantified the relative importance of domestic locations as incursion or spreading hubs for NIS. This approach highlighted several areas that pose high relative biosecurity risk within the national network, but are underrepresented within current surveillance programmes. Synthesis and applications. Our study demonstrates how the movement dynamics of recreational vessels can be quantified at a regional scale to inform proactive management. The identification of spreading hubs and locations at particular risk of NIS introductions, enables managers to design risk-based and effective surveillance and monitoring programmes. Our network-based approach to determine the biosecurity implications posed by recreational vessels is transferable to other parts of the world. It enables managers to understand the distribution of risk within an area of interest (e.g., a jurisdiction) and develop optimised approaches for mitigating impact.

Web-Based Vessel Owner Survey

Because recreational vessels are not tracked by any centralised system, we obtained information on vessel movements through a survey of New Zealand’s recreational vessel owners. A web-based spatial survey platform (https://maptionnaire.com/)(human ethics: CAW-ETH-210609). The survey ran between October 4^th, 2021, and February 26th, 2022, and was able to be completed on a computer, tablet, or smartphone. The survey initially invited participants to provide information on their vessel, including type (yacht/launch), length, which was collected in size ranges (<15 Foot (<5m)", " 15 - 30 foot (5-9m)", " 31 - 45 foot (10-14m)", " 45-60 foot (15-18m)", " >60 foot (>18m)), and their homeport and home region. New Zealand is divided into 16 regions (local government jurisdictions), hereafter referred to as regions. Participants with trailer-borne vessels were not included in the analysis, as these vessels are not in the water for extended periods. Survey participants were then asked to provide details about their most significant trips undertaken in their vessel between 2019 and 2021. These details included when each trip began (month and year), the duration of each trip (number of days), the destinations visited/stops (indicated by placing a virtual marker on an online map), in chronological order, and the time spent at each destination (residency period)). Residency periods (the time a vessel spends at a stop) were reported via a list of range classes (e.g., less than 3 days, 5-7 days, 3-4 weeks, 1-2 months), with the median number of days within each range used. To ensure maximum participation from nationwide boaters, the online survey link was distributed with the assistance of three key entities: the New Zealand Marina Operators Association (NZMOA, whose membership includes 98% of the country’s marinas), Yachting New Zealand (the main national body representing recreational boating, with > 16,000 members owning the types of recreational vessels targeted by the survey), and the regional harbourmasters who manage regional mooring areas throughout the country. The survey was disseminated through these groups via email and social media channels when the survey opened and several more times throughout. This approach aimed to maximise the survey's reach, with an estimated minimum of 60% of the recreational vessel owners contacted. To incentivise participation, two NZ$250 vouchers were offered as prizes for participants and were drawn after the survey closed.

Network Destinations (Potential Vessel Destinations)

The survey had users identify potential destinations by clicking anywhere on a map of New Zealand, rather than specific, named locations (e.g., Westhaven Marina). The points identified by the users had to then be grouped and assigned to known, management-relevant destinations. The first step in this process was therefore to produce a spatial dataset of potential destinations to which these user-defined points could be assigned. Potential destinations included known marinas, mooring areas, aquaculture sites, marine protected areas, fjords, beaches, coastal hydro systems (e.g., harbours) and islands. The positions and attributes of potential destinations were obtained from the NZMOA, harbourmasters, regional councils, the Department of Conservation, the Ministry for the Environment, and the Ministry for Primary Industries.

The list of potential destinations included many that were very close or overlapping. Treating them as separate destinations in the network model would inaccurately dilute relative risk to one location by distributing it among many overlapping points. To focus the analysis on local-to-regional patterns, we condensed these potential destinations into a suite of national network destinations (hereafter referred to as destinations, at a resolution of 10km and within destination type (see supporting information for methodology).

Network Connections (Vessel Connections)

The user-defined virtual markers from the survey were intersected with a 15 km non-overlapping buffer around destinations throughout New Zealand to assign a starting and ending destination to routes undertaken by survey respondents between stops, referred to as connections within the network. Less than 1% of the user-defined virtual markers could not be used as they fell outside these destination buffers, likely due to inaccurate participant entry (e.g., records referring to multi-day visits to unlikely locations on land or in deep offshore environments).

The next step of developing the input dataset is to quantify the number and strength of connections between the defined destinations. Individual journeys by survey respondents were aggregated to quantify the total volume of traffic passing along connections. As only ~60% of the total recreational boater population received the survey with ~10% completing it, the volume for each connection was then scaled up to estimate the full magnitude of recreational vessel traffic in an average year.

This scaling of the survey data was accomplished by converting the number of vessels travelling along each connection to a proportion of the total number of vessels travelling along connections within the survey data, with the number of vessels grouped by their home region. This proportion was applied to the estimated vessel population for each region (based on regional berth and mooring numbers), with the scaled number of vessels from each region travelling along the same connection combined for the total number of vessels on that individual connection. The number of recreational vessels residing at marinas and mooring area throughout the country was estimated using marina berth data supplied by NZMOA, and the number and occupancy of nationwide moorings from regional harbourmasters. The marina berths and moorings were assumed to be occupied, excluding Northland and Marlborough mooring areas, which have occupancy rates of 85% and 45% based on information provided by the regions’ harbourmasters.

Alongside the number of vessels, two proxy measures were used to estimate the relative risk of introducing marine NIS to destinations: (i) incoming hull length (the cumulative, summed total length of all vessels that visited each destination), and (ii) cumulative residency period (the summed total number of days vessels had spent at each destination). For hull length, the mean for all vessels travelling along the connection was calculated from the survey data. This mean hull length was then multiplied by the scaled number of vessels (derived as described above). The same was done for the residency period along each connection. Each connection in the network model includes a ‘from’ or source  destination, a ‘to’ or ending destination, the number of vessels travelling along this connection (over an average year, derived via a temporal analysis of the complete survey data), the combined hull length of these vessels, and the residency period that vessels travelling along the connection collectively spent at the ending destination. We also quantified the month of departure and distances travelled for each trip recorded by the survey respondents. To calculate distances between any pair of destinations, the route between them must not cross over any shoreline as a recreational vessel cannot do this. Therefore, we used a procedure that calculated the shortest distance between two destinations by travelling over water.  These overwater distances between destinations were calculated using the package gdistance v1.6.4 (van Etten 2017) in R. NIWA bathymetry (https://niwa.co.nz/oceans/resources/bathymetry/download-the-data) was used for land/water mask at a resolution of 1km for the cost layer.