At-border interventions are a critical step along the biosecurity continuum, to measure and control the risks associated with the cross-border movement of people and goods. Air passengers are a high-volume pathway for a range of biosecurity risk materials, against which various interventions may be used (e.g., manual searches, detector dogs, x-rays, etc.). Tasmania is an island state, and its environment and industries have benefited from a low level of invasive pests due to their geographic isolation. Therefore, relatively strict at-border interventions are used to prevent the entry of new pests, including some serious invasive pests already present on mainland Australia (e.g., Queensland (Bactrocera tryoni) and Mediterranean (Ceratitis capitata) fruit fly). Using a large interception database for domestic air passengers entering the southern Australian state of Tasmania from mainland Australia, this study applies common statistical modelling tools to assess the efficacy of interventions (namely, dog detectors, and bag searches), and to identify pathway risk factors (e.g., flight origin/route). This analysis considered the effects of interventions on both voluntary declarations by passengers and also detections of undeclared risk material on passengers. The analysis also focused on biosecurity risk items generally (e.g., fruits and vegetables, meat products, cut flowers), and items that are specifically considered to be fruit-fly hosts. The results highlight that active at-border interventions and the presence of biosecurity inspectors capture a significant volume of biosecurity risk items at the border, and detector dogs have particularly strong positive effects on the rate of interceptions, particularly for items detected on passengers. Conducting bag searches also appears to increase interceptions, both by increasing the rate of items being detected and by encouraging voluntary declarations. Sensitivity analyses then test the robustness of results to modelling implementation methods and distributional assumptions. This study demonstrates how statistical modelling can provide robust insights into biosecurity interventions and risk factors along pathways, and further highlights the value of high-quality interception data resources for informing and improving biosecurity systems.

Bactrocera tryoni, border biosecurity, Ceratitis capitata, detector dogs, fruit fly, invasive species, passenger screening, pathway risk analysis

Biosecurity border interventions seek to balance the need for the cross-border movement of goods and people against their biosecurity risks. Although intervention policies inherit from a common framework of international agreements (Outhwaite 2010), there are substantial differences in the policies and implementation of biosecurity interventions across jurisdictions. These differences produce variation in the risk of pest introductions across states, countries and regions (Whattam et al. 2014; Epanchin-Niell et al. 2021). Australia’s biosecurity system tends to have a higher appropriate level of protection and to implement comparatively stricter interventions than many other jurisdictions (Whattam et al. 2014; Black and Bartlett 2020). This is partially due to the opportunity presented by Australia’s historical isolation, which has made its primary industries relatively pest-free on a global scale. However, activities such as tourism and trade have reduced and will continue to reduce this isolation, increasing the risk of pest introductions (Turner et al. 2021; Whattam et al. 2024). Recent studies estimate an aggregated cost of invasive species to Australia to be AU$389.59 billion since the 1960s (Bradshaw et al. 2021), and a net present value of AU$314 billion for the national biosecurity system in terms of the assets that it protects (Dodd et al. 2020; Stoeckl et al. 2023). This highlights the critical importance of ensuring that border interventions act as effective barriers against biosecurity threats.

This is particularly true for the Australian island state of Tasmania (Fig. 1). Due primarily to their geographic and evolutionary isolation, island ecosystems possess a disproportionate level of the earth’s endemic species and support a large percentage of its biodiversity (Kier et al. 2009; Weigelt and Kreft 2013). Tasmania has specifically been identified as a major centre of endemism for Australian flora; for example, more than half of the 30 native Eucalyptus species in Tasmania are endemic (Crisp et al. 2001; Potts et al. 2016). Local industries, communities, and natural ecosystems benefit from the state’s relative isolation and low levels of pests, including species that are present elsewhere in Australia, such as Queensland fruit fly (‘Qfly’, Bactrocera tryoni; Florec et al. 2013), Mediterranean fruit fly (‘Medfly’, Ceratitis capitata; Cook and Fraser 2015), tomato potato psyllid (Bactericera cockerelli; Moir et al. 2022), and grape phylloxera (Daktulsphaira vitifoliae; Skinner 2018). These are potential threats for Tasmania, and a 2018 incursion of Qfly in the state’s north cost millions in direct eradication costs in addition to further indirect costs (e.g., via temporary market access losses; Blake 2019). Tasmania’s low-pest status is therefore a biosecurity challenge, and islands can be particularly vulnerable to impacts from invasive pests and diseases (Keitt et al. 2011; Fraser 2016; Brettell et al. 2021). Furthermore, globalisation is expected to increase movement across borders in both goods and people and to continue to bridge the geographic barriers that once kept Australia and Tasmania isolated (Dodd et al. 2015; Seebens et al. 2017, 2021).

Figure 1.

Location of Tasmania in relation to mainland Australia (inset), including the locations of the seven arrival ports for air passengers, namely (from left to right) 1. King Island, 2. Burnie, 3. Devonport, 4. Launceston, 5. Bridport, 6. Hobart, and 7. Flinders Island. Base map produced via QGIS (v3.24.2; www.qgis.org). (Note, airports have been anonymised for the remainder of the analysis and labelled Airport_A, Airport_B. etc.)

Air passengers are an important high-volume pathway for pest introductions, with around 20 million passengers arriving annually in Australia in 2023–24 (BITRE 2024). Pests may be introduced via luggage (e.g., via infested fruit) or by being attached to passenger clothing or belongings (e.g., soil on shoes or sports equipment; McNeill et al. 2011; Pace et al. 2022; Robinson and McNeill 2022). Air passenger pathways may be particularly important for pest insects, including Medfly (Liebhold et al. 2006; McCullough et al. 2006). Passengers are also a pathway for animal diseases, for example, measures targeting African swine fever detected a yearly average of 33,684 pork products from 2% of screened passengers entering Australia from 2021/22–2023/24 (DAFF 2024). Interventions on these pathways must therefore be able to mitigate a high volume and a diverse range of biosecurity threats.

A range of risk mitigation tools can be employed at multiple points on the air passenger pathway, from pre-departure, in-transit and on-arrival screening/inspection phases of the biosecurity continuum (Sequeira and Griffin 2014; Whattam et al. 2014). For international arrivals into Australia, travellers are subject to pre-arrival risk profiling, and a subset of arrivals are subject to active interventions/screening at the border (Inspector-General of Biosecurity 2019). Common at-border interventions include manual examination, dog detector teams, and x-rays (Inspector-General of Biosecurity 2022). Tasmania also applies similar active interventions for domestic arrivals, including manual bag searches and detector dogs.

Detector dogs may be particularly valuable in air passenger screening, being able to screen large volumes of passengers and luggage efficiently, and able to be trained to target general biosecurity materials as well as specific pests and diseases (Whattam et al. 2014; Moser et al. 2020). Nonetheless, there are limited studies assessing their efficacy relative to other intervention methods. Furthermore, biosecurity interventions are often targeted towards specific flights to maximise the utility of limited resources, e.g., flight-based-traveller profiles used to target international arrivals into Australia (Inspector-General of Biosecurity 2019). Therefore, further empirical evidence about the relative efficacy of different interventions may help target intervention resources towards the highest-risk arrivals.

The analysis of biosecurity interception/surveillance data is prone to some common issues in statistical modelling, including zero-inflation (i.e., where data includes a large proportion of zeros, for example where detections of targeted items are rare), overdispersion (e.g., where variance is much higher than predicted), and censoring (e.g., if data is only recorded where contamination is detected; Kachigunda 2020; Turner et al. 2020; Trouvé and Robinson 2021; Kachigunda et al. 2022). Failing to account for overdispersion or zero-inflation in data can lead to biased or inaccurate parameter or error estimates (Harrison 2014; Campbell 2021; Feng 2021). Although some studies suggest that the outputs of mixed-effects models can be robust to violations of distributional assumptions (e.g., Schielzeth et al. 2020; Knief and Forstmeier 2021), exploring the potential effects of model design and implementation may be important considerations when using interception data to inform biosecurity decision-making.

Focusing on domestic flight arrivals into Tasmania from mainland Australia, the goal of this study is to assess the efficacy of passenger interventions and pathway risk factors on biosecurity interceptions. This focuses on both general biosecurity risk material (‘BRM’) interceptions, and interceptions of BRM material specifically relevant to Qfly and Medfly (collectively referred to as fruit fly, ‘FF’). Preventing FF incursions has been a focus of Tasmania’s border biosecurity system, particularly following the 2018 incursion. This study uses recent air pathway intervention data for Tasmania. These data are rich resources for our study, as they include records for all commercial arrivals, as well as relevant data on the types/amount of BRM intercepted. This provides a valuable opportunity to apply statistical modelling approaches and assess their sensitivity to implementation methods. The specific aims of this analysis were:

1. To determine the relative effects of different interventions (namely, dog detector teams and luggage searches) on the rates of BRM and FF host interceptions, including voluntary declarations by passengers and involuntary detections of items by biosecurity officers. We did not make any specific directional predictions about the effects of searches and detector dogs on interceptions.
2. To identify pathway-risk heterogeneity based on the origin and specific routes of flights. We expected substantial variation in interception rates related to flight origin and route, which may be used to identify high-risk arrivals.
3. To test whether our results are sensitive to overdispersion and zero-inflation by implementing Bayesian mixed models with zero-inflated Poisson and negative binomial distributions. We expected the outputs of models and the estimated effects of intervention methods to be robust to different implementation approaches.

Biosecurity interceptions at Tasmanian airports were strongly influenced by the methods used. For example, the number of bag searches conducted increased both detections and declarations from passengers. This suggests that increased effort in active at-border surveillance by officers will increase the efficacy of interventions by promoting both voluntary compliance by passengers and detections of undeclared risk items that may otherwise have been missed. Similarly, dog detector teams are increasingly deployed for border interventions along high-volume phytosanitary risk pathways, both in Australia and many other countries (Whattam et al. 2014; Inspector-General of Biosecurity 2022). Few studies have quantitatively assessed the efficacy of detector dogs, although a recent study from Williams and Sharp (2023) showed how the presence of a dog versus an officer alone can alter passenger behaviour including eye contact, gestures or interactions with the officer/dog. Our study provides further insights into the effects dogs may have on passenger behaviour and the rate of interceptions at airports. The increase also appears to be largely driven by detections of undeclared items instead of voluntary declarations, suggesting that dogs may be particularly useful for capturing a component of the biosecurity risk material that may otherwise not be found through more passive, voluntary compliance-based methods (e.g., public awareness and education campaigns, biosecurity signage and announcements). Although notably, one inspector alone still appears to capture about half of the biosecurity risk material entering the state relative to a detector dog team. Therefore, although the efficacy of a single person is lower than the detector dog, active human surveillance still effectively mitigates a proportion of risk at the border.

The ability to analyse pathway risk heterogeneity is limited by the type of data collected for arrivals along a biosecurity pathway. Despite the relatively limited set of pathway factors included in models (i.e., origin and route) and the relatively small proportion of total variance explained by these factors, models were able to identify specific flight origins and routes as potentially either high- or low-risk arrivals. This shows how interception data may be useful for supporting risk-based approaches to interventions (e.g., Australia’s flight-based-traveller profiles; Inspector-General of Biosecurity 2019) by identifying higher and lower risk arrivals to allocate limited resources to these arrivals (Trouvé et al. 2024). These analyses can provide important quantitative evidence supporting targeted resource allocations at the border, particularly when combined with further contextual information such as pre-border pest prevalence data or estimates of potential post-border impacts. While Tasmania currently targets 100% of air arrivals, risk-based approaches could include decisions about where and when to allocate their most effective methods (e.g., detector dogs), or to potentially identify a subset of low-risk arrivals based on factors such as origin or route that can be met with less resource-demanding methods (e.g., passive interventions, signage, amnesty bins).

While these data are valuable for identifying how interventions or pathway risk influence the actual interception rates, which is critical information for implementing biosecurity interventions to reduce risk at the border, further information is required to fully quantify the risk of incursions along this pathway. For example, the risk of fruit fly establishing through the air passenger pathway would require us to estimate the actual volume of biosecurity risk material on flights and the proportion of those items infested with fruit fly (i.e., contamination/infestation rates), the proportion of risk material missed (i.e., leakage), or the viability of any individuals or larvae that may infest any of the risk material. These parameters may be estimated using complementary methods, such as endpoint surveys that target a subset of passengers as manual or X-ray searches to estimate contamination rates (Mannix et al. 2024). Samples of intercepted biosecurity risk material may also be further tested to measure their pest contamination/infestation rates. In many cases, particularly in biosecurity, empirical data is lacking. Structured expert judgement may then be used to elicit unknown parameters from relevant experts (e.g., biosecurity managers, entomologists, academics, etc.), using advanced methods to directly elicit uncertainty in parameter estimates and incorporate this uncertainty into the decision-making process (Hemming et al. 2018; Bau et al. 2024). Therefore, although this study highlights the value of interception data for informing biosecurity practices, additional knowledge is required to more completely assess and quantify risk across a biosecurity continuum.

This study used a simple and common approach for count data (i.e., regression modelling based on a Poisson distribution), performed with modelling tools that are accessible, and relatively easy to implement. Sensitivity analysis suggested that the outputs of this approach were qualitatively similar to approaches using more advanced tools (i.e., Bayesian modelling methods), or distributions (e.g., that account for zero-inflation). However, there were some notable differences in outputs. For example, interception rates estimated from negative binomial models tended to be lower, suggesting that failing to account for over-dispersion may lead to overestimates. Furthermore, these models accounting for both zero-inflation and overdispersion (i.e., zero-inflated negative binomial models) were the best-fitting models tested. Also, Bayesian methods tended to lead to higher uncertainty estimates around fixed- and random-effect parameters, so may represent a more conservative approach to modelling pathway risk factors. Therefore, while our conclusions were generally robust to implementation methods, sensitivity analysis may also be a valuable step for providing additional information for decision-makers about the robustness of any conclusions drawn from modelling and future researchers should consider their implementation. In this case, however, the operational interpretation of the simpler models was borne out by the more complex models, i.e. that detector dogs were more effective than biosecurity inspectors.

Finally, effective at-border interventions are a key step in the biosecurity continuum. As at-border interventions become more sophisticated and widely implemented, large interception data sets will inevitably become more available to researchers and biosecurity decision-makers. This study highlights how this data can be a valuable resource for informing management decisions for Tasmania and can provide empirical evidence to support the implementation of risk-based approaches or the use of specific methods such as detector dogs, which can improve resource allocations and lead to more effective interventions at borders.

We acknowledge Biosecurity Tasmania and its staff for supporting this project and specifically thank Rae Burrows, Andrew Bishop, Ryan Wilkinson, Meg Flanagan, Renee Tiggs, Guy Westmore, Emma Hutchinson, Cameron Healey, Jason Paul, Alex Matthews, Liam Harrex, Jesse Siebler, Christian Knapp, Andrew Carter, Deepa Adhikari, Elise Baniowski, Marlene Quinn, and Chris Jones.