It is universally accepted that preventing incursions by taking controlling actions before or at a country’s border is the most cost-effective response to the problem of aquatic and terrestrial bioinvasions (e.g. Bax 2000, Wittenberg & Cock 2001). Prevention protocols use risk-based strategies to assess and prioritise all pathways which can bring unwanted marine species to a nation’s shores and waterways, plus a management system to provide sufficient resources for detecting and responding to border incursions. Risk-based approaches have been adopted since no nation can afford the systems, personnel and transport/trade disruptions for inspecting or treating individual shipments on every vector and route.

2.4.1 Risk Assessment

Risk is the likelihood (probability) of a harmful event occurring, multiplied by the magnitude of the consequences if the event occurs (= level of public harm, ecosystem damage, economic loss, etc). An event which has a low probability of occurring can still pose a large risk if its harmful consequences are deemed severe or catastrophic. A conventional risk analysis involves four steps: (1) hazard identification, which requires a good understanding of the transport systems and vector/s in question; (2) hazard analysis, which takes time and money to perform reliably; (3) an exposure/consequences assessment (i.e. what becomes exposed to what should the identified event/s occur); and (4) the risk calculation and evaluation process, whose reliability and usefulness depends on the quality and completeness of the preceding steps. A fully quantified risk assessment (QRA) also includes measures of the uncertainty of its results.

Risk can be estimated by qualitative, semi-quantitative or quantitative methods, depending on the type and amount of available information and completeness of the model of the hazard under investigation (e.g. US-ANS-RAMC 1996, Hayes 1997, US-NSTC-CENR 1999, Hewitt

& Hayes 2002, Raaymakers & Hilliard 2002, Russell et al 2003, URS-Meridian 2004). Any quarantine, coastguard or fishery agency which uses a risk-based approach for managing marine introductions therefore faces three options for its risk analysis:

(1) Performing qualitative risk assessments, such as Delphi or similar techniques to crunch group-reviewed opinions, guesstimates, score-cards etc. However the outcomes of these relatively inexpensive but essentially subjective exercises can vary according to the backgrounds and influence of the individuals or organisations involved, and the results tend to overestimate low probability/high consequence events while underestimating higher probability/lower consequence events (e.g. Haugom et al, in Leppäkoski et al.

2002; also Section 2.4.2); or

(2) Spending more money on semi-quantitative methods that focus on particular vectors amenable to route assessment (e.g. ballast water, hull fouling, fishing equipment) and use indirect measures of bioinvasion potential. Semi-quantitative methods are aimed at maximising the use of available data to reduce subjective input as much as possible, thereby allowing the information to ‘speak for itself’ (e.g. Hayes 1997, Ruiz & Carlton 2001, Raaymakers & Hilliard 2002, Clarke et al 2003); or

(3) Collecting detailed information on specific taxa with known harmful characteristics to permit (a) quantitative risk assessments of a single key vector such as ballast water, then (b) incorporation of its results into a real-time ‘decision support system’ (DSS), which is used to identify and manage the risk posed by each intended ballast tank discharge (e.g.

Hayes 1998, 1999, Hayes & Hewitt 1998, Colgan 1999, Hewitt & Hayes 2002).

Option (3) is expensive and not easy to achieve owing to the number of key knowledge gaps in


process. By causing rapidly increasing uncertainties with each step of a modelled invasion, these gaps preclude use of conventional QRA methods (e.g. Hayes 1997, Hewitt & Hayes 2002). A major stumbling block faced by conventional QRA has been the inability to identify which combination of species life-cycle characteristics and receiving ecosystem features provide sufficiently reliable predictors of NIS establishment and harmful invasion.

For example, an ideal ‘end-point’ of a ballast water risk assessment (= the level of damaging consequences should a harmful NIS establish and spread) would involve calculating, for each species potentially present in each ballast tank, some measure of its ability to establish in the receiving environment, spread and then reduce native biodiversity, disrupt a key ecosystem process and/or damage pre-identified socio-economic values. Such end-points would allow decision-takers to use risk-acceptance criteria to evaluate the cost-benefits of different management options for each pathway and associated group of species. However the more that information gaps and poorly understood steps in the invasion process force inclusion of conservative assumptions into the risk calculation, the more the risk assessment results become uncertain, uniform and of little value to cost-benefit evaluations.

On the other hand, if the consequences (end-point) of ballast water discharges are simplified and critical data gaps can be filled via research and documented evidence, the risk posed by particular vectors and routes may be estimated with more useful precision. An amenable end-point would be: “Inoculation of any life-cycle stage of a suspected harmful species into a receiving environment, which on current evidence appears capable of permitting its survival and the likely establishment of a harmful invasive population”. Such compromise end-points are required because marine invasion science is an emerging discipline where the name of the game remains ‘ecological roulette’ and where the problem of determining the potential harmfulness of particular species needs to be avoided (Carlton & Geller 1993, Carlton 2002, Hewitt & Hayes 2002; Section 2.4.2).

As noted by Hewitt & Hayes (2002), expressions of risk relating to the potential for unwanted establishments carry “…an implicit assumption that the establishment of any exotic species in the locality is an undesired event. This is equivalent to an expression of environmental value that wishes to preserve ‘natural’ or existing species assemblages”. Such end-points are precautionary as they treat all known and suspected invasive species as potentially harmful and equally unwanted (i.e. ‘risk species’, ‘next pests’; Section 2.2),. They are most suitable when there is no need to predict, compare or evaluate the ultimate consequences from incursions of particular species.


Invasion biologists have frequently resorted to listing species’ traits and characteristics in attempts to predict which are likely to spread and cause harm if introduced. These include:

• reproductive mechanism/s and fecundity,

• duration and toughness of the dispersal stages,

• tolerance to local ranges in temperature, salinity and other physical factors,

• life form and habitat requirements, and

• diet or nutrient needs, growth rate and other pre-adaptive factors.

However the characteristics exhibited by a species in its native range can differ to those displayed by its populations in invaded areas (owing to genetic founder and acclimation effects). In addition, key life cycle and environmental tolerance data are often lacking,


predictions versus outcomes have not been reliable, and assessments which rely heavily on expert subjective opinion can generate a false sense of security (e.g. Simberloff & Alexander 1998, Simberloff 1999, Ruiz & Carlton 2001, Hewitt & Hayes 2002).

The one factor which has shown good correlation with predicted outcomes is whether or not the species in question has been invasive elsewhere. Matches of water temperature/salinity ranges are also useful, but by themselves are not as reliable as the former predictor owing to the wide thermal/salinity tolerances of many estuarine and brackish water taxa. On the other hand, while some terrestrial invaders have expanded into novel habitats once outside their native range, the ability for marine species to occupy divergent habitats is constrained by the powerful hydrodynamic, sedimentary and physico-chemical processes that operate uniformly across the aquatic realm5.

Environmental matching methods trialled for predicting marine bioinvasion potential have ranged from simple overlaps of climate region (one variable) and temperature/salinity range plots (four variables), to comprehensive multivariate similarity analyses using a range of environmental descriptors (e.g. Hilliard et al 1997b, Ruiz et al 2000, Gollasch 2002, Hayes et al 2002, Hayes &

Hewitt 2002, Hewitt 2002, Ruiz & Hewitt 2002, Raaymakers & Hilliard 2002, Clarke et al 2003, URS-Meridian 2004).

Adopting a multivariate approach allows a wider range of coastal and inland habitat types and aquatic regimes to be resolved. For example, simple overlaps of average or extreme salinity ranges will not separate highly seasonal estuaries (those which experience a sudden, major salinity decline but only during a short monsoon or spring-melt season) from those with regular short-term salinity fluctuations caused by more uniform freshwater inputs throughout the year, as found in humid equatorial and temperate maritime regions6.

There is also increasing support for the paradigm of aquatic ‘invader friendly’ receiving environments, since the highest numbers of marine introductions are typically associated with estuaries, harbours or bays dominated by artificial, disturbed and/or eutrophic habitats.

Reduced native biodiversity and increased vacant niche space due to eutrophication, over-fishing (which also increases nutrient availability via reduced grazing pressure and standing biomass), land reclamation/urbanisation, river damming, etc, have been linked to the invasion

‘hot-spots’ and ‘meltdowns’ reported from a range of areas such as the north-west Black Sea, San Francisco Bay, Port Phillip Bay (Melbourne) and Mediterranean localities, including the Berre, Thau (Hérault) and Venice lagoons (e.g. Nichols et al. 1990, Cohen & Carlton 1995, Verlaque 2001, CIESM 2002, CSIRO 2002; B. Alexandrov & G. Parry, pers. comms).

Two other features of the receiving environment have been associated with an increased propensity for both aquatic and terrestrial introductions:

• Its degree of biogeographic isolation and associated percentage of endemic species.

Aquatic environments with low biogeographic connectivity and high endemicity include those in the Ponto-Caspian, Eastern Mediterranean, Laurentian Great Lakes, Mississippi-Missouri Basin, southern Australia, New Zealand, Hawaii and parts of the American Pacific coast (e.g. Cox & Moore 1980, Longhurst & Pauly 1987, Veron 1995, Hilliard et

5 As noted by Veron(1995), “Biogeographic concepts developed from terrestrial biota often have very doubtful relevance to the ocean”.

6 The recent ballast water risk assessments undertaken for six ports by the GEF/IMO/UNDP Global Ballast Water Programme incorporated a multivariate port matching method that used 34 variables to resolve physically divergent marine, brackish and freshwater habitats. Highly seasonally estuarine ports were resolved by including measures of seasonal rainfall intensity, distance to nearest river mouth and size of catchment


al 1997, Leppäkoski & Olenin 2000). Biogeographically isolated regions which also have relatively low native marine biodiversity, such as the Hawaiian Islands, Eastern Mediterranean and Baltic Sea, have also been considered to provide vacant niche space which facilitate successful aquatic introductions (e.g. Por 1978, Baltz 1991, Hilliard et al 1997, Galil 2000, Leppäkoski & Olenin 2000, Hutchings et al 2002).

• Its latitude. The rise in NIS numbers per steps of increasing latitude has been noted along both sides of the Australia and discussed by Hilliard et al (1997), Hewitt (2002) and Coles & Eldredge (2002). Several reasons have been put forward as to why temperate regions may be more prone to marine bioinvasions than tropical and polar coasts (e.g.

Hilliard et al 1997, Hewitt 2002, Hutchings et al 2002, Ruiz & Hewitt 2002, Barnes &

Fraser 2003).

In summary, current best predictors for identifying which and where marine introductions are mostly likely to establish, spread and cause harm may be ranked as follows:

(1) Documented evidence that the species of interest has invaded several region/s and caused demonstrable harm.

(2) Limited evidence and inductive evaluations for suspecting the species of interest is potentially invasive and capable of causing harm.

(3) The degree of similarity between the climate, hydrological characteristics and aquatic habitats of the receiving region and those colonised by the species of interest in its natural and other introduced ranges.

(4) The degree of ‘invader friendliness’ of the waters where the inoculations occur. For example:

- the number of recently established NIS;

- the percentage of artificial, heavily modified or disturbed habitats that offer vacant niches due to absence or immaturity of native assemblages;

- the presence of depauperate native communities owing to eutrophication, pollution, overfishing, river dams or other disruptive processes.

(5) The range of secondary pathways available (i.e. number and frequency of local vectors and their routes that can assist regional spread).

(6) The presence of biogeographically isolated aquatic communities containing a high percentage of endemic taxa and/or offering naturally vacant niche space owing to relatively low biodiversity.

(7) The location of the receiving waters with respect to tropical/polar latitudes (20-60o) However there are insufficient field data to adequately quantify these trends. There are also mechanisms that influence invasiveness on a more ‘case by case’ basis which are difficult to predict. For example, the absence of one or more pathogens, parasites or predators which controls populations in the native range by reducing fitness, fertility and/or longevity. For these reasons there remains no substitute for careful biological research on any species of concern, both in its natural and invaded habitats.