Guidelines for weed-risk assessment in developing countries - Peter A. Williams

INTRODUCTION

The movement of trade goods and aid of one sort or another throughout the world is essential for the well-being of all peoples. Not all these goods and gifts are benign, however, and some come with unwelcome surprises. Invasive species affect agricultural and other systems, and their impacts are second only to habitat destruction in terms of loss of biodiversity. These concerns have generated growing international interest in weed-risk assessment systems to prevent the introduction of new pests and to prioritise existing pests for control. Weed-risk assessment is a new discipline, and the first international symposium on the topic was held only recently in Australia (Groves et al. 2001). This country, along with New Zealand, is at the forefront of developing and implementing strong quarantine protocols. Both countries are relatively isolated from the rest of the world, agriculture is important to their economies, and their citizens value natural landscapes and their ancient indigenous biodiversity.

This report introduces the topic of weed-risk assessment and provides guidelines for countries wishing to strengthen their own quarantine protocols and to use scarce resources efficiently for prioritising existing pests for control. Fortunately, this task is becoming easier because the Internet allows rapid exchange of information and access to detailed databases on pests, e.g. the global compendium of weeds (http://www.hear.org).

INTERNATIONAL FRAMEWORK OF WEED-RISK ASSESSMENT

The actions taken to exclude a plant species from a country because of its weed potential must be consistent with the international standards regulating the movement of trade goods. These obligations are defined under the Agreement on the Application of Sanitary and Phytosanitary Measures (SPS agreement) of the World Trade Organisation (WTO 1994), and the International. Plant Protection Convention (IPPC) (1997 revised edition) deposited with the United Nations Food and Agricultural Organization (FAO, 1996). These two international agreements, whilst allowing countries to specify requirements for the entry of plant material, describe the obligations of countries so that import requirements are not unjustified trade barriers.

A further international convention involving weeds concerns the need to conserve biodiversity. Article 8 (h) of the Convention on Biological Diversity states that: “Each Contracting Party shall, as far as possible and appropriate, prevent the introduction, control or eradicate those alien species which threaten ecosystems, habitats, or species.” Not all countries are signatories to this convention.

A quarantine pest^[1] is defined by the IPPC, as a pest of potential economic importance in an area endangered thereby and not yet present there, or present but not widely distributed and being officially controlled (FAO, 2001a). It is accepted for the purposes of this report, that ‘economic importance’ includes actual or potential effects on the economy of ecosystems and their component species, and that the IPPC definition of a pest is sufficiently broad to include weeds covering the full range of ecosystems, including those covered by the Convention of Biological Diversity (CBD 2001). In fact, international meetings have recently been held to foster collaboration between the IPPC and the CBD (e.g. Bankok, February 6-8, 2001).

Pest Risk Analysis (PRA) is a three-stage, process of evaluating biological or other scientific and economic evidence to determine whether a pest should be regulated and the strength of any phytosanitary measures to be taken against it (FAO, 2001b).

These stages are:

Stage 1. Initiating the process by identifying a pest that may qualify as a quarantine pest, and/or pathways that may allow introduction or spread of a quarantine pest that should be considered for risk analysis in a defined PRA area.

Stage 2. Assessing the pest risk by determining which pest(s) are quarantine pests, and characterising the likelihood of entry, establishment, spread, and economic importance.

Stage 3. Managing the pest risk identified in Stage 2 by developing, evaluating, comparing, and selecting options for dealing with the risk.

The initial steps are to determine the pathway(s), that is, any means that allows the entry or spread of a pest, and correctly identify the pest. The identification of high-risk pathways is an important part of an overall weed-risk assessment process, but this report deals only with the individual pests.

The criteria used to determine the presence or absence of the potential quarantine pest in the area are represented in a flow chart in Figure 1, redrawn from the IPPC standard, Guidelines for pest risk analysis (FAO 1996). Area is defined as an officially defined country, part of a country, or all or parts of several countries (FAO, 2001a). If the species is absent, and has potential economic importance, it can be considered a quarantine pest. If it is already present in an area, then it can be legitimately considered a quarantine pest and evaluated further if it is of limited distribution or under official control. Official is defined as established, authorised, or performed by a national plant protection agency, and control is defined as suppression, containment, or eradication of a pest population (FAO, 2001a). A pest capable of further spread, that is, expansion of the geographical distribution of a pest within an area (FAO, 2001a) (Figure 1), that is not controlled, would require to be put under control to justify quarantine pest status. Species that are controlled but are at the absolute limits of their potential distribution cannot spread further and so cannot be declared quarantine pests either. In reality, most exotic species in most countries have potential for further spread.

Once the quarantine pest status has been confirmed, the next step is to assess the economic (including the environment) importance of the species. This may be high for a pest.

Weed-risk assessment is concerned primarily with the first two stages of the pest risk assessment involving pest categorization, that is, the process for determining whether a pest has, or has not, the characteristics of a quarantine pest or those of a regulated non-quarantine pest (FAO, 2001a). The minimum requirement of any weed-risk assessment system is that it satisfies the international agreements outlined in Section 2. To do this it must be built on explicit assumptions and must use scientific data.

Weed-risk assessment systems designed for use only within a single sovereign state, and which do not have the potential to limit trade, need not comply with international agreements. But to be effective they must be based on similar sound principles.

Figure 1. Pest risk analysis (from FAO, 1996).

PLANT INVASIONS AS A PROCESS

Plant species must cross a series of barriers to reach a new area and spread within it. Initially, these are physical barriers on an intercontinental and/or intracontinental scale. Species that have not crossed these barriers may nevertheless be classified as quarantine pests, primarily on the basis of their pest history elsewhere. Once they have reached the new area, they must overcome a range of abiotic and biotic barriers before establishment. Human activities are important in assisting species to cross these barriers. Species arriving in small numbers by accident have a relatively low chance of establishment. In contrast, those species spread widely as seed contaminants, raised in large numbers within a protected environment for horticulture, or planted out in agricultural or natural environments, e.g. crops or erosion control, have a greater chance of establishment. Once a species is growing in cultivation in a new country it may spasmodically appear in the wild beyond the initial plantings. If it was introduced occasionally as a contaminant of crops it may appear on associated land. The term for such sightings is casual alien or casual exotic. To become a naturalized alien or fully naturalized alien or exotic depending on the definition being followed, a species must then develop self-maintaining populations in the wild. These loci are the points from which it may spread within the area.

Different species characteristics and life-cycle stages (Figure 2) may be important at different barriers. For example, colourful flowers may be the selection criterion for transcontinental transportation in the first place. Rapid reproduction by seed and/or vegetative offspring (e.g. bulbs and tubers) (Figure 2) may then assist its spread once it has been introduced. Its persistence through periods of unfavourable climate may depend on long-lived seed banks. Crossing any of these series of barriers is reversible. A species may be extirpated locally or even driven to extinction within an area, if, for example, there are severe climatic fluctuations or new predators and diseases are introduced. The process of arrival and extinction by natural means or control may be repeated over many years until the species finally becomes fully naturalised.

The spread of a pest may follow a number of patterns in time and space, depending on such factors as its means of dispersal, life cycle, and so on. Many follow a simplified ‘S’ shaped pattern (Figure 3, solid line) that can be illustrated graphically as the proportion of all potential habitat occupied by the pest at any point in time. The essential features are a long tail at the beginning of a species spread as it crosses the first series of barriers, a steep rise as it breaks through these barriers and finds suitable habitats, and then a flattening off as these habitats are saturated. As the pest spreads, the proportion of the uninfested habitat declines at a rate defined by a ‘reverse S’ (Figure 3, dotted line). The process of spread may be continuous, but points are still recognisable (usually only with hindsight) where the rate of change alters markedly from the preceding period. For management purposes the ‘S’ shape can be idealised as stages based on the extent and rate of spread. This concept can be applied at any geographic scale, from a field to a continent.

Figure 2. Life cycle of perennial plants producing both seeds and vegetative organs (from Williams, 1997).

Figure 3. Conceptual phases in the invasion of a weed through time, and the way these relate to the percentage of occupied and unoccupied land (from Williams, 1997).

CONCEPTUAL PHASES IN THE INVASION OF A WEED

Migration phase

The species must first reach the border of the area. Once it has arrived it may, or may not, enter, depending on a variety of factors. Where there are efficient quarantine protocols and risk management procedures it will be detected and, it is hoped, eliminated, if a quarantine pest.

Escape phase

Once inside the area it may escape only occasionally, or finally become fully naturalised. The locations of these naturalisation points are likely to be associated with the pathway of introduction, e.g. in fields planted with contaminated corn, or adjacent to erosion-control plantings. They have been referred to as ‘sentinel sites.’

Establishment phase

During this phase, the plant is able to reproduce in the new environment, and population numbers slowly begin to build up. Virtually all potential habitat is still uninfected.

Expansion phase

Eventually, the number of sites occupied expands beyond the initial loci. Expansion is fastest where there are multiple loci. The causes of this expansion differ among species and are not well documented. Factors are diverse, including particularly favourable growing seasons, the arrival of new pollinators or dispersers, the species becoming adapted to its new environment by the formation of new genotypes. New habitats may be created, e.g. by changes in land use.

Some local areas of habitat are noticeably infested, but most potential habitat is un-infested. It is often only at this stage that the plant begins to be perceived as a pest.

Explosion phase

The period where the pest expands rapidly and often where it begins to attract official concern. Many potential habitats are infested during this phase.

Entrenchment phase

The pest slowly spreads to the last remaining habitats over its full range within the area. This does not mean that it occurs on all suitable land at any one time, but that it has a high chance of occurring there.

Further spread can occur only if more suitable habitat is created, e.g. by fire. Importantly, the pest may be present only in a dormant stage of its life cycle, as shown in Figure 2.

These potential changes in the spread of a pest have implications for weed-risk assessment imperatives:

• The most cost-effective means of avoiding pest impacts is to prevent their introduction or establishment in an area. Failing that, the greatest return for expenditure of money and effort comes from controlling a pest before it has spread.

• Once it has established and begun to spread, the ongoing effort required to eliminate it increases dramatically (Figure 4).

• During the earliest spread phases, when the required funds to extirpate a pest are low, these may be effectively obtained as an adjunct to other pest control programmes. Once the pests begin to spread rapidly, the effort required to obtain the funds may be orders of magnitude greater. To rephrase this, at one extreme, only the person with the spray gun (or spade or slasher) needs to be persuaded to act, while at the other, it may require the approval of government.

• There are no recorded successful weed extermination attempts (A, Figure 4) worldwide where the pest covered more than a few hectares. Species with persistent seed or other regenerative life features that require repeated visits to the site(s) are particularly intransigent.

• The total accumulative costs over time are the effort required obtaining the funds, the money spent on actual control, plus the impacts on the economy and environment. These accumulated costs become progressively greater with time if control attempts are delayed, as illustrated by the differences in the three shaded curves in Figure 4.

• Effective weed-risk assessment systems must be appropriate to: whether or not a pest has established and spread; the pest’s biology and ecology; the values being threatened; the extent to which it has have or has not established in an area; and the technologies and resources available.

Figure 4. Relative combined monetary and environmental costs of undertaking an eradication programme (A), together with those of initiating ongoing control programmes at an early (B) and late stage (C) of the invasion. Arrows indicate programme starting points. The differences in area beneath the curves (B-A, C-A, C-B) represent the benefit of control action at the earlier stage (from Williams, 1997).

Three important factors change as a pest crosses the various barriers:

a) The location and amount of information. Before a species has been introduced to an area, all information needed to undertake pest categorisation will be derived from experience of the species outside the new area. In most cases, this will mean obtaining information from its country of origin and elsewhere. The quality and amount of information will depend largely on whether the species has a history as a pest, or perhaps as a crop or ornamental. Once a species has been introduced, much information can be obtained, particularly about its growth and reproductive biology. Only as it spreads, however, will its environmental tolerances and impacts within the new area be revealed.

b) The certainty of a correct assessment. Resulting from this increasing information as the pest spreads, the reliability of a weed-risk assessment increases, and conversely, the chance of the assessment being incorrect, decreases.

c) Identification of those affected. The ecosystems on which a pest might impact cannot be reliably predicted before it has begun to spread. Neither then, can the individuals or interest groups directly affected by the potential impacts be identified. As a pest spreads through its potential habitats and range (Figure 3), those affected become increasingly identifiable. The corollary of this situation is that those who are likely to benefit from the management of the pest become increasingly identifiable. It means that at one extreme, all those within the area potentially benefit from the detection of a quarantine pest that has not reached there. At the other extreme, only those whose land the pest occupies benefit directly from the local control of a widespread pest.

It is ironic that weed-risk assessments have the greatest chance of being wrong when they are most effective in preventing accumulative impacts and costs of control, and in potentially benefiting the widest range of interest groups. That some risk is acceptable, however, is recognised by the third stage of the pest risk assessment process concerned with managing the risk.

LIMITATIONS OF WEED-RISK ASSESSMENT

Risk assessment systems concerned with both quarantine pests and established pests aim only to predict the potential harmful effects of a species. The weighing of these aspects against any potential beneficial outcomes, e.g. production of a new crop, or food in a shipment of contaminated grain, is an entirely separate exercise involving value judgements. It is not a component of weed-risk assessment as such.

Two important issues are faced in selecting a weed-risk assessment system. These are particularly acute when considering species new to an area as opposed to those that are spreading. First, it is difficult to predict a pest from only the characteristics of the potential pest. Second, that amongst any group of organisms, those attaining pest status do so at a very low rate.

Many studies have attempted to identify the characteristics of pests, as distinct from benign species. Early studies attempted to identify an ‘ideal weed’. More recent studies have concentrated on groups of similar plants within a country, a continent, or throughout the world. A few variables are associated with weediness that is broadly applicable over whole groups of plants, e.g. herbaceous agricultural weeds, or plant families, e.g. pines. These rarely have predictive value when extended to broader groupings, as from agricultural systems to the natural environment, from pines to non-pines amongst the conifers. The consensus appears to be, that no traits are universally important for all species in all habitats. The characteristics of the receiving environment are equally important.

The importance of any particular plant trait in determining the success or failure of invasion becomes discernible only after the species has either established or known to have failed in a new habitat. As the species’ fate becomes apparent through time, the reliability of the prediction will therefore increase. Even then, there may be no suit of endogenous plant characters readily obtainable from the literature that reliably predicts potential weediness. However, the chance of a plant species establishing in a new area is related to the pressure of its propagules on that area. This may be defined as the rate of individual whole plants, or vegetative or sexual reproductive parts, that are dispersed into an area over a given time period. Propagule pressure can operate at any special scale. Examples of low propagule pressure are the infrequent arrival from year to year of an occasional wind-blown seed from a distant source, the occasional seed of an unwanted species in a seed lot, or the infrequent planting of a timber species that produces few seeds. Examples of high propagule pressure are the frequent arrival on a regular basis of wind-blown seeds from a distant source, abundant seeds of an unwanted species in a widely planted seed lot, or abundant viable and readily dispersible seed from widespread plantings of timber trees.

The proportion of imported species that become pests has been calculated as ranging from about 0.01:100 for British angiosperms, 1.3: 100 for grasses into tropical Australia, and 12.0: 100 for pines in New Zealand. However, this wide diversity of ratios, and the effects of time lags between establishment and species acquiring recognised pest status changing the ratios, means the search for any constant ratio is futile. As a consequence of the low proportion of pests amongst a random selection of related organisms, any system designed to detect pests is likely to be wrong as often as it is right. The results of a false positive assessment (excluding a species when in fact it would not become a pest) could have long-term economic consequences for a country, e.g. in the case of a potential new crop. In contrast, a false negative assessment (accepting a species that becomes a pest) could result in serious economic damage.

It is equally important to realise that introductions supposedly for beneficial purposes also have a very low success rate. For example, hundreds of grasses and herbs were introduced into Australia and New Zealand, yet the agriculture of these countries is based on only a handful of species. Similarly, the exotic forest plantations of New Zealand and southern South America are dominated by a single species, Pinus radiata. Furthermore, some estimate of economic benefit can be predicted for agricultural/forestry systems assuming a certain level of uptake, or deleterious effects assuming a certain level of spread as a pest. In contrast, there is no adequate ecological theory to underpin predictions of the future impact of potential environmental weeds. There is great difficulty therefore, in predicting both the positive and negative effects of new introductions.

Despite the apparent theoretical impasse in the prediction of pest status, the task must be undertaken because of the consequences of not detecting potential new pests at the border, of benign species becoming pests, or of the spread of existing pests. Weed screening systems have been devised for woody plants in general (Reichard and Hamilton, 1997), groups of woody plants (Tucker and Richardson, 1995), and water plants (Champion and Clayton, 2001). A few systems are used by national risk assessment authorities for use as quarantine tools, i.e. the United States Department of Agriculture (Lehtonen, 2001) and the Australian Quarantine Inspection Service (AQIS) (Pheloung et al. 1999). A common failing of these systems, is that they do not calculate the probability of the realisation of a predicted impact. Such predictions can be made only with large sample sizes of all individual cases in a class. They would require large databases of known plant histories. In New Zealand for example, where the total number of exotic species is known, the probability of a species naturalising - the first step to having an impact - has been calculated for all families and genera. These data can be incorporated in weed-risk assessments.

Weed-risk assessment systems have limitations, but their widespread use will encourage the international recognition of weed-risk assessment as a discipline.

Risk assessment systems operating at the border of an area to detect quarantine pests not yet established there, and those assessing established pests, have potential fundamental differences related to managing the risk. Managing newly-detected quarantine pests may mean simply prohibiting the entry of the species. There may be costs in doing so, related, for example, to loss of profits from the sale of a shipment of seed containing the pest, or a potential new agricultural crop foregone because of the potential pest status of the species. However, for any additional species undergoing weed-risk assessment there will be relatively little opportunity cost to the administrating authorities unless it involves monitoring a new pathway. In contrast, the outcome of a risk assessment identifying an entirely new pest within an area may require substantial management expenditure to extirpate the infestation. Unless additional funds are available to do this, they will need to be reallocated from elsewhere, usually from other pest-control efforts. Thus there is a need to determine the potential impacts relative to those of existing pests. Internal weed-risk assessment systems therefore have priority of management and expenditure as an important component of their process.

PROCEDURES FOR QUARANTINE WEED-RISK ASSESSMENT

The Australian WRA system

The Weed-Risk Assessment (WRA) system (Pheloung et al. 1999)^[2] was developed in Australia and is the most widely known and applied border weed-risk assessment system encompassing all plant groups. The central ‘argument’ is that if a species has had the opportunity to become a weed in another country, and it has done so, then it should be classed as a weed. Provided, that is, that the climate and environment are compatible with the new country (they are assumed to be so if there is no information). While this argument is essentially circular - it is a weed elsewhere, therefore it will become one here - a history of weediness elsewhere has reliably predicted weediness in several studies (Scott and Panetta, 1993; Reichard and Hamilton 1997; Williamson, 1998; Maillet and Lopez-Gacia, 2000). The WRA system was tested in Hawaii where it was found to be the most successful of those compared (Daehler and Carino, 2000). Until such time as weed-risk theory can contribute greater precision to the practice of weed-risk assessment, the WRA system is a suitable tool for use as a quarantine tool in developing countries. This argument is further reinforced by the imperatives of these countries to protect productive lands of one sort or another from weeds likely to arrive from developed countries (or via their neighbours) in trade goods, e.g. grass seed for sowing, or as a result of aid projects, e.g. re-vegetation schemes fostering legume shrubs. Because the actual or potential weediness of such species is commonly recognised in the developed countries where these goods or schemes originate, the central argument of weed history will be a powerful one in identifying quarantine pests in the developing countries. Ironically, the reverse is increasingly not the case. For example, in the last 20 years, 70 plant species have naturalized in New Zealand that are not known in the wild in any other country outside their native range. These were introduced mostly for urban horticulture and they come mostly from developing countries where their potential weediness has not been realised, e.g. Cotoneaster spp. from China.

The WRA produces a score for weediness and converts this into an entry recommendation for a specified taxon. It also satisfies several other requirements of an acceptable biosecurity assessment system (Hazard, 1988; Panetta, 1993). It can be calibrated and validated against a large number of taxa already present in the recipient country. These should represent the full spectrum of taxa likely to be encountered as imports into that country. It has some success in discriminating between weeds and non-weeds, such that the majority of weeds are not accepted, non-weeds are not rejected, and the proportion of taxa requiring further evaluation is kept to a minimum. The system also identifies which major land use system the taxon is likely to invade. This aids an economic evaluation of its potential impacts. In this respect it appears to be more successful at identifying agricultural weeds than environmental weeds. The WRA attempts to separate economic plants from those unlikely to have any economic benefit in a new country. However, where a taxon may have significant economic benefits, economic value should be scored in a transparently separate exercise and balanced against weediness in appropriate risk assessment evaluations (Walton and Parnell, 1996).

The WRA is not obligatorily computer based, but when operated on a computer it becomes interactive. This allows assessors to measure the influence of different attribute values on the scores generated. Finally, the system has proved to be cost effective to prospective importers and to border control authorities in Australia and New Zealand.

Permitted list approach

The process in which the WRA operates as described here is based on the concept of a Permitted List of plant species (or defined taxa). This system is used by quarantine authorities in Australia (Walton 2001) and New Zealand. The underlying concept is that if a species, or any subspecific taxon, with the potential to be a pest in an area is not on a list of taxa permitted to be in that area, then it will be prohibited until it has undergone pest categorization (Figure 5). Many countries, for example the United States of America, have lists of quarantine species, but they do not have permitted lists. They do not determine if every species new to a country should be on either list. One advantage of the involvement of a permitted list is that it automatically triggers a pest risk analysis in circumstances where there might otherwise not have been an analysis.

Figure 5. A flow chart for screening plant introductions incorporating the permitted list approach (from Panetta et al. 1994).

There are three stages to the prohibited list approach (Walton, 2001).

Tier 1. The first task is to identify a taxon correctly and determine whether it is listed as prohibited or permitted. This requires checking its species, genus, and family names, and whether there are synonyms. The next step is to check its presence in the country, either in cultivation only or in the wild. If a species is neither permitted nor prohibited, its spread within the area and whether or not it is under official control have to be determined. The kinds of information necessary to determine this given are in Appendix 1.

In some cases, the gathering of information on its status in the area may reveal both its presence, and sufficient evidence to justify an internal pest risk analysis. The outcome may result in the taxon being subjected to official control. Those with limited spread would automatically go on the prohibited list. Data should be entered into appropriate databases at all stages, e.g. Appendix 1.

Tier 2. Once a taxon has been classified as a potential quarantine pest, it undergoes a weed-risk assessment. This involves the WRA system, which recommends a species be rejected, accepted, or undergoes further evaluation. Rejected or accepted species are added to the prohibited and permitted lists respectively. Otherwise, further evaluation may be required if the importer wishes to proceed.

Tier 3. A classification of ‘accept’ or ‘reject’ cannot always be obtained from the second tier after the gathering of further information. It may then be appropriate to conduct field or glasshouse trials to further evaluate a species. These would need to be conducted in a secure environment from which the species could not escape via wind-blown seed, for example, or ‘hide’ in persistent seed banks. Whether the prospective importer considers this additional effort is warranted may depend on the potential gains from the plant species.

WRA operation

The WRA is based on the answers to forty-nine questions. These cover a range of weedy attributes in order to screen for taxa that are likely to become weeds of the environment and/or agriculture. The questions are divided into three sections producing identifiable scores that contribute to the total score (Appendix 3).

Biogeography (Section A): encompasses the documented distribution, climate preferences, history of cultivation, and weediness of a plant taxon elsewhere in the world, i.e. apart from the proposed recipient country. Weediness elsewhere is a good predictor of a taxon becoming a weed in new areas with similar environmental conditions. The question concerning the history of cultivation recognizes the important human component of propagule pressure. Such data are obviously never available for the proposed new country. Global distribution and climate preferences, where these are available, are used to predict a potential distribution in the recipient country.

Undesirable attributes (Section B): are characteristics such as toxic fruits and palatability to stock, or invasive behaviour, such as a climbing or smothering growth habit, or the ability to survive in dense shade.

Biology/ecology, (Section C): are those attributes that enable a taxon to reproduce, spread, and persist, such as whether the plant is wind dispersed or animal dispersed, and whether the seeds would survive passage through an animal’s gut.

Availability of information is often limited for new species, and the score system recognises that a minimum of information is required to provide a score and recommendation. The WRA system requires the answers to two questions in Section A, two in Section B, and six in Section C before it will give an evaluation and recommendation. The recommendation can be compared with the number of questions answered as an indication of its reliability. This improves as more questions are answered.

Answers to the questions provide a potential total score ranging from -14 (benign taxa) to 29 (maximum weediness) for each taxon. The total score is partitioned between answers to questions considered to relate primarily to agriculture, to the environment, or common to both (Appendix 3). The total scores are converted to one of the three possible recommendations by two critical score settings. The lower critical score, 0, separates acceptable taxa from those requiring evaluation, and the higher critical score, six, separates taxa requiring evaluation from those that should be rejected. Evaluation could mean either obtaining more data and re-running the model, or undertaking further investigations such as field trials.

The questions within the WRA would ideally be changed slightly for each significantly different area. They need to take into account regional differences in soils and climate. This was done when adopting the model for New Zealand. The critical settings to alter the likelihood of a species being accepted or rejected may be adjusted according to a different level of acceptable risk. This would require testing the new settings against a large number of species in the area.

All details on using the WRA are available on line at the Australian Quarantine Inspection Service site (http://www.affa.gov.au), or from the author.

PROCEDURES FOR INTERNAL WEED-RISK ASSESSMENT

Choice of a system

The objectives in characterizing potential quarantine pests are relatively straightforward because species are uniformly at their migration phase (Figure 3). In contrast, the objectives and information requirements of an internal weed-risk assessment system change as the species spreads. Decisions are made at an increasingly local level. Politics and economics may enter increasingly into the analysis as beneficiaries of control become identifiable and competition for resources between sectors increases (Panetta et al. 2001). These latter issues are not dealt with exhaustively here, but see Wainger and King (2001). The selection of internal weed-risk assessment systems must therefore consider the spread stage(s) of all pests being compared, impacts on the systems they affect, the likely benefits (and beneficiaries) of control efforts, and the quality of the available information. These factors vary widely within countries and between countries. There are numerous internal weed-risk systems in use. Often several are in use simultaneously within one country, even at a national level.

Table 1. The main systems used primarily for internal weed-risk assessment and prioritizing.

A sound and practicable system for application at a local level is that of Randall, (2000). More precise and ecologically defensible systems focus on specific biomes, such as the shrub lands of South Africa (Tucker and Richardson, 1995) or aquatic weeds (Champion and Clayton, 2001). Aquatic systems are almost a class on their own and authorities should be circumspect about any new wetland species. Overall, generalised systems are probably required first in most countries. Besides, the detailed understanding of the relationship between species attributes and the environment that make biome-specific systems effective are poorly understood for most biomes.

A single, widely applicable system cannot be recommended until the objectives of the internal weed-risk assessment system are determined. Countries establishing weed-risk assessment systems at the national level should ensure that the data collected are applicable to a range of spread stages and spatial scales of weed control. National resources for assessment made available by central government should be allocated where the long-term benefits will be greatest. This means first a border screening system, next a system for prioritizing species in the early establishment or expansion phases, and only then to species that have consolidated their spread (Figure 3). For this last group, the detailed system of Virtue et al. (2001) at a national level could be improved only with considerable effort, but would need to be adapted to the area concerned.

Factors to consider

This section describes the factors to be considered, the kind of information needed and likely to be available at different spread stages, and the process of determining what kind of weed-risk assessment system is required in specific sets of circumstances. Many of these considerations are relevant to a range of scales, from a single property to a whole country, and they are always limited by the total amount of resources for pest control within the particular area.

Weed history of congeners

Weed-risk assessment systems developed for the border, e.g. the WRA, usually consider the weediness of an assessed species’ relatives as indicators of potential pest status. This factor has seldom been considered as a risk component of systems designed for species at the earliest spread stages. Exceptions are where this association is implied by the group of species being ranked, e.g. pines (Tucker and Richardson, 1995). The behaviour of species’ relatives (e.g. family, genus) at several levels of taxonomic grouping may usefully be incorporated in internal weed-risk assessments. This applies particularly to species at their earliest invasion stages, where in the absence of much other information, it may contribute to a stated probability of a successful invasion. Weediness is concentrated within certain genera in some families and widely dispersed among many genera in others. Whether these probability estimates can be made at the family level or sub-family level depends on the size of the plant family and genera. Many genera are too small to give statistically reliable ratios.

Weed-led and site-led control

There is a tendency to control only those familiar species that have traditionally been controlled. This inertia demands a prioritising system that reallocates resources away from individual species that have become uncontrollable at a defined scale, to those that are potentially controllable at the same scale. Once attempts to extirpate a species or reduce it to below a defined population density over the entirety of a defined area (species-led control) have failed, then it should be controlled only in specific high value places within the area (site-led control). The concept was developed for conservation weeds in New Zealand (Williams 1997) and the application of this principle to crown owned conservation land is explained by Timmins and Owen (2001). It is relevant to a range of systems, including agricultural systems, and can be used in prioritising pests to be controlled at a national level.

Invasion stage

Some estimate of a species’ stage of infestation, or its surrogate, is required to determine the practicality of control. A clue to spread rate may be indicated by resident time within the area, if this is known, compared with the present distribution. However, unless a plant species is already listed as an unwanted organism for a specific area, often only range expansion (Figure 3) prompts the realisation of a new potential pest. By this stage, most newly recognised pests are well established and spreading. Where historical distributions are unknown, the simplest approach to the infestation stage that avoids the difficult interpretative question of spread rate, is to ask how well the species is established, i.e. the present number, size, and distribution of infestations. This also relates most closely to potential control of the species - those spreading rapidly will usually be well established with many loci, and will be more costly to control if widespread. These factors of range and expansion rate need to be considered within the context of the species’ regeneration time. A species does not necessarily spread to the most favourable or potentially damaging habitats first. Consideration needs to be given to the more favourable habitats and/or more vulnerable land uses it might encounter as it spreads.

Prerequisites for pest extirpation

Systems to determine whether the species is a candidate for weed-led or site-led control need to result in a ‘yes’ or ‘no’ outcome. Predictions of management outcomes may be more reliable than those concerning more complex ecosystem and economic interactions. The question, “Can it be killed?”, is easier to answer than “Will it affect biodiversity?”, or, “What economic impact will it have?” Even for well-established serious weeds, particularly in natural systems, the most meaningful trigger for their management may also be determined primarily by the cost and efficacy of control measures (Panetta and James, 1999).

Species extermination has seldom been achieved over areas of greater than a few hectares anywhere in the world. Irrespective of the area covered or the perceived impacts a pest may be having, there are critical questions for preventing, selecting and determining the level of management:

• Can all individuals of the species be located; and are practicable control measures available?

• Can all individuals be targeted within a defined time period dictated by the life cycle of the plant?

• Are the plant’s responses to control known?

• Are resources to treat new plants, at least as fast as they appear, available to undertake follow up work?

Those corresponding to ‘yes’ answers to these questions are higher priority for weed-led control than those with ‘no’ answers. Information on several aspects of a species is necessary before being able to answer these, and other, questions.

Biological attributes

A cautionary note can be made here. A wide range of biological attributes have been used in attempts to characterise weediness and prioritise species for control. The WRA also considers these factors too (Appendix 3). Most detailed biological attributes are only assumed to equate to invasiveness, even if this is taken to mean spread, as opposed to impacts. While some very general rules relating species attributes to invasiveness are emerging, these apply to only a few groups of plants in specific habitats. These often have particular disturbance regimes, including those determined by human activities. In many natural or semi-natural systems the relative importance of various dispersal modes is unknown. For example, until the relative number of potential wind-dispersed and fleshy-fruited woody plants potentially available to colonise lowland wooded vegetation in New Zealand are known, it is uncertain certain whether the fleshy-fruited syndrome per se has lead to the relative abundance of the latter group. Thus, attributes such as dispersal mode may be used with more integrity if used indirectly in determining management options, such as search frequency, rather than in attempting to predict invasion rates per se.

Ease of eradication

The intensity of weed control can be thought of as the product of the difficulty of killing an individual at the first event, including such factors as non-target effects, multiplied by the frequency of visits to re-treat the infestation. If a plant species has a history of weediness then it is likely to have certain identifiable attributes that make it so, e.g. persistent seed bank, and to have been the subject of control attempts elsewhere. These can help assess the difficulty of control in the new area. Where there is no history even of cultivation, or of weediness in its home region, ease of eradication must be inferred from attributes of the species or its congeners. These attributes could be classified in a variety of ways, but four seem to be critically important.

Time of detection

Detection of new infestations within one or two generations is important if the species is to be eradicated or contained within a small area. This means that the species has to be recognisable as a weed at an early stage. Species cryptic in the wild, such as a short grass or a vine with inconspicuous foliage, are likely to be confused with desirable species by the moderately informed observer. They are likely to spread before they are identified as weeds. They will be more difficult to control than conspicuous species.

Reproductive capacity

The amount of viable seed and vegetative reproduction may be critical components of invasion success. However, there is less certainty about the relative importance of these factors to invasion, or to ease of eradication. Species with persistent seed banks can be just as difficult to eradicate as those lacking seed banks but with vegetative reproduction. There is evidence that species with more than one reproductive system are more invasive, on average, than those with only one system. This is partly because the different strategies may enable the species to cross a wider range of barriers to invasion. As the population increases the barriers change. Species can therefore be ranked according to the number of reproductive strategies they have, without making assumptions about the relative importance of these strategies.

Dispersability

Potential for dispersal is obviously essential but the relative importance of different dispersal mechanisms should not be overstated in assessing weed-risk - most plants have a dispersal system of some sort. Wind-blown seeds are commonly blown long distances, and small seeds, more than large seeds, can be consumed and dispersed by wider-ranging animals. But seed size must always be considered within the context of the range of available dispersers and the potential dispersal mechanisms within the area. Passive dispersal by water, machinery, etc., and through contaminants in produce, may be more important than biological characteristics. Alternatively they may interact, e.g. small seeds are more likely to be carried by machinery than large seeds. In assessing risk from invasion, likely dispersal routes (waterways, farm tracks, randomly) also need to be considered, along with the suitability of the surrounding landscape to the species. In the early stages of invasion, an important contribution of dispersability to weediness is the ability to hide, as discussed above.

People

The attitudes of people towards plant species vary widely. While some species are considered a nuisance by everyone, others are useful to various sectors of the community. The outcomes of human activities involving these useful species, including recognised pests, can have an overriding influence on their spread. Attitudes to a species need to be considered, and as a rule, those species favoured for one reason or another will be the most difficult to extirpate. For a species to qualify for a programme aimed at extirpating it from a defined area, the probability of re-invasions from outside sources should be nil or very low. This is often not possible for those species grown commercially that are also pests. Here it may be possible to minimise the risk to land beyond the plantings by preventing the species regenerating within the defined area. This option may apply when a decision is made that the benefit emerging from a new species outweighs the risk, and pest risk management procedures are put in place while the species is grown commercially.

Climate matching

The utility of climate matching to weed-risk assessment changes as a pest spreads. At the earliest stages only the broadest match between source area (native and or adventive range) and potential range is required to consider the species a potential pest, because climate may or may not be the major barrier limiting spread. Many grasses originally from tropical Africa, for example, are now widespread in temperate regions. At the latter stages and on a local scale, climate matching is less important because the species has shown its potential to spread, and the ranking systems are required only to prioritise amongst known pests. Thus, climate matching between current and predicted range is most useful as a prioritising tool at the intermediate stages of spread, particularly when viewed on a country scale. Climate matching requires thorough distributional data within the areas being considered. On a country-by-country basis this requires comprehensive national databases. If data were collected on a regional basis, e.g. southern South America, perhaps under the umbrella of organisations such as COSAVE [The South Cone Plant Protection Committee (Comité de Sanidad Vegetal del Cono Sur - COSAVE)] it would have greater utility than on merely a national basis. Climate-matching is a specialist activity beyond the scope of this report, but the reader is referred to Kriticos and Randall (2001) for a summary of the applicability of several software packages to this topic in Austrialia.

Impact

Pests have economic, ecological, and or/social impacts, and the assessment method must define which of these it is attempting to assess. Reliable estimates of impact are possible only after the pest has begun to spread. Estimates of impact often involve a calculation of unit impact times a measure of the area covered. Several kinds of impact may be determined, or estimated, for one or more species, and incorporated into a scoring system (Virtue et al. 2001). Impacts may be determined on a very coarse scale and equated merely with presence, e.g. a species is present in ‘x’ number of land use systems in ‘y’ number of regions of a country. Much finer scales may be used, and extrapolated over the potential range of the species, e.g. a weed is sprayed at a cost of ‘w’ dollars on ‘x’ number of ha that would amount to ‘z’ dollars over its potential range.

Species impacts elsewhere may be applicable in the new area. In the absence of history, impact has to be estimated from the attributes of the species. These will differ with the land use likely to be affected. In agricultural systems, the impact of related pests may be relevant. For conservation land however, there is no universally applicable measure of impact. Parker et al. (1999) proposed parameters that might eventually be quantified as: I (overall impact)= R (range) × A (abundance) × E (impact per capita).

Species vary widely in the biomass at maturity that can be generated from a single propagule (seed) or ramett (piece of stem or root). An estimate of the biomass and extent of a species can contribute to a rudimentary estimate of impact. There may be evidence of its growth rate in terms of height and area covered. They are likely to range over tens of orders of magnitude, e.g. from a single grass plant 10 cm tall by 25 cm² (0.002 m³), to a typical perennial herb, 1 m² and 1 m tall (1 m³), to trees 10 m tall and with crowns 10 m diameter (1 000 m³). ‘E’ is likely to be related to the log of the volume of a single individual plant: 1, 10, 100, 1 000, 10 000. These data can be reduced to scores ranging from 1-5. Biomass as a surrogate measure of impact is probably modified by the species’ physical interaction with desirable vegetation. Information is generally available on whether the species co-occurs or replaces the desired vegetation. The long-term effects of weeds in either in the canopy or in a lower regenerative layer are mostly unknown. Intuition suggests replacing the vegetation canopy will, in the short term, displace more species, including invertebrates, than simply occupying a sub-canopy position. This generalization may not hold all species, e.g. for herbaceous ones increasing the effects of fire in natural systems. Similarly, impact is likely to be related to persistence at the site, whether this is via a single generation, or via successive generations.

Designing a scoring format

A scoring or ranking system for internal weed-risk assessments should embody all the principles of a quarantine assessment system, other than the requirement to meet international obligations (unless they were likely to impact on international trade). It should be designed to produce ranks, or other forms of classification. These should be based on the premise that weed management options are a function of the magnitude of risk, that the greatest benefit is achieved by controlling populations at the earliest stages of invasion, and that the score will be modified by the position of the manager to reduce the risk. It should identify the invasion stage(s) targeted and the ecosystems potentially affected. It should be no more comprehensive than is necessary to utilise the available information. Because weed control technology and resources available to manage the risk can change, these should be considered as separate modules and incorporated into the decision-process.

Ranking systems in use differ in the information required to operate them, and also in the structure of their internal rules. The simplest systems give numerical ratings to a set of criteria that may or may not be divided into sections, and are then summed. The questions may have equal or unequal value. The individual scores may, or may not, be modified by the answers to others (Pheloung et al. 1999). The subtotals from one section may be modified by other subtotals (Owen, 1998; Randall, 2000). Aspects of the species sometimes appear twice, as in its innate ability to be a pest and the ease with which it can be controlled (Hierbert, 1997). There may or may not be default scores where questions are not answered, and points may be deducted if answers to certain questions are negative (Pheloung et al. 1999). Others operate via hierarchical decision trees (Reichard and Hamilton, 1997). In a completely different approach, Tucker and Richardson (1995) used an expert system where a series of questions filtered out species of high or low risk before progressing to the next question.

An internal weed-risk assessment system should confirm, more or less, the existing ranking of weeds within an area for predefined spread stage, if this has been undertaken by experts (Hiebert, 1997; Pheloung et al. 1999), rather than produce a reordering of priority species. In other words, the outcomes from any new system must be intuitively sound if the system is to gain acceptance and be applied. This approach then captures all knowledge of the weeds of an area and formalises it within a system that is transparent, repeatable, and applicable to newly recognised species.

The development of these systems on spreadsheets allows the component scores to be adjusted and the effects on species rankings examined.

RESOURCES REQUIRED FOR WEED-RISK ASSESSMENT

A weed-risk assessment is primarily a ‘book exercise’ involving the collection of all available information about a potential pest, interpreting it, and making a decision. There are basically three kinds of information necessary, as well as access to the Internet for gathering it:

• The primary resource required to comply with the IPPC standards is a list of exotic plant species in the country, and whether they are under official control. To collect such information for both cultivated and wild species is a huge task and it has been achieved in very few countries. The first priority for most countries lacking this information is to restrict the list only to exotic plants under official control. Their names must be valid, and ideally there should be a voucher specimen. Once this is achieved, the task can be expanded to include wild exotic plants not under control. These tasks require the close cooperation of a range of national and regional organisations including herbaria, universities, libraries, and agriculture departments.

• Books are an important source of information on plants. Those undertaking weed-risk assessments should have access to all the relevant regional floras and the numerous books on weeds of the world or of regional weed lists. A basic collection of such books is given in Appendix 4. The two most important are Holm et al. (1979) describing the world’s worst weeds, and Huxley (1999) (see Appendix 4), which lists all the families, most of the genera, and many of the world’s species.

• Online information sources are now critical to weed-risk assessment. These are outlined in Appendix 4. They can be used for checking identification and plant names, determining weed history and noxious status elsewhere, and investigating control methods. E-mail is also a critical tool for communicating with numerous experts around the world who are familiar with these resources. In many cases they can also supply information on species not recorded in any form.

• Computer software is not critical for weed-risk assessment on a case-by-case basis but it is potentially a very important tool. It is necessary, however, if climate-matching tools are to be used. Apart from this purpose, the basic software requirements are database software and, if it was decided to use the Australian WRA system, a recent of version of EXCEL.

ACKNOWLEDGEMENT

Thanks to John Hedley and Bruce Trangmar for their helpful comments; to Anne Austin for editing; and to Jemma Callaghan for typesetting.

BIBLIOGRAPHY

CBD. 1993. Convention on Biological Diversity, Montreal. Secretariat of the United Nations Conference on Environment and Development.

Champion, P.D. & Clayton, J.S. 2001: A weed-risk assessment model for aquatic plant weeds in New Zealand. In Groves, R.H., Panetta, F.D., Virtue, J.G., eds. Weed risk assessment. CSIRO Publishing, Canberra. pp. 194-202.

Daehler, C.C. & Carrino, D.A. 2000: Predicting invasive plants: prospects for a general screening system based on current regional models. Biological Invasions 2: 93-102.

Esler, A.E., Liefting, L.W. & Champion, P.D. 1993. Biological success and weediness of the noxious plants of New Zealand. MAF Quality Management, Lynfield, Auckland, New Zealand.

FAO. 1996. Int. Standards for Phytosanitory Measures, Import Regulations: Guidelines for Pest Risk Analysis. Publication No. 2. Rome, Secretariat of the Int. Plant Convention of the Food and Agriculture Organization (FAO) of the United Nations.

FAO. 2001a. Int. Standards for Phytosanitory Measures, Glossary of phytosanitary terms. Publication No. 5. Rome, Secretariat of the Int. Plant Convention of the Food and Agriculture Organization (FAO) of the United Nations.

FAO. 2001b. Int. Standards for Phytosanitory Measures, Pest risk analysis for quarantine pests. Publication No. 11. Rome, Secretariat of the Int. Plant Convention of Food and Agriculture Organization (FAO) of the United Nations.

Groves, R.H., Panetta, F.D. & Virtue, J.G. 2001. Weed risk assessment. Melbourne, CSIRO Publishing.

Hazard, W.H.L. 1988: Introducing crop, pasture and ornamental species into Australia - the risk of introducing new weeds. Australian Plant Introduction Review 19: 19-36.

Hiebert, R. 1997. Prioritising invasive plants and planning for management. In Luken, J.O. & Thieret, J.W., eds. Assessment and management of plant invasions. New York, Springer-Verlag. pp. 11-19.

IPPC. 1997. New revised text of the Int. Plant Protection Convention, 1997. Rome, Secretariat of the Int. Plant Convention of Food and Agriculture Organization (FAO) of the United Nations.

Kriticos, D.J., Randall, R.P. 2001: A comparison of systems to analyze potential weed distributions. In Groves, R.H., Panetta, F.D. & Virtue, J.G., eds. Weed risk assessment. Canberra, CSIRO Publishing. pp. 61-82.

Lehtonen, P.P. 2001. Weed initiated pest risk assessment in the United States: guidelines for qualitative assessments. In Groves, R.H., Panetta, F.D. & Virtue, J.G., eds. Weed risk assessment. Canberra, Australia. CSIRO Publishing. pp. 117-123.

Maillet, J. & Lopez-Gacia, C. 2000: What criteria are relevant for predicting the invasive capacity of new agricultural weeds? The case of invasive American species in France. Weed Res. 40: 11-26.

Owen, S.J. 1998: Department of Conservation Strategic Plan for Managing Invasive Weeds. Department of Conservation, Wellington, New Zealand.

Panetta, F.D. 1993. A system of assessing proposed plant introductions for weed potential. Plant Protection Quarterly 8: 10-14.

Panetta, F.D. & James, R.F. 1999: Weed control thresholds: a useful concept in natural ecosystems. Plant Protection Quarterly 14: 68-76.

Panetta, F.D., Mackey, A.P., Virtue, J. & Groves, R.H. 2001: Weed risk assessment: core issues and future directions. In Groves, R.H., Panetta, F.D. & Virtue, J.G., eds. Weed risk assessment. Canberra, Australia. CSIRO Publishing. pp. 231-240.

Panetta, F.D., Pheloung, P.C., Lonsdale, M., Jacobs, S., Mulvaney, M. & Wright, W. 1994. Screening Plants for Weediness: a Procedure for assessing Species proposed for Importation into Australia. A report commissioned by the Australian Weeds Committee, Canberra, Australia (unpublished).

Parker, I.M., Simberloff, D., Lonsdale, W.M., Goodell, K., Wonham, M., Kareiva, P.M., Williamson, M.H., Von Holle, B., Moyle, P.B., Byers, J.E. & Goldwasser, L. 1999. Impact: toward a framework for understanding the ecological effects of invaders. Biological Invasions I: 3-19.

Pheloung, P. C., Williams, P. A. & Halloy, S. R. 1999: A weed-risk assessment model for use as a biosecurity tool evaluating plant introductions. J. of Environmental Management 57: 239-251.

Randall, R.P. 2000: Which are my worst weeds? A simple ranking system for prioritising weeds. Plant Protection Quarterly 15: 109-115.

Reichard, S. E. & Hamilton, C.W. 1997: Predicting invasions of woody plants introduced into North America. Conservation Biology 11: 193-203.

Scott, J.K. & Panetta, F.D. 1993: Predicting the Australian weed status of southern African plants. J. of Biogeography 20: 87-93.

Timmins, S.M. & Owen, S.J. 2001: Scary species, superlative sites: assessing weed risk in New Zealand’s protected natural areas. In Groves, R.H.; Panetta, F.D.; Virtue, J.G., eds. Weed risk assessment. Canberra, Australia. CSIRO Publishing. pp. 217-227.

Tucker, K.C., Richardson, D.M. 1995: An expert system for screening potentially invasive alien plants in South African fynbos. J. of Environmental Management 44: 309-338.

Virtue, J., Groves, R.H. & Panetta, F.D. 2001: Towards a system to determine the national significance of weeds in Australia. In Groves, R.H., Panetta, F.D. & Virtue, J.G., eds. Weed risk assessment. Canberra, Australia. CSIRO Publishing. pp. 124-150.

Wainger, L.A. & King, D.M. 2001: Priorities for weed risk assessment: using a landscape contex to assess indicators of function, services and values. In Groves, R.H., Panetta, F.D. & Virtue, J.G., eds. Weed risk assessment. Canberra, Australia. CSIRO Publishing. pp. 34-51.

Walton, C. 2001. Implementation of a permitted list approach to plant introductions to Australia. In Groves, R.H., Panetta, F.D. & Virtue, J.G., eds. Weed risk assessment. Canberra, Australia. CSIRO Publishing. pp. 93-99.

Walton, C. & Parnell, T. 1996. Weeds as quarantine pests. Proc. 11^th Australian Weeds Conference. Shepard, R.C.H., ed. Weed Science Society of Victoria Inc., Frankston, Australia. pp. 462-463.

Williams, P.A. 1997. Ecology and management of invasive weeds. Conservation Sciences Publication No. 7. Wellington, Department of Conservation. 67 pp.

Williamson, M. 1998. Measuring the impact of plant invaders in Britain. In Starfinger, U., Edwards, K., Kowarik, I. & Williamson, M., eds. Plant invasions: ecological mechanisms and human responses. Leiden, The Netherlands. Backhuys Publishers.

WTO. 1994: Agreement on the application of sanitary and phytosanitary measures. (also available at http://www.wto.org). World Trade Organization, Uruguay Round of Multinational Trade Negotiations. New York, World Trade Organization.