Feature

Weeding Out the Unwanted

A risk assessment model for screening plant imports into New Zealand may help sort out the weeds from the welcome plants.

Peter A. Williams

Weeds are a major cost to agriculture, forestry and the environment, yet many of these plant species have been introduced into New Zealand and Australia on purpose. This applies particularly to weeds of forestry and the environment which have generally not arrived simply as contaminants amongst other seeds.

A thorough study of this process was conducted in northern Australia where it was found that of the 2,033 species introduced with the hope of improving agriculture, only 5% came to be recommended as useful, whereas 13% became listed as weeds. Seventeen of the useful plants were also weeds, leaving only 1% that were useful without becoming weeds. These were far outnumbered by the 9% that were weeds with no recorded use.

These findings, together with the other major weed problems of that continent, were one of the major stimuli to the Australians developing a computer-based weed risk assessment model (WRA) for screening new species introductions.

Potential Weediness

In New Zealand, the Biosecurities Act requires that all plants proposed to be brought into New Zealand that are not prohibited entry and are not already here, must be evaluated for their potential weediness. This requires an objective, credible, relatively quick and publicly acceptable risk assessment system to screen large numbers of potential new entries.

The WRA system developed in Australia has been tested and modified for MAF Regulatory by scientists at Landcare Research and Crop & Food Research, to see if it would be suitable for use in New Zealand.

The WRA system is set up to run on a computer but can be operated on paper. It is incorporated into a computer-based environment to accelerate the process, help enforce objectivity and automatically maintain a database of assessed taxa.

A complete set of responses for each taxon is stored in a database and can be reviewed at any time and reassessed if details of the scoring system are changed.

A scoring system derived from the answers to 49 questions is used to assign each prospective import into one of three numerical classes which are then converted to a recommendation:

• accept
  -- the plant is allowed in
• reject
  -- the plant is not allowed in
• evaluate
  -- more information is required either to re-run the model, or from such methods as field trials, before a decision can be made

The questions on which the model is based are in three main sections. The first covers biogeography, which includes the documented distribution, climate preferences and weediness of a plant taxon in other parts of the world.

It is widely accepted that a weed in some part of the world has a good chance of being weedy in other areas with similar environmental conditions. The distribution and climate preferences should be used to predict a potential distribution in New Zealand.

Undesirable attributes are attributes related to noxious attributes such as toxic fruits and unpalatability, or invasive behaviour such as a climbing or smothering growth habit and the ability to survive in dense shade.

Biology/ecology are the attributes that contribute to the capacity of the taxon to reproduce, spread, and persist. Included here are questions on whether the plant is wind-dispersed or animal-dispersed, and whether the seeds would survive the passage through an animal's gut.

Because these questions often relate more to one sector or another, the model has some capacity to detect which sector is likely to be effected.

New Zealand Conditions

Several questions in the model were changed to fit New Zealand conditions. For example, the climates in many parts of Australia are quite different to those in New Zealand, and so questions relating to comparisons with the species country of origin, and how this compares with New Zealand, were worded differently.

In Australia, too, there are many seed-eating birds, particularly parrots, which can destroy seeds, and so there are more likely to be effective natural enemies of certain new plants in Australia than in New Zealand.

Good information is required to predict weediness and this must be obtained from throughout the world.

Libraries attached to herbarium collections of plants are the main source, but electronic resources of information are becoming increasingly accessible and useful.

The Internet has great potential in this respect, and there now exists a World Weed Database. These electronic forms provide a means of rapidly obtaining relevant unpublished or obscure information, particularly on taxa that have become weeds in natural ecosystems.

It is also necessary to have access to the most up-to-date information on the presence and distribution of the species, and an accurate idenfication. This last point is first prerequisite to any analysis of weediness.

The model was tested on some 200 plant species currently found in New Zealand, and these were scored as if they had not yet arrived in the country. This step is important, because we had to avoid using information available only in hindsight.

To verify the model, 11 people in various occupations, such as agriculturalists, conservationists and botanists, were asked to rank the species for weediness on a three-point scale: major weed, minor weed, and non-weed. The ranks were averaged and the species assigned to the three classes. These averaged rank scores were compared with the scores from the model.

Virtually all plants ranked as major weeds were rejected (98%), including all species currently classed as noxious plants in New Zealand, and the majority of minor weeds were either rejected (65%) or required further evaluation (27%). Most (64%) non-weeds were accepted.

The variance for the weediness scores from the respondents was high, which was to be expected, and related largely to perceived usefulness.

In contrast, a major benefit of the weed risk assessment system is that it is unbiased as to potential usefulness.

The fact that some useful plants may be potentially rejected in this process, and that many taxa will require evaluation on the basis of the information available, means that ancillary procedures should be in place to gather more information about a taxons biology and ecology, and to undertake cost benefit analysis once this information has been obtained.

The model has potential for use in New Zealand in combination with other systems for evaluating weediness and undertaking cost benefit analysis.

Peter A. Williams is with Manaaki Whenua Landcare Research in Nelson.