Discussion

The scoring approach has several limitations:

  • The summation assumes that the variables are independent. This is unlikely to be the case for current and potential distribution, threatened species and communities, and IBRA regions.

  • The unweighted score (all Wj equal) assumes similarity and linearity of the measurement scales for all ten variables. Again this is unlikely to be the case, as can be seen from the economic information.

  • The previous limitations might be overcome by using differing weights. The assignment of weightings is somewhat arbitrary and the approach taken was to combine the variables into the four criteria according to the agreed framework, recognising the data limitations discussed previously.

  • There is no single criterion which defines when a score is sufficient to justify a weed being nationally significant. Any comparison or test is fraught with distributional dangers and inaccuracies.

  • The overall scoring/rating pattern may be very sensitive to small changes in the weightings and/or the proportions. Any analysis of the sensitivities would be difficult and would require the exploration of the interactions of the 10 variables, which are almost certainly correlated.

The four criteria consisted of data from a range of sources with varying reliability. The final model was devised taking this aspect into consideration in allocating weightings to each criterion and components of criteria. The invasiveness and impacts results were subjected to rigorous development and analysis, and have the highest weightings, reflecting the greater certainty of these sources of data.

The States and Territories nominated their most significant weeds, a process which resulted in a selective sample of invasive plants, or colloquially the worst of the worst weeds. Consequently, most of the scores fall into the higher ranges, resulting in small differences between species scores, with no obvious break points that may be used to segregate taxa.

For most species, the rankings should be seen as approximate rather than absolute. It might be more appropriate to view groups of weeds with some degree of similarity as clusters.

There are several wider potential limitations to the analysis that were beyond the control of the authors:

  • Distribution estimates would benefit from the inclusion of additional information such as soil type, land use and moisture regime that is not available at this time.

  • The analysis used by CLIMATE to estimate distribution is linear, whereas the response to climatic variables is unlikely to be a simple linear function. However, this limitation also applies to other predictive models such as BIOCLIM and ANUCLIM.

  • The data and information collected might have been enhanced if more time and resources had been available.

  • Not all States and Territories provided data as requested or for all weeds in their jurisdiction.

  • The number of weed species requiring assessment (71) meant that data were not always available from throughout Australia. This was particularly so for environmental data.

  • There was variability in the methods of recording and reporting some of the data by State and Territory agencies. This meant that data had to be standardised to provide some consistency at the national level and a degree of relativity between the different sources of State/Territory information. The consistency between State and Territory information sources is difficult to analyse, but was almost certainly quite variable.

  • The subjective analysis of the social values information.

  • The assumption made in these analyses has been that the information provided by the States and Territories is the "absolute best", especially for current distribution and density, environmental impacts, and to a lesser extent, the economic information. By this it is meant that no additional checking of, or removal of, possibly spurious information has been carried out for the current distribution and density information, although the environmental data have been more closely validated. The economic and some of the environmental information has required considerable standardisation to derive some consistency between States and Territories.

The information provided by the various sources varies considerably in quality and consistency. The panel data, based on the questionnaire, were the most consistent from a statistical point of view, followed by current distribution, environmental data, economic data and social values. The appropriateness of these measures is somewhat debatable, with current annual cost of control being a less than satisfactory measure for the economic impact of a weed. However, these were the most suitable measures identified where national data-sets existed or which could be assembled with the time and other resources available.

Of particular note were the limitations to any further analysis caused by the subjectivity of the social values data. This was reflected through the lower weighting and contribution that these data were accorded in the final model. However, in a fully validated model, social impacts would warrant a higher weighting. Examples include the potential threat of mimosa infestations to eco-tourism, and the allergic reactions caused by parthenium weed to local populations. On the positive side, beekeepers earn significant income from honey produced by bees foraging blackberry and Paterson’s curse.

The authors believe that the data sources and analyses undertaken to determine WONS were as comprehensive as they could be, and were sufficiently objective as to be clearly defensible. This satisfied the two prime principles which were to be observed in developing the process and undertaking the analysis for the National Weeds Strategy Executive Committee.

The ranking of a weed will have been affected by the quality of the data available. Therefore, species which may be poorly described by some or all jurisdictions could have scored lower than if all the data had been available for the analysis. Consequently, the taxa rank may well change as further knowledge is obtained about the lesser studied species.

Table of Contents Top