5. Data Analysis and Results
5.1.2 Current and potential distribution 5.1.3 Economic information 5.1.4 Environmental information
5.1.4.2 Number of threatened communities 5.1.4.3 Number of IBRA regions where a weed is present 5.1.4.4 Monoculture potential The Bureau of Rural Sciences and Complexia undertook statistical analysis of the results from the three reference panels, mapped current weed distributions and densities, and calculated the areas occupied by each weed. The Weed Science Group, Agriculture Western Australia generated the potential distribution maps from CLIMATE, and calculated the potential area occupied by each weed. The authors analysed the economic, environmental and social impacts data and developed the final model combining all data into a ranking of weed species.
The following is a summary of the methodology and analysis undertaken by the Bureau of Rural Sciences and Complexia. A full description is provided in appendix 7. Because there was insufficient overlap of weed species between the three panels, the reference panel information was treated as six separate analyses — invasiveness and impacts for each of the three panels. Some weed species were assessed by more than one reference panel, depending on the weed’s climatic and national distribution. Table 1 indicates the number of occurrences of weeds across the reference panels. Only two weeds were common to all panels (cabomba and Noogoora burr). Insufficient replication across the panels confounded the environmental and panel error sources, limiting further analysis. However, the invasiveness and impacts scores of a weed species would generally be expected to vary between panels, because of the effect that changed climatic conditions under a different geographical location would have on the traits exhibited by the weed. appendix 8 provides a table of the common names of weeds considered by each of the three reference panels.
Table 1. Number of occurrences of weeds between the three panels. The reference panel results were analysed using the proportional odds model with a logit link function and cumulative probabilities. The "don’t know" class was eliminated from the analysis — it comprised only 9 of 7084 responses — and the six rating classes were reduced to three because of the large number of zero responses. This is appropriate under the model assumptions. The three reference panels rated most of the weeds as medium to highly invasive. This was to be expected, given the way the weeds were selected and nominated by State and Territory agencies. Table 2 lists low and high invasiveness weeds as assessed by each panel. Mexican poppy, cabomba and mesquite were consistently ranked across two of the reference panels. A different response pattern was observed across the three reference panels in responses to question 2 concerning juvenile weed control tolerance. This was probably because the rating choices for this question were listed on the questionnaire in reverse order to those for the other questions. Overall, the invasiveness scores were medium to high, providing minimal discrimination between the weed species. To a large extent this is to be expected. The reference panels were assessing the most serious weed problems in Australia and it is ‘invasiveness’ that distinguishes the worst weeds.
Table 2. Invasiveness groupings of weeds between the three reference panels. The species placed in each impact category by the three reference panels showed only a small amount of overlap, indicating the strong influence of geographic location. Table 3 lists examples of weeds divided into three impact levels by the reference panels.
Table 3. Impact groupings of weeds between the three reference panels. Notes: The response to questions 9 and 11 were the most variable between the panels. Question 9 examines the effects weeds have on human and animal movement. Question 11 covers possible effects of weeds on animal health. The latter were generally considered to be low, except were they caused poisoning or injury in the case of large thorny weeds (mesquite). Overall, most weeds have minimal harmful effects. 5.1.2 Current and potential distributionThe current distribution data were summarised as the proportion of the continent that the weed affects. This is the ratio of the total presence, comprised of the densities (low, medium and high) to the total number of 0.5-degree grid cells for Australia. The current density distributions were limited, as density distribution data were available only from States and Territories that had nominated the particular weed species. Western Australia provided only absent/present data. The balance of the State and Territory distribution data were derived from Parsons and Cuthbertson (1992) and other sources. The current distribution maps are given in appendix 9. The climate-modeling program CLIMATE was used to predict the distribution of all weed species at four levels, ranging from core to low incidence. This analysis was based on the current distribution data, supplemented where necessary with overseas data. In addition, weed scientists were contacted for their opinions on climatic locations of particular weeds and for supplementary distribution data gained for rice grass and willows. Outliers representing atypical climates or microclimates for two weeds (arum lily and blue thunbergia) were removed from the analysis. Only the proportion of Australia covered by core and high-predicted densities (± 20% either side of the average climate determined by the current distribution of the weed) were used in the final analysis. These are the areas where the weed could be expected to be a significant problem. In the other areas the weeds would be expected to become a problem only where the microclimate was particularly favourable. The potential distributions of aquatic and riparian weeds were modeled without using the rainfall parameter, as these weeds have unlimited access to water. This extends the predicted area and therefore a method was required which established a more realistic distribution for these weeds. No one criterion or geographical information system (GIS) layer was available that adequately described the limits of these weeds. The compromise land-use screen described in section 3.2.2 was used as an overlay to reduce the potential area to permanent water bodies. The aquatic and riparian weeds assessed by the above method were:
Many other weeds (such as athel pine, blackberry, Parkinsonia and rubber vine) also prefer watercourses. However, their generally wide distribution away from water does not warrant them being classified as aquatic or riparian weeds. Rice grass grows in estuaries and an appropriate GIS layer was not available to calculate the potential distribution of the species. For that reason, the predicted distribution was restricted to the coastal 0.5-degree grid cells where there is a possibility of the weed growing. The analysis of potential distribution also provides a method for ordering the weeds, based on the proportion of Australia affected. The potential distribution maps (appendix 10) show the 0–20% prediction that represents the area where a significant weed problem could develop. Lack of data meant that no allowance could be made for soil type or other environmental factors not directly connected with the climate. Only limited allowance was made for physiological requirements such as dormancy, flower initiation, day length responses, exposure to wind, salt and plant competition and other factors; these factors are inferred from the weed’s distribution. The result is that climate models can provide, at best, only a crude approximation of potential distribution and which is almost certainly an over-estimate. Table 4 lists the top twenty weeds for current and potential distribution.
Table 4. The 20 most significant weeds listed in descending order for current and potential distribution. 5.1.3 Economic informationThe costs attributed to agricultural and forestry weeds ranged from $0 to more than $26,000,000. The largest value (blackberry) was an extreme outlier, being 3.5 times the value of the next nearest species (Paterson’s curse). Table 5 summarises estimated costs of the 35 agricultural and forestry weeds for which economic information was provided, broken down into four classes: less than or equal to $100,000; $100,001 to $1,000,000; $1,000,001 to $5,000,000; and greater than $5,000,000.
Table 5. Economic cost groupings of agricultural and forestry weeds. 5.1.4 Environmental informationThe numbers of threatened species and communities for each weed were converted to a proportion of the maximum numbers in each State and Territory. The data from States and Territories on maximum numbers of threatened species and communities are summarised in Table 6.
Table 6. Maximum number of threatened species and communities across States and Territories. Note: (1) Data derived from the Victorian Flora Information System. A generalised linear model and binomial error structure with a logit link were used to analyse the number of threatened species and communities environmental data. A more detailed explanation of the methodology, provided by the Bureau of Rural Sciences and Complexia, is given in appendix 7. The number of IBRA regions where a weed was present was converted to a proportion of the total number of IBRA regions. After analysis, the results of Question 7 relating to the reduction of desired vegetation were extracted from the reference panel data and used as the measure of monoculture potential.
The results of the analysis indicate that many weeds affect the status of threatened species, though there is a strong tendency for this to be State related. The weeds can be conveniently broken into three major groups: Other weeds either have little, if any, effect on the number of threatened species, or it might be that their impacts have yet to be assessed by State and Territory conservation agencies and other bodies. 5.1.4.2 Number of threatened communitiesThe results indicate that a number of weeds have significant effects on the status of threatened communities. There is again a strong tendency for this to be State related. These weeds can be conveniently broken into three major groups: Other weeds either have little, if any, effect on the number of threatened species, or it might be that their impacts have yet to be assessed by State and Territory conservation agencies and other bodies. 5.1.4.3 Number of IBRA regions where a weed is presentThere are eighty IBRA regions. Some weed species populate a significant number of these regions. These weeds can be conveniently broken into three major groups: This measure approximates the degree of plant competition (light, nutrients etc.) imposed on a plant community as a result of invasion by a weed. Table 7 lists three groupings of monoculture potential for the weeds tested.
Table 7. Monoculture potential groupings for all weeds assessed. Broadly, the reference panels identified two competitive mechanisms — for light and for water — on which to base overall scores. Thus, several types of weeds were rated as having high potential to form monocultures. Most weeds in this group, which includes the majority of aquatic weeds, form dense stands that can out-compete native vegetation for light. Bellyache bush, on the other hand, competes strongly for water in dry environments where light is usually not a limiting factor. This results in suppression of native vegetation through water stress. The low monoculture potential group is comprised mostly of herbs, with willows — a tree species — the sole exception. The reference panel discussions of social values for each of the weeds were recorded to identify positive and negative values for each species. This information was subjectively rated as nil, low, medium or high, depending on the number and severity of negative versus positive effects. Discussion and subjective analysis for these rating values included information on social impacts not canvassed in other criteria. No weed achieved a high rating, six were assessed as having a medium social value, 37 were rated as low and 28 scored nil. Table 8 lists the weeds by social value rating.
Table 8. Social value groupings for all weeds. To facilitate the combination of results following data analysis, each of the measures was rescaled to one, based on the highest value for that variable. The exceptions were:
This re-scaling provides up to ten variables indicating the significance of weeds within Australia:
These variables can be combined in a simple sum or, alternatively, a weighted sum of the ten possible measures, which may be calculated for each weed as: Scorei = Si Wj Pij(1) where Scorei is the score for weed i; Pij is the proportion for weed i and measure j (j = 1 to 10); and Wj is the weight or importance assigned to the measure. The data were analysed extensively to investigate how numerous weighting schemes affected the ranking of the weeds and the appropriateness of the model. It was observed that, for additive models, the composition of the first ten species varied marginally with only slight changes occurring in the ranked order. Multiplicative models, on the other hand, tended to have a marked effect on rankings. Greater manipulation of the data was also required, because of the occurrence of zero values for some variables. Weeds having values at either extreme of the range significantly affected the ranking. The use of a multiplicative model implies greater interaction between all of the variables than is actually the case. The general approach was to combine the variables using an additive model. The weightings used were selected to ensure that a criterion did not dominate solely because it was comprised of more variables than another. This was entirely consistent with the key characteristics sought for the criteria as outlined in the introduction. The final ranking was determined by the weighted sum of the four criteria. The final model used and the weightings applied are shown in Figure 3. Figure 3.
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
