for peer review · for peer review 23 non-native plant species are recorded in the leb, with many...
TRANSCRIPT
For Peer Review
Priority threat management of non-native plants to
maintain ecosystem integrity across heterogeneous landscapes
Journal: Journal of Applied Ecology
Manuscript ID: JAPPL-2014-01063.R2
Manuscript Type: Standard Paper
Date Submitted by the Author: n/a
Complete List of Authors: Firn, Jennifer; CSIRO, Land and Water; Queensland University of
Technology, Faculty of Science and Engineering, School of Earth, Environmental and Biological Sciences Martin, Tara; CSIRO , Land and Water; ARC Centre of Excellence Decisions, the NERP Environmental Decisions Hub, Centre for Biodiversity and Conservation Science University of Queensland, Chades, Iadine; CSIRO, Land and Water; ARC Centre of Excellence Decisions, the NERP Environmental Decisions Hub, Centre for Biodiversity and Conservation Science University of Queensland, Walters, Belinda; CSIRO, Land and Water Hayes, John; Queensland University of Technology, Faculty of Science and Engineering, School of Earth, Environmental and Biological Sciences Nicol, Sam; CSIRO, Land and Water; ARC Centre of Excellence Decisions,
the NERP Environmental Decisions Hub, Centre for Biodiversity and Conservation Science University of Queensland, Carwardine, Josie; CSIRO, Land and Water; ARC Centre of Excellence Decisions, the NERP Environmental Decisions Hub, Centre for Biodiversity and Conservation Science University of Queensland,
Key-words: weed control, invasive plants, cost-effectiveness, threatened flora and fauna, biodiversity, cost-benefit analyses, stochastic dynamic programming
Journal of Applied Ecology
For Peer Review
Priority threat management of non-native plants to maintain ecosystem integrity across 1
heterogeneous landscapes 2
Jennifer Firn1,2, Tara G. Martin1,3, Iadine Chadès1,3, Belinda Walters1, John Hayes2, Sam 3
Nicol1,3, Josie Carwardine1,3 4
1 CSIRO Land and Water, Ecosciences Precinct Boggo Road, Brisbane, Australia 5
2Queensland University of Technology, School of Earth, Environmental and Biological 6
Sciences, Brisbane Australia 7
3ARC Centre of Excellence for Environmental Decisions, the NERP Environmental Decisions 8
Hub, Centre for Biodiversity & Conservation Science, University of Queensland, Brisbane Qld 9
4072, Australia 10
Summary: 11
1. Invasive non-native plants have negatively impacted on biodiversity and ecosystem 12
functions worldwide. Because of the large number of species, their wide distributions and 13
varying degrees of impact, we need a more effective method for prioritizing control 14
strategies for cost-effective investment across heterogeneous landscapes. 15
2. Here we develop a prioritization framework that synthesizes scientific data, elicits 16
knowledge from experts and stakeholders to identify control strategies, and appraises the 17
cost-effectiveness of strategies. Our objective was to identify the most cost-effective 18
strategies for reducing the total area dominated by high impact non-native plants in the 19
Lake Eyre Basin (LEB). 20
3. We use a case study of the ~120 million ha LEB that comprises some of the most 21
distinctive Australian landscapes, including Uluru-Kata Tjuta National Park. More than 240 22
Page 1 of 41 Journal of Applied Ecology
For Peer Review
non-native plant species are recorded in the LEB, with many predicted to spread, but there 23
are insufficient resources to control all species. 24
4. LEB experts identified 12 strategies to control, contain or eradicate non-native species 25
over the next 50 years. The total cost of the proposed LEB strategies was estimated at 26
AU$1.7 billion, an average of AU$34 million annually. Implementation of these strategies is 27
estimated to reduce non-native plant dominance by 17 million ha – there would be a 32% 28
reduction in the likely area dominated by non-native plants within 50 years if these 29
strategies were implemented. 30
5. The three most cost-effective strategies were controlling Parkinsonia aculeata, Ziziphus 31
mauritiana and Prosopis spp. These three strategies combined were estimated to cost only 32
0.01% of total cost of all the strategies, but would provide 20% of the total benefits. Over 50 33
years, cost-effective spending of AU$2.3 million could eradicate all non-native plant species 34
from the only threatened ecological community within the LEB, the Great Artesian Basin 35
discharge springs. 36
6. Synthesis and applications. Our framework, based on a case study of the ~120 million ha 37
Lake Eyre Basin in Australia, provides a rationale for financially efficient investment in non-38
native plant management and reveals combinations of strategies that are optimal for 39
different budgets. It also highlights knowledge gaps and incidental findings that could 40
improve effective management of non-native plants; for example, addressing the reliability 41
of species distribution data and prevalence of information sharing across states and regions. 42
43
Page 2 of 41Journal of Applied Ecology
For Peer Review
Introduction 44
The regular movement of people and commerce have breached biogeographic barriers, 45
initiating an unprecedented period of global species migration and homogenization (Elton, 46
1958; Crosby, 1986; Ceballos et al., 2010). Investment in non-native species management is 47
occurring worldwide in an effort to stem the negative impacts on biodiversity, ecosystem 48
function and agricultural production (Levine & D'Antonio, 2003; Sinden et al., 2004; 49
Pimental et al., 2005). The estimated annual expenditure on non-native plant control in 50
Australia is AU $1.5 billion (Sinden et al., 2004), and US $9.6 billion in the United States 51
(Pimental et al., 2005). It is often unclear which management strategies represent the best 52
investments because of the challenge of measuring the impact of non-native species and 53
prioritizing strategies accordingly (Vila et al., 2011). There is also uncertainty around the 54
amount of funding required to manage priority non-native species and how best to spend 55
these funds to minimize spread and establishment. 56
Despite the prolific spread and dominance of many non-native plant species, there is mixed 57
evidence that non-native plants have caused direct extinctions of native species (Davis et al., 58
2011). However, there is substantial evidence that non-native plants can transform 59
ecosystem processes, such as water and nutrient cycling, fire frequency, and sediment 60
transport (D'Antonio & Hobbie, 2005; MacDougall & Turkington, 2005; Simberloff, 2006). 61
Combined with other disturbances such as tree clearing, grazing, irrigation, high fire 62
frequency, soil disturbance or fertilization, non-native plant species are often better able to 63
profit than native species and become dominant (Firn et al., 2008; Davies et al., 2009). 64
Under these multiple stressors, native plant biodiversity often declines (MacDougall et al., 65
2013) and as a result, key ecosystem functions are altered (Vila et al., 2010; Vila et al., 66
Page 3 of 41 Journal of Applied Ecology
For Peer Review
2011). Once established, non-native species can change the structure of ecosystems, such as 67
the conversion of grassland to woodland (Figure 1). 68
In many regions, prompt management action is hampered by a lack of empirical data on the 69
impacts of threats and the responses of species and ecosystems to management strategies 70
to address threats. A growing body of research investigates methods for undertaking 71
conservation management appraisal and prioritization using the knowledge of experts and 72
stakeholders to complement formal scientific data (Burgman et al., 2011; Martin et al., 73
2012b). Expert information has been used to evaluate the cost-effectiveness of a range of 74
strategies to assist decision making for saving threatened species (Possingham et al., 2002; 75
Carwardine et al., 2012; Pannell et al., 2012; Chades et al., 2014; Firn et al., In press). Given 76
the urgency of many conservation issues and the difficulty of managing threats including the 77
invasion of non-native plants, evidence suggests that in many cases it is better to make 78
decisions using expert knowledge, rather than to take no action due to insufficient data 79
(Martin et al., 2012b). 80
These priority threat management approaches evaluate the alternative cost-effectiveness of 81
management strategies, where the expected benefits of each strategy (not measured in 82
dollar terms) are divided by the expected costs (Cullen et al., 2005). The potential benefits 83
of strategies can be measured as the improvement in species habitat protected (Carwardine 84
et al., 2008) or improvement in species persistence (Joseph et al., 2009; Carwardine et al., 85
2012). The costs are usually financial management costs and/or opportunity costs (Naidoo 86
et al., 2006). The expected benefits of the management strategies are the product of the 87
potential benefits multiplied by the feasibility of the management strategies, or the 88
Page 4 of 41Journal of Applied Ecology
For Peer Review
likelihood that the benefit will be achieved. Technical, social and political factors can all 89
influence the feasibility of a management strategy. 90
To date, such priority threat management approaches have not been used to assess the 91
management of threats posed by non-native plants on the integrity of native ecosystems. 92
This is particularly challenging because of the vast distributions of non-native plants and 93
aforementioned knowledge gaps making assessments difficult. Given the negative impacts 94
of non-native plants across large expanses of habitat there is much to be gained by ensuring 95
time and money are spent most efficiently and effectively. 96
Here we provide a rational and transparent framework for cost-effective investment in non-97
native plant species management across a heterogeneous landscape. Our objective is to 98
identify which control strategies are likely to be most cost-effective for reducing the total 99
area dominated by high impact non-native plants in each bioregion of the Lake Eyre Basin 100
(hereafter LEB), central Australia; where there are many non-native species (~240 species), 101
with varying impacts across the region, and limited resources make it impossible to invest in 102
managing them all. Our approach appraises cost-effectiveness of a set of possible 103
management strategies by drawing on empirical data and expert information to estimate 104
the expected benefits and costs of each strategy (Joseph et al., 2009; Carwardine et al., 105
2012; Pannell et al., 2012). While the strategies evaluated for controlling, containing and 106
eradicating non-native species are not new, we develop a unique approach for evaluating 107
their cost-effectiveness, and relative priority for reducing the impact of non-native plants in 108
the LEB over 50 years. 109
We specifically aim to: (i) develop a defensible non-native plant management prioritization 110
framework; (ii) apply our framework to assess a costed suite of management strategies 111
Page 5 of 41 Journal of Applied Ecology
For Peer Review
using a simple but novel approach to estimate biodiversity benefits; (iii) quantify the likely 112
area of plant invasion within geographical subunits of the LEB and the area that can be 113
feasibly managed for non-native species and (iv) appraise the most cost-effective 114
management strategies depending on budgets. 115
Materials and methods 116
Case study 117
The Lake Eyre Basin (LEB) covers approximately 120 million ha of arid and semi-arid central 118
Australia, spanning one sixth of the Australian continent (Figure 2) and multiple states: 119
Queensland, South Australia, New South Wales and the Northern Territory. The LEB is the 120
fifth largest internally draining system in the world and one of the least degraded systems in 121
Australia. LEB habitats include internationally recognized wetlands such as the Ramsar listed 122
Coongie Lakes, grasslands such as the Astrebla Downs National Park and deserts such as the 123
Simpson Desert National Park. The LEB Great Artesian Basin (GAB) discharge springs 124
wetlands (Figure S1 in Supporting Information) are listed as endangered under the 125
Australian Commonwealth Environmental Protection and Biodiversity Conservation Act 126
1999. These permanent springs in an otherwise semi-arid to arid landscape with highly 127
variable rainfall support at least 13 endemic plant species and 65 endemic fauna species 128
(Fensham et al., 2007). 129
More than 240 non-native plants are recorded in the LEB, including 20 Weeds of National 130
Significance (WONS; the federal government classification for the most significant weeds in 131
Australia) (Thorp & Lynch, 2000; Australian Weeds Committee, 2012; CSIRO and QUT, 2013). 132
The current distributions of seven of these WONS are predominantly within the LEB 133
Page 6 of 41Journal of Applied Ecology
For Peer Review
including: Prosopis spp. (mesquite complex: Prosopis pallida (single-stemmed), and Prosopis 134
glandulosa, Prosopis juliflora, Prosopis velutina (multi-stemmed)), Parkinsonia aculeata 135
parkinsonia, Tamarix aphylla athel pine, Opuntia spp. and Cylindropuntia spp. (cacti 136
grouping, more than 14 spp.), Cryptostegia grandiflora rubber vine, Jatropha gossypifolia 137
bellyache bush, and Vachellia nilotica prickly acacia (Figure 1). 138
In addition, a number of high impact non-native plants have limited distributions within the 139
LEB but have the potential to spread in the higher rainfall (semi-arid) regions, e.g. Ziziphus 140
mauritiana chinee apple, and Bryophyllum spp. (mother of millions grouping: Bryophyllum 141
delagoense, Bryophyllum houghtonii and Bryophyllum pinnatum). 142
Buffel grass Cenchrus ciliaris L. has been introduced across the LEB for pasture production 143
and is the most widely distributed non-native plant species across the LEB (Marshall et al., 144
2011). Buffel grass has a “dual impact” in the LEB, being of economic value to many graziers 145
and one of the most serious threats to rangeland biodiversity (Martin et al., 2006; Friedel et 146
al., 2009; Pavey & Nano, 2009; Grice et al., 2012). 147
Data collection 148
We collated existing information on the distribution of 243 non-native plants that are 149
recorded within the LEB from the scientific literature and the Atlas of Living Australia (Atlas 150
of Living Australia website). Further information required for the analysis was gathered from 151
stakeholders and experts (hereafter ‘participants’) at a three-day workshop and follow-up 152
consultations using a structured elicitation approach. Participants during and in follow-up 153
consultations after the workshop included resource managers from local natural resource 154
management boards, non-government organizations, Aboriginal Land Council Members, 155
Page 7 of 41 Journal of Applied Ecology
For Peer Review
land managers, indigenous rangers, park managers, graziers, scientists, state and territory 156
governments, and the Federal Department of Sustainability, Environment, Water, 157
Population and Communities, including several involved with the Lake Eyre Basin Scientific 158
Panel and Community Advisory Committee. Of the 33 experts contacted, 19 contributed in 159
some form, and 11 attended the workshop. Information on the three-day structured 160
elicitation process, biodiversity of concern and invasive plants recorded in the LEB were sent 161
to participants prior to the workshop and an independent facilitator ran the workshop. 162
The first step in the structured elicitation process was to identify a set of management 163
strategies for individual or groups of non-native species that have a significant impact in the 164
Basin. Twelve strategies were identified at the start of the workshop (Table S1). Participants 165
then estimated the potential benefit and cost of implementing each strategy. 166
i. Estimating benefits and costs of strategies 167
Participants provided the information required for the benefit metric by estimating the 168
expected proportion of invadable habitat in each bioregion that each non-native species (or 169
group) is likely to dominate (>30% coverage at a site) in 50 years: 170
• without implementation of any strategy; and 171
• with implementation of the strategy targeted to manage that non-native plant 172
species (or group). 173
For each of these scenarios participants gave their best guess, upper (most optimistic) and 174
lower (most pessimistic) bounds, and a level of confidence that the ‘true’ estimate lays 175
within this range (Martin et al., 2012b; McBride et al., 2012). Participants quantitatively 176
defined dominance and invadable habitat. Dominance was defined as a level of cover 177
Page 8 of 41Journal of Applied Ecology
For Peer Review
exceeding 30%. If a non-native plant was dominant at a site, participants agreed that the 178
site would be considered dramatically altered. The invadable habitat was assumed to be the 179
proportion of suitable habitat for the non-native species in each bioregion, which also 180
considers the likelihood of the success. 181
Participants agreed that the target for managing the threatened ecosystem GAB discharge 182
springs would be to remove all non-native plant species and prevent new introductions. The 183
benefits of the GAB discharge springs strategy was estimated for the entire LEB rather than 184
each bioregion. 185
Participants during the workshop and in follow-up consultations costed each non-native 186
plant control program action within each bioregion over 50 years, and estimated the annual 187
costs/ha of managing (eradicating, containing or controlling) each non-native plant at high 188
and low densities and provided realistic time frames for management over 50 years (e.g. 189
treatment every 2, 10, 20 or 25 years). In all cases the costs of undertaking each strategy i 190
by its component actions in each bioregion j were estimated by considering the costs of 191
previous and current management activities and spatial variants such as land tenure and 192
remoteness. The economic cost Cij was the cost in present day Australian dollars of activities 193
associated with strategy i in bioregion j over 50 years. 194
We used the best available data on the distributions of non-native plant species although 195
only occurrence data with coarse resolutions were available. The potential distributions of 196
most non-native plant species were obtained from the Weeds of National Significance 197
(WONS) program (Thorp & Lynch, 2000). Potential distribution maps for mother of millions 198
and chinee apple were assessed by overlaying 50-km2 squares (a similar approach was used 199
in the WONS mapping) over occurrence maps downloaded from the Atlas of Living Australia 200
Page 9 of 41 Journal of Applied Ecology
For Peer Review
(Atlas of Living Australia website), as it was assumed that the area surrounding an existing 201
population of a non-native plant species will have the highest likelihood of becoming 202
invaded. For buffel grass distribution, we used habitat suitability maps created for this 203
species using key indicative variables e.g. soil moisture, soil type, temperature, rainfall, 204
grazing intensity and fire frequency (Martin et al., 2012a). 205
Individual estimates by experts of strategy benefits in a bioregion were aggregated by 206
averaging. The total expected benefit, ��� of strategy i in bioregion j was defined by: 207
��� = ���∑ (��� �����)
����
�, eqn 1 208
where: 209
• ��� is the total area of invadable habitat (ha) in bioregion j for the non-native species 210
managed under strategy i 211
• ���� is the proportion of invadable area dominated by the non-native species if no 212
action is taken in bioregion j estimated by expert k over the time period (50 years) 213
• ���� is the proportion of invadable area dominated by the weed under strategy i in 214
bioregion j estimated by expert k over the time period (50 years) 215
• � is the number of experts who made an estimate for strategy i in bioregion j. 216
To analyse the sensitivity of our results to participant uncertainty, this process was repeated 217
using the upper and lower bound estimates for the proportion of invadable habitat 218
dominated by the non-native plant species with and without strategies being implemented. 219
Cost-effectiveness ranking approach 220
Page 10 of 41Journal of Applied Ecology
For Peer Review
Cost-effectiveness analysis provides a ranking of strategies based on an expected cost to 221
benefit ratio, where the benefit is not measured in terms of dollars (Levin & McEwan 2001). 222
To determine the total expected cost for a strategy over a bioregion, we summed the cost of 223
all actions required to implement each strategy in a bioregion. For strategies with actions 224
that were costed on a per hectare basis, we used maps of the current distribution of each 225
non-native species and information on the treatment area for each strategy to convert the 226
cost information into an average cost per year over 50 years for each bioregion. In all cases 227
one-off costs, such as building a fence, were counted once, while on-going annual costs, 228
such as maintaining the fence, were summed over 50 years using a rate of 2% per year to 229
calculate net present value, a low discount rate that is considered to be appropriate for 230
inter-generational public goods (Weitzman, 1998; Gollier & Weitzmann, 2010). 231
The cost-effectiveness,����, in ecological terms, of each strategy i in each bioregion j was 232
calculated by: 233
���� =���
��� eqn 2 234
where ��� is the total cost of strategy i in bioregion j. The strategies were then ranked by 235
their cost-effectiveness across all bioregions and within each bioregion. 236
Best spatial sets of strategies for a range of budgets 237
The cost-effectiveness approach ranks the most cost-effective strategies, but may not 238
identify the optimal set of strategies to implement for a given budget. Providing the optimal 239
combinations of strategies for a given budget can be useful information to aid managers to 240
negotiate budgets on the basis of expected outcomes. For example, we may discover that a 241
small increase in budget could lead to a sharp decrease of invaded area, an additional cost 242
that decision makers may find worthwhile to invest in. Alternatively, the optimal set of 243
Page 11 of 41 Journal of Applied Ecology
For Peer Review
solutions may reveal that investing half of the budget leads to almost the same benefits as 244
investing the entire budget. Identifying these thresholds of high or negligible return on 245
investment provides critical information to decision makers when allocating appropriate 246
budgets. 247
Finding the optimal combinations of strategies that minimizes the total area invaded at any 248
given budget across all bioregions requires solving a spatial combinatorial optimization 249
problem (0–1 knapsack problem (Bellman, 1957): 250
max∑ ���!���,�∈#,� subject to ∑ ���!��� ≤ �%&'() eqn 3 251
Where !�� is a binary decision variable that denotes whether (!��=1) or not (!��=0) a strategy 252
i in bioregion j is included in the optimal set of strategies. A vector * ∈ {!,,,, !,,-, … , !|#|,|�|} 253
represents a combination of selected strategies. S represents the set of strategies listed in 254
Table S1 and N the number of bioregions. ��� identifies the benefits of implemented 255
strategy i in bioregion j. We used dynamic programming to find the optimal set of decisions 256
for budgets ranging from $0 to $2.160 million (Bellman 1957). Algorithm details are 257
presented in Text S1. 258
Results 259
Workshop participants defined 12 management strategies (Table S1) focussing on 10 non-260
native plant species considered to significantly impact biodiversity in the LEB now or in the 261
future (Table 1). With the exception of buffel grass, mother of millions and chinee apple, all 262
species are classified as WONS (Thorp & Lynch 2000). Each strategy consisted of one or 263
more supporting actions, many of which were spatially linked to IBRA (Interim 264
Biogeographical Regionalisation of Australia) bioregions. 265
Page 12 of 41Journal of Applied Ecology
For Peer Review
The first strategy was an overarching recommendation for improved mapping, information 266
sharing, education and extension efforts in order to facilitate the more specific non-native 267
plant management strategies. This LEB prevention and monitoring strategy was estimated 268
to cost $17.5 million in total over 50 years. Strategy 12 involved the investment of $2.3 269
million over the next 50 years to eradicate all non-native plant species from the GAB 270
discharge springs, which was estimated at the Basin level. We were not able to compare 271
quantitatively strategies 1 and 12 with strategies 2–11 (Table S1). However when predicting 272
the expected benefits for reducing the dominance of the non-native plants over 50 years, 273
experts assumed that strategy 1 would be implemented. 274
The total cost of implementing all 12 non-native plant management strategies over the next 275
50 years was estimated at $34 million per year (Total = $1.7 billion; Table S1). 276
Implementation of these strategies was estimated to result in a reduction of non-native 277
plant dominance by 17 million ha (a potential 32% reduction), roughly 14% of the LEB (Table 278
1). To put this in perspective, on average, implementing all strategies would cost 279
approximately $100 per hectare for the 50-year period, or $2 per hectare per year. If only 280
targeting WONS, the total cost was estimated to be $2.3 million per year for 50 years (Total 281
= $113 million). The strategy for controlling, eradicating and containing buffel grass was the 282
most expensive representing over 90% of the estimated costs, requiring >$30 million 283
annually for 50 years (Total = $1.5 billion, Figure 3). This strategy would account for 0.06% of 284
the total benefits with an estimated reduction in invaded extent of 1 million ha. 285
Amongst the 10 species-specific management strategies, the top five most cost-effective for 286
the entire LEB were the management of: 1) parkinsonia, 2) chinee apple, 3) mesquite, 4) 287
rubber vine and 5) bellyache bush. These strategies combined would cost approximately 288
Page 13 of 41 Journal of Applied Ecology
For Peer Review
$254 000 per year and are estimated to achieve 30% of the total benefits (Table 1, Figure 3). 289
An additional budget of 1 million per year could see avoidance of 16 million ha (94% of the 290
estimated benefits) of land dominated by prickly acacia, mother of millions and Athel pine 291
(Figure 3). Benefit estimates by the experts varied depending on the strategy with standard 292
deviations being more than half the average best guess estimate for three strategies: chinee 293
apple, athel pine and cacti (Table 1). 294
The cost-effectiveness rankings remained the same for all 10 of the basin-wide strategies 295
when calculated with the best guess and upper estimates for expected benefits (Table 1), 296
but did vary with the lower estimates. Using the lower estimates for benefits, parkinsonia 297
was the most cost-effective strategy, mesquite the second most effective, followed by 298
chinee apple, rubbervine and bellyache bush. Mother of millions changed ranks with prickly 299
acacia so that the two species became the sixth and seventh most cost-effective strategies 300
respectively. 301
The top five most cost-effective strategies within separate bioregions (Figure 2) were the 302
management of (Table 2 & S2): 1) parkinsonia in the Channel Country, 2) parkinsonia in the 303
Desert Uplands, 3) mesquite in the Mitchell Grass Downs, 4) parkinsonia in the Mitchell 304
Grass Downs, and 5) mother of millions in the Desert Uplands. 305
When considering implementing strategies at the bioregional level, we found the best 306
combinations of strategies for all budgets by solving the knapsack problem (see Materials 307
and methods). Because we are seeking to maximize the impact of management and 308
minimize our spending at the same time, the solution is a Pareto frontier where every point 309
of the curve represents an optimal solution. In our case, the Pareto frontier exhibits four 310
distinct thresholds (A, B, C and D, Figure 4). When the budget was below $513 000 per year 311
Page 14 of 41Journal of Applied Ecology
For Peer Review
the total invaded area could reach more than 10 million ha across the LEB (Figure 4A). The 312
benefit of managing non-native plants could be doubled if the yearly budget was increased 313
by about $300 000 per year allowing the implementation of strategy 'S11. prickly acacia' in 314
the Mitchell Grass Downs bioregion (Figures 4 & 5B). Invasion of most of the potential area 315
could be avoided for a yearly budget of about $1.35 million (Figures 4 & 5C). 316
Discussion 317
In the LEB, like many regions of the world, there are more non-native plant species than 318
resources to manage them, and a significant amount of knowledge concerning the location 319
and best control strategies for managing non-native plants is held in the experience and 320
knowledge of natural resource managers (Bart, 2006). Previous attempts at non-native plant 321
species prioritization such as the Australian Weeds of National Significance (Australian 322
Weeds Committee, 2012) were based on the potential ecological and agricultural impacts. 323
The results of our approach, which appraises the benefits and costs of management of these 324
high impact species using a structured elicitation process, provide a prospectus for guiding 325
further investment in non-native species management and in improving information 326
availability and sharing across governmental boundaries. 327
Our analyses suggest that management of several WONS such as parkinsonia and mesquite 328
are likely to be more cost-effective investments than others in large parts of the LEB. 329
Management of parkinsonia was estimated as both relatively cheap and to have high 330
benefits, so control of parkinsonia was consistently ranked as the most cost-effective 331
strategy across the LEB. Other non-native plants will be more costly to manage, but may still 332
be cost-effective to control. Implementation of all non-native plant species was estimated to 333
come at a modest cost ($2 per ha per year on average), but not all strategies are equally 334
Page 15 of 41 Journal of Applied Ecology
For Peer Review
useful, particularly when the budget is limited. We show that there are thresholds in the 335
effectiveness of investment, so that it is important to ensure that the budget allocated to 336
the LEB is realistic to achieve the desired level of impact. While there are cost-effective 337
strategies that can be implemented for less than $513 000 per year, investing less than this 338
amount means that at least 10 million ha of the LEB remains vulnerable to invasion. Our 339
results suggest $900 000 per year could make a substantial improvement on the total area 340
potentially invaded by non-native plants in the LEB by implementing the most effective 341
strategies (control of prickly acacia, Figures 4 &5). Costs per year may appear low given the 342
large spatial extent of the LEB, but strategies were not recommended to be applied to all 343
areas of the LEB, and only where non-native plant control was considered feasible, 344
reasonable or a high priority. 345
Buffel grass management was orders of magnitude more expensive than the management 346
of all other species considered, because of its wide distribution and its high value for 347
rangelands within the LEB. The high costs and low cost-effectiveness ranking for buffel grass 348
control may motivate research into more effective management strategies for buffel grass 349
and also strongly suggest that key sites should be prioritized for protection from buffel grass 350
dominance (e.g. sites of high cultural value such as Uluru National Park (Grechi et al., 2014)). 351
The framework we develop here could be applied to find the most ecologically cost-352
effective solutions for non-native plant control in any management area where multiple 353
stakeholders and landowners reside. A key characteristic of the framework is flexibility, as it 354
can be tailored to a region’s geographical, biological and social characteristics, and the 355
strategies assessed can be updated to suit different perspectives or to account for updates 356
and changes in situations and information. The cost-effectiveness analyses are completed in 357
Page 16 of 41Journal of Applied Ecology
For Peer Review
a spreadsheet that is provided to the managers of a region following completion of the 358
study and can be updated when necessary. There are many uncertainties in future 359
conditions for undertaking non-native plant species control and eradication in the LEB and 360
many other landscapes, such as the consequences of climate change and future 361
developments not considered in the analysis, which may compound existing threats. A 362
precautionary approach suggests that we should increase investment early, then monitor 363
and review the effectiveness of actions and be aware of emerging threats. A flexible 364
approach, like the one presented here, also means that identified priorities can be 365
integrated with existing and future priority setting approaches. 366
We also make some assumptions with our priority threat management approach. The 367
expert-elicited strategies were designed to have an equal chance of being implemented, 368
provided resources were available. Participants chose strategies that were considered 369
feasible if resources are available, with a few exceptions: for example an organic farmer may 370
be reluctant to spray a non-native plant and lose their certification. In these cases, cost-371
effectiveness was likely overestimated. Estimates of the costs and benefits for the strategies 372
may prove to be higher or lower than predicted because they were based on expert and 373
local knowledge. Changes to these estimates could affect priority rankings. This is why the 374
flexibility of our approach is important – policymakers and land managers can update 375
spreadsheets when more information becomes available or situations change. 376
We also assumed that dominance of habitat by one non-native plant has an equal impact to 377
dominance by another non-native plant and that a cover of greater than 30% was a 378
threshold indicating a species has a biodiversity impact. In reality a variable range of impacts 379
would occur prior to and beyond this binary threshold, depending upon non-native plant 380
Page 17 of 41 Journal of Applied Ecology
For Peer Review
type, ecosystem type, existing dominant and threatened species, and other threats. Our 381
approach could be extended to incorporate differential weights to quantify estimates of 382
higher impacts for some non-native plants over others, which could vary depending on 383
characteristics of each bioregion. 384
Our analysis does not include how potential interactions between strategies will change the 385
expected benefits and costs of strategies, although non-native plant management could be 386
carried out for more than one non-native plant at a time. In these cases the cost-387
effectiveness of carrying out management strategies could be underestimated. 388
Although we used the best data available on the current and potential distributions of non-389
native plants in each of the LEB bioregions, it was recognized by all participants that 390
knowledge of the distributions of non-native plant species is patchy and unreliable, 391
evidenced by the recommendation of strategy 1 that included the centralization of 392
mapping, monitoring and surveillance and information sharing for non-native species 393
incursions and extents. 394
Conclusion 395
A major challenge facing managers when controlling non-native plants is that there are 396
numerous species, each with varying levels of impact. Additionally, the values placed on 397
impacts depend on the objectives of management and the resources available to control or 398
eradicate them. Therefore it is critical to evaluate systematically which non-native species, 399
management actions, and locations should be invested in to provide the greatest benefits 400
for ecosystem integrity per unit cost. 401
Our approach is applicable to non-native species management across large landscapes 402
worldwide because it is, systematic, knowledge-based, and can be easily updated as 403
Page 18 of 41Journal of Applied Ecology
For Peer Review
improved information on the costs and expected benefits for invasive plant management 404
become available. 405
Acknowledgements 406
We thank the broad range of stakeholders (policymakers, managers, scientists, community 407
representatives) who generously shared their time and expertise at a workshop and in 408
follow-up consultations. This project was funded by DSEWPaC’s National Environmental 409
Research Program, and a CSIRO Julius Career Award (IC). We thank R. Ponce and H. Murphy 410
for providing valuable comments on earlier version of this manuscript. 411
Data Accessibility 412
Data is archived with CSIRO public data access portal. Please follow this link: 413
https://data.csiro.au/dap/landingpage?pid=csiro:13800. 414
References 415
Atlas of Living Australia website Available at: http://www.ala.org.au (accessed June 2013). 416
Australian Weeds Committee (2012) Weeds of National Significance. In. Department of Agriculture, 417
Fisheries and Forestry, Canberra ACT. 418
Bart, D. (2006) Integrating local ecological knowledge and manipulative experiments to find the 419
causes of environmental change. Frontiers in Ecology and Environment, 4, 541-546. 420
Bellman, R.E. (1957) Dynamic Programming. Princeton University Press, Princeton, New Jersey. 421
Burgman, M., Carr, A., Godden, L., Gregory, R., McBride, M., Flander, L. & Maguire, L. (2011) 422
Redefining expertise and improving ecological judgment. Conservation Letters, 4, 81-87. 423
Carwardine, J., O'Connor, T., Legge, S., Mackey, B., Possingham, H.P. & Martin, T.G. (2012) 424
Prioritizing threat management for biodiversity conservation. Conservation Letters, 5, 196-425
204. 426
Page 19 of 41 Journal of Applied Ecology
For Peer Review
Carwardine, J., Wilson, K.A., Ceballos, G., Ehrlich, P.R., Naidoo, T., Iwamura, S., Hajkowicz, S. & 427
Possingham, H.P. (2008) Cost-effective priorities for global mammal conservation. PNAS, 428
105, 11446-11450. 429
Ceballos, G., Davidson, A., List, R., Pacheco, J., Manzano-Fischer, P., Santos-Barrera, G. & Cruzado, J. 430
(2010) Rapid decline of a grassland system and its ecological and conservation implications. 431
Plos One, 5, e8562 432
. 433
Chades, I., Nicol, S.J., van Leeuwen, S., Walters, B., Firn, J., Reeson, A. & Martin, T.G. (2014) 434
Complementary threat management priorities save more species. Conservation Biology, 29, 435
525-536. 436
Crosby, A.W. (1986) Ecological Imperialism: The Biological Expansion of Europe, 900-1900. 437
Cambridge University Press, Cambridge. 438
CSIRO and QUT (2013) Lake Eyre Basin Invasive Plants Database. In, Brisbane. 439
Cullen, R., Hughey, K.F.D., Fairburn, G. & Moran, E. (2005) Economic analyses to aid nature 440
conservation decision making. Oryx, 39, 327-334. 441
D'Antonio, C.M. & Hobbie, S.E. (2005) Plant species effects on ecosystem processes: insights from 442
invasive species. Species invasions (ed. by D.F. Sax and J.J. Stachowicz). Freeman, New York. 443
Davies, K.W., Svejcar, T.J. & Bates, J.D. (2009) Interaction of historical and nonhistorical disturbances 444
maintains native plant communities. Ecological Applications, 19, 1536-1545. 445
Davis, M.A., Chew, M.K., Hobbs, R.J., Lugo, A.E., Ewel, J.J., Vermeij, G.J., Brown, J.H., Rosenzweig, 446
M.L., Gardener, M.R., Carroll, S.P., Thompson, K., Pickett, S.T.A., Stromberg, J.C., Tredici, 447
P.D., Suding, K.N., Ehrenfeld, J.G., Grime, J.P., Mascaro, J. & Briggs, J.C. (2011) Don't judge 448
species on their origins. Nature, 474, 153-154. 449
Elton, C.S. (1958) The Ecology of Invasions by Animals and Plants. Methuen and CO Ltd, London. 450
Fensham, R., Ponder, W. & Fairfax, R. (2007) Recovery plan for the community of native species 451
dependent on natural discharge of groundwater from the Great Artesian Basin. In: Report to 452
Page 20 of 41Journal of Applied Ecology
For Peer Review
Department of the Environment, Water, Heritage and the Arts. Queensland Parks and 453
Wildlife Sevices, Canberra. 454
Firn, J., Rout, T., Possingham, H.P. & Buckley, Y.M. (2008) Managing beyond the invader: 455
manipulating disturbance of natives simplifies control efforts. Journal of Applied Ecology, 45, 456
1143-1151. 457
Firn, J., Maggini, R., Chades, I., Nicol, S., Walters, B., Reeson, A., Martin, T.G., Possingham, H.P., 458
Pichancourt, J.-B., Ponce-Reyes, R. & Carwardine, J. (In press) Priority threat management of 459
invasive animals to protect biodiversity under climate change. Global Change Biology, 460
Friedel, M., Marshall, N., van Klinken, R.D. & Grice, T. (2009) Quantifying costs and benefits of buffel 461
grass. In. Australian Government Land and Water Australia 462
Gollier, C. & Weitzmann, M.L. (2010) How should the distant future be discounted when discount 463
rates are uncertain? Economic Letters, 107, 350-353. 464
Grechi, I., Chades, I., Buckley, Y.M., Friedel, M.H., Grice, A.C., Possingham, H.P., van Klinken, R.D. & 465
Martin, T.G. (2014) A decision framework for management of conflicting production and 466
biodiversity goals for a commercially valuable invasive species. Agricultural Systems, 125, 1-467
11. 468
Grice, A.C., Friedel, M.H., Marshal, N.A. & van Klinken, R.D. (2012) Tackling contentious invasive 469
plant species: a case study of buffel grass in Australia. Environmental Management, 49, 285-470
294. 471
Joseph, L.N., Maloney, R.F. & Possingham, H.P. (2009) Optimal allocation of resources among 472
threatened species: a Project Prioritization Protocol. Conservation Biology, 23, 328-338. 473
Levine, J.M. & D'Antonio, C.M. (2003) Forecasting biological invasions with increasing international 474
trade. Conservation Biology, 17, 322-326. 475
MacDougall, A.S. & Turkington, R. (2005) Are invasive species the drivers or passengers of change in 476
degraded ecosystems? Ecology, 86, 42-55. 477
Page 21 of 41 Journal of Applied Ecology
For Peer Review
MacDougall, A.S., McCann, K.S., Gellner, G. & Turkington, R. (2013) Diversity loss with persistent 478
human disturbance increases vulnerability to ecosystem collapse. Nature, 494, 86-89. 479
Marshall, N.A., Friedel, M.H., van Klinken, R.D. & Grice, A.C. (2011) Considering the social dimension 480
of invasive species the case of buffel grass. Environmental Science and Policy, 14, 327-338. 481
Martin, T.G., Campbell, S. & Grounds, S. (2006) Weeds of Australian rangelands. Rangeland Journal, 482
28, 3-26. 483
Martin, T.G., Murphy, H. & Liedloff, A. (2012a) Invasive species and climate change: a framework for 484
predicting species distribution when data are scarce. In: 485
http://www.csiro.au/en/Organisation-Structure/Flagships/Climate-Adaptation-Flagship/CAF-486
working-papers.aspx (ed. C.C.a.F.W.P.N. 13g) 487
Martin, T.G., Burgman, M.A., Fidler, F., Kuhnet, P.M., Low-Choy, S., McBride, M. & Mengersen, K. 488
(2012b) Eliciting expert knowledge in conservation science. Conservation Biology, 26, 29-38. 489
McBride, M., Gameti, S.T., Szabo, J.K., Burbidge, A.H., Butchart, S.H.M., Christidis, L., Dutson, G., 490
Ford, H.A., Loyn, R.H., Watson, D.M. & Burgman, M.A. (2012) Structured elicitation of expert 491
judgements for threatened species assessment: a case study on a continental scale using 492
email. Methods in Ecology and Evolution, 3, 906-920. 493
Naidoo, R.A., Balmford, R., Ferraro, P.J., Polasky, S. & Ricketts, T.H. (2006) Integrating economic costs 494
into conservation planning. Trends in Ecology and Evolution, 21, 681-687. 495
Pannell, D.J., Roberts, A.M., Park, G., Alexander, J., Curatolo, A. & Marsh, S.P. (2012) Inegrated 496
assessment of public investment in land-use change to protect environmental assets in 497
Australia. Land Use Policy, 29, 377-387. 498
Pavey, C.R. & Nano, C.E.M. (2009) Bird assemblages of arid Australia: Vegetation patterns have a 499
greater effect than disturbance and resource pulses. Journal of Arid Environments, 73, 634-500
642. 501
Page 22 of 41Journal of Applied Ecology
For Peer Review
Pimental, D., Zuniga, R. & Morrison, D. (2005) Update on the environmental and economic costs 502
associated with alien-invasive species in the United States Ecological Economics, 52, 273-503
288. 504
Possingham, H.P., Ryan, S., Baxter, J. & Morton, S. (2002) Setting biodiversity priorities. In: Sustaining 505
our natural systems and biodiveristy 506
Simberloff, D. (2006) Invasional meltdown 6 years later: important phenomenon, unfortunate 507
metaphor or both? Ecology Letters, 9, 912-919. 508
Sinden, J., Jones, R., Hester, S., Odom, D., Kalisch, C., James, R. & Cacho, O. (2004) The economic 509
impact of weeds in Australia: summary. CRC for Australian Weed Management, Technical 510
Series #8, 1-4. 511
Thorp, J. & Lynch, R. (2000) The Determination of Weeds of National Significance. In. Weeds Strategy 512
Executive Committee, Launceston. 513
Vila, M., Espinar, J.L., Hejda, M., Hulme, P.E., Jarosik, V., Maron, J.L., Pergl, J., Schaffner, U., Sun, Y. & 514
Pysek, P. (2011) Ecological impacts of invasive alien plants: a meta-analysis of their effects 515
on species communities and ecosystems. Ecology Letters, 14, 702-708. 516
Vila, M., Basnou, C., Pysek, P., Josefsson, M., Genovesi, P., Gollasch, S., Nentwig, W., Olenin, S., 517
Roques, A., Roy, D., Hulme, P.E. & Daisi partners (2010) Frontiers in Ecology and 518
Environment, 8, 135-144. 519
Weitzman, M.L. (1998) Why the far-distant future should be discounted at its lowest possible rate. 520
Jounal of Environmental Economic Management, 36, 201-208. 521
Supporting Information 522
Additional Supporting Information may be found in the online version of this article: 523
Table S1: Estimates of the cost of actions that make up the specific non-native plant species 524
strategies. 525
Table S2: Appraisal of key non-native plant management strategy in each of the LEB 526
bioregions. 527
Page 23 of 41 Journal of Applied Ecology
For Peer Review
Figure S1: Map of known locations of the GAB discharge springs within each of the LEB 528
bioregions. 529
Text S1. Dynamic programming algorithm to find the optimal spatial sets of strategies at 530
various budgets. 531
Table S3. Dynamic programming algorithm used to calculate the set of optimal solutions for 532
all budgets. 533
534
Page 24 of 41Journal of Applied Ecology
For Peer Review
Table 1: Appraisal of key non-native plant management strategies across the Lake Eyre Basin – average of estimated expected benefits 535
(reduction in total area potentially invaded by the non-native plant species in ha over 50 years, n = 1–6 participants per estimate, based on 536
best guess, upper benefit estimates and lower benefit estimates), standard deviation of ‘best guess’ benefit estimates, net present value of 537
costs (50 years) and cost-effectiveness (CE). Some actions as indicated in Table S1 were not costed over 50 years, but the values are shown this 538
way for comparison purposes. All costs presented are in Australian dollars; m = millions and t= thousands 539
540 Rank (CE)
Best
guess
Rank
(CE)
Upper
Rank (CE)
Lower
Strategy (S) Average benefits
(dominated area
avoided, ha)
best guess (∆ upper
and ∆ lower)
Average
benefits best
guess
standard
deviation
Average
annual costs
(50 years)
1 (40.8) 1 (42.3) 1 (28.8) S3. parkinsonia 1.7 m (+60 t, - 487 t) 758 t $40 678
2 (19.1) 2 (22.7) 3 (13.3) S9. chinee apple 34 t(+17 t, -10 t) 26 t $1797
3 (18.8) 3 (21.04) 2(14.0) S2. mesquite 2.1 m (+252 t, -546 t) 995 t $112 681
4 (13.5) 4 (14.1) 4 (10.5) S4. rubber vine 1.2 m (+50 t, -277 t) 92 t $92 301
5 (11.1) 5(12.7) 5 (7.8) S6. bellyache bush 71 t (+10 t, -21 t) 6 t $6391
6 (10.6) 6 (12.7) 7 (6.0) S11. prickly acacia 10.1 m (+1.9 m, 4.4 m) 956 t $955 678
7 (10.1) 7 (11.5) 6 (7.2) S8. mother of millions 53 t (+7 t, -14 t) 5 t $5340
8 (6.6) 8 (7.1) 8 (3.30) S10. athel pine 331 t (+18 t, -168 t)
155 t $50 071
9 (0.7) 9 (0.9) 9 (0.5) S7. cacti (e.g. coral,
harissia, devils rope) 713 t (+127 t, -200 t)
455 t $1 004 986
10 (0.03) 10 (0.04) 10 (0.03) S5. buffel grass 1.1 m (+32 t, -174 t) 477 t $30 898 881
Page 25 of 41 Journal of Applied Ecology
For Peer Review
Table 2: Summary of the average annual expenditure on each of the non-native plant 541
species strategies and the proportion spent, where the strategies are ordered into 542
decreasing cost-effectiveness. The buffel grass strategy was not included in the bioregion 543
level analyses 544
Strategy (S) Proportional allocation to each bioregion (%)
MGD DEU CHC BHC STP FLB MCR SSD FIN BTP Tan, MII
S3.
parkinsonia
67% 8% 8% 6% 6% 6% – – –
S1
prevention
and
control
programs
for all non-
native
plants
S9. chinee
apple
100% – – – – – – – –
S2. mesquite 38% 12% 38% 12% – – – – –
S4. rubber
vine
21% 58% 21% – – – – – –
S6. bellyache
bush
– 100% – – – – – – –
S11. prickly
acacia
87% – 13% – – – – – –
S8. mother
of millions
74% 26% – – – – – – –
S10. athel
pine
3% – 13% 9% 9% 9% 16% 9% 16% 16%
S7. cacti (e.g.
coral,
harissia,
devil’s rope)
81% 1% 11% 3% <1% 3% – – – –
MGD = Mitchell Grass Downs, DEU = Desert Uplands, CHC = Channel Country, BHC = Broken 545
Hill Complex, STP = Stony Plains, FLB = Flinders Lofty Block, MCR = MacDonnell Ranges, SSD 546
= Simpson Strzelecki Dunefields, FIN = Finke, BTP = Burt Plains, Tanami = Tan, MII = Mount 547
Isa Inlier 548
549
550
Page 26 of 41Journal of Applied Ecology
For Peer Review
a) 551
b) 552
Figure 1: a) Photo of the endemic Mitchell Grass Downs site (54 million ha across eastern 553
and northern Australia on clay and silt rich soils with a mean annual rainfall of 200–550 mm 554
falling mostly in summer) b) Photo of Mitchell Grass Downs site being invaded by prickly 555
acacia Acacia nilotica, which is converting these natural grasslands to open shrubland 556
(photo credit Jennifer Firn). 557
Page 27 of 41 Journal of Applied Ecology
For Peer Review
558
559
Figure 2: Map of the Lake Eyre Basin (LEB) showing Interim Biogeographic Regionalisation 560
for Australia (IBRA), spanning one-sixth of the Australian continent. 561
562
Page 28 of 41Journal of Applied Ecology
For Peer Review
563
Figure 3: Total dominated area avoided (>30% cover) by the implementing of the ten key 564
non-native plant species management strategies at increasing levels of annual investment 565
into non-native plant species control when assuming each strategy, when selected, was 566
implemented across the Lake Eyre Basin. 567
568
0
2
4
6
8
10
12
14
16
18
Average annual expenditure (millions $AUD)
Cum
ula
tive
be
nefit,
tota
l dom
inate
d a
rea
avoid
ed (
mill
ion
s h
a)
0 1 2 3
Parkinsonia
Chinee apple
Mesquite
Rubber vine Bellyache Bush
Prickly Acacia
Mother of Millions
Athel pineCacti
33 34
Buffel grass
Page 29 of 41 Journal of Applied Ecology
For Peer Review
569
570
Figure 4: Optimal sets of strategies at the bioregion level for reducing the total area invaded 571
with budgets ranging from AUD$0 to $2.2 million per year represented as a Pareto frontier 572
(see Figure 5 for detailed sets). Additional budget between A and B or C and D will be 573
insufficient to invest in additional strategies to decrease the expected invaded area. The set 574
of strategies in area D only brings small additional benefits. A budget = $513 000, B budget = 575
$829 000, C budget = $1.34 million and D budget = $2.04 million. 576
Page 30 of 41Journal of Applied Ecology
For Peer Review
577
Figure 5: Optimal set of strategies for a given budget. Additional budget between A and B or 578
C and D will not lead to the selection of additional strategies and decrease of the expected 579
invaded area. Grey areas represent strategies that are selected and white areas represent 580
strategies that are not selected in the optimal set for the given budget. A budget = $513 581
000, B budget = $829 000, C budget = $1.34 million and D budget = $2.04 million. MGD = 582
Mitchell Grass Downs, DEU = Desert Uplands, CHC = Channel Country, BHC = Broken Hill 583
Complex, STP = Stony Plains, FLB = Flinders Lofty Block, MCR = MacDonnell Ranges, SSD = 584
Simpson Strzelecki Dunefields, FIN = Finke, BTP = Burt Plains. 585
586
A B C D
Page 31 of 41 Journal of Applied Ecology
For Peer Review
Supplementary material 1
Table S1: Estimates of the cost of actions that make up each of the specific non-native 2
plant species strategies, including costs (discounted) for the actions over 50 years, and 3
average annual costs over the 50 years (some actions as indicated below were not costed 4
over 50 years, but the values are shown this way for comparison purposes). All costs 5
presented are in Australian dollars. 6
Strategies Description of actions Net present
value (50 years,
discounted)
Average
annual costs
S1.
Prevention
and
monitoring
program for
all weeds
Develop weed management task/plan basin-wide $160,729 $3,215
Mapping, monitoring and surveillance (ground and
aerial, build on weed spotters network) $217, 651 $4,353
Centralised information sharing for weed
incursion/extents $20,000 $400
Secure positions: two FTEs, one FTE for mapping and
centralised information sharing and one FTE for on-
ground activities $8,333,540 $166,671
mesquite awareness campaign $76,651 $1,533
bellyache bush awareness campaign $76,651 $1,533
chinee apple awareness campaign $76,651 $1,533
buffel grass awareness campaign $217,651 $4,353
S2.
mesquite
(Prosopis
glandulosa)
Eradicate from the following bioregions: Mitchell
Grass Downs (MGD), Desert Uplands (DEU), Broken
Hill Complex (BHC) and the Channel Country (CHC);
$800,000 investment in the first year and then
$300,000 per year for next four years. $3,028,191 $60,564
Periodic suppression of new infestations over time;
investment of $200,000 per year for the first three
years and repeated every 10 years $2,6051873 $52,118
S3.
parkinsonia
Prevent spread into Diamantina National Park by
eradicating from Springcreek and Diamantina river
north of the National Park up to the western river
convergence $471,346 $9,427
Eradicate from the following bioregions: Stony Plains,
Broken Hill (STP) Complex (BHC), Flinders Lofty Block $343,193.00 $6,864
Page 32 of 41Journal of Applied Ecology
For Peer Review
(FLB) (SA)
Control downstream outliers and establish large-scale
buffer zones in the Georgina and Thomson Rivers
(downstream of Boulia and Jundah/Windorah
respectively). Requires initial control program (three
years) then periodic suppression every 10 years.
$716,065
$14,321
Impact reduction through introduction and
proliferation of dieback biological control agents in
established infestation areas of the MGD, DEU and
CHC $503,278 $10,066
S4. rubber
vine
Contain and control in DEU $2,264776
$53,896
Eradicate from MGD, CHC
$1,920,286
$38,406
S5. buffel
grass
Contain in STP, Finke (FIN) and control in all other SA
bioregions
$421,611
$8,432
Control in MacDonnell ranges (MDR) to reduce hot
burns – especially along creeks
$228,386,768
$4,567,735
Control and locally eradicate (including rehabilitation)
and prevent incursions into clean areas in gazetted
conservation areas in Queensland
$1,316,135,691
$26,322,714
S6.
bellyache
bush
Eradicate from DEU $319,543
$6,391
S7. cacti
(e.g. coral,
harissia,
devil’s
rope)
Contain all cacti spp. in southern part of SA, eradicate
to north
$3,613,715
$72,274
Control and contain all cacti elsewhere $46,635,538 $932,711
Page 33 of 41 Journal of Applied Ecology
For Peer Review
S8. mother
of millions
Eradicate from urban areas in all areas of the Basin
$26,948
$539
Control and contain in DEU, MGD and any other
occurrences
$240,033
$4,801
S9. chinee
apple
Eradicate from MGD
$89,826
$1,797
S10. athel
pine
Eradicate weedy and high risk Athel pine from
Queensland (e.g. MGD and CHC), and Northern
Territory, except from the lower FIN
$1,358,496
$27,170
Control in South Australia
$1,145,048
$22,900
S11. prickly
acacia
Eradicate from South Australian part of CHC
$10,000
$200
Eradicate from Northern Territory part of MGD
$20,000
$400
Prevent further spread southwards down the three big
rivers (from Stonehenge on cooper system, converging
Diamantina and Western and Wokingham creek,
Boulia on the Georgina)
$32,052,078
$641,042
Containment and progressive reduction of already
infested areas - DEU, MGD, CHC $12,579,442 $251,589
S12. GAB
discharge
springs
Eradicate all non-native plants from the discharge
springs $2,253,453 $43,069
$1,689,549,071 $33,333,017
Page 34 of 41Journal of Applied Ecology
For Peer Review
Total
7
8
Page 35 of 41 Journal of Applied Ecology
For Peer Review
Table S2: Appraisal of key non-native plant management strategy in each of the 9
bioregions of the LEB – estimated average expected benefits (reduction in total area 10
potentially invaded by the non-native plant species), average costs (discounted) and cost-11
effectiveness (CE). Not all strategies were costed within all bioregions, as not all non-12
native plant species can become established within each of the LEB bioregions, because of 13
low and unpredictable rainfall and low nutrient edaphic conditions. Strategies were 14
ranked based on their cost-effectiveness within each of the bioregions and also across all 15
bioregions. The buffel grass strategy was not included in the bioregion level analyses. All 16
costs presented are in Australian dollars (m = millions and t = thousands). 17
Bioregions Strategies Rank CE
within
bioregion
Rank CE
across
LEB
Average
benefits
between
experts over
50 years (ha)
Average Annual
costs
CE
Mitchell
Grass Downs
S2. mesquite 1 3 1.5m $43,226 35.6
S3. parkinsonia 2 4 931t $27,103 34.4
S10. athel pine 3 6 32t $1,712 19.1
S9. chinee apple 4 8 23t $1,797 13.2
S11. prickly
acacia
7 9 9.6m $829,783 11.6
S4. rubber vine 5 11 187t $19.203 9.8
S8. mother of
millions
6 22 1t $3,968 3.4
S7. cacti 8 33 175t $811,090 0.2
Desert
Uplands
S3. parkinsonia 1 2 162t $3,355 48.3
S8. mother of
millions
2 5 28t $1,372 20.4
S6. bellyache
Bush
3 12 61t $6,800 9.7
S4. rubber vine 4 13 400t $53,896 7.40
S2. mesquite 5 14 79t $13,114 6.1
S7. cacti 6 26 41t $14,189 2.9
Channel
Country
S3. parkinsonia 1 1 450t $3,355 134.1
S10. athel pine 2 7 104t $6,292 16.70
S4. rubber vine 3 10 196t $19,203 10.20
S2. mesquite 4 18 197t $43,226 4.6
S11. prickly
acacia
5 25 373t $125,894 3.00
S7. cacti 6 34 14t $107,43 0.10
Broken Hill
Complex
S3. parkinsonia 1 21 8t $2,288 3.7
S10. athel pine 2 27 13t $4,580 2.9
S7. cacti 3 30 50t $31,424 1.6
S2. mesquite 4 32 20t $13,114 1.53
Stony Plains S10. athel pine 1 17 21t $4,580 4.6
Page 36 of 41Journal of Applied Ecology
For Peer Review
S3. parkinsonia 2 20 8t $2,288 3.8
S7. cacti 3 23 3t $9,427 3.2
Flinders
Lofty Block
S3. parkinsonia 1 15 13t $2,288 5.9
S10. athel pine 2 24 14t $4,580 3.1
S7. cacti 3 31 48t $31,423 1.5
Simpson
Strzelecki
Dunefields
S10. athel pine 1 28 11t $4,580 2.3
Finke S10. athel pine 1 19 32t $7,915 4.1
MacDonnell
Ranges
S10. athel pine 1 16 40t $7,915 5.1
Burt plains S10. athel pine 1 29 14t $7,915 1.8
Tanami
S1 prevention and control programs for all non-native plants Mt Isa Inlier
18
Page 37 of 41 Journal of Applied Ecology
For Peer Review
19
20
Figure S1: Map of known locations of the GAB discharge springs within each of the LEB 21
bioregions. Data sourced from the IBRA and Department of Sustainability, Environment, 22
Water, Population and Communities, Australia Commonwealth Government. 23
24
25
Page 38 of 41Journal of Applied Ecology
For Peer Review
Text S1. Dynamic programming algorithm to find the optimal spatial sets of strategies at 26
various budgets. 27
Intuitively, let us assume we computed an optimal solution � for a given budget b. If we 28
remove a strategy j with benefit Bj from the optimal solution, the remaining solution �-{�� } 29
will remain optimal if the budget available is b-Cj, where Cj is cost. This optimal substructure 30
property allows us to solve knapsack problems using dynamic programming. Dynamic 31
programming solves a small sub-problem of the knapsack and then extends this solution 32
iteratively until the complete problem is solved. 33
max∑ �∈� � subject to ∑ � ≤ ������ 34
Where is a binary decision variable that denotes whether (=1) or not (=0) a 35
strategy j is included in the optimal set of strategies. A vector � ∈ {�, �, … , |�||�|} 36
represents a combination of selected strategies. S represents the set of strategies listed 37
in Table S1 and N the number of bioregion. � identifies the benefits of implemented 38
strategy j. Below we provide the details of the dynamic programming algorithm we used 39
to solve our knapsack problem. The algorithm is in two parts. The first part computes 40
the dynamic programming equations for all budgets (Table S3, Lines 1-18, Figure 4) by 41
applying Bellman recursions. A strategy is either too expensive to be selected (Line 7) 42
or if it isn’t we evaluate the added benefits of selecting this strategy (Line 10). If the 43
strategy j improves upon previous strategies, it is selected. The strategy contributes 44
B(j) at a cost of C(j) (Lines 11-14). The second part of the algorithm retrieves the 45
optimal decisions by exploring the temporary variable A (Table S3, Lines 19-36, Figure 46
5). Interested readers can refer to (Kellerer et al., 2004)for additional details and 47
alternative approach to solving knapsack problems. 48
49
Page 39 of 41 Journal of Applied Ecology
For Peer Review
Table S3. Dynamic programming algorithm used to calculate the set of optimal solutions for 50
all budgets adapted from Hans et al. (2004). 51
Dynamic programming for all budgets
Input variable:
n: total number of strategies
B: vector of benefits of length n
C: vector of costs of length n
Cmax: maximum budget we are willing to invest.
Output variable:
zn(0..Cmax) : optimal allocation of funds for all budgets
X(1..n,0..Cmax) : optimal decisions for all budgets
1
2
A = zeros(0..n,0..Cmax); % we need A to retrieve X
For budget b=0 to Cmax
3 z0(b)=0; % initialise the first value
4 endFor
5 For Strategy j=1 to n
6 For budget b=0 to C(j)-1
7 zj(b) = zj-1(b) % strategy j is too expensive
8 endFor
9 For budget b=C(j) to Cmax
10
11
12
13
14
15
16
17
if zj-1(b-C(j))+ B(j) > zj-1(b) then
% select strategy j by adding benefits to zj(b)
zj(b) = zj-1(b-C(j)) + B(j)
% update A
A(j,b) = 1
else zj(b)=zj-1(b)
endif
endfor
18 endFor
Page 40 of 41Journal of Applied Ecology
For Peer Review
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
% We have calculated all the zj(b) including the optimal solution zn(b) for all
% budget b. We now need to retrieve the optimal decision X from A
For budget b=Cmax to 0 % for all budgets
strategy={};
j=n;
c=b;
While j>= 1 and b>0
While j>= 1 and A(j,c)==0
j=j-1; % strategy j has not been selected
endWhile
If j>=1 and A(j,c)==1
strategy=strategy + {j}; % strategy j was selected
c = c - C(j); % update cost, look for the next strategy selected
j=j-1; % update strategy number
endIf
endWhile
X(strategy,b)=1;
endFor
% We have retrieved all the optimal decisions.
52
References 53
Kellerer, H., Pferschy, U. & Pisinger, D. (2004) Knapsack problems. Springer. 54
55
Page 41 of 41 Journal of Applied Ecology