Research Article |
Corresponding author: David C. Cook ( david.cook@agric.wa.gov.au ) Academic editor: Alain Roques
© 2019 David C. Cook.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Cook DC (2019) Quantifying the potential impact of the European wasp (Vespula germanica) on ecosystem services in Western Australia. NeoBiota 50: 55-74. https://doi.org/10.3897/neobiota.50.37573
|
This paper estimates the ecosystem services return on investment in government control of the introduced European wasp (Vespula germanica) in the state of Western Australia. The predictive model used accounts for uncertainties in the spread and impact of V. germanica on provisioning ecosystem services, represented by pollination, apiculture and viticulture, and cultural ecosystem services represented by households. Results produced by simulating a 20-year period suggest government expenditure on management will generate net benefits of AU$3.2–6.3 million per year, most of which will accrue to producers of pollination-dependent crops. This provides justification for investment from the government’s agriculture portfolio to manage an insect often thought of as an urban pest.
Benefit cost analysis, economic impact assessment, ecosystem service impact, European wasp, German wasp, pest management, pollination services, Vespula germanica, yellowjacket
European wasp (Vespula germanica) is an agricultural, environmental, and urban pest first introduced to the Australian state of Western Australia (WA) in the 1970s. To support the investment of public funds on the control of this pest, this paper estimates ecosystem service benefits attributable to ongoing WA government management activities. Ecosystem services are benefits provided by ecosystems, including provisioning services like pollination and food, cultural services such as outdoor recreation, and regulating services such as flood mitigation (
Since the 1940s, V. germanica has spread from its native range in Europe and the Mediterranean region to North America, Chile, South Africa, New Zealand, and Australia where it has become invasive (
The first WA detection occurred in 1977 when six nests were discovered in the Freemantle port area and eradicated (
As with most government departments, DPIRD activities are highly scrutinised because of opportunity costs created with every funding decision. There is a tendency to consider state government money invested in V. germanica control as only creating social benefits in urban areas at the expense of agricultural and developmental opportunities (
The premise of the paper is that without DPIRD’s activities the population of wasps and their colonies are likely to grow rapidly. Mild winter temperatures and the sandy soil of the Swan Coastal Plain on which Perth is located make the area well suited to nest building. Overwintering nests can reach large sizes by the following summer and produce thousands of new queens (
This paper estimates the difference in ecosystem service costs under two scenarios, one in which V. germanica management in its current form is ongoing and the other in which all government efforts to manage the wasp are halted. A bioeconomic model is used to estimate damages under both scenarios over a 20-year period and, thus, damages avoided by ongoing V. germanica management. Despite being relatively simple, the model sufficies to provide indicative benefits of the management policy. Benefit estimates are then compared to the costs to government of providing management services to indicate the return on investment. All monetary values are stated in Australian dollars.
To predict ecosystem service effects resulting from Vespula germanica spread over time under management and nil management scenarios, impacts on three provisioning ecosystem services and one cultural ecosystem service are considered.
I. Pollination impacts
Apis mellifera was introduced into Australia soon after the arrival of the first Europeans and has become widespread (
Crop | Area†(ha) | Volume†(T) | Gross Value‡($ million) | Pollinator reliance§(%) |
---|---|---|---|---|
Almond | 210 | 145 | 1.5 | 100 |
Avocado | 8506 | 24621 | 118.4 | 100 |
Blueberry | 23 | 81 | 1.8 | 100 |
Canola | 1093647 | 1327849 | 730.0 | 15 |
Citrus | 1436 | 13282 | 27.0 | 30 |
Cucumber | 238 | 4028 | 17.8 | 100 |
Lupin | 331493 | 457262 | 158.4 | 10 |
Mango | 840 | 1424 | 8.1 | 50 |
Melons | 591 | 16076 | 20.4 | 100 |
Pome fruit | 2981 | 38802 | 98.4 | 50 |
Pumpkin | 1114 | 18774 | 16.9 | 90 |
Stone fruit | 298 | 8039 | 26.1 | 70 |
Strawberry | 194 | 5112 | 42.5 | 40 |
TOTAL | 1441571 | 1915495 | 1267.3 |
II. Apiculture
Managed A. mellifera hives are affected by ‘raiding’ behaviour of expanding V. germanica populations. There are approximately 28,500 managed hives in WA producing over 1,600 tonnes of honey worth $4.9 million per year (
III. Viticulture
Vespula germanica damage grapes and introduce foreign yeasts that can interfere with the fermentation process (
IV. Households
Vespula germanica is a serious household pest in warmer climates where breeding and nest construction continue throughout the year, resulting in large summer colonies containing many thousands of individuals (
Vespula germanica impacts over time are approximated using a Monte Carlo simulation model. The main purpose of the model is to provide the benefit component of a benefit cost analysis to inform DPIRD managers of likely returns to investment in V. germanica management activities. However, the model also required sufficient detail to gain traction with these managers, and to produce spread scenarios they considered plausible given their experiences with the pest.
The Monte Carlo model simulates a 20-year period. Uncertain parameters are entered as distributions and a Latin hypercube sampling algorithm used to sample from each using the @Risk software package (Palisade Software, Ithaca, New York). Parameter distribution types used in the model include: (i) PERT, a type of beta distribution specified using minimum, most likely (i.e. skewness), and maximum values; (ii) uniform, a rectangular distribution bounded by minimum and maximum values; and (iii) discrete, a distribution containing several discrete outcomes and their probabilities of occurrence. Biological and economic parameter values appear in Tables
To describe changes in V. germanica impacts across multiple regions, the logistic model of
The model assumes that the proportion of a sector i (i.e. horticulture, apiculture, viticulture, households) affected in period t (Sit) increases over time following the logistic equation:
Here, Simaxis the total size of sector i affected (i.e. in number of ha for horticulture and viticulture, the number of hives for apiculture and the number of residences for households); Iimaxis the maximum proportion of sector i affected; Iiminis the minimum proportion of sector i affected, and; ωi is the rate at which V. germanica moves from Iiminto Iimax.
In the absence of information about ω, a hypothetical impact growth rate is used determined by the number of time periods taken for V. germanica to affect a given proportion, θi , of Simaxsuch that:
Here, θi is a specified proportion of Simaxaffected and tθi is the number of periods (years) taken for V. germanica to reach θi . The values and distributions assigned to each parameter in each sector are provided in Tables
Parameter | Nil management | Management |
---|---|---|
Biological | ||
Infestation growth, ωi (unitless)† | 0.33–0.83 | 0.22–0.33 |
Maximum proportion affected, Iimax(%)‡ | Uniform(20,30) | Uniform(20,30) |
Minimum proportion affected, Iimin(%)† | 0.01 | 0.01 |
Proportion of Iimaxaffected at tθi , θi (%)† | 15–100 | 15–100 |
Time taken for θi to be affected (yr)† | Uniform(10,20) | Uniform(20,30) |
Economic | ||
Demand elasticity, η§ | Uniform(−1.1,−1) | Uniform(−1.1,−1) |
Discount rate, υ (%)¶ | Pert(2,5,7) | Pert(2,5,7) |
Increased variable cost, Vit | 0 | 0 |
Inflation rate, ι (%)†† | Pert(1.5,2,2.5) | Pert(1.5,2,2.5) |
Price of per unit, Pit ($/T) ‡‡ | Almond 10300 | Almond 10300 |
Avocado 4800 | Avocado 4800 | |
Blueberry 22700 | Blueberry 22700 | |
Canola 500 | Canola 500 | |
Citrus 2000 | Citrus 2000 | |
Cucumber 4400 | Cucumber 4400 | |
Lupin 300 | Lupin 300 | |
Macadamia nut 5100 | Macadamia nut 5100 | |
Mango 5700 | Mango 5700 | |
Melons 1300 | Melons 1300 | |
Pome fruit 2500 | Pome fruit 2500 | |
Pumpkin 900 | Pumpkin 900 | |
Stone fruit 3200 | Stone fruit 3200 | |
Strawberry 8300 | Strawberry 8300 | |
Yield loss despite control, Yit (%)§§ | Uniform(8,10) | Uniform(8,10) |
Validation of the model for all sectors in both scenarios is not possible due to a lack of data. No data exists for a nil management scenario as V. germanica has been managed since it was first detected in WA, but data relevant to the management scenario are available from DPIRD for the past 20 years (1999–2018). These data include all reported and detected instances of wasps responded to by DPIRD over time, and given the majority of activity has occurred in the Perth metropolitan area they are used as a proxy for numbers of households affected. This allowed a rudimentary validation of the model to be undertaken as it applied to the household sector using visual assessment and deviance measures.
Parameter | Nil management | Management |
---|---|---|
Biological | ||
Infestation growth, ωi (unitless)† | 0.31–0.61 | 0.2–0.31 |
Maximum proportion affected, Iimax(%)† | Uniform(8,10) | Uniform(8,10) |
Minimum proportion affected, Iimin(%)† | 0.01 | 0.01 |
Proportion of Iimaxaffected at tθi , θi , (%)† | 5 | 5 |
Time taken for θi to be affected (yr)† | Uniform(10,20) | Uniform(20,30) |
Economic | ||
Demand elasticity, η‡ | Uniform(−1.1,−1) | −0.28 |
Discount rate, υ (%)§ | Pert(2,5,7) | Pert(2,5,7) |
Increased variable cost, Vit ($/hive)¶ | Pert(25,30,50) | Pert(25,30,50) |
Inflation rate, ι (%)†† | Pert(1.5,2,2.5) | Pert(1.5,2,2.5) |
Price of per unit, Pit ($/hive) ‡‡ | 170 | 170 |
Yield loss despite control, Yit (%)¶ | 0–10 | 0–10 |
Parameter | Nil management | Management |
---|---|---|
Biological | ||
Infestation growth, ωi (unitless)† | 0.3–0.6 | 0.2–0.3 |
Maximum proportion affected, Iimax(%)† | Uniform(10,15) | Uniform(10,15) |
Minimum proportion affected, Iimin(%)† | 0.01 | 0.01 |
Proportion of Iimaxaffected at tθi , θi , (%)† | Uniform(5,9) | Uniform(5,9) |
Time taken for θi to be affected (yr)† | Uniform(10,20) | Uniform(20,30) |
Economic | ||
Demand elasticity, η‡ | Uniform(−1.1,−1) | Uniform(−1.1,−1) |
Discount rate, υ (%)§ | Pert(2,5,7) | Pert(2,5,7) |
Increased variable cost, Vit ($/ha)¶ | 145 | 145 |
Inflation rate, ι (%)†† | Pert(1.5,2,2.5) | Pert(1.5,2,2.5) |
Price of per unit, Pit ($/T) ‡‡ | 2500 | 2500 |
Yield loss despite control, Yit (%) | 0 | 0 |
Visual assessment involved a graphical display of the data and model simulation output being shown to two experts involved in the DPIRD management project. They were presented with a diagram similar to Figure
Visual validation plotting simulated and observed data of the proportion of households affected by V. germanica over the past 20 years.
Statistical validation of the model is problematic as it is stochastic, producing a distribution for comparison to each observation. Moreover, only a single set of observed time-series data is available to compare the model output against, which introduces an autocorrelation problem. As a simple deviance measure test, the mean absolute error (MAE) and mean absolute percentage error (MAPE) between observed and model output were calculated using the mean of the simulated data. The MAE was 0.14%, indicating predicted values for the proportion of households affected were an average of 0.14% from observed values. The MAPE was 8.3%, indicating prediction error is, on average, 8.3% of the observed value. As a rule of thumb, a 10% MAPE is an approximate maximum limit for model acceptance (
Parameter | Nil management | Management |
---|---|---|
Biological | ||
Infestation growth, ωi (unitless)† | 0.41–0.82 | 0.27–0.41 |
Maximum proportion affected, Iimax(%)† | 1 | 1 |
Minimum proportion affected, Iimin(%)† | 0.01 | 0.01 |
Proportion of Iimaxaffected at tθi , θi , (%)† | 0.9 | 0.9 |
Time taken for θi to be affected (yr)† | Uniform(10,20) | Uniform(20,30) |
Economic | ||
Demand elasticity, η | na | na |
Discount rate, υ (%)‡ | Pert(2,5,7) | Pert(2,5,7) |
Increased variable cost, Vit ($/household)§ | Uniform(200,250) | Uniform(200,250) |
Inflation rate, ι (%)¶ | Pert(1.5,2,2.5) | Pert(1.5,2,2.5) |
Price of per unit, Pit ($/T) | na | na |
Yield loss despite control, Yit (%) | na | na |
The model estimates the ecosystem services damage (d) caused by V. germanica under nil management (dNM) and on-going management (dM) scenarios. The nil management scenario is constructed as a counterfactual to which a management policy can be compared to determine the reduction in damages attributable to the policy over time.
The difference between dNMand dMis simulated over 20 years. The ecosystem services damage cost of V. germanica in sector i in time period t under a nil management policy (ditNM) is calculated as:
where: n is the number of sectors affected by V. germanica; SitNMis the proportion of sector i affected by V. germanica in period t under a nil management policy scenario; Yit is the mean change in yield in sector i attributable to V. germanica in year t; Pit is the world price of product produced in sector i in year t; Nit is the number of “units” (i.e. ha, hives, residences) in sector i potentially affected by V. germanica in year t, and; Vit is the increase in variable cost per unit induced by V. germanica in sector i in year t.
The ecosystem services damage cost of V. germanica in a region i in time period t under an ongoing management policy (ditNM) is calculated as:
where: SitMis the proportion of sector i affected by V. germanica in period t if an ongoing management policy is in place.
For each sector that experiences yield effects from V. germanica, an estimate of price, Pit , is given for the first time step of the model (i.e. Pi0, corresponding to the year 2018). This is the initial price per unit for an affected product, but its price will change over time given that the demand for agricultural products is elastic (i.e. price increases with relative scarcity, and vice versa). The price in periods after t0 will be partially influenced by the impact of V. germanica on production.
This price effect assumes the markets for affected products are protected, preventing perfect substitution of externally produced goods for those damaged by V. germanica. If WA markets were unregulated and open to free trade with suppliers from other states and overseas, and if the WA industries contributed a relatively small amount to global production, local prices of affected agricultural products would remain unchanged in response to V. germanica spread and impact (e.g.
Predicted yield loss, Yit Nit , is used as a proxy for the V. germanica-induced reduction in sectoral output. This is combined with the lagged per unit price, Pt–1, to calculate
Here, Git is the gross value of production divided by 100 and η is the elasticity of demand for the affected commodity (i.e. the ratio of percentage change in quantity demanded over the percentage change in price).
Returning to equations 3 and 4, dNMand dMaccrue over time and are subject to discounting. Discounting has an erosive effect on monetary values that increases with time, meaning that the present value of one unit of damage caused in the present is worth more than the same amount of damage caused in the future.
Applying an exponential discount rate, the present value of benefits anticipated from an on-going management policy in time period t (PVBtM) is estimated by summing ditNM– ditMacross all affected sectors (n) in WA:
where v is the discount rate.
The net present value of the V. germanica management policy (NPBtM) is calculated summing the difference between the present value of costs (PVCtM) and PVBtMover m time periods:
The benefit cost ratio for the on-going V. germanica management option (BCRM) is calculated by dividing the summed PVBtMover m time periods by the summed PVCtMover m time periods. Note that PVBtMrepresents gross (as opposed to net) benefits (i.e. PVBtM= NPVtM+ PVCtM).
In the results section to follow, all costs and benefits are stated in Australian dollars. NPVMand BCRMare given for a range of PVCMbetween $230,000 and $250,000 per annum over a period of 20 years. This range approximates the total amount spent by DPIRD in the past several years, and is indexed to the inflation rate. This means that PVCMis fixed in real terms and nominal costs (CM) increase at the inflation rate (ι) over time (i.e.) .
Ecosystem services damage predicted by the model under the nil management scenario (i.e. dNM, eq. 3) and on-going management scenario (i.e. dM, eq. 4) for each sector are shown in the box-whisker plots in panels A–D of Figure
Predicted damage cost per year associated with V. germanica impacts in WA over 20 years. Panels A, B, C and D show pollination, apiculture, viticulture and household damage costs, respectively, under both scenarios, while panel E shows the summed damage costs across all sectors under both scenarios. Box whisker plots indicate 5th, 25th, mean, 75thand 95thpercentile values, with shaded boxes representing the nil management scenario and hollow boxes the management scenario.
The uncertainty in model predictions is evident in the width of the boxes and length of whiskers in Figure
The benefits and costs of V. germanica management are compared in Figure
Net present value of V. germanica management in WA over 20 years. The box whisker plot indicates 5th, 25th, mean, 75thand 95thpercentile values.
However, there is considerable uncertainty in the model predictions that could lead to a substantially better or worse return on investment than indicated by the mean. Over 10 years, 80% of model iterations produced a present value of benefit of $2.1–5.6 million, suggesting a benefit cost ratio between 8.3 and 22.5. Morover, the uncertainty in model predictions increases as the length of the simulation period increases. Over 20 years, the estimated present value of benefit varies between $6.5–26.2 million, resulting in a benefit cost ratio between 13.8–26.2.
Despite this uncertainty, results of a parameter sensitivity analysis indicate that the return to investment in management remain positive even under worst-case scenarios. To gauge the effect of the parameters on model output, each parameter is sampled across its specified range while holding all other parameters constant in Figure
Sensitivity analysis illustrating how the mean net benefit of V. germanica management in WA 20 years is affected by changes in input parameters.
Results are most sensitive to changes in the discount rate, which is specified as Pert(2%,5%,7%). It is inversely related to the present value of benefit. Lowering the discount rate from its most likely value of 5% to 2% (a change of −60%) increases the present value of benefit by approximately 31% (from $4.9 million to $6.4 million), and increasing it to 7% (a change of 40%) lowers the present value of benefit by approximately 24% (to $3.7 million). Determining an appropriate discount rate is one of the most controversial and important issues in benefit cost analysis since as it has a major impact on the viability of many public projects (
Results are also highly sensitive to the time taken for the indicative proportion θi to be affected under the management scenario. This is also inversely related to the present value of benefit, producing a ±24% change when increased or decreased 20% from the mean value (25 years). As it relates to the effectiveness of DPIRD activities in slowing the spread of V. germanica, the time taken for θi to be affected under the management scenario is a key assumption. Citing the DPIRD time series data used to validate the model, the range 20–30 years is a reasonable approximation for this parameter. Even when at 20 years, the model still produces a present value of benefit of $3.7 million.
Other parameters with relatively high sensitivities mostly relate to the pollination sector, including yield loss despite control, increase in variable costs, maximum proportion affected (Iimax) and the indicative proportion θi . This reflects the large size of pollination sector impacts compared with those in the household, viticulture and apiculture sectors.
The model used in this analysis takes into account multiple ecosystem services and conveys the uncertain future benefits of invasive species controls to decision-makers in relatively simply terms. As the impacts of invasive species change with respect to time, location, and other variables in ways that are difficult to predict, policy-makers need to be informed by predictive (ex ante) analyses that are explicit about the uncertain future effects of decisions made in the present (
Research concerning economic impacts of invasive species has increased in recent decades, but most has involved ex post impact assessments and management evaluation (
Several ex ante studies have used complex, spatially explicit approaches and stochastic simulations to characterise uncertainty in spread patterns over time combining environmental variables and invasive species behaviours (
Economic modelling has seldom been used as part of an invasive species ecosystems service impact assessment.
The future ecosystem service impact predicted in this analysis hint at large returns to investment in the ongoing management of V. germanica in WA, particularly in terms of provisioning ecosystem services to private producers of pollination-dependent crops. This justifies the WA government’s use of DPIRD resources in managing the pest rather than another department since the impacts of the wasp are mainly agricultural. Funding is relatively low (i.e. $200,000–250,000 per year) when compared to the gross value of crops affected (i.e. $1.3 billion, see Table
If the pollination sector is removed from the model, the household sector becomes the largest beneficiary of management activities and the 20-year benefit cost ratio falls from 13.8–26.2 to 3.0–4.3. This might suggest the state’s demand for wasp nest removal could be met by private pest controllers in the Perth metropolitan area rather than government. The main beneficiaries are spatially concentrated in this area and benefits to the apiculture and viticulture sectors are small in comparison. Hence, the positive flow-on effects beyond the household sector would be minimal.
However, if pollination services are included in policy decisions, the situation changes considerably. Beneficiaries of management are now spatially diffuse, consisting of various industry groups, community groups and institutions. This would make it logistically challenging and prohibitively costly to bring all affected parties together to negotiate wasp management plans and control targets and monitoring with private pest control operators. Therefore, government intervention is necessary to ensure an adequate level of management services are provided to all affected groups.
If cultural ecosystem service impacts of V. germanica related to biodiversity are also included in policy decisions, the need for government intervention becomes even stronger because biological diversity is a public good. Public goods are non-rivalrous in consumption (i.e. enjoyment of biodiversity by one person does not affect the quantity available for another) and have benefits that are non-exclusive (i.e. one person cannot prevent another from enjoying the benefits of biodiversity). As such, these goods cannot be provided to a socially desirable level by private providers who are unable to charge for the full benefits their services create, nor prevent people from enjoying benefits they have not paid for.
To the author’s knowledge, no research is currently available concerning the potential for V. germanica to affect biodiversity in WA, but experience elsewhere suggests damage could be considerable. For instance, the introduction of the wasp to Tasmania has resulted in severe local reductions of invertebrates (
The model presented in this paper estimates the return on government investment in continued V. germanica management in WA in terms of provisioning and cultural ecosystem services. Results suggest that the combined ecosystem service benefits of ongoing management over the next 20 years are likely be $3.4–6.5 million per year. With annual costs of management being $200,000–250,000, this indicates a net benefit of $3.2–6.3 million per year. The largest beneficiaries are producers of crops depended on insect pollination. These benefits have a tendency to be overlooked due to the reputation of V. germanica as an urban nuisance, rather than an agricultural pest. If pollination benefits are ignored, households are indeed the largest beneficiaries of wasp control and there may be grounds for turning management over to the private sector. However, if pollination impacts are as large as the results of this analysis suggest, negotiation costs and information constraints are likely to prevent private controllers from providing sufficient management services. If cultural service benefits of V. germanica management are also considered, such as prevented damage to unique species in the south west of WA, the case for government provision is also strengthened.
Thank you to Catherine Webb and Marc Widmer from the Department of Primary Industries and Regional Development for information generously provided. Thanks also to two anonymous reviewers for comments and suggestions that greatly improved the paper.