Handling overheads: optimal multi-method invasive species control
Invasive species are a pervasive problem worldwide and considerable resources are directed towards their control. While there are many aspects to invasive species management, deciding how to allocate resources effectively when removing them is critical. There are often multiple control methods available, each with different characteristics. For example aerial baiting has very high overhead costs, while animal trapping incurs a handling time (the trap must be reset after each capture). Here, we examine a particular challenge that managers commonly face when designing eradication programmes—specifically what type of control measure to rely on at different times during the eradication effort? We solve for optimal resource allocation strategies when there are two control methods available and one has overhead costs and the other has a handling time. We find that, if both controls are being used, the control with overhead costs should be used only at the beginning of a project, the other control should be used in the latter part of the project, and that there is generally an overlap where both controls are used. This contrasts with the strategies employed in many eradication projects, where ground control does not begin until aerial baiting has ceased.
KeywordsOptimal control Overhead cost Handling time Optimal control theory Ecological economics
Invasive species are responsible for considerable environmental and economic damage (Gurevitch and Padilla 2004; Pimentel et al. 2005; Simberloff et al. 2013), and substantial resources are directed towards their control (Veitch and Clout 2002). Although invasive species eradications can result in large conservation gains (Jones et al. 2016), resources to complete eradications are limited (McCarthy et al. 2012). As such, there is a large body of work focusing on allocating conservation resources as effectively as possible. In doing this, managers must decide how to use the different types of available control methods through time.
Multi-method control projects arise across a diverse range of systems. For example, the rabbit eradication on Macquarie Island, an Australian subantarctic island, used a combination of aerial poison baiting and canine-assisted ground hunting (Robinson and Copson 2014; Terauds et al. 2014). The programme started with aerial baiting and switched to ground hunting once a majority of the population was removed. Similar strategies also arise across invertebrate and weed control problems (Baker et al. 2017; Hodgson et al. 2014; Noble and Rose 2013). The conventional wisdom is to use broad-scale aerial treatments during the early stages of the project, while the invasive species abundance is high. This is because these types of methods have very high overhead costs (i.e. the cost of using an aircraft), and the cost is only justified if many invasive individuals can be removed.
We focus on how the cost structure involved in implementing different control actions shapes the best way to combine different control actions during an invasive species eradication programme. Some studies have considered aspects of this problem while pursuing other questions. However, none to our knowledge has focused specifically on the effects of the cost structure of different controls. Many analyses about invasive species control have focused single-species, single-control problems (Baker and Bode 2016; Baxter et al. 2008; Chades et al. 2011; Hastings et al. 2006; Walker et al. 2015). There has also been work in designing management strategies that take into account multiple species (Bode et al. 2015; Carrasco et al. 2010; Epanchin-Niell et al. 2014; Kern et al. 2007; Lampert et al. 2014), and analyses that consider multiple control strategies in invasive species control (Blackwood et al. 2012; Marten and Moore 2011) and epidemiology (Fenichel and Horan 2007; Horan et al. 2008). However, these analyses do not allocate resources dynamically between control methods for invasive species control.
While multi-method control has been somewhat overlooked in the invasive species control literature, the analogous problem in fishery management, where multiple fishers harvest a single resource, has been studied (Armsworth et al. 2011; Cook and McGaw 1996). A key difference between these harvesting problems and invasive species control is that the objective in a harvesting system is often to find an equilibrium solution. However, for invasive species eradication problems,the transient aspect of the problem is the most interesting. That is, how to get from the current state to zero, or very low, abundance while minimising cost (Baker and Bode 2016). Hence, in this paper, we consider the problem of eradicating an invasive species population when there are two different control methods available. In particular, we examine how overhead costs on one control and a handling time on the other affects the optimal control strategy, and we solve to see how managers should dynamically alter their resource allocation to different control methods through time. We begin by treating the optimal control problem, and then we start by solving using only one control and build in complexity, by adding handling times, a second control and finally overhead costs.
Throughout this analysis, we assume that control method one has an overhead cost, while control method two incurs a handling time. Overhead costs appear when there is an additional cost to using a control method, that does not depend on the amount of effort employed. For example, during aerial baiting activities, there is a cost of hiring the aircraft that does not depend on how many hours it is used per day. Hence, there is a fixed cost (cost of hire per day) and a variable cost (cost of labour and baits). A handling time limits the total number of individuals removed per day. This models ground-based control methods such as trapping, where a certain amount of time is required to reset a trap before it can be effective again. For example, if a trap is checked once per day, then any single trap cannot capture more than one individual per day. While any control method would likely have some overhead cost and handling time, we consider a situation where the overhead cost of method one is much greater than that of method two; similarly, the handling time of method two is much greater than that of method one. Hence, we only model overhead costs on method one and handling times on method two, such that we can clearly examine the differences between these two types of control.
To ensure that this is a minimum, we see that the second derivatives of H with respect to e1 and e2 must be positive when evaluated at all control pairs. Hence, solving the system of Eqs. 1, 4, 9 and 11 gives the solution to our optimal control problem.
We solve the system of Eqs. 1, 9, 4 and 11 using boundary value problem software in MATLAB (bvp4c). Some combinations of parameters could not be solved from arbitrary initial guesses. In these cases, we used continuation. That is, we started with a set of parameters that we were able to solve. We then used that solution as the initial guess for a nearby set of parameters, and we repeated this process until we arrived at a solution for our target parameter set.
Our primary focus in the paper is on the role of overhead costs on one control variable when compared to handling time limitations on the other. We chose values for the key parameters governing the size of these cost terms to ensure the relative variation was appropriately captured. Specifically, we assumed that the units of effort to be the number of (full-time equivalent) people working on the project at any given time. As such, we set the cost, c1 and c2, to both be USD$50,000. Eradications typically cost between USD$200,000 and USD$2,000,000 (Holmes et al. 2015), and aircraft hire is often around 35% of the total project cost. Therefore, we consider three values of b1: USD$70,000, USD$300,000 and USD$700,000. Other parameters were chosen to be more indicative of a government programme focused on eradicating a population of a fast-growing species on a small to medium-sized island. Specifically, we assumed a 3% discount rate, a continuous time intrinsic rate of increase of r = 1 and a total population size of 1000. We use a time horizon of T = 2.
There are no clear values to choose for the remaining parameters, and the aim in this paper is to explore how the optimal solution is affected by the control function parameters. Hence, we explore how different solutions arise from different parameter choices. We assume that both control methods should be used for at least part of the project. As the handling time reduces the efficacy of method two, we always choose μ2 ≥ μ1, and specific parameter values are given throughout the results. It would be possible to conduct experiments to determine these control parameters. Though, multiple trials would need to be conducted, with varying effort and at locations with different abundances, to get good estimates of μ2,d2 and h concurrently. Additionally, it is important to know how much altering parameter changes the optimal solution, as the best management strategy can be robust to changes in parameter values (Baker et al. 2016; Li et al. 2017).
We find that the shape of the optimal solution depends on the presence of overhead costs, whether handling times are present, and the number of control methods available. Similar work without handling times has suggested that control effort should increase throughout a project when only one control method used (Baker and Bode 2016). However, we find that introducing handling time changes this, and that, even with a small handling time, the optimal solution changes to peak near the middle of the project. This also leads to a change in the decline of the invasive species. With a handling time, the invasive species declines much more quickly at start of the project and slower at the end of the project, compared to the case without a handling time. While we focused on overhead costs and handling times, there are other interesting aspects to this problem, in particular how the species growth rate and diminishing marginal returns parameter affect the optimal solution, and these interactions are discussed in Baker and Bode (2016).
When a second control without a handling time is added, we find that the optimal solution shifts such that control method two no longer has an internal peak, rather it increases throughout. However, when we added an overhead cost to this method, we found that method one should be used from the start and for potentially only part of the project, while method two is not used immediately. Though, we note that control method two is started while method one is still in use. This result (that there is a period of overlap when both methods should be deployed) is particularly interesting because it is common for eradication projects to begin with broad-scale aerial control, which is followed by ground-based control (Algar et al. 2002; Kessler 2002; Mowbray 2002; Macdonald and Leaman 2002; Veitch 2002). These results indicate that the ground-based control should start while the broad-scale control is ongoing, and this has occasionally been implemented on-ground (McClelland 2002).
There are a range of other aspects to invasive species control that we did not treat here. Our analysis focused on controlling a single species, but, the removal of an invasive species often has an impact on other species present in the ecosystem (Doherty et al. 2015). For example, in some situations, control problems should be formulated with endemic species in the objective to ensure that the strategy will not only remove the invasive species, but also benefit the ecosystem (Bode et al. 2015). Our model is purely temporal and is best suited to island control projects. However, the underlying concepts about allocating resources between different control methods should carry over to more general spatial cases (Baker 2016). Related problems include decisions about when to begin controlling a spreading invader (Sims and Finnoff 2013; Sims et al. 2016), allocating funds towards quarantine (Moore et al. 2010, Rout et al. 2011, 2014) and allocating effort towards detecting species (Epanchin-Niell et al. 2012; Hauser and McCarthy 2009; Mehta et al. 2007; Morrison et al. 2007; Ramsey et al. 2011; Regan et al. 2011). In general, each of these different aspects can have multiple methods, and as our work has shown, deciding how to allocate resources between them is complicated because the optimal temporal allocation for a certain method depends both on its own characteristics (e.g. its cost structure) and what other methods are available.
Invasive species control projects are expensive and high profile (Holmes et al. 2015), and it is important to allocate resources carefully throughout a project. When multiple control methods are being used, cost-effective resource allocation relies on a strong understanding of the costs and benefits of each method. This is complicated as these will change throughout a project as the invasive species population changes. Mathematical modelling and optimisation is a good way to deal with this complexity and identify optimal strategies (Armsworth and Roughgarden 2001). In our analyses, knowing when to cease using the method with large overheads is vital, and getting it right can lead to large savings. Despite the potential benefits, we are not aware of invasive species management projects that have explicitly incorporated this type of analysis when determining the allocation of multiple control methods. There is a strong push for conservation decision-making to be open and transparent (Blomquist et al. 2010; Donlan et al. 2014; Gregory and Long 2009), and mathematical solutions and analysis should be a part of it.
We would like to thank Michael Bode and Michael McCarthy for their input. Christopher Baker is the recipient of a John Stocker Postdoctoral Fellowship from the Science and Industry Endowment Fund. This work was assisted by attendance as a Short-term Visitor at the National Institute for Mathematical and Biological Synthesis, an Institute sponsored by the National Science Foundation through NSF Award #DBI-1300426, with additional support from The University of Tennessee, Knoxville and the Australian Research Council Centre of Excellence for Environmental Decisions.
- Algar D, Burbridge AA, Angus GJ (2002) Cat eradication on Hermite Island, Montebello Islands, Western Australia. In: Turning the tide: the eradication of invasive speciesGoogle Scholar
- Baker CM (2016) Target the source: optimal spatiotemporal resource allocation for invasive species control. Conserv Lett. https://doi.org/10.1111/conl.12236
- Baker CM, Gordon A, Bode M (2016), Ensemble ecosystem modeling for predicting ecosystem response to predator reintroduction. Conserv Biol. https://doi.org/10.1111/cobi.12798
- Baker CM, Hodgson JC, Tartaglia E, Clarke RH (2017) Modelling tropical fire ant (Solenopsis Geminata) dynamics and detection to inform an eradication project. Biological Invasions. https://doi.org/10.1111/cobi.12798
- Bode M, Baker CM, Plein M (2015) Eradicating down the food chain: optimal multispecies eradication schedules for a commonly encountered invaded island ecosystem. J Appl Ecol 571–579. https://doi.org/10.1111/1365-2664.12429
- Cook BA, McGaw RL (1996) Sport and commercial fishing allocations for the Atlantic salmon fisheries of the Miramichi River. Canadian Journal of Agricultural Economics/Revue canadienne d’agroeconomie 44(2):165–171. https://doi.org/10.1111/j.1744-7976.1996.tb00191.x CrossRefGoogle Scholar
- Donlan CJ, Luque GM, Wilcox C (2014) Maximizing return on investment for island restoration and species conservation. Conserv Lett 171–179. https://doi.org/10.1111/conl.12126
- Epanchin-Niell RS, Brockerhoff EG, Kean JM, Turner J (2014) Designing cost-efficient surveillance for early detection and control of multiple biological invaders. Ecol Appl. https://doi.org/10.1890/13-1331.1
- Fenichel EP, Horan RD (2007) Jointly-determined ecological thresholds and economic trade-offs in wildlife disease management. Nat Resour Model 20(4):511–547. https://doi.org/10.1111/j.1939-7445.2007.tb00219.x CrossRefGoogle Scholar
- Hodgson JC, Abbott KL, Clarke RH (2014) Eradication plan “Tropical Fire Ant Solenopsis geminata at Ashmore Reef Commonwealth Marine Reserve”. Tech. rep. Monash UniversityGoogle Scholar
- Jones HP, Holmes ND, Butchart SHM, Tershy BR, Kappes PJ, Corkery I, Aguirre-Muñoz A, Armstrong DP, Bonnaud E, Burbidge AA, Campbell K, Courchamp F, Cowan PE, Cuthbert RJ, Ebbert S, Genovesi P, Howald GR, Keitt BS, Kress SW, Miskelly CM, Oppel S, Poncet S, Rauzon MJ, Rocamora G, Russell JC, Samaniego-Herrera A, Seddon PJ, Spatz DR, Towns DR, Croll DA (2016) Invasive mammal eradication on islands results in substantial conservation gains. Proc Natl Acad Sci p 201521179. https://doi.org/10.1073/pnas.1521179113
- Kessler D (2002) Eradication of feral goats and pigs and consequences for other biota on Sarigan Island, Commonwealth of the Northern Mariana Islands. In: Turning the tide: the eradication of invasive speciesGoogle Scholar
- Lenhart S, Workman JT (2007) Optimal control applied to biological models. Mathematical and computational biology series. Chapman & Hall/CRC, LondonGoogle Scholar
- Li S-L, Bjørnstad ON, Ferrari MJ, Mummah R, Runge MC, Fonnesbeck CJ, Tildesley MJ, Probert WJM, Shea K (2017) Essential information: uncertainty and optimal control of ebola outbreaks. Proc Natl Acad Sci 114(22):5659-5664. https://doi.org/10.1073/pnas.1617482114
- Macdonald N, Leaman AA (2002) An attempt to eradicate feral goats from Lord Howe Island. In: Turning the tide: the eradication of invasive speciesGoogle Scholar
- McCarthy DP, Donald PF, Scharlemann JPW, Buchanan GM, Balmford A, Green JMH, Bennun LA, Burgess ND, Fishpool LDC, Garnett ST, Leonard DL, Maloney RF, Morling P, Schaefer HM, Symes A, Wiedenfeld DA, Butchart SHM (2012) Financial costs of meeting global biodiversity conservation targets: current spending and unmet needs. Science 338(6109):946–949. https://doi.org/10.1126/science.1229803 CrossRefPubMedGoogle Scholar
- McClelland PJ (2002) Eradication of Pacific rats (Rattus exulans) from Whenua Hou Nature Reserve (Codfish Island), Putauhinu and Rarotoka Islands, New Zealand. In: Turning the tide: the eradication of invasive speciesGoogle Scholar
- Morrison SA, Macdonald N, Walker K, Lozier L, Shaw MR (2007) Facing the dilemma at eradication’s end: uncertainty of absence and the Lazarus effect. Front Ecol Environ 5(5):271–276. https://doi.org/10.1890/1540-9295(2007)5[271:FTDAEE]2.0.CO;2 CrossRefGoogle Scholar
- Mowbray D (2002) Eradication of introduced Australian marsupials (brushtail possum and brushtailed rock wallaby) from Rangitoto and Motutapu Islands, New Zealand. In: Turning the tide: the eradication of invasive speciesGoogle Scholar
- Noble M, Rose M (2013) Managing African boxthorn (Lycium ferocissimum) in Australia. Tech. rep. Department of Agriculture, Fisheries and Forestry, TasmaniaGoogle Scholar
- Ramsey DSL, Parkes JP, Will D, Hanson CC, Campbell KJ (2011) Quantifying the success of feral cat eradication, San Nicolas Island, California. N Z J Ecol 35(2):163–173Google Scholar
- Rout TM, Moore JL, McCarthy MA (2014) Prevent, search, or destroy? A partially observable model for invasive species management. J Appl Ecol. https://doi.org/10.1111/1365-2664.12234
- Simberloff D, Martin JL, Genovesi P, Maris V, Wardle DA, Aronson J, Courchamp F, Galil B, García-Berthou E, Pascal M, Pyšek P, Sousa R, Tabacchi E, Vilà M (2013) Impacts of biological invasions: what’s what and the way forward. Trends Ecol Evol 28(1):58–66. https://doi.org/10.1016/j.tree.2012.07.013 CrossRefPubMedGoogle Scholar
- Spulber DF (1982) Adaptive harvesting of a renewable resource and stable equilibrium. In: Mirman L, Spulber DF (eds) Essays in the economics of renewable resources. North-Holland Publishing, Amsterdam, pp 117–139Google Scholar
- Veitch D (2002) Eradication of Norway rats (Rattus norvegicus) and house mouse (Mus musculus) from Motuihe Island, New Zealand. In: Turning the tide: the eradication of invasive speciesGoogle Scholar
- Veitch CR, Clout MN (2002) Turning the tide: the eradication of invasive species (Proceedings of the International Conference on Eradication of Island Invasives). IUCN SCC Invasive Species Specialist Group, IUCN, Gland, Switzerland and CambridgeGoogle Scholar