Public preferences for the management of different invasive alien forest taxa

Abstract

Invasive alien species (IAS) require management to mitigate their impact on ecosystems. The success of management decisions often depends on whether they are socially acceptable and to what extent people are willing to be actively involved in an early warning and rapid response system (EWRR). We administered a nation-wide public poll to assess people’s knowledge on plant, insect and fungal IAS; their perception of IAS as an environmental problem; and their support for different IAS management measures. Most respondents (76%) knew the term IAS, and more than half (62%) provided a correct definition. Species with more media attention and those that are easily visible are more frequently identified correctly. Almost all respondents (97%) support an EWRR system; however, there is heterogeneity in terms of the types of actions people approve of. Non-lethal measures garner more support than lethal ones. Gender and previous knowledge also affect the level of agreement. The willingness-to-pay question largely confirmed this, as people were divided into four classes according to their preferences for either biological, mechanical or chemical measures to control IAS; completeness and location of removal; and having an EWRR established. Mechanical removal is the most preferred treatment in two of the four classes, and complete removal is preferred over partial removal in one of the four classes. Having an EWRR is consistently supported in all classes, and removal in urban areas is preferred over removal in forestland in only one class.

Appendices

Appendix 1: The extended version of the questionnaire (the basic version had no discrete choice experiment part)

Appendix 2: Questions 1–8 comprising the first four parts of the questionnaire with basic descriptive statistics and statistical method used in the analysis of associated responses

Questionnaire part	Question (numbered)	Descriptive statistics				Statistical methods used in analysis
General awareness of IAS-related problems	1. Have you already heard of the term IAS?	76% (yes); 24% (no)				Binomial logistic regression
	2. Do you think IAS pose a problem?	83% (yes); 17% (no)
Support for establishing an EWRR and willingness to act	3. Do you support the establishment of an EWRR?	97% (yes); 3% (no)
	4. If IAS appeared on your land, would you be willing to remove it?	96% (yes); 4% (no)
	5. Would you be willing to report the finding of an IAS to the relevant institution?	89% (yes); 11% (no)
Awareness of IAS and the layperson’s recognition	6. Which IAS have you heard of?					Exploratory factor analysis (extraction by principal component analysis) and Binomial logistic regression
	Chestnut gall wasp (Dryocosmus kuriphilus)	47% (yes); 53% (no)
	Western conifer seed bug (Leptoglossus occidentalis)	9% (yes); 91% (no)
	Japanese knotweed (Fallopia japonica)	47% (yes); 53% (no)
	Goldenrods (Solidago spp.)	49% (yes); 51% (no)
	Eutypella canker (Eutypella parasitica)	24% (yes); 76% (no)
	Tree of heaven (Ailanthus altissima)	11% (yes); 89% (no)
	Common ragweed (Ambrosia artemisiifolia)	68% (yes); 32% (no)
	Chalara ash dieback (Hymenoscyphus fraxineus)	44% (yes); 56% (no)
	7. Which IAS is in the photo?
	Chestnut gall wasp (Dryocosmus kuriphilus)	35% (correct)
	Western conifer seed bug (Leptoglossus occidentalis)	35% (correct)
	Japanese knotweed (Fallopia japonica)	17% (correct)
	Goldenrods (Solidago spp.)	36% (correct)
	Eutypella canker (Eutypella parasitica)	22% (correct)
	Tree of heaven (Ailanthus altissima)	13% (correct)
	Common ragweed (Ambrosia artemisiifolia)	43% (correct)
	Chalara ash dieback (Hymenoscyphus fraxineus)	18% (correct)
Support for IAS-related management measures	8. How strongly do you support different measures of control of IAS?	Do not	Partially	Mostly	Fully	Exploratory factor analysis (extraction by principal component analysis) and Cumulative ordinal logistic regression
	Using poison—animals	52%	29%	12%	6%
	Lethal injection—animals	41%	29%	18%	12%
	Hunting—animals	19%	34%	29%	19%
	Sterilization—animals	8%	19%	34%	39%
	Insecticides—animals	31%	33%	22%	14%
	Shooting—animals	40%	33%	16%	11%
	Transfer into shelters—animals	18%	27%	28%	26%
	Mowing/cutting—plants	4%	19%	16%	62%
	Excavation—plants	2%	16%	11%	71%
	Herbicides—plants	37%	20%	27%	18%
	Natural enemies—plants	11%	20%	27%	42%

Appendix 3: Results of factor analysis (extraction by principal component analysis) for questions 6–8

See Tables 7, 8 and 9.

Table 7 Factor loadings for the general level of information on IAS

Table 8 Factor loadings for the recognition of IAS

Table 9 Factor loadings for the support of control measures

Appendix 4: Theoretical framework of the choice experiment

Lancaster’s consumer theory (Lancaster 1966) provides grounding to the choice experiment technique by establishing that the utility of a good can be broken down into the utilities of its individual attributes. Furthermore, the random utility model (RUM) derived from the work of (Luce 1959) and (McFadden 1973) provides a basis for empirical modelling of respondents’ choices, indicating trade-offs among the attributes of the assessed good (Bateman et al. 2002; Hanley et al. 2001). Accordingly, the goods being investigated are described by bundles of attributes. Levels (quantitative or qualitative values) of those attributes can be varied, and by doing so, different combinations of attribute levels can be generated and grouped into so-called alternatives. Each respondent is presented with a set of alternatives, commonly organized into several consecutive choice sets, and asked to pick the ones he/she likes best (i.e. maximizes his/her utility) from each choice set. This is done via various survey formats. One of the alternatives in each choice set usually presents the current situation indicating a ‘scenario’ without any changes, which is commonly referred to as a business-as-usual (BAU) alternative. Other alternatives represent situations where the attribute levels are changed due to some hypothetical actions which we wish to investigate. Each alternative also has an additional cost attribute indicating the hypothetical amount of money needed to be allocated for implementing the changes of the attributes. Obviously, the BAU alternative has zero cost assigned. A respondent selecting among alternatives is implicitly making trade-offs between the levels of attributes across alternatives (Hensher et al. 2005). Having a cost attribute makes it possible to calculate the marginal values of the changes in attribute levels.

Empirical analysis of choices grounds on RUM, which states that utility U obtained by respondent i by choosing an alternative $j \left( {j = 1, \ldots ,I} \right)$ and conditioned on being in class c can be modelled as a function (indirect utility function) of a deterministic component V, which can be observed by the researcher and related to attributes, and of a random component ε. The latter is an error term comprising non-observable features that affect choices of respondents and is of a type 1 extreme distribution:

$$U_{ij|c} = V_{ij|c} + \varepsilon_{ij|c} = \beta_{c} x_{ij} + \varepsilon_{ij|c} ,$$

where x is the vector of observed attributes, β is a parameter vector (Boxall and Adamowicz 2002).

The deterministic part of utility $V_{ij|c}$ can be divided into two parts, one related to a respondent’s specific characteristics, such as socio-demographic characteristics, perceptions, attitudes etc., and the other two choices of alternatives with respect to levels of the attributes (Boxall and Adamowicz 2002; Swait 1994). The probability of respondent i choosing alternative j conditional on being in class c is:

$$Pr_{ij|c} = \mathop \sum \limits_{c = 1}^{C} \left( {\frac{{e^{{\theta_{c}^{\prime } z_{i} }} }}{{\mathop \sum \nolimits_{c = 1}^{C} e^{{\theta_{c}^{\prime } z_{i} }} }}} \right) \times \left( {\frac{{e^{{\beta_{c}^{\prime } x_{ij} }} }}{{\mathop \sum \nolimits_{k = 1}^{I} e^{{\beta_{c}^{\prime } x_{ik} }} }}} \right),$$

where z_i is a vector of respondent-specific characteristics and $\theta_{c}^{\prime }$ are class-specific coefficients to be estimated. The first part of the right-hand side is the probability of a respondent being in class c, while the second part is the probability of choosing alternative j conditioned on membership in class c.

$\theta_{c}^{\prime }$ and $\beta_{c}^{\prime }$ in the LCLM are jointly estimated by employing the maximum likelihood estimation and are subsequently used to explain respondent choices. The number of classes needs to be determined prior to model estimation, which can be done through different approaches. Setting the number of classes is not straightforward, as there is no precise approach for choosing the optimal number of classes (Milon and Scrogin 2006). Some authors (Boxall and Adamowicz 2002; Scarpa and Thiene 2005) recommend using statistical information criteria such as Bayesian information criteria (BIC) (Schwarz 1978) and Bozdogan-Akaike information criteria (AIC3) (Bozdogan 1987) and also recommend accounting for the plausibility (signs of the parameters and their significance) of the results and the size of classes.

Appendix 5: A representative choice set

Appendix 6: Numerical estimation results of binomial and ordinal regression of questions 1–8

See Tables 10, 11, 12, 13 and 14.

Table 10 Estimation results of binomial regression for the questions ‘Have you already heard of the term alien species?’ and ‘Do you think invasive alien species pose a problem?’

Table 11 Estimation results of binomial logistic regression models for ‘Do you support the establishment of an EWRR?’; ‘If IAS appeared on your land, would you be willing to remove it?’; and ‘Would you be willing to report the finding of an IAS to the relevant institution?’

Table 12 Estimation results of models for ‘Which IAS have you already heard of?’

Table 13 Estimation results of models for ‘Can you recognize the IAS in the photo?’

Table 14 Estimation of the model for ‘How strongly do you support different measures for IAS management?’