Weeds on the web: conflicting management advice about an invasive non-native plant

Summary

1. Invasive non-native plants (INNPs) can have serious and widespread negative ecological and socio-economic impacts. It is therefore important they are managed appropriately. Within domestic gardens management decisions, which will tend to be made by individual members of the public, are likely to vary depending on (a) understanding of problems caused by INNP, and (b) knowledge of best practice.
2. Using content analysis, an approach seldom employed in an ecological context, this study analysed variation in internet-based information sources regarding INNP to determine how this collective discourse might influence risk perceptions and management decisions for domestic garden owners/managers. We used Japanese knotweed Fallopia japonica in the UK, as a case study, as it is one of the most ecologically and economically damaging INNP in the region. Our analysis categorized the types of author disseminating information about Japanese knotweed, the relative frequency of documents between author categories, and variation in content and style between and within author categories.
3. We identified five author categories: environmental NGOs, control companies, government, media and the property market. There was extensive variation in document structure, topics discussed, references and links to other sources, and language style; sometimes this variation was between author categories and sometimes within author categories. The most significant variation in topics discussed between author categories was indirect socio-economic problems, with control companies discussing these most. The number of pieces of legislation referenced and the proportion of militaristic words used were also highly significantly different between author categories. Some documents used neutral terminology and were more circumspect, whilst others were more forceful in expressing opinions and sensational.
4. The author category returning the highest number of documents was the subcategory local government, the shortest of which contained neither links to other information nor referenced any organizations. Further analysis of local government documents revealed conflicting advice regarding the disposal of Japanese knotweed waste material; confusion about this topic could result in decisions being made that spread Japanese knotweed further and are potentially unlawful.
5. The potential implications of our findings for the management of INNP in domestic gardens and societal perceptions of risks posed by INNP are discussed.
6. Synthesis and applications. To help prevent inappropriate management of invasive non-native plants (INNPs), for example Japanese knotweed Fallopia japonica in domestic gardens, we recommend that local and national authorities collaborate and work towards disseminating more consistent messages about (a) the potential socio-economic and ecological problems caused by INNP, whilst avoiding hyperbole, and (b) the most appropriate management techniques.

Introduction

Invasive non-native plants (INNPs) are a cause of significant global concern (CBD 2010; Simberloff et al. 2013). They can lead to biodiversity loss, alter ecological processes and impact ecosystem services (Vilà et al. 2011; Hulme et al. 2013). There are an increasing number of socio-economic impacts of INNP; for example, one study suggested losses from invasive weeds in the USA accumulate to at least $35 billion annually (Pimentel, Zuniga & Morrison 2005).

Planting in domestic gardens is a major introduction pathway for non-native plants, some of which subsequently become invasive (Groves, Boden & Lonsdale 2005; Smith et al. 2006). Invasions are doubtless facilitated by the high areal coverage and proportion of green space contributed by domestic gardens in many Western cities (Gaston et al. 2005; Loram et al. 2007; Gaston & Gaston 2011). Given that gardens can also play important roles in maintaining urban biodiversity and connecting otherwise fragmented urban habitats (Smith et al. 2006; Davies et al. 2008), it is important that INNP are managed in a way that minimizes their negative ecological and socio-economic impacts. In domestic gardens, people are largely free to manage the land ‘as they please’ (Gaston et al. 2005; Qvenild, Setten & Skår 2014). There is considerable variation in public awareness and understanding of INNP (Gozlan et al. 2013), and in the extent and form of INNP management (van Heezik et al. 2013). Furthermore, a recent study found that <20% of a sample of the UK public could identify Japanese knotweed Fallopia japonica (Robinson, Inger & Gaston 2016), an INNP of particular concern in this region (Gozlan et al. 2013). However, research into the challenges associated with managing domestic gardens to restrict INNP impacts is limited compared with equivalent research for public spaces (Qvenild, Setten & Skår 2014).

Knowledge and understanding of INNP and their management can be obtained by members of the public from multiple sources (Defra & GBNNSS 2009), such as government authorities, environmental professionals, media (formal and informal) and word of mouth. These sources are accessible – along with many others – via the Internet. As the Internet is one of the most regularly accessed sources of information in many Westernized societies today for a range of topics (Flanagin & Metzger 2000; Miller & Bartlett 2012), a significant amount of information on INNP in domestic gardens will be obtained in this way. This is reflected in a Google Trends analysis, which shows increasing frequency of Google searches for information on Japanese knotweed, a significant INNP, in the UK (Fig. 1).

Figure 1

Index of number of times ‘Japanese knotweed’ was searched in the UK using Google.

There is a vast amount of information on the Internet, which is likely to be diverse given the broad range of sources, authors (Flanagin & Metzger 2000; Miller & Bartlett 2012) and agendas to disseminate information. It can be trustworthy, accurate and written by specialists (these do not necessarily always correlate); it is also prone to misinformation, selective truths and marketing propaganda (Miller & Bartlett 2012). This can make it hard to find complete and accurate information on a topic, as has been demonstrated, for example, for information relating to human health (Berland et al. 2001). Variation can sometimes be, in part, due to lack of scientific consensus.

Given that a diverse range of authors with different motivations disseminate information about INNP, this raises the questions: How variable (or consistent) is this information? And how might this influence risk perceptions and management decisions by those people responsible for INNP in domestic gardens? Secondary information sources can be particularly influential on risk perceptions when people lack direct experience of phenomena. Variation in how people interpret these secondary sources, combined with their knowledge/experience of similar risks, can create social amplification of risk. This can manifest itself in individual behavioural decisions, which can have social and economic consequences (Pidgeon, Kasperson & Slovic 2003). Variation in source content – for instance, in the topics discussed or signposting to further information – could lead to confusion about the ecology of INNP, the ecological and socio-economic problems they cause, and the most appropriate management techniques. If different Internet sources provide conflicting views on management approaches, this makes it harder for homeowners to assess and decide on appropriate management decisions. If sources vary in their portrayals of the severity of the impacts of INNP, it could be unclear to the reader exactly why – or even if – they should be concerned. Furthermore, Internet sources may vary in the language they use, a subject that has received increasing attention in INNP discourse (Gobster 2005). For instance, militaristic language within invasive species discourse, which is not uncommon outside of scientific literature, may have the positive rhetorical power of motivating action against INNP, might also create inflated or inaccurate perceptions of the problems or limit the reader's confidence in a source's scientific credibility and objectivity (Gobster 2005; Larson 2005).

This study analyses how Internet-based information sources regarding INNP available to those responsible for managing them in domestic gardens vary and then considers how this collective discourse might influence risk perceptions and management decisions. Understanding variation in Internet discourse about INNP can help governmental authorities target and improve their communication, thereby potentially improving societal understanding of current best management practices for INNP in domestic gardens, and subsequently reducing their spread and impacts. Japanese knotweed in the UK is used as a case study as this is a region in which it causes widespread and serious ecological and socio-economic damage (Engler, Abt & Buhk 2011; Gozlan et al. 2013). The method we use, however, can be applied to Internet-based discourse of other INNP, and in different countries.

Japanese knotweed in the UK constitutes a valuable case study for several reasons. Originally introduced for ornamental purposes in the mid-1800s, Japanese knotweed has since become widespread and problematic (Beerling, Bailey & Conolly 1994). It spreads highly efficiently by vegetative reproduction (Engler, Abt & Buhk 2011) and can regenerate from just a few grams of rhizome (Sasik & Pavol 2006). It causes ecological disturbance by out-competing other plants (Engler, Abt & Buhk 2011), and through allelopathy, suppressing their growth (Dommanget et al. 2014). This has consequences for organisms at other trophic levels, by altering habitat structure or modifying availability of food sources (Engler, Abt & Buhk 2011). Japanese knotweed is estimated to cost the UK economy £165 million a year (Williams et al. 2010). It is named in several pieces of UK legislation, most notably the ‘Wildlife and Countryside Act’ (1981), under which it is an offence to plant or cause Japanese knotweed to spread in the wild, and the ‘Environmental Protection Act’ (1990), which requires correct waste material disposal methods (Bailey & Conolly 2000; EA 2006). Recently, Japanese knotweed has received extensive media coverage (Gozlan et al. 2013) relating to problems caused by its rapid spread.

Materials and methods

Document selection

Documents for analysis were obtained using Google (www.google.co.uk) because it is the most popular search engine (Purcell, Brenner & Ranie 2012). Two searches were conducted, first for ‘Japanese Knotweed’ then for ‘Fallopia japonica’. The first 100 websites from each search (omitting sponsored links and duplicates) that provided management advice, general information or news articles were saved – these are referred to as ‘documents’ hereafter. Where websites contained multiple pages relating to Japanese knotweed, and these were immediately identifiable from the website's menu, all relevant pages were combined within one document for analysis. From the total 200 documents collected from both Google searches, only UK-based sources (n = 113) were extracted for inclusion in the final analysis.

Author categories were developed and assigned to each document (Table 1). Documents that did not fit into these categories (n = 9) were omitted from further analysis. For example, Wikipedia has multiple authors and is frequently amended.

Table 1. Number of documents within author categories and subcategories

Author category	Number of documents
1. Environmental NGO	8
2. Control Company	29
3. Government organizations
a) National	8
b) Local	30
4. Media
a) Mainstream – online newspapers	11
b) Other (blogs etc.)	11
5. Property market	8

Document analysis

Methods employed in this study build on content analysis, an approach frequently used in social science research to analyse the content of communications (in this case written text) to ascertain meaning and potential consequences (Bernard 2011; Krippendorff 2013). This method involves identifying ‘themes’ or ‘codes’ within text which are replicable and can therefore be analysed quantitatively (Bernard 2011).

First, the primary coder (BR) read the documents in their entirety to become familiar with the content, and then reread and manually codified the documents systematically and iteratively into themes using Nvivo 10.0 (QSR 2012). Both inductive and deductive approaches were used to develop codes, drawing from existing literature on INNP discourse and environmental management. Codes were then refined through discussions between authors and a focus group of both social and natural scientists. To check for coding bias, 25% of randomly selected documents from each author category were also coded by a second coder (SC) for all elements of analysis. Agreement was assessed using Cohen's kappa, a measure of agreement between the two coders that accounts for agreement that would occur by chance (Bernard 2011). High agreement (>70%) indicated that the method is robust. Coding was subsequently modified through discussion between coders as required.

Data were primarily analysed quantitatively (methods outlined below). These results are illustrated with qualitative data to allow a more focused evaluation of emerging themes and patterns, particularly: the context of the text, relationships between codes, diversity in portrayal of the severity of the impacts of INNP, exploration of the ways this collective information is potentially conflicting and/or confusing, and possible implications for management decisions. Quotes are ascribed to author categories only, as the intention is not to discredit advice given by a particular source.

Analysis of all documents

The first part of the analysis considered all qualifying documents from the Google searches (n = 104) and was structured so that it might be applicable to any problematic INNP in domestic gardens.

The following codes were developed:

1. Number of words

   The total number of words per document was calculated as a coarse measure of the amount of information a document contained. Although not a direct measure, this helps to identify outliers (e.g. very short documents) and was necessary for further analysis (proportion of words of language type). Word count included titles, subtitles and picture captions, but excluded unrelated links and adverts.

2. Topics discussed

   The topics discussed within different documents potentially affect the reader's knowledge and understanding about what the problems of INNP are, whether and why they should be concerned, and the most appropriate management methods. Coded topics were totalled up for each document, but repetition of codes within a document was discounted. See Appendix S1 in Supporting Information for full details of codes.

   1. Problematic traits: These were the biological traits of Japanese knotweed that make it problematic in its introduced range. The codes were as follows: Japanese knotweed… grows fast; grows tall; can regenerate from small fragments; can survive extreme conditions / grow in a wide variety of habitats; has rhizomes that can survive extreme conditions; has rhizomes that can survive extended dormancy periods; and has roots that extend a long way vertically and/or horizontally.
   2. Problems caused: A ‘problem’ was defined as any (real or potential) negative consequence of Japanese knotweed presence. The primary coder identified an exhaustive list of problems identified within the documents. These were then categorized into three broad categories of problem, developed iteratively during focus group discussion (see above).

The three broad types of problem were as follows:

1. Direct socio-economic problems directly affecting the human environment by physically altering either natural or human-made structures. The codes were as follows: Japanese knotweed can… damage gardens; increase flood risk; damage hard human-made structures; reduce visibility; trap litter and vermin; have a negative aesthetic impact; cause a trip hazard; impact recreational activities; cause a fire hazard; and its presence on riverbanks can lead to soil erosion.
2. Indirect socio-economic problems arise because direct socio-economic effects of Japanese knotweed have knock-on second-order impacts on the social environment or have potential associated economic costs. The codes were as follows: Japanese knotweed… is costly to eradicate or control; can reduce land / property value; can cause mortgage problems; can cause legal disputes; can cause delays to planning applications and building development projects; and can cause insurance problems.
3. Negative ecological impacts: These are effects that have a negative impact on other flora and fauna, biodiversity or ecological processes. The codes were as follows: Japanese knotweed can… cause a change in biodiversity; impact trophic reactions; have a negative effect on animals; and have a negative effect on plants.

3. References and links to other sources

   During analysis, it became clear that there was large variation in specific organizations signposted, links provided to further information and INNP legislation referenced. It is valuable to consider these components of a document for two reasons. First, they provide the reader with additional sources of information, allowing them to crosscheck facts and develop their understanding of the topic (Sillence et al. 2006). Secondly, they may be rhetorical tools used by different sources to convince the reader of the validity of the information contained within the document.

   1. Specific organizations signposted and links provided: These were totalled up for each document. Examples include the following: Environment Agency, Defra (Department of Environment, Food and Rural Affairs), control companies and scientific references.
   2. Legislation referenced: The total number of pieces of legislation quoted or referred to (specifically relating to Japanese knotweed) were totalled up for each document.

4. Language Style

   Two types of language style were analysed, informed by literature on INNP and environmental discourses: science/technology and militaristic. For analysis of language, style words were initially drawn from examples used in similar analyses (Webb & Raffaelli 2008 for science/technology; Larson 2005 for militaristic words). Additional words identified within these styles by the coders were added to the analysis. Words were coded as many times as they arose, but omitted if they had negative qualifiers or were part of a name. The proportion of words within a document using these terminologies was calculated.

   1. Science/technology terminology: Words relating to science and technology may contribute towards an increase in perceived legitimacy of the source (Webb & Raffaelli 2008). Words coded were as follows: ecologist, ecology, biologist, biology, scientific, data, professor, research, monitoring, evaluation, evidence, measure, record and rhizome. Scientific names of flora and fauna were also coded for.
   2. Militaristic terminology: Militaristic words are common in invasive species discourse and may influence the readers’ perception of invasive species (Larson 2005). Words coded were as follows: war, enemy, weapon, attack, offensive, battle, solider, fortress, conquer and guerrilla.

Statistical analysis

Differences in codes (the response variable) between the five main author categories (explanatory variable) were analysed using R 3.0.2 (R 2009). Subauthor categories were combined to increase sample size and statistical power of analyses (e.g. mainstream media and other media combined, see Table 1). For the response variables of word count (log-transformed) and proportion of words per document with a scientific or militaristic association (arcsine-square-root-transformed) (elements: 1a, 4a and 4b of part (i) of analysis respectively), linear models were developed. Linear models were interpreted by comparing the full model with the null model using the F-statistic and P-value.

The following response variables were explored using generalized linear models (Poisson distribution) and interpreted by comparing the full model with the null model using the chi-squared statistic and P value: number of occurrences of (i) problematic traits, (ii) direct socio-economic problems, (iii) indirect socio-economic problems, (iv) ecological impacts, (v) specific organizations signposted and links provided and (vi) legislation referenced (elements: 2a, 2b.i, 2b.ii, 2b.iii, 3a and 3b of part (i) of analysis, respectively). Author category was included as a fixed factor and document ID as a random (intercept) factor to account for overdispersion. In all models, environmental NGO was the base (reference) category.

To explore differences in codes within author categories, the error around the mean was assessed. The local government category was chosen as a focus for analysis for within-author variation as it contained the greatest number of documents, and it is reasonable to assume that local government publications are a regularly consulted and trusted source of information for environmental issues as they provide information, advice and authority for a broad range of environmental and societal issues. For the same reasons, local government documents were also chosen as a focus for part two of the analysis.

Analysis of local government documents' disposal advice

It is important to provide clear advice about waste disposal of Japanese knotweed, as it can regrow from small fragments of rhizome (Sasik & Pavol 2006), and incorrect disposal of waste material can result in further spread of this plant. Therefore, the second section of analysis examined discussion of waste disposal within local government documents and whether the information therein was conflicting and/or confusing.

Word searches were used to identify sections of local government documents that contained advice on waste material disposal. Words searched were as follows: disposal, rubbish, landfill and waste. Relevant sections were then reanalysed and codes relating to specific disposal advice were developed iteratively. Authors then evaluated whether and how any disposal advice was conflicting.

Results

Of the 104 documents included in the final analysis, five author categories were identified, with two broken down into subcategories (Table 1; see Appendix S2 for author categories descriptions and list of documents). The local government and control company categories contained the most documents. Sources of documents were diverse, for example mainstream media, contained articles from a range of sources (1 Financial Times, 4 BBC, 1 Daily Mail, 3 Guardian, 2 Daily Telegraph) and local government documents represented a large geographic range, including both urban and rural areas.

Analysis of all documents

Number of words per document was not statistically different between author categories (F_[4,99] = 1·82, P = 0·132; Fig. 2a, see Appendix S3 for raw data). The average length of a document was 1575·6 (SE = 276·3) words: the shortest was 188 words and the longest 22 305. Within documents from local government sources, there was considerable variation in word count, with the shortest only 266 words, the longest 9764. The shortest local government document contained neither links to other information nor referenced any organizations.

Figure 2

Variation between author categories in mean number of (a) words (log-transformed); (b) problematic ecological traits mentioned; (c) direct socio-economic problems discussed; (d) indirect a socio-economic problems discussed; (e) number of ecological problems discussed; (f) specific organizations referenced and links provided; (g) pieces of legislation referenced; (h) proportion of scientific words; and (i) proportion of militaristic words. Stars mark significant differences between author categories.

The numbers of problematic ecological traits mentioned per document was approaching a significant difference between author categories and had large standard errors around the mean ( = 9·05, P = 0·060; Fig. 2b). All local government documents mentioned at least one problematic trait, with a maximum of six (Fig. 3a). The problematic ecological trait mentioned most frequently, in 70·2% (n = 73) of documents, was that Japanese knotweed could regrow from a small fragment of rhizome.

Figure 3

Frequency distributions of coded content or ‘themes’ within local government documents of number of (a) problematic ecological traits discussed; (b) direct socio-economic problems discussed; (c) indirect socio-economic problems discussed; (d) ecological problems discussed; (e) specific organizations referenced and links provided; and (f) pieces of legislation referenced.

The number of direct socio-economic problems discussed per document was marginally significantly different between author categories ( = 9·58, P = 0·048; Fig. 2c). In 30% (n = 9) of local government documents, no direct socio-economic problems were mentioned (Fig. 3b). The most frequently mentioned direct socio-economic problem, in 78·8% (n = 82) of documents, was that Japanese knotweed could cause damage to hard surfaces. However, portrayal of the severity of this issue was diverse. At one end of the spectrum, some authors implied that this is extremely problematic, for example, claiming that Japanese knotweed can ‘burrow into [building] foundations’ (media document). Other authors were more circumspect, explaining that damage caused depends on the type of structure; damage to temporary structures such as greenhouses, for example, was reported as much more likely than damage to the foundations of houses. One property market document discussed this issue in depth and suggested that perceptions are often based on ‘misunderstandings and overreactions’.

The number of indirect socio-economic problems discussed per document was highly significantly different between author categories ( = 27·75, P < 0·001; Fig. 2d). Control companies mentioned these problems most frequently and environmental NGOs least. No indirect problems were mentioned by 30% (n = 9) of local government documents (Fig. 3c). The indirect socio-economic problem mentioned most frequently, in 51·9% (n = 54) of documents, was that Japanese knotweed is costly to control. Some documents were vague about cost – one local government document simply noted ‘it can be expensive’, whereas others attempted to quantify control/eradication costs. Estimates ranged from £1 per m² (local government document), to ‘well over £1000’ per m² (national government document); however, the latter represents costs to development sites and the former to domestic property. Other sources, including multiple control company documents, stated that the presence of Japanese knotweed can add 10% to the budget of a development. The cost of eradicating Japanese knotweed was quoted as ‘£1·56 billion’ by 12·5% (n = 13) of documents, across all except national government, author categories, referring to the calculations of a Defra (2003) report.

The number of ecological problems discussed per document was not significantly different between author categories and had large standard errors around the mean ( = 6·63, P = 0·157; Fig. 2e). No ecological problems at all were mentioned in 16·7% (n = 5) of local government documents (Fig. 3d). The most common ecological problem, mentioned in 56·7% (n = 59) of documents, was that Japanese knotweed out-competes other plants. Some documents communicated this in neutral language; for example, one local government document noted that Japanese knotweed ‘often out-competes existing plant communities’. Others expressed this more sensationally, for example ‘it smothers rival plants’ (mainstream media document).

The number of organizations referenced per document did not differ significantly between author categories ( = 4·59, P = 0·332; Fig. 2f). The most commonly referenced organization, mentioned by 75·9% (n = 79) of documents, was the Environment Agency, a UK government organization responsible for national environmental protection issues. In 26·9% (n = 28) of all documents, reference was made or a link was provided to the guidelines produced by the Environment Agency in 2006 (updated 2013). These guidelines provide advice aimed at the development and haulage industry, but are freely available and also relevant to homeowners. A link to the Environment Agency guidelines was provided by 20% (n = 6) of local government documents. No links to organizations were mentioned or provided by 6·7% (n = 2) of local government documents and 36·7% (n = 11) referenced only one (Fig. 3e).

The number of pieces of legislation referred to per document was highly significantly different between author categories ( = 39·49, P < 0·001; Fig. 2g). Control company and government documents referenced most legislation, and media and environmental NGOs the least. Within local government documents, 10% (n = 3) mentioned no legislation (Fig. 3f). Some documents simply mentioned the name of legislation and a brief description, for example: ‘Wildlife and Countryside Act (1981) – It is an offence to plant, or cause knotweed to grow in the wild’ (local government document). Others provided more detail about the implications of legislation in relation to management decisions and the potential consequences of ignoring this. For example, one local government document explained that offences under this legislation include ‘fly-tipping’ (illegally dumping) material containing Japanese knotweed.

The proportion of science/technology words per document was not statistically significantly different between author categories (F_4,99 = 0·60, P = 0·661, Fig. 2h). However, the two documents with the highest proportion of scientific words contained very few words overall, at 266 and 283 words (local government document and environmental NGO document, respectively). Further examination of these documents revealed that despite the high proportion of scientific language, very little in the way of scientific information was communicated.

The proportion of militaristic words per document, however, was significantly different between author categories (F_[4,99] = 5·23, P = 0·001; Fig. 2i), with media documents containing the most. Examples of more extreme use include ‘reclaiming the war torn landscape’ or ‘more like a guerrilla force than a thug’ (both other media documents). Only 10% (n = 3) of local government documents used militaristic terminology.

Analysis of local government documents' disposal advice

All local government documents provided some advice on disposal of Japanese knotweed (n = 30). Some of this advice, however, related only to prohibited actions, for example ‘Cut material must not be removed from site and cannot be composted’, thereby providing little direction as to best management practice. Other documents gave very detailed advice, subdivided into recommended practices for onsite and offsite disposal. Contact details for companies that dispose of Japanese knotweed were provided by 13·3% (n = 4) of local government documents. One additional document provided locations of regional waste disposal facilities.

Advice about composting was identified as a point of possible confusion. Advice as to whether or not Japanese knotweed should be composted at home was provided by 33·3% (n = 10) of local government documents. Three documents stated it must not be composted at home and one that it can be composted at home, providing, however, no further detail about how and when. The remaining six documents explained that Japanese knotweed can be composted at home, but specified that caution should be exercised (e.g. by only composting dried stems).

Analysis revealed repetition of sentences, or even paragraphs, between local government documents. Some referred the reader to other local government documents. For instance, 16·7% (n = 5) of local government documents directed readers to advice given by Cornwall Council.

Discussion

Management of INNP in domestic gardens is likely to vary depending on understanding of the problems they cause and knowledge of best practice. This study sampled Internet sources of information regarding Japanese knotweed, a particularly problematic INNP occurring within domestic gardens in the UK. Analysis considered the types of authors disseminating such information, their relative frequency, variability of content and style between and within sources and whether the collective discourse is potentially conflicting and confusing. Here, we consider the impact the findings might have on risk perceptions and management decisions about INNP in domestic gardens.

The author category containing the highest number of documents was local government, followed by control companies, then media, suggesting that these sources are likely to be accessed more frequently than other author categories. Whilst local government and control company documents are perhaps more likely to be consulted by people seeking information about Japanese knotweed, previous research suggests public understanding of invasive non-native species more generally, and on a day-to-day basis, is largely based on media reports (McNeely 2001).

The most significant variation in topics discussed between author categories was indirect socio-economic problems (e.g. devaluing property). Many of these are emerging issues at present specific to Japanese knotweed in the UK. It is important to raise awareness about the potential socio-economic problems associated with Japanese knotweed, as this facilitates development and implementation of mitigation strategies. However, incomplete or inaccurate communication about them could inflate or attenuate risk perception. Unfortunately, maintaining clear and accurate messages can be challenging, as scientific research on socio-economic problems associated with Japanese knotweed is lacking. Conversely, perhaps a greater scientific understanding of the ecological problems contributed to the lack of a variation in discussion of this topic between documents

For some authors, focusing on potential socio-economic problems associated with Japanese knotweed could be an advantageous rhetorical tool. For example, control companies referred most frequently to potential indirect and direct socio-economic problems. They also identified a high number of ecological traits, referenced more legislation than other author categories, and multiple control companies highlighted the potential extremities of Japanese knotweed removal costs. The high number of control company documents in the results is likely due to their use of search engine optimization to persuade readers to employ their services. This potentially produces both positive and negative outcomes. If property owners are convinced of the potential severity of the risks associated with Japanese knotweed, they may be more likely to employ professional control companies, thereby limiting the potential for mismanagement, further spread and damage. However, small-scale occurrences of Japanese knotweed in domestic gardens may not require professional attention to be effectively controlled. Consequently, overemphasis of risks may result in unnecessary anxiety and expenditure by householders, and inflate societal perception of the risks posed by INNP.

Some local government documents provided in-depth discussion about the problems caused by Japanese knotweed and detailed advice about how to manage it in domestic gardens. Overall, however, the information provided was highly variable in word count, topics discussed, links to further references and legislation discussed. Furthermore, conflicting management advice was identified within local government documents regarding how to dispose of Japanese knotweed waste material. This raises the concern that if those responsible for domestic gardens consult only the website of their local government authority, the quality and clarity of the information received could vary geographically.

The Environment Agency (EA) guidelines was the longest document in the analysis, and also the source most frequently referenced by others. This is likely to be a reliable source for several reasons. First, the EA works closely with other government agencies and departments, local councils and communities (EA 2014), and indeed, the report was written with input from Defra and National Rail. This collaborative approach to dissemination arguably strengthens the validity and legitimacy of information given. Secondly, the document has undergone several revisions to keep it up to date. Finally, government organizations have a formal responsibility to provide accurate information. The frequent citation of this document is likely due to the extensive information it contains, and because people in the UK generally trust government sources to provide accurate information (Briggs et al. 2002). Although this document may be too detailed for many responsible for managing domestic gardens, useful information can be extracted, and it should be signposted or summarized where possible.

As previously mentioned, the document style adopted depends on the author's motivation and is in part likely due to lack of scientific consensus on the severity of impacts. Some documents discussed the potential problems of Japanese knotweed in neutral terms, whilst others expressed opinions more forcefully, and demonstrated evidence of hyperbole. Media documents used the most militaristic language and stronger rhetoric. Presumably these are journalistic tools to create more stimulating stories; however, concerns have been expressed that inappropriate language could have multiple consequences, including loss of scientific credibility and inaccurate societal perception of risk of INNP (Larson 2005). Media content analyses covering a range of human–wildlife conflicts have highlighted how framing of these issues can influence risk perceptions and behaviours (Jacobson et al. 2012; Sakurai, Jacobson & Carlton 2013). Increased communication between scientists and the media has been suggested as a way to mitigate these impacts (Barua 2010).

Information from such a wide range of sources will undoubtedly be diverse given the broad range of authors, motivations for writing, and economic and political interests. This diversity, combined with variation in visual design, trust and credibility individuals assign to websites (Sillence et al. 2006), the number of articles read on the subject (many Internet users only read one or two; Miller & Bartlett 2012), and that many people will obtain information about INNP from non-Internet sources, all heighten the potential for variation in societal understanding of the problems posed and risk, potentially leading to social amplification of the risks (Pidgeon, Kasperson & Slovic 2003).

Given that information on INNP in domestic gardens provided by many sources cannot be regulated, it is important that government authorities provide clear, detailed and consistent information on this topic. This could be accomplished by providing balanced and neutral discussion of a range of potential ecological and socio-economic problems caused by INNP, communicating clear and consistent messages about appropriate management methods, directing readers towards key management guidelines, highlighting a range of relevant legislation, and providing detailed information on waste disposal methods and local waste disposal facilities. Greater collaboration between local and national governmental departments, and development of a generic template that local government authorities could adapt for their own use, could help to deliver more consistent messages. Some evidence of coordination was observed, such as inputs from multiple stakeholders for some documents, cross-referencing between local and national government documents, and referencing advice from authorities who have invested more in researching, recording and promoting public understanding of INNP (e.g. Cornwall Council). Furthermore, it is likely that a lack of scientific consensus and knowledge gaps are contributing towards variation in Internet-based discourse regarding Japanese knotweed. It is important that the scientific studies regarding INNP that have been done are reviewed regularly and made widely accessible. A good example of this is the series about INNP by the Canadian Journal of Plant Science, which reviews unpublished and published literature.

Despite content analyses being common and well-developed within social scientific research, such analyses in environmental management discourse are relatively scarce (Webb & Raffaelli 2008). This method could be used to analyse Internet-based discourse of other INNP, and in different countries. The results could guide recommendations to improve societal knowledge, understanding and best-practice management, thereby decreasing the impacts and spread of INNP.

Acknowledgements

This project was funded as part of the Wildlife Research Co-Operative between the University of Exeter and the Animal and Plant Health Agency. We are grateful to K. Abernethy and two anonymous reviewers for comments.

Data accessibility

Data are available in Appendix S3 in Supporting Information.