The OECD (Organization of Economic Cooperation and Development) named the Netherlands “an international example” of water resource management in their 2014 report, but warns against “a striking awareness gap among Dutch citizens related to key water management functions, how they are performed and by whom” (OECD, 2014, p. 21). The main causes of this awareness gap are the absence of major water calamities in the past 60 years and the improvement of water quality over the past decades. Dutch citizens take the excellent water resource management for granted (OECD, 2014), which poses social challenges for Dutch water resource management. Citizens' behaviour counteracts efforts of water authorities; flood defences are violated by property development and civic pollution is common (OECD, 2014). Citizens and interest groups do not recognize water threats (Tielrooij, 2000), which causes a decreasing support for investments in flood defence and water quality management (OECD, 2014; Tielrooij, 2000; UvW, 2015a). Other countries also experience a “lack of citizen concern about water policy and low involvement of water users' associations” (OECD, 2011, p. 60). Half of the reviewed OECD countries across the globe face such challenges, including Chile, Italy, Korea and Mexico (OECD, 2011, p. 61). The Dutch water authorities (Unie van Waterschappen, UvW) governing body concluded that collaboration with other government layers, industry, interest groups and citizens is needed (UvW, 2015a). The UvW envisions increased public participation, with citizen science as a form of such participation (UvW, 2015b). In addition to raising awareness, citizen science could contribute to data collection and help water authorities to enhance their monitoring programmes, particularly with respect to the Water Framework Directive.

Definitions of citizen science can be narrow and focussed solely on data collection for academic purposes. Silvertown (2009) describes citizen science in a broader perspective, applicable to citizen science in practitioners' activities considered in this article. “Today, most citizen scientists work with professional counterparts on projects that have been specifically designed or adapted to give amateurs a role, either for the educational benefit of the volunteers themselves or for the benefit of the project. The best examples benefit both” (Silvertown, 2009, p. 467). To prevent confusion, a distinction is made within this category of professional counterparts (Silvertown, 2009). The professional counterparts in Silvertown's definition include scientists, conservation professionals and government practitioners. We define scientists as those involved in academia. Conservation professionals are those working for nature managers or conservation organizations. Government practitioners are defined as those working at a government agency or at the local government level.

Citizen science in water resource management is upcoming but lingering on the verge of a breakthrough (Buytaert et al., 2014; Cohn, 2008; Fraternali et al., 2012). The rise of robust, cheap and low-maintenance sensors enhances opportunities for citizen science in the complex arena of water resource management (Buytaert et al., 2014). New hydrological modelling frameworks (e.g. Clark et al., 2015) and specifically uncertainty quantification (e.g. Shoaib et al., 2016) can allow for more effective design of citizen science campaigns to systematically reduce model uncertainty, as well as more effective use of citizen-collected data. These data are often considered highly uncertain by practitioners, as they indicated in the exploratory interviews held at the start of this research. Fraternali et al. (2012) give an overview of the potential of amateurs taking part in data collection, data analysis and the process of decision-making in water resource management.

This potential is also recognized in the Netherlands. In November 2014 water authority Delfland organized a workshop

Part of the symposium “De fysieke Digitiale Delta” (the Physical Digital Delta; see also http://www.digitaldelta.nu/en/events/item14).

Citizens' motivations have been studied extensively in a diverse set of citizen science projects, such as online crowdsourcing (e.g. Chandler and Kapelner, 2013; Raddick et al., 2010; Rogstadius et al., 2011), environmental monitoring (e.g. Hobbs and White, 2012; Roy et al., 2012) and meteorology (e.g. Gharesifard and Wehn, 2016). It has been acknowledged that the idea of “the public” does not exist (e.g. Varner, 2014), since “the public” consists of a wide variety of people with different backgrounds, interests, traits, values and beliefs. Nevertheless, existing studies of citizen science in field and online projects, despite this diversity, reveal the same dominant motivations over a wide range of projects and participants. The most mentioned reasons for citizens to engage in citizen science are because they think it is fun, because the topic interests them, and because the topic matters to them, e.g. they want to contribute to science or nature conservation (e.g. Chandler and Kapelner, 2013; Hobbs and White, 2012; Raddick et al., 2010; Rogstadius et al., 2011; Roy et al., 2012). Citizens are motivated to continue to contribute by (increasing) the extent of their involvement (Rotman et al., 2012; Roy et al., 2012), offering feedback concerning the work at three levels (individual contribution, group contribution and the use of data) and building a relationship based on trust between scientists and citizens (Rotman et al., 2012). The importance of trust is stressed by authors in the field of water management as well (Buytaert et al., 2014; Gharesifard and Wehn, 2016).

The role of citizens varies depending on the purpose of the citizen science project. Tulloch et al. (2013) studied the purpose of citizen science projects in bird watching, a research field with a century-long history of citizen science. The most common purpose of citizen science is knowledge generation (knowledge also new to the involved professionals), followed by improving monitoring methods and raising awareness among citizens. In water resource management, citizen science can enhance knowledge about the water system, for example by generating knowledge on different spatial or temporal scales. It may also be used to add or test new monitoring methods for the existing monitoring network (mentioned in exploratory interviews held at the start of this study). Citizen science can also be used to increase the water awareness that was dubbed absent in the OECD report (OECD, 2014). According to Tulloch et al. (2013), citizen science can be used to improve management practice as well. In water resource management such improved management can be the result of more frequent monitoring. The literature mentions public education as an important purpose of citizen science (e.g. Cohn, 2008), but Tulloch et al. (2013) found that public education was rarely the main purpose. Other identified, yet also more rare, purposes include doing social research (e.g. on human behaviour), offering recreation and serendipity (i.e. unexpected discoveries). A more recent study in the field of ecology specified the potential of citizen science for the purpose of policy development (Hollow et al., 2015). In early stages of a policy development process, citizen science can be used to discover alternative management actions. In case there is a range of alternatives, citizen science can be used to gauge public opinion. In later stages it can be used to persuade public opinion towards a desired alternative or to provide a legal justification for the chosen policy. Citizen science-based data can be used for decision-making in water resource management as well (Macknick and Enders, 2012). Projects can have one or multiple purposes, as the iSPEX project demonstrates. This project served knowledge generation, public education and method improvement (Land-Zandstraet al., 2015; Snik et al., 2014).

For the purposes of knowledge generation, improving monitoring methods improves management and policy development. Bonney et al. (2009) provide a useful classification of citizens' roles. They suggest there are basically three levels of citizen involvement possible: contribution, collaboration and co-creation. In a contributory project, citizens are mainly involved in data collection; the research question and design are done by scientists or experts. In collaborative projects citizens are involved in the analysis and can be involved in the design and dissemination of results as well. In co-created projects citizens are involved in all steps of the research process and may even initiate the project. The vast majority of studies in the overview presented by Bonney et al. (2009) considered contributory projects. Even the occasionally co-created projects were part of multi-case studies in which contributory projects dominated the results. Citizens' involvement in activities other than data collection may serve different purposes; for example, citizen-based goal-setting could enhance adaptive management practices (Cooper et al., 2007). Bonney et al. (2009) is frequently cited (e.g. Rotman et al., 2012; Roy et al. 2012) and can be considered a typical classification. The levels of involvement align to a large extent with the governance structures defined by Conrad and Hilchey (2011): consultative, collaborative and transformative governance. In transformative governance citizens initiate a project. An example can be found in the global community monitor that measures air and water quality on a global scale (Conrad and Hilchey, 2011).

For the purposes of public education, raising awareness, improving management and policy development, science communication literature (e.g. Varner, 2014) provides a more useful classification of the interaction between citizens and professional counterparts. We use the term science communication here in a broad sense and include communication between public and all professional counterparts referred to by Silvertown (2009), including practitioners. Citizen science is often viewed as a form of informal science education, contributing to public awareness of science (PAS) and public understanding of science (PUS). We think this view is too limited, as it only encompasses the deficit model of science communication. Higher levels of involvement public engagement of science (PES) and public participation in science (PPS) are possible (Van der Auweraert, 2005), particularly in the collaborative and co-created projects in Bonney's definition. An example of such participation is the water quality programme in Rhode Island mentioned in the review by Conrad and Hilchey (2011).

The motivation of professional counterparts to engage in citizen science has been less frequently studied than the motivation of citizens, and to our best knowledge research on professionals' motivation is limited to scientists. Scientists' motivations are primarily to advance science as well as develop their careers (Rotman et al., 2012). This is aligned with citizens' motivational desire to contribute to science and conservation or to engage in exploring a topic of their interest further (e.g. Rotman et al., 2012). Weng (2015) identifies three areas of friction between the vision of scientists and volunteers with regard to citizen science. The first area of friction is often short-term participation of volunteers that conflicts with scientists' interest in long-term processes. The second area concerns the limits of what volunteers can do and their dissatisfaction with the research processes. The third area regards a power hierarchy between citizens and scientists. Rotman et al. (2012) found that while the motivations of citizens and scientists are complementary, they can also change over time. Therefore, continued attention by those who are managing citizen science projects with regard to matching these motivations is crucial.

The motivation of scientists cannot be translated one-to-one to practitioners or (local) government representatives for two reasons. First, scientists are concerned with scientific data collection (Rotman et al., 2012), while practitioners are often interested in improving management practices (Weng, 2015) and government agencies are concerned with policy-making (Hollow et al., 2015). Second, the different role of authorities leads to different expectations. Water authorities believe that citizens see water resource management as a task for authorities only, which implies that citizens do not want to be involved. Nevertheless, most water authorities agree that they need the observations of citizens for their work, as expressed in several studies related to flood risk management (Wehn and Evers, 2014; Wehn et al., 2015).

This study aims to explore perspectives of government practitioners regarding citizen science. We explore Dutch water practitioners' perceptions of citizen motivation, acceptance in their organization, (potential) purposes and level of citizen engagement. We address apparent knowledge gaps regarding the motivation of professionals to embrace citizen science and specifically the link with citizen participation. Perceptions were explored using Q methodology. With Q methodology we identified a set of viewpoints that describe the variation in opinions. This variation we consider meaningful for the design and implementation of citizen science at water boards in the Netherlands. In addition to providing insight into the viewpoints of Dutch practitioners, this study aims to develop a methodology to study perspectives regarding citizen science in a wider range of countries and professionals.

The study uses Q methodology (Van Exel and De Graaf, 2005; Watts and Stenner, 2012) to find viewpoints on citizen science among employees of the Dutch water authorities. Q methodology is a relatively uncommon method in water resource management (e.g. Raadgever et al., 2008), but it is a popular method in social sciences fields, such as political science and psychology (Cools et al., 2009). The strengths of the Q methodology are that it combines qualitative and quantitative aspects and that it is statistically robust, with small samples of 30–40 people (Van Exel and De Graaf, 2005; Watts and Stenner, 2012). For this research Q methodology has the advantage over fully quantitative approaches in that we could explore a wide spectrum of viewpoints, whereas quantitative methods would arrive easily at averaged values. Quantitative methods would likely not reveal the distinguishing elements that are in our opinion crucial to be aware of for effective implementation of citizen science. Q methodology has the advantage over fully qualitative methods that it reduces the variation in opinions to a representative small set of viewpoints in a trackable, relatively little biased way (further discussed below).

This section provides a short description of the methodology and research-specific details. We kindly refer readers interested in more methodological details to Van Exel and De Graaf (2005) for a quick introduction, and Watts and Stenner (2012) or Brown (1980) for an elaboration on the philosophy and statistical basis. In conducting this research we closely followed the guidelines of Watts and Stenner (2012). Our Q methodological research consisted of seven stages summarized in Fig. 1 and elaborated below.

The first stage aimed at collecting as wide a range as possible of opinions on the topic of citizen science for water quality monitoring (discourse) and documenting them in statements (concourse). We held 10 semi-structured interviews with employees of water authorities, nature managers and citizen organizations, from which 181 statements were derived. Additionally, we organized a structured group discussion about citizen science with water professionals at the afore-mentioned workshop, which resulted in an additional 21 statements. To collect a wider range of opinions, we organized a focus group meeting with five middle-aged woman of an informal walking club with high potential to participate in citizen science, which resulted in an additional 20 statements. Finally, benefits and downsides of citizen science were extracted from the literature, which resulted in an additional seven statements. To reduce researcher bias, we based statements as much and as closely as possible on quotes. The final concourse (all possible statements on a topic) consisted of 229 statements. Four themes could be identified in the concourse: (I) citizen motivation, (II) acceptance of citizen science at the water authority, (III) purposes of citizen science for professionals and (IV) level of citizen engagement.

In Stage 2 the concourse was reduced to the so-called Q-set set of 48 statements that still reflected the full discourse. We first reduced the concourse to a preliminary Q-set of 65 statements by excluding or reframing statements that were similar to others, out of scope or ambiguous. Six masters students between the ages of 22 and 25 tested this preliminary Q-set. Two female students had a major in water resource management; the other two female and two male students had a different major (medicine, mathematics, mechanical engineering and management studies). We instructed them to sort the statements and list statements that they did not understand, found similar in meaning or considered irrelevant, and improved the preliminary Q-set based on their feedback. Table 1 contains the final Q-set of 46 statements; the Roman numerals indicate to what theme the statement contributes.

In Stage 3, the P-set (i.e. the group of participants) was sampled using both a structured approach with three criteria and snowball sampling. Flood risk was a first criterion, as water authorities with a high flood risk face different challenges than water authorities with a lower flood risk. Age (expressed in years since the last reform or merger with another water authority) was a second criterion, as a recent reform suggests the organization may be more susceptible to innovation, such as citizen science. Location (within or outside the urban conglomerate Randstad) was a third criterion, because interviewees in the semi-structured interviews held to define the discourse suggested a different relationship between water authority and population in rural and urban areas. The structured approach resulted in eight water authorities representing a combination of these criteria. In each water authority we approached the ecologists because we expected them to be more familiar with the concept of citizen science and we had access to a list of ecologists per water authority. Additionally we used snowball sampling. We asked the participating ecologists to recruit colleagues with a similar opinion, to enhance overlap between individual opinions, and, with different opinions, to increase the diversity of opinions. Also, we asked participants to recommend someone with an opposing opinion, in order to discover as many viewpoints as possible. Participants 20, 24, 25, 30 and 31 out of 33 were recruited with this strategy, with the aim that they would belong to different (new) viewpoints. Two to six people with different positions were interviewed per water authority, which resulted in interviews with 1 politician, 20 policy advisors, 10 ecologists/hydrologists and 2 field staff members.

Next, in Stage 4, the Q-sorts (the actual arranging sorting process) took place in four sub-steps, taking a total time of 60 to 75 min. First, the first author gave three examples of consultative citizen science to all participants, to ensure everyone had a basic level of understanding of citizen science. These examples were

the Dutch garden bird count (www.tuintelling.nl);

Second, participants pre-sorted the statements into three piles: agree, disagree and neutral. Third, they made a final sorting of the statements in a fixed distribution (see Fig. 2). Finally, the first author held a structured post-sorting interview. Post-sorting interviews were included in this study, because they can provide in-depth insight into the beliefs and values underlying the sorts and allow for an analysis based on the participants' rationale rather than on the available literature or the researcher's bias (Gallagher and Porock, 2010). Discussing all statements with participants would have been preferable, but was not feasible given the available time for this study and the geographic spreading of participants. In a structured interview, participants explained their reasoning for the statements in categories +4 and -4 and (if time allowed) any statement of their choice. Participant's afterthoughts were recorded, transcribed and categorized per statement and per factor.

Next, in Stage 5, a factor analysis was performed with software package PQMethod, version 3.2.1. PQMethod is used to perform a factor analysis and is frequently used in Q methodological research (Van Exel and De Graaf, 2005; Cools et al., 2009; Raadgever et al., 2008; Watts and Stenner, 2012). A factor analysis is a statistical method to describe variability in a set of correlated variables, in this case the ranking of statements by individuals, by a smaller number of factors. We included three factors with an eigenvalue above 1 (recommended by Watts and Stenner, 2012) for further analysis. The factor analysis also showed which people load on which factor, i.e. which people have a perspective resembling that factor. People can load on none, one or multiple factors. For this a threshold was used, the significant factor loading (SFL). In this study we followed Watts and Stenner (2012) and used a SFL of 0.38. Factors were optimized using factor rotation, which aims to have as many people as possible load on a single factor rather than loading on two factors simultaneously. The rotating process does not alter the results themselves, but changes the researcher's observation position in order to optimize the loading of each Q-sort on a single factor (Watts and Stenner, 2012, p. 118). A manual rotation was preferred above the built-in Varimax rotation of the PQMethod, because it has a lower inter-factor correlation and thus results in more distinct viewpoints.

Next, in Stage 6, a weighted average was used to create an illusory person with a factor loading of 1.0 per factor, i.e. a hypothetical person who has fully adopted this factor. Following Watts and Steiner (2012) Q-sorts with loadings that exceed the SFL of 0.38 for a factor were incorporated to compute the weighted average for that factor. See Table 2 for the factor arrays and Table 3 for the final factor loadings. It must be noted that the number of people loading on a factor cannot be used to determine the distribution of viewpoints in the total population without additional (quantitative) research.

The final stage, Stage 7, is data analysis, where the factor arrays were translated into a viewpoint narrative. We followed the guidelines of Watts and Stenner (2012), Gallagher and Porock (2010) and Cools et al. (2009), and used distinguishing statements. We created a narrative of the +4 and -4 ranked statements and the statements ranked highest in a single factor array, meaning this statement is ranked lower in all other factors, and vice versa. For example, in Factor A, Statements 2 and 9 ranked with +4 and Statements 10 and 41 with -4, and Statement 34 is an example of a Statement ranked highest in factor A. Factor A ranks it +2, compared to -1 and +1 in Factors B and C. Hence, Statements 2, 9, 10, 41 and 34 were included in the viewpoint narrative of Factor A. In addition to this mechanism of distinguishing statements, two other mechanisms were introduced to reduce researcher bias. First, we conducted post-sorting interviews in order to be able to incorporate participant's underlying values and assumptions in the process of interpretation. During these interviews we also checked whether in their opinion all relevant aspects were covered by the Q-set. Second, we showed all participants an initial version of the narratives and asked whether they recognized themselves in their assigned narrative viewpoint and why (or why not). In addition we presented the results to employees of the Dutch water authorities on two occasions (Delfland Scriptieprijs uitreiking 21 January 2016 and STOWA Monitoringcongres 19 April 2016) to collect feedback. An overview of mechanisms used to reduce bias in all steps of the research is given in Fig. 3.

The 33 Q-sorts resulted in the identification of three factors from which three viewpoints were derived: A “Citizen participation for data application”, B “Water authority in control”, and C “Education and sharing local knowledge”. The choice for three factors was based on the explained variance of the first four factors before rotation that displayed with 53, 8, 6 and 1 % a cut-off after the third factor that was supported by qualitative information from the interviews. Table 1 and the radar charts presented in Fig. 5 show how an individual would rank the items if that person were representing that factor 100 %. For example, Statement 9 (“Citizen Science enables the collection of large amounts of measurements”, Theme III Purposes of Citizen Science) would be placed under most agree (column +4) by a person with Factor A, under agree (column +2) for Factor B and under neutral (column 0) for Factor C. None of the viewpoints disagrees with Statement 9, but the difference between Viewpoints A and C is evident. Except for Statement 2 (“Citizen Science is important, since it contributes to increasing water awareness”), with which all viewpoints fully agree, and Statement 35 (“If citizens are structurally contributing, they should be compensated for that”), differences were found between the viewpoints that will be further discussed below. The factors provided a quite clear separation of the participating practitioners in groups given Table 2. Out of 33 participants, 21 loaded significantly and uniquely on Factor A, 4 on Factor B and 2 on Factor C. Three participants loaded significantly on Factors A and C, one on Factors A and B and one on Factors B and C. One participant did not load significantly on any of the factors.

The remainder of this section contains the three viewpoint narratives. The term viewpoint is used to refer to the factor's interpretation, for which we made use of the quantitative factor arrays in Table 1 and qualitative quotes from participants loading significantly on that factor (see Table 2). The narratives are based on absolute results (agree or disagree), the relative results (an item is ranked higher or lower in Viewpoint A than Viewpoints B and C) and characteristic interview items. Item rankings are presented in the following format: (item number: item ranking) such that (2: +4) means item 02 is ranked +4 in this viewpoint. Interview fragments are integrated in the narratives as a quote followed by the letter Q and a number indicating the source. For example (“quote” – Q1) means the quote comes from the Q-sort and thus participant 1. Figure 4 shows the availability of interview fragments per factor and per statement. Most fragments (125) were available to interpret Factor A as most people loaded significantly to that Factor A and less for Factors B (20 fragments) and C (10 fragments). As an intended result of the interview technique (see Stage 4 in the Method Section), most fragments were available for statements with particularly low or high rankings, and there was a positive relation between the level of disagreement among factors and the number of interview fragments. The higher the total absolute difference between the Factors A, B and C, the higher the number of interview fragments.

This study identified three different viewpoints with regard to citizen science derived with Q methodology. Participants sorted statements about four themes described in the Introduction and Method sections. In this discussion we first (Sect. 4.1) reflect on these four themes and relate the results to the literature, and second (Sect. 4.2) discuss the limitations of the research and make recommendations for future research.

This study contributes to understanding about government practitioners' acceptance and perception of citizen science. A Q methodological approach was applied to identify the viewpoints of practitioners on citizen science, in the case of water quality monitoring at Dutch regional water authorities. Water authority employees sorted a set of 46 statements related to citizen science. Three factors were identified in a factor analysis and translated into corresponding viewpoint narratives.

The first viewpoint, Viewpoint A, is named “Citizen participation for data application”. People with Viewpoint A see more opportunities than challenges when it comes to citizen science. They see applications in practical use of the data, but also in the active engagement of people. The second viewpoint, Viewpoint B, is named “Water authority in control”. People with this viewpoint see a potential for data contributions by citizens in an illustrative way, but are concerned with challenges in organizational capacity, expectation management and motivating citizens as well. The third viewpoint, Viewpoint C, is named “Education and local knowledge”. People with this viewpoint focus on educational goals, such as teaching people about their environment and getting schools involved. They consider data applicability of secondary importance, although the data can be used illustratively.

The outcomes of this study provide strong indications that practitioners at the Dutch water authorities welcome citizen science. These practitioners further believe citizen science can contribute to bridging the awareness gap as identified by the OECD (2014) and the Dutch water authorities (UvW, 2015a). All three viewpoints are positive towards citizen science in the form of contributory projects in which citizens collect data. People with Viewpoint A support collaborative citizen science as well, but none of the viewpoints support co-created projects between citizens and water authorities. Interviews identified low expectations in citizens' motivations and capacities as underlying causes of this low support for higher levels of citizen involvement. This may jeopardize the much-needed trust in the relationship between citizens and practitioners. Although participants recognized the potential of citizen science to change governance structures, the design of citizen science projects in the Netherlands is not expected to move beyond contributory projects in the near future.