Which characteristics and integrations between characteristics in blue–green spaces influence the nature experience?

This study aims to investigate the impact of characteristics in blue–green spaces and integrations between characteristics on people’s preference for nature experience (perception of and interaction with nature). A total of 795 participants performed two sets of twelve choice tasks, each task presenting two photographs where eight characteristics were manipulated. Participants indicated which photograph was preferred for perception of and interaction with nature, respectively. Conjoint analysis was used to analyze the data. Water features were the most important characteristic, followed by bushes, upkeep, and trees. Integrating wildlife, peers, flowers, and paths with vegetation and water features can improve their attractiveness for nature experience. Upkeep should be prioritized in the single green space, but the integration between water features and bushes or trees in the blue–green space appeared to be most important. Through the design and integration of different characteristics, our research results contribute to the construction of nature experience environments in blue–green spaces.


Introduction
Cities undergoing rapid urbanization often destroy or degrade natural areas to make way for built-up areas or other land uses, which can cause compact cities (small size and high density) to show very low per capita green space allocation (Fuller and Gaston 2009). Moreover, urbanization diverges residents' landscape preferences. For example, the more urbanized the area in which a person resides, the more they prefer natural landscapes (Xu, Luo, and Wang 2020). However, rapid urbanization increases the geographical isolation of people from opportunities for nature experience (Turner, Nakamura, and Dinetti 2006). Nature experience, including individuals' perceptions and interactions with stimuli from the natural world through a variety of sensory modalities (Bratman et al. 2019), is closely related to people's health and well-being. There is growing evidence that nature experience in green and blue spaces is vital to improve physical health (Keniger et al. 2013), foster mental well-being (Bratman, Hamilton, and Daily 2012), and promote positive attitudes toward nature . Additionally, as rapid urbanization further restricts the urban natural network, it becomes more difficult for urban residents to visit relatively larger natural environments. Improving the quality of existing green and blue spaces can compensate for the reduced quantity of natural environments (Lau, Yung, and Tan 2021). Faced with the threat of urbanization of natural environments and the health benefits from nature experience, crucial decisions must be made about how to improve the quality of green and blue spaces to increase nature experience.
The quality of green and blue spaces, as an important factor affecting nature experience, can be evaluated by a variety of environmental characteristics. Studies on environmental characteristics have mainly focused on the accessibility, facilities, amenities, aesthetics and attractions, and incivilities in green spaces (Knobel, Dadvand, and Maneja-Zaragoza 2019;Ayala-Azc arraga, Diaz, and Zambrano 2019). In comparison, the characteristics of blue spaces have received less attention (McDougall et al. 2020), mainly the accessibility of water, the presence of water, visual aesthetic, and water quality (Wang and Zhao 2019;Mishra et al. 2020). Additionally, research focusing on freshwater blue space typologies (e.g. lakes, rivers, ponds, and streams) is more limited (V€ olker and Kistemann 2011). Although green and blue spaces tend to be categorized as distinct and separate entities in research concerned with environmental characteristics, sometimes, they cannot be distinctly separated from each other (Korpela et al. 2010). For example, blue spaces are often classified as green spaces or regarded as an attribute of green spaces (Grilli, Mohan, and Curtis 2020). At the same time, terrestrial attributes around blue spaces, such as high-quality paths (Verbi c, Slabe-Erker, and Klun 2016) and easily accessible waterside spaces (De Bell et al. 2017) usually contain green vegetation, which also plays a key role in enhancing the quality of blue spaces (Mcdougall et al. 2020). Blue-green spaces, containing both green vegetation and water bodies (Yang et al. 2020) and having positive interactions between blue and green space characteristics, are more likely to provide good environmental quality than single green spaces or blue spaces. Thus if we want to fully understand the attractiveness of blue-green spaces for nature experience, it is surely necessary to account for the characteristics of blue-green spaces.
Most qualitative and quantitative studies have focused on the importance of environmental characteristics in determining people's use of green spaces. Specific characteristics that encourage the use of green spaces included well-maintained facilities (Van Hecke et al. 2018a), aesthetic attributes (e.g. the presence of nature, large open spaces, water features, and greenness and trees) (Dunton et al. 2014;Edwards et al. 2015), sport and play facilities (e.g. playgrounds, basketball courts, outdoor fitness equipment, slides, and swings) (Veitch et al. 2017), and other characteristics such as the presence of adolescents and friends, walking paths, wildlife or biodiversity, picnic tables, and toilets (Van Hecke et al. 2016;Lovell et al. 2014). Incivilities (e.g. rubbish, graffiti, and presence of undesirable users) (McCormack et al. 2010) may discourage people's use. Besides, good facilities and the presence of wildlife can attract people to visit blue spaces regularly, and longer blue space visits were associated with higher well-being (Garrett et al. 2019). However, little is known about the effect of environmental characteristics and integration between characteristics on the use of other natural environments (e.g. blue-green spaces). Further, most research has focused on the support of environmental characteristics for visiting, physical activity, perceived safety, stress relief, heat relief, and restorative experiences (Van Hecke et al. 2018a;Campagnaro et al. 2020;Ebenberger and Arnberger 2019), but this limits further consideration of other uses, such as nature experience. Although nature experience is diverse and can occur through conditions of real contact, window views, representations, or simulations, Russell et al. (2013) have defined nature experience by characterizing person-nature connections (i.e. knowing, perceiving, interacting with, and living within). Their evidence also indicated that perceiving and interacting with nature make people happier and healthier (Russell et al. 2013). However, we still do not know which characteristics in blue-green spaces are most important for triggering nature experience, especially perception of and interaction with nature.
Given the necessity of addressing reduced nature contact and the opportunities of nature experience provided by blue-green spaces, it is essential to improve our understanding of the relationship between blue-green spaces and nature experience. Previous studies examining the relationship between environmental characteristics and people's use often used self-report questionnaires, in which they asked participants to indicate the environmental characteristics in green spaces, and to link those characteristics to their use behavior. However, the participants had to recall these characteristics, which involve recall bias. The use of a choice-based conjoint (CBC) analysis with manipulated photographs could overcome the aforementioned limitation, as participants who judge photographs do not have to imagine characteristics of the physical environment. Additionally, this method can not only allow the researcher to differentiate the separate influence of each environmental characteristic under controlled conditions, but also examine the influence of integration between these characteristics Van Hecke et al. 2018a;Ghekiere et al. 2018). Therefore, in this study, CBC analysis with manipulated photographs was used to investigate people's general preference for nature experience in blue-green spaces: (1) What are the important characteristics in blue-green spaces for attracting people to experience nature? (2) How can the attraction for nature experience be improved by integration between characteristics in blue-green spaces? (3) What is the difference between single green spaces and blue-green spaces in attracting nature experience? Our findings will inform urban managers and practitioners on which characteristics to prioritize and which integrations between characteristics to consider in the design and renovation projects to facilitate people's nature experience in blue-green spaces.

Participant recruitment and protocol
Most participants were recruited by purposeful convenience using social media, family, and friends. Furthermore, snowball sampling was used to recruit additional participants. First, by using QQ and We Chat software, questionnaires were distributed to undergraduates and postgraduates in the college of Landscape Architecture of Sichuan Agricultural University. Second, with the help of Shanghai Diagaid Marketing Research Company, questionnaires were distributed to college students in Shanghai by E-mail. Finally, students who had completed the survey spread the questionnaire to their friends or relatives living in Sichuan, Shanghai or other cities. A total of 795 participants completed the online questionnaire developed with Sawtooth Software (Lighthouse Studio version 9.8.1, Provo, Utah, USA). The online questionnaire was available from the middle of July 2020 until the end of August 2020. Before actual data collection, a pilot-test with 20 volunteers was performed to test the protocol and ensure the comprehensibility of all questions. Completing the questionnaire took approximately 5 minutes. Informed consent was automatically obtained when participants completed the questionnaire 1 .

Development of manipulated photographs
A set of 6,804 manipulated photographs were developed with Adobe Photoshop software. The photographs were all modified versions of one "basic" photograph representing a typical natural environment within cities, where people experienced nature frequently (Appendix A [online supplementary material]). Eight characteristics of blue-green spaces, including water features, paths, upkeep, trees, bushes, flowers, peers, and wildlife, were manipulated in each photograph (Figure 1), and 2-7 levels were created for each characteristic. These characteristics and their levels were selected Figure 1. Examples of the manipulated photographs. (A) The photograph presenting the characteristic levels that we anticipated to be most preferred. (B) The photograph which we anticipated to receive medium preference scores. (C) The photograph presenting the characteristic levels that we anticipated to be least preferred. based on previous qualitative and quantitative research studying the relationship between environmental characteristics and people's use in green or blue spaces (Table 1). All photographs depicted the same setting under fine weather conditions and differed through the manipulation of at least one characteristic.

The online questionnaire
The online questionnaire consisted of two parts. First, both socio-demographic information (i.e. age, gender, and education level) and use habits in blue-green spaces (i.e. frequency of visits, average duration of visits, and usual accompaniment to blue-green   No Birds and butterflies present Birds and butterflies present spaces) were collected (see Table 2 for the response categories). Second, participants were asked to perform two sets of CBC tasks, each with twelve choice tasks. This CBC method was frequently used in marketing research to examine how different characteristics of a product influence consumer preferences . For example, during a CBC task, people were asked to choose between products (e.g. televisions) that differed in some characteristics (e.g. screen size, screen quality, and price) (Orme 2009). In our study, the "products" were blue-green spaces that differed in eight pre-defined characteristics. This methodology has been used previously to assess restorative components of small urban parks (Nordh, Alalouch, and Hartig 2011) and examine preferences for park attributes (Alves et al. 2008). While these previous studies used a written description of park characteristics, CBC analysis with manipulated photographs, which has been widely used to define the most important characteristics of streets for walking and cycling among children and adults (Van Hecke et al. 2018a), was used in this study. Two photographs with different combinations of characteristics were presented to the participants. In the first set of tasks, participants indicated which scene could most make them perceive nature from the two displayed photographs (Figure 2), while in the second task, they were asked to select the photograph most supportive for interaction with nature ( Figure 3). People experienced ecosystems in a variety of ways, and perceiving (i.e. remote interactions with ecosystem components) and interacting (i.e. physical, active, direct multisensory interactions with ecosystem components) were the two important channels of human experience (Russell et al. 2013). In this study, perception of nature was clarified as perceiving naturalness and biodiversity of blue-green spaces (e.g. viewing natural elements or many kinds of birds and butterflies) and interaction with nature was described as experiencing nature intentionally, such as playing with water, listening to natural sounds and smelling flowers. Participants received the following instructions before starting the first choice task: "Blue-green spaces refer to the space containing both green vegetation (grass, trees, shrubs or other vegetation) and water bodies (streams, rivers, pools, lakes). Please imagine you are going to a blue-green space during daytime in your daily life. Two photographs will be presented to you. Please take your time to look at the photograph, and then select the scene that can make you perceive nature the most. Perception of nature refers to viewing natural elements or many kinds of birds and butterflies in blue-green spaces. There is no good or bad solution, and we are just interested in what you consider as most important when perceiving nature". Prior to the second choice task, the same instruction was given, except that participants were asked to choose the photograph most supportive for interaction with nature. Each participant would complete ten random and two fixed tasks in each set of choice tasks. Since a full-profile design was used in the choice task, the two photographs in one task could differ in one to eight characteristics. The 10 random tasks were different for all participants and were randomly assigned by the Sawtooth program. For the two fixed tasks, the photographs were selected in advance and were the same for all participants (i.e. at the 5th and 10th positions) to be able to check the test-retest reliability. Furthermore, the fixed tasks were also used to check the validity of the statistical model by offering a comparison against the predictions of the final statistical model.

Analyses
We used a logit model to identify how people reacted to the manipulated photographs. In the modeling stage, the dependent variable was the respondent's choice, and the predictors were eight characteristics. A priori power analysis was calculated to establish the appropriate number of participants using the following formula: nta/c > 500 (n ¼ number of participants; t ¼ 12: number of choice tasks; a ¼ 2: number of alternatives per task; c ¼ 21: the largest product of levels of any two factors) (Sawtooth Software Inc 2013). This showed that 438 subjects were sufficient when manipulating eight characteristics in one photograph (with a maximum of seven levels) and presenting 12 choice tasks to each participant. A total of 1,551 people participated in the questionnaire survey, with 795 valid questionnaires. Test-retest reliability of choice tasks about perception of and interaction with nature resulted in 88.7% agreement (90 participants not responding consistently) and 92.2% agreement (62 participants not responding consistently), separately. After excluding the participants not responding consistently on both sets of the fixed choice tasks (n ¼ 137), 658 participants were used for analysis.
Descriptive statistics were calculated using IBM SPSS statistics version 22 (IBM SPSS Inc., Armonk, NY) and the CBC data were analyzed using Sawtooth Software Lighthouse Studio version 9.2.8. CBC analyses yield two types of parameters: partworth utilities and importance scores. Part-worth utilities represent the desirability of each characteristic level and can be interpreted similar to regression coefficients in regression analyses (Orme 2009). Part-worth utilities can be used to determine which level within a characteristic is preferred and greater positive values of a part-worth utility indicate a higher preference for that specific level. However, these results are not comparable between characteristics, only within characteristics. To facilitate interpretation, these part-worth utilities were zero-centered. For example, if the levels "no trees", "a few trees", and "many trees" within the characteristic of "trees" have partworth utilities À10.0, 3.0, and 7.0, respectively, it implies that a blue-green space with many trees is the most preferred and a blue-green space with no trees is the least preferred. Importance scores represent the influence of a characteristic on preference. Higher importance scores imply stronger influences on preferences. Importance scores are directly related to the difference between the least and most favorable attribute level (Orme 2009). This means that the importance scores reflect which characteristic has more (or less) influence on choice, without taking into account what the preferred level of this characteristic is but through calculating the difference between the most and least preferred level.
When analyzing the choice data, we applied Hierarchical Bayesian (HB) algorithm to estimate the logit model parameters at individual level. Hierarchical Bayes (HB) algorithm is regarded as the best method to analyze CBC data (Allenby, Arora, and Ginter 1998). Average part-worth utilities and importance scores were reported for comparison. Of 95% confidence intervals (CIs) were also calculated to determine the significance. Importance scores and part-worth utilities within one characteristic with non-overlapping 95% CIs were significantly different from each other. The fit of the conjoint model was presented by the root likelihood (RLH) which ranges from 0 to 1, with a higher value indicating a better fit of model (Orme 2009). We firstly built and estimated the main effect model, and then added some possible interaction effects to the main effect model to say whether there would be some meaningful interactions. The possible interactions were detected by the "CBC interaction search tool" of the Sawtooth Software (Orme 2016). Separate models were constructed to analyze the interaction effects between different characteristics in blue-green spaces. The interaction effect was also reported for comparison. Finally, given that water features were the most important characteristic in blue-green spaces for nature experience and that four of the nine significant interaction effects were found with water features, the relative importance of all other characteristics in blue-green spaces was calculated within water features and each level of water features.

Descriptive statistics
The socio-demographics and use habits in blue-green spaces can be found in Table 2. Nearly half of the participants (43.3%) were between the ages of 19 and 22. 64.3% were female and 53.5% had obtained undergraduate education. More than half of the participants (66.1%) reported to be regular users of blue-green spaces in the past month (at least 2-3 times a month). Most participants preferred to visit blue-green spaces with friends (59.8%) and most visits to blue-green spaces lasted less than 1 h (91.6%).
Within each characteristic, all part-worth utilities from the different levels of each characteristic significantly differed from each other, except for water features and paths (Table 3). For water features, no significant difference between W4 and W6 was observed. Besides, W2 was not significantly different from W3, W4 and W6. For paths, no significant difference was observed between no paths and paved paths. Within each characteristic, part-worth utilities were in the expected direction, except for water features. The preference in the expected direction for the water features was described as follows: a blue-green space with W7 was preferred over a blue-green space with W6 or W5, which was preferred over a blue-green space with W4 or W3, which was preferred over a blue-green space with W2, and a blue-green space with W1 was the least preferred characteristic level.

Interaction effects
There were 27 possible interaction effects of which only five were significant, including "trees Â peers", "paths Â flowers", "upkeep Â peers", "bushes Â flowers", and "water features Â wildlife" (Appendix B [online supplementary material]). Interaction effects were the additional effects of two or more characteristics that were not explained by the main effect model. For example, the part-worth utilities of "many trees" and "walking peers" in the main effect model for the perception of nature were 32.9 and 14.5, separately (see Table 3). The interaction effects of "many trees" Â "walking peers" were 3.1 (see Table B1 in Appendix B [online supplementary material]), which was the additional effect contributed to the people's preference apart from the main effect of "many trees" and "walking peers". The utilities of interaction effects could be regarded as the interaction "coefficients", and the higher the coefficient, the stronger the interaction effects would be.
First, the significant interaction effect between "trees" and "peers" (chi square ¼ 16.14; p ¼ 0.003) showed that the greatest effect of peers (from no peers or sitting Within one environmental factor, levels with an peers to walking peers) was found with a few trees (see Figure 4; Table B1 in Appendix B [online supplementary material]). The greatest effect from no peers to sitting peers on the preference for perception of nature was found with no trees. Additionally, the greatest effect from sitting peers to walking peers was found with many trees. Second, the significant interaction effect between "paths" and "flowers" (chi square ¼ 10.95; p ¼ 0.004) showed that the effect of flowers was significant for all different paths in the expected direction; only the strength of the effect differed (Table B2 and Figure B1 in Appendix B [online supplementary material]). The greatest effect of flowers was found when there were unpaved paths. Third, the significant interaction effect between "upkeep" and "peers" (chi square ¼ 13.54; p ¼ 0.009) showed that the greatest effect of peers (from no peers to sitting peers or walking peers) was found with moderate upkeep (Table B3 and Figure B2 in Appendix B [online supplementary material]). In addition, the greatest effect from sitting peers to walking peers was found with good upkeep. There was also a significant interaction effect between "bushes" and "flowers" (chi square ¼ 7.92; p ¼ 0.019) (Table B4 and Figure B3 in Appendix B [online supplementary material]). The effect of flowers from absence to presence was greater if there were no bushes present. Finally, the significant interaction effect between "water features" and "wildlife" (chi square ¼ 14.27; p ¼ 0.027) showed that the greatest effect of wildlife from absence to presence was found with W5 (Table B5 and Figure B4 in Appendix B [online supplementary material]). Furthermore, the effect of ponds and lakes was greater than streams and rivers when the wildlife went from from absence to presence.

Relative importance of the characteristics within water features and each level of water features
The relative importance of each characteristic within water features is presented in Table 4. The results showed that bushes were the most important characteristic within water features, followed by upkeep. In addition, the relative importance of each characteristic within the seven levels of water features is presented in Figure 5. The results showed that bushes were the most important characteristic in situations where there were W2, W4, W5, W6, and W7, while as for W1 and W3, upkeep was their most important characteristic. Furthermore, flowers were the least important characteristic among all levels of water features except for W1 and W5.

Interaction with nature 3.3.1. Preferred characteristics and characteristic levels
The most important characteristic for interaction with nature was water features (27.9; 95% CI ¼ 27.4, 28.5) ( Table 3). Bushes (17.0; 95% CI ¼ 16.5, 17.5) were the second most important characteristic while upkeep (11.9; 95% CI ¼ 11.2, 12.6) was the third one. These were followed by trees (11.2; 95% CI ¼ 10.9, 11.5) (not significantly . Average relative importance of the seven characteristics within different levels of water features for perception of nature. W1: no water; W2: streams with grassy slope revetment; W3: rivers with grassy slope revetment; W4: rivers with gravel and cobblestone revetment; W5: ponds with grassy slope revetment; W6: ponds with gravel and cobblestone revetment; W7: lakes with gravel and cobblestone revetment. different from upkeep and wildlife), wildlife (10.5; 95% CI ¼ 10.1, 11.0), peers (8.1; 95% CI ¼ 7.8, 8.5), flowers (7.5; 95% CI ¼ 7.2, 7.8) and paths (5.8; 95% CI ¼ 5.5, 6.1). In line with the results for perception of nature, clear preferences for a specific level of each characteristic were observed, except for water features and paths (Table 3). For water features, no significant difference between W2 and W6 was observed. For paths, no significant difference was found between no paths and unpaved paths. Within each characteristic, the preferences were in the expected direction, except for peers and water features. The preference in the expected direction for peers was described as follows: a blue-green space with walking peers was preferred over a blue-green space with sitting peers, which was preferred over a blue-green space without peers. The preference in the expected direction for the water features has been described (see above).

Interaction effects
The combination of all interaction effects gave 27 possible interaction effects of which four were significant, namely "water features Â peers", "bushes Â flowers", "water features Â trees" and "water features Â paths" (Appendix C [online supplementary material]).
First, the significant interaction effect between "water features" and "peers" (chi square ¼ 32.62; p ¼ 0.001) showed that the effect of peers (from absence to presence or from sitting peers to walking peers) on the preference for interaction with nature was greater when the water features were W4 and W6 (Table C1 and Figure C1 in Appendix C [online supplementary material]). Second, the significant interaction effect between "bushes" and "flowers" (chi square ¼ 12.14; p ¼ 0.002) showed that the effect of flowers from absence to presence was greater when the bushes was less (Table C2 and Figure C2 in Appendix C [online supplementary material]). Third, the significant interaction effect between "water features" and "trees" (chi square ¼ 26.06; p ¼ 0.011) showed that the effect of trees provided a significant increase on the preference for interaction with nature when the water features were ponds (Table C3 and Figure C3 in Appendix C [online supplementary material]). Concretely speaking, the greatest effect of trees (from absence to presence) was found with W6 while the greatest effect of trees (from a few trees to many trees) was found with W5. Finally, there was also a significant interaction effect between "water features" and "paths" (chi square ¼ 25.51; p ¼ 0.013) (Table C4 and Figure C4 in Appendix C [online supplementary material]). The effect of paths from no paths to unpaved paths was greater if there was W5 present. The effect of paths from no paths to unpaved paths was greater in W6 while the effect of paths from unpaved paths to paved paths was greater in W4.

Relative importance of the characteristics within water features and each level of water features
The relative importance of each characteristic within water features is presented in Table 4. The results showed that bushes were the most important characteristic within water features, followed by upkeep. As shown in Figure 6, bushes and paths were the most and least important characteristic separately within five types of water features (W2, W4, W5, W6, and W7) while upkeep and flowers were the most and least important characteristic in situations where there were W1 and W3.

4.1.
What are the important characteristics in blue-green spaces for attracting nature experience?
Our findings suggested that water features were the most important characteristic for nature experience in blue-green spaces centered on a water body, followed by bushes, upkeep, and trees. However, four characteristics, namely wildlife, peers, flowers, and paths, were given relatively low importance. Overall, the physical characteristic related to environmental stewardship (e.g. upkeep) and to natural features (e.g. water features, bushes, and trees) was found to be the important characteristics in blue-green spaces for nature experience.
This study proved that water features were the most important characteristic for perception of and interaction with nature, which is consistent with the findings of previous studies. The presence of water played an important role in perception of nature (Herzog et al. 2000) because water was seen as the survival resource for humans and could satisfied the biological needs during evolution (Ohman 1986). Additionally, opportunities to interact with water were positively perceived (Van Hecke et al. 2018a, 2018b) and many kinds of activities were linked to water (Zhao, Xu, and Ye 2018), which may contribute to the preference for interacting with nature. Ponds with grassy slope revetment (W5) and rivers with gravel and cobblestone revetment (W6) were the two characteristic levels that most promoted nature experience. Undoubtedly, ponds, as still water bodies, could create a sense of tranquility and calmness. When combined with natural water compositions such as grassy slope revetments, they were found to be helpful in creating a sense of tranquility (Sakıcı 2015) and encourage Figure 6. Average relative importance of the seven characteristics within different levels of water features for interaction with nature. W1: no water; W2: streams with grassy slope revetment; W3: rivers with grassy slope revetment; W4: rivers with gravel and cobblestone revetment; W5: ponds with grassy slope revetment; W6: ponds with gravel and cobblestone revetment; W7: lakes with gravel and cobblestone revetment. contemplation, or the pleasure of sitting and watching (V€ olker and Kistemann 2015), which was conducive to the preference for nature experience. Meanwhile, the flow of water reduced tranquility but increased preference because moving water could stimulate and entertain people (Sorvig 1991). As flowing linear water, the flow of a river on gravel and cobblestone revetments created a sense of excitement. Next, using gravel and cobblestones together with rivers offered safer access and more interesting and varied river environments for play and exploration activities (Tunstall, Tapsell, and House 2004). Compared with ponds, streams were less important for attracting nature experience. This may be related to spaciousness, since ponds have a larger and wider body of water than streams and can provide a better-extended view. Previous research has shown that large open water areas are more preferred (Lu and Fu 2019). Thus, in addition to considering the types, revetments and sizes of different water features, using them to create a matching space atmosphere and create activity environments with good accessibility and hydrophilicity may enhance the preference for nature experience in blue-green spaces.
Our results indicated that the relative importance of bushes and trees was ranked second and fourth among the eight characteristics, respectively. It is clear from numerous studies that, as the natural features associated with structural vegetation, bushes and trees were both highly valued in aesthetic preference and visual quality (Akoumianaki-Ioannidou, Paraskevopoulou, and Tachou 2016) and predicted greater use of outdoor spaces (Zhang et al. 2013). Additionally, compared with tall trees, low bushes could reduce visual and physical accessibility, which have often been used as important predictor variables in preference studies (Hofmann et al. 2012). This may be why bushes received higher importance than trees in attracting nature experience. We also found that many trees and bushes had the highest utility values within their corresponding characteristics. These results contrasted with those of some earlier studies, which reported that dense vegetation could evoke the fear of crime (Kuo and Sullivan 2001;Kuo, Bacaicoa, and Sullivan 1998). However, other studies have shown that more vegetation likely increased human preferences and other positive perceptions (Van Den Berg, Hartig, and Staats 2007). Higher density vegetation was also associated with higher levels of perceived naturalness (Suppakittpaisarn et al. 2020). More trees and bushes could provide rich biodiversity, constitute a whole other world, and create a calm environment (Jahani and Saffariha 2020), attracting peoples' attraction and positively affecting their interaction with nature. Our results confirmed that dense trees and bushes increased the preference for nature experience.
Regarding the physical characteristics related to environmental stewardship, our results revealed that upkeep was the third most important characteristic. Previous research has shown that the absence of litter was significantly related to different uses of outdoor space, such as visiting (Van Hecke et al. 2018a), physical activity (Van Hecke et al. 2018b), and social contacts (Kemperman and Timmermans 2014). In this study, upkeep consisted of not only the presence of litter but also the maintenance of grass. Grass was more correlated with subjects' preferences than trees because trees may affect visibility, whereas grass does not (Chou, Lee, and Chang 2016). This may lead to the higher importance of upkeep than that of trees. Good upkeep (no litter and well-maintained grass field) was the most preferred characteristic level. It is clear that signs of disorder such as litter increased the risk of injury and were associated with an unsafe environment (Roberts et al. 2019), which in turn reduced the chances of nature experience. Besides, a well-maintained grass field was regarded as a sign of neatness. Neatness in urban planting was believed to be a prerequisite for attractiveness (Yang et al. 2019) and was regarded as an important aspect associated with the visual qualities (Kemperman and Timmermans 2014).

4.2.
How to improve the attraction for nature experience by integrations between characteristics in blue-green spaces?
By analyzing the interaction effects between different characteristics in blue-green spaces, we found that there were five and four significant interaction effects for perception of and interaction with nature, separately. Most of these interaction effects revealed the relationship between the four characteristics with low importance scores (e.g. wildlife, peers, flowers, and paths) and water features, bushes, and trees. This suggested effective strategies in helping the four characteristics to improve their attractiveness for nature experience.
The presence of wildlife promoted the use of green and blue spaces (Marselle et al. 2016;Garrett et al. 2019) and was considered a valuable attribute , but wildlife in this study was ranked only fifth out of eight characteristics. Previous studies have shown that age affects urban park users' cognition of local wildlife (Sang et al. 2016;Ishibashi et al. 2020). Nearly half of the participants in this study were under the age of 22. Compared with older participants, the younger participants had limited ability to perceive birds and butterflies ) and they usually underestimated the real species richness (Shwartz et al. 2014). The possible reason for this would be that watching wildlife in photographs was different from watching them in the real environment. This was essentially a still and two-dimensional experience, which did not reflect the dynamic experience of a real landscape (Wang and Zhao 2019). People like to experience nature through multiple sensory pathways, such as visual, auditory, and olfactory signals (Franco, Shanahan, and Fuller 2017). Moreover, the size of birds and butterflies was relatively small. Human recognition of animals was influenced by species size and they preferred animal species with large-sized bodies because of a greater likelihood of noticing large-sized species compared with small-sized ones ( _ Zmihorski et al. 2013). Further, interaction effects between "water features" and "wildlife" suggested that the effect of still water bodies (ponds and lakes) on perception of nature was greater than that of moving water bodies (streams and rivers) when the wildlife was from absence to presence. In fact, still water (ponds and lakes) not only promoted tranquility and serenity but also created open spaces (Sakıcı 2014), which were conducive to watching wildlife. Therefore, the questions of how to make people fully recognize the role and importance of wildlife in blue-green spaces through nature experience, and how to strengthen the connection between natural environments and wildlife by creating a good environmental atmosphere would stimulate more related research in the future.
Peers can influence adolescents' physical activity (Fitzgerald, Fitzgerald, and Aherne 2012;Van Hecke et al. 2018a). However, our results were inconsistent with those of the existing literature because peers in this study were ranked sixth out of eight characteristics. This may be because when people saw peers sitting or walking, they may a feel sense of crowding (Arnberger and Eder 2015), thereby reducing their intention toward nature experience. Moreover, the interaction effects between "peers" and "trees" indicated that trees could create a comfortable place for people to perceive nature better. Trees could make people relax (Gardsjord, Tveit, and Nordh 2014) and offer shade to improve temperature regulation (Akoumianaki-Ioannidou, Paraskevopoulou, and Tachou 2016). Moreover, interaction effects suggested that providing water features for peers could promote their interaction with nature, possibly because water offered people many kinds of activities such as cycling or jogging and playing with water (Zhao, Xu, and Ye 2018). Therefore, if we want blue-green spaces to provide opportunities and a high-quality environment for people to engage in nature experience together with peers, the positive physical characteristics related to natural features such as water and trees should be considered.
Regarding the natural features associated with decorative vegetation, flowers were regarded as the reliable predictor of preference because they could evoke a positive response in human beings (Ohman 1986). However, the impact of flowers on nature experience was not as great as we expected in our study. A possible explanation was that colored flowers indicated a more managed environment (planted flowers), whereas managed conditions could be related to less natural ones (Hwang et al. 2019). Part of tall flowers in our photographs would reduce the visibility and legibility of the environment, thus lowering preference ratings. Previous research has shown that low and ordered compositions of brightly colored flowers were the most preferred (Todorova, Asakawa, and Aikoh 2004;Kendal, Williams, and Williams 2012;Jiang and Yuan 2017) because they showed signs of care. This could make people feel more secure, thereby increasing preference (Suppakittpaisarn et al. 2020;Nassauer 2011). Additionally, results from the interaction analysis indicated that when flowers existed in blue-green spaces, the fewer bushes there were, the more favorable the nature experience (perception of and interaction with nature) was. This may be because the scene was less messy when there were fewer or no shrubs. Messy scenes showed signs of neglect and could be perceived as unsafe (Suppakittpaisarn et al. 2020). Adding more bushes would reduce the possibility of obtaining an overall grasp of the surrounding flowers, creating a sense of fear and danger by restricting visual accessibility (Arnberger and Eder 2015). As discussed above, the orderliness of flowers and the visibility of their surroundings should be taken into account when providing a perceived safe blue-green space for people's nature experience.
Our results revealed that paths were regarded as the least important characteristic for nature experience, supporting the existing literature regarding the negative effects of artificial elements on visual qualities and preferences (Arriaza et al. 2004). Additionally, results from the interaction analysis indicated that flowers and water can help paths achieve higher importance in perception of and interaction with nature, respectively. Previous research has shown that pathways with flowers and water at their sides, or with a visual connection to water, were preferred (Zhai and Baran 2017), possibly because flowers and water can attract people's attention and reduce the negative feelings caused by artificial facilities such as paths. Thus, paying attention to pathway design characteristics associated with flowers and water to improve path quality may enhance experience and preference for blue-green spaces.

4.3.
What is the difference between single green spaces and blue-green spaces in attracting nature experience?
We found that in terms of blue-green spaces without water, namely, single green spaces, upkeep was the most highly rated characteristic for perception of and interaction with nature. Although landscapes containing water were considered to have appealing aesthetic qualities, people sometimes considered water areas as unsafe (Van Hecke et al. 2018b). The upkeep of green spaces, such as the presence of litter, affected the visual qualities of the setting and was also an important aspect associated with safety (Kemperman and Timmermans 2014). Additionally, the preference for lawn landscapes was likely innate and developed over the course of human evolution (Falk and Balling 2010). People usually associated well-kept lawns with being active (Deng et al. 2020). This may stimulate people's intention to engage in some activities, such as experiencing nature. Thus, more attention should be paid to upkeep (the presence of litter and the maintenance of grass) in green spaces not only for their aesthetic value and safety (Al-Akl et al. 2018) but also because they could evoke continuous human presence and promote people's participation in activities. For both perception of and interaction with nature, our results indicated that bushes and trees ranked high in importance among streams, rivers, ponds, and lakes, which means characteristics related to structural vegetation (e.g. bushes and trees) are the most important in blue-green spaces. This finding is consistent with that of previous research. For example, the presence of vegetation strongly affected positive perceptions of blue spaces (V€ olker and Kistemann 2011) and plants accompanied by water potentially enhanced the attractiveness of water landscapes (Hu, Yue, and Zhou 2019). This result indicated that what improves peoples' preference for blue-green spaces was not only the water but also the integration between water and its surrounding structural vegetation.
Overall, the characteristics of single green spaces and blue-green spaces in attracting nature experience are different. Giving priority to upkeep in single green spaces could provide an aesthetic and safe space for nature experience. Compared with single green spaces, appropriate integration between water and its surrounding structural vegetation makes blue-green spaces more pleasant. This contributes to the preference for nature experience. The findings further suggested that controlled or refined versions of natural environments could attract people to experience nature because of their preference for a balance between wild nature and human control (Hu, Yue, and Zhou 2019).

Strengths and limitations
The main strength of this study was the method it used (i.e. the choice-based conjoint method using manipulated photographs) to answer the research questions. This method allowed for the study of many characteristics simultaneously. It mimicked real-life situations, where people need to decide which blue-green space they prefer to visit after considering all physical and social characteristics.
Additionally, it did not require participants to recall particular characteristics of blue-green spaces when completing surveys. The method allowed for the study of the effects of variation within and between manipulated characteristics under controlled conditions. Finally, integrating these manipulated photographs into a CBC task did not require participants to rate or rank all 6,804 photographs. Instead, from 12 choices (a major reduction in effort), we could obtain the ranking of different characteristics and their specific levels.
Nevertheless, there are certain limitations to this study. First, the study was only concerned with people's preferences for the two aspects of perception of and interaction with nature. The mechanism of how characteristics influence nature experience was not considered. The study also did not link the preference to actual behavior. Future studies through on-site research could determine how the desired characteristics result in nature experience based on people's psychological feelings and whether blue-green spaces that have the desired characteristics can attract various behaviors related to nature experience. Second, this study did not include other characteristics that had significant impacts on the use and preference for green or blue spaces, such as amenities and natural sounds (Mcdougall et al. 2020). Therefore, future studies should combine conjoint methodology and three-dimensional methods such as virtual reality to explore the effect of these characteristics of blue-green spaces on nature experience. Third, most of the participants were recruited by students using snowball sampling, which resulted in relatively young and less representative participants. This tended to produce a conservative estimate of the true results because frequencies of natural experience were larger for older versus younger participants (McMahan and Estes 2015). Future studies need larger mixed-age samples and participants with a broad range of backgrounds to fully examine the validity of the findings in a more general context. Finally, it remains unclear whether these desired characteristics are more or less important for specific subgroups of participants (e.g. male versus female, younger versus older people, and frequent versus irregular nature experience). Thus, it is essential to study particular subgroups in the future to further understand and promote nature experience in blue-green spaces.

Conclusions
Our study adopted conjoint analysis to rank different characteristics and discussed interaction effects between them. Additionally, it compared the results obtained for green spaces with the those obtained for green-blue spaces. In general, few differences were observed for the relative importance of characteristics for perception of nature compared to the relative importance of characteristics for interaction with nature. Our findings suggested that priority should be given to providing water features to increase the preference for nature experience (perception of and interaction with nature) in blue-green spaces, followed by bushes, upkeep, and trees. For the four less important characteristics (e.g. wildlife, peers, flowers, and paths), integrating them with vegetation and water features can improve their attractiveness for nature experience. In particular, integrating peers, flowers, and paths with vegetation can increase the preference for perception of nature, whereas integrating peers and paths with water features can improve their attractiveness for interaction with nature. For single green spaces, upkeep should be prioritized to attract people to the nature experience, but for blue-green spaces, the integration between water and structural vegetation (e.g. bushes and trees) is most important. In addition to revealing the importance of water, bushes, and trees, our findings highlight the attraction of upkeep to nature experience. Previous studies conducting the same conjoint analysis showed that upkeep was the most important characteristic for park visits and physical activity (Van Hecke et al. 2018a;Veitch et al. 2017). Our results confirmed that upkeep should be considered when improving environmental quality to encourage people's activities. These findings not only contribute to fostering a new understanding about which characteristics of blue-green spaces matter to those people searching for nature experience but also assist us in identifying characteristics that should be prioritized and in further considering the integration between characteristics to increase nature experience.
With rapid urbanization, urban natural spaces including green spaces and open water have decreased sharply, and it is becoming more difficult to develop new natural spaces in compact urban areas for people to use (Lau, Yung, and Tan 2021). Further, this rapid urbanization has led to concerns among policy makers about the potential effects on the quality of human life (Dye 2008), especially in regard to contact with the natural environment. Actually, opportunities for such contact are dramatically limited in urban areas (Barbosa et al. 2007). Thus, with rapid urbanization and declines in human contact with nature globally, crucial decisions must be made about how to preserve and enhance opportunities for nature experience (McEwan et al. 2020). Simply increasing the quantity of green and blue spaces may not be sufficient to enhance the opportunities to experience nature. What is needed is to find ways to enhance their quality. Examining the characteristics of blue-green spaces may contribute to a more profound understanding of this. We advocate focusing on the design of natural features when creating blue-green spaces that are beneficial to nature experience. For example, we can use different types of water features to form a sense of tranquility or excitement, use structural vegetation (e.g. trees and bushes) to create comfortable places, and use decorative vegetation (e.g. flowers) to reduce the negative impact of artificial facilities and support the construction of nature experience environments. Most important, if the blue-green space has good upkeep, its quality is likely to be at its best, providing an aesthetic and safe space to attract people to experience nature. It is also vital to consider the integration between different characteristics. Providing an open waterscape space for watching wildlife, creating waterside activity spaces with good hydrophilicity and accessibility for peers, reducing the number of bushes to ensure visual accessibility of flowers' surroundings, and configuring flowers and water on the roadside are all effective ways to improve people's preference for nature experience in blue-green spaces. Overall, the priority of different characteristics and the importance of integration between characteristics of blue-green spaces should be further considered to ensure that opportunities to experience nature are increased and improved as urbanization continues.