Species richness alone does not predict cultural ecosystem service value

Significance

Sustaining biodiversity and ecosystem services are common conservation goals. However, understanding relationships between biodiversity and cultural ecosystem services (CES) and determining the best indicators to represent CES remain crucial challenges. We combined ecological and social data to compare CES value of wildflower communities based on observed species richness versus revealed social preferences. Using a discrete-choice experiment with images of wildflower communities, we analyzed which aspects of biodiversity were associated with the aesthetic preferences of forest visitors. Although commonly used to indicate biodiversity-based CES, species richness did not predict aesthetic preference. This study suggests that successful management of CES requires understanding stakeholders’ preferences to determine conservation priorities.

Abstract

Many biodiversity-ecosystem services studies omit cultural ecosystem services (CES) or use species richness as a proxy and assume that more species confer greater CES value. We studied wildflower viewing, a key biodiversity-based CES in amenity-based landscapes, in Southern Appalachian Mountain forests and asked (i) How do aesthetic preferences for wildflower communities vary with components of biodiversity, including species richness?; (ii) How do aesthetic preferences for wildflower communities vary across psychographic groups?; and (iii) How well does species richness perform as an indicator of CES value compared with revealed social preferences for wildflower communities? Public forest visitors (n = 293) were surveyed during the summer of 2015 and asked to choose among images of wildflower communities in which flower species richness, flower abundance, species evenness, color diversity, and presence of charismatic species had been digitally manipulated. Aesthetic preferences among images were unrelated to species richness but increased with more abundant flowers, greater species evenness, and greater color diversity. Aesthetic preferences were consistent across psychographic groups and unaffected by knowledge of local flora or value placed on wildflower viewing. When actual wildflower communities (n = 54) were ranked based on empirically measured flower species richness or wildflower viewing utility based on multinomial logit models of revealed preferences, rankings were broadly similar. However, designation of hotspots (CES values above the median) based on species richness alone missed 27% of wildflower viewing utility hotspots. Thus, conservation priorities for sustaining CES should incorporate social preferences and consider multiple dimensions of biodiversity that underpin CES supply.

Sustaining ecosystem services is an emerging priority in sustainability science, and conservation plans increasingly emphasize joint protection or improvement of ecosystem services and biodiversity. Simultaneous concern for biodiversity and ecosystem services led to establishment of the Intergovernmental Platform on Biodiversity and Ecosystem Services as well as multiple national, regional, and local initiatives (1, 2). Despite recognition that the futures of biodiversity and ecosystem services are interconnected (3), understanding the direct links between biodiversity and ecosystem services and determining the best indicators to represent ecosystem services remain crucial challenges (4–7). Biodiversity is defined and measured in a multitude of ways (e.g., species richness, species evenness, genetic diversity, functional diversity, and community distinctness) (8, 9). In biodiversity and ecosystem service research, species richness is the most frequent unit of measure (6) and hypotheses regarding increased biodiversity are often stated in terms of increased species number [e.g., more species confer greater cultural ecosystem services (CES) value] (10, 11).

Studies of biodiversity-based ecosystem services rarely assess alternate metrics of biodiversity and seldom provide empirical links between biodiversity indicators and social preferences for ecosystem services (12). CES—the nonmaterial benefits provided by ecosystems (13)—are among the least-quantified ecosystem services (14–16). Due to their normative nature and often abstract definitions (17), CES can be challenging to study. They represent complex relationships between people and ecosystems, and the definition and valuation of a particular cultural service can vary across stakeholders (18, 19).

For biodiversity-based CES, common practice has been to map species richness as an indicator and use those maps to assess the spatial provision of CES (see, e.g., refs. 17 and 19–22). However, there is little known about whether maps of species richness correspond to actual social preferences for CES. Biodiversity conservation depends on the values that people attach to it (23, 24) and understanding people’s preferences for biodiversity can facilitate communication between the public and land managers and help delineate publicly supported conservation goals (18). In particular, if social preferences can be translated to maps of CES indicators (25) a more complete assessment of conservation objectives targeted at maintaining biodiversity and CES is possible (26, 27).

Aesthetic beauty is a commonly cited CES in amenity-based landscapes (13, 28, 29) and is often assumed to be positively correlated with biodiversity (30, 31). Species-rich, flower-rich views improve the aesthetic value of landscapes, roadsides, field margins, and meadows (32–36), and increased flower color diversity may provide higher CES value, especially in rural landscapes (34, 37–40). Lindemann-Matthies et al. (40) demonstrated that aesthetic appreciation increased with perceived species richness. Moreover, aesthetic appreciation and perceived species richness also increased with evenness (i.e., the equitability of species in a community), suggesting compositional diversity may also be an important driver of aesthetic preference. Additionally, the presence of species with cultural significance or the presence of rare species can increase satisfaction among wildflower viewers (41) and the aesthetic value of particular species has been used as a reason for conservation (42, 43).

We studied the aesthetic preferences of public forest visitors for trailside wildflower communities to test whether species richness predicted CES value. We conducted the study in the Southern Appalachian Mountains of North Carolina, where wildflower viewing and photographing is one of the fastest-growing outdoor recreational activities (44). Wildflower blooms provide important CES to both residents and tourists (45). The region’s high biodiversity, recognized globally, attracts both residents and visitors, many of whom visit public forests to participate in recreation and observe plants and animals (44, 46, 47). We asked three questions about the relationship of biodiversity to CES value: (i) How do aesthetic preferences for wildflower communities vary with components of biodiversity, including species richness?; (ii) How do aesthetic preferences for wildflower communities vary across psychographic groups? and (iii) How well does species richness perform as an indicator of CES value compared with revealed social preferences for wildflower communities?

Public forest visitors were asked to choose among digitally manipulated images of wildflower communities with varied levels of flower species richness, flower abundance, species evenness, color diversity, and presence of charismatic species, as identified from regional tourism websites (SI Text and ref. 48) (Fig. 1). Wildflower community preference was analyzed using multinomial logit models that were then used to predict wildflower viewing utility of actual wildflower communities. This analysis is consistent with Lancaster’s theory of value (49) and random utility theory (50), which assume that individuals prefer goods or services based on the utility derived from the attributes of those goods or services, and that individuals choose options based on their relative utility. Because individual preferences, beliefs, and expertise may affect aesthetic preferences (51), we tested whether the effect of wildflower diversity on aesthetic preferences varied across psychographic profiles. Finally, using data collected from actual wildflower communities in the study region (48), we compared site prioritization for CES based on empirically measured species richness versus predicted aesthetic preference (i.e., wildflower viewing utility).

Fig. 1.

Examples of digitally manipulated images used in the discrete-choice experiment. A total of 48 images were used. Images all used the same background (see large panel) but varied in the flower species richness, flower abundance, the number of colors, and the evenness of the wildflower community. The cutouts displayed here (A–D) are illustrative of variation in the images and were selected from images that varied in the number of colors and evenness; species richness and flower abundance were held constant at five species and 90 flowers in these examples. A shows one color and low evenness, B one color and high evenness, C five colors and low evenness, and D five colors and high evenness. Based on the results of the discrete-choice experiment, D would have the highest predicted wildflower viewing utility. Fig. S2 for uncropped image examples.

Fig. S2.

Four example images showing differences in wildflower community. Images vary only in the number of colors and evenness, with species richness and flower abundance held constant at five species and ninety flowers, respectively. Image A shows one color and low evenness, B one color and high evenness, C five colors and low evenness, and D five colors and high evenness. Based on the results of the discrete-choice experiment, D would have the highest predicted wildflower viewing utility.

Results

We collected usable responses from 293 public forest visitors, representing a cross-spectrum of ages, visitation characteristics, and attitudes (Table S1). Respondents tended to be white (90%) and well-educated (73%), which is representative of recreational visitors in this area (44). Respondents were grouped into segments ranging in size and psychographics based on their attitudes toward forest-based CES, measured along four attitudinal dimensions (Table 1). Thirty-four percent of respondents were generalists, characterized by their high valuation of all forest-based CES (e.g., quiet relaxation, experiences, active escape, and collecting things) (Table 1). The remaining respondents were divided among those that placed high value on active escape (26%), quiet relaxation (27%), or collecting (12%). Nearly half (46%) of respondents reported having visited a forest to view wildflowers within the past year (Table S2).

Table S1.

Characteristics of the sample population

Characteristics	%
Sex
Male	55
Female	45
Age, y
18–24	6
25–34	22
35–44	20
45–54	17
55–64	16
65 and over	18
Education
High School or equivalent	9
Some college/2-y degree	18
4-y college degree	44
Advanced degree	29
Race
African American/Black	2
Asian	2
Caucasian/White	90
Hispanic	4
Native American	2
Pacific Islander	<1
Other	<1
Frequency of visits to public forest annually
One to a few times per year	23
About once per month	17
A couple times per month	18
About once per week	16
Multiple times per week	26

Table 1.

Factor loadings for each forest-based CES and mean factor scores for each psychographic segment

Category	Quiet relaxation	Experiences	Active escape	Collecting
Forest-based CES	Factor loadings
To find solitude	0.80	0.05	0.13	0.16
Spiritual value	0.74	0.17	0.14	0.09
To relax	0.62	0.18	0.31	0.12
To hear nature sounds	0.57	0.50	−0.07	0.07
To see scenic views	0.13	0.73	0.16	−0.14
To be with family and friends	−0.25	0.65	0.48	0.17
To view wildlife	0.27	0.65	−0.04	0.30
To view wildflowers	0.39	0.64	−0.10	0.01
To participate in recreation	0.05	0.02	0.78	0.06
To be physically active	0.24	−0.03	0.68	−0.12
To escape an urban setting	0.42	0.18	0.43	−0.02
Educational value	0.29	0.34	0.40	0.37
To hunt	0.04	−0.03	0.01	0.80
To collect food	0.16	0.09	−0.03	0.77
Psychographic segment (n)	Mean factor score
1: Active/experience seekers (77)	−1.27	0.13	0.16	−0.26
2: Quiet seekers (79)	0.48	−0.70	−0.89	−0.38
3: Collectors (36)	0.11	−0.27	−0.02	2.15
4: Generalists (101)	0.55	0.57	0.53	−0.26

Factor loadings along four interpretable dimensions (quiet relaxation, experiences, active escape, and collecting) of respondents’ attitudes toward forest-based CES. Factors were extracted from survey response data using principal components solution with varimax rotation. The highest factor loadings for each forest-based CES are bolded. Cluster analysis based on the attitudinal factors identified four psychographic segments of respondents. The segments differed in group size (n) and mean scores among the four attitudinal dimensions.

Table S2.

Visitors’ reported participation in a range of activities during visits to public forests within 1 y of being surveyed and motivations for visiting on the day of the survey

Activity	Past year, %	This visit, %
Hiking on trails	92	51
Viewing waterfalls	72	26
Walking for pleasure or exercise	81	24
Viewing scenery	79	17
Nature viewing	83	16
Swimming	65	15
Mountain biking	35	14
Photographing nature	64	10
Camping	60	9
Picnicking	54	8
Wildflower viewing	46	8
Fishing	39	8
Running on trails	33	7
Wildlife watching	54	4
Birdwatching	27	3
Backpacking	29	3
Collecting fruits or berries	30	1
Collecting things (e.g., sticks/rocks)	27	1
Canoe/kayak/boating	42	1
Collecting mushrooms	9	<1
Horseback riding	6	<1
Collecting medicinal plants	4	<1
Hunting	10	0
Geocaching	5	0

Aesthetic Preferences for Wildflower Communities.

People’s aesthetic preferences for wildflower communities varied with components of wildflower diversity but not with flower species richness. Flower species richness had no effect on respondents’ aesthetic preference for images of wildflower communities (Table 2). The abundance of flowers was the most important predictor of aesthetic preference, followed by number of colors and evenness. Photographs displaying wildflower communities with higher bloom abundance, more colors, and higher evenness were more likely to be preferred.

Table 2.

Relative importance of wildflower community attributes from multinomial logit models based on respondent preference for digital photos of wildflower displays

	Relative importance of wildflower community attribute
Model (n)	Species richness	Flower abundance	No. of colors	Evenness	Presence of charismatic species
All respondents (293)	0.02	0.53	0.24	0.12	0.09
		Psychographic segments based on attitudes toward forest CES
1. Active/experience seekers (77)	0.01	0.50	0.27	0.14	0.09
2. Quiet seekers (79)	0.05	0.56	0.18	0.12	0.09
3. Collectors (36)	0.05	0.50	0.28	0.13	0.05
4. Generalists (101)	0.01	0.49	0.28	0.11	0.11
Wald (=)	0.53	1.05	10.34	0.56	1.32
		Segments based on attitude toward wildflower viewing
Flowers less important (78)	0.11	0.54	0.11	0.15	0.08
Flowers mores important (210)	0.01	0.50	0.27	0.11	0.10
Wald (=)	2.96	0.02	8.33*	0.34	0.19
		Segments based on knowledge of local flora
Novice (77)	0.02	0.59	0.18	0.15	0.07
Intermediate (174)	0.02	0.49	0.29	0.12	0.08
Expert (42)	0.03	0.49	0.23	0.08	0.17
Wald (=)	0.23	4.40	5.03	1.30	2.80

The first model is based on all respondents. The remaining models analyzed segments of the respondents based on their attitudes toward forest-based CES, attitudes toward wildflower viewing, and knowledge of local flora. Relative importance values provide a measure of the relative effect of each attribute. Table S3 for full model results and coefficient estimates. Significant Wald (=) values indicate differences in the estimated coefficient of an attribute between segments.

^*P < 0.05.

Aesthetic Preferences Among Psychographic Segments.

Results were remarkably consistent across all four psychographic segments of the respondents, indicating no difference in preference patterns among groups (Table 2 and Table S3). Similarly, preference patterns did not differ based on respondents’ knowledge of local flora (i.e., novice, intermediate, or expert) or the value they placed on wildflower viewing (i.e., flowers more or less important) (Table S3).

Table S3.

Model estimates for the discrete-choice model

Attributes and levels	All respondents	Segments based on attitudes toward forest-based CES					Classes based on attitude toward wildflower viewing			Classes based on knowledge of local flora
		1	2	3	4	Wald (=)	Less important	More important	Wald (=)	Novice	Intermediate	Expert	Wald (=)
Flower species richness	−0.01	0.00	−0.02	−0.02	0.00	0.53	−0.05	0.01	2.96	−0.01	−0.01	−0.01	0.23
Flower abundance	0.02†	0.03†	0.02†	0.02†	0.03†	1.05	0.02†	0.02†	0.02	0.03†	0.02†	0.02†	4.4
No. of colors						10.34			8.33*				5.03
1	−0.61†	−0.79†	−0.29	−0.67†	−0.74†		−0.23†	−0.76†		−0.52†	−0.70†	−0.46†
2	0.08†	0.16†	−0.03	−0.02†	0.16†		−0.15†	0.18†		0.06†	0.11†	0.01†
3	0.24†	0.27†	0.33	0.36†	0.10†		0.20†	0.24†		0.25†	0.18†	0.50†
5	0.29†	0.35†	−0.01	0.33†	0.49†		0.18†	0.34†		0.21†	0.41†	−0.04†
Evenness						0.56			0.34				1.3
Low	−0.24†	−0.30†	−0.21†	−0.23†	−0.25†		−0.29†	−0.23†		−0.31†	−0.24†	−0.16†
High	0.24†	0.30†	0.21†	0.23†	0.25†		0.29†	0.23†		0.31†	0.24†	0.16†
Charismatic species						1.32			0.19				2.8
Present	−0.18†	0.19†	0.15†	0.09†	0.23†		−0.16†	−0.20†		−0.15†	−0.16†	−0.34†
Absent	0.18†	−0.19†	−0.15†	−0.09†	−0.23†		0.16†	0.20†		0.15†	0.16†	0.34†
Pseudo R2	0.25	0.28	0.21	0.23	0.28		0.23	0.26		0.32†	0.23†	0.26
n	293	77	79	36	101		82	211		77	174	42

Model estimates for the discrete-choice model based on all respondents (i.e., single-class) or segments based on respondents’ attitudes toward forest-based CES, attitude toward wildflower viewing, and knowledge of local flora. Significant Wald (=) values indicate differences in the estimated coefficient of an attribute between segments. Significance: ^†P < 0.001 and *P < 0.05.

Species Richness Versus Revealed CES Value in Actual Wildflower Communities.

Empirically surveyed wildflower communities (n = 54) varied in flower species richness, flower abundance, evenness of species in bloom, number of colors, and whether charismatic species were present and blooming (48). Overall flower species richness ranged from 2 to 34 ($\overline{x}=$ 11 and SD =7.3). Wildflower viewing utility calculated using multinomial models of revealed preferences (Table S3) ranged from −0.11 to 13.29 ($\overline{x}=2.2\hspace{0.5em}\text{and}\hspace{0.17em}\text{SD}=2.5$). For surveyed wildflower communities, predicted CES value (i.e., wildflower viewing utility) was correlated with the overall species richness observed at a site (Spearman’s rho = 0.66, Fig. 2). Species richness was also correlated with aesthetic traits of flower abundance, evenness, color diversity, and number of charismatic species present (Pearson’s r 0.48–0.77, all P < 0.001). When sites were classified as CES hotspots (CES values above the median) based on either wildflower viewing utility or overall species richness, classification broadly agreed, with 34 hotspots identified by both indicators. However, site classification based on species richness alone missed 27% (seven) of the sites predicted to have high wildflower viewing utility (Fig. 2). Similarly, ranking sites based on wildflower viewing utility alone missed 29% (eight) of sites predicted to have the highest flower species richness.

Fig. 2.

Predicted wildflower viewing utility correlated with overall flower species richness of 54 wildflower communities in the southern Appalachian Mountains. Dashed lines show the hotspot classification thresholds (median values) for each indicator. Sites were either not hotspots (below the median, bottom left quadrant), agreed-upon hotspots (above the median, top right), or hotspots based on either wildflower viewing utility (top left) or species richness (bottom right) but not the other. Closed circles indicate agreement, and open symbols indicate disagreement in hotspot classification. Wildflower community data were collected over an 18-wk period between April and August 2014 (see ref. 48 for details); discrete-choice data used to predict wildflower viewing utility were collected in 2015.

Discussion

Conservation planning and management increasingly require consideration of both ecosystem services supply and maintenance of biodiversity. However, despite calls for holistic management of a full suite of ecosystem services to achieve landscape sustainability (52–54) CES have been largely absent from biodiversity and ecosystem service literature. We linked stakeholders’ revealed preferences with empirical measurements of wildflower community diversity and demonstrated that only partial overlap exists between high species richness and high CES supply. Species richness per se was not a significant predictor of aesthetic preference, and site rankings based on empirically measured wildflower communities showed that the use of observed species richness as a CES indicator does not fully encompass sites with high predicted CES value. Thus, management of biodiversity-based CES and conservation of species diversity should be considered complementary, but different, goals when developing landscape conservation targets (55).

People’s aesthetic preference for images of trailside wildflower communities was driven primarily by the abundance of flowers and not by species richness of flowers. However, people preferred wildflower communities with more colors, suggesting that although respondents may not distinguish between flower species if they are the same color (Fig. 1 A and B) they recognize diversity in colors (Fig. 1 C and D). Our models suggest people respond to a complex combination of these floral traits, which were generally correlated with species richness in our study area, but not perfectly so. Because perceived species richness has been linked to aesthetic value and support for biodiversity conservation (40), misperceptions of the species richness in wildflower communities with lower color diversity could lead to biases in people’s attitudes toward these wildflower communities. Our study did not test explicitly whether people judged wildflower communities with more colors to be more species-rich, which limited our ability to judge whether visitors preferred wildflower communities that they perceived as more species-rich. If people’s perception and preferences are closely linked (56), and people’s perception of species richness does not match actual species richness (57), promoting education that emphasizes knowledge about species diversity could increase appreciation of sites with high flower diversity but low color diversity.

Our study suggests that targeting management at sites with high wildflower viewing utility will yield benefits across a spectrum of visitors. People value nature for many different reasons including intrinsic, economic, emotional, spiritual, or psychological values that are not mutually exclusive (24). Landscape aesthetic preferences can vary based on age (56), gender (58), cultural and social groups (59–61), and recreation patterns (58, 60). However, preferences for wildflower communities in this study were remarkably similar across demographic, attitudinal, and recreational groups and were instead driven by the attributes of the wildflower communities. These results suggest that variation in aesthetic preference is greater among sites than across public forest visitors (61, 62). Because aesthetic appreciation and scenic beauty are desired conditions in recreation and outdoor tourism in amenity-based landscapes (44, 63, 64), understanding how to manage aesthetic CES can have positive impacts for residents and visitors to these areas.

Aesthetic preference varies among persons. Whereas preferences among psychographic groups were similar, the discrete-choice models explained only about 30% of the variation among individual respondents. Cultural preference theories contend that the attitudes of each individual are in constant flux and are shaped by cultural and personal experiences (e.g., ref. 65). Both biophysical and personal–social situational context affects aesthetic experience (30). In our study, we tested both the biological factors (i.e., wildflower community traits) and cultural factors (i.e., age, gender, and botanical knowledge). Unmeasured factors related to a person’s attitude and ethnic and cultural background could explain the remaining variation, but this information was beyond the scope of this study and impractical to collect under the field conditions.

In conservation and sustainability science, determining how to best conserve the biosphere while meeting the needs of humans has led to vigorous debate. Although increasing recognition of ecosystem services and the contribution of ecosystems and biodiversity to human well-being has the potential strengthen conservation (1, 28, 66–69), some authors have suggested that increased emphasis on ecosystem services as a conservation goal may lead to unintended losses and inadequate protections for biodiversity (8, 70, 71). In part, this debate stems from lack of clarity about the multiple relationships between biodiversity and ecosystem services (66). Studies have revealed both positive and negative relationships between priority areas for biodiversity and priority areas for the provision of ecosystem services, complicating landscape conservation planning (56–59). To preserve aesthetic beauty and the CES provided by wildflower communities, some maintenance of species diversity, which allows for a diversity of flower forms and colors, is important. However, despite correlations between richness and CES value in wildflower communities, conservation and management priorities based solely on maintaining species richness may not adequately conserve sites that supply biodiversity-based CES. Conservation priorities targeted at achieving sustainability of CES should use suitable indicators, beyond measures of species richness, that incorporate social preferences and recognize the multiple ways that biodiversity may contribute to the provision of ecosystem services.

Methods

Study Area.

We collected empirical data on wildflower communities (48) and people’s aesthetic preferences in the French Broad River Basin (FBRB) in western North Carolina during the summers of 2014 and 2015. The FBRB, located within the southern Appalachian Mountains, covers an area of 7,330 km² (Fig. S1). The region is characterized by complex terrain and is known for its high biodiversity and popular for ecotourism (46) (see SI Text for more detail). Approximately 75% of the FBRB is forested, mainly second growth, with spruce-fir (Picea-Abies) and northern hardwoods at high elevations, mixed hardwood species at lower elevations, and mixed mesophytic forests on lower slopes and coves (46). The regional economy changed in the last century from resource extraction (e.g., timber) and agricultural production to a nature-based, amenity-driven economy, leading to altered patterns of land use and increased exurban development (72–74). Land-use changes have altered plant communities within the region (73, 75–77) and likewise affect the location and abundance floral resources within the study area (48).

Fig. S1.

Study area and public forest locations in the FBRB, North Carolina.

Aesthetic Preferences for Wildflower Communities.

We surveyed 295 public forest visitors using a convenience sampling approach at trailheads on national forest and state forest properties. Face-to-face surveys were conducted at trailheads and visitor information points during the summer of 2015. We varied the day of the week and time of day that each trailhead was visited, and individual surveys generally lasted less than 5 min. Once a survey was completed, the next visitor encountered was asked to participate in the study. At remote trailheads with limited use, we posted signs asking people to complete an online version of our survey. Online respondents accounted for 5% of our respondents. This study was approved by the University of Wisconsin–Madison Education and Social/Behavior Science institutional review board (IRB no. 2015-0384). All interviewees gave their informed consent to participate in the study.

The survey (SI Text) consisted of three parts: (i): respondents’ attitudes toward a set of CES provided by public forests, (ii) respondents’ recreational patterns and social and demographic data, and (iii) a discrete-choice experiment to determine preferences for different wildflower communities. Respondents’ attitudes toward forest-based CES were measured with the help of 15 statements about forest uses (Table S4). Respondents indicated their personal valuation of each service on a five-point Likert-type scale (1: unimportant, 2: somewhat important, 3: important, 4: very important, and 5: extremely important). Respondents were asked to provide an estimate of the frequency with which they visited public forests in the last year and what activities they participated in while visiting public forests. We also asked them for their gender, age, race, highest level of education, and a self-assessment of their knowledge of plants in the area (1: no knowledge, 2: novice with some knowledge, 3: intermediate knowledge, and 4: expert knowledge).

Table S4.

Respondents’ rating of the importance of different forest-based ecosystem services

Forest-based CES	Mean importance rating	SE
Be outdoors	4.51	0.04
Be physically active	4.25	0.06
See scenic views	4.25	0.05
Participate in recreation	4.15	0.06
Hear nature sounds	4.12	0.06
Escape an urban setting	4.11	0.06
Relax	4.02	0.06
See wildlife	3.93	0.06
Find solitude	3.87	0.07
Be with family and friends	3.72	0.07
Spiritual value	3.47	0.08
Educational value	3.42	0.08
See wildflowers	3.36	0.07
Collect food	1.55	0.06
Hunt or fish	1.29	0.05

All values were measured on five-step rating scales, with 1: unimportant, 2: somewhat important, 3: important, 4: very important, and 5: extremely important. Mean scores and SEM were derived from raw data. The survey question was phrased to read “How important is it to you personally that a public forest is a place to…?”

Preferences for wildflower communities were obtained using a discrete-choice experiment where respondents were shown 8.5- × 11-inch photographs of near-view forest wildflower communities manipulated to contain different levels of diversity (i.e., flower abundance, flower species richness, number of colors, evenness or the distribution of abundance among species in a community, and presence of charismatic species). Respondents were asked to indicate their preferred alternative between pairs of digital images of wildflower communities. Each respondent was shown four pairs of images, or choice sets. The images were created using Adobe Photoshop and choice sets varied according to a D-efficient sampling design (78, 79), which maximizes the amount of information about each parameter through the most efficient number of choice sets. The choice model included 48 images organized in six blocks of four choice sets (Table S5).

Table S5.

Wildflower community characteristics of the 48 images used in the discrete-choice experiment

		Alternative 1					Alternative 2
Choice set	Block	Species richness	Abundance	No. of colors	Evenness	Charismatic species	Species richness	Abundance	No. of colors	Evenness	Charismatic species
1	1	5	90	3	0	1	10	10	1	0	0
2	1	3	35	3	1	1	5	90	2	0	0
3	1	3	5	3	1	0	10	90	2	1	1
4	1	3	35	3	1	0	5	5	2	0	1
5	2	3	10	3	0	1	5	5	5	1	0
6	2	5	90	1	0	1	3	5	2	1	0
7	2	5	5	2	0	0	3	10	3	1	1
8	2	10	10	3	1	1	1	35	1	0	0
9	3	1	90	1	0	1	3	35	3	0	0
10	3	1	5	1	0	0	5	10	2	1	1
11	3	10	90	2	0	0	5	10	3	0	1
12	3	3	10	2	0	1	5	90	5	0	0
13	4	10	90	1	1	0	1	90	1	0	1
14	4	3	90	2	1	0	10	90	1	0	1
15	4	10	35	3	0	0	5	90	1	1	1
16	4	1	10	1	1	0	3	90	3	1	1
17	5	3	5	1	1	1	1	10	1	0	0
18	5	10	90	5	1	1	3	90	2	0	0
19	5	5	10	5	0	0	3	35	3	1	1
20	5	3	35	1	0	0	1	5	1	1	1
21	6	5	90	5	1	0	10	35	3	0	1
22	6	3	5	3	0	1	5	35	3	1	0
23	6	3	35	2	0	1	10	10	3	1	0
24	6	1	35	1	1	1	3	5	3	0	0

Wildflower community characteristics varied per a d-efficient design. Respondents were randomly assigned to a block, which consisted of four choice sets. Each choice set included a pair of images of which respondents indicated their preferred alternative. Evenness had two levels: low (<0.5) coded as 0 and high (>0.5) coded as 1. Charismatic species presence is binary where 1 = present, 0 = not present.

We used factor analysis to identify interpretable dimensions of attitudes toward forest-based CES. Factor analysis has been used previously to study psychographics of survey respondents in nature recreation, ecotourism, and ecosystem service research (33, 80). Exploratory factor analysis identified a four-factor structure describing people’s attitudes toward forest-based CES and accounted for 60% of the variance in the dataset (Table 1). We performed K-means cluster analysis to identify segments of respondents with different psychographic profiles based on their attitudes toward forest-based CES, represented by their scores along the four factors (80, 81). The effect of biodiversity attributes on aesthetic preference for wildflower communities was modeled using multinomial logit models (SI Text). We first analyzed the preferences of all respondents, without regard for psychographics or demographics, including only the wildflower community attributes. We tested whether the inclusion of interactive effects between the wildflower community attributes and respondents’ attitudes toward forest-based CES improved the model fit. We then ran multinomial logit models for groups based on people’s preference of different CES, knowledge of plants, demographics, and recreational patterns to determine whether the effect of biodiversity attributes varied across segments. See SI Text for more details.

Indicators of CES Value in Actual Wildflower Communities.

We used wildflower community data recorded in the study region (48) to evaluate differences between the designation of CES hotspots based on empirically measured flower species richness versus wildflower viewing utility predicted by the revealed preference models. Fifty-four forested sites were visited over the course of an 18-wk growing season (April–August 2014) and richness, abundance, and evenness of plants in flower were recorded, as well as the number of flower colors and presence of charismatic species (see SI Text for more details). Sites were visited multiple times, either weekly or triweekly. For each site visit we calculated the predicted wildflower viewing utility, based on the discrete-choice multinomial logit models above. The maximum of the predicted wildflower viewing utility for each site was used as an indicator of CES value. We calculated overall flower species richness for each site using species accumulation curves, which allowed us to account for differences in observed species richness due to survey effort (e.g., weekly sampling versus triweekly sampling).

Finally, we identified sites with the highest CES value, or “hotspots.” A variety of methods have been used to define ecosystem service hotspots (4, 81–84). We defined hotspots to be sites above the median value for flower species richness or wildflower viewing utility. We evaluated hotspot congruence based on the two CES indicators: overall flower species richness and maximum wildflower viewing utility. We compared site rankings and hotspot classifications based on these alternate indicators using Spearman rank correlation, Cohen’s kappa coefficient, and percent agreement.

SI Text

Study Area Description.

From 1976 to 2006 the human population of the FBRB increased by 48% (85), accompanied by increased exurban, low-density housing development and increased forest land cover. Recent stakeholder interviews indicate that area residents strongly value biodiversity and are concerned for the futures of ecosystem services, particularly CES (86). Several large tracts of public (e.g., federal, state, county, and municipal) land within the study area have trails, viewpoints, and other access areas for visitors and residents (Fig. S1).

The North Carolina tourism office estimates that tourism’s impact increased from $269 million in 1991 to $901 million in 2013 in one urban center in the region, with combined visitor expenditures for 2014 over $1.33 billion for the FBRB (87, 88). Although no data specifically report dollars generated by ecotourism, overnight visitors to the North Carolina Mountain Region reported participating in rural sightseeing (26%), visiting state/national parks (23%), wildlife viewing (14%), hiking/backpacking (10%), nature/ecotouring (9%), other nature (8%), and birdwatching (4%) during 2014 (89). Guidebooks specifically focused on wildflower hikes have been published for the area (90) and 2013 was named the “Year of the Wildflower” by North Carolina’s State Parks. An informal survey of tourism websites for the region revealed that 7 out of the top 10 website results mention wildflowers at least once in their tourism and marketing materials.

Visitor Survey Analysis.

For each forest-based CES the “average importance” attributed to that service was determined by calculating an ordinal mean across all respondents (Table S4). We used factor analysis to identify interpretable dimensions of attitudes toward forest-based CES. Only factors with an eigenvalue >1 were retained. The level for interpretation of factor loadings was 0.40, based on a significance level of 0.05 and a power of 80% (91). Items that loaded on more than one factor were included in the factor for which they had the highest factor-loading score. One item (“forests are important as a place to be outdoors”) added no information and was dropped from the analysis based on communalities <0.4 (92). We performed K-means cluster analysis with the four factors to identify groups of respondents, representing segments with different psychographic profiles, based on their attitudes toward forest-based CES (80, 81). Factor and cluster analysis were conducted in R using packages psych and vegan (93, 94).

Discrete-Choice and Multinomial Logit Models.

The effect of biodiversity attributes on aesthetic preference for wildflower communities was assessed using a discrete-choice experiment and modeled using multinomial logit models. Discrete-choice experiments are a quantitative technique for eliciting preferences from individuals by asking them to state their preference over alternative scenarios, goods, or services. In this study, the discrete-choice experiment included alternative images of wildflower communities and asked individuals to indicate which image they preferred from each set of two images. A full factorial design consisting of all possible combination of wildflower community attributes was not feasible; with five attributes, three with four levels and two with two levels, a full factorial design would consist of 4³2² = 256 experimental conditions. Because a full orthogonal array is not possible, we selected an efficient design, with trade-offs between the degree of orthogonality and balance (79). We assessed the sampling design using a measure known as D-efficiency. D-efficient sampling designs maximize the amount of information about each parameter through the most efficient number of choice sets (79). The sampling design was created using NGene 1.02 (www.choice-metrics.com). The final sampling design included 48 images organized in six blocks of four choice sets and had a D-error of 0.13 (Table S5). Each choice set was a decision between two images.

The images were created using Adobe Photoshop and were representative of wildflower communities commonly encountered in the region. The base image was a photograph from a field site within the study area (48), and composition of the wildflower community was altered by adding or removing flowers. All species included in the digitally manipulated images are found within the study region. Images were shown to respondents as pairs of 8.5- × 11-inch images.

Each respondent was randomly assigned one of six choice blocks and was asked to evaluate four choice situations. According to random utility theory (49), an individual, n, chooses alternative, i, from a choice set, C. The utility derived from alternative, i, is assumed to be greater than any other choice (j) within the choice set. The formula below specifies the utility derived from any particular option (e.g., wildflower community) as

$$U_{ni}>U_{nj}\to V_{ni}+{\epsilon }_{ni}>V_{nj}+{\epsilon }_{nj}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\forall \hspace{0.17em}j\ne i;\hspace{0.17em}i,j\in C_n$$

$$V_{ni}=\beta x_{ni}+\gamma z_{ni}.$$

Here, U_ni represents the latent utility of a chosen wildflower community, i, for respondent, n. V_ni represents the explainable, or systematic, component of utility and ε_ni is a random, or unexplainable, component of utility. Moreover, V_ni can be a function of the wildflower community attributes and their levels (x_ni) as well as other covariates (z_ni) thought to influence aesthetic preference (e.g., demographic information, recreation patterns, and respondent characteristics). Finally, β and $\gamma$ are the vectors of coefficients associated with x_ni and z_ni. Multinomial logit analysis was conducted using Latent Gold software (95). These models can be thought of as simultaneously estimating binary logits for all comparisons among the alternatives. Each choice is treated as an observation, with a binary response variable and alternative specific explanatory variables.

We first analyzed the preferences of all respondents, without regard for psychographics or demographics (i.e., a single-class model). Flower species richness and abundance were modeled as continuous variables whereas number of colors, evenness, and charismatic species presence were effect-coded as categorical values. We tested whether the inclusion of interactive effects between the wildflower community attributes and respondents’ attitudes toward forest-based CES improved the model fit. We then ran multinomial logit models for groups based on people’s preference of different CES, knowledge of plants, demographics, and recreational patterns to determine whether the effect of biodiversity attributes varied across segments. Models based on segments were compared using the Wald statistic (95), which tests the restriction that parameter estimates in one segment are equal to the corresponding estimates within each of the other groups. In other words, the Wald statistic tested the equality of the regression effects across all of the groups.

Characteristics of Public Forest Visitors.

We collected usable responses from 293 public forest visitors. Respondents were 55% male, predominantly white (90%), and ranged across age groups (Table S1). The largest group of respondents were the 25- to 34-y age group (22%). Respondents were well-educated; 44% had completed at least a 4-y college degree and 29% had advanced degrees. Respondents ranged in the frequency that they visit public forests from visiting public forests only a few times a year (22%) to visiting several times a week or more (26%). The respondents participated in a range of activities and most commonly reported hiking on trails and viewing waterfalls as the main motivations for visiting the public forests (Table S2). Nearly half (46%) of respondents reported having visited a forest to view wildflowers within the past year and 8% of respondents participated in viewing wildflowers during the visit in which they were surveyed.

Exploratory factor analysis identified a four-factor structure describing people’s attitudes toward forest-based CES and accounted for 60% of the variance in the dataset (Table 1). Based on factor composition, we labeled the factors “quiet relaxation,” “experiences,” “active escape,” and “collecting.” Respondents’ gender was not related to their scores along either the quiet relaxation or active escape axis. However, females scored higher on the experiences axis (t = −2.93, P = 0.004) and males scored higher along collecting (t = 2.89, P = 0.004). Respondents who visited public forests more frequently tended to have higher ratings for quiet relaxation and collecting and lower ratings along the experiences factor than those who visited public forests less frequently. Older people tended to score lower along the axis active escape (Kruskal–Wallis H = 26.07, P < 0.001) and higher on experiences (Kruskal–Wallis H = 20.53, P < 0.001). Respondents’ education level was unrelated to factor scores. Respondents’ preferences for images were strongly influenced by flower abundance, evenness, and number of colors. Preference did not vary across segments of public forest visitors (Table S3).

Wildflower Community Field Data Collection.

Wildflower community data were collected at 54 forested sites in the FBRB during the summer of 2014. We stratified the study area by elevation, building density, and land use. Sites were in forested areas and within 150 m of trails or roads to characterize floral resources likely to be visible to people. Sites were located on both public and private property, and sampling design is described in detail in another study (48). In brief, we established a 50- × 2-m belt transect at each site and surveyed for wildflower blooms at least three times between April 1 and August 8, 2014. During each visit, we tallied the number of flowering individuals, identified each flowering individual to species, and estimated percent cover of flowers along the transect. We classified each species as charismatic or not (48) and recorded the color. For each site visit we tallied the number of species in flower (flower species richness), the number of flowers (flower abundance), and the number of flower colors. Charismatic species presence was recorded as a binary variable (1 = present, 0 = not present). We calculated evenness of the flower community using Simpson’s evenness (E) (96). Overall flower species richness (for the whole sampling period April–August 2014) was calculated using species accumulation curves and the vegan package in R (93), which allowed us to account for differences in observed species richness due to survey effort (e.g., weekly sampling versus triweekly sampling).