Introduction

Because landscape means different things to people depending on their context and background (Antrop 2005; Fairclough et al. 2018; Griffiths 2018), perspective and preference of landscape varies greatly and there is not a universally accepted theory for landscape aesthetics (Kaymaz 2012). Literature searches show that there is limited information on methods of assessing users’ perception and preference in the USA (Palmer and Hoffman 2001; Smardon 2016; Zube, et al. 1982). Daniel and Boster (1976; as cited in Peng and Han 2018) introduced a scenic beauty estimation (SBE) method to quantitatively measure aesthetic preferences for alternative wildlife management systems, which was probably the earliest objective method that used photographs.

Using photographs as substitutes for on-site viewing of landscapes has been a commonly accepted approach in assessing public preference for a number of years. In another previous study by the authors, a contemporary method of using technology to assess users’ preferences for landscape and landscape attributes in the pAGP was implemented. In that study, a combined method of geographic information system (GIS) and machine learning (ML) approach was employed by using photographs that were uploaded in the Flickr crowdsource, an indirect approach. Previous studies have used density of photographs as an indicator to assess users’ interest and cultural ecosystem services (Callau et al. 2019; Tieskens et al. 2018; Wood et al. 2013). However, the researchers also pointed out that the density of photographs from a specific location alone may not necessarily relate to the public interest in the location (Wood et al. 2013) or be sufficient for measurement of cultural ecosystem services (Richards and Tunçer 2018). They also discussed biases inherent with such crowdsource data as presented in the previous paper (Malik et al. 2016; van Zanten et al. 2016; Wang et al. 2018; Wood et al. 2013). Previous studies sought a combination of different methods to assess public preference.

In addition to the previous approach of implementing Flickr photographs in combination with GIS and ML, we executed two more approaches to assess users’ preferences. Previous studies have used various theories and models. These include phenomenological (e.g., Lynch), ecological (e.g., McHarg & Mumford), prospect-refuse (e.g., Appleton), affordance (e.g., Gibson), psychology (e.g., Kaplan & Kaplan), and psychophysical models (e.g., Daniel & Boster). This research was conducted based on psychophysical aspects of landscape and assumed that the preference was about viewers’ perception and cognition (Peng and Han 2018). Photograph rating and questionnaire survey methods were employed. The aim was to investigate how participants interpret landscapes and landscape attributes visibly (pictorial and sense of sight based) and invisibly (cognitive and thought process based). This is a semi-direct approach as we provided photographs of landscape scenes and questionnaires using anonymous survey links. The potential participants were then asked to rate those photographs and answer the questions. This study seeks answers for the following research questions: Can we relate specific landscape settings and features with users’ preferences? What landscape features are preferred by the users? Is the demography of users associated with preference for landscape and landscape features?

Background

Objective and Subjective Philosophies of Landscape

Human perspective of land resources has increasingly been considered not just as a plot of land that provides food and shelter, but also as a visual resource for its aesthetic, scenic, and recreational values since the middle of the nineteenth century. Since then, keeping the best and unique landforms and landscapes aside for protection and recreation began by designating such areas as national parks, reserves, areas of outstanding beauty and wilderness areas. One approach is through the designation of a Landscape Park as implemented in Poland (Dmytrowski and Kicińska 2022). Two major philosophies, viz., objective and subjective have been widely discussed, particularly in the field of visual quality assessment of the landscapes (Brooks and Lavigne 1985). In the objective philosophy, the importance is given to the landscape’s biophysical aspects and beauty in terms of forms, shape, texture, color, vividness, uniqueness, pattern, etc. (Daniel 2001). “Biophysical landscape interactions are those biotic and abiotic processes in a landscape that have an influence on the developments within and evolution of a landscape” (van der Ploeg et al. 2018, p. 1167). The subjective philosophy is dominated by human cognitive constructs such as needs, wants, feelings, observations, behaviors, attitudes, perspectives, etc. (Brooks and Lavigne 1985; Dearden 1981; Zube et al. 1982) and landscape’s use values (Tveit et al. 2006) for agriculture, industry, transportation, refuge, etc. For instance, the European settlers viewed the minerals of American lands as objective values (e.g., gold of California), whereas the desire of the European settlers for gold (e.g., seeking for economic benefits) is a subjective aspect of the landscape (Langschwager 2018). Both of these philosophies of beauty are equally important. However, the objective philosophy has dominated in landscape evaluations and studies for a number of years compared to the subjective or user-centered landscape evaluations (Kvan 2013). According to Brooks and Lavigne (1985), “Neither the extreme subjective nor the objective cognitive philosophies of beauty is completely wrong.” As Swanwick (2002, p.10) said, “There has been long-standing debate about the role of objectivity and subjectivity in dealing with landscape. The search for supposedly objective approaches has reflected a desire, in some quarters, to remove the element of personal judgement from the process. In Landscape Character Assessment it is accepted that there is a role for subjective inputs, but these must be made in a systematic and transparent way.”

Landscape preference depends on what the landscape offers (Gibson 1979; Herzog 1987). The “affordances are features and configurations” of landscapes. The greater the positive objectivity or higher the beauty of the landscape, the greater the subjectivity of the landscape or preference is (Peng and Han 2018). This is valuable information for planning and development of the landscape.

Experts vs. Citizens in General in Landscape Studies

The subjective aspect of the landscape “can be measured objectively through surveying people’s likes and dislikes- their aesthetic preference” (Lothian 2017). Landscape assessment, planning and evaluation depends largely upon human judgment. However, the human judgement is often discussed over expert vs. citizen perspectives in general (Daniel 2001). Usually, the expert’s judgement is based on training and experience. This type of assessment is often based on certain established standards and frameworks (Kaplan and Kaplan 1989) where experts/planners usually compare a particular example that is to be assessed for its value (Daniel 2001) or decide using their own intuition (Daniel and Boster 1976). It is mostly a top-down approach and lacks psychophysical aspects (Liu et al. 2018). Expert domination is generally found in objective resource assessment and is often focused on scientific, ecological, and physical aspects of the landscape (Liu et al. 2018). Citizen opinion is a perception-based approach (Daniel 2001) where neither training nor experience and standards or frameworks are required. It simply deals with human sentiments, stimulation, and interaction with visible landscape characteristics (perception, aesthetic, artistic, etc.) (Antrop 2000; Daniel 2001; Kaplan 1989; Kvan 2013; Zube et al. 1982). So, it is a bottom-up approach. Herzog et al. (1982) pointed out a number of ways in which “citizens’ perceptions and preferences differ from those of planners/[experts].” Daniel (2001) pointed out inconsistences in opinions even between the experts for the same landscape and found that perceptions greatly vary from person to person (U.S. Department of the Interior n.d.). Public perspectives and preferences play a vital role in making right decisions, policies, and plans (Shafer and Brush 1977). A shortcoming of research due to the exclusion of the public was experienced by Lothian in his study in Australia (Lothian 2012; Kvan 2013). Daniel and Vining (1983; as cited by Daniel 2001) criticized expert-based landscape quality assessment “for having inadequate levels of precision, reliability and validity.” Jensen (2007) also pointed that “expert-led approaches to landscape assessment are no longer viable as local people might view their landscape very differently from landscape specialist.” According to Wartmann et al. (2021), even though there has been significant work done for measuring physical aspects of landscape, there is a lack of information about measuring public perspective and preference.

Photography in Landscape Studies

“People are particularly aware of information that is visual, that concerns what they see” (Kaplan and Kaplan 1989, p. 4). Since the introduction of photography during the nineteenth century, photographs became an important tool in the studies of environmental conditions, including landscapes (Daniel and Meitner 2001; Bergen et al. 1995; Steiner 2009). Photographs are two-dimensional visual representations of landscapes which have been increasingly used as surrogates for an on-site assessment of the landscape (Lothian 2017). “Few studies provide guidance for photographs” (e.g., Kaplan and Kaplan 1989; Lothian 2017). Both color and black-and-white photographs have been used in the studies of landscapes (Kaplan and Kaplan 1989). Kaplan and Kaplan (1989) have presented both advantages and disadvantages of color and black-and-white photographs in their book The Experience of nature: Psychological Perspective (p. 16-17). According to them, the participants’ ability to rate photographs was not thwarted by the color or black-and-white photographs. Kellomäki and Savolainen (1984) also noted that “black-and-white photographs may[be] reliable tool[s] for estimating scenic value of forest landscapes.” However, Lothian (2017) emphasized color photographs in horizontal view. According to Lothian, “black-and-white photographs emphasizes formalist qualities but lose the life-giving quality that color conveys.” This statement is supported by Shuttleworth’s (1980; as cited by Lothian 2017) finding that “black and white photographs gave more extreme rating and had lower correlations with field assessments than color.”

Different Approaches of Using Photographs in Landscape Studies

Subsequently, different approaches of obtaining and using landscape photographs have evolved in landscape studies. In some studies researchers and professionals take photographs of their interest area and analyze them (Svobodova et al. 2015). In other instances, individuals are provided landscape photographs to rate based on their preferences (Kalivoda et al. 2014; Lothian 2017). In other cases, the individuals are given cameras and asked to take landscape photographs of their interest to get their perspective on landscapes and landscape attributes (Goldberg 2019; Steiner 2009). Dorwart et al. (2010) provided disposable 24-exposure digital cameras to intended visitors and asked them to take photographs of features along the Appalachian Trail, North Carolina. Landscape studies also used other visual imaging techniques such as photomontage, simulation (Mahdjoubi and Wiltshire 2001), graphically enhanced images (Lee et al. 2019; Liu et al. 2018; Lothian 2017, 2012) and videos to present animated landscape scenes such as water motion and public participation geographic information systems (PPGIS) (Brown and Weber 2011). Today, photographs uploaded in the social domains have been widely considered as suitable substitutes for measuring public interest and preference in landscape studies (Bubalo et al. 2019; Callau et al. 2019; Hansen 2016; Richards and Tunçer 2018; Ros-Candeira et al. 2020; Stepchenkova and Zhan 2013; Tenerelli et al. 2016; van Zanten et al. 2016; Wood et al. 2013). These techniques may all be used to study and manage lands.

Other studies also used photographs in combination with other techniques to assess the visual quality as well as preference for landscapes and landscape attributes. Kellomäki and Savolainen (1984) used photographs in combination with questionnaire surveys to assess the scenic value of forest landscapes. They used on-site forest stands and photographs of those stands to see the correlation between the scenic values. They found that the mature stands of moderate density had the greatest scenic value. Hull IV and Stewart (1992) carried out a comparative study of scenic beauty judgements based upon photographs of landscapes and on-site experiences of landscapes in national forests and wilderness areas in Colorado by using the same persons, but in different time periods. They observed insignificant correlations between photo-based and on-site scenic beauty assessments. They concluded that there was an uncertainty in the validity of photo-based scenic beauty assessments. Daniel and Meitner (2001) performed a comparative study of photographs of the forest landscape and computer-rendered photographs of the same forest landscape to assess the representational validity by using different groups of viewers. They noticed a low correlation between the ratings of the two visualizations and uncertainty about the representational validity of the computer-rendered visualizations. A similar approach was used by Bergen et al. (1995) by creating computer images with vantage points. They found a lack of correlation between the two visual representations due to the omission of important landscape elements in the vantage point images. They concluded that the photographic images were promising over the computer-rendered landscape images and noted that this might be useful for preliminary assessment through modification, especially in landscape designing.

However, Scott and Canter (1997) questioned the method of using photographs for individuals to analyze, pointing out that people have limited choices to focus on because the photograph is imposed by the researcher. According to the researchers, photographs are inadequate for eliciting potentially important humanistic factors (actions, thoughts, feelings, etc.) which constrain or do not accurately measure the viewers’ preferences for the landscape and landscape features while using photographs (Shuttleworth, 1980a as cited by Kaplan and Kaplan 1989). Daniel and Meitner (2001) also reported that still photographs may not adequately represent potentially important actions. Therefore, they may not be a valid medium for perception studies.

The researches cited above pointed out both pros and cons concerning the use of photographs in landscape studies. This study attempted to investigate public preference through visual as well as cognitive stimulation exercises. This was not a new approach; however, it was used to support and supplement our machine learning approach. This would contribute to an investigation of the findings from different approaches.

Methodology

Study Region

The study region is the proposed Appalachian Geopark (pAGP) and is located in the Appalachian region covering Fayette, Greenbrier and Raleigh counties in WV. The pAGP covers an area of 5963 sq. km. which is approximately 10% of the total area of the state (62,258 sq. km.). Of the total pAGP area, approximately 5923 sq. km. is land and 40 sq. km is water (rivers and lakes) (Fig. 1).

Fig. 1
figure 1

(a) The whole state of West Virginia lies in the Appalachian Region. The study region is the proposed Appalachian Geopark encompassing three counties: Fayette, Greenbrier, and Raleigh. (b) The study region consists of mountains and valley and ridges. Greenbrier County has more flat areas compared to other two counties

Full size image

The proposed Appalachian Geopark’s landscape is characterized by two major physiographic provinces,Footnote 1 namely, Appalachian Plateau and Valley and Ridge. The Appalachian Plateau is composed of a higher rugged section and a relatively flat section which are named as the Appalachian High Plateau Province (Allegheny Mountain) and the Appalachian Low Plateau Province. Of the three counties of the pAGP, whole counties of Fayette and Raleigh lie in the Appalachian Low Plateau Physiographic Province. This province is characterized by flat-lying to gently folded rocks of Pennsylvanian periodFootnote 2 and a dendritic stream pattern. The relief of the region is a few hundred feet with the hills between major streams possessing rounded summits and relatively shallow slopes. Greenbrier County lies both in the Appalachian Low Plateau and the Valley and Ridge provinces. The Valley and Ridge Province consists of folded and faulted rocks ranging in age from the late PrecambrianFootnote 3 to the early Mississippian period.Footnote 4 “The steeply dipping folded beds, the harder layers-particularly the sandstone tend to form ridges whereas the softer shale beds with limestone tend to form valleys” (Cardwell 1975). This province is characterized by long folded mountains and parallel valleys made of asymmetric to overturned anticlines exposing rocks from the Ordovician ageFootnote 5 (the oldest rocks of this region) and with a trellis stream pattern. There is a structural break differentiating the Appalachian Plateau and Valley and Ridge provinces which is known as the Allegheny Structural Front and is represented by a prominent escarpment/cliff. These fundamental structures along with the action of wind, water, heat, glaciation, vegetation, wildlife, and anthropogenic activities determine landscape characters in the pAGP (Figs. 2 and 3).

Fig. 2
figure 2

Appalachian low plateau and valley and ridge provinces (source: 2016 Geologic History of WV John J. Renton & Thomas Repine. The red circles represent the physiographic provinces of the study region)

Full size image
Fig. 3
figure 3

Distinctive geology of landscape due to sandstone and limestone-rich areas

Full size image

“The Appalachian Low Plateau is composed of sedimentary rocks including sandstone, conglomerates and shales deposited during the late Paleozoic era”Footnote 6 (National Park Service 2018). A large part of the study region is represented by sandstone (69%) followed by shale (19.5%), limestone (4%), and others such as alluvium, and a combination of sandstone and shale as well as sandstone and limestone. The majority of Raleigh and Fayette counties is made up of sandstone with some shale and siltstone making the area unique with high and magnificent rock cliffs, canyon walls, gorges, and valleys. In addition, the plateau region consists of coal beds due to the deposition of plants and animals in between the hard sedimentary rocks, whereas the distribution of shale and limestone became the major bedrock in the Greenbrier County, featuring wide valleys, gently rolling hills, and flatter land. The imperviousness of shale and its deposition made the Greenbrier County suitable for karst landscapes in the form of caves, caverns, sinkholes, etc. According to Dasher (2000), in WV, if there is coal, there is generally no significant karst, and if there is karst, there is no coal (Fig. 5).

Coal towns in sandstone and karsts in limestone-rich areas are represented by green triangles. Caves and caverns are represented by red polygons. The very different landscapes in eastern and western ends are evident.

The remarkable landforms and landscapes such as deep gorges, ancient rivers, mountains, and karsts along with mixed-mesophytic forests, land-use history, and archeological significances are some influential characteristics that have provided recreational, educational, scientific, aesthetic, and economic values in the pAGP. These attributes are the potential nature merchandizing assets of this region that support a Geopark concept. Even though the USA was a founding member of the United Nations Educational, Scientific and Cultural Organization (UNESCO), it withdrew the membership in 2018. At the time of this writing, there is no avenue forward for gaining Geopark designation. However, there are three aspiring Geoparks in the U.S. There is currently no regional geopark network in North America. The New River Gorge is one of the most prominent landscape features in the pAGP and was designated as the New River Gorge National Park and Preserve (NRGNP&P) in 2020. The Gauley River National Recreation Area, Bluestone National Scenic River, and the underground wonders of caves and caverns are other attractions in the pAGP (National Park Service 2012). According to the NPS, about 1.4 million people visited the New River Gorge and two other scenic areas nearby in 2019. There was an increase of 20% in visitors after receiving a national park designation. We can expect more visitors if the pAGP receives an official Geopark designation.

In the pAGP, approximately one-third of land resources is owned by private sectors (corporates, individuals). More land in the study region is owned by private absentee owners. Kayden (2000) reported that the “states have constitutional authority under so called ‘political powers’ to plan and regulate the use and development of land.” The public land/water resources in the pAGP have been managed by agencies under federal, state, and local principles of conservation, protection, multiple use, and sustainable development. The National Park Service (NPS) manages the NRGNP&P. The National Forest Service (USFS) manages federally owned forests. The West Virginia Department of Natural Resources (WVDNR) manages wildlife management areas and the West Virginia Division of Forestry (WVDOF) manages state owned forests.

Research Instrumentation

Initially, the research instruments were designed to utilize in the study region for face-to-face interviews and group discussion as direct approaches. Because of the unforeseen situation of COVID-19, instruments were developed in the form of anonymous links. In this paper, two methods of assessing public preferences of landscapes and landscape attributes were implemented. They were as follows:

Photograph Rating

This is a photo-based, visual stimulation exercise. We took approximately 1000 photographs during several field trips. These photographs were taken using smart phones and digital cameras. Taking photographs during field visits helped us become more familiar with the study region. They were taken from the grounds, hill tops and overlook outlets, etc. Seventy selected photographs that represented various types of landscapes including natural, natural looking, and historical and cultural landscapes of the study region were used to develop a Photograph Rating survey instrument using Qualtrics software. Some photographs consisted of a single attribute that helped make up the landscape scene. Color photographs in landscape format (i.e., horizontal) were used assuming that the color photographs and horizontal view conveyed a natural look. We selected photographs that exhibited disturbance free objects (e.g., pole, animals, people, etc.) as best as possible. For instance, photographs were not modified digitally by removing grazing animals in the pasture landscape scene.

A preference rating scale was designed with a sliding scale bar consisting of 1–10, where 1 is the lowest and 10 is the highest preference score. The respondents were asked to slide the sliding bar to a point of their interest to show their preference for landscape and landscape attributes (i.e., the degree of preference value need not necessarily be an integer). We chose to use the sliding bar to get more granularity in the preference rating. However, this approach is not immune to the shortcomings of the traditional 5–7 Likert scale. It too may fail to measure the true attitude of respondents. People’s answers can be influenced by preceding questions, and individuals can be reluctant to choose the extreme options even when they are the most accurate.

Questionnaire Survey

This survey is a cognitive-based elicitation. An anonymous link of the questionnaire survey was constructed by using ArcGIS Survey123 which is a map and form-based product useful for creating, sharing, and analyzing surveys. The majority of questions were structured to keep the method simple, clear, and relevant with approximately 20% open-ended questions. Most questions were developed using a Likert scale from 1 to 5.

Data Collection

The target population for the study was the public in general including both residents and visitors. A list of emails of different government agencies, public organizations, and individuals was prepared. The accessible population were all the populations that we contacted via emails/social media. We planned to obtain data using simple random sampling. However, this was impossible due to COVID-19. Thus, the sample was obtained via a self-selective statistical process, typically termed to be “convenience” sampling (Babbie, 2010 in Kalivoda et al. 2014). Anonymous links of Photograph Rating Method and Questionnaire Surveys were distributed through emails and social media. Respondents were invited to participate in the surveys and to share the survey instruments with those who were interested in participating in the survey. The survey links were also distributed among friends, colleagues, and associates. Because we distributed anonymous survey links and encouraged email recipients to share the links for participation and most of the respondents were not associated with the researchers, our only option was to assume that the sample was representative of the population of visitors and residents.

Two online surveys were employed to collect input from respondents. Both the Survey123 and Qualtrics surveys were publicly available on the internet. Both products collect responses on their webservers and made the survey results available to us as developers. We imported the results into spreadsheets and a database for aggregate analysis and statistical processing. The survey instruments provided instructions on how to take the survey.

Data Analysis

A data matrix was prepared in Excel. The data was then analyzed using the statistical software John’s Macintosh Project (JMP). JMP is a powerful statistical analysis software and a subsidiary of the Statistical Analysis System (SAS)Footnote 7 Institute. The scenes were categorized into different visual types, and means were calculated. The mean scores represent preferences for the scenes. Ratings of the landscape scenes (photographs) represent the preference assigned to them by the viewers (Kaplan and Kaplan 1989; Lothian 2017). Landscape scenes contain physical attributes that remain constant regardless of who is viewing the scenes. The ratings given to the landscape scenes by the viewers refer to the observed variables which may largely be influenced by characteristics associated with the viewers (Kaplan and Kaplan 1989; Lothian 2017) such as gender, age, emotions, familiarity, and benefits. The data was analyzed by comparing the means that each visual type scored. We also assumed that the higher the mean for a scene/visual type or attribute, the higher the preference of the viewers to that particular landscape scene/visual type or landscape attribute.

Categorization of Photographs

The photographs were chosen to cover the representative landscape types which included rivers, creeks, forests, trails, buildings (commercial, historical, residential), townscapes, parks, highways, country sides, etc. The selected photographs were categorized into different landscape character types for further analysis.

First, the photographs were categorized into three major visual types such as:

  1. 1.

    Natural visual type: this type consists of only natural landscape visual types that were without any human induced or built features in them.

  2. 2.

    Natural appearance visual type: in this type, those photographs that were natural looking, but are not natural in-deed, were considered. Man-made lake landscapes, for example, represent this type of landscape.

  3. 3.

    Historical/cultural/built visual type: as the name suggests, this type of landscape consisted of historical and cultural (human induced features) attributes in the photographs.

Second, the photographs were categorized based on landscape features such as water, vegetation, and built features.

  1. 1.

    Water feature

  2. 2.

    Vegetation feature

  3. 3.

    Built features

Third, the photographs were categorized based on unique natural and built features because they act as the center of attraction, and viewers compose landscape based on these features (Karjalainen and Tyrva 2002).

  1. 1.

    Unique natural feature

  2. 2.

    Unique built feature

We set up hypotheses as follows:

  • Null hypothesis (H0): there is no difference in the preferences of each of the different categories of visual types of landscapes (i.e., preference to all landscape categories of visual types is equal).

  • Alternative hypothesis (H1): there is at least some variation between the preferences of each of the categories of different visual types (i.e., at least two of the preferences are not equal).

We then tested three sets of hypotheses, each set corresponding to one of the visual category types to investigate if the landscapes and landscape attributes matter to respondents by comparing the mean preference that each of the categorized landscape types obtained. A generalized analysis of variance (ANOVA) was performed to assess the difference between the preference means of these different landscape character types by using statistical software JMP. The general assumption was that the observations are independent of one another and the sample data was representative of the population of residents and visitors.

Figure 4 shows a stepwise process of photograph rating method.

Fig. 4
figure 4

Photograph rating method (adopted and modified from Lothian, 2017)

Full size image

Results

General Findings

Unsurprisingly, as previous studies found, we received the highest mean preference for natural landscapes with forest and water features indicating that these features were mostly preferred by the respondents. The bar chart (Fig. 5) showed an order of mean preferences for landscape and landscape attributes. The natural features including New River and forest landscapes received the highest mean scores between 7 and 8. The lowest mean score was obtained for a commercial landscape scoring between 2 and 3 (Fig. 6).

Fig. 5
figure 5

Order of respondents’ mean preferences for landscapes and landscape features (the numbers next to the pictures are the photograph identifier number)

Full size image
Fig. 6
figure 6

Most preferred (left) (number 2) and least preferred (right) (number 65) landscape scenes/photographs (photography by March G. Nakarmi, 2020)

Full size image

Preference Using Landscape Visual Types

Upon testing the hypothesis, the JMP output showed that the p-value was less than 0.05,Footnote 8 i.e., p ≤ 0.0001 (Table 1). This p-value indicated that the null hypothesis must be rejected. This means the alternative hypothesis was more likely to be correct rather than the null hypothesis. This revealed that there was some variance in the preferences of the three visual landscape types.

Table 1 Preferences using landscape visual types
Full size table

For post hoc analysis, we performed multiple comparisons using Student’s t-test to determine how participants’ preferences differ for different visual types. The JMP output showed all p-values being less than 0.05 (Table 2). This test showed that preferences for the three visual types were statistically significantly different from each other. Figure 7 presented discernible display of the mean preferences of the respondents.

Table 2 Multiple comparison of preferences using landscape visual types
Full size table
Fig. 7
figure 7

Mean preference scores for natural, natural looking, and historical and cultural visual types

Full size image

Preference Using Landscape Feature Types

We also looked at major landscape features such as water, vegetation, and built features to investigate how participants’ preferences differ from each other. The resulting p-value of 0.0001, which was less than significance level (α = 0.05), indicated that the null hypothesis must be rejected and the alternative hypothesis was more likely to be correct. This illustrated that there was a statistically significant difference in the mean preferences of the three types of landscape features (Table 3).

Table 3 Preferences using landscape features
Full size table

We then conducted post hoc multiple comparisons using Student’s t-test. This resulted in all p-values of less than 0.05 (Table 4). These provided sufficient evidence to conclude that water, vegetation, and built features were preferred differently (Figs. 8 and 9).

Table 4 Multiple comparison of preferences using landscape features
Full size table
Fig. 8
figure 8

Landscape visual types: a natural. b Natural looking. c Historical/cultural (photography by G. Nakarmi, March 2020)

Full size image
Fig. 9
figure 9

Landscape features: a water. b Vegetation. c Built (photography by G. Nakarmi, March 2020)

Full size image

Preference Using Unique Landscape Feature Type

Studies (e.g., Karjalainen and Tyrva 2002) noted that when viewing a landscape, people focus their attention on unique features forming the landscape. We tested how viewers’ preferences for unique natural feature and unique built feature differ in the study region. The ANOVA resulted in a p-value of 0.0554 (Table 5) which was slightly greater than α = 0.05. This indicated that the two means were not statistically significantly different.

Table 5 Preferences using unique features
Full size table

The pooled t-test table exhibited a result similar to the ANOVA. Notice that the p-values were the same (p-value = 0.0554) (Table 6). This was to be expected when our effect variable in the ANOVA had only two possible levels, i.e., unique natural feature and unique built feature.

Table 6 Pooled t-test using unique natural feature–unique built feature assuming equal variances
Full size table

This result was further supported by the 95% confidence interval on the difference of the two mean preferences. This confidence interval contains the value 0. This implies that 0 is a plausible or believable value for the difference of these two means, further implying that the two mean preferences were not statistically significantly different. Unique natural feature had a mean of 7.25 compared to the unique built feature mean of 6.01. This indicated that the respondents were attracted by the natural unique feature approximately equally with the built unique feature. New River Gorge Bridge, Grist Mill of Babcock State Park, unique rock cliffs, and historic buildings in the town center were some examples that supported Karjalainen and Tyrva’s (2002) statement in the study region (Fig. 10). These unique features provided conspicuous vistas with which people form landscapes.

Fig. 10
figure 10

Unique natural and built features that make landscape: a rock cliff, b historical mill, and c bridge (photography by G. Nakarmi, March 2020)

Full size image

Gender, age, and residency were chosen as anonymous demographic classification. Where the respondents reside and who they are influence their preferences. The sections below expand our findings on each of these classifications (Table 7).

Table 7 Attributes of the respondents
Full size table

Preferences by Demographic Measures

We tested for consistency of respondents’ preferences with respect to some demographic measures by calculating standard deviations. Standard deviation is the spread of the data from the mean value that helps measure consistency of opinions among respondents. A low standard deviation suggests preferences are fairly similar whereas a high standard deviation suggests diverse preferences (Lothian 2017). Table 8 shows that rivers and forested mountains have smaller standard deviations, i.e., more clustered around the mean preferences which indicates for consistency.

Table 8 Preferences for landscape character types using the entire sample
Full size table

Testing Preference for All Landscape Character Types Using Gender

We performed several two-tailed tests using a significance level of 5% for all landscape character types. We investigated whether the preferences differed according to gender. The results revealed all p-values greater than 0.05, so we cannot say that there is a statistically significant difference in mean preferences for males and females (Table 9).

Table 9 Differences in mean preferences according to gender (male–female)
Full size table

Testing Preference for All Landscapes Character Types Using Residency Status

Upon performing a two-tailed test for preferences for visitors and residents under the significance of 5%, the p-values that were greater than 0.05 show that we cannot say there was a statistically significant difference in mean preferences for visitors and residents. For example, the mean preferences for rivers were not statistically significantly different between visitors and residents (Table 10). The p-values less than 0.05 showed that we can say there was a statistically significant difference in mean preferences for visitors and residents. For example, the mean preferences for forested mountains were statistically significantly different between visitors and residents.

Table 10 Differences in mean preferences according to residence status (visitor–resident)
Full size table

Interaction of Gender and Age Group

We performed a two-way ANOVA and constructed an interaction plot (least square means plot) (Fig. 11) to investigate the relationship between age group and gender for both natural/scenic and historical/cultural landscape types. Figure 11 a indicated that there was an interaction between gender and age with respect to natural/scenic mean preferences. The preference for natural/scenic landscape for females’ age 18 to 30 years was less than that of males. However, the preference for 31–50-year-olds was found to be the opposite. The preference remained the same for females when age increased from 51 to 70, but males’ preference increased along with the increase of age from 51 to 70 years. Likewise, preference decreased drastically as females became older, 70 + , whereas the males’ preferences did not change with the increased years. Figure 11 b indicated there was not an interaction between age and gender with respect to historical/cultural mean preferences. Males’ preference for historical/cultural landscape was higher than that of females regardless of age.

Fig. 11
figure 11

Preference ratings by gender and age groups. a Gender preference by age group for natural/scenic landscapes. b Gender preference by age group for historical/cultural landscapes

Full size image

Interaction of Gender and Residency Status

We also tested preferences with respect to gender and residency status. In the interaction plot, Fig. 12a, there was an interaction between residents and visitors with respect to natural/scenic landscapes. The male residents’ preference for natural/scenic was relatively less than that of female residents. The male visitors’ preference was higher than the female visitors. For historical/cultural landscapes, there was not an interaction between gender and residency status. Males’ preference for historical/cultural landscapes was higher than that of females regardless of age (Fig. 12b).

Fig. 12
figure 12

Preference ratings by gender and residency status. a Gender preference by residency status for natural/scenic landscapes. b Gender preference by residency status for historical/cultural landscapes

Full size image

The participants reported the New River Gorge and Bridge as the highest preferred places of the pAGP, followed by Greenbrier River Trail, Babcock State Park, Lewisburg, and others (Fig. 13).

Fig. 13
figure 13

Most preferred places reported by the respondents (vertical axis represents the frequency of responses)

Full size image

The respondents’ preferences for landscapes and landscape features for recreational activities is the highest for trails, rivers, woodland forests, and campsites. They preferred those landscapes and landscape features mostly for hiking, biking, canoeing, kayaking, swimming, fishing, hunting, vacationing, climbing, and others (Fig. 14). Surprisingly, the preference was less for karst landscape such as caves and caverns.

Fig. 14
figure 14

Recreation preferences (bars represent number of responses for each score)

Full size image

Discussion

Using photographs as surrogates of landscapes and landscape attributes and questionnaire surveys are common processes in landscape studies and plannings, and have been used in evaluations for years. Photographs are two-dimensional visual representations of landscapes while the questionnaire survey is designed to elicit information from participants where cognitive processes are often used. Also, the photography depends on what the landscape offers, what catches viewers’ eyes, how pleasant and unique the location is. and so on. We used color photographs. However, Kaplan and Kaplan (1989) and others reported that small black-and-white printed photographs were more reliable over the badly colored photographs and concluded that preference rating by viewers was not hampered. In addition, it was cost effective.

Content in the photographs has a strong impact on preference (Herzog et al. 1982), and this statement aligned with Gibson’s theory of affordance (as cited by Herzog 1987). As was true in the previous studies, naturalness was the highest preferred landscape in the pAGP. The landscape visual types with vegetation and water features were the highest preferred landscapes. The highest preference was obtained for densely forested mountains, hills, valleys, and woodlands. However, Daniel (2001) noted that a relatively flat landscape with dense forest vegetation does not offer vista perspectives and was less preferred as a result. According to Herzog et al. (1982), vegetation is a potential content category that increases preferences, and thus, the presence of vegetation can improve urban landscape. We found rivers, waterfalls, and creeks as the most important water features probably due to the significance of the New, Gauley, Kanawha, and Greenbrier Rivers and the associated waterfalls and creeks for scenic and recreational values. Peng and Han (2018) reported a similar result of preference for vegetation and water. Arriaza et al. (2005) used photos and regression analysis to assess public preferences on agricultural landscapes. According to them, degree of “wilderness and positively evaluated man-made features play a key role in determining the visual quality of the rural scenes.” The respondents also opined that the geology, distinctive rock and cliff, rock types, forests, and wildlife made this region different from other regions. They also reported that increased tourism would be the best way to improve the economic activities of the region. Therefore, implementing a geopark and geotourism may have a potential role in developing this region.

Surprisingly, although the study region abounds with karst landscapes such as caves and caverns, the respondents reported their preference relatively low. This may be because the photograph rating instrument did not consist of many photos of caves and caverns since they are subordinate landscapes or people have very limited or no information about this attribute in the study region. The reason for this limited access may be due to a lack of publicity because of White Nose syndrome or some safety issues. Only Organ Cave and Lost World Caverns were famous among the visitors. There may be a need for prior contact with cave owners to visit the caves and caverns. We visited Organ Cave and Lost World Cavern with a prior contact with the cave owners. More consultation and strategies should be developed to promote the karst landscapes in the study region because this is a strength of this region that possesses potential for geopark and geotourism development.

People tend to visit popular areas (Antoniou et al. 2010) that are easily accessible (Gill et al. 2015; Muñoz et al. 2020) and possess elements of interest (Gill et al. 2015; Karjalainen & Tyrva 2002). This information is true in the case of the study region. Despite being potentially attractive, only a few locations are popular among the public (Fig. 13).

We found different preferences given the age and residency status for natural/scenic and historical/cultural landscapes. For instance, females of 70 + years were less attracted to natural/scenic landscapes. This may be because at that age, they tend to go out in nature less than males do.

Since landscape is an end product of human-interaction with land resource, landscape management is basically a management of people, their perspective and behavior which is not an easy task. Because landscape means different things to people and can be described in varying perspectives, there is not a concrete or universally accepted single approach in terms of its studies, planning, evaluation, and management. While talking about aesthetic preference, objectivity, and subjectivity, there is a concern that “if everyone is different, then preference is a matter of taste and irrelevant to policy” (Michigan Law Review 1973).

The findings contained herein would be beneficial to the planning authorities in the counties, cities, and towns to raise awareness about potential utility of the landscapes and geological heritages for tourism through new concepts of geopark and geotourism. Although visiting geological heritage sites in general has been considered a part of tourism development for a long time, it has emerged in its own importance since the development of the Geopark concept and the network between the participating countries called the Global Geopark Network (GGN). The GGN member countries get benefits through knowledge sharing in terms of raising awareness about earth sciences, geological values, consequences, and economic development through tourism development (Ólafsdóttir and Tverijonaite 2018).

There are certain limitations to this study, especially for the photograph ratings part of the survey. Some respondents reported that they could download the survey only partially. Therefore, it is of utmost importance to ensure adequate internet facilities, especially for a photograph rating method that is accessed through an anonymous link.

Conclusions

Questionnaire and photograph ratings are both simple and straightforward methods for assessing public preferences of landscape and landscape features. A combined approach of photographs and questionnaires provided greater potential for understanding respondents’ preferences for landscape character types. We looked at the character types from different perspectives such as visual, feature, and individual and unique elements. A combined approach enabled us to look at preference in relation to demographic aspects. This approach can be applied in other areas at any scale and with any kinds of variables.

Overall, the study revealed that the participants’ preference increases with the increase of degree of naturalness and unique built features in the landscape scenes. Naturalness has a strong impact on preference. Natural and natural looking landscapes were preferred over cultural landscapes. Also, forest and water features were highly preferred landscape attributes. Such studies could provide a basis for comparing preferences for landscape character types that enrich the decision-making process. We also concluded that a combination of different methods could be helpful/effective depending on objective, resources, and time. We believe that the most effective methods are the in-person interviews, workshops, public meetings, and landscape analyses. They should be based on real world or three-dimensional settings because a researcher can use all of his/her senses and can obtain more information to come to sensible conclusions.

We found that naturalness of the landscape and the features were shown to be highly preferred landscapes as were reported by previous studies. Therefore, future research can exclude already well-established features in order to avoid general or obvious findings. Careful research design is needed for specific details of preferences for landscape type and feature.

The significance of the study is that the local authorities and individuals can locate regions of natural and cultural heritages. Public knowledge regarding landscapes and landscape attributes would significantly impact landscape planning and decision-making which also aligns with a bottom-up approach of the concept of a Geopark (Dowling 2017). The significance of this study also lies in the Landscape Character Assessment (LCA) guideline of “people’s perspective” which plays a vital role in landscape planning and management. This would help the identification of potential geosites through peoples’ involvement and preferences. The findings can be helpful for students, researchers, locals, government, and other stakeholders to communicate about important locations and most preferred landscapes so that the land management policies incorporate this information in their plannings and managements. It also advocates for geoparks and geotourism as new approaches to benefit from the landscapes of this region. This study connects community and conservation through non-exploitive, educational and appreciative ways of diversifying lands.

A bigger sample size would be helpful in gaining more confidence in the accuracy of the results. Due to COVID-19, on-site surveys could not be conducted which can be considered for future research. It would also be interesting to analyze the amount of natural or cultural characteristics in each picture. Likewise, the degree of lighting and respondents’ familiarity with the region maybe considered to investigate how they affect preferences.