Spatial analysis of landscape social values in multifunctional landscapes of the Upper Missouri River Basin

The Upper Missouri River Basin (UMRB) was used as a study region to examine how the spatial distribution of social values (SVs) people hold for their landscapes overlaps with ecoregions of the UMRB and how the results can be interpreted to inform regional-scale strategic landscape management for optimizing the production of ecosystem services. The methods include a survey of residents about their landscape SVs across the UMRB and a multistage spatial analysis of eleven georeferenced SVs linked to ecosystem services and human activities in the region. The results show agriculture, recreation, and conservation are, respectively, the most significant activities that people value in landscapes of the UMRB. An overlapping area was identified between hotspots of recreation, tourism, and cultural activities on the one hand and conservation on the other hand. A large overlap was found between landscapes of social importance (i.e., landscapes where residents value the land) and ecoregion junctions (i.e., the meeting points of ecoregions). More than one-fourth of the mapped points were no more than 5 km from the meeting points of ecoregions, meaning ecoregion junctions are highly attractive to residents, and therefore, subject to a variety of human activities. It was also shown that although distance to the place of residence may influence the spatial distribution of mapped points, proximity to ecoregion junctions is a stronger factor for predicting the locations people value on landscapes. These findings have important implications for food and energy production, cultural ecosystem services, and biodiversity across landscapes of the UMRB.


INTRODUCTION
Regional-scale strategic landscape management is essential to ensure sustainability and the healthy functioning of ecosystems (Wu 2013, McDonough et al. 2017. This is mainly because mega-trends like climate change, biodiversity loss, and human-induced land-cover change have profound regional impacts on the environment (Martay et al. 2017, Northrup et al. 2019, Zhang et al. 2019. The social values (SVs) people hold for their landscapes, among other factors, play an indispensable role in changing the structure and function of landscapes (Botequilha Leitão et al. 2006, Wagner and Gobster 2007, Plieninger et al. 2015, Varela et al. 2018. Social values can determine where, how, and to what extent human activities can occur on the landscape and this, in turn, can provide key information for regional-scale strategic landscape management. Understanding values people assign to the land at smaller scales (e.g., landscapes) can be used as a vehicle to evaluate how humans interact with the land at broader scales (e.g., ecosystems and ecoregions). Brown and Reed (2012) introduced the concept of social landscape metrics to quantify human perceptions of landscapes using GIS. Public Participation Geographic Information Systems (PPGIS ;Sieber 2006), Social Values for Ecosystem Services (SolVES; Sherrouse et al. 2011Sherrouse et al. , 2014, and Landscape Values Mapping (LVM; Biedenweg et al. 2019) are also among recent initiatives for incorporating social-value information into land analysis, assessment, and management at varying scales. Mapped SVs are correlated with people's attitudes toward land use, can be predictive of land-use conflicts, and seem to be valid at various scales (Brown et al. 2020). Such models can support collaborative land-use planning and management (Tyrväinen et al. 2007), help to better understand land-use and land-cover change patterns (Hu et al. 2019), provide a basis for ecosystem services assessment (Zhang et al. 2019), and underpin public involvement in policymaking in protected areas (Johnson et al. 2019). Although mapped SVs are generally employed to study multidimensional and complex interactions between humans and the land (Alessa et al. 2008, Cerveny et al. 2017, incorporating this information into actual decision support is, to a large extent, undeveloped (Brown and Fagerholm 2015), more specifically at the broad scale (Brown et al. 2014). Therefore, research is needed to develop and test methods through which interactions between SVs and landscapes can be best modeled in a spatially explicit manner in a way that the outputs can be understood and used by a wider range of stakeholders at broader spatial scales.
Using the Upper Missouri River Basin (UMRB) as the study region, the aim of this study was to examine (1) how the spatial distribution of SVs people hold for their landscapes overlaps with ecoregions, (2) what areas can be considered as landscapes of social importance (i.e., landscapes where residents value the land), and (3) how the results stemming from the spatial analysis can be interpreted to inform regional-scale strategic landscape management. Two novel aspects of this study are addressing an extensive land area in a relatively rural, largely agricultural landscape and performing a novel multistage spatial analysis in relation to the spatial extent of ecoregions. Approximately half of the UMRB is covered by grassland ecosystems, as an important mosaic of the Great Plains of North America. These grasslands provide a wide range of ecosystem services for their inhabitants (Derner et al. 2006, Gleason et al. 2011, Schulte et al. 2017. Grasslands occupy about 40% of the earth's surface, and more than 800 million people live in this ecosystem, worldwide (White et al. 2000). Therefore, this work provides an important baseline for future research into social values of low population density grasslands worldwide (e.g., the prairies of North America, Australian rangelands, the pampas, savannas of Africa, and tussock grasslands of New Zealand).

Study region
The UMRB encompasses large areas of Montana, North Dakota, South Dakota, and Wyoming, as well as small parts of Nebraska, Iowa, and Minnesota, cumulatively covering more than 746,000 km 2 of the United States. The UMRB is among the world's least populated areas, and its landscapes significantly contribute to the production of energy and food in the United States (Stoy et al. 2018). Since the beginning of Euro-American colonization of the United States (Glenn 2015), landscapes of this region have undergone widespread land-cover change (Auch et al. 2011, Sleeter et al. 2013). The UMRB is contained within the North American grasslands, which are the largest continuous biome in North America (Samson and Knopf 1994). The continued conversion of grasslands to agricultural lands in the eastern UMRB (Wright and Wimberly 2013) has had far-reaching consequences for ecosystems and wildlife (Sieg et al. 1999, Wimberly et al. 2018. Primary ecosystem services in the UMRB involve food and energy production, cultural services (e.g., tourism, recreation, and aesthetics), water supply, as well as the provision of habitat and food sources for wildlife.

Social values
The study was designed to convert SVs of people to spatially explicit patterns applicable to regional-scale strategic landscape management.
The process included a multistage GIS-based spatial analysis based on a set of SVs data collected through a survey of residents of the UMRB during 2018 and 2019 (Jarchow et al. 2018, Carnes 2019. Eleven SVs were selected to measure where and for what reasons people value landscapes of the UMRB (Table 1). The list of social values was finalized after reviewing the literature and conducting 33 semi-structured interviews with stakeholders about SVs and landscapes of the UMRB. Detailed information about these interviews is available in Carnes (2019). The chosen SVs are strongly connected to fundamental ecosystem services that are necessary for supporting landscape sustainability and livelihoods in the region and beyond (Atwell et al. 2009, Genareo and Filteau 2016, Jarchow et al. 2020. A questionnaire was designed in Survey123 based on these SVs and circulated across the study region during 2018 and 2019. We conducted door-to-door surveys in 22 cities and rural communities ranging in size from more than 109,000 residents to less than 400 residents (i.e., residential areas; Table 2). We chose which residential areas to survey based on the built-up area's size and location. In residential areas with more than 2500 residents, we used stratified random sampling to select neighborhoods throughout the area to survey and then surveyed most houses within those neighborhoods. In residential areas with fewer than 2500 residents, there generally were not neighborhoods, so sampling occurred throughout the area. The sampling intensity was based on residential area size and ranged from 8 h for small rural communities to up to 20 h for the largest cities.
As part of a more comprehensive project, a digital map of the UMRB was provided for participants and they were asked to indicate locations around their area of residence (i.e., 80 km) that they associate with their SVs of interest. In this study, the focus was placed on this mapbased question. Results derived from other parts of this survey are reported in other publications. Participants were asked to select up to four SVs and to pinpoint on a digital map the locations of interest with respect to the selected SVs. In total, 5481 points were specified by 1009 participants  Note: Detailed information about the population data is available in U.S. Census Bureau (2018). across the region (total population of the residential areas =~430,000 people; U.S. Census Bureau 2018; Tables 2, 3). About 91% (n = 917) of participants identified themselves as White. More information about the survey is available in Jarchow et al. (2018).

Spatial data
The meta-dataset Level III Ecoregions of the Continental United States (U.S. Environmental Protection Agency 2013) was used as the basic spatial framework in this study. It maps 105 ecoregions in North America, providing ecological information about the characteristics of similar ecosystems in each region, including vegetation cover, wildlife, hydrology, soil, and climate. Applying this meta-dataset to regional planning and management has been widely recommended (Omernik 1987, Wiken et al. 2011, Omernik and Griffith 2014. As Level III Ecoregions comprise many landscapes and ecosystems with similar ecological, environmental, and climatic features, they provide a coherent and robust regional framework for a more integrated management of socio-ecological interactions (Omernik and Griffith 2014).

Socio-spatial analysis
Several multi-scale spatial analyses were undertaken using ArcMap v.10.5 (Esri Inc., Redlands, CA, USA), including an overlay analysis of SVs and ecoregions, a hotspot-coldspot analysis of landscapes of social importance, and a weighted overlay analysis of SVs, ecoregions, and residential areas in the UMRB. To prepare the data, the points (n = 5481) collected during the survey were georeferenced in the GIS environment. These points were overlaid with the UMRB's boundaries in order to exclude points that were indicated outside of the study region. As a result, 5046 points were identified across the UMRB. These points were regarded as locations where people value landscapes of the UMRB based on one of the SVs listed in Table 1. The meta-dataset Level III Ecoregions of the Continental United States was used to map the spatial extent of ecoregions across the region. To depict how the UMRB is covered by these ecoregions, this dataset was overlaid with the boundaries of the study region. Finally, SVs were overlaid with ecoregions of the UMRB to examine how the points indicated by participants are linked to ecoregions.

Hotspot-coldspot analysis
To map landscapes of social importance, a hotspot-coldspot analysis was conducted. For doing this analysis, an individual layer was created for each SV. Considering the extent of the UMRB, a x,y tolerance of 10 km was assigned to integrate nearby points for each SV. A weight was also allocated to each integrated point based on the number of points collected in each location. The incremental spatial autocorrelation tool was run to calculate z scores for each integrated point. The values of z scores indicate standard deviations for each point and can be interpreted with respect to P values (Mitchell 2005). It means statistically significant hotspots and coldspots in the GIS environment are validated where the values of z scores and P values are, respectively, very high and very low. According to the methods employed in this study, an extremely high or low density of mapped points in a certain location leads to a high value of z scores. If the z score of a SV in a location is greater than +1.96, it can be interpreted as a hotspot with the P value of <0.05 (confidence level = 95%). If the z score of a SV in a location is less than −1.96, it can be interpreted as a coldspot with the same P value. To detect and visualize clusters of social values in space, the Getis-Ord Gi * statistic was utilized (Getis and Ord 1992). This statistic was calculated using the following formula: v www.esajournals.org where x j is the attribute value for feature j, w i,j is the spatial weight between feature i and j, n is equal to the total number of features. More information about these calculations is available in https://pro.arcgis.com/en/pro-app/latest/tool-refe rence/spatial-statistics/h-how-hot-spot-analysisgetis-ord-gi-spatial-stati.htm. The resultant spots were interpolated based on the value of z scores using inverse distance weighting. As a result of this process, 11 maps were produced corresponding to the SVs listed in Table 1. Applying the quantities of z scores and P values, nine classes were determined, on a spectrum from extremely cold to extremely hot. The maps represented spatially explicit patterns of hotspots and coldspots of the SVs across the study region.

Weighted overlay analysis
The 11 maps generated in the hotspotcoldspot analysis were used to perform a weighted overlay analysis of the extent and density of mapped points in the study region. A unique score (S i ) was assigned to each pixel (1 ≤ S i ≤ 9) based on the nine classes determined in the hotspot-coldspot analysis (from 1, extremely cold to 9, extremely hot). In addition, a normalized weight was dedicated to each SV using its relative importance calculated based on the responses received from participants during the survey (Table 4). The weighted overlay analysis determined the importance of the landscape based on both the weight of SV and point density.
For each map, the normalized weight of the SV was multiplied by the unique score of each pixel (pixel size = 30 m 2 ) to measure the weighted score of that pixel. This process was performed for the 11 maps produced in the hotspot-coldspot analysis. Next, maps were overlaid using the weighted overlay analysis tool. As a result of this process, five classes were defined from most important to not important, depicting the density of values assigned by participants to landscapes of the UMRB. Detailed information about the application of weighted overlay analysis in landscape planning and management is available in Hill et al. (2005), Wang and Hofe (2007), and Rastandeh (2018).
This map was overlaid with ecoregions to examine to what extent hotspots and coldspots of social values overlap with the meeting points of ecoregions in the UMRB, termed in this study as ecoregion junctions. Ecoregion junctions, like ecosystem junctions (i.e., ecotones), are biologically important areas, where different types of ecoregions converge and meet, and therefore, these areas are believed to be critically important for biodiversity conservation, as well as protecting wildlife-dependent ecosystem services such as seed dispersal, pollination, pest control, and soil and water protection (Rastandeh 2018). Three bands of 5, 10, and 25 km were applied from the edge of each ecoregion to define three buffer widths of ecoregion junctions, responding to a variety of ecological processes among abiotic and biotic resources, including multi-scale predator-prey interactions, species dispersal patterns, pollination, and seed dispersal (Dramstad et al. 1996, Bowman et al. 2002, Feldhamer et al. 2003). In addition, 10 5-km buffers were applied from the center of residential areas to examine whether proximity to residential areas is a determinant factor in the spatial distribution of mapped points.

Social values
More than half of the mapped points were for three SVs: agriculture (19.99%, n = 1009), Fig. 1. The spatial distribution of ecoregions, sampled residential areas, and points indicated by participants across the study region (cf. Table 5). Casper was sampled because it is the only metropolitan area in Wyoming close to the UMRB. The spatial extent of ecoregions was extracted from the meta-dataset Level III Ecoregions of the Continental United States (U.S. Environmental Protection Agency 2013). recreation (19.21%, n = 969), and conservation (13.79%, n = 695; Table 5). Conversely, development (3.88%, n = 196) and spiritual (3.98%, n = 201) were the least important SVs associated with landscapes of the UMRB. No point was detected in Idaho Batholith (ER16) and Nebraska Sand Hills (ER44). More than half of the points were in Northwestern Great Plains (ER43), which covers approximately 46% of the study region. In the Middle Rockies (ER17), 1179 points were specified by participants, which was more than 23% of mapped points. This ecoregion encompasses about 13% of the study region. While more than 22% of the UMRB is dominated by Northwestern Glaciated Plains (ER42), less than 14% of points (n = 694) were indicated within this ecoregion.

Landscapes of social importance
Hotspots and coldspots of each SV differed in terms of size and location, which revealed that landscapes of the study region were not equally valued by people (Fig. 2). Note: SVs have been coded as Aesthetics (I), Agriculture (II), Community (III), Conservation (IV), Cultural (V), Development (VI), Economy (VII), Energy (VIII), Tourism (IX), Spiritual (X), and Recreation (XI).
The highest value of z scores was for energy (+2.78) and aesthetics (+2.58; Table 6). Hotspots of these two SVs mainly cover the southern and central landscapes of the study region. Energy has the highest range difference of z scores (4.48) meaning the difference between the coldest and hottest spots of landscapes of the UMRB is extremely high for energy. Hotspots of aesthetics, conservation, cultural, recreation, and tourism mainly overlap; however, the extent, density, and size of their hotspots are not the same. The largest hotspots were for recreation and aesthetics. Although agriculture was the most commonly identified SV, a very small part of the study region had the z score that was greater than +1.83, revealing that agriculture was valued throughout the UMRB rather than in particular spots. Fig. 2. Hotspots and coldspots of the SVs across the UMRB; z score >+2.58 or <−2.58, P value of <0.01 (confidence level = 99%); z score >+1.96 or <−1.96, P value of <0.05 (confidence level = 95%); z score >+1.65 or <−1.65, P value of <0.1 (confidence level = 90%). The minimum and maximum z scores of each SV are provided in Table 6. 3.70 † z score >+2.58 or <−2.58, P value of <0.01 (confidence level = 99%); z score >+1.96 or <−1.96, P value of <0.05 (confidence level = 95%); z score >+ 1.65 or <−1.65, P value of <0.1 (confidence level = 90%). v www.esajournals.org

Human activities at ecoregion junctions
Five classes were created for landscapes of social importance in the UMRB, ranging from most important to not important (Fig. 3).
There are two hotspots that cumulatively cover about 3% of the study region. The larger hotspot overlaps with the Middle Rockies (ER17) adjacent to Northwestern Great Plains (ER43). The smaller hotspot is at the meeting point of Northwestern Glaciated Plains (ER42), Northern Glaciated Plains (ER46), and Western Corn Belt Plains (ER47) located in the southeastern part of the UMRB. In addition to these two hotspots, there is a relatively wide linear hotspot that encompasses about 17% of the UMRB and continues along the western edge of Northwestern Great Plains (ER43) adjacent to other ecoregions. These three hotspots are within or around ecoregion junctions. Areas not in these  three hotspots form a large crescent from the northwest to the southeast landscapes of the study region (Fig. 3).
The detected hotspots are within or around three buffer widths of ecoregion junctions (i.e., 5, 10, and 25 km). The size of overlapping areas between hotspots and ecoregion junctions is decreasing, from a band of 5 km to a band of 25 km, showing a higher concentration of mapped points at the meeting point of ecoregions. Half (n = 11) of the residential areas sampled in this study fall in a band of 25 km from the meeting points of ecoregions, but only seven residential areas are in a band of 5 km. Despite this, 25% (n = 1281) of the mapped points fall within ecoregion junctions defined by a band of 5 km. This band covers only 12% of the entire region ( Table 7). The Euclidean distances between the residential areas and their nearest ecoregion junctions vary from 0.1 to 119.8 km (mean, 38.4 km), and about 38% of the mapped points, excluding Casper, were between 0 and 5 km from the center of sampling residential areas (Table 8).
The correlation between the number of mapped points and distance to the place of residence was negative for all residential areas, and for 7 out of 22 residential areas, it was statistically significant (Table 9). The distance of these seven residential areas to their nearest ecoregion junctions varied from 0.4 to 58.8 km (mean: 24.8 km; Table 8). In addition, there was no correlation between the number of mapped points that fell within 5 and 10 km from the center of residential areas and the Euclidian distance of a residential area to its nearest ecoregion junction. Only one statistically negative correlation was observed between these two factors, where mapped points were between 10 and 25 km from the center of residential areas (Fig. 4).

DISCUSSION
The inclusion of non-market values, including SVs, into decision-making processes is a requirement for multi-scale land planning and management (Sherrouse et al. 2014, Bagstad et al. 2016. To meet this requirement, the first step is to devise methods through which SVs can be converted to measurable spatial data. In this study, a multistage spatial analysis of SVs v www.esajournals.org associated with landscapes of the UMRB was performed to bring social and spatial data together in order to inform strategic landscape management for the production of ecosystem services, and provide a basis for addressing complex socio-environmental issues linked to landscapes of the region. The results derived from this study have important implications for regional-scale strategic landscape management. These implications are briefly discussed in relation to major ecosystem services of the UMRB.

Food and energy production
Landscapes of the UMRB are highly valued by people for agriculture. Despite this, no major  Fig. 4. The relationship between the distance of residential areas to the nearest ecoregion junction (DNERJ) and the proximity of mapped points to residential areas in terms of the Pearson correlation coefficient (−1 ≤ r ≤ +1), at the confidence level of 95%. The black points on the plots represent residential areas sampled in this study. (Fig. 4. Continued) v www.esajournals.org hotspot was observed for this SV (Fig. 2). It is mainly because the distribution of points dedicated to this SV is not limited to certain locations of the UMRB. The economy and energy hotspots in Montana were, to some extent, surprising given the importance of the oil shale extraction in North Dakota and coal in Wyoming (Fig. 2). This can be due to the impacts of the sector participants were employed, an influence of the methods used to sample residential areas in these two states, or a combined impact of both.
A classification of SVs in terms of the number of mapped points in ecoregions showed that in the Middle Rockies (ER17), recreation, conservation, and aesthetics were prioritized over agriculture, and in the Wyoming Basin (ER18) and the Western Corn Belt Plains (ER47), agriculture was selected as a SV less frequently than in other ecoregions ( Fig. 5; cf. Table 5). Areas covered by agricultural land play a pivotal role in supporting food security, not only in the UMRB, but also in the United States. For example, about onethird of wheat production in the United States is harvested in the UMRB (Stoy et al. 2018). This level of food supply contributes to the employment of many residents as well (U.S. Census Bureau 2018). At the same time, agricultural practices in the UMRB may have adverse impacts on water quality (Secchi et al. 2011) and biodiversity (Samson and Knopf 1994).
Areas used for agriculture also can be regarded as an alternative source of energy, especially cornbased ethanol in the eastern UMRB; and concurrently as a potential tool for mitigating the impacts of climate change through BioEnergy with Carbon Capture and Storage (BECCS; Obersteiner et al. 2001). BECCS is based on a climate policy aimed at limiting climate change to 2°C or less (Beck and Mahony 2017). While this policy helps to remove CO 2 from the atmosphere, it concurrently contributes to the sustainable provision of energy. The UMRB may have the potential for implementing BECCS (Stoy et al. 2018). In this study, an overlap was detected, from the west to southwestern parts of the region, which is valued for both agriculture and energy (Fig. 2). Parts of this overlapping area are also covered by landscapes that are valued for economy. This overlapping area can be considered as a candidate site for future research to examine whether applying BECCS in areas dedicated to agriculture is socially acceptable, ecologically feasible, and economically justifiable. Because agriculture is strongly linked to U.S. traditions, history, and culture in the UMRB (Carnes 2019), it strongly affects land-use choices in the region.

Cultural services
A broad spectrum of services related to recreational activities, spiritual inspiration, and cultural identity is provided by ecosystems of the UMRB. Cultural ecosystem services are essential to ensure human well-being (Tengberg et al. 2012, Plieninger et al. 2015. Expanses of landscapes that are valued for these activities are not identical (Fig. 2). In three ecoregions, recreation was found to be the most demanding SV, and in the Middle Rockies (ER17), aesthetics was among the most demanded SVs people hold for their landscapes (Fig. 5). Landscapes of the southern part of the study region, in particular, are largely valued for recreation. In this area, there is an extensive overlap between hotspots of recreation, tourism, and cultural activities on the one hand and conservation on the other hand. This overlapping area mainly covers the Middle Rockies (ER17), including the Black Hills, a strategic and iconic place for the provision of habitat and food sources for biodiversity. Spiritual values are low in places of high values to Native Americans (e.g., the Black Hills), perhaps because Native Americans were not intentionally surveyed in this study. As identified, recreation and conservation are, respectively, ranked the second and third SVs (Table 5). One implication of this finding is that conservation efforts and recreational activities may contradict in overlapping areas, mainly because recreational activities require infrastructures and facilities that may be harmful to natural areas. For example, an assessment of the values of public lands in Australia showed there exists the potential management conflict in national parks, which are of conservation importance (Brown et al. 2014). Even light recreational activities in nature (i.e., hiking) can alter daily patterns of wildlife activity (Gaynor et al. 2018). Conflicts could also be driven around issues of huntable and fishable species that are highly valued for recreation in the region. In addition, noise and light pollutions are other factors that can negatively affect wildlife (Gaston et al. 2013, Slabbekoorn et al. 2018. Understanding these contradictions and conflicts is an essential step toward achieving environmental sustainability. Spatially explicit information, similar to what is depicted in Fig. 2, can be regarded as a means to inform regional-scale strategic landscape management.

Ecoregion junctions and social values
Most recently, a GIS-based mapping of SVs in landscapes of southeastern Arizona showed areas along streamlines (i.e., the meeting point of rivers and terrestrial ecosystems) were highly valued by residents (Petrakis et al. 2020). In our study, in particular; the overlap between landscapes of social importance and ecoregion junctions was detected (i.e., more than 25% of the mapped points were within a band of 5 km from ecoregions; Table 7). This means ecoregion junctions were of greater interest to residents for various human activities compared to locations not at ecoregion junctions. One implication of this overlap can be an increase in the risk of habitat degradation and biodiversity loss in the long run, because human activities are inherently associated with varying environmental impacts. For example, the concentration of human activities (e.g., food production, energy extraction, urban development, recreation, etc.) at ecoregion junctions can be associated with changes in wildlife activity patterns, leading to more widespread environmental pollutions in areas of conservation importance (Gaynor et al. 2018, WWF 2018. Changes in natural patterns of wildlife activity can affect biodiversity as a whole, and a wide range of ecosystem services, accordingly (Balvanera et al. 2006, Cardinale et al. 2006. The most important SVs in terms of the number of points counted in ecoregions of the UMRB. The Idaho Batholith (ER16), High Plains (ER25), Canadian Rockies (ER41), and Nebraska Sand Hills (ER44) were excluded due to the limited number of points indicated by participants (n < 10). Detailed information about all SVs in the 10 ecoregions is available in Table 5.
v www.esajournals.org et al. 2012), and this, in turn, requires stakeholders to develop a more comprehensive understanding of multidimensional land-use issues at ecoregion junctions, and take more effective measures in ways that mitigate the impacts of dense and diverse human activities on abiotic and biotic resources.
This coincidence in the UMRB can be explained in two different ways. First, people often value landscapes that are close to their settlements; perhaps because access to remote areas for benefiting from ecosystem services can be costly, if not impossible (cf. Lovelock 2010). Second, the meeting points of ecoregions are more attractive to people due to the diversity of abiotic and biotic resources, which naturally leads to a higher level of ecosystem services, including aesthetics values. The latter explanation is compatible with the findings observed by Lindemann-Matthies et al. (2010) and Schirpke et al. (2013). Globally, most human settlements have been established at ecosystem junctions (Kühn et al. 2004, Stewart et al. 2004, Alvey 2006, Elmqvist et al. 2013, Güneralp et al. 2013, Mayer-Pinto et al. 2015, Ives et al. 2016. This study showed urban development in the UMRB has occurred at ecoregion junctions (i.e., 7 out of 22 residential areas sampled in this study were less than 5 km from the ecoregion junctions; Table 8).
The density of mapped points around ecoregion junctions observed in this study was first assumed to be due to the proximity of residential areas to ecoregion junctions and/or the natural attractiveness of these areas (Fig. 6). Further analysis, however, showed that where residential areas were not in close proximity to ecoregions, the number of mapped points was greater around residential areas, but where residential areas were closer to ecoregion junctions, the distance to residential area was not a determinant factor to the spatial distribution of mapped points (Fig. 7). In addition, it was revealed that proximity to residential areas was not necessarily correlated with the number of mapped points (i.e., the location of landscapes of social importance). Accepting this, point density around ecoregion junctions cannot be necessarily attributed to proximity to the residential areas established at ecoregion junctions. Rapid City, for example, is located at the confluence of the Middle Rockies (ER17) and the Northwestern Great Plains (ER43; Fig. 1 and Table 8). At least 617 points were mapped in a radius of 50 km from the center of this city. However, there was no statistically significant relationship between the spatial distribution of mapped points and proximity to this city (Fig. 7). These observations also indicate that although the instruction applied during the survey to indicate the location of SVs on the map (i.e., within a radius of 80 km from the participants' places of residence) led to an aggregated pattern of mapped points around residential areas, this instruction had no significant impact on the point density observed at a band of 5 km from the meeting point of ecoregions. Overall, these findings suggest that proximity to ecoregion junctions is a stronger factor in predicting the spatial distribution of mapped points.
Two major hotspots were detected in the south and southeastern parts of the UMRB, overlapping with four residential areas with populations of more than 10,000 people (i.e., Yankton, Rapid City, Sheridan, and Billings; Fig. 3). Landscape fragmentation triggered by human activities has consequences for wildlife habitats in the region (Wimberly et al. 2018, Jarchow et al. 2020. This confluence (i.e., urban development at ecoregion junctions) may threaten vulnerable ecosystems and species, more specifically where SVs linked to agriculture, development, energy, and economy are highly valued (cf. Bernath-Plaisted and Koper 2016, Correll et al. 2019). Fig. 7. Examples of the spatial distribution of mapped points in terms of the distance to the center of residential areas in 10 bands of 0-5, 5-10, 10-15, 15-20, 20-25, 25-30, 30-35, 35-40, 40-45, and 45-50 km.

Limitations
The accuracy of social data was fundamental to ensure that spatially explicit patterns of the SVs associated with landscapes of the UMRB are as detailed as possible. During the survey, the highest level of assistance was offered to participants in order to enable them to pinpoint their locations of interest in a proper manner. In practice, however, it was not possible to estimate how participants comprehended SVs, their landscapes, spatial scales, and land-cover classes, because the spatial literacy of people can affect this process (cf. Escobedo et al. 2020). In addition, the number of points assigned to each SV, as well as the locations where participants live and/or work could affect their choices. Ideally, a higher number of informed participants who indicate more accurate points can help to generate more detailed maps in the future. In addition, the spatial resolution of this analysis, while likely acceptable for such a large and sparsely populated area, introduces uncertainties when using interpolation methods that do not account for underlying environmental and sociopolitical variables (cf. Sherrouse et al. 2011). These issues must be taken into consideration for future research.
About 6% of the total population of the UMRB are Native Americans ) and more than 10% of the study region is sovereign nations representing the Native American reservations. Areas in the southern part of the UMRB highlighted for recreation and other values raise the question of recreation and values for whom, given the large expanses of Native American lands are located in western South Dakota, in particular. Although more than 20 Native American tribes manage the land across the UMRB (Stoy et al. 2018), Native American reservations were not surveyed because, as sovereign nations, human-subjects research must be approved by each tribal nation. Of the total number of locations indicated by participants as landscapes of social importance (n = 5046), less than 2% (n = 100) were within Native American lands. This limitation should be compensated in future research with reference to previous successful studies (Watson et al. 2009). Very little value was ascribed to iconic places like Yellowstone National Park, perhaps because in this study only residents of UMRB participated in the survey. It is expected that visitors, including international tourists, who are not residents of the region, may have different attitudes toward landscapes of iconic places (e.g., national parks and reserves).
This study is a starting point for further research on SVs, landscapes, and ecoregions. There was no specific focus on any certain species and habitats in this study; however, for future research, biodiversity datasets such as the Gap Analysis Project Species Habitat Maps (U.S. Geological Survey 2018) can be used to address how the confluence of ecoregion junctions with landscapes of social importance may impact particular habitats and species of interest.

CONCLUSION
Mapping the SVs people hold for their landscapes provides useful information for optimizing the production of ecosystem services through regional-scale strategic landscape management. The overlap between these SVs and ecoregion junctions was observed for the first time in this study. Agriculture, recreation, and conservation are the most significant SVs in landscapes of the UMRB. These SVs are strongly related to the primary services that are provided by ecosystems of this region. More than onefourth of the mapped points were no more than 5 km from the meeting points of ecoregions, meaning ecoregion junctions are highly attractive to residents, and therefore, subject to a variety of human activities. It was also shown that although distance to the place of residence may influence the spatial distribution of mapped points, proximity to ecoregion junctions is a stronger factor for predicting the locations people value landscapes. The results derived from this study combined with limitations discussed suggest that more case studies are needed to better understand how SVs are linked to landscapes at the regional scale.

ACKNOWLEDGMENTS
This work was supported by the National Science Foundation through the EPSCoR Track II cooperative agreement OIA-1632810 and NSF DBI-1560048. We thank Darius Semmens, Selena Ahmed, Julia Haggerty for feedback on the survey instrument and methods;