RouteLAND: An Integrated Method and a Geoprocessing Tool for Characterizing the Dynamic Visual Landscape Along Highways

Abstract

Moving away from a static concept for the landscape that surrounds us, in this research article, we approach the visual landscape as a dynamic concept. Moreover, we attempt to provide an interconnection between the domains of landscape and cartography by designing maps that are particularly suitable for characterizing the visible landscape and are potentially meaningful for overall landscape evaluation. Thus, the present work mainly focuses on the consecutive computation of vistas along highways, incorporating actual landscape composition—as the landscape is perceived from an egocentric perspective by observers moving along highway routes in peri-urban landscapes. To this end, we developed an integrated method and a Python (version 2.7.16) tool, named “RouteLAND”, for implementing an algorithmic geoprocessing procedure; through this geoprocessing tool, sequences of composite dynamic geospatial analyses and geometric calculations are automatically implemented. The final outputs are interactive web maps, whereby the segments of highway routes are characterized according to the dominant element of the visible landscape by employing (spatial) aggregation techniques. The developed geoprocessing tool and the generated interactive map provide a cartographic exploratory tool for summarizing the landscape character of highways in any peri-urban landscape, while hypothetically moving in a vehicle. In addition, RouteLAND can potentially aid in the assessment of existing or future highways’ scenic level and in the sustainable design of new highways based on the minimization of intrusive artificial structures’ vistas; in this sense, RouteLAND can serve as a valuable tool for landscape evaluation and sustainable spatial planning and development.

1. Introduction

The visual landscape is a vital resource for human well-being. People’s landscape preferences are crucially and positively affected by the presence of natural landscape elements, such as water (e.g., [1,2,3,4,5]); (woody) vegetation (e.g., [5,6,7,8]); land of particular/mountain morphology (e.g., [3,9]); and overall elements of wilderness (e.g., [2]). On the contrary, certain landscape elements associated with the urban settings, such as industrial units, paved roads, electric power lines, and, in general, urban sprawl, arouse negative impressions (e.g., [2,5,10,11,12]). Yet, not all built elements are negatively associated; for instance, well-preserved rural man-made elements such as villas of vernacular architecture and traditional farm buildings can cause positive impressions [2,13].

Urbanization has profoundly transformed prevailing rural landscapes into urban ones over the last decades, inducing a dramatic alteration of land cover/use composition and structure [14]. Urban and peri-urban landscapes are composed of a great amount of artificial (i.e., non-natural), built elements; thus, natural landscape elements are typically sparse and, therefore, are particularly coveted in such landscape areas. Environmental psychology has long ago shown the positive effects of natural elements on human health and psychological well-being (e.g., [15]). The beneficial, restorative effects of nature, emerging from natural landscapes [6,16,17], are diminished in urban settings; in such settings of the built environment, urban residents tend to be more prone to mental diseases, suffering from increased stress levels [18,19,20]. As an effect, sustained interaction with the non-natural, built environment leads to the depletion of cognitive, affective, and physiological human functions [19,21].

Since 1950, the rapid growth of the world urban population—from 0.8 billion (30%) to 4.2 bullion (55%) [22]—is an alarming indicator regarding the decreasing exposure of human population percentages to natural landscape elements. By 2050, approximately 6.7 billion people (68%) will be estimated to live in urban settlements [22]—a fact implying a further reduction in the world population percentage that is expected to be exposed to natural landscapes.

One important aspect when approximating the visual urban and, particularly, peri-urban landscapes refers to the fact that a large number of people move or drive in certain linear routes, i.e., roads and highways, on an everyday basis. The peri-urban landscapes that are being produced due to increasing urban sprawl tend to be increasingly “cluttered, unattractive, and monotonous” [12]. In this sense, highways can be considered as critical markers of landscape transformation in peri-urban regions, directly influencing viewers’ visual perception and experience [12]. Given the dire necessity of urban residents to have contact with nature and natural elements in order to lead a healthy life [23], it is essential to study the visual experience of observers while moving in such routes.

The visual landscape can be modeled under two perspectives: either from the human (ground) standpoint, i.e., the way people experience landscape as they stand, move, etc., or in the way a landscape is cartographically represented, i.e., in 2D maps [24]. The first perspective constitutes the egocentric, while the second one constitutes the exocentric perspective (e.g., [24,25,26,27,28,29]). Furthermore, there is a crucial need to quantitatively describe not only the static but also the dynamic perspective when perceiving the visible landscape, especially along linear components, such as highways [12,30]. However, several issues are raised when it comes to the integration of the two available perspective modes of the visual landscape, especially when the dynamic visual landscape is to be modeled (see Section 2.1 and Section 2.2 for a more thorough investigation).

In this research work, our endeavor focuses on how to overcome these challenges from a conceptual but chiefly technical point of view. Hence, this article deals with modeling the dynamic landscape along highway routes from a human perspective. In this sense, it mainly aims at the development of both a method and a software tool for automatically and objectively characterizing highway routes depending on the (vehicles’) movement speed, the dynamic (human) field of view (dFoV) at all segments of the highway routes, and the actual relative contribution (in %) of each landscape element within the dFoV. Aside from developing and implementing this method, the article also aims at providing as final outputs interactive web maps visualizing the prevailing landscape elements within the dFoV attributed to all predefined segments of highway routes. In this context, a use case of the method and the tool is provided that refers to a selected study area. This case study does not lie at the core of this research work; it has been selected to display and showcase the usefulness and the usability of the developed geoprocessing tool in a peri-urban region.

2. Background and Related Work

2.1. Describing the Perceived Size of Landscape Elements from the Egocentric (Human) Perspective: Concepts and Issues

Describing the visual landscape experience as apperceived by humans, i.e., from the egocentric perspective, is a core matter in landscape studies [28,29,31,32,33]. Various methods, techniques, and technologies have been employed for registering/recording and measuring observers’ landscape visual perception, such as eye tracking, EEG, and fMRI (see [24] for a list of research studies). However, these approaches, no matter how valid they indeed are, are ‘plagued’ with subjectivity, in that the apperceived landscape impressions—in both behavioral (i.e., gaze patterns), and neuronal/cognitive (i.e., brain activity) terms—may significantly vary across potential observers. On the other hand, typical exocentric perspectives objectively represent the landscape but do not involve the observer, since they constitute vertical views of an area from above.

Therefore, having as a starting point the egocentric perspective, while moving a step back, a prerequisite is to define what parts of the landscape and which landscape elements can be visually perceived within a potential observer’s field of view (e.g., [28,31]). The quantitative description of the landscape elements that lie within the human field of view (hFoV) at any position (viewpoint) is a crucial prerequisite for further studying the visual perception and evaluation of the landscape, since this description provides an objective information of the landscape composition at the level of the hFoV.

In this line of thought, Brabyn [32] developed a GIS-based method for combining space, time, and landscape character: for a given track (i.e., route), they have calculated the contribution of different landscape character types (e.g., “High Hill, Indigenous Scrub”) in certain distance zones (0–2, 2–5, 5–10, and 0–10 km) by introducing “experions”, which are a new unit for measuring visual landscape experience. Despite the conceptual merits of this method, aspects related to the inadequacy of precisely calculating the actually perceived size of each landscape type and its relative contribution within the hFoV from the observer’s viewpoint have limited the utility of the method. Moreover, the relative time during which each landscape type is exposed to an observer poses another challenge. Other innovative, GIS-based modeling methods such as those developed in the research work of Schirpke et al. [31] and Misthos and Menegaki [28] are also indirect approximations of the perceived landscape composition from the egocentric perspective.

Some other research studies have addressed the issue of miscalculating the landscape elements’ perceived size by employing different means. For instance, Minelli et al. [34] calculated the perceived size of an object (i.e., wind turbines) as a proportion of its area in a varying field of view. Nutsford et al. [35] developed an index (VVI) to rectify standard viewshed analysis outputs by calculating the vertical angle (vertical angular size) between an observer’s eye level and the top and bottom points of each visible cell within a viewshed output; by also indirectly correcting the horizontal angular size, the VVI became a determinant of the visual significance of each visible cell on a terrain from certain observers (viewpoints), providing an accurate representation of landscape visibility from a human (egocentric) perspective.

2.2. Describing the Dynamic Landscape Along Higways: Concepts and Issues

Some decades ago, in the influential work of J.J. Gibson, the visual perception of the environment has been acknowledged as a dynamic process—in that humans, like animals, perceive the world around them as a sequence of changing ambient optic arrays—that persists during long acts of locomotion [36]. In today’s everyday life, and as urbanization keeps increasing, moving in vehicles is a common practice. So, people typically perceive a cityscape chiefly by means of the dynamic visual experience [37]. The dynamic visual landscape is a concept which specifies the landscape’s visual experience from the perspective of moving observers (e.g., [27,38,39]) and/or vehicles. Roads and highways, in particular, have expanded in order to meet increased urban area sprawl [40]. Provided that many highways have been designed without adequately taking into consideration visual landscape preferences or quality, visual landscape degradation is anything but rare [40]. Since highways are crucial physical linkages enabling transportation [12,40] but also offering to large numbers of moving observers the potential to witness impressive scenery [12,30], the quantitative description of the dynamic visible landscape along such linear components is a key issue.

In the last decades, some previous research studies have focused on visualizing and quantitatively analyzing the visible parts of a terrain from the perspective of a moving observer along linear routes in different topographic feature types (ridgelines, valleys, pass-lines, etc.) [27,38,39]. In these research studies, viewsheds in different types of routes have been visualized in animated maps, while statistical correlations and inferences among elevation/topography and viewshed area (acreage) have been established. Nevertheless, these studies have not dealt with the qualitative aspect (i.e., landscape composition) of the area viewed from each viewpoint/observer and have not taken into consideration the actual size of visible/not visible cells within the hFoV. In another research study, a general framework is provided which, by combining various indicators (e.g., sky expanse, landmark distribution, etc.) at multiple dimensions and emphasizing the continuity of the linear urban landscape (LUL), serves to quantify dynamic visual perception [41]. This study has taken into consideration the continuous and dynamic features of visual perception, quantitatively approximating the continuous dynamic visual perception of consecutive viewpoints along linear routes (LULs) and also under different movement modes (walking, cycling, and boating). Nonetheless, the emphasis in this study has been placed on modeling the dynamic/changing views of the cityscape expressed/extracted in the form of photographs or perspective drawings and also on the simultaneous utilization of visual evaluation indicators in order to assess LULs based on distinct scenario preference criteria and patterns.

3. Materials and Methods

3.1. Overall Rationale

The overall approach of this research refers to the selection of initial highway routes (polyline features) in a peri-urban area in Attica region in order to derive the dFoV’s landscape elements’ composition information, which is ultimately attributed to each segment of the route and visualized in properly designed interactive web maps. The implementation of the computations is realized via the execution of an autonomous geoprocessing Python tool, introduced and developed within the presented work, which receives as inputs open geospatial data, while the tool’s outputs are visualized using JavaScript and html technologies.

3.2. Input Data (Case Study)

For the implementation of the methodology analyzed below (Section 3.4), three different geospatial data inputs are required: (i) a polyline vector road network layer for creating the highway route (of the case study), (ii) a raster digital elevation/surface model (DEM/DSM) for deriving terrain information (elevation, slope, aspect), and (iii) a polygon vector land cover layer for deriving landscape elements’ information.

The presented case study relies on the exclusive use of open geospatial data. More specifically, the road network vector features used in the case study were adapted from OpenStreetMap (OSM) platform (https://www.openstreetmap.org/, accessed on 15 October 2024) and downloaded freely from BBBike Extract Service (https://extract.bbbike.org/, accessed on 15 October 2024). The crowdsourcing platform of OSM provides information stored in different fields, among others, for the road type (field: fclass), for the maximum allowed speed (field: maxspeed), etc. Regarding the road type, seven main road type categories are present, including motorway, trunk, primary, secondary, tertiary, residential, and unclassified networks; the OSM also includes special road- and path-type categories (e.g., pedestrian, path, footway, cycleway, raceway, etc.). The utilized polyline layer was extracted both spatially (only segments within the study area (see Section 3.3 for the description of the study area) were retained) and thematically (only segments/records classified as ‘trunk’ were retained) in order to meet the aims of this research study.

Furthermore, the European Digital Elevation Model (EU-DEM), provided by OpenDEM portal (https://www.opendem.info/, accessed on 15 October 2024), was employed in the performed analysis. EU-DEM has been derived by the Shuttle Radar Topography Mission (SRTM) and the Advanced Spaceborne Thermal Emission and Reflection radiometer (ASTER) Global Digital Elevation Model (GDEM) having a ~30 m (1 arc per second) spatial resolution. EU-DEM has been preserved by the Copernicus Land Monitoring Service (CLMS) implemented by the European Environmental Agency (EEA) and Joint Research Center of the European Commission.

In addition, the Urban Atlas Land Cover/Land Use 2018 (LU/LC 2018) (https://land.copernicus.eu/en/products/urban-atlas/urban-atlas-2018, accessed on 15 October 2024), also provided by the CLMS, was used as an input for implementation of the developed software tool. Urban Atlas constitutes a pan-European dataset that involves seventeen (17) urban classes of land use/land cover (LU/LC) with the Minimum Mapping Unit (MMU) of 0.25 ha and ten (10) rural classes with 1ha of MMU (https://land.copernicus.eu/en/products/urban-atlas/urban-atlas-2018, accessed on 15 October 2024). Urban Atlas is a satellite product with six-year update cycle, which is distributed as a vector dataset.

Aside from the three initial input datasets, two other datasets were derived from the EU-DEM—also required for performing the tool, namely, the terrain slope and the terrain aspect products—having the same spatial resolution as the EU-DEM (~30 m). In total, the input layers that were utilized are presented in

Table 1. Input datasets or layers accompanied by their formats and corresponding geometries, and by their source geospatial datasets.

Input Dataset (Layer)	Format—Geometry	Open Geospatial Dataset
highway routes	feature class/shapefile—polyline vector	OSM roads
terrain elevation	GeoTIFF—raster	EU-DEM
landscape elements	feature class/shapefile—polygon vector	Urban Atlas LU/LC
terrain slope *	GeoTIFF—raster	EU-DEM
terrain aspect *	GeoTIFF—raster	EU-DEM

* derived product.

.

3.3. Study Area

The area of concern is a peri-urban region in the southeastern Attica region, lying almost exclusively within the boundaries of the Lavreotiki municipality (whereas the selected route is exclusively within the Lavreotiki municipality). The study area consists of a high variety of LU/LC types (over 20 different types), including urban fabric (of varying density), roads, industrial areas, agricultural land, pastures, herbaceous vegetation, forests, water, sea, etc. This variety in LU/LC is directly connected with the landscape elements’ diversity or complexity in the study area. This area, being a peri-urban region, displays a high landscape complexity, since it comprises a mixture of elements found in urban, agricultural, and forest areas. Other landscape regions, for instance rural/agricultural landscape regions, display a much lower complexity [42].

The selected route—which was characterized in terms of landscape composition within the dFoV—is part of a highway (i.e., main road but not motorway) with the generic name “Lavriou” and “Markopoulo-Lavrio”. This route is an important dual carriageway road, i.e., divided road with two directions connecting Lavrio town and Eleochori settlement. The route length is approximately 5 km, while its general direction (from Lavrio toward Eleochori) is south–north; more precisely, for the first 1.5 km (from Lavrio), the general direction is southeast–northwest, while for the rest of the 3.5 km, the general direction is southwest–northeast. The reference map of the study area with the selected route is presented in Figure 1. As displayed in Figure 1, since the route has two directions, the vehicles can also move toward the opposite direction (from Eleochori toward Lavrio). In addition, points upon this route have been placed at a certain spacing interval (50 m) for the needs of the methodology implementation (see Figure 1 and Section 3.4.2).

Moreover, this route was divided in segments according to the maximum allowed speed. As will be shown in the following sections, the human field of view varies with movement speed. Therefore, the horizontal field of view for a hypothetical driver along this route was calculated based on the maximum allowed speed or speed limits, ranging from 50 to 80 km/h per segment of the route (and per movement direction) for the study area (harnessing the OSM field maxspeed).

3.4. Methodology

3.4.1. Overview

In order to attain the aims of the research study and address the issues and challenges pointed out in Section 2.1 and Section 2.2, two complementary tasks were accomplished. At first, a new method was developed and introduced for integrating a sequence of geoprocessing computations and geometric calculations in order to quantify the dominant landscape elements within the selected highway route’s dFoV. Secondly, an integrated algorithmic procedure, namely, a geoprocessing, Python-based software tool—named “RouteLAND”—was developed and implemented on the basis of the introduced method. Both the method development and the tool implementation can be considered to consist of certain semantically and sequentially distinct sub-routines (‘sub-models’). In the rest of this section, the five ‘sub-models’ are described separately in the respective sub-sections because of their autonomous rationale within the method’s and the geoprocessing tool’s algorithmic sequence. The five sub-models are summarized as follows:

• Sub-model 1: Generation of viewpoints along highway routes and computation of route segments’ viewing directions;
• Sub-model 2: Computation of (start and end) horizontal viewing directions based on the route segments’ mean viewing directions and speed limits (max speed) for further viewshed analyses computation;
• Sub-model 3: Iterative computation of viewsheds and other relationships or attributes (distance, elevation, slope, aspect, LU/LC) of the visible/viewshed cells (or points) associated with each respective viewpoint;
• Sub-model 4: Computation of viewing significance and actual landscape composition, at the level of the highway route’s viewpoints dFoV, based on the spatial/thematic relationships and analyses among the viewshed points’ attributes and the associated viewpoints’ positions;
• Sub-model 5: Qualitative and quantitative characterization of the highway routes regarding the dominant landscape element ‘occupying’ the dFoV per route segment (Figure 2).

3.4.2. Method Development: Rationale and Computations (Geoprocesses and Calculations)

Before implementing the main procedure (divided in the aforementioned five sub-models), some pre-processing tasks were executed. First, a 6 km buffer zone around the highway route was produced in order to restrict both the computation process within certain boundaries and the volume of the input data. Hence, the DEM dataset, along with its derivatives—slope and aspect datasets—and the Urban Atlas LU/LC dataset were clipped within this 6 km buffer zone. Furthermore, the initial highway route was dissolved in order to generate a single-record feature.

For the first sub-model, the required input data were (a) (i) the initial highway route (OSM polyline feature layer) and (ii) the dissolved highway route (polyline feature layer) and (b) the DEM (raster layer). At the beginning, the dissolved polyline feature was segmented, and point features were generated at a 50 m interval. This interval was selected for approximating the dynamic change of the visual landscape along a route for observers moving upon vehicles with relatively high speeds (>50 km/h), since other research studies have suggested 10 to 40 m intervals for ‘walking’ observers (e.g., [38,39]), and since the DEM’s resolution is approximately the same size (30 m). The previously generated points were ascribed (via spatial join) all the attributes (e.g., max speed, type, etc.) existing in each record of the initial OSM polyline feature layer. The dissolved highway route was further split at the 50 m interval points in order to adjust the geometry of the polyline route according to the generated points, while all the attributes of the points were transferred back (via spatial join) to the split highway route. A new field (“road_str”) was created in the new split highway route to retain each FID corresponding to each 50 m highway segment, and the mean direction (azimuth) was calculated for each 50 m segment and stored in a new field (“COMPASS”). The vertices of this directed polyline were converted to point features by retaining only the starting vertex of each directed 50 m segment. In addition, the elevation from the DEM was attributed to each viewpoint (“RASTERVALU” field), and the initial information of the max speed (or speed limit) was also transferred to these point features. In total, among others, useful information about (i) the elevation, (ii) the max speed, and (iii) the mean azimuth (i.e., mean viewing direction) of each segment of the route was finally stored in the 50-meter-spaced viewpoints produced (upon the route).

The second sub-model received as inputs the previously produced viewpoints (point feature layer). The horizontal field of view was calculated under a certain condition, becoming narrower as the (max) speed increased—based on the relevant literature (e.g., [43,44]). So, the horizontal fields were set to: 120° (azimuth) for very low speeds, 110° for speeds up to 30 km/h, 100° for speeds greater than 30 km/h and up to 65 km/h, 60° for speeds greater than 65 km/h and up to 80 km/h, 40° for speeds greater than 80 km/h and up to 100 km/h, and 20° for speeds greater than 100 km/h (Table 2). In this respect, the left-/right-hand azimuth angles relative to the mean (viewing) direction (“COMPASS”) were the half of the aforementioned values (horizontal half-angles), i.e., 60, 55, 50, 30, 20 and 10°, respectively, and were stored in a new field (“hlf_angle”). Two new fields were created (“st_angle” and “end_angle”) to store the horizontal start/end azimuth angles for each one of the 50-meter-spaced viewpoints. The start azimuth angles (“st_angle”) were calculated by abstracting the horizontal half-angles (“hlf_angle”) from the mean (viewing) direction (“COMPASS”) of each segment, while the end azimuth angles (“end_angle”) were calculated by adding the half-angle left-/right-hand azimuth angles (“hlf_angle”) to the mean (viewing) direction (“COMPASS”) of each segment, correspondingly. It is worth noticing that the final left-/right-hand azimuth angles were recalculated in order to fall in the interval [0, 360]°, satisfying the following conditions. The calculations of “st_angle”, “end_angle”—under conditions—are specified below as pseudocode:

• if st_angle < 0°: $st\_angle=(360-\left|st\_angle\right|)$°;
• else if 0° <= st_angle <= 360°: $st\_angle=(st\_angle)$°;
• else (if st_angle > 360°): $st\_angle=\left(360-st\_angle\right)$°;

  $$st\_angle=Compass-hlf\_angle;$$

  (1)

  if end_angle < 0°: $st\_angle=(360-\left|end\_angle\right|)$°;
• else if 0° <= end_angle <= 360°: $end\_angle=(end\_angle)$°;
• else (if end_angle > 360°): $end\_angle=\left(360-end\_angle\right)$°;

  $$end\_angle=Compass+hlf\_angle;$$

  (2)

Table 3. Start/end azimuth angles occurring by indicative (max) speeds and mean directions.

Max Speed (km/h)	Mean Direction (°)	Half-Angle (°)	Start Angle (°)	End Angle (°)
20	355	55	300	50
20	5	55	310	60
50	265	50	215	315
50	95	50	45	145
70	175	30	145	205
70	185	30	155	215
90	85	20	65	105
90	275	20	255	295
110	10	10	0/360	20
110	350	10	340	0/360

presents an illustration of calculating the start/end azimuth angles with regard to certain, indicative (max) speeds and mean directions along different highway segments.

Table 2. Speed limits per highway segment accompanied by their horizontal and horizontal half-(azimuth) angles.

Speed Limit (km/h)	Horizontal Angle (°)	Horizontal Half-Angle (°)
0	120	60
0 < sl <= 30	110	55
30 < sl <= 65	100	50
65 < sl <= 80	60	30
80 < sl <= 100	40	20
>100	20	10

Table 2. Speed limits per highway segment accompanied by their horizontal and horizontal half-(azimuth) angles.

Speed Limit (km/h)	Horizontal Angle (°)	Horizontal Half-Angle (°)
0	120	60
0 < sl <= 30	110	55
30 < sl <= 65	100	50
65 < sl <= 80	60	30
80 < sl <= 100	40	20
>100	20	10

Within the third sub-model, the geoprocesses and calculations were executed in an iterative manner. The input data were (i) the LU/LC (polygon feature layer); (ii) the DEM (raster layer); the two other DEM/terrain derivatives, i.e., (iii) the slope (raster layer) and (iv) the aspect (raster layer); and (v) the viewpoints (point feature layer) produced in the previous sub-model (2). The fundamental procedure here was the iterative selection of each one of the viewpoints within a loop; in this loop, both the FID and the elevation of each viewpoint were stored (fields: “fid_value” and “elev_value”), and the visibility/viewshed analyses were executed for each viewpoint. The parameters for this iterative viewshed analyses were specified either from the information calculated in the previous models and stored in the fields of the viewpoints (e.g., horizontal start/end viewing angle) or from information set at will (manually) in the RouteLAND tool (e.g., observer offset: 1 m; outer analysis radius: 5 km). The horizontal field of view had been already calculated in sub-model 2; the observer offset was set to 1 m, taking into consideration a typical height of drivers’ eyes in a conventional vehicle; the radius for computing the viewshed was confined to 5 km in order to diminish the extensive computation load, since landscape elements that are far away (>5 km) from a moving observer are considered to have a diminished or even negligible visual impact compared to elements that lie in an observer’s foreground. The latter assumption is endorsed in other research works and technical reports [29,45,46,47]. Now, for every viewpoint of the route, a ‘cloud’ of visible terrain cells was produced; cells that were not visible were excluded from the rest of the geoprocessing to significantly reduce the subsequent computation load. Additionally, for every iteration, a unique code number (FID field: “fid_value”) characterizing each viewpoint had already been stored in order to associate each one viewpoint with its ‘cloud’ of produced visible or viewshed cells—along with other outputs in subsequent geoprocesses. The several ‘clouds’ of viewshed terrain cells were converted into viewshed points (features) to enable the subsequent management of both descriptive and geometric information. In the next stage, the three DEM derived values, i.e., terrain elevation, terrain slope, and terrain aspect of the visible cells, were transferred into the viewshed points as separate fields, namely, “ELEV”, “SLOPE”, and “ASPECT”. By utilizing the LU/LC, information about the type of the visible landscape elements (fields: “code_2018”, “class_2018”) was also transferred—in a similar manner—in the ‘clouds’ of viewshed points. Moreover, the distance and the horizontal angle/azimuth from each viewshed point of the cloud to each associated viewpoint upon the route were computed as well, while this information was stored and two new fields as “NEAR_DIST” and “NEAR_ANGLE”, respectively. Finally, two attributes of every viewpoint of the route were joined with each associated ‘cloud’ of viewshed points, that is, the viewpoint FID (“VPoint” = “fid_value”) and the viewpoint elevation (“VPElev” = “elev_value”). In doing all the previous tasks, the several ‘clouds’ of viewshed points were attributed to the information of (a) elevation (i), slope (ii), aspect (iii), land cover (iv), and distance from viewpoints (v)—attributes unique for each viewshed point in every viewshed point ‘cloud’—as well as (b) FID (vi) and elevation (vii)—attributes shared/common in every viewshed point ‘cloud’. This information was stored in different fields of the several ‘clouds’ of viewshed points. It should be noted that each viewshed point ‘cloud’ was descriptively (i.e., non-spatially) distinguished from any other viewshed point ‘cloud’ on the basis of its unique FID—‘carrying’ the FID of the associated viewpoint along the route. The entire algorithmic procedure of the third sub-model described above is depicted in the following summarized flowchart (Figure 3).

In the fourth sub-model took place all the calculations within each viewshed point ‘cloud’ regarding the actually perceived size (perceived visual significance) of each visible landscape element and its contribution or percentage (landscape composition) within the field of view. The most important part of calculations refers to the Perceived Visual Significance Index (PerVSI). PVSI calculation can be subdivided to the calculations of (A) Perceived Adjusted Vertical (Viewing) Angle (PAVA) and (B) Perceived Adjusted Horizontal (Viewing) Angle (PAHA).

• For the calculation of PAVA, several individual physical quantities were calculated:
  • The vertical (positive/negative) viewing angle (VerAng) at which each viewpoint of the route is connected to each viewshed point of the cloud (sharing a common FID) was calculated as a function of the viewpoint elevation (VPElev), viewshed elevation (ELEV), and horizontal distance among viewpoint and viewshed points (NEAR_DIST), which is defined as follows:

    $$VerAng={\tan }^{-1}\left({\displaystyle \frac{VPElev-ELEV }{NEAR\_DIST}}\right);$$

    (3)
  • The actual (i.e., not just horizontal) distance (ActDist) from each viewpoint of the route to each viewshed point of the cloud (sharing a common FID) was also calculated as a function of the viewpoint elevation (VPElev), viewshed elevation (ELEV), and horizontal distance among viewpoint and viewshed points (NEAR_DIST), which is defined in the following:

    $$ActDist=\sqrt{{\left(VPElev-ELEV\right)}^2+{NEAR\_DIST}^2};$$

    (4)
  • The relative vertical viewing angle (RelVAng) that occurs between (i) the vertical angle at which each viewpoint of the route is connected to each viewshed point of the cloud (VerAng) and (ii) the slope of the surface of the viewshed cells (SLOPE), which is calculated as follows:

    $$RelVAng=\left|\left|VerAng\right|-SLOPE\right|;$$

    (5)
  • The actual height (ActHei)—due to the relative viewing vertical angle—of each viewshed cell of the cloud relative to each viewpoint of the route (sharing a common FID), which is herein computed as follows:

    $$ActHei=\sin \left(RelVAng\ast cell size\right);$$

    (6)
  • The perceived adjusted vertical viewing angle (PAVA) as a function of ActHei (Equation (6)) and ActDist (Equation (4)), which is based on the following formula:

    $$PAVA={2\ast \tan }^{-1}\left({\displaystyle \frac{ActHei}{2\ast ActDist}}\right);$$

    (7)
• On the other hand, for the calculation of PAHA, the following calculations took place:
  • The relative horizontal viewing angle (RelHorAng) that occurs between the following: (i) the horizontal angle (azimuth) at which each viewpoint of the route is connected to each viewshed point of the cloud (NEAR_ANGLE) and (ii) the aspect of the surface of the viewshed cells calculated under certain conditions in two stages (a, b):

    (a)

    In order to delimit the relative horizontal viewing angles within the [0–180] range, the following condition was enforced, obtaining the following two arguments (NEAR ANGLE and ASPECT): if ASPECT = = −1 (flat surface): $ASPECT=90$°,

    else if 0 <= ASPECT <= 180°: $ASPECT=\left|\left|NEAR\_ANGLE\right|-ASPECT\right|$,

    else (if ASPECT > 180°): $ASPECT=\left|\left|NEAR\_ANGLE\right|-\left|ASPECT-360°\right|\right|$;

    and then an intermediate variable, RelHorA, was calculated as a function of the two arguments:

    $$RelHorA=f(NEAR\_ANGLE, ASPECT);$$

    (8)

    (b)

    In order to delimit the relative horizontal viewing angles within the [0–90] range, the following condition was enforced:

    if RelHorA <= 90°: RelHorA remains unchanged;

    else (if RelHorA > 90°): $RelHorA=RelHorA-90$°;

    and, so:

    $$RelHorAng=RelHorA;$$

    (9)

  • The actual width (ActWid)—due to the relative viewing horizontal angle—of each viewshed cell of the cloud relative to each viewpoint of the route (sharing a common FID) was computed using the following equation:

    $$ActWid=\sin \left(RelHorAng\ast cell size\right);$$

    (10)
  • The perceived adjusted horizontal viewing angle (PAHA) as a function of ActWid (Equation (10)) and ActDist (Equation (4)) was calculated based on the following formula:

    $$PAHA={2\ast \tan }^{-1}\left({\displaystyle \frac{ActWid}{2\ast ActDist}}\right);$$

    (11)
  PAVA and PAHA indices measure the perceived angular sizes of visible/viewshed cells in the two (vertical and horizontal) dimensions, i.e., perceived angular height and width, while their angle values range in the interval [0, 180)°. The Perceived Visual Significance Index (PerVSI) is a compound index that measures the two-dimensional (solid) angle (measured in square degrees) and occurs as the product of PAVA and PAHA, namely,

  $$PerVSI=PAVA\ast PAHA$$

  (12)

The fifth sub-model received as inputs all the viewshed points in each ‘cloud’. Each one of these viewshed points has been ascribed several attributes from all the aforementioned geoprocesses, such as the previously calculated PerVSI, the LU/LC, the FID of the initial viewpoint (along the route), etc. Now, by spatially and thematically aggregating (dissolving) the PerVSI based on the same LU/LC type (field: “class_2018”), the total (sum) PerVSI per LU/LC type was calculated inside each ‘cloud’ of viewshed points and stored in a new field (“SUM_PerVSI”). By summing all the SUM_PerVSIs for all different LU/LC types existing in each ‘cloud’ (summing the PerVSI for each viewshed point is equivalent), a new field was created (“SUM_SUM_PerVSI”), defining the total sum of the two-dimensional angles of visible terrain within the dFoV. LandComp, which represents the relative contribution (percentage) of each LU/LC type or landscape element, i.e., the landscape composition normalized in the [0, 100] within the within the dFoV, was calculated as a division:

$$LandComp=\left({\displaystyle \frac{SUM\_PerVSI}{SUM\_SUM\_PerVSI}}\right)\ast 100$$

(13)

The dominant LU/LC or landscape element (“new_class_2018”) and its respective relative contribution (percentage) (“MaxLand”) were computed by further aggregating (dissolving) the “LandComp” values; by selecting the features/rows having the maximum “LandComp” value, for each viewshed points ‘cloud’, only one point–row was retained, creating a set of multipart point features, while the number of this set equaled the number of the initial viewpoints (along the route). After this aggregation, the multipart point features were joined to their associated initial viewpoints. As an effect, all the attributes of the multipart point features (e.g., “MaxLand”, “LandComp”, etc.) were transferred to the initial viewpoints. These viewpoints–features were merged into a single-part point feature, and so each row represented the viewpoints—having been attributed a variety of fields. In this last stage, all the attributes stored in the viewpoints were transferred to the segments of the highway route via spatial join processing. Therefore, in these polyline features (viewlines), multiple lines–rows were occupying the same highway segment, since more than one point was spatially joined to each segment. These lines–rows were aggregated (i.e., dissolved) into a unique ID under the following condition: if the “new_class_2018” (LU/LC) was the same, e.g.,: “Complex and mixed cultivation patterns”, with “MaxLand” values, e.g., 54.67 and 62.85, respectively, then the LU/LC was retained (“Complex and mixed cultivation patterns”), and the “MaxLand” mean value was calculated by averaging the two values (58.76); if the “new_class_2018” (LU/LC) was different, e.g.,: “Complex and mixed cultivation patterns” and “Other roads and associated land”, with “MaxLand” values, e.g., 44.31 and 52.25, then the LU/LC corresponding to the larger value would be retained (“Other roads and associated land”), with the respective “MaxLand” value (52.25). So, in the final output, only the initial segments of the initial segments are retained, while two new fields are created, storing (i) the description of the dominant LU/LC (“final_class_2018”) and the quantitative values of the dominant LU/LC (“DomLand”). One last field (“LandChar”) was created suitable for storing an information fusion mixture of two fields, namely, the Urban Atlas LU/LC coding, “code_2018” and the “DomLand”, by concatenating string values; thence, “LandChar” contains a verbal description of both qualitative and quantitative information about the dominant landscape character corresponding to each highway route segment (e.g., 24000: 58.76%; 12220: 52.25%, etc.).

3.4.3. Tool Implementation: Employing ArcPy Toolboxes, Toolsets, and Tools

The whole methodology was implemented via a software tool particularly developed for the needs of this research work. More specifically, a Python (version 2.7.16) script-based geoprocessing tool, named “RouteLAND”, was developed and executed within the ArcMap 10.8, ESRI^®, by making use of the ArcPy Python site package, provided by ESRI^®, obtaining as inputs freely available, open geodata (described in Section 3.2). The input data were imported into the script, and then the path to the output data was set. The geoprocesses and calculations were implemented by employing several geoprocessing ArcPy toolboxes, toolsets, and tools. In

Table 4. Employed (ArcPy) tools (column 4), classified per toolset (column 3) and toolbox (column 2), along with the datasets serving as inputs within each toolbox (column 1). The last row is differentiated in that it constitutes a class-function, enabling iterative selection (SearchCursor) of feature records.

Input Layer	Toolbox/Data Access Module—ArcPy Function	Toolset	Tool/Class-Function
1. Initial highway route (OSM roads); 2. Landscape elements (UrbanAtlas LU/LC); 3. Intermediate highway route	Analysis	Extract	Clip; Select
		Overlay	Identity; SpatialJoin
		Proximity	GenerateNearTable; Buffer
		Statistics	Statistics (Summary Statistics)
1. Initial highway route (OSM roads); 2. Landscape elements (UrbanAtlas LU/LC); 3. Terrain elevation; 4. Terrain slope; 5. Terrain aspect (EU-DEM); 6. Intermediate highway route	Management	Fields	AddField; CalculateField; DeleteField
		General	Merge
		Generalization	Dissolve
		Features	CopyFeatures; FeatureVerticesToPoints
		Joins	JoinField
		Layers and Table Views	MakeFeatureLayer; SelectLayerByAttribute
		Raster Processing	Clip
		Sampling	GeneratePointsAlongLines
1. Initial highway route (OSM roads); 2. Terrain elevation; 3. Terrain slope; 4. Terrain aspect (EU-DEM)	Spatial Analyst	Surface	Aspect; Slope; Viewshed2
		Extraction	ExtractMultiValuesToPoints
1. Initial highway route (OSM roads); 2. Terrain elevation; 3. Terrain slope; 4. Terrain aspect (EU-DEM)	Conversion	From Raster	RasterToPoint
1. Initial highway route	Spatial Statistics	Measuring Geographic Distributions	DirectionalMean
1. Initial highway route	Classes/Cursors	-	SearchCursor

, the employed tools are summarized and classified based on their corresponding toolsets and toolboxes, while the layers involved are also summarized based on the toolboxes, harnessing them as input data; moreover, the last row of the table constitutes a somewhat separate one, providing information about the “SearchCursor” element—a class enabling the iterative data access (selection) of each viewpoint separately for further proceeding to the computations per viewpoint. After leveraging all the displayed tools within the overall geoprocessing sequence, the final output was a polyline feature layer, whereby each segment–record was characterized by the dominant landscape element within the highway’s dFoV.

3.4.4. Web Map Design

The final output of the presented case study was used to design an interactive web map that visualizes the prevailing landscape elements, within the dFoV, for each one of the predefined segments of the highway routes. More specifically, the web map portrays the segments of the highway routes, categorized by the dominant land uses/land covers, on the cartographic backgrounds of ESRI Gray (dark) and OpenStreetMap (OSM). The web map serves as an exploratory tool that allows users to reach the attributes of LU/LC, the DomLand values, and the LandChar descriptions for each route segment. Additionally, the users can filter the depicted routes’ segments based on both DomLand values and LU/LC attributes, as well as measure path distances and areas on the map. Moreover, we have restricted the zoom levels of the web map from 13z to 16z to help potential users to efficiently explore the selected highway routes. The development of the interactive web map was based on the utilization of open-source geospatial software. In more detail, highway route symbolization was implemented in QGIS software (Version 3.34.12 LTR), utilizing the standard UrbanAtlas legend for the different LU/LCs (https://land.copernicus.eu/en/map-viewer?dataset=70903c20fc2a4a90ad200bc95a7557d4\ accessed on 10 December 2024), while different cartographic backgrounds were imported from the corresponding web map services using the QuickMapServices QGIS plugin (https://plugins.qgis.org/plugins/quick_map_services/, accessed on 10 December 2024). Finally, for the production of the interactive web map, the qgis2web QGIS plugin (https://plugins.qgis.org/plugins/qgis2web/, accessed on 10 December 2024) and LeafLet JavaScript Library (https://leafletjs.com/, accessed on 10 December 2024) were utilized.

4. Results

4.1. Presentation

The results of this article are both the Python-based geoprocessing tool (“RouteLAND.py”) and the web map. The source code of the developed tool has been freely distributed to the scientific community under the third version of the GNU General Public License (GPL v3) on the GitHub platform (https://github.com/), as presented below, in the Supplementary Materials’ section. As regards the interactive web map, it is freely accessible on the web via the webpage https://routeland.geo.uniwa.gr/ (created on 29 November 2024). An indicative snapshot of the web map is illustrated in Figure 4. In the illustrated snapshot, a popup emerges, showing the produced attributes’ information (stored in the respective fields)—i.e., LU/LC, DomLand, and LandChar—as the user interactively explores and hovers along different segments of the highway route (Eleochori–Lavrio).

In

Table 5. UrbanAtlas LU/LC types expressed as landscape composition percentage (%) within the route’s 6 km zone—exocentric perspective (column 2); route’s composition percentage (%) mainly viewing a specific LU/LC type—egocentric perspective (column 3); mean/max DomLand values in each group of route’s segments’ mainly viewing a specific LU/LC type—egocentric perspective (column 4).

UrbanAtlas LU/LC Type	Area of LU/LC Within the Route’s 6 km Zone (%)	Number of Segments with Dominant LU/LC in Route’s dFoV (%)	Mean/Max DomLand for Each Segment (%)
Arable land (annual crops)	2.10	1.03	46.37/46.37
Complex and mixed cultivation patterns	6.61	9.23	61.75/88.39
Construction sites	0.01	-	-
Continuous urban fabric (S.L.: >80%)	0.29	-	-
Discontinuous dense urban fabric (S.L.: 50–80%)	2.02	0.51	62.08/62.08
Discontinuous low-density urban fabric (S.L.: 10–30%)	0.88	-	-
Discontinuous medium density urban fabric (S.L.: 30–50%)	2.35	-	-
Discontinuous very-low-density urban fabric (S.L.: <10%)	0.04	-	-
Forests	1.45	-	-
Green urban areas	0.02	-	-
Herbaceous vegetation associations (natural grassland, moors...)	34.92	47.18	55.13/86.21
Industrial, commercial, public, military, and private units	1.47	1.54	78.67/80.09
Isolated structures	1.23	-	-
Land without current use	0.11	-	-
Mineral extraction and dump sites	0.09	-	-
Open spaces with little or no vegetation (beaches, dunes, bare rocks, glaciers)	0.01	-	-
Other roads and associated land	2.74	34.36	66.12/93.66
Pastures	3.22	4.10	41.90/49.15
Permanent crops (vineyards, fruit trees, olive groves)	1.35	-	-
Port areas	0.67	-	-
Sea	35.84	1.03	41.45/41.45
Port areas	0.67	-	-
Water	2.38	1.03	34.24/37.30
Total	100	100	-

, we further elaborate on the derived attributes’ information. More precisely, the second column of the table presents the percentage area (%) covered by each LU/LC within the route’s 6 km buffer zone—i.e., study area—while the third column displays the percentage of segments (%) dominated by the vista of each LU/LC along the highway route. Thence, the second column provides the landscape composition from an exocentric perspective, whereas the third column provides the route’s dominant landscape element (LU/LC) composition from the egocentric perspective of an observer moving on a highway, implementing computations utilizing the RouteLAND tool. There are some striking cases where the percentages of these two columns/perspectives differ dramatically. For instance, while the study area is exocentrically dominated by the “Sea” LU/LC (35.84%), only in 1.03% of the totality of the route does the “Sea” LU/LC dominate moving observers’ vista (i.e., from an egocentric perspective). Another very indicative case that refers to the “Other roads and associated land” LU/LC is the following: while within the study area, this LU/LC can be hardly found (2.74%), it constitutes one of the most prevailing segments (34.36%) of the route in terms of the LU/LC dominating the vistas of moving observers. In total, out of the 23 UrbanAtlas LU/LC types existing in the study area, only 9 LU/LC types have been computed to constitute dominant LU/LC types viewed from the highway route.

In the fourth column of the table, the mean and maximum (max) values of the “DomLand” are summarized for the nine group segments dominated by each separate LU/LC type. By comparing the third and the fourth columns, it becomes clear that for some LU/LC types, such as “Complex and mixed cultivation patterns”, their maximum “DomLand” values are very high (88.39%) but do not constitute segments that significantly contribute as LU/LC types dominating the moving observers’ vistas in the totality of the route (9.23%). On the other hand, the “Herbaceous vegetation associations” LU/LC type is found to prevail along the route (47.18%), yet it does not present equally high max “DomLand” values (86.21%).

The previous deviations show the great magnitude of differences occurring between (i) the two perspectives and (ii) the spatial and arithmetic occurrences of the landscape composition.

The first deviation occurs because the landscape elements existing in the study area (exocentric perspective) may be present in very different percentages or may be absent within the dFoV of the moving viewers (exocentric perspective)—as computed with RouteLAND. Computing the number/percentage at which one LU/LC type prevails within the hFoV within a highway route is based on the calculation of the sum of each viewed cell that represents the particular LU/LC type, which is expressed as an angular size/diameter visually perceived by an observer. What is more, when dealing with the dFOV of a moving observer, the horizontal hFoV varies with the observer’s motion speed along a highway. As a consequence, the percentage of highway segments principally viewing the particular LU/LC type may be very different compared to the percentage of this particular LU/LC type within the highway’s 6 km buffer zone—especially in areas where the landscape element diversity is relatively high (e.g., peri-urban areas).

The second deviation arises because a large percentage of (segments of) a highway may be mainly viewing a specific LU/LC type, yet the magnitude of the percentage within the dFoV may be not very high (e.g., 55%)—albeit higher than any LU/LC type within the dFoV. This means that a specific LU/LC type (e.g., “Herbaceous vegetation associations”) may be the most frequent type in spatial terms (traced in the majority of segments along the highway (e.g., 90%)), while at the same time (compared to other segments of the highway where their dFoVs are dominated by other LU/LC types) the mean and max values of this specific LU/LC type—along the segments where it is dominantly viewed—may be relatively low.

4.2. Evaluation

The results produced by our modeling approach cannot be evaluated in a direct fashion; hence, the evaluation can only be based on an indirect method. For this reason, we generated (3D) perspective geovisualizations at representative viewpoints, imitating the egocentric perspective of the visual experience. By doing so, we were able to visually compare the mean DomLand values (produced by the RouteLAND’s main export) with the dominant LU/LC landscape elements (produced by the perspective geovisualizations) at specific points/segments of the highway route.

Specifically, we utilized two indicative instances from the RouteLAND’s exocentric views; at these two segments, the route was characterized by DomLand values corresponding to “Herbaceous vegetation associations” (52.07%) and “Complex and mixed cultivation patterns” (44.515%), respectively (Figure 5). In the same Figure, we also exclusively display the LU/LC landscape elements dominating in the previous perspective (i.e., “Herbaceous vegetation associations” and “Complex and mixed cultivation patterns”), albeit this time, these elements have been visualized under the egocentric perspective—viewed from the same segments of the route (as in the exocentric perspective). The horizontal viewing angle for both points/segments is 100°; for the point/segment at the top position of the figure, the mean viewing direction (azimuth) is approximately 189°, whereas for the point/segment at the bottom position, it is approximately 310°.

From the visual inspection of these two alternative geovisualizations, the main output produced by the developed method/tool appears to significantly correspond to the results produced by the egocentrically based perspective. More precisely, according to the top-left geovisualization results (utilizing RouteLAND), the specific route segment’s vista is mainly characterized as (i.e., 52.07%) “Herbaceous vegetation associations”; this characterization is visually validated by the top-right geovisualization, where the aforementioned LU/LC is indeed a dominant element of the visible landscape. In the same vein, the results depicted at the bottom-left geovisualization (utilizing RouteLAND) are also visually validated by the generated bottom-right geovisualization.

5. Discussion and Conclusions

5.1. Synopsis

The present article had as a main goal to provide a new, integrated method for characterizing highway routes based on landscape composition and visual significance (i.e. actually perceived sizes of landscape elements), along with a software tool (RouteLAND) through which the sequence of the required geocomputations and calculations were built and implemented. The application of the method through the tool produced specific results for a selected use case; these results were cartographically visualized in an interactive web map—also developed for the needs of this article. By summarizing the results of executing RouteLAND in a selected peri-urban area in Attica region, Greece, it was found that the most frequently occurring (within the route) prevailing landscape elements were “Herbaceous vegetation associations”, while the quantitatively most dominant landscape elements—albeit less frequently occurring—were “Other roads and associated land”. The web map generated provides an overall visualization of the results of the use case.

5.2. Assumptions and Limitations

The introduced method has been built upon previous attempts to quantify the landscape composition within the dFoV along highway routes. Toward this end, there have been some assumptions made referring either to the utilized datasets or to certain conceptualization issues. Some additional limitations are also associated with the developed software.

The UrbanAtlas LU/LC dataset—a categorized geodata layer—has been considered to appropriately represent the exocentric perspective of the visual landscape. It is apparent that using this dataset constitutes a simplification of the actual visual landscape (compared to, e.g., a high spatial and spectral resolution satellite image), yet this is a reasonable one, taking into account the purpose and scale of the produced maps. The elevation dataset used also constitutes a simplification of the actual morphological conditions: the 30 m EU-DEM is a medium-scaled approximation of the topographic relief compared to a very precise and detailed model of the actual surface morphology, including vegetation and buildings’ heights (e.g., fine-scaled DSM). However, for the requirements of this landscape-based application and provided that the main focus was on the development of the method and the tool, the usage of this elevation geospatial dataset is assumed to be adequate. All the previous assumptions are supported by the pertinent literature (e.g., [28,31,39]).

Additionally, the maximum viewing distance, which was set to 5 km, poses a concern; however, it is a reasonable assumption that elements positioned in the far background of the viewing field are considered to have a negligible contribution within the hFoV. Moreover, the calculation of the horizontal field of view according to the maximum allowed speed of each segment of the highway, namely, the ‘narrowing’ of the horizontal hFoV with the increase in the movement speed, is also based on assumptions, yet they are plausible and empirically derived ones; what is more, the hFoV applied from each viewpoint is associated with a more simplified (rectangular) version of the actual hFoV. Another point of concern is the spacing intervals at which the highway route was ‘sampled’; extracting viewpoints for every 50 m of the route and computing viewsheds from these viewpoints was considered an appropriate viewpoint interval for specifying the dFoV and executing all the associated computations. At this point, it should be noted that related research work also supports the abovementioned assumptions (e.g., [29,38,39,43,44,45,46,47]).

Regarding the software tool, the first released version of RouteLAND has two main limitations. First, although RouteLAND has been freely distributed to the scientific community, its functionality depends on the utilization of the commercial ArcPy Python site package provided by ESRI^®. Second, the developed tool was tested and executed in Python version 2.7.16, which is compatible with the ArcMap 10.8 (ESRI^®) environment in which the presented case study was implemented.

5.3. Future Research and Outlook

The present methodology executed through the RouteLAND tool has enabled the automatic characterization of any potential route according to the landscape element that dominates the dFoV. Thus, through this tool, we modeled the vistas apperceived from vehicles moving upon highway routes, hence managing to cartographically represent the actual composition of landscape elements that drivers and passengers are experiencing in their everyday transportation. Since highways are crucial transportation links involving large numbers of observers, this method/tool can be immediately embedded and utilized in cases of comparative landscape analyses in peri-urban regions, further feeding public decision support systems for landscape evaluation and sustainable spatial planning. This first endeavor can be further amended, ameliorated, and expanded to meet the needs and goals of relative applications and (scientific) fields.

As regards the software tool itself, there are some enhancements that can be made: In order to be more generic and less dependent on existing GIS software and commercial Python site packages, future versions of the tool could be based on the exclusive use of open-source libraries and/or modules. Future versions of the tool could be developed in Python version(s) 3.x, also taking advantage of the core improvements of the Python programming language. In addition, RouteLAND can be further extended by incorporating an automatic process for generating a web map that visualizes the final output and supports effective exploration of the produced results. Finally, the existing source code constitutes a solid groundwork toward developing a new standalone software or a QGIS plugin with an integrated graphical user interface (GUI).

Aside from the future work toward enhancing the tool’s usability and functionality, the applications of this integrated method can be further extended and expanded. In the realm of landscape studies and research, this method/tool can be adopted in relation to Landscape Character Assessment (LCA). As LCA is concerned primarily with attributing the landscape character—i.e., the distinct and consistent pattern of elements in the landscape—to an area [48] viewed from the exocentric perspective [24], this method/tool can enable a shift for LCA from the exocentric to the egocentric perspective—that is, computing the LCA along routes of great importance. A step further could be the evaluation of landscape character according to certain criteria, such as naturalness, complexity, disturbance (see, e.g., [49] for such criteria), or the evaluation and rating of a landscape’s visual quality/preferences along routes. Dramstad et al. [50] have long ago investigated the relationship between visual landscape preferences and map-based indicators (indices and metrics) of landscape composition and structure by encompassing viewshed analyses into their investigation. RouteLAND has the potential to accurately measure and calculate indicators of landscape composition and structure from the egocentric perspective and within the dFoV along routes.

The RouteLAND integrated method could also address issues related to visual impact assessment (VIA) and ‘feed’ alternative design scenarios toward the optimization of spatial/landscape planning. For instance, the visual impacts from intrusive surface mining activities (i.e., quarries) have already been quantified and mapped through a geospatial model (GEMMELIM) [29]. However, this model does not take into consideration the dFoV. RouteLAND could be adjusted to serve as a tool for the dynamic VIA of quarries or other technical works related to energy projects (e.g., wind farms, solar farms, etc.) along routes. From a somewhat different point of view, RouteLAND could be utilized as a tool for spatial planning supporting decisions about either the assessment of alternative routes or the design of new roads (routes) by creating outputs that ‘feed’ cost path/surface geoprocesses; the computation of the least cost path regarding the contribution of an intrusive energy project within the dFoV can support urban planners to qualify a certain route (which minimizes the visual nuisance) among a multitude of alternative ones. Even more fundamentally, the introduced method of this article could support existing legislation concerned with VIA, providing a scientifically sound basis for quantifying the impacts from existing or future technical projects (see also [51]). The other side of the coin of VIA is associated with the maximization of exposure to vistas of landscape elements that are either positively rated by people (e.g., lakes, forests, buildings of traditional architectures, etc.), or are of particular importance (e.g., monuments, mountain peaks, various landmarks, etc.). In this sense, RouteLAND can assess existing (land or sea) routes or prompt the design of new routes that maximize their vistas to such valuable landscape resources. As an effect, RouteLAND can support spatial planning decisions aiding in the amelioration of the touristic experience along routes and, more generally, in further promoting sustainable landscape design.

Finally, the introduction of this method/tool furnishes a new opportunity to fuse GISystems and GIScience with the fields of eye movement analysis and visual perception. This outlook has been introduced and empirically explored by Schirpke et al. [52] and revisited by Misthos et al. [24]. RouteLAND computes the visible landscape elements within the dFoV in both a quantitatively and a spatially explicit manner. Thence, the embedment of eye movement recordings and analyses appears to be a very promising ground for coupling map (2D) geospatial information with information derived from the visual perception of actual observers. This fusion becomes even more intriguing when it is synthesized in a dynamic sequence in cases of routes upon the ground and observers moving upon these routes.