With a rising emphasis on sustainability, stakeholder interest in preserving and enhancing the road environment is growing. According to Matijošaitienė and Stankevičė (2014) , road users are like the traditional consumer, looking for aesthetics as a way to enhance road usage utility. In Maslow’s hierarchy of needs, cognitive and aesthetic needs are necessary for societal growth (McLeod, 2007, 2018; Mozer, 1988) . As a result, the need to create high-quality roadside design components contributes to the achievement of this goal.

The aesthetics of roadways is the primary link to the environment since the urbanized global population primarily engages with natural landscapes visually through their movement along roads. Multiple variables have been used to define this connection, with Armstrong et al. (2013) advancing natural terrain features, human-enhanced topography, design of road signs and road-safety measures. Therefore, planners must seek to balance the natural world and human-built forms to produce aesthetic roads. These would have to consider the demands of highway users, the roadway owners, and the preservation or improvement of the natural and cultural surroundings. Since creating excellent and pleasant road designs is vital for road planning professionals, there is a need to identify clear and concise definitions of attractive roads from the perspectives of road development experts and users.

Road aesthetics encompasses not only qualities of beauty that relate to its form, the proportions of spaces, lines, and mass, but also the preferences of road users (Blumentrath & Tveit, 2014; Qin et al., 2023) . Blumentrath and Tveit (2014) identified three aspects of road design that are applied in defining the visual attributes of a road: the visual characteristics of the road as an independent structure, the visual characteristics of the surroundings, and the traveller’s movement. In past visual assessment studies (Chon & Shafer, 2009; Clay & Smidt, 2004; Mok et al., 2006) , different characteristics in various social and cultural contexts have been identified and used to incorporate aesthetic dimensions into the planning and development of infrastructure. The studies apply social science methods to construct descriptive models representing people’s responses to the environment’s visual aspects. Therefore, this paper synthesizes existing empirical studies on the visual aspects of the road through content analysis and text mining. The intention is to identify the most commonly used dimensions for evaluating road aesthetics in different social and cultural contexts.

The nine articles that met the inclusion criteria were reviewed. A synthesis of the articles through manual systematic coding revealed aesthetic dimensions used in assessing roads, as depicted in Table 1 . These articles highlighted studies undertaken in diverse regions and settings, whose main objectives were to evaluate the aesthetics of roads and their surrounding environments. For instance, Blumentrath (2016) identified characteristics of importance for reviewing the aesthetic quality of roads from expert and layperson perspectives.

Using ATLAS.ti v8, we collated the list of identified descriptors of road aesthetics and used a single term to record variations of terms that were deemed similar. For instance, coherency and coherent design were recorded as coherence;

harmonious was recorded as harmony, natural as naturalness, skittish as skittishness and sophisticated as sophistication. An analysis using OrangeTM, an open-source data mining software, produced the term frequencies as depicted in Figure 3 and Table 2 ; naturalness was the most used term in the evaluation of road aesthetics with six occurrences, followed by coherence and imageability, which appeared in five and four instances, respectively.

Furthermore, we assessed descriptions of terms used in the various studies. Some terms used in the various studies had similar conceptual meanings; hence, we undertook word association and categorization and reviewed term descriptions provided in the selected articles with the aid of Roget’s (2022) thesaurus. Terms with similar descriptions and meanings were recorded using a surrogate term, as shown in Table 3 . This assessment revealed meanings associated with the terms and aided in creating a new corpus used to run the term frequency operation and produce a term-by-document matrix ( Table 4 ).

An inspection of the term-by-document matrix revealed that more than 50%

Table 3. Assessment of aesthetic dimensions used in the articles.

Note: 1 = existence of the term in an article; 0 = term does not exist in the article.

of the studies reviewing road aesthetics discussed coherence, imageability, naturalness and variety. Legibility, maintenance, and visibility were also common terms, with 30% - 44% of the articles considering them as descriptors critical in reviewing aesthetics ( Figure 4 ).

The percentage occurrence of terms per document and the concept of optimal weighting using Euclidean regularisation were used to process and identify terms most relevant in evaluating road aesthetics. As a result, we generated the term frequencies, inverse document weighting and the TF-IDF (term frequency-inverse document frequency). These operations efficiently determined objects of similarity and reduced the dimensionality of the identified road aesthetics descriptors. First, the textual data obtained from evaluating meanings created a term frequency matrix ( Table 5 ). Each term within an article was assigned a weight ( $t{f}_i$) determined by Equation (1).

$\text{Term frequeney}\left(t{f}_i\right)=\frac{\text{Number of occurrences of a term}}{\text{Total relevant terms reviewed in the articles}}$ (1)

In this regard, the more relevant terms within an article had higher weighting. For example, Figure 5 indicates that eight articles accessed roads for coherence. Therefore, this term had the highest cumulative term frequency weighting. Legibility had the second-highest cumulative weighting and was assessed in four articles. Imageability ranked third, having been reviewed in seven articles, while naturalness and variety had similar weighting. However, six articles reviewed naturalness, compared to five that reviewed variety.

Each term’s Inverse Document Frequency (IDF) was calculated using Equation (2), and their weights were noted ( Table 6 ). IDF measured the probability of relevance of the terms related to the corpus, where a higher weighting indicated that the term was occurring less frequently, while a lower IDF weight was obtained for frequently occurring terms in the corpus.

${\text{IDF}}_{i,j}=\log \left(\frac{N}{d{f}_i}\right)$ (2)

Where,

i = term used in an article to evaluate road aesthetics

j = article discussing the evaluation of road aesthetics

$d{f}_i$ = Number of articles containing ‘i’

N = Total number of articles

Finally, the TF-IDF formula, as shown in Equation (3), was used to determine the TF-IDF for each term per article and to develop the TF-IDF matrix ( Table 7 and Figure 6 ).

${\text{TF-IDF}}_{i,j}={\text{TF}}_{i,j}\times \log \left(\frac{N}{d{f}_i}\right)$ (3)

The TF-IDF matrix revealed that legibility had the highest cumulative weight and was evaluated in four articles. It was followed by imageability and maintenance, which were evaluated in seven and three articles, respectively. The study determined the final list of relevant terms by considering the three highest-ranking terms according to the percentage occurrence of terms per article, term frequencies, inverse document frequency weighting, and the TF-IDF (term frequency-inverse document frequency) outputs ( Table 8 ).

Based on the four ranking criteria, the study identified coherence, imageability,

legibility, naturalness, and maintenance as the most relevant descriptive terms in evaluating road aesthetics.

This synthesis sought to identify the most relevant terms or descriptors useful in clarifying how to achieve aesthetics for transportation planners in Kenya. As a result, the synthesis carried out in this study provided a meaningful list of relevant and appropriate aesthetic dimensions for evaluating road environments and outlined them as follows: coherence, which is the ability to see and comprehend the pattern within a scene related to scale and proportion (Tveit et al., 2006) , imageability; the expression connoting the ability of a scene to create a strong visual image in an observer (Clemente & Ewing, 2005) , naturalness; described by the type of nature content and measured in road environments by the appearance of natural elements such as vegetated roadside areas (Blumentrath & Tveit, 2014) , legibility, which refers to the ability of a road to be understood by the observer, who then makes sense of it to allow for safe functioning, and finally maintenance, represented by the tidiness of the road environment, the quality of materials used, and excellent workmanship (Blumentrath et al., 2015; Blumentrath & Tveit, 2014) .

The assessment of meanings revealed a similarity in some dimensions. For instance, coherence, described as the quality of design and layout that creates a clear and understandable urban environment for its inhabitants (Lynch, 1960) , was also described as the degree of harmony, consistency, and logical connection between different elements and aspects of an environment. It encompasses the urban environment’s overall organization, layout, and design, aiming to create a sense of unity and cohesion. Hence, coherence relates to dimensions such as order, harmony, uniformity, integration, proportion and good fit.

Imageability, which is the quality of an environment’s elements that trigger lucid images in an observer and the ease with which people can form a mental image of a place (Appleyard et al., 1964; Lynch, 1960; Nasar, 1998) , related to the descriptions provided for identity, sense of place, character, distinctness, uniqueness, distinguishability and memorability. In road planning, imageability refers to the design and layout of road networks that enhance the visual appeal and legibility of a city or urban area. It involves creating streetscapes and transportation systems that are aesthetically pleasing, easily navigable, and conducive to human-scale activities (Gehl, 2013; Van der Hoeven et al., 2008) .

Legibility is defined as the clarity and comprehensibility of an urban environment’s spatial layout and organization. It focuses on how easily people can navigate and understand a city’s physical structure, landmarks, and transportation networks (Lynch, 1960) . In road planning, legibility refers to the clarity and ease with which road users can understand and navigate a road network. It encompasses clear signage, well-designed intersections, logical route numbering, and easily comprehensible road markings. For instance, Pasetto & Barbati (2012) showed that improved road markings led to a better perception of the road and a general reduction in average speed. Legibility in road planning aims to provide a road network that minimizes confusion and enhances safety for all users (Appleyard et al., 1964; Wilson, 2019) . This dimension relates to simplicity, aesthetics of flow, consistency, functionality, orientation, rhythm and repetition.

Naturalness related to greenery and intactness was described as an environment’s closeness to a perceived natural state (Kaplan & Kaplan, 1989; Orians, 2022; Tveit et al., 2006) . It is an important aesthetic dimension in road planning that integrates natural elements, such as trees, vegetation, and open spaces, which can enhance the visual appeal of roadways and create a more pleasant and harmonious environment for drivers and pedestrians. Finally, maintenance, linked to stewardship and upkeep, is the ongoing care and upkeep of urban spaces to ensure their visual appeal and functionality. It involves cleaning, repairing, and regularly maintaining infrastructure, public spaces, and landscapes (Gehl, 2013; Jacobs, 1961; Lynch, 1984) .

The results of the TF-IDF shown in Figure 6 suggest that despite coherence being the most used descriptive term in studies exploring road aesthetics, legibility had a higher weighting, suggesting a link between the legibility of a road and the need for self-explaining roads, as discussed by Theeuwes (2021) and Chowdhury (2014) . The link between self-explaining roads (SERs) and road legibility lies in the shared goal of promoting efficient and safe navigation within road environments. Road legibility contributes to the nature of SERs through the provision of visual cues and navigational aids that help road users understand road layouts and anticipate reactions. Moreover, SERs result in high road legibility as this principle prioritizes simplicity, clarity and ease of understanding of road scenes (Chaudhary, 2020; Cunningham, 2018) . Imageability and maintenance are then ranked second and third regarding the TF-IDF weighting. However, this result should be interpreted cautiously, as more research into how different users perceive these dimensions is needed to assess their importance in the Kenyan context better.

Moreover, it is important to consider conflicts and synergies between aesthetics and other sustainability indicators in road development to create a balanced and effective infrastructure. One major potential conflict is the impact on the environment, particularly in terms of the naturalness of the road. Greening requires significant water and maintenance resources (Liu & Jensen, 2018) . However, integrating aesthetic values can enhance environmental benefits. For example, using vegetated swales and native plants that support local biodiversity can reduce maintenance needs and costs (O’Sullivan et al., 2017; Phillips et al., 2020; Säumel et al., 2016) . In addition, we identify notable synergies in community engagement and user experiences. Understanding road coherence, imageability, and legibility from the user’s perspective enhances user experiences and increases safety. Therefore, it is critical to adopt an integrated design approach that considers all sustainability indicators (Suprayoga et al., 2020; Yang et al., 2020) and implements adaptive management practices for ongoing monitoring and adjustment of road design to address emerging conflicts, such as the recent flooding experience in Kenya.

While the provision of aesthetics is legally recognized as one of the elements concerning the environment in Kenya, there is no clear demonstration of criteria that planners can use to assess it. For road planning entities, existing criteria are limited to strategies involving tree planting and grassing road reserves. This approach impedes sustainability in providing aesthetics to the users of roads. Road aesthetics is a product that road consumers should receive, hence the need for clarity in its achievement. Clarifying what constitutes road aesthetics provides planners with a quantifiable way of assessing it, which, in turn, improves the health and safety of the population and increases the sustainability of planning transport infrastructure. By reviewing the terms used in measuring road aesthetics from previous empirical studies, this paper establishes that road aesthetics is the character a road has in terms of its spaces’ ability to be seen and understood by the users.

The study integrated multiple techniques to synthesize and evaluate the study data. It leveraged the statistical properties of term frequencies (TF) and inverse document frequencies (IDF) and a qualitative assessment of terms to comprehensively determine the most frequently used and relevant aesthetic dimensions for evaluating road environments. Through this, the study systematically identified critical descriptors of aesthetics for future analyses of the beauty of a road environment, such as coherence, imageability, naturalness, legibility, and maintenance. Based on these findings, this study proposes recommendations that will ensure that road projects are functional, sustainable, and visually appealing, thus contributing to the overall wellbeing of communities and road users. First, planners should seek to develop further guidelines that establish consistent design themes, ensuring visual coherence across road elements such as barriers, lighting, and greening.

Second, there is a need to identify unique landmarks and incorporate them to enhance distinctive visual elements and make roads more memorable and easily recognizable. Third, road designs should ideally follow the natural contours while integrating natural landscapes and features. Additionally, native plants should be used for landscaping to promote biodiversity and reduce maintenance costs. Fourth, road agencies should prioritize improving navigational clarity by ensuring that road layouts, signage, and markings are intuitive and easy to understand for all road users. Fifth, rigorous and regular maintenance schedules are crucial, addressing both functional and aesthetic aspects.

Moreover, relevant agencies should allocate adequate resources for cleaning, repair, upkeep, signage, and landscaping. Sixth, it is important to incorporate sustainable practices, ensuring that aesthetic improvements contribute to the reduction of carbon footprints and enhance biodiversity. Finally, comprehensive policies must be developed that prioritize aesthetics in road development. These policies should mandate the inclusion of aesthetic considerations in all stages of road planning, design, and construction. These policies can be achieved by incorporating and outlining aesthetic requirements in regulatory frameworks and funding criteria for road projects and providing training and resources for planners and road designers to integrate aesthetics into their work effectively.

However, the study notes that the meta-synthesis did not indicate the most crucial dimension or the degree of combination that results in the most aesthetic road; it identifies the most relevant in evaluating road aesthetics. As such, future inquiry into such concerns is necessary. Nevertheless, the findings provide a starting point for reviewing the aesthetic dimensions and determining the dimensions of interest for planning road aesthetics in Kenya. Moreover, a limitation of this study is the low number of articles used in the corpus generation. The study only included articles published since 2010. Future studies should consider reviewing this criterion and determining whether including articles published prior to 2010 would yield the same results.