The overall concept of walkability has, in recent years, evolved into a multidimensional concept and describes the interaction between pedestrians, public space and its function for people to perform social practices (Blecic et al., 2020).

Blecic et al. (2020) trace ideas of walkability back to the mid-twentieth century and detects close links to debates about quality of life, social justice and the right to the city. In the last two decades, the theoretical discussion has been extended by the operationalisation of the concept of walkability (Blecic et al., 2020). Based on the need for consistent and efficient methods to collect walking environment qualities, a lot of research concentrated on operationalisation and developing decision-making tools for policymakers (Clifton et al., 2007; D’Orso and Migliore, 2020). As a concept relating to the friendliness of the environment for walking (Litman, 2003; Manzolli et al., 2021), walkability is associated with encouraging a shift from motorised transport to active mobility by creating pedestrian-friendly spaces where activities are accessible by short walking distances (Bödeker et al., 2018; D’Orso and Migliore, 2020; Papa and Bertolini, 2015; Su et al., 2017; Van Cauwenberg et al., 2016).

As one important indicator of walkability, walking accessibility is often used in urban planning and transportation research. Since Hansen introduced accessibility as the “opportunities for interaction” (Hansen, 1959, p73) focusing on the transport and land-use system, recent definitions have concentrated on the individual, for example, the ease with which people can access “[…] desired goods, services, activities and destinations” (Litman, 2003, p28). Walking accessibility measures, like the Walk Score® or Neighbourhood Destination Accessibility Index, used to determine the shortest distance (or effort, cost, travel time) to a facility, are mostly based on the shortest-path algorithm on a graph. The Walk Score® was originally developed to promote the real estate market in the USA. It targets homebuyers in more middle-class neighbourhoods. In its original form, the Walk Score® takes only limited account of urban planning and policy aspects. More detailed explanations of Walk Score® methodology follow in section 3.1.

Accessibility measures are typically static, seem to imply homogeneous accessibility among people living in one place and say nothing about how that accessibility is experienced by people living in places defined as (in-)accessible. The term ‘perceived accessibility’ is intended to counter this criticism and to include the perceptions of individuals into accessibility measurement. Perceived accessibility describes “how easy it is to live a satisfactory life using the transport system” (Lättman et al., 2016, p257). Pot et al. (2021, p1) define it in a recent study as “the perceived potential to participate in spatially dispersed opportunities”. Recent studies empirically capture perceptions of accessibility surveys with the intention of integrating perceptions into accessibility analyses. This has resulted in a number of measurement methods, like measuring the overall perceptions of accessibility (Curl, 2013), the Perceived Accessibility Scale (Lättman et al., 2016) and measuring perceived walking accessibility (van der Vlugt et al., 2019). Van der Vlugt et al. (2022) found that perceived walking accessibility has a significant direct positive impact on walking, even after accounting for objectively measured accessibility. This is backed up by the recent work of Ma and Cao (2019), Tuckel and Milczarski (2015), Lättman (2018) and Gebel et al. (2011). These findings have not yet been integrated into spatially measured accessibility.

In their recent work, Blecic et al. (2020) identified three strands of walkability studies. The first examines the relationship between the built environment and walking. The second concentrates on individuality and the perception of pedestrians. The third strand explores walkability from a human capabilities’ perspective. In the following, the focus is on the first two strands, but the third strand is also briefly outlined due to its importance.

In most studies, built environment features dominate when measuring walkability. However, in the last two decades, attempts have been made to combine several strands of walkability studies. In order to represent the influence of the environment on travel, Cervero and Kockelman (1997) identified the 3Ds: diversity, density and design. By developing a pedestrian strategy for London, Gardner et al. (1996) developed the 5Cs, which identified five qualities of the walking environment with the categories ‘connected’, ‘convenient’, ‘comfortable’, ‘convivial’ and ‘conspicuous’. This was followed by studies that included objective and subjective qualities in walkability. For example, Sarkar (1993) evaluated six criteria of a walking environment: safety, security, comfort and convenience, continuity, system coherence and attractiveness. Pikora et al. (2002), on the other hand, identified four categories that have an influence on walking: functionality, aesthetics, destinations and safety. Ewing (1999) also formulated ten influential factors for a walking friendly environment (medium-to-high densities, mix of land use, short-to-medium length blocks, transit routes every half-mile, two-to-four lane streets, continuous pavements wide enough for couples, safe crossings, appropriate buffering from traffic, street-oriented buildings and comfortable and safe places to wait). Ewing and Cervero (2010) expanded the 3D into 5D: diversity, density and design, destination to facilities and destination to transit. Galanis and Eliou (2011) identified five basic factors for a walkable urban environment: road safety, attractiveness, personal safety, accessibility and convenience broken down into road segment indicators, including the features of the street (pavement width, street furniture), corner indicators describing features of pavement corners (ramps) and crossing indicators (length, width). Furthermore, D’Alessandro et al. (2016) identified twelve indicators and split them into three categories: namely, safety (crossing protection, protection from vehicles, road lighting), pleasantness (vehicular traffic, building context and green spaces) and practicability (obstacles, pavement surface, road slope). Blecic et al. (2020) identify four walkability indicator categories based on their literature review: efficiency and comfort, safety, security and certainty as well as pleasantness and attractiveness.

Recent walkability perception studies already described in 2.2.2 like van Dyck et al. (2013) and Suarez-Balcazar et al. (2020), used the Neighbourhood Environment Walkability Scale (NEWS), developed by Saelens et al. (2003). The NEWS contains 83 items, including walking time to 23 different facilities and also 17 satisfaction-related items. These items cover residential density, land-use mix access, land-use mix diversity, satisfaction, walking and cycling facilities, aesthetics, traffic safety and crime safety. Due to the immense length, shortened and target-group specific NEWS versions were developed. The Neighbourhood Environment Walkability Scale (abbreviated as NEWS-A) no longer includes walking time and excludes residential related questions. It now consists of only 37 items, used in many studies (Bartshe et al., 2018; Berry et al., 2017; Brown and Jensen, 2020; Consoli et al., 2020; Jensen et al., 2017; Lui and Wong, 2021; Perez et al., 2011). The NEWS-Y (for young people) is also based on the NEWS-A and was used, for example, by Hinckson et al. (2017) to investigate the perceived walkability of adolescents. Alongside NEWS, the Physical Activity Neighbourhood Environment Scale (PANES) measures perceived density, walking infrastructure, traffic safety and accessibility to amenities based on 17 statements (Sallis et al., 2010). Furthermore, the Leyden Walkability Instrument (LWI) measures perceived walkability on a scale basis (Bias et al., 2010; Leyden, 2003).

These studies often use multi-method study design to operationalise and collect data for these categories. In addition to objective characteristic data and open data, Google Street View data, expert inventory surveys, general surveys and in-depth interviews are used to collect data.

The literature review has shown that walking accessibility as a dimension of walkability is not sufficient to analyse the local provision of facilities and services for different target groups. Therefore, walking accessibility perception as well as the physical features and qualities of the walking route seem particularly relevant for walking and can determine whether a route is taken on foot or by another transport mode. It seems essential to adapt an existing accessibility analysis instrument with walkability indicators and perceptions in order to provide a more nuanced picture of the local provision of facilities and services for different population groups. Various studies responded to the critique of one-dimensional walkability assessments by measuring distances and the facilities to be reached (Blecic et al., 2020) and made efforts to combine different strands of walkability, such as the CANVAS (Bader et al., 2015), the IAAPE Measure of Walkability (Moura et al., 2017), the OS-WALK-EU tool (Fina et al., 2022) and the Walkability Explorer (Blecic et al., 2015). Although a large number of tools is available, there is still a lack of studies on a holistic view of walkability with regard to: firstly, target group-specific adaptation in terms of weighting and number of facilities to be considered; secondly, simultaneously considering multiple indicators of the physical characteristics of an environment and the quality of a walking route; and, thirdly, the integration of perceptions into existing tools. Most studies have compared different methods but show deficits in terms of spatial modelling and integration of all elements into one instrument or tool.

Based on the literature review and the research gap identified, we developed a conceptual framework that provides the basis for developing an adapted instrument, the Perceived Environment Walking Index (PEWI) (Figure 1, to be read from top to bottom). Adaption takes place in three steps. First, since we paradigmatically place our approach in strands one and two of walkability research (Section 2.2), we calculated the Walk Score® that represents objective walking accessibility (strand one) and adapted it to the specific target groups (strand two). To do so, we adjusted the weighting and variety of facilities included in the accessibility calculation (step 1). Second, we integrated 18 indicators which give a more nuanced description of the influence of the built environment on the walkability of different target groups, representing physical features (step 2a) and the quality of the walking route (strands one and two) (step 2b) broken down into the categories of efficiency and comfort, safety, security and certainty as well as pleasantness and attractiveness according to Blecic et al. (2020). We renamed the category ‘efficiency and comfort’ as ‘functionality’ as well as the category ‘safety, security and certainty’ as ‘safety and security’ because we feel this fits the characteristics of its variables better. The third and final basis was on our preliminary work on factors influencing walking accessibility perception and their effect on walking (step 3).

As a first step, objective accessibility was calculated according to the well-known Walk Score® (Carr et al., 2010; Duncan et al., 2011; Walk Score®, 2011). The Walk Score® is an efficient, cost-effective valid measurement to assess walking accessibility (Carr et al., 2010, 2011; Duncan et al., 2016). Since it is a common and widely accepted methodology approach which has been used and adapted in several academic papers (Vale et al., 2016; Hall and Ram, 2018; Fina et al., 2022), we decided to use the Walk Score® as the basis of our analysis. The analysis required different basic datasets: geometries (as a starting point for distance calculation), facilities (as an end point for distance calculation) and the road network (to calculate the shortest distances). For our analysis, we decided to choose a 100x100 metre grid, allowing the calculation of very detailed objective walking accessibility measures and making the units independent of administrative boundaries (Göddecke-Stellmann, 2013; Kaup and Rieffel, 2013; Neutze, 2015).

A comprehensive and thematically diverse database was set up for facilities. In the process, locations with particularly important facilities for the local provision of facilities and services of the population in Bahrenfeld and Barmbek-Nord were identified via open geodata from the city of Hamburg, enquiries, internet research via maps, OpenStreetMap (OSM) and the Yellow Pages. These locations were geocoded in the reference system in order to use them in GIS as destinations in an accessibility analysis (van der Vlugt and Welsch, 2020).

In this first step, we base the calculation of objective walking accessibility principally on the Walk Score® methodology but adapt it in certain aspects. This includes the decay function and the integration/selection of facilities. Table 1 shows the facilities included. The first nine indicators, based on the original Walk Score® (light grey), were supplemented by two additional indicators (pharmacies, general practitioners) (dark grey). These were selected since recent studies underline the importance of healthcare as an essential component of public services and they are important destinations, especially for the two target groups (Ahlmeyer and Wittowsky, 2018; Chudyk et al., 2015). The facilities were included in the calculation with varying frequency and weighting. The distance to the nearest destination of various facilities was calculated on this basis. The absolute distances were converted into a distance function value. The distance decay function determined what percentage of a full score a category received based on the distance of the calculated walking route, that is, between origin and destination. We used a cumulative Gaussian distance decay function (Vale and Pereira, 2017). A distance within 300m to the facility got the full score. After this, scores decreased gently with distance. At a distance of 1.5km, facilities received only about 25% of the score. After 1.5km, the scores decreased less quickly with greater distance until they reached 2.5km, after which they did not enter the final score (Walk Score®, 2011). The distance decay function converted the distances into point values based on the distances and weightings of the facilities. The maximum value is 25. According to the Walk Score® calculation logic, a linear expansion is made on a scale from 0 to 100 (Walk Score®, 2011). For this purpose, the values are multiplied by a factor of 4.

As a second step, a small-scale neighbourhood audit was conducted to collect walkability indicators representing the qualities of the living environment and of urban walking routes. By using the methodology of a neighbourhood audit on a small-scale level, we are able to collect detailed information about specific neighbourhood characteristics that cannot be obtained from secondary data (for example, the number of trees, pavement width, obstruction due to parking). Besides the availability of data, some indicators need to be directly observed or felt. Therefore, the neighbourhood audit is a suitable method to record specific walkability indicators, which include impressions as well as just the recorded data. Using an assessment form, walkability indicators per street segment were assessed by a trained team in both study areas. In one week, a total of 95 street segments in Bahrenfeld and 85 street segments in Barmbek-Nord were examined with regard to walkability indicators. The segments were assessed by two people to limit survey bias and ensure a certain objectivity (Figures 2 and 3). As already explained in the conceptual framework, four broad categories of walkability indicators were analysed following Blecic et al. (2020), complemented by the perception criteria. Based on the literature review the overarching criteria were further subdivided into thematic segments (pavement, street design, barrier-free environment, green space, noise, attractiveness of the environment, subjective fear-causing points, perceived walking accessibility) and broken down into a total of 18 walkability indicators (Table 2).

Table 2 illustrates the analysed walkability indicators, their measurement and weighting factor for adapting the Walk Score®. We decided to apply an additive equal weighting of the walkability indicators in the form of deduction points. Deduction points count any aspects on the route that negatively affect walkability. Equal weighting means that we do not weight one indicator higher than another. For example, the presence of trees is not weighted higher than the presence of seating or obstruction due to parking. We have deliberately chosen a deficit analysis with the aim to reveal the potential for intervention of the individual indicators and to critically examine whether a pedestrian paradise really exists. This means that any variable affecting walkability will result in a 2 percent deduction (such as an obstacle on the pavement or a 50 km/h speed limit). This applies to all variables except noise pollution and perceived walking accessibility. This is justified by the fact that average exposure to noise above 55 dB seriously harms health (WHO, 2010) and can trigger high blood pressure and heart diseases (European Commission). It is not sufficient to integrate just one level of noise pollution, since an increase in decibel levels has exponential negative effects on human health (European Environment Agency, 2020) resulting in cardiovascular and psychophysiological effects (WHO, 2010). We therefore differentiated between four decibel levels starting in line with Otsuka et al. (2021) with penalties above 55 dB having a 2% deduction. In the case of perceived walking accessibility (the only variable reflecting the individual perception of the target groups), it is not useful to include just one characteristic either. By classifying the variable into five categories, we tried to include heterogeneous human perceptions.

Accordingly, the deduction points were summed up per street segment in an additive equal weighting and integrated into the PEWI calculation.

As a third step we developed a household survey in the two investigation areas in 2015 to collect and integrate subjective perceptions. A random sample of 400 households of the relevant target groups, 200 per neighbourhood was randomly selected based on a stratified sample of the two target groups retrieved from the resident’s registration office. The survey documents were placed in the letterboxes of the selected households together with a stamped return envelope. All adults in the household were encouraged to complete the survey about their perceptions of accessibility, mobility requirements, usual transport modes, time-dependent safety perceptions, barrier-free environment as well as their travel and neighbourhood related attitudes. We incorporated data previously presented in van der Vlugt et al. (2019; 2022).

The participants were on average 60.51 (sd 19.12) years old with a slight predominance of women among the respondents (59.9%) in Bahrenfeld (60.3%) as well as in Barmbek-Nord (59.5%). More than half of the respondents did not have a university degree (59.3%) and had an average monthly household net income of less than 3,000 € (55.6%).

As the main indicator to capture individual perceptions, perceived walking accessibility was calculated using a factor analysis and subsequent averaging of three items measured on a five-point Likert scale:

Based on a high-level correlation between these items (α = 0.739), the factor of perceived walking accessibility was created as an average scale. For more details, we refer interested readers to van der Vlugt et al. (2019).

The perceptions of accessibility differ by neighbourhood and age group. Both older people and young families perceive accessibility better in Barmek-Nord (3.93; sd 0.75) than in Bahrenfeld (3.27; sd 0.87). Since we do not distinguish between the target groups in this paper, we rather focus on the difference between the two target areas.

The perceived walking accessibility scale has a value range of 1–5 and is integrated in the PEWI in 2 percentage steps (see weighting factor in Table 2). Since the survey is based on a random sample, some grid cells include several respondents while others have zero observations. Therefore, the values were interpolated over the entire study area and averaged to the grid cells.

The household survey method was chosen in order to be able to quantitatively survey perceptions of accessibility. Through the preliminary work of van der Vlugt et al. (2019), it was possible to integrate the factor of perceived accessibility into an instrument, which would not have been possible with qualitative data.

The PEWI was calculated based on the description in Section 3.1. In this process, the road segments investigated from the walkability audit were intersected with the population grid (100m × 100m). As a result, 233 inhabited grid cells each in Bahrenfeld and Barmbek-Nord were included in the calculation. A network analysis was used to re-calculate the shortest route to the different facilities, taking into account the above-described indicators. Each route was loaded with deduction points, depending on the respective walkability indicator assessment as well as the perceived walking accessibility score (Table 2).

As a special characteristic of the neighbourhood audit, each route consisted of different segments and regularities determined the conditions under which deduction points were assigned per route: A distinction was made between indicators for which deduction points were assigned for the entire route as soon as they occurred in a single segment (e.g., steps, obstacles, available seating) and indicators that must be present in the majority of the segments for deduction points (count or length) to be assigned for the entire route (separation due to road width, trees, lanes, street greenery) (Table 2, right column).

The loaded routes were then included in the PEWI according to the variety and weighting of each facility.

To visualise this approach, Figure 4 shows a simplified illustration of this calculation. It shows the deductions on the Walkscore based on two trips (ATM and the grocery store) and three walkability indicators (speed, noise pollution, and available seating).

The first route (green) leads to a grocery store with an infrastructure weighting of 3: a large part of this route goes along a main road with a maximum speed of 50 km/h and a noise level of 65 to 75dB, which leads to a deduction of 6%. Equation (1) shows the entire calculation. The plain value after the distance weighting is 2.07; by deducting noise and speed the value is only 1.97.

The second trip (blue) to the ATM is only 350m but is not equipped with benches. Without the deductions the plain value is 0.99 (see equation 2). The 2% deductions reduce the value to 0.98.

With the intention of developing an objective information, planning and decision-based instrument, the PEWI is presented in Table 3 as the outcome of integrating and considering indicators of the built environment, qualities of the walking route as well as individual walking accessibility perceptions. It depicts the descriptive results of the Walkscore and PEWI categories by target area. For the Barmbek-Nord case study, the average of objective walking accessibility (Walkscore) (Figure 9) within the target area resulted in 90 (sd = 6.78). This figure falls into a category of ‘walker’s paradise’ (Walk Score®, 2011). Our adapted version, the PEWI considering all 18 indicators shows a more nuanced picture of current walkability with an average of 80 (sd = 5.15), a very walkable category. In order to find out whether the Walkscore differs on average from the PEWI, a comparison of mean values was carried out. Accordingly, a t-test was used to compare the mean values of the Walkscore with the mean values of the PEWI. The t-test shows a significant difference between the Walkscore and the PEWI t (44.63) = 208 (p < 0.01).

By comparing the two figures, it can be seen that Barmbek-Nord’s PEWI and therefore its walkability drops extensively from the highest category to very walkable. In some parts, walkability in Barmbek-Nord has changed to the category somewhat walkable (north-east). In the south, a small part remained in the walker’s paradise category. In the Bahrenfeld case study, the average of objective walking accessibility (Figure 9) was 76 (sd = 12.08) and fell into the very walkable category. If we take PEWI into account (Figure 9), walkability deteriorates on average to 68 (sd = 10.38), somewhat walkable. The Walkscore shows a small walker’s paradise in the centre of Bahrenfeld. This category is no longer present in the PEWI (very walkable). In the southwest, walkability changes from very walkable into somewhat walkable. In the northeast, even grid cells fall in terms of their walkability from somewhat walkable into car dependent. The comparison highlights the limitation of common methods based on distances to the nearest facilities. They do not consider qualities of the walking routes which can decrease the walkability of an area. In line with D’Orso and Migliore (2020), the results show the importance of integrating a variety of walkability indicators to prevent people from preferring to take their car, instead of walking, and creating noisy and unattractive walking environments themselves. Finally, it is possible with the PEWI to identify priorities for interventions, spatially represent and contribute to developing walkable environments and ultimately promote active mobility for all population groups (D’Orso and Migliore, 2020).