Assessment of Outdoor Lighting: Methods for Capturing the Pedestrian Experience in the Field

: This study assessed whether methods for capturing the pedestrian experience of outdoor lighting, previously evaluated in a full-scale laboratory, were applicable in a real-world setting. It applied an approach capturing the human response to outdoor lighting in a systematic way, by assessing perception, evaluation and behaviour in the lit environment. The study involved 81 participants from two age groups (Young—n: 48, mean age: 26, 63% women; Elderly—n: 33, mean age: 69, 67% women) and was carried out on a pedestrian path in a park in the centre of Malmö, Sweden, in the evenings during wintertime. Two LED lighting applications, differing in light distribution, uniformity and horizontal illuminance, were presented, and the pedestrians’ perception (facial expression recognition and sign reading), evaluation (arousal, valence and perceived outdoor lighting quality) and behaviour (pedestrian ﬂow) were assessed. The results from the perceptual tasks differed signiﬁcantly between the lighting applications, in favour of the lighting application with greatest uniformity and horizontal illuminance. There was a signiﬁcant difference in sign reading distance between the two age groups. The methods applied in this study are feasible to administer and could be used to assess lighting solutions in order to capture the needs of vulnerable groups.


Introduction
Today, the majority of the world's population lives in cities, and further urbanisation is expected globally [1]. This trend accentuates the need to develop sustainable cities, with safe and healthy living environments, and energy-efficient transportation systems [2]. The pedestrian is central in sustainable urban design, since nearly all journeys undertaken in the urban environment incorporate walking in one way or the other. Walking is therefore an essential component in sustainable intermodal transportation systems [3][4][5][6], and benefits public health, by reducing the risk of chronic diseases such as cancer, diabetes and heart disease [7].
In the Nordic countries (and other countries at similar latitudes), daylight hours are very limited in winter and pedestrians depend on outdoor lighting to provide functional levels of visual accessibility and perceived safety when getting to and from work.
The social-ecological model of walking [8] suggests a hierarchical structure of pedestrian needs consisting of five levels (ranging from feasibility via accessibility, safety and comfort to pleasurability) that people consider when deciding to walk. The existence of artificial outdoor lighting impacts the evaluation of accessibility and safety needs [9][10][11], and is fundamental for people's decisions as to whether to walk or not after dark. The importance of outdoor lighting for pedestrians was highlighted in a recent focus-group study in Malmö, Sweden, where the participants were encouraged to speak freely about the perceived design qualities of neighbourhoods in relation to walking. In most cases, despite the discussions taking place during daytime, outdoor lighting was mentioned spontaneously, in relation to perceived safety and intention to walk after dark [12]. This is in line with the previous research indicating that the presence of outdoor lighting improves the perceived neighbourhood quality [13] and increases the amount of walking after dark among all age groups: adolescents [14], adults [15][16][17] and elderly [18][19][20].
Unfortunately, the benefits of outdoor lighting are associated with adverse ecological consequences [21], as well as considerable energy use, generating financial and environmental costs [22]. One way of reducing energy use is to replace or retrofit existing outdoor lighting applications with light sources with greater energy efficiency [22][23][24].
In Sweden, many municipalities are actively promoting walking as a means of transportation and have developed pedestrian plans [25]. In these plans, improved outdoor lighting for pedestrians is stressed as a key factor. Municipalities on the verge of upgrading the lighting infrastructure are looking to LEDs to reduce energy use. However, LED lighting may have different photometric properties (e.g., spectral power distribution) than previous lighting technologies, such as high-pressure sodium and ceramic metal halides, and vary greatly depending on the quality and design [26]. Before committing to a costly, long-term investment, the needs of the users should be considered. Both technical aspects of lighting and the pedestrians' experience should be taken into consideration, in order to find lighting solutions adapted to user needs while minimising energy use. Moreover, the needs of users from vulnerable groups, such as the elderly and the visually impaired, must be considered [9,27,28]. Most components of the visual system deteriorate with age, which affects visual performance negatively. Elderly therefore are more sensitive to glare and may have difficulties performing certain visual tasks important for pedestrians (such as detecting obstacles [29,30] and recognising facial expressions [30]) at low illuminance levels [31]. For people with low vision, it has been deemed difficult to provide general illuminance recommendations, due to large individual differences [31]. Therefore, in order to design lit environments more suitable for people of all ages, with varying levels of vision, it is important to identify which parameters are central for the pedestrian experience, and to develop tools that can be used to assess them.
In many municipalities, for economic reasons, it is likely that only the luminaire will be exchanged, while the old lampposts are retained. This might bring about an unsatisfactory lighting solution and indicates the need to assess the pedestrian lighting experience in the field before initiating large-scale retrofits.
To identify and evaluate the methods that assess pedestrians' experience of the lit environment, Rahm and Johansson [30] conducted a full-scale laboratory study, with three different lighting applications. The identified methods for perception (obstacle detection, facial expression recognition distance and sign reading distance) and evaluation (level of arousal [70,71] and Perceived Outdoor Lighting Quality scale (POLQ) [10]) differentiated between the different lighting applications. The lighting application with the greatest mean horizontal illuminance (E H ) (32 lx, compared to 28 and 18 lx), the widest light distribution, and highest correlated colour temperature (CCT) (3810 K, compared to 2912 and 2890 K) achieved the best results on the perception tasks and was perceived as significantly different on the Perceived Comfort Quality (PCQ) and the Perceived Strength Quality (PSQ) dimensions of the POLQ scale, as well as on the composite arousal measure. No significant difference in the behaviour measure, walking speed, was found between the different lighting applications. However, other studies indicate that differences in illuminance level may result in changes in walking speed (0.05-0.11 m/s for differences in horizontal illuminance ranging from 14 to 290 lx [62][63][64]) and in pedestrian flow when improving the lighting conditions [52,61].

Aim and Hypotheses
The aim of this study was to assess whether the methods for capturing the pedestrian experience of outdoor lighting, previously evaluated in a full-scale laboratory, are applicable in a real-world setting. A second aim was to investigate whether the group of elderly (60-75 yrs.) experienced the lit environment differently than the group of younger participants (20-35 yrs.), and whether the lighting applications were sufficient for both age groups. The motive for this research is to develop an approach that captures the human response to outdoor lighting in a systematic way, by assessing pedestrians' perception, evaluation and behaviour in the lit environment. Such an approach could be used by municipalities to differentiate between lighting applications before initiating large-scale retrofits. The observer-based environmental assessment using a mobile method, along with the assessment of the three dimensions of human response to outdoor lighting, served to contribute to the understanding of the impact of outdoor lighting on the walkability of a neighbourhood.
Previous research, along with the results from the laboratory study, gave rise to the following expectations: The methods are expected to, in a field setting, discriminate between two lighting applications, deemed equivalent by the local municipality, on the dimensions of perception, evaluation and behaviour. Further, due to the decline in night vision associated with increased age [72], the lighting applications are expected to be less supportive to the needs of the group of elderly with regard to the perceptual tasks.

Participants
The sample consisted of 81 participants (mean age: 43, 64% women) belonging to a younger population (n: 48, mean age: 26, 63% women) and an older population (n: 33, mean age: 69, 67% women) (see Table 1). The participants usually walked outside after dark at least a few times a week (Total: 74%, Young: 79%, Elderly: 67%) and most declared that they could see well or very well with the help of the visual aid they normally used (Total: 93%, Young: 98%, Elderly: 83%). No medical problems relating to eyesight were reported. The participants were recruited through organisations for the elderly, information meetings on the university campus, and through personal networks. All participants had good command of both spoken and written Swedish. This study was carried out in accordance with the rules and regulations laid down by the Ethics Committee of the Swedish Research Council [73]. Information about the aim of the study was given and written informed consent was obtained from all participants. The participants were informed of their right to withdraw at any time without giving an explanation. Personal information was anonymised to retain the privacy of the participants, who received approximately 41 EUR after participation as remuneration.

Setting
The study was carried out in Pildammsparken, an urban park in the centre of Malmö, Sweden, in wintertime (November and December, 3-11 degrees centigrade, no snow cover on the ground), between 5 and 8 pm in the evenings (The sun set at approximately 4 pm). The participants walked a 90-m long and 3.4-m wide pedestrian gravel path (Figure 1). The study was located in a dark and secluded part of the park, which was being considered for new lighting applications by the local municipality. The part of the park where the study was located is frequently used by pedestrians passing through the park, people walking for recreation, and by joggers for exercise. The park is surrounded by residential areas, and there are no commercial areas in the vicinity.

Setting
The study was carried out in Pildammsparken, an urban park in the centre of Malmö, Sweden, in wintertime (November and December, 3-11 degrees centigrade, no snow cover on the ground), between 5 and 8 pm in the evenings (The sun set at approximately 4 pm). The participants walked a 90-metre long and 3.4-m wide pedestrian gravel path ( Figure 1). The study was located in a dark and secluded part of the park, which was being considered for new lighting applications by the local municipality. The part of the park where the study was located is frequently used by pedestrians passing through the park, people walking for recreation, and by joggers for exercise. The park is surrounded by residential areas, and there are no commercial areas in the vicinity.

Lighting Applications
On the right-hand side of the path, lampposts were placed at intervals of 21.5 metres, with luminaires at a height of 4 metres. Two LED lighting applications were used in the study: lighting application I (CCT: 3060, Colour Rendering Index, CRI: 74, Scotopic/Photopic ratio, S/P: 1.25) and II (CCT: 3028, CRI: 80, S/P: 1.28) (for photometric data, see Table  2). The lighting applications were selected by the municipality of Malmö on the basis of economic feasibility, fulfilment of technical specifications according to the national standards and relevance for use on pedestrian paths in the city. To attain greater ecological validity, the most common CCT (3000 K) for outdoor use in Malmö was used. The horizontal illuminance on the path varied between 3 and 58 lx (lighting application I,  The reflection factors for the facial expression (0.8) and sign reading (0.9) stimuli were estimated by using a Hagner reflection reference, and a Hagner S4 universal photometer. The vertical illumination levels for the face and the sign were then calculated based on the luminance values from Table 2.

Lighting Applications
On the right-hand side of the path, lampposts were placed at intervals of 21.5 m, with luminaires at a height of 4 m. Two LED lighting applications were used in the study: lighting application I (CCT: 3060, Colour Rendering Index, CRI: 74, Scotopic/Photopic ratio, S/P: 1.25) and II (CCT: 3028, CRI: 80, S/P: 1.28) (for photometric data, see Table 2). The lighting applications were selected by the municipality of Malmö on the basis of economic feasibility, fulfilment of technical specifications according to the national standards and relevance for use on pedestrian paths in the city. To attain greater ecological validity, the most common CCT (3000 K) for outdoor use in Malmö   Lighting application I had the greatest luminous efficacy, since it provided greater illuminance levels at lower power. The horizontal illuminance measurements were conducted with a Hagner S4 universal photometer, using a grid with a distance of 2.15 m between data points along the length of the path, and at a distance of 1.14 m between data points across the path. CRI, CCT and spectral power distribution were measured with an Avantes AvaSpec 2048 spectroradiometer combined with an Avantes FC-UV400-2-ME detector. Luminance was measured with a Hagner S4 universal photometer, and vertical illuminance was calculated based on the measured luminance and the reflectance factors. Lighting application I had the greatest luminous efficacy, since it provided greater illuminance levels at lower power. The horizontal illuminance measurements were conducted with a Hagner S4 universal photometer, using a grid with a distance of 2.15 metres between data points along the length of the path, and at a distance of 1.14 metres between data points across the path. CRI, CCT and spectral power distribution were measured with an Avantes AvaSpec 2048 spectroradiometer combined with an Avantes FC-UV400-2-ME detector. Luminance was measured with a Hagner S4 universal photometer, and vertical illuminance was calculated based on the measured luminance and the reflectance factors.
The accuracy of the Hagner S4 universal photometer is stated to be better than ±3% for all common light sources. To get more precise measurements, a spectral mismatch correction factor could be applied, to compensate for the spectral mismatch between the measured light sources and incandescent light, for which the photometer is calibrated [74]. However, since the light sources evaluated in this study have similar Spectral Power Distribution, SPD (Figure 3), the inaccuracy of the measurements may be assumed to be similar. Since the purpose of Figure 2 is to illustrate the relative difference between the lighting applications, the illuminance levels have not been mismatch corrected. However, when designing road lighting applications, spectral mismatch correction factors ought to be considered in order to obtain correct simulations [75]. The lampposts were located on the right-hand side, at the start and end of the depicted segment of the path. The measurements were conducted from midpoint to the second lamppost and extrapolated for illustrative purposes.
The accuracy of the Hagner S4 universal photometer is stated to be better than ±3% for all common light sources. To get more precise measurements, a spectral mismatch correction factor could be applied, to compensate for the spectral mismatch between the measured light sources and incandescent light, for which the photometer is calibrated [74]. However, since the light sources evaluated in this study have similar Spectral Power Distribution, SPD (Figure 3), the inaccuracy of the measurements may be assumed to be similar. Since the purpose of Figure 2 is to illustrate the relative difference between the lighting applications, the illuminance levels have not been mismatch corrected. However, when designing road lighting applications, spectral mismatch correction factors ought to be considered in order to obtain correct simulations [75].

Perception
To evaluate whether differences between the lighting applications affected the visual accessibility of the environment, two tasks were administered: facial expression recognition and sign reading (Figure 4). Facial recognition is deemed an important visual task for  To evaluate whether differences between the lighting applications affected the visual accessibility of the environment, two tasks were administered: facial expression recognition and sign reading (Figure 4). Facial recognition is deemed an important visual task for pedestrians after dark, since judging the intent of oncoming pedestrians from a safe distance is thought to influence the perception of safety [66,67,76,77]. In the facial expression recognition task, the participants were instructed to walk along the path towards a photograph of a woman's face (175 × 200 mm; positioned at a height of 1.65 m; printed on non-glossy paper), placed on the right-hand side of the path 17.5 m from the first lamppost. The participants were instructed to stop when they could discern the facial expression of the woman (the expression was anger (p. 127), from the Emotions Revealed photo set [78]). They were then asked to give a verbal statement of the perceived emotion.

Perception
To evaluate whether differences between the lighting applications affected the visual accessibility of the environment, two tasks were administered: facial expression recognition and sign reading (Figure 4). Facial recognition is deemed an important visual task for pedestrians after dark, since judging the intent of oncoming pedestrians from a safe distance is thought to influence the perception of safety [66,67,76,77]. In the facial expression recognition task, the participants were instructed to walk along the path towards a photograph of a woman's face (175 × 200 mm; positioned at a height of 1.65 m; printed on non-glossy paper), placed on the right-hand side of the path 17.5 m from the first lamppost. The participants were instructed to stop when they could discern the facial expression of the woman (the expression was anger (p. 127), from the Emotions Revealed photo set [78]). They were then asked to give a verbal statement of the perceived emotion.  The second task, street sign reading, is considered an important task for orientating oneself in the environment [79]. For the sign reading task, the participants were asked to continue along the path towards a street sign placed 2 m to the left of the path and 5 m from the next lamppost in the walking direction, positioned at a height of 2.10 m. The street sign was placed to the left of the path in order to simulate conditions in a realworld setting. The street sign was of similar type and with equivalent number of syllables (Rosenlundsgatan; Tratex font; Size: 212) as in the laboratory study [30]. The task of the participants was to stop when the text on the street sign was legible. The distance was then measured, and the participants were asked to read out the street name aloud in order to validate the legibility.

Evaluation
To assess whether the differences between the lighting applications had an impact on how the participants experienced the lit environment, the participants assessed their emotional state (arousal and valence) at the instance when they had walked halfway down the path, in a two-dimensional grid (consisting of 5 × 5 cells with the labels Active placed above, Passive below, Negative to the left and Positive to the right) [70,71]. Moreover, after the participants had walked down the path, they responded to the POLQ scale [10]. The scale consists of ten items factored into two dimensions: Perceived Strength Quality (PSQ, Cronbach's alpha, α = 0.841) (subdued-brilliant; strong-weak; dark-light; unfocusedfocused; clear-drab) and Perceived Comfort Quality (PCQ, α = 0.791) (warm-cool; naturalunnatural, glaring-shaded; mild-sharp; hard-soft) rated on a 7-point scale; items in italics were reversed before indices were calculated. The participants also rated how well they could see under the present lighting application, using a 7-point scale (1 = very poorly, to 7 = very well) and they rated their perceived visual accessibility by the following statements: I would have been able to (a) detect objects on the ground; (b) read a street sign and (c) recognise the people's faces. Responses were given on 5-point scales (1= totally disagree, to 5 = totally agree). The mean score of the perceived visual accessibility scale was calculated and its internal consistency was evaluated using Cronbach's alpha (α I = 0.854; α I I = 0.788).

Behaviour
To evaluate the potential impact of lighting conditions on pedestrian flow (number of pedestrians/min), an observation was conducted between 7 and 8 pm on four different occasions (Sunday, Tuesday, Wednesday and Thursday) for each lighting application. The pedestrian path was observed from a secluded spot some distance away and the number of pedestrians was recorded.

Design and Procedure
The study employed a between-subjects design, where the first group conducted the tasks under lighting application I and the second group under lighting application II (Tables 1 and 2). The luminaires were mounted on opposite sides of the lamppost, and the top of the lampposts were rotated to give each group the right lighting condition. The participants arrived in groups of five at the meeting point, located a short distance from the starting point of the path. Before starting the study, they were informed, according to the process of informed consent, about the aim, procedure and their rights to abort their participation without stating a reason before they signed a consent form.
The participants were then instructed on how to use the POLQ scale and the affect grid. For the POLQ scale, the participants were asked to assess the perceived lighting quality of a pedestrian path depicted in a photo [80] (17 × 21 cm) with the objective of establishing a common starting point, as well as giving an opportunity to ask questions about how to use the scale.
The participants were then asked to complete questionnaires surveying background data and individual characteristics. The meeting point had the same type of lighting applications as the path, and the participants spent approximately 15 min under these conditions. The next step was to, one by one, walk to the starting point and then walk along the footpath. The procedure was based on the structured walk methodology developed by Johansson et al. [71]. The participants were instructed to stop when they reached markings on the ground indicating they had reached the location to fill in the affect grid. When they had completed the affect grid, the participants continued to the end of the path where they completed the POLQ scale, rated how well they could see, and responded to the visual accessibility items. When one participant had finished the walk, the next started. When all participants had completed their walks, they returned to the meeting point while the sign reading and facial expression recognition tasks were prepared. Then the participants individually returned to the starting point and performed the facial expression recognition task, followed by the sign reading task. When all participants had completed the tasks, there was a debriefing and the participants were thanked for their participation.
The results were analysed by IBM SPSS 22, using two-way independent ANOVA for analysing differences on measures of perceptual tasks and evaluation measures between the groups experiencing different lighting applications and between the age groups, whereas a Mann-Whitney U-test was used for analysing differences in pedestrian flow between the two lighting applications. To avoid potential problems with violations of assumptions underlying the use of ANOVA, parallel analyses were conducted using a robust two-way independent ANOVA on the trimmed means (10%) using the t2way function of the WRS2 package in R. The results from the robust trimmed means two-way ANOVA supported the findings from the original two-way ANOVA.

Perception
Descriptive data for the assessments of perception, evaluation and behaviour in the lit environment are reported in Table 3, and the results from the statistical analyses are presented in Table 4. The results for the visual tasks used to assess perception of the lit environment differed significantly between the lighting applications. For facial expression recognition, lighting application I enabled the participants to feel confident of recognising the facial expression at an average distance of approximately 1.5 m greater than for lighting application II (F(1, 77) = 5.256, p = 0.025, η 2 p = 0.064) (Table 3). Similarly, for sign reading, the first lighting application enabled the participants to read the street sign at a greater distance (15.79 m) compared to lighting application II (12.24 m) (F(1, 77) = 18.325, p < 0.001, η 2 p = 0.192). For sign reading, there were also significant age differences (F(1, 77) = 36.953, p < 0.001, η 2 p = 0.324). The young group was able to read the street sign at a distance of about 16 m, whereas the older group averaged at about 11 m. Further, there were significant interaction effects between age and lighting application for the sign reading task (F(1, 77) = 7.201, p = 0.009, η 2 p = 0.086), where the younger group managed to read the street sign at a relatively greater distance for lighting application I compared to application II (Table 4).

Evaluation
The two lighting applications were evaluated similarly for emotional state, seeing conditions, the POLQ scale and perceived visual accessibility (Tables 3 and 4). The results from the composite arousal measure showed that the younger group were significantly less aroused than the elderly (F(1, 77) = 4.853, p = 0.031, η 2 p = 0.059) and for perceived visual accessibility, there were significant interaction effects (F(1, 77) = 4.343, p = 0.040, η 2 p = 0.053). The young group rated lighting application I as providing the best visual accessibility, whereas the elderly preferred lighting application II (Table 4).

Behaviour
For the behaviour measure, observations of pedestrian flow, there was no significant difference between the lighting applications (U = 1, p = 0.057, r = −0.71) (Table 4). However, the effect size of −0.71 indicates a large effect, and the non-significant result may be a type II error, due to the limited number of observations (N = 8). As seen in Figure 5

Discussion
This field study evaluated two retro-fit LED outdoor lighting applications that differed in light distribution, uniformity and horizontal illuminance, but that were similar in other aspects. The study had two aims: to assess whether the methods capturing pedestrian experience of outdoor lighting, previously evaluated in a full-scale laboratory, were applicable in a real-world setting, and to investigate whether the elderly experienced the lit environment differently to the young, and if so, whether the lighting applications were sufficient for both age groups.
The study shows that the perceptual tasks (facial expression recognition and sign reading) differentiated between the two lighting applications that, while providing satisfying horizontal illuminance levels according to international standards, produced markedly different seeing conditions. While the differences in lighting conditions resulted in significant differences for the perceptual tasks, no significant differences were found between the lighting applications for the evaluation and behaviour measures. The lighting application with the greatest illuminance and most uniform distribution allowed the participants to identify facial expressions and read street signs at greater distances. This was expected, since these conditions resulted in greater luminance on the photograph and the street sign, so they were brighter and easier to distinguish. Similar results would likely have been obtained had the visual tasks been performed at another part of the path, since lighting application I provided greater illuminance on all parts of the path (Figure 2).

Discussion
This field study evaluated two retro-fit LED outdoor lighting applications that differed in light distribution, uniformity and horizontal illuminance, but that were similar in other aspects. The study had two aims: to assess whether the methods capturing pedestrian experience of outdoor lighting, previously evaluated in a full-scale laboratory, were applicable in a real-world setting, and to investigate whether the elderly experienced the lit environment differently to the young, and if so, whether the lighting applications were sufficient for both age groups.
The study shows that the perceptual tasks (facial expression recognition and sign reading) differentiated between the two lighting applications that, while providing satisfying horizontal illuminance levels according to international standards, produced markedly different seeing conditions. While the differences in lighting conditions resulted in significant differences for the perceptual tasks, no significant differences were found between the lighting applications for the evaluation and behaviour measures. The lighting application with the greatest illuminance and most uniform distribution allowed the participants to identify facial expressions and read street signs at greater distances. This was expected, since these conditions resulted in greater luminance on the photograph and the street sign, Energies 2021, 14, 4005 10 of 15 so they were brighter and easier to distinguish. Similar results would likely have been obtained had the visual tasks been performed at another part of the path, since lighting application I provided greater illuminance on all parts of the path (Figure 2). However, even though the perceptual tasks differed between the lighting applications, the participants were not aware of the differences in seeing conditions. The different designs of the luminaires may have resulted in different glare properties, but no significant differences were found for the POLQ scale, which contains an item assessing glare. The participants reported that they could see well and rated the perceived accessibility as satisfactory for both lighting applications, despite differences in horizontal illuminance and uniformity. However, there was a significant interaction effect for perceived visual accessibility where the young group experiencing lighting application I rated the perceived visual accessibility greater than the group experiencing lighting application II. The opposite was true for the two groups of elderly.
There were also significant differences between the age groups. The younger participants could read street signs at greater distances than the older participants. This result was also in line with the expectations, since both acuity and night vision deteriorate with age [72]. The results from the group of elderly are therefore of special interest, since they have greater needs in terms of providing pedestrian-friendly outdoor lighting.
Despite both lighting applications providing horizontal illuminance levels far above the minimum requirements for Swedish pedestrian paths [81], the older participants struggled with identifying facial expressions at the recommended minimum distance of four metres [46,68,76] for both lighting applications, whereas the young group managed to identify facial expressions at an average distance of six metres under lighting application I. However, the vertical illuminance levels at the signpost and at the photograph were below the lowest recommended level (2.5 lx) for pedestrian paths, according to the Swedish standards [81], which might explain the poor performance of the elderly. The relatively short distances, compared to the recommended minimum distance, may also be a sign of the task being more difficult than detecting the facial expression of a real person. Both the elderly and the young performed much better on the sign reading task despite lower luminance levels, which may reflect the difference in difficulty between the two perceptual tasks. The facial expression detection task relied upon correct perception of relatively minor changes in expressions, while the sign reading task relied upon correctly reading black letters on a white background, using a large font size and a font designed for legibility.
The perceptual tasks corresponded to different needs according to Alfonzo's needs hierarchy [8], where sign reading is related to accessibility and facial expression recognition to perceived safety. From that perspective, both lighting applications provided lighting conditions good enough to satisfy the accessibility need, but were not sufficient in supporting facial expression recognition, which is deemed relevant for the perception of safety. Unfortunately, the experimental situation did not allow for an unbiased assessment of perceived safety, due to the presence of the researchers and the group of participants in the nearby surroundings. However, in a future study, it would be of interest to assess the perception of safety of pedestrians walking alone in differently lit environments. The PCQ dimension of the POLQ scale may be used for predicting perceived safety [10]. Using it as a proxy for perceived safety, the results do not indicate significant differences between the two lighting applications.
There are several plausible explanations for why the evaluative methods did not capture the differences in illuminance levels and uniformity. First, the difference between the two lighting applications might not have been sufficiently pronounced to detect the differences with the selected methods. In other studies, paired comparisons have been employed, both in the laboratory [33,34] and in the field [37]. Paired comparison could possibly have been successful in differentiating between the lighting applications used in this study. However, such an approach was not suitable, due to our intention to capture the direct exposure of walking in the lit environment, in contrast to standing still and alternating the view between two different lighting applications. In future studies, a newly developed method, Random Environmental Walking [82], might be used for differentiating between the lighting applications. It uses a forced-choice technique to tap into pedestrians' relative preferences regarding lighting applications. However, the method is developed for a simultaneous presentation of multiple lit environments, which was not a viable option for this study.
Second, the illuminance levels of both lighting applications might have reached a plateau where an increase in illuminance did not contribute noticeably to the pleasantness of the lit environment. Third, in comparison to the laboratory study, where the evaluative methods were successful in differentiating between the lighting applications, this study had to resort to a between-subjects design for practical reasons. With a weaker research design, potential differences between the different lighting applications might have been missed. Lastly, the differences in the laboratory study might have been partly due to differences in CCT, a factor that was held constant in this field study. In the field, the most common CCT for outdoor use in Malmö was picked, to attain greater ecological validity.
The observational method applied did not discover any significant differences regarding pedestrian flow. However, the limited number of observations may have concealed existing differences indicated by the large effect size. Nonetheless, there are many reasons why people choose a specific path, and it is possible that something other than perceived lighting quality determined where people walked. Habit or preference for the most direct route to the destination may have outweighed the impact of the differences in uniformity and illuminance. The lighting applications might also have been too similar to generate detectable differences in pedestrian flow. Therefore, it would have been valuable to employ an additional behavioural assessment method. A possible complement to direct observation could have been to film the path. The use of video recordings, possibly in combination with eye-tracking, could have cast light on where pedestrians placed themselves on the path, where they directed their gaze, and which strategies were employed with regard to glaring luminaires.
In this study, the lighting applications were chosen in collaboration with lighting designers from Malmö Municipality, with the aim to discriminate between lighting applications that were to be integrated with the municipality's normal lighting scheme. This meant that the lighting applications differed in light distribution, uniformity and horizontal illuminance, while being similar in other aspects. If the objective would be to evaluate the impact of a certain parameter in isolation, all but that parameter should be kept constant. Such an approach would for example be suitable for determining how much lighting applications need to differ for a difference to be detected, or for evaluating lighting applications at varying levels of illuminance.
The presence of outdoor lighting is an important factor when people are considering if and where to walk after sunset. With recent technological advances in LED lighting, there is potential for achieving outdoor lighting much more tailored to the needs of the users. To accomplish this, an understanding of how different user groups perceive the lit environment is needed. To date, the research on how to assess the human response to outdoor lighting is scarce, and methods and tools for field use must be developed. This study is an attempt to develop a systematic approach to the assessment of pedestrians' experience of outdoor lighting, intended for use by municipalities prior to large-scale retrofits. It evaluates the generalisability of measures that have been shown to differentiate between lighting applications and age groups under controlled conditions, by applying them in a field setting. The results show that the perceptual tasks of sign reading and facial expression recognition can be used to differentiate between different lighting applications in the field, and to identify where lighting applications are inadequate with regard to groups with special needs. The methods applied in this study are feasible to administer and could be part of a municipality's analysis prior to a retrofitting or before a decision on lighting schemes for a new development. The methods could be used to capture the needs of vulnerable groups, to create better lighting conditions and provide accessibility for all users. Outdoor lighting is associated with considerable energy use, and the methods should be applied along with considerations of energy-efficiency. In this study, the lighting application providing the best seeing conditions was also the most energy-efficient, affording the municipality a solid foundation for their decision.
The perceptual tasks correspond to the pedestrian accessibility and perceived safety needs that, according to Alfonzo [8], contribute to the walkability of a neighbourhood. However, established measures of walkability, such as the Systematic Pedestrian and Cycling Environmental Scan [83] and the Neighborhood Environment Walkability Scale [84], treat outdoor lighting superficially, or leave it out entirely [85]. We suggest that, in order to capture the complex impact of outdoor lighting on walkability, all three dimensions (perception, evaluation and behaviour) of the pedestrian experience of the lit environment needs to be assessed. Further research is necessary to advance the understanding of how LED outdoor lighting impacts walkability in an urban context. It is especially important to investigate how lower illuminance levels might affect the perception, evaluation and behaviour of the elderly and the visually impaired, to identify lighting solutions that cater to user needs while minimising energy use.

Conclusions
In this field study two retro-fit LED outdoor lighting applications that differed in light distribution, uniformity and horizontal illuminance were evaluated by pedestrians. The methods applied were feasible to administer in field and can be used by municipalities in parallel with lighting guidelines to obtain qualitative information about different lighting schemes to improve decision-making regarding new investments. In this study the lighting application with the highest illuminance and most uniform light distribution allowed the participants to succeed with important perceptual tasks at greater distances. For the sign reading task, the younger participants performed better, but for facial expression recognition there were no significant age differences, and both groups of elderly and one of the young groups struggled with identifying facial expressions at the recommended minimum distance of four metres. This shows that the perceptual tasks can be used to differentiate between different lighting applications in the field and highlights the importance of considering the needs of vulnerable groups. However, the evaluation and behaviour measures did not detect any statistically significant differences between the two lighting applications. Further research is necessary to advance the understanding of how LED outdoor lighting impacts walking in an urban context with special consideration to the elderly and the visually impaired, to identify lighting solutions that cater to user needs while minimising energy use.   2021))). The APC was funded by Lund University.

Institutional Review Board Statement:
This study was carried out in accordance with the rules and regulations laid down by the Ethics Committee for the Swedish Research Council after consultation with the Regional Ethical Review Board. The Board concluded that approval according to the Swedish Ethical Review Act was not needed for this study.

Informed Consent Statement:
Information about the aim of the study was given and written informed consent was obtained from all subjects involved in the study in accordance with the Declaration of Helsinki. The participants were informed of their right to withdraw at any time without providing an explanation. Data Availability Statement: Data is available upon request from the authors. Acknowledgments: The authors would like to thank the city of Malmö for providing and installing the lighting applications. Thanks also to Lina Haremst and Rifa Maliqi for assisting with the data collection and to all participants who, despite the dark and cold Swedish winter evenings, volunteered to spend time in the park assessing outdoor lighting.