3.2. Comfort Evaluation Model Results
Mplus software was used for structural equation model analysis in this research. Since the Mplus software for structural equation model analysis lacks the ability of data management, it needs to be processed in SPSS (
SEM1109.sav in Supplementary Materials) [
44].
In this paper, SPSS software was used to conduct descriptive statistical analysis and normal distribution test on the survey report results of lighting attributes filled in by respondents from 65 assessment points in 4 residential areas and the field measurement data collected by the researchers.
In the questionnaire, a Likert four-point scale was used, and the value of each light attribute variable was between 0 and 3 points. All factors should be standardized to represent the relative importance between observable variables and latent variables. Since different variables and indicators have different dimensions and dimensional units, which will affect the results of data analysis, in order to eliminate the impact and meet the application conditions of structural equation model, the min–max standardization method was used to standardize the data of each variable to 0~3 (Equation (5)).
where,
represents the result after standardization, and
represents the original data;
represents the maximum value of all data of this item. Through the standardized formula, the field measurement data and questionnaire data are unified.
The hypothetical structural equation model diagram shown in
Figure 12 includes measurement model and structural model, in which uniformity and color temperature (temperature in
Figure 12) are exogenous latent variables, security and comfort are endogenous latent variables, Un1, Un2, Uni, S1–S3, Sec, T1, T2, Tem, C1, C2 and Com are all observable variables, and e1–e21 are variable residuals (unexplainable variables). Un1 is the standard deviation of the illuminance in the eight directions; Un2 is the ratio of the minimum to the maximum illuminance in the eight directions; Uni is the mean value of the respondents’ score on the brightness uniformity of the evaluation point. S1 represents the mean value of measured illuminance; S2 represents the measured value of glare index; S3 represents the number of lamps illuminated at the evaluation point; Sec is the mean value of respondents’ security score at the evaluation point. T1 represents the mean value of x-color temperature measurement of the light at the measuring point; T2 represents the mean value of y-color temperature measurement of the light at the measuring point; Tem is the mean color temperature score value of the respondents at the evaluation point. C1 represents the mean illuminance score of the respondents at the evaluation point; C2 represents the mean glare index score of the respondents at the evaluation point; Com represents the mean overall feeling of comfort score of the respondents at the evaluation point.
The probability of significant confirmatory analysis of structural equation model analysis is relatively low. The analysis needs to assume and draw the structural equation model in advance, and then verify it through the questionnaire data, and fit the structural equation model before it can be considered that the hypothesis holds. The key to the establishment of this model depends largely on the structure of the initial hypothetical model.
Figure 13 is the syntax of structural equation model construction in Mplus software based on the hypothetical model.
Because this study uses the questionnaire data, the observed Likert scale is a four-point scale, and the field measurement data are also standardized to a four-point scale. In the process of model construction, if the observed Likert scale is less than the five-point scale data, Mplus software will use the WLSMV (weight least square with mean and variance) estimation method for conversion and estimation, and whether it is significant can be read through the schematic diagram of the results.
The model results (
Figure 14 and
Table 3) show that there are 19 continuous dependent variables and 4 continuous latent variables (
comfort1-20220104.inp in Supplementary Materials). Uniformity and temperature (color temperature) are exogenous latent variables, and security and comfort are endogenous latent variables. BY in
Table 3 represents the measurement model, ON represents the structural model, and WITH represents the relationship between the exogenous latent variables.
Table 3 shows that the absolute values of signal-to-noise ratio (=Est./S.E. in
Table 3) on Un1 and Un2 are so low that the two variables do not significantly contribute to the model results. In addition, the evaluation weights of Un1 and Un2 directions are too small, and the P-values are greater than 0.05, indicating that these two factors should be removed. It can also be seen from the results that uniformity and temperature were not suitable for directly acting on comfort, so the model needed to be modified.
The model also shows that the influence of endogenous potential variables of security is less affected by uniformity and color temperature. The evaluation of comfort mainly depends on the variable of security, and the influence of uniformity and color temperature can be ignored. At the same time, there is correlation between color temperature and uniformity.
The model is adjusted according to the hypothetical model result until the best fitting model result is obtained. The best fitting results were obtained after several revisions of the model (
Figure 15) (
adjust4.inp in Supplementary Materials).
The results of the comfort evaluation model can be expressed separately as:
where, parameters are the same as in the model diagram.
The final SEM for comfort is:
or:
where, parameters are the same as in the model diagram.
According to the sample analysis results of the structural equation model analysis, when the public comfort level cannot be directly obtained, the comfort level of the study area can be obtained through other observable variables and corresponding model parameters.
The results (
Figure 15,
Table 4 and Equation (10)) show that the feeling of security and the observable variables affect the comfort of residents. Among them, the color temperature affects the feeling of security, and the feeling of security affects the feeling of comfort. Among the parameters of the comfort model results, C1, sec, and tem have absolute weight advantages in the model. C1 is the evaluation score of residents on light illumination, sec is the evaluation score of residents on feeling of security, and tem is the evaluation score of residents on color temperature. The results show that the residents’ evaluation results have a great impact on the final comfort model, which shows that the model can reflect the residents’ feeling of comfort in the public space lighting environment.
Based on the results of comfort model, the public space lighting comfort of four typical residential areas was calculated, respectively (
Figure 16).
It can be seen from the comfort results of four typical residential areas that Chongwenmenwai street and Jinrongjie street are more comfortable than Beitaipingzhuang street and Chunshu street, indicating that the areas with high light intensity are more comfortable than the areas with low light intensity.
In order to further explore the feedback of the luminous remote sensing data from the measured lighting data of the four routes and the public perception survey data, this paper used a multi-factor analysis method to evaluate the lighting score on each route separately. The selected factors were illuminance (Ev) in the measurement data, x color temperature (x), y color temperature (y), glare index and the six items of public perception survey data. The six items are illumination, color temperature, uniformity, glare perception, feeling of security, feeling of comfort.
The data used passed the KMO (Kaiser–Meyer–Olkin) and Bartlett significance test, and then the variance of the original variables explained by each component in the principal component analysis was used to determine the number of common factors, and the factors with a large cumulative variance contribution rate were retained (
Table 5).
The component score coefficient matrix was obtained to express the relationship between each index variable and the extracted common factors (
Table 6). If the score on a common factor is higher, it means that the relationship between the index and the common factor is closer.
Through the interpretation of
Table 5 and
Table 6, the following results can be drawn:
(a) The nighttime light image reflects that Chongwenmenwai street has high red light. The light information of this route is mainly concentrated on the first principal component. In its public space lighting evaluation score, the larger weight is the illumination perception, uniformity perception, security and comfort in the public perception data, and the weights of these four perception variables are almost the same. The second is glare index and glare perception, and the smallest weight is color temperature. This shows that in an area with high light intensity, people’s perception of the public space lighting environment is the main part of the lighting evaluation in this area. People pay more attention to the feeling brought by the lighting environment in this area, but they pay less attention to whether the color temperature is mild or not. This shows that the comprehensive lighting conditions in this area are relatively good, the light is relatively uniform, and people have a good feeling of security and comfort.
(b) The nighttime light remote sensing image reflects that the blue light intensity of Jinrongjie street is very high. The lighting comprehensive evaluation score of the route in this area is composed of three principal components. Among them, the weight of glare perception and glare index is the largest, indicating that the residents have a high degree of feedback on the dazzling degree of the route in this area, which greatly affects the comprehensive score of this area. However, due to the high intensity of blue light, the glare index and glare perception greatly affect the lighting evaluation of the route. The color temperature is still the lowest in the comprehensive score of this area.
(c) The nighttime light image reflects that Beitaipingzhuang street has low blue light. Due to the lack of light, residents pay high attention to the color temperature, which shows that the degree of cold and warm is the key to the comprehensive lighting score of the area where the lighting is insufficient, and the change of color temperature is particularly prominent in this case. Moreover, the weights of feeling of comfort and security are reduced relatively. This shows that in an area with low light, people need more basic lighting than security and comfort. In this case, residents pay more attention to whether the public space lighting can meet the normal lighting needs of pedestrians.
(d) The nighttime light image reflects that Chunshu street has low red light. The weights of glare index and glare perception are relatively high in this route. Followed by the weights of illumination, uniformity, security and comfort. This shows that the illumination degree of this area is not bad. However, people can feel a certain degree of glare. Therefore, on the basis of meeting the basic lighting conditions, the glare problem in this area should be solved first.
Through the analysis, it can be concluded that the performance of the lighting phenomenon reflected by each typical residential area is consistent with that reflected by the luminous remote sensing image within the allowable error range. When people live in the outdoor public space light environment with good lighting conditions, they will pay more attention to the feelings of light uniformity, comfort and security. Whereas when people live in the outdoor public space light environment with poor lighting conditions, the feelings of color temperature and glare are more obvious. In addition, the blue light radiation in the area corresponding to the dazzling glare is obvious. The blue light with short wavelength not only makes people feel visual discomfort, but also does great harm to people’s physical and mental health. From the above concerns, we can further discuss the impact of blue light pollution in the future research.