Research on the Difference Between Recreational Walking and Transport Walking Among the Elderly in Mega Cities With Different Density Zones: The Case of Guangzhou City

Walking is the easiest method of physical activity for older people, and current research has demonstrated that the built environment is differently associated with recreational and transport walking. This study modelled the environmental characteristics of three different building density zones in Guangzhou city at low, medium, and high densities, and examined the differences in walking among older people in the three zones. The International Physical Activity Questionnaire (IPAQ) was used to investigate the recreational and transport walking time of older people aged 65 years and above for the past week, for a total of three density zones (N = 597) and was analysed as a dependent variable. Geographic Information Systems (GIS) was used to identify 300, 500, 800, and 1,000 m buffers and to assess differences between recreational and transport walking in terms of the built environment [e.g., land-use mix, street connectivity, Normalised Difference Vegetation Index (NDVI) data]. The data were processed and validated using the SPSS software to calculate Pearson's correlation models and stepwise regression models between recreation and transit walking and the built environment. The results found that land use mix and NDVI were positively correlated with transport walking in low-density areas and that transport walking was negatively correlated with roadway mediated centrality (BtE) and Point-of-Interest (PoI) density. Moreover, recreational walking in medium density areas was negatively correlated with self-rated health, road intersection density, and PoI density while positively correlated with educational attainment, population density, land use mix, street connectivity, PoIs density, and NDVI. Transport walking was negatively correlated with land-use mix, number of road crossings while positively correlated with commercial PoI density. Street connectivity, road intersection density, DNVI, and recreational walking in high-density areas showed negative correlations. Moreover, the built environment of older people in Guangzhou differed between recreational and transport walking at different densities. The richness of PoIs has different effects on different types of walking.


BACKGROUND Ageing Trends
The problem of population ageing is becoming more and more serious worldwide, and the World Population Prospects 2019 programme predicts that the elderly population in China will reach 300 million by 2025, with the ageing rate exceeding 20% and entering a deeply ageing society. In 2019, China issued the National Medium and Long-term Plan for Actively Coping with Population Ageing and proposed the construction of a "high-quality service and product supply service system for the elderly, " with the home as the basis and the community as the backbone. During 2021, they saw the authorisation of the Xinhua News Agency to release the 14th Five-Year Plan (1), which clearly states in Chapter 45, Section 3, "[i]mproving the elderly care service system, " that the system of elderly care at home should be improved, the urban environment should be adapted to ageing, and the community elderly care system should be improved. According to the statistics released by the Guangzhou Municipal Health Commission (GMHC), by the end of 2019, the population aged 60 and above in Guangzhou was 1,755,100, with the ageing population accounting for 18.40% of the total household population GMHC (2). Guangzhou has already entered an ageing society, and the health conflicts caused by ageing are increasing. Therefore, how to improve the participation and level of daily physical activity of the elderly to maintain their physical health has become an important topic of research and a hot topic in many fields at home and abroad.

Benefits of Walking
Inactive behaviours such as prolonged sitting are the cause of a high incidence of chronic diseases that threaten the health of the people (3,4). Studies have shown that as the body ages, physical performance decreases, with maximum oxygen uptake decreasing by 8-16% for every decade after the age of 30 (5). Muscle strength decreases by 10-15% every year. Moderate physical activity has a positive impact on the health of older people, and physical activity is beneficial in reducing the incidence of chronic diseases and some sudden illnesses (6). Based on the results of epidemiological studies, it has been demonstrated that walking has health benefits for older people, with overall results similar to those of physical exercise.
In addition, a study has shown that an increase in walking time increases the maximum oxygen uptake of older people (7). More walking promotes cardiovascular health and may prevent associated cardiovascular chronic diseases. Walking increases aerobic capacity, reduces Body Mass Index (BMI) and body fat, and gradually improves diastolic blood pressure caused by prolonged sedentary time (8).
Overall, research has shown that regular walking not only has a positive impact on body fat and cardiorespiratory fitness, but that walking could also improve physical strength and flexibility. Given that declines in strength and flexibility could affect the independence of older people, this finding suggested that reasonably regular walking can help older people maintain their independence in life.

Built Environment and Walking
Research on walking environments increasingly shows that walking in residential settings is influenced not only by population density and the social characteristics of the population (such as age, gender, and income) but also by the built environment around an own residence (9)(10)(11)(12)(13). Previous research has demonstrated that good accessibility to usable green space is positively associated with walking behaviour and that better green space environments motivate older people to walk (12,14,15). Meanwhile, recent studies show that street greening has a positive effect on walking propensity and time (16). Factors of the built environment include land use diversity, street connectivity, building density, and accessibility (17).
However, there are some studies that show inconsistent results for the correlation between factors of the environment and walking. The study of Thornton et al. (18) found a positive correlation between land use mix and walking, while the work of Zhang et al. (19) found a negative correlation, with inconsistent results likely due to differences in objective perceptions of the study sample (10). Other later studies have used Geographic Information Systems (ArcGIS) to analyse the sample sites by delineating a specific radius of 200 to 2,000 m with residents as the central point for environmental analysis, which more objectively evaluated environmental factors, and re-did the study between the subdivided environment and walking, obtaining results that were more different than those from the previous non-subdivided environment (20)(21)(22). Some research has explored the application of environmental relevance (e.g., land use mix, street connectivity) of autonomous openly provided geographic information (e.g., open street maps), but the accuracy and reliability of their geographic information still need to be validated (23,24).

Transport Walking and Recreational Walking
The impact of the built environment on walking also varies depending on the purpose of walking and can be broadly divided into two types of walking, namely, transport walks motivated by walking to a destination (e.g., to work, to an event), and recreational walks specifically for exercise or to a recreational area such as a park., both of which are physical activities beneficial to the health of an individual (10,11,25). Although there are some inconsistencies in the theoretical results, research suggests that transport walking is primarily promoted through road density, land use, and accessibility to public transport, while recreational walking is related to community neighbourhoods, community pavements, and the ease of access and number of entrances to parks and activity venues (10,26). The walking behaviour of older people in high-density urban settings also varies depending on the time of day. Fewer studies have examined differences in the relationship between environmental and recreational and transport walking among older people in Guangzhou, China. To fill these research gaps, this study investigated how different densities of the built environment affected recreational walking and transport walking in Guangzhou. As an old first-tier city in China, Guangzhou has a complex and diverse urban fabric, with a high-density old city, a low-density new city, and other medium-density areas. Our research questions were as follows: 1. How does the built environment differ in association with different types of walking at different densities in the same city? 2. Are differences in the built environment and demographic characteristics associated with differences in transport walking times and recreational walking times?

Research Design and Study Population
This study was based on a survey of parks, selected to represent the surrounding community based on the different population density zoning of the Guangzhou city plan, 1 , where density zones 1, 2, and 3 corresponded to the low, medium, and highdensity zones of this article. The areas were sampled according to the price of the area in March 2021 divided into low Socio-Economic Status (SES) and high SES, with areas below 30,000 RMB/m 2 being low SES and areas >30,000 RMB/m 2 being high SES, divided into six groups in total. We considered people aged 65-74, 74-84, 85, and over. Respondents were surveyed in the spring to avoid seasonal effects. The survey considered demographic characteristics (e.g., income, age, education) and travel information (e.g., broad questions related to the mode of transport, travel time). In order to maximise the sample size, 600 data were pooled and after excluding incomplete data records, our representative sample of older people in Guangzhou was 597 respondents in total (see Figure 1 below). Our preliminary research of the 12 sample areas that we studied has revealed that the built environment varied significantly between the samples at different population densities in Guangzhou, so much so that during the initial modelling process it was found that the data for the built environment in the low-density areas significantly reduced the data for the more distinctive features of the built environment in the medium-and high-density areas. In order to prevent the model from overlooking some environmental features in its calculations, the low, medium, and high-density areas were modelled and calculated separately to ensure sufficient objectivity. A summary of the selected cases was shown in Table 1.

Walking Behaviour
The International Physical Activity Questionnaire-Long Version (IPAQ-LC) was used to group transport walking and recreational walking in the questionnaire based on the questionnaire and the purpose of travel from the interviews. The total daily duration of these two types of walking in minutes per person was used as the dependent variable. Additionally, areas outside the residential environment (>1,000 m) were excluded.

Built Environment
In order to simulate the walking environment of older people, we applied buffers generated at the scale of the residential neighbourhood of the respondents. To incorporate the current walking environment, buffers of 300, 500, 800, and 1,000 m radius were used and compared (20,27 (32), we categorized the total land-use types in terms of the actual use by older people into seven categories, namely residential, recreational, commercial, educational, medical, and service. An index was calculated based on the proportion of area in each land use category. Its value ranges between 0 and 1, with higher values implying greater diversity. 3. Street connectivity was related to the design of the street layout (27). Higher street connectivity usually leads to more walking, while the opposite is true for more cul-de-sacs (10). Intersection data for this study were obtained from Sky Map data China (33) and calculated through GIS software. 4. BtE (mediated centrality) and MED (proximity centrality) were calculated in the software Spatial Design Network Analysis (sDNA 2 ) for 300, 500, 800, and 1,000 m as control variables.

Control Variables
We adjusted for several demographic and socio-economic characteristics of the respondents. Age was divided into three categories, which were 65-74, 75-84, and >85. Education level was divided into low (none), medium (primary education), and high (secondary and above). Self-rated health status was categorised as very good, fair, and bad.

Statistical Analysis
Firstly, in order to ensure that the grouped studies were meaningful, a test of variance between groups was required, using SPSS to test for variance between the independent variables, as statistically significant between the independent variables under the three categories of the low, medium, and high (see Table 2 below). Descriptive statistics summary data. We checked for covariance between multiple different environmental variables using SPSS correlation tests and demonstrated no covariance in the data by VIF < 5. Walking time for recreational and transport was used as the dependent variable in our stepwise regression analysis (35,36). Since some respondents reported only partial walking time, in correlation tests, we assessed the association between walking and environmental variables with buffer sizes of 300, 500, 800, and 1,000 m. The significance of each variable was assessed based on 95% CIs. Table 3 presents 597 valid questionnaires after removing invalid questionnaires, of which 86.6% were aged 65-74, 11.89% were aged 75-84 and 1.51% were aged 85 or above. The educational level of those with higher education was 8.71%, those with secondary education was 79.06%, and those with no education were 12.23%. Self-rated health was excellent at 20.44%, good at 75.54%, and bad at 4.02%. Respondents spent more total time walking for recreation than for transport, street connectivity gradually decreased with increasing density of the sample cities, road intersection density gradually increased with increasing urban density, the number of bus and metro stations increased step by step with increasing density of the sample cities, and NDVI decreased step by step with increasing density of the sample cities. Table 4 shows the analysis between the independent variables in the low-density areas, wherein there was a correlation between the built environment characteristics and transport walking in the low-density areas, while land-use mix, number of bus stops, distance to the nearest bus stop, and NDVI were positively correlated with transport walking. Finally, transport walking was positively correlated with MED 300 m, BtE 300 m, BtE 1,000 m,       4 also shows the analysis between independent variables in high-density areas, wherein there was a correlation between environmental characteristics recreational walking in high-density areas, with street connectivity, road intersection density, MED 300 m, number of bus stops, and DNVI showing a negative correlation with recreational walking. There was a positive correlation with the distance to the nearest bus stop. Finally, no significant correlations were found between transport walking and built environment features in high-density areas.

Regression Analysis
There were also significant differences in the correlation between environmental architectural features and recreational walking in high-density buffer zones of different sizes, with street connectivity (R = −0.219 * * ; P = 0.009, R = −0.251 * * ; P = 0.003) being negatively correlated with recreational walking in the 300 and 500 m buffers, while street connectivity (R = 0.187 * ; P = 0.028) was positively correlated in 1,000 m. The correlation was positive in the 1,000 m buffer and did not show a correlation in the 300 m buffer. Road intersection density (R = −0.271 * * ; P = 0.001, R = −0.266 * * ; P = 0.006) was negatively correlated with recreational walking in the 500 and 1,000 m buffers, but not in the 300 and 800 m buffers. The number of bus stops (R = −0.201 * ; P = 0.018) was negatively correlated with recreational walking within the 300 m buffer while no negative correlation was observed within the 500, 800, and 1,000 m buffers. Table 5 summarises the stepwise regression results for recreational walking and transport walking at low, medium, and high densities. In particular, at low densities, there was no correlation between recreational walking and built and natural environment features, and within the 300, 500, and 1,000 m buffers, transport walking was negatively influenced by a regression value of −0.197 * (t = −2.450; p = 0.015 < 0.05) with MED 300 m, and the number of buses stops regression value of.195 * (t = 2.429; p = 0.016 < 0.05) had a positive influence relationship with transport walking.

Regression Analysis
For the medium density areas, the BtE 500 m regression coefficient of −0.312 * * (t = −4.122; p = 0.000 < 0.01) for the 500, 800, and 1,000 m buffers and the regression coefficient of −0.191 * (t = −2.522; p = 0.013 < 0.05) for the health self-report within all buffers were negatively associated with recreational walking. The MED 300 m regression coefficient of.167 * (t = 2.209; p = 0.028 < 0.05) within the 500, 800, and 1,000 m buffers had a positive impact relationship with transport walking. The Commercial PoIs regression coefficient of −0.182 * (t = −2.410; p = 0.017 < 0.05) within the 300 m buffer had a negative relationship with transport walking.
Transport walking at high densities was not correlated with built and natural environment features, and recreational walking was negatively associated with NDVI regression values of −0.237 * * (t = −2.851; p = 0.005 < 0.01) within the 300 m buffer, intersection density regression values of −0.271 * * (t = −3.299; p = 0.001 < 0.01) and −0.251 * * (t = −3.036; p = 0.003 < 0.01) for the regression value of street connectivity within the 800 m buffer had a negative effect.

Key Findings
This study examined the correlation and consistency between built and natural environment features and walking in megacities at different density buffers. Previous studies have demonstrated a positive correlation between land-use diversity and transport walking (18,31), however in this study the situation was different at different densities, with land-use diversity only positively correlated with transport walking at low densities, land use diversity not encouraging more transport walking at medium densities but weakening, and no correlation at higher densities. At medium densities, land-use mix encouraged more recreational walking, but at low and high densities there was no such relationship. This phenomenon might be due to the fact that low-density areas are sparsely populated with a land-use mix that is most attractive for transit walking, while medium-density areas have a higher distribution of all types of PoIs compared with low-density areas resulting in a land-use diversity that is not more attractive to older people for transit walking and less demand for transit trips. However, the abundance of PoIs in medium-density results in a land-use diversity that is more attractive to seniors for recreational walking. The reason why high-density land-use diversity does not correlate with walking may be that high-density PoIs are dense enough that seniors tend to be able to reach the functional areas they need to meet their daily needs on the ground floor of their own residence. Therefore, there was no correlation with walking. The finding that NDVI was positively associated with transport walking at low densities and with recreational walking at medium densities was consistent with previous studies, but the finding that NDVI was negatively associated with recreational walking at high densities differed from previous studies (37)(38)(39). This phenomenon could be explained by the fact that recreation sites (e.g., pocket parks) with poorer environments lacking shelter space in high-density areas are located in areas with higher levels of vegetation, resulting in less recreational walking (40). We found that street connectivity and road crossing density were negatively correlated with recreational walking in small buffer sizes at high densities but positively correlated with medium density areas at high densities with buffers of 1,000 m, which is not consistent with previous studies (18). This could be explained by the fact that higher street connectivity in small areas at high densities may reduce the perceptions of safety of people or due to individual differences in perceptions of the neighbourhood, and secondly due to the fact that people are reluctant to walk in unsafe and uncomfortable areas due to the density of motorways (21,41,42).
We added MED and BtE to the independent variables of the built environment, where MED tended to respond to the importance of the global location and BtE to the centre of gravity of a small area, i.e., the density of the network. The results showed that at low densities MED 300, BtE 300, and BtE 1,000 m were all negatively correlated with transport walking. This phenomenon could be explained by the fact that at low densities the road network was relatively sparse, and the most important global zones were in residential areas, the areas with the highest network densities were adjacent to roads where the elderly lack a sense of security and therefore could lead to less transport walking. However, at medium densities, the MED 300 m was positively correlated with transport, which could be explained by the fact that the most important sites in the global location overlap with commercial sites, thus promoting transport walking for older people. Furthermore. 300 m BtE and 1,000 m BtE were negatively correlated with transport walking, as is the case at low densities where the network density was also adjacent to a motorway, resulting in a negative impact on transport walking. At higher densities, MED 300 m was negatively correlated with recreational walking. The high-density core areas were not the type of parks that seniors would choose for recreation, resulting in the MED having a negative effect on recreational walking for seniors. Although the indices of the built environment at different densities were relatively consistent across buffer sizes, we found that the correlation for some environmental factors varied by geographic scale, which was consistent with the findings of Etman et al. (43) and Zang et al. (15). For example, land use diversity was positively correlated with transport walking in the low-density 300 m buffer zone, while no correlation was found at 500, 800, and 1,000 m. In contrast, in the medium-density zone at 500, 800, and 1,000 m, there was no correlation with transport walking. In contrast, transport walking was negatively correlated with land-use diversity in the medium density areas at 500, 800, and 1,000 m buffers, with no correlation found in the smaller 300 m buffers. No correlation between transport walking and land use diversity was found in any of the buffers in the high-density areas. This suggested that older people preferred to transport walking in areas of moderate density and that there was little need to walk for transport when densities were high enough. Furthermore, for other reasons such as time constraints and physical limitations, seniors chose not to walk long distances for transportation. In low-density areas, 500 and 1,000 m buffer zones transit walking was positively correlated with the number of bus stops, and 300 and 500 m buffer zones transit walking was positively correlated with the closest distance to a bus stop. Recreational walking in the medium density 500 m buffer was negatively correlated with the number of bus stops. High density 300 m buffers had a negative correlation between recreational walking and the number of bus stops, but all buffers of the same size have a positive correlation with the closest distance to a bus stop. The number of metro stops and the distance to the nearest metro stop were not correlated with walking in the low-and high-density areas, while the number of metro stops and the distance to the nearest metro stop were positively correlated with recreational walking in the mediumdensity areas for all buffer sizes. The remaining density areas transit stations and walking were not correlated across buffer sizes. In the lower density areas, the number of metro stations was low and the demand for transport was mainly by bus, which was consistent with the current study. At medium densities, the metro coverage was already more comprehensive and due to time constraints, the elderly rarely chose public transport as a transport option so no correlation was presented. At high densities, due to the fact that the density was sufficiently high, even though the metro network coverage was dense enough, it did not affect the mode of travel of the elderly, and the bus was more convenient than the metro, and it was only a short walk to the entertainment venues. The results of this study were currently different (44).

Advantages and Limitations
This study has several strengths. Firstly, we used the most recent geographical sample of Guangzhou as a reference, the questionnaire and geographical data were collected more recently and are more up to date. Secondly, unlike previous studies on environmental perception (41,45), we used GIS for more objective measurements and calculations of the built environment. Thirdly we distinguish between multiple buffersized environmental scales at different densities to examine the impact of multiple geographical scales on recreational walking and transport walking, a study that has rarely been used before.
There were also some shortcomings in that the volume of our data could have been further improved, and there was a lack of additional data on human subjects to minimise the objective effects of respondents. Secondly, our current study was limited to a walking study of older people in Guangzhou, Guangdong Province, China (a mega-city with extremely high building and population density) and some of the data and findings might not be transferable to other countries. Thirdly, due to the lack of GPS walking data, walking areas could only be delineated by buffer size. This could lead to an overestimation of walking times for older people. Fourthly, due to the lack of streetscape data for low-density neighbourhoods, we used the latest NDVI data and tried to reduce the amount of error by increasing the resolution of the NDVI to 10 m to ensure accuracy, but some unusable green space was still accounted for. This was expected to be addressed by the subsequent addition of street view data. Finally, all streets in Guangzhou were walkable, but there were a few expressways that were included, which we believe would have minimal impact on our model.

CONCLUSIONS
In this study, we examined the differences between the residential environments of older people in Guangzhou at different densities in relation to recreational and transport walking, using a more comprehensive sample of data that provides credible evidence that low, medium, and high densities are different from each other. The environmental characteristics of the medium density areas correlated with both recreational walking and transport walking, while high and low densities did not. The correlation between various environmental characteristics (e.g., land use diversity) and walking in different buffer zones in medium density areas was also significant, indicating that medium density environment might play a more significant role in affecting physical activity, including walking among older residents than other densities. Our findings might be of better help to cities and older people for studies in cities of different densities in the future.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
PZ: conceptualisation. SM and FX: resources. XZ and YZ: supervision. HQ: validation. PZ and HQ: writing-original draft. PZ: writing-review and editing. All authors contributed to the article and approved the submitted version.