Analyses of Multiple Factors Affecting Residents′ Walking to rail Transit Stations in Los Angeles Metropolitan， United States

XU Junping

( College of Architecture， Huaqiao University， Xiamen 361021， China )

Analyses of Multiple Factors Affecting Residents′ Walking to rail Transit Stations in Los Angeles Metropolitan， United States

XU Junping

( College of Architecture， Huaqiao University， Xiamen 361021， China )

Taking Los Angeles Metropolitan, United States as the case study, depending on the data from the regional household travel survey conducted during 2011-2013 by the Southern California Association of Governments， the logistic regression models is used to find significant factors that affect residents′ walking to the rail transit stations. Results show that the distance to stations， the continuity of sidewalks， density of street lights， density of street trees， station parking and land use mix are the significant environment factors; meanwhile， the travel destinations， household income， the number of household vehicles and ethnicity are also significantly factors influencing residents′ walking to rail transit stations.

walking； rail transit oriented development； metropolitan； multiple factors analysis； logistic regression models； Los Angeles City

Car dependence can have detrimental effects on the environment and public health， such as increasing green house gas (GHG) emissions， traffic congestion， oil price vulnerability， and physical inactivity[1-2]. Reliant on public transit is one of the success policies toreduce car travel and car dependence[3]. Public transit is generallynot a point-to-point mode of travel， which may incorporate regular physical activities into daily life. A large body of cross-sectional studies found that transit users have higher levels of walking compared to those who do not use transit[4-7]. Numerous studies have found that people are more likely to walk in the neighborhoods with certain environmental characteris tics， especially for transportation purpose[6-9]. Walkable neighborhoods are often characterized by medium-to-high population density， a mix of land uses， high connectivity， and presence of pedestrian in frastructure[10-14].

With more concerns on the transit orient development (TOD)， walking to transit also get more attention than before[15-19]. There were very few researches of examining walking to transit， and most of them have involved the similar conceptual models as other travel behavior research[20-23]. To fill in the research gap， this study would introduce groups of predictors， including socioeconomic factors of station areas， built environment factors of station areas and socio-demographic factors of individuals， as well as other factors to predict walking to transit.Meanwhile， not as most previous studies using subjective recall questionnaires， this study employed the travel data that were collected through time diaries， which could avoid recall bias and social-desirability bias[24-25]. The results from this study would provide meaningful suggestions for future TOD practice in metropolitan areas not only limited in the North America， but for worldwide.

Fig.1 Metro rail system in Los Angeles City圖1 洛杉磯市地鐵軌道系統(tǒng)

1 Research Design

1.1 Study Area

The study area is the city of Los Angeles， which is the most populous city in the state of California and the second largest city in the United States with a population of 3 792 621 from the 2010 census[26]. Based on the number of daily riders， the city′s subway system is the ninth busiest in the United States and its light rail system is the country′s second busiest[22]. The rail system includes the subway lines (red and purple) and the light rail lines (gold， blue， expo， and green)[22](Figure 1).

As one of the most economically and ethnically diverse regions in the country， Los Angeles′s transit station areas encompass a wide range of demographic， physical， and economic characteristics[16]. The transit network of Los Angeles City extends to various neighborhoods with different household income levels， different rates of car ownership and diverse ethnic populations[16]. Table 1 illustrates the demographic characteristics among regions， cities and transit station areas (half-mile buffers of stations). (Source:CenterforTransit-OrientedDevelopment， 2011). It indicates that households with lower incomes and lower rates of car ownership tend to live closer to transit stations and take more transit trips or other non-motorized trips than other households.

Tab.1 Regional， city and station area demographic characteristics of Los Angeles City in 2010表1 2010年洛杉磯市的區(qū)域、城市和車(chē)站地區(qū)的人口特征

1.2 Unit of Analysis and Data Source

Most previous researches used 400 meters (0.25 miles) or 800 meters (0.5 miles) as the walking distance to rapid transit stations， which means that the unit of analysis is often centered by the station with 400 meters or 800 meters as the radius[8，21，23，27]. Based on the literature review and the characteristics of data source， this research defines 400 meters (0.25 mile) radius buffer centered by each station as the spatial unit of analysis.

The socioeconomic variables and socio-demographic variables were obtained through regional household travel survey conducted from 2011 to 2013 by the Southern California Association of Governments (SCAG) survey and census data. The other variable was achieved through SCAG survey. The built environment variables were all objective ones and measured using geography information system (GIS)， which were gotten through multiple data sources， including Los Angeles County GIS portal， Los Angeles County sheriff， City of Los Angeles Department of Transportation (LADOT)， and U.S. Geological Survey (USGS). The network analyst tool would be employed to measure the connectivity of streets； the proximity tool (buffer) and extract tool (clip) extract the attributes in 400 meter/quarter mile buffers； and the summarize function in the attribute table get the results we need.

1.3 Research Method

There are four groups of predictors in the analysis， including built environment attributes (both continuous and categorical variables)， individuals′ socio-demographic attributes (both continuous and categorical variables)， socioeconomic attributes of station areas (continuous variables) and other variable (travel destination) (categorical variable). Here， one group of predictors was added in the new model in a stepwise approach and finally four logistic regression models were produced. The models (1)， (2)， (3) and (4) would be stated as follow.

(1)

(2)

(3)

(4)

In here，Nmean walking to transit，Ameans other variable (travel destination)，Bmeans built environment variables，Cmeans socioeconomic variables of units of spatial analysis，Smeans socio-demographic variables of individuals，μ=regression error term.

The first model only has other variable， the second model has both other variable and socioeconomic attributes of station areas， the third one has three groups of predictors while adding the socio-demographic factors of individuals in， and the final model adds the group of built environment predictors. In the final model， walking to transit is regressed on four groups of independent variables.

2 Data Analysis and Results

2.1 Descriptive Analysis

Total number of 745 individuals′ records in the Southern California Association of Governments (SCAG) household travel survey are completed and valid for this research and total 55 transit stations are involved for the records above. Descriptive analysis is performed for the spatial unit of analysis (400 meters distance from the rail stations). Mean and standard deviation (SD) are calculated for the 20 independent variables and they are displayed in Table 2. In table 2， “(c)” means categorical variables，N=745 is total number of individuals，N=55 is transit stations involved.

2.2 Results

To determine the significant factors that impact the walking behavior to transit stations， four binary logistic regression models were employed to do the analysis (see table 3). In table 3， “(c)”

Tab.2 Descriptive statistics of independent variables表2 自變量的描述性統(tǒng)計(jì)表

means categorical variables， OR representing odd ratio， Coeff. representing coefficient. The first model only has travel destination as the predictor， which is significant to predict the walking behavior to stations. Traveling to utilitarian destinations decreased the likelihood of walking to stations by

0.260 times compared with traveling to recreational destinations. Traveling destination maintained statistical significance in all of the four models.

The socioeconomic variables of station areas include black percentage， Hispanic percentage and median household income. While adding the socioeconomic variables of station areas in the second model， none of them were significant. The median household income variable turned into significance in model 3 and one level increased in the median household income would increase the likelihood of walking to stations by 1.313 times. It became more significant in the final model， with a one level increasing the median household income increasing the likelihood of walking to stations by 1.636 times. However， the percentage of black and percentage of Hispanic was not significant in the following models.

Tab.3 Results of four logistic regression models predicting walking to transit stations表3 4個(gè)邏輯回歸模型預(yù)測(cè)步行到站點(diǎn)的結(jié)果

*:P<0.05;**:P<0.01;***:P<0.001.

In model 3， vehicle number of household， household income， and ethnicity are the significant indicators to impact walking behavior to stations. Here the ethnicity was a dummy variable (white=1).

While one vehicle increased in the household， the likelihood of the individual walking to stations decreased by 0.714 times. The availability of cars in household had been tested as an important variable for encouraging driving and decreasing walking in early studies.

There are total six significant built environment factors to predict walking to stations. The distance and percentage of sidewalk completeness were the two most significant ones. The distance is the spatial distance from the departure origin to the station destination and the unit in this analysis is 100 feet. With one hundred feet increasing distance， it decreased the likelihood of walking to stations by 0.922 times. While one percentage increased in the sidewalk completeness， the likelihood of walking to stations increased by 1.020 times. Consistent with previous findings， the availability of sidewalks to stations decided the possibility of walking to stations.

The street lights density， trees coverage density， transit station parking and land use mix were other four built environment factors that impact the walking to stations significantly. Street lights are essential street facilities for the safety of walkers at night and trees shade is essential for walking in summer. While adding one street light per mile， the likelihood of walking to stations could increase by 1.028 times. While adding one street tree per mile， the likelihood of walking to stations could increase by 1.007 times. Land use mix was reported as a critical indicator in a great number of previous studies for encouraging walking， the same findings in this study. Every 0.1 increase in the land use mix index (0-1)， it increased the likelihood of walking to stations by 1.145 times. Transit station parking was indicated as a significant negative indicator in early researches and it is also a negative significant factor in this analysis. The stations with parking would decrease the likelihood of walking to stations by

0.588 times compared with the stations without parking.

Generally， under model summary， -2 log likelihood statistic measures how poorly the model predicts the decisions， the smaller the value the better the model. In model 1， -2 log likelihood statistics is 942.61， and it decreased in model 2 (916.402) after adding socioeconomic factors of station areas. It continually decreased in model 3 (859.704) while adding socio-demographic variables of individuals. When added the built environment attributes in model 4， the -2 Log Likelihood decreased to 751.809. It is obvious that the models are continually improving the predictive power for the dependent variable.

The maximum value of NagelkerkeR-square is equal to 1.0. Overall， high values are better than low values， higher values suggesting that the model fits increasingly well. In model 1， NagelkerkeR-square is 0.081， which means that 8.1% of the variation in dependent variable (walking to stations) could be explained by travel destination. In model 2， NagelkerkeR-square is increasing to 0.125， which means that after adding in socioeconomic predictors of station areas， the variations of dependent variable (walking to stations) could be explained 12.5% by the model 2 and increased 4.4% compared with model 1. The NagelkerkeR-square in model 3 is 0.216， which explained 21.6% of the variations of dependent variable (walking to stations) after adding socio-demographic factors of individuals and increased 9.1% compared with model 2. In the final model (model 4)， NagelkerkeR-square is 0.370. The final model incorporated built environment predictors in and explained 37% variation of the dependent variable， which increased 15.4% compared with model 3.

3 Discussions and Conclusions

3.1 Limitations of This Study

The survey population for the present survey was households with telephones in the Southern California Association of Governments (SCAG) region； however， Census 2010 data indicates that

1.6% of occupied housing units in the SCAG region are without telephones. This survey has conducted through phone， thus some potential respondents were ignored. Meanwhile，the overall response rate was low， only 25 percent， which is primarily due to the complex of interview processes. An important determinant of data quality is the accuracy of the reported trips. To enhance reporting accuracy，this survey relied on diary instruments in which respondents are asked to record each trip for a specific time period (e.g.， 24-hours， 48-hours)， however， the accuracy of the records are case by case. The NagelkerkeR-square of final model is 0.37， which means that the model can explain 37% variation of the dependent variable. The value is not so high due to other reasons， such as self-selection of residents， which do not matter if they have walkable environment but their preferences.

3.2 Conclusion

The findings of this study indicate that the built environment of station areas has significantly impact on residents′ walking to transit. Improving the pedestrian environment of station areas could increase the likelihood of walking to transit， such as increasing sidewalk completeness to make walking possible， adding more street lights for walking safely at night， adding more street trees for walking comfortably in summer， increasing mixed land use for convenient shopping and decreasing parking lots around stations to avoid driving. These findings would be the potential suggestions for policy makers to enhance transit oriented development in future. This research highlights not only built environment indicators， but emphasizes that some variables of socioeconomic characteristics of station areas and socio-demographic variables of individuals also influence walking to transit. It is interesting to find that the households with higher income would have less opportunity walking to stations due to owning the cars， however， the station areas with higher average household income would have more walkable environment. Although the households with high income intend to live in a livable neighborhood， most of them still prefer to using a car instead of walking. Thus， self-selection is very important for individuals if they can afford cars， and the walkable environment is not sufficient for them to choose walking to transit. There need more policies to encourage walking plus taking transit， such as economic incentives.

[1] FRANK L D，SCHMID T L，SALLIS JF，etal.Linking objectively measured physical activity with objectively measured urban form: Findings from SMARTRAQ[J].American Journal of Preventive Medicine，2005，28(Suppl 2)：117-125.

[2] SALLIS J F.Measuring physical activity environments： A brief history[J].American Journal of Preventive Medicine，2009，36(4)：S86-S92.

[3] SALLIS J F，F(xiàn)RANK L D，SAELENS B E，etal.Active transportation and physical activity： Opportunities for collaboration on transportation and public opportunities health research[J].Transportation Research Part A: Policy and Practice，2004，38(4)：249-268.

[4] BOARNET M，CRANE R.Travel by design: The influence of urban form on travel[M].New York: Oxford University Press，2001.

[5] BROWN B B，WERNER C M.A new rail stop-tracking moderate physical activity bouts and ridership[J].American Journal of Preventive Medicine，2007，33(4)：306-309.

[6] CERVERO R B.Walk-and-ride: Factors influencing pedestrian access to transit[J].Journal of Public Transportation，2001，3(4)：1-23.

[7] MACDONALD J M，STOKES R J，COHEN D A，etal.The effect of light rail transit on body mass index and physical activity[J].American Journal of Preventive Medicine，2010，39(2)：105-112.

[8] DURAND C P，TANG X，GABRIEL K P，etal.The association of trip distance with walking to reach public transit: Data from the California household travel survey[J].Journal of Transport and Health，2016，3(2)：154-160.

[9] EWING R，CERVERO R.Travel and the built environment: A synthesis[J].Transportation Research Record，2001，1780(1)：87-114.

[10] EWING R，CERVERO R.Travel and the built environment[J].J Am Plan Assoc，2010，76(3)：265-294.

[11] FRANK L D，ENGELKE PO.The built environment and human activity patterns： Exploring the impacts of urban form on public health[J].Journal of Planning Literature，2001，16(2)：202-218.

[12] HANDY S L，BOARNET M G，EWING R，etal.How the built environment affects physical activity: Views from urban planning[J].American Journal of Preventive Medicine，2002，23(2)：64-73.

[13] MOUDON A V，LEE C.Walking and bicycling: An evaluation of environmental audit instruments[J].American Journal of Health Promotion，2003，18(1)：21-37.

[14] SAELENS B E，HANDY S L.Built environment correlates of walking: A review[J].Medicine and Science in Sports and Exercise，2008，40(Suppl 7)：S550-S566.

[15] BOARNET M，SARMIENTO S.Can land-use policy really affect travel behavior: A study of the link between non-work travel and land-use characteristics[J].Urban Studies，1998，35(7)：1155-1169.

[16] LOO B P Y,CHEN C,CHAN E T H.Rail-based transit-oriented development: Lessons from New York City and Hong Kong[J].Landscape and Urban Planning,2010,97(3):202-212.

[17] FRANK L D，SALLIS J F，SAELENS B E，etal.The development of a walkability index： Application to the neighborhood quality of life study[J].British Journal of Sports Medicine，2010，44(13)：924-933.

[18] MILLWARD H，SPINNEY J，SCOTT D.Active-transport walking behavior: Destinations， durations， distances[J].J Transp Geogr，2013，28：101-110.

[19] SAELENS B E，SALLIS J F，BLACK J B，etal.Neighborhood-based differences in physical activity: An environment scale evaluation[J].American Journal of Public Health，2003，93(9)：1552-1558.

[20] HOBACK A，ANDERSON S，DUTTA U.True walking distance to transit[J].Transportation Planning and Technology，2008，31(6)：681-692.

[21] LACHAPELLE U，NOLAND R B.Does the commute mode affect the frequency of walking behavior: The public transit link[J].Transport Policy，2012，21(4)：26-36.

[22] WIKIPEDIA.Los Angeles County metropolitan transportation authority[EB/OL].(2014-03-12)[2016-02-12].https://en.wikipedia.org/wiki/Los_Angeles_County_Metropolitan_Transportation_Authority.

[23] PARK S.Defining， measuring， and evaluating path walkability， and testing its impacts on transit users′ mode choice and walking distance to the station[D].[S.l.]：University of California，2008.

[24] FRANK L D，BRADLEY M，KAVAGE S，etal.Urban form， travel time， and cost relationships with tour complexity and mode choice[J].Transportation，2008，35(1)：37-54.

[25] MARSHALL J D,BRAUER M,FRANK L D.Healthy neighborhoods: Walkability and air pollution[J].Environ Health Perspect,2009,117(11):1752-1759.

[26] WIKIPEDIA.2010 United States census[EB/OL].(2011-03-12)[2014-09-23].https://en.wikipedia.org/wiki/2010_United_States_Census.

[27] MAGHELAL P K.Healthy transportation， healthy communities： Developing objective measures of built-environment using GIS and testing significance of pedestrian variables on walking to transit[D].[S.l.]：Texas A and M University，2007.

(責(zé)任編輯： 黃仲一 英文審校：方德平)

2017-04-20

許俊萍(1980-)，女，高級(jí)教員，博士，主要從事低碳城市的研究.E-mail:ggxxxu@126.com.

國(guó)家自然科學(xué)基金青年基金資助項(xiàng)目(51508208)； 福建省自然科學(xué)基金面上資助項(xiàng)目(2015J01637)

美國(guó)洛杉磯大都市影響居民步行到軌道交通站點(diǎn)的多因素分析(英文)

許俊萍

(華僑大學(xué) 建筑學(xué)院， 福建 廈門(mén) 361021)

以美國(guó)加利福尼亞州的洛杉磯市為例，根據(jù)2011-2013年美國(guó)南加州政府聯(lián)盟提供的居民出行數(shù)據(jù)，應(yīng)用邏輯回歸模型尋找影響居民步行到快速公交站點(diǎn)的顯著因子.結(jié)果表明：到達(dá)站點(diǎn)的距離和人行道的連續(xù)性、路燈密度、行道樹(shù)密度、站點(diǎn)周邊停車(chē)和土地混合度是顯著影響的居民步行到站點(diǎn)的環(huán)境因子；而出行目的地、家庭收入、家庭擁有私家車(chē)數(shù)量，以及種族等其他因子也顯著影響到居民是否步行到站點(diǎn).

步行； 公交導(dǎo)向型發(fā)展； 大都市； 多因素分析； 邏輯回歸模型； 洛杉磯市

10.11830/ISSN.1000-5013.201704009

U 491.17 Document Code： A Article Number： 1000-5013(2017)04-0489-08