BART @ 20 Series Rail Access Modes and Catchment Areas for the BART

Home , Ashby station (BART), Concord station (BART), Daly City station, Fruitvale Station, Glen Park station, Hayward station (BART)

BART@ 20 Series

Rail Access Modes and Catchment Areas for the BARTSystem

Robert Cervero Alfred Round Todd Goldman Kang-Li Wu

Working Paper UCTCNo. 307

TheUniversity of California TransportationCenter Universityof California Berkeley, CA94720 The University of California Transportation Center

The University of California Center activities. Researchers Transportation Center (UCTC) at other universities within the is one of ten regional units region also have opportunities mandated by Congress and to collaborate with UCfaculty established in Fall 1988 to on selected studies. support research, education, and training in surface trans- UCTC’seducational and portation. The UCCenter research programs are focused serves federal Region IX and on strategic planning for is supported by matching improving metropolitan grants from the U.S. Depart- accessibility, with emphasis ment of Transportation, the on the special conditions in California Department of RegionIX. Particular attention Transportation (Caltrans), and is directed to strategies for the University. using transportation as an instrument of economic Based on the Berkeley development, while also ac- Campus, UCTCdraws upon commodatingto the region’s existing capabilities and persistent expansion and resources of the Institutes of while maintaining and enhanc- Transportation Studies at ing the quality of life there. Berkeley, Davis, Irvine, and Los Angeles; the Institute of The Center distributes reports Urban and Regional Develop- on its research in working ment at Berkeley; and several papers, monographs, and in academic departments at the reprints of published articles. Berkeley, Davis, Irvine, and It also publishes Access, a Los Angeles campuses. magazine presenting sum- Faculty and students on other maries of selected studies. For University of California a list of publications in print, campuses may participate in write to the address below.

University of California Transportation Center

108 NavalArchitectuie Building Berkeley, California 94720 Tel: 510/643-7378 FAX:510/643-5456

Thecontents of this report reflect the viewsof the author whois responsible for the facts and accuracyof the data presentedherein. Thecontents do not necessarilyreflect the official viewsor policiesof the State of Californiaor the U.S.Department of Transportation.This report doesnot constitute a standard, specification,or regulation. Rail Access Modes and Catchment Areas for the BART System

Robert Cervero Alfred Round Todd Goldman Kang-Li Wu

Departmentof City and Regional Planning Institute of Urbanand RegionalDevelopment University of California Berkeley, CA 94720-1850

Working Paper September 1995

UCTCNo. 307

The University of California Transportation Center University of California at Berkeley Table of Contents

1. INTRODUCTION 1 2. WHYSTUDY ACCESS TRIPS AND CATCHMENTZONES? 1 3. RESEARCH FOCUS 3 4. DATA SOURCES AND RESEARCH APPROACH 5 5. STATION CLASSIFICATIONS 6 5.1. GroupingVariables 6 5.2. Classification 8 5.3. Station Classes 10 6. ANALYSIS OF MODESOF ACCESS ACROSSSTATION CLASSES 14 7. PREDICTORS OF ACCESS AND EGRESS MODES 18 8. ACCESS MODESAS A FUNCTIONOF DISTANCE TO STATIONS 23 9. RIDERSHIP CATCHMENTAREAS 40 9.1. CatchmentAreas for All Modesof Accessand Egress 41 9.2. CatchmentAreas for WalkingAccess Modes 42 9.3. Comparisonof CatchmentsAcross Station Classes, for all Modes 42 9.4. Comparisonof CatchmentsAcross Station Classes, for WalkingTrips 45 9.5. CatchmentsAmong Station Types with Different Parking Supplies 46 10. CONCLUSION 49 Notes 41 References 53 APPENDIX 55 FIGURES

1° Hypothesized Effect of NeighborhoodDesign on Pedestrian Access to Stations 2. Hypothesized Distributions of Rail Station Access Modes as a Function of Distance to Stations Dendogramfor Classifying BARTStations by Land-Use Environment , 4. Modeof Egress for CommuteTrips from San Francisco Office Center BARTStations to Work 25 Modeof Access for CommuteTrips , from Hometo San Francisco Office Center BARTStations 25 Modeof Access/Egress for All San Francisco Office Center BARTCommute Trips 26 , 7. Modeof Egress for CommuteTrips from San Francisco Commercial/Civic Center BARTStations to Work 26 Modeof Access for CommuteTrips , from Hometo San Francisco Commercial/Civic Center BARTStations 27 Modeof Access/Egress , for All San Francisco Commercial/Civic Center BARTCommute Trips 27 10. Mode of Egress for CommuteTrips from DowntownOakland BARTStations to Work 28 11. Mode of Access for CommuteTrips from Hometo DowntownOakland BARTStations 29 12. Mode of Access/Egress for all DowntownOakland BARTCommute Trips 29 13. Modeof Egress for CommuteTrips from Urban District BARTStations to Work 31 14. Modeof Access for CommuteTrips from Hometo Urban District BARTStations 31 15. Modeof Access/EgressFor all Urban District BARTCommute Trips 32 16. Mode of Access for CommuteTrips from Hometo Suburban Center BARTStations 32 17. Mode of Egress for CommuteTrips from Suburban Center BARTStations to Work 33 18. Modeof Access/Egress for All Suburban Center BARTCommute Trips 33 19. Mode of Access for CommuteTrips from Hometo Low-Density-Area BARTStations 35 20. Modeof Egress for CommuteTrips from Low-Density-Area BART Stations to Work 35 21. Mode of Access/Egress For All Low-Density-Area BARTCommute Trips 36 22. Mode of Access/Egress For Terminal/Near-Terminal BARTCommute Trips 36 23. Modeof Access for CommuteTrips from Hometo All BARTStations 37 24. Modeof Egress for CommuteTrips from All BARTStations to Work 37 25. Modeof Access/Egress for All BARTCommute Trips 38 26. Mode of Access/Egress for All Home-to-Work BARTCommute Trips, Longer Distance Plot 38 27. Comparison of Mean Catchment Radii and Land Areas AmongStation Classes 43 28. MeanRadii and Land Areas as Percent of the "San Francisco Office Center" Catchment 44 29. Comparison of Mean Catchment Radii and Land Areas AmongThree Station Types 48 30. Mean Radii and Land Areas as Percent of the "Urban/No Parking" Catchment 48

ii TABLES

° Candidate Variables for Classifying BARTStations 7 2. Station Characteristics: BARTSystem 8 3. Characteristics of the Six BARTStation Classes 12 4. ANOVAin Percent of Home-End Access Trips by Automobile Across Station Classes 15 5. ANOVAin Percent of Home-End Access Trips by Drive-Alone Automobile Across Station Classes 16 6. ANOVAin Percent of Home-EndAccess Trips by Kiss-and-Ride, Across Station Classes 16 7. ANOVAin Percent of Home-EndAccess Trips by Transit, Across Station Classes 16 8. ANOVAin Percent of Work-EndEgress Trips by Transit, Across Station Classes 17 9. ANOVAin Percent of Home-EndAccess Trips by Walking, Across Station Classes 18 10. A.NOVAin Percent of Work-EndEgress Trips by Walking, Across Station Classes 18 11. Regression Modelfor Predicting Percentage of Access Trips to BARTStations by Automobile 19 12. Regression Modelfor Predicting Percentage of Egress Trips from BARTStations by Transit 2O 13. Regression Modelfor Predicting Percentage of Access Trips to BARTStations by Walking 20 14. Regression Modelfor Predicting Percentage of Egress Trips from BARTStations by Walking 21 15. Midpoint Elasticities of Access and Egress ModalShares as Functions of Land-Use and Other Variables 22 16. SummaryStatistics for Access and Egress Trips by Station Class 23 17. Summaryon Influence of Distance on Modesof Access and Egress AmongClasses of BARTStations 39 18. Comparison of Catchment Radii, Land Areas, Populations, and Densities for Six Station Classes, for all Home-AccessModes 43 19. Regression Modelfor Predicting Size and Radius of CatchmentAreas for BARTStations, Walking Modes 45 20. Comparison of Catchment Radii, Land Areas, Populations, and Densities for Six Station Classes, for Home-AccessWalk Trips 46 21. Regression Modelfor Predicting Size and Radius of CatchmentAreas for BARTStations, Walking Modes 47 22. Comparison of Catchment Radii, Land Areas, Populations, and Densities for Three Station Types, for All Home-AccessModes 47 23. Comparison of Catchment Radii, Land Areas, Populations, and Densities for Three Station Types, for Home-AccessWalk Trips 49

MAPS Map 1. Map of the BARTSystem 11

°°o III Rail Access Modes and Catchment Areas for the BARTSystem

Robert Cervero

1. INTRODUCTION To date, far moreresearch has been conductedon the effects of the built environmenton transit demandalong mainline corridors than in the catchmentzones surrounding transit stops. Pushkarevand Zupan(1977), for example,correlated transit ridership for the line-haul segmentof trips as a function residential densities, distance to downtown,and size of downtown;however, they ignored howaccess trips to transit stops were influenced by such factors. Seminalwork by Meyer,Kain, and Wohl(1965) studied factors influencingbus andrail transit demandfor three segmentsof trips mresidential collection- distribution, line-haul, and downtowncirculator -- however,their workdid not examinethe direct effects of land-usevariables. For example,in the case of access trips fromhome to rail stations, or whatthey call the residential collection-distribution segment,the numberof "trip origins per city block" wasused as the predictor of access demand.Standard trip generationrates wereused to directly estimate access demand. As part of the BART@20study, this report studies the influence of the built environmenton two aspects of transit demand:(1) modesof access to and fromrail stations; and (2) the sizes and shapes of the ridership catchmentareas. Variations in both modesof access and catchmentarea sizes are studied for different classes of stations, defined mainlyin terms of the land-use environment.Also, both descriptive statistics and analytical models(ANOVA and regression) are used for examiningthese relationships.

2. WHY STUDY ACCESS TRIPS AND CATCHMENT ZONES? It is importantto study transit access trips and catchmentareas for a numberof reasons. One, in manysuburban areas served by rail systems, the private automobileis predominantlyused to reach stations. Froman air quality standpoint,transit riding does little goodif mostpeople use their cars to reach stations. For a three-mile automobiletrip, the typical distance driven to access a suburbanBART park-and-ride lot, around84 percent of hydrocarbon(HC) emissions and 54 percent of nitrogen oxide (NOx)emissions are dueto cold starts (inefficient cold enginesand catalytic convertersduring the first few minutesof driving) and hot evaporativesoaks (Barry and Associates,1991). That is, a sizeable share of the tailpipe emissionsof the two main precursors to photochemicalsmog formation occur from turn- hag the automobileengine on and off. Drive-aloneaccess trips to rail stations, regardless howshort they are, emit levels of pollutants that are not too muchbelow those of the typical 10-milesolo commute. Rail trips that rely on park-and-rideaccess do verylittle to improveair quality. It is clear that the built environment,along with parkingprovisions, has a significant influence on rail access trips, but to whatdegree remains unclear since little systematicwork has been carried out to date on this question. In general, we knowthat as densities fall and distances to downtownsincrease, people increasingly rely on mechanizedmeans to reach stations. In downtowns,most people reach transit stops by foot. In 1992, for instance, over 60 percent of rail users walkedto downtownBART stations (BART,1993). As one leaves downtownstations and heads outward,the share of walk-ontrips falls steadily, replaced by access trips madeby somemechanized mode -- normally, park-and-ride, kiss-and- ride, and bus-and-ride.In the case of BARTstations, like Ashbyand Glen Park, that lie in fairly built- up, urbanizedareas but that have park-and-ridefacilities, around50 percent of customersreach statibns by car, 8 percent ride a bus, and 37 percent arrive by foot. Andat suburbanstations, like WalnutCreek and Fremont,upwards of 85 percent of access trips by BARTare by passengercar, and fewer than 5 percent are by foot or bicycle travel. Studies in greater Washington,D.C., metropolitanToronto, and the BayArea show that beyondone mile of a suburbanrail station, around60 to 80 percent of access trips are by automobile,with the share steadily rising with access distance (Stringham,1982; JHK and Associates, 1986, 1989; Cervero, 1993; BART,1993). Creatingplaces that are moreconducive to pedestrain, bicycle, and bus transit access to rail stops wouldmake environmental contributions beyondimproving air quality. Reducedautomobile access trips wouldalso help reduce the emissionof greenhousegases and such precursors to acid rain as sulfur diox- ide (SO2). Morenon-motorized access trips wouldalso reduce energy consumptionto the same degree they eliminate tailpipe emissions. The objectives of linking land developmentand transportation pro- gramsfor the purposesof promotingnational air quality, energy, and related environmentalbenefits are embodiedin a series of recently adoptednational legislative acts, includingthe 1990Clean Air ActAmend- ments, the 1991Intermodal Surface Transportation Efficiency Act (ISTEA),and the 1992National EnergyPolicy Act. The encouragementof alternatives to the automobilefor accessing transit stops is also consonant with emergingcommunity design concepts, such as traditional neighborhooddevelopment (TND) and "newurbanism." A growinglegion of urban designers, such as Solomon(1992), Calthorpe (1993), Katz (1993), call for designingnew communities like those of yesteryear. Theycontend that encouraging morehuman-scale, compact development, returning to conventional gridiron street forms, narrowing street widthsand building setbacks, and landscapingfor pedestrian movementswill reduce the dominance of the automobileand reduce dependenceon it. The goal of linking transit facilities to community developmentis also embodiedin the Liveable Communitiesprogram recently initiated by the Federal Transit Administration. Overall, research that establishes relationships betweenbuilt environmentsand modesof access could prove valuable in influencing future land planning, neighborhoodand pedestrian facilities designs, and bus transit service deploymentin the vicinity of rail stations. Recentlaws and mandateshave created a receptive policy environmentfor promotingstronger linkages betweenneighborhood development and transit services, makingresearch in the areas all the moretimely.

3. RESEARCH FOCUS Thefollowing research questionsare addressedin this study: 1. Modesof access to and fromrail stations, and the shapes and sizes of ridership catchment areas, vary systematicallyby class of transit station area. Catchmentareas increases in size and automobileaccess increases in shares in lower-density, moresuburban areas. Stationcatchment areas increaseas densities decline, park-and-ridefacilities increase, distances fromthe CDBincrease, and wherevera station is a terminus. 2. Station catchmentareas are larger whensurrounded by residential than by commercial/ office land uses. Mixedland use environmentsare associated with morecompact catchmentareas. 3. Walkingis the dominantmode for access trips under 2,500 feet, or aroundone-half mile. Busaccess is used mostfrequently for access trips 1.5 to 4 miles in length. Beyondone- half mile, the private automobileis the dominantaccess modefor stations with park- and-ridefacilities. 4. For egress trips fromstations in downtownsand highly urbanizedareas, walkingis the dominantmode up to a distance of aroundthree-quarters mile. Beyondthis distance, bus transit is the predominantegress mode. 5. Theshares of non-autoaccess and egress trips to and fromrail stations increase with density and levels of land-usemixture, controlling for other factors like parkingsupply. Thehypotheses tested on the effects of the built environmenton the spatial dimensionsof the catchmentarea is reflected by the drawingin Figure1. In this figure, for a single class of station (such as suburbanones with no parkingfacilities, as suggestedin Figure 1), the catchmentradius for walk-access trips is thoughtto extendfarther out as residential densities increase, the amountof non-residentialacti- vities increases, and the quality of the pedestrianenvironment rises. Thehypotheses on the effects of the built environmenton the modaldistributions of access trips are reflected by Figure 2. In both drawings,walk shares are presumedto drop precipkiouslywith distance. For accesstrips fromresidences to rail stations, park-and-rideaccess is assumedto eclipse walkaccess at distances of aroundone-half mile. In lowerdensity settings, auto access mightdominate at an even shorter distance, like one-quarter mile. Thesebenchmarks are based on research by Untermann(1984) and others whohave recorded the maximumdistances Americansare willing to walk -- generally in the range of one-quarterto a half of a mile, thoughresearch showsacceptable walking distances can be stretched con- siderably (perhapsas muchas doubled)by creating pleasant, interesting urban spaces and corridors. an access mode,bus transit’s marketniche is thoughtto be primarily in the one- to 2.5-milerange. For (%Trips)

Dense, Mixed-Use, Pedestrian-

I I I I I I I I I I I I I I I I Transit 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,00~ 5.50’0 6,000 6,500 7,000 7,500 8,000 Slalion OISTAHCEFROM STATIOH (IN FEET}

Figure 1. HypothesizedEffect of NeighborhoodDesign on Pedestrian Access to Stations

Residential Areas: Employment Areas: HypothesizedDistributions HypothesizedDistributions r -. , Of Access Modes For Egress Modes %Trips %Trips ToTransit ToTransit Slation Station

~Walk

f Car Passenger/ OlherTransil , __

=’::,~’o’I.’~~ "5 ~ ,h ,~ !h I ,~s ,!5 ,h MJL.ESFROM STA~IOH MILESFROM STATION

Figure 2. HypothesizedDistributions of Rail Station Access Modes as a Functionof Distance to Stations egress trips froma rail station to one’s worksite, Figure2 showsthat bus transit or other surface modes are thought to predominatebeyond three-quarters of a mile. Theabsence of private car accessibility at exit stations leaves manywith few other options than to walkor take bus transit to their destinations? For the mostpart, these are original research questionswhich have not been systematically addressedto date. In a study of access trips to rail stations in Torontoand Edmonton,Stringham (1982) carried out the most in-depth workon these topics so far. Stringhamfound the "walkingimpact zone" to be as far as 4,000 feet fromrail stations in Toronto,suggesting more compact, mixed-use urban environmentsare indeed associated with relatively large walk-oncatchment areas. His workestablished that shares of access trips madeby park-and-ridemodes increased with distance. Stringhamdid not, however, studyhow access modesvaried as a function of class of station area or directly as a functionof land-use environments.

4. DATA SOURCES AND RESEARCH APPROACH Theprimary data source used for carrying out this analysis was a survey of trips madeby around 35,000 BARTpassengers, conductedin late-October 1992. The on-board passenger survey compileddata for individual BARTtrips, includingthe origin and destination station, trip purpose,time-of-day, fare paid, andvarious characteristics of the access trip. Informationwas available on the nearest street intersection whereaccess trips to BARTstations originated as well as the nearest intersection to whichegress trips fromBART stations ended. This fine-grain resolution of origin-destination data allowedfairly precise estimatesof the straightline distance of access andegress trips to be estimated.Additionally, data on the modes,purposes, and times-of-dayof access and egress trips werealso available. For further information on the 1992 BARTpassenger profile survey, see BART(1993). Theprincipal land-use data used in this research wasa 1990digital inventory of dominantland uses within hecatre grid cells (100xl00meters), compiledby the Association of BayArea Governments (ABAG)for the entire San Francisco Bay Area. Using the Archinfo GeographicInformation Systems (GIS)package, buffers werecreated to generatefairly precise estimates of the compositionof land uses wkhina one-half-mile radius of all 34 BARTstations. 2 Whilethe ABAGinventory compiles data for over 40 individual land uses, these categories werecollapsed into six majorones: residential, commercial, industrial/office, public, vacant, and other. In testing the research hypotheses,several approacheswere taken. BARTridership data supported the analysis of access and egress trips and fromstations as functionsof trip distance. For purposesof studyinghow access and egress trips varied by land-use environment,cluster analysis was used to classify stations wkhsimilar land-use characteristics, defined in terms of dens@and land-use composition.Resi- dential and employmentdensities wereestimated for block groupsand census tracts surroundingeach station, using 1990 census data from SummaryTape File (STF) 3-A and the CensusTransportation Planning Package (CTPP)for the San Francisco Bay Area.3 Land-use compositionswere based on the 1990 ABAG inventories. These data were supplementedby other information compiledfrom various primary and secondarysources, such as the supply of parking at each station (from BARTrecords) and whethersta- tions weresituated in freewaymedians (supplied fromfield observations). Oncestations wereclassified, differencesin access modeswere studied across classes of stations. In examiningareas aroundwhich transit ridership is drawn, catchmentareas weredefined as con- tiguous census tracts whichencompass the origins of 90 percent of all access trips to BARTstations (or destinations of 90 percent of all egress trips fromstations). Additionally,catchment areas for walk-on trips weredefined as the contiguouscensus tracts encompassingthe origins ofg0 percent of all access trips madeby foot. This secondcatchment represents, we believe, an area that correspondsto the zone of transit-oriented developmentaround stations. Thevery fact that rail customersare willing to walkto stations fromthese areas suggests that, whetherbecause of design, proximity,or a combinationof the two, developmentis reasonablywell-oriented to rail stations. 4 Definedcatchment areas for each station are portrayed using GISoutputs. These outputs provide someunderstanding of howcatchment areas vary in shape and size. Additionally, the average radius of catchmentareas wereestimated and compared acrosss each class of station, Last, for testing the hypotheseson howbuilt environmentsinfluence access trips, several regression equationswere estimated that predict the percent of variousaccess modesas functions of urbandensi- ties, land-usecompositions, and variouscontrol variables (e.g., supplyof parking,distance of station to downtownSan Francisco, presence of freeway median). In estimating modes,cases consisted of BART stations and a data base wasconstructed using the variousdata sourcesdescribed in this section.

5. STATION CLASSIFICATIONS The analyses of howaccess modesand catchmentareas vary relied on constructing a meaningful typology of BARTstation environments.The land-use settings of BARTstations are not alike. Some are in dense, downtownareas, someare in predominantlyresidential suburbancommunities, some are in the mediansof freeways, someinclude acres of parking, and somehave no parking. At least in part, modesof access and egress to and fromBART stations and the influences of distance on rail access trips will dependon features of the built environmentaround rail stops. Theprocess of classifyingobjects, be they rail station areas, cities, or insects, involvestwo steps: (1) selecting a set of variables whichdefine the dimensionsalong which stations areas will be grouped(e.g., densities, parkingsupplies); and(2) applyinga clustering algorithm.Each of these steps is discussedbelow.

5.1. Grouping Variables Variables whichdefined the land-use and physical environmentsaround BARTstations wereused for groupingstations into classes. Table1 lists the variablesinitially considered.Land use variablesgauged Table 1. Candidate Variables for Classifying BARTStations Land use characteristics Resdens Residential density, in dwellingunits per acre in 1990.Measured for census tracts and block groups that encompassa one-half-mile radius aroundstation. Source: 1990census STF3-A. Popdens Populationdensity, in populationper acre in 1990. Measuredfor censustracts and block groups that encompassa one-half-mile radius aroundstation. Source: 1990 census STF3-A. Empdens Employmentdensity, in employeesper acre in 1990. Measuredfor census tracts and block groups that encompassa one-half-mile radius around station. Source: 1990 CensusTransportation Planning Package, Part II, Metropolitan Transportation Commission. CommercialProportionof land area in commercialuse for one-half-mileradius aroundstation. Source: 1990 Association of Bay Area Governmentsland use inventory. Industrial Proportionof land area in industrial or office use for one-half-mileradius aroundstation. Source: 1990 Association of BayArea Governmentsland use inventory. Residential Proportionof land area in residential use for one-half-mileradius aroundstation. Source:1990 Association of Bay Area Governmentsland use inventory. Entropy Indexof land-usemixture. Relative entropy- {Y"i[Pi * ln(p.~]}/ln(k) wherePi - proportionof land area in land-usecategory i, and k - numberof land-use categories; ranges between0 and 1, where0 signifies land devotedto a single use and 1 signifies all land area evenlyspread among all uses. Domlan Dominantland use category: 1- residential, 2- commercial,3 - industrial/office, 4- public, 5- other. Source: 1990 Association of BayArea Governmentsland use inventory and field observations. Vclnd Vacant/developableland within one-half-mile of station: 1 -low (< 10 percent of land area, 2-medium(10-25 percent land area), 3-high (> 25 percent of land area). Source: 1990Association of BayArea Governmentsland use inventory and field surveys. Station Characteristics Furypx Freewayproximity, where limited-access freewaylies the followingdistances fromstations: 1 - 0-0.5 miles, 2 - 0.5-1.0 miles, 3 - 1.0-2.0 miles, 4 - > 2 miles. Source: ThomasBrothers Maps,1994. Fwymd Freewaymedian station location: 1-yes, 0-no. Source: Field observations. Parking Park-and-ridespaces at station, surface and structured. Source: BARTSystemwide Parking Inventory, 1993. Stnfn Station function: 1-transfer, 2-terminal, 3-other. Source: BARTsystem map. Ridership Characteristics Dayexits Averageweekday exists, 1992 (january-December). Source: BARTplanning department. BARTcm BARTcommutes as a percent of total journeys-to-workmade by employed-residentsliving within one-half-mileradius of station. Measuredfor censustracts and block groups that encompassa one- half-mile radius aroundstation. Source: 1990 census STF-3A. Neighborhood Characteristics Income Annualhousehold income for householdswithin one-half-mile radius of station, 1990. Measuredfor census tracts and block groups that encompassa one-half-mile radius aroundstation. Source: 1990 census STF-3A. Red/s Redevelopmentdistrict encompassesstation: 1 - yes, 0 - no. Source: interviews with local planning departments. Speczone Special zoningin station area: 0 - none,1 - incentive zoning(e.g., density bonuses),2 - restrictive zoning (e.g., downzoningof densities). Source: local planningdepartments.

the densities, compositions, and levels of mixture of activities, generally for a one-half-mile radius around stations. Other grouping variables measuredcharacteristics of stations (e.g., parking supplies), ridership (e.g., rail modalsplits), and neighborhoods(e.g., householdincomes). Table 2 presents a matrix of values for the grouping variables for each of the 34 stations.

7 Table 2. Station Characteristics: BARTSystem Com- Indus- Resi- Res- F~.T- Fwy- Day- BART-Dora-Park- Emp-Popden- Pec- Station density ~ rnd Smfn exits cm lan ing Vclnd density fi.~Redls zone merci~ trial dential Rockridge 8.6 1 1 3 4)016 10.7 1 889 3 7.0 18.0 0 2 0.0900 0.0604 0.7162 0.5456 MacArthur 8.1 1 1 1 4,407 9.5 1 609 3 5.5 18.9 0 0 0.1808 0.0687 0.5467 0.7524 W. Oakland 5.5 2 0 3 3,722 8.5 1 424 2 2.4 15.9 0 0 0.2010 0.1519 0.2365 0.9033 19th St. Oak. 7.9 1 0 3 7,855 15.2 3 0 3 64.8 11.5 0 0 0.3429 0.1004 0.3829 0.8029 Oak.City Ctr. 7,3 1 0 1 9,534 6.6 -3 0 3 52.2 20.4 1 0 0.3032 0.1164 0.2753 0.9257 LakeMerritt 12.1 1 0 3 3,549 11,5 3 205 3 23.4 21.8 0 0 0.3637 0.1312 0.2669 0.8723 Fruitvah 5.0 1 0 3 5,741 7.7 2 1,103 3 4.4 18.1 0 0 0.3405 0.0903 0.5118 0.6349 Coliseum 3.6 1 0 3 5,571 3.5 3 1,059 2 2.6 12.5 0 0 0.2289 0.0848 -.3252 0.9013 N. Berkeley 10.1 3 0 3 3,181 10.0 1 840 3 7.7 20.8 0 2 0.1523 0.0601 0.7274 0.1533 Berkeley 14.1 4 0 3 10,055 10.8 2 0 3 24.4 23.0 0 0 0.1534 0.1862 0.6443 0.5304 Ashby 11.3 3 0 3 3,104 9.5 2 626 3 4.1 23.4 0 2 0.1762 0.0525 0.7259 0.4905 San Leandro 6.0 2 0 3 3,937 9.9 3 1,295 3 4.8 12.1 1 0 0.1233 0.0732 0.4480 0.7133 BayFair 6.3 2 0 3 5,247 7.2 1 1,903 3 3.5 14.7 0 0 0.2063 0.1005 0.6136 0.6250 Hayward 4.1 3 0 3 4,890 3.7 2 1,061 2 7.2 10.4 1 0 0.2084 0.1266 0.5912 0.6241 S. Hayward 4.0 3 0 3 2,845 9.5 1 1,307 3 1.1 14.2 0 0 0.1294 0.0992 0.4252 0.7332 Union City 2.2 4 0 3 3,807 1.1 3 1,218 2 2.1 6.5 1 0 0.0933 0.0521 0.4587 0.7899 Fremont 4.9 4 0 2 5,674 4.9 5 2,494 2 1.5 12.7 0 0 0.2221 0.1585 0.3399 0.7530 Pleas. Hill 4.8 1 0 3 6,088 16.9 1 3,245 1 4.1 8.6 1 1 0.1391 0.0671 0.7477 0.4548 Concord 2.8 2 0 2 7,730 13.0 1 1,975 3 1.6 7.6 1 0 0.1993 0.1211 0.6325 0.5911 Walnut Creek 5.3 1 0 3 5,308 13.7 3 1,518 3 19.0 9.0 0 0 0.2517 0.0392 0.6017 0.6105 Lafayette 0.7 1 1 3 3,179 13.6 1 1,521 3 0.5 1.7 0 0 0.1027 0.0482 0.5633 0.6494 Orinda 2.0 1 1 3 2,951 5.3 1 1,380 3 0.2 4.2 0 0 0.0334 0.0148 0.5058 0.5854 Richmond 5.9 2 0 2 2,704 10.7 1 796 1 4.3 17.7 1 1 0.1777 0.0909 0.6453 0.5686 ECdel Notre 4.9 1 0 3 7,387 14.4 1 2,516 3 2.2 12.3 1 0 0.1089 0.1474 0.6318 0.6455 El Cerrlto (EC) 6.6 2 0 3 3,769 15.6 1 795 3 4.9 14.1 1 0 0.1153 0.1303 0.6834 0.5774 Embarcadero 11.4 2 0 3 26,966 2.4 3 0 3 156.0 20.3 0 0 0.4456 0.0438 0.2046 0.7953 MontgomerySt. 4.8 2 0 3 28,080 2.3 3 0 3 234.0 9.7 0 0 0.4109 0.0361 0.2489 0.7967 PowellSt. 23.6 2 0 3 17,413 4.8 3 0 3 86.0 46.9 0 0 0.4503 0.0492 0.2105 0.7705 S.F. CivicCtr. 42.1 2 0 3 12,931 6.0 3 0 3 75.0 75.7 0 0 0.4406 0.0382 0.2414 0.7537 Mission16thSt. 22.0 2 0 3 5,963 15.2 2 0 3 22.6 53.2 0 2 0.2548 0.0402 0.4685 0.7287 Mission24th St. 21.6 2 0 3 8,659 12.0 1 0 3 16.1 63.6 0 2 0.1154 0.0445 0.7226 0.5529 Glen Park 10.3 1 0 3 5,795 15.4 1 55 3 2.4 27.4 0 2 0.0320 0.0276 0.8036 0.4319 BalboaPark 8.5 1 0 3 10,001 13.5 1 0 3 4.4 26.7 0 2 0.0440 0.0772 0.8067 0.4167 Daly City 7.8 1 0 2 10,250 8.7 1 2,228 2 2.5 28.6 0 0 0.0895 0.1328 0.5941 0.6828 Sources: BAR:l, ThomasBros, Maps, 1990 US Census, ABAG, MTC.

5.2. Classification The groupingof the 34 BARTstations into homogenousclasses was carried out using cluster analysis. Theprocess involvedcombining cases into clusters on the basis of their "nearness"to each other whenexpressed as squaredEuclidean distances. 6 Usingthe technique of agglomerativehierarchi- cal clustering, clusters weresequentially formed by groupingcases into evenlarger clusters until all cases were7 members of a single cluster. A numberof combinationsof variables wereattempted in creating decipherableand intuitive appealingclusters. Becauseof high collinearity amongvariables, employingall variables wouldhave introduced unnecessaryredundancy and overemphasizedcertain variables. Themost satisfactory results wereobtained by using the followingvariables: ¯ Employmentdensity (workers/acre) ¯ Residential density (households/acre) ¯ Percent of station area devotedto residential land uses ¯ Entropy index of land-use mixture ¯ Parkingsupply, based on an ordinalscale of 0 to 4. ¯ Annualhousehold income, in $1,000s ¯ Percentof commutesby station-areaemployed-residents by rail All of these variableswere drawn directly fromthe data base shownin Table2 exceptfor the variable measuringparking supply. Because of the large variationin parkingsupplies, with aroundone- third of stations havingnoparking and some stations havingseveral thousandspaces, the use of original parkingvariable dominated all other variablesin the formationof clusters,s Therevised ordinal parking variablewas scaled as follows: 0 = no parking,1 - 1 to 1,000 spaces, 2 1 1,001 to 2,000 spaces, 3 - 2,001 to 3,000 spaces, and4 = > 3,000 spaces. Theresults of the dusteranalysis are summarizedin the hierarchicalgraph, called a dendogram, showin Figure3. This showsthe clusters beingsequentially combined and the normalizedvalues of the

Dendrogram using Average Linkage (Between Groups)

Rescaled Distance Cluster Combine

CASE 0 5 i0 15 20 25 Label Num + ...... + ...... + ...... + ...... + ......

W. Oakland 3 Richmond Fruitvale 7 Coliseum 8 Rockridge 1 Balboa Park 33 N. Berkeley 9 -- Pleas. Hill 18 -- Glen Park 32 EC del Norte 24 E1 Cerrito (EC) 25 -- Mac Arthur 2 -- San Leandro 12 Ashby ii Union City 16 -- Fremont 17 -- Bay Fair 13 -- S. Hayward 15 -- Da!y City 34 -- i Hayward 14 -- Concord 19 Lafayette 21 Orinda 22 Walnut Creek 20 Lake Merritt 6 Berkeley -i Mission 16th St. ¯ 3o _1 Mission 24th St. Embarcadero 26 Montgomery St. 27 ] 19th St. Oak. a Oak. City Ctr. Powell St. 28 S.F. Civic Ctr. 29 --~ Figure 3. Dendogramfor Classifying BARTStations by Land-UseEnvironment

9 coefficients (i.e., squaredEuclidean distances) at each step. Thejudgemental part of cluster analysis decidingat whatstage to stop joining clusters. This is normallydone when the distance coefficients dra- matically increase fromon agglomerationto another, or whenan intuitive number,normally 4 to 6, of clusters have been formed.For this analysis, six station classes wereconsidered to be the maximum acceptable. Six classes wereformed between the 24th and 25th stages of mergingclusters. This provided . an intuitive and interpretative groupingof stations. Thefollowing six station classes wereformed, with the BARTstations that groupedinto each class also listed: San Francisco Office Center: Embarcaderoand Montgomery San Francisco Commercial/CivicCenter: Powell and Civic Center DowntownOakland: City Center (12th Street) and 19th Street UrbanDistricts: LakeMerritt, Berkeley, Mission16th Street, and Mission24th Street SuburbanCenters: Lafayette, Orinda, WalnutCreek, Pleasant Hill, 9 and Concord Low-DensityAreas: MacArthur,West Oakland, Rockridge, Fruitvale, Coliseum, San Leandro, Bay Fair, Hayward,South Hayward,Union City, Freemont, Ashby, North Berkeley, E1 Cerrito Center, E1 Cerrito del None, Richmond,Glen Park, BalboaPark, and Daly City. (See Map1 for a mapshowing the locations of each station on the BARTsystem.) Table3 suggestswhy these particular titles werechosen for describingthe six station classes; it presents the means,standard deviations, and low-to-highranges of the sevenvariables used in forming clusters. Thehomogeneity of cases in each cluster is reflected by the low standarddeviations relative to means(i.e., lowcoefficients of variation) for mostvariables. Thedistinctiveness of clusters is reflected by the relative large differencesin meansfor variablesacross the six groups.10

5.3. Station Classes Thefollowing six station classes are presentedin hierarchical order basedon their level of urbanization. Level of urbanization is perhapsbest reflected by the descendingemployment densities across these station classes. Otherdistinguishing land-use features of eachstation class are also highlightedin this section. ¯ San Francisco Office Center: Thesetwo stations -- Embarcaderoand Montgomery--serve the heart of downtownSan Francisco’shigh-rise office and financial district, surrounded by the tallest buildings in the BayArea. Theyare characterized by extremelyhigh employmentdensities, with relatively little housingnearby (reflected by the low per- centageof residential land area). Therelatively modestresidential densities reflect relatively few dwellingunits per gross acre. (Ona net residential acreagebasis, densities wouldbe fairly high.) Theyhave a moderatelevel of mixeduses owingto the presence of somerestaurants and retail services in the area. Averagehousehold incomes within a half mile of the stations are fairly high. Thesestations haveno parking; however,they havethe highest levels of connectingtransit services, includingdiesel andtrolley buses, cable cars, light rail transit, trams, andferry services. Relativelyfew employed-residents in the area commuteby rail in large part becausemany can walkto their jobs.

10 / :; ..’ C~.~ord "-~#t w" * e !: RICHHOND l ’ Ric CORRIDOR CONCORD , e ’: "%/ CORRIDOR ¯ "° PleasantHill El CerritoDel Norte ~ ," ...... ~-° ..o..°mS d o-*" a/aJnutCreek

:RANCISCO,’ \ / 16~thSt/Mission 24 SdMission~, ,i Glen Park,, FREMONT N.-"1"- Balboa:Park," CORRIDOR

CORRIDOR H~. ard .¯¯: ", ...... :~,;-~- .,. SouthHayward"

~:".. JnionCity ¯ ¯ d~~*:~-.~ ~,’~-%~.

~? ".. --~-"..... T! %~./%. "’"~ d J

BayArea Rapid Transit Station 10miles I I 10kin o.

Map1. Mapof the BARTSystem

11 Table 3. Characteristics of the Six BARTStation Classes

Class of BART Station San Francisco Low- San Francisco Commercial/ Downtown Urban Suburban Density Office Center Civic Center Oared Districts Centers Areas Density EmploymentDensity (workers/acre) - Mean(std.-dev.) 195.0(58.2) 80,5(7,8) 58.5(8.9) 21.6(3.8) 5.1(7.9) 3.9(1.9) Range 156.0-234.0 75.0-86.0 52.1-64.8 16.1-24.4 0.2-18.9 1.1-7.7 Residential Density (dwellingunits/acre) Mean(std. dev.) 8.1(4.7) 32.8(13.1) 7.6(0.5) 17.4(5.1) 3.1(1.9) 5.7(2.7) Range 4.8-11.4 23,6-42.1 7.2-7.9 12.1-21.9 0.7-5.3 2.2-11.3 Land Use Percent LandArea Residential Mean(std. dev.) 22.7(3.1) 22.6(2.2) 32.9(7.6) 52.6(20.3) 61.0(9.1) 57.5(16.1) Range 20.5-24.9 21.0-24.1 27.5-38.3 26.7-72.3 50.6-74.8 23.6-80.7 Mixture1) of Use(relative entropy Mean(std. dev.) .796(.001) .762(.011) .864(.087) .671(.168) .578(.073) .647(.139) Range ,795-.796 .753-.771 .803-.926 .534-.872 .454-.649 .417-.903 OtherCharacteristics ParkingSpaces at Station Mean(std. dev.) o(o) o(o) o(o) 51(103) 1928(77~ 1116(725) Range 0-0 0-0 0-0 0-205 1380-3245 0-2516 Annual Household Income ($1000,1990) Mean(std. dev.) 39.4(5.6) 31.0(4.2) 15.1(1.4) 20.2(3.6) 37.4(7.1) 22.4(7.8) Range 35.4-43.5 28.2-34.1 14.4-15.7 16.8-25.3 26.4-44.1 9.2-35.1 Percent Residents Commutingby BART,1990 Mean (std. dev.) 2.38(.16) 5.42(.85) 10.90(6.07) 12.36(1.98) 12.49(4.32) 9.16(3.98) Range 2.27-2.50 4.82-6.01 6.61-15.20 10.82-15.16 5.26-16.88 1.11-15.60 1relative.entropy - { El, p * ln(p.O]}/ln0i)where.’eE" " "Idlr°P ortion of landarea. in land-usecate. ~o.ryi, and k - numberof land-use categories; rangesbetween 0 and1, where0 slgni/ies land devotedto a single use and 1 signifies all land area evenlyspread amongall uses.

San Francisco Commercial/CivicCenter: These two stations -- Powell and Civic Center -- represent the remainingdowntown San Francisco stations, serving the region’s major shoppingdistrict (Powell) and institutional-cultural complex(Civic Center). They relatively high employmentdensities (though muchlower than the Office Center stations) andby far the highestresidential densities of all station classes. Still, relatively little land area aroundthese stations is devotedto housing.As part of the downtown, these stations rate fairly high in terms of the levels of mixeduses. Theyhave no parking but like the Office Center enjoyhigh levels of surface transit connections. DowntownOakland: These two stations -- City Center (12th Street) and 19th Street serve the Bay Area’s second-tier urban center, downtownOakland. Employmentdensi- ties in downtownOakland fall belowthose of downtownSan Francisco, but are consid- erably above those of the remaining BayArea. DowntownOakland is less segmented than downtownSan Francisco, with office, retail, and services intermingled;this is reflected by the high relative entropy index, signifying a rich mixtureof land uses. Compared to downtownSan Francisco, downtownOakland has more housing in the immediate vicinity, thoughaverage household incomes are low. The City Center station lies in a redevelopmentdistrict; the redevelopmentauthority has recently used tax increment financing and other incentives to attract newdevelopment, including a mixedretail-

12 office plaza withattractive landscapingthat ties directly into the station anda large federal building complex.These stations have no parking, but are the major terminuses of buses operated by ACTransit, whichserves the urbanized parts of Alamedaand Contra Costa Countiesin the East Bay. ¯ UrbanDistricts: Thesefour stations -- Berkeley(downtown), Lake Merritt, Mission-16th Street, and Mission-24thStreet -- lie outside of the region’s two big CBDs,but are in highly urbanizedareas. In the urban hierarchy, they represent third-tier centers. They are maturedistricts, with considerablenumbers of jobs (in low- to mid-rise buildings) and significants amountsof housing. Amongall station classes, they have the highest gross residential densities andrelatively high shares of land devotedto residential uses. Thesestation areas are also mostbalanced in terms of jobs and housing.Retail is prominent aroundall except the Lake Merritt station. DowntownBerkeley has the most mixed office-retail-residential development.The Lake Merritt station area is predominantlya governmentemployment district surrounding by mid-rise housing and a sprinkling of retail uses. Oakland’sChinatown, cultural complex,and LaneyCollege also flank the LakeMerritt station. The two Missionstations, serving the traditional Hispanicdistrict of San Francisco, feature very similar mixesof small, independentlyowned retail outlets interspersed by moderate-incomehousing. Only the Lake Merritt station has parking (just 205 spacesthat cost a quarter per day to park), andall four stations are well-served by bus transit connections.Relatively high shares of residents aroundthese stations commuteby rail transit. ¯ SuburbanCenters: Thesefive stations -- Orinda, Lafayette, WalnutCreek, Pleasant Hill, and Concord-- are surroundedby fourth-tier commercialcenters in the eastern suburbs of the BayArea. As shownin Figure 4, they are also aligned along the Concordcorridor serving Contra Costa County. The three outermost stations -- WalnutCreek, Pleasant Hill, and Concord-- are surroundedby mid-rise office towers, while the Orindaand Lafayettestations flank commercial-retaildistricts with relatively few offices nearby. All five stations haveapartments nearby (especially Pleasant Hill, whichhas over 1,600 apartmentunits within a quarter-mile of the station); middle- and upper-middle-income single-famil)~ detachedhousing generally lies beyondthese apartments.Overall, gross residential densities are fairly low in these station areas and averagehousehold incomes are comparativelyhigh. In general, householdincomes and rent gradients decline along the Concordcorridor as one travels outwardfrom the Orinda station -- Orlndabeing a fairly well-to-do communityand Concordhaving muchlarger shares of moderate- incomehouseholds. What most distinguishes these stations are the large volumeof park- and-ride.... spaces -- ranging from1,380 at Ormdato 3,24511 at Pleasant Hill. Large shares of residents living within one-half mile of these stations commuteto workby rail transit -- on average,around 12.5 percent. This reflects the relatively high shares of residents in Orinda and Lafayette workingin downtownSan Francisco. All five stations lie near a freeway;both Orindaand Lafayette stations lie in the medianof Highway24, thus inhibiting developmentimmediately at the station. ThePleasant Hill station is distin- guishedfrom the other stations for being in an unincorporatedarea and being part of a redevelopmentdistrict. The formationof a redevelopmentdistrict in the early-1980sat the Pleasant Hill station has helpedleverage over 1.5 million squarefeet of newoffice space constructionand five large apartmentcomplexes within a quarter-mile of the station in the past sevenyears (see Cervero,Bernick, and Gilbert, 1994). ¯ Low-DensityAreas: The remaining19 BARTstations form a station class of low-density development.What most distinguishes these station areas is their comparativelylow

13 employmentand residential densities. All lie in low-rise, suburban-like settings. Most are surrounded by predominantly residential development (e.g., Glen Park and Balboa Park neighborhoodsin San Francisco), though somehave prominent retail districts nearby (e.g., Rockrldge in Oakland and Bay Fair in San Leandro) and others are surrounded industrial and vacant land uses (e.g., Coliseum in Oakland and South Hayward). general, these areas have relatively low levels of land-use mixing. Moststations in this class have moderate supplies of parking, ranging from none at Balboa Park in San Fran- cisco to 2,516 at the El Cerrito del Norte station on the Richmondllne. Bus transit connections tend to operate at lower service levels at these stations, except at the MacArthur station in Berkeley which functions as a transfer station between the three East Bay BART lines. Several of the stations on the Richmondllne (El Cerrito, E1 Cerrito del Norte, and Richmond)and the Fremont line (San Leandro, Hay-ward, and Union City) lie within redevelopmentdistricts. The most significant redevelopmentactivities have been near the E1 Cerrito del Norte station, where new housing and retail projects have opened in recent years (see Cervero, Bernick, and Gilbert, 1994). Five of the station areas -- Rock- ridge, North Berkeley, Ashby, Glen Park, and Balboa Park -- are notable for the restrictive zoning introduced after BARTwas opened, aimed at limiting preserving the single- family residential characters of these neighborhoods.

6. ANALYSIS OF MODES OF ACCESS ACROSS STATION CLASSES This section examinesdifferences in modesof access across these six classes of BARTstations. Accesstrips to and egress trips from classes of BARTstations are also stratified as either home-end(going from or to home) or work-end(going from or to a workplace). Thus, four submarkets of access trips are examined: 1. Home-endaccess trips: trips from hometo a BARTstation. 2. Home-endegress trips: trips from a BARTstation to home. 3. Work-endaccess trips: trips from a workplace to a BARTstation. 4. Work-endegress trips: trips from a BARTstation to a workplace. For the typical commute,the first and fourth trips would occur in the morning and the second and third wouldoccur in the afternoon/evening. In general, modalsplits should be similar for the two home-endtrips and the two work-endtrips. Whatis expected to differ most as a function of land-use environment are differences in home-endtrips versus work-endtrips. Accordingly, this section empha- sizes differences in modesof access and egress (combined) between home-endand work-endtravel. To the degree that land-use environments vary amongstation classes, modal splits for access and egress trips should also vary amongthe classes, as postulated by the research hypotheses. In this section, Analysis of Variance (ANOVA)results comparingdifferences amongstation classes in the mean percentages of access and egress trips by modesare presented. Automobileaccess trips are first examined,followed by access trips by kiss-and-ride and then access and egress trips by transit and walking (at both the home- and work-ends). Table 4 shows the ANOVAresults for home-endaccess trips by automobile, revealing statistically significant differences amongstation classes. (Results for home-endegress trips were virtually identical,

14 indicating symmetryin modal splits of access to and egress trips from home; also, few work-endaccess or egress trips are by automobile, so these submarkets are not examined.) As expected, few home-end access trips to BART’sdowntown stations are by automobile, indicating that dense, mixed-use settings, along with the absence of park-and-ride facilities, discourage auto access. As station areas becomemore suburban-like in character, automobile access shares increase. Percentages of home-endaccess trips by auto seem strongly related to densities and even more strongly related to parking supplies. 12 Suburban centers, with the largest average parking supplies and comparatively high average incomes in surrounding households, average the highest share of park-and-ride trips. Thus, while land-use environments clearly seemto influence auto access trips, so do non-land-use factors, particularly parking supplies.

Table 4. ANOVAin Percent of Home-End Access Trips by Automobile Across Station Classes DependentVariable: Percent of Home.EndAccess Trips by Automobile Standard Summary.Statistics Station Class Mean Deviation F Statistic Probability San Francisco Office Center 1.5 0.71 26.22 .000 San Francisco Commercial/ Civic Center 5.0 1.41 Downtown Oakland 9.0 1.41 UrbanDistricts 15.5 12.02 SuburbanCenters 71.2 11.97 Low-DensityAreas 53.9 13.49 TOTAL 42.9 26.24

Tables 5 and 6 break downhome-end automobile access trips by the two major types: drive- alone and klss-and-ride (passenger drop-off). Relationships for drive-alone trips were very similar those of total automobile trips: virtually no one reaching downtownstations from homedrive alone, only around 7 percent of those going to urban district stations (which also have few parking spaces) drive alone, and betweenone-third and one-half of those going to all remaining stations (all of which have low densities) drive alone. Table 6 shows that for the entire BARTsystem, only around one of ten home-endaccess trips are by kiss-and-ride; for access trips to suburban centers and low-density station areas, around 13-14 percent are by kiss-and-ride. The ANOVAresults for home-end access trips by transit are shown in Table 7.13 No strong patterns are evident. While San Francisco’s Commercial/CivicCenters had the largest share of transit access trips, the nearby San Francisco Office Center had the lowest share. Transit service levels, which were not used to classify stations, 14 as well as density probably explain someof the differences in home- end access trips. San Francisco’s Commercial/CivicCenter stations receive intensive feeder bus, streetcar, and cable car services, and thus average high share of transit access trips. While San Francisco’s

15 Table 5. ANOVAin Percent of Home-End Access Trips by Drive-Alone Automobile Across Station Classes DependentVariable: Percent of Home.EndAccess Trips by Drive.Alone Automobile Standard Summary.Statistics Station Class Mean Deviation F Statistic Probability

San Francisco Office Center 1.00 0.00 14.65 .000 San Francisco Commercial/ Civic Center 2.00 0.00 Downtown Oakland 2.50 1.29 Urban Districts 7.05 7.07 Suburban Centers 49.20 12.32 Low-Density Areas 33.84 12.56 TOTAL 27.03 29.58

Table 6. ANOVAin Percent of Home-EndAccess Trips by Kiss-and-Ride, Across Station Classes DependentVariable: Percent of Home-EndAccess Trips by Kiss.and.Ride Standard Summary.Statistics Station Class Mean Deviation F Statistic Probability San Francisco Office Center 0.50 0.71 26.68 .000 San Francisco Commercial/ Civic Center 2.00 1.41 Downtown Oakland 5.55 0.57 Urban Districts 6.47 3.53 Suburban Centers 13.80 1.30 Low-Density Areas 13.32 2.58 TOTAL 10.65 4.99

Table 7. ANOVAin Percent of Home-EndAccess Trips by Transit, Across Station Classes DependentVariable: Percent of Home.EndAccess Trips by Transit Standard Summary.Statisti~ Station Class Mean Deviation F Statistic Probability San Francisco Office Center 8.00 0.00 1.55 .207 San Francisco Commercial/ Civic Center 23.50 9.19 Downtown Oakland 12.50 2.12 Urban Districts 16.00 6.27 Suburban Centers 12.20 7.46 Low-Density Areas 20.16 9.68 TOTAL 17.53 9.05

16 Office Center stations also enjoy intensive feeder services, the area is so dense and compactthat most nearby residents can more easily reach the station by foot. Manysuburban stations have higher transit modal splits for home-endaccess trips than more urbanized ones. Similarly insignificant results were found for the share of work-endegress trips by transit. Table 8 shows low-density areas averaged more work-endegress trips by transit than downtownstations. Overall, it appears transit plays a relatively modest feeder role in more urbanized station settings, partly because destinations can be more easily reached by foot. Its role as an access and egress modeis most vital in low-density station areas, where manyorigins and destinations are beyond a convenient walk.

Table 8. ANOVAin Percent of Work-End Egress Trips by Transit, Across Station Classes DependentVariable: Percent of Work-EndEgress Trips by Transit Standard Summary.Statistics Station Class Mean ~ F Statistic Probability San Francisco Office Center 12.00 2.82 2.07 .098 San Francisco Commercial/ Civic Center 16.50 4.95 Downtown Oakland 12.50 6.36 UrbanDistricts 18.00 7.48 SuburbanCenters 16.60 9.02 Low-DensltyAreas 27.42 12.23 TOTAL 22.29 11.63

These results are reinforced by the ANOVAresults, shown in Tables 9 and 10, on market shares of home-endaccess and W0rk-endegress stations by walking. Shares of pedestrian access and egress are highest in the densest, most mixed-use settings. Aroundnine of ten access trips from hometo San Fran- cisco Office Center stations (ostensibly as part of a reverse commute)are by foot. For suburban center and low-density station areas, fewer than one of five BARTusers walk from their homesto stations. Higher shares of those going to and from low-density station areas walk than their counterparts using suburban center stations, ostensibly because suburban center stations have the most generous supplies of parking. Overall, the ANOVAresults support the research hypotheses. Dense, urban station areas have relatively high shares of walk access trips. Stations in suburbansettings are reached mainly by private automobiles. The land-use environment appears to have the weakest influence on transit access and egress. Transit service levels are likely moreimportant (see next section).

17 Table 9. ANOVAin Percent of Home-End Access Trips by Walking, Across Station Classes DependentVariable: Percent of Horne.EndAccess Trips by Walking Standard Summary.Statistics Station Cla~ Mean ~ F Statistic Probability San Francisco Office Center 89.50 0.71 39.27 .000 San Francisco Commercial/ Civic Center 70.00 11.31 Downtown Oakland 70.50 10.61 UrbanDistricts 71.75 6.61 SuburbanCenters 15.40 4.82 Low-DensityAreas 18.84 11.81 TOTAL 35.76 26.87

Table 10. ANOVAin Percent of Work-EndEgress Trips by Walking, Across Station Classes DependentVariable: Percentof Work-EndEgress THpsby Walking Standard Summary.Statistics Station CI~ Mean Deviation F Statistic Probability San Francisco Office Center 86.50 3.54 32.73 .000 San Francisco Commercial/ Civic Center 87.00 5.65 Downtown Oakland 78.00 0.00 UrbanDistricts 76.00 7.81 Suburban Centers 19.60 7.26 Low-DensityAreas 28.05 12.89 TOTAL 42.73 26.56

7. PREDICTORS OF ACCESS AND EGRESS MODES This se~:tion builds upon the previous one by presenting regression models which directly predict the influences of land-use variables as well as other factors (e.g., parking supplies) on the percentages access and egress trips by different modes. Stations and their surrounding one-half mile areas served as thes cases for estimating regression models-- 34 cases in all) Table 11 presents a multiple regression modelthat predicts the percentage of access trips (home- end and work-end combined) to BARTstations by automobile. The hypotheses postulated in the research are supported by these results. Market shares of automobile access trips fall with both employmentand residential densities and mixedland uses around stations, and rise with park-and-ride spaces. The model suggests, for instance, that an increase in station-area residential densities of 10 householdsper acre lowers the share of access trips by automobileby 10 percent, ceteris paribus. Fromthe coefficient on the entropy

18 Table 11. Regression Model for Predicting Percentage of Access Trips to BARTStations by Automobile DependentVariable: Percent of Access Trips by Automobile Variable: Coefficielat ~alag[argl~.xx~

Employees/acre within one-half mile of station -0.198 0.064 .003 Households/acre within one-half mile of station" -1.00 0.311 .003 Percent of land area within one-half mile of station in commercial use -0.729 0.304 .023 Percentof land area withinone-half mile of station in residentialuse -1.029 0.339 .005 Entropyindex of land-usemixture within one-half mile oft station -85.334 33.894 .032 Park-and-ridespaces at station 0.014 0.003 .000 Constant 169.691 45.058 .001 SummaryStatistics: R2 - .869 F - 29.88, prob. - .000 No. of cases - 34 Note: 1Relativeentropy - {Zi[Pi * ln(p.~]}/ln(k) wherePi " proportionof land area in land-use categoryi and k - numberof land-use categories rangesbetween 0 and1, where0 signifies land devotedto a single use and1 signifies 1and’areaevenly spread among all uses. variable, wesee that a station area that has an evendistribution of acreageamong land uses could be expectedto have85 percent feweraccess trips by automobilethan a station area with just one use, like offices, assumingall other characteristics of the station areas, includingdensities, wereidentical. Parking supply, on the other hand, inducesautomobile access -- every 100 spaces is associated with 1.4 percent moreaccess trips by automobile,all other things held constant. Overall, the modelpredicts around87 percent of the variation in access trips by automobilefor the 34 stations. In the case of egress trips from BARTby transit, Table 12 shows, consistent with the ANOVA results, that land-usevariables exerted a relatively weakinfluence. Transit modalshares increasedas station-area densities declined, especially residential densities. This is consistent with the ANOVAfind- ings that suburbancenter and low-densityarea stations averagedthe highest transit access and egress modal splits. Far moresignificant in explainingtransit egress modalsplits is transit service levels, measuredin route miles per 1,000 householdswithin one-half mile of stations. Wherebus feeder connectionsare good and densities are low, transit egress modalsplits tend to be high. Theexceptions, again, are the downtown stations, wheremost destinations are close enoughto stations that mostegress trips are by foot, despke the very high quality of transit connectionsto downtownstations in San Francisco and Oakland.Over- all, the modelshown in Table12 explains two-thirdsof the variation in transit egress modalsplits. (The modelfor transit access modalsplits wasnearly identical, andis thus not presented.) The modelsfor access versus egress walkingtrips varied somewhat;thus both are shown.Table 13 reveals that the percentof accesstrips by foot rises sharplywith densities (especiallyresidential densi

19 Table 12. Regression Model for Predicting Percentage of Egress Trips from BARTStations by Transit DependentVariable: Percent of Egress Trips by Transit Y_ariah~ Coefficient ~t~aifla,rA_F.Im~ Probability Employees/acre within one-half mile of station -0.157 0.029 .000 Households/acre within one-half mile of station -0.669 0.174 .000 Transit service levels, in route miles per 1,000 households1 within one-half mile of station 4.984 0.707 .000 Constant 12.482 2.611 .000 Summary.Statistics: R2 - .667 F - 20.11, prob. - .000 No. of cases - 34 No re: tRoutemiles of all surfacetransportation, including bus transit, streetcar trams, light rail transit, andcable car services, within one-halfmile of rail station, excludingBART services.

Table 13. RegressionModel for PredictingPercentage of AccessTrips to BARTStations by Walking DependentVariable: Percent of A ccess Trips by Walking Variable: Coefficient ~mchaX[a~.ZJL~ Probability Employees/acre within one-half mile of station 0.330 0.057 .000 Households/acre within one-half mile of station 1.130 0.314 .001 Percent of land area within one-half mile of station in residential use 0.532 0.312 .100 Entropy index of land-use mixture within one-half mile1 of station 55.746 35.308 .127 Park-and-rlde spaces at station -0.020 0.004 .000 Transit service levels, in route miles per 1,000 households2 within one-half mile of station -3.121 1.099 .009 Terminal3 or near-terminal station (0-no, 1 ,-yes) 19.569 6.886 .009 Constant -18.664 42.474 .664 SummaryStatistics: R2 - .887 F - 29.30,prob. - .000 No.of cases- 34 Notes: tRelativeentropy - {Y-i[Pi ’* ln(p3]}/ln0i ) wherep.t - proportionof land area in land-usecatel~ory i andk - numberof land-use categories. Rangesbetween 0 and 1, where0 sigriifies lanc] devotedto a single use and 1 signil%sla.dd area evenlyspread among ~l Llse.$. Routemiles of all surfacetransportation, including bus transit, streetcar trams, light rail transit, andcable car services, within pne-halfmile of rail station, excludingBART services. . . . _ aNear-terminalrepresents stations towardthe end of the llne that funettonlike te .rt~mals becausethey .are closer to freeways than the actual terminals and thus serve a larger catchmentarea. BART’snear-termm.al statmns:El Cerrito del Norteand PleasantHill, havelarger suppliesof parkingthan terminalstations since they are easter to reach by freeway.

ties) andmixed land use levels, and declinesas substitutes to walkingare moreplentiful -- i.e., lots of parking and goodtransit connections.According to the model,an increase in residential densities of 10 householdsper acres is associatedwith an 11.3 percentincrease in the share of access trips by walking,

20 controlling for parking supplies and other explanatory variables. Interestingly, Table 13 showsthat once parking supplies and other factors are controlled for, terminal and near-terminal stations tended to have higher levels of access trips by foot, despite their freeway and highwayorientations. This finding largely reflects the presence of several large apartment complexesin the vicinity of the two near-terminal stations, Pleasant Hill and E1 Cerrito del Norte, yielding high shares of walking-accesstrips to these two stations. Overall, the model shownin Table 13 produced the best-fitting modelof all the regression modelsesti- mated, explaining around 89 percent of the variation in walk-access modalsplits. Table 14 showsthat the relationships for explaining walk egress trips were fairly similar, though land-use variables exerted an even stronger influence in this model. Controlling for densities, parking supplies, and other factors, for instance, Table 14 suggests that a station area with an even mix of land uses will average 73 percent more egress trips by walking than one with a single land use. Adding 10 more jobs per acre can be expected to induce 3.33 percent more egress trips by foot, holding other factors constant. Workingagainst walk egress trips are parking supplies, transit service levels, and interestingly, the presence of freeway medians. Table 14 suggests that BARTstations in freeway medians average around 7 percent fewer egress trips by foot, controlling for densities and other factors. This finding suggests that the quality of walking environment could have some bearing on walking egress (and, by exten- sion, walking access, though this variable was not very significant in the access model shownin Table

Table 14. RegressionModel for PredictingPercentage of EgressTrips from BARTStations by Walking DependentVariable: Percent of Egress Trips by Walking Variable: Coefficient ~lkclkrALF.ir~ Probability Employees/acrewithin one-half mile of station 0.333 0.057 .000 Households/acrewithin one-half mile of station 1.556 0.313 .000 Percent of land area within one-halfmile of station in residential use 0.637 0.310 .050 Entropyindex of land-use mixturewithin one-half milet of station 73.577 37.090 .058 Park-and-ridespaces at station 43.012 0.003 .000 Transit service levels, in route miles per 1,000 households2 within one-half mile of station -3.629 1.054 .002 Station located in freewaymedian (0-no, 1-yes) -6.892 4.323 .131 Constant -35.37 42.293 .411 Summary.Statistics: R- .886 F - 28.91, prob. - .000 No. of cases - 34 Notes: 1Relativeentropy - {lgi[pl* in(p.~]}/In(k)where Pi - proportionof landarea in land-usecategory i, andk - numberof land- usecatel~ories. Ranges between 0 and1, where0 signi/ieslanddevoted to a singleuse and1 si~fifies land areaevenly spread ~i~nongan uses. Routemiles of all surfacetransportation, including bus transit, streetcartrams, light rail transit, andcable ear services,within one-halfmile of tall station, excludingBART services.

21 13). Freeway medians represent barriers to movementin manyways -- physically, visually, psychologi- cally, and symbolically. The vibrations created by heavy freeway traffic and shadowscast by elevated freeway structures can also discourage foot travel. To summarizethe results of this section, Table 15 presents the regression results as midpoint elasticities - revealing the relative sensitivity of access and egress modalsplits to changesin land-use and other variables.16 Thetable showsthat, in general, access and egress trips are inelastically related to the land- use environment,though the influences of land-use variables are generally as strong as other factors. Access and egress modal splits were most sensitive to land-use mixtures -- highly mixed uses around rail stations encourage foot travel and discourage automobile usage. Having considerable amounts of compact residential developmentwithin one-half mile of stations also significantly induces walk access and reduces automobileaccess. In general, access and egress modalsplits were moresensitive to residential than employmentdensities. Parking supplies exerted stronger influences on access and egress modal splits than did land-use densities, though less of an influence than land-use mixtures. Parking supplies had their strongest effect on reducing walk trips from homesto stations. Transit access and egress was strongly influenced by feeder service levels; land uses played a relatively modestrole in affecting transit travel to and from stations. Lastly, besides parking lots, the physical characteristics of stations, such as lying in the medianof a freeway, had a fairly weakinfluence on access and egress modalsplits.

Table15. MidpointElasticities of Accessand Egress Modal Shares as Functionsof Land-Useand OtherVariables Independent Automob~e Transit Walk Walk Variable: &cicu/.F.gr. ^cces Employees/acrewithin one-half mile of station -.116 -.177 .220 .196 Households/acrewithin one-half mile of station -.209 -.270 .269 .328 Percent of land area withinone-half mile of station in commercialuse -.339 ------Percent of land area within one-halfmile of station in residential use -1.167 -- .733 .775 Entropyindex of land-use mixture within one-half1 mile of station -1.281 -- .989 1.153 Park-and-ridespaces at station .300 -- -.484 -.257 Transit service levels, in route miles per 1,000 households2 within one-half mile of station -- .888 -.328 -.337 Terminalor near-terminalstation (0-no, 1 -yes)~ -- -- .093 -- Station located in freewaymedian (0=no, i,.yes) .... ¯ 029 Notes: 1Relativeentropy .- {~:i[Pi* ln(pi)]}/ln(k)where pl " proportionof landarea in land-usecategory i, andk .. numberof land- use categiories.Ranges between 0 and 1, where0 signities’landdevoted to a singleuse and1 signifiesland area evenly spread i~nongan uses. Routemiles of all surfacetransportation, indudlng bus transit, streetcartrams, light rail transit, andcable car services,within 9he-halfmile of rail station, excludingBART services. "Nearterminal represents stations toward the endof the line that functionllke terminalsbecause they are closerto freeways than the actualteiminals and thus servea larger catchmentarea. BART’snear terminalstations, H Cerrltodel Notreand PleasantHill, havelarger suppliesof parkingthan terminal stations since they are easier to reachby freeway.

22 The results presented in Tables 11 to 15 are also noteworthyfor the variables which did not enter the models as significant predictors. Averagehousehold incomes in the vicinity of stations, for instance, had no significant effect on modesof access and egress. Nor did factors like station function (e.g., transfer stations) or proximity of freeways. Far more important were the land-use environment -- mixtures of use and densities -- as well as two supply-s~devariables -- parking spaces and transit service levels.

8. ACCESS MODES AS A FUNCTION OF DISTANCE TO STATIONS In this section, the influences of distance on modesof access to and egress from the six classes of BARTstations are examined, for both home-endand work-endtrips. This analysis allows the interactive effects of the land-use setting and distance on access and egress modalsplits to be explored. Table 16 showsthe meansand standard deviations of access and egress trips to and from each station class. For each station class, access trips tend to be muchlonger than egress trips, particularly in the case of the downtownSan Francisco and Oakland stations, wherein egress trips are around 10 to 12 times as long. While home-endaccess trips to the downtownSan Francisco and Oakland stations tended to be the longest amongstation classes, egress trips from these stations to workplacestended to be the shortest. (It should be noted that downtownstations average far more work-end egress trips than home-endaccess trips -- typically by a ratio of 10:1 -- so, overall, trip connections to and from these stations tend to be short.) For the non-downtownstations, home-end access trips (which by far predominate over work- end egress trips) are most often in the 2- to 3-mile range.

Table 16. Summary Statistics for Access and Egress Trips by Station Class Ratio of Mean AccessTrip Length to Home-End AccessTrips Work-end EgressTrips MeanEgress Station Class Mean S2~[..D~ffialJ~a Mean ?~2¢3fiafima San Francisco Office Center 3.62 2.19 0.31 0.14 11.7:1 San FranciscoCommercial/ Civic Center 4.68 2.71 0.38 0.18 18.3: 1 Downtown Oakland 3.86 2.54 0.36 0.21 10.7: 1 UrbanDistricts 2.35 2.24 0.38 0.17 6.1 : 1 SuburbanCenters 3.00 2.11 0.51 0.30 5.9 : 1 Low-DensityAreas 2.87 2.08 0.72 0.43 4.0 : 1

In the figures that follow, access and egress modalsplits are plotted for distances of up to 2 to 3 miles. While longer access and egress trips are made, there tends to be relatively few samplecases over these longer distances. Including longer access and egress trips wouldhave generated less interpretable plots due to sampling bias (and seemingly randomappearance) of modal splits for distances of 4 miles and longer. Over such longer distances, the influences of sampling errors would have overly dominated the

23 graphs)7 Beyond4 or so miles, moreover, modal splits tend to vary muchless (e.g., automobile travel dominatesfor home-endaccess trips to stations in low-density areas), so extending plots out any fa,’ther usually reveals little newinformation.

San Francisco Office Center Figure 4 showsthat walk trips are tl~e predominant egress modefrom San Francisco Office Cen- ter stations (Embarcaderoand Montgomery)to workplaces up to around 9/16 of a mile (3,000 feet). Beyondthis distance, surface transit (bus, tram, light rail, and cable cars) serve the majority of trips. Beyondone mile, virtually all egress trips to workplacesare by someform of feeder transit. Figure 5 indicates access trips from hometo San Francisco Office Center stations follow a similar pattern, with transit becomingthe dominant access modeat a distance of around 5/8 of a mile (3,300 feet). Wherestations are are around two miles from a BARTcommuter’s home, around 20 percent of access trips to San Francisco Office Center stations are by automobile. Comparedto work-endegress trips, a wider variety of access modesare used for home-endtrips. Figure 6 combinesall home-endand work-endaccess and egress trips. Overall, the Bay Area’s densest, most urban rail station areas are associated with two distinct access features: (1) high levels transit riding for longer access trips -- over 80 percent for distances beyonda mile; and (2) relatively long walk trips, with walking being dominantaccess modefor trips as far as 4,000 feet to and from stations.

San Francisco Commercial/Civic Center Figures 7, 8, and 9 reveal similar relationships betweendistance and access/egress modalsplits for the San Francisco Commercial/CivicCenter stations (Powell Street and Civic Center). In these fairly dense, very mixed-use settings, walking is the dominant access and egress modeup to around 3,200 feet. Comparedto the Office Center stations, walking predominates over slightly longer distances for access trips and shorter distances for egress trips. The hillier terrain separating nearby residential areas and the PowellStreet and Civic Center stations, as well as the high levels of feeder transit services into these two stations, likely account for transit’s popularity as access/egress modesto these stations. Figure 8 shows transit’s market share exceeds 95 percent for all access and egress trips to and from the commercial/civic center stations that are in the 2- to 3-mile range; at 3 miles, however, over 10 percent of BARTpatrons access or leave these stations as automobilepassengers or by bicycle.

Downtown Oakland Up to 3,500 feet, the majority of egress trips from downtownOakland’s 12th Street and 19th Street stations are by foot (Figure 10). Public transit captures virtually all egress journeys beyondthree- quarters of a mile. These relatively long walking-egress distances in downtownOakland could reflect the influences of two factors: (1) Oakland’sdowntown is less compactand its feeder transit services are

24 0.9

0.8

0.7 :>, GI 0.6 / I- - - Bicycle (3. ! I" .... OriveAtone

I_. _.~_T,?es~!...... C O 0.4 ~2 0 ~ 0.3 / 0.2

0.1 /

"°"i .... ~r zo o* o~ ,o Distancefrom Hometo BARTStation, Miles

Figure 4. Mode of Egress for Commute Trips from San Francisco Office Center BARTStations to Work

0.9

0.8

0.7 fa 0.6 D. I------Bicycle Inn Drive AloneI J ...... 0.5 Shared Ride [ .... Kiss&Ride I ] .... t- O 0.4

O Q. 2 0,3 I1.

0,2

0.1 \

¢:) Distancefrom Hometo BARTStation, Miles Figure5. Modeof Accessfor CommuteTrips fromHome to SanFranczsco Office CenterBART Stations

25 o9 ~ J~----~~

0,7

" Walk 0.6 Bicycle / I ...... o,,~eA,oooI 0 5 X/ I .... s.~,edR,~e I Kiss & R,ae/P,ck-Up /~ I .... Transil J

0.20.30.4 ~ ~__

0.1 /--~ ~-.... o -.~==~-~ ...... ~ --’.:;=-’-----.’-.-’=’"~’~’~" ~ ..’: .... ".~’~’--~ ......

Distancefrom Home to BARTStation, Miles

Figure 6. Modeof Access/Egress for All San Francisco Office Center BARTCommute Trips

0.9

0.8

0.7

Wa~ 0.6 ------Bicycle ...... Drive Alone .... Sha~ed Rk~e 0.5 .... PassengerPick.Up Transil 0.4

0,3

0.2.

0.1

o Distancefrom Home to BARTStation, Miles Figure 7. Modeof Egress for CommuteTrips from San Francisco Commerclal/Civic Center BARTStations to Work

26 0.9

0.7

m o...... Drive Alone 0.5 .... SharedRide Kiss & Ride c Transit O 0.4 ~2 0 O 0.3

0.2 /’’~’,. ." \

/ \ 0.1 ,, / ¯ \ ,’" ..... 0 to .~ co

Distance from Hometo BARTStation, Miles

Figure 8. Modeof Access for CommuteTrips from Hometo San Francisco Commercial/CivicCenter BARTStations

0.9

0.7 ¢¢ r~ Walk ~Q. 0.6 ...... OdvaA(ona es Shared.~J i. 0.5 I .... Kiss~ ~idelVi~.Up /I ~ ] .... J Transit I 0,4

0,3 O.

0.2

0.1

Distance from Hometo BARTStation, Miles

Figure 9. Modeof Access/Egress for All San Francisco Commercial/Civic Center BARTCommute Trips

27 I

0.8

0.7 .... :- W; ------8icicle 0.6 ...... Drive A~one .... Shared R,de 0.5 I.... PassengerP ¢[,,,-Up Transit

0.4

0.3

0.2 /

o Distance from Home to BART Station, Miles

Figure 10. Mode of Egress for Commute Trips from Downtown Oakland BART Stations to Work less frequent, resulting in people walkingmore to reach destinations that tend to be farther fromstations (e.g., Jack LondonSquare, Kaiser Center complex); and (2) Oakland’sdowntown has the widest variety of mixeduses amongstation classes (see Table 3), whichcould encouragesome people leaving downtown stations to walkto their jobs andother destinations. For access trips to downtownOakland stations, Figure 11 showstransit eclipsing all other modes for distances beyond5/8 of a mile. Beyond11/16 of a mile, Figure 11 showsa greater variety of access modesthan was the case for the downtownSan Francisco stations. Themodal split patterns are not tidy, however.Kiss-and-ride modalsplits exceed those for walkingbeyond one mile. For access trips over 1/2 mile, somedegree of park-and-rideand kiss-and-ride activities can be found at the downtownOakland stations. Theavailability of surface parkingat daily rates of $2-$3 in downtownOakland likely has attrac- ted somepark-and-ride activities. The compositeplot of access and egress trips to downtownOakland is shownin Figure 12. Overall, transit becomesthe dominantmode at 5/8 mile and kiss-and-ride modalsplits exceed those of walkingat 7/8 mile.

28 0.9

0,8 i" ..... Drive Alone .... Shared Ride .... Kiss & R/de 0.7 Transit

..~

/ ; I | i i" 0.2 , . ,~ o * o"

oO°’°’’,¯ / /-..~ 0.1 ,o° ¯ : ¯ / "

Distance from Hometo BARTStation, Miles

Figure 11. Modeof Access for CommuteTrips from Hometo DowntownOakland BARTStations

0,9

0,8

0.7 r~

¢n 0.6 n ~. 0.5

0.4

0 D. 0 0.3. I1.

0.1 "-~.o...

Distance from Hometo BARTStation, Miles

Figure 12. Modeof Access~Egress for all DowntownOakland BARTCommute Trips

29 Urban District Stations J For egress trips fromurban district stations (Berkeley,Lake Merritt, Mission-16thStreet, Mission- 24th Street), Figure 13 showsthat modalsplit patterns weresimilar to those of downtownBART stations for distances up to aroundone mile. Beyond1-1/2 miles, there wereso few samplecases of egress trips fromurban district stations as to makethe plotted informationfairly meaningless. Far moreaccess trips from hometo urban district stations occurred beyondone mile, producing the moreinterpretable patterns shownin Figure 14. Walkingis again shownto be the predominantaccess modeto urban district for distances up to 5/8 mile. Transit access generally dominatesover the 5/8-mile to 1-3/4-miledistance, beyondwhich most access trips are by private automobile(drive-alone and kiss- and-ride). Bicycleaccess plays a minorrole for access trips up to around1-1/2 miles, foundmainly at the Berkeleystation. Thecomposite graph of access and egress trips to and fromurban district stations is shownin Figure 15. Thedominance of home-to-stationaccess trips on modalsplits is revealed by this figure.

Suburban Center Stations A starkly different pattern of modalsplits for home-endaccess trips wasfound for suburbancenter stations (Figure 16). Whilewalking predominates for all access trips up to one-half mile, beyondthat distance park-and-rideaccess dominates.Even at a 1/8-mileaccess distance (660 feet), over one-quarter home-endaccess trips are by drive-alone motorists or kiss-and-riders. At distances beyondone mile, well over half of access trips are by park-and-ride.Kiss-and-ride’trips occur mostfrequently over the 1/2- to 1-mile distance. Beyondtwo miles, transit is the secondmost common access mode.Over the 1- to 3- mile range, riding as a passengerin a car accountsfor around10 to 15 percentof accesstrips. Amongegress trips from suburbancenter stations to workplace,Figure 17 showsthat mosttrips under 3,000 feet wereby foot. Beyondthis distance, transit predominates,though some level of passenger pick-up occurs over the 1/2- to 3-mile distance. There are so few egress trips to workplacesmore than 1-1/4 miles fromsuburban center stations that the right side of the graphis no doubtdistorted by samplingerrors. Thecomposite graph for access and egress trips to and fromsuburban center stations is shownin Figure 18. Here, the patterns seemclearest. Upto distances of around3,000 feet, walkingaccess and egress predominate.Beyond one mile, fewer than one out of ten access and egress trips are by foot. Beyond5/8 mile, park-and-rideaccess predominates.Passenger pick-up and drop-off (kiss-and-ride) are mostcommon over the 1/2- to 2-mile range. Beyondaround 3/4 mile, transit access and egress captures a 15-20 percent marketshare.

30 . .Wa~ - - - e~vcSe I 0.9 ...... Drive Alone I .... Shared Ride .... PassengerPick.Up JI 0,8 I , Transit J

0.7

0.2

0,1 o,°°" { ¯ ° I, ,o°° ..... I ", l 0

o Distance from Hometo BARTStation, Miles

~irgUreom Urban 13. DistrictModeof EgressBART Stationsfor Commute to WorkTrips

Walk 0.9 --- -- Bicycle - - ¯ Drive None .... Shared Ride 0,8 Kiss & Ride Transit

.o°-o°.. ° oO’°" 0.2 ¯ ’" " I "%--. ~’. _ /

,o , ° ¯ _....~,~’’~’\" , ~., ¯ o ~ ¯ ¯ %° ~.. 0.1 o

.... ,/ \ 0 .

o Distance from Hometo BARTStation, Miles Figure 14. Modeof Access for CommuteTrips from Hometo Urban District BARTStations

3I 0.9

0.8

0.7

I--

’E 0 oel 0.

Distancefrom Home to BARTStation, Miles

Figure 15. Modeof Access/Egress For all Urban District BARTCommute Trips

0.8 , Waft( o¯°° ..... ," ,. o°¯° ------Bicycle 0.7 o°*°¯.... ¯ ¯ ° ...... Drive Alone -- ¯ SharedRide ¯ - ~ Kiss & Ride ,¯°,°°* 0.6 Transit o--°..¯¯ ’ ¯ ¯ ¯, ~ ° ° o"

t. 0.4 ° oo

O °° 0.3 O , ° .oel \ 0. 0.2 .°°°°" /° , °o° ~.=o \ 0.1 /

Distancefrom Home to BARTStation, Miles Figure 16. Modeof Access for CommuteTrips from Hometo Suburban Center BARTStations

32 o.g

0.8

0.7. >,, ro C:Z Walk 0.6 O. ------Bicycle ...... D,’/ve A/one ~ne~ "- 0.5 SharedRide I-- .... PassengerPick-Up Transit e, o 0.4 /. o n~ 0.3 /, r, ’ , ,,,, /* " ~ I ’. - ...... o" , 0.2 / ’" " I ; : ,, : ~ - ,o I’ ,,, ,,’"’, / , , 0.1 °° o, ¯ ". , ~..~ " " 7"x~-", . ~. _~/ ~ ’, ¯ ; 0 i i i I

Distance from Home to BART Station, Miles

Figure 17. Modeof Egress for CommuteTrips from SuburbanCenter BARTStations to Work

0.9

-. Walk Bicycle 0.8 [ ¯ ¯ Orive’AIone [ .... Shared Ride I 0.7. .... Kiss & Ride/Pick-Up J ° ¯ °.-o ¯ Transit J ¯ ° ,. r~ 06 ¯° °¯ ...... ¯.¯.-°’’°° 0.5 n o:- ¯oO" I’- 0.4 G ¯#¯°

0 0.3 PQ.

0.2

Distance from Hometo BARTStation, Miles Figure 18. Modeof Access/Egress for All SuburbanCenter BARTCommute Trips

33 Low-Density Areas Themodal split profiles for accessand egress trips for low-densityarea stations are shownin Fig- ures 19, 20, and 21. For access trips from home,Figure 19 showswalking predominates up to a distance of around 2,500 feet, beyondwhich park-and-ride becomesthe main modeof access. Beyond1/2 mile, transit is the secondmost common access mode.At 3 miles, comparableshares of access trips are by transit andkiss-and-ride. For egress trips, Figure20 reveals a modalsplit pattern similar to other station classes. Asa market share, transit egress eclipses walkingat slightly shorter distancesat low-densitystation areas -- around 2,800feet. Thecomposite graph, shownin Figure 21, reinforces the findings. The private automobileis predominantlyused to reach low-densitystations for distances beyondone mile. Overintermediate distances, transit is the secondmost commonaccess and egress mode,followed by kiss-and-ride/passengerpick-up.

Other Station Groupings Accessand egress trips werealso studied for other station groupingsto gain a better understanding of howmodal splits varied over longer distances. Thestation groupingwhich produced the mostdiscerna- ble patterns for access and egress trips up to 10 miles in length was Terminal/Near-TerminalBART stations (Fremont, Daly City, Richmond,E1 Cerrito del None, Concord,and Pleasant Hill)) s Figure 22 showsthat park-and-ride was the predominantaccess modeover mostdistances except the very shortest (under 3/8 mile) and the intermediate distance of 6 to 6-3/4 miles. Walkingpredominated over the very short distances, andtransit wasthe mainfeeder for the 6- to 6-3/4-miledistance. Overall, the patterns were not as smoothas one mightexpect. This could reflect the undersamplingof access trips beyond1- 219 milesin length as well as the influencesof localizedfactors (e.g., terrain, freewaylocations).

All Stations Combined A final set of graphs wereproduced on access and egress modesfor all 34 BARTstations combined. Figure23 reveals that the dominantmodes of home-endaccess were: walking-- 5/8 mile or less; transit -- 5/8 to one mile; and park-and-ride-- beyondone mile. Kiss-and-ridecaptured nearly 20 percent of home- end access trips over the 4,000- to 5,000-footdistance range. For work-endegress trips, Figure 24 shows transit eclipsed walkingat a distance of around3,600 feet; over the 1- to 3-milerange, no other modecap- tures morethan 7 percentof all egress trips. Combiningaccess and egress trips for all stations, Figure25 indicates that transit is the mostpopular connectingmode over the 5/8-mile to 1-3/4-miledistance range; over longer distances, drive-alonetravel predominates.Lastly, Figure 26 presents the summarymodal splits for combinedaccess and egress modesplotted over a muchlarger distance -- 10 miles. This figure reveals that while general patterns hold over longer distances, the modalsplit plots are certainly not smoothor

34 0.9

tie 0.8 e Alone red Ride s & Ride nslt

oO *. oo ,o "--.° .oO ¯ " ", o°~ "".

°.° ...... o,O°° oo.O° ooO° ° o. oO° °°

0.2 oO oo oO 0.1 o

Distancefrom Home to BARTStation, Miles

Figure 19. Mode of Access for CommuteTrips from Home to Low-Density-Area BARTStations

0.9

0.8 ~ck-Up

0.7

¢g CZ 0.6

U~ 0.5

c- O 0,4

O O ~. 0.3

0.2

0.1 .. ~ 0 ~~--....’~ ...... ~ ., , ,.-,~ , -

Distancefrom Home to BARTStation, Miles Figure 20. Mode of Egress for Commute Trips from Low-Density-AreaBART Stations to ~ork

35 0g

08 Wal~ ------B~c~ 07 ...... Drive Alone -- ¯ ~- - SharedRide .... Kiss & RideJPic~-Up II [ . Transit

¯ " -¯ ¯oo°° .¯° ...... o "°-.°.¯-°" ° ° ° ° o . ° ° ¯o

0 .... ~ ......

Distancefrom Home to BARTStation, Miles Figure 21. Modeof Access~Egress For All Low-Density-Area BARTCommute Trips

O9O V./alk . ’,

ol ...... Shaled Olr.’e O 70 ~KJSS & Ride . .,"’’. ~ ,° ° ¯

~" 05o , , ," ’. : " /\ ," ¯ .,, ," ", ."

~---~’~ ~_/’" o,ot..:,: ~ ~F,1 ", i".."\t..-I l I:it". ,"" .- ," \ ’ / " . ~ - ¯ ~" ", ’ ¯ x ¯ ., ,/--.’,,, ,/ ..,, ,, ,~..o . ,,, ,. / \, .I \ I o~o b;’\ " " ;, "\- ’ " " ":( ~ " I I" A’",, ."’-’~.~._’":~.S" ", I ~. " "~-" . \ ,I.1"~"-, ~ I L ._-_,_x-~c" ~..;,~,. x?~- ,.." .. ;I’~¯,~ ’. \ I ooo - " " " I

Distancebetween BART Station and Origin/Destination, Miles Figure 22. Modeof Access/Egress For Terminal/Near-Terminal BARTCommute Trips

36 0.9

0.8 Walk ------Bicycle ...... Olive Alone 0.7 - -- - SharedRide Kiss & Ride Transit

° ° ¯ ° ,°...4° ¯, ,°... ¯ ¯°°" "’-°...°°. ¯¯#

.° ° °°.°-- .o° °°°

°.¯ ° °°

°¯o

0.2

0.1

- "~- --~--- H --- ~ ......

Distance from Home to BART Station, Miles Figure 23. Modeof Accessfor CommuteTrips from Hometo All BAKTStations

0.9 -.J

0.8

O.eA,eoe // I"..... [ Trensll

0.2

0.1

°° ¯

Distance from Home to BART Station, Miles Figure 24. Modeof Egress for CommuteTrips from All BARTStationsto Work

37 0.9 WaTk ] B,cycle I ...... Dr~’e Alone 0.8 I .... Shared RKJe I II.... Kiss & Ride/Pc~k-Up 0.7 Transit J a o6 ¢L

’’" I-" 0.5 °°...... , ...°o,-" ...... O 0.4 .¯ ° o ~ 0.3 o, °° °¯ 0.2

...... ~¯~ ......

Distance from Hometo BARTStation, Miles Figure 25. Modeof Access/E~;ress for All BARTCommute Trips

------Bicy~e 0.9 I I...... DriveAlone .... Shared Ride ’ 0.8 ....Kiss & R~letPk:~..U I .... _T!a"S""- ..... :

0.7 .,. ., ¯ - o ,, : : ,,. ¯ ., :. 0.6 ," .. : : : ", ¯ "..¯," ,,, , ¯, ’ ’,o" , , . 0.5 : .! 8 ~ 0.4 2 n. 0.3

0,2

0.1 ._,". ~, ::% .ovoo

Distance from Hometo BARTStation, Miles Figure 26. Modeof Access/Egress for All Home-to-WorkBART Commute Trips, Longer Distance Plot tidy, likely reflecting the combinedeffects of smaller samplesizes for longer distance access trips and localized factors (e.g., the presence of water bodies reducing transit travel over certain distance ranges).

Summary The salient findings on the influence of distance on modesof access and egress are summarizedin Table 17. One major distinction betweenstation classes is the distance at which walking is eclipsed by other modesas the predominate meansof access and egress. In general, people seem willing to walk farther to and from stations in denser, mixed-use settings. The largest walkshed was for downtownOak- land stations, which also have the greatest mixture of land uses. DowntownSan Francisco and urban district stations had the second-largest walksheds, with the exception of home-endaccess trips to San Francisco Office Center stations. In general, net residential densities are so high near these stations that most resident accessing them are fairly close, manyunder 1,000 feet. For the two lowest-density station classes, transit and park-and-ride eclipsed walking as the predominantmodes at shorter distances. People seemmost averse to walking to suburban center stations, the station class with the largest size parking lots, the least land use mixes, and the lowest surroundingresidential densities.

Table 17. Summary on Influence of Distance on Modes of Access and Egress Among Classes of BART Stations Distance up to which Modeof Access Beyond Modeof Egress Beyond Walking Predominates Home- Work- End End Station Class Access Egr_¢~ Dominant Secondary. Dominant Secondary San Francisco Office Center 3,000 ft. 4,000 ft. Transit -- Transit -- SanFrancisco Commercial/ Civic Center 4,000 ft. 3,300 ft. Transit Kiss-n-ride Transit Downtown Oakland 3,800 ft. 3,600 ft. Transit Kiss-n-ride Transit Urban Drive-alone/ Districts 3,300ft. 3,600 ft. Transit Kiss-n-ride Transit Bicycle Suburban Park Kiss-n-ride/ Passenger Centers 2,700ft. 3,300 ft. n-Ride Transit Transit Pick-up Low-Density Park- Transit/ Passenger Areas 2,900ft. 2,900 ft. n-Ride Kiss-n-ride Transit Pick-up

In general, more dense, mixed-use settings seem to stretch walking access distances by 200-300 feet and walking egress distances by 200-600 feet. Overall, the land-use environments immediately around stations appear to have a modest yet measurable influence on walking access/egress distances. Quality of walking environment(e.g., amountof landscaping, sidewalk and street furniture provisions,

39 extensivenessand attractiveness of ground-floorretail) could very well have an evenbigger effect on the size of walksheds;there is likely not enoughvariation in walkingquality amongthe 34 BARTstations (oncevariables like land-usemixture are accountedfor) to statistically gaugewhether this is the case. Oneother finding is that walkingtends to predominateover a longer distance for egress trips to workplaces than for access trips fromhome. This suggests that people mightbe less inclined to wait for a bus at exit stations whenthey can walkto their workplacesin 5 to 10 minutes.And in the case of suburban center and low-densityarea stations, the existence of park-and-ridelots encouragessome nearby residents to drive whenaccess distancesare less than 1,000feet. Table 17 also reveals that for access trips beyondwalking distance, modesvary markedlybetween higher-densityurban stations and lower-densitysuburban ones. For downtownand urban district stations, whereaverage employment densities exceed20 workersper acre and average residential densities exceed 17 dwellingunits per acre, transit is the dominantaccess modefor trips beyondone-half mile. Kiss-and- ride is the secondmost commonaccess modebeyond this distance. For suburban-centerand low-density area stations, park-and-rideis the dominantmode of access beyond3,000 feet distance. Giventhe abun- danceof parkingspaces at these stations, this is no great surprise. Bothkiss-and-ride and transit functionas secondarymodes for longer-distanceaccess trips to these suburban-likestations. In general, kiss-n-ride is morecommon as a secondarymode at suburbancenter stations, likely becauseof their goodfreeway accessibility. Transit, on the other hand, is the morecommon secondary mode for low-densityarea stations. Thesmallest differences wererecorded for egress trips madebeyond walking distance. Regardless of station class, transit was the predominantmode for egress trips beyondone-half mile. For downtown stations, there is little option to walkingor riding transit fromthe exit BARTstation to one’s workplace. In lower density settings, a secondaryegress modewas passenger pick-up. Overall, the hypothesesposed by this research are supported. Theland-use environmentexerts a significant influenceon modesof access and egress, especially with regard to the distance peoplewill walk to or froma BARTstation. Beyondnormal walkingdistances, factors like the supply of parking have a far stronger influenceon access modalsplits.

9. RIDERSHIP CATCHMENT AREAS This section presentsfindings on the physicalshapes, ranges,and other characteristics of the catchment areas for access and egress trips to and from BARTstations. In examiningthe areas aroundwhich transit ridership is drawn,catchment areas weredefined as contiguouscensus tracts whichencompass the origins of 90 percentof all accesstrips to BARTstations (or destinationsof 90 percentof all egress trips fromstations). The use of the 90 percent rank to demarcatecatchment areas was chosento represent the distances at whichthe vast majorityof access trips are drawn.It wasselected largely basedon visual inspections of the cumulativedistances of access trips to all BARTstations. Beyond90 percent, most access trips fall towardthe extremetail of the distance distribution. That is, somepeople make fairly

4o long access trips of 20 or moremiles fromthe exurbanand rural fringes to reach suburbanBART stations; however,these access trips represent statistical outliers. Invokingthe criterion that a zonehad to be con- tiguousto at least oneother zonein a cluster to becomea memberof that cluster ensuredthat a reasonably well-defined territorial representation wasobtained. The catchmentareas for walk-ontrips to BART stations weredefined as the censustracts encompassingthe origins of 90 percent of all access trips made by foot. This secondcl/tchment represents, webelieve, an area that reflects the degree of urbanization aroundstations. As already shown,people seemwilling to walk farther in denser, moremixed-use environs. Mappingof these walk-oncatchment areas shouldfurther corroborate this finding. Still, walking catchmentsof lower-densityareas can be expectedto be larger simplyby virtue of the greater distances that2° must be overcomeby foot. In this section, the catchmentareas for each station are portrayed using GeographicInformation System(GIS) tools. The mapsproduced by GIS provide someunderstanding of howcatchment areas vary in shape and size. Additionally, the averageradii of catchmentareas wereestimated and compared across eachclass of station. It shouldbe addedthat using census tracts to define catchmentboundaries is not ideal. In more urbanizedsettings, census tracts tend to be small enoughthat compositesof censustracts tend to be rea- sonable representations of catchments.In lower-densityareas, however,census tracts generally increase in size, meaningthat catchmentswill tend to be larger in these settings. In somesuburban and exurban parts of the BayArea, for instance, large amountsof openspace are within several census tracts, producing large territorial units. Evenif oneaccess trip origin is in one of these zones,using our criteria, the zone will be addedto the catchment,thus skewingthe estimate of land coverage.From visual inspections of our results, this turned out not to be a serious problem.Virtually all censustracts that met the catch- mentcriteria had far moreland that wasdeveloped than undeveloped.Still, ideally, smaller geographic units, like block groupsor even blocks, wouldbe used in defining catchments.

9.1. Catchment Areas for All Modes of Access and Egress This section presents the catchmentareas that capture 90 percentof access and egress trips for all modescombined, with census boundaries used to delineate areas. The GISmaps are presented in the appendixorganized around the six station classes. Mapsare first presented for homeaccess trips by all modes,grouped by station classes. This is then followedby series of GISmaps for homeaccess trips by walkingonly, also groupedby station classes. Visual inspection of these mapsreveals somepatterns. Perhapsone is first struck by the finding that catchmentsdo not follow simple, tidy concentric patterns. This is not surprising. Becauseof natural features, historical patterns of development,zoning restrictions, and other factors, residential develop- mentitself tends to be spread out and spatially uneven.As expected, the catchmentsfor all modesare manyorders of magnitudelarger than those for walking access. Walkingcatchments, however,tend to

41 be more compactand concentric-like in shape. Lastly, as hypothesized, catchments tend to enlarge in suburbanstation settings, particularly those with large park-and-ride lots and especially those that are terminals or near-terminals (i.e., function like terminals). The Dilly City, Pleasant Hill, and El Cerrito del Notre stations, in particular, have huge catchments. Someof the intermediate stations closer in, like Lafayette, Balboa Park, and Hayward,as well as major transfer stations, like MacArthur,also have spa- cious catchments. The very large catchments, however, are mainly limited to the two lowest-density classes of stations: "suburban centers" and "low-density areas." Andwe see that being an end-of-the-line station does not alway gaurantee a large catchmentarea. The Richmondstation, for instance, has a fairly small catchmentfor all access trips, encompassingjust 23 square kilometers, comparedto the next-in station, E1 Cerrito del Norte, whosecatchment is 146 square kilometers in size. This difference is partly explained by differences in freewayaccessibility -- E1 Cerrito del Notre station sits right off of Interstate 80 whereas the Richmondstation is several miles awayfrom the freeway. It is also explained by differences in parking supplies (which are likely also influenced by level of freewayaccessibility) -- the Rich- mondstation has 796 parking spaces, comparedto 2,516 at E1 Cerrlto del Notre. For all access trips, the largest catchmentareas were along the Concordline: Lafayette (628 sq. kms), Concord(420 sq. kms), Pleasant Hill (335 sq. kms), and Orinda 013 sq. kms). This partly reflects the vast amountsof suburbanresidential developmentthat lies east of these stations as well as the existence of somelarge census tracts in the suburban East Bay. Lake Merritt (5.8 sq. kms), Mission 16th Street (9.86 kms), and Embarcadero(14.6 sq. kms) have the smallest catchmentsfor all trips, reflecting their high-density surroundings and absence of free parking (though Lake Merritt does have 205 paid parking spaces at 25 cents per day).

9.2. Catchment Areas for Walking Access Modes As noted, 90 percent catchments for walk-on trips tended to be smaller and more concentric-like than those for all modes. Differentials in the catchmentareas for all modesversus walk modesvaried significantly by stations -- from the all-mode catchmentbeing just 2.8 times larger in the case of the downtown Berkeley station to over 55 times as large for the E1 Cerrito del Notre station. In general, the differentials were smallest for urban district stations (Berkeley, the Missionstations, and Lake Merritt) and were largest for terminal and suburban center stations. Someof the densest stations, however, also had large differentials -- the all-mode catchment was 17 times larger for the MontgomeryStreet station and 13 times larger for Oakland’s12th Street station.

9.3. Comparison of Catchments Across Station Classes, for all Modes Significant variation was found in the size, radius, service population, and density of catchments across the six classes of stations. For all modesof access, Table 18 and Figure 27 showthat the Suburban Center stations tended to have catchments with the largest land coverage and average radii. Relative to

42 Table 18. Comparisonof CatchmentRadii, Land Areas, Populations, and Densities for Six Station Classes, for all Home-AccessModes Land Radiust Area2 Population3 D2mi~~ San Francisco 2.60 21.78 164,836 7,793 Office Center (0.62) (10.09) (57,416) (972) San Francisco 2.77 24.29 200,089 8,165 Retail/Civic Center (0.41) (7.06) (69,914) (506) Downtown 2.94 28.90 102,623 3,898 Oakland (1.05) (19.35) (45,513) (1,036) Urban 2.29 18.40 98,448 6,138 Districts (0.91) (13.75) (53,693) (2.565) Suburban 10.94 386.99 220,570 598 Centers (2.06) (150.01) (77,295) (200) Low 5.59 113.55 172,938 1,523 Density (2.26) (84.64) (71,488) (1.247) F-Statistic 11.58 11.47 1.84 20.49 (F probability) (.000) (.000) (.137) (.000) Notes: t Meanradius of catchment- {Xi a](L,sn,d Area)/H} / ni, whereZ i = sumover all casesin catchmenti and ni = numberof [rasesin catchmenti (std. dev.in parentheses), 3Meanland area (squarekilometers) of catchment(std. dev.in parentheses), 4Meanpopulation of catchment(std. dev. in parentheses). Meanpopulation density of catchment(people/square kilometer) (std. dev. in parentheses).

Mean Radius Mean Land Area I ¯ SFOffice Center SF Office Center

SFRetail/Civic SF Retail/Civic

Downtown Oak. Downtown Oak. ! UrbanDistricts L UrbanDistricts n ! I Sub. Centers Sub. Centers i l

l i I i Low-Den. Areas Low-Den. Areas ~ , i i I i i ~ i i I , , L I ’ I I t i ’ i ’ 0 2 4 6 8 10 0 100 200 300 400 500 Kilometers Square Kilometers IiWalkl~AII Modes I I.Walkl~AIIModes I Figure 27. Comparisonof MeanCatchment Radii and LandAreas AmongStation Classes

43 the station class with the smallest average catchmentarea and radius, the San Francisco Office Centers (Embarcaderoand MontgomeryStreets), the Suburban Center stations are, on average, 1,677 percent larger in area and 321 percent larger in radius (Figure 28). Next in average catchmentsize was the Low- Density station class, which were around 421 percent larger than the San Francisco Office Centers stations. Surprisingly, the average catchmentareas and radii of the six remaining stations were fairly similar -- average land areas in the range of 18.4 to 28.9 square kilometers, and average radii from 2.29 to 2.94 lineal kilometers. This likely reflects the fact that nearly all stations in these four station classes contain no parking, as was shownpreviously in Table 3. It is parking supplies, we suspect, that is the principal explainer of differences in catchmentarea, muchmore so than land-use variables like density or mixtures of use. The simple correlation between catchment area (for all modesof access) and parking supply for the 34 BARTstations was 0.63; the correlation between employmentdensity and catchment area was -.46. Of course, density and parking supply are highly correlated amongthemselves, so these relationships are interconnected.

Radius Land Area

42 100 SFRetail/Civic SFRetail/Civic 2 57 144 Downtown Oak. DowntownOak.

145 516 UrbanDistricts UrbanDistricts 1177 Sub. Centers Sub.Centers

Low-Den. Areas Low-Den.Areas

-100 0 100 200 300 400 -500 0 500 1000 1500 2000 Percent Difference Percent Difference

HIWalkNIAII Modes I.Walkr~AIIModes I I

Figure 28. MeanRadii and LandAreas as Percent of the "SanFrancisco Office Center"Catchment

To further explore these associations, regression models were produced that predict catchment size and radius (Table 19). As expected, catchmentareas were generally smaller for stations in dense settings (e.g., downtownSan Francisco), and were generally larger in those with high parking supplies (e.g.,

44 Table 19. Regression Model for Predicting Size and Radius of Catchment Areas for BART Stations, Walking Modes DependentVariable: DependentVariable: CatchmentSize (sq. km) CatchmentRadius (km) Coefficient Coefficient Employees/acrewithin -0.803"- -.019"* one-halfmile of station (.417) (.008) Households/acre within -3.679* -.095" one-halfmile of station (2.402) (.046) Median household 7.185"** 0.161"** income($1,0000 (1.951) (.037) Park-and-ridespaces 0.052** 0.0014"** at station (.025) (.0005) Constant -52.23 1.465 (56.13) (1.89) R-squared .607 .720 F Statistic 11.19 18.61 F probability .000 .000 Notes: Valuesin parenthesesare standarderrors. Numberof cases - 34 ***= significant at .01 probabiltylevel ** -- significant at .05 probabilitylevel * = significant at .10 probability level

Pleasant HAll) and faiHy high incomes (e.g., Lafayette, Orinda). Incomes and parking were most highly associated with catchment size. Table 18 indicates there were statistically significant differences in land area and radius among the six station classes (ANOVAF statistics). Even more significant were differences in the catchment densities. The catchmentdensities for urban districts, for example, were ten times as dense as those of suburban centers. Catchmentpopulation sizes, however, varied less amongthe six classes.

9.4. Comparison of Catchments Across Station Classes, for Walking Trips As noted, there was less variation in catchmentsizes for walk trips. Table 20 and Figure 27 reveal walking catchments were in downtownSan Francisco were 1.5 to 3 square kilometers in size with radii of 0.7 to 1 kilometers. As one moveddown the urban hierarchy, walking catchments generally enlarged and catchment densities fell. The largest walking catchments were for suburban centers, reflecting their low surrounding residential densities as well as vast inventories of parking (which spatially separates walkers from rail stations). Figure 28 shows the typical walking catchments around suburban center stations were more than 11 times larger and had 2.5 times longer radii than those in downtownSan Francisco.

45 Table 20. Comparison of Catchment Radii, Land Areas, Populations, and Densities for Six Station Classes, for Home-AccessWalk Trips Land RadiusI Area2 Population3 ~4 San Francisco 0.69 1.52 18,715 12,047 Office Center (0.04) (0.18) (35372) (4,637) San Francisco 0.98 3.04 45,890 15,080 Retail/Civic Center (0.02) (0.11) (7,828) (2,013) Downtown 1.08 3.70 23,028 6,086 Oakland (0.09) (0.58) (9,782) (1,684) Urban 1.69 9.37 62,200 6,662 Districts (0.39) (4.05) (28,541) (1,799) Suburban 2.43 19.41 23,600 1,412 Centers (0.61) (9.69) (12,741) (1,006) Low 1.95 13.58 35,372 3,527 Density (0.75) (13.85 (16,907) (1,803) F-Statistic 3.33 1.25 3.07 23.43 (F probability) (.017) (.311) (.025) (.000) Notes: l Meanradius of catchment,- {Z i 31(L’3n_d Area)/H} / n/, whereY- i " sumover all cases in catchmenti and n/- numberof ~asesin catchmenti (std. dev. in parentheses), 3Meanland area (square kilometers)of catchment(std. dev. in parentheses), Meanpopulation of catchment(std. dev. in parentheses). 4Meanpopulation density of catchment(people/square kilometer) (std. dev. in parentheses).

Theregression modelproduced for estimating the size and radii of walkingcatchments did not haveas gooda fit (Table21). Parkingsupplies increase catchmentsizes, while, controlling for parking, terminal and near-terminal stations averagedsmaller walkingcatchments. Walkingcatchment radii tended to decline with denser station environs.

9.5. Catchments AmongStation Types with Different Parking Supplies Since parking supplywas found to havea significant bearing on sizes of catchmentareas, these relationships werealso studied for three other types of stations: (1) Urban/NoParking; (2) Suburban/ Parking (but not a terminal or near terminal); and (3) Terminal/Near-Termlnal.Table 22 and Figure 29 showthat catchmentareas varied even morestrongly as a function of these classificatons. Theurban stations with no parking averagedcatchments which were around2.6 kmsin radius and 22 square kilometers in land area. Terminal/Near-Terminalstations averagedcatchments that were nearly ten times as large (Figure30). Relationshipswere similar for walktrips (Table 23). However,suburban stations with parking generallyhad the largest pedestriancatchments. This is consistent with the findings of the previoussec- tion that, controlling for parkingsupplies, terminal/near-terminalstations generally had smaller walking catchmentsthan other suburbanstation areas.

46 Table 21. Regression Model for Predicting Size and Radius of Catchment Areas for BART Stations, Walking Modes Dependent Variable: Dependent Variable: Catchment Size (sq. km) Catchment Radius (km) Coefficient Coefficient

Park-and-ride spaces 0.0091"** 4.885** at station ~002~ (.002) Terminal or near-terminal -17.672~* -I.087"** (0-no, l-yes) ~.658) (.3793) Employees/acre within one-half mile of station -~0024) -.0057"* Constant 6.875 1.695 ~.747) (0.19~ R-squared .423 .497 P Statistic 7.63 5.08 F Probability .001 .012 Notes: Values in parentheses are standard errors. Numberof cases - 34 *** - significant at .01 probabilty level ** - significant at .05 probability level * - significant at .10 probability level

Table 22. Comparison of Catchment Radii, Land Areas, Populations, and Densities for Three Station Types, for All Home-Access Modes Land . I Area2 Radius Population3 ~4 Urban/ 2.58 22.35 160,832 6,498 No Parking (0.73) (11.78) (68,196) (2,225) Suburban/ 6.08 146.47 160,832 2,119 Parking 0.20) (154.87) (68,197) (1,407) Terminal/ 8.59 242.66 248,947 1,153 Near Terminal (2.02) (114.53) (46,621) (363) F-Statistic 11.60 6.42 6.44 30.43 (F probability) (.000) (.005) (.005) (.000) Notes: 1 Meanradius of catchment- {Zi ~/(L,)nd Area)Ill } / n/, whereZ i " sumover all cases in catchmenti and ni - numberof ,,~es in catchmenti (std. dev. in parentheses), ~Meanland area (square kilometers)of catcliment(std. dev. in parentheses), Meanpopulation of catchment(std. dev. in parentheses). Meanpopulation density of catchment(people/square kilometer) (std. dev. in parentheses).

47 Radius Land Area

Urban/no parking Urban/no parking

Suburban/Parking Suburban/parking 147

Terminal 243 Terminal I ’ I ’ I ’ I ’ i ’ I ’ I 0 100 200 300 I I I I 0 2 4 6 8 10 50 150 25O Kilometers Square Kilometers

iiWalk []All Modes J I RWalkDAII Modes

Figure 29. Comparisonof MeanCatchment Radii and LandAreas AmongThree Station Types

Radius Land Area

Suburban/parking Suburban/Parking ~136 k555

38 77 Terminal Terminal ~ 233 ~ 986

, I , I ¯ I , I , I , 0 50 100 150 200 250 300 0 200 400 600 800 10001200 Percent Difference Percent Difference

I "Walk []AII Modes InWalk[]AII Modes

Figure 30. MeanRadii and LandAreas as Percent of the "Urban/NoParking" Catchment

48 Table 23. Comparison of Catchment Radii, Land Areas, Populations, and Densities for Three Station Types, for Home-AccessWalk Trips Land RadiusI Area2 Population3 Density4 Urban/ 1.23 5.40 42,407 9,307 No Parking (0.48) (4.21) (26,142) (4,327) Suburban/ 2.16 16.54 32,177 2,860 Parking (0.80) (14.70) (15,097) (1,857) Terminal/ 1.70 9.55 35,147 3,766 Near Terminal (0.41) 0.74) (22,043) (1,893) F-Statistic 6.38 3.35 0.84 17.51 (F probability) (.01) (.048) (.441) (.000) Notes: 1 Meanradius of catchment- {E i ~](L~dArea)/H } / ni, whereY~ i "* sumover all eases in catchmenti and nl - numberof ~asesin catchmenti (std. dev. in parentheses), 3Meanland area (square kilometers)of catchment(std. dev. in parentheses), 4Meanpopulation of catchment(std. dev. in parentheses). Meanpopulation density of catchment(people/square kilometer) (std. dev. in parentheses).

10. CONCLUSION Thehypotheses posited at the outset of this report werelargely substantiated by empiricalevi- dencefrom the San FranciscoBay Area. In short, densities and, to a lesser extent, land-use mixturesdo matter in determininghow people access rail stations and whatthe general size of catchmentareas are for accesstrips. The key findings of the research are summarizedbelow: ¯ Dense, mixed-useurban areas averagehigh shares of walk access trips. Mostsuburbanites access stations by private automobile. ¯ Landuse mixture wasmost strongly associated with access and egress modalsplits. This suggests that creating communitiesaround rail stations that provide services to walk-on users and that are morepedestrian-friendly in design could encouragehigher shares on non-motorizedaccess trips. ¯ The built environmentappears to have the weakest influence on whetheraccess and egress trips to andfrom rail stations are by bus transit. Qualityof transit servicesis a far moreimportant factor. ¯ Parkingsupplies had their strongest influence on reducing walktrips from homesto stations. Whenfacing the prospects of wadingthrough seas of surface parking, some would-bewalkers opted to park-and-ride instead. ¯ Peopleare willing to walkfarther to and fromstations in denser, mixed-usesettings. Peopleare mostaverse to walkingto suburbancenter stations, the station class which,in the BayArea, has the largest-size parkinglots, the least land use mixes,and the lowest surroundingresidential densities. ¯ Moredense, mixed-usestation settings seemto stretch acceptablewalking access distances by 200-300feet and walkingegress distances by 200-600feet.

49 ¯ Walkingtends to predominateover a longer distance for egress trips to workplacesthan for access trips fromhome. ¯ For downtownand dense, urban stations, transit is the dominantaccess modefor trips beyondone-half mile. For suburbanstations, park-and-ride is the dominantaccess mode beyonda half a mile froma station. For egress trips beyondone-half mile, transit is the dominant mode. * Home-accesscatchment areas varied sharply by land-use environment.The lowest- density station areas and suburbancenters generally had catchmentareas that were6 to 10 times larger in size than those of downtownand urbanstation areas. ¯ Parkingsupplies, densities, and neighborhoodincomes were the strongest predictors of catchmentarea. Catchmentswere smaller for stations in denser settings and larger for those with large parking lots and in affluent communities. ¯ There wasless variation in catchmentsizes for pedestrian access trips. Suburbanwalk- on catchmentstended to be 8 to 10 times larger than those of downtownstation areas and3 to 4 timeslarger than those of urbanizedareas.

5o Notes tThe San Francisco Bay Area is somewhatunique in that there are a range of common-carrieraccess modesbeyond bustransit availablein SanFrancisco. These include light-rail transit, bus trolleys, jitneys, and cablecars. 2Summinginformation on the dominantland use for each hectare over the numberof hectares within a half-mile radius of rail stations providedcounts of the total squaremeters of land area devotedto eachland use withina circle of one-milediameter aroundeach BARTstation. 3Dataon residential densities wereobtained from STF 3-A for 1990census tracts and block groupsthat mostclosely correspondedto a half-mile radius surroundingeach station. Data on employmentdensities were obtained from the 1990 CTPP,provided by the Metropolitan Transportation Commission,for 1990 ncensus tracts that most dosdycorresponded to a half-mile radius of each station. 4Theuse of censustracts for defining catchmentareas posedproblems for studying walk-ontrips in suburbanareas wherecencus tracts can be large, often with dimensionswell beyondthe one-quarter-to one-half-miledistance nor- really considered to be the maximumdistance Americanswill walk. In moreurbanized areas, especially downtown SanFrancisco, census tracts (sometimesas small as four or five city blocks)are moresuitable for studyingthe catchmentsfor pedestrianaccess trips. STheaverage radius for the catchmentof each station wasestimated by assumingthe calculated land area represents a circle. Theradius of a circle of equivalentarea wasthen calculated. 6Themeasure used for joining clusters wasthe average linkage betweengroups, often called UPGMA(unweighted pair-group methodusing weightedaverage [see Everitt, 1980~). Here, the distance measuredbetween two dusters is the averageof distances betweenall pairs of cases in whichone memberof the pair is fromeach of the clusters. rUnderthis approach,all cases are initially consideredas separateclusters, i.e., there are as manyclusters as cases. As the secondstep, the two cases with the mostcomparable squared Euclidean distances (i.e., the ones whosesum of squaredfactor scores are the msotalike) are combinedinto a single duster. At the third step, either a third case is addedto the cluster already containingtwo cases, or twoadditional cases are mergedinto a newduster. The process continuesuntil all cases are groupedtogether. See Everitt (1980)for further discussionsof this approach. 8Thisis becausethe squaredEuclidean distances betweenstation cases for the parkingvariable wasso hugethat the distance metrics for other variables werecomparatively small and thus playeda small role in fusing together cases in the clustering algorithm. 9ThePleasant Hill station was~ssigned to a different groupthan that shownin Figure 1, becauseof its uniquedevel- opmentcharacteristics. Specifically, Pleasant Hill wasassigned to the cluster of SuburbanCenters, with Walnut Creek, Concord,Lafayette, and Orinda. This is because Pleasant Hill emergedas a suburbancenter mainlyafter 1990. "From1990 to 1992, aroundone million square feet of office floorspace and over 1,000 dwellingunits were addedto the one-half-milering aroundthe Pleasant Hill station. Thus, Pleasant Hill’s developmentpost-dated the 1990 land-use data compiledfrom the census and ABAGinventory. By 1992, the year for which the BARTpas- senger surveywas conducted, Pleasant Hill clearly had the character of a suburbancenter, similar to WalnutCreek and Concord. 1°Thesesix duster groupsof BAR.T stations are reasonablysimilar to those created by Johnstonand Tracy(1982). Their classifications werefar moreimpressionistic, however, based on "windshieldsurveys" of land uses near stations. Theygrouped all downtownSan Francisco, Oakland,and Berkeleystations in a single group (failing to acknow- ledge large differences in employmentdensities across these stations. Their classifications of "urbanhigh-density residential" stations matchedour classification of "urbandistricts," with the exceptionthat the MacArthurstation wasplaced in this groupin lieu of the Berkeleystation. In addition to WalnutCreek and Lafayette, they grouped San Leandro,Hay’ward, and Richmondas suburbanstations. Remainingstations were placed in three different levels of suburbandensities: medium,low, and low-density/rural. Their general criteria for judgementallyclassi- fying stations werebased on densities and land use activities. Otherfactors usedin our analysis, suchas related to levels of land-usemixture, parking supplies, and rail modalsplits, werenot explicitly used. 11Six hundrednew spaces were addedto the Concordstation in the summerof 1994, bringing the total up to 2,575 spaces. In 1992, the year for whichthe BARTpassenger data were compiled, however,the parking supplies shown

51 in Table 3 existed. With the newparking supply at the Concordstation, the averagenumber of parking spaces at the SuburbanCenters class of BARTstation is currently 2,048. 12Asresidential densities decline amongthe six station clases, so do auto accessmarket shares. Therelationship is not quite as tidy as a function of employmentdensities in that suburbancenters, thoughthey have higher employmentdensities than low-densityareas, averagehigher auto aeccess trips than low-densityareas. Changesin home- end access trips seemmost closely correlated with changesin parkingsupplies. l~Modalsplits for home-endegress trips by transit werenearly identical to those of home-endaccess trips, and thus are not presented. *’~¢hiletransit servicelevels (e.g., vehicle milesof transit service per daywithin catchment zone) were not directly usedto classify stations in the cluster anazlysis,their influencesare generallycaptured by the land-usevariables. Transit agenciesgenerally focus their services in the mosturbanized areas, meaningdense, mixed-usesettings often receive the mostintensive services. ~SForsome variables, suchas parkingsupplies and the dependentvariables (modalsplits), data are recordeddirectly for stations themselves.For land-usevariables (e.g., density, percentland area), the areas encompassedby a one- half-mileradius aroundthe stations servedas cases. l~vfispoint elasticities are measuredat the meanvalues of both the dependentand independentvariables. lZPlots of distances up to 5 to 10 miles wereinitially generated;however, the results werenot particularly deciph- erable for longerdistances. As a result, the plots generatedin this sectionare only for distancesup to 2 to 3 miles. 18Near-terminalrepresents stations towardthe end of a line that functionlike terminals becausethey are closer to freewaysthan the actual terminal stations and thus serve a larger catchmentarea. BART’snear-terminal stations E1Cerrito del Norteand PleasantHill havelarger supplies of parkingthan terminal stations since they are easier to reach by freeways. 19Mis-samplingof longer-distance access and egress trips couldreflect the tendencyof those traveling longerdistance to have less time available to completea questionnaireform. This mightparticularly be the case for those pressed to reach job sites that are relatively far formtheir exit BARTstations. 2°Theway membership was determined was that zones that were closest to the station weregrouped into a catchment. Workingoutward from the station, if a zone contained trip ends that were within the 90 percentile rank of access distances to a station and shared a boundaryor a point with a zone already assignedto the catchment,then the zone wasadded. Sharinga point meansthat the two zonesonly had to connect at one point, and not neces- sarily share a boundary,to be consideredcontiguous.

52 References

Barry and Associates. 1991. Air Quality in California. Sacramento:California Air ResourcesBoard. BART.1993. 1992 BARTPassenger Profile Survey. Oakland: Bay Area Rapid Transit District, unpublishedreport. Calthorpe, P. 1993. The Next AmericanMetropolis. Princeton: Princeton Architectural Press. Cervero, R. 1993. RidersbipImpacts of Transit-FocusedDevelopment in California. Berkeley:Institute of Urban and Regional Development,Monograph No. 54. Cervero, R., M. Bernick, and J. Gilbert. 1994. MarketOpportunities and Barriers to Transit-Based Developmentin California. Berkeley: Institute of Urban and Regional Development,Working Paper No. 621. Everitt, B. 1980. Cluster Analysis. NewYork: Halstead Press, secondedition. JHKand Associates. 1986. Development-RelatedRidersbipSurvey I. Washington, D.C.: Washington Metropolitan Area Transit Authority. JHKand Associates. 1989. Development-RelatedRidershipSurvey IL Washington, D.C.: Washington Metropolitan Area Transit Authority. Johnston, R., and S. Tracy. 1982. The RelationshipBetween Local LandUse ControlPolicies and Regional Transit SystemEffectiveness: The San FranciscoBay Area Rapid Transit (’BAR7~ Experience.Davis, California: UCDavis Appropriate TechnologyProgram, Division of EnvironmentalStudies, unpublishedreport. Katz, P. 1993. The NewUrbanism: Towardan Architecture of Community.New York: McGraw-Hill. Meyer,J., J. Kain, and M. Wohl.1965. The UrbanTransportation Problem. Cambridge:Harvard UniversityPress. Pushkarev, B., and J. Zupan. 1977. Public Transportation andLandUse Policy. Bloomington:Indiana UniversityPress. Solomon,D. 1992. ReBuilding. Princeton: Princeton Architectural Press. Stringham,M. 1982. Travel BehaviorAssociated with LandUses Adjacent to Rapid Transit Stations. ITE Journal52(4): 18-22. Untermann,R. 1984. Accommodatingthe Pedestrian: AdaptingTowns and Neighborhoodsfor Walking and Bicycling. NewYork: Van Nostrand Reinhold. 54 90% Catchment Area for

SAN FRANCISCO OFFICE CENTER

AII 3lodes Embarcadero Station - 90% Catchment Area Home Access by All Modes

1 0 4000 8000

j--- Montgomery Station - 90% Catchment Area - Home Access by All Modes

0 4000 8000 90% Catchment Area for

SAN FRANCISCO COMMERCIAL/CIVIC CENTER

All Modes Powell Station 90% Catchment Area - Home Access by All Modes

4000 8000 Civic Center Station - 90% Catchment Area - Home Access by All Modes

I P 0 4000 8000

::::::::::::::::::::::::: iiii!i!!!!i!!i!ii!ilI...... ~i]~!{~]{~iii!iiiiiiiiiii

:.:.:.:.:.:+:.:.:.:.~::::::::::::::::::::::::: :::::: ~i~iii~i~i~iii~ii!~i~i iii~iiiiiiiiiiii~iii~i! .-.-.-.-.....-.-.,::::::::::::::::::::::::: i!iii!i ~i~,#,~iii~i,~,~,~,...... :::::::::::::::::::::::.:,:.:-:+:+:.:.:.:.: iiiiiiiiiiii!!ii~ii~i! i i i:~iiiiiiiii~!!~!!!!!-:::::::::::::::::::::::: iiii[iii!iii!iiii~!iii!~ 90% Catchment Area for

DOWNTOWN OAKLAND

A I1 Modes 12th St. Oakland Station 90% Catchment Area - Home Access by All Modes

I r 0 5000 10000 19th St. Oakland Station 90% Catchment Area - Home Access by All Modes

0 5000 10000 90% Catchment Area for

URBAN DISTRICTS

A II Modes Mission-16th Street Station 90% Catchment Area - Home Access by All Modes

o 4000 8O0O Mission-24th Street Station 90% Catchment Area Home Access by All Modes

0 4000 8000

I Berkeley Station 90% Catchment Area Home Access by All Modes

6000 12000 Lake Merritt Station - 90% Catchment Area Home Access by All Modes

0 5000 10000 90% Catchment Area for

SUBURBAN CENTERS

A I1 Modes Orinda Station - 90% Catchment Area - Home. Access by All Modes

0 10000 20000

J Lafayette Station - 90% Catchment Area - Home Access by All Modes

0 10000 20000

\ \. Walnut Creek Station 90% Catchment Area - Home Access by All Modes

0 10000 20000 Pleasant Hill Station - 90% Catchment Area - Home Access by All Modes

1 ’ I 0 10000 20000 oncord Station - 90% Catchment Area - Home Access by All Modes

o 10000 20000 90% Catchment Area for

LOW-DENSITY AREAS

All Modes MacArthur Station - 90% Catchment Area - Access by All Mode

0 5OO0 1O0OO Ashby Station - 90% Catchment Area - Home Access by All Mode

I 0 6000 12000

\ North Berkeley Station 90% Catchment Area Home Access by All Modes

6000 12OOO El Cerrito Station - 90% Catchment Area HomeAccess by All Modes

0 6000 12000

\. El Cerrito del Norte Station - 90% Catchment Area Home Access by All Modes

12000 0 6000 Richmond Station - 90% Catchment Area Home Access by All Modes

0 6000 12O00

\ \ \ \. Fruitvale Station 90% Catchment Area - Home Access by All Modes

I T 0 7000 14000 Coliseum Station - 90% Catchment Area Home Access by All Modes

I t 0 8000 16000

L/ \ San Leandro Station - 90% Catchment Area -.Home Access by All Modes

16000 0 8OO0 Bayfair Station 90% Catchment Area - Home Access by All Modes

I f 0 8000 16000 Hayward Station - 90% Catchment Area - Home Access by All Modes

0 8000 16000 South Hayward Station 90% Catchment Area Home Access by All Modes

0 8000 16000 Union City - 90% Catchment Area - Home Access by All Modes

16000 0 8000

t_/ Fremont Station - 90% Catchment Area - Home Access by All Modes

0 8000 16000

L/ \ West Oakland Station - 90% Catchment Area Home Access by All Modes

0 5000 10000 Rockridge Station - 90% Catchment Area - Home Access by All Modes

0 10000 20000 Glen Park Station - 90% Catchment Area - Home Access by All Modes

4000 8000

:::::::::::::::::::::::::::::::

..-6..:;:;:;:-:; :::::::::::::::::::::::::::::::::: :.:;A:.:.:.:.:.:.:.:, !iii::iii!ii~!!!!i::j Balboa Station - 90% Catchment Area - Home Access by All Modes

0 7000 14000 Daly City Station - 90% Catchment Area - Home Access by All Modes

7OOO 14000

\ 90% Catchment Area for

SAN FRANCISCO OFFICE CENTER

Walking Mode Embarcadero Station - 90% Catchment Area - HomeAccess by Walking

0 4000 8000 Montgomery Station - 90% Catchment Area Home Access by Walking

0 4000 8000 90% Catchment Area for

SAN FRANCISCO COMMERCIAL/CIVIC CENTER

WalMng Mode Powell Station - 90% Catchment Area - Home Access by Walking

8000 0 4000 Civic Center Station - 90% Catchment Area - Home Access by Walking

0 4000 8000 90% Catchment Area for

DOWNTOWN OAKLAND

Walking, Mode 12th St. Oakland Station - 90% Catchment Area Home Access by Walking

0 5000 10000 19th St. Oakland Station - 90% Catchment Area - HomeAccess by Walking

0 5000 10000 90% Catchment Area for

URBAN DISTRICTS

Walking Mode Mission-16th Street Station - 90% Catchment Area - Home Access by Walking

0 4000 8000 Mission-24th Street Station - 90% Catchment Area - Home Access by Walking

0 4000 8000 Berkeley Station 90% Catchment Area Home Access by Walking

600O 12000

\ Lake Merritt Station - 90% Catchment Area - Home Access by Walking

0 5000 10000

\ 90%Catchment Area for

SUBURBAN CENTERS

WalMn~ Mode Orinda Station - 90% Catchment Area - Home Access by Walking

0 10000 20000 Lafayette Station - 90% Catchment Area Home Access by Walking

0 I0000 20000

\ \ )

I Walnut Creek Station - 90% Catchment Area - Home Access by Walking

0 10000 20000

\ Pleasant Hill Station 90% Catchment Area - HomeAccess by Walking

0 1OOOO 20o00 Concord Station - 90% Catchment Area - Home Access by Walking

10000 20000

\ 90% Catchment Area for

LOW-DENSITY AREAS

Walking Mode MacArthur Station - 90% Catchment Area - Access by Walking

10000 0 5000

\ Ashby Station - 90% Catchment Area - Home Access by Walking

0 6000 12000

\ North Berkeley Station 90% Catchment Area - Home Access by Walking

6000 12000 El Cerrito Station - 90% Catchment Area - Home Access by Walking

1 0 6OO0 120OO

-\ \ El Cerrito del Norte Station - 90% Catchment Area - Home Access by Walking

0 60O0 12OOO

\ Richmond Station - 90% Catchment Area Access by Walking

0 6000 12000

\ Fruitvale Station - 90% Catchment Area - HomeAccess by Walking

t i t 0 7000 14000 Coliseum Station - 90% Catchment Area - Home Access by Walking

I 0 8000 16000

L/ San Leandro Station - 90% Catchment Area - Home Access by Walking

0 8000 16000 Bayfair Station - 90% Catchment Area - Home Access by Walking

0 8000 16000 Hayward Station - 90% Catchment Area - Home Access by Walking

0 8000 16000

\ South Hayward Station - 90% Catchment Area Home Access by Walking

I, I 0 8000 16000

L./

\. Union City Station - 90% Catchment Area - Home Access by Walking

0 8000 16000 Fremont Station - 90% Catchment Area - Home Access by Walking

0 7000 14000

L_/

\ Rockridge Station - 90% Catchment Area Home Access by Walking

"’" r 0 10000 20000 West Oakland Station 90% Catchment Area - Home Access by Walking

0 5ooo 10000 Glen Park Station - 90% Catchment Area - HomeAccess by Walking

4000 8000 Daly City Station 90% Catchment Area - Home Access by Walking

7000 14000