California statewide model for high-speed rail

Ou wa e , Ma en e al. A icle Cali o nia s a ewide model o high-speed ail Jou nal o Choice Modelling P o ided in Coope a ion wi h: Jou nal o Choice Modelling Sugges ed Ci a ion: Ou wa e , Ma en e al. (2010) : Cali o nia s a ewide model o high-speed ail, Jou nal o Choice Modelling, ISSN 1755-5345, Uni e si y o Leeds, Ins i u e o T anspo S udies, Leeds, Vol. 3, Iss. 1, pp. 58-83 This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/66842 S anda d-Nu zungsbedingungen: Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen Zwecken und zum P i a geb auch gespeiche und kopie we den. Sie dü en die Dokumen e nich ü ö en liche ode komme zielle Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich machen, e eiben ode ande wei ig nu zen. So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen (insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en, gel en abweichend on diesen Nu zungsbedingungen die in de do genann en Lizenz gewäh en Nu zungs ech e. Te ms o use: Documen s in EconS o may be sa ed and copied o you pe sonal and schola ly pu poses. You a e no o copy documen s o public o comme cial pu poses, o exhibi he documen s publicly, o make hem publicly a ailable on he in e ne , o o dis ibu e o o he wise use he documen s in public. I he documen s ha e been made a ailable unde an Open Con en Licence (especially C ea i e Commons Licences), you may exe cise u he usage igh s as speci ied in he indica ed licence. h p://c ea i ecommons.o g/licenses/by-nc/2.0/uk/ Jou nal o Choice Modelling, 3(1), pp. 58-83 www.jocm.o g.uk Cali o nia S a ewide Model o High-Speed Rail Ma en Ou wa e 1,* Ke in Tie ney2,† Ma k B adley3,Ŧ Elizabe h Sall4,§ A un Kuppam5,¶ Vamsee Modugula6,β 1 Resou ce Sys ems G oup, 917 E. Spoone Rd, Fox Poin , WI 53217 2 206 B oad Meadow Rd, Needham, MA 02492 3 Ma k B adley Resea ch and Consul ing, 524 A oyo A enue, San a Ba ba a, CA 93109 4 San F ancisco Coun y T anspo a ion Au h, 100 Van Ness A e, San F ancisco, CA 94102 5 Camb idge Sys ema ics, 9015 Moun ain Ridge, Sui e 210, Aus in, TX 78759 6 Ci ilabs, A-2/14, Vanash ee CHS, Plo No 1&2, Sec o 58A, Ne ul, Na i Mumbai, 400706 Recei ed 19 Ma ch 2008, ecei ed e sion e ised 22 Decembe 2008, accep ed 19 Sep embe 2009 Abs ac The Cali o nia High Speed Rail Au ho i y (CHSRA) and he Me opoli an T anspo a ion Commission (MTC) ha e de eloped a new s a ewide model o suppo e alua ion o high-speed ail al e na i es in he S a e o Cali o nia. The app oach o his s a ewide model explici ly ecognizes he unique cha ac e is ics o in a egional a el demand and in e egional a el demand. As a esul , in e egional a el models cap u e beha io impo an o longe dis ance a el, such as induced ips, business and commu e decisions, ec ea ional a el, a ibu es o des ina ions, eliabili y o a el, pa y size, and access and eg ess modal op ions. In a egional a el models ely on local highway and ansi cha ac e is ics and beha io associa ed wi h sho e dis ance ips (such as commu ing and shopping). Keywo ds: disc e e choice models, in e egional, high-speed ail, induced demand, ide ship o ecas ing, ip equency, des ina ion choice, mode choice * Co esponding au ho , T: +1 425 269 9684, F: + 011 802 295 1006, mou wa e @ sginc.com †T: +1 617 839 5295, ke in ie ney@ ocke mail.com Ŧ T: +1 805 564 3908, F: + 011 805 564 3927, ma k_b adl[email p o ec ed] § T: +1 919 469 3609, F: + 011 919 302 0265, [email p o ec ed] ¶ T: +1 512 691 8503, F: + 011 512 691 3289, [email protected] β T: +91 022 27525025, F: + 011 510 523 9706, [email p o ec ed]m Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 1 In oduc ion By 2030, Cali o nia’s popula ion will g ow o 50 million people, which will nea ly double in e egional a el o one billion ips pe yea . High-speed ains a e being conside ed o alle ia e he need o build – a a cos o nea ly $100 billion – abou 3,000 miles o new eeway, plus i e ai po unways, and 90 depa u e ga es o e he nex wo decades. Elec ic and ully sepa a ed om au omobile a ic, Cali o nia’s high-speed ain would p o ide a new anspo a ion op ion a ailable o mo e han 90 pe cen o he esiden s o he s a e. A ull build-ou , he sys em will un om San Diego no h o Sac amen o and San F ancisco. The p ojec objec i es we e o de elop a new ide ship o ecas ing model ha would se e a a ie y o s a ewide planning and ope a ional pu poses: • To e alua e high-speed ail ide ship and e enue on a s a ewide basis; • To e alua e po en ial al e na i e alignmen s o high-speed ail in o and ou o he San F ancisco Bay A ea; and • To p o ide a ounda ion o o he s a ewide planning pu poses and o egional agencies o be e unde s and in e egional a el. The co e model design ea u e is he ecogni ion ha in e egional and u ban a ea a el is dis inc and should be modelled sepa a ely o cap u e hese dis inc ions accu a ely. This led o ou app oach o de elop sepa a e, bu in eg a ed, in e egional and in a egional models, as demons a ed in Figu e 1. The e a e wo p ima y easons o de eloping sepa a e models o in e egional and u ban a ea a el: i s , he ip pu poses a e di e en and second, he in e egional a el models need o explici ly es ima e induced demand. These models a e applied o bo h peak and o -peak condi ions o an a e age weekday. Weekend a el demand and annual ide ship es ima es a e de eloped using annualiza ion ac o s de eloped om obse ed da a on high-speed ail sys ems a ound he wo ld. Figu e 1. In eg a ed Modeling P ocess 59 Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 60 The e a e ou een egions es ablished in he s a e ha de ine in e egional and in a egional a el: • In e egional models es ima e ip equency, des ina ion choice, and mode choice s a i ied by ip pu pose (business, commu e, ec ea ion, and o he ) as well as by dis ance ( ips g ea e han o less han 100 miles) and by ip ype ( ips made by esiden s o he ou la ges ci ies in Cali o nia e sus o he ips). The in e egional ip equency models es ima e induced a el based on imp o ed accessibili ies due o high speed ail op ions. The in e egional models a e simila in s uc u e o models de eloped o he Aus alian Ve y Fas T ain (VFT) p ojec , excep ha he e is mo e spa ial de ail (i.e. many mo e zones) and all models we e es ima ed based on e ealed p e e ence wi h ull and consis en nes ing o he accessibili y measu es om access/eg ess models up h ough ip equency models. The e a e also some simila i ies o he No wegian and Swedish na ional models, which combine long-dis ance and sho -dis ance a el and ha e simila modeling s uc u es bu ha e es ima ed simul aneous des ina ion and mode choice wi h sampled des ina ions. • In a egional models a e based on ip ables gene a ed om he Me opoli an Planning O ganiza ion (MPO) models and es ima e mode choice o u ban a ea ips. These mode choice models e lec local u ban a ea highway and ansi sys ems as well as op ions o high speed ail wi hin he egion. U ban a el is s a i ied by ip pu pose (wo k, school, college, o he , and non-home-based). The in e egional and in a egional a ea models a e based on a el su ey da a collec ed o hese pu poses. 2 Li e a u e Re iew Al hough ew in e ci y high speed g ound anspo a ion (HSGT) sys ems ha e been implemen ed in he Uni ed S a es, planne s ha e been pe o ming ide ship o ecas ing analyses and bene i s assessmen s o p oposed sys ems o mo e han hi y yea s. Almos all o he ide ship o ecas s ha e been in suppo o analyses o one o he 11 designa ed in e ci y co ido s au ho ized by he Fede al Rail oad Adminis a ion in he In e modal Su ace T anspo a ion E iciency Ac (ISTEA) o 1991 and he T anspo a ion Equi y Ac o he 21s Cen u y (TEA-21) o 1998, including (Fede al Rail oad Adminis a ion 2006): • No he n New England; • No heas Co ido ; • Empi e Co ido ; • Keys one Co ido ; • Sou heas Co ido ; • Gul Coas Co ido ; • Flo ida Co ido ; • Chicago Hub Ne wo k; • Sou h Cen al Co ido ; • Paci ic No hwes Co ido ; and • Cali o nia Co ido . Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 61 The i s U.S. HSGT ide ship o ecas s we e de eloped o suppo in es men s o imp o e he No heas Co ido (Koppelman e al. 1984). Since hen, a a ie y o o ecas ing echniques ha e been used in co ido easibili y analyses, ide ship and e enue e alua ions, and en i onmen al impac s udies. The li e a u e desc ibing he analy ical echniques employed in hese a ious s udies and he cu en s a es o o ecas ing p ac ice o e ime ha e been e iewed by Koppelman e al. (1984), Mille (1992, pp. 378-389), Fo inash (1992), and Ho owi z (2006). The ini ial o ecas ing e o s elied on agg ega e di ec demand models, such as he Quand and Baumol abs ac mode model (1966), he Rand Co po a ion model de eloped by DeFe an i e al. (1973), and he Pee s and Be ilacqua s uc u al in e ci y model (1976). As u ban modele s adop ed he ou -s ep modeling p ocess o egional o ecas ing, in e ci y models also became mul i-s ep p ocesses whe e he in e ci y ip ables a e es ima ed and o ecas , and sepa a e mode choice models a e used o de e mine modal sha es unde di e en se ice scena ios. Mos o he mo e ecen U.S. HSGT ide ship o ecas ing e o s can be cha ac e ized by he speci ic analyses used o de elop in e ci y ip ables, de e mine modal spli , and he le el o which he di e en model s eps a e in eg a ed. Modele s ha e employed many ip gene a ion and ip dis ibu ion echniques o o ecas u u e yea in e ci y ip ables. Fo an ea ly analysis o he Flo ida high speed ail co ido , PRC Voo hees used ip pu pose speci ic g ow h ac o models o o ecas ip ables (1983). O he modele s, such as Thaku iah e al. (1999) and Cohen e al. (1978), ha e elied on g a i y models o ip able p edic ion. One esea che iden i ied by Ho owi z used a ime se ies model o o ecas u u e in e ci y a el demand (Yu 1970). Many o he s ha e used a c oss-sec ional eg ession-based o al a el demand (di ec demand) o mula ion, including o ins ance, Booz-Allen and Hamil on (1989), TMS/Benesch (1991), and B and e al. (1992). The p e ious analysis o he Cali o nia high speed ail co ido elied on his modeling app oach (Cha les Ri e Associa es 2000). Since he ea ly 1980’s, se e al di e en o ms o disagg ega e mode choice models ha e been de eloped o o ecas ing U.S. HSGT ide ship. Ea ly e o s, such as hose by G ayson (1981), employed mul inomial logi models, bu subsequen models ha e expanded he echnique in di e en di ec ions. Cohen e al. (1978, pp. 21-25), B and e al.(1992, pp. 12-18), and Cha les Ri e Associa es (2000) o mula ed mode choice as sepa a e bina y di e sion models whe e pe cen ages o au o, ai , and bus passenge s a e di e ed o HSGT indi idually h ough bina y models. This app oach seeks o elimina e he oublesome ou come o he mul inomial logi model’s IIA p ope y. O he esea che s ha e elied on he use o nes ed logi models o in e ci y mode choice (TMS/Benesch 1991 and Fo inash and Koppelman 1993). Resea che s, mos no ably Chand a Bha , ha e expe imen ed wi h a wide a ie y o nes ed logi model o mula ions and a iable combina ions (1995) (1997) (1998). Wi h ad ancemen s in e ec i e ways o combining e ealed p e e ence and s a ed p e e ence su ey da a (Ben-Aki a and Mo ikawa 1990) (B adley and Daly 1997), in e ci y modele s ha e also begun o use s a ed p e e ence su ey me hods o a g ea e ex en o o ecas ing HSGT mode choice. Mos in e ci y anspo a ion demand models ha e ea ed a ele ’s decisions ega ding ip equency, des ina ion choice, mode choice, and ou e selec ion as sepa a e sequen ial choices. The model sys em desc ibed in his pape seeks o connec he di e en model componen s by passing in o ma ion om one choice componen o he o he s du ing model de elopmen . When he mul inomial logi and nes ed logi model componen s a e applied, hey a e ully consis en wi h each o he and a e sensi i e o each o he ’s changes. In eg a ed modeling app oaches ha e been p oposed Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 62 by and implemen ed by Koppelman and Hi sh (1986), Koppelman (1989), P oussaloglou and Tie ney (1999) and, ou side he U.S., by Alge s (1993) and Gunn e al. (1992). The la e app oach was de eloped o an Aus alian VFT s udy and uses a e y simila app oach o he models de eloped o Cali o nia, including he use o bo h s a ed and e ealed p e e ence da a and linking mode, des ina ion and ip equency models h ough he use o accessibili y measu es. 3 In e egional Models The in e egional models a e comp ised o ou se s o models: ip equency, des ina ion choice, main mode choice, and access/eg ess mode choice. The s uc u e and con en s o he in e egional modelling sys em is p esen ed in Figu e 2. The ip equency model componen p edic s he numbe o in e egional ips ha indi iduals in a household will make based on he household’s cha ac e is ics and loca ion. The des ina ion choice model componen p edic s he des ina ions o he ips gene a ed in he ip equency componen based on zonal cha ac e is ics and a el impedances. The mode choice componen s p edic he modes ha he a ele s would choose based on he mode se ice le els and cha ac e is ics o he a ele s and ips. The mode choice models include a main mode choice, whe e he p ima y in e egional mode is selec ed, and access/eg ess componen s, whe e he modes o access and eg ess o he ai and ail ips a e selec ed. A combined des ina ion and mode choice model was ini ially conside ed, bu he choice se o des ina ion choice was all zones (4,667) and he combina ion o des ina ions and modes would ha e made his choice se oo la ge. 3.1 Da a o Es ima ion The e we e h ee ypes o da a compiled o he s udy: a el su eys, ne wo ks, and socioeconomic da a. Some o he a el su eys we e collec ed speci ically o his s udy, h ee we e a ailable om MPOs a ound he s a e (Sou he n Cali o nia Associa ion o Go e nmen s (SCAG), San F ancisco Me opoli an T anspo a ion Commission (MTC), and Sac amen o Associa ion o Go e nmen s (SACOG)), and he e was a Cal ans s a ewide su ey a ailable. The in e egional models we e based on e ealed and s a ed p e e ence su eys, collec ed speci ically o his s udy, o ai and ail a ele s, as well as addi ional households in he s a e o cap u e au o a ele s. These new da a we e collec ed in ou een egions in Cali o nia. These we e combined wi h e ealed p e e ence su eys o households ac oss he s a e collec ed by Cal ans and in e egional a el ex ac ed om he MPO egional a el su eys (San F ancisco, Sac amen o, and Los Angeles). In a egional mode choice models we e based on u ban a ea a el su eys in combina ion wi h a s a ed p e e ence su ey o high speed ail conduc ed in Los Angeles. By combining he a ious a ailable da a sou ces, we we e able o p o ide mo e obus da a se s o model es ima ion han was o he wise possible. A summa y o hese sou ces shows he ip eco ds de i ed om each: Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 T io F equency/Day •Household Cha ac e is ics •T ip Pu pose/Dis ance Class •Le el o Se ice (Logsum & Accessibili y •Region •Pa y Size (Fo Sho Dis ance) Des ina ion Choice •Le el o Se ice (Logsum & Accessibili y •Employmen & Household Cha ac e is ic s •Region and A ea Type •T ip Pu pose/Dis ance Class •Pa y Size (Fo Long Dis ance) Main Mode Choice •Le el o Se ice •Household Cha ac e is ics •Pu pose/Dis ance Class •Pa y Size (Fo Long Dis ance) •Access & Eg ess (Logsum) Access Mode Choice •Le el o Se ice •Household Cha ac e is ics •Pu pose/Dis ance Class •Pa y Size (Fo Long Dis ance) •Main Mode (Rail/HSR/Ai ) Eg ess Mode Choice •Le el o Se ice •Household Cha ac e is ics •Pu pose/Dis ance Class •Pa y Size (Fo Long Dis ance) •Main Mode (Rail/HSR/Ai ) One T ip Two-Plus T ips No T ips Zone 1 Zone 2 Zone N-1 Zone N Ca Rail HSR Ai D i e and Pa k D op O Ren al Ca T ansi WalkTaxi T ansi WalkTaxi Ren al CaPicked Up Unpa k and D i e Figu e 2. In e egional Model S uc u e Figu e 2. In e egional Model S uc u e 63 Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 • Ai , ail, au o passenge su eys 2678 ips • Ai , ail, au o passenge su eys 2678 ips • Cal ans a el su eys 2820 ips • Cal ans a el su eys 2820 ips • SCAG a el su eys 343 ips • SCAG a el su eys 343 ips • MTC a el su eys 723 ips • MTC a el su eys 723 ips • SACOG a el su eys 318 ips • SACOG a el su eys 318 ips A e combining hese su eys, 6,882 comple ed su eys we e a ailable o use o model es ima ion, as shown in Table 1. The e was di e en es ima ion da ase s used o each model componen , depending on he equi emen s o he model. This is desc ibed in mo e de ail in he In e egional Model Sys em De elopmen Repo (Camb idge Sys ema ics 2006). A e combining hese su eys, 6,882 comple ed su eys we e a ailable o use o model es ima ion, as shown in Table 1. The e was di e en es ima ion da ase s used o each model componen , depending on he equi emen s o he model. This is desc ibed in mo e de ail in he In e egional Model Sys em De elopmen Repo (Camb idge Sys ema ics 2006). The e a e highway, ai , ail, and local ansi ne wo ks o suppo bo h he u ban a ea and in e egional a el models. The socioeconomic da a includes household da a in ou classi ica ions (household size, income g oups, numbe o wo ke s, and ehicle owne ship) and employmen da a by ype. The e a e highway, ai , ail, and local ansi ne wo ks o suppo bo h he u ban a ea and in e egional a el models. The socioeconomic da a includes household da a in ou classi ica ions (household size, income g oups, numbe o wo ke s, and ehicle owne ship) and employmen da a by ype. 3.2 Accessibili y Measu es 3.2 Accessibili y Measu es In he de elopmen o he ip equency models, accessibili y measu es we e es ima ed o all ips o app oxima e he des ina ion choice logsum measu e. In he inal models, accessibili y measu es we e e ained o in a egional ips because he in a egional models main ained by he MPOs do no include des ina ion choice models, which a e necessa y o p oduce logsum measu es. Accessibili y measu es o in e egional ips we e eplaced wi h logsum measu es om he des ina ion choice models in he inal models, as desc ibed below. The e we e ou accessibili y measu es calcula ed, as ollows: In he de elopmen o he ip equency models, accessibili y measu es we e es ima ed o all ips o app oxima e he des ina ion choice logsum measu e. In he inal models, accessibili y measu es we e e ained o in a egional ips because he in a egional models main ained by he MPOs do no include des ina ion choice models, which a e necessa y o p oduce logsum measu es. Accessibili y measu es o in e egional ips we e eplaced wi h logsum measu es om he des ina ion choice models in he inal models, as desc ibed below. The e we e ou accessibili y measu es calcula ed, as ollows: • Au o peak wo k ip accessibili y • Au o peak wo k ip accessibili y ⎥ ⎥ ⎦ ⎤ ⎢ ⎢ ⎣ ⎡∑ ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ −+= d TimeTimeTo alEmp LN A mean p ea k au o p ea k d au o p ea k _ _ _ / × 2exp × 1 Table 1. To al o Su ey In e egional T ips by Mode, Dis ance, and Pu pose Table 1. To al o Su ey In e egional T ips by Mode, Dis ance, and Pu pose D i e D i e Ai Ai Rail Rail Bus Bus O he O he To al To al Long T ips Business 314 620 27 18 17 996 Commu e 263 15 9 1 74 362 Rec ea ion 1114 228 80 3 23 1448 O he 365 85 17 8 91 566 Sho T ips Business 381 14 48 3 15 461 Commu e 1136 0 168 9 108 1421 Rec ea ion 873 2 29 3 52 959 Sho O he 591 1 10 23 44 669 To al 5,037 965 388 68 424 6,882 Sou ce: Bay A ea/Cali o nia High Speed Rail Ride ship and Re enue Fo ecas ing S udy In e egional Model Sys ems De elopmen Repo , Table 2.6, Augus 2006. 64 Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 • Au o o -peak non-wo k ip accessibili y • Au o o -peak non-wo k ip accessibili y ( ) ⎥ ⎦ ⎤ ⎢ ⎣ ⎡−+++= ∑ d meano pea k au oo pea k d d d au o o pea k TimeTime Se iceEmp R e ailEmp H ouseholds LN A _ _ _ / × 2exp × )(1 • Non-Au o peak wo k ip accessibili y • Non-Au o peak wo k ip accessibili y ⎥ ⎥ ⎦ ⎤ ⎢ ⎢ ⎣ ⎡ ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ −+ =∑ d mean p ea k nonau o p ea k d nonau o p ea k TimeTimeTo alEmp LN A _ _ _ / × 2exp × 1 • Non-Au o o -peak non-wo k ip accessibili y • Non-Au o o -peak non-wo k ip accessibili y ( ) ⎥ ⎦ ⎤ ⎢ ⎣ ⎡−+++ =∑ d meano pea k nonau oo pea k d d d nonau o o pea k TimeTimeSe iceEmp R e ailEmp H ouseholds LN A _ _ _ / × 2exp*)( 1 whe e: whe e: To alEmpd = o al employmen a he des ina ion zone; To alEmpd = o al employmen a he des ina ion zone; Householdsd = o al households a he des ina ion zone; Householdsd = o al households a he des ina ion zone; Re ailEmpd = e ail employmen a he des ina ion zone; Re ailEmpd = e ail employmen a he des ina ion zone; Se iceEmpd = se ice employmen a he des ina ion zone; Se iceEmpd = se ice employmen a he des ina ion zone; Timepeak_au o = highway a el ime du ing he peak (based on conges ed ime) om he o igin zone o he des ina ion zone; Timepeak_au o = highway a el ime du ing he peak (based on conges ed ime) om he o igin zone o he des ina ion zone; Timepeak_nonau o = ansi a el ime du ing he peak (based on conges ed ime) om he o igin zone o he des ina ion zone; Timepeak_nonau o = ansi a el ime du ing he peak (based on conges ed ime) om he o igin zone o he des ina ion zone; Timeo peak_au o = highway a el ime du ing he o -peak (based on ee- low a el ime) om he o igin zone o he des ina ion zone; Timeo peak_au o = highway a el ime du ing he o -peak (based on ee- low a el ime) om he o igin zone o he des ina ion zone; Timeo peak_nonau o = ansi a el ime du ing he o -peak (based on ee- low a el ime) om he o igin zone o he des ina ion zone; Timeo peak_nonau o = ansi a el ime du ing he o -peak (based on ee- low a el ime) om he o igin zone o he des ina ion zone; Timepeak_mean = a e age a el ime om he o igin zone o all possible des ina ion zones du ing he peak pe iod, calcula ed om he a e age o su ey esponden s a el ime based on peak ne wo k imes; and Timepeak_mean = a e age a el ime om he o igin zone o all possible des ina ion zones du ing he peak pe iod, calcula ed om he a e age o su ey esponden s a el ime based on peak ne wo k imes; and Timeo eak_mean = a e age a el ime om he o igin zone o all possible des ina ion zones du ing he o -peak pe iod, calcula ed om he a e age o su ey esponden s a el ime based on o -peak ne wo k imes. Timeo eak_mean = a e age a el ime om he o igin zone o all possible des ina ion zones du ing he o -peak pe iod, calcula ed om he a e age o su ey esponden s a el ime based on o -peak ne wo k imes. 3.3 Logsum Measu es 3.3 Logsum Measu es Logsum measu es a e a means o es ima e a weigh ed a e age o a el ime and cos ha can be ed back om one componen o ano he . A summa y o he logsum measu es o each model componen is as ollows: Logsum measu es a e a means o es ima e a weigh ed a e age o a el ime and cos ha can be ed back om one componen o ano he . A summa y o he logsum measu es o each model componen is as ollows: • T ip equency models use “logsum” measu es om he des ina ion choice models, which a e in ended o cap u e he ac ha i is easie o make ele an in e egional ips om some zones han om o he zones. Fo ini ial model es ima ion, a syn hesized ne wo k zone accessibili y measu e was used. • T ip equency models use “logsum” measu es om he des ina ion choice models, which a e in ended o cap u e he ac ha i is easie o make ele an in e egional ips om some zones han om o he zones. Fo ini ial model es ima ion, a syn hesized ne wo k zone accessibili y measu e was used. • Des ina ion choice models use logsum measu es om he main mode choice models ha a e in ended o p o ide measu es o he composi e impedance ac oss • Des ina ion choice models use logsum measu es om he main mode choice models ha a e in ended o p o ide measu es o he composi e impedance ac oss 65 Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 72 Table 4. Access and Eg ess Mode Choice Models o Long T ips1 Access Models Eg ess Models Business/Commu e Rec ea ion/O he Business/Commu e Rec ea ion/O he Obse a ions 1,500 2,724 1,466 2,668 Final log-likelihood -1,662.3 -2,519.4 -2,121 -3,066.6 ρ2(0) 0.276 0.365 0.075 0.231 ρ2(cons) 0.003 0.068 -0.023 0.053 Coe . ( -s a ) Coe . ( -s a ) Coe . ( -s a ) Coe . ( -s a ) Le el o Se ice Cos ($) -0.075 cons ained -0.120 cons ained -0.075 cons ained -0.120 cons ained In- ehicle ime (min) -0.060 cons ained -0.030 cons ained -0.060 cons ained -0.030 cons ained Ou o ehicle ime (min) -0.147 (-6.4) -0.083 (-2.5) -0.139 (-6.2) -0.060 cons ained VOT IVT ($/hou ) $48.00 $15.00 $48.00 $15.00 Ra io OVT/IVT 2.45 2.76 2.33 2.00 D i e and (un)pa k T a el alone -1.925 (-3.0) Fewe ca s han pe sons -1.547 (-2.2) -1.903 (-2.8) Low income -2.741 (-1.8) -1.960 (-2.8) -18.006 (-2.5) -1.263 (-1.1) High income 0.709 (1.6) 0.339 (1.4) To/ om con en ional ail -9.490 (-2.5) To/ om high-speed ail -2.251 (-1.8) Ai po is LAX -3.128 (-3.8) -1.275 (-1.7) Ai po is SFO -4.082 (-4.4) -3.036 (-2.6) Ai po is SJC -1.479 (-2.1) Ai po is SAN -1.410 (-2.3) -1.370 (-2.3) Ren al ca To/ om con en ional ail -5.0002 cons ained -5.000 cons ained -3.522 (-2.4) -1.176 (-3.1) To/ om high speed ail -0.552 (-2.4) No ca s in HH 5.110 (3.2) High income 2.953 (2.4) T a el alone -2.588 (-4.7) Low income -2.082 (-0.9) -1.891 (-3.7) Ge d opped o /picked up In- ehicle ime (min) -0.014 (-2.5) -0.031 (-3.1) -0.015 (-3.9) Household size 0.606 (2.9) 0.478 (2.8) 0.974 (2.8) Taxi Au o dis ance -0.084 (-4.8) -0.071 (-3.8) -0.126 (-7.9) -0.052 (-6.6) To/ om con en ional ail -2.827 (-2.6) -2.265 (-2.4) To/ om high-speed ail -1.092 (-2.1) 2.507 (3.6) T a el alone -0.877 (-1.8) -2.768 (-4.6) Low income -3.010 (-1.9) -3.002 (-2.3) -1.038 (-2.3) High income 0.849 (1.9) T ansi No walk eg ess -4.836 (-4.6) -1.807 (-1.9) Rail used in pa h 3.689 (5.2) 1.727 (2.4) 2.960 (5.0) To/ om con en ional ail 3.580 (5.2) 1.830 (2.8) To/ om high-speed ail 0.592 (0.7) 1.032 (1.9) T a el alone 1.569 (2.3) No ca s in HH 1.439 (1.7) Fewe ca s han pe sons 1.480 (2.1) Low income 0.846 (1.0) 1.216 (1.9) Walk To/ om ai po -5.0002 cons ained -2.634 (-1.0) -2.074 (-2.0) Nes ing and scaling Nes - ansi , walk, axi 0.387 (5.9) 0.451 (3.3) 0.280 (6.9) 0.470 (5.3) Scale on hypo he ical choices 0.682 (15.9) 1.000 cons ained 0.516 (9.8) 1.000 cons ained 1 Does no include al e na i e speci ic cons an s. 2 These we e la e educed o -3.0 du ing model calib a ion. Sou ce: Bay A ea/Cali o nia High Speed Rail Ride ship and Re enue Fo ecas ing S udy In e egional Model Sys ems De elopmen Repo , Table 3.12 and 3.13, Augus 2006. Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 • The long segmen s, axi, pa king, and en al ca s a e gene ally less desi able o ail s a ions han o ai po s, while ansi is mo e desi able om ail s a ions. Walking is e y a e o o om ai po s, cap u ing accessibili y a ec s ha a e no cap u ed well in he zone sys em. • D i e and pa k access is less likely a he busies ai po s – San F ancisco (SFO), Los Angeles (LAX), and San Diego (SAN) – and somewha a San Jose (SJC) as well. This may cap u e bo h cos and incon enience e ec s a hose ai po s. • Fo mos segmen s, hose in la ge households a e mo e likely o be d opped o . • In gene al, high income a o s en al ca , axi, and d i e and pa k, and low income sligh ly a o s ansi in some segmen s. • The e is a logsum coe icien less han 1.0 on he nes ha includes ansi , walk, and axi. Each o he o he h ee al e na i es is in i s own “nes ,” and scaled by he same logsum pa ame e o p ese e equal scaling a he elemen al le el. • The scale ( he in e se o he esidual e o a iance) o he hypo he ical choices ela i e o he ac ual choices was signi ican ly lowe han 1.0 o mos o he Eg ess model segmen s. This esul indica es ha many esponden s ha e di icul y making an accu a e assessmen o mode choice op ions in less amilia su oundings a he non-home end o hei ip, so ha hypo he ical choices should be weigh ed less in es ima ion han ac ual ones. The main mode choice models p oduce p obabili ies ha each ip will choose one o he main modes (au o, ai , con en ional ail, and high-speed ail). Se e al nes ing s uc u es we e es ed o he main mode choice models and he inal nes ing s uc u e chosen is shown in Figu e 4, wi h all he non-au o modes in a single nes . This s uc u e p o ided he mos logical and s a is ically sound nes ing s uc u e o he mode choice models. 73 Figu e 4. Main Mode Choice Model S uc u e Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 74 The main mode choice models we e based on s a ed p e e ence (SP) su ey da a. The o e all choice sha es in he SP da a we e a ound 50% o high speed ail, wi h mos o he o he choices o he esponden s’ ac ual chosen modes. The HSR choice sha e was highes o business ips and long ips, gi ing a i s indica ion ha HSR subs i u es mo e closely wi h ai han wi h ca . To p epa e he da a o es ima ion, he access and eg ess mode choice models we e i s applied o calcula e access and eg ess mode logsums o each al e na i e. Then, a nes ed logi model was es ima ed ac oss he ou main modes o each o he segmen s (only h ee al e na i es o he Sho segmen s, as ai was no a ailable o hose segmen s). The es ima ion esul s o he Long Segmen s a e shown in Table 5. Some esul s o no e include he ollowing: • The cos and in- ehicle ime pa ame e s we e no cons ained du ing model es ima ion and p oduce easonable alues o ime. In gene al, he alue o ime o he longe , mo e expensi e ips is highe han o he sho e , mo e equen ips. This is a ypical esul . • The alue o equency (headway) is signi ican o all segmen s, bu was only abou 20 pe cen as la ge as he in- ehicle ime coe icien . I wai ime we e hal he headway and alued wice as highly as in- ehicle ime, hen we would expec he same coe icien on headway and in- ehicle ime. Fo hese modes, and pa icula ly ai , headway is less ela ed o wai ime han i is o scheduling con enience. Because none o he le els used in he SP had headways highe han a ew hou s, he implica ions o scheduling may no ha e been la ge enough o g ea ly in luence mode choice. This coe icien was cons ained o ma ch in- ehicle ime based on commen s om he pee e iew panel. • The alue o eliabili y is ai ly low o all segmen s, al hough wi h he co ec sign. I is e y di icul o measu e he e ec o eliabili y in a la ge-scale mailou SP su ey, so we decided o use a somewha highe e ec o eliabili y in applica ion, based on e idence om o he models ha his was easonable. • Those a eling wi h o he s a e mo e likely o use ca and less likely o use ai . This e ec was also es ed on he cos coe icien s and no ound o be signi ican , so his ela i e mode p e e ence appea s o be ela ed o mo e han jus cos – such as he ac ha people can sha e d i ing o long ips. Pa y size models we e es ima ed o gene a e hese da a, bu a e no included he e o b e i y. • People in la ge households a e mo e likely o use ca . E en hough we al eady ha e he g oup/alone segmen a ion, people in la ge households a e likely o be in la ge g oups. • Highe income gene ally a o s ai and high-speed ail e sus au o. • Low au o a ailabili y wi hin he household is ela ed o a lowe chance o choosing he au o. • A nes wi h ai , ail, and HSR, (wi h ca in i s own “nes ”) p oduced a logsum coe icien below 1.0 o all segmen s, indica ing ha his was a easonable nes ing s uc u e o in e egional ips. • The access mode choice logsums we e es ima ed wi h posi i e coe icien s in he ange o 0.14 o 0.46 o all segmen s. • Fo he long ips, he eg ess mode accessibili y seems o ha e somewha mo e in luence on mode choice han does he access mode. T a ele s may be less Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 75 cons ained a he home end, whe e hey know he op ions and can use hei own au o, han hey a e a he des ina ion end. Table 5. Main Mode Choice Models1 1 Does no include al e na i e speci ic cons an s. 2 These we e la e cons ained du ing model calib a ion o ma ch in- ehicle ime based on commen s om he pee e iew and he modeling eam. 3A e he headway coe icien was cons ained, his a io becomes 1. Sou ce: The model was e-es ima ed when he headway measu es we e cons ained and does no ma ch p e iously published e sions o his model - Bay A ea/Cali o nia High Speed Rail Ride ship and Re enue Fo ecas ing S udy In e egional Model Sys ems De elopmen Repo , Table 3.15, Augus 2006. Long T ip Business/Commu e Rec ea ion/O he Obse a ions 2,918 5,075 Final log-likelihood -1,998 -3,936 ρ2(0) 0.380 0.309 ρ2(cons) 0.151 0.154 Coe . ( -s a ) Coe . ( -s a ) Main Mode Cha ac e is ics Le el o Se ice Cos ($) -0.017 (-12.8) -0.035 (-18.5) In- ehicle ime (min) -0.018 (-13.4) -0.011 (-14.2) Se ice headway (min) -0.0042 (-3.9) -0.0032 (-3.6) Reliabili y (% on ime) 0.023 cons ained 0.005 (1.9) Access Mode Choice Logsum 0.136 (3.4) 0.204 (3.7) Eg ess Mode Choice Logsum 0.171 (3.9) 0.399 (7.1) Implied Value o Time IVT ($/hou ) $63.64 $18.45 Ra io Headway/IVT3 0.21 0.24 T ip Cha ac e is ics Ca – T a el in a G oup 2+ 1.086 (4.6) 1.43 (9.1) Ai – T a el in a G oup 2+ -0.356 (-2.8) -0.505 (-3.7) Household Cha ac e is ics Ca – Household Size 0.182 (1.2) 0.296 (4.4) Ai – High Income 1.18 (4.6) Con en ional ail – High Income 0.613 (1.4) High-speed ail – High Income 1.147 (4.8) Ca – Less han 2 Ca s pe 2+ Household -0.308 (-2.3) Nes ing Nes – ai , ail, high-speed ail 0.692 (10.4) 0.738 (13.0) Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 76 4 In a egional Models In a egional models will be used o o ecas high speed ail ips wi h bo h ends wi hin an u ban a ea ha has mo e han one p oposed high-speed ail (HSR) s a ion. These a eas a e he San F ancisco Bay A ea, G ea e Los Angeles, and San Diego egions. Regional a el o ecas ing models in hese a eas will be modi ied o o ecas u ban high-speed ail ips o he San F ancisco and Los Angeles a eas. The ma ke segmen s o in a egional a el include ypical ip pu poses such as home-based wo k, school, uni e si y, shopping, social- ec ea ional, and o he ips as well as wo k- and non-wo k- ela ed non-home-based ips. San Diego is he only o he egion ha con ains he possibili y o in a egional high-speed ail ips, bu he es ima e o hese ide s is e y low ela i e o he o he egions; and he le el o e o o de elop, calib a e, and apply he egional mode choice model is e y high, so we decided o de elop in a egional ide ship o San Diego using a popula ion-based es ima e a he han a adi ional mode choice model. To model in a egional ips, we elied on he ip gene a ion and dis ibu ion models in each o he u ban a eas and modi ied exis ing mode choice models. The u ban mode choice models include a a ie y o ansi modes, bu no speci ically a high-speed ail mode. The MTC u ban mode choice models we e modi ied o inse a high-speed ail mode based on coe icien s and cons an s om he commu e ail mode, as a conse a i e es ima e. The SCAG u ban mode choice model was buil om he MTC amewo k. Following is a b ie desc ip ion o he model implemen a ion o each o he u ban a eas: • San F ancisco Bay A ea - The San F ancisco egional model was enhanced o include ansi submodes (BART, commu e ail, ligh ail, e y, local bus, and exp ess bus) in he mode choice model. This allowed o easie inclusion o he high-speed ail mode in he model. The new mode choice model was alida ed a he egional le el o ma ch obse ed ide ship numbe s by ope a o . • Sou he n Cali o nia Associa ion o Go e nmen s Region - The Sou he n Cali o nia Associa ion o Go e nmen s (SCAG) mode choice models we e de eloped using he pa ame e s and s uc u e o he MTC model in combina ion wi h he SCAG ne wo ks and ip ables. This model was alida ed a he egional le el o ma ch obse ed ide ship numbe s by ope a o . U ban ip ables om he MTC and SCAG me opoli an a eas we e added o he in e egional ips o he assignmen . 5 Model Applica ion 5.1 Model Valida ion The alida ion o he combined in e egional and in a egional (u ban) models was comple ed o he yea 2000, because he a ailable obse ed da a o 2000 was mo e obus han o any o he yea . This s a ewide model was es ima ed om a combina ion o exis ing and new household and in e cep a ele su eys collec ed in Cali o nia and combined wi h in a egional ips gene a ed om egional and s a ewide sou ces. Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 77 The alida ion wo k included he calib a ion p ocess, de elopmen o da a used o obse ed a el beha io , and documen a ion o he esul ing calib a ion pa ame e s o he in e egional ips. In addi ion, his wo k included summa ies and easonableness checks on he in a egional ips de i ed om he MPO ip ables. These we e no sepa a ely alida ed o calib a ed, because each MPO has p o ided assu ances ha hese ip ables we e alida ed. T ips by mode om he in e egional models we e combined wi h in a egional ips by mode o assign o he highway, ai , and ail ne wo ks. Table 6 p esen s a summa y o he 2000 in e egional ips by mode and ma ke . Highway ips we e con e ed om pe son ips o ehicle ips using ehicle occupancy ac o s de i ed om he Cal ans S a ewide T a el Su ey. In addi ion, highway ips we e sepa a ed in o peak and o -peak ime pe iods so ha peak and o - peak ip ables could be assigned sepa a ely o he highway ne wo k. This ensu es ha peak-pe iod a el imes would mo e accu a ely e lec conges ion ha occu s in he peak pe iod. Following he de elopmen o peak and o -peak au o ehicle in e egional ips, hese we e combined wi h he au o ehicle in a egional ips. These in a egional ips come om ou sou ces: MTC, SANDAG, SCAG, and Cal ans. The Cal ans S a ewide Model is used o es ima e in a egional ips o all he o he egions (excep MTC, SANDAG, and SCAG) so ha he au o ip able will be ep esen ing all s a ewide a el. This ensu es ha conges ion wi hin each smalle u ban a ea is adequa ely ep esen ed. Valida ion o he base yea assignmen s by mode in ol ed de ailed e iew o obse ed and modeled olumes. Fo ai , hese e iews ocused on assignmen s o he majo ma ke s. Fo ail, hese e iews ocused on assignmen s by ope a o . Fo highway, hese e iews ocused on assignmen s by ga eway and by egion. A summa y o he assignmen s by mode is p o ided in Table 7. Table 6. 2000 Daily In e egional T ips by Mode Ma ke Au o Ai Rail To al Pe cen o To al LA o Sac amen o 7,479 4,935 - 12,414 1% LA o San Diego 257,441 100 5,395 262,936 17% LA o SF 28,031 26,867 - 54,898 4% Sac amen o o SF 137,739 25 1,816 139,580 9% Sac amen o o San Diego 175 2,858 - 3,033 0% San Diego o SF 4,630 10,309 - 14,939 1% LA/SF o SJV 205,205 3,393 926 209,524 14% O he o SJV 281,750 243 344 282,337 19% To/F om Mon e ey/ Cen al Coas 275,794 3,532 1,105 280,431 19% To/F om Fa No h 184,506 3,005 16 187,527 12% To/F om W. Sie a Ne ada 59,192 668 11 59,871 4% To al 1,441,942 55,935 9,613 1,507,490 100% Pe cen o To al 95.7% 3.7% 0.6% 100% Sou ce: Bay A ea/Cali o nia High Speed Rail Ride ship and Re enue Fo ecas ing S udy Final Repo , Table 5.1, July 2007. Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 78 Table 7. 2000 Daily Assignmen s by Mode Mode Uni s Obse ed Model4 Di e ence Pe cen Di e ence Ai Boa dings 54,2711 54,876 605 1% Rail Boa dings 16,7102 17,743 1,033 6% Au o Vehicle Coun s 27,145,3003 25,206,373 (1,938,927) -7% 1Sou ce: U.S. Depa men o T anspo a ion FAA O&D en-pe cen sample da abase 2Sou ce: In e egional ail ope a o s and MTC 3Sou ce: Cal ans, MTC and SCAG a ic coun da abases Sou ce: Bay A ea/Cali o nia High Speed Rail Ride ship and Re enue Fo ecas ing S udy Final Repo , Table 5.2, July 2007. E en hough he ai and ail assignmen s we e e y small compa ed o au o, hese we e c i ical o he e alua ion o high-speed ail, so a g ea a en ion o he alida ion o hese modes was impo an . Fo he majo ma ke s and ope a o s, hese compa ed e y well wi h obse ed numbe s. Au o assignmen s we e p ima ily alida ed based on ga eways along he high-speed ail co ido s. These compa ed e y well o obse ed a ic coun s. Addi ional alida ion e o o e ine and imp o e he highway assignmen s is ecommended i his model we e o be used o highway planning pu poses. Compa ison o he 2030 o ecas o a No-Build scena io was comple ed o alida ion o ensu e ha he 2030 o ecas s a e easonable o each model componen . O e all, he e is a 42 pe cen inc ease in households and a 51 pe cen inc ease in employmen , and he e is a 62 pe cen inc ease in in e egional ips. The 2030 in e egional ip able is p esen ed in Table 8. The highe pe cen o in e egional ips compa ed o s a ewide household and employmen g ow h is a e lec ion o he expansion o he egions beyond hei egional bo de s, causing mo e a ele s o make in e egional a el ins ead o in a egional a el. The au o assignmen s ( ep esen ed by o al ehicle miles a eled) inc ease by 73 pe cen om 2000 o 2030, which is also caused by a ele s ha ing o go u he o each hei des ina ions. These a e p esen ed in Table 9. Table 8. 2030 Daily In e egional T ips by Mode Ma ke Au o Ai Rail To al LA o Sac amen o 12,636 8,105 – 20,741 LA o San Diego 340,862 96 25,898 366,856 LA o SF 30,253 25,351 – 55,604 Sac amen o o SF 174,844 26 11,798 186,668 Sac amen o o San Diego 164 5,258 – 5,422 San Diego o SF 5,038 18,259 – 23,297 LA/SF o SJV 360,177 9,609 6,237 376,023 O he o SJV 553,466 1,944 4,792 560,202 To/F om Mon e ey/Cen al Coas 426,056 5,886 2,077 434,019 To/F om Fa No h 320,667 5,957 962 327,586 To/F om W. Sie a Ne ada 96,404 1,177 335 97,916 To al 2,320,567 81,668 52,099 2,454,334 Sou ce: Bay A ea/Cali o nia High Speed Rail Ride ship and Re enue Fo ecas ing S udy Final Repo , Table 5.3 July 2007. Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 79 Table 9. 2000 and 2030 Assignmen s by Mode Mode Uni s 2000 Model 2030 Model Di e ence Pe cen Di e ence Ai Boa dings 54,876 80,643 25,767 47% Rail Boa dings 16,430 30,653 14,222 87% Au o Vehicle Miles T a eled 748,606,510 1,297,116,168 548,509,657 73% Sou ce: Bay A ea/Cali o nia High Speed Rail Ride ship and Re enue Fo ecas ing S udy Final Repo , Table 5.4, July 2007. Rail boa dings inc ease a a highe a e han au o, indica ing ha as conges ion inc eases; mo e a ele s a e aking ail, as expec ed. Ai boa dings do no inc ease as as as ail o au o because he ai a es inc eased and equencies dec eased be ween 2000 and 2005, making ai a less a ac i e op ion. The 2005 obse ed ai le el o se ice was kep cons an h ough 2030. The p ima y eason o signi ican changes in ai se ice om 2000 o 2005 was he Sep embe 11 e o is a acks in 2001, which a ec ed ai a el mo e han o he modes. 5.2 Fo ecas Resul s Table 10 p esen s a summa y o he ips by mode and mode sha es o he base yea (2000) and he u u e yea (2030) wi h and wi hou he high-speed ail p ojec . High- speed ail cap u es o e 7 pe cen o he ips and d aws om all o he modes. 5.3 Sensi i i y Tes s A se ies o sensi i i y es s we e conduc ed o es he impac s o changes in le el o se ice on high-speed ail ide ship and e enue. These es s we e designed o assis in de eloping an imp o ed ope a ing plan, op imum a es, and o unde s and he impac s o po en ial changes in assump ions o he ai and au o modes. The esul s o he sensi i i y es s a e p o ided in Table 11. Table 10. Summa y o T ips and Mode Sha es o Base and Fu u e Condi ions 2000 Base Yea 2030 wi hou HSR 2030 wi h HSR 2030 Di e ence T ips Mode Sha e T ips Mode Sha e T ips Mode Sha e T ips Pc o To al Au o 1,441,942 95.7% 2,320,567 94.5% 2,193,248 89.2% -127,319 -71% Ai 55,935 3.7% 81,668 3.3% 53,823 2.2% -27,845 -16% Rail 9,613 0.6% 52,099 2.1% 31,790 1.3% -20,309 -11% HSR 179,482 179,482 100% To al 1,507,490 100.0% 2,454,334 100.0% 2,458,343 100.0% 4,009 No e: The 4,009 di e ence in 2030 ips wi h and wi hou HSR demons a es how much induced a el is a esul o HSR. Sou ce: Bay A ea/Cali o nia High Speed Rail Ride ship and Re enue Fo ecas ing S udy Model Valida ion Repo , Table 7.5, July 2007 and Bay A ea/Cali o nia High Speed Rail Ride ship and Re enue Fo ecas ing S udy Ride ship and Re enue Fo ecas s, Table 2.2, Augus 2007. Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 80 Table 11. Sensi i i y Tes s o High-Speed Rail Sensi i i y Tes Change in Le el o Se ice Pe cen Change om Base Boa dings Re enues High-Speed Rail Le el o Se ice Tes s Highe HSR Fa es 25% inc ease -13% 2% A e age Daily Headways HSR headways1 -15% -14% Highe HSR F eq 100% inc ease 15% 16% Exp ess Se ice SF/LA Double F eq SF/LA o SJV, SD/SF o SAC 22% 24% Ai and Au o Le el o Se ice Tes s Highe Ai /Au o Times 6% inc ease2 6% 6% Highe Ai /Au o Cos s 50% inc ease 46% 53% Combined Le el o Se ice Tes s Highe HSR Fa es and Highe Ai /Au o Cos s 25% inc ease in a es, 50% inc ease in cos s 13% 19% Highe HSR Fa es and Highe Ai /Au o Cos s 50% inc ease in bo h 31% 40% Highe HSR Fa es and Highe Ai /Au o Cos s 100% inc ease in a es, 50% inc ease in cos s -6% 1% 1 A e age daily headways assume ha he headway in he peak and o -peak pe iods a e equal. This e ec i ely inc eases peak headways and dec eases o -peak headways. 2The 6 pe cen inc ease in a el ime was based on a 30-minu e inc ease in a el ime om San F ancisco o Los Angeles by ca . Sou ce: Bay A ea/Cali o nia High Speed Rail Ride ship and Re enue Fo ecas ing S udy Final Repo , Table 7.1, July 2007. The esul s show ha imp o emen s in high-speed ail equencies can suppo much highe high-speed ail ide ship; inc eased high-speed ail equencies in he majo co ido s (San F ancisco o Los Angeles, Los Angeles o San Joaquin Valley, San Diego o Sac amen o, and San F ancisco o Sac amen o) we e hen e ained o he al e na i es analysis. These esul s also show ha aising high-speed ail a es will no signi ican ly inc ease e enues, unless his is combined wi h di e en assump ions o ai and au o cos s. Assump ions ega ding ai and au o cos inc eases emain a di icul issue, gi en he ola ili y in hese cos s in he pas 5 yea s alone. The sensi i i y es s do show ha high-speed ail ide ship is highly sensi i e o he assump ions o ai and au o cos s and can inc ease as much as 46 pe cen wi h a 50 pe cen inc ease in ai and au o cos s, which seems qui e easonable compa ed o cu en ends in hese cos s. 6 Summa y The a el o ecas ing models de eloped o p edic ing high-speed ail al e na i es o he s a e o Cali o nia ha e se e al immedia e bene i s o e p e ious ide ship o ecas ing me hods used in he s a e: hey a e ne wo k-based and p o ide mo e accu a e assessmen s o ime and cos adeo s wi h o he modes, modal choices a e sensi i e o eliabili y, pa y size, and de ailed access and eg ess op ions, induced a el is assessed based on changes in le el o se ice o all modes, and in a egional a el is es ima ed based on de ailed u ban a ea models whe e in e egional a el is es ima ed based on s a ewide models es ima ed om obse ed a el beha io . The in a egional and in e egional models a e in eg a ed o assess impac s o conges ion on o he modes and o e lec di e ences in peak and o -peak condi ions. Ou wa e , Tie ney, B adley, Sall, Kuppam, Modugula, Jou nal o Choice Modelling, 3(1), pp. 58-83 81 The p ima y ad ancemen in his model is he addi ional le el o de ail (4,600 zones used o all modeling componen s wi hou sampling), he inclusion o peak and o -peak assignmen s, and he consis en use o logsum accessibili y measu es a all le els o he models ( om access and eg ess models up o ip equency models). These models we e es ima ed using e ealed p e e ence da a and by combining mul iple su ey da ase s, a mo e obus es ima ion da ase was possible. The e a e some a eas whe e hese models may be imp o ed o o he s a ewide and egional planning ac i i ies. The ip equency models could bene i om addi ional da a on weekly o mon hly long dis ance a el, because a one-day snapsho does no p o ide as s ong a basis o a el decisions as longe - e m da a would p o ide. The des ina ion choice models could also be imp o ed by including da a on special gene a o s, such as Disneyland. Las ly, he mode choice models could bene i om a ou -based me hodology, ecognizing ha decisions on mode a e a ec ed by bo h he ou bound and e u n po ions o he ip. In hese cases, he models could bene i om addi ional da a and esou ces ha we e beyond he o iginal scope o he p ojec . These in eg a ed s a ewide models o e a comp ehensi e ool o o ecas long and sho dis ance a el in Cali o nia. The sepa a ion o a el in o ma ke segmen s based on dis ance (sho and long), pu pose (business, commu e, ec ea ion and o he ) and a el ma ke s (in e - and in a egional) p o ide a obus and accu a e assessmen o mul imodal a el a he s a ewide le el. 7 Re e ences Alge s, S., 1993, In eg a ed S uc u e o Long-Dis ance T a el Beha iou Models in Sweden, T anspo a ion Resea ch Reco d 1413, T anspo a ion Resea ch Boa d, 141-149. Ben-Aki a, M. and Mo ikawa, T., 1990, Es ima ion o Swi ching Models om Re ealed P e e ences and S a ed In en ions, T anspo a ion Resea ch Pa A, 24(6), 485-495. Bha , C.R., 1995, A He e oscedas ic Ex eme Value Model o In e ci y T a el Mode Choice, T anspo a ion Resea ch Pa B, 29(6), 471-483. Bha , C.R., 1998, Accomoda ing Va ia ions in Responsi eness o Le el-o -Se ice Measu es in T a el Choice Modeling’, T anspo a ion Resea ch Pa A, 32(7), 49-57. Bha , C.R., 1997, Co a iance He e ogenei y in Nes ed Logi Models: Econome ic S uc u e and Applica ion o In e ci y T a el, T anspo a ion Resea ch Pa B, 31(1), 11-21. Booz-Allen & Hamil on, 1989, Demand ‘Model Es ima ion: Final Repo ’, p epa ed o AMTRAK. B adley, M.A. and Daly, A.J., 1997, Es ima ion o logi choice models using mixed s a ed p e e ence and Re ealed P e e ence In o ma ion’, in S ophe , P.R. and Lee-Gosselin, M. (Eds.) Unde s anding T a el Beha iou in an E a o Change, Ox o d: Pe gamon, 209-232. B and, D., Pa ody, T.E., Hsu, P.S. and Tie ney, K., 1992, Fo ecas ing High-Speed Rail Ride ship’, T anspo a ion Resea ch Reco d 1342, 12-18. Camb idge Sys ema ics wi h Ma k B adley Resea ch and Consul ing, 2006, Bay A ea/Cali o nia High-Speed Rail Ride ship and Re enue Fo ecas ing S udy: In e egional Model Sys em De elopmen Repo ’, p epa ed o he Me opoli an T anspo a ion Commission and he Cali o nia High-Speed Rail Au ho i y.