Evaluation of a Support System for Large Area Tourist Evacuation Guidance: Kyoto Simulation Results

Transcription

1 CI-II 2012, Kobe, Japan, November 20-24, Evaluaton of a upport ystem for Large Area Tourst Evacuaton Gudance: Kyoto mulaton Results ek Knugasa 1, Tomoko Izum 2, Yosho Nakatan 3 Abstract Most studes on provdng evacuaton gudance have targeted resdents, wth lttle consderaton for evacuaton gudance for vstors to the area, such as toursts and busnesspeople. Accordngly, ths study targets the development of a system that supports the safe and effcent evacuaton of toursts from dsaster areas to specfc safe destnatons. The system models the evacuaton behavor of toursts and then smulates an evacuaton process n whch a specfc evacuaton gudance method s utlzed. A major characterstc of toursts n dsasters s that they tend to converge on the lmted number of ralway statons, whch may result n severe crowdng and panc. The system therefore makes t possble to compare and evaluate the effectveness of varous evacuaton gudance methods. The effectveness of the system was tested by smulaton of evacuaton processes that utlze a phased evacuaton gudance method to be ntroduced n Kyoto, the most popular tourst destnaton cty n Japan. Index Terms evacuaton smulaton, tourst, evacuaton gudance I. INTRODUCTION N Japan, whch has recently suffered the unparalleled ITohoku earthquake, there s a possblty that further large earthquakes, such as Tonanka and Nanka Earthquakes or a major earthquake n the Tokyo Metropoltan Area, wll occur n the near future. Despte ths, the government has postoned toursm as a major economc growth area [1]. The Toursm Naton Promoton Basc Law came nto effect n 2007, and the number of toursts has subsequently been ncreasng, although any major future earthquakes wll result n a decrease of toursts n the short term. Manuscrpt receved July 1, 2012; revsed eptember 5, Ths work was supported n part by a grant from the trategc Research Foundaton Grant-aded Project for Prvate Unverstes made avalable by the Mnstry of Educaton, Culture, port, cence, and Technology, Japan (MEXT), ( ), and as part of Kyoto Cty funded research. 1. Knugasa belongs to the Graduate chool of cence and Engneerng, Rtsumekan Unversty, Kusatsu cty, hga, Japan (phone: ;fax: ;e-mal:cc006064@ed.rtsume.ac.jp). 2 T. Izum belongs to the Graduate chool of cence and Engneerng, Rtsumekan Unversty, Kusatsu cty, hga, Japan (phone: ;fax: ;e-mal:zum-t@s.rtsume.ac.jp). 3 Y. Nakatan belongs to the Graduate chool of cence and Engneerng, Rtsumekan Unversty, Kusatsu cty, hga, Japan (phone: ;fax: ;e-mal:nakatan@s.rtsume.a c.jp). Most countermeasures aganst dsasters target resdents and rarely target non-resdents, such as toursts and commuters. When regonal dsaster preventon plans were examned at the prefectural level n Japan, 11 out of 30 prefectures were dentfed as ncludng toursts, but n very lttle detal. In Japan, as a dsaster-prone country, dsaster countermeasures amed at toursts not only need to protect vstors, but also the resdents of tourst areas. Toursts may decrease f there s a rsk of sufferng severe damage from dsasters. The tme needed for recovery of tourst ndustres wll take a far longer perod than most people may assume. Countermeasures aganst dsasters n place for toursts can therefore protect not only toursts, but also the local toursm ndustry and ndeed the toursm reputaton of Japan. The possblty of a secondary dsaster occurrng depends not only on the scale of the orgnal dsaster or the countermeasures n place, but also the state of evacuaton procedures. Plans for regonal dsaster countermeasures need to nclude accurate gudance of evacuees to ther destnatons n addton to the materal aspects of dsaster preventon facltes and equpment. Thus t s very mportant that approprate evacuaton gudance and desgn gudelnes are well prepared beforehand. Devsng evacuaton gudance for toursts, however, s much more dffcult than gudance for resdents, whch does not need to consder long dstance evacuaton n a large-scale area. Evacuaton experments are mpractcal because of the large numbers of partcpants and areas, and thus computer smulaton appears to be the most effectve method. Wth the am of solvng these problems, the purpose of ths study s to provde admnstratve staff wth an envronment n whch they can plan and verfy safety gudance by computer, for quck gudance of toursts to safe destnatons, n a broad range of areas. II. EVACUATION GUIDANCE FOR TOURIT Ths secton outlnes some problems n exstng evacuaton gudance, and ntroduces phased evacuaton gudance, a new method of evacuaton gudance. A. Exstng evacuaton gudance Few systems have been developed to support evacuaton gudance for toursts. Among them, one example developed by Kyoto Unversty n Japan nvolves an nstructor who conducts evacuatons for evacuees va moble phones [2]. Locatons of evacuees are constantly supervsed based on locaton data provded by GP (Global Postonng ystem) and are /12/$ IEEE 440

2 CI-II 2012, Kobe, Japan, November 20-24, replcated n a vrtual cty called FreeWalk, whch provdes a brd's-eye vew of the area. The nstructor constantly observes the behavor of evacuees n the FreeWalk vrtual world whle provdng them wth nstructons va moble phones. Another system, developed by Wakayama Unversty n Japan, enables evacuees to send dsaster nformaton and receve evacuaton gudance va dsaster nformaton statons nstalled n a cty [3]. A number of wreless devces therefore need to be nstalled throughout the area, and Bluetooth s used for communcaton between the users moble phones and wreless devces. B. Phased Evacuaton Gudance The evacuaton gudance method proposed by Nakatan et al. of Rtsumekan Unversty, Japan nvolves evacuees beng guded to temporary tourst s desgnated near sghtseeng spots n a phased manner, n order to prevent concentraton at any one tme n the central part of a cty [4]. It features stagng posts as temporary refuges of toursts on the way to ther destnaton such as ralway statons. Cooperators n dsaster areas, such as souvenr shop staff and tourst agents, are requested to send dsaster nformaton to the emergency management center. The emergency management center then determnes and changes evacuaton nstructons based on ths nformaton before sendng the nstructons to the cooperators. The cooperators then gude evacuees to the nearest tourst accordng to the nstructons they receve. The stagng posts functon as a buffer zone to prevent toursts from rushng nto ralway statons n the central parts of ctes and thereby enablng fast evacuaton of toursts from dangerous areas to safer areas, and reducng the fatgue caused by evacuaton on foot. The specfc effectveness of ths method, however, has not yet been verfed. III. YTEM CONTRUCTION A. Approach The admnstratve staff needs to be supported wth nformaton systems, as t s dffcult for vstors to evacuate by themselves wthout accurate evacuaton gudance because they are strangers n the area. The desgn of an approprate evacuaton gudance method thus requres understandng of the evacuaton stuatons that can actually occur when specfc gudance methods are adopted. Because large-scale evacuaton experments are mpractcal, our approach s to model tourst behavor tendences by computer and smulate ther behavor to fnd the most effectve evacuaton gudance method. mulatng evacuaton behavor has been nvestgated n varous felds. However, no smulator has been devsed that can be used to verfy the effectveness of evacuaton gudance methods for toursts. Recently, RoboCup Rescue smulaton systems as large-scale mult-agent systems have been challenged [5]. Most RoboCup Rescue systems cannot smulate evacuaton behavor of toursts to target resdents and rescue methods n ctes that are tourst areas. We therefore developed a new system that smulates the evacuaton behavor of toursts n a large area usng a specfc method and then quanttatvely evaluated the results. B. ystem creen Confguraton Ths secton outlnes the specfcatons of ths system n reference to Fgure 1. Fg.1: ystem screen Evacuaton gudance methods nput space: The evacuaton gudance method s entered here. The detals of ths nformaton are shown n Table 1. The orgn s the start pont of the evacuaton and the destnaton s the end pont. The stagng post s an ntermedate pont between the orgn and the destnaton. The watng tme s the tme spent watng at the stagng post. The stagng post s entered many tmes, and the watng tme has to be entered wth the stagng post. The evacuaton route s changed by draggng the route lne on the map. The number of evacuees equals the number of evacuees at the orgn. These are consdered one data set and can be entered many tmes. The evacuaton gudance method s then determned usng ths nput. In bref, ths process decdes whch evacuees should go to, and va whch route, for each destnaton. TABLE 1 CONTENT OF INPUT INFORMATION Evacuaton nfo. Input detals Orgn 1 tagng post 2* Watng tme Destnaton Evacuaton route Number of evacuees 2 Address or longtude and lattude Address or longtude and lattude or nothng Watng tme at the stagng post Address or longtude and lattude Draggng the route lne on the map Number of evacuees at the orgn mulaton dsplay space: Ths space shows the evacuaton route and evacuaton behavor graphcally. In Fgure 1, shows the screen durng enterng gudance methods on, and * shows the enlarged screen of after the start of the smulaton. Ths fgure shows the transton of evacuees from Kyomzu Temple to Kyoto taton. Evacuees n transt are shown as a thck lne. One mnute n real tme s one second n the smulaton. tuaton dsplay space: Ths space shows the evacuaton stuaton, ncludng the number of evacuees n the area. Operaton button: Ths space s used to start, stop, pause, and resume the smulaton /12/$ IEEE 441

3 CI-II 2012, Kobe, Japan, November 20-24, C. ystem process The followng flowchart represents the process after the smulaton starts (Fgure 2). Fg.2: Flowchart of process after start of smulaton Move Evacuees represents the transton of evacuees from each step to the next. A step s the mnmum unt from the orgn to the destnaton. Each step ncludes the next step poston, dstance, and wdth of the road. The move s calculated based on ths nformaton and transports evacuees n each step from the last step poston to the next step poston. Ths move results n a lne beng drawn from the last step to the next step, thus depctng how the evacuees move n a bold lne. Evacuaton Complete shows that the evacuees n the last step have reached ther destnaton. The system ncludes condtons that change ther peed of Movement, wth that speed dependng on densty of the crowd. The crowd densty shows the number of persons per 1 m2, and s calculated based on the wdth, dstance and number of evacuees n a step area. A hgh crowd densty results n a decrease n the walkng speed and thus a delay n the evacuaton. Below are the speed equaton (1) and a flowchart of the process after the start of the smulaton (Fgure 3). no tart Move Evacuees Evacuaton Complete yes End V ( ρ ) = 1.1ρ (1) ρ:crowd densty (person/ m2 ) v:walkng speed (meter/second) Change Wdth of Road or Meet yes Change Crowd Crowd Densty <4 persons/ m2 yes Calculate peed (1) no Move peed Change (Fgure 4) peed:0.3652(m/s) Fg. 3: Flowchart of change n speed Togawa revealed the relatonshp between the densty and the walkng speed. Frun [6] dscovered that 4 persons per 1 m2 no leads to a sharp decrease n the walkng speed. In ths system the walkng speed s calculated usng ths nformaton, wth Equaton (1) beng based upon t. In the system the walkng speed s changed when the wdth of the road wdens or when two evacuee groups meet and merge (Fgure 3). A narrow road results n an ncrease n the crowd densty. The sze of the step then changes, but the number of persons n each step does not. Evacuaton groups mergng also ncreases the densty. The number of persons n a step changes, but the sze of the step does not. If the densty s lower than 4 (persons/ m2 ) the walkng speed depends on Equaton (1), but f the densty s hgher than 4, the walkng speed s (meter/second), and the densty remans at 4. The length of crowd then ncreases, revealng that the crowd extends backward when t has passed the lmt. The current verson of the system does not completely take nto consderaton changes n the wdth of the road and the mergng of two groups. When two groups merge, the walkng speed does not slow down. The length of a group also does not change even when the wdth of the road changes. These defcences are currently beng mproved upon, thus obtanng more realstc results. IV. YTEM EVALUATION The effectveness of the system was examned usng Japan s most popular tourst destnaton cty, Kyoto. As the phased evacuaton gudance method was proposed by Nakatan for use n the tourst evacuaton gudance feld n Kyoto, we examned the method by usng t n our system and obtaned an evaluaton of the results from experts. The system does not currently nclude all features n full, but s at a level where t can be evaluated. A. Kyoto Kyoto s world-famous tourst destnaton cty and was vsted by 50,210,000 toursts n The number of toursts durng the autumn folage season n November was the largest n the year, numberng 6,590,000 toursts. The followng lst outlnes the features of Kyoto s toursts. These features are consdered to mply a tendency for Kyoto to be crowded wth many evacuees n a dsaster stuaton and to cause panc at statons, because most toursts are from the Knk regon, ncludng Osaka, Hyogo, Nara and hga Prefectures, and vst by tran. A prompt decrease n travel and toursm after a dsaster s also expected for many elderly female and ndependent vstors. Many female toursts: 63.5% of toursts are female Many elderly toursts: 63.5% are over age 40 Many toursts from the Knk regon: Knk regon: 61.2%(from wthn Kyoto Prefecture: 10.5%) Kanto regon, ncludng Tokyo: 14.0% Many ndependent toursts: 90.3% not wth tour groups Many day toursts: 73.6% are day trppers Many repeaters: 54.5% of toursts have vsted more than 10 tmes(24.5% sad 5-9 tmes.) Many ralway users: 61.8% vst by tran /12/$ IEEE 442

4 CI-II 2012, Kobe, Japan, November 20-24, The possblty also exsts that an earthquake wll occur n Kyoto. Accordng to a report from the commttee of the Headquarters for Earthquake Research Promoton of the Japanese government, the probablty of ground moton equal to or larger than a sesmc ntensty of 6 Lower occurrng wthn 30 years n Kyoto Cty s more than 3% [7]. Ths probablty s a hgh level even n earthquake-prone Japan. The probabltes of Tonanka and Nanka Earthquakes occurrng are 60% and 55%, respectvely, whch are very hgh when compared to the probabltes of other earthquakes. It s predcted that such an earthquake wll cause a ground moton of a sesmc ntensty of 6 Lower n Kyoto. B. mulaton Condtons Evacuaton gudes, when large-scale dsasters occur, have to gude evacuees to safe evacuaton areas chosen as temporary s n Kyoto Cty. After resumpton of transportaton ncludng tran servces, many evacuees n some temporary s wll smultaneously head for safe destnatons ncludng Kyoto taton because they hope to return to ther homes quckly. A safe and effcent gudance method therefore s needed between temporary s and the destnaton. Ths smulaton verfes the evacuaton stuaton after temporary evacuaton s complete and tran servces resume. The stuaton mples that the evacuees from each temporary go through a stagng post to the specfc destnaton. When amng for early arrval, drectly gong to the destnaton, wthout gong through the stagng posts, s of course more effectve. However, long-dstance walkng becomes a burden for elderly female evacuees, and tends to cause crowdng of toursts at the destnaton. It s therefore requred to verfy how the stagng posts should be used for the mplementaton of approprate and safe evacuaton gudance at the tme of dsaster. We verfy the evacuaton gudance n cases n whch the number of stagng posts s 0, 1 and 2, and compare those evacuaton stuatons. Ths smulaton assumes the envronment of Table 2. A Tonanka or Nanka Earthquake, whch have the hghest probablty of occurrence, s envsoned to occur around noon n November when the number of toursts s the largest. Ths stuaton wll requre long-dstance walkng and cause confuson n the crowd. TABLE 2 AUMED ENVIRONMENT OF IMULATION Purpose of evacuaton Gudance to statons for toursts gong home gudance Targeted cty Kyoto Tme of year Assumed earthquake esmc ntensty Tran servce stuaton Evacuaton gudance method Noon, weekend, November Tonanka / Nanka Earthquake esmc ntensty 5 Upper to 6 Lower JR Lne ncludng Kyoto taton s resumed after 6 hours Phased evacuaton gudance The specfc temporary, stagng post, destnaton, number of evacuees, and evacuaton route n ths smulaton are explaned below. Fgure 7 shows these specfc postons. Table 3 shows whch temporary to go to va whch stagng post for the destnaton, and letters a-k n Table 3 correspond to those on Fgure 7. When there are zero stagng posts, the gude of the temporary would drectly go to destnaton va the shortest route (Condton A). When there s 1 stagng post, the gude would go to tagng Post 1 va the shortest route and, after arrvng, take a 60-mnute break and then go to the destnaton (Condton B). When there are 2 stagng posts, the gude would go through tagng Post 1 and tagng Post 2 lke n Condton B, and then go to the destnaton va the shortest route (Condton C). Locatons a, c, e, j, and k of the temporary s are the nearest to each tourst ste. Number of evacuees n the temporary s ndcates when toursts, n the ten most popular tourst stes n Kyoto, are evacuated to the nearest. The number of toursts at each ste was calculated based on the number of vstors per hour. tagng Posts 1 and 2 denote safe evacuaton areas chosen as temporary s n Kyoto Cty and are near the route between each temporary and the specfc destnaton. Destnaton was envsoned to be JR Kyoto taton. Other statons also were consdered, but the effectveness of phased evacuaton gudance was consdered to be obvous for Kyoto taton, as t s surrounded wth stagng posts. e does not go through the stagng posts as the dstance between the and the destnaton s short, and j does not go through tagng Post 2, as the dstance from to s short and safe evacuaton areas are not nearby. (number of evacuees) TABLE 3 GOING THROUGH TAGING POT tagng post 1 tagng post 2 a (9,500) b d c (8,400) e f e (4,000) j (20,300) k (8,500) h g V. REULT Destnaton A. Result of the smulaton Tables 4, 5 and 6 each dsplay the elapsed tme from the start of gudng to the next poston n the cases of A, B, and C. For example, n the case of Table 6, the gude reaches tagng Post 1 after 80 mnutes and takes a 60-mnute break there, then moves to tagng Post 2 n 69 mnutes and takes a 60-mnute break agan, and fnally reaches the destnaton n 25 mnutes. Fgures 4 and 5 show the evacuaton behavor dsplayed n 2 of Fgure 1, 95 mnutes after start of the smulaton. Ths smulaton showed the occurrence of a long lne from the orgn to the destnaton n all results. Moreover, n the result of Condton A, the ntersecton n front of j has a tendency to be crowded wth many evacuees n the dotted lne n Fgure 4. In Condton B, crowded space dd not occur as shown n Fgure 5, and also dd not occur n Condton C. Fgure 6 compares n the form of a graph Condtons A, B, and C. Ths shows the change n the number of evacuees reachng at ten mnutes ntervals. The horzontal axs shows the elapsed tme from the evacuaton start. Evacuees begn to reach 76 mnutes after the start of the evacuaton, and fnsh after /12/$ IEEE 443

5 CI-II 2012, Kobe, Japan, November 20-24, mnutes n Condton A, 314 mnutes n Condton B and 375 mnutes n Condton C. TABLE 4 REULT OF CONDITION A a 167 c 148 e 76 j 73 k 121 TABLE 5 REULT OF CONDITION B tagng post 1 a 80 b 90 c 85 e 76 e 76 j k 40 h 88 Mnutes Destna ton Mnutes Destna ton TABLE 6 REULT OF CONDITION C Mnutes tagng post 1 tagng post 2 Destna ton a 80 b 69 d 25 c 85 e 55 f 18 e 76 j k 40 h 64 g 14 e f g Fg.4: creen after 95 mnutes n Condton A e h h k j k j Persons ~70 ~90 ~110 ~130 ~150 ~170 ~190 ~210 ~230 ~250 ~270 ~290 ~310 ~330 ~350 ~370 A B C Mnutes Fg.6: Change n the number of evacuees reachng B. Consderaton The results of the smulaton wll now be dscussed. Condton C takes the shortest tme on foot. The elapsed tmes of Condtons A, B and C are 167, 90 and 85 mnutes respectvely. In Kyoto, n whch most toursts are mddle-aged and older women, duraton s very mportant. The lnes of evacuees dd not take the same route at the same tme n Condtons B and C. Amng at dfferent stagng posts thus can avod the occurrence of dangerous secondary dsasters. The most stable change n these 3 condtons occurs n C (Fgure 6). The volume of passengers carred by tran s unchanged, and stable change s therefore requred for avodng secondary dsasters. The most unstable change occurs n B. Havng stagng posts does not depend on stable change, but gves respte from movng on foot. A gudance method that consders both of these s requred. From the above, the effectveness of phased evacuaton gudance can be seen n the followng aspects: Establshng stagng posts avods a long crowd of evacuees, and rest tmes are provded. table evacuaton gudance s enabled n some stagng posts and at the last destnaton. The rsk of secondary dsaster due to crowdng of evacuees s mtgated by proper selecton of stagng posts. C. Evaluaton by Experts The results were evaluated by two persons from Kyoto Cty Fre Department wth regard to the effectveness of the system, wth the result beng that they consdered t effectve. Table 3 summarzes ther opnons. f g Fg.5: creen after 95 mnutes n Condton B /12/$ IEEE 444

6 CI-II 2012, Kobe, Japan, November 20-24, TABLE 3 OPINION OF KYOTO CITY FIRE DEPARTMENT TAFF No. Opnons 1. Change evacuaton route dependng on envronmental dsaster. 2. Consder evacuaton method usng emergency transportaton route. 3. Consder both wdth of road and shortest path. 4. Change the envronmental dsaster dependng on scale of earthquake. 1. It s rather dffcult for admnstratve staff to be able to completely predct what wll happen n earthquakes, and therefore the system needs to take varous stuatons nto consderaton n verfyng varous evacuaton gudance methods. 2. The admnstratve staff proposed the use of an emergency transportaton route for evacuaton, because emergency transportaton routes are only for publc use and unavalable to general vehcles. Ths assures the safety of the evacuaton to a certan level. Ths s also related to the next pont (3). 3. The wdth of roads must be more than 6 meters to avod congeston. The wdest and shortest paths need to be desgnated. For example, a road may not be useable because of the collapse of buldngs or fre. The system needs to take such possble dangers nto consderaton. VI. CONCLUION Ths paper shows the mportance of dsaster countermeasures for toursts, the necessty of a system that consders approprate evacuaton methods n large dsaster areas, and presents the structure, and evaluaton of such a system. In the future we would lke to develop a detaled behavoral model that ncludes meets, splts, stops and slowng of an evacuee lne wth psychologcal factors taken nto consderaton. We then hope to develop and evaluate the system wth the help of an expert. Although the system s applcable not only to Kyoto Cty but also other tourst destnaton ctes, frst of all, we are consderng applyng the system n Kyoto Cty. Fnally, our heartfelt apprecaton goes to Kyoto Cty Fre Department offcals. REFERENCE [1] Japan Toursm agency, Toursm Naton Promoton Basc Law. (Accessed July ) [2] H. Itou, H. Nakansh,. Kozum and R. Ishda, Transcendent Communcaton : Locaton-based Navgaton for a Large cale Publc pace, Proceedngs of Computer Human Interacton 2004, CHI2004, Aprl, 2004, Venna, Austra, pp [3] K. Nozak, et al., A Proposal for Dynamc Emergency Navgaton ystem n Dsasters, Proceedngs of The pecal Interest Group Techncal Reports of IPJ, 2007, 2007-DB-141 (29), pp (in Japanese) [4] Yosho Nakatan, A verfcaton and research report on collectng and makng assstance nformaton avalable on tourst s dsaster preventon, March, 2011, pp (In Japanese) [5] RoboCup Rescue, RoboCup Rescue Agents mulaton Project. (Accessed July ) [6] John J. Frun, Pedestran plannng and desgn, 1971, pp [7] The Headquarters for Earthquake Research Promoton: Natonal esmc Hazard Maps for Japan, km 1mle c h e k a j b d f g Fg.7: Postons of temporary s, stagng posts and destnatons /12/$ IEEE 445