IMPACT OF AIRPORT NOISE REGULATIONS ON NETWORK TOPOLOGY AND DIRECT OPERATING COSTS OF AIRLINES

Transcription

1 27 TH INTERNATIONAL CONGRESS OF THE AERONAUTICAL SCIENCES IMPACT OF AIRPORT NOISE REGULATIONS ON NETWORK TOPOLOGY AND DIRECT OPERATING COSTS OF AIRLINES Prksh N. Dksht*, Dnel A. DeLurents*, nd Wllm A. Crossley* *Dept. of Aeronutcs nd Astronutcs, Purdue Unversty, W. Lfyette, IN Keywords: rport, nose regultons, rlnes, network topology, opertng cost Abstrct Due to growng demnd for r trnsportton, rport nose cn be resonbly expected to ncrese. Ths, coupled wth n ncresng wreness of rport nose ssues, suggests tht the number nd restrctveness of nose regultons wll ncrese. Nose regultons ffect rlnes, nd compel them to lter ther opertons to mnmze the mpct of the regultons on ther opertng cost. The mn opertonl chnges for rlnes re the network of rports they servce (network topology), nd the fleet utlzton to servce ths network. Ths pper presents frmework to study the mpcts of nose regultons on the network topology nd drect opertng costs for rlnes. The study uses ths frmework to exmne four types of nose regultons, nd compres the effectveness, dvntges, nd dsdvntges of these regultons on rlnes from 2008 to Introducton There s suffcent scentfc evdence tht nose exposure cn nduce herng mprment, hypertenson, schemc hert dsese, nnoynce, sleep dsturbnce, nd decresed school performnce [1]. The growng wreness of the detrmentl spects of vton nose hs led to the formton of ctve nose control groups tht nclude communty, rlne, nd rport representtves. These groups hve mposed rport nose regultons to mnmze the dverse mpct of nose exposure. Opertonl restrctons, opertonl procedures, nose txes, nd nose exposure lmts re some common exmples of nose regultons. Becuse the nose regultons force rlnes to dpt ther network nd fleet utlzton, the rlnes devte from ther optml mnmum drect opertng cost (DOC). The DOC for n rlne ncludes costs tht re drectly ttrbutble to the rlne s opertons such s fuel costs, mntennce costs, crew costs, etc. Consderng the sme revenue, lower DOC mples hgher proft. Rther thn studyng the mpct of nose regultons only t the regulted rport, ths pper explores the effect of nose regultons t the ntonl rspce level. Thus, ths study ttempts to understnd the opertonl chnges due to nose regultons, nd the costs of these chnges to rlnes. 2 Methodology Ths study ms to combne the regultory nd opertonl spects of vton to llustrte the mpct of nose regultons on rlne opertons. The study uses clbrted network forecstng lgorthm, resource llocton problem, nd n rport nose model to nvestgte the mpct of nose regultons. Fg. 1 llustrtes the concept of the smulton model. Fg. 1: Smulton concept overvew 1

2 PRAKASH N. DIKSHIT, DANIEL A. DELAURENTIS, WILLIAM A. CROSSLEY The smulton begns by forecstng the next yer s network structure by usng the exstng network structure nd the projected demnd. Ths s followed by optmlly lloctng the fleet to meet the demnd over the network, whle stsfyng the specfed nose constrnts. The nose t ech rport s computed usng the results of the resource llocton. The cycle strts over when the new network s determned usng the exstng network, projected demnd, nd current nose levels. Secton 3 presents the network forecstng lgorthm, whle Sectons 4 nd 5 detl the resource llocton process nd the nose model, respectvely. 3 Network Forecstng Algorthm A network forecstng lgorthm predcts the ddton nd removl of lnks n the network. To study the opertonl chnges due to nose regultons on n rlne s network, t s crtcl to be ble to predct the chnge n n rlne s network. Integrtng ths cpblty nto the smulton helps understnd the nture nd extent of n rlne s response to regulton. To lmt the scle of the problem ths study uses the FAA s OEP-35 rports to represent the r trnsportton network. The OEP rports re commercl rports tht serve mjor metropoltn res nd serve s hubs for mjor crrers. More thn 70 percent of pssengers move through these rports. * These rports re chosen, becuse they cover wde geogrphc re, they support sgnfcnt percentge of the pssenger demnd, but smll enough subset to be computtonlly nexpensve. 3.1 Exstng Model DeLurents, et l. [2] presented nd compred severl network forecstng models. Amongst the models presented, the ftnessfuncton forecstng lgorthm ws selected for ths study becuse the model s node-bsed, whch s pproprte for rport-relted studes, * offces/to/publctons/oep/fq/arports/ndex.cfm nd externltes (e.g. rport nose) cn be esly ncorported nto the model. In ths model, the exstence of lnk n the next terton depends on the ftness vlue of the lnk. Ths network forecstng lgorthm computes ftness vlue for ech node () n the network bsed on node degree (k), egenvector centrlty (x), clusterng coeffcent (CC), populton (p), nd nodl weght (w). Eq. (1) presents the ftness functon formulton presented by DeLurents, et l. [2]. The ftness vlue of ech lnk n the system s product of the ftness vlues of the nodes, whch defne the lnk. Ths pproch follows the fundmentls of scle-free network, [3] becuse nodes wth hgher ftness vlues hve hgher probblty of constructng new lnk. CC CC x x NAS NAS k k p p NAS NAS w w NAS externlt es (1) Another dvntge of ths lgorthm s tht t only uses the prevous yer s network to predct the subsequent yer s network topology. Thus, less dt s requred to ntte the model. DeLurents, et l. [2] used ths lgorthm to predct the ddton of new lnks n the network wth n verge ccurcy of 16.64% for the yers 1990 to Here, ccurcy ws defned s the rto of the number of correctly predcted routes to the number of ctul new routes. 3.2 Modfed Network Forecst Model Although rlnes both dd nd remove lnks every yer, DeLurents, et l. [2] dd not predct the removl of lnks from the network. Moreover, the prmeters used n the ftness functon were eqully weghted. There s possblty tht weghted ftness functon mght provde better results. Ths pper explores these two possble mprovements. The orgnl model used probblty threshold to determne f prtculr route wll be dded to the network. Ths study specfed the number of lnks to be dded (deleted), nd 2

3 IMPACT OF NOISE REGULATIONS ON NETWORK TOPOLOGY AND DIRECT OPERATING COSTS OF AIRLINES selected the lnks wth the hghest (lowest) ftness vlues for ddton (deleton). To renforce the scle-free network structure, rport demnd growth ws ncorported s prmeter nto the nodl ftness functon. Moreover, n ddton to the ftness functon of the two nodes, the strength of n exstng lnk (.e. demnd between the two nodes) ws used to evlute node-prs for ddton nd deleton of new lnks. To summrze, the mprovements to the model were usng weghted ftness functon, ncludng rport demnd growth n the ftness functon, nd usng the strength of lnk to determne the lnk s ftness. Ths study tested these mprovements usng hstorcl dt from 1990 to network structure. Whle the FRATAR lgorthm s the most wdely used method of genertng trp dstrbutons [4], ths lgorthm hs sgnfcnt lmttons whch mke t unusble n our nlyss. Our study explores new network structures, nd the FRATAR lgorthm does not hve bss to determne the demnd on new routes becuse t determnes the optml forecst trp dstrbuton bsed on the current trp dstrbuton. Fg. 2 presents flow chrt of new lgorthm developed to llocte demnd over n evolvng network Hstorcl Dt To compute the prmeters n the ftness functon, the model requred network nformton, nd dt on demnd nd populton from 1990 to The Bureu of Trnsportton Sttstcs (BTS) mntns comprehensve dtbse of rlne-reported domestc pssenger demnd nformton begnnng from Ths dtbse ws used to obtn network nd pssenger demnd nformton. For ech OEP-35 rport, the county-level decennl census reports were used to compute the populton for the correspondng metropoltn res n 1990 nd The Census Bureu lso provdes yerly populton estmtes bsed on the decennl Unted Sttes census, whch ws used to compute the metropoltn populton n The popultons n the ntervenng yers were nterpolted bsed on the popultons n 1990, 2000, nd Cre ws tken to use the sme re for populton estmtes despte chnges n the reportng formt nd clssfcton of res Demnd dstrbuton An mportnt spect of network forecstng model s to dstrbute the demnd on Fg. 2: New demnd dstrbuton lgorthm In ths lgorthm, the new demnd for exstng lnks s equl to the product of the old demnd vlues nd the demnd growth rte. The unllocted demnd s defned s the dfference 3

4 PRAKASH N. DIKSHIT, DANIEL A. DELAURENTIS, WILLIAM A. CROSSLEY between the totl new demnd nd the new demnd llocted to the exstng lnks. Ths unllocted demnd s dstrbuted mongst the new lnks n proporton to ther ftness vlues. Ths lgorthm works wth both ncresng nd decresng demnd scenros. 3.3 Clbrton The network forecstng lgorthm s clbrted by determnng the weghtng of the ftness functon tht produces the mxmum predcton ccurcy of the lgorthm s defned n Secton 3.1. Ths pper uses n optmzton pproch to dentfy the optml weghtng for the prmeters n the ftness functon. Fg. 3 presents the decomposton scheme for the optmzton problem. Fg. 3: Decomposton scheme for optmzton Becuse n unconstrned nonlner optmzton technque (smplex serch method) ws unble to hndle the numerous locl mnm present n the soluton spce, ths study used the genetc lgorthm presented by Crossley, et l. [5] to fnd the optml weghtng. Intl experments showed tht route ddton nd route deleton emphszed dfferent components of the ftness functon. Therefore, developng seprte weghtng functons cheved the best predcton ccurcy. Tble 1 presents the optml weghtng for the route ddton nd deleton processes. These vlues show tht the clusterng coeffcent, egenvector centrlty, nd demnd growth were more mportnt to predct the routes tht should be dded, whle node weght, populton, nd lnk weght were more mportnt for forecstng route deleton. Tble 1: Prmeter weghts for ftness functon Prmeter Addton Deleton Node Degree Node Weght Clusterng Coeffcent Egenvector Centrlty Populton Demnd Growth Lnk Weght Fg. 4 presents the yerly ccurcy of the weghted nd un-weghted ftness functons n forecstng chnges n the network. Addton ccurcy s defned s the percentge of ctul ddtons predcted by the lgorthm. Deleton ccurcy s defned smlrly. The plot shows tht the ftness functons were better t forecstng ddton of routes thn deleton of routes. The verge predcton ccurcy mproved from 62.93% to 80.10% for route ddton, nd from 23.8% to 55.69% for route deleton. Ths study used the weghted ftness functon model obtned from the clbrton process, becuse t forecsts the network chnges better thn the un-weghted functon. Fg. 4: Predcton ccurcy The consderble dfference n predcton ccurcy of the un-weghted model presented here compred to DeLurents model [2] s ttrbuted to the dfference n the sze of the network. The gns from weghtng the prmeters re sgnfcnt, but cnnot be 4

5 IMPACT OF NOISE REGULATIONS ON NETWORK TOPOLOGY AND DIRECT OPERATING COSTS OF AIRLINES drectly compred to the vlues presented by DeLurents, et l. [2]. 3.4 Nose externltes DeLurents, et l. [2] brefly dscussed the possblty of ntroducng externltes nto the forecstng lgorthm. In the smulton model, the effect of nose on the network structure ws nvestgted by tretng rport nose levels s externltes n the fleet forecstng lgorthm. In ths study, the proporton of n rport s contrbuton to the network s nose re ws subtrcted from the rport s ftness vlue. The nose externlty s un-weghted, whch penlzes rports wth hgher nose levels. The nose model s descrbed n Secton 5. 4 Resource Allocton As stted erler, rlnes wll dpt to nose regultons by chngng the wy they utlze rcrft to stsfy demnd nd lower opertng costs. A resource llocton model provdes wy to determne the nture nd extent of n rlne s response to the nose regultons. Thus, resource llocton provdes wy to pproxmte the behvor of rlnes under these new constrnts. Ths pper used scled-down dptton of the resource llocton pproch presented by Zho, et l. [6]. To smplfy the resource llocton problem, ths pproch does not mke dstnctons for ndvdul rlnes, so t ssumes tht benevolent monopolstc rlne serves the network to stsfy ll the pssenger demnd. Ths study lso used the fleet bstrcton presented by Zho, et l. [6], where the entre fleet s dvded nto sx set-bsed clsses. A combnton of the most flown rcrft n 2005 n ech clss (representtve-n-clss rcrft), nd the rcrft wth the newest Entry-n- Servce (EIS) dte s of 2005 (best-n-clss rcrft) represent ll rcrft opertons n prtculr clss. The two ctegores provde dstncton between the stndrd nd the ltest technology n ech clss. Tble 2 presents the clss defntons nd the selected rcrft types n ech ctegory. Tble 2: Representtve nd best n clss rcrft Clss Sets Rep.-nclsclss Best-n CRJ 200 ERJ CRJ 700 ERJ B B B B B A B B Eq. (2) presents the formulton for the resource llocton problem. Objectve : Constrnt s : 12 1 C X, j 1,,35 ; Rnge dstnce, j 1,,35 ; Feld Length Mx Runwy 1,, 35; 1,, BT, j 1 MT 1 j1,, j 1,, 12; D Nose constrnt mnmze, j j, j ( DOC) j ; 1,,12 where,, j = rport ndces, = rcrft type, C = rcrft effectve cpcty, D = demnd, TA = turnround-tme, BT = block-tme, MT = mntennce-tme, X = number of round-trps, n = number of vlble rcrft (2) Whle Zho, et l. [6] used revenue mxmzton s ther objectve, ths study 12 1 TA 24 n 5

6 PRAKASH N. DIKSHIT, DANIEL A. DELAURENTIS, WILLIAM A. CROSSLEY mnmzed the DOC of the rlne. The objectve functon of the resource llocton problem reflects the rlne s prortes. Becuse lower DOC wll result n ncresng proft for the sme revenue, ths objectve ws consdered resonble surrogte for rlne behvor. Constrnt 1 checks tht the demnd between cty-prs s stsfed. Constrnts 2 nd 3 ensure tht n rcrft llocted to prtculr route cn servce tht route by trckng the rnge of the rcrft, nd the runwy length of the rports. Constrnt 4 prevents the overutlzton of rcrft. Constrnt 5 s the plceholder for ny nose constrnts. The nose model, specfclly developed for fleet-level studes, uses the nose energy equvlent of the FAA-publshed nose levels t the tkeoff, sdelne, nd pproch certfcton ponts. The nose model ccounts for the dfference between deprture nd rrvl opertons, the effect of tkeoff gross weght (TOGW) on the tkeoff nose, nd correltes well (normlzed RMSE = 4.79%) wth the predctons of FAA s Integrted Nose Model (INM). The model s lnerty wth respect to the number of rcrft opertons llows ts use s n objectve or constrnt n resource llocton problem. 5 Nose Model The nose model s n ntegrl prt of the smulton process. The nose model uses the opertonl nformton to compute the nose t ech rport n the network. The FAA s Integrted Nose Model (INM) s the stndrd rport nose model. INM s unsutble for fleetlevel studes becuse of ts long setup nd computton tmes [7]. Ths study uses the nose model developed by Dksht nd Crossley [7]. Ths nose model uses weghted-lner equton tht estmtes the re wthn the 65 db Dy-Nght Level (DNL) contour round n rport s lner functon of the number of rcrft opertons t tht rport. Eq. (3) presents ths nose model. Are 12 TO TO TO TO P Q 1 N X, 1 Arr Arr Arr Arr P Q 1 N X , where, X = number of opertons, = rcrft type, = rport ndex, δ = dy rto, TO = tkeoff, ARR = rrvl, N TO = 10 (EPNL/10)-7, N ARR = 10 ((EPNL-10)/10)-7, P = dytme rcrft coeffcent, Q = nghttme rcrft coeffcent..(3) 6 Dt The smulton model requres nformton on future populton chnges nd pssenger demnd s nputs to the network forecstng lgorthm. The study needs exmples of currently-mplemented nose regultons to smulte such regultons n the model. The followng prgrphs descrbe the source of ths dt. 6.1 Future Populton Estmtes The U.S. Census Bureu projected sttelevel populton estmtes from 2004 to The populton growth round ech rport ws pproxmted to be the sme s the populton growth n the stte(s) tht contned the metropoltn res correspondng to ech rport. Ths computton used the percentge chnge vlues for the perod to estmte the popultons from , nd the chnge vlues from to estmte the popultons from Future Pssenger Demnd The 2007 nd 2008 pssenger demnd ws extrcted from the BTS dt mentoned n secton The Termnl Are Forecst (TAF) s the offcl forecst of vton ctvty used to meet the budget nd plnnng needs of projectonsgesex.html 6

7 IMPACT OF NOISE REGULATIONS ON NETWORK TOPOLOGY AND DIRECT OPERATING COSTS OF AIRLINES the FAA **. The 2009 TAF dt ws used to obtn the pssenger demnd t the OEP-35 rports for the yers 2009 to Nose Regultons Nose regultons re mndted by the rport uthorty to reduce the negtve mpcts of rport nose on the surroundng communty. Over the yers, severl dfferent forms of nose regultons hve been mplemented. The Boeng Corporton mntns n up-to-dte lst of rport nose regultons. Ths study explored the mpct of mplementng nose regultons t three rports ATL, JFK, nd PHL. These rports were selected becuse they were well connected to the other OEP-35 rports, nd ny opertonl chnges t these rports would lkely cscde throughout the network. Another reson ws tht between these three rports, they spnned wde rnge of pssenger demnd levels. Ths study consdered four types of nose regultons 1) rport nose restrctons, 2) nose txes, 3) rcrft opertons quots, nd 4) rcrft nose restrctons. For ech type of regulton, sx levels of severty were smulted. The followng prgrphs dscuss ech of these regultons, nd ther mplementton n the smulton Arport Nose Restrctons Arport nose restrctons specfy the mxmum nose levels t rports. Ths nose level my be specfed usng ny metrc (e.g. re exposed to nose, number of nght-tme wkenngs, etc.). Ths study used the Are wthn the 65 db DNL contour s ts metrc, becuse the nose model mesures rport nose usng ths metrc. Becuse the bse yer of ths forecst model ws 2007, the rport nose regulton ws ted to the 2007 nose levels t the selected rports (.e. the rport nose for ny yer could not be more thn specfed percentge of the 2007 rport nose level). The sx levels of ** severty tested for rport nose restrctons were 95%, 100%, 105%, 110%, 115%, nd 120% of the 2007 nose levels Nose Txes Nose txes re reltvely new form of nose regulton. Nose txes levy fee (fxed or vrble) bsed on the nose exposure of ndvdul rcrft. The rport determnes threshold vlue, nd ll rcrft tht re bove the threshold hve to py the nose tx. Chngng ether the fee or the threshold vlue cn vry the tx. Ths study used n dptton of the nose tx leved by the Adelde Arport. The threshold vlue t Adelde, whch s computed s the sum of the deprture, sdelne, nd rrvl certfcton nose, s equl to 265 db. Eq. (4) presents the formul to compute the nose tx leved on rcrft exceedng ths threshold. The severty of the regulton ws vred usng the rte prmeter. The sx levels of nose txes tested n order of decresng severty were $1200, $1000, $800, $600, $400, nd $200. tx rte Arcrft Opertons Quots (4) Arcrft opertons quots lmt the dly number of opertons t ny rport. Ths type of regulton my ether hve dfferent quots for dfferent rcrft, or my only be pplcble to certn types of rcrft. Ths regulton ws mplemented smlrly to the mxmum rport nose regulton. The rcrft opertons quots lmted the totl opertons t the selected rports to specfed percentge of the 2007 vlue. The sx levels of severty tested for opertonl quots n order of decresng severty were 75%, 80%, 85%, 90%, 95%, nd 100% of 2007 opertons. 7.4 Arcrft Nose Restrcton Arcrft nose restrctons prohbt the use of certn rcrft t the selected rports bsed ( nose_ level 265)

8 PRAKASH N. DIKSHIT, DANIEL A. DELAURENTIS, WILLIAM A. CROSSLEY on ther nose exposure. Ths regulton my be consdered specl cse of rcrft opertons quots. Ths pper restrcted rcrft types on the bss of ether ther rrvl or deprture certfcton nose levels. Becuse the nose model uses vrous prevously stted fleet bstrctons, the number of types of rcrft n the fleet ws lmted. Therefore, the rcrft nose levels were dscrete, whch lmted the effectveness of ths regulton. Furthermore, due to the multple prmeters (rrvl nd deprture nose level lmts), t ws not lwys possble to compre two regulton levels. For purposes of brevty, the regultons re specfed usng the notton, [deprture_nose_lmt; rrvl_nose_lmt]. The regulton levels (n EPNdB) tested were [91; 97], [91; 98], [91; 100], [93; 97], [93; 98], nd [93; 100]. Here, [91: 97] s the most strngent, whle [93; 100] s the lest strngent regulton. convenence, the re wthn the 65 db DNL contour s referred to s the nose re. 8.1 Nose t Regulted Arports An mportnt mesure of the effectveness of ny nose regulton s the nose level t the regulted rports. Fg. 5 presents the mpct of regultons t ATL (lrgest regulted rport) n 2011 (md-wy pont of the smulton), whle Fg. 6 presents the correspondng plot for ATL n 2015 (end pont of the smulton). 8 Results The smulton ws run from 2007 to 2015 wthout nose regulton for the bselne cse nd for the vrous types nd levels of nose regultons detled n Secton 7. It s mportnt to note tht the selected levels of severty for ech type of regulton my not be equvlent. Whle the rtes for the nose txes were chosen rbtrrly, the most strngent regultory level for the other regultons reflects the most severe regulton tht llowed the resource llocton module to fnd soluton. Snce the reltve severty of ll rcrft nose restrctons could not be determned, only the results of the lest nd most severe regultons were plotted. The scenro wthout ny nose regultons s consdered the bselne scenro, nd ll regulted scenros were compred to the bselne scenro. Becuse the network s set up to dpt to the nose regulton, the results my not be compred drectly (becuse the vlues my correspond to dfferent network structures), but the bselne vlue provdes reference for evluton. For Fg. 5: Impct of regultons t ATL n 2011 Fg. 6: Impct of regultons t ATL n 2015 At the lest strngent level, none of the nose regultons were effectve n curbng the nose t the gven rport. At the most strngent level, nose txes nd opertonl quots were more effectve n curbng nose t the trgeted 8

9 IMPACT OF NOISE REGULATIONS ON NETWORK TOPOLOGY AND DIRECT OPERATING COSTS OF AIRLINES rports compred to the other restrctons. The mpct of ncresng the severty of the regultons ws lso more evdent n nose txes nd opertonl restrctons. These trends were dscernble n both 2011 nd In comprson wth the bselne, the opertonl restrctons nd nose txes were more effectve n 2015 thn n Ths cn be ttrbuted to combnton of chnges n fleet utlzton nd network topology. Secton 8.4 presents the mpct on network topology. To study the chnges n the fleet servcng regulted rport, Fg. 7 compres the chnges n utlzton of the fleet t ATL from 2011 to 2015 of the bselne scenro wth the level 2 opertons restrcton. 8.2 Nose t Other Arports n the Network In ddton to the nose t the regulted rports, ths pper nvestgted the mpct of these regultons on the other rports n the network. Fg. 8 presents the sum of the nose wthn the 65 db DNL contour t the 32 unregulted rports n 2011, nd Fg. 9 presents the correspondng vlues n Fg. 8: Nose t non-regulted rports n 2011 Fg. 7: Dfference n fleet utlzton Snce the Boeng ws the representtve rcrft, s well s the best-nclss rcrft, n clss 6, the opertons were grouped together. The bselne scenro showed n ncrese n the utlzton of best-n-clss rcrft n clsses 1, 4, nd 5. It lso showed reducton n the use of ll clss 6 rcrft, nd the representtve rcrft n clss 5. On the other hnd, the regulton forced the rlne to utlze lot more representtve rcrft n clsses 5 nd 6, whle sgnfcntly reducng the number of clss 4 representtve rcrft. Whle the bselne scenro smply chose newer technology rcrft to reduce DOC, the regulton forces the rlne to lower the number of opertons t the rport, resultng n n ncresed use of lrger rcrft rther thn the most cost-effectve. Ths exmple llustrtes the effect of the restrctons on the rlne s opertons. Fg. 9: Nose t non-regulted rports n 2015 Whle opertonl quots hd the most detrmentl effect on the nose t non-regulted rports n 2011, ther dverse mpct on nose t these rports ncresed exponentlly n Ths suggests tht t ws hrder for rlnes to cope wth opertonl quots n subsequent yers. A lrger proporton of noser 9

10 PRAKASH N. DIKSHIT, DANIEL A. DELAURENTIS, WILLIAM A. CROSSLEY rcrft were llocted to the non-regulted rports due to the bsence of nose regultons. Thus, the non-regulted rports bore the brunt of ths sub-optml llocton. In contrst to opertonl quots, the mpct of nose txes on nose t the nonregulted rports remned consstent n the 2011 nd 2015 scenros. Arport nd rcrft nose restrctons dd not hve lrge dverse mpct on these non-regulted rports n ether scenro. In 2011, there ppered to be correlton between the extent of the regultons nd the mpct on the nose t the non-regulted rports for nose txes nd opertonl quots. Ths trend contnued nd ws more vsble n 2015, even for rport nose regultons. 8.3 Drect Opertng Cost Whle nose regultons re mportnt, they should not be prohbtvely expensve for rlnes. The mpct on the rlne DOC mesures the reltve burden mposed on the rlne by these regultons. Fg. 10 presents the DOC for the rlne n 2011 under vrous nose regultons, whle Fg. 11 presents the vlues for the 2015 scenro. Becuse nose regultons result n suboptml llocton of resources, most of the regulted cses showed hgher DOC compred to the bselne. Ths dfference ws lrger n 2015 compred to Becuse the nose txes drectly mpct the rlne s DOC, the nose tx scenros hd the hghest penlty on the DOC n both 2011 nd Opertonl quots hd the second lrgest mpct on the rlne DOC. Thus, the two nose regultons tht were most effectve n reducng the rport nose were lso the most severe on n rlne s opertng cost. Moreover, for the nose tx nd opertonl quot scenros, there ws drect correlton between the extent of regulton nd the ncrese n DOC. Becuse the smulton s objectve ws to mnmze the DOC, the fleet llocton s n mportnt prt of the smulton. Fg. 12 compres the dfference n rcrft utlzton under the severest regultons of ech knd wth the bselne vlues n Fg. 10: DOC for OEP-35 rports n 2011 Fg. 11: DOC for OEP-35 rports n 2015 Fg. 12: Arcrft utlzton n

11 IMPACT OF NOISE REGULATIONS ON NETWORK TOPOLOGY AND DIRECT OPERATING COSTS OF AIRLINES The rport nose restrcton ncresed the utlzton of the representtve rcrft n clss 4 nd the best-n-clss rcrft n clss 3, becuse these rcrft hve lower contrbutons to rport nose per pssenger. Snce nose txes penlze the rlne on the bss of certfcton nose levels, under ths constrnt the smulton lowered the utlzton of the clss 4 representtve rcrft, nd ncresed the use of the best-n-clss rcrft n clss 3 nd the clss 2 representtve rcrft. Interestngly, the opertons quot restrcton ncresed the use of ll clss 1 nd 2 rcrft. Snce the smulton forced the rlne to use ll the lrger rcrft to servce the regulted rport (to mnmze opertons), the rlne requred mny smller rcrft to stsfy the demnd t other non-regulted rports. Ths lso explns the ncresed nose t nonregulted rports. The rcrft nose restrctons dd not produce ny sgnfcnt fleet-level utlzton trends. 8.4 Network Topology Network topology s the network structure creted by the lnks connectng the nodes of the network. For ech tme-step of the smulton, the network forecstng lgorthm dds nd removes lnks n the network. Becuse nose s ncluded s n externlty, the nose regultons hve n nfluence on the network topology of the rlne. The degree of node s the number of other nodes n the network connected to t. Over tme, rlne networks grvtte towrds scle-free structure. The degree dstrbuton plot fclttes the study of the mpct of nose regultons on the tendency to form scle-free networks. Fg. 13 presents degree dstrbuton of the network n 2011 for the most severe regultons, nd Fg. 14 presents the equvlent dstrbuton n The bselne network s smlr to sclefree network structure [3], but t s not true scle-free network due to the lmted sze nd hgh nter-connectvty between the nodes. The 2015 bselne network hd steeper slope nd shorter tl compred to the 2011 bselne network. In ll regulted 2011 scenros, except rcrft nose, the number of nodes wth degree 34 decresed, nd number of nodes wth degree 33 nd 32 ncresed. Ths pttern ws gn repeted wth the number of nodes of degree 31 decresng, nd number of nodes wth degree 30 ncresng. Fg. 13: Network structure n

12 PRAKASH N. DIKSHIT, DANIEL A. DELAURENTIS, WILLIAM A. CROSSLEY lrger mpct on the network thn the other regultons n All regultory scenros showed one rport wth low node degree vlue compred to the others. On nlyss, the nomlous rport ws dentfed s MIA. MIA ws consstently droppng lnks under nose regultons compred to the bselne. Ths decrese ws ccompned by correspondng ncrese n the connectvty of STL. Tble 3 presents the node degree vlues for both rports for ech scenro n 2011 nd Tble 3: Node degree: MIA nd STL MIA STL MIA STL bselne rport nose nose tx opertons quot rcrft nose Fg. 14: Network structure n 2015 Ths pttern ndctes tht ncludng nose s n externlty n the ftness functon lowered the ftness of the rports wth hever trffc, nd ncresed the probblty of lnkng of nodes wth lower trffc. Thus, these nose regultons opposed the trend towrds scle-free network structure. Ths trend ws lso evdent n the 2015 scenro, even for rcrft nose regultons. Arport nose nd nose tx regultons hve In comprson to the bselne, the nose regultons forced the smulton to ssgn suboptml rcrft to non-regulted rports. As explned n Secton 8.2, ths often ncresed the nose t the non-regulted rports. Snce nose ws ncluded s n externlty n the network forecstng model, the hgher nose lowered the ftness vlues of the non-regulted rports. Non-regulted rports tht hd low ftness vlues to strt wth (e.g. MIA hs low ftness vlue due to low demnd nd low clusterng coeffcent) were most ffected. As the smulton progressed, the mpct on MIA ws compounded. Other rports n the network, such s STL, were the unntended benefcres of the nose regultons. 9 Concluson In the ner future, mny rports wll consder enctng nose regultons due to the ncresng demnd for r trnsportton, nd the growng wreness of the ll-effects of rport nose. In ddton to the locl beneft, t s mportnt to consder the system-level effects of these regultons. 12

13 IMPACT OF NOISE REGULATIONS ON NETWORK TOPOLOGY AND DIRECT OPERATING COSTS OF AIRLINES Ths study nvestgted four types of nose regultons - rport nose lmts, nose txes, opertonl quots, nd rcrft nose restrctons - usng system-of-systems pproch. The smulton model used network forecstng lgorthm, resource llocton module, nd nose model to study the mpct of mplementng these nose regultons on the system. Nose txes nd opertons quots were the most effectve t regultng nose t the trgeted rports. Arport nd rcrft nose restrctons prevented further ncrese n nose, but dd not lower the nose sgnfcntly. Nonregulted rports n the network were most ffected by opertonl quots, but ths regultory pproch dd not ncrese DOC prohbtvely. On the other hnd, nose txes sgnfcntly ncresed the rlne s DOC, but hd smller mpct on the non-regulted rports. Although rport nd rcrft nose restrctons dd not lower the nose t the regulted rports, they prevented ny sgnfcnt ncrese n the nose re t these rports. Moreover, rport nd rcrft nose restrctons dd not sgnfcntly mpct ether the nose t the non-regulted rports, or the drect opertng cost of rlnes. Thus, these restrctons my be resonble mddle ground. Nose txes nd rcrft nose restrctons my be more effectve f they were bsed on the ctul nose contrbuton, rther thn certfcton nose levels. Mny rports currently employ combnton of nose regultons, nd such customzed mesures my be needed to ddress the ndvdul needs of ech rport. As seen n secton 8.4, nose regultons cn ffect other rports n the network n unntended wys. In ddton to the extent of the regultons, the proporton of the network tht s regulted s bound to ffect ll spects of the system. All rports n the network wll be ffected by nose regultons t ny rport n the network, nd t s mportnt to nlyze these ntrcte nterctons usng systems pproch. Whle nose regultons re mportnt, t s crtcl for the sustnble growth of vton tht these regultons re studed n frmework, such s the one presented n ths pper, before mplementton. 10 Contct Author Eml Address Prksh Dksht. pdksht@gml.com Copyrght Sttement The uthors confrm tht they, nd/or ther compny or orgnzton, hold copyrght on ll of the orgnl mterl ncluded n ths pper. The uthors lso confrm tht they hve obtned permsson, from the copyrght holder of ny thrd prty mterl ncluded n ths pper, to publsh t s prt of ther pper. The uthors confrm tht they gve permsson, or hve obtned permsson from the copyrght holder of ths pper, for the publcton nd dstrbuton of ths pper s prt of the ICAS2010 proceedngs or s ndvdul off-prnts from the proceedngs. References [1] Psscher-Vermeer, W. nd W.F. Psscher, Nose exposure nd publc helth. Envron Helth Perspect, Suppl 1: p [2] DeLurents, D., et l., Utlzton of Network Theory for the Enhncement of ATO Ar Route Forecst n 8th AIAA Avton Technology, Integrton, nd Opertons Conference (ATIO). 2008: Anchorge, AK. [3] Brbás, A.-L. nd R. Albert, Emergence of Sclng n Rndom Networks. Scence, (5439): p [4] Long, D., et l., A Method for Forecstng Commercl Ar Trffc Schedule n the Future. 1999, LMI for NASA Lngley: Hmpton, VA. [5] Crossley, W.A., et l., Usng the two-brnch tournment genetc lgorthm for multobjectve desgn n Structures, Structurl Dynmcs, nd Mterls Conference nd Exhbt. 1998, AIAA: Long Bech, CA. [6] Zho, J., et l., Assessng New Arcrft nd Technology Impcts on Fleet-Wde Envronmentl Metrcs ncludng Future Scenros, n 48th AIAA Aerospce Scences Meetng. 2010: Orlndo, FL. [7] Dksht, P.N. nd W.A. Crossley, Development of n Arport Nose Model Sutble for Fleetlevel Studes, n 9th Avton Technology, Integrton, nd Opertons Conference. 2009, AIAA: Hlton Hed, SC. 13