MODELING THE OPERATIONAL IMPACT OF AIR TRAFFIC CONTROL AUTOMATION TOOLS:

MODELING THE OPERATIONAL IMPACT OF AIR TRAFFIC CONTROL AUTOMATION TOOLS: A Case Study f Traffic Management Advisr Megan Smirti, Mark Hansen University f Califrnia at Berkeley NEXTOR Berkeley, CA, USA msmirti@berkeley.edu Xing Chen CSSI, Inc Arlingtn, VA, USA Abstract Traffic Management Advisr (TMA) is a decisin supprt tl develped t assist Traffic Management Units (TMU) in metering and sequencing arrival traffic. This study examines the use and impact f TMA during its early stages f deplyment at Chicag Center (ZAU). Determining impacts f use presents a methdlgical challenge because usage may depend n weather and traffic cnditins, pssibly leading t spurius results if simple with/withut cmparisns are made. In an effrt t islate the impact f TMA, this study therefre emplys an alternate methd. A preliminary understanding f TMA use is established thrugh summary statistics. This enables the develpment and use f detailed statistical mdels t islate the impact f TMA at ZAU. We find evidence thrugh these detailed mdels that TMA use increased capacity in specific cnditins and capacity variability was reduced in all scenaris. A simulatin f these results n delay at Chicag O Hare Internatinal Airprt (ORD) shwed that TMA use can decrease delay by 33%. Keywrds: Air Traffic Management, Capacity, Traffic Management Advisr I. INTRODUCTION The Federal Aviatin Administratin (FAA) develped the Free Flight Phase 1 (FFP1) prgram with the gal f autmating certain functins f air traffic cntrl t imprve perfrmance f the Natinal Airspace System (NAS). The FFP1 prgram established metrics used t evaluate system deplyments, which assisted the FAA in perfrming tests and evaluatins befre undertaking widespread deplyment f the tls. Tls analyzed in recent years include User Request Evaluatin Tl (URET) and Traffic Management Advisr (TMA), which is the fcus f this paper. TMA is part f a suite f tls that was planned t increase the efficiency f flight peratins in all five dmains f the NAS [1]. As discussed by Hansen [2], air traffic cntrl system evaluatins present a unique challenge. Because the NAS is affected by many diverse factrs, such as weather and demand, islating the impact f a specific air traffic cntrl enhancement is cmplicated. The challenge is even mre difficult during early stages f deplyment when the tl is used nly in selected time perids, which may be different frm nn-use perids in sme systematic ways. In this study, t islate the impact f TMA n airprt peratinal capacity, we extend an ecnmetric mdeling methd develped in [3] that cnsiders capacity as a randm variable. Our wrk cntributes t the develpment f cnsistent and credible evaluatin methds fr autmatin tls, which will becme increasingly imprtant as NAS mdernizatin prceeds. Sectin II f this paper prvides backgrund n TMA, describes its functinality, and discusses previus benefit studies. Sectin III intrduces summary statistics t aid in understanding hw TMA is used, and describes the data used in the analysis. Sectin IV defines an ecnmetric mdel used t determine the impacts f TMA implementatin and presents estimatin results. Sectin V islates the capacity and variance f capacity effects f TMA t determine a change in delay frm TMA use. Sectin VI cncludes the research with discussin and recmmendatins. II. TMA BACKGROUND The rle f TMA is t crdinate the transitin between center and cntrl airspace fr arrivals. TMA was designed fr decisin supprt fr the metering psitin f the Traffic Management Crdinatrs (TMC). Hwever, as discussed by Blic [4], the adaptatin, r actual use instead f intended use, f systems develped fr air traffic cntrllers (ATC) and traffic management crdinatrs (TMC) ften diverges frm the intended purpse. Fr example, at Ls Angeles center, TMA was initially used t display traffic in a larger area than was previusly available [2]. This increased shared situatinal awareness generated cnsiderable peratinal benefit even when the decisin supprt functinality was nt in use. TMA began initial daily use (IDU) at ZAU in June 2005. Adaptatin als tk place at ZAU, as TMA was used exclusively t facilitate the release f internal departures thse bund fr an airprt within the same Air Rute Traffic Cntrl Center (ARTCC) airspace. The TMA display screen is well suited t this functin because f a detailed arrival 127

schedule fr the majr airprts in the Chicag TRACON ORD and Chicag Midway. Implementatin at ZAU fllwed the successful implementatin f TMA at eight ARTCCs, with the first implementatin in 1996 at Frt Wrth. Later implementatins were supprted by studies finding benefits frm TMA implementatins at Frt Wrth and ther centers. These benefit studies relied n befre and after analysis, including summary statistics and regressin mdeling. Tw examples f such studies are belw. A. TMA at Minneaplis Center (ZMP) Thrugh a cmparative analysis f airprt acceptance rates (AAR) befre and after TMA deplyment, the FFP1 prgram ffice determined that TMA increased AAR at ZMP. A regressin analysis was then perfrmed t islate the impact f TMA. By defining AAR as a functin f TMA, metrlgical cnditin, and runway interactin, it was fund that the increase in the AAR mean was nt statistically significant. This regressin treated TMA as a dummy variable which was set t 1 t signify a time perid after TMA was deplyed. A similar study was perfrmed regarding the ttal peratins rate, r the sum f the airprt acceptance and airprt departure rates. This analysis fund a statistically significant increase in the peratins rate after TMA was deplyed. It was cncluded that ptimized arrivals flws under TMA allwed the cntrllers t release mre aircraft [5]. B. TMA at Ls Angeles Center (ZLA) The impact f TMA n internal release departures t LAX frm ther airprts within ZLA was examined after the June 2001 TMA implementatin. Similar t ZAU, TMA allwed the Traffic Management Unit (TMU) at ZLA t ptimize the release f these departures by fitting them in t the arrival stream withut causing delays. By calculating the mean delay befre and after the deplyment f TMA, it was fund that bth gate and airbrne delay decreased after TMA deplyment. It was cncluded that because ther airprts experienced increases in gate and airbrne delay fr the same time perid, TMA was able t reduce delay at LAX [6]. This study did nt include a regressin analysis and did nt cnsider ther factrs which culd have cntributed t a decrease in delay, such as changes in demand. III. EXPLORATORY TMA ANALYSIS Fr the purpse f mdeling the impact f TMA n airprt runway capacity, the peratinal impact at Chicag O Hare Internatinal Airprt (ORD) was chsen fr case study. Data were cllected fr the study perid f July 2005, immediately after IDU f TMA, t mid-march 2006. 1 Data were gathered frm the FAA s Aviatin System Perfrmance Metrics (ASPM) database. The Airprt Efficiency prtin f this database prvides variables n quarterly-hur arrival and departure cunt and demand at ORD, which will be explred in greater detail in Sectin IV. Each entry includes crrespnding infrmatin abut the meterlgical cnditin 1 The perid frm December 19 t 25 was excluded, because schedules and peratins are substantially changed by large vlumes f hliday travel. (MC), ther weather related infrmatin, and runway cnfiguratin. A TMA usage lg was cllected frm ZAU t match the perids in ASPM with the perids when TMA was explicitly being used by the TMCs. During the study perid, TMA was pwered n and available fr use frm 6AM t 8PM daily. Hwever, TMA was referred t spradically by the TMCs; the times when TMA was assisting TMCs was recrded in a usage lg [7]. T cmbine these data with ASPM data, time stamps n each f the data sets were matched. A. TMA Use at ZAU The fllwing summarizes TMA usage data with the gal f gaining a general understanding f the factrs affecting use f TMA during the study perid. Discussins with TMCs, managers, and cnsultants supprting TMA implementatin at ZAU revealed the plicies and prcedures affecting TMA use was spradic; therefre, this study will fcus n TMA usage perids rather than befre and after TMA deplyment perids. T determine the best mdel frmulatin, crrelatins between TMA use, meterlgical cnditins, and runway cnfiguratin are explred. 1) Meterlgical Cnditins Table I summarizes TMA use in terms f visibility cnditins at ORD. The three meterlgical cnditins classified are visual meterlgical cnditins (VMC), marginal visual meterlgical cnditins (MVMC), and instrumental meterlgical cnditins (IMC) [8]. Each quarter-hur data entry in ASPM is identified as either VMC r IMC. We further subdivided VMC int MVMC and full VMC, based n visibility criteria defined in [8]. TABLE I. CEILING AND VISIBILITY AVERAGES, BY METEOROLOGICAL CONDITIONS AND TMA USE IMC MVMC VMC OFF ON OFF ON OFF ON Celiling (100's Ft) 8.95 16.02 19.96 27.64 12.92 13.2 Visibility (statute mi) 3.14 2.06 7.93 7.94 9.57 9.71 n f bs. with TMA 54 165 1254 ttal n f bs. 1694 2445 19362 Frm Table I it can be seen that during the study perid there were very few bservatins f TMA use under IMC. Out f the 1473 perids that TMA was used, nly 3.67% (54 perids) were during IMC. Fr thse few perids when TMA was used during IMC, it was typically during high ceiling cnditins. The average ceiling cnditin under IMC and TMA use was almst duble that f the average ceiling cnditin under IMC with n TMA use. Cnditins under MVMC and VMC when TMA was and was nt in use are mre similar, althugh under MVMC the ceiling is cnsiderably higher when TMA is in use. 128

2) Runway Cnfiguratins Chicag O Hare Internatinal Airprt has 6 active runways in 3 pairs f parallel runways. There are a number f pssible runway cnfiguratins at ORD fr arrivals and departures that can be used at any given time. The five mst frequently used runway cnfiguratins fr arrivals and departures are shwn Table II, alng with the prprtin f time each is used and prprtin f TMA use. The cnfiguratin 4R, 9L, 9R 4L, 9L, 32L, 32R is knwn as the default cnfiguratin fr VMC and MVMC. TABLE II. Cnfiguratin FIVE MOST COMMON RUNWAY CONFIGURATIONS AT ORD % f Perids Cnfiguratin Used % f perids TMA Used 22R, 27L, 27R 22L, 32L, 32R 40.19 6.64 4R, 9L, 9R 4L, 9L, 32L, 32R 36.22 6.95 22R, 27L 22L, 32L, 32R 6.96 10.49 14R, 22L, 22R 9L, 22L, 27L 13.07 3.39 9R, 14L, 14R 4L, 9L, 22L 3.56 1.61 During the study perid, TMA was certified n tw runway cnfiguratins: 22R, 27L, 27R 22L, 32L, 32R and 4R, 9L, 9R 4L, 9L, 32L, 32R (referred t as cnfiguratin 1 and 2, respectively). This means that fr these cnfiguratins TMA predicts the time when flights reach ORD entry fixes with sufficient accuracy. TMA was mst likely used fr these cnfiguratins and fr 22R, 27L 22L, 32L, 32R. This tw arrival runway cnfiguratin was favred fr tw pssible reasns. First it is very similar t the certified cnfiguratin 1. Secnd, TMCs nted that TMA did nt recgnize the third runway in cnfiguratin 22R, 27L, 27R 22L, 32L, 32R when scheduling internal departures, a prblem that did nt arise when just tw arrival runways were in use. IV. ECONOMETRIC MODELING OF CAPACITY UNDER TMA The fllwing sectin intrduces the ecnmetric mdeling technique used t mdel and determine the impact f TMA n peratinal capacity at ORD. This technique is based n the mdel develped by Hansen [3] t determine the capacity impact f new runway develpment. A. Cunt and Demand Data Analysis T accurately determine the capacity impact f TMA, the peratins rate (peratin cunt per unit time) is cmpared with peratin demand per unit time. The data are divided int tw grups based n TMA use; data fr perids when TMA was in use are separated frm data cllected when TMA was nt in use. Data frm ASPM were used fr this analysis. The variable arrival (departure) cunt in the ASPM database indicates the number f arrivals (departures) in a time perid (defined as a 15 minute interval). The variable arrival (departure) demand represents the number f aircraft scheduled t arrive (depart) in a specific time perid. While demand fr an peratin ften leads t that peratin ccurring, scenaris exist where the arrival (departure) demand exceeds the arrival (departure) capacity, r the maximum number f aircraft that can perfrm the peratin in a given perid. In this case, sme aircraft will be queued. Aircraft cunting tward the demand in a given perid that d nt actually arrive (depart) in that perid are cunted tward demand in the subsequent perid. Thus the difference between cunt and demand in a given perid is essentially the size f the queue at the end f that perid. T measure demand, ASPM determines the expected arrival time f an aircraft by adding the en-rute time t the wheels-ff time. An arrival in a time perid befre the calculated time is cunted twards the demand in the earlier perid in which it arrives; an arrival at the calculated time is cunted tward the demand fr that perid; and an arrival after the calculated time is cunted tward the demand in all time perids between the calculated arrival and the actual arrival time. Departure demand is calculated similarly, based n the actual pushback time plus an airprt-specific unimpeded taxi time, r when a flight is subject t a grund delay prgram (GDP), the estimated time when the flight will be cleared fr departure under the GDP. The mdel develped fr this study will use the data t determine the change in capacity fr arrivals nly due t TMA use. A mdel is cnstructed which treats capacity as a randm variable, by calculating capacity as a functin whse distributin depends n weather, runway cnfiguratin, demand, and TMA use. This methdlgy uses statistical prcedures that estimate the relatinship between these factrs and capacity. T islate the impact f TMA, the capacity functin includes a dummy variable which is set t 1 if TMA is in use in time perid t, and it is set t zer therwise. The parameter f primary imprtance is the cefficient n the dummy variable representing TMA use. This parameter is the cntributin t capacity f TMA. If the cefficient is negative, it can be cncluded that TMA reduces capacity; if it is psitive, it can be cncluded that TMA increases capacity. This cefficient fr peratin type O (where O= arrivals nly fr this study) will be termed 0. The example in Fig. 1 depicts 0. The slid curve is a sample prbability distributin f runway capacity. The secnd dashed curve is a sample prbability distributin fr runway capacity when TMA is in use, but when ther cnditins (weather, etc.) are similar. The difference in the mean values f these curves, represented by the curve peaks, is 0. Fig. 1 depicts a case when TMA use affects nly the mean f the capacity distributin. TMA use may als affect the variance f the capacity distributin by cnsistently feeding traffic t the airprt at a mre cnsistent rate. Bth effects are cnsidered belw in sectin B. 129

prbability distributin f capacity withut TMA f(capacity) prbability distributin f capacity with TMA The mdel is estimated using a maximum likelihd methd, which will find the parameters that best fit the data. Mainly, we are interested in 0 and 0, the effects f TMA n the mean and the variance f the capacity distributin. The detailed mdel estimatin technique is discussed in great depth by Hansen [3]. 1) Illustratin f Censred Regressin Mdel Results Fr illustrative purpses, the full mdel results fr ne data set will be described in detail. We chse the mdel fr VMC cnditins and runway cnfiguratin 1 fr this illustratin. Estimatin results appear in Table III. capacity (peratins/hur) Figure 1. Depictin f 0, the cntributin f TMA t capacity. B. Operatinal Impact: Censred Regressin Mdel The mdel t be used in this sectin is a censred regressin, r tbit, mdel, which measures the difference in capacity due t TMA use. A censred regressin mdel is apprpriate because it is impssible fr a cunt value t exceed a demand value. Thrughput, r runway peratins per unit time, is therefre censred by demand. The tbit mdel frmulatin is belw. The mdel will calculate the capacity based n the knwn peratin demand and the knwn peratin cunt. T islate the impact f runway cnfiguratin and meterlgical cnditin, there are separate mdels fr each cnfiguratin and cnditin. We estimated the mdel fr 4 different data sets. Mdels were estimated fr VMC and MVMC and fr runway cnfiguratins 1 (22R, 27L, 27R 22L, 32L, 32R) and 2 (4R, 9L, 9R 4L, 9L, 32L, 32R). Each mdel cnsiders capacity as a functin f demand, windspeed, and TMA use. Each mdel als captures the variance f capacity, and analyzes the impact f TMA n this variance. The mdel specificatin is belw. Q ( t) min( D ( t), C ( t)) C ( t) A( t) W ( t) D ( t) Where: (t) Q (t) D (t) C A(t) is the cunt fr peratin f type (either arrivals r departures) in 15-minute time perid t; is the demand fr peratins f type in time perid t; is the ORD capacity fr peratins f type in time perid t; is equal t 1 if TMA is in use in time perid t and 0 therwise; W (t) is the windspeed in time perid t; is a stchastic errr term, assumed t be IID nrmal with mean 0 and variance + A(t);,,,, 2, 0 2 are parameters t be estimated. (1) Intercept TABLE III. Parameter Effect f TMA n capacity Effect f Windspeed Effect f Demand Variance Effect f TMA n Capacity Variance CENSORED REGRESSION MODEL RESULTS 2 0 Symbl Estimate (Standard Errr) T-Statistic 26.164 (0.333) 78.517 1.720 (0.412) 4.17214-0.201 (0.026) -7.797 0.000 (0.000) -0.055 5.994 (0.101) 59.338-1.267 (0.310) -4.089 The mdel results shw that the baseline quarter-hur capacity fr arrivals at ORD is 26.164 arrivals, which is the equivalent f 104.656 arrivals per hur. This is very clse t the benchmarked 100 arrivals per hur determined by the FAA [9]. The results als shw that when TMA is being used by the TMCs, arrival capacity is increased by 1.720 arrivals per quarter hur, r 6.880 arrivals per hur. This is equivalent t a 6.6% capacity increase. The results shw that windspeed decreases arrival capacity by -.201 arrivals per quarter hur, and that demand has n impact n capacity. The estimated variance is 5.994 arrivals per quarter hur squared, which is decreased by -1.267 when TMA is in use. All parameters except demand are significant at the 0.05 level (dented by the bldface type). 2) Mdel Results fr the Impact f TMA n Arrival Capacity and Variance f Capacity The impact f TMA n the capacity mean, measured by 0, and capacity variance, 0, fr the fur sets f MC and runway cnfiguratin are shwn in Table IV. 130

TABLE IV. MC & RW Cnfiguratin VMC, RW 1 VMC, RW 2 MVMC, RW 1 MVMC, RW 2 THE EFFECT OF TMA ON CAPACITY MEAN AND CAPACITY VARIANCE Values (Standard Errr) T-Statistic 1.720 (0.412) 4.172.302 (.357).846.822 (.914).899-1.318 (.961) -1.372 Values (Standard Errr) T-Statistic -1.267 (0.310) -2.976-1.892 (.345) -5.479-1.554 (.642) -2.420-1.784 (.718) -2.485 The 0 values are nt significant in three f fur meterlgical cnditins and runway cnfiguratin cases. Under VMC and runway cnfiguratin 1, capacity mean is significantly higher due t TMA. There is a pssible selfselectin bias in this case because it represents favrable cnditins, which culd encurage TMA use. The 0 values indicate the estimated change in capacity variance when TMA is in use. The results suggest that arrival capacity variance did decline when TMA was in use. We als nte that these results are cnsistent with the FFP1 LAX study [6], which fund less dispersin between arrival cunts and thrughput after TMA was implemented. If TMA usage at ZAU did in fact reduce arrival capacity variance, this wuld have an imprtant benefit. It wuld reduce delay, because a negative capacity deviatin is mre likely t have an adverse effect than is psitive deviatin t have a beneficial effect. In many cases, psitive deviatins cannt be fully explited because there is insufficient demand. While a negative deviatin can als be incnsequential, it is mre likely t cntribute t a queue ging int the next perid. The fllwing sectin explres hw the use f TMA can affect delay due t its capacity and variance impacts. V. DELAY IMPACT ESTIMATION T illustrate the ptential f TMA use t save minutes f flight delay, a simulatin was emplyed. The peratinal cunt if TMA was in use 100% f the time was simulated and cmpared with peratinal cunt if TMA had never been in use during the study perid. T further islate the capacity and variance effects f TMA, tw ptential peratinal cunt scenaris were calculated: ne with the capacity effect f TMA calculated alne ( 0 =0), and anther with bth the capacity and capacity variance effect. Operatinal demand was kept cnstant ver all scenaris t fully illustrate the delay changes due t TMA. A. Delay Calcuatin withut TMA Using demand and cunt data fr all quarter hur perids at ORD cllected fr January 2006, a cumulative cunt curve was cnstructed. A cumulative curve in this case is a plt f cumulative peratinal cunt n the y-axis and time n the x- axis. In the first perid, cumulative peratinal cunt (n 1 ) is equal t the cunt f peratins in perid ne (n 1 ). In the secnd perid, cumulative peratinal cunt (n 2 ) is the cunt in perid tw (n 2 ), plus the cunt in perid ne (n 1 ). Therefre the cumulative peratinal cunt in perid tw is n 2 =n 1 +n 2. The cunt in perid three is n 3 =n 2 + n 3, and s n fr all remaining perids. Cumulative demand is determined similarly. The hrizntal distance between any tw pints n the curves is equal t the wait time in queue that an peratin (arrival) was delayed. The area between the tw curves is the delay in flight-minutes fr the time perid f study. T illustrate hw this methd can be used t determine the delay savings ptential f TMA, the study perid f January 6, 2006 frm 13:15-21:15 was chsen. The first step was t cnstruct the curves f cumulative demand and cumulative cunt in the withut TMA scenari fr this perid. These curves can be seen in Fig. 2. Figure 2. Cumulative Demand and Cunt: Withut TMA Scenari The area between the tw curves, r the study perid delay in flight-hurs, is equal t 225.9 flight-hurs. B. Delay Calcuatin with TMA T simulate and islate the capacity effect and the variance f capacity effect f TMA, cumulative curves were cnstructed fr the tw scenaris. The estimated parameters f the capacity functin frm (1) were used t calculate the new capacity. The parameters f the best fit mdels are in Table V. 131

TABLE V. CAPACITY ESTIMATION EQUATION PARAMETERS Capacity Mean Capacity Variance 0 2 0 VMC, 1 26.164 1.720-0.201 0.000 5.994-1.267 VMC, 2 21.696 0.302 0.114 0.112 6.333-1.892 MVMC, 1 28.151 0.822-0.560-0.021 6.028-1.554 2) Simulatin f TMA Variance f Capacity Effect T simulate the variance f capacity effect, the capacity effect alng with the variance f capacity effect was calculated. The same methd was used as fr the TMA capacity effect nly scenari. Capacity was assumed t be nrmally distributed with mean µ C and variance 2 + 0, where 0 is the assciated value fr each MC and runway cnfiguratin frm Table V. The cumulative peratinal cunt fr the TMA capacity and capacity variance scenari can be seen in Fig. 4. MVMC, 2 25.949-1.318-0.126-0.054 5.740-1.784 The fllwing sectins describe hw the capacity effect and the variance f capacity effect were determined. 1) Simulatin f TMA Capacity Effect T islate the effect n capacity f TMA, capacity was calculated as a functin f the parameters in Table V depending n the MC and runway cnfiguratin. Capacity in each perid was assumed t be nrmally distributed with mean µ C = 0+ 0 A(t)+ 0 W(t)+ 0 D 0 (t) (2) and variance 2 + 0, where 0 = 0. Capacities fr each quarter hur perid in the study perid were then drawn frm this distributin. Next, as in (1), the peratinal cunt was calculated as the minimum f the capacity and the peratinal demand. The unserved peratins in any perid were added t the peratinal demand f the next perid. The simulated cumulative peratinal cunt curve represents the peratinal cunt that wuld have been achieved if TMA was in use during the entire study perid, but nly the capacity effect f TMA was realized. The cumulative cunt f peratins with the TMA capacity effect is shwn belw in Fig. 3, alng with the cumulative cunt withut TMA and the cumulative demand. Figure 3. Cumulative Demand and Cunt: TMA Capacity Effect Only Scenari The delay calculated fr the TMA capacity effect nly scenari was 147.7 flight-hurs which is a delay savings f 78.1 flight-hurs ver the scenari when TMA is never in use. Figure 4. Cumulative Demand and Cunt: TMA Capacity Variance Effect The delay fr the TMA capacity and capacity variance effect was 121.0 flight-hurs, which is a savings f 26.7 flighthurs as cmpared with the TMA capacity effect nly scenari and an verall delay savings f 104.9 flight-hurs. Using the same methd fr the entire mnth f January 2006, if TMA had been in use 100% f the time, TMA wuld have saved 750 flight-hurs f delay fr arrivals cmpared t the withut TMA scenari. Of these 750 flight-hurs, 500 flight-hurs f savings were due t capacity effect, and 250 flight-hurs f savings were due t variance effect. This finding generalizes t a savings in delay f 9,000 flight-hurs per year and abut 10 secnds per flight. VI. CONCLUSIONS This study fund that the use f TMA fr releasing internal departures appears t have decreased capacity variance and in sme cases increased capacity mean. Using the mdel results, it was fund that increased use f TMA culd lead t decreased delay f abut 10 secnds per flight. Additinally, we have furthered the use f censred regressin applied t ASPM data as an evaluatin methd fr ATM tls. In particular, we have shwn hw this methd can be used t investigate the effect f new tls n the variance f capacity as well as its mean. In ur particular case, we find that TMA use, even thugh it was restricted t releasing internal departures, had a measurable impact n arrival capacity variance at ORD. Further study is necessary t assess the impact f TMA when it is used fr time based metering. Time based metering went int effect in June 2007 at ZAU, and culd decrease the variance in capacity by allwing cntrllers t effectively 132

manage capacity especially during high traffic perids. Understanding the impact f TMA n capacity and capacity variance due t time based metering, and cmparing these findings with thse in this study, wuld prvide insight int the benefits f TMA when it is emplyed fr its full range f uses rather than used nly fr mre limited, adapted purpsed. REFERENCES [1] United States General Accunting Office. (2001). Natinal Airspace System: Free Flight tls shw prmise, but implementatin challenges remain. Available at: http://www.ga.gv/new.items/d01932.pdf Accessed n January 4, 2007. [2] Hansen, M., Mukherjee, A., Knrr, D., and Hwell, D. Effect f T-TMA n capacity and delay at Ls Angeles Internatinal Airprt, Transprtatin Research Recrd 1788, 2002, pp. 43-48. [3] Hansen, M. (2004). Pst-deplyment analysis f capacity and delay impacts f an airprt enhancement: Case f a new runway at Detrit. Air Traffic Cntrl Quarterly, 12 (4). [4] Blic, T. and Hansen, M. User Request Evaluatin Tl (URET) adptin and adaptatin; Three center case study, in Prceedings f 4th USA/Eurpe Air Traffic Management R&D Seminar, 2005. [5] Free Flight Prgram, Perfrmance metrics results t date. June 2001 Reprt. Received frm FAA. [6] Free Flight Prgram, Perfrmance metrics results t date. June 2003 Reprt. Received frm FAA. [7] Michalak, D. Traffic Management Lead at Chicag Center. Persnal Interview. December 1, 2006. [8] de Neufville, R. and Odni, A. 2003 Airprt systems planning, design and management. McGraw-Hill. [9] Federal Aviatin Administratin. Airprt capacity benchmark reprt 2004: Chicag O'Hare. Available at: http://www.faa.gv/events/benchmarks/. Accessed n January 4, 2007. 133