2. FAA Arbitrarily and Capriciously Rejected NE-1a Based on Computer Estimates of Capacity and Delay
As one looks more closely at FAA’s bases for determining that NE-1a was not a reasonable alternative, it becomes apparent that the decision was erroneous. Ultimately, FAA relies on the results of computer-generated capacity and delay to justify its rejection of NE-1a. Specifically, FAA relied on two computer models to project capacity and delay levels, the Runway Capacity Model to estimate capacity and the Annual Delay Model to estimate delay (collectively, "RCAD"). In addition, FAA relied on a more sophisticated computer model, called SIMMOD, to verify W-1W’s performance.
Careful analysis of the results of these analyses demonstrates that the computer analyses do not support FAA’s decision to reject NE-1a as unreasonable. Moreover, comparing the initial RCAD results to more recent RCAD results demonstrates that FAA did not maintain a consistent basis for evaluating alternatives. Close scrutiny reveals FAA’s decisionmaking as inconsistent, incomplete and indisputably arbitrary and capricious.
St. Louis’s original RCAD analysis projected that NE-1a would have an average delay per operation of 10.44 minutes plus 5.5 minutes of average taxi-time per operation. B.App. 2:380. The same analysis projected that W-1W would have an average delay per operation of 4.7 minutes plus 6.9 minutes of average taxi-time per operation. Id. Looking behind the average delay numbers, however, reveals that there was no rational reason to exclude NE-1 but include W-1W.
The average delay numbers are an average of the projected delay per operation for each of the five weather conditions (VFR-1, VFR-2, VFR-3, IFR-1 and IFR-2). Because VFR-1 and VFR-2 (good weather) conditions prevail for over 90% of the year, any averaging of delays for each weather condition will tend to give less weight to delays in poor weather conditions, because they prevail for less than 10% of the time. B.App. 1:174. As a result, poor delay performance in poor weather conditions could mean that an alternative may not provide acceptable capacity and delay performance even though the average delay number appears acceptable. That is critical here because additional runway capacity, and in particular the claimed need for dual independent simultaneous IFR approach capability, is needed primarily to address projected delays in poor weather conditions. B.App. 4:864.
Here, reliance on average delay per operation performance masks the fact that during the poor weather conditions W-1W is projected to have levels of delay comparable to the projected delays for NE-1a. As shown on the table at B.App. 6:2162, the RCAD analysis indicates that NE-1a, W-1W and S-1 all provide identical delay levels in good (VFR-1 and VFR-2) weather conditions but different delay levels in poor weather conditions (VFR-3, IFR-1 and IFR-2).
The significance of this is two-fold. First, it is clear that NE-1a could perform as well as W-1W in good weather conditions, which prevail over 90% of the year. Second, although there are significant differences between W-1W and NE-1a in VFR-3 and IFR-2 in east flow, the delay levels for W-1W in IFR-1 and IFR-2 in west flow are so extraordinarily high that it is unclear whether W-1W could provide adequate capacity/delay performance in IFR conditions. For example, the delay levels for S-1 remain well below 10 minutes in IFR-1 and IFR-2 in west flow, while delay levels for W-1W skyrocket to 59 and 76 minutes per operation.
The only significant differences between NE-1a and W-1W are in VFR-3 and IFR-2 in east flow. W-1W and NE-1a are equal in VFR-1 and VFR-2 and W-1W’s delays in IFR-1 and IFR-2 in west flow are comparable to NE-1a’s delays. Critically, FAA makes no effort to explain why NE-1a’s delay levels in VFR-3 (which occurs 4.2% of the year) and IFR-2 in east flow (which occurs 0.24% of the year) are unacceptable, but why comparably high delay levels for W-1W in IFR-1 (which occurs 5% of the year) and IFR-2 in west flow (which occurs 0.36% of the year) are acceptable. B.App. 1:174. Nothing in the Record justifies that facially arbitrary decision. Although the differences in delay levels may be a factor in deciding which alternative is preferable, the difference is not great enough to justify labeling NE-1a "unreasonable."
The arbitrary nature of the rejection of NE-1a is underscored by the SIMMOD analysis, which indicated that the projected delay levels for W-1W in IFR-1 conditions fell from 76.2 minutes per operation using RCAD to 10.2 minutes per operation using SIMMOD. B.App. 1:225; 6:2162. Both W-1W and NE-1a had high delay levels in certain poor weather conditions. Given the dramatic improvement in projected delay levels for W-1W, it was arbitrary and capricious not to also analyze NE-1a with the more refined SIMMOD model.
FAA cannot have it both ways. Either the RCAD results control, in which case it is arbitrary and capricious to determine that W-1W is acceptable but NE-1a is unacceptable. On the other hand, if the SIMMOD results control, the decision to reject NE-1a is also arbitrary and capricious because no SIMMOD analysis was done for NE-1a. There is nothing in the Record to explain why a SIMMOD analysis of NE-1a, which had RCAD results comparable to W-1W, would not demonstrate that W-1W and NE-1a would have comparable delays.
The results of these analyses, as reported in the ROD, show that the changes in assumptions caused projected delays to increase for all alternatives tested, but that projected delays for W-1W increased more than the projected delays for NE-1a. B.App. 6:2163. The difference in projected delay (including taxi time) between W-1W and NE-1a dropped from 4.3 minutes in the MPSS to between 1.8 and 3.64 minutes in the June 1998 analysis. Id. The results of the June 1998 analysis are significant in two ways.
First, the projected delay per operation (including taxi time) for W-1W increased to between 14.48 and 16.32 minutes per operation. Id. Yet in the FEIS, FAA determined that 15.94 minutes of delay per operation (including taxi time) made NE-1a unable to meet purpose and need. There is no rational basis, and FAA does not offer one, to explain why 15.94 minutes of delay per operation is unacceptable for NE-1a while 14 to 16 minutes of delay per operation is acceptable for W-1W.
Second, although these projected delay levels still favor W-1W, the difference is not so great as to make a decisive difference. For example, in the FEIS the difference in projected delay per operation between W-1W and S-1 was 3.3 minutes in favor of S-1, yet W-1W was selected. B.App. 2:380. The June 1998 data shows that the difference in delay between W-1W and NE-1a is between 1.8 and 3.6 minutes, yet NE-1a was "unreasonable." A.R. 89:2; B.App. 6:2163. Nothing in the Record explains why W-1W is acceptable with 3.3 minutes more delay than S-1, but NE-1a is unacceptable with 2 - 3 minutes more delay than W-1W.
FAA’s attempts to justify its exclusion of NE-1a in the face of this evidence merely accentuate the arbitrariness of FAA’s decision that NE-1a was an unreasonable alternative. FAA concluded that because W-1W "increases capacity and reduces delays better than NE-1a and the No-Action Alternative" W-1W was preferable. B.App. 5:1610. Again, FAA ignores its duty under NEPA to compare W-1W to all reasonable alternatives, not just perfect alternatives. Far from validating FAA’s decision to eliminate NE-1a from detailed consideration, the relatively small difference in projected delay levels between NE-1a and W-1W demonstrates conclusively that NE-1a is a reasonable alternative to W-1W that should have been carried forward for detailed analysis, as Bridgeton and others had repeatedly told FAA.
E. THE DECISION TO ACCEPT W-1W AS REASONABLE WAS ARBITRARY AND CAPRICIOUS
1. In Practice W-1W Will Not Permit Independent Simultaneous IFR Operations
The selection of W-1W as a reasonable alternative was arbitrary and capricious under the standards FAA set for itself, because the manner in which W-1W will be operated will not permit the independent simultaneous IFR operation capability FAA claims is so critical. Because independent simultaneous IFR operation capability was applied strictly to eliminate NE-1a at Tier 1, FAA should have eliminated W-1W at Tier 1 as well.
Independent simultaneous IFR operations are permitted when the centerlines of runways are separated by at least 4500 feet, or 3400 feet if a Precision Runway Monitor is used, as is the plan at Lambert. B.App. 5:1597. Although the centerline of W-1W will be 4300 feet away from the centerline of the existing northern runway (designated 30R-12L), making independent simultaneous IFR operations theoretically possible, St. Louis and FAA already have committed to operate the post-W-1W airport in a manner that will not permit independent simultaneous IFR operations.
FAA defines independent simultaneous IFR operations as "parallel operations when aircraft on one runway are spaced independently (at random) of those on the adjacent runway . . . ." In contrast, dependent simultaneous parallel operations are defined as "operations when aircraft spacing is dependent (determined) on the position of aircraft on the adjacent runway." B.App. 6:2166.
Operations at Lambert after the construction of W-1W will be dependent simultaneous parallel operations. Arrivals will use W-1W and 30R, the two "outboard" runways while most departures will use the existing southern runway (designated 30L-12R), which will become the center runway. B.App. 1:189. Because neither W-1W nor 30R are far enough from 30L to operate independently from 30L, operations on both W-1W and 30R are dependent on operations on 30L. In other words, there is a "double dependency" created because departures on 30R are dependent on arrivals to W-1W and on arrivals to 30R. B.App. 6:1680.
Because of this double dependency, the two arrivals must be coordinated in order to ensure that both arrivals are spaced far enough from the departure on 30L to permit the departure to be released. Id. As a result of this "pairing," spacing of aircraft approaching W-1W will be determined by the position of aircraft approaching 30R, or vice-versa. Based on FAA’s own definition of independent and dependent simultaneous parallel arrivals, therefore, arrivals on W-1W and 30R will be dependent on each other.
Bridgeton, NATCA and ALPA raised this point repeatedly throughout the EIS process, and each time FAA denied that the operations were dependent. E.g., B.App. 6:1816; 1858-1872. Before the ROD was issued, however, St. Louis, in response to a comment from Bridgeton, explained why it considered such operations independent in terms that make it plain the operations are dependent. "During mixed-operations periods (e.g., 50% arrivals and 50% departures) in northwest flow, departures are dependent on arrivals, but arrivals remain independent. Indeed the "pairing" method described by Bridgeton is possible only because W-1W does provide simultaneous independent arrival capability." B.App. 6:2150.
First, to insist that the departure is dependent on arrivals but that arrivals remain independent is to ignore reality. In order to achieve "pairing", aircraft approaching W-1W will have to be placed in the arrival stream based on the position of aircraft in the 30R arrival stream. That, by definition, is a dependent operation. The fact that the cause of that dependency is the departure on 30L (rather than inadequate separation between runways) does not make it any less of a dependency. Second, to say that the otherwise independent nature of the two runways "enables" pairing is another way of admitting that the act of pairing makes the two arrival streams dependent. Put another way, if arrivals on W-1W and 30R were dependent because of inadequate spacing between the two runways, St. Louis would not have to "pair" the arrivals because they would already be "paired" through normal dependent operation procedures. Here, pairing is necessary because the operations would be independent but for the double dependency and the resulting need to "pair" arrivals.
FAA also relies on the ATC Manual to argue that the arrivals on W-1W and 30R will be independent. B.App. 6:1875. FAA makes the formalistic argument that because the ATC Manual authorizes independent simultaneous IFR operations on runways with specified minimum spacing requirements, and because W-1W and 30R meet those spacing criterion, operations on the two runways are by definition independent.
This argument is facially illogical. The ATC Manual authorizes independent simultaneous IFR operations in certain circumstances, but it does not define all circumstances that may make otherwise independent operations dependent. One must look to the definition of independent operations quoted above to determine if a given operation is actually independent. It is clear here that the pairing procedure converts otherwise independent operations into dependent operations.
Under FAA’s own statement of purpose and need, W-1W should have been rejected at Tier 1, because, like NE-1a, it will not permit independent simultaneous IFR operations. If FAA is using independent simultaneous IFR operation capability as a basis for rejecting alternatives, it was arbitrary and capricious to reject NE-1a but accept W-1W.
2. It was Arbitrary and Capricious to Select W-1W Without Performing
a Real-Time Simulation
It was also arbitrary and capricious for FAA to approve W-1W without determining whether W-1W’s unique configuration would in fact operate as FAA assumes. B.App. 6:1961. Specifically, it was arbitrary and capricious for FAA to refuse to conduct a "real-time simulation" test to determine (1) whether W-1W would be practicable given that FAA previously had concluded that "staggered" runways like W-1W were not practicable and (2) whether the operational assumptions upon which the RCAD and SIMMOD capacity and delay analyses depended were valid.
FAA acknowledged, when rejecting a staggered runway at Seattle, that staggers require "greater coordination" between dependent aircraft and "produce inefficiencies." B.App. 6:1936. Indeed, the magnitude of those complexities led FAA to conclude that "the stagger option is not practicable." Id. Yet at Lambert, FAA embraced a stagger without any empirical analysis demonstrating practicability. In fact, FAA justifies this inconsistency by stating that it does not know if the concerns at Seattle apply to Lambert. B.App. 6:1898. It is inherently arbitrary and capricious for FAA to reject a staggered runway option at one airport as "impracticable," yet approve a stagger at another airport with no empirical study to show that the stagger is now practicable. This error is only magnified by relying on the uncertainties of comparing the two cases to justify not conducting the empirical study necessary to resolve those uncertainties.
Second, the results of the RCAD and SIMMOD analyses, like any computer model, are only valid if the assumptions underlying the input data are valid. Here, those assumptions include how the difficult and uncertain pairing procedure actually will work. B.App. 6:1963. Although the pairing procedure is theoretically possible, it has never been performed under these circumstances and the controllers who must actually perform the procedure have serious doubts that it can work at all, and if so at what penalty to capacity and delay performance. Id. Specifically, it may be necessary to increase the spacing between aircraft, or slow aircraft speeds in the arrival streams, in order to release the departure on the center runway. Id.
These factors effectively define the capacity and delay levels because greater spacing between aircraft, or slower aircraft speeds, mean lower capacity (fewer aircraft will land in a given period of time) and higher delays. Until the precise spacing, speed and other parameters of the pairing procedure are defined, it is impossible to calculate W-1W’s capacity and delay.
The consequences of this lack of empirical data is apparent in the SIMMOD simulation upon which FAA relies. In the simulation, allowance is made for the pairing procedure by the crude expedient of having the aircraft come to a dead stop in mid-air for approximately one minute to account for the time FAA assumes it would take the controllers to achieve pairing. A.R. 89:2. Obviously, that is not an accurate model of reality. Moreover, the time allowed to achieve the pairing is based on an estimate, not on empirical data. In reality, it may take more time, or prove to be infeasible.
It is primarily to resolve these two issues, as well as a large number of related safety and capacity issues, that the Lambert air traffic controllers who must implement the pairing procedures and make W-1W work safely and efficiently and the TWA pilots who must use W-1W hundreds of time daily have insisted that FAA conduct a real-time simulation before approving W-1W. E.g. 5:1599; 6:1909. Unlike a computer simulation, which calculates the consequences of a given set of assumptions, a real-time simulation provides an empirical basis for a given assumption. In a real-time simulation, controllers and pilots "act out" in real-time airfield operations using computers to model the airspace, air traffic patterns, airfield and traffic demands. B.App. 5:1599.
Such an empirical study is the only way to determine the spacing and time requirements of implementing the novel pairing procedure. And until those spacing and time requirements are defined precisely, it is impossible to determine with any degree of precision the capacity and delay capabilities of W-1W, or indeed whether W-1W is practicable.
FAA’s excuses for not performing a real-time simulation only underscore the need to perform a real-time simulation. First, FAA claims that real-time simulations performed for closely spaced independent arrivals validate the pairing procedure. B.App. 5:1599-1600. Those studies are inapposite for the simple reason that none of them simulate the pairing procedure at issue here and there was no double dependency in those simulations. B.App. 6:1993-2147.
Second, FAA claims that a real-time simulation is unnecessary because W-1W will utilize existing and authorized procedures. FAA contends that a real-time simulation is necessary only to test new procedures; it is not a capacity and delay calculator. B.App. 5:1600. This argument misses the point.
Although independent simultaneous IFR approaches are a recognized procedure, using the pairing procedure to resolve a double arrival-departure dependency is not a recognized procedure. B.App. 6:1963. Moreover, the capacity and delay calculators (SIMMOD and RCAD) require data that can only be obtained from experience, either real-world experience, which is unavailable, or a real-time simulation. The real-time simulation is a necessary predicate for the capacity and delay analysis, not a substitute as FAA argues.
In the absence of a real-time simulation, the capacity and delay results upon which FAA’s decision to select W-1W depends are tantamount to speculation. Under any understanding of arbitrary and capricious decisionmaking, to allow such a large and expensive project as W-1W to proceed, knowing that it will destroy Bridgeton and impact thousands of people, when there is a real likelihood that W-1W will not provide the benefits desired and when a test is at hand that would resolve that issue, is the very definition of arbitrary and capricious.
F. CONCLUSION
FAA’s alternatives analysis is fatally flawed because it arbitrarily and capriciously rejected reasonable alternatives without the complete environmental analysis required by NEPA. Specifically, FAA
Rejected N-1 and NE-1 as not constructible, with no evidence in the Record to support that conclusion, even though St. Louis had concluded that they were constructible.
Rejected NE-1a because the 1995 Capacity and Delay Model analysis predicted high annual average delays but accepted W-1W even though the 1998 Capacity and Delay Model analysis predicted that W-1W would have virtually identical delay levels.
Rejected NE-1a, despite the fact that it was projected to provide delay performance equal to W-1W for over 90% of the year.
Rejected NE-1a because it allegedly could not provide "sufficient" capacity to reduce delays to "acceptable" levels even though FAA never established how much capacity was needed and how much delay was acceptable.
Rejected NE-1a because it was projected to have unacceptably high delay levels in poor weather conditions but accepted W-1W despite projections of comparably high levels of delay in poor weather conditions.
Rejected NE-1a based on high RCAD delay projections in poor weather but accepted W-1W with similarly high RCAD projections in poor weather based on a SIMMOD analysis of W-1W even though no SIMMOD analysis was performed for NE-1a.
Rejected NE-1a because it would not provide independent simultaneous IFR operation capability, but accepted W-1W even though W-1W will not utilize independent simultaneous IFR operations.
The arbitrary and capricious decision to reject the Northern alternatives as unreasonable are matched by the arbitrary and capricious decision to accept W-1W:
Accepted W-1W based on RCAD and SIMMOD computer projections even though FAA lacked empirical data to define the operating assumptions upon which those computer analyses depended.
Accepted W-1W without performing a real-time simulation to determine if staggered runways rendered W-1W :impracticable" even though FAA had determined that staggered runways had rendered a similar configuration at another airport impracticable.
The net result is an EIS that exploits every opportunity to reject reasonable and environmentally preferable alternatives to W-1W, but that bends over backwards to describe W-1W as "reasonable." This is the very definition of arbitrary and capricious decisionmaking. Because the EIS violates NEPA, the Record of Decision must be vacated.
IV. FAA HAS VIOLATED SECTION 4(f)
Although FAA exercises some discretion in approving projects, Congress has severely limited that discretion when the proposed project would "use" a 4(f) resource. Congress has decided that resources protected under §4(f) should be given "paramount importance." Overton Park, 401 U.S. at 412-13. Despite §4(f)’s express language and the Supreme Court’s admonition, FAA has violated both the spirit and the express requirements of §4(f).
A. FAA’s Section 4(f) Analysis Is Unreasonable
Section 4(f) bars FAA from using a 4(f) resource for a project absent a determination that there is no prudent and feasible avoidance alternative and requires FAA to implement mitigation measures if the use can not be avoided. 49 U.S.C.
§303(c)(1). FAA initially must determine whether the proposed project will "use" a 4(f) resource. 49 U.S.C. §303(c). Notwithstanding the NHPA §106 terminology used in the ROD, it states that W-1W will "use" 4(f) resources. B.App. 5:1558. Specifically, the FEIS states that W-1W will use certain 4(f) parks - Freebourn Park, Cardinal Park, Bridgeton Memorial Park, and a portion of Oak Valley Park - for the construction of the runway and associated mid-field terminal development areas. B.App. 3:617. In addition, the FEIS states that W-1W will constructively use four parks - O’Connor Park, Berry Hill Golf Course, Oak Valley Park, and Carrollton Buffer Zone. Id. The FEIS further states that W-1W "may result in the physical destruction of all or part of the following eligible historic buildings - the Bridgeton Inn, the Airport News Building, the Emmanuel Blum House, the Blum Store, the Village á Robert Cemetery and the DeHatre House. B.App. 3:621. In sum, FAA has determined that W-1W will use six historic resources and seven parks.