Effect of Adding Gateways on Air Traffic Flows Over the North Pacific Ocean

The North Pacific (NOPAC) oceanic airspace has a high traffic demand between Asia and North America, and its capacity is limited by the non-radar air traffic control environment. Improvements in satellite-based navigation and datalink services have resulted in a reduction in the aircraft minimum separation standard from 185.2 km (100 NM) to 42.6 km (23 NM); moreover, with the adoption of more capable systems, airspace design and flight operations are being modified to take advantage of the lower minimum separation. The NOPAC airspace is currently being restructured based on the 42.6 km (23 NM) standard. However, the “gateway” points between the oceanic and radar-controlled airspaces continue to be at 111.1 km (60 NM) intervals in the Japanese airspace. In this study, we seek to design a gateway configuration that improves efficiency for both individual flights and the oceanic airspace. Concretely, we propose additional gateways to exploit the 42.6 km (23 NM) standard. We conducted a fast-time simulation study to examine the effect of additional gateways using three models of the adjacent NOPAC airspace, viz., a baseline configuration with three parallel oceanic airways at 92.6 km (50 NM) intervals and the existing gateways, and two configurations with our proposed gateway alignment: four parallel airways at 46.3 km (25 NM) intervals reflecting current NOPAC restructuring plans, and a proposed free route airspace (FRA) model without fixed airways. The results showed that increased flight-planning flexibility due to the addition of gateways and the removal of fixed airway constraints allowed for more optimal flight routes with lower fuel consumption and flight time; thus, the FRA model provided the highest average fuel benefit for individual flight routes. However, flights choosing optimal routes tended to favor a small number of gateways, thereby increasing the concentration of westbound traffic in NOPAC. Nonetheless, the overall fuel benefit derived from the selected routes outweighed the penalties incurred from the necessity of cruising at lower-than-planned altitudes for deconfliction.

SECTION I. Introduction

The demand for air transport between East Asia and North America is considerably high, with a large number of flights operating over the North Pacific Ocean. In addition to passenger transportation, there is a high demand for cargo transportation between Asian cargo hubs and Anchorage International Airport. Although passenger traffic declined in 2020 because of the COVID-19 pandemic, the reduction in cargo traffic was not as significant, partly because of the demand for COVID-related supplies [1], [2], [3], [4]. Nonetheless, traffic over the North Pacific has recovered and growth has resumed, with the traffic in late 2023 exceeding the 2019 levels [5]. Therefore, it is necessary to expand the capacity of the airspace and to improve the efficiency of both individual flight operations and the airspace as a whole to reduce operational cost and alleviate the effect of the increasing air traffic on climate change.

This study considers the traffic flow between Asia and North America. Fig. 1 shows the airspace through which the traffic flows considered in this study operate: Fig. 1(a) shows a Federal Aviation Administration en route chart [6], and Fig. 1(b) shows a simplified diagram indicating the airspaces relevant to this study. The traffic flows pass through the following oceanic flight information regions (FIRs), shown in light blue in Fig. 1(b): Fukuoka FIR oceanic control area, Oakland Oceanic FIR, and Anchorage Oceanic FIR. Despite the vast area of North Pacific oceanic airspace, traffic becomes concentrated due to the effects of winds aloft, peaks in the daily traffic flow, airspace design, and for geopolitical reasons. The winds aloft over the North Pacific are dominated by the westerly jet stream, and eastbound flights plan their routes to take advantage of the jet stream tailwind while westbound flights avoid the headwind area. Consequently, operators plan similar oceanic routes that change in pattern from day to day. Flights between geographical regions tend to depart and arrive at approximately the same time of the day because of passenger convenience, airport restrictions, and flight connections, resulting in peaks in traffic flows. Furthermore, to assist the air traffic control (ATC) in managing the flow and establishing separation, flights must pass between radar-controlled airspace and oceanic airspace via defined waypoints (referred to here as “gateways,” indicated by red diamonds in Fig. 1(b)). In contrast to the well-distributed gateways along the west coast of North America, gateways in the Japanese airspace are concentrated in a relatively small area. Finally, some operators may choose to avoid certain airspaces for reasons such as service charges or geopolitics. Consequently, the traffic flow exhibits a tendency to concentrate at specific areas, times, and gateways in the North Pacific airspace.

FIGURE 1.

Target airspaces of this study.

The increasing adoption of satellite-based navigation and datalink services for operations over the North Pacific has helped reduce the separation intervals between aircraft [7]. The NOPAC (a term taken from “North Pacific” and used as a proper noun to describe both the North Pacific airspace and the fixed airways within it) was established with five parallel airways between Alaska and Japan at approximately 92.6 km (50 NM) intervals in the 1980s, when most aircraft used HF voice communication and inertial navigation. With the retirement of these older fleets, satellite-based navigation and datalink equipage in the airspace has increased to more than 90% [8]; moreover, the introduction of performance-based communication and surveillance (PBCS) [9] has allowed a phased restructuring of the NOPAC from 2020 to increase capacity and flight planning flexibility [10]. However, the restructured NOPAC retains the current oceanic gateway configuration.

In previous studies, we explored the effect of completely removing the NOPAC fixed airways and allowing operators to plan flex tracks throughout the NOPAC airspace, creating the so-called free-route airspace (FRA) environment. Fast-time simulations confirmed reductions in fuel consumption and flight time. However, the airspace complexity, as indicated by the potential loss of separation (PLOS), increased with 55.6 km (30 NM) PBCS separations; this was offset by further reducing the separation to 27.8 km (15 NM), assuming a technology such as space-based automatic dependent surveillance–broadcast (ADS-B) [11], [12], [13].

In the aforementioned NOPAC FRA study, existing gateways were retained to elucidate the effects of removing the fixed airways. Although FRA is a long-term goal, medium-term improvements can be achieved using the 42.6 km (23 NM) separation standard to create additional gateways, smoothing the transition between radar-controlled and oceanic airspaces, and paving the way towards seamless flight operations. In this paper, we report the results of a fast-time simulation study that examines the effect of additional gateways on the restructured NOPAC and the NOPAC FRA. The trade-off between the fuel benefits due to better flight planning flexibility and the increase in traffic complexity and the concomitant decrease in flight efficiency is considered. To the best of our knowledge, the degree of the effect of PLOS on flight efficiency has not yet been quantified.

The remainder of this paper is organized as follows. Section II describes the proposed gateway design and simulation experiments. Section III reports the experimental results and discussion. Finally, Section IV presents the conclusions.

SECTION II. Airspace Design and Simulation Experiments

This section describes the concept of the proposed airspace design and the methodology of the fast-time simulation experiment.

A. Airspace Design

The airspace over the North Pacific Ocean is a non-radar control environment, and the lack of real-time communication and surveillance means that the required separation between two aircraft (separation standard) is greater than that in radar-controlled airspace (generally 9.3 km (5 NM) horizontally for aircraft less than 305 m (1,000 ft) apart vertically in radar airspace [14]). To assist ATC to manage the traffic flow between the oceanic and radar-controlled airspaces and establish lateral separation, flights are required to pass through gateways on the boundary between radar-controlled airspace and oceanic airspace. The largest oceanic horizontal (lateral) separation standard, 185.2 km (100 NM), is established for aircraft that use dead-reckoning inertial navigation and report position periodically via HF voice communication for surveillance. Satellite datalink-based surveillance, communication, and satellite navigation are now commonly used. When the PBCS performance and monitoring criteria are satisfied, the current international civil aviation organization (ICAO) regulations state that the shortest horizontal (lateral) control interval must be 42.6 km (23 NM) [9], [14].

This study proposes a gateway design between radar-controlled airspace and oceanic airspace in the Fukuoka FIR, wherein gateways are added approximately half-way between existing gateways; thus, the current gateway spacing of approximately 111.1 km (60 NM) is reduced to approximately 55.6 km (30 NM). This gateway spacing is intended to provide the minimum applicable separation between two aircraft passing simultaneously through adjacent gateways at the same altitude. We evaluated the impact of the proposed gateway design and various NOPAC configurations on traffic flows through simulation experiments using three airspace models, M1, M2, and M3 (shown in Fig. 2). In the figure, the gateways are indicated by diamonds (♦) and NOPAC fixed airways are represented by black lines with arrows indicating their directions. The gray line represents the boundary between the Fukuoka FIR oceanic and radar-controlled airspaces and continues to the northeast along the boundary with Russian Federation airspace. The baseline for the comparison was M1 (Fig. 2(a)), which is based on the NOPAC airspace and gateway configuration as of February 2023 [10]. Three one-way airways were set at intervals of ~92.6 km (50 NM), approximately parallel to the Russian Federation FIR boundary. The two northmost airways are westbound-only, and the southmost airways are eastbound-only. The NOPAC restructuring model (M2), shown in Fig. 2(b), combines our proposed ~55.6 km (30 NM)-spaced gateway design with the NOPAC fixed airways that are planned in the final phase of the current NOPAC restructuring [10]. Four airways are set at intervals of ~46.3 km (25 NM), with the two northmost airways dedicated to westbound traffic and the remainder dedicated to eastbound traffic. With reference to actual operations, operational rules were established for the M1 and M2 models to prohibit flight plan routes from merging with, branching from, or crossing the NOPAC fixed airways. The M3 model, shown in Fig. 2(c), combines the NOPAC FRA from our previous study (with no NOPAC airways) [11], [12] and the proposed gateway design, with the aim of better utilizing the PBCS separation standard.

FIGURE 2.

Airspace models used in the simulations.

B. Evaluation Metrics

Fast-time simulation is often used to evaluate changes to airspace design and air traffic control and management procedures, including changes pertaining to oceanic airspace [15], [16]. In addition to the impact on each flight operation (flight distance, flight time, and fuel burn), airspace metrics, such as capacity and efficiency, must be considered in the evaluation. Airspace capacity is related to air traffic controller workload, which cannot be measured via fast-time simulation. Controller workload is incurred by each potential conflict, involving detection, resolution, and monitoring. Accordingly, we calculated the four-dimensional flight trajectories from flight plans and used the PLOS between those trajectories as a proxy metric for airspace capacity. PLOS was used as an indicator when FRA was introduced in Europe [17]. A common conflict-resolution technique in oceanic airspace is to change the cruising altitude of one of a conflicting pair of aircraft; therefore, in this study, the occurrence of a PLOS reflects a deviation from the planned cruise altitude. The impact of this possibility of conflict resolution by ATC on the flights involved in each PLOS event (to be precise, changes in fuel burn) was estimated as described in Section II-C6.

C. Simulation Experiment

The effect of the airspace models was examined via a fast-time simulation experiment using a realistic traffic demand scenario under different operational conditions to improve the reliability of the results. We first explain the overall design of the experiment and subsequently, detail each step. A flowchart of the experiment is shown in Fig. 3. The independent variables are summarized in Table 1. The dependent variables were the individual flight performance metrics (flight duration, flight distance, and fuel consumption) and PLOS.

TABLE 1 Independent Variables in the Simulation Experiment

FIGURE 3.

Flowchart of the simulation experiment.

The traffic demand scenario comprises a set of scheduled air transportation services between Asia and North America based on historical flight plan data, including city pairs, aircraft types, and departure times. The initial weight of a flight affects its planned vertical profile (initial cruising altitude and step-climbs); hence, three variants of the scenario with different initial weights for each flight, S₁, S₂, and S₃, were created.

The flight plan route of each service depends not only on the airspace configuration but also on the winds aloft, which vary based on the seasons as well as on a day-to-day basis. A set of 11 wind datasets, W₁, …, W₁₁ were selected so that the experimental results represented the wind conditions over a year [18]. A realistic flight plan route for each flight service (a minimum fuel consumption route considering the winds aloft) was calculated using a wind-optimal trajectory generator, and the performance (flight distance, flight duration, and fuel consumption) for each flight was calculated. Sets of flight plans FP_S,M,Wwere created for the weight variant, airspace model, and wind condition of each scenario.

Each flight plan set was then simulated using the AirTOp fast-time simulator [19] to generate a corresponding set of four-dimensional trajectories T_S,M,W,D that were used to calculate PLOS. PLOS is detected from the relative positions of pairs of aircraft at each time instant, and hence, it is sensitive to the aircraft departure times. Therefore, the simulations were repeated with four different flight departure time conditions D₁, …, D₄, which corresponded to random variations to the departure time.

Finally, the aggregated results of the per-flight benefit, benefit over airspace, and PLOS were compared to evaluate the performance differences between the airspace models.

1) Traffic Scenarios

The traffic demand scenario comprised 749 flight services that flew through the target airspace and was based on historical plans for scheduled air transportation flights. The parameters were aircraft type, departure and destination airports, departure time (derived from the estimated off-block time), and initial cruise altitude. Flights that operated in the Fukuoka FIR during the UTC day (24-hour period from 00:00–24:00 UTC) in 2019 with the highest number of oceanic flights in the Fukuoka FIR (a day in August) were selected. Some of these flights departed on the previous day and some others landed the following day. Therefore, to obtain complete (end-to-end) trajectory data, the fast-time simulations were conducted over a period spanning three days. Fig. 4 shows the histograms of departures per hour in the traffic demand scenario for eastbound (red) and westbound (blue) flights, where day 2 as the target day.

FIGURE 4.

Histograms of departures per hour in the traffic demand scenario for eastbound (red) and westbound (blue) flights.

The traffic scenario comprised 239 combinations of departure and destination airports. Table 2 shows the combinations aggregated by region, and Table 3 shows the top 10 combinations of departure and destination airports (shown by their four-letter ICAO codes) ranked by the number of flights. Hereafter, each region is denoted by two letters as shown below, and the number of airports in each region in the scenario is indicated in parentheses.

<Asian side>

Japan: jp (8), South Korea: kc (2), Others: es (18)

<North America side>

North American continent: na (29), Alaska: pa (1), Hawaii: ph (2)

TABLE 2 Breakdown of Departure/Destination Regions by Scenario

TABLE 3 Breakdown of Departure/Destination Airports by Scenario

2) Wind-Optimal Flight Plan Calculation

Flight plan routes over the Pacific Ocean were calculated based on the forecast winds aloft. The influence of strong winds aloft coupled with long flight distances implies that the ideal route (typically the minimum flight time route or minimum fuel route) may deviate significantly from the shortest distance (great circle) path, as shown in Fig. 5. Eastbound flights tend to follow the jet stream core axis (which is often at around the latitude of the gateways in Japanese airspace) for some distance east of Japan to take advantage of the tailwind, whereas westbound flights tend to avoid the jet stream strong headwind area to the north or south. Flight plans are also constrained by the airspace (waypoint locations, airways and formal planning rules, restricted airspaces). In this study, crossing the NOPAC fixed airways was prohibited to reflect the current operations [20].

FIGURE 5.

Seasonal influence of the jet stream on traffic flows over the north pacific ocean.

The seasonal variations in the jet stream must be considered in the evaluation of the oceanic airspace design. A weather day selection method that uses a clustering technique was developed in our previous study [18]; this technique was used to select winds on the following 11 representative days, labeled as wind conditions W₁, …, W₁₁.

<11days of weather data reflection; yyyy/mm/dd>

2017/01/28, 2017/03/28, 2017/06/05, 2017/06/24,

2017/09/23, 2017/10/17, 2017/10/25, 2017/11/02,

2017/11/12, 2017/12/10, 2017/12/22

These weather conditions were then used to calculate the routes for each flight service on each day. The gridded numerical meteorological forecast published by the Japan Meteorological Agency [21] was used as the data source. Forecast data are published at 6-hour intervals, and the “now-cast” data of each forecast (forecast look-ahead time of 0 hours) were used in the flight route calculation. The atmospheric conditions at arbitrary latitudes, longitudes, altitudes, and times, were determined via linear interpolation between latitude and longitude grid points and pressure levels in the forecasts straddling the target time, then linearly in time between forecasts. Although operators plan routes to avoid forecast phenomena such as convective weather, turbulence, or icing conditions, weather avoidance was not considered in this study because it does not have a systematic effect on the results.

The wind-optimal routes for each flight were calculated using a wind-optimal route calculation tool [22]. Given the start time, cruise altitude, and initial weight as the initial conditions, the tool calculates a minimum-cost route between the start and end points using a node-link graph that represents the possible route segments for a specified aircraft type and weather data. These costs can be fuel, flight time, or operating cost. Aircraft performance calculations were based on the point-mass model, and parameters from the EUROCONTROL BADA3 database were used [23]. The tool also outputs flight time, flight distance, and total fuel consumption.

Flight durations over the North Pacific Ocean typically exceed 6 h and involve 1–3 step-climbs as weight reduction due to fuel burn-off allows flight at higher, more efficient altitudes. At each successive node during the graph search, the tool evaluates whether the aircraft performance at its current weight allows it to execute a step-climb; if so, it searches the graph from that node forward at both the current and higher altitudes, and thereby calculates a wind-optimal route including step climbs [12]. In this study, the tool was configured to calculate the minimum fuel consumption route with 610 m (2,000 ft) step climb increments.

3) Initial Weight Determination

The step-climb profile is sensitive to weight. However, the initial weight parameter has fixed and variable components, not all of which can be estimated. The basic empty weight of the aircraft type is taken from the BADA 3 database, and trip fuel can be estimated from the planned flight time; however, the payload (passengers and cargo) varies from flight to flight. Obtaining the actual initial weight data from operators is difficult because the data are commercially sensitive information. Although methods have been proposed to estimate the takeoff weight from the takeoff and climb trajectories and meteorological conditions [24], such approaches were considered impractical in this study because of the large number of flight samples and analyses required to empirically model the takeoff weight. Instead, we randomly sampled the initial masses from a normal distribution with the aircraft type’s reference mass from the BADA performance table file (PTF) the mean value, and 32% of the difference between the PTF maximum and nominal mass values as the standard deviation. Wind-optimal flight routes were calculated for each set of three randomly-sampled initial weight conditions. Fig. 6 shows the distributions of the initial weights for each scenario weight variant binned at 2,000 kg intervals. The shape of the distribution is similar for all three weight conditions because they are based on the same aircraft type population.

FIGURE 6.

Distribution of initial weights for each weight condition scenario variant.

4) Trajectory Generation

The AirTOp fast-time air traffic simulator generates trajectories in simulated traffic flows based on wind-optimal flight plans. For each set of flight plans FP_S,M,W with a given weight variant, wind condition, and airspace model, the flight from the departure to the destination airport via the waypoints of the flight plan route and step-climb schedule was simulated. The resulting trajectories were output at 1-second intervals to produce a corresponding trajectory set T_S,M,W,D. PLOS events between pairs of trajectories were searched for in the trajectory sets T_S,M,W,D. The wind conditions used for the flight trajectory calculations were identical to those used in planning each flight; that is, the effect of the forecast error was ignored because it would not have a systematic effect on the results.

As mentioned earlier, the PLOS between a pair of flights is sensitive to departure times. Therefore, to ensure that the simulation results do not depend on a single set of departure times, each flight plan was simulated four times with the departure times varying from the scheduled times with variation patterns D₁, …, D₄. To generate the departure time patterns, the planned departure time of each flight was perturbed by an amount within ±1 h drawn at random from a normal distribution with a standard deviation of 20 min.

5) PLOS Extraction

The PLOS between the pairs of trajectories in each traffic scenario generated by the fast-time simulator was calculated. Two PLOS metrics were calculated: PLOS count and PLOS time. The PLOS count denotes the number of trajectory pairs between which a loss of separation occurred, and the PLOS time is the sum of the durations of the loss of separation as shown in Fig. 7.

FIGURE 7.

Representation of PLOS extraction from trajectory data used as an airspace metric.

For simplicity, a single parameter was used for detecting horizontal loss of separation in previous studies [11], [12]; that is, PLOS was considered to occur when another aircraft intruded into a cylindrical virtual protection area centered on the “ownship” with a vertical separation of less than 305 m (1,000 ft). Although this is suitable for radar-controlled airspaces, the applied horizontal separation standard in oceanic airspace is divided into longitudinal and lateral separations. Therefore, we developed a more accurate method PLOS detection method that considers lateral and longitudinal separations using a rectangular protection area. We validated the developed method against the judgment of a former oceanic air traffic controller (one of the authors of the present study) using surveillance data [25]. The proposed method was used in this study with reference to the reduced separation standard currently applied over the Pacific Ocean (longitudinal separation of 55.6 km (30 NM), lateral separation of 42.6 km (23 NM), and vertical separation of 305 m (1,000 ft)) to obtain a more precise evaluation compared to our previous studies on the same airspace.

PLOS occurrences in the target airspace between each pair of trajectories were detected at 1-min intervals from the aircraft positions (latitude, longitude and altitude) generated by the AirTOp simulator, and two metrics were extracted: PLOS count, which is the number of trajectory pairs with at least one PLOS event, and PLOS time, which is the sum of one-second time steps for PLOS was detected between each trajectory pair. For example, if PLOS between flights A and B was detected 30 times within the target airspace, it contributes 1 count to the total PLOS count for the airspace and 30 min to the PLOS time.

6) Calculating Flight Efficiency Change Associated with PLOS

PLOS indicates only the potential loss of separation between planned trajectories because in actual operations, ATC intervenes to maintain separation. Therefore, each PLOS count indicates an intervention by ATC, which typically corresponds to changing the altitude of one of the conflicting pair (lateral avoidance maneuvers typically cannot be applied in a non-radar control environment such as an oceanic airspace). It is assumed that air transportation flights are planned to operate at the most fuel-efficient altitude (which tends to be the highest altitude allowed by weight) consistent with the flight-level allocation rules. Therefore, inability to operate at the planned altitude due to a potential conflict tends to reduce flight efficiency. For example, the increase in fuel burn for an off-optimum altitude for the B737 is approximately 1–2 % for a 610 m (2,000 ft) difference [26], and the values for other aircraft types are similar. The PLOS time indicates the duration of such conditions due to ATC intervention. The PLOS time in the case of a conflict between a crossing pair of trajectories could be short, whereas the PLOS time in the case of flights traveling on a common route portion at the same time could be long. Therefore, the PLOS time is an indicator of the loss of flight efficiency.

The flight plan generation tool calculates the fuel consumption for the original flight route. The vertical profile was calculated using the fuel flow from the BADA3 PTF database for the corresponding aircraft type, linearly interpolating the tabulated cruise fuel burn rates at each altitude as necessary. To quantify the changes in the efficiency of a flight associated with a PLOS event, we assumed that the fuel consumption during the PLOS time associated with the PLOS event was at the fuel flow rate at the next lower altitude; that is, 2,000 ft lower than the planned altitude. Thus, the change in fuel consumption due to the PLOS event for an aircraft, $\Delta F$ , was calculated as shown in (1): $$\begin{equation*} \boldsymbol {\Delta }F={PLOS}_{TIME}\left ({{ F_{-2000}-F_{org} }}\right) \tag {1}\end{equation*}$$ where,

• $F_{org}$ is the fuel flow rate at the planned altitude, and

• $F_{-2000}$ is the fuel flow rate at an altitude 610 m (2,000 ft) lower than the planned altitude.

Among the potentially conflicting pairs of aircraft, the choice of aircraft chosen for intervention depends on the operational circumstances and the individual judgment of the air traffic controller. In this study, the change in the fuel consumption due to each PLOS event was assumed to be $$\begin{align*} & \text {Fuel consumption change per PLOS event} \\ & \qquad \qquad \qquad \qquad \qquad \qquad = \left ({{ {\Delta F}_{1}+{\Delta F}_{2} }}\right) \mathord {\left /{{\vphantom {\left ({{ {\Delta F}_{1}+{\Delta F}_{2} }}\right) 2}}}\right. \hspace {-1.2pt} } 2 \tag {2}\end{align*}$$ where the subscripts 1 and 2 denote each aircraft of the PLOS pair.

SECTION III. Results and Discussion

A. Flight Performance

Table 4 shows the average values of flight distance, flight time, and fuel consumption for individual flights for each airspace model. Eastbound and westbound flights are shown separately because they exhibit different trends owing to the prevailing winds aloft [11], [12]. For example, as shown in Fig. 5, eastbound flights tend to follow the jet-stream core axis, resulting in a longer flight distance, but shorter flight time and lower fuel consumption than westbound flights. The values of the performance indicators for the eastbound flights were smaller than those of the westbound flights for all three models, and the difference between the performance indicators increased in the order of M1, M2, and M3. Thus, the addition of gateways and the expansion of the flex-route operations airspace via removal of fixed airways enabled a more efficient route design.

TABLE 4 Various of Different Metrics for Each Airspace Model (Average Values Per Flight)

Fig. 8 shows the distributions of the differences in flight performance metric values between the baseline airspace M1 and airspace models M2 and M3. That is, Fig. 8 shows the variation in the benefits (negative values) and losses (positive values) for flights that operate in airspace configurations M2 or M3 compared to operations in M1. The histogram bin widths w of the flight distance, flight time, and fuel consumption differences were 2,000 m, 600 s, and 500 kg, respectively. Each bin excludes the lower endpoint and includes the upper endpoint, such that zero values are placed in the bin corresponding to the interval (−w, 0]. Table 5 presents the proportion of flights in the traffic scenario with benefits, no change, or losses for each metric compared with M1. The increased routing flexibility in M2 and M3 allows the design of flight plan routes that better reflect the characteristics of the winds; that is, the routes that are closer to the unconstrained ideal wind-optimal trajectories. There were flights with losses compared to the M1 model. This is attributed to the differences in the connection edges from the radar-controlled airspace to the gateways for eastbound flights, as well as to subtle gaps in the node-link route search graphs created within the FRA. However, the proportions and magnitudes of the losses were small and did not affect the overall conclusions.

TABLE 5 Proportions of Flights in Inter-Model Differences

FIGURE 8.

Distribution of flight benefits and losses for the traffic scenario given by airspace models M2 and M3 compared to the baseline airspace model M1.

The effects of changes in airspace configuration on flights depend on their origin and destination region or city pair: the changes benefit some flights and have possibly adverse effects on others. Fig. 9 shows a comparison of the average change in fuel consumption for airspace models M2 and M3 over the baseline M1 for each pair of regions. For eastbound flights, an increase in the number of gateways resulted in the greatest benefits to flights from Korea to Hawaii, and the NOPAC FRA offered the greatest benefits to flights from Japan to Alaska. For westbound flights, the benefits were greater for flights from North America than for flights from other regions. A major factor that distinguishes M3 from other models is the absence of fixed airway constraints in the NOPAC area in M3. In M2, some NOPAC fixed airways were retained to structure the traffic flow, and flight plan routes were prohibited from joining, branching from, or crossing the fixed airways within the NOPAC area to reduce complexity. The greater fuel benefit in the M3 model for westbound flights was attributed to the elimination of these constraints.

FIGURE 9.

Difference in average fuel consumption for flight trajectories of M2 and M3 models relative to M1 (baseline) model.

B. Changes in Gateway Traffic Density

A larger number of gateways was expected to increase the dispersal of flight routes because the flight plan generator can use the additional gateways to better approximate the ideal wind-optimal trajectories. The reduced traffic concentration resulting from dispersal was also expected to reduce PLOS and so increase the likelihood of flights being able to reach their planned cruise altitudes, thereby improving the overall efficiency of the traffic flow. In this section, we present the changes in traffic flow resulting from the addition of gateways.

Fig. 10 shows the gateways used in this study; the existing gateways are shown in black and the proposed additional gateways are shown in red. Fig. 11 and Fig. 12 respectively show the eastbound and westbound traffic volumes passing through each gateway for each airspace model. Region pairs are color-coded using the scheme shown in Fig. 9. The labels “Other_n” and “Other_s” refer to flights that do not pass through the gateways. Depending on the upper-level winds, some flights may pass through Russian Federation airspace to the west and north of the NOPAC area instead of traveling via the North Pacific; such flights are labeled ‘Other_n’ (Fig. 13). Furthermore, some southeast Asia flights fly along the southern area of the Fukuoka Oceanic Airspace without passing through the gateways, and they are labeled ‘Other_s’.

FIGURE 10.

Proposed gateway design. New waypoints (shown in red) are added between existing waypoints (shown in black). NOPAC areas (which cannot be crossed by traffic not on airways) are shaded: Model M1 = green + blue: Model M2 = green area only. Solid lines are fixed airways for the M1 and M2 models, dashed lines are fixed airways for the M2 model only, and dotted-lines are fixed airways for the M1 model only.

FIGURE 11.

Eastbound traffic passing through each gateway.

FIGURE 12.

Westbound traffic passing through each gateway.

FIGURE 13.

Traffic flows not passing through gateway area.

The gateways with the highest traffic concentrations in each model were ADNAP in M1 and PUTER in M1 and M2 for eastbound flights, and NWP31 in M3 for westbound flights. We now focus on these gateways of interest.

For eastbound flights, Fig. 11 shows that traffic through PUTER in M2 increases by approximately 10% compared to M1, whereas the ADNAP traffic decreases sharply. To help explain this change, the shift of traffic through these gateways in M1 to northern and southern gateways in M2 and M3 was examined, and the results are shown in Table 6(a). In M2, nearly two-thirds (64%) of ADNAP traffic moved north, and 39% transitioned to PUTER. In M1, PUTER is a gateway to a NOPAC fixed airway for traffic towards the Anchorage Oceanic FIR, which is deleted in models M2 and M3 (see Fig. 10). Thus, PUTER became available to other traffic, some of which originally transited via ADNAP. This can be seen, for example, in the traffic towards Hawaii, as indicated by the appearance of the red/orange/pink colored traffic at PUTER in Fig. 11 (b) and Fig. 11 (c). This contributed to an increase in the volume of traffic through PUTER and a decrease in traffic through ADNAP.

TABLE 6 Shifts in Traffic Between Gateways for Different Models

For westbound flights, the most marked change in Fig. 12 is the concentration of traffic at NWP31 in M3, which is close to the boundary of the NOPAC with Russian Federation airspace (Fig. 10). Table 6(b) shows the shifts in traffic from M1 and M2 to NWP31 in M3, and lists the gateways with the largest shifts. The gateways with the largest shifts were ADNAP in M1 and NWP21 in M2, which are the gateways closest to the eastern boundary of the NOPAC airspace in their respective models. As shown in Fig. 14, to avoid the jet stream headwind area, westbound traffic from North America tends to select routes that approach Japan from the northeast tending towards along a direction roughly parallel to the Kurile Islands. However, the most eastern fixed airways in the NOPAC airspace are eastbound only, and the NOPAC operating rules do not permit westbound flights to cross the fixed airways. Thus, in M1 and M2, such flights either enter the NOPAC airspace from the north via the Anchorage Oceanic FIR boundary, or they enter Fukuoka FIR radar-controlled airspace at the gateway closest to the eastern edge of the NOPAC airspace (ADNAP in M1 or NWP21 in M2). In contrast, M3 allows the latter flights to cut across the NOPAC area and shift to NWP31.

FIGURE 14.

Route trends of westbound flights from north America to Japan and south Korea. Dashed green line indicates the ideal route that avoids jet stream core area (shaded blue) and is available in M3. In M1 and M2, flights cannot cross the NOPAC eastbound fixed airway (pink line) and hence, use the solid green line route.

We conclude that the proposed gateway design generally disperses the routes of the eastbound flights. The addition of gateways increases the number of two-dimensional route options (called the “2D (two-dimensional) route effect” here) and makes it possible to plan routes with lower fuel consumption. However, if the ideal routes for multiple flights favor the same gateway (as seen in the case of westbound traffic through NWP31 in the M3 airspace model), and given that flights between the same region pairs tend to operate during the same time periods, the resulting traffic concentration will increase PLOS and conflict resolutions via altitude change; thus, it may be difficult to achieve a fuel benefit from the two-dimensional routing effect.

C. Changes in PLOS

We now compare the PLOS of each model to evaluate the effects of additional gateways and the relaxation of NOPAC flight planning constraints on airspace management, airspace efficiency, and flight benefits.

As shown in Table 1, the airspace evaluation traffic scenario had 11 combinations of winds, three initial weights, and four departure times, resulting in a sample size of 132 flights. Fig. 15 and Fig. 16 respectively show box-and-whisker plots of PLOS count and PLOS time for each airspace model; the averages over all flights are shown as bar heights, and the error bars indicate standard deviation. Compared to the baseline model M1, M2 reduced the PLOS count by 7.4% for eastbound flights and 1.8% for westbound flights, whereas M3 increased the PLOS count by 1.1% for eastbound flights and 36.6% for westbound flights. Compared to the PLOS times of the M1 model, the M2 and M3 models reduced the PLOS times of eastbound flights by 11.2% and 6.9%, respectively; for westbound flights, M2 reduced the PLOS times by 3.5% and M3 increased the PLOS time by 6.1%. The mean values of both PLOS count and PLOS time of westbound flights were significantly higher in M3 (confirmed via the Mann-Whitney U-test, p < 0.05).

FIGURE 15.

Average per-flight PLOS counts for each airspace model. Range of one standard deviation is shown by error bars.

FIGURE 16.

Average per-flight PLOS times for each airspace model. Range of one standard deviation is shown by error bars.

The factors that might increase the PLOS of westbound flights in M3 were investigated. First, we compared the PLOS counts and PLOS times by departure region of the M2 and M3 models with the M1 model. The percentage changes in PLOS count and PLOS time relative to the M1 model are shown in Tables 7 and 8, respectively. In the M3 model, PLOS counts increased considerably for both Alaskan flights (+70%) and North American flights (+40%), and PLOS times for the Alaskan flights also increased considerably compared to other regions (+25%). The geographic locations of PLOS occurrences were then visualized using heat maps, and the results are shown in Fig. 17. The figure shows a heat map of PLOS times for westbound flights integrated over a grid of 5° longitude $\times 1$ ° latitude cells, and the total time in each cell is indicated according to a color scale. The heat maps are plotted from 140°E to 235°E (125°W) and from 20°N to 60°N. The lines in each plot show the airspace boundary lines, and oceanic airspaces are delimited by red lines. M3 increased the PLOS time concentration in the northernmost part of the NOPAC area along the boundary with Russian Federation airspace, which is consistent with the concentration of traffic from North America at NWP31, as shown in Fig. 12(c).

TABLE 7 Percentage Change in PLOS Count for Westbound Flights Relative to the M1 Model (by Departure Region)

TABLE 8 Percent Change in PLOS Time for Westbound Flights Relative to the M1 (2023) Model by Departure Region

FIGURE 17.

PLOS time heat maps for westbound flights. Color scale indicates cumulative PLOS time (minutes) within a 5° longitude $\times 1$ ° latitude grid of cells.

D. Trade-off Between 2D Route Effect and PLOS Time

By eliminating the prohibition on crossing, branching, and merging associated with fixed airways in the NOPAC area and increasing the number of gateways, the M3 model enabled flight plan routes with the highest fuel savings (Section III-A). However, owing to this flexibility, more flights from North America could reach more favorable northern gateways. Consequently, the traffic concentration at those gateways increased, leading to increases in the PLOS count and PLOS time of westbound traffic (Sections III-B and III-C). This section examines the trade-off between the increase in fuel benefit due to flight planning flexibility (2D route effect) and the reduction in fuel benefit due to the increase in PLOS (due to operations at less efficient altitudes).

We first estimated the reduction in fuel benefit due to the increase in PLOS as described in Section II-C6. Table 9 shows the estimated average change in fuel consumption per flight due to PLOS events in each airspace model calculated using equation (2). Fig. 18 shows the changes in fuel consumption in the M2 and M3 models compared to the baseline M1 model resulting from the 2D route effect and PLOS. A positive value indicates a fuel benefit (i.e., lower fuel consumption) compared to the baseline. The changes in fuel consumption resulting from the 2D route effects are derived from the results shown in Table 4, and those resulting from the conflict resolution of PLOS are derived from Table 9. Westbound flights in the M3 model show high traffic flow concentration and the highest number of PLOS events. The extra fuel consumption due to altitude changes to resolve PLOS was greater than that in other models, leading to a negative value; however, the positive benefit resulting from the 2D route effect significantly outweighed the loss.

TABLE 9 The Average Change in Fuel Consumption Per Flight Due to PLOS Events in Each Airspace Model for Eastbound Flights and Westbound Flights

FIGURE 18.

Fuel consumption benefits for the M2 and M3 models compared to the M1 model.

SECTION IV. Conclusion

Among the three airspace models investigated in this study, the M3 model (the NOPAC FRA model with the proposed additional gateways) gave the lowest per-flight fuel consumption. Although our brief assessment does not take into account any new “knock-on” PLOS that may arise after an altitude change to resolve an initial PLOS, the fuel benefit resulting from the increased lateral route flexibility was significantly greater than the penalty incurred because of altitude changes for PLOS resolution. The M3 model gives a greater increase in PLOS count than PLOS time, which implies that compared to M2, there are fewer conflicts occur between flights flying on the same route and more crossing conflicts. An increase in the number of intersecting flight routes corresponds to an increase in complexity, which in turn necessitates more frequent altitude changes and a consequent increase in ATC workload. Therefore, the practicability of the proposed FRA under actual operational circumstances at the current traffic density depends on whether air traffic controllers can handle the greater workload. However, workload could be alleviated by increased automation, such as more sophisticated conflict probes and better conflict resolution support. Therefore, it is necessary to examine the applicability of the proposed system in actual operational scenarios.

ACKNOWLEDGMENT

The authors thank the Japan Civil Aviation Bureau (JCAB) for providing flight plan data for the flight schedules generated in this study.