Nomenclature

ADP

aircraft departure problem

ALP

aircraft landing problem

ASSP

aircraft sequencing and scheduling problem

MILP

mixed integer linear programming

MIP

mixed integer programming

NM

nautical miles

PMS

point merge system

RSP

runway assignment problem

TMA

terminal manoeuvring area

2.0 Literature Review

Increasing traffic and congestion in the air traffic system have led to various solutions and tools to improve efficiency. Decision support mechanisms such as the arrival manager and departure manager, models aimed to prevent traffic conflicts, the design of standard instrument departures and standard terminal arrival routes that provide procedural separation, the collection point system established by eurocontrol in 2009, etc., are some examples. While some studies examine ground movements and terminal manoeuvring area (TMA) operations separately, some studies aim to optimise ground movements by considering ground movements and TMA operations. These studies include aircraft sequencing, runway assignment, and gate assignment. In the literature, solutions to these problems are commonly sought through mathematical modelling. Mathematical models provide solutions by considering the constraints in line with the specified objective. In aircraft sequencing and scheduling, attention to runway and gate assignments allows more aircraft to be served per unit time. Aircraft sequencing and scheduling problem (ASSP) is generally divided into aircraft landing problem (ALP) and aircraft take-off problem (ATP), which consider unique physical constraints involving arrival or departure operations. If both landing and take-off traffic is analysed in the study, it is called the ASSP. Since ASSP evaluates both landing and take-off aircraft, it reflects the reality more. ALP and ATP aim to determine the optimal arrival and departure sequencing to optimise the objective function given the operational constraints. The constraints used for the mathematical formulation of ALP and ATP are the separation values between two consecutive aircraft required for a safe landing, the maximum time constraint for the aircraft to land due to fuel, and priority constraints. The objective function used varies depending on the decision maker. The decision maker can set different objective functions by considering the various stakeholders of the airport system. However, there are differences between what is in theory and what is realised due to the human factor [Reference Ikli, Mancel, Mongeau, Olive and Rachelson8].

Flow optimisation in air traffic management has been the subject of many studies from different perspectives. These studies address conflict resolution, sequencing and scheduling issues to achieve flow optimisation in the airspace. In contrast to previous studies, Hong et al. presented a single and multi-objective optimisation model for the ASSP with a limited number of sequence changes by considering the working principle of an actual air traffic controller. This study paved the way for studies that will take more into account the human factor [Reference Hong, Cho, Kim and Choi9]. Cecen presented a single and multi-objective optimisation model that minimises the total delay and number of conflict resolution manoeuvres by identifying that sequencing and conflict resolution is the most time-consuming task for air traffic controllers. A solution to the ASSP was presented with mixed integer programming (MIP) [Reference Cecen and Kursat10]. Cecen et al. proposed a multi-objective optimisation model for the ALP to minimise emissions and total flight time. They used the GAMS CPLEX solver, which gives exact solutions. Their study obtained Pareto-optimal solutions for each air traffic situation [Reference Cecen, Saraç and Çetek11].

In other studies, Ma et al. presented a model that handles airspace and ground operations together, aiming at both traffic flow optimisation and the proper functioning of ground operations, with the idea that the air traffic system, airspace operations and airport operations should be considered. With the presented model, a solution for resolving conflicts in the airspace was offered, while at the same time, runway assignment was proposed [Reference Ma, Delahaye, Sbihi, Scala and Mujica Mota12]. Cecen et al. proposed a stochastic MILP model to solve ASSP using the simulated annealing algorithm. The model, which aims to minimise the total aircraft delay for an airport runway serving mixed operations, provided an appropriate aircraft sequencing considering the wind direction uncertainties, which are critical in the decision-making process [Reference Cecen, Cetek and Kaya13]. Ghoniem et al. addressed the static aircraft sequencing problem for single runway or closely interacting parallel runways with mixed operation modes. The model utilised the basic structure of the time window asymmetric peddler problem, and its effectiveness was demonstrated with simulated examples using real data from Doha International Airport [Reference Ghoniem, Sherali and Baik14]. Cecen presented a mathematical model to minimise the total fuel consumption for ALP. The path stretching method and the point merge system (PMS) were compared using vector manoeuvre and speed reduction techniques. The results showed that the path-stretching approach increases the number of continuous descent operations and reduces fuel consumption [Reference Cecen15]. Some studies provided various solutions to RSP using single and multi-objective models. Runway assignment studies aim to provide a more efficient traffic flow by assigning landing and take-off aircraft to the most suitable runway. When runway assignment is considered from the perspective of the airline carrier, the parking position taxi time between the runway and the parking position is included in the objective function. In contrast, safety, shorter flight time, and less workload are considered from the air traffic controller’s perspective. Liu et al. introduced a mathematical model for the TMA operations using a tailored selective simulated annealing algorithm. They compared the trombone procedure and the PMS system. Their outputs showed that the PMS system improved the performance of arrival air traffic [Reference Liu, Delahaye, Cetek, Zhao and Notry16]. Salehipour mentioned that the ALP is a problem that includes both the landing sequence of aircraft and the runway assignment. For this reason, he presented a heuristic model that can provide a solution to the ALP in a short time. He tested his algorithm on 124 examples. According to the results, the presented model produced a satisfactory solution quickly [Reference Salehipour17]. Dönmez et al. discussed the importance of landing sequencing in a collection point system and the reassignment of runways to ensure maximum efficiency in landing sequencing. They showed that assigning aircraft to a different runway can significantly reduce delays and fuel consumption due to the difference in taxi entry/exit times. The study proposed single and multi-objective programming models for a TMA with multiple collection point merge systems to minimise total fuel consumption, flight time and delay, including taxi entry/exit times. According to the results obtained from applying the proposed model with Istanbul Airport data, it reduced delays by 77.5% and fuel consumption by 8.7% due to the difference in taxi entry/exit times [Reference Dönmez, Çetek and Kaya18]. Dönmez et al. presented deterministic single-objective, deterministic multi-objective, stochastic single-objective and stochastic multi-objective mathematical models for parallel runway airports with multiple collection point systems considering wind uncertainty and taxi times, respectively. The model, which aims to minimise total fuel consumption, total flight time, and total delays, provides arrival-departure sequencing, scheduling and runway assignments and shows how wind and taxi times affect fuel consumption, flight times, delays and runway preferences [Reference Dönmez, Çetek and Kaya19].

Guépet et al. presented a model to resolve terminal airspace conflicts, reduce airside capacity overload and reduce delays, as airspace and ground operations were considered the main bottlenecks affecting capacity. In this study, an optimisation model was produced by using speed, arrival time, departure time and runway assignment, and then the optimisation model is tested with data from Paris Charles De Gaulle Airport using the simulation annealing method. A four-hour period was considered for the test during one of the airport’s peak days, and runway assignments were made for landings and take-offs. As a result, a 37% improvement in landing delays and a 36% in take-off delays were achieved. The presented model showed that balanced use of runways and reduced delays can be achieved [Reference Guépet, Briant, Gayon and Acuna-Agost20]. Rodríguez-Sanz et al. introduced the integration of runway scheduling for arrival and departure operations with a mathematical model to minimise delays in the tactical phase while maximising runway utilisation. The presented model was tested with data from Alicante-Elche airport. The results showed that the presented model reduces the delays by 50% on average [Reference Rodríguez-Sanz, Arnaldo Valdes, Pérez-Castán, López Cózar and Comendador21].

It has been observed that environmental factors have been ignored in the studies conducted for runway efficiency. Sölveling et al. presented an integrated approach to ensure efficient runway use and minimisation of environmental impact with a first-come, first-served (FCFS) logic. The study aimed to minimise cost and environmental factors while introducing runway scheduling and aircraft sequencing. Instead of including environmental factors directly, they were included in the objective function in the form of fuel consumption. The presented model was analysed at Detroit Metropolitan Wayne County Airport, which has two parallel runways with intersecting taxiways. The results showed that optimisation-based scheduling that considers environmental costs significantly saves airlines and society. Furthermore, even if environmental components are not directly included in the optimisation but a fuel consumption-based target is used, ecological savings through a FCFS policy are still significant [Reference Sölveling, Solak, Clarke and Johnson22]. Lieder et al. presented an optimisation model for the ASSP based on generic runway configurations. The dynamic programming heuristic model minimised the delay-related costs of aircraft and generated a runway assignment solution by considering the minimum diagonal separation between dependent runways. The model’s validity was tested on the runway configurations currently used at Frankfurt Airport, which has four runways [Reference Lieder and Stolletz23]. Ng et al. presented a model that provided solutions to dynamic runway configuration planning, aircraft sequencing, and scheduling problems by considering the criteria considered by the air traffic controller in determining the runway configuration to ensure efficient and balanced use of runways during peak and off-peak periods of landing or take-off air traffic. The model, which provides a systematic approach to runway operations, was tested at Hong Kong International Airport and reduced delays by 71.6% and 37.08% for the two modes of using the separated runway system, respectively [Reference Ng, Lee, Zhang and Keung24]. Cecen presented a mathematical model for a terminal traffic flow problem. Extended TMA boundary and conventional TMA boundary with PMS were compared using vector manoeuvre, the PMS and speed reduction techniques. The result revealed that extended TMA operations using speed reduction reduced fuel consumption [Reference Cecen25].

Weiszer et al. presented an optimisation model to provide solutions to airport ground movements and runway scheduling problems, emphasising the efficient use of time and the consideration of fuel consumption and related emission values while investigating the optimisation of airport surface operations. To evaluate the performance of this model, experimental studies using data from Doha, Manchester and Beijing Capital International airports showed that the model could be used as a decision support mechanism [Reference Weiszer, Chen, Stewart and Zhang26]. Fritzsche et al. studied the dynamic assignment of runways to landing aircraft, recognising that flexible and demand-driven use of runways could be vital to meeting future capacity, efficiency and environmental sustainability. The study aims at balanced runway utilisation to reduce landing and take-off delays and taxi times, which cause fuel consumption and emissions. The results showed that providing a balanced runway assignment impacts delays, taxi times and fuel consumption [Reference Fritzsche, Günther and Fricke27]. Delsen examined the problem of RSP for landing and take-off aircraft at Amsterdam Schiphol International Airport from a different perspective. Since he observed a trade-off between reducing noise and increasing capacity in the current runway assignment but that fuel consumption and resulting emissions were not taken into account in the runway assignment, he presented a model using MIP that aims to optimise for fuel and noise [Reference Delsen2].

3.0 Model description

To test the validity of the mathematical model developed in this study, which provides a solution to the runway assignment problem, actual data from Istanbul Airport was utilised.

3.1. İstanbul airport structure

On October 31, 2018, shortly after the first commercial flight was launched, Istanbul Airport became the busiest airport in Europe, surpassing Paris Charles de Gaulle, Amsterdam Schiphol, Frankfurt and London Heathrow airports, which are the busiest airports in the world [28]. Istanbul Airport served 200 m passengers with 262 destinations in 2021. Istanbul Airport, where the first phase has been completed [29], has 5 parallel runways (34L, 34R, 35L, 35R and 36) and 186 taxiways. While runways pairs 34L-34R and 35L-35R, are dependent parallel runways, 34-35-36 runways are operated independently.  For instance, runways 34R and 35L are used as independent parallel runways while 34L is used for arrival and 34R is used for departure operations. Depending on the wind direction and different variables, aircraft generally use two different configurations: North and South. The North and South configuration is shown in Fig. 1.

Figure 1. Configuration of the parallel runways at Istanbul Airport.

Only the North configuration and its operation modes were studied in this study. It assumed that 34L and 35R serve for landing, 34R and 35L serve for take-off, and runway 36 serves for both landing and take-off.

3.2. Istanbul airport traffic demand

According to the State Airports Authority’s annual report for 2021, traffic volume in 2021 increased by 39% compared to 2020 but remained below the traffic volume in 2019 due to the travel restrictions imposed during the pandemic. In 2021, the number of passengers carried increased by 57.1% compared to 2020, reaching approximately 128 m passengers. The Covid-19 crisis has also significantly affected airports in Turkey. Although it took a long time for many airports to recover from this impact, Istanbul, Sabiha Gökçen and Antalya Airports were close to their previous performance in a short time. Among these, Istanbul Airport became the second airport in Europe in terms of average daily movements [30].

Istanbul Airport served 280,108 traffic in 2021. While 94% of the aircraft served are passenger aircraft, cargo aircraft are the second most common type, with a rate of 4%. State, military and other aviation aircraft accounted for 1% of the aircraft served. In September 2021, 29,818 aircraft (14,955 landings and 14,863 take-offs) used Istanbul Airport. In September 2021, a daily average of 499 landings and 496 take-offs, and a daily average of 993 traffic was served, according to the data obtained from the State Airports Authority, the airport, which is mainly used by passenger planes, cargo planes, military planes and private jets, which constitute the majority of the traffic, served an average of 985 passenger planes per day in September. While 30% of these passenger flights were domestic, 70% were international. It was figured out that there were 304 domestic flights and 681 international flights on a daily average.

For the analysis mentioned above, the aircraft types used at Istanbul Airport in September 2021 and the average taxi time between each parking position and each runway were obtained. The aircraft types are obtained using the one-month traffic data of Istanbul Airport. Since the engine types directly affect the fuel consumption value, the engine types used for 1 jumbo category, 20 heavy categories and 27 medium categories of aircraft types in the table were obtained from Eurocontrol data and examined in detail. Fuel consumption data for the alternative engine types listed for each aircraft type were obtained from the ICAO Engine Emission Data Bank [7]. Listing all possible engine types was essential to see the differences in fuel consumption. If more than two engines exist for an aircraft type, the engine that provides less fuel consumption values was used for an aircraft type.

Since Istanbul Airport covers a large area, taxi times vary considerably depending on the runway and the gate assignments. To calculate the average taxi time between Istanbul Airport runways and parking positions, the taxi times of the aircraft landing and taking off from the airport in one month, the data of five runways and 364 parking positions used were analysed carefully, and the average taxi time between each runway and parking position was obtained.

According to the analysis of the average taxi time of landing traffic for each runway in a previous academic study, an aircraft landing on runway 34L has an average taxi time of 17.97 minutes, an aircraft landing on runway 35R has an average taxi time of 10.20 minutes and an aircraft landing on runway 36 has an average taxi time of 10.45 minutes. According to this analysis, an Airbus 300-600 aircraft landing on runway 35R instead of 34L can taxi 7.77 minutes less on average and save 191kg of fuel, provided the assigned gate remains the same. For this reason, the study examines the improvement that can be achieved by the runway assignment through fuel savings [Reference Güven and Çetek31]. According to the data, in September 2021, it had an average of 1,006 daily traffic. In September, eight different days with higher-than-average September traffic are selected. In the model, six-hour traffic operations are considered. The amount of traffic in the scenarios is shown in Table 1.

Table 1. The amount of traffic in the scenarios

3.3 Mathematical model

This section introduces the sets, parameters, decision variables, constraintsand objective function equation used in the proposed mathematical model (PMM). This problem aims to minimise the total fuel consumed in ground movements for the RSP. Equations (1b), (2)–(13) are used for the fixed runway assignment approach (FRAA) regarding actual data, and Equations (1a), (2)–(13) are used without FRAA.

Sets:

1. $I = {\rm{\;}}\left\{ {1,{\rm{\;}}2,{\rm{\;}} \ldots .,n} \right\}{\rm{\;}}$ the aircraft set

2. $J = {\rm{\;}}\left\{ {1,{\rm{\;}}2,{\rm{\;}} \ldots .,r} \right\}{\rm{\;}}$ the runway set

3. $K = {\rm{\;}}\left\{ {1,{\rm{\;}}2,{\rm{\;}} \ldots .,s} \right\}{\rm{\;}}$ the park position set

Indices:

1. $i,\;{i_1},\;{i_2} \in \;I$ are indices to donate aircraft

2. $j,\;{j_1},\;{j_2}\;$ are indices to donate runway

3. $k\; \in \;K$ is indices to donate the park position

Parameters:

Decision Variables:

Constraints

(1a) \begin{equation} \mathop \sum \limits_{j|r{a_{o{x_i},j}} = 1} {x_{i,j}} = 1 \qquad\qquad \forall \;i \in I \end{equation}

(1b) \begin{equation}\mathop \sum \limits_{j|j = {r_i}} {x_{i,j}} = 1 \qquad\qquad \forall \;i \in I \end{equation}

(2) \begin{equation} ru{t_i} = r{t_i}\qquad\qquad \forall \;i \in I,o{x_i} = 1 \end{equation}

(3) \begin{equation} gu{t_i} = r{t_i} - {t_i}\qquad \forall \;i \in I,o{x_i} = 2 \end{equation}

(4) \begin{equation}ru{t_i} = r{t_i} - {t_i} + \mathop \sum \limits_j {x_{i,j}} \cdot taxiou{t_{j,k}} + g{w_i}\qquad \forall \;i \in I\;,\forall \;k \in K,\;o{x_i} = 2, \;k = {g_i} \end{equation}

(5) \begin{equation}gu{t_i} = ru{t_i} + \mathop \sum \nolimits_j^{} {x_{i,j}} \cdot taxii{n_{j,k}} + a{w_i}\qquad \forall \;i \in I\;,\forall \;k \in K,\;o{x_i} = 1,\;k = {g_i} \end{equation}

(6) \begin{align} & ru{t_{{i_2}}} - ru{t_{{i_1}}} \geq tse{p_{j,{p_{{i_1}}},{p_{{i_2}}},o{x_{{i_1}}},o{x_{{i_2}}}}}\nonumber\\[5pt] & - \left( {1 - {e_{{1_{{i_1},{i_2}}}}}} \right) \cdot M - \left( {2 - {x_{{i_1},{j_1}}} - {x_{{i_2},{j_2}}}} \right) \cdot M\qquad \forall \;{i_1},{i_2} \in I,\forall \;{j_1},{j_2} \in J,\;{i_1} \ne {i_2},dp{r_{{j_1},{j_2}}} = 1 \end{align}

(7) \begin{align} & ru{t_{{i_1}}} - ru{t_{{i_2}}} \geq tse{p_{j,{p_{{i_2}}},{p_{{i_1}}},o{x_{{i_2}}},o{x_{{i_1}}}}}\nonumber\\[6pt]& - \left( {{e_{{1_{{i_1},{i_2}}}}}} \right) \cdot M - \left( {2 - {x_{{i_1},{j_1}}} - {x_{{i_2},{j_2}}}} \right) \cdot M\qquad\qquad \qquad \forall \;{i_1},{i_2} \in I,\forall \;\;{j_1},{j_2} \in J,\;{i_1} \ne {i_2},dp{r_{{j_1},{j_2}}} = 1 \end{align}

(8) \begin{equation}gu{t_{{i_2}}} - gu{t_{{i_1}}} \geq B - \left( {1 - {e_{{2_{{i_1},{i_2}}}}}} \right) \cdot M\qquad \qquad \qquad \forall \;{i_1},{i_2} \in I,\;{g_{{i_1}}} = {g_{{i_2}}} \end{equation}

(9) \begin{equation}gu{t_{{i_1}}} - gu{t_{{i_2}}} \geq B - \left( {{e_{{1_{{i_1},{i_2}}}}}} \right) \cdot M\qquad \qquad \qquad \forall \;{i_1},{i_2} \in I,\;{g_{{i_1}}} = {g_{{i_2}}} \end{equation}

(10) \begin{equation}taxi{a_i} = \mathop \sum \nolimits_j^{} {x_{i,j}} \cdot taxii{n_{j,k}}\qquad \qquad \qquad \forall \;i \in I\;,\forall \;k \in K,\;o{x_i} = 1,\;k = {g_i} \end{equation}

(11) \begin{equation}taxi{d_i} = \mathop \sum \nolimits_j^{} {x_{i,j}} \cdot taxiou{t_{j,k}}\qquad \qquad \qquad \qquad \forall \;i \in I\;,\forall \;k \in K,\;o{x_i} = 2,\;k = {g_i} \end{equation}

(12) \begin{equation} min\mathop \sum \nolimits_i^{} \left( {taxi{a_i} + taxi{d_i} + a{w_i} + g{w_i}} \right) \cdot {f_i} \end{equation}

Equation (1a) ensures that each aircraft is assigned to a suitable runway according to the type of operation. In contrast, Equation (1b) ensures that each aircraft is given to the runway, considering the actual data. According to real data, Equation (2) provides the runway use time for each aircraft that arrives. Similarly, Equation (3) calculates each departure aircraft’s parking position leaving time. Equation (4) calculates the runway use times of the aircraft relative to taxi duration, using the parking position leaving time while considering the departure waiting time of each departure aircraft. Likewise, Equation (5) calculates the parking position reaching time for each landing aircraft using landing time, taxi duration and parking position waiting time durations. Equations (6) and (7) maintain wake turbulence separations between two consecutive aircraft using the same runway or two parallel dependent runways. Similarly, Equations (8) and (9) provide the time separation between pairs of aircraft using the same park position. Equations (10) and (11) calculate the taxi times obtained in ground operations for each landing and departing aircraft. Equation (12) minimises the total fuel consumption, considering taxi duration and waiting times. The parameters in Equations (6) and (7) were obtained using the minimum separation values published in Turkey by DHMI Aviation Information Publication [32]. These values are presented in Table 2. In addition to these published values, the wake turbulence separation criteria between the aircraft in the ICAO Procedures for Air Navigation Services - Air Traffic Management document were also considered [33]. Wake turbulence separation values are given in Table 3.

Table 2. Runway separation minima in nautical miles (NM)

Table 3. Wake turbulence separation

The minimum radar separation in Istanbul Terminal Airspace is 3NM. However, the minimum separation value that reduces for runways is 4NM. It has been applied in a way that meets the requirements of Table 3. The model assumes that the engine consumes fuel in the idle configuration during the taxi movements and the gate waiting periods, the landing aircraft can only wait when approaching right before the park position, and the departure aircraft can wait just before entering the runway. The wind speed is assumed to be calm at 0kt.

4.0 Computational results

To test the proposed MILP model, scenarios were created using the busiest six-hour part of the eight-day traffic data of Istanbul Airport for September 2021. These scenarios were analysed according to the runway assignments presented in the actual traffic data and the runway assignments made by the mathematical model. The number of arrival and departure aircraft and the total number of aircraft in the traffic data used are shown in Table 4.

Table 4. Arrival and departure aircraft number for scenarios

While a minimum of 194 and a maximum of 241 arrival aircraft occur in the scenarios, the number of departure aircraft is at least 217 and at most 248. The maximum traffic in scenarios with the busiest six hours of a day’s traffic is 478. The results of the study can be evaluated in two criteria. In comparison, the first part evaluates the examination of operation types of aircraft in terms of taxi times and fuel consumption values. The second one assesses taxi times and fuel consumption, considering their performance categories. The average taxi times and enhancement rates (%) for aircraft operation types comparing runway use for the FRAA and PMM are presented in Table 5.

Table 5. Average taxi times of arrival and departure traffic (min.)

In the scenarios for arrival traffic using FRAA, the average taxi time of the landing traffic varies between 9.5 minutes and 12.4 minutes, while the average taxi time varies between 8.1 and 10.1 minutes in the PMM. The PMM provided a minimum of 11.3% and a maximum of 23.1% improvement in the average taxi time of landing traffic in eight scenarios. While the average taxi time of departure traffic varies between 13.5 and 16.7 minutes in the FRAA, the average taxi time varies between 13.3 and 14.9 minutes in the PMM. It is seen that the proposed model provides a minimum of 0.8% and a maximum of 10.8% improvement in the average taxi time of arrival traffic in a total of eight scenarios. According to the results, departure air traffic takes longer in taxis. It has been observed that waiting on the ground is preferred since arriving aircraft have priority in landing. In addition, the difference in the taxi routes used by the arrival and departure aircraft on the ground directly affected the taxi times. This situation can be seen in Table 5. The average fuel consumption values of the arrival and departure traffic in the taxi operations are given in Table 6.

Table 6. Average fuel consumption of arrival and departure traffic during taxi operations (kg)

Arrival traffic consumed a minimum of 241.8 and 307.1kg of fuel during the taxi duration. The improvement rates offered by the PMM, which significantly reduces the average fuel consumption of the arrival traffic in the taxi, vary between 13.7% and 24.7%. In four scenarios, it is seen that the average amount of fuel consumed by the landing traffic in the taxi has improved by over 24%. Departure traffic consumes a minimum of 337.6kg and a maximum of 431.4kg of fuel during the taxi duration. It is seen that the model’s departure traffic improves the average fuel consumption in the taxi between 0.9% and 11.1%. Scenario 1 has the highest enhancement rate. Considering the traffic distribution in the scenarios, it is seen in Table 7 that the aircraft in the heavy category do not exceed 30% of the total traffic. For this reason, exploring fuel consumption according to the aircraft performance categories is necessary to understand the study’s importance.

Table 7. Number of traffic according to performance categories (count)

Table 8 shows the average taxi times of aircraft set in the medium performance category (MPC) and heavy performance category (HPC) aircraft for each scenario.

Table 8. Average taxi times for medium and heavy performance categories traffic (min.)

The average taxi time for medium category traffic in FRAA is at least 11.5 and at most 14.5 minutes. In the PMM, this value is at least 10.7 minutes and, at most, 12.5 minutes. With the PMM, the minor improvement was in Scenario 6, with 3.7%. In this scenario, the average taxi time was reduced from 11.5 to 11.1 minutes. The highest improvement rate was 14% in the first scenario. In this scenario, the average taxi time was reduced from 14.5 to 12.5 minutes. It is seen that the average taxi time for heavy category traffic is at least 12.2 minutes and at most 15.1 minutes in essential scenarios. In the PMM, it is seen that this value is at least 10.9 minutes and, at most, 13 minutes. The highest improvement rate was observed in the eighth scenario, with 15%. In Scenario 8, the proposed model reduced the average taxi time from 13.1 to 11.1 minutes. The amount of fuel consumed by medium and heavy category traffic in a taxi is given in Table 9.

Table 9. Average fuel consumption of medium and heavy category traffic (kg)

It was observed that a minimum of 5.2% and a maximum of 14.4% improvements were achieved for MPC. For the first scenario, which showed a gain of 14.4%, it is seen that the average fuel consumption values were reduced from 324.7 to 278kg. In Scenario 6, where the lowest recovery rate is seen, fuel consumption in the average taxi decreased from 268.4 to 254.5kg. While the average amount of fuel consumed by HPC traffic in a taxi was at least 388.4kg in Scenario 6, the average fuel consumption was reduced to 354.6kg with the proposed model. A minimum of 7.5% and a maximum of 15.1% improvement has been demonstrated. In Scenario 8, with the highest recovery rate, it is seen that there are 134 heavy-category aircraft, and the average fuel consumption values have been reduced from 424.4 to 360.4kg. These values saved more than 8 tons of fuel with the model proposed for the eighth basic scenario. When the changes in the taxi times for MPC and HPC aircraft and the decrease in total taxi fuel consumption are examined with the model presented, it reduces at least 4.5 tons, at most 13.7 tons in medium category scenarios. In comparison, it decreases to at least 3.7 tons and, at most, to about 8.5 tons in the heavy category. The average fuel consumption enhancements among all scenarios are 9.18% and 6.15% for the MPC and HPC, respectively. The distribution of the number of aircraft using the runways in the scenarios is given in Table 10.

Table 10. The distribution of the number of aircraft using the runways (count)

When all scenarios are examined, when the 34L runway is used only for arrival operations, it is seen that the utilisation rate has increased in all proposed scenarios. It is seen that an assignment has yet to be made to runway 34R, which is used only as a take-off runway. It is understood from the tables that assigning departures to the 35L runway, which is used only as the take-off runway, reduces taxi time and, accordingly, fuel consumption. When the usage frequency and usage rates of the 35L runway are examined, the usage rate of the 35L runway has increased in all scenarios except the fourth scenario. Considering the use frequency and usage rates of the 35R runway, which is used only for arrivals, the use rate was increased in the three scenarios suggested. In comparison, it was decreased in five scenarios. While the usage rate of the 36 runways used as landing and take-off runways decreased in seven scenarios, the usage rate was increased in one scenario. When the tables of runway usage frequency and rates are examined, the proposed model makes suggestions to improve the use of runway 34L, which is used only for landing, to eliminate the use of runway 34R, which is used only for take-off, to increase the use of runway 35L, which is used only for take-off, and to reduce the use of runway 36, which is used for both landing and take-off. It is seen that the use of the 34L runway more for landing and the more preferred use of the 35L runway for take-off will improve fuel consumption. The total fuel consumption of landing and take-off aircraft and the total fuel consumption values in each scenario are given in Table 11.

Table 11. Total fuel consumption values

The proposed model reduced overall fuel consumption in all scenarios. The improvement in the total fuel consumption of the landing traffic is relatively higher than the improvement in the total fuel consumption of the departure traffic. The presented model showed an improvement between 6.6% and 14.4%. The table shows that the most significant improvement was achieved in the first scenario, with 14.4%.