Comparative Review of ICAO and EUROCONTROL Flight Carbon Emission Approximators

Abstract

While airlines can directly quantify carbon emissions based on flight-specific fuel burn data, such data, along with data on other gaseous emissions that do not scale linearly with fuel consumption, are often unavailable to external stakeholders, necessitating the reliance on estimation models. Emissions are thus approximated from known quantities, with most usually from the fuel burned and distance travelled. Emission approximators developed for the aviation industry thus involve some degree of approximation and assumptions, as well as different exogenous and endogenous factors. As a result, such solutions differ primarily due to the significant methodological variations they incorporate. This paper assesses carbon emission approximators developed to valorize emissions generated by flight operations. It reveals the significance and sources of the misestimation of emissions by focusing on the ICAO Carbon Emission Calculator (ICEC), ICAO CORSIA CO₂ Estimation and Reporting Tool (CERT) and EUROCONTROL’ Advanced Emission Model (AEM) and Small Emitters Tool (SET). Thereby, the main research findings indicate considerable estimation uncertainty among the reviewed solutions, ranging from 1.77% to 27.95% on average compared to the baseline, which translates to statistical confidence levels ranging from 15% to 77.50% on average with respect to a 95% confidence threshold.

1. Introduction

The aviation industry is a key driver of the global economy, supporting over a third of world trade shipments by value and 65.5 million jobs [1]. Connecting businesses around the world, the aviation industry contributes to global economic wealth development. In addition, connecting people in a way that a hundred years ago people could only have dreamed of, the aviation industry contributes to better mutual understanding and appreciation of various cultures and traditions. However, like all other transportation modes, air transport has an adverse effect on the environment and climate change. Accordingly, despite the enormous benefits it brings to society and the global economy, the role of aviation and its environmental implications are currently frequently the subjects of greater scrutiny by the general public, most notably in relation to carbon emissions [2,3].

The issue of measuring the carbon emissions of the aviation industry stems from several factors. The first factor is a public misunderstanding of the operational and organizational settings of the aviation industry, which largely contributes to the increase in populism [4] and the belittling of the efforts made by the aviation industry so far [5]. Therefore, the general public frequently does not perceive itself as a source of generation of carbon emissions, but rather aircraft operators and the industry as a whole. Another problem is the multiplicity of greenhouse gasses (GHGs) resulting from aircraft fuel burned, and the difficulty to compute the global warming potential of GHG emissions in the upper troposphere and lower stratosphere. This issue was raised back in 1999 [6], while currently there are more detailed studies in this regard [7,8,9]. The other main uncertainty lies in the approximation of the actual carbon emissions generated per distance travelled—representing the key metric used to estimate the carbon emissions of each horizontal flight profile.

Currently, carbon emission approximators are used by a range of decision-makers, including governments, non-government organizations, international trade bodies, etc., for international emissions reporting. Also, such tools are used by many businesses, mainly for corporate and public reporting purposes, thus promoting social awareness and responsibility. However, in recent years there has been an aspiration for greater consistency between approximators, as the plethora of solutions makes reporting inconsistent and confusing [10]. In that respect, this paper provides a review of four solutions developed to estimate gate-to-gate CO₂ emissions. It presents a comparative review of their approximations articulated in the sense of fuel burned per distance travelled and resulting environmental implications. In doing so, it addresses the research question of the applicability of four solutions developed by prominent aviation stakeholders, the ICAO and EUROCONTROL, by studying the influence of methodological discrepancies between tools on emission estimation. The main research objective was to determine and quantify differences in the performances among the solutions reviewed. The secondary objective was to approximate the repercussions of using these solutions within the domain of the strategic planning and development of the aviation industry on a global scale. Therefore, apart from bringing transparency regarding the adverse effect of air transport on the environment and climate change, the relevance of the research carried out also stems from the fact that reviewed solutions were developed by two highly respective stakeholders of the aviation industry. This consequently contributes to both the greater applicability of the reviewed solutions at the regional and global level, and to the greater relevance of the main research findings obtained within the research carried out.

2. Research Background

Total man-made emissions of carbon dioxide in 2019 were approximately 43.1 Gt. Like most business activities, directly or indirectly, the aviation industry produces carbon dioxide emissions. A comparative review of the global and aviation industry-generated CO₂ emissions during 2019 indicates that it accounts for around 2.1% of CO₂ emissions globally [11]. And the level of aviation’s CO₂ emissions has remained at around 2% of total global emissions for years [12], primarily because the efficiency of aircraft engines is improving continuously, and a flight taken currently produces half the CO₂ compared to flights in the 1990s [13]. On the other hand, the industry is also growing rapidly to meet the demand for air travel, which consequently leads to a rise in overall emissions [14]. Moreover, 80% of aviation emissions were generated by flights over 1500 km, i.e., at a distance where for passengers, there are a lack of alternatives in the forms of utilizing other transportation modes [15]. Most of the aviation-generated emissions, around 62%, occur during the en-route flight phase [16]. A comparative review of 2019 data also points out that aviation generates less than the shipping sector and around the same as the servers and transmission cables of the Internet (not including the computers and tablets accessing the Internet) [17]. Furthermore, if the aviation industry were a country, according to 2019 data [13] it would be the 11th or 12th largest emitter of carbon emissions. However, despite these numbers, the increasing number of complaints correlating the aviation industry with air pollution, poor air quality and noise in areas near airports, combined with the rise in concentrations of fine (PM_2.5) and coarse (PM₁₀) particles, currently represents a growing source of public concern [18].

With a goal to contribute to the minimization of its environmental footprint, on a global level, the aviation industry strives for decarbonization through four pillars of climate action, including the development and implementation of new technologies, substitute fuels, optimization in the domain of aircraft operations, and infrastructure improvements [19,20,21]. The industry also strives towards the reduction of its environmental impact by offsetting carbon emissions. For instance, offsetting carbon dioxide emissions has been included in the European Union Emission Trading System (EU ETS) since 2012 [22]. Under the EU ETS, airlines are required to monitor, report and verify their emissions and to acknowledge allowances against those emissions [23]. On a global scale, the ICAO has also introduced a scheme for offsetting carbon dioxide emissions, named the Carbon Offsetting and Reduction Scheme for International Aviation (CORSIA) [24]. It requires airlines to offset their emissions from routes in the scheme by purchasing emissions units generated by projects that reduce emissions in other sectors [25]. In addition, currently, airlines are progressively introducing so-called voluntary carbon offsets that are available for purchase by their passengers—giving them the opportunity to compensate for the emissions generated by their travel. For instance, in 2020, 44 airlines offered voluntary carbon offsetting programmes to their passengers. In other words, over 30% of the overall number of passengers had flown with an airline that offered some kind of carbon offsetting programme. However, the uptake by passengers was generally very low—as it had been recorded that only 1–3% of passengers had purchased offsets for their flights [26]. Therefore, since there is no consensus on the methodological framework that can be used for the approximation of CO₂ emissions on a global scale [27], the rhetorical question arises of what passengers are actually purchasing when voluntarily purchasing offsets to compensate for the carbon emissions generated by their flight.

In addition to industry-driven initiatives and regulations targeting decarbonization, the literature review indicates that there are few studies focusing on modelling and estimating carbon emissions within the transportation sector [28,29,30], with even fewer dedicated specifically to aviation [31,32]. It also indicates that the rapid development of artificial intelligence (AI) and significant advancements in its capabilities have profoundly transformed the process of carbon emission estimation. For instance, Masood and Ahmad [33] present an overview of the AI-based methods most widely used for air pollution prediction, also outlining the advantages and disadvantages of various AI-based methods. Similarly, Tesfai et al. [34] have conducted a comparative analysis of various forecasting models and evaluated the most effective AI-driven methodologies for estimating carbon emissions in the United Arab Emirates. A detailed overview of the AI-based methods and Machine learning (ML) algorithms used for forecasting various major pollutants is also provided in the study by Subramaniam et al. [35], emphasizing the better performance of a hybrid model over single AI models. Apart from overviews, there are also examples of utilizations of AI-driven solutions developed to provide approximations of carbon dioxide emissions. As such, the study by Nassef et al. [36] utilizes three artificial intelligence tools, a feed-forward neural network, an adaptive network-based fuzzy inference system and long short-term memory, to forecast the yearly amount of CO₂ emissions in Saudi Arabia. Alam et al. [37] employ deep learning techniques to develop a CO₂ emission prediction model, utilizing a lightweight multilayer perceptron architecture.

Considering the above information, although none of the reviewed solutions generates outputs using AI-based methods, given the progressive trend of AI deployment, it is expected that in the coming years, new solutions will likely emerge that leverage AI to accurately quantify carbon dioxide emissions from flight operations, especially as carbon emission forecasting has become a prominent topic in the field of artificial intelligence [38].

3. Flight Carbon Emission Approximators

Among other factors, carbon dioxide emissions are monitored to evaluate the environmental footprint of the aviation industry. Over the last two decades, many so-called carbon emission calculators were developed to assess the carbon dioxide emissions from flight operations. Currently, such tools are used by individuals, private companies, international organizations and non-profit organizations to estimate carbon inventories for offsetting purposes [39]. Frequently, these tools require minimum inputs from the user. Undoubtedly, that contributes to a better user experience and makes these solutions more user-friendly. Even though most of these solutions are presented as carbon emission “calculators”, including the ones studied and presented in further content, within this paper, the term “approximators” is used. According to the Cambridge English Dictionary [40], the term “approximation” denotes a guess of a number that is not exact but close. In that sense, from a methodological viewpoint, the output of a carbon emission calculator should provide the exact carbon emission amount for each flight. However, as practice shows, since insights on carbon dioxide emissions are often unavailable to external stakeholders, estimation models become necessary. Moreover, since the direct quantification of other pollutants is not possible, then consequently developed “calculators” are also not capable of providing realistic measurements—but rather approximations. In addition, the relevance of using the term “approximators” stems from the fact that all carbon emission “calculators” are developed based on the adoption of some degree of methodological assumptions, interpolation and coefficients—mainly used to convert fuel burned into a carbon footprint.

3.1. ICAO Carbon Emission Calculator

The International Civil Aviation Organization has developed environmental tools, available to states and the general public, to support initiatives to reduce the carbon footprint of the aviation industry. One such solution is the ICAO Carbon Emission Calculator (ICEC) [41]. Developed in the form of a web application, the ICEC allows interested parties to estimate the carbon dioxide emissions from a given flight for use in carbon offsetting programmes [42]. Since 2009, the ICEC has been used by the entire UN system for computing their annual air travel emission inventories in support of the UN Climate Neutral UN initiative. Yet since 2008, the general public has also had the ability to use the ICAO Carbon Emission Calculator. In fact, it has become the most popular ICAO tool being consulted daily by the travelling public [43].

From a methodological viewpoint, the framework integrated into the ICEC is similar to the approach taken by Montlaur et al. [44]. Concisely, they made an analytical model that calculates gate-to-gate emissions of carbon dioxide and travel time based on the flight distance and the number of available seats.

The methodological framework integrated into the ICEC was designed to minimize user input. Once the user specifies the origin and destination airports for a direct flight, the Flight Emission Estimation (FEE) is computed using the following equation:

$$FEE=3.157 \times (total fuel \ast PCF)/(y_{seats} \ast PLF)$$

(1)

where the constant 3.157 denotes the number of metric tons of CO₂ emitted per ton of jet fuel burned, while the total fuel represents the weighted average of fuel consumption across all flights serving the airport pair, based on the Great Circle Distance (GCD). The Passenger-to-Cargo Factor (PCF) accounts for the ratio between passengers and freight/mail carried. The Passenger Load Factor (PLF) marks the ratio of passengers transported to the available seat capacity, while yseats denotes the total number of available seats on flights operating between the specified airport pair [45].

This methodological framework integrates multiple data sources to support the estimation process. Airport pair data are sourced from the Airlines Multilateral Schedules Database (AMSD), and GCD values are obtained from the ICAO Location Indicators (LOCIND) database, while load factor data are derived from the Traffic by Flight Stage (TFS) database. Fuel consumption per equivalent aircraft is estimated using the ICAO Fuel Formula (IFF), while data on yseats is derived from the Manual on Airplane Characteristics for Airport Planning (MACAP). Figure 1 presents the conceptual framework underpinning the ICEC, illustrating the relationship between databases, computational components and the resulting flight emissions reported.

Observed from the aspect of its conceptual framework, user input is compared with the published scheduled flights to obtain the aircraft types used to serve the two airports concerned. Each aircraft is then mapped into one of the 312 equivalent aircraft types to calculate the fuel consumption [46]. The ICAO calculation of fuel consumption is based on the distance travelled, i.e., on the GCD between the specified airports [47]. The system then calculates the average fuel consumption for the journey weighted by the frequency of departure of each equivalent aircraft type [48]. Data obtained is then divided by the total number of passengers, giving an average fuel burn per class passenger. The estimate also accounts for different service classes (business and economic cabin class) [49]. Therefore, as Baumeister [50] specifies, the ICEC mainly relies on average data within its data processing framework. The 2021 version of the ICEC was used within this study.

3.2. ICAO CORSIA CO₂ Estimation and Reporting Tool

Emissions of carbon dioxide are mostly approximated by converting the amount of fuel burned during every flight operation. However, in certain circumstances, aircraft operators may use simplified monitoring and reporting options. As such, operators eligible for simplified monitoring are not required to monitor actual fuel use but can calculate their emissions using the ICAO CORSIA CO₂ Estimation and Reporting Tool (CERT), a tool that was developed by the ICAO [51]. According to Annex 16, vol. IV, Part II, 2.2.1.2 [52], during the baseline period (2019–2020), aircraft operators with annual CO₂ emissions (from all international flights) below 500,000 tons were eligible for simplified monitoring. Following Annex 16, vol. IV, Part II, 2.2.1.3., from 2021, an aircraft operator may use the CERT for flights not subject to offsetting requirements, and for flights subject to offsetting requirements, if the total annual CO₂ emissions from these flights is below 50,000 tons [53]. If an airline’s CO₂ emissions increase above the 50,000 ton threshold for two consecutive years, it will no longer be eligible to use the CERT. Accordingly, the use of the CERT depends on the level of emissions generated—which represents the main eligibility criteria to use the CERT. Apart from aircraft operators, the administrating authority may also use the CERT, for instance, to estimate the typical fuel consumption for a flight and to compare it with the data reported by the aircraft operator [54].

The CERT, which comprises a Microsoft Excel spreadsheet, was developed to provide practical support to users to facilitate their compliance with the CORSIA monitoring, reporting and verification requirements [55]. In doing so, it allows for the monitoring of CO₂ emissions with minimum input requirements from the users. The underlying methodological framework requires users to input data related to the aircraft operator and flight profile, including the aircraft type and airport pair referenced using ICAO Doc 7910, or, alternatively, the flight operating time (i.e., block time). In this way, the tool supports both single-flight and aggregated data inputs. Based on these parameters, i.e., the application of CO₂ Estimation Models (CEMs), the tool approximates fuel burn as a function of either GCD or block time, depending on the selected estimation method [56,57].

The CEMs incorporate a set of variables drawn from multiple sources. For instance, aircraft-specific data are primarily sourced from ICAO Doc 8643. However, because it does not include Maximum Take-Off Mass (MTOM) data, supplementary sources such as the European Union Aviation Safety Agency (EASA) Aircraft Noise Certificate Database (ANCdb) and the U.S. Federal Aviation Administration (FAA) Type Certificate Data Sheets (TCDS) are used to map MTOM values to ICAO aircraft designators. The CERT also integrates the CCG Operations and Fuel Database (COFdb), which contains data on actual flight operations, including aircraft type, GCD, fuel burn, block time and the year of operation. Additionally, due to the commercial sensitivity of fuel consumption data, the COFdb is anonymized to prevent the identification of individual aircraft operators or specific flights. Also, the CERT methodological framework is reviewed and updated annually to expand the range of aircraft types supported, especially the new generation of aircrafts entering the market, and to ensure that the tool reflects evolving performance characteristics and operational settings. In that respect, for the purposes of this study, the 2021 version of the ICAO CORSIA CERT was employed. Its conceptual framework, illustrating the relationship between databases and computational components, is presented in Figure 2.

3.3. Advanced Emission Model

The Advanced Emission Model (AEM) is a standalone application, developed and maintained by EUROCONTROL, designed to approximate aircraft emissions through the advanced modelling of fuel consumption. It determines the fuel consumption of a specific aircraft type that is equipped with a defined type of engine, operating along a designated four-dimensional (4D) trajectory [58]. It successfully passed ICAO’s stress tests in 2008–2009 [59,60] and has since become part of the approved suite of assessment models used by ICAO’s Committee on Aviation Environmental Protection (CAEP) [61].

Unlike other reviewed solutions, it also reports approximations on water vapour (H₂O), the oxides of nitrogen (NOx) and sulphur (SOx), unburned hydrocarbons (HC), carbon monoxide (CO), volatile organic compounds (VOCs), and other organic gases (OGs).

The methodological framework supporting the AEM relies on modular configuration. For flight operations below 3000 feet, it relies on ICAO Engine Certification specifications, where fuel burn calculations are based on the Landing and Take-off (LTO) cycle [62]. The LTO cycle, covering four distinct engine operating modes, is used to approximate fuel burn during six operational phases: taxi-out, taxi-in (idle), take-off, climb-out, approach, and landing (approach). For flight operations above 3000 feet, flight emissions are estimated based on aircraft engine indices, fuel flow rates and pollutant-specific emission factors, which are adapted to standardized atmospheric conditions at altitude. Each emission index (EI) represents the mass of a specific pollutant emitted per unit of fuel burned, expressed for a given engine thrust setting. The Flight Pollutant Estimation (FPE) for each thrust level is calculated using the following equation:

$${FPE}_{\left(\% of thrust\right)}={EI}_{(\% of thrust)}\times {FFref}_{ (\% of thrust)}$$

(2)

where $FPE$ denotes the pollutant emission rate (kg/s), $EI$ marks the emission index (kg of pollutant per kg of fuel burned), while $FFref$ represents the reference fuel flow rate (kg/s). This estimation process is supported through the use of pre-calculated aircraft performance tables and databases. For instance, the AEM uses Base of Aircraft Data (BADA) and the EUROCONTROL Flight Operations (FOPS) database, which provide aircraft-engine mappings and performance characteristics. Additionally, pollutant-specific emission indices are sourced from the ICAO Aircraft Engine Emissions Databank (AEED), while airport taxi time values are derived from the EUROCONTROL Central Office for Delay Analysis (CODA) database. In addition, the AEM implements atmospheric conditions at altitude by using Boeing Fuel Flow Method 2 (BFFM2) [63], which is described within Aerospace Information Report 5715 [64]. As it assumes static and idealized atmospheric conditions, the AEM introduces discrepancies when compared to real-world operations. Combined with static fuel burn rates, this affects the accuracy of the estimates, particularly in scenarios that are sensitive to variations in ambient conditions. In that respect, for the purposes of this study, the 2021 version of EUROCONTROL’s AEM was employed within this study. Its conceptual framework is presented in Figure 3.

3.4. Small Emitters Tool

Since the 1st of January 2010, the European Commission [65] has required aircraft operators to monitor and report their annual emissions of carbon dioxide. To facilitate the administrative workload of airlines operating a limited number of flights, such operators may use the Small Emitters Tool (SET) [66]. More precisely, the SET may be used by airlines that are categorized as “small emitters” (according to the definition in European Commission regulations related to Emissions Trading System), concerning their monitoring and reporting obligations pursuant to Article 14(3) of Directive 2003/87/EC [67] and Part 4 of Annex XIV to Decision 2018/2066/EC (the monitoring and reporting guidance) [68]. Also, the SET can be used by all aircraft operators, pursuant to Part 5 of Annex XIV to Decision 2007/589/EC to estimate the fuel consumption of particular flights covered by the EU ETS, where the data necessary to monitor the emissions of carbon dioxide are missing as a result of circumstances beyond the control of the aircraft operator and which cannot be determined by an alternative method defined in the operator’s monitoring plan [69].

The SET, developed by EUROCONTROL on behalf of the European Commission, supports CO₂ emission reporting under the EU Emissions Trading System (EU ETS). It is developed in the form of a Microsoft Excel-based tool that estimates fuel consumption and resulting emissions across all phases of flight. Its internal fuel burn and emissions models are founded on a statistical methodology, which utilizes real-world fuel burn data. Based on these fuel burn samples, the SET provides estimates of CO₂ emissions for any given distance and aircraft type [70]. Therefore, even though this data is collected from volunteer aircraft operators, the methodological framework and equations integrated into SET are proprietary, so attempts to replicate them will likely yield inconsistent results [71].

Fuel consumption estimates are generated through linear regression that is specific to each aircraft type in the sample to consider the fuel dependency from the distance flown [72]. Accordingly, to approximate emissions at the aircraft-operator level, it only requires user input on the aircraft type, the route distance and the number of flights.

In addition, the SET only considers a linear correlation between energy consumption and flight distance. This simplification is legitimate since most flights are long-haul flights where take-off and landing phases do not dominate the overall energy consumption of the whole flight [73]. Last but not least, to keep the SET up to date, EUROCONTROL regularly gathers and updates data samples on fuel burn with respect to flight profiles. Apart from that, annual updates of the SET also cover the introduction of a wide range of aircrafts [74]. For the purposes of this study, SET version 5.11, released on 20 December 2021, was employed. This version does not include any new features concerning previous versions, but the estimation model has been updated based on the latest data on the air traffic covered by the EU Emissions Trading System.

4. Main Research Determinants

A comparative review of carbon emission approximators has included the application of both quantitative and qualitative research approaches. The applicability of all solutions had been tested by their application, while their findings have been cross-checked. Accordingly, to facilitate a comparative review of carbon emission approximators, within this research, EUROCONTROL’s tool named the Network Strategic Tool (NEST) [75] was used. The NEST has been used primarily to extract the actual lengths of studied flight profiles. Furthermore, within the study, the 1st of August 2022 has been randomly selected as the reference period of study. Accordingly, the study and its conclusions are derived from a representative sample, after the processing of which, obtained findings are scaled to a greater observation scale. In terms of geographical determinants, the selection of Europe as a reference geographical scope of the study has contributed to the scaled evaluation of comparative findings.

Even though all reviewed solutions were developed to approximate carbon emissions from flight operations, neither of them take into account the impact of ground or en-route Air Traffic Flow Management (ATFM) delay. Hence, since all studied solutions omit the impact of ATFM delay on the overall generation of carbon emissions, the consideration of actual ATFM restrictions on flight operations has been excluded from the study to maintain consistency with the tools’ modelling approaches. There are several other assumptions which have not been considered within this study as they are also not reflected in the reviewed tools’ input requirements or algorithms. This includes meteorological conditions en-route, aircraft age, aircraft maintenance and payload factors, all of which may have an influence on the accuracy of the reviewed solutions. Therefore, these omissions do not reflect a limitation in study design, but rather a deliberate alignment with the scope and capabilities of the tools evaluated. In that regard,

Table 1. Overview of the basic differences among the reviewed tools.

Emission Approximators’ Elements		ICEC	CERT	AEM	SET	NEST
Operational data of actual flights	Horizontal flight profile	☓	☓	✓	☓	✓
	Vertical flight profile	☓	☓	✓	☓	✓
	Meteorological conditions	☓	☓	☓	☓	✓
	Air Traffic Control restrictions	☓	☓	✓	☓	✓
	Aircraft Maximum Take-Off Mass	☓	☓	☓	☓	✓
	Aircraft subtype differences	☓	☓	✓	☓	✓
	Aircraft fuel type loaded	☓	✓	✓	☓	☓
	Airspace (sector) availability	☓	☓	✓	☓	✓
	Number of passengers onboard	✓	☓	☓	☓	☓
	Flight classes differentiations	✓	☓	☓	☓	☓
Technical features	Integrated exporting option	☓	✓	✓	☓	✓
	Manual user input required	✓	☓	✓	✓	☓
	Integrated database	☓	☓	✓	☓	✓
	Geospatial visualizations	✓	✓	☓	☓	✓
	Modular framework	☓	☓	✓	☓	✓
	Web-based application	✓	☓	☓	☓	☓
	Reporting interoperability	☓	✓	✓	☓	✓

Symbol ✓ denotes an available and integrated feature, while ☓ indicates that the feature is not available or integrated within solution.

summarizes and highlights the basic distinctions between the reviewed tools, indicated by their operational and technical settings. Finally, to test their applicability within our study, the findings derived from carbon emission approximators were assessed from an airline-centric viewpoint. Taking into account actual data on fuel burned per airport pairs, the assessment process was complemented with insights into the statistical confidence of reviewed solutions. Accordingly, for the purpose of a comparative review, horizontal flight profiles of one regional airliner were analyzed. To comply with the Terms of Use of the reviewed solutions, the main research findings had to be aggregated, while specific information on aircraft operators and flight numbers had to be omitted.

5. Main Research Findings

Although reviewed solutions were developed to approximate carbon emissions, within the study, the main research findings have also been articulated in the form of cross-products that derive from aircraft fuel burn. In that respect, apart from the aircraft fuel burned and the generated carbon dioxide, the main research findings are complemented and interpreted in the sense of the operational cost of the fuel burned, carbon monoxide, unburned hydrocarbons, water vapour, nitrogen oxides, nitrous oxide, sulphur oxides, sulphur dioxide, particulate matter, black carbon, Non-Methane Volatile Organic Compounds (NMVOCs), and emissions of Total Organic Gases (TOGs).

On 1 August 2022, the studied airline operated 89 flights connecting 26 aerodromes across Europe. Figure 4 in continuation displays a spatial overview of horizontal flight profiles considered within the assessment, represented by red lines. Accordingly, a sample covering 89 flight profiles had been used for airline-centric assessment. Concerning an overall figure, 41.30% operated as a round trip, while the remaining operated in one direction. The most-visited aerodrome was the airline’s main hub with 28.26% in the overall distribution, followed by the other two regional airports with 21.74% and 6.52% in the overall distribution. The breakdown of studied flights per aircraft type indicates that 26 flights were operated by the Airbus A319-100 (A319), 14 by the Airbus A320-200 (A320) and 49 flights were operated by Dash 8-Q400 (DH8D). The shortest route length equalled 79.27 NM, while the longest route length equalled 1475.55 NM. The average route length of the studied sample equalled 414.34 NM, which corresponds to the average route length in Europe (415 NM) as defined by Reynolds [76]. The average flight duration equalled 1 h and 16 min. The overall distance flown by the airline on 1 August 2022 equalled 34,839.75 NM.

The main research findings point out a considerable range of approximated values. As such, approximations of the quantity of fuel burned by the airliner on the reference day vary between 200,788.72 kg to 278,698.40 kg, depending on the tool employed. A gap between 3628.28 kg and 77,909.68 kg in absolute terms, i.e., from 1.77% to 27.95% in relative terms, thus may be perceived as estimation uncertainty among reviewed solutions.

Table 2. Overview of relative and absolute differences in fuel burn estimation among solutions.

Fuel Burn Approximator		ICEC	CERT	AEM	SET
		Absolute Differences
ICEC	Relative differences		77,909.68 kg	62,047.42 kg	74,281.40 kg
CERT		27.95%		15,862.26 kg	3628.28 kg
AEM		22.26%	7.90%		12,233.98 kg
SET		26.65%	1.77%	5.99%

displays relative (dark blue) and absolute (light blue) differences among reviewed solutions.

A comparative analysis of fuel burn approximations across the reviewed solutions also reveals substantial variation in the derived products, particularly in terms of carbon dioxide emissions. The corresponding estimations of CO₂ emissions range from 9421.20 kg to 243,407.61 kg, depending on the solution applied.

Furthermore, the economic implications of differences in fuel quantity can be assessed by converting fuel burn discrepancies into monetary terms. Expressing emissions in monetary terms aims to bridge the gap between technical findings and broader policy-relevant and environmental considerations. However, unlike CO₂ emissions, the financial impact is subject to temporal variations due to fluctuations in fuel prices and the associated price elasticity of demand. For instance, the cost of fuel ranged from 0.31 EUR/kg in 2014 to as high as 1.00 EUR/kg in 2022 [77,78]. To provide a consistent basis for evaluating financial effects of estimation error, this study employs the average fuel price from 2019, amounting to 0.43 EUR/kg [79].

Table 3. Overview of financial and emission effects arising from fuel burn estimation discrepancies.

Fuel Burn Estimation Discrepancies		ICEC	CERT	AEM	SET
		Emission Effects
ICEC	financial effects		243,407.61 kg of CO2	193,282.86 kg of CO2	233,986.41 kg of CO2
CERT		33,501.16 EUR		50,124.74 kg of CO2	9421.20 kg of CO2
AEM		26,680.39 EUR	6820.77 EUR		40,703.54 kg of CO2
SET		31,941.00 EUR	1560.16 EUR	5260.61 EUR

presents an overview of the fuel burn estimation discrepancies across the reviewed solutions, expressed in terms of both CO₂ emissions and corresponding monetary effects (dark blue).

Estimation uncertainty among the reviewed solutions may be further elucidated by examining the correlation between fuel burn approximations and the corresponding flight distance. Considering the latter, a comparative review of research findings indicates that approximations made by ICEC are less dependent on changes in route length, and vice versa for approximations made by the CERT, AEM and the SET. In addition, for all studied flight profiles, the ICEC provided overestimated figures compared to other solutions. Figure 5 shows a consolidated overview of approximations made by reviewed solutions placed with respect to the route length of 89 horizontal flight profiles. These profiles served as a test sample, and thus are referred to as the flight testing sample. Therefore, the flights reviewed have been listed by flight length, from shortest to longest flight in terms of distance flown.

As expected, the correlation between route length and the approximation of fuel burned indicates that as flight distance increases, fuel burn also increases. More importantly, the findings indicate that the rate of increase is not uniform across reviewed solutions. A review of the result distribution by means of a linear regression $y=ax+b$, where $y$ marks fuel burned in kilograms, $x$ denotes route length in Nautical Miles (NM), while $a$ and $b$ represent the slope and the intercept, indicates that, on average, the ICEC considers the longest unit route length segment to be 70.533 NM. Simultaneously, about 80.2% (R² = 0.802) of the variability in ICEC fuel burn is explained by route distance alone. This also points out the existence of a strong linear relationship between distance and fuel burn, meaning that it is more sensitive to increases in route length.

Furthermore, findings indicate that in events when different aircraft types have served the same airport pairs, approximations made by the ICEC did not consider the change in aircraft types. More precisely, approximations made by the ICEC were mostly erroneous, both in the frequency of event occurrence and in the scale of assessment inaccuracy, mostly for flight profiles that were operated by Dash 8-Q400. In absolute numbers, the inaccuracy ranged from a minimum of 2.90 NM up to a maximum of 3136.30 NM for DH8D. The misestimation for A320 was within a range from 114.20 NM to 2470.90 NM, while for A319 it ranged from 17.18 NM to 1538.30 NM. As a result, this had led to erroneous approximations of fuel burned, and consequently to differences between approximations made by the ICEC and those made by other reviewed solutions.

To address estimation uncertainty among the reviewed solutions, approximations they produced were benchmarked against actual data obtained by reviewing the fuel burn of 17,493 flights. As a result, this enhances the interpretability of the results by shifting the emphasis from estimation uncertainty to a measurable estimation error. Figure 6 illustrates the distribution of minimum, average, and maximum fuel consumption across 67 unique airport pairs, along with the corresponding flight sample size for each pair.

To quantify estimation errors, the Root Mean Square Error (RMSE) was employed, calculated as follows: $RMSE=\sqrt{{\displaystyle \frac{1}{n}}\sum _{i=1}^n{\left(y_i-x_i\right)}^2}$, where $y_i$ represents the actual value of the fuel burn for the ith flight profile, $x_i$ denotes the approximated fuel burn for the ith flight profile and n is the sample size.

Research findings in this context outline that the RMSE associated with the ICEC (1038.36 kg) is significantly higher than the estimation errors of other solutions. The RMSE of the ICEC’s approximations is 5.57 times higher than that of the SET, 4.98 times higher than that of the CERT and 3.83 times higher than that of the AEM. Accordingly, the ICEC provides a notably lower accuracy compared to other solutions. Figure 7 illustrates the differences in fuel burn, represented by the absolute error, with data plotted in order of increasing flight distance, from the shortest to the longest.

A review of outputs classified as either overestimated or underestimated approximations indicates that the ICEC (94.12%) and the AEM (81.18%) provide mostly overestimated figures, with a small proportion being underestimated (5.88% and 18.82%, respectively). In contrast, the CERT predominantly produces underestimated figures, accounting for 57.65% of the data distribution. Ultimately, the SET has a relatively balanced distribution between overestimated (57.65%) and underestimated approximations (42.35%). Figure 8 displays four histograms which provide a summary of the distribution of average errors concerning flight distance.

A review of the statistical deviation between the approximated and actual average fuel consumption figures, applying a 95% confidence level for the standard normal distribution, indicates significant differences in the z-scores of the outputs from the reviewed solutions. As such, the SET’s and the CERT’s z-scores exhibit a high degree of normality, with 93.75% and 83.75% of estimates falling within the confidence interval, suggesting greater stability. In contrast, the z-scores for the ICEC and AEM show more variability, with only 17.5% and 78.75% of estimates within the confidence interval, respectively.

In addition to the view of statistical deviation, the obtained approximations were statistically validated against the 95% confidence level, denoted by two-tail p-values. A review of approximations which are within 95% confidence indicate that the ICEC has the strongest statistical significance, with 85.00% of its values falling below the confidence threshold of p < 0.05. In contrast, the CERT, AEM and SET demonstrate higher variability, with only 22.50%, 23.75%, and 31.25% of their p-values falling below the confidence threshold, respectively. This suggests a lower degree of statistical confidence of their approximations on fuel burned. Figure 9 displays the distribution of statistical significance and confidence in fuel burn approximations.

The application of basic methods from differential statistics offers insight into the distribution and variability of confidence in the outputs reported by the reviewed solutions.

Table 4. Distribution overview of z-scores of fuel burn approximations per solution.

z-Score	μ	M	σ	∧	∨	[∧, ∨]
ICEC	6.25493	5.20704	5.14136	−2.94756	20.40253	23.35010
CERT	0.21417	0.28498	1.09913	−2.31372	4.80971	7.12342
AEM	−0.22499	−0.31785	1.21140	−3.79927	4.54503	8.34429
SET	1.25096	1.15621	1.29167	−1.19127	6.01357	7.20484

and

Table 5. Distribution overview of two-tailed p-values for fuel burn approximations per solution.

p-Value	μ	M	σ	∧	∨	[∧, ∨]
ICEC	0.06545	0.00000	0.17063	0.00000	0.81729	0.81729
CERT	0.47043	0.43653	0.26765	1.51152 × 10−6	0.96991	0.96991
AEM	0.48093	0.52192	0.30820	5.49281 × 10−6	0.97148	0.97148
SET	0.31833	0.24016	0.29153	1.81478 × 10−9	0.98564	0.98564

present a statistical review of data, denoting the confidence of fuel burn approximations reported per reviewed solution, where $\mu$ represents the average value, $M$ marks the median value, $\mathsf{\sigma }$ marks the standard deviation, $\wedge$ denotes the minimum value, and $\vee$ marks the maximum value, while $[ \wedge , \vee ]$ denotes the range of the observed data set.

Considering the presented research findings, the CERT demonstrates the most balanced and accurate performance. It shows a low mean z-score (μ = 0.21417) and a median close to zero (M = 0.28498), with a relatively small standard deviation (σ = 1.09913), suggesting limited deviation from the expected values and consistent estimation behaviour. In contrast, the ICEC solution, despite having a high z-score mean and a broad confidence interval, exhibits extremely low p-values, suggesting significant deviations and lower estimation reliability. While the AEM and the SET show moderate performance, their z-score statistics reflect less consistency and higher estimation bias than the CERT. Hence, from a statistical accuracy perspective, the CERT outperforms the other solutions in approximating fuel burn and, consequently, in approximating carbon dioxide emissions.

6. Discussion

Due to the increase in situational awareness of ongoing climate change, currently, many efforts are being made to monitor, report and minimize the amount of carbon dioxide emissions. Among other generators, airlines are one of those facing mounting pressure from the general public, governments and media to reduce their fast-growing CO₂ emissions [80]. However, airlines are not solely responsible. For instance, due to the lack of ATC capacity in Europe during the summer of 2018, airlines had to frequently fly at suboptimal Flight Levels (FLs) [81]. As a result, that led to greater fuel consumption and consequently additional CO₂ generation. Moreover, apart from the FL requested and the one operated at, the emissions of each flight also depend on the aircraft type operated, the distance covered, and other operational parameters, including aircraft weight and the vertical and speed profile (known as Cost Index), weather conditions (e.g., wind lift), etc. Since direct and exact measurement of emissions is not possible, in recent years there has been the widespread development and application of carbon approximators that are used to provide approximations of the CO₂ emitted by airlines.

Currently, various carbon approximators are simultaneously used by governments and their agencies, international organizations and NGOs, trade bodies and carbon offset companies for international emission reporting. However, rapid development followed by the slight validation and verification of carbon emission approximators, unfortunately, has led to inconsistency between approximators as no two methodologies are identical. This point aligns with the research findings obtained within this study.

When contrasting the conceptual and methodological frameworks of reviewed solutions, the ICEC presents the advantage with respect to other reviewed solutions as it is sensitive to the number of available seats, a parameter generally not explicitly considered. Models that correlate emissions generated, distance flown and passengers transported are crucial for a deeper understanding of performance trade-offs. Such solutions also provide airlines with information that can support their network planning tasks, particularly for short- and medium-haul flights. Moreover, except for information about the number of passengers onboard, all reviewed solutions can be further updated with information on airline-specific key performance indicators.

Aside from its advantages, the main research findings outline that the ICEC provides overestimated findings with respect to other reviewed solutions. The first reason for this is because it does not take into account the accurate distance flown between the origin and the destination. The second reason is the application of the weighted average of total fuel burned (based on the frequencies of the scheduled flights) for a few generic types of aircraft instead of the specific aircraft that is operating the actual flight. As a result, it provides an inaccurate output on fuel burned per flight which leads to difficulties in assessing the actual impact of a given flight. Also, the ICEC only considers the larger aircraft series but does not consider the difference between the aircraft sub-series [82].

Furthermore, the main research findings point out that the ICEC is not able to compute the effects of non-scheduled passenger and cargo flights. Cui and Li [83] suggest that this is due to the fact that the ICEC has its limitations, such as incomplete aircraft coverage and incomplete applicability to routes within the specific country. As a result, within the conducted study, 3.26% of overall studied flight profiles, mostly including a few charters, cargo, general and business aviation flights, were not included in the conducted study, despite the fact that these flights did take place.

Fuel combustion follows stoichiometric principles [84], meaning that CO₂ emissions directly correlate with fuel burn. Hence, the difference in approximations does not stem from the conversion of fuel burned into carbon emissions generated, but from the approximation of fuel burned. Within studied solutions, aircraft fuel burn is determined based on the distance travelled and the aircraft operated. In that respect, findings indicate the differences in several aspects, including the difference between the actual and assumed distance, the difference between the actual and assumed aircraft operated, and consequently the difference between the actual and assumed amount of fuel burned.

Apart from conceptual and methodological frameworks, differences between reviewed solutions also include the use of different technologies, access options, and other legal restrictions. For instance, while other solutions may be used free of charge on a large scale, the ICEC can be used in the form of a paid service (through the use of the ICAO Application Programming Interface (API)). Furthermore, tools developed by the ICAO do not require user input on the distance flown between aerodromes, and this is also true for tools developed by the EUROCONTROL. To obtain information on carbon emissions via the ICEC and the CERT, users do not need to use any auxiliary solutions or sources, and this is also true for other solutions. In addition, the CERT provided 7.01% and 6.31% errors in distance flown with respect to the actually flown distance of the overall studied flight samples. Therefore, a comparative review of benchmarked flight samples points out that the error equals 7.28%. Also, the ICEC does not provide any information on the distance flown as part of its outputs.

Through the use of statistical inference, the obtained findings were scaled from the regional to the global level, i.e., from the sample to the population scale. Statistical inference is the process of drawing conclusions about an underlying population based on a sample or subset of the data [85]. For that purpose, apart from already addressed main research determinants, a normal distribution of the following indicators was assumed: the cumulative aircraft distance flown, the number of aircraft departures, the number of passengers carried and the average Passenger Load Factor (PLF).

Table 6. Breakdown of the main performance effects of air transport reported per ICAO region [86].

ICAO Region	Distance Flown [NM]	Aircraft Departures	Passengers Carried	PLF [%]
Europe	6,172,000,000	8,308,000	927,757,000	82.00
Africa	673,000,000	1,016,000	73,979,000	68.00
Middle East	1,651,000,000	1,228,000	186,705,000	76.00
Asia and Pacific	7,532,000,000	9,788,000	1,205,703,000	79.00
North America	7,370,000,000	10,817,000	878,458,000	84.00
Latin America and the Caribbean	1,557,000,000	2,859,000	260,172,000	79.00
Total	24,955,000,000	34,017,000	3,532,774,000	80.00

presents an overview of key performance indicators of air transport, categorized by ICAO regions. These regional baselines, combined with the main research findings, were used to scale regional insights to a global level. However, it is important to emphasize that the scaling process is inherently dependent on these baselines and assumptions adopted in the study, and it cannot fully compensate for the limitations arising from operational patterns and other differences among ICAO regions, which ultimately affects the accuracy of global-scale estimates.

Furthermore, to scale the regional to the global effects, it was first necessary to standardize them, thereby enabling their comparison. Since all indicators used within the scaling process were reported on an annual level, there was no need for temporal scaling among indicators. However, it was necessary to temporally scale the reference date from its representation of a typical traffic volume during the summer period with respect to annual traffic distribution. Furthermore, since the main research findings address insights referring to the traffic distribution in Europe, and since there are differences in traffic volumes among ICAO regions, these findings had to be further processed to obtain spatially scaled insights shifted from a regional to a global level. In that regard, Figure 10 displays normalized distributions which have been considered within the process of scaling research findings from the regional to the global level. The left-hand side illustrates the normalized temporal distribution considered within the scaling process, with the reference date indicated in relation to the overall distribution. On the right-hand side, the visualization depicts spatial normalization where the main performance effects of air transport in Europe are highlighted with dashed lines.

Table 7. Comparative overview of cumulative differences in population study and effects.

	ICEC	CERT	AEM	SET
Operational cost (bn EUR)	85.84	61.84	66.73	62.96
Fuel burn (Mt)	199.63	143.82	155.19	146.42
Carbon dioxide (Mt)	630.23	454.04	489.93	462.25
Carbon monoxide (Mt)	0.12	0.09	0.10	0.09
Hydrocarbons (Mt)	0.06	0.04	0.05	0.04
Water vapour (Mt)	245.74	177.04	191.97	180.24
Nitrogen oxides (Mt)	3.02	2.18	2.30	2.22
Nitrous oxide (Mt)	0.02	0.01	0.01	0.01
Sulphur oxides (Mt)	0.24	0.17	0.13	0.18
Sulphur dioxide (Mt)	0.17	0.12	0.13	0.12
Particulate matter (Mt)	0.01	0.01	0.01	0.01
Black carbon (Mt)	0.01	~ 0	~ 0	~ 0
NMVOCs (Mt)	0.23	0.17	0.18	0.17
TOGs (Mt)	0.20	0.15	0.16	0.15

concisely presents the main research findings. By scaling outputs from the regional to the global level, aggregated research findings point out a difference ranging from 74.88 to 130.69 Mt in fuel burned, 235.88 to 411.67 Mt in CO₂ emitted or EUR 23.21 to 40.51 bn additionally spent by the aircraft operators. In addition to differing from each other, the figures referenced also differ from other values referenced within the literature review. For instance, approximations of carbon dioxide emissions are between 21.27% and 43.28% lower, with an average of 33.41%, than 800.48 Mt, a value earlier reported for 2022 on this merit [87], mainly because all reviewed solutions approximate flight emissions based on gate-to-gate operations, and do not consider emissions generated by supporting flight activities or emissions generated by other stakeholders of the aviation industry. Therefore, one should not be surprised by the disparity identified given that neither of the reviewed solutions were developed for large-scale approximations. Moreover, for emissions reported in 2022, the UK Department for Transport identified differences between the findings of the Department for Energy Security and Net Zero (DESNZ), OECD and EUROCONTROL. Comparisons showed that approximations from the latter two were 11% and 6% lower, respectively, than those from DESNZ [88]. All of the above may also be perceived as one of the reasons why the European Commission initiated the Flight Emissions Label (FEL), a regulation aimed at addressing the representativeness and transparency of existing carbon emission approximators to provide passengers with clear information about flight emissions [89].

Finally, concerning the practical implications of the identified discrepancies in emissions estimates, they may be perceived from two standpoints. On the one hand, discrepancies identified do not directly affect airlines as carbon offsetting programmes mostly consider the actual fuel burned as reported by airlines. Accordingly, the actual consumption is monetized rather than estimated. On the other hand, discrepancies identified do indirectly affect airlines as all those who are not included in the source of information, i.e., all those who have no insight into actual figures, may be deceived. Combined with progressive populism and straw man argumentations, this in return may contribute to the development of novel, more restrictive politics and agendas, and as such reduce already thin profit margins of airlines. In the long run, this may result in the elimination of weaker competitors from the market and prevent new entrants into the market—in favour of flag carriers and financially better-off low-cost carriers. Ultimately, to offset additional expenses, airlines—regardless of their business model—will transfer these costs to passengers through ticket prices, resulting in higher fares for end-users. Hence, there is a clear need for greater transparency and accuracy in emission reporting to ensure fair market competition and to mitigate unintended economic consequences that threaten further market development.

7. Conclusions

The relevance of carbon emission approximators developed to support the decision-making processes increases with the increase in public interest in the issues concerning the environmental implications of the aviation industry. Since decisions made through the utilization of carbon emission approximators can have significant economic repercussions on all aviation-related stakeholders, including but not limited to aircraft and airport operators, it is crucial to ensure that the information supporting decision-making processes is highly reliable. Nevertheless, it is a frequently omitted fact that such solutions are based on the integration of methodological frameworks which often involve some degree of methodological approximation and assumption.

The research performed included a review of four carbon emission approximators. A comparative review of the main research findings indicates that the CERT, AEM and SET provide slightly more accurate approximations of fuel burned per flight than the ICEC does. Even though the ICAO has developed an API for the ICEC in the form of a paid service, allowing automatic integration with other software, web or mobile applications, the main research findings point out that the ICEC does not distinguish the result as a function of the aircraft type. This consequently leads to differences in approximations of carbon emissions generated by every flight. Considering that all relevant publications within the aviation industry apply a 95% confidence level for data processing to meet accuracy requirements, any such publication correlating the aviation industry and its environmental implications from the aspect of any stakeholder that may qualify to use the ICEC should not use it for operational planning nor as the support to strategic planning and development, regardless of the observation area and time scale.

This research confirms that currently there are solutions which have been developed to be highly user-friendly but at the expense of information accuracy. As a result, those who are less familiar with the settings of the aviation industry, the relevant inputs and the incomplete methodological framework behind the solution can use it to generate widespread misleading and inaccurate information on the impact of the aviation industry on the environment, climate change and society in general. The prominence of this taking place is downgrading and marginalizing of all the achieved effects thus far, followed by belittling the relevancy of the aviation industry to the global economy and its further aspirations for the development of a sustainable business environment. Moreover, the difference in approximations in a range from 1.77% to 27.95% in relative terms and from 22.50% to 85.00% in terms of accuracy on average calls out the need and urgency for standardization and further improvement in the field of study.

Considering the improvement areas, it has been demonstrated that the reviewed solutions can be further upgraded so that they provide information on other emissions, in addition to carbon dioxide. Also, additional efforts can be made in the areas of making computation processes more sophisticated, so that any change in user inputs is reflected in a follow-up change in the approximation of the carbon footprint.

To sum up, it has been demonstrated that the reviewed solutions are limited to operational emissions generated by aircraft operators. Although this includes en-route and ground phases of flight operations, emissions related to airport facility operation, ground handling, a lack of aerodrome or Air Traffic Control capacity, aircraft manufacturing or maintenance are not being considered. Hence, the reviewed tools should be framed as limited approximators, not comprehensive calculators. Accordingly, as long as solutions such as those reviewed here do not integrate the above variables into their computation framework, it is not relevant to use any of the reviewed solutions to approximate the overall implications of the aviation industry in the context of carbon dioxide emissions. In addition, within future works, regardless of their geographical scope, special emphasis should be placed on comprehensive verification and validation, including the transparent and strict review of conceptual and methodological frameworks integrated into carbon emission approximators that are being developed with the intent to monitor, assess and valorise the environmental impacts of the aviation industry.