Feasibility of climate-optimized air traffic routing for trans-Atlantic flights

To examine the relationship between changes in the climate impact and changes in costs for routing options, we apply a four step procedure. First, for each case study day, radiative forcings resulting from locally-confined unit emissions over the north Atlantic are calculated (section 2.2) by employing the detailed chemistry-climate model EMAC (Jöckel et al 2010, 2016) and then globally averaged. Second, the climate change induced by these emissions is calculated by applying emission metrics to these RF, resulting in climate-change functions (section 2.3, referred to in previous publications as climate-cost functions). The emissions represent a change in the routing strategy towards climate-optimized routing. Third, these CCFs are input to an air traffic simulator (section 2.4). For each flight lateral and vertical variations in the flight profile are considered, resulting in a number of possible flight paths for each flight. The climate impact of each possible flight path of the roughly 800 trans-Atlantic flights (figure 1) is calculated from the CCFs; in addition the cost of each flight is also considered. Finally, for this large set of possible traffic realizations, the optimal relation between climate impact reduction and cost increase relative to the minimum cost situation is derived by successively replacing the flight trajectories with the highest ratio of climate impact reduction versus costs increase. Details of the underlying methodology have been reported previously (Grewe et al 2014a) as have results for one individual case study day in winter (Grewe et al 2014b).

Zoom In Zoom Out Reset image size

The climate impact of local emissions, i.e. the climate-change functions, are determined for five winter and three summer days, which represent frequently-occurring North-Atlantic winter and summer weather patterns (WP and SP) using the classification of Irvine et al (2013). They are classified by their similarity to the North Atlantic Oscillation and East Atlantic teleconnection patterns, which describe main circulation patterns and impact the location and strength of the jet stream. For example WP1 is characterized by a strong zonal jet stream, whereas WP3 is a blocking situation. Winter weather patterns 1–4 each occur on average between 16%–18% of the time, whereas WP5 has the highest frequency of occurrence of 26%. The summer pattern 2 (SP2) is most frequently occurring weather pattern during summer with 60% of the time and SP1 and 3 occurring each 20% of the time.

For each weather pattern, one day was selected, which best represents the location and strength of the jet stream of that weather pattern. For each of these days, unit amounts of nitrogen oxide, water vapor, and carbon dioxide are released at about 500 different points in the atmosphere. At each of these 500 points 50 air parcels are started in which chemical perturbations are calculated. In addition, a flight distance is accounted for in every air parcel to simulate the effect of contrails. These emission points are located in an area over the north Atlantic, at multiple levels which cover the main North Atlantic flight levels, and the unit emissions occur at multiple times during the day. The processes simulated in the air parcels include effects from emissions of CO₂, H₂O, NO_x, and the formation of contrail-cirrus. Emitted nitrogen oxides are converted to HNO₃, which is rained out depending on the simulated cloud physics. HNO₃ can also be reconverted into NO_x. The nitrogen oxides concentration in the air parcel further affect ozone production via the reaction of NO+HO₂ → NO₂ + OH, where NO₂ easily photolyses and the resulting O atom recombines with molecular oxygen (O₂) to form ozone. In addition the emitted NO_x impacts the OH to HO₂ ratio in favor of OH (see reaction above). The latter is important since the increased OH reduces the background methane concentration. The same effect also results from the produced ozone, which is destroyed by photolysis. The resulting atomic oxygen reacts with water vapor to form OH, which again leads to methane loss. Water vapor is, in a similar way to NO_x, emitted into the air parcels, which are transported with the winds in the atmosphere. Whenever precipitation occurs, the water vapor in the air parcels is rained out. Contrails form within the air parcels whenever the Schmidt-Appleman criterion is fulfilled. They are persistent if the air is saturated with respect to ice. Persistent contrails increase the cirrus coverage and, depending on the available water vapor and the prevailing temperature, the contrail coverage and contrail ice water content can grow due to water vapor uptake or shrink due to sublimation. Contrail ice particles sediment and sublimate whenever the air parcel is transported to warmer atmospheric layers. Wind shear acts on the contrails and results in a spreading of the contrail and hence increases the contrail coverage. The contrail coverage, however, is limited by the potential contrail coverage, which is the fraction of an EMAC grid box, which can maximally be covered by contrails. All regarded quantities are transformed from the air parcels onto the EMAC grid, where radiation changes are calculated, and then globally averaged. The resulting changes in the concentrations of—and RF from—CO₂, ozone, methane, water vapor and contrail-cirrus then provide the relationship between a local emission and the resulting global-average RF. Further details of this modelling approach are reported in Grewe et al (2014a).

The simulation of chemical and contrail effects were validated in Grewe et al (2014a) by analyzing key parameters: The temporal course of the response of nitrogen oxide, ozone and methane to an initial unit emission agrees well with results from Stevenson et al (2004). The frequency distribution of ice water content of the simulated contrails agrees well with results from in situ measurements by Voigt et al (2011), whereas the frequency of optically thin (with respect to the visible spectral range) contrails is larger in our simulation. The specific radiative forcing of contrails has been verified with the Myhre et al (2009) benchmark test, which determines the RF for a 1% cirrus increase at 11 km with an optical depth of 0.3. The results fall well within the range of other model results.

To choose an appropriate climate metric we pose the question, what potential reduction in climate impact could be achieved by steadily applying a climate optimizing aircraft routing strategy, especially in the next few decades? From this objective we derive an adequate climate metric (Grewe and Dahlmann 2015). We consider a business-as-usual future air traffic scenario as a reference and compare that to a scenario where we daily fly trans-Atlantic routings with a low climate impact. We use the global and temporal average near-surface temperature response over 20 years after introducing the climate-optimized routing strategy. This metric enables the different climate relevant emissions to be placed on a common scale and thus be directly compared. Other metrics, which are suitable to assess a continuous change in the routing strategy, were investigated without significantly altering the conclusions (Grewe et al 2014b). We note that had we adopted the more frequently used pulse-based metrics (e.g. Fuglestvedt et al 2010), we would have found much stronger sensitivity, and more contrast between the short- and long-lived emissions—however, these would not have been best suited to quantifying the sustained impact of a permanent change in routing strategy on near-term climate change, which is the aim here. Applying this metric to the calculated RF (section 2.2) we then obtain a relation between locally and temporarily specified emissions and the global-average impact on climate in terms of future temperature changes. We call these 4-D response patterns 'climate-change functions' (CCFs). They comprise, e.g. the contribution from NO_x emissions to the global-mean climate impact via ozone. The climate impact of NO_x emissions depends significantly on where they are emitted with higher impacts for emissions in the jet stream and lower values for emissions north of it (Grewe et al 2014b).

The CCFs are included in a state-of-the-art air traffic simulator (SAAM, Eurocontrol 2012). We evaluate 85 alternative routings (17 horizontal and 5 vertical) for each of the roughly 400 flights crossing the North Atlantic in either direction each day with respect to climate impact and costs (figure 1). This simulator includes wind effects on flight trajectories. Additionally necessary safety margins in terms of separation of individual flights are taken into account. The traffic simulation provides a huge number of flight combinations, each forming a set of around 800 flights. We calculate an optimal relation between climate impact reduction and cost increase (figure 2). Starting point for this calculation is the set of flight trajectories, which minimizes costs (lower right). Here we consider fuel and crew costs. Then we calculate for every alternative flight trajectory the ratio between climate impact reduction and cost increase and pick that with the largest ratio, i.e. the most eco-efficient trajectory change. This procedure is repeated as long as we arrive in the climate optimal situation (upper left). The costs (M€), presented here, refer to one day of operation and the climate impact reduction refers to mean temperature change over the next 20 years for a change in the routing strategy towards a climate impact reduction, relative to the mean temperature change from aviation of all summer and winter situations (weighted with their frequency of occurrence).

Zoom In Zoom Out Reset image size

To account for safety issues 4 further optimizations were performed along the line of optimal relations between climate impact reduction and cost increases, which include a separation of aircraft following standard safety margins. The results hardly differ, showing that over the Atlantic air traffic density is not? a limiting factor for climate-optimized routing (Grewe et al 2014b). But it may well be, that climate optimized routing and safety would be in conflict over more congested areas like Europe and north-east of the USA.

From the large set of possible combinations of flight options, the minimum cost and minimum climate impact combinations are derived as are the most cost-effective intermediate solutions (figure 2). Costs and climate impacts are presented with respect to the emissions from one day. Optimal climate-cost relations are then obtained by successively changing individual aircraft trajectories, which have the highest ratio of climate impact reduction to costs. This reduces the climate impact of the whole one day trans-Atlantic air traffic and increases the costs. Westbound (red) and eastbound (blue) climate-cost relations are given separately, because the tailwinds have a large impact on routings and hence on these relations. Numbers at the endpoint of each curve indicate the climate impact reduction (%) relative to the climate impact of aviation flown at minimum costs, separately for each weather pattern and flight direction (rather than relative to all summer and winter situations and both flight directions, as given on the x-axis); thus, a saving for an individual pattern may be high, but if that pattern occurs less frequently, it contributes less to the total gain.The daily maximum climate impact reduction for a specific weather pattern ranges between 5% and 56% compared to the climate impact of the fleet of aircraft flown at minimum economic costs in that respective weather pattern and flight direction. These reductions are associated with increases in costs ranging from 0.6 M€ to 2.8 M€ per day, which is around 3.5% to 11%, depending on the weather situation. Note that the climate impact of an individual flight is almost negligible. However, if these climate impact reductions were achieved every day the total climate impact of air traffic could be reduced significantly. A clear difference in the climate-cost relations between westbound and eastbound air traffic is found between the winter patterns WP1 and WP2 and the other patterns. WP1 and WP2 are characterized by a strong zonal jet stream. The eastbound traffic takes advantage of tailwinds within the jet stream. Leaving the jet stream implies large penalties on fuel demand, emissions and climate impact; hence eastbound winter flights show less potential to reduce climate impact. By contrast, for WP4, an intense blocking situation, where the direction of the jet stream is more northward or southward, the difference between east- and westbound traffic is smaller. The summer patterns show a larger variability and a larger potential to reduce the climate impact from aviation by climate-optimized routing. The weather situation SP1 with a fairly zonal jet stream is similar to WP1 and shows a large difference between the flight directions, also similar to WP1.

Most of the potential climate impact reductions can be achieved at relatively low additional costs resulting in a 10% decrease in climate impact for only a 1% cost increase when all weather situations with their climatological frequency of occurrence are taken into account (figure 3, top). The maximum climate impact reductions are around 20% and relate to the maximum climate impact reductions shown in figure 2 by adding the weighted averages of the winter and summer values and adding the results for both flight directions. This cost efficient reduction in the climate impact is mostly achieved by avoiding the formation of warming contrails and by producing cooling contrails. However, it is also important to include the effect from NO_x emissions, since they often counteract the climate gain from the contrails (Grewe et al 2014b).

Zoom In Zoom Out Reset image size

Such cost increases in the order of 1% may not be acceptable to aviation stakeholders. Overall operational cost is the principal driver for routing strategies. Under the current regulations there is no incentive to fly routes that minimize the total climate impact of CO₂ and non-CO₂ emissions. However, if a market-based measure were in place, which included the climate impact of contrail formation and NO_x-emissions, airlines would have an incentive to fly routes that minimize the total climate impact. Here, we use the total 'equivalent CO₂ emissions (eq. CO₂)' which is the amount of CO₂ that would produce the same averaged temperature response as the combination of CO₂ and non-CO₂ emissions over the 20 year period. We add the costs for CO₂ (currently ∽5–10 € per ton) to the operating costs (figure 3, bottom) and also consider an upper and lower limit of 2 and 25 € per ton of eq. CO₂ emission, respectively. Reducing the climate impact by climate-optimized routing implies that lower eq. CO₂ costs are incurred, which reduces the operational costs. This results in a roughly 10% decrease in climate impact and a decrease of the cash operating costs by 5% for the trans-Atlantic air traffic.

Our results clearly show that under appropriate framework conditions and regulations, cost-effective climate-optimized routing has the potential to significantly reduce the climate impact from aviation. However, our study also raises questions and potential concerns, regarding the maturity level, scientific uncertainties, and also political and ethical questions. Clearly, our approach needs further development, for which we have formulated a road map (figure 4), before it can be used operationally. It includes systematic studies between the climate-change functions and weather systems to derive relationships between weather forecasts and the climate impact of a local emission. These algorithm-based CCFs and the climate impact reduction would need to be tested and verified. A feasible self-consistent way is to implement an air traffic simulator in an Earth-System Model and to optimize the climate reduction within this model (Yamashita et al 2016).

Zoom In Zoom Out Reset image size

Here, we concentrated on one specific region and showed that under current regulations safe climate—optimized routing is feasible. More studies are required to investigate the feasibility of climate beneficial routing and the impact on ATM over areas like Europe or America with a much larger traffic density.

The CCFs have to be mature enough to be used operationally. For example, we do know under which conditions contrails form (Schmidt 1941, Schumann 2000). But, are we able to predict the locations which support the formation of persistent contrails with sufficient accuracy? Even predicting those regions is not sufficient to determine the climate impact of contrails. This is because interactions between (micro-) physical processes, such as transport, dispersion, sedimentation, growth of particles, shape of contrail ice particles and their impact on radiation are important in determining the contrail RF values. The large atmospheric variability impacts these interactions and leads to large RF ranges. In addition, aircraft and especially fuel characteristics influence the contrail RF. For instance, the number of ice crystals formed initially is a strong function of the number of soot particles emitted (Kärcher and Yu 2009). Condensation of water vapor on soot surfaces proceeds on certain chemical compounds, so-called functional groups which due to their polarity can attach water molecules. Water soluble substances on the soot surfaces can greatly enhance absorption of water vapor. Sulfur can contribute to both the functional groups and the water soluble material, but according to laboratory studies (Popovicheva et al 2004, 2008, Demirdjian et al 2007) it is not the main constituent. Thus, we think, if sulfur-free fuel is burned, contrail formation will largely be unaffected.

In a case study we have analyzed the sensitivity of the cost-benefit analysis and related air traffic responses with respect to potential errors in the climate-change functions (Grewe et al 2014b). For example, we changed the weighting between the climate impact from NO_x and contrail-cirrus. In this case, the air traffic system still behaves similarly during the optimization. These sensitivity studies indicate a stable response in the shape of the cost benefit analyses and the way air traffic is routed for climate impact reduction. There are other non-CO₂ impacts not included in our present study, as they are much more uncertain, for example, the effect of aviation soot and sulphate aerosols on lower-altitude cloud characteristics. Recent studies (e.g. Gettleman and Chen 2013, Righi et al 2016, Kapadia et al 2016) indicate that these effects may be substantial, but also that there are large uncertainties in their quantification. We have enough information and knowledge about major atmospheric processes to start the implementation and demonstration in an operational environment. Once the uncertainties in the other non-CO₂ components are reduced, they could be incorporated within the same methodological framework as that presented here.

The different timescales of the impacts of CO₂ and contrails makes the comparison dependent on the chosen time horizon in the CCF (Fuglestvedt et al 2010). Here we defined a metric, reflecting a mean temperature change within a short time scale (20 yr) for implementing the climate-optimized routing on a daily basis. The results do not vary strongly for longer time horizons (100 yr). Different emission assumptions, especially pulse emissions, might have a strong impact on the results, since they affect the weighting between short-term, such as the less certain contrail effects, and long-term effects, such as the more certain CO₂ impacts. However, emission assumptions such as pulse emissions are not focusing on assessing the climate impact of a strategy change and hence are not appropriate to analyze our operational climate-optimized routing strategy.

There are open questions in ATM research and the financial and political framework. For example, what implications does climate-optimized routing have on ATM, e.g. the controller's work load? Which market-based measures are required? What kind of metric and what time horizon is most relevant for a pursuit climate target? How should the bookkeeping be performed for the individual contributions to climate change from non-CO₂ emissions?

In our approach, the routes which reduce the climate impact avoid regions where warming contrails are formed or the ozone impact is large. However, routes are also favored, where contrails contribute to cooling or the emission of NO_x leads to a methane reduction which cools more than the increase in ozone warms. This raises the question, to what extent should additional contrail formation be allowed, which—over a chosen time span—cools the global climate more than the additional CO₂ emitted by climate optimized-routing warms. These questions have to be considered carefully for any climate-optimized routing. In the light of increasing air traffic and the goal of climate policy to stabilize climate change below 2 °C, there is the need to mitigate aviation's climate impact. The approach and road map introduced here may allow a large reduction in aviation's climate impact in 20 yr' time using the current aviation system and can be complemented by advances in air frame and engine design.

The idea of avoiding climate sensitive regions was formulated previously (Sausen et al 1994, Mannstein et al 2005), but until now previous studies have only considered a trade-off between fuel consumption and contrails for a limited air traffic sample rather than the detailed analysis of all major aviation climate impacts for a complex air traffic sample presented here. Sridhar et al (2010) investigated the air traffic in the United States on 1 August 2007 and found that 53% of the contrails could have been avoided with an increase in fuel consumption of around 3%. For a more eco-efficient situation they found a reduction in contrail formation of 35% and a cost increase of 0.23%. In another study Sridhar et al (2011) analyzed the 24 May 2007 and flights between 12 city pairs and found that the time spent in contrail forming regions can be reduced by more than 70% associated with an increase in costs of 2%. Hartjes et al (2016) investigated fuel cost changes when optimizing a flight from Amsterdam to Washington DC and found an increase in costs of 1% for avoiding contrails formation. Schumann et al (2011) investigated the global fleet on the 6 June 2006 and calculated integrated radiation changes from contrails and CO₂ and found that by vertical shifts of individual flights the radiation changes can be reduced by 97% at an increase of costs of 0.2%. These results cannot directly be compared to our findings. However, they show the same tendency, i.e. small changes in routings and costs in the order of less than 0.5% potentially lead to a substantial reduction in the impact on climate.