Data-driven insights for air traffic control


Traffic forecasting will be essential to efficient operations as air travel returns. But making accurate predictions has become more difficult than ever.

The impact of the coronavirus pandemic on air traffic numbers is scarcely believable. At one point, passenger traffic was down more than 80%. More recent figures are encouraging – with flight numbers globally up 20% on average from February 2021 to March 2021 – but continued growth is far from predictable.    


Though the drastic drop in flight numbers is a concern for air navigation service providers (ANSPs), the sheer unpredictability of what the future will bring is in many ways every bit as concerning.                           

How will travel restrictions, quarantines, and new virus strains affect demand going forward? Even in a best-case scenario, do people want to travel internationally in the current environment? What will be the lingering economic impact of the pandemic? How much business travel has migrated permanently online? And can airlines and airports adapt their business models and networks sufficiently quickly?

CANSO has released Educated Guess: Interpreting traffic forecasting during the pandemic to provide an overview of the current state of traffic forecasting.

ANSPs need accurate information on demand to plan staffing, capacity enhancements and anticipated income. But as the paper states: “It is clear that a lack of medium-term forecasts is a significant hindrance to the financial and capacity planning required to enable ANSPs to successfully navigate the current crisis. It may be some time after the pandemic before medium term forecasting is reliable again”.

Tradition makes way

Traditionally, aviation forecasting was broken down into short, medium and long term. Short-term forecasts cover approximately six months or even shorter periods and will account for specific noteworthy events, such as a soccer world cup. In the medium term, perhaps a year or more out, forecasts are used to plan budgets and resource requirements. Long-term forecasts, meanwhile, are used in connection with strategic planning. Trends are analysed to determine key decisions and goal setting.

All forecasts combine qualitative and quantitative elements to arrive at a reasonable conclusion.

By March 2020, it was evident that this traditional methodology was obsolete. COVID-19 disrupted too many of the normal data sources that feed traffic forecasts and many of the assumptions used to create forecasts no longer held. Prior to the pandemic, airlines flew 98%-plus of their schedules published more than six months in advance. As the virus spread, schedule planning was reduced to a matter of weeks and changes, once minimal, could occur just days out. Cancellations became commonplace.

Consequently, all the major forecasting organisations changed forecasting frequency as the dynamic situation unfolded.

The CANSO paper advises caution in using forecasts issued while the pandemic was in full swing. “ANSPs should aim to identify more direct local data to support decision-making,” it says and suggests sensible caveats would include rechecking the available forecasts on a weekly basis to make sure no major changes are introduced.

“The planning ahead process should be adapted to account for spontaneous changes and forecasts should be trusted for short-term outlooks rather than drive medium-term strategic decisions,” says Educated Guess.

New assumptions

Traditional forecasting techniques might become valid again should vaccination programmes continue their impressive rollout. But it is unlikely that the situation will be stabilised in the near term. And a return to 2019 traffic levels could be five years away.

For the foreseeable future, therefore, a new set of assumptions will be needed for predicting traffic. These will range from the loss of GDP to changing travel patterns. Undoubtedly there will be higher uncertainty in the forecasts and that will lead to difficulties in planning for best and worst-case scenarios.

As forecasting models necessarily evolve, which assumptions are valid and which are not will become obvious. How the virus develops, and the extent of the economic recovery are particularly important. Getting these parameters correct will generate better forecasts that prevent short-term problems and generate longer-term certainty.