Forecasting a Severe Convectively Induced Turbulence Event of a Passenger Aircraft using the ECMWF IFS model

Atmospheric clear air turbulence (CAT) poses a major risk to aviation. Besides airframe damage, it can cause injuries of passengers and crew when encountered unexpectedly. CAT is usually found close to mountains, near the atmospheric jet streams, or convective clouds.

Of these types, convectively induced turbulence (CIT) continues to pose a challenge for forecasting models because its generation mechanism including initiation and development of convection is not fully resolved in models and needs to be parameterized. Despite a general trend to increase the resolution in operational forecasting models, turbulence itself on the scale that affects aircraft is also not resolved in state-of-the-art models. As grid sizes are larger than the size of turbulent eddies, operational forecasts typically use a collection of empirical turbulent indices to predict turbulence relevant to aviation. These indices that cover different sources for turbulence are combined to derive the cube-root of the energy dissipation rate (EDR), a common, aircraft-independent parameter to indicate the strength of turbulence. EDR can be translated to turbulence intensity depending on aircraft size, i.e. moderate turbulence for EDR > 0.18 m^2/3 s^-1 and EDR > 0.27 m^2/3 s^-1 for light and heavy aircraft, respectively.

The Singapore Airlines flight SQ321 was on its way from London to Singapore when severe turbulence was encountered over Myanmar on 21 May 2024. Abrupt rapid changes in flight altitude in an area of developing convection including a drop of 178 ft left one passenger dead and dozens injured.

In our work, we analyzed how well the encountered turbulence was predicted by the CAT-forecast index that was implemented in the European Centre for Medium-Range Weather Forecasts (ECMWF) integrated forecasting system (IFS) in recent years. The CAT-forecast index is based on the total dissipation rate (DISS) of the IFS in units of EDR which is derived from the model’s physical tendencies for horizontal momentum. It includes contributions from the vertical diffusion scheme, the convective momentum transport and the convective gravity wave drag (GWD).

It was found that ECMWF IFS was able to predict the developing convection and associated convectively induced turbulence (CIT) 24 hours in advance. However, knowledge about the characteristics of the IFS forecasts for convection (i.e. the uncertainty in time and space) is needed in order to properly interpret such turbulence forecasts and to make proper use of the probabilities for moderate and severe turbulence determined from the IFS ensemble. The probability for EDR > 0.18 m^2/3 s^-1 determined from the IFS ensemble members is 12 % at the grid-point of the incident at 240 hPa and increases to 34 % when the EDR maxima of +/-2 surrounding grid-points (+/-0.2° in latitude and longitude) are taken into account.