10:00 AM - 10:15 AM
[ACG47-05] High-Resolution Numerical Simulation of a Clear Air Turbulence Event over Kanto Region Validated by In-Situ Flight Data
Keywords:Clear Air Turbulence, Large Eddy Simulation, Flight Data, Flight Simulation
Turbulence, a sudden shift in the air, sometimes becomes a serious threat to flying airplanes. Most turbulence can be avoided relatively easily by seeing what is visually found around turbulence, such as clouds, mountains, or other airplanes. However, Clear Air Turbulence (CAT), which is a kind of turbulence, cannot be avoided easily, because it is invisible.
To investigate the mechanisms of CAT generation, one effective method is performing case studies with high-resolution numerical simulations. Recent exascale supercomputers such as Fugaku, the fastest supercomputer in Japan, allow us to perform very high-resolution atmospheric simulations in a computational domain sufficiently large, which is capable to simulate the full processes of CAT generation. A lot of numerical simulations including Large Eddy Simulation (LES) have been performed to investigate and understand the processes of CAT generation in both realistic and ideal atmospheric conditions.
However, it is still uncertain how well turbulent eddies are realistically reproduced in LES and how dangerous they are to aircraft. In most case studies, the onset of CAT in simulations is validated by in situ observation data such as PIREPs (Pilot Reports) or EDR (Eddy Dissipation Rate) data. Generally, PIREPs are not inadequate for validating the intensity of the simulated turbulence eddies, as they are manually reported by pilots and often have large uncertainty in the reported location, time, and strength of shaking. EDR data includes EDR measured at a 1-minute interval. Therefore, it could be effective to use FDR (Flight Data Recorder) data for validation, in which atmospheric variables are recorded at sub-1 second intervals. Therefore, we can observe a wide variety of spatiotemporal atmospheric variables from FDR data. Unfortunately, however, we have not found similar works, because of the high difficulty of acquiring FDR data and high-computational costs for LES to capture aircraft-sized turbulence eddies.
We address a CAT event that affected aircraft at a relatively lower part of the free atmosphere (3-4 km) by straightforwardly advancing the resolution of a regional weather forecasting model, and the results are verified by the aircraft themselves. We perform a 4-domain simulation and the atmosphere is simulated by LES with 35 m resolution in the innermost domain. Every simulation is performed on Fugaku, and the resulting wind fields are validated by a flight simulator and flight data on the day. We also check how close the simulation on a finer computational grid can get to the flight data. During 7:00-11:00 UTC on Dec. 30, 2020, several encounters with CAT were reported over Tokyo. Most of the airplanes were about to land at Tokyo International (Haneda) Airport.
Kelvin-Helmholtz (KH) instability waves were reproduced over Tokyo in the coarsest domain at 10:00 UTC on Dec. 30, 2020. The waves grow from west to east and finally break to dissipate. The region and altitude of the waves seem close to the location of the aforementioned PIREPs, suggesting the breaking of the KH waves is the major cause of the generation of dangerous turbulent eddies on the day. The simulation in the finest domain reproduces details of the breaking of the KH waves in the atmosphere.
We utilize three flight data recorded by three airplanes that encountered CAT on Dec. 30, 2020, to validate the simulated results. To compare the simulations and the flight data, we performed virtual flight simulations to simulate aircraft response to the reproduced turbulence. Flight simulations are performed on each computational domain to check the resolution sensitivity to the simulated aircraft response. We compare power spectra of the simulated and observed vertical acceleration histories. As a result, the response to the turbulence in the finest domain reproduces a large part of the flight data. A spectral peak at the natural frequency of the pitching motion of the aircraft (0.14 Hz) is found in both the finest domain and the flight data. However, the simulated spectra decay at a frequency higher than 1 Hz. Several reasons can be considered: insufficient grid spacing, no pitching stability argumentation systems, and the assumption of a rigid body.
In summary, the results of the flight simulations show the 35-m LES can resolve turbulence eddies shaking aircraft at the natural frequency of the aircraft motion, which is also recorded in the FDR data. We also confirm that the smaller the grid spacing of the atmospheric simulation is, the closer the simulation can get to the flight data.
To investigate the mechanisms of CAT generation, one effective method is performing case studies with high-resolution numerical simulations. Recent exascale supercomputers such as Fugaku, the fastest supercomputer in Japan, allow us to perform very high-resolution atmospheric simulations in a computational domain sufficiently large, which is capable to simulate the full processes of CAT generation. A lot of numerical simulations including Large Eddy Simulation (LES) have been performed to investigate and understand the processes of CAT generation in both realistic and ideal atmospheric conditions.
However, it is still uncertain how well turbulent eddies are realistically reproduced in LES and how dangerous they are to aircraft. In most case studies, the onset of CAT in simulations is validated by in situ observation data such as PIREPs (Pilot Reports) or EDR (Eddy Dissipation Rate) data. Generally, PIREPs are not inadequate for validating the intensity of the simulated turbulence eddies, as they are manually reported by pilots and often have large uncertainty in the reported location, time, and strength of shaking. EDR data includes EDR measured at a 1-minute interval. Therefore, it could be effective to use FDR (Flight Data Recorder) data for validation, in which atmospheric variables are recorded at sub-1 second intervals. Therefore, we can observe a wide variety of spatiotemporal atmospheric variables from FDR data. Unfortunately, however, we have not found similar works, because of the high difficulty of acquiring FDR data and high-computational costs for LES to capture aircraft-sized turbulence eddies.
We address a CAT event that affected aircraft at a relatively lower part of the free atmosphere (3-4 km) by straightforwardly advancing the resolution of a regional weather forecasting model, and the results are verified by the aircraft themselves. We perform a 4-domain simulation and the atmosphere is simulated by LES with 35 m resolution in the innermost domain. Every simulation is performed on Fugaku, and the resulting wind fields are validated by a flight simulator and flight data on the day. We also check how close the simulation on a finer computational grid can get to the flight data. During 7:00-11:00 UTC on Dec. 30, 2020, several encounters with CAT were reported over Tokyo. Most of the airplanes were about to land at Tokyo International (Haneda) Airport.
Kelvin-Helmholtz (KH) instability waves were reproduced over Tokyo in the coarsest domain at 10:00 UTC on Dec. 30, 2020. The waves grow from west to east and finally break to dissipate. The region and altitude of the waves seem close to the location of the aforementioned PIREPs, suggesting the breaking of the KH waves is the major cause of the generation of dangerous turbulent eddies on the day. The simulation in the finest domain reproduces details of the breaking of the KH waves in the atmosphere.
We utilize three flight data recorded by three airplanes that encountered CAT on Dec. 30, 2020, to validate the simulated results. To compare the simulations and the flight data, we performed virtual flight simulations to simulate aircraft response to the reproduced turbulence. Flight simulations are performed on each computational domain to check the resolution sensitivity to the simulated aircraft response. We compare power spectra of the simulated and observed vertical acceleration histories. As a result, the response to the turbulence in the finest domain reproduces a large part of the flight data. A spectral peak at the natural frequency of the pitching motion of the aircraft (0.14 Hz) is found in both the finest domain and the flight data. However, the simulated spectra decay at a frequency higher than 1 Hz. Several reasons can be considered: insufficient grid spacing, no pitching stability argumentation systems, and the assumption of a rigid body.
In summary, the results of the flight simulations show the 35-m LES can resolve turbulence eddies shaking aircraft at the natural frequency of the aircraft motion, which is also recorded in the FDR data. We also confirm that the smaller the grid spacing of the atmospheric simulation is, the closer the simulation can get to the flight data.