Evaluation of multimodel-based ensemble forecasts for clear-air turbulence

To test more consistent and reliable upper-level turbulence forecasts, seven global numerical weather prediction (NWP) model outputs are used to construct the multimodel-based ensemble forecasts for clear-air turbulence (CAT). We used the updated version of the well-known Ellrod index, the Ellrod–Knox index (EKI), which is currently an operational CAT diagnostic for the significant weather chart at one of the World Area Forecast Centers. In this study, we tested two types of ensemble forecasts. First is an ensemble mean of all EKI forecasts from the NWP models. Second is a probabilistic forecast that is computed by counting how many individual EKI values from the seven NWP models exceed a certain EKI threshold at each grid point. Here, to calibrate the best EKI thresholds for the moderate-or-greater CAT intensity, the individual EKI thresholds, which vary depending on the resolutions and configurations of the NWP models, are selected using the 95th, 98th, and 98th percentiles of the probability density functions for the EKIs derived from the seven NWP models for a 6-month period. Finally, performance skills of both the ensemble mean and probabilistic forecasts are evaluated against the observations of in situ aircraft eddy dissipation rate and pilot reports. As a result, the ensemble mean forecast shows a better performance skill than the individual EKI forecasts. The reliability diagram for the probabilistic forecast gives a better reliability when using high-percentile EKI values as the threshold although it still suffers overestimation of CAT events likely due to the lack of observation and ensemble spreads.