Speaker
Description
Knowledge of humidity in the upper troposphere and lower stratosphere (UTLS) is of special interest due to its importance for cirrus cloud formation and the resulting climate impact. However, the accurate description of the UTLS water vapor distribution in current weather models is subject to large uncertainties. Here, we develop a dynamic-based humidity correction method using an artificial neural network (ANN) to improve the relative humidity over ice (RHi) of numerical weather predictions from the European Center for Medium-Range Weather Forecast (ECMWF). The ANN model is trained over Europe, Africa and the East Atlantic with measured humidity data from the In-service Aircraft for a Global Observing System (IAGOS) in 2020 and with 8 time-dependent thermodynamic and dynamical variables from ECMWF ERA5. Previous (timesteps at -6 h and -2 h) and current atmospheric variables within ±2 ERA5 pressure layers around the IAGOS flight altitude between 400 and 200 hPa are used for the ANN training. Amongst them, RHi, temperature and geopotential exhibit the highest impact on ANN results, while other dynamic variables are of minor importance.
The ANN shows excellent model skills and the predicted RHi in the UT has a mean absolute error MAE of 6.6% and a coefficient of determination R² of 0.93, which is significantly improved compared to ERA5 RHi (MAE of 15.7%; R² of 0.66). The ANN model also improves the prediction skill for ice supersaturation in the all sky LS, as well as the cloudy and the clear sky UTLS, with MAE between 4.33% and 6.77% and R² of 0.92 to 0.95. Also, the erroneous peak at 100% RHi in the ERA5 data sets is corrected by the ANN. For a specific contrail cirrus region over the East Atlantic on 14 April 2021, the contrail predictions using the ANN are in better agreement with MSG satellite observations of ice optical thickness than the results without the ANN humidity correction. Our results suggest that the ANN method can be applied to other weather models to improve humidity predictions and to support climate research applications.
Reference
Wang, Z., Bugliaro, L., Gierens, K., Hegglin, M. I., Rohs, S., Petzold, A., Kaufmann, S., and Voigt, C.: Machine learning for improvement of upper tropospheric relative humidity in ERA5 weather model data, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2024-2012, 2024.
| You plan to attend | On site (at FZ Jülich) |
|---|
