Felhőalap és felhőborítottság rövid távú előrejelzése neurális hálózat segítségével
In one sentence. Neural network post-processing of three years of WRF output improves short-term cloud ceiling and cover forecasts for aviation use.
Abstract
Numerical weather prediction systems do not provide cloud information directly, these information can be assessed during post processing only in an indirect manner. However accuracy of these forecasts is not sufficient for operational usage, therefore appropriate prediction of cloud ceiling is a real challenge for the forecasters. Our research focuses on whether short term forecasts of cloud ceiling could be improved using model outputs statistic based artificial neural network method. Our research based on three years of WRF numerical model output which was initiated by GFS forecast as initial boundary and lateral condition. The performance of artificial neural network highly depends on network configuration, therefore choosing the appropriate topology, transfer function and the right learning algorithm is a crucial element. The results of the best artificial neural network configurations have been compared to some old cloud assessments methods.
Labels
- Topic
- Post-processing & calibration
- Type
- New model or method
- Method
- Classical ML (trees, GP, linear, kernels)
- Time scale
- 6 h–3 days
- Domain
- Local / station
- Data
- NWP analysis or forecasts (IFS, GFS, ...)
- Status
- Research
- Identifiers
- OpenAlex:W7218152603
- Added
- 2026-10-22
Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.