| << | Validation of Measurements and Models | >> |
Abstract
Wind profiler data collected during a 3 months field campaign in late summer 2008 have been used as an independent measurement source to validate the high resolution numerical model COSMO-2 of MeteoSwiss. This action has taken place within a larger project aiming at the development of an improved high resolution weather prediction for the Swiss Plateau related to the safety of nuclear power plants. A companion contribution by Ruffieux et al. presents this project and the remote sensing network used to improve the short range forecast at the micro-scale.
COSMO-2 is run routinely eight times a day since February 2008. Analysis and forecast model data have been verified against independent wind profiler measurements.
Vertical profiles of wind direction and speed have been compared between wind profiler and model. The results of this verification over the entire measurement period show a bias close to zero for both parameters. This confirms that the model generally reproduces the air flow as observed at the location of the wind profiler. However standard deviation and quantile values of the model error are quite large, indicating that the model forecast is inaccurate over short time periods. Analysis of the model error versus time shows that a few short episodes are responsible for a large fraction of the forecast error and cause the loss of reliability of the model forecast. An experiment with local assimilation of the wind profiler data has shown the impact of the measurement on the short term and local forecast.