Diploma Thesis

WRF-Modellierung meteorologischer Größen über Wald- und Grasökosystemen

Christoph Thieme (03/2009-12/2010)

Support: Andrei Serafimovich, Thomas Foken

The aim of this thesis is the evaluation of a WRF-ARW-model (WRF: Weather Research and Forecasting, ARW: Advanced Research WRF) over a forested and a grassland ecosystem. The model run was initialized with three nested domains (grid-spacing: 9,3 km, 3,1 km and 1 km). Time of the model run was the IOP2 (Intensive Observation Period) of the EGER-Project (ExchanGE processes in mountainous Regions).

To evaluate the model, following statistical parameters were calculated (MBE: Mean bias error, RMSE: Root mean square error, MAE: Mean average error, IA: Index of agreement, E: Coefficient of efficiency) for the whole period of the IOP2 as well as for each single day. These difference measures were calculated with the “nearest” gridpoint and with an “interpolated” gridpoint of the model. Pressure and wind direction were evaluated with their absolute values.

It was proved that the third domain provided better results for five of six meteorological parameters (temperature, wind speed, wind direction, sensible and latent heat flux). This shows that a resolution of 1 km in space and half an hour in time makes sense. Furthermore, differences in model results and measured values caused by micrometeorological phenomena (e.g. cold air flowing off mountains or mountain-valley-wind-systems) could be shown. These micrometeorological phenomena could not be simulated because the resolution of the terrestrial input is too low. The influence of large-scale weather patterns on the quality of the forecast could be observed as well as the fact, that there is no significant difference between the two ways to calculate the difference measures (“interpol” or “nearest”).

For more accurate weather forecasts in the Fichtelgebirge mountains, it is necessary to evaluate further model runs with different model-physics and model dynamics. Also more accurate terrestrial input can lead to better forecast results.

There are additional file downloads belonging to this publication


last modified 2010-12-13