Verification and sources of uncertainity

Temperature and wind speed is widely and regularly observed and the numerical weather models can also predict these well. In contrast, the detection of clouds and their microphysical properties is not as straightforward as with temperature. Therefore, the modelling clouds deterministically is also a challenging field of science. Misplaced cloud field will certainly influence on icing prediction. However, the misplacement error is relatively random by its nature, and therefore, the possible errors in instantaneous icing modelling results will not degrade the value of long term average presented in this work.

The existing icing observations from three locations (Luosto, Puijo and Riutunkari) were used for verification of the icing model. Two of these sites provided only "on - off" type of observations. This means that the instrument is only able to define if ice is present or not, but it cannot give any information about the intensity or severeness of icing event. From the third location (Luosto) ice mass measurements were also available. All of these measurements were compared against ice model results in terms of the duration, location and magnitude of icing event. Figure 1 shows that the dynamics of ice mass event measured (5m height) match relatively well with the modelled ice mass. The overestimation of ice mass can be partly explained by the hight difference between observations (5m) and model data (30-50m). Cloud water most likely exist more on higher levels than near the surface.



Figure 1. Timeseries (13 - 16 Oct 2007) of ice mass aggregated on a standard cylinder. Observations (light blue) on 5 m height, model results on 30 and 50 m height.