Bringing modelled and observed surface winds together

The 10 m‐wind over flat terrain in models is often lower than observations indicate. Here we show that this negative bias is mainly caused by the larger roughness lengths that models use than the World Meteorological Organisation (WMO)‐advised roughness length of short grass (0.03 m). The concept of potential wind, which transforms the original wind to the theoretical wind under the assumption of a local roughness of 0.03 m, is well suited to distinguish this effect from other causes. It shows that the yearly averaged bias in Hirlam Aladin Research on Meso‐scale Operational NWP (numerical weather prediction) in Euromed (HARMONIE) is largely due to roughness effects. The year‐to‐year variations are caused by misrepresentations of the weather situations by HARMONIE. This can be concluded from the analysis of the geostrophic wind, which is insensitive for roughness effects.

Although the yearly averaged potential wind speed is well represented, HARMONIE reveals a strong seasonal cycle in the bias, being too low in summer, and slightly too high in winter. This is mainly caused by a too strong seasonal cycle in the representation of the leaf area index.

The HARMONIE data set points out to underestimate the average wind over The Netherlands up to 10% in the period before 1995. Even after correction for roughness effects, an underestimation of 6% remains. It is likely that the disappearance of the bias after 1995 is caused by the gradually improved quality of the ERA‐Interim data set with time.