AbstractA new study has been made of the Risø profiles with the following conclusions:(i)roughness lengths determined from neutral wind profiles selected by gradient Richardson number are biased towards high roughness lengths. Selecting the profiles by using a bulk Richardson number criterion yields average profiles with lower roughness lengths. This roughness length is relatively independent of stability while that determined from profiles using a gradient Richardson number is greater than the roughness lengths for either stable or unstable cases;(ii)it is doubtful whether ‘double kinks’ in some of the Risø wind profiles are caused by local terrain inhomogeneities, as has previously been suggested;(iii)three‐dimensionality and unknown mesoscale and baroclinic effects make it difficult if not impossible to compare current change‐of‐terrain theory with the Riø 123