Significant uncertainty in the representation of orography in numerical weather prediction and implications for atmospheric drag and circulation

The representation of orographic drag remains a major source of uncertainty for numerical weather prediction (NWP) and climate models. Its accuracy depends on contributions from both the model grid‐scale orography (GSO) and the subgrid‐scale orography (SSO). Different models use different source orography datasets and different methodologies to derive these orography fields. This study presents the first comparison of orography fields across several operational global NWP models. It also investigates the sensitivity of an orographic drag parameterisation to the inter‐model spread in SSO fields and the resulting implications for representing the northern hemisphere winter circulation in a NWP model. The inter‐model spread in both the GSO and the SSO fields is found to be considerable. This is due to differences in the underlying source dataset employed and in the manner in which this dataset is processed (in particular how it is smoothed and interpolated) to generate the model fields. The sensitivity of parameterised orographic drag to the inter‐model variability in SSO fields is shown to be considerable and dominated by the influence of two SSO fields: the standard deviation and the mean gradient of the SSO. NWP model sensitivity experiments demonstrate that the inter‐model spread in these fields is of first‐order importance to the inter‐model spread in parameterised surface stress, and to current known systematic model biases. The revealed importance of the SSO fields supports careful reconsideration of how these fields are generated, guiding future development of orographic drag parameterisations and re‐evaluation of the resolved impacts of orography on the flow.