Dynamical downscaling and variable resolution in limited‐area models

Abstract

Limited‐area models (LAMs) are widely used in numerical weather prediction and regional climate modelling to obtain high‐resolution results which would be too expensive computationally to acquire using a global model. This is often referred to as downscaling. To run LAMs requires lateral boundary conditions (LBCs) which are normally obtained from a global model or a lower‐resolution LAM containing the domain of the smaller LAM. Using a large‐domain high‐resolution LAM as the control, smaller‐domain LAMs were run using LBCs from a low‐resolution LAM run over the large domain. The results from the smaller‐domain LAMs were used to assess the errors due to the LBCs by taking differences against the large domain control. These differences were also used to understand how downscaling occurs and to assess the impact of increasing the domain size. Tests using variable‐resolution LAMs were conducted to determine whether errors arising from LBCs could be reduced. Investigations were made into the appearance of noise over steep orography at boundaries and possible mitigation strategies are discussed.