A simple method to assess the added value using high-resolution climate distributions: application to the EURO-CORDEX daily precipitation

Soares PMM, Cardoso RM
Climate Dynamics, (2018),

Download PDF


Regional climate models (RCMs) are used with increasing resolutions seeking to represent in an improved way regional to local-scale atmospheric phenomena. The EURO-CORDEX simulations at 0.11 and simulations exploiting finer grid spacing approaching the convective-permitting regimes are representative examples. These climate runs are computationally very demanding and do not always show improvements, which depend on the region, variable and object of study. The gains or losses associated with the use of higher resolution in relation to the forcing model (global climate model or reanalysis), or to different resolution RCM simulations, are widely known as added value. Its characterization is a long-standing issue, and many different added-value measures have been proposed. In the current study, a new method is proposed to assess the added value of finer-resolution simulations, in comparison to its forcing data or coarser-resolution counterparts. This approach builds on a probability density function (PDF) matching score, giving a normalized measure of the difference between diverse resolution PDFs, mediated by the observational ones. The distribution added value (DAV) is an objective added-value measure that can be applied to any variable, region or temporal scale, from hind cast or historical (non-synchronous) simulations. The DAVs metric and its application to the EURO-CORDEX hind cast daily precipitation data are presented here. Generally, the EURO-CORDEX simulations at both resolutions (0.44 and 0.11) display a clear added value in relation to ERA-Interim, with maximum values around 30% in summer and 20% in the intermediate seasons. When both RCM resolutions are directly compared, only three of five models (0.11) show added value, with a maximum of ~10%. The regions with the larger DAVs are areas where convection is relevant, e.g. Alps and Iberia. When looking at the extreme PDF tail, the higher-resolution improvement is generally greater than the low resolution for seasons and regions.