Climate impact assessments often target a smaller area of interest, such as a city or local province. The climate information is often not applicable to that level, and the results might suffer from “noise” related to chaotic effects at weather scales which leaves the climate statistics uncertain at the local level. Further, model results can have large uncertainty also at the resolution (grid scale) of the models, which further increases the uncertainty. Therefore, the representation of the results is more reliable when combining statistics from grid boxes within a larger region. In the Climate Information Portal, no single grid-box result from models is used, but all data are first remapped to a common grid, by considering at least four model grid boxes at its original resolution.
How to get local information?
There are several methods and techniques available to extract local information from climate model output (see IPCC-TGICA 2007). Three well-recognized approaches are:
- Dynamical downscaling
- Statistical downscaling
- Simple interpolation (averaging the number of grid boxes surrounding and including the study area grid box).
Overall, dynamical and statistical downscaling methods are the most sophisticated methods to extract local information from regional or global models. If dynamical and statistical downscaling is not possible due to limited computational resources or limited observational data availability, simple interpolation methods can be useful in some applications.
It has been suggested that the minimum effective spatial resolution of a site should be defined by at least 4 model grid boxes for coarser resolution models (GCMs) and 9 model grid boxes for higher resolution models (RCMs) (Stocker et al., 2010, EURO-CORDEX guidelines, 2017). However, in complex topography, averaging may result in the loss of specific (orographic induced) grid box results (Maraun and Widmann, 2015). For sites by the coast or in mountainous areas other approaches might be considered to define representative regions.