Distribution-Based Scaling (DBS)

Distribution-Based Scaling (DBS) is a tool for bias correction of climate model results in order to make them suitable for hydrological climate change impact assessment. The approach of DBS is to match observed and simulated frequency distributions by assuming variable-dependent theoretical distributions.

In order to use the output from Regional Climate Models (RCMs) for hydrological impact studies, post-processing of the raw RCM data (for example temperature and precipitation) is required. This is because the RCM results contain systematical errors (bias) that invalidate the subsequent hydrological modeling. Such bias may manifest in e.g. overestimated winter temperatures or too many days with light precipitation.

In DBS, observed and simulated values of a specific variable (for example, temperature) are fitted to a suitable theoretical frequency distribution (for example, Gaussian distribution). Then percentiles of the simulated distribution may be mapped to the observed distribution, that is, the simulated climate may be interpreted in terms of the observed climate.
By assuming that this mapping is valid also in the future, simulated future climate projections may be corrected. The changes in both mean values and variability estimated by the climate model will be preserved in the bias-corrected data.

The figure below shows an example of bias correction by the DBS method. After correction the simulated values agree well with the observations. It shows how the overestimated number of days with light precipitation in the climate model becomes adjusted to the observed level.

Om analysen, bild 4

Observations (black), raw RCM output (red) and RCM output after post-processing with the DBS method (green). Cumulative distribution of daily temperatures (left) and frequency of days with different precipitation intensity (right).

DBS was originally developed for variables temperature and precipitation but also routines for bias correction of wind speed, relative humidity and radiation are being developed. The method has been applied on both local scale (station data), national scale (Sweden) and continental scale (Europe).

The method is being developed for applications in also other climate types, such as semi-arid and monsoon climates, where climate model biases may differ substantially from the ones encountered in Europe.


Dahné, J., Donnelly, C., and J. Olsson (2013) Post-processing of climate projections for hydrological impact studies, how well is reference state preserved?, IAHS Publications, in press.

Yang, W., Andréasson, J., Graham, L.P., Olsson, J., Rosberg, J., and F. Wetterhall (2010) Distribution-based scaling to improve usability of regional climate model projections for hydrological climate change impact studies, Hydrol. Res., 41, 211-229.