Bias-corrected climate projections

Bias correction has become a common practice to improve numerical weather forecasting. However, bias correction (BC) methods for ensembles typically deteriorate the relation across space, time, and different variables, thereby leading to physically unrealistic atmospheric situations. Complicated additional steps based on copula methods are therefore required to correct these deficits. A simple member-by-member (MBM) post-processing method was developed that required no additional correlation-preservation steps. In this project, this MBM post-processing method will be adjusted to serve as a method for BC in the context of climate projections. 

In order to encompass uncertainty, three greenhouse gas emission scenarios used in the IPCC assessment reports will be considered (i.e., RCP2.6, RCP4.5 and RCP8.5) and bias-corrected over Belgium. In order to serve as input for the HETEROFOR forest dynamics simulation experiments, a subset of ensemble members will be identified as representative scenarios.

The correction scheme will be compared with the reference quantile mapping (QM) method. 

Hourly-adjusted and bias corrected time series will be provided for six surface variables (two components of wind, radiation, temperature, rainfall and specific humidity).

Besides, it has been shown that global climate models cannot reliably reproduce some aspects of the climate change trends of the past decades. Therefore, the aforementioned MBM and QM methods will be adjusted and applied in the context of climate projections with the aim of increasing the reliability of climate trends from the existing datasets from the CORDEX and ensembles.