Addressing the assumption of stationarityin statistical bias correction of temperature

Bias correction of climate variables has become a standard practice in Climate Change Impact (CCI) studies. While various methodologies have been developed, their majority assumes that the statistical characteristics of the biases between the modeled data and the observations remain unchanged in time. However, it is well known that this assumption of stationarity cannot stand in the context of a climate. Here, a method to overcome the assumption of stationarity and its drawbacks is presented. The method is presented as a pre-post processing procedure that can potentially be applied with different bias correction methods. The methodology separates the stationary and the non-stationary components of a time series, in order to adjust the biases only for the former and preserve intact the signal of the later. The results show that the adoption of this method prevents the distortion and allows for the preservation of the originally modeled long-term signal in the mean, the standard deviation, but also the higher and lower percentiles of the climate variable. Daily temperature time series obtained from five Euro CORDEX RCM models are used to illustrate the improvements of this method.