Abstract

Climate models are the major tools to study the climate system and its evolutions in the future. However, their simulations often present statistical biases that have to be corrected against reference data (e.g., observations) before being used in impact assessments. Several bias correction (BC) methods have therefore been developed in the last 2 decades, to adjust simulations according to historical records and obtain climate projections with appropriate statistical attributes. Most of the existing and popular BC methods are univariate, i.e., correcting one physical variable and one location at a time and, thus, can fail to reconstruct inter-variable, spatial or temporal dependencies of the observations. These remaining biases in the correction can then affect the subsequent analyses. This has led to recent developments of “multivariate bias correction” (MBC) methods to restore/adjust multidimensional dependencies. However, these methods often have differences in their technical aspects, assumptions and applicability, which can lead to different results. In this talk, I want to give a personal view on bias correction techniques, focusing on multivariate ones, how they can be categorized, the choices required for their configuration, and some of their underlying assumptions. For illustration purposes, a short (non-exhaustive) intercomparison of four existing MBCs will be provided. The main conclusions will help to ask still open questions as well as to outline some perspectives for future developments and applications of MBC