Is time a variable like the others in multivariate statistical downscaling and bias correction?

Abstract. Bias correction and statistical downscaling are now regularly applied to climate simulations make then more usable for impact models studies. Over the last few years, various methods were developed account multivariate – inter-site or inter-variable properties in addition usual univariate ones. Among such methods, temporal either neglected specifically accounted for, i.e. differently from other properties. In this study, we propose a new approach called “time-shifted bias correction” (TSMBC), which aims correct dependency marginal aspects. TSMBC relies on considering initial variables at times (i.e. lags) as additional be corrected. Hence, dependencies (e.g. auto-correlations) corrected viewed adjusted an existing (MBC) method can used answer need. This is first evaluated synthetic data vector auto-regressive (VAR) process. second evaluation, work “perfect model” context where regional model (RCM) plays role of (pseudo-)observations, its forcing global (GCM) downscaled For both evaluations, results show large reduction biases properties, while spatial dependence structures still correctly adjusted. However, increasing number lags too much does not necessarily improve overly strong increase dimensions dataset even imply some potential instability and/or results, calling reasoned use datasets.