Effects of Univariate and Multivariate Bias Correction on Hydrological Projections in Northern Tunisia

The purpose of this study is to evaluate the contribution of multivariate bias correction versus univariate correction to post-process climate simulation for the case of Oued El Abid catchment, situated in northern Tunisia. A recently developed multivariate bias correction method (MBCn) was used to post-process climate simulations (daily precipitation and temperature) from the recent EURO-CORDEX experience (CMIP5). Two climate models that have shown their effectiveness in the Maghreb region were used. The results were compared to those obtained by univariate quantile mapping (QDM) over a 30-year reference period (1970–2000). Several metrics were used to compare bias-corrected climate variables by QDM and MBCn. We found that there is not much difference between QDM and MBCn to reproduce correctly the seasonality of temperature and precipitation. Hydrological projection was then performed for the period 2040–2070 under RCP 8.5, using HBV-light conceptual rainfall-runoff model. This study has little conclusive evidence that multidimensional approaches to correcting climate change are generally preferable. It clearly demonstrates that integration or ignorance of inter-variable relationships between temperature and precipitation data did not substantially influence the conclusions drawn from hydrological studies on the impacts of climate change in a semi-arid area.