Uncertainty of uncertainty decomposition approaches for projections of hydrological extremes

The quantitative description of uncertainty in future projections of hydrometeorological variables provides valuable information for a better interpretation of climate change impact for informed policy decisions and actions to mitigate the associated risk. Several methods have been developed and used for decomposing the uncertainty in projections. The relative importance of the uncertainty associated with the choice of uncertainty decomposition methods compared to the other sources of uncertainty has however never been quantitatively investigated. We scrutinize where and to what extent the rate of fractional uncertainties could vary across the globe depending on the choice of uncertainty decomposition methods. We characterize drought by the standardized precipitation evapotranspiration index (SPEI) using a large ensemble of the Coupled Model Intercomparison Project Phase 5 (CMIP5) and Phase 6 (CMIP6) general circulation models (GCMs). Flood and extreme precipitation are quantified by fitting a generalized extreme value distribution (GEV) to the annual maxima time series of the ISIMIP2b global hydrological model and CMIP5/CMIP6 simulations. The uncertainty in future projections of extreme precipitation, flood, and drought is then split in the variance contributions using the traditional ANOVA, quasi-ergodic ANOVA (QE-ANOVA), HS09, and variance decomposition-same sample size (VD-SSS) methods. Finally, the uncertainty arising from the choice of uncertainty analysis methods is quantified and compared across different types of hydrological extremes and the IPCC reference regions.