Abstract
Climate projections contain the uncertainty due to the internal variability of the climate system, including its chaotic nature. While the uncertainty due to the internal variability can be theoretically mitigated by executing large ensemble simulations with perturbed initial conditions, only a limited number of large-ensemble experiments are available in CMIP6 future scenario dataset. Here we propose a method that increases the effective ensemble sampling size in evaluations of future projection by integrating multiple SSP-RCPs for a period corresponding to a specific increase in temperature from the preindustrial level (i.e., X°C warming). The success of the method was assessed by investigating whether the uncertainty due to small number of ensemble members could be reasonably reduced. First, we confirmed that the spatial distributions of the future flood magnitude change were similar under a 2 °C warming in all SSP-RCP scenarios. Additionally, the uncertainty due to the different SSP-RCPs (5-10%) was smaller than the differences between different warming levels such as between 2 and 3 °C (around 20-50%), suggesting differences among SSP-RCPs as to future flood discharge change are relatively small. These results suggested that integrating SSP-RCPs to increase the effective ensemble size was a reasonable approach, reducing unbiased variance among GCMs in about 70% of land grid points comparing to the result using SSP5-RCP8.5 alone.