Abstract: Long-term regional climate simulations are computationally very costly. One way to improve their computational efficiency is to split them into overlapping time slices, which can then be run in parallel. Although this procedure reduces the cost, sufficient spin-up must be left at the start of each slice. In any case, discontinuities will occur due to internal variability where two different slices join. In this study, we explore the relative role of spin-up time and internal variability in the discontinuities of overlapping time slice simulations and their effect on the simulated climate. This analysis has implications also for non-overlapping time slices, commonly used in very high resolution climate modelling, where long transient simulations cannot be afforded. We show that discontinuities are negligible for surface and upper-air variables, but they are noticeable in variables with long response times, such as soil moisture or snow depth. For these variables, differences between the slices are mainly attributed to internal variability, but also to insufficient spin-up time, depending on the region. In general, the results show that the overlapping time slice approach is valid to accomplish long term regional climate simulations.

Authorship

Download reference