The CLM4.5 runs driven by picontrol climate (i.e., experiment I) slightly differ from the protocol: Basically a different time window was used for the historical run. Because CLM4.5 mixes Option 1 (for pre-industrial) and Option 2 (for historical) (cf. ISIMIP2b protocol for water sector). The corresponding societal drivers are 1860soc from 1661 to 1860, and 2005soc from 1861 to 2099, which is not entirely consistent with the protocol and creates a jump in land use data in 1861. To address this issue, the historical runs (Option 2) now branches off from the extended picontrol runs (with 2005soc) in the year 2100 so that the land-use jump that occurred during the transition from preindustrial to the historical period (jump from 1860soc to 2005soc) is no longer an issue because the model already had time to stabilize from the jump. The main issue is that the branching in the protocol should happen in 1861 and now happens in 2101, which is not a big issue because the climate is a picontrol climate. But year-to-year comparisons are not valid in the other time windows anymore.