Understanding and projecting ecosystem dynamics is a key goal for both basic science and applied ecosystem management. In ecosystems dominated by long-lived organisms such as forests, space-for-time substitutions (SFT) inferring temporal development from observations of study objects differing in age across space at a given time, also called chronosequences, are widely used as an alternative approach for long-term monitoring, in particular, to understand forest dynamics after stand-replacing disturbances. Uncertainties and challenges of SFT substitution, however, arise from violations of the underlying assumption that everything but time since disturbance is identical among the spatial sampling sites. Direct comparisons between true time series and SFT substitutions are still surprisingly rare. I resampled a 20-year-old SFT substitution study to check if this method accurately projects vegetation changes after major disturbances in primary subalpine conifer forests and to evaluate its conclusions for forest management. Two decades and a spatial scale of about 600 km² constitute spatial and temporal conditions for which little empirical evidence for the validity of SFT substitution exists. SFT suggested a strong shift away from old-growth plant community compositions after clearcutting with, misleadingly, no trend back over two decades. True temporal development, however, differed significantly from this expectation as it showed a directional trend back towards old-growth plant community compositions after the initial displacement. Compositional recovery is predicted to take 100-200 years in the studied ecosystem. Accordingly, several late seral forest floor herbs and bryophytes were still missing completely even 50 years after clearcutting. Non-constant climate, species pools, and management actions compromise the ability to reliably project temporal dynamics in plant community composition and stand structure by space-for-time substitutions, likely also for other scenarios than the studied subalpine conifer forests after stand-replacing disturbance. Long-term monitoring therefore is indispensable for understanding ecosystem dynamics, in particular in slow-growing and stressful environments. Management needs to consider the slow regeneration in its planning of rotation times and is advised to monitor and retain species indicative of old-growth conditions at the landscape level.

This dataset contains plot-level information and plant species occurrences from 48 plots from a subalpine, boreal forest region in southern British Columbia, Canada, sampled as a chronosequence of primary forests and development after stand-replacing disturbance (clearcutting, with young and old clear-cuts differing by 20 years in time). Initial sampling took place in 2003, and resampling of all plots took place in 2023. The 20 years of true temporal development can therefore be compared to a 20-year chronosequence (space for time) approach. See the main publication for more details.