Dataset properties affect the performance of the ordination regression-based approach (ORBA) in predicting time to recovery

The globally increasing pace of ecosystem degradation makes it necessary to develop efficient restoration programmes. A reliable tool for estimation of the remaining time to recovery is an important component of such a programme. In this study, we explore the performance of the novel, ordination regression-based approach (ORBA) for prediction of time to recovery. We applied ORBA to a dataset of plant species composition data from ten spoil heap sites in S Norway, resulting from hydropower development. Five study sites were situated in the alpine and five in the boreal bioclimatic zone, the latter group showing large between-site variability of climate, time since disturbance (age) and restoration treatments. The vegetation of spoil heaps was recorded at two time-points, while that of the surroundings of each spoil heap was recorded once, to serve as a reference. We obtained time-to-recovery predictions based on ordination of the full dataset and of two data subsets that represented alpine and boreal sites, respectively. The vegetation of all spoil heaps became more similar to the vegetation of their surroundings over time. In many cases, linear models described the temporal pattern less well than asymptotic models. The latter models predicted slow recovery for all sites, but showed large variation in time to recovery among the boreal sites. In some of the boreal sites, we observed extremely species-poor plots and many infrequent species. Moreover, in one site we observed apparent mismatch between observed vegetation development and the underlying model of increasing similarity to surroundings. We ascribed these cases to suboptimal restoration practises, and for them, non-significant ORBA models resulted. Our findings support the conclusions of previous studies; that ORBA is a robust method for predicting time to recovery over a wide range of restoration projects. Its output – predicted time (years) to recovery – is easily communicated and may be useful in evaluating the speed and direction of restoration programmes. However, our results also demonstrate the importance of carefully assessing the properties of datasets (e.g. plotwise species richness) subjected to ORBA analyses.