Aim
The diversity and composition of natural communities are rapidly changing due to anthropogenic disturbances. Magnitude of this compositional reorganization varies across the globe, but reasons behind the variation remain largely unknown. Disturbances induce temporal turnover by stimulating species colonizations, causing local extinctions, altering dominance structure, or all of these. We test which of these processes drive temporal community changes, and whether they are constrained by natural environmental gradients. Moreover, we assess to what degree identity shifts translate to changes in dominance structure.

Location
Finland

Time period
Observations 1985-2006, disturbance history >140 years.

Major taxa studied
Vascular plants.

Methods
We investigated temporal turnover of boreal forest understory in response to disturbance, here forest management, along a soil fertility gradient. We disentangle the roles of species gains, losses and abundance changes in driving temporal turnover in response to and after disturbance by comparing turnover rates in different forest age categories along fertility gradient. We quantify temporal turnover richness-based complement of Jaccard’s similarity index and proportional-abundance based dissimilarity index. We also test whether disturbance history or fertility influence the relationship between identity shifts and dominance structure.

Results
We found that the impact of disturbance on temporal turnover depends on soil fertility. The greatest turnover occurred in most fertile forests immediately after disturbance. There, species gains and losses strongly altered dominance structure leading to high turnover, whereas undisturbed old forests and nutrient-poor habitats were characterized by stable dominant species even when the majority of species shifted their identity.

Main conclusions
Our results suggest that human impacts on temporal biodiversity change vary along environmental gradients. In boreal forests, the fertile habitats have higher probability than nutrient-poor sites to change their composition in response to anthropogenic disturbances. Resource availability and disturbance history may thus influence consequences of temporal turnover for ecosystem functioning.

Vascular plants in forest understory were surveyed on a systematic network of 1700 sites established on mineral-soil in forested land in 1985–86. These sites are a part of a systematic sampling network of the 8^th Finnish National Forest Inventory (Reinikainen et al., 2000). This network consists of clusters, which were located 16 km from each other in Southern Finland, and 24 and 32 km apart in Norther Finland along east-west and north-south axes, respectively. Each cluster consists of four linearly located sampling sites 400 m apart from each other in Southern Finland and three sampling sites 600 m apart from each other in Northern Finland. All 1700 sites were resurveyed in 1995, and a subset of 443 of them (max 1 site per cluster) were resurveyed in 2006 (Fig. 1). The spatial extent of this subset was comparable to previous surveys covering whole country. The survey performed on 443 sites in 2006 was part of the BioSoil project carried out under the Forest Focus scheme, which is a subset of the pan-European UN-ECE ICP Forests extensive monitoring site network (Level I, Lorenz & Fischer, 2013).

The following files contain a two subsets of these understory surveys: 'SubsetA.csv' includes sites, with NO forest management between two surveys, while 'SubsetB.csv' includes sites, which encountered some type of forest management between 1985 and 1995.

Species abundances are given as an average percent cover across four 2m² study quadrats within a 'Site'.

Data files SubsetA and SubsetB contain the necessary data to perform the analyses in the linked manuscript. Both files comprise of species abundance data, environmental covariates (explained in the methods of the article) and timing of the latest forest management action. 'Site_year_pair' is an unique identifier for each 10 years sampling period, while 'Site' identifies the location. Missing values are indicated with 'NA'.

Please see the attached Metadata_Subsets_A_B.doc for a detailed description of the variables.

SubsetA.csv contains sampling periods WITHOUT forest management between the 'Start_year' and the 'End_year' and known 'Time_since_disturbance' before the 'Start_year'. In SubsetA, some 'Sites' were re-surveyed twice (both in 1995 and 2006) and have thus two 'Site_year_pairs'.

SubsetB.csv contains sampling periods WITH known timing and type of forest management during the period, i.e. between the 'Start_year' and the 'End_year'.