Questions: Managed forests often show a more homogeneous age structure compared to untouched forests, but also a higher share and frequency of sun-exposed conditions due to harvest operations. Frequent, small-scaled forestry operations may therefore lead to elevated species richness, compared to undisturbed forests. When non-intervention is introduced in formerly managed forests,

• Is there a significant decrease in species richness of vascular plant species?
• What species are particularly affected? Are there nonrandom shifts in species composition?
• Are also typical shadetolerant forest species jeopardized due to prolonged deep shade ?

Location: Four recently installed strict reserves in Atlantic to Sub-Atlantic lowland oak and beech forests on fertile loess soils east and south-west of Brussels (Belgium).

Methods: We compared vegetation relevés in permanent plots (183 plots of 16x16m) with a 10 year interval. Total species richness per site was derived from rarefaction curves; significance of differences in species richness and composition at plot level were tested using paired t-tests and Wilcoxon signed rank tests. NMDS was used to visualize non-random shifts in species composition.

Results: We registered significant declines in species richness both at reserve and plot level. This decline was not random, but strongly depending on ecological traits and strategies, with strong declines in light-demanding gap phase-associated species like Lonicera periclymenum and Deschampsia cespitosa, and species requiring disturbed or bare soils for their recruitment, such as Juncus effusus and Carex sylvatica. Shade-tolerant mesic species like Anemone nemorosa and Allium ursinum clearly increased in frequency and cover, despite strong declines in previous decades.

Conclusion: The combined effect of continued closed canopy and slow soil recovery from euthrophication apparently support the development of a less species rich vegetation, but with higher dominance of characteristic species of mesic lowland oak and beech forests.

The study was performed in four strict forest reserves, located in ancient woodland sites in Central Belgium, South-West and East of Brussels.

At all sites, a systematic grid of permanent plots (mesh size 70m) was installed. Location of the plot center was registered and permanently marked, so exact relocation was guaranteed. This resulted in a total of 183 plots on fertile loess soils over the four sites that can be used in the analysis.

At every plot, vegetation relevés and dendrometric measurements were performed between 2003 and 2005, and repeated 10 years later. Vegetation relevés were carried out on 16x16m squares, oriented with the corners in the cardinal directions. One survey consisted of two samplings in late April and June, in order to register both spring and summer species at their optimal development. Species presence was recorded and coverage estimated using the extended decimal scale of Londo, and later transformed into percentage values using the midpoint value of the respective class. Both records are integrated to one result, taking into account the record with the highest cover (cfr. Mölder et al., 2014). Like Van Calster et al. (2008a) and Baeten et al. (2009), we excluded the woody seedlings from the analysis to keep irregular mast events from clouding the trends in the vegetation composition. At all plots, dendrometric surveys were performed in the winter preceding the vegetation survey, and during the first survey, also soil samples were taken and analyzed.

Forest affinity classes are attributed to each species according to the assignments by Schmidt et al. (2011) and Heinken et al. (2019). Five classes of forest affinity are applied: 1 = taxa mainly found in closed canopy forest; 2 = taxa mainly typical for gaps and edges in forests and open forests; 3 = taxa which can be found both in forest as well as open vegetation; 4 = taxa which can be found partly in forests, but mainly in open vegetation; 5 = no forest affinity, species of open countryside.

For all species, a set of indicator values and characteristics are compiled and assigned (see: supplementary information S2). Ellenberg indicator values for light (L), soil reaction (R), soil moisture (F) and nitrogen (N) were derived from Ellenberg et al. (2001) (=le, re, fe, ne) and Hill et al. (1999) (=lh, rh, fh, nh). When different values are given for a species in both publications, the average value is assigned (l_av,...). Species with Ellenberg L-indicator values of 6 and higher are clustered as ‘light demanding species’. A subset of ‘vernal geophytes’ is selected combining the Raunkiaer life forms classification according to Ellenberg et al. (2001), and flowering period March to May according to Florabank (Van Landuyt et al., 2012). Based on the functional signature scores according to the CSR-strategies of Grime (1979) assigned to each species according to Hunt et al (2004), a set of ‘stress-tolerant’ species and ‘competitors’ are derived (scores S>0.33 and C ≥ 0.5, respectively). Species richness and cumulative cover for these different subsets was also calculated.

Dendrometric surveys were done at the same plots using nested circular plots of 9 and 18m radius. Tree species, position and diameter are registered with DBH threshold of 40 cm (large circle) and 5 cm (small circle). Overall basal area per ha are calculated, and volumes per ha using two-entry tariffs of Dagnélie etal. (1999; 2013). Volume of standing and lying deadwood are also calculated using the calculation tools in FieldMap(TM). Total aboveground biomass is the sum of living volume, standing and lying deadwood.

Litter Quality (lq) for each plot is calculated as weighted average of litter quality scores of the main tree species (Quercus spp. and Fagus sylvatica = 1, Populus = 4; other species = 3) and their relative share in the basal area at each plot. Cumulated canopy cover (CCC) is calculated as a combined value for tree layer and shrub layer cover excluding double count for overlap, as used by Mölder et al. (2014) and Heinrichs & Schmidt (2017)

During the first sampling cycle, also soil samples were taken and analyzed. Mineral soil was sampled using a 3cm diameter auger, taking 5 subsamples within each plot. These subsamples were aggregated to one soil sample, dried to constant weight at 40°C. Samples were analyzed on texture (Sand, Loam, Clay fraction), pH-CaCl₂ , CEC and Kjelldal-N using the Standardized Operation Procedures described by Cools & Devos (2016).

The dataset consists of three files. Headers in every file are described below

Datafile herblayerchanges.csv : gives the raw vegetation relevé data with following columns

- plotid : unique code identifying the sampling plot (see also other files)

- recordid : unique code of a record= plantcover at one survey

- survey : 1 or 2 referring to first survey and second survey, 10 years later

- speciescode : internal coding of species

- coverpercentage: value (% cover) for coverage of the species, derived from Londo scale, using mindpoint value of respective classes

-year : year of sampling

- speciesnamecode :abbreviation of species name, also used in the NMDS figure in the manuscript

- species : Species names follow EURO+MedPlantBase: http://ww2.bgbm.org/EuroPlusMed/query.asp [accessed april 15, 2020]

File plotdata_loc_soil_dendro

- plotid, survey: see before

- lq : litter quality (see methods section for calculation)

- cover = cumulated canopy cover (CCC) : see methods section for calculation

- cover_class = subdivision in open (< 66%), intermediate (66-89%), closed (90% and more)

- site : one of four forest reserves: T= Terrijst, E= Everzwijnbad, P=Pruikenmakers, J=Jansheideberg

- x and y : coordinates of the plot centre, using Belgian Lambert72 co-ordinates (in meters)

- sand_fraction : percentage of particles over 50 micrometer

- Kjeldahl_N, Organic_content, P_plantavailable, CEC_BaCl2 : soil analysis results according to the methods described above

- basalarea, livingvolume, deadstanding, deadlying, abovegroundbiomass: dendrometric parameters calculated as explained above; all values in m² per ha and m³ per ha

File species_characteristics

- speciescode, speciesnamecode, name : see above

- aw_hermy1999 : Ancient woodland species (1/0) according to Hermy et al. (1999)

- woodlandassociation_schmidtetal_2011 and woodlandassociation_Heinkenetal_2019: woodland associated species according to 5-scale scoring according to Schmidt et al (2011) and Heinken et al (2019) respectively. For the scores: see methods section

-le, lh, fe, fh, re, rh, ne, nh : Ellenberg indicator values according to Ellenberg (2001) and Hill etal (1999) respectively (see methods section)

- l_av, f_av, r_av, n_av :  average score for L, F, R and N- indicator value based on both previous publications

- CSR: CSR-strategies of Grime (1979) assigned to each species

- C, S, R : scores according to Hunt et al (2004)

- herbspecies : herb species (1) excluding the woody tree and shrub species (0)

- geophyte : (1/0) is the species a geophyte using Raunkiaer life forms classification according to Ellenberg et al. (2001)

- floweringfrom; floweringto : flowering period according to Florabank (Van Landuyt et al., 2012); values 1-12 according to the months of the year