The aim of this resurvey study is to check if herbaceous vegetation on the forest floor exhibits overall stability at the stand-scale in spite of intensive dynamics at the scale of individual plots and stand dynamic events (driven by natural fine scale canopy gap dynamics). In 1996, we sampled a 1.5 ha patch using 0.25 m² plots placed along a 5 m × 5 m grid in the best remnant of central European montane beech woods in Hungary. All species in the herbaceous layer and their cover estimates were recorded. Five patches representing different stand developmental situations (SDS) were selected for resurvey. In 2013, 306 plots were resurveyed by using blocks of four 0.25 m² plots to test the effects of imperfect relocation.

We found very intensive fine-scale dynamics in the herbaceous layer with high species turnover and sharp changes in ground layer cover at the local-scale (< 1 m²). A decrease in species richness and herbaceous layer cover, as well as high species turnover, characterized the closing gaps. Colonization events and increasing species richness and herbaceous layer cover prevailed in the two newly created gaps. A pronounced decrease in the total cover, but low species turnover and survival of the majority of the closed forest specialists was detected by the resurvey at the stand-scale. The test aiming at assessing the effect of relocation showed a higher time effect than the effect of imprecise relocation.

The very intensive fine-scale dynamics of the studied beech forest are profoundly determined by natural stand dynamics. Extinction and colonisation episodes even out at the stand-scale, implying an overall compositional stability of the herbaceous vegetation at the given spatial and temporal scale. We argue that fine-scale gap dynamics, driven by natural processes or applied as a management method, can warrant the survival of many closed forest specialist species in the long-run.

Nomenclature: Flora Europaea (Tutin et al. 2010) for vascular plants; Soó 1968–1980 for syntaxa

Deciduous forest, temperate forest, resurvey, forest reserve, forest developmental stage, Fagus sylvatica , ancient forest herbs

The significance of the herbaceous layer in forest biodiversity and ecosystem functioning has been widely appreciated (Whigham 2004, Gilliam 2014), generating an interest in its dynamics. During the last decade, remarkable changes in the herbaceous layer have been shown by several studies (e.g. reviewed in Verheyen et al. 2012). Resurvey studies investigated the understorey to detect potential long-term changes and to assess the role of different environmental variables as driving forces in deciduous forests. The observed changes were attributed to different background causes, such as combined effect of temperature increase and canopy opening (De Frenne et al. 2013, 2015), abandonment of former forest management practices (Hédl et al. 2010, Heinrichs et al. 2014), lack of disturbance (Brewer 1980, Taverna et al. 2005), deer herbivory (Rooney and Dress 1997, Taverna et al. 2005, Wiegmann and Waller 2006), increased soil acidity (Falkengren-Grerup and Tyler 1991, Durak 2010, Šebesta et al. 2011), nitrogen deposition (Verheyen et al. 2012) and often the intermingled influence of several factors (Heinrichs et al. 2014, Vanhellemont et al. 2014, Bernhardt-Römermann et al. 2015, Naaf and Kolk 2016). The majority of the observed changes in the herbaceous layer show an explicit trend, as a consequence of direct or indirect anthropogenic effects. A marked successional shift towards species-poorer herbaceous layer communities was detected as a result of abandonment of former management practices (Baeten et al. 2010, Hédl et al. 2010), lack of disturbance (Brewer 1980) and herbivory (Rooney and Dress 1997). A loss of rarer native plants (Hédl 2004) and the homogenization of the herbaceous layer was detected, and was attributed to high browsing pressure by deer herbivory (Wiegmann and Waller 2006), acidification (Durak 2010) and light deficit (Davison and Forman 1982, Heinrichs et al. 2014).

Fine-scale herbaceous layer dynamics is also often linked to stand-scale dynamics. Gap formation significantly increases solar radiation and soil moisture (Collins et al. 1985, Moore and Vankat 1986, Gálhidy et al. 2006) leading to intensive alteration of composition and cover of the herbaceous layer (Gálhidy et al. 2006, Kelemen et al. 2012) in large (Degen et al. 2005) and small gaps (Mountford et al. 2006). Stand dynamic events were matched with certain vegetation changes in different Slovakian primeval forests (Ujházy et al. 2005, 2007, Martináková and Martinák 2012). The herbaceous layer was best developed at the beginning of the decaying stage, and its cover sharply decreased when the tree seedlings and saplings overgrew them at the end of the decaying stage and in the growing stage. Herbaceous layer vegetation change can have a cyclic pattern, especially in forest stands that do not experience the strong successional effect of abandonment or other marked environmental change. Ujházy et al. (2007) detected low species turnover and cyclic changes in a Slovakian primeval forest at the stand-scale. Changing abundance or dominance of herb species characterized the cyclic changes, which was interpreted as internal community dynamics (Van Der Maarel 1996, Ujházy et al. 2005). Carøe et al. (2000) could not detect a successional trend either, only changes in species abundance in a managed Danish beech forest. Neither could Martináková and Martinák (2012) in a Slovakian natural fir-beech forest; however, the time interval was short in both investigations.

These short- and long-term changes in the herbaceous layer are most often detected by resurveys of phytosociological relevés or smaller quadrats after a few years (Ujházy et al. 2007, Martináková and Martinák 2012), one or a few decades (Davison and Forman 1982, Taverna et al. 2005, Łysik 2008, 2009, Vanhellemont et al. 2014) or even after a half-century interval (Brewer 1980, Hédl 2004, Durak 2010, 2012, Hédl et al. 2010, Šebesta et al. 2011). In some cases the relocation can be precise, when the time span is short or permanent corner markers were used, but more often former plots are relocated only approximately. The error resulting from relocation uncertainty can be high, especially in the case of small plots (Kopecký and Macek 2015). However, if relocation is done carefully, the resurvey can be robust to localization uncertainty, and can provide evidence of long-term changes in the herbaceous layer (Chytrý et al. 2014, Kopecký and Macek 2015). Ross et al. (2010) also demonstrated that if temporal change in the vegetation is greater than recent spatial heterogeneity, results of the resurveys can be interpreted with some confidence.

In our study, we were especially interested in the dynamics of the herbaceous layer at multiple spatial scales in relation to simple stand dynamic events, such as the opening and closure of smaller and larger gaps of a primeval beech forest remnant in Hungary. The exact location of the original sampling plots surveyed in 1996 was unknown, but the sampling plots could be relocated with 1 m accuracy after 17 years.

We hypothesized that in this virgin woodland remnant: 1) fine-scale changes in the herbaceous layer are evened out at the stand-scale, where no significant changes can be detected in the species pool (low species turnover) and total cover of the herbaceous layer at this temporal and spatial scale; 2) fine-scale changes in the herbaceous layer are governed by dynamic events in the tree canopy, such as opening and closing of the canopy gaps; 3) small gap stand dynamics warrant the survival of closed forest specialist herbs at the stand-scale.

After 17 years we found low species turnover and a general decrease in herbaceous cover at the stand-scale. Therefore, our first hypothesis is only partly supported. We found an overall stability of the species pool in the herbaceous layer at the stand-scale, whereas the abundance of vegetation (measured as total herb cover) changed considerably.

Several analyses pointed out that a decrease in herbaceous cover in a temperate forest is attributed to light deficit caused by a denser canopy. This can be a consequence of less intensive forest management or even abandonment, as well as the lack of natural disturbance including the activities of grazers (Brewer 1980, Davison and Forman 1982, Vera 2000, Verheyen et al. 2012, Kopecký et al. 2013). A drastic decrease in herbaceous species cover was found by Łysik (2008) in a primeval beech forest in Poland and by Ujházy et al. (2007) in a primeval fir-beech forest in Slovakia. Both studies connected this sharp decrease with the massive recruitment of woody regeneration. The observed decrease in our case can also be related to the partial closure of the canopy, and to the infilling of the two main former canopy gaps with young woody species (see below). The majority of the reduction in the herb cover was due to a few forest species (Galium odoratum, Mercurialis perennis) and to species (Urtica dioica, Dryopteris filix-mas, Athyrium filix-femina) that were abundant in gaps in 1996. Hence, we argue that stability of herbaceous cover might be realized on a larger study site, where all stand dynamic situations are represented in natural proportions. We also assume that the extremely dry years of 2012 and 2013 as a potential climate factor could have contributed to the observed decrease.

The majority of the temperate forest resurvey studies document high species turnover, as a consequence of environmental change. Denser canopies effect not only the cover of the herbaceous layer, but also trigger trait-based reactions as the decrease in light-demanding species and increase in species with good abilities for living in shade (Brewer 1980, Hédl et al. 2010, Heinrichs et al. 2014). A direct trend in species trait shift can be observed in spite of the formerly assumed or postulated stability of a site (Brewer 1980, Taverna et al. 2005, Łysik 2008). Despite the expected stability, results of Taverna et al. (2005), for example, showed systematic changes in herbaceous species richness in a climax hardwood forest in North-Carolina, as a potential result of the elimination of ground fires and widespread grazing typical at the beginning of the twentieth century. Invasive species can significantly change the composition of the herbaceous layer even in primeval forests (Łysik 2008). One of the few examples, where the ground vegetation showed an overall stability, is reported from a Danish beech forest by Carøe et al. (2000), but the temporal scale was very short (5 years). Similar stability of species pool was found by Woods et al. (2012) in unmanaged northern hardwood forests over a three-decade period. These results highlight the rarity of the observed low species turnover of Kékes Forest Reserve after almost two decades, where all extinct species were represented with very low abundances in 1996; similarly, all colonization happened with low abundance and many of the colonizing species were present in the species pool of the close surroundings.

Our results support the concept that the experienced intensive fine-scale dynamics is profoundly governed by the stand dynamic events of the studied forest stand. Significant differences were found between the individual SDSs and between years in herbaceous cover, species richness and species turnover at the SDS-scale. Highest species richness, mean species number/plot and highest herb cover characterized the two main gaps (OCO and OBG) in 1996. An expressed decrease in the herbaceous layer cover was observed after 17 years in these two, gradually closing gaps, where the saplings have overgrown and started to cast shadow on the herbaceous plants. Lateral canopy expansion of bordering trees has also contributed to gap closure. A very high number of extinctions with high absolute species turnover were detected in these situations. In the old collapse (OCO) the cover of Dryopteris filix-mas, Athyrium filix-femina, and Urtica dioica was drastically reduced. In the old beech gap (OBG) pronounced recession of Galium odoratum, Dryopteris filix-mas, Fagus sylvatica (young beech individuals have grown out of the herbaceous layer) and Mercurialis perennis was detected. Competition between the herbaceous plants and the saplings, especially with the very competitive beech saplings, and the resulting drop in herbaceous plant cover, was observed by several authors (Davis et al. 1998, Łysik 2008, Ujházy et al. 2007, 2013). On the other hand, the two younger gaps, created in 2010 and 2012 (3YG, 1YG) had low species richness, low mean species/plot and low herbaceous cover at the beginning, as the canopy was closed in 1996. The resurvey detected a sharp increase in both species richness and cover in the case of the three-year-old gap (3YG), where the massive colonization process was performed by gap specialists (Urtica dioica, Solanum dulcamara), by responsive forest species (Dryopteris filix-mas, Galium odoratum) and by tree seedlings (Fagus sylvatica, Acer pseudoplatanus). All these changes were more moderate in the case of 1YG, due to the shorter time-span. Despite the high absolute species turnover, the cover of the herbaceous layer and species richness changed only slightly in the control SDS. These results support our second hypothesis that stand dynamic events govern the herb cover, and drive the herbaceous layer changes, at least during this nearly two decades in our study site.

The Slovakian fir-beech forest investigated by Ujházy et al. (2005, 2007) had a very similar species pool in the herbaceous layer as our studied site. They found relatively low species turnover – compared to commercial forests – in the three developmental stages; rather the abundance or dominance values differentiated the growing, optimum and decay stages (Ujházy et al. 2005). The investigated short-term (4 year) changes were attributable to stand dynamics; they recorded a rapid change in the herb layer during the four-year period established in young gaps (Ujházy et al. 2007). However, they did not analyse colonisation and extinction patterns for all species.

Although the four subsamples of each plot were relocated only with a 1 m accuracy in 2013, statistical tests proved that the time effect (change in vegetation) was significantly higher than the relocation effect. With this we managed to assure that the interpreted temporal changes were not artefacts resulting from imperfect relocation.

We found that intensive local-scale extinction and colonization episodes were balanced at the stand-scale, resulting in overall stability of the species pool in the herbaceous layer vegetation. These results have important implications for both forest management and conservation.

The observed stand dynamics driven changes of vegetation indicate that the use of small regeneration areas in forest management can prevent competitive ruderal species (e.g. Calamagrostis epigejos) from causing serious problems in regeneration (Gálhidy et al. 2006, Kelemen et al. 2012).

From a conservation viewpoint, one of the most important species-related implications is that the long-term survival of closed forest specialists, including ancient forest indicators (Peterken and Game 1984, Hermy et al. 1999), could be guaranteed by management techniques mimicking natural small gap dynamics. This is important for several of those species that are seriously dispersal limited, meaning that once they disappear it is extremely hard, or it takes decades to centuries, to recolonize secondary forests (Kelemen et al. 2014). We assume that the observed stability in the species pool was secured by the primeval character of Kékes Forest Reserve. By primeval we refer to both structurally rich old-growth character and long forest continuity. The former is illustrated by the presence of all important forest developmental phases produced by natural stand dynamics (nat-man report 2002) and the high amount and diversity (size and decay phase) of deadwood (Christensen et al. 2005) and also of forest specialist organisms (e.g. Ódor 2000, Ódor and Standovár 2001, Ódor et al. 2006, Standovár et al. 2006, Heilmann-Clausen et al. 2014). Long continuity is convincingly documented by Czájlik (2009). Similar relationships between the richness in ancient forest indicator plants and that of other organism groups were shown, for example, by Hofmeister et al. (2014) for macrofungi.

Several studies documented that habitat specialist species occur in the highest numbers in forested landscapes with long continuity of forest structures and habitats (for a review see e.g. Nordén et al. 2014). This emphasizes the specific conservation role of even small remnants of virgin or ancient forest, as they secure the long-term survival of habitat specialists, and they might also serve as a source for dispersal from these remnants into the surrounding landscape.

In areas, where nature conservation is not the sole purpose of management, managers have started to apply retention forestry as a means of integrating conservation concerns into forest management. A thorough review (Fedrowitz et al. 2014) showed that retention cuts supported greater abundance and higher species richness of forest species than clear-cuts. This positive effect increases with the proportion of retained trees and time since harvest. However, they also showed that retention cuts cannot be a substitute for uncut forest because it had negative impacts on mostly closed forest specialist species.

All these suggest that for successful conservation of forest biodiversity we still need to preserve existing remnants of woodlands with high conservation value and also to apply a mix of management approaches in their immediate surrounding that support natural processes and the creation of important habitats.

The authors are grateful to all those friends and colleagues who helped in any phase of the fieldwork during the original or the resurvey, especially to Péter Ódor, Erzsébet Szurdoki and Ilona Paszterkó and László Gálhidy. We thank the constructive comments and suggestion of an anonymous reviewer, which helped to improve our paper. We also thank Zoltán Botta-Dukát for his statistical advice. This work was partially supported by a grant from Switzerland through the Swiss Contribution (SH-4/13).