The importance of small natural features in forests—How the overgrowth of forest gaps affects indigenous flower supply and flower‐visiting insects and seed sets of six Campanula species

Abstract The abandonment of historical land‐use forms within forests, such as grazing or coppicing, and atmospheric nitrogen deposition, has led to an increasing overgrowth of forest gaps and canopy closure in forest ecosystems of Central Europe. From 1945 to 2015, 81% of the forest gaps greater than 150 m2 within the study area transitioned into a closed forest. This study investigated how the overgrowth process affects flower supply, flower visitors, and reproduction of Campanula species. Six native Campanula species with different light requirements were used as phytometers. The forest gaps in the studied area are a feature of the historical European cultural landscape. We compared large gaps caused by human activities, small gaps caused by habitat conditions, and closed forests. In eight blocked replicates, each with the three habitat categories, we recorded the flower cover and number of indigenous flowering species in the immediate surroundings, and, of six Campanula species, flower visitors and seed production. Forest gaps and their size positively affected the number of flowering plant species in the surrounding area, the number of all flower visitor groups, and the number of seeds produced by all six Campanula species. Flower cover in the surrounding area was higher in large gaps, but there was no difference between small gaps and closed forests. Among flower visitors, small bees varied the most between the three habitat categories, and flies varied the least. The effect on the number of seeds produced was particularly strong for three light‐demanding Campanula species. The overgrowth of forest gaps negatively affected flower supply, flower‐visiting insects, and seed sets of six Campanula species. Forest gaps should be managed to maintain the reproduction of open forest plants and their pollinators.


| INTRODUC TI ON
The pollination process is critical for natural and agricultural systems (Guntern et al., 2014;IPBES, 2016;Klein et al., 2006;Naturkapital Deutschland TEEB, 2018). However, this process is threatened in many parts of the world (Leonhardt et al., 2013;Potts et al., 2010;Sanchez-Bayo & Wyckhuys, 2019). In Central Europe, changes in land use are considered the most important cause of the decline in pollinators (MEA, 2005;Zurbuchen & Müller, 2012). Multiple studies in Central Europe (Carrié et al., 2017;Hopfenmüller et al., 2014;Tscharntke et al., 2005;Zurbuchen et al., 2010) assessed open landscapes with regard to management intensity, landscape composition, and configuration or distances between nesting and foraging sites, as well as changes in agricultural landscapes. However, changes in forest landscapes and the associated impacts on pollinating insects have received less research attention, even though forests cover 35% of the area of Europe (MCPFE, 2020).
As light is a prerequisite for many plants to flower, many forest plants flower in spring before the trees unfold their leaves.
Unprecedented human-induced changes in the nitrogen (N) cycle in Western Europe in the last century resulted in enormous N deposition from the atmosphere into forest ecosystems (BMEL, 2018;Sutton et al., 2011). Furthermore, the increased atmospheric CO 2 content promoted optimal use of the available N, thereby increasing tree growth rates over the last 50 years (Ciais et al., 2008;Laubhann et al., 2009;Thomas et al., 2010). The analysis of 23 studies investigating the effect of atmospheric N deposition on plant communities in understory forests showed densification of the canopy and a shift toward shade-tolerant plant species (Verheyen et al., 2012). In addition to anthropogenic N deposition, the abandonment of traditional land-use practices has caused a significant improvement in the forest nutrient supply, especially within thin soil layers. Historical land-use practices, such as forest litter removal, forest pasture, intensive coppicing, or dead wood collection, led to nutrient impoverishment and, thus, low tree growth rates. Open forests with gaps resulted (Gatter, 2004;Poschlod, 2017;Rubner, 1967). Today, approximately 73% of European forests' net annual wood increment is utilized by fellings (MCPFE, 2020). For example, in Germany, forest gaps account for only 2% of the forested area (Hampicke, 2018;Schmalfuß & Aldinger, 2012). In a study of the development of forest gaps in a deciduous forest, our research group (Braun-Reichert & Poschlod, 2018) revealed an 81% decrease in forest gaps of more than 150 m 2 from 1945 to 2015 in the study area. Specifically, historical meadows or pastures are no longer used, and clear cuttings and many small sites naturally treeless due to drought and nitrate deficiency are now overgrown because of recent N deposition.
Nevertheless, little is known about the consequences of the overgrowth of forest gaps on the reproduction of entomophilous plants. A suitable approach to experimentally assess habitat quality or the effects of landscape changes on pollinators is to use attractive potted food plants as phytometers (Steffan-Dewenter et al., 2002;Woodcock et al., 2014;Zurbuchen et al., 2010). Members of the genus Campanula are attractive to flower visitors and suitable as phytometer plants due to the quantity and volume of pollen grains and the amount of nectar they produce (Müller et al., 2006). Furthermore, the genus Campanula attracts the highest number of oligolectic bee species in Central Europe (Zurbuchen & Müller, 2012). Additionally, flies, especially hoverflies, frequently visit this genus (Hansen & Totland, 2006;Janzon, 1983).
Therefore, this study asked the following questions: • Is the flower supply as food resources for pollinating insects poorer in a closed forest than in small and large gaps?
• Is the number of Campanula flower-visiting insects lower in a closed forest than in small and large gaps?
• How does the overgrowth of forest gaps affect the seed set of six Campanula species with different light demands?

| Study area
The study area was "Jochenstein," the easternmost section of the nature reserve "Donauleiten from Passau to Jochenstein." The nature reserve is located in the Danube valley in southeast Germany, where it borders Austria. The Danube River cuts into the paragneiss rock within the reserve at a depth of approximately 300 m, and the average slope is 30° (LDBV, 2012).
On the south-facing slopes, there is a mosaic of different forest communities. In 36% of the study area, forests with Fagus sylvatica (Luzulo-Fagetum) grow on mesophilic sites, and in 30%, forests with Carpinus betulus and Quercus petraea (Galio-Carpinetum) grow on dry and warm sites. According to aerial photographs of ADBV (Office for Digitisation, Broadband and Surveying Vilshofen, 2013, unpublished data), forest gaps constituted three per cent of the study area. Openness and gaps of the forests may result from the dry site conditions, exposure to the south in combination with thin layers of soil, scree slopes, and exposed rock outcrops. Identified species include Origanum vulgare, Teucrium scorodonia, and Hylotelephium maximum (Geranio-Trifolietum with a transition to Sedo-Scleranthetea).
Historical use, such as cuttings, coppicing, removing forest litter, establishing forest pastures, and harvesting leaf fodder, led to further depletion of soil nutrients. Between 2013 and 2015, the study area received 17-19 kg of additional nitrogen deposition per hectare and year (Schaap et al., 2018). Based on extensive surveys of the nearby power plant and the BIOKLIM-monitoring project (Bässler et al., 2015;Donaukraftwerk Jochenstein, 2012), the insect fauna of the area is known to be particularly species-rich and thermophilic.

| Study design
We defined three habitat categories: large gaps (clearings), small gaps (glades), and closed forest. The forest gaps were classified based on their size, measured using aerial photographs (ADBV, Office for Digitisation, Broadband and Surveying Vilshofen, 2013, unpublished data) and QGIS software (QGIS.org, 2018). Small gaps ranged between 32 m 2 and 375 m 2 and large gaps between 1,344 m 2 and 6,008 m 2 (Table 1). Large gaps were cleared by humans, they resulted for example from road embankments, recent cuttings, nature conservation management measures, or use as meadows (Table 1).
Small gaps represent the character of open forest in the study area caused by habitat conditions, namely exposure to the south in combination with thin soil layers, scree slopes, and exposed rock outcrops.
We selected sites containing each of these three habitat categories with a maximum distance apart of 210 m to maintain continuity of environmental factors (He et al., 2012). Each site with these three habitat categories formed a statistical block and was replicated eight times ( Figure 1). Five blocks were studied in 2015 (blocks A-E) and three blocks in 2016 (blocks F-H; for dates, see Table 2).
Experimental approaches with potted, attractive food plants have proven to be effective in detecting the effects of and on flower visitors (Steffan-Dewenter et al., 2002;Woodcock et al., 2014;Zurbuchen et al., 2010). Six native Campanula species occurred in the study area. To demonstrate their utility as phytometers, the list of plants below provides the light requirements, most important habitats, and habitat category in which the species occurred in the study area. The Ellenberg indicator value of light (L) represents the ecological characteristics of Central European plants in relation to the relative light availability in the field, ranging from one (occurring only in deep shadow) to nine (occurring only in full sunlight) (Ellenberget al., 1991). The list is arranged according to this indicator value. The most important habitats of the species are specified by BfN (1999). Finally, the occurrence in one of the habitat categories in the study area was described. The immediate surrounding area was defined as a square with a 30 m × 30 m side length centered on the pots containing Campanula.
• C. patula: L = 8; habitats are fresh meadows and pastures; flowered in the surrounding area in large and small gaps.
• C. glomerata: L = 7; habitats are dry and semi-dry grasslands; flowered in the study area in large gaps, but not in the immediate surrounding area. We grew these six Campanula species from autochthonous seeds in pots in an open greenhouse under standardized field conditions: one plant per pot, same pot size, same soil substrate in the pots, and the same water and light conditions within a given  In the garden, the species received light according to their needs.

Reason for openess of large gap
As a result, the pots with the Campanula plants were only exposed to the influence of the different habitat categories during the flowering period.
Six pots of each species per site were considered sufficient to lure flower-visiting insects (Sowig, 1989). All pots were placed adjacent to each other in the center of the study sites. We controlled predation, especially by slugs, with slug pellets.

| Flower supply
Every site block was inspected eight times per year. During these eight inspection passes between May and the beginning of August in 2015 and 2016, we recorded the flower supply in the surrounding area (for specific dates, see Table 2). The surrounding area was

| Flower visitors
During

| Pollination success
The pots with the Campanula species were returned to controlled conditions in a garden after the last flower had faded and before the first seed capsule had burst open. The timing of this return also minimized the influence of herbivores (snails). To test the pollination success in the different habitat categories, from the six pots in a plot, we collected ten fruits of each Campanula species and counted the number of seeds per fruit.
The six Campanula species in the study area have been deemed self-incompatible (Gadella, 1964;Stephenson et al., 2000). However, depending on the presence of pollinators, self-compatible and even spontaneous selfing plants of the genus Campanula may exist (Inoue & Amano, 1986;Stephenson et al., 2000). Therefore, we verified the self-incompatibility of our Campanula species. For this purpose, we excluded flower visitors on one flower of each species with nylon stockings. Later, we determined if these capsules contained seeds.

| Statistics
We performed two separate linear mixed-effect models to estimate the habitat effect on the flower resources and the number of flowering species in the surrounding area. We added the eight site blocks as random effects in the models due to the experimental design.
Similarly, we analyzed the habitat effect on the occurrence of the four different flower visitor groups. The first group included all recorded flower visitors, which we then divided into the number of small bees, large bees, and flies. For each visitor group, we used a generalized linear mixed-effect model (family="poisson") We used the closed forest as the intercept because the deviation from the closed forest to the forest gaps is the effect of interest.

| Flower supply
Flower cover in the area surrounding large gaps was significantly greater (estimate, 5,606) than in the closed forest. In contrast, flower cover in the area surrounding closed forest and small gaps did not differ significantly (Table 3). The number of flowering plant species in the surrounding area was significantly greater in the forest gaps than in the closed forest (estimated at 12.28 and 2.25, respectively; Table 3).
When we compare the absolute numbers, we see that the mean flower cover in the large gap was 6,138 cm 2 in 900 m 2 surrounding area, while it was lower in small gaps (477 cm 2 ) than in closed forest

| Flower visitors
The numbers of all flower visitor groups were significantly more frequent in two-gap habitat categories than in closed forest (Table 4).
The deviation between the forest gaps and the closed forest was greatest for the small bees (estimated at 4.1 and 3.7, respectively) and smallest for the flies (estimated at −0.62 and 0.98, respectively).
The large gaps were a greater distance from the closed forest than

| Pollination success
The capsules excluded from pollination did not produce any seeds, confirming that spontaneous selfing did not occur. Forest gaps and their size positively influenced the number of seeds in all six studied Campanula species (Table 5). All Campanula species had the lowest number of seeds in the closed forest. Most species had the highest number of seeds in large gaps, whereas C. rotundifolia and C. rapunculoides produced more seeds in the small gaps ( Figure 4).
The deviation of the seed set in forest gaps from that in the closed forest was larger for the light-demanding species C. patula, C.
glomerata, and C. rotundifolia (estimated from 2.41 to 5.77) than for the shade-tolerant species C. rapunculoides, C. persicifolia, and C. trachelium (estimated from 0.78 to 1.14; Table 5). Species with an Ellenberg indicator value of light of 7 or 8-C. rotundifolia, C.
glomerata, and C. patula-produced a mean of fewer than 10 seeds in the closed forest, while C. rapunculoides (L = 6), C. persicifolia (L = 5), and C. trachelium (L = 4) produced a mean of 41 and 83 seeds, respectively. The highest number of mean seeds was 259 for C. persicifolia in large gaps (Figure 4).

| Flower supply
It is well known that light positively affects flower formation even in forest plants (Cao et al., 2017;Cunningham, 1997;Killkenny & Galloway, 2008). This observation is consistent with expectations that the number of flowering species would be higher in small gaps (glades) than in closed forest and highest in large gaps (clearings). However, flower cover was only significantly higher in the large gaps than in the closed forest, but not in the small gaps.
We attribute this to the high flower cover of spring geophytes in the "closed deciduous forest" in the first inspection pass in May, in addition to a large standard deviation in all inspection passes ( Figure 2a). Thus, when forest gaps are overgrown, flower supply declines.

TA B L E 3
The linear mixed-effect models shows the effect of habitat category on the flower cover in the surrounding areas and on the number of flowering species in the surrounding areas. We set the closed forest as intercept (ic) throughout to show the deviation (estimate) to the large and the small gap

| Flower visitors
If forest gaps become overgrown into closed forests, all the groups of flower visitors studied showed declines in their flower visits.
These declines were particularly clear for small bees while hardly noticeable for flies (Table 4). Temperature also varied in each habitat category (shading). We assume that temperature explains the differences in abundances between small bees, large bees, and flies.
In warm temperatures, bees are dominant flower visitors, while flies are abundant at more moderate temperatures (Adedoja et al., 2018;Corbet et al., 1993;Herrera, 1997;Hodkinson, 2005;Ssymank, Keams, et al., 2011). However, large bees are not a uniform group with regard to their temperature requirements. Bumblebees may also fly at very cool temperatures. In contrast, certain large bees such as Megachile spec. or Melitta haemorrhoidalis require warmer temperatures (Westrich, 2018).
Similarly, flower visitors were found to more frequently visit Campanulastrum americanum (Killkenny & Galloway, 2008) and Hosta ventricosa (Cao et al., 2017) in sunny and open patches than in shaded and forested ones. In contrast, Hansen and Totland (2006) found no difference in the number of flower visitors of Campanula persicifolia between forest and meadow habitats in Norway. However, they had counted mainly hoverflies, muscoid flies, and few bumblebees.
We interpret this as an indication that temperatures in Norway were similarly cool in both forest and meadow, in contrast to the habitats we studied (Adedoja et al., 2018;Ssymank, Kearns, et al., 2011).

| Pollination success
The results clearly showed that the overgrowth of forest gaps into closed forests negatively affected the seed production of all studied Campanula species. As expected, the negative influence was TA B L E 4 The generalized linear mixed-effect models shows the effect of habitat category on the number of flower visitors in different groups *p < .05; **p < .01; ***p < .001.

F I G U R E 3
The number of flower visitors (mean and standard deviation) in different groups in the three habitat categories. All flower visitors also included butterflies, beetles, and other insect groups, which were only recorded in small numbers. In all groups of flower visitors, the numbers of small and large gaps differed significantly from those of closed forest

TA B L E 5
The generalized linear mixedeffect models shows the effect of habitat category on the number of seeds of the six Campanula species. We arranged the species in decreasing order of Ellenberg indicator value of light (Ellenberg et al., 1991) F I G U R E 4 The number of seeds produced in 10 capsules (mean and standard deviation) of the six Campanula species in the three habitat categories. We arranged the Campanula species in decreasing order of their Ellenberg indicator value of light (Ellenberg et al., 1991). In all species, the number of seeds in small and large gaps differed significantly from that of closed forest stronger in light-demanding species, such as C. patula, C. glomerata, and C. rotundifolia, than in less light-demanding species, such as C.
persicifolia, C. trachelium, and C. rapunculoides. Although C. patula has a higher Ellenberg indicator value for light than C. glomerata and C.
rotundifolia, it demonstrates greater plasticity with respect to light and a generally higher seed number. If forest gaps are overgrown into closed forests, the extremely low seed production becomes a limiting factor for reproduction (extinction debt). Moreover, spontaneous selfing resulting in any seed set could not be detected when excluding pollinators, supporting the validity of our results. Goodell et al. (2010) revealed similar effects for Lonicera maacki in edge and interior forest habitats, Killkenny and Galloway (2008) for In contrast, no significant difference was detected in the number of Campanula persicifolia seeds between forest and meadow habitats in Norway (Hansen & Totland, 2006). In the Norwegian study, the number of seeds per fruit was strongly pollen-limited in both habitat categories, possibly indicating that flower visitors were crucial for pollination. In contrast, regarding groups of flower visitors, our study showed differences in pollination success, whereas the Norwegian study did not.  (Poschlod & Braun-Reichert, 2017). In addition, an opening of the tree canopy positively affects the number of other arthropod species (Bussler, 2016;Müller et al., 2007). It is known that butterflies, in particular, have a high diversity in open and coppice forests (Fartmann et al., 2013). Furthermore, light forests and forest gaps are important habitats for other animal groups such as birds or bats (Dietz et al., 2016;Gatter, 2004). Hilmers et al. (2018) illustrated the importance of light in the initial stage of forest succession for the diversity of various organism groups. Forest gaps increase habitat diversity, structural complexity, and faunal and floral species diversity (Muscolo et al., 2014). Historically open forests should be preserved as a part of the European cultural landscape and as important habitat for flora and fauna, especially pollinators.

| Relevance to conservation
Only a few decades earlier, open forests and forest gaps were much more common in the cultural landscape of Central Europe (Poschlod, 2017). Now, historical forms of forest use have been abandoned, and nitrogen deposition has increased (Hampicke, 2018;Poschlod, 2017;Stuber & Bürgi, 2011;Verheyen et al., 2012). Even naturally formed gaps are growing over faster than they would without the heavy nitrogen inputs. Especially on marginal sites like rocky heads, where only certain tree species could grow very slowly, are now colonized by atypical tree species with very dense canopies like Fagus sylvatica. Therefore, the political demand for a reduction of nitrogen inputs must be continuously asserted (Sutton et al., 2011).
Maintenance is necessary to preserve forest gaps of high ecological value because, without human intervention, they will close (Braun-Reichert & Poschlod, 2018;Bussler, 2016). However, simple thinning through logging often leads to undesirable effects such as strong growth of Rubus spp. or other nutrient-indicating plants. A significant reduction of N deposition is necessary for the long term to preserve forest gaps (Sutton et al., 2011;Verheyen et al., 2012).
Historical forms of land use in forests such as forest grazing or coppice are complex (in terms of target species, type of grazing animal, and intensity and duration of grazing) and labor-intensive and therefore not easy to implement (Bärnthol, 2003;Liegl & Dolek, 2008;Poschlod, 2017;Rackham, 2003;Rupp & Michiels, 2020;Zahn et al., 2014). However, grazing animals and coppicing would remove nutrients which contributes to the openness of the forest (Bärnthol, 2003;Berendse, 1985;Marrs et al., 2020). The new development of pristine forests is often set as a conservation goal, which is easier to manage but seemingly contradicts forest gap management.
Only seemingly, because forest gaps are small natural features that do not occupy large areas in contrast to pristine forests. However, the implementation of both concepts would greatly increase the diversity of an area and its ecological and nature conservation value.

ACK N OWLED G M ENTS
We thank Volker Debus and the gardeners of the Botanical Garden of the University of Regensburg for cultivating a multitude of Campanula plants, Cristian Odar for the help with data entry, Florian Hartig for the support with statistics, and Martin Lüthke for improving the writing in English. We thank the Government of Lower Bavaria for granting access and seed collecting permits and the Land Survey Office in Vilshofen for providing maps.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data of flower supply, flower visitors, and number of seeds that support the findings of this study are openly available in Dryad at http://doi.org/10.1002/ece3.7965.