The importance of tree species identity and trait-based winter foraging ecology of bark-foraging bird species in a large Central European floodplain forest

The study was conducted in the Donau-Auen National Park, Eastern Austria, which contains the longest free-flowing section of the Danube with one of Europe’s largest semi-natural riparian forest areas of approximately 9300 ha. The national park stretches from Vienna to the eastern border of Austria (Fig. 1). It was established in 1996, and since then, no commercial forest management operations have been carried out (Hönigsberger 2017). Some parts of this forest have been sealed off from floods by a flood protection dam, which prevents adjacent settlements from flooding (Adrion 2016). We conducted our study in both the areas that are still flooded regularly and the sealed-off areas, which can still be affected by flooding events due to a rising groundwater table.

The forest cover of the national park is a mixture of softwood and hardwood forest stands that include 30- to more than 100-year-old stands, with a high tree and shrub species diversity (Nagl and Schulze 2017). We carried out our surveys throughout a variety of floodplain forest types and age classes while making sure to avoid monoculture plantations. Our transects were located in the vicinity of the settlements of Orth an der Donau, Eckartsau, Witzelsdorf, and Stopfenreuth, with a total length of ca. 30 km (Fig. 1). This study was focused on woody species with a diameter at breast height greater than 10 cm. Based on monitoring of available tree species (see Methods), the most frequent tree species were white poplar (Populus alba, ca. 25%), common ash (Fraxinus excelsior, ca. 17%), field maple (Acer campestre, ca. 10%), common walnut (Juglans regia, ca. 7%), black poplar (P. nigra, ca. 5%), and pedunculate oak (Quercus robur, ca 5%). The trees with the largest trunk diameter mainly included pedunculate oak, common ash, and black poplar. Such trees with a diameter at breast height over 150 cm were scattered mainly throughout the oldest (> 100 years of age) stands of the study area but can also be found in younger stands. There was a low frequency of dead trees in the study area (ca 6%), and most of them had a diameter at breast height (DBH) < 40 cm. There were sporadic (< 5%) native floodplain trees such as white elm (Ulmus laevis), white willow (Salix alba), and crack willow (S. fragilis), and mountainous species, such as common hornbeam (Carpinus betulus) and sycamore maple (Acer pseudoplatanus). We also found old individuals of shrub species, such as Cornelian cherry dogwood (Cornus mas), common dogwood (C. sanguinea), common hazel (Corylus avellana), black elderberry (Sambucus nigra), and common hawthorn (Crataegus monogyna) and non-native trees such as the tree of heaven (Ailanthus altissima) and black locust (Robinia pseudoacacia).

Data were collected in two periods: from December 2019 to January 2020 and from January to early March 2021, under winter weather conditions below or slightly over 0 °C. We visited each transect only once in each of the two periods, walked at a steady pace, and within a 100 m buffer of where we recorded foraging behavior of each bird that spent a minimum of 3 s in a particular spot gleaning, probing, pecking or excavating (Czeszczewik 2010; Nappi et al. 2015; Osiejuk 1998) on a woody species with a DBH > 10 cm. To avoid autocorrelation of data, we used only “first foraging observations” (Filek et al. 2018). In the first period, we observed only obligate bark-foraging species, such as woodpeckers, Eurasian nuthatches, and treecreepers, while during the second period, we additionally included facultative bark-foraging birds as great tits (Parus major), Eurasian blue tits (Cyanistes caeruleus), and marsh tits (Poecile palustris). We took great care in the direction of flights of the birds encountered to minimize the possibility of pseudo-replication (i.e. double counting). Observations were carried out only during periods without heavy rain or strong winds from one hour after sunrise through the morning hours and sometimes into the late afternoon in the case of long transects.

Based on the protocols of similar studies (Czeszczewik 2010; Duron et al. 2018; Filek et al. 2018; Lorenz et al. 2016; Nappi et al. 2015; Osiejuk 1998; St-Amand et al. 2018; Török 1990), during every encounter, we recorded the following variables: tree species, tree condition, and trunk DBH.

Tree or shrub species were classified at the species level, except white willow and crack willow (collectively classified as “willow trees” since they have very similar architecture and wood structure and are difficult to distinguish in leafless conditions). These two willow species also hybridize (Triest 2001). White poplar often hybridizes with common aspen (Populus tremula) (Castiglione et al. 2010), while black poplar crossbreeds with planted hybrid Euramerican poplars (Populus × canadensis) (Gencsi and Vancsura 2002). These two groups can be clearly identified and are referred to as “white poplar” and “black poplar”. The two poplar groups have different bark structures, as the bark of white poplars is smoother compared to black poplars, which may lead to differences in their bark-dwelling arthropod communities (Ónodi et al. 2021).

The condition of trees was classified as living (less than half of their branches are decayed), decaying (more than half of their branches are decayed but still have living branches), or dead (all of the branches are dead or branchless).

Stem DBH was assigned at 30-cm intervals, e.g. > 10–40 cm, > 40–70 cm, > 70–100 cm, and > 100 cm.

We recorded tree species, condition, and stem DBH of the tree the focal individual bird foraged upon. Furthermore, we also collected data on unused trees, with a slight modification of the ‘random tree’ approach (Pierson et al. 2010). We registered one tree west and one east of the focal tree within a 15 m distance. We randomly selected a tree if there was no tree within 15 m of a given direction.

Further, the bark roughness of each tree species was defined in a posteriori mode, based on the characteristics of middle-aged and older trees as smooth (no crevices = 1), intermediate (< 2 cm deep crevices = 2), and rough (> 2 cm deep crevices = 3) after Buba and Danmallan (2019). Values are shown in Table 1.

Throughout our field sessions, we collected 234 foraging records of eight species altogether. We only included species in our analysis that had at least 20 encounters, which resulted in 231 foraging observations of seven bird species. We excluded encounters of lesser spotted woodpeckers (Dryobates minor, three encounters) and analyzed foraging data of 52 great spotted woodpeckers (Dendrocopos major), 20 middle spotted woodpeckers, 57 Eurasian nuthatches, 20 treecreepers, 39 Eurasian blue tits, 22 great tits, and 21 marsh tits. Following earlier studies (Skórka and Wójcik 2003; Brauze and Zieliński 2006), we grouped the common treecreeper (Certhia familiaris) and the short-toed treecreeper (C. brachydactyla) into one category, since these sibling taxa have highly similar morphology (Clouet and Gerard 2019). Most of the time, it was not possible to tell the two species apart for non-singing individuals, especially in winter conditions, while they are spiraling quickly on the trunk and gleaning on the bark. Along the transects, we observed a few black woodpeckers (Dryocopus martius) and Eurasian green woodpeckers (Picus viridis) only flying and calling, thus, we did not include them in our study. We had data on 231 foraging trees of 20 species and 462 adjacent trees. We encountered 27 tree species in total throughout our study (Table 1).

To compare the diversity of tree species used, a species accumulation curve was calculated for each bird species using iNEXT (Chao et al. 2014, 2016). The curves visualize how the number of tree species used by individual bird species increases with the number of first foraging observations. For direct comparison between bird species, we then used a rarefaction approach calculating the number of tree species (± 95% CI) predicted to be used for the largest common number of first foraging observations of 40. For each bird species, the number of tree species used was estimated either by rarefaction or the extrapolation for a common number of foraging observations of 40 since the smallest number of foraging observation of a species was 20, and Chao et al. (2016) recommend extrapolating species accumulation curves to no more than twice the sample size. This corresponds to a sample size of 40 observations.

We made a two-mode network on the tree species preferences of the studied bird species. We only included the tree species with more than four registered individuals throughout the study to mitigate the bias of sporadic but over-preferred tree species with, for example, occasionally highly decayed wood. We only worked with bird-plant relationships of a positive Jacobs’ preference index value. This index ranges from -1 (total avoidance) to + 1 (strong preference) (Loehle and Rittenhouse 1982; Swamidoss et al. 2012). Any ties were weighted by Jacobs’ preference index values. We computed network metrics on the macro- (connectedness, fragmentation, and transitivity) and micro-scale (normalized degree, closeness centrality, and betweenness centrality) as well (Martínez-García et al. 2020). We also computed key-player metrics to determine the most important bird species and the five most important tree species in this network via the distance-weighted fragmentation criterion (Borgatti 2006). These analyses were made with UCINET 6 for Windows (Borgatti et al. 2002) and KeyPlayer 1.45 (Borgatti 2003). The network was visualized with NetDraw 2.176 (Borgatti 2002).

To compare the use of tree characteristics between species, we used Kruskal–Wallis analysis followed by Wilcoxon pairwise post hoc tests with Monte Carlo resampling. P values were considered significant under 0.1 (Pasinelli 2000). We made this analysis in PAST version 2.17c (Hammer et al. 2001). The visualization of these particular results was made in R version 4.2.2 (R Core Team 2022) with packages tidyverse (version 2.0.0, Wickham et al. 2019), rstatix (version 0.7.2, Kassambara 2019), devtools (version 2.4.5, Wickham and Bryan 2023), ggpubr (version 0.6.0, Kassambara 2019), GmAMisc (version 1.2.1, Alberti 2020), multcompView (version 0.1–10, Graves et al. 2024) and ggplot2 (version 3.4.4, Wickham 2016).

To reveal which tree traits were crucial in the foraging tree choice of each bird species, we computed Wilcoxon rank sum tests with Monte Carlo resampling, comparing the used and unused trees (Western and Eastern). P values were considered statistically significant under 0.1 (Pasinelli 2000). For this analysis, we used PAST version 2.17c (Hammer et al. 2001).

Twenty species of tree were used by the target species for foraging. The tree species accumulation curves indicate that the majority of bird species used a wide range of species for foraging. Further, all species accumulation curves indicate that the number of records of tree species used would continue to increase as the number of first foraging observations increased. However, the treecreepers appeared to use more tree species than, e.g., the Eurasian blue tit (Fig. 2). Still, when standardizing the number of first foraging observations to a shared sample size of 40, the estimated number of tree species used did not significantly differ among the seven bird species when considering their overlapping 95% confidence intervals (Fig. 3).

The seven bird species used 20 tree species in this study. Among woodpeckers, the great spotted woodpeckers used 13 tree species, while the middle spotted woodpeckers used 9 species. Eurasian nuthatches used 14 tree species, treecreepers used 10 species. Great, Eurasian blue and marsh tits used 8, 10, and 8 tree species, respectively (Fig. 4).

According to the Jacobs' preference indices, the two tree species most preferred by individual bird species were Betula pendula and Quercus robur for great spotted woodpecker, Ailanthus altissima and Quercus robur for middle spotted woodpecker, Juglans nigra and Quercus robur for Eurasian nuthatch, Robinia pseudoacacia and Salix spp. for treecreepers, Populus alba and Quercus robur for great tit, Populus alba and Salix spp. for Eurasian blue tit and Acer campestre and Carpinus betulus for marsh tit, respectively. The full spectrum of tree species preferred by individual bird species is presented in Fig. 4.

The bimodal network of bird species preferring woody species was a completely unfragmented network, with a connectedness of 1 and a fragmentation of 0, with a transitivity of 0.662. According to the normalized degree, closeness, and betweenness centrality, the most connected tree species was Ulmus laevis, although Populus alba, P. nigra, Quercus robur, and Fraxinus excelsior also had high values (Table 2). The most connected bird species were the great spotted woodpecker, Eurasian nuthatch, and treecreeper (Table 3). According to the distance-weighted fragmentation criterion, the most important bird species was the marsh tit; hence only this species preferred Corylus avellana and Carpinus betulus. The top five key tree species from the most to the least important were Ulmus laevis, Quercus robur, Populus nigra, P. alba, and Fraxinus excelsior.

Among the three tree characteristics, we only found statistically significant differences between species in terms of tree condition (χ² = 25.559, df = 6, p = 0.00027). Based on the pairwise post hoc tests, great spotted woodpeckers used significantly more decayed trees than great tits, Eurasian blue tit, and marsh tit. Middle spotted woodpeckers and Eurasian nuthatches used significantly more decayed trees than Eurasian blue tits (Fig. 5).

We found statistically significant correspondence in the tree choice of six species, great spotted woodpecker, middle spotted woodpecker, Eurasian nuthatch, great tit, Eurasian blue tit, and marsh tit. Excluding the marsh tit, which only preferred trees with higher DBH, the other species preferred trees of higher DBH and rougher bark structure, while Eurasian blue tits also preferred less decayed trees (Table 4).

Floodplain forests are highly degraded habitats globally, yet are among the most important forest types with high conservation significance, both as green corridors for forest species and as breeding habitats for floodplain forest specialists (Havrdová et al. 2023; Machar et al. 2019; Mölder and Schneider 2011). As one of the largest free-flowing floodplain areas of Europe, Donau-Auen National Park has a crucial role in conservation as a biodiversity hotspot in the region with a high species richness of both bird and tree species, it provides an exceptional opportunity to study tree species preferences of tree-foraging bird species in such complex ecosystems (Nagl and Schulze 2017).

Although we hypothesized that particular bird species would be more specialist while others would be more generalist in terms of how many tree species each use, we did not find substantial differences among bird species in this regard. In fact, all bird species used a variety of tree species for foraging. Nevertheless, considering preference, we found substantial differences between the studied bird species. It can be stated that the number of visited tree species can be fairly higher than the number of preferred tree species (Gabbe et al. 2002; Korňan and Adamík 2017).

As we hypothesized, the bird-tree preference connections in our study revealed an unfragmented network with key members of both bird and tree species.

As we recorded, great spotted woodpeckers used a wide range of tree species, being the most generalist woodpecker species in the Western Palearctic (Gorman 2004; Ónodi et al. 2021; Stański et al. 2020, 2021a). In a study on its foraging behavior in an oak-lime-hornbeam forests of Białowieża National Park, they also preferred a wide range of tree species, e.g., aspen (Populus tremula), small-leaved lime (Tilia cordata), Norway maple (Acer platanoides), common hornbeam and pedunculate oaks (Stański et al. 2020).

Similarly to other studies (e.g., Pasinelli 2000 Stański et al. 2021b), middle spotted woodpeckers highly preferred pedunculate oaks, as old oak trees with decaying parts can have a variety of foraging sources for this woodpecker species. In spite of this, they showed the greatest preference for girdled individuals of the invasive tree of heaven. These trees, along with non-native box elder trees (Acer negundo), were killed by horizontal chainsaw cuts around the stems in a national park-wide effort from 2010 to 2014 (Quadt et al. 2016). Due to these efforts, a substantial proportion of snags and logs found nowadays in the National Park originate from this invasive tree species. We found no information in the literature for middle spotted woodpeckers foraging on Ailanthus altissima. However, there are a few records of wood-boring insect species found in the wood of this species in Hungary that can be prey of woodpeckers (Kovács 1995, 1997; Kovács and Gebei 2021). In the oak-lime-hornbeam (Tilio-Carpinetum) forests of the Białowieża National Park, Stański et al. (2021b) found that apart from pedunculate oaks, middle spotted woodpeckers also preferred Norway maple (Acer platanoides) trees, which were only found scarcely in our study area and we did not record any foraging data from this tree species. Pasinelli (2000) also found preferential use of oaks in Swiss lowland forests characterized by oak, hornbeam and lime.

As already documented, the Eurasian nuthatch occupies a broad niche in terms of tree species use (Adamík and Kornan 2004, Korňan and Adamík 2017). This was also confirmed by our study. Treecreepers are known to use a wide range of tree species (Ceia and Ramos 2016a; Korňan and Adamík 2017), similarly to our observations. Adamík and Korňan (2004) and Korňan and Adamík (2017) also reported a preference for sycamore maple by both common treecreeper and Eurasian nuthatch, for which we did not find evidence.

The unique tree preference exhibited by marsh tits may perhaps be explained by its subordinate role among Parid species. Haftorn (1993) found that marsh tits are subordinate to both great tits and also Eurasian blue tits. This may explain why we did not observe this species foraging on the scarcely found pedunculate oaks. At the same time, other studies highlight its significant role in habitat selection and winter foraging ecology of marsh tits (Broughton et al. 2014). The competitive exclusion by great tits and Eurasian blue tits relative to pedunculate oaks was also suggested by Carpenter (2008). In a coniferous-deciduous mixed mountain forests of Slovakia, in the absence of great tits and Eurasian blue tits, but in the presence of coal tits (Periparus ater) and crested tit (Lophophanes cristatus), marsh tits preferred to forage on sycamore maples (Adamík et al. 2003). In our study, we could not find any preference for sycamore maples by marsh tits, but great and Eurasian blue tits did show a preference for this particular tree species. This preference may be due to social dominance or difference in prey availability of the same tree species at different geographical locations in different seasons (Osborne and Green 1992; Travis 1977; Unno 2002). Great tits frequently used oaks (Quercus spp.) and sycamore maple trees as well in a study conducted in spring in parks with common ash, oak, sycamore maple, beech, elder, hawthorn, hazel, blackthorn (Prunus spinoza) and willow (Salix spp.) trees in Oxford (Franzreb 1984). In that study, Eurasian blue tits foraged most frequently on common ash and oak trees, while in our study they did not use oaks but preferred common ash trees. Marsh tits in Franzreb’s (1984) study foraged mostly on black elderberry and common ash trees; we also found a preference for the latter tree species. Rolando (1982) studied the autumn–winter foraging behavior of these three particular tit species in parks in the Po River valley, with dominant tree species such as pedunculate oak, common ash, and poplars. He found that marsh tits foraged on oak trees the least among tit species and foraged the most on hazel. We also found a preference for hazel.

We found the European white elm to be the most connected and essential key species for this bird-tree network, as all bird species showed preferences for this particular tree species. The European white elm is symbolic of temperate European oak-ash-elm floodplain forests. As a late successional tree species, it is susceptible to the drier climate caused by climate change, but also to the effects of river regulations with deeper riverbeds and thus a lower groundwater table (Allen et al. 2010; Hemery et al. 2010; Venturas et al. 2014). Studies on bark-foraging birds using elm trees are rare, especially in the case of the European white elm. In North America, downy woodpeckers (Dryobates pubescens) prefer dead American elm trees (Ulmus americana) in winter (Jackson 1970). In a study on the tree preferences of tree-foraging insectivorous bird species in a primeval beech-fir forests of Slovakia, the wych elm (U. glabra) was most preferred by species also considered in our study, namely Eurasian nuthatch, Eurasian treecreeper, and marsh tit (Korňan and Adamík 2017). According to Holmes and Robinson (1981) and Holmes and Schultz (1988), elms, besides other tree species such as willows and sycamores (Platanus spp.), are disturbance-dependent species that grow fast and generally have lower defenses against herbivory and therefore have higher abundance and biomass of insects. As elms have relatively low population densities but have a significant role in bird foraging, it is crucial to take the conservation of the European white elm into consideration among other secondary tree species in floodplain forest habitats (Venturas et al. 2014).

The high importance of Quercus robur is equivocal with other studies, as oak species provide a high diversity of microhabitats with essential food sources for bark-foraging birds (Domokos and Cristea 2014; Ónodi et al. 2022; Proença et al. 2010; Robles et al. 2011; Stański et al. 2021b).

In our study, species that are particularly specialized on searching for insects on and under bark showed a clear preference for decayed and dead trees. This was also the case in the study of Korňan and Adamík (2017), which was conducted in a primeval mountain mixed forest in Slovakia, where among the numerous tree-foraging bird species, only the Eurasian nuthatch and the studied two woodpecker species, namely the three-toed woodpecker (Picoides tridactylus) and the white-backed woodpecker (Dendrocopos leucotos), showed preference for dead trees. Also, in our study, besides the importance of tree species identity, there was at least a weak preference for more decayed trees in both woodpecker species compared to other bark-foraging birds. Excluding treecreepers, all species studied showed a preference for larger DBH.

In this study, apart from treecreepers, all other bird species preferred trees with a larger diameter and a rougher bark structure. In an earlier study by Ónodi et al. (2021) on the winter intersexual foraging segregation of great spotted woodpeckers in softwood willow-poplar forests in Hungary with a high presence of invasive non-native tree species, such as green ash (Fraxinus pennsylvanica) and box elder (Acer negundo), this woodpecker species preferred the rougher bark willow and poplar trees with larger DBH, compared to the invasive tree species with smoother bark surfaces. In those forests, willow and poplar species provided more foraging sources compared to non-native trees. In contrast, the floodplain forests of the Donau-Auen National Park have a high number of native species and, therefore, a wider variety of food sources. Great spotted woodpeckers also chose the larger-diameter trees with rough bark structures. Similarly to our results, in the highly diverse oak-lime-hornbeam stands of Białowieża National Park in Poland, Stański et al. (2020) found a preference of this woodpecker species for more decayed foraging trees with a greater DBH. Nevertheless, on the level of bird species, we did not find a preference for decayed trees. On the preference for larger-diameter trees, Stański et al. (2020) hypothesized that trees with larger trunk diameters can have a higher number of invertebrates than smaller-diameter trees.

Middle spotted woodpeckers preferred dead trees over living trees in the work of Stański et al. (2021b) in the oak-lime-hornbeam forest stands of the Białowieża National Park in the non-breeding season. As in that particular study, middle spotted woodpeckers preferred the rough-barked pedunculate oak and Norway spruce the most, thus, bark roughness could also be an important factor for the species, as also documented by our study.

In a study on post-fledging middle spotted woodpeckers in the Cantabrian mountains of Northern Spain, in forests where oaks were the major species with a significant amount of poplars, willows, ashes, and pines, birds used only larger diameter decayed oak, willow, and poplar trees of rougher bark (Ciudad et al. 2009), similarly to our study.

In a study on the foraging of middle spotted woodpeckers made in the Taurus mountain range in Turkey during April/May (Bergner et al. 2016), in an area of mainly pollarded individuals of four oak species, among other tree characteristics, middle spotted woodpeckers preferred to forage on trees with a greater circumference and bark roughness, as we also found.

In contrast to Korňan and Adamík (2017), who found a preference of foraging Eurasian nuthatches for dead trees in a beech-fir mixed forest in spring, we could not confirm this phenomenon in our study. This difference may be attributed to habitat and seasonal differences (Michielsen et al. 2024). In the former study, nuthatches also preferred tree species with rougher bark structure (such as sycamore maple and Norway spruce), similar to our study.

In addition to species-specific preferences for certain tree species, we found that great, Eurasian blue and marsh tits preferred to forage on trees with rougher bark. Similar results were reported by Franzreb (1984) and Rolando (1982) for the same tit species in similar forests consisting predominantly of pedunculate oak, common ash, and poplars. The three species almost exclusively used trees of intermediate or rough bark structure, such as common ash, sycamore maple, and pedunculate oak.

Studies by Fraticelli and Guerrieri (1988) carried out in a Mediterranean oak-dominated forest and that conducted by Karpińska et al. (2023) in a temperate primeval lime-oak-hornbeam and ash-alder forests showed that tree trunks and bark contributed less to the foraging properties of tit species. In the floodplain forests of our study area, all tit species were observed foraging on both the limbs and the trunk. The fact that bark roughness had a significant effect on both great and blue tits suggests that this variable plays an important role in some forest types.

According to MacFarlane and Luo (2009), who measured bark roughness of individual trees belonging to different species in an old-growth natural hardwood forest in Michigan, USA, bark roughness of individual tree species correlated with the stem DBH, suggesting that tree aging increases the amount of microhabitats for bark-living arthropods. As more fissured bark structures can harbor more arthropod prey for bark-foraging birds (Jackson 1979), it is crucial to conserve old forest stands or apply forest management practices for retaining old, aging tree individuals in commercial forests, such as green tree retention (Machar et al. 2019).