Variation in Deadwood Microsites in Areas Designated under the Habitats Directive (Natura 2000)

: The continuing decline in biodiversity presents a major environmental protection challenge. The conservation of su ﬃ ciently extensive and diverse habitats requires an array of coordinated actions, often involving large areas. While a set of conservation objectives have been deﬁned for the Natura 2000 network, no universal methods of accomplishing them have been speciﬁed, and so they must be designed by individual Member States. Deadwood volume and the density of large deadwood pieces are widely used for evaluating the quality of forest habitat types designated under the Habitats Directive. In the present study, data from 5557 sample plots were used to evaluate the mean values of the two deadwood indicators as well as the ratio of deadwood volume to living tree volume for each of the 13 habitat types in Poland. In addition, a logistic regression model was constructed to evaluate the e ﬀ ects of terrain, site, and tree stand characteristics as well as protection type on deadwood volume in Natura 2000 areas. Mean deadwood volume varied greatly between habitat types, with the lowest values found for Central European lichen Scots pine forests (91T0–2.5 m 3 ha − 1 ) and Old acidophilous oak woods (9190–4.4 m 3 ha − 1 ), and the highest for Riparian mixed forests (91F0–43.1 m 3 ha − 1 ) and Acidophilous Picea forests of the montane to alpine levels (9410–55.4 m 3 ha − 1 ). The ratio of deadwood volume to living tree volume ranged from approx. 1%–17%. Additionally, the presence of large deadwood di ﬀ ered among habitat types: in some, there were no deadwood pieces with a diameter of ≥ 50 cm, while their maximum density was 6.1 pieces ha − 1 . The logistic regression model showed that the likelihood of a habitat type to have a ‘favorable conservation status’ as deﬁned by deadwood abundance (a threshold of at least 20 m 3 ha − 1 according to Polish manuals on habitat type evaluation) increased with sample plot elevation, site fertility, and moisture, as well as stand age and volume. Positive e ﬀ ects were also observed for forests under strict and active protection versus managed forests. Planned e ﬀ orts are necessary to enhance the quality of habitats with insu ﬃ cient deadwood, especially in managed forests. Special attention should be given to areas that are readily accessible due to gentle terrain and low site moisture. Furthermore, younger stands on less fertile sites may require intervention to promote deadwood accumulation. We recommend retaining a certain proportion of mature stands until natural death and decomposition. Increasing the density of large deadwood is currently one of the most pressing conservation needs in most habitat types.


Introduction
Europe boasts the largest network of coordinated conservation areas in the world, known as Natura 2000, which covers more than 18% of the land area of EU Member States and almost 10% The criteria concerning deadwood volume and the density of large deadwood may differ depending on the specific characteristics of a given habitat. Those criteria are specified in manuals on habitat evaluation in Poland [10,27]. In most cases, a favorable conservation status requires a deadwood volume of >20 m 3 ha −1 and a density of at least 3-5 large deadwood pieces per ha; large pieces are understood as those having a diameter/DBH of >50 cm (or in some cases >30 cm) and a length/height of >3 m ( Table 2). Medium-sized trees are defined as those with a diameter/DBH in the range of 30-49.9 cm, and large trees as those with a DBH of ≥ 50 cm.
In some cases, the deadwood volume threshold in a given habitat may be expressed as a proportion of stand volume rather than in absolute values. Furthermore, in some habitat types, such as 91T0, the presence of deadwood is generally undesirable given the adopted conservation priorities. This is due to the fact that large amounts of deadwood on the forest floor lead to rapid enrichment of the substrate in biogenic substances, thus increasing the competition of bryophytes and herbaceous plants to the detriment of terricolous lichens. While small amounts of deadwood are not harmful, large quantities of twigs and branches left, e.g., in the aftermath of management procedures, may cause habitat degradation [27]. Asperulo-Fagetum beech forests 566 9140 1 Medio-European subalpine beech woods with Acer and Rumex arifolius 4 9160 Sub-Atlantic and medio-European oak or oak-hornbeam forests of the Carpinion betuli 52 Forest ecosystems were evaluated on the basis of data obtained in the years 2010-2014 from NFI sample plots located within Natura 2000 sites under the Habitat Directive, also known as special areas of conservation (SACs). NFI measurements were conducted only for plots located on sites classified as forest areas pursuant to Polish regulations. In the study period each sample plot was measured once [23]. To determine which sample plots should be included in the study, the authors used a spatial dataset in the form of ESRI Shapefile layers including: • vector data for the location of NFI sample plots; • SAC database containing vector and descriptive data concerning forms of nature conservation [24]; and • vector and descriptive data for tree stands from the Forest Data Bank [25].
These input data were integrated using Qgis 2.14 software to ensure information compatibility and generate an information layer for tree stands and sample plots within the boundaries of Natura 2000 sites. As a result, 312 Natura 2000 sites (SACs) with an overall area of 3,102,247 ha, and with a total of 5557 NFI sample plots located on them, were available for the study. Those SACs included the habitat types listed in Annex I and the habitats of species listed in Annex II. The next step involved the identification of the location and type of the various Natura 2000 habitats on the aforementioned SACs [7], as well as the NFI sample plots within their boundaries. For that purpose, we used Standard Data Forms providing information about the habitat types present on a given site, as well as conservation plans (or drafts of those plans) specifying their location. For Natura 2000 sites without protection plans, natural habitats were identified using methodological keys based on taxonomical descriptions of tree stands [26], as well as the available literature, naturalists' notes, and manuscripts concerning a given Natura 2000 site. In total, the 312 selected Natura 2000 SACs were found to contain 15 habitat types (including four priority types) with a total area of 711,306 ha. While 1620 NFI sample plots were located within the boundaries of 14 of those habitat types (see Table 1), as many as 3937 sample plots present in Natura 2000 SACs were found in habitats not listed in Annex I to the Habitat Directive [7]; in this work the latter were designated as "no habitat type" (NHT). In addition, the Web Map Service made available by the Main Office for Surveying and Cartography was used with the C-GEO software package (with web connection) to assess elevation above sea level for each sample plot. Elevation was determined in accordance with the Polish system PL-KRON86-NH by means of interpolation algorithms prepared on the basis of the Numerical Terrain Model with a 1 × 1 m grid.
The criteria concerning deadwood volume and the density of large deadwood may differ depending on the specific characteristics of a given habitat. Those criteria are specified in manuals on habitat evaluation in Poland [10,27]. In most cases, a favorable conservation status requires a deadwood volume of >20 m 3 ha −1 and a density of at least 3-5 large deadwood pieces per ha; large pieces are understood as those having a diameter/DBH of >50 cm (or in some cases >30 cm) and a length/height of >3 m ( Table 2). Medium-sized trees are defined as those with a diameter/DBH in the range of 30-49.9 cm, and large trees as those with a DBH of ≥ 50 cm.
In some cases, the deadwood volume threshold in a given habitat may be expressed as a proportion of stand volume rather than in absolute values. Furthermore, in some habitat types, such as 91T0, the presence of deadwood is generally undesirable given the adopted conservation priorities. This is due to the fact that large amounts of deadwood on the forest floor lead to rapid enrichment of the substrate in biogenic substances, thus increasing the competition of bryophytes and herbaceous plants to the detriment of terricolous lichens. While small amounts of deadwood are not harmful, large quantities of twigs and branches left, e.g., in the aftermath of management procedures, may cause habitat degradation [27].

Data Analysis
Data from sample plots were used to determine the mean deadwood volume for the studied habitat types. The volume of living trees and the proportion of deadwood volume to living tree volume were assessed for each habitat type to account for variation in site productivity. Then, the density and variability of medium-sized and large deadwood pieces were evaluated. Due to NFI methodology and the definition of medium-sized and large deadwood pieces (Table 2), the diameter of logs at the thicker end was calculated based on measurements taken halfway along their length (adopting a mean taper of 1 cm per 1 m). Statistical differences between habitat types were evaluated by analysis of variance and the Kruskal-Wallis test implemented in Statistica 13 software (StatSoft, Kraków, Poland).
The factors affecting deadwood volume in the studied habitat types were evaluated using a logistic regression model [28]. The choice of that statistical tool was dictated by the uneven distribution of deadwood, whose volume varied greatly among sample plots and which was absent from approx. half of them. The logistic regression model used a dichotomous dependent variable [28]. Sample plots with a deadwood volume of >20 m 3 ha −1 were assigned the value of 1, with 0 assigned to other plots. The 20 m 3 ha −1 threshold corresponds to the favorable conservation status as defined for most of Natura 2000 habitat types. The adopted independent variables were factors that may affect deadwood volume, such as stand, terrain, and site characteristics, as well as protection type, also obtained from the NFI. The protection type variable assumed three values: active, strict, or managed forest. For the purposes of this work, 'managed forests' are defined as forest areas that are managed with no active or strict protection plans. In turn, active and strict protection plans are most often used in nature reserves and national parks. Terrain was described by two variables: elevation above sea level (m a.s.l.) and the percentage slope of sample plots (%). Tree stands were characterized by the age of the dominant tree species (years), the volume of living trees (m 3 ha −1 ), and tree density (trees ha -1 ). The model also included site fertility (dystrophic, oligotrophic, mesotrophic, eutrophic) and moisture (mesic, moist, boggy), as those parameters varied considerably among the studied habitat types. Another independent variable was habitat type, operationalized by assigning one of 13 Natura 2000 habitat codes, or "no habitat type" (NHT) for sample plots in habitats not included in Natura 2000.   3 This value refers exclusively to downed deadwood in relation to stand volume, e.g., <5% means that downed deadwood volume amounts to less than 5% of stand volume. 4 In relation to stand volume. 5 d.n.a.-not available.
The model was built using the step-wise forward method (a model constructed using the step-wise backward method arrived at the same set of significant variables). The odds ratio was calculated to characterize the effects of independent variables on the dependent variable. In the case of quantitative independent variables, an increase or decrease by one unit increased or decreased the probability for the dependent variable to assume the value of 1 by the odds ratio. In the case of qualitative variables, Forests 2020, 11, 486 6 of 13 the odds ratio was adopted in the form of reference values for each of them. The independent variables were tested for intercorrelations. The quality of the model was evaluated by means of Nagelkerke values and the Hosmer-Lemeshow test [31]. A successful classification test was carried out on the basis of observations used to estimate the parameters of the model [32].

Results
The mean deadwood volume for the entire Natura 2000 area was 12.7 m 3 ha −1 , with very large differences between habitat types (Kruskal-Wallis H = 235.7; p < 0.05). The lowest deadwood volume was found for Central European lichen Scots pine forests (91TO-2.5 m 3 ha −1 ) and Old acidophilous oak woods (9190-4.4 m 3 ha −1 ), and the highest for Riparian mixed forests (91F0-43.1 m 3 ha −1 ) and Acidophilous Picea forests of the montane to alpine levels (9410-55.4 m 3 ha −1 ). Furthermore, substantial variation was recorded for sample plots located within one habitat type, as reflected by very high standard error values (Table 3). The mean volume of living trees for the entire Natura 2000 area was 305 m 3 ha −1 . Among the habitat types, the lowest values were found for bog woodland (91D0-238 m 3 ha −1 ) and Central European lichen Scots pine forests (91T0-282 m 3 ha −1 ). Volumes in the range of 300-400 m 3 ha −1 were recorded for seven habitat types, with the highest value for Sub-Atlantic and medio-European oak or oak-hornbeam forests of the Carpinion betuli (9160-478 m 3 ha −1 ), with differences between habitats often reaching statistical significance (Kruskal-Wallis H = 301.7; p < 0.05, see Table 3). The ratio of deadwood volume to stand volume ranged from approx. 1% in 91TO and 9190 to 11.4% in 91F0 and 17.0% in 9410 (Table 3).
Euro-Siberian steppic woods with Quercus spp. (91I0) and in Central European lichen Scots pine forests (91T0). In the other habitat types its mean density ranged from 2.5 pieces ha −1 (9190) and 2.7 pieces ha −1 (91D0) to 16.3 pieces ha −1 (9130) and 31.1 pieces ha −1 (9410) (Figure 3). Differences were also found when analyzing the density of large living trees only (DBH of ≥50 cm, Kruskal-Wallis H = 878.1; p < 0.05), which ranged from 4 trees ha −1 for 91T0 to 61 trees ha −1 for 9180, with a mean of 16 trees ha −1 (Figure 2). Large deadwood pieces were absent in a total of five habitat types. In the other habitat types their mean density ranged from 0.4 pieces ha −1 (9170) and 0.8 pieces ha −1 (91E0) to 5.6 pieces ha −1 (91F0) and 6.1 pieces ha −1 (9410) (Figure 3).    The inclusion of factors other than habitat type in the analysis (Table 4) substantially changes the picture of deadwood accumulation compared to analysis based exclusively on habitat types ( Table 3). The logistic regression model indicates that deadwood volume is mostly determined by factors associated with terrain and site accessibility, type of protection, as well as soil and stand parameters (Table 4). From among the nine variables entered in the model, the slope of sample plots and living tree density failed to reach statistical significance, while elevation above sea level,  The inclusion of factors other than habitat type in the analysis (Table 4) substantially changes the picture of deadwood accumulation compared to analysis based exclusively on habitat types ( Table 3). The logistic regression model indicates that deadwood volume is mostly determined by factors associated with terrain and site accessibility, type of protection, as well as soil and stand parameters (Table 4). From among the nine variables entered in the model, the slope of sample plots and living tree density failed to reach statistical significance, while elevation above sea level, protection type, site fertility and moisture (water abundance), the age of the dominant tree species, and the volume of living trees were significant. An increase in the quantitative variables (elevation, stand age, living tree volume) was associated with an increase in the odds ratio, or the likelihood of finding a favorable deadwood volume (>20 m 3 ha −1 ) in a given habitat type. In terms of site conditions, the reference value corresponded to sites poorest in nutrients. The odds ratio increased with both site fertility and moisture; in the latter case the highest odds ratio was recorded for boggy sites. There was a substantial difference in the odds ratio between managed forests and areas subjected to either active or strict protection. A significant effect was also found for habitat type, which was the last element entered into the model. No habitat type exhibited a significant difference as compared to the reference value in the model (Central European lichen Scots pine forests-91T0).

Discussion and Conclusions
It is crucial to develop appropriate strategies for Natura 2000 sites to aid policymakers and managers in reaching biodiversity targets [33,34]. Despite an increase in Europe's afforestation, it is estimated that only 15% of its woodland qualifies for a favorable conservation status [35]. The study provides a general overview of the conservation status of forest habitats in Poland. The use of a large number of sample plots made it possible to determine mean values for 13 habitat types. However, it should be noted that the status of a given habitat type may vary between different SACs. The statistical method applied by the NFI, employing a random, evenly distributed network of sample plots, precludes the evaluation of individual SACs due to an insufficient number of representative sample plots in each of them. Nevertheless, it is an excellent, objective tool providing a general characterization of Natura 2000 sites. In addition, logistic regression analysis revealed the site and stand characteristics that have a positive or negative effect on deadwood accumulation in areas designated under the Habitats Directive. Knowing the characteristics of individual SACs and the factors conducive to deadwood accumulation, one can predict deadwood volume for the various areas.
In Poland, deadwood thresholds adopted for most habitat types are 20 m 3 ha −1 for favorable conservation status and 10-20 m 3 ha −1 for unfavorable-inadequate status. While this is supported by some publications, those thresholds represent the lower limits of deadwood ranges proposed for European forests. Indeed, papers on the conservation of various saproxylic species or groups of species tend to suggest thresholds of 30-50 m 3 ha −1 , or even more [17,36,37]. In addition to the quantitative criterion, it is also necessary to ensure variability in deadwood types [38] as well as an adequate spatial distribution of deadwood microsites [16,39]. In some cases, a deadwood volume threshold may be expressed in terms of its proportion relative to stand volume. In the present study, the deadwood volume threshold of 20 m 3 ha −1 amounted to only a few percent of the mean stand volume (approx. 300 m 3 ha −1 ).
National parks and nature reserves, which are almost exclusively subjected to strict or active protection, revealed a markedly higher likelihood of reaching the threshold deadwood volume. However, the overall area of parks and reserves is relatively small (approximately 4% of the afforested area of Poland) as compared to that of managed forests. Given the well-established differences between managed and unmanaged woodland [40][41][42], it is little wonder that favorable volumes of deadwood as defined under Natura 2000 are usually found in the latter. In turn, in managed forests, deadwood volume mostly depends on the adopted management principles and their implementation. Taking into account the specific features of a given site, management procedures are determined by the species composition of the stand, its functions, as well as management objectives [43]. The implementation of different felling systems, management interventions, and regeneration patterns may result in significant differences in deadwood volume between sites [44][45][46]. In the present study, the average deadwood volume on Natura 2000 sites was 12.7 m 3 ha −1 , which is more than twice higher than the mean volume reported for all Polish forests (5.9 m 3 ha −1 , NFI 2014). This is attributable to the fact that the Natura 2000 network primarily encompasses the best preserved woodlands in the country, including protected areas. A general assessment of Natura 2000 sites (not only forests) conducted in the years 2017-2018 as part of a periodic monitoring program revealed a declining proportion of sites with a favorable status and an increase in unfavorable-inadequate and unfavorable-bad sites [47]. In the case of some forest habitats (e.g., 91F0), general conservation status deteriorated substantially due to adverse quantitative and qualitative changes in the floristic composition, the presence of alien species, as well as hydrological disturbances [47,48]. Furthermore, it has been reported that the conservation status of many sites has been affected by excessive deadwood removal; of particular concern is the scarcity of large deadwood pieces [47].
While the management difficulty indices calculated for Polish montane and lowland forests are highly varied [49], the terrain factor was found to be significant in the model, suggesting that the higher the site elevation the higher the likelihood of finding more deadwood. A large proportion of Natura 2000 woodland sites are located in mountainous areas. Habitat types that are in part or in their entirety represented by such sites and those which are otherwise associated with steep slopes (9130, 9180, 9410) exhibited higher deadwood volumes. Additionally, monitoring reports have indicated a much better quality of forest habitats in the Alpine biogeographical region as compared to the continental region. While the conservation status of mountainous sites is usually classified as favorable or unfavorable-inadequate, that of sites in the continental region is more often deemed unfavorable-inadequate or unfavorable-bad [47]. In managed forests, higher deadwood accumulation is significantly promoted by harvesting and skidding difficulty as well as by a less dense road network [50,51], entailing higher operating costs. Site accessibility also plays a role in lowland areas, but probably to a lesser extent [52].
While the mean deadwood volume varied considerably between different habitats, terrain and stand characteristics were of primary importance. The presented model indicated a significant contribution of stand age: the older the stand, the higher the likelihood of the site reaching the deadwood volume threshold. Since the Natura 2000 network has a relatively short history in Poland, the age structure of stands at the time of their inclusion continues to play a major role in habitat evaluation. Indeed, in the case of some habitat types this may partially explain the low deadwood volume and density of large dead trees. A good case in point are boggy coniferous forests, which were often represented by young stands at the time of their inclusion in Natura 2000, and so they have not had the time to accumulate enough large deadwood [27]. Nevertheless, stand structure analysis indicates quite high current mean densities of living trees with DBH ≥30 cm for all habitat types. Although boggy coniferous forests still reveal lower values, in the coming years they should add more large deadwood as long as they are appropriately managed. Moreover, although trees with DBH ≥50 cm are found in all habitat types, their distribution is much more irregular than that of medium-sized trees. This may be attributable to many factors, such as the adopted rotation period in managed forests or site conditions that determine the growth capacity of trees in individual habitat types. Large deadwood, which is particularly important for supporting biodiversity, is scarce or absent in many habitats. Since deadwood is deemed a crucial structural forest indicator [12], an improvement in that parameter is a crucial target that should be pursued with a view to enhancing the quality of Natura 2000 forest habitats.
In the present study, the habitat type with the greatest mean deadwood volume was Acidophilous Picea forests of the montane to alpine levels, although the high standard error points to an irregular distribution pattern with local aggregations attributable to frequent biotic and abiotic disturbances [53]. It should be noted that disturbances fulfill an important role in biodiversity promotion as long as the dying and dead trees are retained in the ecosystem, which is often the case in this habitat type due to its location in poorly accessible mountainous regions within the boundaries of Polish national parks. The diverse array of niches afforded by deadwood provide suitable microhabitats, shelter, and nutrition for a variety of species, increasing their numbers in a given area [54,55]. Other habitat types also contain trees which currently tend to exhibit high mortality, such as Fraxinus excelsior L., which occurs as an accompanying species in habitat types 91E0 and 91F0 [52,56].
The above notwithstanding, it should be noted that deadwood is not desirable in some habitats (91T0 and 91I0) as it may interfere with the conservation of the priority species occurring in them [27] due to the chemical properties of decomposing wood and its role in nutrient cycling and soil forming processes [57][58][59]. Indeed, both in 91T0 and 91I0 the mean deadwood volume is among the lowest with no medium-or large-sized deadwood despite the presence of large trees. However, a decision to protect certain species (e.g., rare lichens) by removing deadwood from a given habitat to prevent site eutrophication should be compensated on other sites as some saproxylic organisms have very specific requirements concerning deadwood type and other site conditions, such as insolation [60,61].
The applied regression model indicates that, in addition to protection type, deadwood volume is mostly influenced by terrain conditions, site fertility and moisture, stand age, and living tree volume. Analysis involving a dichotomous dependent variable for deadwood volume with a threshold value of 20 m 3 ha −1 shows that appropriate deadwood management should mitigate the effects of the aforementioned independent variables, or at least decrease their odds ratio. The factors that were found significant in the model were generally attributable to the "forces of nature." No sizable effects of management interventions were found for readily accessible terrain and for sites characterized by low growing stock. Thus, it is necessary to design a strategy for those habitats where deadwood is desirable and where standard management procedures and natural disturbances are insufficient to ensure favorable conservation outcomes. Particularly problematic is the scarcity of large deadwood. Therefore, in managed forests fragments of saw timber stands should be left to die naturally and decay. Further monitoring is necessary as the evaluation of the Natura 2000 network depends both on its duration in individual Member States and on the adopted conservation principles for the included areas.