Nothing else matters? A nationwide study of microhabitats drivers at the tree scale

Managing forest structure to preserve biodiversity requires a good knowledge of the elements that actually support biodiversity as well as the driving factors of their dynamics. Tree-related microhabitats (cavities, cracks, conks of fungi) are tree-borne features that are reputed to support specific biodiversity, linked to microhabitats for at least a part of their life cycle. While several studies have analysed the drivers of microhabitats number and occurrence at the tree scale, they remain limited to a few tree species located in relatively narrow biogeographical range. Here, we used a nationwide database of forest natural reserves comprising more than 22000 trees where microhabitats have been inventoried since 2005. We analysed the effect of tree diameter and live status (alive or dead) on microhabitat number and occurrence per tree, taking into account biogeoclimatic variables and tree genus. We confirmed that larger trees as well as dead trees bore more microhabitats than their smaller or alive counterparts, and extended these results to a wider range of tree genus and conditions. Contrary to expectations, these relationships varied neither much with tree genus, with slightly higher accumulation levels for broadleaves than for conifers, nor with biogeographical context. We observed these results both for the total number of microhabitats per tree and for the occurrence of individual microhabitat types. However they were more marked for microhabitats linked with wood decay processes (e.g. dead branches or woodpecker feeding holes) than for other, epixylic, microhabitats such as epiphytes (ivy, mosses and lichens). Promoting large living and dead trees of several tree species seems a good and quite universal way to promote microhabitats and enhance potential substrates to support specific biodiversity. In addition, a better understanding of the drivers of microhabitats at the tree scale may help to better define them as biodiversity indicators for large scale monitoring.

several studies have analysed the drivers of microhabitats number and occurrence at the tree 23 scale, they remain limited to a few tree species located in relatively narrow biogeographical 24 range. Here, we used a nationwide database of forest natural reserves comprising more than 25 22000 trees where microhabitats have been inventoried since 2005. We analysed the effect of 26 tree diameter and live status (alive or dead) on microhabitat number and occurrence per tree, 27 taking into account biogeoclimatic variables and tree genus. 28 We confirmed that larger trees as well as dead trees bore more microhabitats than their smaller 29 or alive counterparts, and extended these results to a wider range of tree genus and conditions. 30 Contrary to expectations, these relationships varied neither much with tree genus, with slightly 31 higher accumulation levels for broadleaves than for conifers, nor with biogeographical context. 32 We observed these results both for the total number of microhabitats per tree and for the 33 occurrence of individual microhabitat types. However they were more marked for microhabitats 34 linked with wood decay processes (e.g. dead branches or woodpecker feeding holes) than for 35 other, epixylic, microhabitats such as epiphytes (ivy, mosses and lichens). 36 Introduction 8 these results, bearing in mind that our results were conservative despite the large number of 174 observations we analysed. In addition, we focused our interpretations on magnitude of the 175 results rather than statistical significance (see e.g. [29]). We processed all the analyses with 176 the R software v. 3.4.3 [30]. 177 178

Number of microhabitats per tree 180
Single parameters estimates were significant in the model (apart from PCA second axis 181 coordinates), while second order and third order interactions were less often and less 182 significant (Supplementary Materials, Table S1). All tree genus except Pine had higher 183 microhabitat number on dead than living trees. Overall, the difference was higher for Oak and 184 Pine (resp. 50% and 43% more on dead trees for a mean DBH, Table 3), than for the other 185 genus (around 30% more on dead trees). Globally, number of microhabitats per tree increased 186 with tree diameter both for live and dead trees. However, the accumulation of microhabitat with 187 diameter varied with genus, with higher accumulation levels for broadleaves (Beech and Oak) 188 than for conifers (Fir, Pine, Spruce), but also for dead compared to living trees (except for Pine; 189  Table S2). 190

Occurrence of microhabitat types per tree 191
Five microhabitats out of twenty showed generally higher occurrence on standing deadwood 192 than on living trees, but not systematically for all species or for all live status: trunk cavities 193 (broadleaves), woodpecker feeding holes, rot (broadleaves), conks of fungi (except Pine) and 194 bark characteristics (except Pine and Spruce, Table 3 and Appendix S3). Conversely,injuries,195 dead branches whatever their size, and forks (broadleaves) showed higher occurrence on 196 living trees. The strongest interpretable differences were observed for woodpecker cavities 197 (e.g. they occurred around 300% more often on standing dead Beech, Oak and Pine, for a 9 mean DBH = 44cm). Magnitudes for microhabitats which occurred more on living trees were 199 smaller, e.g. for small branches or injuries (around 60% to 90% more on living trees, Table 3). 200 For most microhabitats, probability of occurrence increased with DBH, either for living or dead 201 trees with the remarkable exceptions of woodpecker cavities, cracks and crown skeletons 202 (Supplementary Materials: Figure S2, Table S3). However, the magnitude of the relation varied 203 with tree genus and live status, the increase in occurrence with DBH being higher for dead 204 than for living trees (e.g. 30% more base and trunk cavities on dead Beech, 22 to 44% more 205 woodpecker feeding holes, except on Pine). For living trees, the magnitude was generally 206 smaller, except for occurrence of small and medium branches (e.g. 70% more medium dead 207 branches on living Pine) and to a lesser extent for mosses on Beech and Fir (18% and 23% 208 more respectively). All other magnitudes were smaller, generally below 10%. Note that in some 209 cases, due to the very small occurrence of some microhabitats on some tree genus (e.g. 210 canopy cavities on Spruce), the estimations proved unreliable in these cases (Supplementary 211 Materials: Figure S2, Table S3). 212 213

214
Numerous recent studies in various contexts showed that the number of microhabitats per tree, 215 as well as the occurrence of some types increase with tree diameter [10,13,15] and showed 216 higher levels on dead than living trees [11,12] . Our nationwide study based on a large tree 217 database confirmed these relationships and extend them to a larger range of tree species in 218 various biogeographical conditions than before. Indeed, our results concerned at least five tree 219 genus (eleven if we take only living trees into account, Supplementary Materials: Figure S1). 220 221 Dead trees bear more microhabitats than living trees 222 Standing dead trees contribute significantly to the supply of microhabitats, as they overall bore 223 30 to 50% more microhabitats than their living counterparts in our dataset. Dead trees could 224 even bear a lot more microhabitats than living trees when individual types are analysed (e.g. 225 woodpecker feeding holes or bark characteristics). Previous studies comparing microhabitat 226 number between living and dead trees almost all found higher microhabitats numbers on dead 227 trees (see [17]). However, this difference varied across studies, from 1.2x more microhabitats 228 in Mediterranean forest [15], 2x more in five forests in France [12], to 4x more on habitat trees 229 in south-western Germany [31]. Our results ranged from 1.3x to 1.5x more microhabitats on 230 dead than living trees, which is of a slightly lower order of magnitude than what was observed 231 before, but on a larger geographical gradient. Once dead, standing trees are affected by 232 decomposition processes that initiate and develop microhabitats [14,32,33]. Such trees could 233 also constitute privileged foraging grounds for a number of species [5,7,19], including for 234 example woodpeckers [33,34]. In particular, insect larvae or ants that live below the bark of 235 more or less recently dead trees constitute a non-negligible part of some birds' diet [7,35,36]. 236 As living trees also bear microhabitats, it seems logical that many of them persist when the 237 tree dies and continue to evolve, or even condition the presence of other microhabitats linked 238 with the decay process [14]. Injuries caused by logging, branch break or treefall could slowly 239 rot and evolve in decayed cavities [5,32]. These successions likely explain why these 240 microhabitats are more numerous on dead trees. The only exception to this global pattern 241 concerned epiphytes and forks with accumulated organic matter, that tend to be more 242 numerous on living trees. Ivy, mosses and lichen are likely to benefit from bark characteristics 243 and conditions (e.g. pH, [37]) likely to occur only on living tree. In addition, epiphytes require a 244 relatively stable substrate to grow or anchor, especially when they grow slowly like some 245 species of mosses or lichens [38]. Such property is lost when the tree dies as the bark loosen 246 and falls more rapidly than on living trees, which could cause epiphytic community replacement 247 as well as lower levels of detection due to the absence of individuals. In a nutshell, decay 248 processes linked to the tree death makes a clear difference between microhabitats that are 249 linked to it (i.e. saproxylic microhabitats, sensu [5]) and those that are notor lesslinked to 250 those phenomena (i.e. epixylic microhabitats). 251 252 Number and occurrence of microhabitats increase with tree diameter 253 We confirmed that microhabitat number and occurrence increase with tree diameter but, 254 contrary to expectations, tree genusas well as abiotic factorshad a limited effect on this 255 relationship, with slightly higher microhabitat accumulation levels on broadleaves than conifers 256 ([10-12], but see [13]). At the individual microhabitat level, almost all types showed the same 257 trend, but also with considerable variations in terms of magnitude. Larger (living) trees have a 258 generally longer lifespan than smaller ones, and are consequently more prone to damages 259 due to meteorological events (storms, snowfall), natural hazards (rockfalls) or attacks and use 260 by different tree-and wood-dependent species (woodpeckers, beetles, fungi, e.g. [12,39]). 261 Depending on the studies, for a comparable increase in tree diameter (from 50 to 100cm), 262 number of tree microhabitats was roughly multiplied by two in several studies [12,16,17], but 263 can be multiplied by four [31] up to five [11] in certain cases. Our results showed magnitudes 264 below the lower end of this range (the multiplication coefficient ranged from 1.2 to 1.4). This is 265 probably linked to the fact that the large trees in our dataset may be younger that those in the 266 other studies, especially compared to studies located in near-natural or long-abandoned 267 forests [11,12]. At the individual microhabitat scale, dead branches were more prone to occur 268 on large trees than smaller trees, which seems quite obvious but has rarely been quantified 269 before: larger trees have more, but also larger, branches likely to die from competition with 270 neighbours, especially broadleaves [40]. Indeed, Oak and Beech were the tree genus that 271 showed higher large dead branches accumulation rates in our analyses, while conifers showed 272 almost no large dead branches. Cavity bird and bats are reputed to choose preferentially larger 273 trees to nest or roost [41,42], since larger wood width around the cavity provides buffered and 274 more stable conditions [43]. However, this relationship was not the best shown in our results, 12 since the accumulation rate of woodpecker cavities with tree diameter was very slow. This 276 absence of relationship between tree diameter and woodpecker cavities seems hard to prove 277 in the context of temperate European forests (see [12] at the tree scale, or [6] at the stand 278 scale) and probably require more targeted examination [33,44]. This could also be linked to 279 non-linear dynamics [10] of this particular microhabitat (some cavities in living trees can close 280 back when they are not used anymore) but also for other microhabitats with specific phenology 281 like conks of fungi [45]. The number and occurrence of microhabitats also increased with 282 diameter of standing dead trees, sometimes at a higher rate than for living trees. In this case, 283 the longer persistence of large dead trees compared to smaller ones [46,47] may combine the 284 effects of increased hazard and damage risks with the decay processes described above. This 285 probably explains the higher accumulation levels we observed in many cases, especially for 286 saproxylic microhabitats (e.g. rot, feeding holes, trunk cavities). Once again, the only exception 287 to this rule was epiphytes: their probability of occurrence tended to increase with tree diameter 288 but in a very noisy and unclear way, both for living and for dead trees. For such epiphytic 289 organisms, larger scale processes and biogeoclimatic (e.g. soil fertility, precipitation) context 290 is probably more important than local tree characteristics [48,49]. 291 292

Limitations and research perspectives 293
We showed a limited effect of biogeoclimatic variables on the relationship between 294 microhabitats, tree diameter and living status. However, the way we controlled for them in the 295 models remains rather imperfect. Some specific interactions may exist, especially in the case 296 of epiphytes [49], but could not be detected by our approach with aggregated variables. In 297 addition, it was rather difficult to disentangle the effects of tree genus with that of biogeoclimatic 298 variables, since distribution range of most tree species we analysed is linked to a climatic 299 niche, apart from Beech and more marginally for Pine that occur over larger gradients. Still, 300 the fact that we did not highlight any clear interaction with biogeoclimatic variables during 301 exploratory analyses tend to confirm that the relations we observed are valid for a wide range 302 of species. However, further analyses are required to assess the effects of biogeoclimatic 303 13 variables on microhabitat patterns for the species with a large ecological amplitude (especially 304 Beech species, that occur over temperate and Mediterranean Europe and beyond). 305 Our data is issued from nature reserves with a potentially larger anthropogenic gradient than 306 managed forests. Some of these reserves have not been harvested for several decades and 307 exhibit characteristics of overmature forests (see e.g. [22], who analysed some of the reserves 308 included in this paper), but their overall structure reflects a relatively recent management 309 abandonmentif anyprobably marked by previous intensive harvesting and use over the 310 past centuries characteristic of western European forests [50]. This is testified by the relatively 311 rare occurrence of dead standing trees, in particular those with a large diameter, in the dataset 312 we analysed: standing dead trees represented a mere 10% of the total dataset while very large 313 individuals (DBH > 67.5cm) only 1% ( Table 2). As a consequence, despite the fact that we 314 worked on an extended management gradient including unmanaged strict reserves, we still 315 lack a part of the elements characteristic of old-growth and overmature forests, especially large 316 dead trees [20,51], which cause our relationships to be truncated and imprecise for the larger 317 diameter categories. Further research on the last remnant of old-growth primeval forests in 318 Europe [52] is thus needed to bridge this gap and better understand microhabitats dynamics 319 over a whole life of a tree. 320 Compared to recent developments [5,21], the typology we used (Table 1) appears rather 321 coarse and imprecise. But, on the one hand, it allowed us to have a sufficient number of 322 occurrences in each types to analyse the combined effects of diameter and species for almost 323 all microhabitat types in the typology. On the other hand, it is also likely that we were not able 324 to confirm some effects mentioned in the literature due to imprecise distinctions between types, 325 for example different woodpecker cavity types. The current approach should then be viewed 326 as a compromise between sufficient occurrence of each microhabitat type in the dataset and 327 specificity of the typology. Current developments mentioned above [5] will certainly help to 328 homogenize data in a near future and to build larger shared databases on common and 329 comparable bases. 330 Finally, our models assumedunrealisticallymicrohabitat number to increase exponentially 331 with diameter. Recent studies [17], as well as ecological theory (e.g. species-area 332 relationship), tend to rather show a saturated (e.g. logarithmic or sigmoid) relationship between 333 microhabitats and diameter. Models allowing for different link functionsprobably within a 334 Bayesian frameworkremain to be tested to see whether they perform better than the current 335 ones used (see e.g. [10]). 336 337 Implications for forest management and biodiversity conservation 338 Among small natural features, large and old trees are considered a keystone in forest and 339 agro-pastoral landscapes because of their disproportionate importance for biodiversity 340 relatively to their size [3]. This functional role for biodiversity seems further enhanced by the 341 'smaller' natural featuresmicrohabitatsthey bear. In this large-scale analysis, we confirmed 342 and extended some of the results already observed locally: most microhabitats occur 343 preferentially on living large trees and even more on dead ones. This relationship seems true 344 for several tree species included in this analysis, and across a large gradient of ecological 345 conditions, with minor differences in terms of accumulation rates. As a consequence, 346 conserving and promoting large trees in daily forest management is likely to enhance the 347 structural heterogeneity at the stand scale [20,53], including a variety of tree-borne 348 microhabitats, that could further help to better conserve specific forest biodiversity [5,54]. We are in debt to the reserve managers who feed the database and made this study possible. 360 Without their commitment and implication in their daily management, such results would not 361 have been achievable. We thank C. Bennemann for her work on GIS data extraction and R. 362 Andrade for his invaluable help with ggplot issues.   Ivy (>50) 7.9 Small branches (5-10cm) Dead branches with a diameter comprised between 5 and 10cm and a length higher than 1m 28.4 Medium branches (10-30cm) Dead branches with a diameter comprised between 10 and 30cm and a length higher than 1m 13.3 Large branches (>30cm) Dead branches with a diameter higher than 30cm and a length higher than 1m         Figure S1: Relationship between total number of microhabitats per tree and Diameter at Breast Height (DBH) 548 by species and live status (living vs. dead standing trees). The line represents the estimation issued from a 549 generalized mixed effect mode with Poisson error distribution. The ribbon represents the 95% confidence 550 interval of the mean. Principal component analysis eigenvalues were hold constant for the representation. 551 552 Figure S2: Relationship between occurrence of microhabitats per tree and Diameter at Breast Height (DBH) 553 by species and live status (living vs. dead standing trees). The line represents the estimation issued from a 554 generalized mixed effect mode with Binomial error distribution. The ribbons represent the 95% confidence 555 interval of the mean. Principal component analysis eigenvalues were hold constant for the representation.