Continuous cover forestry is a cost-efficient tool to increase multifunctionality of boreal production forests in

All material supplied via JYX is protected by copyright and other intellectual property rights, and duplication or sale of all or part of any of the repository collections is not permitted, except that material may be duplicated by you for your research use or educational purposes in electronic or print form. You must obtain permission for any other use. Electronic or print copies may not be offered, whether for sale or otherwise to anyone who is not an authorised user. Continuous cover forestry is a cost-efficient tool to increase multifunctionality of boreal production forests in Fennoscandia Peura, Maiju; Burgas Riera, Daniel; Eyvindson, Kyle; Repo, Anna; Mönkkönen, Mikko


INTRODUCTION 39
Forests are crucial in delivering ecosystem services for human wellbeing. During the last decades many 40 forests in the boreal zone have been managed for intensive timber production applying conventional even-41 solely on timber production reduces carbon storage in boreal forests compared with optimal forest 49 management (Triviño et al., 2016). Focusing on timber production can also be in conflict with other 50 economically beneficial forest uses, such as recreation and harvest of non-timber forest products (e.g., 51 berries and mushrooms) (Peura et al., 2016). Earlier research has shown that diversifying forest initial forest data was provided by the Finnish Forest Centre, and are based on laser scanned data with 117 ground-truthing (Maltamo et al., 2007). The data contain forest characteristics, such as forest site type, 118 age, or tree species compositions. In the initial data, Scots pine (Pinus sylvestris) was the dominant tree 119 species on 23% of the stands, Norway spruce (Picea abies) on 63% of the stands and birches (Betula 120 pendula and B. pubescens) on 14% of the stands. Mixed stands, i.e., where none of the tree species 121 accounted for more than 75% of the total volume, represent 45% of all the stands. The variation in the 122 site type and initial age of stands are given in Appendix S1: Figure S1. 123 The development of each stand was simulated 100 years into the future using SIMO-forest 124 simulator (Rasinmäki et al., 2009) under three different forest management regimes: CCF, RFM, and no 125 silvicultural management (set aside, SA). We chose a time scale of 100 years since it is long enough to 126 cover an entire rotation, and thereby to reveal the long-term impacts of silvicultural practices. The forest 127 simulations create forest structural data at 5 year intervals. 128 In CCF, a selection of the largest trees is removed from the forests approximately every 15 years. 129 Through natural regeneration, the composition of tree species becomes more mixed (Appendix S1; Figure  130 S2). Over time, CCF changes the forest age structure to uneven-aged containing different age classes of 131 trees. No retention trees were left (trees retained permanently through 100 years). The management rules 132 for cuttings are given in Appendix S1: Table S1 (according to the good practice guidance for forestry in 133 consists of species-specific individual-tree diameter increment and survival models, and a stand level 138 model for ingrowth. 139 RFM is currently the recommended and the most common forest management practice in 140 Fennoscandia (Äijälä et al., 2014). In Finland RFM includes several silvicultural actions: soil preparation, 141 seeding or planting trees, one to three thinnings, and the final clear cut, where approximately five 7 retention trees per hectare are retained (according to the good practice guidance for forestry from Äijälä et 143 al., 2014). The management rules for regeneration cutting are given in Appendix S1 : Table S2. The 144 average rotation length of RFM is approximately 80 years in our study region (Appendix S1; Figure S2). 145 RFM creates forest stands, which are often very homogenous in tree species composition as well as in the 146 age structure. The growth models of Hynynen et al. (2002) for even-aged stands were applied for this 147

regime. 148
In SA regime, forests are allowed to grow without human intervention (Appendix S1; Figure S2). 149 In SA, forests are denser, grow slower and there is more tree competition compared with managed forests

Ecosystem services 156
Different ecosystem categories (provisioning, regulating, and cultural) were considered with a set of 157 ecosystem services (Table 1). Timber production is the economically most important provisioning service 158 in boreal forests (Vanhanen et al., 2012). The net present value (NPV, €) of sawlogs and pulpwood for 159 each tree species across 100 years was estimated. The timber NPV consists of three components: the 160 revenues from harvesting (clear-cuts, thinnings, selective loggings; Appendix S1: Table S3), the value of 161 standing timber at the end of simulations, and the value of spare land at the end of simulations (Pukkala, 162 2005). In addition, timber NPV accounted for costs resulting from silvicultural actions related to 163 regeneration and young stand (Appendix S1: Table S4). The stumpage prices of harvest revenue 164 components and the prices for the silvicultural costs were calculated from the historical averages in 165 Finland (Peltola 2014, Appendix S1: Table S3). The stumpage prices included costs from harvesting. The 166 harvesting costs are higher in partial cuttings in CCF than in final fellings in RFM (e.g., Pukkala, 2016, 167 Tahvonen et al., 2010) so the prices from second thinnings (also called intermediate felling) were used for 8 CCF. The interest rate varied between 1% and 5% in discounting the timber revenues and costs during the 169 100-year period. In addition, to study the sensitivity of timber NPV with different costs and prices in CCF 170 and RFM, we calculated NPV without regeneration costs in RFM and using the same prices for CCF and 171 RFM. Timber revenues can be seen as a service for private forest owners, and thus, we also estimated the 172 amount of harvested timber biomass separately for pulpwood and sawlogs, which can be considered as a 173 provisioning service for the whole society since the forest industry is dependent on biomass.  The marketed mushroom yields for spruce dominated stands were calculated using the model of 185 Tahvanainen et al. (2016). In addition, scenic beauty of forests was estimated to describe their 186 recreational values. The scenic beauty index was calculated based on forest age, density and tree species 187 composition according to Pukkala et al. (1988) (Table 1). 188 189

Biodiversity 190
Biodiversity is a multi-faceted phenomenon, which can be measured using indices derived from forest 191 structural data (Table 1). Dead wood is a critical resource in boreal forests (Siitonen, 2001). In boreal 192 Fennoscandia, 20-25% of the forest-dwelling species are dependent on dead-wood habitats, and species 193 dependent on dead wood constitute 60% of the red-listed species (Rassi et al., 2010). Association between 9 dead-wood volume and biodiversity is well established (Gao et al., 2015). The capacity of a stand to 195 maintain populations of dead-wood associated species was estimated by multiplying total dead-wood 196 volume by the diversity of deadwood across tree species, diameter and decay stage categories (Triviño et 197 al., 2016). Thus, a stand with large total deadwood volume distributed evenly across deadwood types will 198 receive high values of deadwood availability. In addition, large diameter living trees are an essential 199 structural feature in boreal forests that has become a limiting factor for biodiversity in production forests 200 sub-utility functions based on expert knowledge and known species habitat that translate characteristics of 211 each stand into a habitat suitability index between 0 (unsuitable habitat) and 1 (the most suitable habitat). 212 For each species, we calculated habitat availability across the entire landscape as a sum of products 213 between stand specific HSI-values and the area of a stand. 214 215

Comparison between silvicultural practices 216
To estimate the performance of silvicultural practices in maintaining ecosystem services and biodiversity 217 (Table 1) where SMFi indicates a scaled multifunctionality value (between 0 and 1) of a single ecosystem service or 238 biodiversity measure i in a stand over 100 years (Table 1) value 1 if either of the SMF in a group was larger than the threshold value t. Finally, multifunctionality 248 score of a stand was calculated as the sum GMFs that had a value above a threshold t as follows: 249 where GMFi indicates a scaled multifunctionality value (between 0 and 1) of a grouped ecosystem 251 (provisioning, regulating and cultural) or biodiversity measure i in a stand where threshold t was 252 continuous between 0 and 1. The maximum multifunctionality score is 4 when a stand is able to provide 253 all services and biodiversity above the threshold level, and minimum is 0 indicating that all indicators 254 remain below the threshold. We calculated average multifunctionality score over time and space in the  Table S1,S2). Carbon storage was moderately higher 264 and sequestration remarkably higher in landscapes consistently managed using CCF in comparison with 265 RFM (Figure 1a; Appendix S2: Figure S1). In terms of the regulating services, CCF outperformed RFM 266 in approximately 75% of the stands (Figure 1b). Bilberry yields and scenic beauty were higher with CCF 267 (Figure 1a) and CCF outperformed RFM in 70-90% of the stands (Figure 1b). In contrast, cowberry and 268 marketed mushroom yields were higher in RFM (Figure 1a) and it outperformed CCF in approximately 269 90% of the stands (Figure 1b). 270 RFM provided more harvested timber (Figure 2a; Appendix S2: Table S1) outperforming CCF in 271 60% of the stands (Figure 2b). There were differences in the shares of sawlogs and pulpwood between 272 CCF and RFM (Appendix S2: Figure S2). In CCF, 79% of the harvested timber was sawlogs and 21% 273 pulpwood, whereas in RFM the share of sawlogs was 65% and the share of pulpwood 35%. 274 The discount rate affected the economic performance of CCF and RFM (Figure 2a; Appendix S2: 275 Table S2). With a 1% discount rate, the NPV of both strategies was the same. The timber NPV was 276 greater in CCF than in RFM when the discount rate was 2% or larger. The optimal share of CCF and 277 RFM was 50% of each when the discount rate was 1% (Figure 2b). The optimal share of CCF increased 278 with the discount rate and was 80% at 5% discount rate. Only when the costs related to regeneration and 279 young stands were not taken into account and the discount rate was 1%, the NPV was greater in RFM 280 than in CCF (Appendix S2, Figure S3c). 281 The optimal combination of CCF and RFM provided higher levels of ecosystem services than 282 either of them separately in all cases except in cowberry yields (Figure 1a, 2a). The benefit of applying 283 both regimes in landscapes was the largest for harvested timber and timber NPV (Figure 2a). 284 SA forests provided higher values than CCF, RFM or their combination particularly for the climate 285 regulating services but also for scenic beauty (Figure 1). Managed forests tended to provide a higher 286 delivery of collectable goods than SA forests. However, SA performed as well as CCF in marketed 287 mushroom production and as well as RFM in bilberry production. 288 289

Biodiversity 290
Consistent application of CCF in landscapes yielded higher values than RFM for five out of eight 291 biodiversity indicators (Figure 3a; Appendix S2: Table S3). For three biodiversity indicators (Lesser 292 spotted woodpecker, Long-tailed tit and number of large diameter trees) CCF outperformed RFM in 293 almost 100% of stands (Figure 3b) and the difference in favour of CCF was remarkably large (Figure 3a). 294 CCF also yielded clearly higher values than RFM in the cases of Hazel grouse and Three-toed 295 woodpecker when CCF outperformed RFM in between 70% to 80% of the stands. In contrast, RFM 296 13 yielded slightly higher dead wood indicator values outperforming CCF in 67% of the stands. Moreover, 297 for the Capercaillie and flying squirrel consistent application of RFM clearly performed better than CCF 298 ( Figure 3a) providing higher habitat suitability index in between 50% to 97% of the stands (Figure 3b). 299 In five cases, a combination of CCF and RFM provided higher scores than either of them separately 300 (Figure 3a). The benefits of combining the two management practices in landscapes were particularly 301 pronounced for the Capercaillie and the Hazel grouse. For the Capercaillie, the combination of CCF and 302 RFM (Figure 3b) yielded 55% larger HSI value than consistent application of CCF and 30% larger than 303 consistent application of RFM (Figure 3a). For the Hazel grouse, the combination (Figure 3b) yielded 304 65% larger HSI value than consistent application of RFM and 20% larger than consistent application of 305 CCF (Figure 3a). 306 From biodiversity perspective, SA was always clearly better than RFM (Figure 3a). For two 307 biodiversity indicators (habitat availability for the Hazel grouse and the Lesser-spotted woodpecker), CCF 308 and SA performed equally well, while in two cases CCF outperformed SA: CCF provided ten times 309 higher number of large trees, and more than two times higher habitat availability for Long-tailed tit than 310 SA (Appendix S2: Table S3). 311 The frequency of unsuitable habitats varied among species (Figure 4)

Multifunctionality 318
At all threshold levels, the average forest multifunctionality was larger when forests were managed with 319 CCF than forests managed with RFM ( Figure 5) indicating a larger capacity of CCF to simultaneously 320 provide services from different categories. An optimal combination of CCF and RFM always produced 321 higher multifunctionality scores than RFM alone, and slightly higher scores than CCF alone when the 322 threshold was larger than 0.4. This indicates that even though CCF in general is better from the 323 multifunctionality point of view, there are some stands where RFM has higher potential to provide 324 multifunctionality. SA provided the lowest multifunctionality below 0.4 threshold levels, which is 325 primarily due to the lack of timber harvesting. When the demand for multifunctionality is high (threshold 326 >0.6) SA forests provided the highest multifunctionality scores. This means that when high levels of 327 ecosystem services and biodiversity are simultaneously desired, leaving forests unmanaged (set aside) is 328 more desirable than managing forests (both CCF and RFM). characteristics; an aspect that should be further studied. 340 In terms of regulating services, while there were no large differences between CCF and RFM in 341 carbon storage, CCF outperformed RFM in carbon sequestration. In RFM, a stand is periodically a source 342 of carbon after the clear-cut. This is mainly because the decomposing forest harvest residues release 343 carbon more than is sequestered in the biomass growth. In CCF, changes in soil carbon stock are smaller 344 after harvests than in RMF because the litter input from harvest residues is smaller. The carbon balance of 345 forests critically depends on the final use of timber biomass after it is removed from the forest ecosystem 346 (Lundmark et al., 2016; Pukkala, 2016b) but we did not take into account the carbon storage in wood 347 products or emissions from the procurement chain. However, as the proportion of sawlogs compared with 348 pulpwood is higher in CCF than in RFM, the carbon retention time would be longer for timber produced 349 in CCF (Pukkala, 2014). Therefore, inclusion of carbon storage in the wood products would not change 350 the main findings. The superior capacity of unmanaged forests to sequestrate carbon is explained by the 351 initial state of the forest stands and their management history. Intensively managed forest landscapes in 352 Finland have a high proportion of young stands holding large potential for carbon sequestration. Although 353 in SA the rate of carbon sequestration decreases with the increasing age of the forest (Pukkala, 2016b), 354 we show that unmanaged forests can have a remarkable role in climate change mitigation for several 355

decades. 356
Our results considering timber NPV support earlier findings (Pukkala et al., 2011;Tahvonen et 357 al., 2010) where discount rates larger than 1% make CCF more profitable than RFM. Higher profitability 358 of CCF is related to the higher price of sawlogs versus pulpwood (CCF provides more sawlogs) as well as 359 the large costs of regeneration and thinnings in RFM. However, in some cases RMF provided also larger 360 economic profits than CCF. This is the case, for stands that are mature, i.e. ready for final harvesting at For four out of the six umbrella and indicator species, CCF provided higher habitat availability 370 than RFM. This is not surprising since many species in this study are dependent on tree cover and 371 deciduous trees, which CCF provides. Moreover, the frequency where the stands were totally unsuitable 372 for the species was often highest in RFM. Thus, one benefit of CCF for mature forest species is the less 373 severe temporal fluctuations in habitat quality. However, for early successional species RFM may 374 actually provide more habitats (Calladine et al., 2015). Nevertheless, habitat availability for species 375 dependent on high tree volume and dead wood availability in forests under both CCF and RFM were far 376 from those in unmanaged forests. Thus, some species habitat availability and dead wood availability more 377 critically depend on the amount of harvested timber than on the silvicultural practice used in harvesting 378 provided only about 25% of that. To further improve the ability of CCF to promote biodiversity we 381 recommend a similar kind of green tree retention that is applied in RFM (leaving permanently behind, at 382 least, 5 trees per ha) to be included in the CCF management regime. Interestingly, CCF provided the 383 greatest number of large trees resulting from the larger resource availability of individual trees (space and 384 light). In contrast, for unmanaged forests, tree growth is lower. This is likely due to the development and 385 transition of very young and planted stands at the beginning of the simulation. For these stands, transition 386 to uneven aged stands will take longer than 100 years and the large trees will likely be harvested during 387 the next (or following) CCF cutting. However, if the simulation time had been longer than 100 years, the 388 number of large diameter trees would have been larger in SA regime and on the other hand smaller in 389 CCF since the transition from even-aged to uneven-aged forestry allows large diameter trees in our CCF 390

simulations. 391
In general, our multifunctionality results indicate that CCF has greater potential than RFM to 392 simultaneously produce multiple benefits in forests, which supports earlier findings (Pukkala 2016a; 393 Sharma et al. 2016). With a moderate demand level for services (threshold value 40%), CCF 394 simultaneously provided services from all categories when RFM provided services only from three 395 categories. Moreover, the relative multi-functionality performance of SA increased with the demanded 396 level of services. Therefore, when discussing the delivery of ecosystem services and maintaining 397 biodiversity, their demanded levels should be taken into account. If society demands high 398 multifunctionality in forest landscapes, more resources must be allocated to unmanaged set aside forests 399 since their role in delivering high levels of biodiversity and regulating services is often indispensable. 400 In the simulation process, all management options were based on decision rules instead of 401 optimizing the specific management at stand level. Both management alternatives that we used could be 402 changed substantially by altering the specific decision rules (i.e. delaying final felling, restricting the 403 frequency of harvests, or requiring green tree retention following a clear felling) to increase the delivery 404 of multiple benefits in forest stands (e.g., Liski  Third, in practice management decisions depend on the choices of the forest owner, whose preferences 413 may more likely be based on the personal economic situation rather than on specific stand characteristics 414 (Brazee, 2003). In such cases, simple decision rules may better provide guidance to decisions than 415 knowledge about stand level optimal management. 416 Our comparison among CCF, RFM and unmanaged forests has some limitations. Even though the 417 planning horizon was long (100 years), even set-aside forests do not provide a natural-state benchmark for 418 managed forests. For example, the amount of dead wood in the natural forest state is approximately 60-90 419 m 3 ha -1 (Siitonen, 2001) but in our data the amount of dead wood under set-aside was significantly lower. 420 Moreover, our simulation did not include natural disturbances, such as storms and diseases, which may 421 substantially change the forest characteristics, e.g. dead wood volumes (Kuuluvainen, 2002). Therefore, 422 our simulations probably underestimate the delivery of some ecosystem services and biodiversity values 423 in all management regimes, and on the other hand, overestimate timber production in both CCF and RFM.

CONCLUSION 442
Our results indicate that continuous cover forestry has greater potential than rotation forest management 443 to maintain multifunctional forests. However, continuous cover forestry was not the best for all ecosystem 444 services or biodiversity indicators. Furthermore, the combination of different forest management practices 445 provided higher levels of services and indicators than single practices applied consistently over the 446 landscape. Moreover, we show that commercially managed forests, if set aside, may provide important 447 resources for biodiversity and regulating services. Thus, it is not reasonable to rely on one single practice 448 and careful landscape planning is needed. Continuous cover forestry does not itself guarantee the 449 maintenance of all ecosystem services and biodiversity in commercial forests but it can be an important 450 part of a successful progression towards more sustainable forestry. 451 452