Manuscript Click here to download Manuscript Manuscript_JBARBA.docx Click here to view linked References 1 Strong resilience of soil respiration components to drought-induced die-off 2 resulting in forest secondary succession 3 4 Running head: Soil respiration partitioning fluxes 5 6 Barba J. (Josep Barba)1,2*, Curiel Yuste J. (Jorge Curiel Yuste)3, Poyatos R. (Rafael 7 Poyatos)1, Janssens IA. (Ivan A. Janssens)4 and Lloret F. (Francisco Lloret)1,2 1 8 CREAF, Cerdanyola del Vallès, 08193 (Barcelona), Catalonia, Spain; 2 9 UnivAutònoma Barcelona (UAB), Cerdanyola del Vallès, 08193 (Barcelona), 10 Catalonia, Spain; 3 11 Museo Nacional de Ciencias Naturales (MNCN), CSIC, Madrid, 28006, Spain. 4 12 Biology Department, Universiteit Antwerpen, Wilrijk, Antwerpen, B-2610, Belgium. 13 * Corresponding author: Josep Barba; e-mail: j.barba@creaf.uab.cat; Phone: 34 93 i p t 14 5813345; Fax: 34 93 5811312 u s cr 15 a n dm 16 Keywords: Soil respiration, Heterotrophic respiration, Autotrophic respiration, 17 pt e Partitioning fluxes, Resilience, Mediterranean forest, Pinus sylvestris, Quercus ilex. c ce 18 a or’s 19 Globally increasing drought-induced forest die-off and its associated vegetation shifts 20 may have direct impacts on soil respiration. Here, we found that soil respiration and its 21 u t h autotrophic and heterotrophic components remained unaffected 3-11 years following 22 23 A drought-induced Scots pine die-off. Despite this post-disturbance functional resilience, the replacement by holm oak was associated with a strong reduction in the 24 heterotrophic respiration component, producing an important drop in total soil 25 respiration. 26 27 Author contributions: JB, JCY and FLL conceived and designed the experiment; JB, 28 JCY and RP performed the experiment; JB, JCY, IJ and RP analysed the data; JB wrote 29 the paper and all authors edited the manuscript. 30 Post-print of: Barba, J. et al. “Strong resilience of soil respiration components to drought-induced die- off resulting in forest secondary succession” in Oecologia (Ed. Springer), published online 15 February 2016. The final versión is available at DOI 10.1007/s00442-016-3567-8 31 Abstract 32 How forests cope with drought-induced perturbations and how the dependence of soil 33 respiration on environmental and biological drivers is affected in a warming and drying 34 context are becoming key questions. 35 The aims were to determine whether drought-induced die-off and forest succession were 36 reflected in soil respiration and its components and to determine the influence of climate 37 on the soil respiration components. 38 We used the mesh exclusion method to study seasonal variations soil respiration (RS) 39 and its components: heterotrophic (RH) and autotrophic (RA) (further split in fine root 40 [RR] and mycorrhizal respiration [RM]) in a mixed Mediterranean forest where Scots 41 pine (Pinus sylvestris L.) is undergoing a drought-induced die-off and is being replaced 42 by holm oak (Quercus ilex L.). 43 Drought-induced pines die-off was not reflected in RS nor in its components which i p t 44 s cr denotes a high functional resilience of the plant-and-soil system to pines die-off. u 45 a n However, the succession from Scots pines to holm oaks resulted in a reduction of RH dm 46 and thus in an important decrease of total respiration (RS was 36% lower in holm oaks 47 pt e than in non-defoliated pines). Furthermore, RS and all its components were strongly c ce 48 regulated by SWC-and-temperature interaction. a or’s 49 Since Scots pine die-off and Quercus species colonization seems to be widely occurring 50 on the driest limit of the Scots pine distribution, the functional resilience of the soil 51 u t h system over die-off and the decrease of RS from Scots pine to holm oak could have 52 53 A direct consequences on the carbon balance of these ecosystems. 54 1. Introduction 55 Drought-induced forest die-off episodes are being increasingly reported globally (Allen 56 et al. 2010), raising concerns on their possible association with recent global warming, 57 at least in some parts of the globe (Carnicer et al. 2011). Increased tree defoliation and 58 mortality rates have been related to chronic and episodic drought in the Mediterranean 59 region (Peñuelas et al. 2001; Carnicer et al. 2011), where the projected increase in the 60 frequency and intensity of droughts (IPCC, 2013) may enhance the risk of extensive 61 forest die-off. Widespread drought-induced tree mortality can have dramatic effects on 62 forest carbon cycling (Reichstein et al. 2013) which may differ from those reported for 63 other disturbances such as fire or pest outbreaks (Anderegg et al. 2013; Borkhuu et al. 64 2015). Forest ecosystems store over two thirds of the carbon in their soils (Dixon et al. 65 1994), but the impact of drought-induced tree mortality on soil carbon fluxes and their 66 post-disturbance dynamics remains poorly understood (Allen et al. 2015). 67 The trajectory of ecosystem C cycling in stands where the dominant overstorey species 68 has been severely affected by a die-off event is complex in space and time (Edburg et al. 69 2012; Reed et al. 2014; Borkhuu et al. 2015). At subdecadal timescales, the trend in 70 forest C uptake largely depends on disturbance intensity and the degree of overstorey 71 canopy loss, the spatial pattern of tree dieback (i.e. clustered, diffuse), the compensatory 72 responses by surviving dominant trees, and the response of understorey vegetation to 73 competition release (Amiro et al. 2010; Brown et al. 2010; Gough et al. 2013). With 74 regard to forest soil C losses, die-off episodes immediately curtail root and mycorrhizal 75 respiration and reduce exudate supply from roots to soil, but they also stimulate 76 decomposition of litter, roots and dead wood (Nave et al. 2011). Moreover, die-off 77 episodes may increase soil moisture (Redding et al. 2008), which could enhance soil i p t 78 s cr organic matter (SOM) decomposition rates in water limited ecosystems. Nevertheless, u 79 a n our knowledge on post-dieback ecosystem (and soil) C fluxes is still incomplete, dm 80 because (1) it largely arises from studies where pests or wildfires, not drought, were the 81 pt e main disturbance drivers and (2) it lacks a detailed understanding of the post-dieback c ce 82 trajectories of soil respiration components (heterotrophic respiration, RH; autotrophic a or’s 83 respiration RA; fine roots respiration, RR; mycorrhizal respiration, RM) and their 84 responses to environmental drivers. 85 u t h Drought-induced tree mortality episodes often result in a complex spatial pattern of 86 87 A standing dead trees and partially defoliated individuals, together with apparently unaffected trees. These episodes may cause vegetation shifts within years or decades if 88 adult mortality and recruitment of the dominant and most affected species are not 89 balanced through time (Lloret et al. 2012). Given that the substituting species will likely 90 be more drought-resistant, such species replacement could have persistent effects on 91 both stand C uptake and release processes. The spatial distribution of soil CO2 effluxes 92 under non-limiting soil moisture conditions is sensitive to local changes in vegetation 93 composition following tree mortality (Barba et al. 2013), but we do not know whether 94 the response of soil CO2 effluxes to the main abiotic drivers (i.e. soil temperature and 95 moisture) differs along the stages of an ongoing vegetation shift. In Mediterranean 96 forests, the seasonal correlation between plant productivity and moisture-and- 97 temperature course hinders to determine the effect of these variables on RS (Tedeschi et 98 al. 2006) because of the inherent seasonal variability of plant photosynthetic activity 99 and belowground C allocation (Reichstein et al. 2002; Keenan et al. 2009). 100 The variability of soil CO2 efflux associated to the environmental drivers may also be 101 mediated by the differential metabolic response of microbes, roots and rhizosphere to 102 moisture and temperature changes (Uren 2000; Kuzyakov 2006; Moyano et al. 2010). 103 Changes in microbial community composition have been observed following drought- 104 induced forest die-off and succession (Curiel Yuste et al. 2012), but how these shifts in 105 microbial community composition modify the contribution of autotrophic and 106 heterotrophic respiration to total soil respiration remains poorly known. 107 Drought-induced dieback episodes, characterised by increased crown defoliation and 108 mortality rates, have been reported in several Scots pine (Pinus sylvestris L.) 109 populations in the northeast of the Iberian Peninsula (Martínez-Vilalta and Piñol 2002; 110 i p t Galiano et al. 2010; Hereş et al. 2012). In particular, extreme drought events together 111 s cr with a lack of forest management have caused several Scots pine die-off episodes at the u 112 a n Prades Mountains in the 1990s and 2000s (Martínez-Vilalta and Piñol 2002; Hereş et al. dm 113 2012) and the associated holm oak (Quercus ilex L.) colonisation (Vilà-Cabrera et al. 114 pt e 2013). These processes have resulted in important changes not only at above-ground c ce 115 component of the forest (Aguadé et al. 2015) but also at the below-ground part, altering a or’s 116 soil microbial diversity and structure (Curiel Yuste et al. 2012), litter decomposition and 117 nutrients cycling (Barba et al. 2015) and spatial variability of soil respiration (Barba et 118 u t h al. 2013). In this study, we aim to quantify the effects of this Scots pine drought-related 119 120 A die-off and the consequent holm oak colonisation on soil respiration and its components, at seasonal and annual timescales. During one year, we measured the 121 seasonal variation of soil respiration and its heterotrophic and autotrophic (roots and 122 mycorrhiza) components, associated with the different stages of this drought-induced 123 die-off and species-replacement process: non-defoliated pines [NDP], partially 124 defoliated pines [DFP], dead pines [DP] and holm oaks [HO]. 125 We hypothesized that: 1) Heterotrophic respiration would show a gradual increase 126 across this die-off gradient (from NPD to DFP and DP) and autotrophic respiration 127 would decrease because of reduced above-and-belowground plant activity in DFP and 128 DP. Given that holm oak is a more drought-tolerant species compared to Scots pine, for 129 this drought-exposed site we expected (2) higher soil respiration in holm oak and lower 130 sensitivity to decreasing soil moisture during summer drought compared to Scots pine. 131 Finally, we hypothesized that (3a) the spatial and temporal variation in heterotrophic 132 soil respiration would be mainly dependent on environmental variables (soil 133 temperature and moisture), whereas these variables would explain little variability for 134 autotrophic respiration. Since vegetation has the capacity to modify soil environmental 135 conditions, we also hypothesized that (3b) the vegetation die-off and the ongoing 136 species succession would largely influence soil CO2 effluxes via modifications in 137 environmental conditions such as SWC. 138 139 2. Materials and Methods 140 2.1. Study site 141 The study was conducted in a mixed forest in the Titllar Valley (Prades Mountains, NE 142 Iberian Peninsula). The climate is Mediterranean, with a mean annual temperature of 143 11.2°C and precipitation of 664 mm (Ninyerola et al. 2007a; Ninyerola et al. 2007b). 144 i p The experimental area was located on the northwest face of the valley at an elevation t 145 s cr between 1,010 and 1,030 m a.s.l. with a steep slope (33°). Soils are xerochrepts with u 146 a n clay loam texture (49% sand, 32% silt, 19% clay) and relatively high gravel content of dm 147 46% (Barba et al. 2013; Sus et al. 2014). The substrate consists of fractured 148 pt e metamorphic schist that outcrops on a large part of the study area. Mineral soil is c ce 149 slightly acidic (pH is 6.2) and contain 5.9% of C, 0.27% of N and 0.044% of P (Curiel a or’s 150 Yuste, unpublished results). No differences are found in C, N or P availability or in soil 151 carbon pools (both quality and quantity) between soils associated with the different type 152 u t h of trees (Curiel Yuste et al. 2012). For more information related to the study area, see 153 154 A Hereter and Sánchez, 1999; Barba et al., 2013. This mixed forest is mainly dominated by Scots pine (Pinus sylvestris L.) in the 155 overstorey and holm oak (Quercus ilex L.) in the understorey, with total stem density of 156 2235 stems ha-1 (Poyatos et al. 2013). Severe drought events since the 1990s (Martínez- 157 Vilalta and Piñol 2002) and particularly in 2001-2003 and 2005-2008 (Hereş et al. 158 2012) have affected the Scots pine populations, inducing a mean crown defoliation of 159 52% and standing mortality of 12% (Vilà-Cabrera et al. 2013). This situation, coupled 160 to the low regeneration of pines (Vilà-Cabrera et al. 2013) is currently driving the 161 replacement of pines by oaks as the dominant overstorey species. 162 163 2.2. Experimental design 164 2.2.1 Experiment scheme 165 Soil respiration fluxes were measured close to 12 trees (less than two meters from the 166 tree stem) belonging to the different stages along the vegetation shift following the 167 Scots pine die-off process (from now, type of tree): three non-defoliated Scots pines 168 [NDP], three living defoliated Scots pines [DFP], three dead Scots pines [DP] and three 169 Holm oaks [HO]. The maximum distance between measuring points was ca. 200 m. 170 Dead Scots pines were devoid of needles and small branchlets, and only the main bole 171 and primary branches were still standing. Hereş et al. 2012 found that 86% of the 172 standing mortality resulted from the drought events in 2001-2003 and 2005-2008. 173 Therefore we estimate that these trees have been dead for 3-11 years. The degree of pine 174 defoliation was visually estimated as the percentage of green needles relative to a non- 175 defoliated canopy of a similar sized tree from the same population (Galiano et al. 2010). 176 Defoliated pines had less than 50% of green leaves. DBH for each type of tree was 37±7 177 cm in NDP, 59±9 cm in DFP, 58±7 cm in DP and 14±8 cm in HO (mean ± sd). The i p t 178 s cr Hegyi competition index was significantly higher for HOs with respect to pines, u 179 a n whereas no significant differences were found between NDP, DFP and DP (Curiel dm 180 Yuste et al. 2012). 181 pt e The root excision method (Subke et al. 2006) was used for studying total, autotrophic c ce 182 and heterotrophic soil respiration, using the protocol proposed by Heinemeyer et al., a or’s 183 (2007). Three different PVC collars (treatment collars) of 63 cm in diameter were 184 installed within 3 m of each tree (Figure 1). The first type of collar (A) was 5 cm in 185 u t h height and was inserted only 2 cm into the ground and fixed with three metal sticks. 186 187 A Thus, these A-type collars did not interfere with fine roots, mycorrhizal or soil microbial dynamics and growth. The second and the third types of collar (B and C 188 respectively) were 50 cm in height and were inserted up to a depth of 45 cm into the 189 soil. Collar B had four rectangular windows (17 x 5 cm) at 10 cm from the top, covering 190 33% of the total collar perimeter. Windows were covered with nylon fabric of 41 µm 191 mesh size, allowing ingrowth of fungal hyphae but not of roots (Ek 1997). The deeper 192 B-type collars prevented the ingrowth of fine roots (which are concentrated mainly in 193 the upper centimetres of the soil (Jackson et al. 1997). C-type collars were the same size 194 and were installed at the same depth as B-type collars, but they did not have windows, 195 so it prevented the ingrowth of both, fine roots and mycorrhizal hyphae. Hence, we 196 assumed that in C-type collars, only the non-rhizospheric microbial community 197 remained active. 198 Due to the great stoniness of the soil, two modifications were made to the Heinemeyer 199 et al., (2007) experimental design. First, treatment collars were bigger (20 cm in 200 diameter in the original protocol), and second, we dug a hole in the soil previous to the 201 installation of the deeper collars. After placing the collar inside the hole, we filled the 202 space inside the collar with the previously removed soil material, preserving the original 203 soil-horizon order. Two smaller PVC collars (sampling collars; 10 cm in diameter and 4 204 cm of height) were installed as replicate collars within each treatment collar. Sampling 205 collars were inserted 1 cm into the ground and fixed with polyurethane foam. These 206 collars delimited the reference surface of our soil respiration measurements. Collar 207 installation was made 14 months before the start of measurements in order to minimize 208 the effects of the soil disturbance on soil respiration measures. 209 210 2.2.2 Measurements i p t 211 s cr Soil respiration rates were measured with a close-path infra-red gas analyser (IRGA) u 212 a n coupled to a closed dynamic chamber (EGM-4 and SRC-1, PP-Systems, HITCHING, dm 213 UK) in the sampling collars (the soil chamber fitted well with the small collars because 214 pt e it had the same diameter). Soil respiration was measured every two weeks during one c ce 215 year, from June 2012 to June 2013 (25 campaigns). For each campaign, five rounds of a or’s 216 measurements were made at all 72 small collars during 24 hours, in order to capture the 217 soil respiration variability associated to daily cycles. One round was started two hours 218 u t h before sunrise, another was performed after sunset and the others were equally 219 220 A distributed during the daytime. Since soil respiration was manually measured, the effort for integrate the spatial variability of soil respiration in this extremely rocky location 221 and also the daily soil respiration course did not allow us to increase the number of 222 measured trees. 223 Soil temperature at 10 cm was measured once per big collar and per round of 224 measurements using a thermometer (OMEGA, HH806AU, Stamford, USA). Soil water 225 content (SWC) in the top 15cm was measured each campaign at each tree by time 226 domain reflectometry (TDR) (Tektronix 1502C, Beaverton, Oregon, USA). One 15 cm 227 long TDR probe was permanently installed vertically in the upper soil close to each tree 228 throughout the experiment. In order to correct the SWC measurements for the stoniness, 229 gravimetrical SWC measured in soil samples close to the TDR probes were regressed 230 against TDR measurements (for more information, see Poyatos et al., 2013). 231 One litterfall trap (555 x 355 mm) was installed within 2 m of each tree and litterfall 232 was collected during every campaign, dried during 24 h at 70°C and weighed. 233 Air temperature and relative humidity were continuously measured in a meteorological 234 tower installed less than 100 m from the farthest tree. Continuous SWC was also 235 recorded in the upper 30 cm of soil using four frequency domain reflectometers (CS616, 236 Campbell Scientific INC) randomly distributed among the trees. A data acquisition 237 system (CR1000 datalogger and AM16/32 multiplexers, Campbell Scientific Inc., 238 Logan, UT, USA) was used to store 15-min means of soil moisture and meteorological 239 variables sampled every 30 s. 240 241 2.3. Data analysis 242 2.3.1. Soil respiration calculation 243 The five soil respiration measurements recorded for each small collar were time- i p t 244 s cr averaged in order to obtain mean daily soil respiration. Then, both replicates were u 245 a averaged at the big collar level (A, B and C; Figure 1). n dm 246 Total soil respiration (RS), heterotrophic soil respiration (RH), autotrophic soil 247 pt e respiration (RA), fine roots respiration (RR) and micorrhyzal respiration (RM) were c ce 248 calculated following the protocol proposed by Heinemeyer et al., (2007). RS and RH a or’s 249 were estimated directly as the soil respiration rates measured in collars A and C, 250 respectively. RA was calculated by subtracting C from A; RR by subtracting B from A 251 u t h and RM by subtracting C from B. 252 253 A 2.3.2. Soil respiration drivers. 254 Mixed-effects models were used to analyse the relationships between soil respiration 255 and type of tree, soil temperature, SWC, litterfall and season with campaign data. The 256 limits of seasons were adjusted from environmental variables (i.e. Summer ended with 257 the first rainfall events in Fall, which changed drastically SWC and temperature). Linear 258 and exponential relationships between soil respiration and temperature were tested in 259 the models, as well as linear and quadratic relationships between soil respiration and 260 SWC. In all cases, the linear relationships performed better than the nonlinear 261 transformations according to the AICc (corrected Akaike information criterion) statistic 262 (data not shown). 263 Different models were fitted for each soil respiration component (RS, RH, RA, RM and 264 RR). As all variables were measured near the same trees throughout the campaigns, tree 265 identity was included as a random factor in all models. Models with all combinations of 266 predictor variables and their second-order interactions were performed, and the best 267 model in terms of AICc was selected. To determine the variability explained by each 268 mixed model, we calculated the coefficient of determination using the log likelihood of 269 both, the studied model and the null model (which did not include any predictor 270 variables). In the root exclusion method, the PVC collars could interfere with soil 271 temperature and moisture (Kuzyakov 2006). Unfortunately, due to instrumental 272 limitations, SWC was only measured at tree level (and not at collar level) at each 273 campaign. Although, SWC was gravimetrically measured at samples inside the 72 small 274 collars at the end of the experiment (July 2013) and one-way ANOVA was performed 275 to test for differences in SWC between the different treatment collars. 276 277 2.3.3. Annual SR i p t 278 s cr We used the models fitted in 2.3.2 to estimate daily values of RS, RH, RA, RM and RR for u 279 a n a whole year (from mid-June 2012 to mid-June 2013). Apart from fixed predictors dm 280 (season, type of tree), daily-aggregated values of soil temperature, SWC and litterfall 281 pt e were needed as inputs for the model. For each tree and collar type, linear regressions c ce 282 were fitted between daily-averaged soil temperature measured during the campaigns and a or’s 283 simultaneous air temperature measured at the meteorological tower (R2=0.93±0.02, 284 across-trees mean±sd). Likewise, daily SWC for each tree was estimated from linear 285 u t h regressions against mean SWC measured with the four frequency domain reflectometers 286 287 A near the meteorological tower (R2=0.66±0.27, across-trees mean±sd). Daily litterfall across-trees was linearly interpolated from biweekly-measured litterfall. Modelled 288 values of daily RS, RH, RA, RM and RR were then aggregated to obtain annual values. 289 Mixed-effects models, including tree identity as a random factor, were then used to 290 analyse the differences in annual soil respiration and its components between types of 291 tree. Given the high spatial variability of soil respiration, especially in this ecologically 292 complex site (Barba et al., 2013), and the limited number of replicates (3 trees per type), 293 we considered marginally significant differences among means when 0.05 < p < 0.1. 294 Additionally, Friedman test and its post hoc analysis were applied to the daily-averaged 295 data from the 25 campaigns, to test for possible differences in soil respiration and its 296 components between types of tree. The Friedman test is a non-parametric repeated 297 measure ANOVA. Its procedure involves ranking soil respiration from the different tree 298 types and then considering the values of ranks by campaigns. 299 To test whether the relative contribution of RH to RS increases along the die-off process, 300 mixed-effects models with tree type (for testing annual differences) or with the 301 interaction between tree type and season (for testing differences over seasons) were 302 used, including tree identity as a random factor. Heterotrophic relative contribution data 303 was log transformed to achieve normality. 304 All the analyses were carried out using R 3.0.3. (R Foundation for Statistical 305 Computing, Vienna, Austria). The mixed-effects models were performed using the R 306 package nlme (Pinheiro et al. 2009) and the step-wise model selection was performed 307 with MuMIn package (Bartón 2014). 308 309 3. Results 310 3.1. Seasonal course of environmental variables and soil respiration components. 311 i p Over the study period, climate was typical of a low elevation Mediterranean mountain, t 312 s cr with mean air temperature of 11.2°C, annual precipitation of 703 mm and relatively dry u 313 a n summer (93 mm from June to September and mean SWC below 15% from mid-July to dm 314 the end of September) (Figure 2b and 2c). The seasonal pattern of litterfall was not as 315 pt e clear as those of temperature and SWC, but it seemed to peak at the end of fall (Figure c ce 316 2d). a or’s 317 No significant differences were found for soil temperature, SWC and litterfall among 318 types of trees (Table 1). While type of collar did not influence soil temperature during 319 u t h the experiment (p=0.87, repeated measures ANOVA), SWC measured gravimetrically 320 321 A at the end of experiment (July 2013) was higher for deeper collars (B and C) than for surface ones (A) (p=0.03, one-way ANOVA), 3.4% on average. 322 Total soil respiration (RS) flux peaked during late spring and early summer (up to 6 323 µmol m-2 s-1; Fig 3, top panel) and then decreased over the course of the summer to 33 324 % of peak values. During fall and winter RS was quite low (between 0 to 2 µmol C m-2 325 s-1). RH showed a similar annual pattern as RS with values ranging between 1 and 4 326 µmol C m-2 s-1 (Figure 3, RH panel). No seasonal pattern was observed either for RA, or 327 for its components (RM and RR) (Figure 3, bottom panels). Soil CO2 efflux measured in 328 HO was the lowest in 88% of the campaigns for RS, 76% for RH, 68% for RA, 56% for 329 RR, but only in 24% of the campaigns for RM (Figure 3). Indeed, the non-parametric 330 Friedman test applied to the campaign data indicated that respiration rates under HO 331 were lower than under the three types of pines for RS, RH, RA and RR (p<0.001 for RS 332 and RH, p=0.001 for RA and p=0.002 for RR), but not for RM. 333 The relative contribution of RH to RS did not show differences among tree types, either 334 at annual (p=0.968) or at seasonal scales (p=0.325) (GLM models with tree identity as a 335 random factor). Nonetheless, the relative contribution of the different soil respiration 336 components considering all tree types together varied during the year. The contribution 337 of RH to RS was highest during late winter, spring and summer, and decreased during 338 fall and early winter. In contrast, the contribution of RR to RS increased during fall and 339 early winter. The contribution of RM to RS did not show any seasonal pattern, remaining 340 low throughout the year (Figure 4). 341 342 3.2. Soil respiration drivers. 343 A mixed-effects model including the interactions between season-SWC-temperature and 344 temperature-type of tree as predictor variables (Table 2, RS) explained 51% of the 345 seasonal variability in RS. The higher the SWC, the larger the positive effect of i p t 346 s cr temperature was on RS, especially during spring and summer. However, at SWC values u 347 a n below 10%, the temperature effect on RS was negligible in summer (Figure 5; i, j, k and dm 348 l). Additionally, SWC had higher effect on RS under HO than under pines. Similarly, 349 pt e tree type interacted with SWC to determine RH, while the rest of the RH predictors were c ce 350 almost the same as those in the RS model (Table 2, RH and Figure 6). RH model a or’s 351 explained 56% of the total variability in RH. The interaction between SWC and 352 temperature had the same positive effect than in the RS model (higher effect of 353 u t h temperature at high SWC values). Temperature had the lowest effect on RH during fall 354 355 A and winter, a higher effect during summer and the highest effect during spring (Table 2, RH). 356 Models of the autotrophic components of RS explained much less variability than the 357 ones fitted for RS and RH. The selected RA model was able to explain only 15% of the 358 autotrophic respiration variability, and included the positive effect of litterfall and the 359 positive interaction between SWC and temperature (Table 2, RA). Likewise, the selected 360 RR model was able to explain only 13% of the variability in fine root respiration and 361 contained the interactions season-SWC and season-temperature (Table 2, RR). Finally, 362 the selected RM model was able to explain 24% of the variability in mycorrhizal 363 respiration and contained only the interaction between SWC-temperature (Table 2, RM). 364 Type of tree had no effect on RA, nor on its fractions (RR and RM). 365 366 3.3. Annual soil respiration and its components. 367 For the overall set of sampled trees, modelled annual RS (mean±SD) from July 2012 to 368 July 2013 was 2.6±0.6 µmol m-2s-1, RH was 1.7±0.3 µmol m-2s-1, representing 65% of 369 RS, while RA was 1.0±0.5 µmol m-2s-1 (36% of RS). RR and RM, as components of RA, 370 were 0.6±0.6 µmol m-2s-1 (23% of RS) and 0.3±0.7 µmol m-2s-1 (13% of RS), 371 respectively. 372 The mixed-effects models showed that annual RS under holm oak trees (HO) was 373 marginally lower than under non-defoliated pines (NDP) (p-value: 0.074) (64% on 374 average), whereas defoliated pines (DFP) and dead pines (DP) did not show differences 375 with NDP or HO. Likewise, annual RH under HO was significantly lower than under 376 NDP (p-value: 0.030; 36% lower), and marginally significantly lower than under DFP 377 (p-value: 0.089; 23% lower) and DP (p-value: 0.054; 33% lower). Mixed-effects models 378 for annual RA, RR and RM did not show differences among types of trees (Figure 7). 379 i We used the model in section 3.2 to assess the effect of the higher soil moisture in p t 380 s cr deeper collars (B and C) due the collar effect (3.4% higher on average at July 2013), u 381 a n because this increased SWC could have stimulated RH, overestimating the heterotrophic dm 382 contribution to RS and consequently, underestimating RA. Since SWC was not measured 383 pt e at the different collar types during the campaigns, it was not possible to asses this effect c ce 384 over the year. However, if we assumed that the difference observed in July persisted a or’s 385 throughout the year, annual RH and its contribution on RS would only have been 0.92% 386 and 0.60% higher than the reported values, respectively (models estimations with an 387 u t h increment of 3.4% in SWC for all campaigns). Therefore, we considered the collar 388 389 A effect on the estimation of respiration fluxes to be negligible. 390 4. Discussion 391 4.1. Annual soil respiration and its fractions 392 Mean annual RS at the study site was 2.6 µmols m-2 s-1, similar to values reported for a 393 parallel valley in the same nature reserve (2.3 µmols m-2 s-1, Asensio et al., 2007). RH 394 was the most important fraction of RS, representing about 65%. This relative importance 395 agrees well with estimates obtained in other forest ecosystems (Rey et al. 2002; 396 Tedeschi et al. 2006; Heinemeyer et al. 2007; Subke et al. 2011), confirming the 397 important role of microbial soil respiration in RS. The RR and RM contribution to RS 398 (23% and 13%, respectively) were also similar to values reported in other partitioning 399 studies (Malhi et al. 1999; Rey et al. 2002; Subke et al. 2006; Heinemeyer et al. 2007; 400 Ruehr and Buchmann 2010). t 401 402 4.2. Soil respiration and its components following Scots pine die-off cr i p 403 n u s Scots pine die-off, from non-defoliated pines to defoliated pines and to dead pines, was a dm 404 not associated with major changes in soil respiration (both heterotrophic and autotrophic e 405 activity, see Figure 7), contrary to what we had hypothesized (H1). Therefore, also the pt ce 406 relative contribution of heterotrophic respiration did not increase from NDP to DFP and 407 a c DP. Defoliation did not affect RS or its components with respect to non-defoliated pines or’s 408 (NDP). Moreover, 3-11 years after tree death, we observed that RS (and all its t h 409 components) associated with dead Scots pines has either completely recovered or 410 411 A u remained unchanged (Figure 7). Even the autotrophic components (RR and RM) of RS, directly dependent on the substrate inputs from plant photosynthetic activity (e.g. 412 Högberg et al., 2001), were not affected along this die-off gradient. These effects are 413 surprising, given that soil autotrophic respiration had been reported to be extremely 414 sensitive to decreases in photosynthetic activity associated with defoliation and die-off 415 (Levy-Varon et al. 2012; Moore et al. 2013; Levy-Varon et al. 2014; Borkhuu et al. 416 2015). Indeed, decreases in plant productivity have been associated with lower fine root 417 biomass and hence lower root metabolic activity, lower belowground substrate 418 allocation and lower root exudation, all contribution to lower RA and its fractions 419 (Högberg and Read 2006). 420 We did not find higher heterotrophic respiration (RH) under DP than under NDP, which 421 is consistent with the fact that tree mortality did not apparently result in higher soil 422 moisture, soil temperature or litterfall under dead pines, factors which are known to 423 stimulate RH. Although the relative contribution of RH to RS (and thus, also RA 424 contribution) showed a clear seasonal pattern, this pattern was preserved across the 425 entire pine die-off gradient (NDP, DFP and DP). 426 The stability in RS and in its fractions along the pine die-off gradient denotes a high 427 degree of resilience of soil processes with respect to aboveground perturbations. Some 428 studies have also shown that RS remained stable after a perturbation (Binkley et al. 429 2006; Levy-Varon et al. 2014), suggesting that this RS resilience could be produced by 430 mobilization of reserve carbohydrates (Levy-Varon et al. 2012), by a higher growth 431 rates of non-disturbed trees (Levy-Varon et al. 2014) speculating that an increment of 432 RH due to an increment in organic matter availability could be compensating a decrease 433 in RA (Borkhuu et al. 2015). While this has been partially shown in some studies (Levy- 434 Varon et al. 2012; Moore et al. 2013; Levy-Varon et al. 2014; Borkhuu et al. 2015), 435 i here we present, to the best of our knowledge, the very first evidence that all the p t 436 s cr components of RS remain apparently unaffected after 10 yr of drought-related mortality u 437 processes. a n dm 438 Our results might be partially explained by the disturbance recovery findings in Nave et 439 pt e al. (2011), which suggest that a short perturbation of forest C cycling due to partial c ce 440 canopy disturbance could be rapidly recovered (within a few years), thereby stabilizing a or’s 441 the C cycle. They hypothesize that shortly after the perturbation, the expected decrease 442 in ecosystem production and soil respiration (due to decrease fine root biomass and non- 443 u t h structural carbohydrate concentrations in roots) could be compensated by higher growth 444 445 A rates from the remaining healthy trees due to reduced competition for limiting resources. At the study site, after drought-induced pine mortality in the 1990’s higher 446 growth rates of the remaining healthy trees were observed (Martínez-Vilalta and Piñol 447 2002), consistent with this hypothesis. 448 The mechanism underlying the observed resilience of RS following Scots pine die-off is 449 likely more related to belowground colonization by HO than to the remaining, 450 unaffected, pines. Results obtained from previous studies in the same site support this 451 idea. Firstly, spatial variability of RS close to dead pines has been mainly associated 452 with the spatial distribution of HO basal area (Barba et al. 2013) suggesting a functional 453 colonization by HO rhizosphere. Moreover, it has been observed that rates of R R 454 measured in living roots under DP were similar to those measured for HO (Pereira- 455 Blanco 2014), which suggests that fine roots colonization of the disturbed gap is taking 456 place by HO. Finally, Curiel Yuste et al. 2012 found that there was a similarity in the 457 most abundant bacterial taxon (i.e. Actynimycetes, Rhizobiales, Xantomonadales) between 458 rhizosphere from DP and HO, indicating that HO colonization is also occurring at the 459 microbial level. 460 461 4.3. Drought-induced substitution of Scots pine by holm oak causes a decrease in soil 462 respiration 463 Changes in vegetation could produce changes in the whole plant-to-soil system, such as 464 root biomass and distribution, nutrients and water balances, net primary production, 465 carbohydrate allocation patterns, litter quantity and quality, decomposer community or 466 microbial diversity (Jackson et al. 1997; Binkley and Giardina 1998; Palacio et al. 2007; 467 Strickland et al. 2009; Curiel Yuste et al. 2012), which in turn, could modify RS and its 468 heterotrophic and autotrophic fractions (Uren 2000; Janssens et al. 2001; Kuzyakov 469 i p t 2006; Cornwell et al. 2008; Vivanco and Austin 2008). Despite the strong resilience of 470 s cr RS and its components along the Scots pine die-off (NDP, DFP and DP) (see section u 471 a n 4.2), changes in RS associated with the succession from Scots pine to holm oak were dm 472 observed. Annual RS was 36% lower in HO compared to NDP, contrary to our 473 pt e hypothesis H2. Although the analysis of annual values did not show differences in RA, c ce 474 RR or RM between NDP and HO (Figure 7), seasonal data analysed with the non- a or’s 475 parametric Friedman test suggested lower values for RA, and RR under HO. However, 476 the magnitude of these differences was small (Figure 3) and the differences in RS 477 u t h between HO and NDP were apparently more related to differences in RH (36% lower in 478 479 A HO than in NDP). But the vegetation effect on RS was not indirect via modifications in the environmental conditions (as we expected in H3b), since no differences were found 480 in SWC or soil temperature along the die-off stages or between species. These 481 differences in RS were probably driven by changes in microbial community composition 482 and functional diversity. This statement is supported by (i) the observed species-specific 483 microbial community under each type of tree at the study site and the lack of differences 484 in soil environmental conditions and in soil C pools (both quality and quantity) (Curiel 485 Yuste et al. 2012) and by (ii) the different functional diversity of the decomposer 486 community between type of trees as observed in a litter decomposition experiment in 487 the same study site (Barba et al. 2015). 488 This shift towards lower RS under holm oak following Scots pine drought-induced 489 mortality could have crucial implications for the carbon balance of this particular 490 ecosystem and, by extension, for the carbon cycling in Mediterranean drought-exposed 491 Scots pine forests where a gradual replacement by Quercus species is increasingly being 492 reported (Martínez-Vilalta et al. 2012; Galiano et al. 2013; Vilà-Cabrera et al. 2013; 493 Carnicer et al. 2014). However, we aware that this study has been performed with 494 limited number of replicates and studies addressing this question at ecosystem level 495 should be required in order to make more reliable projections. Additionally, selecting 496 individual trees as the experimental unit allows the comparison of soil respiration and 497 its components between different types of trees, but hinders the extrapolation of the 498 results to the forest level. 499 500 4.4. Environmental controls of autotrophic and heterotrophic soil respiration along a 501 die-off gradient and species succession. 502 The interactions between season-SWC-temperature and temperature-type of tree were 503 i able to explain 51% of the RS variability. The positive interaction between SWC and p t 504 s cr temperature indicated that the higher the SWC, the higher the temperature effect, and u 505 a n had similar effects on RH, RR and RM fluxes. However, only during part of the spring- dm 506 time were there simultaneous high SWC and temperature values (see figure 2). We did 507 pt e not find support for our hypothesized higher sensitivity of RS and RA to SWC under c ce 508 Scots pine (H2). On the contrary, we observe a higher sensitivity of RH to SWC under a or’s 509 HO (Table 2, RH model; Figure 6), which could be due to a better adaptation to the 510 strong seasonal changes in water availability of the microbial community associated 511 u t h with this typical Mediterranean species (Curiel Yuste et al. 2014) or to higher microbial 512 513 A biomass under HO resulting in a higher response to changes in SWC. While environmental variables have been described as major drivers of RA at global 514 scale (Piao et al. 2010), vegetation activity has been usually described as the most 515 determinant factor in RS at ecosystem scale (Janssens et al. 2001; Tang et al. 2005; 516 Högberg et al. 2009). Therefore, the limited explanatory power of environmental 517 variables on autotrophic component models obtained in this study (and expected in 518 H3a) could be explained by the lack of variables reflecting properly the photosynthetic 519 activity of trees. Indeed a parallel study in the same area has shown that variations in RR 520 at tree level were positively correlated with the sap flow of the same trees (Pereira- 521 Blanco 2014). This highlights the need for further studies on the dependency of RS and 522 its components on aboveground plant productivity (and not only on environmental 523 variables), both measured at stand level, to understand the ecosystem mechanisms to 524 cope with climate-driven disturbances. 525 526 6. Acknowledgments 527 The authors thank I. Azcoitia, G. Barba, J. Estrada, I. Ourêlo, P. Pellicer and I. Urbina 528 for their help in fieldwork and S. Vicca for her valuable comments. The insights from 529 two reviewers helped to improve the manuscript. This study was supported by the 530 Spanish Government projects SECASOL (CGL2009-08101), DRIM (CGL2010-16373), 531 VULGLO (CGL2010-22180-C03-03), SECADIN (CGL2012-32965) and VERONICA 532 (CGL2013-42271-P) by the Government of Catalonia grants (2009-SGR-00247 and 533 2014-SGR-453) and by a Community of Madrid grant REMEDINAL 2 (CM 534 S2009/AMB-1783). JB was supported by FPI (BES-2010-036558) and EEBB (EEBB-I- 535 13-07002) scholarships from the Spanish Ministry of Economy and Competitiveness. 536 i p t u s cr a n t e dm ce p a c t h or’s A u 537 7. References 538 Aguadé D, Poyatos R, Rosas T, Martínez-Vilalta J (2015) Comparative Drought 539 Responses of Quercus ilex L. and Pinus sylvestris L. in a Montane Forest 540 Undergoing a Vegetation Shift. Forests 6:2505–2529. 541 Allen CD, Breshears DD, McDowell NG (2015) On underestimation of global 542 vulnerability to tree mortality and forest die-off from hotter drought in the 543 Anthropocene. Ecosphere 6:8, art129. 544 Allen CD, Macalady AK, Chenchouni H, et al (2010) A global overview of drought and 545 heat-induced tree mortality reveals emerging climate change risks for forests. For 546 Ecol Manage 259:660–684. 547 Amiro BD, Barr AG, Barr JG, et al (2010) Ecosystem carbon dioxide fluxes after 548 disturbance in forests of North America. J Geophys Res. p t 549 Anderegg WRL, Kane JM, Anderegg LDL (2013) Consequences of widespread tree 550 cr mortality triggered by drought and temperature stress. Nat Clim Chang 3:30–36. i 551 n u s Asensio D, Peñuelas J, Ogaya R, Llusia J (2007) Seasonal soil and leaf CO2 exchange 552 a rates in a Mediterranean holm oak forest and their responses to drought conditions. dm 553 Atmos Environ 41:2447–2455. pt e ce 554 Barba J, Lloret F, Curiel Yuste J (2015) Effects of drought-induced forest die-off on c 555 litter decomposition. Plant Soil doi: 10.1007/s11104-015-2762-4. a or’s 556 Barba J, Curiel Yuste J, Martínez-Vilalta J, Lloret F (2013) Drought-induced tree h 557 species replacement is reflected in the spatial variability of soil respiration in a 558 559 A u t mixed Mediterranean forest. For Ecol Manage 306:79–87. Bartón K (2014) MuMIn: Multi-model inference. R package version 3.1-96. 560 Binkley D, Giardina C (1998) Why do tree species affect soils? The warp and woof of 561 tree-soil interactions. Biogeochemistry 42:89–106. 562 Binkley D, Stape JL, Takahashi EN, Ryan MG (2006) Tree-girdling to separate root and 563 heterotrophic respiration in two Eucalyptus stands in Brazil. Oecologia 148:447– 564 454. 565 Borkhuu B, Peckham SD, Ewers BE, et al (2015) Does soil respiration decline 566 following bark beetle induced forest mortality? Evidence from a lodgepole pine 567 forest. Agric For Meteorol 214-215:201–207. 568 Brown M, Black TA, Nesic Z, et al (2010) Impact of mountain pine beetle on the net 569 ecosystem production of lodgepole pine stands in British Columbia. Agric For 570 Meteorol 150:254–264. 571 Carnicer J, Coll M, Ninyerola M, et al (2011) Widespread crown condition decline, 572 food web disruption, and amplified tree mortality with increased climate change- 573 type drought. Proc Natl Acad Sci U S A 108:1474–1478. 574 Carnicer J, Coll M, Pons X, et al (2014) Large-scale recruitment limitation in 575 Mediterranean pines : the role of Quercus ilex and forest successional advance as 576 key regional drivers. Glob Ecol Biogeogr 23:371–384. 577 Cornwell WK, Cornelissen JHC, Amatangelo K, et al (2008) Plant species traits are the 578 predominant control on litter decomposition rates within biomes worldwide. Ecol 579 Lett 11:1065–1071. 580 Curiel Yuste J, Barba J, Fernandez-Gonzalez AJ, et al (2012) Changes in soil bacterial 581 community triggered by drought-induced gap succession preceded changes in soil 582 C stocks and quality. Ecol Evol 2:3016–3031. 583 Curiel Yuste J, Fernandez-Gonzalez AJ, Fernandez-Lopez M, et al (2014) Strong t 584 functional stability of soil microbial communities under semiarid Mediterranean i p 585 conditions and subjected to long-term shifts in baseline precipitation. Soil Biol cr 586 Biochem 69:223–233. n u s a 587 Dixon RK, Solomon AM, Brown SE, et al (1994) Carbon pools and flux of global forest dm 588 ecosystems. Science (80- ) 263:185–90. 589 pt e Edburg SL, Hicke JA, Brooks PD, et al (2012) Cascading impacts of bark beetle-caused ce 590 tree mortality on coupled biogeophysical and biogeochemical processes. Front a c 591 Ecol Environ 10:416–424. or’s 592 Ek H (1997) The influence of nitrogen fertilization on the carbon economy of Paxillus t h 593 involutus in ectomycorrhizal association with Betula pendula. New Phytol A u 594 135:133–142. 595 Galiano L, Martínez-Vilalta J, Eugenio M, et al (2013) Seedling emergence and growth 596 of Quercus spp. following severe drought effects on a Pinus sylvestris canopy. J 597 Veg Sci 24:580–588. 598 Galiano L, Martínez-Vilalta J, Lloret F (2010) Drought-Induced Multifactor Decline of 599 Scots Pine in the Pyrenees and Potential Vegetation Change by the Expansion of 600 Co-occurring Oak Species. Ecosystems 13:978–991. 601 Gough CM, Hardiman BS, Nave LE, et al (2013) Sustained carbon uptake and storage 602 following moderate disturbance in a Great Lakes forest. Ecol Appl 23:1202–1215. 603 Heinemeyer A, Hartley IP, Evans SP, et al (2007) Forest soil CO2 flux: uncovering the 604 contribution and environmental responses of ectomycorrhizas. Glob Chang Biol 605 13:1786–1797. 606 Hereş AM, Martínez-Vilalta J, Claramunt López B (2012) Growth patterns in relation to 607 drought-induced mortality at two Scots pine (Pinus sylvestris L.) sites in NE 608 Iberian Peninsula. Trees 26:621–630. 609 Hereter A, Sánchez JR (1999) Experimental areas of Prades and Montseny. In: Rodà F, 610 Retana J, Gracia CA, Bellot J (eds) Ecology of Mediterranean Evergreen Oak 611 Forests. Springer, Berlin, pp 15–27 612 Högberg P, Löfvenius MO, Nordgren A (2009) Partitioning of soil respiration into its 613 autotrophic and heterotrophic components by means of tree-girdling in old boreal 614 spruce forest. For Ecol Manage 257:1764–1767. 615 Högberg P, Nordgren A, Buchmann N, et al (2001) Large-scale forest girdling shows 616 that current photosynthesis drives soil respiration. Nature 411:789–92. 617 Högberg P, Read DJ (2006) Towards a more plant physiological perspective on soil 618 ecology. Trends Ecol Evol 21:548–554. 619 Jackson RB, Mooney HA, Schulze ED (1997) A global budget for fine root biomass, 620 surface area, and nutrient contents. Proc Natl Acad Sci U S A 94:7362–7366. t 621 Janssens IA, Lankreijer H, Matteucci G, et al (2001) Productivity overshadows i p 622 cr temperature in determining soil and ecosystem respiration across European forests. 623 Glob Chang Biol 7:269–278. n u s a 624 Keenan T, García R, Friend AD, et al (2009) Improved understanding of drought dm 625 controls on seasonal variation in Mediterranean forest canopy CO2 and water e 626 fluxes through combined in situ measurements and ecosystem modelling. 627 Biogeosciences 6:1423–1444. pt 628 a c ce Kuzyakov Y (2006) Sources of CO2 efflux from soil and review of partitioning or’s 629 methods. Soil Biol Biochem 38:425–448. t h 630 Levy-Varon JH, Schuster WSF, Griffin KL (2012) The autotrophic contribution to soil A u 631 respiration in a northern temperate deciduous forest and its response to stand 632 disturbance. Oecologia 169:211–220. 633 Levy-Varon JH, Schuster WSF, Griffin KL (2014) Rapid rebound of soil respiration 634 following partial stand disturbance by tree girdling in a temperate deciduous forest. 635 Oecologia 174:1415–1424. 636 Lloret F, Escudero A, Iriondo JM, et al (2012) Extreme climatic events and vegetation: 637 the role of stabilizing processes. Glob Chang Biol 18:797–805. 638 Malhi Y, Baldocchi DD, Jarvis PG (1999) The carbon balance of tropical, temperate 639 and boreal forests. Plant Cell Environ 22:715–740. 640 Martínez-Vilalta J, Aguadé D, Banqué M, et al (2012) Las poblaciones ibéricas de pino 641 albar ante el cambio climático: con la muerte en los talones. Rev Ecosistemas 642 21:15–21. 643 Martínez-Vilalta J, Piñol J (2002) Drought-induced mortality and hydraulic architecture 644 in pine populations of the NE Iberian Peninsula. For Ecol Manage 161:247–256. 645 Moore DJP, Trahan NA, Wilkes P, et al (2013) Persistent reduced ecosystem respiration 646 after insect disturbance in high elevation forests. Ecol Lett 16:731–737. 647 Moyano FE, Atkin OK, Bahn M, et al (2010) Respiration from roots to 648 mycorrhizosphere. In: Kutsch WL, Bahn M, Heinemeyer A (eds) Soil Carbon 649 Dynamics. An integrated methodology, 1st edn. Cambridge University Press, 650 Cambridge, pp 127–156 651 Nave LE, Gough CM, Maurer KD, et al (2011) Disturbance and the resilience of 652 coupled carbon and nitrogen cycling in a north temperate forest. 653 Ninyerola M, Pons X, Roure JM (2007a) Objective air temperature mapping for the 654 Iberian Peninsula using spatial interpolation and GIS. Int J Climatol 27:1231– 655 1242. 656 Ninyerola M, Pons X, Roure JM (2007b) Monthly precipitation mapping of the Iberian 657 Peninsula using spatial interpolation tools implemented in a Geographic t 658 Information System. Theor Appl Climatol 89:195–209. 659 cr i Palacio S, Maestro M, Montserrat-Martí G (2007) Seasonal dynamics of non-structural p 660 u s carbohydrates in two species of mediterranean sub-shrubs with different leaf n a 661 phenology. Environ Exp Bot 59:34–42. e dm 662 Peñuelas J, Lloret F, Montoya R (2001) Severe drought effects on Mediterranean 663 pt woody flora in Spain. For Sci 47:214–218. 664 a c ce Pereira-Blanco E (2014) Response of fine root respiration to variations in biotic and or’s 665 abiotic factors in a mixed Mediterranean forest affected by drought induced 666 secondary succession. Universitat Autònoma de Barcelona u t h A 667 Piao S, Luyssaert S, Ciais P, et al (2010) Forest annual carbon cost: a global-scale 668 analysis of autotrophic respiration. 91:652–661. 669 Pinheiro J, Bates D, DepRoy S (2009) Linear and Nonlinear Mixed Effects Models. R 670 package version 3.1-96. 671 Poyatos R, Aguadé D, Galiano L, et al (2013) Drought-induced defoliation and long 672 periods of near-zero gas exchange play a key role in accentuating metabolic 673 decline of Scots pine. New Phytol 200:388–401. 674 Redding T, Winkler R, Teti P, et al (2008) Mountain pine beetle and watershed 675 hydrology. BC J Ecosyst Manag 9:33–50. 676 Reed DE, Ewers BE, Pendall E (2014) Impact of mountain pine beetle induced 677 mortality on forest carbon and water fluxes. Environ Res Lett 9:105004. 678 Reichstein M, Bahn M, Ciais P, et al (2013) Climate extremes and the carbon cycle. 679 Nature 500:287–95. 680 Reichstein M, Tenhunen JD, Roupsard O, et al (2002) Severe drought effects on 681 ecosystem CO2 and H2O fluxes at three Mediterranean evergreen sites: revision of 682 current hypotheses? Glob Chang Biol 8:999–1017. 683 Rey A, Pegoraro E, Tedeschi V, et al (2002) Annual variation in soil respiration and its 684 components in a coppice oak forest in Central Italy. Glob Chang Biol 8:851–866. 685 Ruehr NK, Buchmann N (2010) Soil respiration fluxes in a temperate mixed forest: 686 seasonality and temperature sensitivities differ among microbial and root- 687 rhizosphere respiration. Tree Physiol 30:165–176. 688 Stocker TF, Qin D, Plattner GK, et al (eds) (2013) IPCC, 2013: Climate Change 2013: 689 The Physical Science Basis. Contribution of Working Group I to the Fifth 690 Assessment Report of the Intergovernmental Panel on Climate Change. 691 Strickland MS, Lauber C, Fierer N, Bradford MA (2009) Testing the functional 692 significance of microbial community composition. Ecology 90:441–451. 693 i p t cr Subke J-A, Inglima I, Francesca Cotrufo M (2006) Trends and methodological impacts 694 in soil CO2 efflux partitioning: A metaanalytical review. Glob Chang Biol 12:921– 695 943. n u s a dm 696 Subke J-A, Voke NR, Leronni V, et al (2011) Dynamics and pathways of autotrophic e 697 and heterotrophic soil CO2 efflux revealed by forest girdling. J Ecol 99:186–193. pt ce 698 Sus O, Poyatos R, Barba J, et al (2014) Time variable hydraulic parameters improve the a c 699 performance of a mechanistic stand transpiration model. A case study of or’s 700 Mediterranean Scots pine sap flow data assimilation. Agric For Meteorol 198- 701 199:168–180. u t h A 702 Tang J, Baldocchi DD, Xu L (2005) Tree photosynthesis modulates soil respiration on a 703 diurnal time scale. Glob Chang Biol 11:1298–1304. 704 Tedeschi V, Rey A, Manca G, et al (2006) Soil respiration in a Mediterranean oak forest 705 at different developmental stages after coppicing. Glob Chang Biol 12:110–121. 706 Uren NC (2000) Types, amounts, and possible functions of compounds released into the 707 rhizosphere by soil-grown plants. In: Pinton R, Varanini Z, Nannipieri P (eds) The 708 Rhizosphere: Biochemistry and Organic Substances at the Soil-Plant Interface, 2nd 709 edn. Marcel Dekker, New York, pp 19–40 710 Vilà-Cabrera A, Martínez-Vilalta J, Galiano L, Retana J (2013) Patterns of Forest 711 Decline and Regeneration Across Scots Pine Populations. Ecosystems 16:323–335. 712 Vivanco L, Austin AT (2008) Tree species identity alters forest litter decomposition 713 through long-term plant and soil interactions in Patagonia, Argentina. J Ecol 714 96:727–736. 715 716 Tables 717 Table 1. Environmental variables during the study period summarized by type of tree 718 (mean and standard deviation). No differences has been found between type of trees 719 (p<0.05, mixed effects model with tree as random factor). NDP: non-defoliated pines; 720 DFP: defoliated pines; DP: dead pines; HO: holm oaks. Units: soil temperature in °C; 721 SWC in cm3cm-3; litterfall in g m-2 d-1. NDP DFP DP HO Mean SD Mean SD Mean SD Mean SD Temperature 11.68 0.29 11.70 0.68 12.00 0.40 11.21 0.14 SWC 20.20 0.04 20.21 0.06 20.21 0.02 20.20 0.05 Litterfall 1.84 0.67 1.19 0.11 1.92 0.89 1.96 0.19 i p t cr 722 n u s a t e dm ce p a c t h or’s A u 723 Table 2. Summary of the selected model of RS, RH, RA, RR and RM. NDP: non- 724 defoliated pines; DFP: defoliated pines; DP: dead pines; HO: holm oaks. Holm oak and 725 summer are used as the reference categories and are included in the intercept. Different 726 lowercase letters indicate significant differences between levels of predictor variables. RS Variables Estimate SE t-value p-value Intercept 6.971 2.016 3.458 0.001 SWC -0.563 0.145 -3.888 <0.001 Temperature -0.412 0.109 -3.798 <0.001 Summer - - - - Season Fall -8.275 3.410 -2.427 0.016 t Winter -5.869 2.640 -2.223 0.027 Spring -4.132 3.392 -1.218 0.224 cr i p HO - - - n - u s a Type NDP 1.102 0.669 1.647 0.138 dm DFP 0.426 0.673 0.633 0.545 t e DP 0.287 0.664 0.433 0.677 SWC*Temp 0.044 ce p 0.008 5.356 <0.001 a c or’s Summer*SWC - - - - Season* SWC Fall*SWC 0.596 0.182 3.269 0.001 u t h Winter*SWC 0.536 0.162 3.305 0.001 A Spring*SWC 0.423 0.196 2.161 0.032 Summer*Temp - - - - Season* Temp Fall*Temp 0.840 0.339 2.480 0.014 Winter*Temp 0.476 0.328 1.450 0.148 Spring*Temp 0.347 0.270 1.285 0.200 Temp*Type ab Temp*HO - - - - a Temp*NDP -0.024 0.027 -0.881 0.379 ab Temp*DFP 0.016 0.027 0.608 0.544 b Temp*DP 0.051 0.026 1.963 0.051 Season*SWC* a Sum*SWC*Temp - - - - b Temp Fall*SWC*Temp -0.050 0.015 -3.453 0.001 b Win*SWC*Temp -0.043 0.015 -2.875 0.004 ab Spr*SWC*Temp -0.029 0.015 -1.949 0.052 727 728 RH Variables Estimate SE t-value p-value Intercept -0.734 0.578 -1.270 0.205 SWC -0.032 0.029 -1.088 0.278 Temperature 0.027 0.033 0.826 0.410 SWC*Temp 0.009 0.002 1.146 0.000 Summer - - - - Season Fall 1.344 0.617 2.179 0.030 Winter 1.455 0.574 2.536 0.012 Spring -0.079 0.586 -0.133 0.894 HO - - - - i p t Type NDP 1.526 0.535 2.853 0.021 s cr DFP 1.586 0.511 3.102 0.015 DP 1.667 0.479 3.480 a n u0.008 dm Season*Temp a Summer*Temp - - - - e b t Fall*Temp -0.112 0.051 -2.185 0.030 ce p b Winter*Temp -0.127 0.054 -2.345 0.020 c c Spring*Temp 0.101 0.040 2.535 0.012 a or’s a SWC*HO - - - - SWC*Type h b SWC*NDP -0.050 0.022 -2.289 0.023 A u t SWC*DFP SWC*DP -0.061 -0.059 0.021 0.019 -2.960 -3.191 0.003 0.002 b b 729 RA Variables Estimate SE t-value p-value Intercept 1.611 0.586 2.747 0.006 Litterfall 0.076 0.035 2.204 0.028 SWC -0.062 0.026 -2.379 0.018 Temperature -0.095 0.037 -2.577 0.011 SWC*Temp 0.007 0.002 3.850 <0.001 730 731 RR Variables Estimate SE t-value p-value Intercept -1.329 0.767 -1.731 0.0845 SWC 0.093 0.028 3.331 0.001 Temperature 0.041 0.028 1.463 0.145 a Summer - - - - Season 2.213 1.236 1.790 0.075 ab Fall 1.868 1.109 1.684 0.093 ab Winter 3.978 1.216 3.271 0.001 b Spring Season*Temp Season*SWC a Summer*SWC - - - - -0.123 0.053 -2.329 0.021 b Fall*SWC -0.093 0.045 -2.077 0.039 b Winter*SWC t -0.128 0.046 -2.789 0.006 b Spring*SWC Summer*Temp - - - - a cr i p u s 0.043 0.060 0.709 0.479 ab Fall*Temp a n -0.070 0.063 -1.097 0.273 ab Winter*Temp dm -0.182 0.045 -4.036 0.000 b Spring*Temp pt e ce 732 RM a c t h Variables or’s Estimate SE t-value p-value A u Intercept SWC 1.286 -0.054 0.544 0.023 2.366 -2.395 0.019 0.017 Temperature -0.081 0.032 -2.511 0.013 SWC*Temp 0.005 0.002 3.076 0.002 733 734 Figure captions 735 Figure 1. Root-exclusion experimental design. Large collars depict the different 736 exclusion treatments for soil CO2 efflux measurements, small collars are 737 pseudoreplicates within treatments and small rectangles in B represent the mesh 738 allowing mycorhizae ingrowth. A, roots and mycorhizae included; B, roots excluded 739 and mycorhizae included; C, roots and mycorhizae excluded. 740 741 Figure 2. Seasonal course of environmental variables over the study period (June 2012 742 – June 2013). a) daily soil temperature (mean of all trees and treatments ± sd); b) daily 743 i p t precipitation; c) daily SWC (mean of all trees ± sd); d) daily litterfall (mean of all trees 744 u s ± sd). Daily soil temperature for each tree and treatment was modelled with soil cr a n dm 745 temperature campaign data and daily air temperature. Similarly, daily SWC for each 746 pt e tree was modelled with SWC campaign data and daily SWC measured continuously at 6 747 c ce points randomly distributed in the study site (see section 2.2.2). Litterfall data showed at a or’s 748 panel d) are the campaigns data. 749 u t h 750 A Figure 3. Annual soil respiration evolution and its components. a) Total soil respiration 751 [RS]; b) Heterotrophic respiration [RH]; c) Root respiration [RR]; d) Micorrhyzal 752 respiration [RM]. Each dot represents the average of the 5 measurements within a 24-h 753 cycle and the 3 trees of each type. Campaign-specific error bars were not drawn for 754 better clarity. NDP: non-defoliated pines; DFP: defoliated pines; DP: dead pines; HO: 755 holm oaks. 756 757 Figure 4. Soil heterotrophic respiration (RH), micorrhyzal respiration (RM) and root 758 respiration (RR) contribution (in %) to total soil respiration (RS) during campaigns (from 759 June 2012 to June 2013) considering all tree types together. 760 761 Figure 5. Modelled soil temperature responses of RS at different levels of soil moisture, 762 for each season and type of tree, according to the RS model in Table 2. Modelled 763 responses have been drawn only for the observed soil temperature range for each 764 season. 10, 20 and 30% levels of SWC have been selected to cover the whole range of 765 SWC on the study site. 766 i p t 767 u s cr Figure 6. Modelled soil temperature responses of RH at different levels of soil moisture, a n dm 768 for each season and type of tree, according to the RH model in Table 2. Modelled 769 pt e responses have been drawn only for the observed soil temperature range for each 770 c ce season. NDP: non-defoliated pines; DFP: defoliated pines; DP: dead pines; HO: holm a or’s 771 oaks. 772 u t h 773 A Figure 7. Annual RS and its components (mean ± sd) modelled for each type of tree. 774 NDP: non-defoliated pines; DFP: defoliated pines; DP: dead pines; HO: holm oaks. 775 Different normal font letters indicate significant differences between types of tree 776 within each respiration component (p<0.05, mixed-effects model). Different letters in 777 italics indicate marginally significant differences (p<0.1, mixed-effects model). 778 779 Figure 1 780 i p t u scr a n t edm ce p a c t h or’s A u 781 Figure 2 i p t u scr a n t edm ce p a c t h or’s A u 782 783 Figure 3 i p t u scr a n t edm ce p a c t h or’s A u 784 785 Figure 4 i p t cr 786 n u s a t edm ce p a c t h or’s A u 787 Figure 5 i p t u scr a n t edm ce p a c t h or’s A u 788 789 790 Figure 6 i p t u scr a n t edm ce p a c t h or’s A u 791 792 Figure 7 793 i p t u scr a n t edm ce p a c t h or’s A u