Beneath the canopy: Linking drought-induced forest die off and changes in soil properties

Forest Ecology and Management 422: 294-302 (2018) 1 1 Beneath the canopy: linking drought-induced forest die off and changes in 2 soil properties 3 4 Antonio Gazol1, J. Julio Camarero1, J. José Jiménez2, David Moret-Fernández3, M. Victoria 5 López3; Gabriel Sangüesa-Barreda1 and José-Mariano Igual4 6 7 1Pyrenean Institute of Ecology (IPE-CSIC), E-50059 Zaragoza, Spain 8 2Pyrenean Institute of Ecology (IPE-CSIC), E-22700 Jaca, Spain 9 3Aula Dei Experimental Station (EEAD-CSIC), E-50080 Zaragoza, Spain 10 4Institute of Natural Resources and Agrobiology of Salamanca (IRNASA-CSIC), E-37008 11 Salamanca, Spain 12 13 Corresponding author: Dr. Antonio Gazol 14 Instituto Pirenaico de Ecología (IPE-CSIC); Avda. Montañana 1005, 50059, Zaragoza, Spain 15 E-mail: [email protected] Telephone: (+34) 976369393 (ext. 880039) 17 Fax: (+34) 974363222 18 19 Forest Ecology and Management 422: 294-302 (2018) 2 Abstract 20 Climate warming and the occurrence of more severe dry spells are causing widespread 21 drought-induced forest die-off events. Despite research on drought-triggered die-off 22 processes is rapidly increasing, little is known on how soil conditions and rhizosphere 23 features are affected by canopy dieback and tree death. We studied the soils in the 24 rhizosphere of three coniferous forests where die-off was induced by a severe drought in 25 2012. We selected three forest types subjected to contrasting climatic and edaphic conditions 26 dominated by three different tree species: silver fir (Abies alba; temperate conditions), Scots 27 pine (Pinus sylvestris; continental and Mediterranean conditions) and Aleppo pine (Pinus 28 halepensis; semi-arid and Mediterranean conditions). In each forest, we analyzed soil 29 physical characteristics such as water retention capacity and soil texture, nutrient availability 30 and microbial community structure (Phospholipid fatty acids, PLFA) below non-declining 31 and declining or dying trees. We did not observe differences in nutrient availability between 32 the two vigor classes. Conversely, we found strong differences in soil microbial community 33 structure below non-declining and declining trees in the Silver fir and Aleppo pine stands. 34 Soils in the Scots pine stand presented extremely low values of soil saturated sorptivity 35 indicating a reduction of soil water infiltration after prolonged dry periods which could 36 exacerbate drought stress. We conclude that forest dieback impacts the soil microbial 37 community structure in the short term. Further research is required to understand the linkages 38 between a reduced capacity of soil water infiltration after prolonged droughts, short-term 39 changes in the soil microbiota, long-term nutrient imbalances and tree death. Soil conditions 40 shall be considered as an important part of forest management strategies after drought41 induced die off. 42 Keywords: forest dieback, PLFA, rhizosphere, soil hydrophobicity, soil microbial community 43 structure, tree death. 44 Forest Ecology and Management 422: 294-302 (2018) 3 1. Introduction 45 The reports of forest die-off events triggered by dry spells have increased 46 considerably during the last decades (Allen et al., 2010). Such rising trend is of major 47 concern for many forests worldwide if global warming amplifies drought stress (Anderegg et 48 al., 2013; Allen et al., 2015; Camarero et al., 2015). The consequences of drought-induced 49 forest die-off can be complex, acting at several temporal and spatial scales (Ruthrof et al., 50 2016), namely from the simple replacement of some trees to large compositional changes at 51 the community level or even shifts in productivity at the ecosystem level due to the increase 52 of soil surface temperature and decline in evapotranspiration (Royer et al., 2011; Anderegg et 53 al., 2012). While catastrophic and widespread die-off events that cause the death of most 54 trees have more drastic effects on forests, gradual die-off processes affecting some 55 individuals, while other neighbouring trees survive, have less obvious consequences on tree56 soil interactions. How soils will respond to the forecasted increase in drought and related 57 forest die-off disturbances is a pivotal question in current global-change ecology (Curiel 58 Yuste et al., 2011; Brunner et al., 2015; Phillips et al., 2016; Baldrian 2017). 59 Many ecosystem goods and services provided by forests depend on ecosystem 60 processes such as carbon (C) and nitrogen (N) cycling which are part of forest belowground 61 processes (van Der Heijden et al., 2008; Curiel Yuste et al., 2011; Baldrian 2017). For 62 example, the C stock in forest soils exceeds by far that in aboveground vegetation (Jobbágy 63 and Jackson, 2000), and it is estimated that 40-70% of the C photosynthetically assimilated 64 by forests is transferred to the rhizosphere (Hopkins et al., 2013). Along these lines, the C 65 balance between aboveground vegetation and soil microbial respiration is controlled by N 66 supply which controls C uptake by plants (Thomas et al., 2015; Wurzburger and Brookshire 67 2017). Thus, nutrient cycles largely depend on the interaction between physicochemical soil 68 properties, tree species identity and microbial communities. 69 Forest Ecology and Management 422: 294-302 (2018) 4 The occurrence of abrupt disturbances such as forest die-off events caused by drought 70 or pest outbreaks can have important influences on soil microbial communities and 71 belowground nutrient cycling (Stursova et al., 2014; Pold et al., 2015). Pathogens causing 72 forest die off (e.g. Phytophthora and Heterobasidion species) develop a substantial part of 73 their life-cycles in the soil (Brasier et al., 1993; Oliva and Colinas 2007). In addition, soil 74 physical and chemical properties such as low water-holding capacity largely affected by soil 75 hydrophobicity (water repellence), and nutrient deficiency can exacerbate the negative 76 consequences of drought stress on tree vigour leading to forest die-off (Doerr et al., 2000; 77 Pinto and Peñuelas 2007; Hallett et al., 2011). However, these changes in soil conditions can 78 arise because of forest die-off since tree species can modify soil organic matter content and 79 water retention capacity (Lebron et al., 2007). 80 The influence of drought-related die-off on forest soil conditions may have different 81 pathways. If phloem transport within declining trees is weakened, the composition of soil 82 nutrients below the tree may change since C and N organic concentrations in the soil decrease 83 (Dannenmann et al., 2009) while inorganic N concentration increases (Kreuzwieser and 84 Gessler 2010). This has been shown by girdling experiments causing tree death and altering 85 the quantity and quality of C and N compounds reaching the rhizosphere (Stursova et al., 86 2014; Pold et al., 2015). In dry regions, drought can also reduce phosphorous (P) and 87 potassium (K) uptake by trees increasing their concentration in the soil (Sardans and Peñuelas 88 2007). If rhizodeposition, i.e. root exudates (organic compounds released into the soil by 89 plant roots), is depressed then there may be a decrease of microbial biomass or activity in the 90 rhizosphere (Dannenmann et al., 2009) or changes in the diversity of the soil microbiota 91 (Schulze et al., 2005). 92 The processes involving the interaction between trees and soil microbes are very 93 dynamic in space and time (Bahram et al., 2015; Baldrian 2017). For example, soil microbes 94 Forest Ecology and Management 422: 294-302 (2018) 5 are able to utilize root exudates in minutes or few hours, but they will need several years to 95 decompose dead wood (Baldrian 2017). The decline and death of a tree reduces or stops the 96 supply of carbohydrates from the roots to the soil via root exudation (Brunner et al., 2015). 97 However, the subsequent fall of dead needles and branches could feed soil microorganisms 98 with different carbon sources through decomposition (Kana et al., 2012; Mikkelson et al., 99 2016). However, the use of this readily available organic matter by microorganisms depends 100 on the composition and activity of microbial communities as well as on environmental 101 conditions (Pisani et al., 2014; Kuzyakov and Blagodatskaya, 2015). Thus, it is expected that 102 drought-triggered forest die-off is anticipated to induce changes in soil microbial structure 103 (Curiel Yuste et al., 2012; Stursova et al., 2014; Lloret et al., 2015). Nevertheless, it is 104 unclear if these post-drought changes in the soil microbiota are similar among forests 105 showing die-off but presenting different soil types and subjected to contrasting climate 106 conditions. 107 Drought could also alter physical and chemical soil features (structure, water holding 108 capacity, pH, hydrophobicity, nutrient availability) and decrease the enzymatic activity of the 109 soil microbiota (Dannenmann et al., 2009; Curiel Yuste et al., 2011; Baldrian et al., 2013), 110 leading to a negative feedback on tree nutrition by reducing nutrient mineralisation by 111 bacteria (Kreuzwieser and Gessler, 2010). For instance, in Eucalyptus forest showing die-off 112 this was associated with changes in the soil bacterial functional diversity linked to a decrease 113 in the utilization of carbohydrates, amino acids and amines by the soil bacterial communities 114 in sites with declining tree health (Cai et al., 2010). Furthermore, forest die-off usually leads 115 to a reduction in canopy cover, which could increase the radiation reaching the ground and 116 enhance evaporation further exacerbating soil dryness and contributing to more extreme 117 microclimatic conditions in the uppermost soil, where most fine roots are found, thus 118 reducing the rhizosphere microbiota activity (Sardans et al., 2008). 119 Forest Ecology and Management 422: 294-302 (2018) 6 Tree species composition has important influences on forest soil physical and 120 chemical characteristics (Augusto et al., 2015). As a consequence, there can be a large 121 variation in soil microbial community composition between different forest types (Baldrian 122 2017). Drought may impact forest soil differently depending on the forest type studied and its 123 prevailing climatic conditions (Aponte et al., 2013). For example, while climate warming 124 may enhance soil enzyme activity in temperate hardwood forests of central Europe (Baldrian 125 et al., 2013), it might have contrasting influences on drought-prone Mediterranean forests 126 (Sardans and Peñuelas 2007; Sardans et al., 2008). However, the soil physical properties as 127 well as soil nutrient content and microbial community structure may also vary at the scale of 128 centimetre in forest soils (Koorem et al 2014; Nacke et al., 2016). Thus, forest soils may 129 differ between neighbouring declining and non-declining trees growing in the same stand 130 (Curiel Yuste et al., 2012; Mikkelson et al., 2016). 131 Here we compare belowground soil properties (physical features, nutrient availability, 132 microbiota composition) of coexisting declining or recently dead vs. non-declining trees in 133 three forest types which experienced a drought-induced die-off in 2012. These forests are 134 dominated by Scots pine (Pinus sylvestris), Silver fir (Abies alba) and Aleppo pine (Pinus 135 halepensis), respectively, and they are subjected to contrasting climatic conditions. After the 136 drought in 2012, we monitored the canopy cover, a proxy of tree vigour (Dobbertin 2005), of 137 surviving trees. In 2015, we measured soil physical and chemical characteristics and 138 characterized the soil microbial structure. We hypothesize that despite strong differences in 139 nutrient supply rates and soil microbial structure linked to different forest characteristics (soil 140 type, climate condition), die-off will be followed by a shift in nutrient availability and 141 microbial community structure in the rhizosphere of declining and recently dead trees. 142 143 2. Materials and methods 144 Forest Ecology and Management 422: 294-302 (2018) 7 Study sites and sampling protocols 145 We studied the populations of three conifers inhabiting three sites situated in Aragón (north146 eastern Spain) and subjected to contrasting climatic conditions (Camarero et al., 2015). 147 Particularly, we studied: a Scots pine (Pinus sylvestris L.) forest located in the Iberian System 148 (Corbalán, Teruel) subjected to a continental Mediterranean climate (Table 1), a silver fir 149 (Abies alba Mill.) forest situated in the Pyrenees (Paco Ezpela, Ansó) where temperate 150 conditions prevail, and an Aleppo pine (Pinus halepensis Mill.) forest situated in the Middle 151 Ebro Basin (Peñaflor, Zaragoza), close to the Monegros steppe, characterized by a semi-arid 152 Mediterranean climate (see sites’ features in Table 1). The Iberian system and the Pyrenees 153 represent the southernmost distribution limit of Scots pine and Silver fir in Europe. The Ebro 154 Basin is one of the driest regions in Europe. Climate warming and the occurrence of severe 155 droughts during the late 20th century and early 21st century (1986, 1994-1995, 2005, 2012) 156 have caused die-off events affecting the study forest types (Camarero et al., 2015; Vicente157 Serrano et al., 2010; Supporting Information, Fig. S1). 158 The Corbalan Scots pine forest present no evident signs of human management during 159 the last 50 years (Camarero et al. 2015). The Paco Ezpela silver-fir forest present signs that 160 intense logging activity was undertaken in the past such as stumps and wood trails (Sangüesa161 Barreda et al., 2015). Nevertheless, these management activities ceased in the early 1950s 162 (Camarero et al., 2015). The Peñaflor Aleppo pine stands were part of a large recreational 163 forest which in the past was reserved for the sole use of the aristocracy for hunting and 164 recreation purposes. No evident signs of management were observed in this forest since the 165 1950s. 166 Soils were of the loam and loamy sand types in all sites (Supporting Information, 167 Table S1). The soil pH varied from 6.79 in the Silver fir forest to 7.33 in the Scots pine forest 168 and 7.77 in the Aleppo pine forest. The percentage of clay was higher in the Silver fir forest 169 Forest Ecology and Management 422: 294-302 (2018) 8 than in the pine forests (13.22% more). Conversely, the percentage of sand was higher in the 170 Scots pine forest. 171 172 173 174 175 Forest Ecology and Management 422: 294-302 (2018) 9 Table 1. Main features of the three study conifer forests showing drought-induced forest die-off in northeaster Spain. The name of each site (and 176 region), the dominant tree species, together with size features (diameter at breast height – dbh; tree height) of monitored trees (mean ± SD) and 177 the percentage of dead trees in 2012 and 2015 are shown. MAT and TAP stand out for mean annual temperature and total annual precipitation, 178 respectively (data correspond to the period 1950-2012). 179 Site (nearby locality) Tree species Latitude (N) Longitude (W) Bedrock Soil type MAT (ºC) TAP (mm) Dbh (cm) Height (m) Mortality in 2012 (%) Mortality in 2015 (%) Corbalán (Teruel) Scots pine (Pinus sylvestris) 40º26’ 0º58’ Limestones Cambisol, loamy 12.0 371 27.3 ± 1.3 8.3 ± 0.3 24 82 Paco Ezpela (Ansó) Silver fir (Abies alba) 42º45’ 0º52’ Marls, limestones Cambisol, loamy 9.5 1153 36.7 ± 1.3 23.7 ± 0.6 37 47 Peñaflor (Zaragoza) Aleppo pine (Pinus halepensis) 41º47’ 0º44’ Marls, gypsum Regosol, loamy 15.2 324 32.7 ± 1.6 8.1 ± 0.3 16 29 Forest Ecology and Management 422: 294-302 (2018) 16 When the soil nutrient supply rates for the different elements were analyzed using NMDS, we 316 found significant differences in soil nutrients between forest types (Fig. 2a), but no 317 significant interaction between forest and tree vigour using PERMANOVA. Along these 318 lines, we found significant differences between forests for the supply rates (estimated using 319 PRS resin probes) of most of the mineral elements studied, but not between declining and 320 non-declining trees (Table 2; Fig. 3). This was the case of NO3, Ca, Mg, K, P, Fe, Al, Mn, 321 Zn, Cu, S and Pb but not for NH4, B and Cd. Thus, we found little support for the existence 322 of differences in soil nutrient content and nutrient supply rates between declining and non323 declining trees within each forest. 324 325 Table 2. Summary of the models and analyses of soil nutrient contents and soil nutrient 326 supply rates estimated using resin probes. Models were fitted using Generalized Least Square 327 models. A first model containing, forest, tree vigour (declining vs. non-declining trees) and 328 their interaction was fitted. Multimodel selection based on information criteria was applied to 329 select the best model. The model showing the lowest corrected Akaike Information Criteria 330 (AICc) was selected as the best model. Some of the variables were log-transformed 331 (log(x+1)) to reduce skewness. For each variable, we show whether the variables were 332 transformed (Y) or not (N), the most parsimonious model, AICc, and the model Akaike 333 weight (Wi) or probability that the selected model is the best one, and coefficients associated 334 to each variable and their significance (* P < 0.05; ** P < 0.01). The inclusion of forest in the 335 final model is based on the existence of differences between forests. 336 337 Variable Logtransformation AICc Wi Forest type Tree vigour Forest type * tree vigour Soil nutrient content C Y 3.05 0.82 9.16** N N 1.42 0.66 6.44** C/N N 3.59 0.86 71.71** Soil nutrient supply rate NO3 Y 7.88 0.98 26.54** NH4 Y 3.92 0.88 Ca N 7.11 0.97 12.05** Mg N 7.72 0.97 7.37** K Y 5.89 0.92 6.49** P Y 7.22 0.97 7.02** Fe Y 9.16 0.99 13.41** Mn Y 9.00 0.99 12.1** Cu Y 6.82 0.97 11.65** Zn Y 7.74 0.97 8.89* B Y 2.15 0.62 S Y 9.34 0.99 19.52** Forest Ecology and Management 422: 294-302 (2018) 17 Pb Y 7.25 0.96 7.04** Al Y 5.73 0.90 6.01* Cd Y 1.28 0.64 338 339 340 3.3.Soil microbial community structure 341 We observed a significant interaction between forest and tree vigour in soil microbial 342 community structure derived from PLFA analyses, indicating that there exist differences 343 between declining and non-declining trees (PERMANOVA main effect of forest F = 25.75; p 344 < 0.01; interaction between forest and status F = 4.77, p < 0.01 Fig. 2b). In the Silver fir and 345 Aleppo pine forests, declining and non-declining trees showed marked differences in soil 346 microbial community structure whereas no differences were found in the case of the Scots 347 pine forest (Fig. 2b). We found a strong correlation between the first axis of the soil 348 microorganism’s ordination (NMDS axis 1) and the PRS resin probes estimated content of 349 NO3 (r = 0.52; p < 0.01), K (r = 0.58; p < 0.01) and S (r = 0.50; p < 0.01) in the soil, and a 350 negative correlation with Fe (r = -0.53; p < 0.01), Cu (r = -0.53; p < 0.01) and Al (r = -0.48; p 351 = 0.01) soil estimated content. The second axis (NMDS 2) was related with the concentration 352 of NO3 (r = 0.49; p < 0.01), P (r = 0.53; p < 0.01) and Cd (r = 0.44; p = 0.02) in the soil. 353 354 Table 3. Summary of the PLFA samples. Models were fitted using Generalized Least Square 355 Models. A first model containing, forest, tree vigour (declining vs. non-declining) and their 356 interaction was fitted. Multimodel selection based on AIC criteria was applied to select the 357 best model. The model showing the lowest AICc was selected as the best model. Some of the 358 variables were log-transformed (log(x+1)) to reduce skewness. For each variable, we show 359 the best fitted model,  AICc, Wi or model weight (probability that the selected model is the 360 best one), and coefficients associated to each variable and their significance (* p < 0.05; **; p 361 < 0.01). The inclusion of forest in the final model is assumes differences between forest soils, 362 whereas the inclusion of the interaction between forest and tree vigour assumes differences in 363 soil type between tree vigour classes within each forest. 364 AICc Wi Forest type Tree vigour Forest type * tree vigour Total Biomass 5.25 0.93 23.56** 0.09 5.63* Eukaryote 2.65 0.79 7.17** 0.61 4.45* Gram Negative 8.83 0.99 21.22** 1.22 7.49** Gram Positive 4.94 0.92 12.43** 0.47 5.73** Forest Ecology and Management 422: 294-302 (2018) 18 Actinomycetes 7.14 0.97 35.58** 0.31 7.00** Fungi 3.92 0.88 20.71** AM Fungi 1.48 0.60 3.90* 365 Table 4. Comparisons of the abundances of the major groups of soil microorganisms between 366 declining and non-declining trees within each forest type. A separate two-way ANOVA was 367 performed to compare the major groups of microorganisms in which we found a significant 368 interaction between forest type and tree vigour (see Table 3). For each comparison, the F 369 statistic and its associated probability level (p) are shown. 370 Scots pine Silver fir Aleppo pine F p F p F p Total Biomass 2.64 0.11 5.26 0.03 3.66 0.07 Eukaryote 0.03 0.67 6.66 0.02 3.36 0.08 Gram Negative bacteria 3.66 0.07 5.35 0.03 5.74 0.03 Gram Positive bacteria 3.25 0.08 5.37 0.03 3.54 0.07 Actinomycetes 1.99 0.20 5.01 0.03 5.06 0.04 371 372 Fig. 2. Non-Metric multidimensional scaling (NMDS) biplots of a) soil nutrient supply rates 373 estimated using resin probes; b) and soil microbial composition based on PLFA 374 (Phospholipid-derived fatty acids) for the three forest types. The correlation between PRS 375 (Plant Root Simulator resin probes) soil nutrient composition and PLFA NMDS axes was 376 projected in the ordination diagram. The polygons indicate the centroid of the distribution of 377 the trees’ groups. Declining and non-declining trees are indicated by open and filled circles, 378 respectively, for the three studied forests. 379 380 Forest Ecology and Management 422: 294-302 (2018) 19 When the major groups of soil microorganisms were analysed separately we found 381 significant differences between declining and non-declining trees (Table 3, Fig. 4). In most 382 cases, we found a significant interaction between forest and tree vigour which suggest 383 differences between declining and non-declining trees (Table 3). The only exception to this 384 rule were the groups of fungi and AM fungi (Table 3). When separate analyses were 385 performed for each forest we found that the total biomass of microorganisms, eukaryotes, 386 gram positive and negative bacteria, and actinobacteria differed significantly between 387 declining and non-declining trees in the silver fir forest (Table 4). In the Aleppo pine forests 388 only Gram negative bacteria and actinobacteria showed significant differences between 389 classes of tree vigour. However, none of the analyzed microorganism types differed between 390 declining and non-declining trees in the Scots pine forest (Table 4). 391 392 Forest Ecology and Management 422: 294-302 (2018) 20 393 Fig. 3. Differences in soil nutrient supply rate (shown as g/10cm2/91 days) across forest 394 types and between declining (open boxes) and non-declining (solid boxes) trees. 395 396 Forest Ecology and Management 422: 294-302 (2018) 21 397 Fig. 4. Univariate analysis of the PLFA (Phospholipid fatty acids) showing the relative 398 abundance of major groups of soil microbes across the three forest types and between 399 declining (open boxes) and non-declining trees (solid boxes). The total biomass of microbes 400 is expressed in nanomoles per gram of dry soil; the abundance of the rest of the groups is 401 expressed as a log-ratio of the proportion. 402 403 404 Forest Ecology and Management 422: 294-302 (2018) 22 4. Discussion405 We found evidence that drought-induced forest die-off is linked to changes in soil 406 characteristics below coexisting trees of different vigour. Despite we found no differences in 407 soil nutrient content and soil nutrient supply rates between declining and non-declining trees 408 (Table 2, Fig. 2), we detected that soil microbial community structure differed significantly 409 between declining and non-declining trees in two of the three forests studies (Tables 3 and 4, 410 Figs. 2 and 4). Only in the Scots pine forest, where mortality rates peaked after the drought 411 and die-off was more widespread than in the other sites (Table 1), we did not find any 412 differences in soil microbial community structure between declining and non-declining trees 413 (Table 4; Fig. 2). Importantly, differences between coexisting declining and non-declining 414 trees are specific to each forest type and tree species. These results indicate that soil 415 conditions might change because of tree decline across entire stands, and this suggests that 416 differences between declining and non-declining trees are more prone to be detected in soil 417 microbiota than in soil nutrients. Notably, we found that a high proportion of soil samples in 418 two of the three studied forests, the Scots pine and the Aleppo pine forest, presented very low 419 saturated sorptivity (S) values (Fig. 1). Since these S values were smaller than those reported 420 in the literature for similar texture (Minasny and McBratney, 2000; Moret-Fernández et al., 421 2017) they could be related to a soil hydrophobicity phenomenon. Hydrophobicity could 422 drastically reduce the affinity of soil for water after prolonged drought periods (Hallett et al., 423 2011). These results suggest that the studied forests diverge in their ability to process rainfall 424 after prolonged droughts, which might exacerbate drought impacts on tree vigour. 425 Saturated sorptivity (S) was significantly lower in the Scots pine and Aleppo pine 426 forests than in the Silver fir forest but no differences were found in hydraulic conductivity 427 (Ks) between forests. The results obtained point to the existence of soil hydrophobicity or 428 water repellence in the pine forests studied (Doerr et al., 2000, Hallett et al., 2011). Soil water 429 Forest Ecology and Management 422: 294-302 (2018) 23 repellence have been found to be related with soil texture, dissolved organic compounds, dry 430 periods and some other natural perturbations such as fire or drought (Hallett et al., 2011). 431 Deciphering whether the observed soil water repellence is cause or consequence of the forest 432 die off is out of the scope of this study, but it should be a research priority. Soil water 433 repellence could exacerbate the effects of drought on tree health by reducing soil water 434 infiltration following the first rainfall events after prolonged dry spells often characterized by 435 high temperatures thus increasing drought stress and predisposing to forest die off (Goebel et 436 al., 2011). However, the death of trees can also favour soil water repellency by changing the 437 composition of soil organic contents and increasing light availability due to needles loss 438 (Lebron et al., 2007). We found the lowest values of sorptivity in the Scots pine forest (Fig. 439 1) in which only 9 out of 38 trees were marked as dead trees in 2012 whereas in 2015 only 6 440 trees remained alive. In conifer forests, higher soil organic matter is often linked to increased 441 soil water repellency during droughts as has been shown in Aleppo pine forests under semi442 arid conditions (Mataix-Solera et al., 2007). In previous works, we attributed the strong Scots 443 pine mortality observed in the region to the rise in air temperatures during early 2012 444 associated to climate warming but soil water repellence could be also a contribution factor to 445 explain tree death (Camarero et al., 2015). Low values of sorptivity were also found in 446 Aleppo pine forest studied, but this site is in a semi-arid region where both trees and soil 447 microbes might be better adapted to low soil water contents, which characterizes some of 448 these gypsum-rich substrates (Gazol et al., 2017). Furthermore, mortality rates were lower 449 there than in the most affected Scots pine forest (Camarero et al., 2015). Overall, longer 450 drought and warmer soil conditions enhance soil water repellency (Goebel et al., 2011), thus 451 further research is requited to investigate whether such low values of soil sorptivity and 452 strong soil water repellency are cause or consequence of tree death. 453 Forest Ecology and Management 422: 294-302 (2018) 24 We found no differences in soil nutrient concentrations and soil nutrients supply rates 454 between declining and non-declining trees. This pattern can be explained by different 455 reasons. Most of the nitrogen in the soil is contained in dead organic matter as insoluble 456 polymers and some microorganisms are responsible of transforming it to dissolved organic 457 nitrogen forms accessible to trees (van Der Heijden et al., 2008). However, trees and free-458 living soil microorganisms, particularly bacteria, can compete for the same sources of organic 459 nitrogen (van der Heijden et al., 2008; Dannenmann et al., 2009). Thus, tree death interrupts 460 nutrient uptake and carbon exudation by roots (Brunner et al., 2015), inducing in a short-term 461 change in soil microbial composition that might consume the available nitrogen that trees are 462 not using anymore. Similarly, and due to the short distance between the studied non-declining 463 trees and other non-studied healthy trees, there is a chance that roots from neighbouring trees 464 maintain the balance of soil nutrients in the surroundings of declining trees (Curiel Yuste et 465 al., 2012). Nevertheless, the most plausible explanation is that more time is required to detect 466 changes in soil nutrients after forest die-off or tree death (Dannenmann et al., 2009; Curiel 467 Yuste et al., 2012). For example, organic matter from death trees can have large amounts of 468 lignin and other compounds that need time to be decomposed (van Der Heijden et al., 2008; 469 Pold et al., 2015; Baldrian 2017). In conclusion, differences between declining and non-470 declining trees in soil microbial community structure can be detected rapidly after tree death 471 but more time might be required to detect differences in soil nutrient content. 472 We also found marked changes between declining and non-declining trees in two of 473 the study sites, which highlight the strong specificity of the rhizosphere at fine spatial scales 474 (Curiel Yuste et al., 2012; Koorem et al., 2014; Mikkelson et al., 2016), despite soil 475 microbiota also varies strongly from local to regional scales (Aponte et al., 2013; Lloret et al., 476 2015; Gazol et al., 2016; Tedersoo et al., 2016). Stursova et al. (2014) found drastic changes 477 in soil microbial communities, particularly in fungi species, after a rapid bark beetle-induced 478 Forest Ecology and Management 422: 294-302 (2018) 25 tree dieback. They attributed most of the changes in the soil microbiota to the drastic increase 479 of soil organic matter amount as a consequence of the raid fall of dead needles and branches. 480 However, in our study, we were comparing declining and non-declining trees at small spatial 481 scales (a few meters) and thus we do not expect major changes in canopy-derived organic 482 matter. In contrast, rhizodeposition rates can be shifted by droughts (Preece and Peñuelas 483 2016). Along these lines, Curiel Yuste et al. (2012) found that soils beneath death, declining 484 and non-declining Scots pine trees differed in soil bacterial community composition while no 485 major differences in soil nutrient contents were found. These results highlight that soil 486 microbial communities are able to respond rapidly to forest die-off and tree decline induced 487 by severe droughts (Cai et al., 2010; Lloret et al., 2015), although more time might be 488 required to detect changes in soil nutrient content. 489 Despite strong differences in soil microbial community structure between declining 490 and non-declining trees in the Aleppo pine and Silver fir forests, we found no differences at 491 all in the Scots pine forest where declining trees presented the lowest values of saturated 492 sorptivity (Table 4, Figs. 1 and 2). We hypothesize that this paticular behaviour is due to the 493 strong mortality observed in the Scots pine forest after 2012 (76% of trees tagged in 2012 494 were dead in 2015). While in the Aleppo pine and Silver fir forests tree decline, and die-back 495 is spatially heterogeneous, in the Scots pine forest the die-off event was widespread and some 496 of the individuals classified in 2012 as non-declining trees show evident signs of decline or 497 have dead three years later. 498 499 5. Conclusions 500 The links between drought, soil nutrient availability, shifts in the soil microbiota and forest 501 die-off are still little understood. Here we provide relevant information to start filling some of 502 these research gaps. The composition of soil microbiota seems to be a more rapid proxy and 503 Forest Ecology and Management 422: 294-302 (2018) 32 compounds in soils with increasing mean annual temperature. Organic Geochemistry 646 76:118–127. 647 Phillips, R.P., I. Ibáñez, L. D’Orangeville, P.J. Hanson, M.G. Ryan, and N.G. McDowell. 648 2016. A belowground perspective on the drought sensitivity of forests: towards improved 649 understanding and simulation. Forest Ecology and Management 280:309–320. 650 Pinheiro, J., D. Bates, S. DebRoy, and D. Sarkar. 2016. nlme: Linear and Nonlinear Mixed 651 Effects Models. R package version 3.1-131, https://CRAN.R-project.org/package=nlme. 652 Pinto, J., and J. Peñuelas. 2007. Drought changes phosphorus and potassium accumulation 653 patterns in an evergreen Mediterranean forest. Functional Ecology 21:191–201. 654 Pold, G., J.M. Melillo, and K.M. DeAngelis. 2015. Two decades of warming increases 655 diversity of a potentially lignolytic bacterial community. Frontiers in Microbiology 6:480. 656 Preece, C., and J. Peñuelas. 2016. Rhizodeposition under drought and consequences for soil 657 communities and ecosystem resilience. Plant and Soil 409:1–17. 658 R Development Core Team. 2017. R: A Language and Environment for Statistical 659 Computing. In. R Foundation for Statistical Computing, Vienna, Austria. 660 Royer, P.D., N.S. Cobb, M.J. Clifford, C.Y. Huang, D.D. Breshears, H.D. Adams, and J.C. 661 Villegas. 2011. Extreme climatic event-triggered overstorey vegetation loss increases 662 understorey solar input regionally: primary and secondary ecological implications. 663 Journal of Ecology 99:714–723. 664 Ruthrof, K.X., J.B. Fontaine, G. Matusick, D.D. Breshears, D.J. Law, S. Powell, and G. 665 Hardy. 2016. How drought induced forest die-off alters microclimate and increases fuel 666 loadings and fire potentials. International Journal of Wildland Fire 25:819–830. 667 Sangüesa-Barreda, G., Camarero, J.J., Oliva, J., Montes, F. and Gazol, A. 2015. Past logging, 668 drought and pathogens interact and contribute to forest dieback. Agricultural and Forest 669 Meteorology 208:85–94. 670 Forest Ecology and Management 422: 294-302 (2018) 33 Sardans, J., and J. Peñuelas, 2007. Drought changes phosphorus and potassium accumulation 671 patterns in an evergreen Mediterranean forest. Functional Ecology 21:191–201. 672 Sardans, J., J. Peñuelas, and R. Ogaya. 2008. Drought’s impact on Ca, Fe, Mg, Mo and S 673 concentration and accumulation patterns in the plants and soil of a Mediterranean 674 evergreen Quercus ilex forest. Biogeochemistry 87:49–69. 675 Schulze, E.-D., C. Wirth, D. Mollicone, and W. Ziegler. 2005. Succession after stand 676 replacing disturbances by fire, wind throw, and insects in the dark Taiga of Central 677 Siberia. Oecologia 146:77–88. 678 Stursova, M., J. Snajdar, T. Cajthaml, J. Barta, H. Santruckova, and P. Baldrian. 2014. When 679 the forest dies: the response of forest soil fungi to a bark beetle-induced tree dieback. 680 ISME Journal 8:1920–1931. 681 Taubner, H., B. Roth, and R. Tippkotter. 2009. Determination of soil texture: Comparison of 682 the sedimentation method and the laser-diffraction analysis. Journal of Plant Nutrition and 683 Soil Science 172:161–171. 684 Tedersoo, L, et al. 2016. Tree diversity and species identity effects on soil fungi, protists and 685 animals are context dependent. ISME Journal 10 :346–362. 686 Thomas, R.Q., E.N.J. Brookshire, and S. Gerber. 2015. Nitrogen limitation on land: how can 687 it occur in Earth system models? Global Change Biology 21:1777–1793. 688 van Der Heijden, M.G.A., R.D. Bardgett, and N.M. van Straalen. 2008. The unseen majority: 689 soil microbes as drivers of plant diversity and productivity in terrestrial ecosystems. 690 Ecology Letters 11:296–310. 691 Venables, W.N. and B.D. Ripley. 2002. Modern Applied Statistics with S. Springer. 692 Vicente-Serrano, S.M., Beguería, S., and López-Moreno, J.I. 2010. A Multi-scalar drought 693 index sensitive to global warming: The Standardized Precipitation Evapotranspiration 694 Index - SPEI. Journal of Climate 23: 1696–1718. 695 Forest Ecology and Management 422: 294-302 (2018) 34 Wurzburger, N. and E.N. Jack. Brookshire. 2017. Experimental evidence that mycorrhizal 696 nitrogen strategies affect soil carbon. Ecology 98:1491–1497. 697 Zelles, L. 1997. Phospholipid fatty acid profiles in selected members of soil microbial 698 communities. Chemosphere 35:275–294. 699 Zelles, L. 1999. Fatty acid patterns of phospholipids and lipopolysaccharides in the 700 characterisation of microbial communities in soil: a review. Biology and Fertility of Soils 701 29:111–129. 702 703 Forest Ecology and Management 422: 294-302 (2018) Graphical Abstract Differences in soil microbial community structure of declining (open symbols) and nondeclining (filled symbols) trees in three forest types subjected to contrasting climatic conditions (Silver fir, black symbols; Scots pine, green symbols; Aleppo pine, red symbols). The graph shows the Non-Metric multidimensional scaling (NMDS) biplot of soil microbial composition based on PLFA (Phospholipid-derived fatty acids). The correlation between soil nutrient composition and NMDS axes was projected in the ordination diagram. The polygons show the centroid of the distribution of the trees’ groups. The two trees with different vigour correspond to the Silver fir forest. Supporting Information for the manuscript “Beneath the canopy: linking droughtinduced forest die off and changes in soil properties” by Gazol et al. Supporting information captions Table S1. Soil texture in the three studied forests Figure S1. August 12-months long Standardised Precipitation–Evapotranspiration Index (SPEI). Figure S2. Soil, Carbon (C), Nitrogen (N) and C/N ratio values observed in the three studied forests. Table S1. Soil texture of the three studied forests. Site – Tree species Sand (%) Lime (%) Clay (%) Paco Ezpela - Silver fir (Abies alba) 45.58 41.20 13.22 Corbalán - Scots pine (Pinus sylvestris) 76.49 21.13 2.38 Peñaflor - Aleppo pine (Pinus halepensis) 71.18 23.76 5.05 Figure S1. August 12-months long Standardised Precipitation–Evapotranspiration Index (SPEI; Vicente-Serrano et al. 2010). The SPEI is a climatic proxy widely used for drought quantification (see Vicente-Serrano et al. 2010). Positive and negative values indicate wet and dry conditions, respectively. The red arrows indicate the SPEI values observed in 2012. Different colours and symbols are used for each site: Corbalán (P. sylvestris stand; green lines and squares); Paco Ezpela (A. alba stand; black lines and circles); Peñaflor (P. halepensis; red lines and triangles). Figure S2. Differences in soil carbon, nitrogen and C/N ratio across forest types and between declining (open boxes) and non-declining (solid boxes) trees. The C and N values represent the percentage of the soil elements in the sample. Highlights  Drought-induced forest die-off may impact soil nutrients and microbiota.  Die-off alters soil microbial community structure substantially.  Soil nutrient availability does not change as a function of tree vigor.  Soil water repellence amplifies drought stress.  The composition of soil microbiota is a monitor of forest die-off impacts.