This is the accepted version of the article: Liu, Daijun; Zhang, Chao; Ogaya, Romà; [et al.]. «Effects of decadal experimen- tal drought and climate extremes on vegetation growth in Mediterranean forests and shrublands». Journal of Vegetation Science, Vol. 31, issue 5 (Sep. 2020), p. 768-779. DOI 10.1111/jvs.12902 This version is avaible at https://ddd.uab.cat/record/232637 under the terms of the license 1 Effects of decadal experimental drought and climate extremes on vegetation growth in 2 Mediterranean forests and shrublands 3 4 Short running title: Vegetation growth to decadal drought 5 6 Daijun Liu1,2,3*, Chao Zhang1,2*, Romà Ogaya1,2, Marc Estiarte1,2, Josep Peñuelas1,2 7 1 8 CSIC, Global Ecology Unit CREAF-CSIC-UAB, Bellaterra (Catalonia) E-08193, Spain 2 9 CREAF, Cerdanyola del Vallès (Catalonia) E-08193, Spain, 3 10 School of Geography, Earth and Environmental Sciences, University of Birmingham, B15 11 2TT, United Kingdom 12 * 13 Equal contribution: 14 Daijun Liu, E-mail: d.liu@creaf.uab.es, ORCID: https://orcid.org/0000-0002-0993-0832 15 Chao Zhang, E-mail: c.zhang@creaf.uab.es, ORCID: https://orcid.org/0000-0001-7327-9477 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 1 35 Abstract 36 Increased drought combined with extreme episodes of heatwaves, are triggering severe 37 impacts on vegetation growth, particularly for plant communities in arid and semiarid 38 ecosystems. Although there is an abundance of short-term field drought experiments in 39 natural ecosystems, remaining knowledge gaps limit the understanding and prediction of 40 vegetation growth to ongoing and future climate scenarios. Here, we assessed the impacts of 41 long-term (1999-2016) experimental drought (ca. -30% rainfall) on the vegetation growth of 42 a Mediterranean high (H-) and low (L-) canopy forests and an early successional shrubland, 43 as indicated by aboveground biomass increment (ABI) and standing density, respectively. We 44 found habitat context (historical climate change impact, soil depth and successional status) of 45 the study sites significantly affected the magnitude of climate impacts; there were synergistic 46 effects of experimental drought and meteorological drought (Standardised Precipitation- 47 Evapotranspiration Index, SPEI) as well as extreme dry years on vegetation growth. Long- 48 term experimental drought decreased the ABI for the two forest canopy types and the 49 standing density for the shrubland. Water availabilities in winter-spring (SPEIs) were 50 positively correlated with the ABI and standing density. Moreover, the experimental drought 51 decreased the vegetation growth in extreme dry years for the shrubland. We propose that 52 future work not only study the vegetation dynamics with physiological, phenological and 53 demographical changes at long-term process and across climate gradients, but also should 54 explore the changes of simultaneous multiple functions (e.g. multifunctionality) under long- 55 term process and extremes. This type of analysis of long-term data is essential to understand 56 and predict biodiversity loss, composition shifts, declines in ecosystem function and carbon 57 budgets at temporal and spatial scales, to enable policymakers to design and implement 58 strategies for the maintenance of sustainable ecosystem function under future climate change 59 scenarios. 60 61 Key words: Mediterranean ecosystems; decadal drought experiment; vegetation growth; 62 aboveground biomass increment (ABI); standing density; habitat context; 63 64 65 66 67 68 2 69 Introduction 70 Drought, largely due to rapid warming and shifts in precipitation have increased water stress 71 across the terrestrial ecosystems, resulting in ecological consequences for the dynamics of 72 plant community structure and ecosystem functions (Walther et al. 2002; Bellard et al. 2012; 73 Peñuelas et al. 2013; Seddon et al. 2016). Mediterranean ecosystems are particularly 74 susceptible to declines in soil water availability and the more frequent combination of climate 75 extremes such as heatwaves and drought events during the summer (Myers et al. 2000; 76 Peñuelas et al. 2017; 2018; Cramer et al. 2018). Indeed, unprecedented impacts of water 77 deficit on vegetation have been widely documented, such as episodes of tree mortality and 78 forest die-off (Peñuelas et al. 2001; Lloret, Siscart, et al. 2004; Carnicer et al. 2011; Saura- 79 Mas et al. 2015), shifts in community composition (Peñuelas & Boada, 2003; Peñuelas et al. 80 2007; Ruiz-Benito et al. 2017) and declines in ecosystem carbon uptake and storage (Ciais et 81 al. 2005; Vicente-Serrano et al. 2013; Papadopoulos et al. 2017). By the end of this century, 82 water deficits in the Mediterranean Basin are expected to increase in severity and frequency 83 under the projected scenarios of 2-4 °C increases in temperature and the concurrent decreases 84 of up to 30% in precipitation (Cramer et al. 2018; Lionello & Scarascia, 2018). These 85 increases in frequency and intensity of climate extremes will further trigger forest 86 degradation and land desertification, especially within the overarching context of land-use 87 change, habitat fragmentation and biotic invasions (Doblas-Miranda et al. 2017; Peñuelas et 88 al. 2017; 2018; Cramer et al. 2018). Therefore, understanding and predicting vegetation 89 dynamics to future climate change is essential, especially if the impacts are to be effectively 90 mitigated through the design and implementation of ecosystem restoration in future. 91 The effective method to measure the vegetation response to climate regime is still limited. 92 Manipulating drought experiments (i.e. reduced rainfall) in natural ecosystems are considered 93 as one of the useful methods to track the climate-caused vegetation responses, such as 94 physiological changes, plant growth, community composition and ecosystem functioning 95 (Luo et al. 2011; Beier et al. 2012; Liu et al. 2015; 2017). These studies provide an essential 96 opportunity to explore the response of a local community to climatic change. However, the 97 result of plant responses and other ecological processes by drought regime is based on short- 98 term studies (less than 5 years). It is still unclear whether the direction and magnitude of the 99 effects of drought on vegetation will continue over time. Short-term experiments have shown 100 immediate and transient vegetation responses to simulated climate change, and the effects 101 could be accumulative and lag behind at long-term processes (Smith et al., 2009; Peñuelas et 102 al. 2013; 2018). A recent study has reported that drought effect on aboveground net primary 3 103 productivity (ANPP) of Mediterranean forest is increasing over 15 years under 27% rainfall 104 exclusion (Gavinet et al. 2019), which is not detected in the short-term. It is also likely that 105 plants rely on their plasticity to tolerate the shifts in climate regimes at short-term climate 106 regimes. It has been reported no net change in standing biomass and species diversity for the 107 shrublands across 9 years of reduced rainfall in Israel (Tielbörger et al. 2014; Metz & 108 Tielbörger, 2016). Moreover, the impacts of climate on plant growth can be decreased or 109 dampened over time (Leuzinger et al., 2011; Barbeta et al., 2013). A number of studies have 110 reported that plants can adapt to drought through physiological, phenological and 111 morphological adjustments, community recruitment (i.e. regrowth and new seedling) and 112 biotic interactions (i.e. increased mycorrhizae-plant associations) (Bellard et al. 2012; 113 Peñuelas et al. 2013; 2018; Liu, et al. 2017). Thus, gaining knowledge on longer temporal 114 scale of vegetation responses to climate change is urgently required. 115 Long-term (decadal) drought experiment is a promising method to evaluate vegetation 116 growth over time that can capture the results of plant adjustments and ecological interactions 117 (Luo et al., 2011; Liu et al., 2017, Liu, Ogaya, et al. 2018). The magnitude of long-term 118 drought effects could be especially responsive in such already water-deficit ecosystems. 119 Strong responses in plant species diversity and composition have been found in decadal 120 drought experiments for the shrublands in the Mediterranean ecosystems, whereas nearly no 121 responses have been observed in the shrublands of wet sites in UK (Clocaenog) and Denmark 122 (Brandbjerg) (Kröel-Dulay et al. 2015; Estiarte et al. 2016; Reinsch et al. 2017). Moreover, 123 unprecedent and severe climate extremes may increase the drought intensity for the 124 experimental plot, leading to severe negative impacts on carbon assimilation, water transport 125 and growth limitation beyond the bounds of plasticity and endurance of the Mediterranean 126 plants (Limousin et al., 2012; Liu, et al. 2016; 2017). Some studies have reported that 127 experimental drought and combined extreme droughts significantly affect flower and fruit 128 development, which may decrease the new recruitment (Peñuelas et al. 2004; Misson et al., 129 2011). In recent extreme dry years, higher stem mortality and crown defoliation (i.e. leaf and 130 branch litterfall) in a holm oak forests were reported in experimental drought plots than in 131 control plots (Barbeta et al. 2015; Liu et al., 2015; Liu, Ogaya, et al. 2018). Interestingly, the 132 effect of long-term drought may also relate to the climate, soil depth and vegetation 133 composition of the study site (Martin-Stpaul et al., 2013; Liu, Ogaya, et al. 2018). These 134 issues have not been addressed by long-term drought experiments, hindering an accurate 135 prediction of vegetation dynamics to climatic regime. 4 136 The aim of this study is to analyze plant biomass dynamics from two long-term (1999- 137 2016) drought experiments in the Mediterranean region, one in a holm oak forest (with a high 138 and low canopy, details in Method and Materials) and the other in an early successional 139 shrubland. These two experiments have provided evidence for the impacts of short-term 140 drought on plant physiological (carbon assimilation) processes, especially during summer 141 (Llorens et al. 2003; Prieto et al. 2009; Liu et al. 2016: Zhang et al., 2017), phenology 142 (advanced and delayed leaf and flower opening) (Peñuelas et al. 2004; Ogaya & Peñuelas, 143 2007a; Bernal et al. 2011) and genetic responses (increased genetic differentiation) (Jump et 144 al. 2008; Rico et al. 2014). We hypothesize that long-term experimental drought combined 145 with meteorological drought (Standardized Precipitation-Evapotranspiration Index, SPEI) 146 will significantly decrease the plant biomass accumulation for the Mediterranean forest 147 (especially in low canopy) and shrubland. The primary objective of our study is to analyze 148 the vegetation growth (aboveground biomass increment, ABI in the forest and standing 149 density in the shrubland) in response to the 18 years of experimental drought and explore the 150 responses with the effects of meteorological drought (SPEIs) and extreme drought episodes. 151 This study aims, thus to improve the understanding of the responses of plant growth to 152 ongoing and future climate change, and inform the design of effective strategies to maintain 153 ecosystem carbon sink capacity. 154 155 Method and Materials 156 Long-term drought experiments 157 1) Mediterranean forest 158 The forest site is located at a south-facing slope (ca. 25o) in Prades mountains in southern 159 Catalonia, northeastern Iberian Peninsula (41°21´N, 1°2´E; 930 m above sea level; Fig. 1C). 160 This forest is a typical Mediterranean holm oak forest, dominated by the evergreen tree 161 Quercus ilex, and usually accompanied with the shrubs Phillyrea latifolia and Arbutus unedo. 162 The forest has not been disturbed by anthropogenic thinning and copping since 1940s. The 163 vegetation structure and height are naturally variable due to the heterogeneity of nutrient 164 availability, soil depth and bedrock type (Liu, Ogaya, et al. 2018). The vegetation at deep soil 165 sites is usually composed by tall and big stands and dominated by Q. ilex individual 8-10 m 166 tall (High, indicated as H-canopy hereafter), whereas at the shallow soil sites Q. ilex is mixed 167 with several shrubs including P. latifolia reaching 4-6 m in height (Low, indicated as L- 168 canopy hereafter) (detailed feature for the two canopy types in Table 1) (Liu, Ogaya, et al. 169 2018). We simulated a drier climate by excluding 30% of annual precipitation (similar to the 5 170 predicted scenarios of precipitation reduction in coming decades) with a nonintrusive system 171 based on the installation of transparent plastic long strips (0.5-0.8 m wide) within drought 172 plots (Fig. 1A, B). Four plots (10 x 15 m per plot) were respectively established in both H- 173 and L-canopy, and half of them (two replicates per each type) were randomly selected for 174 drought treatment whereas the other two were assigned as control treatment. At the upper 175 edges of the drought plots, we built ditches (0.8-1 m deep) to intercept the water runoff into 176 the plots. Throughout the study period of 1999-2016, annual soil moisture in the H-canopy 177 and L-canopy was decreased by ca. 15% (see details in Table S1). 178 In 1998, all living trees (Diameter > 2 cm) were labelled at a height of 50 cm (D50) and 179 since then their diameter was measured in winter, a growth-less season. Aboveground 180 biomass (AB) at plot level was estimated from stem diameters by allometric relationships 181 (Ogaya & Peñuelas, 2007b; Liu et al. 2015). The aboveground biomass increment (ABI) was 182 calculated annually using the differences in AB between years for the two canopy types. This 183 growth metric is a good proxy for exploring the forest growth dynamic in response to 184 experimental drought and natural droughts. 185 A meteorological station installed at the study site automatically recorded daily 186 temperature and precipitation since the pre-treatment in 1998. Soil moisture at 0-15 cm depth 187 was measured seasonally throughout the experimental period by Time Domain Reflectometry 188 (TDR, Tektronix 1502C, Tektronix, Beaverton, OR, USA) using probes permanently 189 installed in the soil at four randomly selected locations in each plot. Monthly temperature and 190 precipitation for 1975-2001 were obtained from the nearby meteorological station of Poblet 191 Monastery (5.6 km away from the study plots and at 510 m a.s.l.) (linear regression for 1998- 192 2001: R2=0.97 for temperature and R2=0.75 for precipitation). Based on the data collected at 193 the site and at the nearby meteorological station, we calculated the Standardized 194 Precipitation-Evapotranspiration Index (SPEI), a drought index that takes into account both 195 precipitation and temperature fluctuations, and identifies the drought severity according to the 196 intensity and duration (Vicente-Serrano et al. 2013). Thus, a database of monthly difference 197 between precipitation and the potential evapotranspiration by temperature changes (D values) 198 was constructed from 1975-2016, and then applied to calculated the SPEI with different 199 timescales (1-36 months) by SPEI package (version 1.6). The SPEI index is widely used to 200 explore the relationships between plant responses and water deficits at different temporal 201 scales. The range of this drought index is from -3 to 3, lower values represent more drought 202 stress. SPEI with different timescales stand for the levels of drought-sensitivity, higher 203 drought-sensitivity usually at shorter temporal scale while more drought-resistance at longer 6 204 temporal scale (a detailed calculation was described in Barbeta et al. (2013) and Liu et al. 205 (2015, 2017)). We used SPEI to evaluate the drought severity from 1975 to 2016. At the 206 forest site, mean annual SPEIs and August SPEI-3 decreased (R2 = 0.09, p < 0.1 and R2 = 207 0.26, p < 0.001; respectively) during the period of 1975-2016 (Fig. 2A, C). Moreover, we 208 defined the extreme dry years based on the lowest 10 percentile mean annual precipitation 209 (MAP) (Knapp et al. 2015) during the period of 1975-2016 and we identified 2001, 2005-06 210 and 2015 as extreme dry years during the study period 1999-2016. 211 2) Mediterranean shrubland 212 The shrubland site is located at a south-facing slope (ca. 13o) hill of Garraf Natural Park in 213 central coastal Catalonia (northeastern Iberian Peninsula) (41º18´N, 1º49´E; 210 m above sea 214 level; Fig. 1C). The soil in the study site is typically calcareous, and the substrate is mainly 215 composed of marls and limestone with rocky outcrops. The dominant vegetation changed to 216 short perennial shrubs after the two wild fires in 1982 and 1994 eliminated coniferous forest. 217 The main plant species are the shrubs Globularia alypum and Erica multiflora whose 218 composition is higher than 50% of the total community and are accompanied by other 219 Mediterranean shrubs such as Rosmarinus officinalis, Dorycnium pentaphyllum and Ulex 220 parviflorus and herbs such as Euphorbia serrata (see details in Liu et al. 2017). To simulate 221 future drought regimes, we excluded precipitation with a nonintrusive system that 222 automatically extends a transparent waterproof cover to intercept precipitation during spring 223 and autumn growing season (May-September) during 1999-2015 (Peñuelas et al. 2007; Liu et 224 al. 2016; 2017) and transparent V-type plastic to intercept the rainfall since the damage of 225 automatic system after 2015. Six plots (4 x 5 m) were built along the same slope, three plots 226 randomly selected as drought treatment while the rest were used as control (Fig. 1D, E; Table 227 1). Throughout the study period of 1999-2016, the drought treatment reduced annually the 228 precipitation around 30%, and the soil moisture at 0-15 cm depth around a 15% compared to 229 control. 230 From the pre-treatment in 1998 to 2016, the vegetation growth in the shrubland was 231 measured once per year in end July, when growth is stopped, using the point-intercept 232 method. Vegetation measurements were conducted at the central area (3 x 4 m) of each plot 233 using five parallel 3-m long transects permanently fixed at the beginning of the study. Along 234 each transect 61 points were evenly distributed totaling 305 points per plot. We lowered a 235 long steel pin (3 mm diameter) through each sampling point and every time the pin hit a plant 236 we recorded plant species, organs (e.g. leaf, reproductive structure or stem) and state (live or 237 dead). The growth metric is calculated as the stand biomass (or density) for this study site. 7 238 Daily air temperature and precipitation during the study period were recorded by a 239 meteorological station at the study site. Since the start of this experiment in 1998, soil 240 moisture at of 0-15 cm depth was measured biweekly using Time Domain Reflectometry 241 (TDR) by probes permanently inserted at three points per plot. Historical climate data since 242 1975 was collected from a nearby meteorological station (10 km away from the study plots 243 and at 200 m a.s.l.) to get the background climate regimes for the study site (Liu et al. 2017). 244 The SPEI drought index was calculated based on the site climate data from 1998-2016 and 245 historical climate (linear regression with our station for 1998-2010: R2=0.99 for temperature 246 and R2=0.83 for precipitation). The calculation of SPEI in the shrubland site is similar with 247 the forest site described above. No trends in mean annual SPEIs was detected in the 248 shrubland, but August SPEI-3 decreased significantly (R2 = 0.09, p < 0. 05) across the period 249 of 1975-2016 (Fig. 2B, D). As described for the forest site, in the shrubland we identified 250 2001, 2005-06 and 2016 as extreme dry years based on the lowest 10 percentile MAP (Knapp 251 et al. 2015) during the period 1975-2016. 252 Statistical analysis 253 To analyze the vegetation responses to drought throughout the period from the pre- 254 treatment year 1998 to 2016, vegetation growth of ABI for the forest and the standing density 255 for the shrubland were tested respectively. In the forest we analyzed separately the two 256 canopy types. In the shrubland, first we log-transformed the abundance or frequency for each 257 species, and then the values of all the species were summed and given equal weight for all 258 species at community level. This method reduces the overrepresentation of highly dominating 259 species such as G. alypum and E. multiflora (Šmilauer & Lepš, 2014; Liu, Peñuelas, et al. 260 2018). The vegetation growth for forests and shrubland were taken as the responsive 261 variables to the drought treatment combined with natural droughts (SPEIs). The SPEIs were 262 calculated for different timescales ranging from 3 to 12 months, and the growing season was 263 considered during the rainfall accumulative period from December to June in Mediterranean 264 region (Barbeta et al. 2013; Liu et al. 2015; 2017). Then we applied these different SPEIs to 265 test the correlations with plant growth for forest and shrubland. The best-fitted SPEIs with the 266 vegetation growth for forest (separately for high and low canopies) and shrubland were 267 selected based on the stronger correlations (the details were described in the Appendix). We 268 used a linear mixed model for growth responses (Growth=Drought*SPEI) at both forest and 269 shrubland sites according to the lowest Akaike information criterion (AIC). The plot nested 270 with year (plot+year) was selected as random factor, to explain the spatial and temporal 271 autocorrelation. Moreover, significant differences for the slopes between control and drought 8 272 for the vegetation growth were statistically tested by least-Squares Means (lsmeans package). 273 The vegetation responses of control and drought at extreme dry years were assessed by 274 repeated measures ANOVA. All the analyses were performed with the R program (R version 275 3.5.0). 276 277 Results 278 Vegetation responses to experimental drought 279 Long-term experimental drought decreased either aboveground biomass increment (ABI) in 280 both the Mediterranean forests or standing density in the shrubland. Experimental drought 281 decreased the ABI values for both H-canopy and L-canopy (t-value = -3.52, p < 0.001 and t- 282 value = -2.41, p < 0.05; respectively) (Fig. 3A, B). The effects of experimental drought on 283 ABI were stronger for the first 5 years (1999-2004), especially for L-canopy. Experimental 284 drought decreased the standing density for the whole 1999 to 2016 period in the early- 285 successional shrubland (t-value = -7.71, p < 0.001), but the effect tended to be more stable 286 after the year 2002 (Fig. 3C). 287 The changes of vegetation growth to natural drought (SPEIs) 288 The values of vegetation growth were also correlated positively with the natural 289 meteorological droughts (SPEI) during spring-winter season (Fig. 4 and Table 2). In the 290 forest, ABIs increased with May SPEI-7 for the H-canopy (t-value = 2.77, p < 0.05) and with 291 May SPEI-5 in the L-canopy (t-value = 2.41, p < 0.05) (Fig. 4A, B). Interestingly, we found 292 there was a significant interaction between May SPEI-5 and experimental drought treatment 293 for the ABI in the L-canopy according to the best fitted model (t-value = 2.07, p < 0.05) (Fig. 294 4B). In the H-canopy, the slope of the ABI versus May SPEI-7 correlation did not differ 295 among control and drought plots (common slope = 1.20, ns) (Fig. 4A). In the L-canopy, ABI 296 correlation with May SPEI-5 was stronger in drought (R2 = 0.51, p < 0.001) than control plots 297 (R2=0.32, p < 0.01), and the slopes were marginally steeper for drought than control plots 298 (common slope = 1.69, p < 0.1) (Fig. 4B). 299 In the shrubland, the standing density was positively correlated with May SPEI-4 during the 300 study period (t-value = 3.53, p < 0.001). The positive correlation was stronger in control (R2= 301 0.34, p < 0.01) rather than in drought plots (R2=0.15, p<0.1). The slopes of the density versus 302 May SPEI-4 correlation differed among control and drought (common slope = 2.29, p < 0.05) 303 (Fig. 4C). 304 The severe effect of extreme dry episodes 305 9 306 Vegetation growth to extreme dry years were strong and apparent both in the forest and the 307 shrubland. In the forest, the ABI decreased more in drought than in control plots in extreme 308 dry episodes in both H- and L-canopy types (Fig. 5A, B). In the shrubland, there was 309 significant lower biomass accumulation for the drought at extreme dry years (Fig. 5C, 310 ANOVA, F = 11.02, p < 0.01). 311 312 Discussion 313 Decreases in vegetation growth in response to the long-term drought experiments and the 314 progressive meteorological drought 315 The declines in vegetation growth were accelerated by natural extreme droughts. ABI 316 decreased in response to the experimental drought in the two canopy types (Fig. 3A, B) as 317 reported already in previous studies (Barbeta et al. 2015; Liu et al. 2015; Liu, Ogaya, et al. 318 2018). However, the decrease in ABI appeared earlier and stronger in L-canopy due to 319 shallowest soil site. Indeed, soil depth is a determinant of soil water holding capacity and is 320 likely a modulator of the intensity of plant responses to climate change. Continuous drought 321 reduces the ground water levels limiting the uptake of ground water that is vital for the 322 maintenance of water balance in the hot and dry Mediterranean summer, which leads to 323 hydraulic failure (Barbeta et al. 2015; Liu et al. 2015; Liu, Ogaya, et al. 2018). For this 324 reason, tree mortality and forest die-back have been observed in disturbed Mediterranean 325 forests with shallow soils (Lloret, Siscart, et al. 2004; Carnicer et al. 2011; Saura-Mas et al. 326 2015). In the early-successional shrubland, we also found negative effects of experimental 327 drought on standing density over time (Fig. 3C). Nowadays, the anthropogenic changes in 328 land-use changes such as habitat fragmentation have affected most vegetation ecosystems 329 which may increase ecosystem instability and push back to early-successional status (Seddon 330 et al. 2016; Peñuelas et al. 2013; 2018). Rapid climate regimes could also trigger sharp and 331 strong effects on species diversity and composition shifts (Peñuelas et al. 2007; Kröel-Dulay 332 et al. 2015; Liu et al. 2017). Thus, habitat context may be a key driver of vegetation 333 responses to climate change, so we recommend that future climate experiments account for 334 the factors, such as soil depth and successional status, to improve the understanding and 335 prediction of vegetation dynamic under continuous climate change. 336 Plant growth was closely correlated with SPEIs at growing season. Generally, ABI of 337 forest and standing density of shrubland were positively related with water availability in the 338 winter-spring seasons (Table 2), which is consistent with the findings that rainfall in winter- 339 spring is the most essential for plant growth and establishment in Mediterranean ecosystems 10 340 (Barbeta et al. 2013; 2015; Liu et al. 2017; Gavinet et al. 2019). The SPEI timescales for 341 forests varied differentially, with longer timescales in H-canopy (Fig. 3, 4). The result 342 indicates higher stem water storage in H-canopy forest than L-canopy in response to water 343 limitation (Vicente-Serrano et al. 2013; Barbeta et al. 2013; Liu et al. 2015; Liu, Ogaya, et al. 344 2018). In addition, H-canopy can successfully access deep soil water and groundwater with 345 longer root depth, allowing it to buffer shorter drought events than L-canopy (Barbeta et al. 346 2015; Liu, Ogaya, et al. 2018). On the contrary, L-canopy with shallow soil likely have 347 shallow roots and shrubland stores less water in stems, may suffer more severe and negative 348 impacts by meteorological droughts in future. Indeed, the mean annual SPEIs and summer 349 SPEIs of August SPEI-3 also suggest that water deficit is becoming more severe and stronger 350 over the latest four decades (Fig. 2), which may increase soil water deficit and cause episodes 351 of tree mortality and forest canopy die-off (Carnicer et al., 2011; Saura-Mas et al. 2015). 352 Furthermore, the trend of summer water deficit is continuing, which will potentially decrease 353 the ecosystem carbon sink and drought recovery in the most immediate future. 354 The correlations with the SPEIs varied greatly between drought treatment and control. For 355 L-canopy, the sensitivity was higher in the drought plot, i.e. the slope of the ABI and SPEI 356 was steeper in drought than that of control, which could be related with the limitation on stem 357 growth and higher mortality rate in the experimental drought (Barbeta et al. 2015; Liu et al. 358 2015; Liu, Ogaya, et al. 2018). For shrubland, there was a strong (steeper slope) and 359 significant relationship between SPEI and standing biomass in control than that of drought 360 plots, indicating the plants in control was more responsive to meteorological drought. 361 However, we observed that the correlations to meteorological drought were similar between 362 control and drought plots for H-canopy, suggesting a consistent sensitivity to natural drought 363 (Fig 3A). Therefore, a new climate status simulated by experimental drought (30% rainfall 364 exclusion) had negative impacts on L-canopy and shrubland, which may also result in 365 different species assembles and composition compared to control plots. 366 Strong decreases in vegetation growth were found in drought plots at extreme dry years. 367 Throughout the extreme dry years, the average ABI in forests and standing biomass of 368 shrubland were more reduced in the drought treatments than in the control. The results might 369 be related to the limitations on leaf carbon sequestration, stem/branch hydraulic failure and 370 higher mortality rates in drought plots than in control plots for the holm oak forests (Barbeta 371 et al. 2015; Liu et al. 2015; Liu, Ogaya, et al. 2018). These frequent and intense extreme 372 episodes might significantly change forest productivity over time due to the decreases in 373 plant growth, higher stem mortality and composition shifts (Allen et al. 2015; Martinez- 11 374 Vilalta et al. 2017). Indeed, a continental-scale reduction in net primary productivity across a 375 large area of Europe was caused by extreme water deficit in summer 2003 (Ciais et al. 2005). 376 Significant differences in standing density between control and drought plots were also found 377 in the extreme dry years in the shrubland (Fig. 5C). Therefore, such plant community in the 378 arid and semi-arid ecosystems may be more sensitive to extreme droughts than more mesic 379 ecosystems in the near future, thus presenting more changes in plant density, community 380 diversity and evenness. 381 382 Contrasting results in other long-term drought experiments in Mediterranean ecosystems 383 Plant responses to long-term drought varied among experiments cited in the literature. The 384 high risk for limitation of carbon uptake could come from lower photosynthetic activities 385 and/or increase higher respiration rates (Wu et al. 2011; Beier et al. 2012; Reinsch et al. 386 2017). However, it has been shown that decadal drought did not affect species richness and 387 aboveground biomass in Mediterranean shrublands in Israel, possibly as a result of ecosystem 388 resistance with stable species diversity and composition (Tielbörger et al. 2014; Metz & 389 Tielbörger 2016). An experimental drought that started in 2004 in another Mediterranean 390 holm oak forest (Puébachon State Forest, South France) had no effect on stem growth 391 (Martin-Stpaul et al. 2013), but strongly decreased stem sap flow, leaf and acorn production 392 (Limousin et al. 2009; 2012; Gavinet et al. 2019). Recently, it was reported that long-term 393 (15 years) drought have decreased 10% ANPP possibly owing to the lower leaf and acorn 394 production over time (Gavinet et al. 2019). The varying drought effect on stem growth was 395 possibly related to the site differences between Prades and Puébachon, such as in climate of 396 (650 mm vs 936 mm), stem number (stem ha-1:13273 in H-canopy and 19005 in L-canopy vs 397 5933) and basal area (m2 ha-1: 47.1 in H-canopy and 35.2 in L-canopy vs 24.9). From the 398 above, we can infer our study sites could be more sensitive to drought due to more severe 399 water limitation and stronger competition. However, longer experimental drought (15 years) 400 significantly decreased forest ANPP possibly owing to the lower stem sap flow, leaf and 401 acorn production (Limousin et al. 2009; Martin-Stpaul et al. 2013; Gavinet et al. 2019). Thus, 402 the contrasting responses of drought effect on vegetation across study sites are probably 403 linked to site-specific climate, vegetation structure and composition. Therefore, future studies 404 should focus not only on the physiological, phenological and morphological responses to 405 long-term process, but also on the different context of the study sites in comparison. 406 Implication for the future experiments 12 407 Our results have shown that long-term drought (30% rainfall reduction for 18 years), 408 combined with meteorological drought and extreme dry years are decreasing vegetation 409 growth in the water-limited Mediterranean forest and shrubland ecosystems. Such 410 experiments provide an opportunity to estimate ecosystem function dynamic under future 411 climate change and also a validation of the predictions by dynamic global vegetation 412 modelling (DGVM). However, knowledge gaps still remain that hinder our understanding of 413 future carbon storage in natural ecosystems in a changing climate. Firstly, the magnitude of 414 plant responses to drought differs for temporal processes (short-term vs long-term), and are 415 strongly associated with extremes. Secondly, the response of vegetation growth to climate 416 change may vary greatly with study sites depending on soil depth, water availability and 417 community composition as well as successional status. Since most of previous studies rarely 418 tested these responses, understanding of the dynamics at regional and continental scales 419 remains limited. Thirdly, plant growth to climate change is correlated with multiple 420 physiological, phenological and demographical adjustments that should also be taken into 421 consideration. Fourthly, the responses of vegetation dynamics are complex and 422 multidimensional. For example, drought regimes may decrease plant aboveground biomass 423 and productivity, but increase biomass allocation to belowground plant organs (e.g. root 424 mass), increase fungal and/or bacterial activities. Future studies should also probe the 425 multiple ecological functions (e.g. multifunctionality), such as vegetation growth, seed 426 production, nutrient use and decomposition by the combination of long-term process and/or 427 extremes. Conducting long-term field experiments in contrasting natural ecosystems under 428 different climate regimes and exploring multivariate data analysis of interactions and trade- 429 offs is thus necessary. 430 431 Acknowledgements 432 We thank the people who managed and maintained the long-term experiments and collected 433 data. Funding for this research was provided by the Spanish government project CGL2016- 434 79835-P, the Catalan government project SGR2017-1005, the European Research Council 435 Synergy grant ERC-2013-SyG-610028 IMBALANCE-P and the European Research Council 436 (ERC) under the European Union’s Horizon 2020 research and innovation program (grant 437 agreement No 758873, TreeMort). 438 439 13 440 Author contributions 441 D.L and C.Z analysed the data and wrote the paper and J.P conceived the idea and provided 442 in-depth suggestions for the manuscript. J.P, M.E and R.O designed the experiment, 443 performed the research and revised the manuscript. 444 445 References 446 Allen, C.D., Macalady, A.K., Chenchouni, H., Bachelet, D., McDowell, N., Vennetier, M., … 447 Gonzalez, P., 2010. A global overview of drought and heat-induced tree mortality 448 reveals emerging climate change risks for forests. Forest Ecology and Management, 259: 449 660–684. 450 Barbeta, A., Mejía-Chang, M., Ogaya, R., Voltas, J., Dawson, T.E., & Peñuelas, J. 2015. The 451 combined effects of a long-term experimental drought and an extreme drought on the 452 use of plant-water sources in a Mediterranean forest. Global Change Biology 21: 1213– 453 1225. 454 Barbeta, A., Ogaya, R., & Peñuelas, J. 2013. Dampening effects of long-term experimental 455 drought on growth and mortality rates of a Holm oak forest. Global Change Biology 19: 456 3133–3144. 457 Beier, C., Beierkuhnlein, C., Wohlgemuth, T., Peñuuelas, J., Emmett, B., Korner, C., de … 458 Hansen, K. 2012. Precipitation manipulation experiments-chanllenges and 459 recommendations for the future. Ecology Letters 15: 899–911. 460 Bellard, C., Bertelsmeier, C., Leadley, P., Thuiller, W., & Courchamp, F. 2012. Impacts of 461 climate change on the future of biodiversity. Ecology Letters 15: 365–377. 462 Bernal, M., Estiarte, M., & Peñuelas, J. 2011. Drought advances spring growth phenology of 463 the Mediterranean shrub Erica multiflora. Plant Biology 13: 252–257. 464 del Cacho, M., Peñuelas, J., & Lloret, F. 2013. Reproductive output in Mediterranean shrubs 465 under climate change experimentally induced by drought and warming. Perspectives in 466 Plant Ecology, Evolution and Systematics 15: 319–327. 467 Carnicer, J., Coll, M., Ninyerola, M., Pons, X., Sanchez, G., & Peñuelas, J. 2011. Widespread 468 crown condition decline, food web disruption, and amplified tree mortality with 469 increased climate change-type drought. Proceedings of the National Academy of 470 Sciences 108: 1474–1478. 471 Ciais, P., Reichstein, M., Viovy, N., Granier, A., Ogée, J., Allard, V., … Valentini, R. 2005. 472 Europe-wide reduction in primary productivity caused by the heat and drought in 2003. 14 473 Nature 437: 529–533. 474 Cramer, W., Guiot, J., Fader, M., Garrabou, J., Gattuso, J.P., Iglesias, A., … Xoplaki, E. 2018. 475 Climate change and interconnected risks to sustainable development in the 476 Mediterranean. Nature Climate Change 8: 972–980. 477 Doblas-Miranda, E., Alonso, R., Arnan, X., Bermejo, V., Brotons, L., de las Heras, J., … 478 Retana, J. 2017. A review of the combination among global change factors in forests, 479 shrublands and pastures of the Mediterranean Region: Beyond drought effects. Global 480 and Planetary Change 148: 42–54. 481 Estiarte, M., Vicca, S., Peñuelas, J., Bahn, M., Beier, C., Emmett, B.A., … Janssens, I.A. 482 2016. Few multiyear precipitation-reduction experiments find a shift in the productivity- 483 precipitation relationship. Global Change Biology 22: 2570–2581. 484 Gavinet, J., Ourcival, J.M., & Limousin, J.M. 2019 Rainfall exclusion and thinning can alter 485 the relationships between forest functioning and drought. New Phytologist. 2019 Apr 21. 486 Jump, A.S. & Penuelas, J. 2005. Running to stand still: adaptation and the response of plants 487 to rapid climate change. Ecology letters 89: 1010-1020. 488 Jump, A., Penuelas, J., Rico, L., Ramallo, E., Estiarte, M., Martnez-Izquierdo, J.A., & Lloret, 489 F. 2008. Simulated climate change provokes rapid genetic change in the Mediterranean 490 shrub Fumana thymifolia. Global Change Biology 14: 637–643. 491 Van Kleunen, M., Dawson, W., Essl, F., Pergl, J., Winter, M., Weber, E., … Pyšek, P. 2015. 492 Global exchange and accumulation of non-native plants. Nature 525: 100–103. 493 Knapp, A.K., Hoover, D.L., Wilcox, K.R., Avolio, M.L., Koerner, S.E., La Pierre, K.J., … 494 Smith, M.D., 2015. Characterizing differences in precipitation regimes of extreme wet 495 and dry years: implications for climate change experiments. Global Change Biology 21: 496 2624-2633. 497 Kröel-Dulay, G., Ransijn, J., Schmidt, I.K., Beier, C., De Angelis, P., de Dato, G., … 498 Penuelas, J. 2015. Increased sensitivity to climate change in disturbed ecosystems. 499 Nature Communications 6: 1–7. 500 LaForgia, M.L., Spasojevic, M.J., Case, E.J., Latimer, A.M., & Harrison, S.P. 2018. Seed 501 banks of native forbs, but not exotic grasses, increase during extreme drought. Ecology 502 99: 896–903. 503 Leuzinger, S., Luo, Y., Beier, C., Dieleman, W., Vicca, S., & Ko, C. 2011. Do global change 504 experiments overestimate impacts on terrestrial ecosystems ? Trends in ecology & 505 evolution 26: 2–7. 506 Limousin, J.M., Rambal, S., Ourcival, J.M., Rocheteau, A., Joffre, R., & Rodriguez-Cortina, 15 507 R. 2009. Long-term transpiration change with rainfall decline in a Mediterranean 508 Quercus ilex forest. Global Change Biology 15: 2163–2175. 509 Limousin, J.-M., Rambal, S., Ourcival, J.-M., Rodríguez-Calcerrada, J., Pérez-Ramos, I.M., 510 Rodríguez-Cortina, … Joffre, R. 2012. Morphological and phenological shoot plasticity 511 in a Mediterranean evergreen oak facing long-term increased drought. Oecologia 169: 512 565–577. 513 Lionello, P., & Scarascia, L. 2018. The relation between climate change in the Mediterranean 514 region and global warming. Regional Environmental Change 18: 1481–1493. 515 Liu, D., Estiarte, M., Ogaya, R., Yang, X., & Peñuelas, J. 2017. Shift in community structure 516 in an early-successional Mediterranean shrubland driven by long-term experimental 517 warming and drought and natural extreme droughts. Global Change Biology 23: 4267- 518 4279. 519 Liu, D., Llusia, J., Ogaya, R., Estiarte, M., Llorens, L., Yang, X., & Peñuelas, J. 2016. 520 Physiological adjustments of a Mediterranean shrub to long-term experimental warming 521 and drought treatments. Plant Science 252: 53–61. 522 Liu, D., Ogaya, R., Barbeta, A., Yang, X., & Peñuelas, J. 2015. Contrasting impacts of 523 continuous moderate drought and episodic severe droughts on the aboveground-biomass 524 increment and litterfall of three coexisting Mediterranean woody species. Global 525 Change Biology 21: 4196–4209. 526 Liu, D., Ogaya, R., Barbeta, A., Yang, X., & Peñuelas, J. 2018. Long-term experimental 527 drought combined with natural extremes accelerate vegetation shift in a Mediterranean 528 holm oak forest. Environmental and Experimental Botany 151: 1-11. 529 Liu, D., Peñuelas, J., Ogaya, R., Estiarte, M., Tielbörger, K., Slowik, F., Yang, X., & Bilton, 530 M.C. 2018. Species selection under long-term experimental warming and drought 531 explained by climatic distributions. New Phytologist 217: 1494–1506. 532 Llorens, L., Penuelas, J., & Estiarte, M. 2003. Ecophysiological responses of two 533 Mediterranean shrubs, Erica multiflora and Globularia alypum, to experimentally drier 534 and warmer conditions. Physiologia Plantarum 119: 231–243. 535 Lloret, F., Peñuelas, J., & Estiarte, M. 2004. Experimental evidence of reduced diversity of 536 seedlings due to climate modification in a Mediterranean-type community. Global 537 Change Biology 10: 248–258. 538 Lloret, F., Peñuelas, J., Prieto, P., Llorens, L., & Estiarte, M. 2009. Plant community changes 539 induced by experimental climate change: Seedling and adult species composition. 540 Perspectives in Plant Ecology, Evolution and Systematics 11: 53–63. 16 541 Lloret, F., Siscart, D., & Dalmases, C. 2004. Canopy recovery after drought dieback in holm- 542 oak Mediterranean forests of Catalonia (NE Spain). Global Change Biology 10: 2092– 543 2099. 544 Luo, Y., Melillo, J., Niu, S., Beier, C., Clarks, J.S., & Classen, A.T. 2011. Coordinated 545 approaches to quantify long-term ecosystem dynamics in response to global change. 546 Global Change Biology 17: 843–854. 547 Martin-Stpaul, N.K., Limousin, J.-M., Vogt-Schilb, H., Rodríguez-Calcerrada, J., Rambal, S., 548 Longepierre, D., & Misson, L. 2013. The temporal response to drought in a 549 Mediterranean evergreen tree: comparing a regional precipitation gradient and a 550 throughfall exclusion experiment. Global Change Biology 19: 2413–2426. 551 Metz, J., & Tielbörger, K. 2016. Spatial and temporal aridity gradients provide poor proxies 552 for plant-plant interactions under climate change: a large-scale experiment. Functional 553 Ecology 30: 20–29. 554 Misson, L., Degueldre, D., Collin, C., Rodriguez, R., Rocheteau, A., Ourcival, J.M. and 555 Rambal, S., 2011. Phenological responses to extreme droughts in a Mediterranean forest. 556 Global Change Biology 17: 1036–1048. 557 Myers, N., Mittermeier, R.A., Mittermeier, C.G., da Fronseca, G.A.B., & Kent, J. 2000. 558 Biodiversity hotspots for conservation priorities. Nature 403: 853–858. 559 Ogaya, R., & Peñuelas, J. 2007a. Species-specific drought effects on flower and fruit 560 production in a Mediterranean holm oak forest. Forestry 180: 351–357. 561 Ogaya, R., & Peñuelas, J. 2007b. Tree growth, mortality, and above-ground biomass 562 accumulation in a holm oak forest under a five-year experimental field drought. Plant 563 Ecology, 189: 291–299. 564 Papadopoulos, A., Carreira, J.A., Hevia, A., Camarero, J.J., Pasho, E., Andreu-Hayles, L., … 565 Sánchez-Salguero, R. 2017. Climate extremes and predicted warming threaten 566 Mediterranean Holocene firs forests refugia. Proceedings of the National Academy of 567 Sciences 114: E10142–E10150. 568 Peñuelas, J., & Boada, M. 2003. A global change-induced biome shift in the Montseny 569 mountains (NE Spain). Global Change Biology 9: 131–140. 570 Peñuelas, J., Filella, I., Zhang, X., Llorens, L., Ogaya, R., Lloret, F., Comas, P., Estiarte, M., 571 & Terradas, J. 2004. Complex spatiotemporal phenological shifts as a response to 572 rainfall changes. New Phytologist 161: 837–846. 573 Peñuelas, J., Lloret, F., & Montoya, R. 2001. Severe drought effects on mediterranean woody 574 flora in Spain. Forest Science 47: 214–218. 17 575 Peñuelas, J., Prieto, P., Beier, C., Cesaraccio, C., de Angelis, P., de Dato, G., … Tietema, A. 576 2007. Response of plant species richness and primary productivity in shrublands along a 577 north-south gradient in Europe to seven years of experimental warming and drought: 578 Reductions in primary productivity in the heat and drought year of 2003. Global Change 579 Biology 13: 2563–2581. 580 Peñuelas, J., Sardans, J., Estiarte, M., Ogaya, R., Carnicer, J., Coll, M., … Jump, A.S. 2013. 581 Evidence of current impact of climate change on life: A walk from genes to the 582 biosphere. Global Change Biology 19: 2303–2338. 583 Peñuelas, J., Ciais, P., Canadell, J.G., Janssens, I.A., Fernández-Martínez, M., Carnicer, J., … 584 Sardans, J. 2017. Shifting from a fertilization-dominated to a warming-dominated period. 585 Nature Ecology and Evolution 1: 1438–1445. 586 Peñuelas, J., Sardans, J., Filella, I., Estiarte, M., Llusià, J., Ogaya, R., … Peguero, G., 2018. 587 Assessment of the impacts of climate change on Mediterranean terrestrial ecosystems 588 based on data from field experiments and long-term monitored field gradients in 589 Catalonia. Environmental and Experimental Botany, 152:49-59. 590 Prieto, P., Peñuelas, J., Llusià, J., Asensio, D., & Estiarte, M. 2009. Effects of long-term 591 experimental night-time warming and drought on photosynthesis, Fv/Fm and stomatal 592 conductance in the dominant species of a Mediterranean shrubland. Acta Physiologiae 593 Plantarum 31: 729–739. 594 Reinsch, S., Koller, E., Sowerby, A., de Dato, G., Estiarte, M., Kovács-Láng, E., … Emmett, 595 B.A. 2017. Shrubland primary production and soil respiration diverge along European 596 climate gradient. Scientific Reports 7: 1–7. 597 Rico, L., Ogaya, R., & Barbeta, A. & Peñuelas, J. 2014. Changes in DNA methylation 598 fingerprint of Quercus ilex trees in response to experimental field drought simulating 599 projected climate change. Plant Biology 16: 419–427. 600 Ruiz-Benito, P., Ratcliffe, S., Zavala, M.A., Martínez-Vilalta, J., Vilà-Cabrera, A., Lloret, 601 F., … Jump, A.S. 2017. Climate- and successional-related changes in functional 602 composition of European forests are strongly driven by tree mortality. Global Change 603 Biology 23: 4162–4176. 604 Saura-Mas, S., Bonas, A., & Lloret, F. 2015. Plant community response to drought-induced 605 canopy defoliation in a Mediterranean Quercus ilex forest. European Journal of Forest 606 Research 134: 261–272. 607 Seddon, A.W.R., Macias-Fauria, M., Long, P.R., Benz, D., & Willis, K.J. 2016. Sensitivity of 608 global terrestrial ecosystems to climate variability. Nature 531: 229–232. 18 609 Šmilauer, P., & Lepš, J. (2014). Multivariate analysis of ecological data using CANOCO 5. 610 Cambridge university press. 611 Smith, M.D., Knapp, A.K., & Collins, S.L. 2009. A framework for assessing ecosystem 612 dynamics in response to chronic resource alterations induced by global change. Ecology 613 79: 3–24. 614 Tielbörger, K., Bilton, M.C., Metz, J., Kigel, J., Holzapfel, C., Lebrija-Trejos, E., … 615 Sternberg, M. 2014. Middle-Eastern plant communities tolerate 9 years of drought in a 616 multi-site climate manipulation experiment. Nature Communications 5: 5102 617 Vicente-Serrano, S.M., Gouveia, C., Camarero, J.J., Beguería, S., Trigo, R., López-Moreno, 618 J.I., … Morán-Tejeda, E. 2013. Response of vegetation to drought time-scales across 619 global land biomes. Proceedings of the National Academy of Sciences 110: 52–57. 620 Walther, G., Post, E., Convey, P., Menzel, A., Parmesan, C., Beebee, T.J.C., … Bairlein, F. 621 2002. Ecological responses to recent climate change. Nature 416: 389-395 622 Wu, Z., Dijkstra, P., Koch, G.W., Peñuelas, J., & Hungate, B.A. 2011. Responses of 623 terrestrial ecosystems to temperature and precipitation change: A meta-analysis of 624 experimental manipulation. Global Change Biology 17: 927–942. 625 Zhang, C., Filella, I., Liu, D., Ogaya, R., Llusià, J., Asensio, D., & Peñuelas, J. (2017). 626 Photochemical reflectance index (PRI) for detecting responses of diurnal and seasonal 627 photosynthetic activity to experimental drought and warming in a mediterranean 628 shrubland. Remote Sensing, 9(11), 1189. 629 630 631 632 633 634 635 636 637 638 639 640 641 642 19 643 Tables Table 1. Characteristics of forests (H- and L- canopies) and shrubland sites. Soil depth and mean tree height represent the range of minimum and maximum and the others show the mean and standard error values. Aboveground biomass for forest and shrubland were measured at the pre-treatment year of 1998 (n=4 and n=6; respectively). Characteristic High canopy Low canopy Shrubland Soil depth (cm) 30-50 10-30 5-30 Mean height (m) 8-10 4-6 0.5-1.5 Mean stems at pre-treatment (stem ha-1) 13273±17.5 19005±29.8 - Mean basal area and aboveground biomass (m2/Ha and g/m2, respectively) 47.1±3.1 35.5±7.9 36.6±2.4 Soil moisture in control plots for 1999- 2016 (%) 17.19±0.58 16.32±0.6 19.02±0.8 Soil moisture in drought plots for 1999- 2016 (%) 14.79±0.6 13.96±0.6 15.88±0.8 644 645 646 647 648 649 650 651 652 653 654 655 656 20 Table 2. The community responses of above ground biomass increment ABI and density to long-term drought and natural droughts for the forest and shrubland sites. The responses were statistically analysed by linear mixed models based on the lowest AIC values. The variables of experimental drought, SPEI (May SPEI-7 and May SPEI-5) and interaction of experimental drought and SPEI were applied into the models for the two canopy types. And the variables of experimental drought and May SPEI-4 were applied into the model for shrubland. The significance and non-significance were listed. Plant communities Variables t-value p-value Experimental drought -3.52 p < 0.001 H-canopy May SPEI-7 2.77 p < 0.05 Experimental drought x May SPEI-7 1.01 ns Experimental drought -2.41 p < 0.05 L-canopy May SPEI-5 2.42 p < 0.05 Experimental drought x May SPEI-5 2.07 p < 0.05 Experimental drought -7.71 p < 0.001 Shrubland May SPEI-4 3.53 p < 0.001 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 21 674 Figures with legends 675 Fig. 1 Two long-term drought experiments in a Mediterranean forest and shrubland. Plastic 676 cover system used to rainfall runoff (A) and exclusion (B) from the forest plots; (C) The 677 location of the two long-term drought experiments; (D, E) And plastic cover system used to 678 exclude rainfall from the shrubland study site 679 680 Fig. 2 Meteorological drought at summer and spring for the forest (A, C) and shrubland (B, D) 681 sites. The SPEI (Standardized Precipitation Evapotranspiration Index) for summer (August 682 SPEI-3) and growing season (May SPEI-5 and May SPEI-4) are presented. And the trends 683 were tested by linear regression, and the significance (p < 0.05) were highlighted with lines. 684 685 Fig. 3 Vegetation responses of aboveground biomass increment (ABI) for two canopy types 686 of forest and standing density (sum ln) for the shrubland. The bars are the best-correlated 687 SPEI (Standardized Precipitation Evapotranspiration Index) for H- and L-canopy forests 688 (May SPEI-7 and May SPEI-5; respectively) the shrubland (May SPEI-4). The vertical bars 689 stand for the standard error (se) of the mean for forests and shrubland (n=2 and n=3, 690 respectively). 691 692 Fig. 4 The comparations for the slopes between control and drought experiments of 693 aboveground biomass increment (ABI) for two canopy forests and standing density of the 694 shrubland. The differences between control and drought are analyzed by smatr test. 695 696 Fig. 5 Effects of experimental drought on vegetation responses of forest aboveground 697 biomass increment (ABI) and shrubland standing density in extreme dry years. The years 698 2001, 2005-2006 and 2015 were extreme dry years for the forest and 2001, 2005-2006, 2016 699 were extreme dry years for shrubland based on the definition of mean annual precipitation 700 (MAP) <10th percentile during the period 1975-2016. There were no significant differences 701 between control and experimental drought during extreme years for the two forest canopies. 702 But we observed the significant differences for the shrubland site (F = 11.02, p < 0.01). 703 704 705 706 707 22 708 709 Fig. 1 23 710 711 Fig. 2 712 713 714 715 716 717 718 719 720 721 722 24 723 724 Fig. 3 25 725 726 727 Fig. 4 728 729 730 731 732 733 26 734 735 736 Fig. 5 27