Holocene aridification trend interrupted by millennial-and centennial-scale climate 1 fluctuations from a new sedimentary record from Padul ( Sierra Nevada , southern Iberian 2 Peninsula ) 3

Holocene centennial-scale paleoenvironmental variability has been described in a 14 multiproxy analysis (i.e. lithology, geochemistry, macrofossil and microfossil analyses) of a 15 paleoecological record from the Padul basin in Sierra Nevada, southern Iberian Peninsula. This 16 sequence covers a relevant time interval hitherto unreported in the studies of the Padul 17 sedimentary sequence. The ca. 4700 yr-long record has preserved proxies of climate variability, 18 with vegetation, lake levels and sedimentological change the Holocene in one of the most unique 19 and southernmost peat bogs from Europe. The progressive Middle and Late Holocene trend 20 toward arid conditions identified by numerous authors in the western Mediterranean region, 21 mostly related to a decrease in summer insolation, is also documented in this record, being here 22 also superimposed by centennial-scale variability in humidity. In turn, this record shows 23 centennial-scale climate oscillations in temperature that correlate with well-known climatic 24 events during the Late Holocene in the western Mediterranean region, synchronous with 25 variability in solar and atmospheric dynamics. The multiproxy Padul record first shows a 26 transition from a relatively humid Middle Holocene in the western Mediterranean region to more 27 aridity from ~4700 to ~2800 cal yr BP. A relatively warm and humid period occurred between 28 ~2600 to ~1600 cal yr BP, coinciding with persistent negative NAO conditions and the historic 29 Iberian-Roman Humid Period. Enhanced arid conditions, co-occurring with overall positive 30 NAO conditions and increasing solar activity, are observed between ~1550 to ~ 450 cal yr BP 31 (~400 to ~1400 CE) and colder and warmer conditions happened during the Dark Ages and 32 Medieval Climate Anomaly, respectively. Slightly wetter conditions took place during the end of 33 the MCA and the first part of the Little Ice Age, which could be related to a change towards 34 negative NAO conditions and minima in solar activity. Evidences of higher human impact in the 35 Padul peat bog area are observed in the last ~1550 cal yr BP. Time series analysis performed 36 from local (Botryococcus and TOC) and regional signals (Mediterranean forest) helped us 37 determining the relationship between southern Iberian climate evolution, atmospheric, oceanic 38 dynamics and solar activity. 39


Introduction 42
The Mediterranean area is situated in a sensitive region between temperate and subtropical 43 climates making it an important place to study the connections between atmospheric and oceanic 44 dynamics and environmental changes. Climate in the western Mediterranean and the southern 45 Iberian Peninsula is influenced by several atmospheric and oceanic dynamics (Alpert et al.,46 2006), including the North Atlantic Oscillation (NAO) one of the principal atmospheric 47 phenomenon controlling climate in the area (Hurrell, 1995

Padul peat bog 126
The Padul basin is situated at approximately 725 m of elevation in the southeastern part of the 127 Granada Basin, at the foothill of the southwestern Sierra Nevada, Andalucía, Spain (Fig. 1). This 128 is one of the most seismically active areas in the southern Iberian Peninsula with numerous faults 129  machine equipped with a hydraulic piston corer from the Scientific Instrumentation Centre of the 175 University of Granada. The sediment cores were wrapped in film, put in core boxes, transported 176 and stored in a dark cool room at +4ºC. 177

Age-depth model (AMS radiocarbon dating) 178
The core chronology was constrained using fourteen AMS radiocarbon dates from plant remains 179 and organic bulk samples taken throughout the cores (Table 1). In addition, one sample with 180 gastropods was also measured for AMS radiocarbon analysis, although it was rejected due to 181 important reservoir effect, which provide a very old date. Thirteen of these samples came from 182 Padul-15-05 and one from the nearby Padul-13-01 (Table 1). We were able to use this date from 183 Padul-13-01 core as there is a very significant correlation between the upper part of Padul-15-05 184 and Padul-13-01 cores, shown by identical lithological and geochemical changes (Supplementary 185 information 1; Figure S1). The age model for the upper ~3 m until 21 cm from the surface was 186 built using the R-code package 'Clam 2.2' (Blaauw, 2010) employing the calibration curve 187 IntCal 13 (Reimer et al., 2013), a 95 % of confidence range, a smooth spline (type 4) with a 0.20 188 smoothing value and 1000 iterations (Fig. 2). The chronology of the uppermost 21 cm of the 189 record was built using a lineal interpolation between the last radiocarbon date and the top of the 190 record (Present; 2015 CE). The studied interval in the present work are the uppermost 115 cm of 191 the record that are constrained by six AMS radiocarbon dates (Fig. 2). 192

Lithology, MS, XRF and TOC 193
The length for the Padul-15-05 core is ~ 43 m. In this study, we focus in the uppermost ~ 115 cm 194 from that core. Padul-15-05 core was split longitudinally and was described in the laboratory 195 with respect to lithology and color (Fig. 3).

196
Magnetic susceptibility (MS) was measured with a Bartington MS3 operating with a MS2E  197 sensor. MS measurements (in SI units) were obtained directly from the core surface every 0.5 cm 198 (Fig. 3). 199 Elemental geochemical composition was measured in an X-Ray fluorescence (XRF) Avaatech 200 core scannerâ at the University of Barcelona (Spain). A total of thirty-three chemical elements 201 were measured in the XRF core scanner at 10 mm of spatial resolution, using 10 s count time, 10 202 kV X-ray voltage and a X-ray current of 650 µA for lighter elements and 35 s count time, 30 kV 203 X-ray voltage, X-ray current of 1700 µA for heavier elements. Thirty-three chemical elements 204 were measured but only the most representative with a major number of counts were considered 205 (Si, K, Ca, Ti, Fe, Zr, Br and Sr). Results for each element are expressed as intensities in counts 206 per second (cps) and normalized (norm.) for the total sum in cps in every measure (Fig. 3).

207
Total organic carbon (TOC) was analyzed every 2 or 3 cm throughout the core. Samples were 208 previously decalcified with 1:1 HCl in order to eliminate the carbonate fraction. The percentage 209 of organic Carbon (OC %) was measured in an Elemental Analyzer Thermo Scientific Flash 210 2000 model from the Scientific Instrumentation Centre of the University of Granada (Spain).

211
Percentage of TOC per gram of sediment was calculated from the percentage of organic carbon 212 (OC %) yielded by the elemental analyzer, and recalculated by the weight of the sample prior to 213 decalcification (Fig. 3). 214

Pollen and NPP 215
Samples for pollen analysis (1-3 cm 3 ) were taken every 1 cm throughout the core. Pollen 216 extraction methods followed a modified Faegri and Iversen (1989) methodology. Processing 217 included the addition of Lycopodium spores for calculation of pollen concentration. Sediment 218 was treated with NaOH, HCl, HF and the residue was sieved at 250 mm previous to an acetolysis 219 solution. Counting was performed using a transmitted light microscope at 400 magnifications to 220 an average pollen count of ca. 260 terrestrial pollen grains. Fossil pollen was identified using 221 published keys (Beug, 1961) and modern reference collections at University of Granada (Spain). 222 Pollen counts were transformed to pollen percentages based on the terrestrial pollen sum, 223 excluding aquatics. The palynological zonation was executed by cluster analysis using twelve 224 different pollen taxa-Olea, Pinus undifferentiated, deciduous Quercus, evergreen Quercus, 225 Pistacia, Ericaceae, Artemisia, Asteroideae, Cichorioideae, Amaranthaceae and Poaceae 226 (Grimm, 1987) (Fig. 4). Non-pollen palynomorphs (NPP) include fungal and algal spores, and 227 thecamoebians (testate amoebae). The NPP percentages were calculated and represented with 228 respect to the terrestrial pollen sum (Fig. 4). Furthermore, some pollen taxa were grouped, 229 according to present-day ecological bases, in Mediterranean forest and xerophytes (Fig. 4). The 230 Mediterranean forest taxa is composed of Quercus total, Olea, Phillyrea and Pistacia. The 231 xerophyte group includes Artemisia, Ephedra, and Amaranthaceae. 232 4 Results 233

Chronology and sedimentation rates 234
The age-model of the studied Padul-15-05 core (Fig. 2) shows that the top 115 cm continuously 235 cover approximately the last ca. 4700 cal yr BP, being the age constrained by fourteen AMS 14 C 236 dates (Table 1). Five distinct sediment accumulation rate (SAR) intervals can be differentiated 237 between 0 and 122.96 cm based on the linear interpolation between radiocarbon dates in the 238 studied core (Fig. 2). 239

Lithology, MS, XRF and TOC 240
The lithology of the upper ca. 115 cm of the Padul-15-05 sediment core was mainly deduced 241 from a visual inspection together with the element geochemical composition (XRF) and the 242 correlation of these data, and the MS of the split cores. In addition, this information was 243 complemented with the TOC (Fig. 3).

244
A Linear r (Pearson) correlation was calculated for the XRF data, the correlation for the 245 inorganic geochemical elements show us two different groups of elements that covary ( in siliciclastic linked to a strong increase in MS. 265

Pollen and NPP 266
A total of seventy-two pollen taxa were identified but only the most representative taxa are here 267 plotted in a summary pollen diagram (Fig. 4). Selected NPP percentages are also displayed in 268 using the program CONISS (Grimm, 1987). Pollen zones are described below: zone is subdivided into subzones-1a, 1b and 1c (Fig. 4) is also observed in this period. Other herbs such as Plantago, Polygonaceae and Convolvulaceae 311 show moderate increases. This zone 4 is subdivided into subzones 4a and 4b (Fig. 4). The top of 312 the record, which corresponds approximately 1830 CE to Present, is characterized by the 313 subzone 4b, the main characteristic that differentiate subzone 4b from the previous 4a is a 314 decrease in some herbs such as Cichorioideae. However, an increase in some xerophytic herbs 315 such as Amaranthaceae occurred. The increase in Plantago is also significant during this period.

316
A noteworthy increase in Pinus (from an average of ca. 14 to 27 %) and a slight increase in Olea 317 and evergreen Quercus are also characteristic of this subzone. With respect to NPP, there is an 318 increase in thecamoebians such as Arcella type and in the largely coprophilous sordariaceous 319 (Sordariales) group. This zone also documents the decrease in fresh-water algal spores, in 320 Botryococcus concomitant with Mougeotia and Zygnema type. lithological, geochemical, and palynological variability and the water lake level oscillations. 326 Sediments with higher values of TOC (more algae and hygrophytes), rich in Ca (related with the 327 occurrence of shells and charophytes remains) most likely characterized a shallow water 328 environment. The absence of aquatic shells, decreasing Ca and a lower TOC and/or a higher 329 input of clastic material (higher MS and Si values) into the lake, could be related with lake level 330 lowering, and a shallower wetland environment (increase in Poaceae) (Fig. 5). 331

Spectral analysis 332
Spectral analysis was performed on selected pollen and NPP time series (Mediterranean forest  333 and Botryococcus), as well as TOC in order to identify millennial-and centennial-scale 334 periodicities in the Padul-15-05 record. The mean sampling resolution for pollen and NPP is ca. 335 50 yr and for geochemical data is ca. 80 yr. Statistically significant cycles, above the 90, 95 and 336 99 % of confident levels, were found around 800, 680, 300, 240, 200, 170 (Fig. 7). 337 increases in clastic input in lake sediments have been interpreted as due to lowering of lake level 373 and more influence of terrestrial-fluvial deposition in a very shallow/ephemeral lake (Martín-374 Puertas et al., 2008). We used the variations between those proxies to estimate water level (Fig.  375  5). 376 Nitrophilous and ruderal pollen taxa (Convolvulus, Plantago lanceolata type, Urticaceae type 377 and/or Polygonum avicularis type) are also very useful as proxies for human impact (Riera et al., of grass cultivation in other Iberian records (Carrión et al., 2001b). In this study we also used the 384 NPP mycorrhizal fungus Glomus sp. as a proxy for erosive activity. This interpretation comes 385 from a study from van Geel et al. (1989), who correlated erosive events with elevated 386 percentages of Glomus cf. fasciculatum. 387

Late Holocene aridification trend 388
This study shows that a progressive aridification trend occurred during at least the last ca.

Millennial-and centennial-scale climate variability in the Padul peat bog during the 419
Late Holocene 420 The multi-proxy paleoclimate record from Padul-15-05 shows an overall aridification trend. 421 However, this trend seems to be modulated by millennial-and centennial-scale climatic 422 variability. However, a discrepancy between the local and regional occurs between 3000-2800 cal yr BP, 430 with an increase in the estimated lake level and a decrease in the Mediterranean forest during the 431 late Bronze Age until the early Iron Age (Figs. 6 and 7). The disagreement could be due to 432 deforestation by humans during a very active period of mining in the area observed as a peak in The high-resolution Padul-15-05 record shows that climatic crises such as the one occurred at ca. 451 4200 cal yr BP, which seems to be recorded worldwide (Booth et al., 2005), are the product of 452 the sum of more than one single climatic event (i.e., ca. 4500, 4300, 4000 cal yr BP) and thus are 453 affected by climatic variability at centennial-scales. 454

Iberian-Roman Humid Period (~2600 to 1600 cal yr BP) 455
High relative humidity is recorded in the Padul-15-05 record between ca. 2600 and 1600 cal yr  Figs. 6 and 7). In 459 addition, there is a simultaneous increase in Botryococcus algae, which is probably related to 460 higher productivity during warmer conditions. Evidence of a wetter climate around this period 461 has also been shown in other regional records and several alpine records from Sierra Nevada. period, seems to point to a common feature of centennial-scale climatic variability in many 480 western Mediterranean and North Atlantic records (Fig. 6). Humid climate conditions at around 481 2500 cal yr BP are also interpreted in previous studies from lake level reconstructions from 482 Central Europe (Magny, 2004 with the decrease in Drift Ice Index. In addition, persistent positive solar irradiance occurred at 491 that time. The increase in Botryococcus algae reaching maxima during the IRHP also seems to 492 point to very productive and perhaps warmer conditions in the Padul peat bog area (Fig. 6). Arid and colder conditions during the Dark Ages (around 1680 to 1000 cal yr BP) are also 518 suggested for the central part of the Iberian Peninsula using a multiproxy study of a sediment 519 record from the Tablas de Daimiel Lake (Gil García et al., 2007). 520 A second period that we could differentiate within this overall arid phase occurred around 1100 521 to 600 cal yr BP/900 to 1350 CE, during the well-known MCA (900 to 1300 CE after Moreno et 522 al., 2012). During this period the Padul-15-05 record shows a slight increasing trend in the 523 Mediterranean forest taxa with respect to the DA, but the decrease in Botryococcus and the 524 higher abundance of herbs still point to overall arid conditions. This change could be related to 525 an increase in temperature, favoring the development of temperate forest species, and would 526 agree with inferred increasing temperatures in the North Atlantic areas, as well as the increase in 527 solar irradiance and the increase in SSTs in the Mediterranean Sea (Fig.7). This hypothesis 528 would

The last ~600 cal yr BP: LIA (1350-1850 CE) and IE (1850 CE-Present) 541
Two climatically different periods can be distinguished during the last ca. 600 cal yr BP (end of 542 Zone 3b to Zone 4; Fig. 4) in the area. However, the climatic signal is more difficult to interpret 543 due to a higher human impact at that time. The first phase around 1350-1450 CE was 544 characterized by increasing relative humidity by the decrease in xerophytes and the increase in 545 Mediterranean forest taxa and Botryococcus after a period of decrease during the DA and MCA, 546 corresponding to the LIA. The second phase is characterized here by the decrease in the 547 Mediterranean forest around 1700-1850 CE, pointing to a return to more arid conditions during 548 the last part of the LIA (Figs. 4 and 7). This climatic pattern agrees with an increase in 549 precipitation by the transition from positive to negative NAO mode and from warmer to cooler 550 conditions in the North Atlantic area during the first phase of the LIA and a second phase 551 characterized by cooler (cold event 0; Bond et al., 2001) and drier conditions (Fig. 6). A stronger 552 variability in the SSTs is described in the Mediterranean Sea during the LIA (Fig. 6). Mayewski  (Fig. 7). In contrast, the spectral analysis 581 performed in the Mediterranean forest time series from Padul peat bog record, pointing to 582 cyclical hydrological changes, shows a significant ~800 yr cycle that could be related to solar 583 variability (Damon and Sonett, 1991) or could be the second harmonic of the ca. ~1600 yr 584 oceanic-related cycle (Debret et al., 2009). A very similar periodicity of ca. 760 yr is detected in 585 the Pinus forest taxa, also pointing to humidity variability, from the alpine Sierra Nevada site of 586 Borreguil de la Caldera and seems to show that this is a common feature of cyclical 587 paleoclimatic oscillation in the area. 588 A significant ~680 cycle is shown in the Botryococcus time series most likely suggesting 589 recurrent centennial-scale changes in temperature (productivity) and water availability. A similar 590 cycle is shown in the Artemisia signal in an alpine record from Sierra Nevada (Ramos-Román et 591 al., 2016). This cycle around ~650 yr is also observed in a marine record from the Alboran Sea, 592 and was interpreted as the secondary harmonic of the 1300 yr cycle that those authors related 593 with cyclic thermohaline circulation and sea surface temperature changes (Rodrigo-Gámiz et al., activities in the mountain areas as well as the abandonment of irrigated terrace systems during 635 the Christian Reconquest. Enhanced soil erosion at this time is also supported by the increase in 636 Glomus type (Fig. 4).

637
An important change in the sedimentation in the environment is observed during the last ca. 300 638 cal yr BP marked by the stronger increase in MS and Si values. This was probably related with 639 the Padul peat bog water drainage by humans using canals in the late XVIII century for 640 cultivation purposes (Villegas Molina, 1967). The increase in wetland vegetation and higher 641 values of Poaceae could be due to cultivation of cereals or by an increase in the population of 642 Phragmites australis (also a Poaceae), very abundant in the Padul peat bog margins at present 643 due to the increase in drained land surface. 644 The uppermost part (last ca. 100 cal yr BP) of the pollen record from Padul-15-05 shows an 645 increasing trend in some arboreal taxa at that time, including Mediterranean forest, Olea and 646 Pinus (Fig. 4). This change is most likely of human origin and generated by the increase in Olea 647 cultivation in the last two centuries, also observed in many records from higher elevation sites 648 from Sierra Nevada, and Pinus and other Mediterranean species reforestation in the 20 th century 649 BP. However, this trend was slightly superimposed by a more arid but warmer event coinciding 671 with the MCA and a cold but wetter event during the first part of the LIA. Besides natural 672 climatic and environmental variability, there seems to be intense human activities in the area 673 during the last the last ca. 1550 cal yr BP. This suggests that the natural aridification trend during 674 the Late Holocene in the western Mediterranean region could have been intensified due to the 675 higher human activity in this area. 676 Furthermore, time series analyses done in the Padul-15-05 record show centennial-scale changes 677 in the environment and climate that are coincident with the periodicities observed in solar, 678 oceanic and NAO reconstructions and could show a close cause-and-effect linkage between 679 them. 680

Acknowledgement 681
This work was supported by the project P11-RNM-7332 funded by Consejería de Economía, 682 Innovación, Ciencia y Empleo de la Junta de Andalucía, the project CGL2013-47038-R funded 683 by Ministerio de Economía y Competitividad of Spain and fondo Europeo de desarrollo regional