Late Neogene and Early Quaternary Paleoenvironmental and Paleoclimatic Conditions in Southwestern Europe: Isotopic Analyses on Mammalian Taxa

Climatic and environmental shifts have had profound impacts on faunal and floral assemblages globally since the end of the Miocene. We explore the regional expression of these fluctuations in southwestern Europe by constructing long-term records (from ∼11.1 to 0.8 Ma, late Miocene–middle Pleistocene) of carbon and oxygen isotope variations in tooth enamel of different large herbivorous mammals from Spain. Isotopic differences among taxa illuminate differences in ecological niches. The δ13C values (relative to VPDB, mean −10.3±1.1‰; range −13.0 to −7.4‰) are consistent with consumption of C3 vegetation; C4 plants did not contribute significantly to the diets of the selected taxa. When averaged by time interval to examine secular trends, δ13C values increase at ∼9.5 Ma (MN9–MN10), probably related to the Middle Vallesian Crisis when there was a replacement of vegetation adapted to more humid conditions by vegetation adapted to drier and more seasonal conditions, and resulting in the disappearance of forested mammalian fauna. The mean δ13C value drops significantly at ∼4.2−3.7 Ma (MN14–MN15) during the Pliocene Warm Period, which brought more humid conditions to Europe, and returns to higher δ13C values from ∼2.6 Ma onwards (MN16), most likely reflecting more arid conditions as a consequence of the onset of the Northern Hemisphere glaciation. The most notable feature in oxygen isotope records (and mean annual temperature reconstructed from these records) is a gradual drop between MN13 and the middle Pleistocene (∼6.3−0.8 Ma) most likely due to cooling associated with Northern Hemisphere glaciation.


Introduction
Profound paleoenvironmental and paleoclimatic events in the late Cenozoic affected life on Earth and gave rise to modern climate regimes and biomes. Progressive cooling, which began in the middle Miocene , ultimately led to the onset of Northern Hemisphere glaciation ,2.7 Ma [1][2][3]. This cooling was not monotonic, however. For example, reorganized ocean circulation, perhaps associated with initial restriction of circulation between the Pacific and Atlantic, contributed to the Pliocene Warm Period between ,4.7 and 3.1 Ma [4]. Shifts in temperature and ocean circulation were associated with shifts in the global water budget, though impacts varied by region. Furthermore, terrestrial environments were transformed from the end of the Miocene to the beginning of the Pliocene (,8-3 Ma) by the worldwide expansion of C 4 plants [5][6]. C 4 plants evolved repeatedly from C 3 plants, most likely as a response to low atmospheric pCO 2 , higher temperatures and increasing waterstress [7].
In southern Europe, our focus here, tectonic closure of the Mediterranean Basin reduced circulation from the Atlantic, likely exascerbated by a drop in sea level associated with increased Antarctic ice volume, culminating with the formation of thick evaporite deposits (Messinian Salinity Crisis or MSC) between ,6.0 and 5.3 Ma [8][9].
As one of the few locations in southern Europe with a relatively complete (albeit low resolution) late Cenozoic stratigraphic succession, a number of recent investigations have reconstructed regional paleoclimatic and paleoenvironmental conditions on the Iberian Peninsula. Based on the bioclimatic analysis of Plio-Pleistocene fossil rodent assemblages, Hernández Fernández et al. [10] argued there was a cooling trend, from subtropical temperatures in the early Pliocene to temperate conditions for the rest of the studied period. Study of palynological records from different Iberian sections led Jiménez-Moreno et al. [11] to suggest that warm temperatures of the Early to Middle Miocene gave way to progressively cooler temperatures in the remainder of Miocene and Pliocene. By the end of the Pliocene and beginning of the Pleistocene, the Iberian palynological record showed the development of steppes, coincident with cooler and drier conditions at the start of glacial-interglacial cycles in the Northern Hemisphere. Van Dam [12] investigated precipitation rates in the Iberian Peninsula using micro-mammal community structure. The most striking features are a decrease of mean annual precipitation (MAP) in the beginning of the Late Miocene (,1128.5 Ma), an increase in MAP in the middle part of the Late Miocene (,8.526.5 Ma) and a drop in MAP between the end of the Late Miocene and the Late Pliocene (,6.523 Ma). Böhme et al. [13] reconstructed MAP using herpetological assemblages between the end of the Early Miocene and the Early Pliocene in the Calatayud-Daroca Basin. Their MAP record differed from that of van Dam Mammalian tooth enamel is a reliable source of isotopic data that can be used to explore past environmental and climatic changes. Here, the stable carbon and oxygen isotope compositions of fossil tooth enamel from different genera of herbivorous mammals spanning from late Miocene to middle Pleistocene (,11.1-0.8 Ma) were analyzed. Our objectives are twofold: 1) to infer the paleoecology of the selected taxa over the study interval, and 2) to reconstruct paleoenvironmental and paleoclimatic trends in Iberia from the late Miocene to the middle Pleistocene.

Materials and Methods
The Iberian Cenozoic basins ( Fig. 1) were formed as a consequence of Alpine compression between the African and Eurasian tectonic plates [14][15]. Most of the basins are located on basement comprising Precambrian and Paleozoic metasediments or granitoids and Mesozoic detrital and carbonate rocks. These basins constitute 40% of the total surface area of the Iberian Peninsula and they offer a complete sedimentary record that spans most of the Cenozoic. Most fossil sites selected for this study (La Roma 2, Masía de la Roma 604B, Puente Minero, Los Mansuetos, Cerro de la Garita, El Arquillo 1, Las Casiones, Milagros, La Gloria 4) are in the Teruel Basin in the northeastern Iberian Peninsula. The name, age and taxonomic composition for localities in the Teruel Basin and other Neogene and Quaternary sites are supplied in Table 1.
The stable carbon and oxygen isotope composition of tooth enamel was analyzed for proboscideans, suids, giraffids, cervids, bovids, and equids from 18 localities from the Iberian Peninsula spanning from 11.1 to 0.8 Ma (late Miocene-middle Pleistocene) (Table S1). Chronological ages of the studied localities are from Domingo et al. ( [16] and unpublished data). Although ages are assigned for each fossil site, the MN (Mammal Neogene) biochronology is used in order to allow comparisons among localities [17][18][19][20][21]. Since all the basins studied here belong to the same biogeographic province [22], the use of the MN units to aggregate fossil sites is assumed to be an appropiate approach, despite the fact that the Mammal Neogene biochronological system has been challenged as a true biozonation at larger scales [22][23][24].
Tooth enamel was sampled using a rotary drill with a diamondtipped dental burr. Fossil teeth for this study are housed in the Museo Nacional de Ciencias Naturales-CSIC (Madrid, Spain) and Fundación Conjunto Paleontológico de Teruel-Dinópolis (Teruel, Spain), after being recovered in excavations carried out with public funding. Sampling was performed with the permission of both institutions.
Measurement of d 13 C values of fossil tooth enamel allows for characterization of the diet of extinct taxa, providing a means to reconstruct past landscapes and habitats [25][26][27][28][29][30][31]. For herbivorous mammals, the d 13 C value of tooth enamel (d 13 C enamel ) has a direct relationship to the d 13 C value of the diet (d 13 C diet ), which varies depending on plant photosynthetic pathways (C 3 , C 4 , CAM), as well as ecological factors (aridity, canopy density, etc.) that affect fractionation during photosynthesis [32][33]. The d 18 O values in the carbonate and phosphate fractions of mammalian tooth enamel record the d 18 O value of body water (d 18 O bw ), which in turn is a reflection of oxygen uptake (inspired O 2 and water vapor, drinking water, dietary water, oxygen in food dry matter) and loss (excreted water and solids, expired CO 2 , and water vapor) during tooth development [34][35]. Carbon and oxygen isotope results are reported in d-notation d H X sample = [(R sample -R standard )/R standard ]61000, where X is the element, H is the mass of the rare, heavy isotope, and R = 13 C/ 12 C or 18  Tooth enamel samples (n = 149) were analyzed for the carbon and oxygen isotope composition of carbonate in bioapatite (d 13 C and d 18 O CO3 , respectively). Carbonate analyses were conducted at the stable isotope laboratories of the University of California Santa Cruz using a ThermoScientific MAT253 dual inlet isotope ratio mass spectrometer coupled to a ThermoScientific Kiel IV carbonate device and of the University of Minnesota using a ThermoScientific MAT252 dual inlet isotope ratio mass spectrometer coupled to a ThermoScientific Kiel II carbonate device. Approximately 5-6 mg of tooth enamel were sampled and treated with 30% H 2 O 2 for 24 h. Samples were rinsed 5 times in deionized (DI) water and soaked for 24 h in 1 M acetic acid buffered to ,pH 5 using Ca acetate solution. After 5 rinses with  for each MN, x a and x b are mean isotopic values for taxa a and b, and n a and n b are the number of selected teeth for taxa a and b. We opted to use the weighted mean since the number of analyzed teeth differs among taxa and therefore, they do not contribute equally to the final average. The application of the weighted mean when constructing temporal trends allows to avoid biases due to differences in physiological and ecological traits among taxa. MAP was estimated following the work of Kohn [38] after a modern equivalent of diet composition (d 13 C diet, meq ) had been calculated using the following equation: where d 13 C leaf = d 13 C tooth -14.1% [39], d 13 C modern atmCO2 is 28%, and d 13 C ancient atmCO2 is the mean d 13 C atmCO2 values from Tipple et al. [40] considering the following time bins: late Miocene, Pliocene and Pleistocene (  (Table S3). Equations were selected according to the closest living relative of the fossil taxa assuming there were no significant differences in the d 18 O PO4 -d 18 O w fractionation between modern and fossil mammals.
Finally, we used a regression equation between MAT and weighted d 18 O w estimated using meteorological data included in Rozanski et al. [41]: Equation 3 was selected because it uses data from meteorological stations worldwide, hence all existing climate regimes are represented. Tectonic reorganization including the closure and opening of sea gateways (e.g., closure of the Panama Isthmus and the passage between the Indian Ocean and the Tethys, opening of the Drake passage and Bering Strait), the uplift of mountain chains (e.g., Himalaya, Andes, Alps) along with shifts in the orbital cycles have exerted an important control on global ice volume and distribution as have perturbations in the atmospheric CO 2 concentration and, by extension, in the carbon cycle. These factors have given rise to different climate regimes since the late Miocene and have culminated in modern climate configuration. In general, Cenozoic climates were globally warmer than at present as corroborated by different proxies [1,[42][43][44]. Warmer conditions have also been recorded in Western Europe during the Miocene and most of the Pliocene based on palynology, vertebrate fossils and General Circulation Models [11,42,[45][46] with the definitive establishment of the Mediterranean climate regime at some point between 3.4 and 2.5 Ma [10][11]. Hernández Fernández et al. [10] and van Dam [12] highlighted the migration of the atmospheric cells, with the subtropical high pressure belt (between the Ferrel and Hadley cells) fluctuating since the late Miocene and profoundly affecting the distribution of Iberian ecosystems. Biome analyses carried out in the Iberian Peninsula between the Miocene and Pleistocene based on macro-and micro-mammals assemblages [10,[47][48] detected a shift in biomes from tropical deciduous woodland, savanna and subtropical desert during the Miocene and Early Pliocene, to nemoral broadleaf deciduous forest for the Late Pliocene, to the modern Mediterranean conditions characterized by schlerophyllous woodland-shrubland since the end of the Pliocene. Due to the different climate regimes and biomes that existed in the Iberian Peninsula during the period under study (late Miocene-middle Pleistocene), it is necessary to use a MAT-d 18 O w relationship that considers data from a wide range of climate regimes and biomes.
Statistical analyses were performed using SPSS PASW Statistics 18.0 software. Analysis of covariance (ANCOVA) was used to compare linear regressions. Analysis of variance (ANOVA) and Student-t tests were used to detect significant differences in isotopic data among taxa within MN intervals, whereas ANOVA and posthoc Tukeys analyses were used to analyze the variability of the isotopic record among MNs.

Diagenesis
The potential for diagenetic alteration should be assessed before accepting paleoecological or paleoenvironmental interpretations based on stable isotope results from fossil bioapatite. Here, only tooth enamel was analyzed, as it is the mineralized tissue least likely to experience isotopic alteration during diagenesis [49]. Phosphate oxygen is more resistant to inorganic isotopic exchange than carbonate oxygen, but carbonate oxygen is more resistant to microbially-mediated exchange [50].
Modern, unaltered bioapatites exhibit a linear relationship between d 18 O CO3 and d 18 O PO4 with a consistent difference (d 18 O CO3 -d 18 O PO4 = ? 18 O CO3-PO4 ) of 8.6-9.1% for co-occurring CO 3 22 and PO 4 23 formed in isotopic equilibrium with body water at a constant temperature [51][52][53]. In this study, the mean ? 18 O CO3-PO4 was 8.261.3% (VSMOW), close to the expected value. Figure 2 shows the d 18 O PO4 -d 18 O CO3 regression from this study. Zazzo et al. [50] suggested that the slope of the regression line between d 18 O CO3 and d 18 O PO4 is close to 1 in modern (unaltered) bioapatite. Slopes higher than unity suggest more extensive alteration of d 18 O CO3 by inorganic mechanisms, whereas slopes lower than unity indicate a higher degree of microbially-mediated isotopic exchange of phosphate. Our slope is close to unity, but slightly higher (1.07). This slope is not as high as those observed by Zazzo et al. [50] in samples affected by intense diagenesis (see their Fig. 4) and no significant differences were detected by an ANCOVA test between our d 18 O PO4 -d 18  These results suggest that our samples have experienced minimal isotopic alteration of either phosphate or carbonate oxygen. There are no comparable tests for carbon isotopes, but the fact that species cluster in bivariate isotope space, and that the relative positions of these clusters are consistent for some taxa, suggest that animal paleobiology, and not diagenesis, is the main driver of isotopic variation.

Paleoecology of the Iberian Fossil Mammalian Taxa
In terrestrial settings, the dominant control on the d 13 C value of plants is photosynthetic pathway [54][55][56][57][58]. Plants following the C 3 or Calvin-Benson photosynthetic pathway (trees, shrubs, forbs and cool-season grasses) strongly discriminate against 13 C during fixation of CO 2 , yielding tissues with d 13 C values averaging 227% (VPDB) (ranging from 236 and 222%). The most negative d 13 C values of this range (236 to 230%) reflect closedcanopy conditions due to recycling of 13 C-depleted CO 2 and low irradiance. The highest values (225 to 222%) correspond to C 3 plants from high light, arid, or water stressed environments. C 4 plants (Hatch-Slack photosynthetic pathway) comprise grasses and sedges from areas with a warm growing season and some aridadapted dicots. C 4 plants discriminate less against 13 C during carbon fixation, yielding mean d 13 C value of 213% (ranging from 217% to 29%). Crassulacean acid metabolism (CAM) is the least common pathway, occurring chiefly in succulent plants. CAM plants exhibit d 13 C values that range between the values for C 3 and C 4 plants. Using the expected d 13 C ranges for C 3 and C 4 plants and a typical diet-to-enamel fractionation of +14.160.5% [39], we can estimate the expected d 13 C values for pure C 3 feeders in different habitats (closed-canopy, 222 to 216%; woodlandmesic C 3 grassland, 216 to 211%; open woodland-xeric C 3 grassland, 211 to 28%) and pure C 4 feeders (23% to +5%). Enamel d 13 C values between 28% and 23% represent mixed C 3 -C 4 diets. When considering fossil taxa, however, it is necessary to account for shifts in the d 13 C value of atmospheric CO 2 (the source of plant carbon), including anthropogenic modification due to fossil fuel burning, which has decreased the d 13 C value of atmospheric CO 2 from 26.5 to 28% since onset of the Industrial Revolution [59][60]. Using isotopic data from marine foraminifera, Tipple et al. [40] reconstructed the d 13 C value of the atmospheric CO 2 since the Cretaceous. In order to calculate vegetation d 13 C end-members, we considered the following time bins: late Miocene, Pliocene and Pleistocene. Table 2 shows a summary with d 13 C atmCO2 and d 13 C cut-off values for the transition between diets composed of different types of vegetation  Among Miocene proboscideans, Gomphotherium angustidens had brachyo-bunodont dentition, suggesting a browsing behaviour, which is in agreement with d 13 C values pointing to consumption of woodland or woodland/C 3 grassland vegetation. The gomphothere Tetralophodon longirostris replaced Gomphotherium angustidens. Tetralophodon was larger and more hypsodont than Gomphotherium, but also probably a browser [64]. Its d 13 C values shift from lower values similar to Gomphotherium in older localities (Nombrevilla and Los Valles de Fuentidueñ a, MN9) to ,0.5% higher values in younger sites (Puente Minero, MN11 and Cerro de la Garita, MN12). The mammutid Zygolophodon turicensis from the Cerro de la Garita locality had a zygodont dentition with sharp, transverse ridges and d 13 C values similar to those for the youngest Tetralophodon. Overall, the slight trend of increasing d 13 C values toward the end of the Miocene in these proboscideans points to consumption of plants from increasingly open, drier habitats. Since proboscideans are obligate drinkers [34,65] (Fig. 3, Table 3).
In the case of Miocene bovids, the boselaphine Tragoportax is the best-represented genus. It had relatively long limbs suggesting cursorial adaptations and preference for open habitats [64]. Microwear studies performed on the teeth of this bovid suggest it was a mixed feeder with strong grazing habits [66][67]. This is consistent with its d 13 C values, which are the highest for any taxon in all the MNs in which Tragoportax occurs (Fig. 3), and in most MNs are close to values expected for animals foraging in open woodlands or dry C 3 grasslands. In the MN13 fossil sites, Tragoportax d 13 C values were ,1-2% lower, most likely due to a shift towards more humid conditions (see next section and Fig. 4). Using dental microwear, Merceron et al. [68] showed that a species of the bovid Hispanodorcas from the Neogene of northern Greece (H. orientalis) had strong similarities to extant browsers and mixed feeders; that reconstruction is also consistent with the d 13 C values of H. torrubiae from Los Mansuetos (MN12; Fig. 3). According to Merceron et al. [69], Tragoportax was likely an obligate drinker based on a low inter-individual d 18 O variability among species, and therefore its high d 18 O CO3 and d 18 O PO4 values when compared to the rest of taxa (including the bovid H. torrubiae) in MN10-12 (Fig. 3, Table 3) are consistent with ingestion of evaporated water in open environments.
Cervids have the lowest d 13 C values of the late Miocene mammalian assemblage (Fig. 3), consistent with membership in the browsing guild as indicated by tooth morphology and microwear analyses [64,70] (Table 3 (Fig. 3, Table 3) may indicate that this sivatherine obtained much of its water from highly evaporated leave water as suggested by Cerling et al. [71] for the extinct Palaeotragus and Levin et al. [65] for modern giraffids.
Finally, the suid Microstonyx major has intermediate d 13 C values in the Puente Minero (MN11) and Cerro de la Garita (MN12) fossil sites. Suids are more omnivorous and according to Agustí and Antón [64], M. major had a cranial morphology suggesting a strong and highly mobile muzzle disk (like in modern pigs) interpreted as an adaptation to digging roots and tubers, although other sources of dietary intake such as fruits, insects and even carrion cannot be discarded, the combination of which may have given rise to the observed intermediate d 13 C values.

Pliocene (La Gloria 4, MN14-Huélago, MN16)
The gomphothere Anancus arvernensis has d 13 C values indicative of browsing in a woodland to woodland-mesic C 3 grassland (Fig. 3), which is consistent with observations by Agustí and Antón [64] and Tassy [72] who argued that its dentition was similar to that of other tetralophodont gomphotheres. Low d 18 O CO3 and d 18 O PO4 values may relate to ingestion of water in closed areas or flowing water not subject to significant evaporation (Fig. 3, Table 3).
The Pliocene bovids Gazella and Protoryx were ubiquitous taxa as far as occupancy of different habitats is concerned and are considered browsers to mixed feeders [67][68]70,[73][74]; the relatively low d 13 C values for these taxa are more supportive of a browsing habitat (Fig. 3, Table 3). Rivals and Athanassiou [70] argued that the antelope Gazellospira torticornis was a mixed feeder that grazed on seasonal or regional basis. Although this antelope has ,1 to 1.5% higher d 13 C values than Gazella and Protoryx, these values are consistent with woodland browsing and do not point to a substantial proportion of grass in the diet. The bovid cf. Hesperidoceras merlae has similar d 13 C values to G. torticornis (Fig. 3,

Pleistocene (La Puebla de Valverde, MN17-Huéscar 1)
Filippi et al. [76] and Palombo et al. [77] studied microwear on Elephas antiquus of the Middle Pleistocene and suggested a browsing to mixed feeding behaviour; our d 13 C data are consistent with woodland browsing but do not point to a substantial proportion of grass in the diet (Fig. 3). Mammuthus meridionalis has been considered to be a mixed feeder to grazer based on microwear and previous stable isotope analyses [78][79][80]. Our M. meridionalis d 13 C value is more indicative of a mixed feeder occupying a woodland (Fig. 3).
The bovid, Gallogoral meneghini from La Puebla de Valverde (MN17) has higher d 13 C values, close to those expected for an animal foraging in an open woodland (Fig. 3, Table 3). According to Guérin [81], Agustí and Antón [64] and Brugal and Croitor [82], G. meneghini was a mixed feeder with a robust skeleton and short limbs adapted to locomotion on mountainous uneven areas similar to modern gorals from Asia. Fakhar-i-Abbas et al. [83] studied the feeding preferences of the gray goral and found out that it relies mainly on grasses, although it can browse too; this is in agreement with our G. meneghini d 13 C values situated towards the high cut-off for open woodland and mesic C 3 grassland. Lower d 13 C values in the case of Gazella borbonica are similar to those for this bovid in the Pliocene and again these values are consistent with woodland browsing and do not point to a substantial proportion of grass in the diet.
The cervid Croizetoceros ramosus also shows low d 13 C values indicative of a woodland. The equid Equus stenonis has higher d 13 C values near those expected for animals feeding in an open woodland (Fig. 3). This might be indicating ingestion of C 3 grasses not subject to water stress. Slightly higher d 18 O CO3 and d 18 O PO4 values for the equid E. stenonis and the cervid C. ramosus in comparison to the elephantids and bovids may suggest ingestion of water in more open areas (in the case of the equid) or consumption of more evaporated water in leaves (in the case of the cervid) (Fig. 3, Table 3). Figure 4 shows d 13 C and modern equivalent d 13 C values (d 13 C diet, meq ), which can be related to MAP (see material and methods section and Table S2) between MN7/8 and the middle Pleistocene.

Changes in d 13 C Values
A prominent faunal turnover event, known as the Middle Vallesian Crisis (ca. 9.6 Ma) [84] occurred in Western Europe between MN9 and MN10. This event is recognized by the replacement of humid-adapted taxa with taxa more adapted to drier conditions, and is associated with the replacement of evergreen subtropical woodlands by a seasonally adapted deciduous woodland as observed by Agustí and Moyà-Solà [85] and Agustí et al. [84] in the Vallès-Penedès Basin (North Eastern Iberian Peninsula). This event coincides with the Mi7 positive shift in benthic foraminifera d 18 O values interpreted to reflect global cooling [86][87]. In Figure 4A, d 13 C values of herbivorous mammals in the Iberian Peninsula increase between MN9 (Nombrevilla 1 and Los Valles de Fuentidueñ a) and MN10 (La Roma 2 and Masía de la Roma 604B), which may be related to a change towards drier conditions. d 13 C diet, meq values mirror tooth enamel d 13 C values, with an increase observed between these MNs (Fig. 4B). MAP values (estimated after Kohn, [38]) dropped from ,410 mm/yr to ,200 mm/yr between MN9 and MN10. Böhme et al., [13]), who used the ecophysiological structure of herpetofaunas in the Calatayud-Daroca Basin of Spain to estimate changes in MAP over the Miocene, also recognized a decrease in precipitation at 9.7-9.6 Ma. However, the decrease in the study of Böhme et al. [13] is greater than 1000 mm/yr in comparison with the ,200 mm/yr decrease estimated here. The explanation for this large difference is unclear, but we note that the Kohn [38] method has relatively large error.
During MN13, the Messinian Salinity Crisis (MSC) in the Mediterranean Basin resulted from a sharp decrease in the marine water circulation from the Atlantic and culminated in the formation of thick evaporite deposits [8]. The lack of significant differences in mammal tooth enamel d 13 C values between MN12 and MN13 (t = 21.285, p = 0.204) suggests that the MSC did not cause substantial modifications to terrestrial ecosystems, although a post-hoc Tukeys test places the MN13 in groups a, b, c, and d (versus groups c and d for MN12) pointing to more humid conditions. However, and since we cannot unequivocally determine the synchrony between the chronology assigned to the MN13 localities considered in this study and the MSC, we regard this conclusion as preliminary pending more accurate datings. Ongoing paleomagnetic analyses in the MN13 Venta del Moro fossil site may modify the current chronology, which places this locality as contemporaneous to the MSC (J. Morales, pers. comm. 2013). Fauquette et al. [88] carried out an analysis of 20 pollen sequences in the Mediterranean realm and found no differences when comparing data before, during and after the MSC.
Mean tooth enamel d 13 C values decrease sharply from MN13 to MN14, and the mean value in MN15 is lower still (Fig. 4A). The statistically significant drop in d 13 C values during MN14 and MN15 may be related to the Pliocene Warm Period which began at ,5 Ma and brought about more humid conditions in Europe [1,64]. Figure 4B also shows a drop in d 13 C diet, meq , which corresponds to an increase in MAP values of ,400 mm/yr between MN13 (, 410 mm/yr) and MN14 and MN15 (, 800 mm/yr). The decrease in d 13 C values in MN14 and MN15 is not biased by the type of taxa sampled, since in La Gloria 4 and Layna ubiquitous taxa such as Gazella and Protoryx were chosen and therefore, an isotopic change in these generalistic bovids [67][68]70,[73][74] points towards real paleoenvironmental variations.
After MN15, d 13 C values increase in MN16, MN17 and middle Pleistocene, but do not reach values as high as those observed in MN10, MN11 and MN12 (Fig. 4A). This increase in d 13 C values corresponds to global and regional climatic changes and to faunal and environmental changes in Europe. The beginning of MN16 (,3.2 Ma) [89] predates the onset of Northern Hemisphere glaciation [1,90]. At that time, the modern Mediterranean climatic regime was established and aridity in Europe was enhanced, which led to changes in mammalian fossil assemblages in such a way that, according to Agustí et al. [89], the Villanyian mammal turnover occurred at this time with an increase in grazers, the appearance of morphological features associated with a highly cursorial lifestyle in some ungulates, and the diversification of pursuit carnivores. All of these changes point towards the development of prairies and grasslands in Europe [64,89]. Fortelius et al. [91] estimated hypsodonty index in mammalian herbivores between the Late Miocene and the Pliocene in Eurasia and found out that browsing taxa in MN15 were replaced by grazers in MN16 and MN17. Another important event occurred at ,2.6 Ma, when there was a replacement of forests by tundra-like vegetation in northern and central Europe, while in northwestern Africa, savanna biome shrunk in favour of desert biome [64]. The Iberian Peninsula also experienced a shift towards the development of more herbaceous vegetation, such as the well-documented increase of Artemisia [11,92]. The increase in mammal tooth enamel d 13 C values observed in MN16, MN17 and the middle Pleistocene may reflect this episode.  (Fig. 5B) estimated using the taxon-specific relationships (Table S3) and equation (3)  Jiménez-Moreno et al. [11] argued that during the Messinian, there were not major variations in climate before, during and after the MSC. The pollen assemblage from the Carmona section suggests a MAT between 20.5uC and 22.5uC during the Messinian in southwestern Spain. In our study, MN13 fossil sites that correspond to the Messinian suggest a warmer MAT of 23.865.0uC (Fig. 5B) [41]. MAT values based on pollen and micro-mammal data are from Fauquette et al. [88,95], Hernández Fernández et al. [10] and Jiménez-Moreno et al. [11]. Chronology according to 1 Domingo et al. ([16], unpublished data), 2 Agustí et al. [89], 3  complicated by glacial-interglacial dynamics, which may have produced large shifts in temperature in relatively short periods of time.

Temperature Record
Overall, the MAT values estimated here using mammalian tooth enamel are in good agreement with data from palynology and rodent assemblage analyses. Other isotopic studies on mammal tooth enamel from the Iberian Peninsula [93][94] showed consistently lower MAT values compared to those obtained here. This may be due to the use of different equations relating MAT and d 18 O w . We use the equation (3)

Absence of C 4 Vegetation in Southwestern Europe
Our d 13 C record offers no evidence of the high d 13 C values typical of C 4 consumers (Figs. 3 and 4, Table 2) and the calculation of the percentage of C 4 vegetation points to a low C 4 dietary intake (,20%) in most of the analyzed taxa. This percentage of C 4 vegetation may reflect either an actual small fraction of C 4 plants in mammal diets or it may be an artifact related to the ingestion of C 3 plants from open areas subject to water stress (which therefore have higher d 13 C values). The lack of a significant expansion of C 4 plants in the Iberian Peninsula is intriguing. The expansion of C 4 plants took place between 9 and 2 Ma in different regions [6]. C 4 photosynthesis is favored under conditions of low atmospheric CO 2 , when growing seasons experience high temperature (i.e., summer rainfall), in arid regions, or in soils with high salinity. The combined effects of fires and herbivory may also lead to open environments where C 4 grasses may thrive. Given the high temperatures suggested by our isotopic analyses (Fig. 5) and other proxy data, conditions in the late Miocene and early Pliocene would seem conducive to a regional C 4 expansion if habitats were relatively open and there was adequate summer precipitation.
Palaeoclimatic studies of Iberian mammalian assemblages from late Miocene to middle Pleistocene (,11.1 to 0.8 Ma) indicate that the most likely biomes at some of the fossil sites studied here (Puente Minero, Los Mansuetos, Cerro de La Garita, El Arquillo, Venta del Moro, La Gloria 4, Layna and Huéscar 1) were tropical deciduous woodland with perhaps occasional savanna and subtropical desert environments, prior to the development of the sclerophyllous woodland-shrubland at the start of the Pleistocene [10,48]. By definition, a woodland supports woody cover of .40% and ,80% with the remaining patches often dominated by grasses, either C 3 or C 4 [96][97]. In a study of the isotopic composition of individual pollen grains from ,20 to 15 Ma in the Rubielos de Mora Basin, Urban et al. [98] showed that while the overall abundance of grass pollen was low and in the range expected for a woodland (10-15%), C 4 grasses comprised 20-40% of the grains. Since there are no isotopic studies on pollen grains in the time interval selected for our study, we assume that C 4 grasses were potentially present in the flora of the Iberian Peninsula since at least the Early Miocene.
While a detailed analysis of the ultimate cause/s for the low abundance of C 4 plants in southwestern Europe after their expansion elsewhere is beyond the scope of this paper, there are several potential explanations. At middle latitudes, only regions with summer rainfall are suitable for C 4 grasses. A seasonality of rainfall similar to the modern Mediterranean precipitation pattern, with precipitation occurring chiefly during the winter, would lead to very low abundance of C 4 plants on the Iberian Peninsula. Several studies have questioned the age of 3.4 and 2.5 Ma for the onset of the Mediterranean climate and proposed that such a climate regime may have been present much earlier (e.g., [99]). For example, Axelrod [100] studied fossil leaves in the Mediterranean area and argued that sclerophyllous evergreen woodlands with chaparral undergrowth were present throughout the Miocene. Yet there is no way to determine if these species were dominant on the landscape, and Axelrod ([100]: p. 325) himself noted that sclerophyllous species might constitute part of the tropical-subtropical woodlands understory but that the ''existence of chaparral and macchia over wide areas as climax vegetation in the Tertiary seems unlikely''.
Tzedakis [99] reviewed evidence for the onset of the Mediterranean climate regime and noted that seasonality similar to the summer-dry and winter-wet pattern may have appeared intermittently before the onset of the ''true''-Mediterranean climate regime. The occasional occurrence of Mediteranean-like climate in the Iberian Peninsula in the early Pliocene has also been suggested by studies of rodent faunas and has been linked to the presence of bimodal precipitation regimes, which may produce a short summer dry season in addition to the winter dry season typical of tropical climates [10]. The prevalence of these short summer dry periods is probably not sufficient to explain the absence of C 4 -dominated landscapes.
An alternative is that C 4 plants were somewhat more abundant, but that mammals selectively foraged on C 3 plants, perhaps avoiding C 4 plants because of their lower nutritional value [101]. Paleoecological studies from other regions suggest that this explanation is unlikely. In North America, South America, Asia and Africa (see a review in Strömberg [6]), when C 4 plants became available (as determined by soil carbonates and other lines of evidence), they came to comprise a substantial part of the diet of at least some mammalian grazers. Indeed, once C 4 grass became abundant, different taxa began to specialize on them. There is no reason to assume that some genera of Miocene mammals (e.g., Tragoportax, a mixed feeder with strong grazing habits) in the Mediterranean region would not have used a new dietary resource such as C 4 grasses had they been abundant.
It seems that the most likely cause for a limited C 4 vegetation development may be related to the biome configuration of the late Miocene-Pliocene in the Iberian region. Pollen records indicate low percentages (10-15%) of grasses, belonging to the Poaceae family, during the late Miocene and the Pliocene (Jiménez-Moreno, pers. comm. 2012). Pollen analyses are not able to distinguish between C 3 and C 4 grasses, but if we assume that the percentage of C 4 plants estimated by Urban et al. [98] for the early Miocene Rubielos de Mora Basin (20-40%) was maintained in the late Miocene and Pliocene, the final percentage of C 4 grasses may have not been enough as to be recorded on mammalian tooth enamel d 13 C values.

Conclusions
Long stratigraphic sequences of isotopic data from mammalian tooth enamel are not frequently analyzed due to gaps in the terrestrial fossil record. Such studies are important since they can reveal modifications in paleoenvironmental and paleoclimatic factors in terrestrial settings during critical intervals in Earth history. Here, we used stable isotope analysis of a succession of mammals from 18 localities in Spain ranging in age from 11.1 to 0.8 Ma to reconstruct environmental and climatic changes during the late Neogene and early Quaternary. In general, tooth enamel d 13 C values indicate that analyzed taxa may have occupied woodland to mesic C 3 grassland and in some cases, open woodland to xeric C 3 grassland, with no evidence of significant C 4 consumption in any of the genera we studied. An increase in d 13 C values between MN9 and MN10 appears to correspond to the Middle Vallesian Crisis, a faunal turnover that led to the replacement of humid-adapted taxa by taxa more adapted to drier conditions. A significant decrease in d 13 C values during MN14 and MN15 is probably linked to the Pliocene Warm Period (with an associated increase in moisture), whereas the higher d 13 C values from MN16 onwards may have been a consequence of the increased aridity in Europe related to the onset of Northern Hemisphere glaciation. The MAT pattern estimated using tooth enamel d 18 O PO4 values agrees well with the thermal trend based on palynological records, rodent assemblage structure, and other isotopic studies from the Iberian Peninsula, with a gradual drop in MAT from MN13 onwards in response to the progressive cooling observed since the Middle Miocene and culminating in the Northern Hemisphere glaciation.