Long-term dynamic topographic support during post-orogenic crustal thinning revealed by stable isotope (δ18O) paleo-altimetry in eastern Pyrenees

Understanding the geodynamic and Earth surface processes at the origin of post-collisional surface uplift in mountain ranges requires reconstruction of paleo-elevation. Here, we focus on the topographic evolution of the Cerdanya Basin in the eastern Pyrenees formed by post-orogenic extension during the Late Miocene. Stable isotope (δ18O) analyses of small rodent teeth and biogenic carbonates show the basin uplifted by 500 m since 6.5 Ma. These new paleoaltitudes constraints when combined with the regional geology and geophysical data reveal the anomalously high topography of the region is the result of density changes in the sublithospheric mantle associated with crustal thinning and then opening of Gulf of Lion during the Chattian-early Burdigalian.

dynamically supported. This is further indicated by the large negative Bouguer anomaly of about −100 mGal ( Fig. 1) that led 3,17 to propose the removal of the dense lower crust or/and ascent of an abnormally buoyant and hot lithospheric mantle. For comparison, the rest of the Pyrenees exhibits a homogenous mean elevation with the highest peaks above 3 km that are isostatically compensated by a 44 km crustal root 3 .
The central and eastern Pyrenees are further characterized by the presence of late Oligocene-early Miocene highly-elevated low relief surfaces, considered remnants of a single composite planation surface recently dissected 18 . There is a great debate about how these surfaces were created and especially at which elevation. Two main hypotheses are proposed. The first one assumes a formation at low elevation and an uplift of the eastern Pyrenees since the early Miocene 5,19 . In this interpretation, mantle thinning is thought to have caused uplift of summit peneplain from low-elevation near sea-level 20 or at ca. 750 m to present-day 2.4-2.9 km since 12 Ma 5 . The second hypothesis postulates a control by the piedmont sedimentation and the development of a planation surface at high elevation, implying only a limited surface uplift of 400 m in response to post-orogenic erosional rebound 21,22 . Note that because late Miocene normal faulting accommodated little extension, it is considered to have played a subordinate role.
In the present study, we aim to provide new calibration of the post-orogenic paleo-elevation evolution of the Pyrenees. To this aim, we compare the stable isotope composition of material mineralized from meteoric water at different elevations 23,24 . We targeted the Roussillon Basin that remained at low elevation since the Miocene and the Cerdanya basin which current elevation ranges between 1000 and 1200 m ( Fig. 1) and that are both characterized by contemporaneous Neogene sedimentary filling. Given the lack of material classically used for paleo-elevation reconstruction like soil carbonates or authigenic minerals, we developed an unconventional approach based on δ 18 O measurements of mammal teeth 25 and on freshwater algae (charophytes oogonia) and terrestrial gastropod shells.
Climate models have shown that modern atmospheric circulations in western Europe, characterized by dominant moisture source from the north Atlantic, were established during the late Miocene 29 . Thus, the modern isotope lapse rate established for the eastern Pyrenees 30 is used to extrapolate the δ 18 O lw values to estimate a ∆-elevation paleogradient between the two sites during the late Miocene. Measurements from modern small rivers yielded a gradient of −3.76‰/km for the δ 18 O 30 . We infer a mean ∆δ 18 O lw of −1.5‰ for the Miocene samples that corresponds to an altitude difference of ∆H = 396 ± 50 m according to the modern isotope lapse rate.
The modern difference of elevation is 910 m between the two sites ( Fig. 3). This result therefore suggests that the Cerdanya Basin uplifted by about 500 m since 6.5 Ma. The basin is currently at 1100 m, we thus estimate a paleo-elevation of 600 m consistent with altitude inferred from pollen floras 16 . The concordant results obtained based on two independent approaches emphasize the robustness and the accuracy of the calculated paleo-elevation. We derive a surface uplift rate of 0.07 mm/a since 6.5 Ma in the range of 0.06-0.12 mm/yr obtained by 16 for the same basin and close to uplift rates of 0.08-0.19 mm/yr obtained for the Central Pyrenees 31 . Such a rate is also close to incision rates of 0.05-0.09 mm/yr since 5 Ma obtained in the Têt river canyon (current www.nature.com/scientificreports www.nature.com/scientificreports/ elevation ~400 m) 32 . Pollen data further suggested the relief between the Cerdanya Basin and the surrounding high mountains has not changed since 10 Ma 16 . This inference is supported by the good preservation of the late Miocene sediment infill of the Cerdanya Basin at high elevation and the low rate of erosion on the flank of the mountain range. With a maximum altitude of the region close to 2-2.5 km during the Messinian, the eastern Pyrenees were virtually in the same isostatic state as today, that is the topography was not isostatically compensated by a crustal root. This argues that the major geodynamic changes at the origin of post-orogenic uplift must have started before 10 Ma. This rifting phase ended in the Burdigalian as indicated by a regional erosional surface recognized in the Gulf of Lion onto which the transgressive shallow-marine post-rift Burdigalian series were deposited 35 . This places an additional elevation constraint near sea-level in the Chattian-Early Burdigalian (Fig. 4). The mapping of the Burdigalian erosional surface offshore of the Gulf of Lion 35,36 reveals subaerial erosion occurred on a crust that was moderately to extremely thinned in the SE direction (30 < h c < 5 km; stretching factor 1.4 < β < 9). The elevation of the Gulf of Lion rifted margin was therefore anomalously high and flat in the late Oligocene-Early Burdigalian. In the eastern Pyrenees, the present-day crustal thickness below the Têt Fault ranges between 30 and 40 km. It is thicker below the Cerdanya Basin and thinner below the Conflent Basin, and is only 22 km in the Roussillon Basin. Geophysical data therefore indicate that crustal thinning in the eastern Pyrenees and in the Gulf of Lion did not lead to the subsidence predicted by the McKenzie's model 37 , otherwise the whole region would have been buried several km below sea-level. Pre-break-up surface uplift that does not fit the subsidence effect of thinning the crust (McKenzie stretching model) is documented on many rifted margins. This requires processes leading to density reduction like serpentinization of the exhumed mantle, mantle phase transitions to lighter mineral phases and the trapping of melt in the rising asthenosphere before breakup are required 38 . We infer that similar processes did occur in the eastern Pyrenees and the Gulf of Lion in order to keep the region close to sea level.
Following the early Burdigalian, however, the Gulf of Lion recorded a rapid post-rift subsidence coeval with oceanic spreading in the Ligurian-Provençal Basin and rotation of Sardinia occurred between 20.5 and 15 Ma 39 . The paleo-elevation constraints obtained in this work show that after the onset of oceanic spreading in the Gulf of Lion, the eastern Pyrenees continued to be uplifted. Differential vertical movements between the Gulf of Lion and the eastern Pyrenees likely triggered post-rift normal faulting that led to the development of the Cerdanya Basin during the Tortonian (12-9 Ma). The Late Miocene reactivation of the Têt Fault as a right-lateral strike-slip fault 40 was contemporaneous with the deposition of 400-800 m of non-marine sediments in the Cerdanya Basin 41 . In the Roussillon Basin, a maximum of 800-900 m of post-Messinian sediments is preserved 42 . The formation of the Cerdanya Basin was synchronous with normal faulting along the oblique NNW-trending Transverse Fault system in the Sierras Transversales 43 , volcanism in Emporda (10-9 Ma) and Selva (7-2 Ma) region, North-East Catalonia. Magmatism continued with the Olot (Garrotxa) volcanic system (0.7-0.11 Ma), an intraplate alkaline basaltic volcanism with close affinities to the volcanic system of the French Massif Central and Calatrava, Central Spain 44 .
Because late Miocene normal faulting occurred when the Cerdanya Basin was at elevation, the Tortonian-Messinian extension appears to be a consequence rather than a cause of the regional uplift. The post-Messinian uplift of 500 m resolved from this study therefore represents a fraction of the long-term regional uplift that initiated in the Aquitanian-Late Burdigalian (20 Ma) when the eastern Pyrenees were close to sea-level (Fig. 4). This result reveals that the short-lived (5 Myr) initial back-arc rifting event was the main driver of the dynamic support of the topography. Because the region was close to sea-level in the late Oligocene-early Miocene then uplifted in the Late Miocene, a post-orogenic piedmont sedimentation could hardly be maintained, thus precluding the preservation of pre-Late Miocene planation surface. Other factors such as flexural uplift in the conclusion Paleo-elevation constraints resolved from stable isotopic analyses indicates that the Cerdanya Basin, one of the main valleys of eastern Pyrenees, was at 600 m above sea level during the Messinian, 500 m below its current elevation. Because most of the relief was established at this time, we argue for a moderate late Miocene uplift of the summit planation surface of 500 m. Tectonic-stratigraphic relationships further indicate the pre-6.5 Ma topography was built on an older landscape inherited from the Chattian-early Burdigalian rifting episode that gave birth to opening of the Gulf of Lion. The non-isostatic processes required to support the current topography are therefore the consequence of a short-lived but major geodynamic event at the origin of both crustal thinning and density changes in the mantle. These new paleo-elevation constraints together with other geological data in the region suggest the uplift was a long-term process initiated in the Late Burdigalian in response to pre-breakup uplift in the Gulf of Lion. Following oceanic spreading of the Gulf of Lion and the rotation of Sardinia, the Tortonian extension associated with transcurrent deformation and volcanism was responsible for the last stage of topographic growth of the eastern Pyrenees.

Method
paleoaltitude and paleotemperature reconstructions. The method relies on the comparison of the stable isotope signature of Late Miocene mammal teeth preserved in two basins at the eastern termination of the Pyrenean range: the Roussillon Basin that remained at low elevation since the Miocene and the Cerdanya basin which current elevation ranges between 1000 and 1200 m. The basic principle of isotopic paleoelevation reconstruction lies on the direct dependency of the δ 18 O and δD of rain with elevation, following the Rayleigh distillation behavior 45 . Paleoelevation can thus be quantified from the analysis of mineralization that precipitate from meteoric waters 46   www.nature.com/scientificreports www.nature.com/scientificreports/ basins, clay minerals from fault zones or authigenic minerals mineralized at different elevation 24 . However, such material is rarely preserved syn-orogenic deposits and paleoelevation are thus often difficult to reconstruct. In this work, we adopt an approach combining mammal remains (rodent and lagomorph teeth) with biogenic carbonates. Small mammals are homoeothermic animals living in small areas so the δ 18 O of their biominerals reflects both the life-long δ 18 O composition of their body water 47 and the surface water of their living area δ 18 O w 25 . From the local δ 18 O w , the δ 18 O analyses of non-homoeothermic biogenic carbonates allow constraining paleotemperatures 48 . One of main challenge when reconstructing paleoelevation is to carefully take into account climatic parameters changes through time 49 . To minimize this impact we have compared the δ 18 O of rodent teeth of two sites, one that remained at low elevation and one that was potentially uplifted. We derive a paleo-∆δ 18 O w that could be converted to ∆-elevation 23 . Geochemical results are provided in Supplementary Dataset 1. Fossils were sampled from two contemporaneous deposits located in the eastern Pyrenees. First, we analyzed rodent incisors of the Castelnou 3 cave (n = 5), located in the Roussillon Plain at low elevation (~200 m), which preserved sediments deposited near the shoreline and attributed to the late Miocene (~6.5 Ma) by biostratigraphic approach 50 (Fig. 1). We also sampled fossils from late Miocene alluvial to lacustrine deposits of the Can Villela section of the Cerdanya extensional Basin 41 (Fig. 1). These deposits have been attributed by magnetostratigraphy and biostratigraphy to Chron C3An.2n or C3An.2n, i.e. 6.5 or 6.1 Ma 41 . Teeth from the species Prolagus michauxi (Lagomorpha) (n = 4) and undetermined rodent incisors (n = 5) were analyzed (Fig. 2). This outcrop also yielded charophyte oogonia (freshwater green algae) of the species Lychnothamnus barbatus (n = 8). Oogonia are the female reproductive organs and they are preserved as small calcitic spheres biomineralized in small lakes or ponds during the warmer weeks. Their δ 18 O allow constraining summer freshwater temperatures 51,52 . We also analyzed gastropods from the family Clausilidae (n = 3), which are small terrestrial gastropods frequently observed around the Mediterranean Sea. We also obtained land snails from the genus Testacella (n = 5), corresponding to small slugs living in soils with a reduced shell located at the posterior end of their bodies. Description of the sampling sites and photographs of the samples analyzed in this work are provided in Supplementary Dataset 2.
Uncertainties are provided both for paleoelevation and paleotemperature estimations. For paleoelevation values, we took into account the standard deviation of the δ 18 O PO4 values obtained from the analysis of the rodent teeth and the uncertainty related to the modern isotope lapse rate 30 . Concerning the paleotemperature values, we consider the standard deviation of the mean of all δ 18 O values for a given species.
Geochemical analyses. Mammal teeth stable isotope analyses were performed at the Biogéosciences Laboratory of the University of Burgundy (Dijon, France). The teeth were ultrasonically cleaned and residual sediment was removed with a Dremel © tool. The teeth were crushed into powder in an agate mortar and pestle, and aliquots of powdered apatite (1 mg) were dissolved in nitric acid and chemically converted to Ag 3 PO 4 using the method described by 53 . Oxygen isotope ratios were measured on CO using a High Temperature Pyrolysis Analyzer (Elementar Pyrocube) connected online to an Elementar Isoprime mass spectrometer. All δ 18 O values are reported in per mil relative to V-SMOW (Vienna Standard Mean Ocean Water) by attributing a value of 21.7‰ to NBS120c 54 . Accuracy and reproducibility (≤±0.3‰, 2 σ) were monitored by multiple analyses of Ag 3 PO 4 from NBS120c.
Charophyte oogonia and gastropods were measured at the Institut des Science de la Terre de Paris (ISTeP, Sorbonne University, Paris, France). Each oogonium was observed and crushed under a binocular glass to prevent any recrystallization or sedimentary filling. Gastropod shell preservation was tested by X-Ray diffraction. Each individual carbonate powder sample (80 µg) was reacted with a 100% anhydric orthophosphoric acid at 70 °C in a Kiel IV carbonate device. Stable isotope analyses were performed on a DELTA V mass spectrometer. Isotope values are reported in conventional delta (δ) notation relative to the Vienna Peedee Belemnite (VPSB) standard. We used an internal standard (marble) calibrated to the international standard NBS-19. Precision is ±0.1‰ for δ 18 O. Geochemical results are provided in Supplementary Dataset 1.