Integration of stress indicator data, gravity data, crustal kinematics data, and analysis of topography and recent vertical motions demonstrates the occurrence of consistently oriented spatial patterns of large-scale Alpine to recent intraplate deformation in Iberia. The inferred upper crustal and lithospheric deformation patterns and the timing of the associated expressions at or near the surface support the existence of a close coupling with plate boundary processes operating at the margins of Iberia. Patterns of lithosphere and upper crustal folds are oriented perpendicular to the main axis of present-day intraplate compression in Iberia inferred from structural analysis of stress indicator data and focal mechanism solutions. These findings suggest the presence of lithospheric folds, with wavelengths compatible with theoretical predictions of folding wavelengths of Variscan lithosphere. Stress-induced intraplate deformation set up by plate interactions is compatible with indications for the absence of present-day deep mantle-lithosphere interactions inferred from seismic tomography.
 In this paper we integrate new data on the Iberian stress field and the record of recent vertical motions. We demonstrate through numerical modeling that lithosphere folding could be a major contributor to the current anomalous topography of the Iberian continent. The focus of the present paper is on the Late Neogene - recent intraplate deformation of Iberia. As pointed out recently by several authors [Vegas et al., 1990; Van Wees et al., 1996; Andeweg, 2002], large-scale intraplate deformation of Iberia has also occurred from Early Tertiary onward. In the eastern part of the Iberian Peninsula the structural grain of the mechanical coupling with the Pyrenean orogeny is still recognizable with indication for small-scale lithosphere folding [Verges et al., 1998; Waltham et al., 2000].
 The subsequent superposition of the intraplate deformation induced by the Betic orogeny (see Figure 1) has resulted in a complex interference pattern of deformation in the northeastern segment of Iberia being resolved by current field studies [Andeweg, 2002] and geothermochronology [Juez Larre and Andriessen, 2002]. More constraints are needed on this area to model the interference patterns. As in the central and western part of Iberia the relation between the Betic orogeny and induced intraplate deformation is more straightforward, this area forms the focus of our recent modeling effort.
 This succession of polyphase deformation is an important factor in understanding the Late Neogene to present-day record of intraplate deformation. We have incorporated the large-scale effects of Early Tertiary tectonics in the lithospheric configuration adopted in the modeling of the response of Iberia to recent plate reorganizations.
2. Iberia: Record of Intraplate Deformation in the Absence of Mantle Dynamics?
 Seismic tomography data of the western Mediterranean [Wortel and Spakman, 2000] provide evidence for a major Neogene retreat of plate subduction in an eastward direction, essentially leaving the main part of the Iberian Peninsula absent to intensive deep mantle-lithosphere interaction with a remnant of deep seismicity only restricted to the southeastern margin of Iberia [Blanco and Spakman, 1993]. At first sight this appears incompatible with the widespread occurrence of intraplate deformation continuing up to present day.
 A first-order question is whether the above mantle processes are overprinted by a contribution from intralithospheric mechanisms. Below we demonstrate that recent evidence on the stress field and the gravity data of Iberia support an important role of continental lithosphere folding in the Late Cenozoic to recent tectonic evolution.
3. Constraints on Vertical Motions
 Following the advent of quantitative subsidence analysis, a large number of studies have addressed the record of vertical motions of different basins in the Iberian continent. Most of these studies have been focusing on either the Mesozoic record of central Atlantic breakup within areas such as the Iberian Chain [Van Wees et al., 1998; Casas and Salas, 1993], the Lusitanian Basin [Stapel et al., 1996] or the pre-Betic [Peper and Cloetingh, 1992] or on the marine stages of flexural basins related to Alpine thrust belt emplacement in the Betic foreland [Van der Beek and Cloetingh, 1992; Garcia-Castellanos et al., 2002] and the Ebro Basin [Gaspar-Escribano et al., 2001]. Oligocene and Miocene deformation left a major imprint in the eastern margin of Iberia, at first sight not affecting the Variscan core of the peninsula. Compared to the large amount of work focusing on the earlier record, the attention paid so far on the tectonic controls of more recent vertical motions have been relatively minor. Backstripping analysis and forward modeling of the record of the Neogene pull-apart basins of eastern Iberia (Figure 3a) has, however, revealed the existence of a large Pliocene-Quaternary uplift of a typical magnitude of 500 m over a large part of Iberia [Cloetingh et al., 1992; Janssen et al., 1993; Docherty and Banda, 1995]. Quaternary uplift in the Betic foreland can be quantified by Late Miocene marine sediments that are elevated to more than 1200 m in several sierras, related to the continuous convergence between the African and Eurasian plates [Andeweg and Cloetingh, 2001]. Similarly, fission track studies have been able to assess the record of differential vertical motions for a number of areas in the Iberian Peninsula, including the Betics [Zeck et al., 1992], the southeastern margin of Iberia [Stapel, 1999], the Western margin of Iberia [Stapel, 1999], the Pyrenees [Fitzgerald et al., 1999] and the Spanish Central System [De Bruijne and Andriessen, 2000, 2002]. These studies have revealed a long and complex record of polyphase deformation of Iberia with a concentration of major rift-related early Mesozoic thermal rejuvenation concentrated at the western and eastern margins of Iberia. Apatite fission track studies [Stapel, 1999; De Bruijne and Andriessen, 2000, 2002] have also demonstrated a simultaneously occurring rapid post-Miocene cooling phase concentrated in structural highs in both the western margin of Iberia as well as the Spanish Central System (SCS) (Figures 3b and 3c).
 In the central part of Iberia the topography was around sea level in Late Cretaceous time, prior to the onset of north-south oriented compressional stress from the Pyrenean collision, which was leading to local deformation in the area, accompanied by a first phase of accelerated cooling around Late Eocene to Early Oligocene times (around 36 Ma). This was followed by a phase of significant accelerated cooling in central Spain around Middle Miocene times (15 Ma), followed by a very pronounced phase of accelerated cooling from Early Pliocene (5 Ma) onward [De Bruijne and Andriessen, 2000] (also see Figure 3b). The largest amount of cooling occurred in the Sierra de Guadarrama (eastern SCS) during the Pliocene (Figure 3b), and has been shown to be the result of a major phase of up to 6 km Pliocene uplift accompanied by erosion [De Bruijne and Andriessen, 2000; de Bruijne and Andriessen, 2002]. These authors relate this to the far-field effect of the ongoing Betic compression. Most denudation in the Sierra de Gredos (western SCS) occurred from the Middle Eocene to the Lower Miocene and can be related to a N-S stress field induced by the Pyrenean compression. The difference in trend of the preexisting lineaments (E-W for Gredos; NE-SW for Guadarrama) explains the difference in the relative importance of these far-field stresses on vertical motions in the area [De Bruijne and Andriessen, 2002]. This idea is supported by a combined study of structural elements and sedimentary sequences along the border between the NE-SCS and the Madrid Basin [Andeweg, 2002].
 Thermal modeling of the fission track data supports a scenario of enhanced simultaneous uplift and erosion from Pliocene to present day, providing independent support for an ongoing Betic compression [de Bruijne and Andriessen, 2002]. Other scenarios, including a mid-Miocene uplift phase creating a mountain range eroded during Pliocene due to a change of drainage pattern, are incompatible with the observed accelerated cooling in the area during the Pliocene and Quaternary [de Bruijne and Andriessen, 2002].
 Recent data from geodetic leveling and VLBI laser ranging [Rutigliano et al., 2000] point to current vertical uplift of the Spanish Central System of the order of 1 mm/yr. The above findings suggest that a long-lasting and still active tectonic control is affecting the topography of Iberia. The first indications from GPS networks in the Iberian Peninsula [Fernandes et al., 2000; Castellote et al., 2000] are pointing to consistently oriented horizontal motions, pointing to the question whether these recent vertical motions are a byproduct of continuing crustal shortening in the peninsula [Andeweg and Cloetingh, 2001], amplifying preexisting differential topography.
 In line with the focus of this paper on large-scale intraplate deformation we refrain from a detailed discussion of structural features on field scale and spatial dimensions at kilometer scale not addressed by the modeling.
 In this context it should be noted that the focus of this paper is on the far-field effects of the Betic collision on intraplate deformation and stress propagation in the Iberian micro continent. In this context we explicitly do not address the internal structuring of the Betic system itself (see, e.g., Ziegler et al. [1998, 2002] for a general discussion of the fine structure of the mechanical coupling between orogen and foreland lithosphere).
4. Stress Indicator Data and Present-Day Stress Regime of Iberia
 Being close to active plate boundaries, the southern margin of Iberia is characterized by a high level of seismicity. The seismicity, however, is not restricted to this area (see Figure 4a) as manifested by a high level of neotectonics in the western margin of Iberia, including the Tajo Valley and the Galicia margin. The same can be said about the central part of Iberia, with significant seismicity detected in the Madrid basin. Therefore it appears that the seismicity, which is concentrated along preexisting structural weakness zones, exhibits a typical intraplate type of lithosphere deformation. A large number of focal mechanism studies have been carried out in the last few years as well as a considerable number of breakout studies in basins in the peninsula explored for hydrocarbons. At the same time structural field studies, till recent times primarily focusing on the Variscan structures in the northern and northwestern peninsula, are now increasingly addressing the record of neotectonics in these areas. Recently, newly available stress indicator data for Iberia [Ribeiro et al., 1996; Jurado and Mueller, 1997; de Vicente et al., 1996; Herraiz et al., 1996; SIGMA, 1998; Borges et al., 2001] have filled the gap in information on the present-day state of stress of data of the Iberian Peninsula, as is apparent from the database for the World Stress Map [Zoback, 1992].
 Inspection of Figure 4b demonstrates the presence of a consistently oriented intraplate stress field in South and Central Iberia, with variations in orientation toward the west and northeast. The stresses set up by the collision between Africa and Eurasia, as well as the ridge push forces induced by spreading in the Atlantic appear to propagate into the interior of Iberia, producing a fanning pattern of stress in the Iberian Peninsula. Detailed field studies of Middle Miocene to Quaternary tectonics [de Vicente et al., 1996] and seismic activity have demonstrated an active phase of inversion in the Madrid Basin. The stress field in the neotectonic period in this region has been relatively consistent with regard to the N140–160E direction of maximum horizontal compression direction (Figure 4b). As proposed by Andeweg et al. , the existence of contemporaneous parallel compression and extension in areas as the Madrid Basin [de Vicente et al., 1996] and the Betic Cordillera [SIGMA, 1998] might be the result of folding in the upper crust. More detailed studies are presently carried out to get a better understanding of the relation between focal depth, focal mechanism and location.
5. Present-Day Topography and Gravity
 The topography of Iberia displays a high elevation of the peninsula with a succession of highs and lows trending roughly NE-SW in the interior (see Figure 5a). Inspection of the topographic map (Figure 5a) shows that the topography of a large part of Iberia (the area west of 357° and north of 37°) is characterized on a first-order scale by an alternating series of E-W to NE-SW trending mountain ranges and river basins. From northwest to southeast this series consists of the Miño river, the Cantabrian/Leon mountains, the Duero-Douro river basin, the Central System, the Tajo river basin, the Toledo mountains, the Guadiana River and the Sierra Morena. The topographic highs and lows are oriented perpendicular to the main axis of present-day intraplate compression in Iberia (Figure 4b).
 Similar trends can be observed in the Bouguer gravity map (Figure 5b). As discussed above, the lithosphere of Iberia, although largely of Variscan age, has been subject to subsequent intraplate deformation. The orientation of the anomalously high areas and the magnitude of the post-Miocene vertical motions suggest that the most recent phase of intraplate deformation has exerted a first-order control on the present topographic configuration. The alternating series of topographic highs and lows is more clearly demonstrated in Figure 6a, which shows three topography and Bouguer gravity anomaly profiles transecting Iberia (locations are shown in Figures 5a and 5b). The profiles show a close relationship between magnitude of the gravity anomalies and the makeup of the lithospheric and crustal structure of Iberia.
 Variations in Bouguer gravity anomalies reflect the long wavelength variations of the geometry of deep intralithospheric density interfaces, such as the Moho boundary [Cloetingh and Burov, 1996]. The wavelength of the Bouguer gravity anomaly thus mirrors the spatial scale of mantle lithosphere deformation. Consequently, the existence of a correlation between periodic long-wavelength Bouguer gravity anomaly and topography indicates the presence of whole-crustal and mantle lithosphere folding, whereas the lack of such correlation is in support of decoupled folding [Cloetingh et al., 1999].
 Results of a spectral analysis reveal the presence of a dominant wavelength of 200 (±50) km in both the topographic and gravity undulations along profiles A and B (Figure 6b). The power spectra for the gravity also show a large wavelength of at least 500 km, which probably reflects the high average elevated base level of Iberia, with a steep contrast at the borders of Iberia between elevated continental topography and deep oceanic basins. There is no conclusive evidence from seismic tomography for a perturbed upper mantle under Iberia [Wortel and Spakman, 2000]. A key question therefore remains whether the high average elevation of Iberia is due to a deep-seated asthenospheric low density source associated with large-scale thermal anomaly reflected in this very long wavelength gravity signal (wavelength longer than 500 km).
 The presence of dominant long-wavelength undulations in both topography and gravity has been observed in other intracontinental areas around the world, for instance in central Asia [Burov and Molnar, 1998] and in Brittany [Bonnet et al., 2000], and has been explained by these authors as stress induced folding of the continental lithosphere. This folding mode of intracontinental deformation appears to be more important in the large-scale deformation of continental lithosphere than till hitherto thought [Cloetingh et al., 1999].
6. Numerical Models for Lithospheric Folding
 The hybrid (finite element/finite differences) explicit code Paravoz [Poliakov et al., 1993] in the version modified by Burov et al.  and Burov and Poliakov  used in this study allows for direct implementation of various possible geodynamic scenarios. We can, for example, directly introduce lithological and rheological structures derived from seismic and gravity profiles and from thermal and rock mechanics data, using known horizontal convergence rates as the boundary conditions (Figure 7).
 The code solves the Newtonian equations of motion in a continuum mechanics formulation coupled with heat transport and surface erosion equations:
where vi are respectively components of the velocity vector and v is the velocity tensor, σij is stress tensor components, xj are coordinate vector components, gj are the components of the gravity vector, and ρ is the density. T is temperature, the parameter Cp is the specific heat, k is the thermal conductivity tensor, H is the radiogenic heat production per unit volume (here we use the commonly inferred values adopted, for example, as given by Burov et al. ), ke is the coefficient of erosion, and h is the surface topography. The Lagrangian mesh moves with the material, and at each time step the new positions of the mesh grid nodes are calculated from the current velocity field and updated in large strain mode accounting for stress axis rotation. Paravoz can handle rheologically complex behaviors, including localization and propagation of non-predefined faults (shear bands), power law creep and various kinds of strain softening and work hardening behaviors. The modified version of Paravoz allows the incorporation of the erosion/sedimentation models employed by Burov and Cloetingh , rheological and lithological models similar to that of Burov and Cloetingh , and an initial temperature field for continental lithosphere [Burov and Diament, 1995].
6.2. Rheology and Physical Properties
 Brittle-elasto-ductile nonlinear rheology is assumed for all materials (Table 1). The crustal and sedimentary segment of the lithosphere is represented by a quartz-dominated lithology; the mantle lithosphere is represented by an olivine-dominated lithology. The rheological parameters (elastic moduli, ductile creep material constants, activation energies and power exponents) used here are the same as given by Burov et al. . The brittle part is approximated by Mohr-Coulomb plasticity with friction angle 30° and cohesion of 20 MPa for all materials except sediments, which have smaller cohesion of 5 MPa. The densities are 2300, 2700, 2800, 3330, and 3250 kg/m3 for the sediment, upper crust, lower crust, mantle and asthenosphere, respectively. The thermal conductivities are 1.6 W m−1 K−1, 2.5 W m−1 K−1, 2 W m−1 K−1, 3.5 W m−1 K−1, 3.5 W m−1 K−1 for the sediment, upper crust, lower crust, mantle, and asthenosphere, respectively. The thermal expansion coefficient used is 3.1 × 10−5 K−1. These values are averages of the range of values presented in a compilation by Turcotte and Schubert .
Table 1. Creep Parameters for Crustal and Upper Mantle Rocks and Mineralsa
 We used linear diffusion erosion for the short-range surface processes and flat deposition outside the elevated topography range [Avouac and Burov, 1996]. The adopted coefficient of erosion, ke, is scale-dependent and varies from 0 to 8000 m2/yr to keep the basins filled with sediments.
6.4. Boundary Conditions
 The lateral boundary conditions are prescribed in velocities. We assume free surface on the top and Winkler restoring forces on the bottom of the model. The Winkler restoring force is equivalent to isostatic buoyancy forces used for lithospheric scale models to simulate lithosphere-asthenosphere interactions. It implies a vertical restoring force proportional to vertical displacement of the lower boundary of the model. The coefficient of the proportionality is given by the isostatic condition for the density contrast at the bottom interface of the model. In numerical models the Winkler forces are implemented by a series of ideal elastic springs attached to the bottom of the model. At the upper surface of the model a free surface boundary condition is implemented.
6.5. Setup of Numerical Experiments
 The initial thermal structure was imposed using a conventional thermal distribution for a specified thermotectonic age [Burov et al., 1998]. We have carried out a set of numerical experiments employing representative scenarios corresponding to two end member thermal structures of the Iberian lithosphere. The “hot” end-member case has a thermal age of 120 Ma, corresponding to Mesozoic rifting of Iberia [Van Wees et al., 1998], and has a thermal thickness of 250 km. The “cold” end-member case has a thermal age of 350 Ma, corresponding to Variscan deformation of Iberia [Banda, 1988; Vegas and Banda, 1982; Ordonez Casado et al., 2001; Simancas et al., 2001], and has a thermal thickness of 250 km. The concept of thermal age of continental lithosphere is explained by Burov and Diament  and Cloetingh and Burov . In our experiments we used the same set of rheological and thermal parameters as in the studies by Burov et al.  and Cloetingh et al. . Parameters adopted for brittle-elasto-ductile rheology including quartz-rich upper and diabase lower crust and olivine-controlled mantle are specified in Table 1. The shortening rates used in the numerical experiments correspond to the values commonly inferred from plate motion and plate reconstruction studies for the area (4 mm/yr) [see, e.g., Dewey et al., 1989; DeMets et al., 1994]. The initial crustal thickness was 32 km with 16 km thick upper crust and 16 km thick lower crust. The length of the numerical box was 1500 × 120 km, or 350 × 30 grid elements (Figure 7).
 The goal of the experiments is not to reproduce the fine structure of the present-day topography, but to investigate some of the basic controls on the development of the Neogene to present topography of Iberia. This implies in particular to the interplay of large-scale intraplate compression and surface erosion in Iberia.
7. Far-Field Intraplate Deformation in Iberia: Modeling Results
7.1. The 350 Ma Thermotectonic Age, Symmetric Compression, Erosion
 The first set of numerical experiments replicates the geotectonic situation for the westernmost profile experiencing the combined effects from Africa-Europe collision on the south and the Atlantic ridge push in the northwestern segment of the cross section. The models incorporate symmetric shortening, during 20 Myr, of 350 Myr old lithosphere subjected to simultaneous erosion (∼3 km of erosion during the model period, which roughly corresponds to the fission track data [Stapel, 1999; De Bruijne and Andriessen, 2002]. The results (Figure 8a) demonstrate the development of two superimposed large deformational wavelengths, the largest of which is about 350–420 km and the shortest is about 180–200 km. Simultaneously one very short wavelength of 50–60 km develops, apparently resulting from folding and faulting in the uppermost crust, with a fault spacing largely controlled by the thickness of the brittle layer. The very long wavelengths of 350–420 km are not very well expressed in the surface topography as the latter is strongly modified by simultaneous erosion and sedimentation, but are very clearly seen in the displacement velocity patterns and in the geometry of the subsurface layers. The predicted topography patterns with wavelengths of 180–200 km resemble those actually observed (Figures 5 and 6). The presence of two different large wavelengths of deformation indicates partly coupled and partly decoupled folding (described by Cloetingh et al. ). In most cases it is difficult to discriminate between folding of the lower crust and upper crust, as their folding wavelengths are quite comparable. Furthermore, the lower crustal folding will not be well expressed in the gravity signal due to insufficient density contrast with the upper crust. We therefore only discriminate between crustal and mantle folding. A partial crustal-mantle coupling can be observed in some plate segments, due to a strong deformation of the upper crust, which is transmitted to the mantle layer. At the same time, the crust-mantle decoupling results in anti-phase folding of the crustal and mantle lithosphere, in which areas of maximum uplift of the crustal basement are laterally shifted with respect to those of maximum mantle depression. Although this geometrically resembles boudinage, it is produced by contra-phase folding of two competent layers. This geometry might lead to incorrect estimates of lithospheric strength. The induced amplitude of the mantle depressions roughly correspond to those expected for local isostatic compensation of the overlying topography, traditionally interpreted in terms of low lithospheric strength. For the short-wavelength upper crustal mode of deformation (wavelength of 50–60 km) the numerical resolution of the problem is just about the limit of resolution for this wavelength (4 elements per crustal layer). To address the resolution problem, we have carried out a number of numerical simulations with double resolution (Figure 8b). These basically reproduce the same wavelength and allow to resolve upper crustal faulting occurring at about 50 and 200 km spacing, consistent with observations (see Figure 6), and high strain rate zones at the crustal/mantle and mantle/asthenosphere boundaries, demonstrating decoupling of the competent lithospheric layers from the weak ductile matrix.
 Note that the induced folding patterns display some asymmetry since folding itself is an unstable process. Folding of nonlinear layers is even more unstable, especially at large strains. Such lateral variations of both folding amplitude and wavelength were previously demonstrated in a number of analytical and numerical studies [e.g., Hunt et al., 1996; Cloetingh et al., 1999].
7.2. The 350 Ma Thermotectonic Age, Symmetric Compression, No Erosion
 As was pointed out by Avouac and Burov , syndeformational erosion coupled with tectonic processes may accelerate localization and then stabilize the evolution of the tectonic deformation. However, if the erosion is too rapid, it may just wipe out all newly created topography and influence the wavelength of folding. For this reason, a series of experiments without erosion were implemented to test the relative importance of simultaneously operating surface processes during intraplate deformation in the Iberian lithosphere. In particular, to assess the importance of the interplay of folding and surface processes, we have repeated the previous experiments without erosion (Figure 8c). A comparison of the results demonstrates that erosion tends to suppress shorter (50–60 km) wavelengths of deformation, but leaves intact the longer wavelengths. It can also be observed that the erosion accelerates vertical uplift rates. Both numerical experiments produce compatible vertical topography amplitudes on the same time scale. However, they differ in the sense that in the erosional case the erosion rate should be added to the surface uplift rate, thus yielding rock uplift rates several times higher than predicted in the absence of erosion.
7.3. The 350 Ma Thermotectonic Age, Asymmetric Compression, Erosion
 The second set of experiments with a compressional stress regime initiated at 50 Ma corresponds to the central and the eastern profiles, experiencing the influence from the push exerted by ridge-push and the closure of the Bay of Biscay (starting at 50 Ma) at the northern border of Iberia (Figure 8d). It is analogous to the previous experiments with the difference that the shortening is applied only on the northern side of the plate whereas the southern margin was kept fixed. The folding patterns demonstrate a modest lateral variation of the deformation wavelengths, which are slightly shorter in the south and longer in the north of Iberia.
7.4. The 350 Ma Thermotectonic Age, Asymmetric Compression, No Erosion
 This set of experiments repeats the previous one but with no erosion included (Figure 8e). Note that the short wavelength of deformation is well present in this case and that the surface geometry basically resembles some typical features of the Iberian topography. The longest folding wavelength appears to be the same as in the previous case.
7.5. The 120 Ma Thermotectonic Age, Symmetric Compression, Erosion
 The third set of experiments (Figure 9) considers a thermally younger lithosphere with a thermotectonic age (that is the time elapsed since the last major thermal event) of 120 Ma. At this time the eastern margin of Iberia underwent major extension and reheating [Van Wees et al., 1998]. At the same time stratigraphic studies of the Duero basin indicate that it was only mildly affected by this phase of reheating suggesting that the western part of Iberia did not underwent any major reheating at that time. Figure 9 shows that in this case the folding instability develops only in the cold upper crustal domain and that the deformation is concentrated at the boundaries of the plate. It is thus likely that such a weak rheological structure is not plausible in the particular tectonic configuration of Iberia [Faccenna et al., 2001].
7.6. Northward Increasing Lithospheric Strength (Thermotectonic Age Increases From 120 Ma to 350 Ma), Symmetric, Erosion
 A major question is whether the amount of shortening induced by the Africa-Europe collision has been primarily accommodated in the Alboran block characterized by relatively high heat flow [Fernandez et al., 1998] and major thermal rejuvenation associated with Upper Jurassic to Early Cretaceous rifting. Such an inferred weak zone could have a strong effect on stress propagation due to forces exerted at the southern boundary of this Alboran block on the adjacent stronger Iberian lithosphere. To investigate this scenario, we have constructed a model involving an important lateral variation in strength. These models (Figure 10) aim at testing the hypothesis of lateral variations in the thermal state of the lithosphere, as suggested by surface heat flux data indicating an increase of the heat flow to the south [Fernandez et al., 1998]. To account for this scenario, we introduced a lateral variation in the thermal field, which corresponds to a 120 Ma thermotectonic age in the 500 km long southern segment of the lithosphere and a 350 Ma thermotectonic age in the 1000 km long northern segment. The general basement deformation patterns are similar to those produced in the experiments with laterally homogeneous lithosphere, but at depth the model demonstrates more intensive mantle and crustal folding in the southern part (with wavelengths of about 320–350 km on the south and 420 km to the north). Mantle folding in the southern part of the Iberian plate results in localization of the deformation in a single down warped mega fold resembling subduction patterns inferred from analogue modeling [Faccenna et al., 2001].
8.1. Spatial Variations in Rheology
 The topographic and lithospheric structures predicted by the modeling in the case of Variscan-type lithosphere, are strongly similar to those actually observed. It should be realized that the data available (rheology data, heat flow) only allow for a first-order evaluation of the input rheological profiles and the thermal field. For this reason we have implemented a number of experiments in which we have varied the internal mechanical structure of the lithosphere from those corresponding to thermotectonic ages of 120 Ma to 350 Ma. Irrespective of the actual thermotectonic age of the lithosphere, it appears that the models for continental lithosphere of 350 Ma reproduce very well the overall behavior of the Iberian lithosphere. This result is additionally supported by previous estimates of 30 km for the effective elastic thickness (EET) obtained for the Ebro basin in northeastern Spain [Gaspar-Escribano et al., 2001], which also well corresponds to the theoretical prediction given by Burov and Diament  for 350 Ma lithosphere. For central Spain, Van Wees et al.  provide a lower bound estimate for the upper crustal EET of only 7 km. This value is somewhat less than the 12–15 km estimate for upper crustal EET, predicted from rheological profiles of the thermal and lithological structure adopted in this paper. Locally, crustal detachment accompanied by brittle deformation in the upper crust and ductile thickening of the lower crust could have played a role in the localization of uplift in the Spanish Central System [de Bruijne and Andriessen, 2002].
 As pointed out by Andeweg et al. , the spatial distribution of present-day seismicity and Quaternary deformation indicates a present amplification of the initial foreland-bulge of the Madrid basin-Spanish Central System due to intraplate compression. Similar conclusions were reached by Garcia-Castellanos et al.  on the basis of an integrated analysis of the Betic foreland basin. These authors pointed out that a major amplification of the flexural foreland loading by intraplate compression was required to yield the observed history of basement stress and basement deflection evolution and the adjacent uplift in the Sierra Morena area. The obtained low values of EET estimates in this area (EET = 18 km) are interpreted in terms of flexural decoupling of crust and mantle and/or stress-induced weakening due to horizontal compression. Flexural decoupling is also inferred from structural restorations in the area [de Vicente et al., 1996].
 The above observations suggest that the wavelength of deformation increases to the north, which can be explained either by the proximity to the active plate boundaries (Figure 4), or by the hotter thermal state of the lithosphere in the south (Figure 2).
 The models used imply basically laterally homogeneous crustal structure. There is no technical difficulty to introduce various lateral variations and preexisting faults in the model [see also Beekman et al., 1996; Gerbault et al., 1999]. As was shown by Cloetingh et al. , preexisting crustal faulting and segmentation in general do not have a significant affect on the developing folding wavelength unless these faults crosscut the entire crust and mantle lithosphere.
8.2. Interplay Between Surface Transport and Vertical Motions
 As pointed out above, the thermal modeling of fission track data favors a Pliocene phase of simultaneous uplift and erosion rather than a Miocene uplift phase followed by Pliocene erosion [de Bruijne and Andriessen, 2002]. This is important, as the erosion following the creation of lithospheric folds will redistribute mass at the surface into the topographic lows, thus enhancing the flexure of the underlying lithosphere. In this context, it should be noted that our modeling approach is by its nature 2-D, whereas drainage patterns characteristic of some of the major rivers of Iberia, including the Tajo, the Duero and the Guadiana rivers, trending in a NE-SW direction, remove sediments out of the plane of cross section toward the coastal areas. A noteworthy feature (see Figure 5a) is the progressive development of the drainage systems toward the Betics, resulting in a fully developed drainage system in the Guadalquivir River, whereas the Duero River only very recently crosscut the coastal topography, establishing an opening of the Duero Basin toward the Atlantic. Numerical experiments of drainage network evolution [Garcia-Castellanos, 2002] demonstrate that surface transport processes effectively enhance the tectonically induced large-scale continental topography.
 In 2-D (cross section) modeling studies including the one presented above, erosion is often approached as a diffusive process. This technique permits to reproduce the basic effects of surface transport, such as transport from elevated areas to relief depressions and the eventual peniplanation of an initially rough topography. However, diffusivity is not a process-based approach for large-scale transport. In the temperate climatic conditions of Iberia during most of the Tertiary, most of this surface mass transport is undertaken by rivers. River catchments have typically strong 3D asymmetry since their drainage is hierarchically organized to collect the waters and sediment from a widespread area into a single outlet. This property implies that whereas erosion is distributed along a large area of the catchment's headwaters, deposition takes place relatively localized along the lower riverbed and around the river mouth in the Atlantic Ocean and possible intermediate lakes.
 The pre-Alpine paleogeography of the Iberian Massif corresponds to a relatively flat topography resulting from the erosional planation during the long period elapsing between the Hercynian and the Alpine orogenies. Vertical movements modify river profiles and the distribution of drainage basins [e.g., Kooi and Beaumont, 1996; Bonnet et al., 1998]. Despite the intrinsic nonlinear nature of drainage networks, moderate vertical movements seem also to be capable to organize the drainage patterns in relatively flat areas where drainage is not well-organized or incised [Garcia-Castellanos, 2002].
 In the studied region of central and western Iberia, drainage is mostly oriented westward, implying that whereas widespread erosion takes place in the whole region, deposition is restricted to the westernmost edge of the Iberian Peninsula: at the lower part of the valleys and offshore in the Atlantic Ocean. However, the diffusive models used above assume erosion at high areas and deposition at topographic minima along the studied cross sections. Therefore models adopting diffusive transport should be considered as a first-order approximation.
8.3. Lithosphere Folding and Drainage Patterns
 To further evaluate the interplay between surface transport and lithospheric folding, we simulate fluvial transport via a drainage network in which runoff water flows along the maximum slope with a sediment transport capacity proportional to slope and water discharge. For the implementation of this approach, which intrinsically predicts planform hierarchical organization of drainage networks, we use the relationships by Beaumont et al. , subsequently adopted in several studies [e.g., Kooi and Beaumont, 1996; Braun and Sambridge, 1997; Van der Beek and Braun, 1998]. It should be noted that there is an ongoing discussion on the optimal empirical relationships relating the actual amount of erosion/deposition to slope and water discharge [e.g., Willgose et al., 1991; Howard et al., 1994; Tucker and Slingerland, 1995; Whipple and Tucker, 1999]. This discussion is of secondary importance for this study, as our analysis focuses on the first-order features of the interplay between fluvial transport and folding, rather than on the properties of fluvial transport themselves. The model applied here incorporates additionally the deposition in topographic minima (lakes) and in the Atlantic Ocean, which allows for modeling closed transport systems where virtually all eroded material is deposited within the model (further details are given by Garcia-Castellanos ).
 Folding is calculated as the response of a thin homogeneous visco-elastic 2-D (plan form) thin-plate to tectonic loading (horizontal) and surface mass redistribution (vertical load). The equation governing this plate is derived by applying the principle of correspondence between elasticity and visco-elasticity [Lambeck, 1983] to the equivalent elastic equation [e.g., Van Wees and Cloetingh, 1994]. It should be noted that this approach of the lithospheric rheological behavior is less sophisticated than the ones governing the 2-D (cross sectional) models presented above. Whereas the 2-D models have the full capacity to incorporate a true depth-varying elasto-ductile-viscous rheology, the planform models are restricted to an approximate depth-independent visco-elastic rheology. The planform models, however, are superior to the 2-D models in terms of the treatment of surface transport as well as their capability to address the 3D nature of the lithospheric deformation of Iberia. The cross section and the planform models therefore should be considered as complementary, thus allowing to investigate both the main characteristics of the depth-dependent response of the lithosphere as well as the plan view of the first-order pattern of intraplate deformation and topography evolution.
 We have calculated the 3-D response to lithosphere shortening adopting an effective elastic thickness (EET) of 30 km. This value corresponds to a mechanically coupled crust and mantle lithosphere with a thermotectonic age of 350 Ma, and is consistent with results from a flexural analysis of the Ebro Basin in northeastern Spain [Gaspar-Escribano et al., 2001]. A viscous relaxation time of 1.2 Myr is adopted from a recent flexural analysis of the Guadalquivir Basin (southern edge of the Iberian Massif) [Garcia-Castellanos et al., 2002]. The initial configuration of the synthetic models is a flat square continent elevated at 400 m above sea level with a random perturbation between −10 and +10 m. The ocean surrounding this square region is 2000 m deep. The deflection in the boundaries of the model is set to zero. We assume a runoff (water going to the drainage system) distribution of 200 + 300*altitude (m) mm yr−1 and adopt standard values for the transport parameters [Kooi and Beaumont, 1994, 1996]: constant of river transport capacity Kf = 0.01; length scale of fluvial erosion Lfe = 120 km (60 km for sediment erosion); length scale of fluvial sedimentation Lfs = 25 km.
Figure 11a shows the drainage networks resulting after a period of 12 Ma of N-S directed horizontal compression. For initial random perturbations of the topography lower than ∼20 m, the resultant drainage pattern is clearly controlled by a ∼350 km wavelength of lithospheric folding. A reduction to an EET of 18 km, closer to the values found in the Guadalquivir Basin [Garcia-Castellanos et al., 2002] reduces this characteristic wavelength to 220 km. During the initial stages of compression, the peniplane geometry limits erosion and deposition only to the vicinity of the shoreline, whereas the interior of the continent remains almost unperturbed. The vertical movements associated with this initial mass transport near the shoreline play a role as the necessary perturbations that trigger folding. As a consequence, maximum uplift occurs along the north and south shores, as they are perpendicular to the compression axis.
 Inversely, the organization of drainage has also important effects on the vertical motions. This is illustrated by a comparison of Figure 11a with an identical model where transport is diffusive (Figure 11b), adopting a value of ke = 33.0 m2/yr. Although we chose the diffusive constant to produce a total transport of sediments similar to the previous model (71.9 × 103 km3), the hierarchical organization of the drainage pattern has a strong influence on the spatial redistribution of mass. Deposition is localized near the main river mouths and in phase with the vertical motions, since drainage is forced along the E-W trending folding minima (Figure 11a). It therefore accentuates the folding pattern off the west shore, at x < −420 km. Instead, a diffusive approach to surface transport does not introduce relevant asymmetries in the x direction (Figure 11b) except for overprinting the folding pattern along the western Atlantic margin of Iberia. Therefore river transport is less efficient in amplifying vertical motions because erosion and sedimentation occur in distant and larger areas.
 As in the 2-D dynamic models presented above, the folding wavelength of Iberia is here related to the mechanical thickness of the plate. It is remarkable that the used elastic thickness derived independently for the Ebro Basin and Guadalquivir Basin successfully predict the range of observed wavelengths of folding of Iberia, thus revealing consistency between the flexural models of the northern and southern plate edge and the intraplate deformation model described here.
 In summary, transport along the river network contributes to enhance and localize vertical movements, and introduces lateral asymmetries in spatial patterns of erosion and sediment accumulation, as actually observed in Iberia. The study of paleoplanation surfaces has been often used to detect the presence of regional vertical movements potentially related to processes of crustal or lithospheric scale [e.g., Bonnet et al., 2000; Van Balen et al., 2000]. These types of studies may help significantly to further improve the understanding of present relief in Iberia.
8.4. Other Examples of Folding in Variscan Lithosphere
 An important factor in favor of the lithosphere folding scenario is a resemblance of the wavelength of deformation, thermotectonic age and the total amount of shortening to other known cases of continental lithospheric folding. A prominent example of folding can be found in the Western Goby area in central Asia with a thermotectonic age of 400 Ma. In this area, mantle and crustal wavelengths are 360 km and 50 km, respectively, with a shortening rate of ∼10 mm/yr and a total amount of shortening of 200–250 km during 10–15 Myr [Burov et al., 1998; Burov and Molnar, 1998].
 Another case of recently detected folding in Variscan lithosphere is the Armorican Massif of Brittany at the western margin of the Paris Basin [Bonnet et al., 1998, 2000]. The wavelength of the folds is 250 km, pointing to a lithospheric mantle control of the deformation. As pointed out by Bonnet et al. , the spatial pattern and the timing of the uplift inferred from river incision studies of Brittany is incompatible with a glacio-eustatic origin. These authors link the observed patterns of vertical motions in NW France to the present-day NW-SE oriented main axis of compression of the intraplate stress field of NW Europe. The stress-induced uplift pattern appears to control the amount of fluvial incision in the area as well as the location of the main drainage divides. The area located at the western margin of the Paris Basin and the Atlantic rifted margin of France has been subject to thermal rejuvenation during Mesozoic extension related to North Atlantic rifting and subsequent compressional intraplate deformation [Ziegler et al., 1995], also affecting the Paris Basin [Lefort and Agarwal, 1996]. Leveling studies of the area [LeNotre et al., 1999] also point to an ongoing deformation of the area.
 The inferred wavelengths for neotectonic lithosphere folds in Variscan lithosphere are consistent with the general relationships established between folding wavelength and thermotectonic age of the continental lithosphere, recognized in a number of folded areas of the globe [Cloetingh and Burov, 1996]. In Figure 12 we have plotted the wavelength of folding of Variscan lithosphere, inferred for Iberia and Brittany. In a number of other areas of continental lithosphere folding also smaller wavelength crustal folds have been detected, for example in central Asia [Nikishin et al., 1993; Cobbold et al., 1993]. This notion opens the perspective for a possible existence of a smaller, hitherto overlooked system of shorter wavelength upper crustal folds in Iberia.
 The following points summarize our conclusions.
The overall modeled patterns of lithosphere deformation are consistent with Alpine to recent reactivation of Hercynian lithosphere by development of compressional instabilities. It should be noted that because of slowly increasing growth rate, at low rates of shortening these instabilities take some time to develop important amplitudes (1 km). This time ranges from 5 to 10 Ma and depends on many factors such as rheology, structure, and geometry of the lithosphere and surface process activity.
It appears that partial coupling/decoupling between the crustal and mantle lithosphere may explain the existence of different wavelengths of deformation observed in Iberia. The modeling suggests that the observed short wavelengths (∼50 km) can partly result from folding of the upper crustal brittle layer, and partly from distributed faulting controlled by the effective thickness of the brittle crustal layer (Figure 8b).
Models that include weakness zones in the southern margin of Iberia suggest that shortening from the south in itself is not capable to induce large-scale propagation of deformation over the whole Iberian lithosphere. Models subjected to symmetric shortening, however, produce very realistic deformation patterns, in particular reproducing lateral variation in deformation.
The modeling suggests that shortening from the north might be more important than previously thought, not only for the Early Tertiary but also for the Late Neogene evolution of Iberia. This is consistent with the preservation of the earlier (50 Ma) folding in the relatively cold (thermotectonic age of 350 Ma) northern part of Iberia until the onset of the new folding stage at 20 Ma associated with Betics compression from the south.
Erosion appears to be a largely overlooked factor in the discussion of the intraplate lithosphere deformation in Iberia. It plays a major role in the attenuation of the amplitude of the shorter (50 km) wavelength of surface topography. Erosion also accelerates vertical movements on the newly created and preexisting faulted structures (Figure 8b) in all bandwidths. In turn, the self-organization of the drainage network during folding determines the asymmetric surface mass redistribution.
 Contribution NSG20011201 of the Netherlands Research School of Sedimentary Geology. Funding through the PRIOR Project (CSN-ENRESA-IGN) to G. de Vicente and funding of E. Burov through the INSU IT program are gratefully acknowledged. We thank Marlies ter Voorde and Gerco Stapel for helpful discussions on an early version of the manuscript. We are thankful to both reviewers and to the associate editor for helpful and in-depth comments on the paper. The numerical code used in this paper is a modification of the Paravoz code, which kernel is developed by A. Poliakov and Y. Podladchikov, who are acknowledged for helpful and fruitful discussions.