A 3-D seismic data set was acquired in summer 2010 in Hontomín, across an area of 36 km2 centred according to the target dome-shaped structure (Alcalde et al., 2013a) (Fig. 2). The acquired seismic data set was processed down to 1500 ms up to post-stack time migration (Alcalde et al., 2013b). Critical steps in processing were related to the existence of an unexpected velocity inversion near the surface, which decreased the general quality of the data. However, the final migrated volume provided clear images from the subsurface structures down to the anhydrite level (i.e. the bottom of the dome structure). Further details on the seismic data acquisition and processing can be found in Alcalde et al. (2013a, b, 2014).

The seismic data were integrated with well-log data from four hydrocarbon exploration boreholes (H1 to H4) and three shallow groundwater sampling wells (GW1 to GW3) (Fig. 2). Results indicate that the structure of the Hontomín dome includes eight different sedimentary packages, from Triassic to Cenozoic and four sets of faults. The target Mesozoic dome structure has a NW–SE orientation and is bounded by two major faults, the south (S) and the east (E) faults, showing vertical offsets of up to 450 and 250 m respectively. Another two sets of faults have been singled out: they trend N–S and E–W and have been related to extensional episodes that occurred during the opening of the Bay of Biscay. The dips of the dome flanks decrease upwards suggesting a protracted and discontinuous growth. The joint interpretation of the seismic and well log data allowed for inferring three evolutionary stages for the Hontomín structure (Alcalde et al., 2104): (1) the development of N–S trending faults as a consequence of differential loading around Triassic E–W faults could have allowed the development of the lowermost Jurassic units and probably produced forced folding of the Hontomín dome. (2) Later on, the development of E-trending faults during the opening of the Bay of Biscay is recorded by the accumulation of Purbeck deposits. Simultaneously, the migration of the Triassic salts would have further increased the dome's growth. (3) Finally, the Pyrenean orogeny caused the reactivation of the south and east faults and the final development of the dome structure.

Microgravity data were acquired between August and December 2010 by Implemental System Company. A high-density mesh covering an area of 4 × 4 km2 was recorded with measurements taken at every 100 m, which resulted in 1600 total measured points (Fig. 2). The measurements were carried out using a Scintrex CG-5 2008 gravity meter with a resolution of 0.001 mGal. The acquisition parameters ensured a resolution lower than 4 µGal, with over three measurements per grid point on average. Seventy-one of those gravimetric measurement points were concreted, along with the seven gravimetric stations used to calibrate the gravity-meter, delivering an accuracy of 0.5 cm. The positioning of the 1600 measured points was performed with a LEICA GPS9000 and has an accuracy of 1cm or less. The calibration of the topographic instrument was carried out with respect to a geodesic vertex located within the study area. This ensures that measurement points were within the high-quality range of the device.

The gravity data were provided without the effects of moon/sun/earth tides and instrumental drift, which were already corrected by the contractor. The data were then processed using Geosoft Oasis Montaj^™ to produce the complete Bouguer anomaly (BA) map of the area. This includes carrying out the Bouguer slab, free-air and terrain corrections (Blakely, 1995). The resulting Bouguer anomaly map is shown in Fig. 3. The reduction density used was 2400 kg m-3, although several values were tested to assess the better reduction density for performing the Bouguer slab correction. Anomaly maps obtained using densities of 2000, 2400 and 2600 kg m-3 were compared with the topography, being 2400 kg m-3 which showed less correlation with the relief.

Several procedures may be applied to Bouguer anomaly data in order to enhance features that may later on help in the interpretation. Accordingly, the following processes were carried out in the present data set: (1) separation of the long and short wavelength components, (2) calculation of the gravity derivatives (vertical and total horizontal derivatives, THD), (3) construction of two 2-D gravity models and (4) inversion of gravity data. To carry out a qualitative assessment of the source depth of the Bouguer anomaly data, the short and long wavelength number components had to be separated. This separation was performed by applying an upward continuation to the gravity field (Jacobsen, 1987) of up to 350 m. The filter was selected in order to minimize distortion and ringing effects (Fig. 4a). The regional component was then subtracted from the complete Bouguer anomaly to generate a residual anomaly map presented in Fig. 4b.

Edge enhancements are often used to aid in the potential field data interpretation (Verduzco et al., 2004). Accordingly, two sets of derivatives were performed to our data set aiming to highlight the gradient zones and accordingly the structural setting of the area. The vertical derivative (Fig. 4c) (Blakely and Simpson, 1986) and total horizontal derivative (THD, Fig. 4d) show a coherent result between them. Both maps highlight coincident areas with maximum gradient, that is, areas where the contacts between materials of different density exist. However, the vertical derivative also enhances those that are shallow.

Finally, two 2-D gravity profiles were extracted from the Bouguer anomaly map in order to produce a model of the structure of the area along them (see location in Fig. 3): Model 1, striking NE–SW and Model 2 striking almost E–W (Fig. 5). First, these models were computed with the assistance of the information provided by wells H1 and H2 but without the input of 3-D seismic data (Andrés, 2012) (Table 1). The models were then revisited to include the variations in layer depth and fault offset observed in the seismic model (Fig. 5a). It is worth highlighting that changes introduced to the models were not significant and provide a good match with the seismic model and 2-D sections derived from the inversion procedure (Fig. 6).

Gravity data can be inverted in order to obtain the topography of a desired layer (Oldenburg, 1974). However, inversion results are more reliable when the number of variables to invert decrease. Inverting just the topography of one layer can provide us with excellent results if we can preliminarily establish the density of every layer and the thickness of every density interval. The method utilized in this study is based on the Parker's algorithm (Blakely, 1996; Parker, 1972). It works in the frequency domain and is integrated in GM-SYS 3-D platform. Here, we inverted the microgravity data in order to obtain the geometry of the basement.

The large amount of information available in the Hontomín area has granted us the basis to build a well-constrained model that allows us to perform a successful inversion. Well data, seismic horizons and 2-D gravity models were used to construct an initial model. The inversion process, which was then applied, consisted of isolating all the gravity contributors, i.e. different sources (sediments and regional component) and calculating their gravity responses. That was then subtracted from the Bouguer gravity anomaly (BA) map (Fig. 3) and the residual was inverted.

Accordingly, building the gravity model for the inversion consisted of the following steps:

Calculation and removal from the BA map of the gravity anomaly generated by the stratigraphic succession below 0 m (Fig. 7a and c), assuming a constant density for each layer (Table 1).

The sedimentary succession was interpreted and modelled from the 3-D seismic cube generated by Alcalde et al. (2014). Eight layers were used to characterize the signal of the sediments. Three of them, corresponding to the Cenomanian–Maastrichtian, Early Cenomanian (Utrillas Fm.) and Aptian–Albian (Escucha Fm.) layers, lie above the sea level. They were used to compare the BA obtained using a reduction density of 2400 kg m-3 with that obtained using a model based on Alcalde et al. (2014). The latter map had a higher wavelength number component that needed to be filtered out, worsening the resolution of the final results and accordingly was discarded. The remaining five layers are below sea level and correspond, from shallower to deeper, to Late Jurassic to Early Cretaceous Purbeck, Dogger, marly Lias, Lias limestones and Lower Jurassic anhydrites.

The forward calculation of the sediment gravity anomaly was performed using a constant density (Table 1) for each of the five layers described above, including the topography, down to the top of the anhydrites. To solve the ambiguity generated by the unknown depth of the base of the anhydrites we used the constraints provided by the 2-D gravity models and the seismic profile shown in Fig. 5a. These data sets support the use of a constant thickness (100 m) layer of anhydrites. Furthermore, the seismic profile presented by Alcalde et al. (2014) shows a general homogeneity in the thickness of the layers for the study area. A bottom flat imaginary boundary for the Keuper rocks was used in order to avoid cutting the overlaying strata so no errors were carried into the inversion process. Finally, the long wavelength/deeper contribution to the BA was removed by using the simplest first-order polynomial.

Three cross sections (Fig. 6) were extracted from the results of the gravity inversion. These include the sedimentary layers down to the anhydrites and the basement. Two of them coincide with the location of the 2-D gravimetric models, NE–SW and WNW–ESE (Figs. 3, 5a, c) and the third one has a NW–SE direction (Fig. 3). The cross sections have been used to make a comparison with the previous 2-D models shown in Fig. 5 and to assess the basement structure crossing the main tectonic features that are also highlighted by the derivatives (Fig. 4c, d).

The Hontomín dome has been described as an extensional forced fold (Tavani et al., 2013) resulting from the extension and reactivation of Permo–Triassic normal faults affecting the basement. The presence of a detaching level represented by the Keuper Triassic salts produced the forced folding in the upper sedimentary cover by migration of these salts (Tavani et al., 2013; Carola et al., 2015). For the area of study, Carola et al. (2015) have suggested a thin-skinned configuration based on surface geology, well data and vintage seismic lines.

Here, we present a comprehensive model, from surface to basement, centred in the surroundings of the spot where the CO2 storage site is currently being developed. The new model suggests a local thick-skinned deformation style for the area with two major faults affecting the basement, namely the south and the east faults, already described by Alcalde et al. (2014). This model defines an area divided into three main blocks, south, centre-north-west and north-east.

The THD presented in Fig. 9 delineates high-gradient areas affecting the residual gravity data set of Fig. 7d. These zones are interpreted as faults (discontinuous lines in Fig. 9) and are compared with the ones described by Alcalde et al. (2014). Among these, one striking ∼ E–W stands out in the southern area and is interpreted as a major fault, coinciding with the south fault as defined by Alcalde et al. (2014). The gravity gradient of this fault is observed in most of the gravity maps produced in this paper, including the residual gravity map and derivatives shown in Fig. 4 as well as those in shown in Fig. 9, indicating that the south fault affects sediments and basement. Furthermore, the cross sections derived from the 3-D model (Fig. 6) show an offset in the basement of 150 m for the south fault. Another fault parallel to the south fault is shown in Fig. 5a, close to the thickest interval of Keuper evaporites, and features a normal offset of 200 m over the basement. These two faults combined create a downward displacement of the basement of about 350 m. The south fault is thought to be a branch of the right-lateral Ubierna Fault affecting the basement and all the stratigraphic succession and conditioning the structural setting of the area. This fault has been affected by two deformation stages: a Late Jurassic–Early Cretaceous extensional stage (Tavani and Muñoz, 2012) and a later Cenozoic compression related with the formation of the Pyrenees. The kinematics and sedimentary history of this fault were interpreted by Alcalde et al. (2014). They suggest a flower-like structure associated with the strike-slip movement of the Ubierna faults. The Jurassic–Lower Cretaceous succession shows a thicker sedimentary record to the NW of the south fault, suggesting a normal displacement of the hanging wall during this period.

Another fault interpreted from the map in Fig. 9 is located to the NE, strikes NW–SE and is correlated with the east fault of Alcalde et al. (2014). This fault affects the basement and the sedimentary succession up to the Cenozoic packages, which are not affected by it. The fault has two minor associated faults (Fig. 9) that strike in the same direction but have less extension. The vertical motion of this fault was also described by Alcalde et al. (2014). Here, a downward displacement of the SW block during the Jurassic period has been assigned by the information extracted from the exploration wells. The 3-D model created after the inversion of the gravity data shows a SW dipping fault with a normal sense of motion and offsets of around 400 m, which are comparable to the offset accumulated by the southern fault. The east fault, as interpreted in the seismic data set, is not as clear in the derivatives of the BA data (Fig. 4c and d), although it is visible and clear in the residual maps (Fig. 4b), in the basement depth maps (Fig. 8a) and in the THD performed over the final anomaly grid used for the inversion (Fig. 9). This might be due to the fact that it does not affect the entire sedimentary sequence. In general, the south and east faults seem to have been reactivated during the Jurassic–Lower Cretaceous extension, generating a sunken basement block.

Figure 9 also shows good agreement existing between the south and east faults as deduced from gravity and 3-D seismic data. Good correlation exists too between minor fractures associated with these faults, like an E–W fault located to the north of the south fault and interpreted from both data sets.

Finally, the two 2-D transects modelled from the Bouguer Anomaly map (Fig. 5a and c), both of which run over one or two of the Hontomín 1 and 2 boreholes, show a similar set of faults and the same reservoir-dome-like structure. However, these models cannot be compared to the seismic lines at depth since the seismic data do not reach the basement level.

The faults interpreted in the central sunken block by Alcalde et al. (2014) are not clearly recognizable in the THD map (Fig. 9), which shows an irregular anomaly pattern in the central domain. This could be due to either the confinement of faulting within the cover succession or to the minor offsets associated with these faults not generating strong enough gravity gradients to be recognized with this method.

Another set of faults can be interpreted from the THD of the reduced BA map (Fig. 9). These strike NNW–SSE and have limited length. Even though they have a weak signature, they are similar to those interpreted from the seismic data striking in an almost N–S direction (Alcalde et al., 2014), which were described as active during the Liassic period. However, there is a discrepancy regarding their area of influence, as the latter were interpreted to be associated with a possible Triassic extensional normal faulting affecting the basement, which triggered the movement of the Triassic evaporates. Furthermore, either they do not affect the basement topography (Figs. 8a, 9 and 10), except locally to the SW, or at least they do not produce a basement offset identifiable in the gravity data. This suggests that this set of faults may only affect the Jurassic and Triassic succession and that they can be associated with the movement of the Triassic evaporites towards the basement wall, causing the dome growth.

Despite the general good agreement between the basement model presented here and that presented by Alcalde et al. (2014), there are still some differences, especially concerning the NW sector. Here, the basement topography appears to be higher when deduced from gravity data than in the seismic model. Three hypotheses can be proposed to explain this discrepancy (Fig. 11). The first (Fig. 11a) is that the basement high deduced by the gravity study is real. This basement high could be explained as a rollover anticline generated by the listric geometry of the south fault. The remaining hypotheses imply that the gravity high observed in the NW sector is not only produced by a basement high, but also by a complex organization of the stratigraphic succession. The second hypothesis (Fig. 11b) has a reduced Keuper thickness in this area, which would imply a general increase of the gravity signal. This could be explained as the result of the migration of the Keuper evaporites towards the dome and would imply a thickening of the Jurassic–Recent stratigraphic succession in this zone. Figure 11c shows a third possible scenario where the positive gravity anomaly is related to the occurrence of a dense stratigraphic unit, which could correspond to Triassic anhydrites laterally passing to the less dense Keuper evaporites of the dome domain. Nevertheless, we cannot rule out the possibility that this structure is related to the edge effects generated after filtering a regional component in such a small area.

The basement high (2100 m b.s.l.) at the NW end of the gravimetric model also contrasts with the position of the basement top at -2800 m b.s.l. 5 km to the north as interpreted by Carola et al. (2015) based on seismic data (Fig. 12). These authors propose a thin-skinned model detached at the top of the basement and associated with a large offset on top of a low angle footwall ramp. However, the staircase geometry of the basement top across the Hontomín structure (Fig. 12) is in our opinion in disagreement with this model, since the cover does not show a structural culmination associated with the NW basement high, which should be expected in a thin-skinned model. Besides, the monoclinal attitude of the Upper Cretaceous limestones on the southern border of the Hontomín structure above the south fault and the homogeneous uplift of this layer north of this point suggests that this feature is mainly associated with the inversion of the south fault and hence is a thick-skinned model. In this scenario, the basement step NW of the gravimetric model would be associated with a new normal fault (Fig. 12), which was not inverted during the compressive stage. The NW fault would be mainly responsible for the extensional forced folding occurring during the rifting stage as already proposed by Tavani et al. (2013).

At a regional scale, the thickness of the Triassic salt layer in the Plataforma Burgalesa is very inhomogeneous from north to south. Exploration wells have drilled different thicknesses, ranging from 2000 to just 300 m (Carola et al., 2015). This is also the case in the Hontomín area, where the average thickness of the Keuper evaporites is 1660 m with a maximum of 2020 m and a minimum of 1200 m. This thickening generates a gravity low observed in the BA map, which appears somehow displaced from the south fault, as is also clearly pictured in the 2-D models and the cross sections derived from 3-D models (Figs. 5a, c, 6). These offsets of the gravity minima provide evidence of the mobilization of the evaporites towards the walls of the principal faults (south fault and the east fault) related to the migration pathways forced by the Cenozoic compressional stage (Tavani et al., 2013) but also to some extent to salt mobilization during the rifting stage as a consequence of the sedimentary load of the syn-rift succession on the hanging wall of the basement faults.