Unlocking the correlation in fluvial outcrops by using a

8 Adequate characterization of depositional architecture is of great importance when 9 studying fluvial outcrops as reservoir analogs. The complex three-dimensional (3D) 10 distribution and lateral/vertical relationships of sandstone bodies require a high degree of 11 stratigraphic control in order to make a proper assessment of the distribution and connectivity 12 of the reservoir facies. This demands the use of reliable correlation datums. Unfortunately, 13 clear marker beds (e.g., ash/coal layers and paleosols) are not always available in fluvial 14 outcrops, and when present they are often covered by vegetation or debris that prevents their 15 tracking over long distances. 16 A new method to achieve highly accurate and semiautomatic correlations within 17 fluvial DOMs (Digital Outcrop Model) is presented in response to the need for further 18 correlation procedures, especially in the absence of suitable datums. The method is based on 19 the hypothesis that the average depositional paleosurface of a sedimentary system can be 20 represented by a plane at outcrop scale. If this assumption is valid, this plane can be used as a 21 Virtual Datum to identify along the DOM the sediments that were deposited simultaneously. 22 The method was tested and applied successfully within two kilometer-scale outcrops 23 of the Huesca Fluvial Fan (Early Miocene, N Spain), where the Virtual Datum provided 24 accurate correlations regardless of stratigraphic or topographical complexities. Moreover, the 25 entire sedimentary successions were automatically subdivided into the desired stratigraphic 26 intervals by only moving the Virtual Datum vertically. These intervals can be subsequently 27 isolated to facilitate the detection of subtle variations and trends of their fluvial properties. 28 Consequently, a Virtual Datum is the equivalent of having a marker bed crossing the 29 stratigraphic succession of an outcrop at any desired position. 30 The advantages provided by a Virtual Datum proves to be especially useful in large 31 and topographically complex outcrops that previously could not have been studied with such 32 a high degree of 3D stratigraphic control. 33 INTRODUCTION 34 Fluvial sandstones represent some of the commonest reservoir rocks, and a great deal 35 of the hydrocarbon production worldwide is extracted from sediments that were deposited by 36 ancient river systems. However, because this type of reservoir has a very complex 37 distribution in the subsurface, accurate field studies to characterize outcrops as analogs of 38 buried fluvial systems are needed. 39 The three-dimensional (3D) architecture of sandstone bodies is not only the most 40 complex and unpredictable but also the most important fluvial attribute that must be taken 41 into account when characterizing reservoirs located within ancient fluvial sedimentary 42 successions (North and Prosser, 1993). As noted by Miall (1996), these deposits are difficult 43 to map in detail because of the high lateral heterogeneity of their facies and the poor 44 definition of individual beds in successions consisting of repeated channel and overbank 45 units. This author also argued that the restricted lateral and vertical dimensions of the 46 paleochannels and their associated sandy deposits, together with the nonlinear evolution of 47 facies belts through time and space, hamper our understanding of the 3D architecture and 48 distribution of petrophysical properties within these sedimentary systems. 49 Methods employed to study hydrocarbon reservoirs are mainly seismic surveys, well 50 logs and cores, which lack sufficient spatial resolution to properly characterize the geometries 51 and sedimentological properties of the discrete elements composing fluvial reservoirs (Li et 52 al., 2012). To this end, a number of studies have been focused on the detailed characterization 53 and modeling of outcropping analogs as good approximations for understanding the behavior 54 and spatial arrangement of fluvial reservoirs (Willis and White, 2000; Martinius and Næss, 55 2005; Miall, 2006; Pranter et al., 2009; Li et al., 2012). However, most of the available 56 outcrops are composed of 2D sections, and geological expertise is needed to design accurate 57 3D reconstructions to determine parameters such as channel sinuosity, connectivity and 58 continuity (Pringle et al., 2006). 59 Correlation uncertainty 60 Strong stratigraphic control is required when working in fluvial outcrops in order to 61 perform accurate 3D characterizations of the geometries and stratigraphic architecture of 62 sandstone bodies. The high lateral and vertical heterogeneity of facies in fluvial environments 63 causes the uncertainty of the correlations to increase with the number of sandstone bodies and 64 the distance between them. 65 Li et al. (2012) made an experiment to quantify the relationship between the density 66 of data and the accuracy of correlations. First, these authors established a base case 67 performing a high-resolution stratigraphic analysis based on 58 sections of a 30 Km-wide 68 Cretaceous fluvio-deltaic outcrop. Subsequently, they designed three different datasets with 69 progressively fewer sections than the base case in order to compare the interpretations made 70 for each dataset. The results showed how, in extreme cases, overcorrelation led to the 71 identification of only 40% of the existing fluvial bodies, whose widths and thicknesses were 72 exaggerated by about 400% (Li et al., 2012). These results indicate how inaccurate 73 correlations may result in very different stratigraphic frameworks, profoundly affecting the 74 sizes, geometries and connectivities of the reservoir facies during subsequent modeling. 75 Geologists working in the characterization of large fluvial outcrops have used 76 different methods to obtain a stratigraphic control over sedimentary successions, of which the 77 use of marker horizons provides the most accurate and reliable correlations. Typical marker 78 horizons are coal, paleosol or volcanoclastic levels, which are assumed to have been 79 generated at a specific time and can extend tens of kilometers along the surface of fluvial 80 systems (Miall, 1996). Similarly, the presence of major erosional surfaces (if flat) or of large 81 tabular sandstone bodies can also be used as good datums. Unfortunately, in many outcrops 82 the use of marker horizons is not possible because these either are absent or covered by debris 83 or vegetation. In such cases, the correlation criteria will be largely based on the identification 84 of similarities between the characteristics of the sandstone bodies (e.g., size, geometrical 85 proportions, internal architecture and elevation) and/or on the recognition of distinctive 86 sequential arrangements (e.g., amalgamated intervals, coarsening/thickening trends and 87 prograding/retrograding sequences). However, given that these methods are strongly 88 conditioned by subjective interpretations, the resulting correlations will continue to be 89 uncertain. Moreover, the degree of uncertainty increases when correlating between nearby 90 outcrops or between several outcrop faces that cannot be observed from the same location 91 because of topographical constraints (e.g., if they are located in parallel valleys or on 92 opposite slopes of the same hill), which rules out direct visual correlation or the use of 93 photomosaics. 94 Background studies using TLS for characterizing geological outcrops 95 In recent years, improvements in digital data collection techniques and processing 96 software have led to significant advances in the field of outcrop characterization (Pringle et 97 al., 2004; Enge et al., 2007; Jones et al., 2011). This evolution is based on the premise that 98 the greater the quantity, quality (accuracy) and speed of data collection, the better constrained 99 the deterministic models derived from them (McCaffrey et al., 2005; Buckley et al., 2008; 100 Jones et al., 2008; Faubel-Pérez et al., 2010). In this regard, Pringle et al. (2006) provide a 101 review of the different digital data collection methods, highlighting Terrestrial Laser 102 Scanning (TLS) as the preferred technique of geologists. TLS is based on lidar technology, 103 which although developed in the early 1960s, has only recently been incorporated into the 104 study of geological outcrops. Lidar typically uses the two-way travel time of a laser pulse to 105 determine the distance to a target as sonar uses sound waves or as radar uses radio waves, but 106 with a much higher resolution and accuracy. The word lidar has been commonly attributed to 107 the acronym for light detection and ranging in the literature. However, according to the 108 Oxford English Dictionary and the first paper that refers to this technology (Ring, 1963), it is 109 a portmanteau word for light + radar. 110 The main advantages of TLS over the rest of digital data collection techniques are the 111 following: (1) very rapid collection of high amounts of 3D data (thousands of points per 112 second); (2) high resolution (few centimeters) and accuracy; (3) acquisition of information 113 about the scanned materials through the intensity of the returned pulse (Burton et al., 2011); 114 and (4) photorealistic 3D data visualization obviating the need to create a mesh from the 115 point cloud, avoiding thereby the generation of extra geometries (Kreylos et al., 2013). 116 In the last decade, TLS has been used to characterize geological outcrops with diverse 117 purposes. Examples include the following: study of dinosaur footprints (Bates et al., 2008); 118 characterization of folds, faults and fracture networks (Baker et al., 2008; Olariu et al., 2008; 119 Jones et al., 2009; Wilson et al., 2009; García-Sellés et al., 2011; Pearce et al., 2011; Wilson 120 et al., 2011); and geometrical depiction of carbonate platforms (Phelps and Kerans, 2007; 121 Verwer et al., 2009). 122 Regarding the study of channelized bodies, several works have been focused on the 123 detailed description, characterization and modeling of sandstone bodies (Labourdette and 124 Jones, 2007; Pranter et al., 2007; Faubel-Pérez et al., 2009; van Lanen et al., 2009; Pyles et 125 al., 2010; Olariu et al., 2011; Olariu et al., 2012; Rittersbacher et al., 2014; Sahoo and Gani, 126 2015); performing flow simulations (Klise et al., 2009; Nichols et al., 2011); the study of the 127 alluvial architecture (Labourdette, 2011; Hajek and Heller, 2012); and building of geocellular 128 and seismic 3D models (Enge et al., 2007; Janson et al., 2007; Buckley et al., 2010; Faubel129 Pérez et al., 2010; Pringle et al., 2010; Tomasso et al., 2010). In these works, TLS data were 130 mainly employed to characterize the internal/external geometries and spatial arrangements of 131 sandstone bodies, whereas correlations were carried out by merely recognizing in the DOM 132 the sedimentary features that have been traditionally used for correlation (e.g., marker beds, 133 extensive sandstone bodies, major erosional surfaces or characteristic architectural 134 arrangements). However, we consider that these correlation procedures do not exploit all the 135 possibilities that an exhaustive analysis of TLS data can offer. 136 The main aim of the present paper is to provide a new TLS-based methodology 137 leading to the creation of a Virtual Datum that furnish the degree of stratigraphic control 138 needed to perform highly accurate correlations at outcrop scale and help solve some of the 139 issues regarding the characterization of the fluvial reservoirs mentioned above. 140 OUTCROPS UNDER STUDY 141 Fluvial outcrops selected for testing the suitability of using a Virtual Datum as a 142 correlation tool are located near Huesca, NE Spain (Fig. 1). Their sediments were deposited 143 in the early Miocene by rivers that flowed through the Huesca Fluvial Fan (Hirst and Nichols, 144 1986; Hirst, 1991; Nichols and Hirst, 1998; Jones, 2004; Luzón, 2005; Fisher and Nichols, 145 2013). This fluvial system was developed adjacent to the northern boundary of the Ebro 146 Foreland Basin under endorheic conditions (Puigdefàbregas and Souquet, 1986; 147 Puigdefàbregas et al., 1992; Barnolas and Gil-Peña, 2001), and have been classified as 148 pertaining to the Sariñena Fm. (Quirantes, 1969). 149 The closure of the connection between the Ebro Foreland Basin and the Atlantic 150 ocean during the Priabonian (Costa et al., 2010) marked the onset of a widespread deposition 151 of thick continental sequences throughout the basin, resulting in the development of a series 152 of large distributive fluvial systems (Hartley et al., 2010) spreading out from the surrounding 153 mountain ranges, e.g., Montsant, Guadalope-Matarranya, Caspe, Luna and Huesca (Allen et 154 al., 1983; Hirst, 1991; Puigdefábregas et al., 1991; Möhrig et al., 2000; Jones et al., 2001; 155 Luzón and González, 2003; Luzón, 2005; Nichols, 2005; Cuevas et al., 2007; Barrier et al., 156 2010). The Huesca Fluvial Fan was the largest one, with a radius of about 60 Km, covering 157 an area of around 4500 Km and presenting thicknesses exceeding 1000 m (Hirst and 158 Nichols, 1986; Hirst, 1991; Nichols and Hirst, 1998). This fluvial system evolved between 159 the late Oligocene and the lower Miocene (Luzón and González, 2003) adjacent to the 160 External Sierras, which were formed by the southward propagation of the South Pyrenean 161 Frontal Thrust (SPTF in Figure 1). Its sediments were sourced from the Pyrenean axial zones 162 and from the exhumed south Pyrenean piggy-back basins (Jupp et al., 1987; Vincent and 163 Elliott, 1997; Vincent, 2001; Yuste et al., 2004), and were transferred towards a perennial 164 lake located at the basin center (Cabrera and Sáez, 1987; Arenas and Pardo, 1999; Cabrera et 165 al., 2002; Cabrera et al., 2011) (Fig. 1). 166 The Huesca Fluvial Fan developed after the main phases of deformation in the 167 adjacent Pyrenees (Fisher and Nichols, 2013) in a context where the aggradation rates 168 exceeded those of basin subsidence (Nichols, 2004, 2007). This suggests that tectonic 169 controls did not play a significant role in the evolution of the system. Climatic controls can 170 also be ruled out because of the lack of clear cyclical sequential arrangements in the vertical 171 architecture of the available outcrops (Fisher and Nichols, 2013). Owing to the lack of 172 significant allogenic forcings, the evolution of the Huesca fluvial Fan was largely controlled 173 by autogenic processes, especially by the major avulsions triggered by cycles of channel 174 backfilling and plugging (Nichols, 2007; Fisher and Nichols, 2013; Ventra et al., 2014). The 175 resulting fluvial architecture largely consists of isolated to amalgamated sandstone lenses and 176 sheets surrounded by fine-grained floodplain sediments, which is the characteristic facies 177 arrangement of labyrinthine-type reservoirs (Webber and van Geuns, 1990). 178 Montearagón and Piracés outcrops (Fig. 1) are located approximately 45 Km away 179 from the estimated apex of the Huesca Fluvial Fan (Jupp et al., 1987) and have been 180 interpreted as belonging to the medial part of this fluvial system (Hirst, 1991). Despite being 181 about 16 km apart, they are assumed to be located in similar stratigraphic positions like most 182 of the outcrops of fluvial fan deposits in the zone (Cuenca et al., 1992). This is due to a 183 slightly tilted sedimentary succession (typically <1.5) and to the relatively smooth structural 184 relief (<100 m) existing across the whole fan area. In the absence of chronological data for 185 the Montearagón and Piracés outcrops, biostratigraphic and geochronological datings in the 186 proximity suggest a lower Miocene age (Álvarez-Sierra et al., 1990; Odin et al., 1997). 187 The Montearagón outcrop is located adjacent to the Flúmen river, 5 Km NE of 188 Huesca (Fig. 1). It is composed of two parallel and unconnected slopes of kilometric length 189 (Montearagón in the south and Barranco Hondo in the north, Fig. 2A) that present a fluvial 190 succession about 80 m thick. The Piracés outcrop is located in the surroundings of the village 191 of the same name (Fig. 1), and comprises more than 6 Km of steep and continuous slopes of 192 about 100 m. This outcrop can be subdivided into two sectors (Fig. 2B): an amphitheater 193 opened towards the SE (located to the N of Piracés), and a NW-SE trending cliff facing SW 194 (located to the NW of the same village). 195 The two outcrops present several cliffs oriented towards the SW and/or NE (Fig. 2), 196 which together with main paleocurrents towards the W-SW (Friend et al., 1986; Friend et al., 197 1989; Hirst, 1991) theoretically should provide numerous cross-sections of paleochannels. 198 However, they differ in three-dimensionality and physiographic complexity as well as in the 199 proportion and size of paleochannels (Hirst, 1991). 200 Sedimentary facies 201 Seven detailed stratigraphic logs (1:50 scale, more than 550 m in length, see location 202 in Figure 2) were measured in the Montearagón outcrop to characterize the facies and verify 203 the quality of the correlation results. Lithofacies described in Montearagón can be 204 extrapolated to Piracés since both outcrops are located at similar radial positions of the same 205 fluvial system (Fig. 1). Earlier studies carried out in the area (Friend et al., 1989; Hirst, 1991; 206 Donselaar and Schmidt, 2005; Luzón, 2005) and observations made during the different TLS 207 acquisition campaigns support this premise. Outcropping lithologies largely consist of fine to 208 medium-grained sandstones embedded in siltstones and mudstones with scarce occurrences 209 of centimeter-thick limestone levels (Fig. 3), and have been classified into channel-fill and 210 overbank facies 211 Channel-fill facies are mostly medium grained, although coarse sandstone and/or 212 pebbles are occasionally found forming basal lags, and typically exhibit a fining-upward 213 granulometric trend to fine and very fine sandstone at the top. Most paleochannels show flat 214 erosional basal surfaces that grade laterally to well-defined cut banks and are poorly incised 215 into older deposits owing to the characteristic aggradational trend that prevails in endorheic 216 basins (Nichols, 2004, 2007, 2012; Fisher and Nichols, 2013; Ventra et al., 2014). Trough 217 cross-bedded sedimentary structures are commonly present in the lower parts of the 218 paleochannels, whereas horizontally stratified and ripple cross-laminated fine sandstone 219 dominate their upper parts. Clay plugs, which are also common, are the product of the passive 220 infill of abandoned channels with sediments transported as suspended load during flooding 221 events. Paleochannels in the medial zone of the Huesca Fluvial Fan have been interpreted as 222 the deposits of braided, meandering and straight channels, which distally show a tendency to 223 reduce their dimensions and increase their lateral stability as a result of a decrease in stream 224 power (Hirst, 1991; Nichols and Fisher, 2007). In outcrop, sandstone beds stand out as steep 225 rock faces owing to the lower erodibility of this lithology with respect to the surrounding 226 fine-grained sediments. This contrast in erodibility is enhanced by a late-diagenetic calcite 227 cementation of sandstones (Donselaar and Schmidt, 2005). 228 Overbank facies are composed of variable amounts of sand, silt and clay-rich 229 sediments with an average content of carbonate of 30%, and were deposited from the 230 suspended load during floods (Nichols and Hirst, 1998). The coarsest overbank sediments are 231 found adjacent to the paleochannel margins in the form of levee deposits (Fig. 3, A). They 232 consist of inclined beds of alternating sandstone/mudstone that extend from tens to hundreds 233 of meters towards the floodplain, forming the characteristic channel “wings” (Fig. 3, B). 234 Crevasse splays consist of extensive sheets of fine sandstone that typically show thicknesses 235 exceeding one meter, non-erosive bases, coarsening-upwards sequences and a predominance 236 of planar and ripple laminations. The feeder channels of these crevasse splays are constituted 237 by small-scale ribbons (<1 m thick) of fine sandstone. The finest sediments, which were 238 deposited by decantation in the waning stages of floods, constitute the bulk of the floodplain 239 facies. Thin limestone levels occasionally occurring within the fine-grained intervals and the 240 top of sandstones are attributed to the precipitation of carbonate from ponded waters. 241 Evidence of pedogenic processes associated with incipient paleosol development is 242 found in the form of reddish decolorations, light yellow levels with versicolor mottlings 243 (usually associated with rhyzoliths), gray levels with iron nodules and carbonate and gypsum 244 concretions in the finest sediments. The development of these paleosol horizons has been 245 associated with periods of non-deposition since their degree of maturity increases with 246 distance from the active channels (Hamer et al., 2007a). Trace fossils are widely present in 247 both overbank and channel fill facies as rhyzoliths, burrows and ant/termite nests. These 248 traces do not always reach the top of the sandstone beds, which in the case of the channel fills 249 suggests a discontinuous water regime since the fauna could not have colonized the bed of 250 the channels had these been continuously active. 251 Friend et al. (1986) proposed a classification for the sandstone-bodies of the Huesca 252 Fluvial Fan, which was later applied regionally by Hirst (1991). This classification is based 253 on the cross-sectional external geometry and on the internal architecture of the paleochannels. 254 Further descriptions and interpretations of paleochannel types and their internal architectures 255 can be found in Nichols and Hirst (1998), Donselaar and Schmidt (2005) and Luzón (2005). 256