The use of lidar (light detection and ranging) 3-D photorealistic outcrop models, combined with traditional sedimentological and structural field data, improves the accuracy and efficiency of qualitative and quantitative characterization of outcrops, which in turn can be used as analogs for reservoir modeling and other geologic purposes. This paper illustrates how geological data extraction from 3-D photorealistic outcrop models can be exploited, and presents some novel workflows that reduce the time needed for postprocessing. The extracted data are calibrated with conventional outcrop studies and allow extensive quantitative analyses and detailed statistical examinations of the distribution, dimension, and shape of geological features that can be used to define and build geological models. We present the first statistical characterization based on lidar of a set of geological outcrops at centimeter resolution (bed scale) over a distance of 45 km (basin scale). These innovative methods of outcrop visualization and characterization are applied to the Eagle Ford Formation, an important unconventional hydrocarbon play in Texas. The Eagle Ford Formation consists of alternating organic-rich mudstone, limestone, and bentonites; mudstones represent the source and reservoir of the hydrocarbons, limestones control the rock’s brittleness, and bentonites provide time lines for dating and correlating sections. The presented analyses provide empirical relationships that can be applied to better understand geologic processess, to build geologic models, and to reduce uncertainties in exploration and development of hydrocarbon systems.


In recent decades, lidar (light detection and ranging) technologies have been developed for acquisition of aerial and terrestrial acquired three-dimensional (3-D) spatial data (e.g., Miller, 1965; Kraybill et al., 1984; Wilson and Hawkes, 1987; Bosch and Lescure, 1995; Axelsson, 1999; Ackermann, 1999; Baltsavias, 1999; Xu et al., 1999; Amann et al., 2001; Shan and Toth, 2008; Mallet and Bretar, 2009; Heritage and Large, 2009; McCaffrey et al., 2010; Vosselman and Maas, 2010; Lemmens, 2011; Hodgetts, 2013). Lidar involves laser pulses hitting a target and returning to a detector, thus defining distances and the generation of precisely located point clouds that delineate 3-D surfaces. New digital acquisition techniques have been developed, including the ability to accurately and rapidly capture 3-D digital data of rocks exposed in outcrops. These techniques permit linking of geological measurements and descriptions of outcrops to their correct geographic position for building a spatial framework. Numerous groups apply lidar to field geology problems, mainly to bridge the resolution gap between seismic and well data (Jones et al., 2004; Pringle et al., 2004, 2010; Bellian et al., 2005; McCaffrey et al., 2005; Clegg et al., 2005; Trinks et al., 2005; Buckley et al., 2006, 2008; Oldow et al., 2006; Labourdette and Jones, 2007; Redfern et al., 2007; Fabuel-Perez et al., 2009; Sima et al., 2012; Wang et al., 2012). Lidar acquired in the outcrops provides the ability to study depositional systems with similar characteristics in the subsurface at a wide range of scales (kilometers to centimeters), offering nearly 3-D accessibility and better data continuity (e.g., Bracco Gartner, 2000; Enge et al., 2007; Lanen et al., 2009; Buckley et al., 2010). Some types of high-resolution data can only be obtained from outcrops or mines to properly interpret subsurface rocks, for example, lateral facies variability, joint systems, fault offsets, and stratal terminations. Accurate description, mapping, and measurement of outcrops are extremely important for understanding basin evolution, including depositional environments and structure. Traditionally, geologists have mapped outcrops with sketches, using 2-D photography (including photo panels), compasses, and distance-measuring devices such as tapes and theodolites. These traditional methods are time consuming and force 3-D real-world data into 2-D representations.

Among the various techniques employed by lidar systems since the late 1990s (Bellian et al., 2005, and references therein), the system used here is the one that combines terrestrial laser scanning (TLS) (static in this case from tripods), digital cameras, and survey grade (1–2 cm) global navigation satellite systems (GNSS; including the global positioning system satellites), which allows the generation of georeferenced 3-D photorealistic outcrop models (models herein). These models first fit the collected points to triangle meshes such as a triangulated irregular network (TIN) and then drape the digital photos upon the mesh (Fig. 1), as described in Xu et al. (2000) and Aiken and Xu (2003). Any measurable attribute on the model (e.g., bedding dip and strike, fault offset, joint orientations and spacing, bed shape and thickness) can be spatially captured and correlated with other types of georeferenced geological data (e.g., stratigraphic sections, cores behind the outcrop, well logs, seismic maps and profiles) to create or improve quantitative geological models.

The novelty of our work includes (1) a refined workflow to construct high-resolution and high-accuracy 3-D photorealistic outcrop models that consist of a fast long range scanner, high-resolution photos, close spacing scanner setups, and rapid photo to model alignment; (2) a pilot experiment with a dedicated tool to interpret models using ArcGIS (http://www.esri.com/software/arcgis) and extract quantified geologic features; and (3) a variety of techniques applied for improved visualization and analysis, including superimposing diverse data on the models. As a consequence, these novelties allow the first statistical characterization, based on lidar, of a set of geological outcrops (100–800 m long) at centimeter resolution (bed scale) over an extent of 45 km (basin scale). The purpose of this paper is to illustrate how an appropriate combination of techniques, equipment, and software was designed and used, considering the logistics and characteristics of the outcrops, ultimately facilitating geologists’ analyses and interpretations using 3-D photorealistic outcrop models.


Lidar acquisition and processing techniques have been established and documented in the past few decades; building on this, a postprocessing workflow was developed that improves the efficiency and accuracy of the construction of 3-D photorealistic outcrop models at a rate several times faster than related previous workflows (i.e., Xu et al., 2000; Hodgetts et al., 2004; Thurmond et al., 2005; Olariu et al., 2008, 2011; Alfarhan et al., 2008; Buckley et al., 2008; Alfarhan, 2010). This total workflow increases the reliability of the positioning of images onto the terrain to 1–2 mm, very cost effectively, and reduces work time from months or years to days or weeks. The postprocessing time per outcrop ranged from 2.5 to 5 days, including TIN mesh assembly, photo draping, georeferencing, and conversion to an ESRI ArcGIS compatible format.

Coherent orientation parameters were generated by bore sighting a scanner-mounted camera to the scanner beam. These coherent parameters, consisting of camera location, camera rotation angles, effective focal length, scale factor, and image principle point (Xu, 2000), were used to drape the photos onto the TIN mesh model and to transform the various coordinate systems. Such parameters have been available in lidar technology for several years but have been mostly applied to coloring point clouds (multiple 3-D spatial points). Because the mounting parameters of the camera are calibrated, the pixels from the photos are related accurately to the scan point clouds, therefore the time consuming effort in the field or in the office of positioning the images onto terrain models is eliminated. Consequently, the link between the photos and the 3-D spatial points becomes automatic (Fig. 1).

This concept was used to build models generated by scanning at 9 outcrops distributed over 45 km of roadcuts with 183 scan positions and more than 2000 photos generated by using scan setups spaced at ∼15 m operating at 20–30 m ranges, necessary to obtain the intended high resolution and high accuracy. By operating one of the faster scanners, a Riegl VZ-400i, with a coaxial camera at 100,000 points per second such a large number of scan sites could be done relatively fast, 15 minutes per scan position, compared to scanners operating at 10,000 pts/s, such as a Riegl 620i, which would have required 10 times the scanning time. The combination of several outcrops, a high-speed scanner, closely spaced scan positions, and high-resolution photography produced an extremely large amount of data for postprocessing. The use of the coherent parameters from the Riegl coaxially mounted camera system greatly facilitated the creation of the photograph to model transformation parameters that are needed to drape the photographs onto the TIN mesh. The draping process (Xu, 2000) is based on the colinearity, pinhole camera model, and perspective projection, which are commonly used in machine vision (Brown, 1992; McGlone et al., 2004). The coaxial camera provides the .ini and the .cam files needed to link the photos to the scanned points. The .ini and the .cam files are exported directly from the Riegl software. The .cam file provides the interior image distortion correction parameters for the camera. This is available by characterizing the camera distortion using Riegl’s RiScanPro software. The .ini file provides the camera orientation information for each of the photos. The two files are read by the Geological & Historical Virtual Models, LLC (GHVM; http://www.ghvmodels.com/) software used in this study to drape the photos onto the surface models.

Once created, the models are loaded into ESRI’s ArcGIS, converted to multipatch format and visualized through the ArcGIS application ArcScene (http://resources.arcgis.com/en/help/main/10.1/index.html#//00q8000000p0000000; Animation 1). The models are then georeferenced in global coordinates and integrated with data collected in the field and analyzed in the laboratory (e.g., lithologs, spectral gamma ray curves, petrography, geochemistry, biostratigraphy, X-ray diffraction data). The remote sensing data set (lidar processing and creation of the model) is validated and calibrated through conventional field work (description of outcrops, collection of samples, and measurements of geologic features) (Animation 2). Eventually, data from the calibrated models are extracted using GeoAnalysis Tools (White, 2010), a commercial ArcGIS plug-in software developed by GHVM. The extracted quantitative data (orientation and dimension of fractures, bed geometries, mudstone:limestone ratios) are then assembled, statistically analyzed, and used to build geologic models and to better define geological processes. In turn, these outputs are useful in populating subsurface reservoir models (e.g., defining bedding frequency, stratigraphic architecture, fracture patterns) to reduce uncertainties in hydrocarbon exploration.


A Riegl VZ-400i scanner was used to collect 183 scans in 10 days along the roadcuts of U.S. Highway 90 in western Texas between Comstock and Langtry. Scanner-to-target distances ranged from 20 to 30 m for the outcrops and 100 m for the control points. Angular resolution of scanning was 0.03°, resulting in point intervals of ∼2 cm at 10–30 m range. Photographs, acquired with a Nikon D700 with an 85 mm lens, resulted in a pixel size of 1 mm. A local base station, using a Topcon Hyperlite GNSS, was utilized at each outcrop with a roving receiver positioning all the control reflectors relative to the local base; the raw base station data were postprocessed with the closest continuous operating reference station (CORS) to globally georeference the local base station and therefore all the associated data (Soler, 2011). Using a scan position interval of 15 m, ∼350 m of lateral outcrop were captured per day, given an overshoot of the target area on both ends (see Fig. 2 for locations). The scan-position intervals were determined by taking into account the geometries and the distances to the outcrops in order to determine the best overlap of the scans and the photos. Specifically, the scan-position intervals were 20 m for outcrop DR7, 15 m for outcrops DR6, DR12, and DR15 (Figs. 3–5), and <10 m for outcrop DR1 (Fig. 6). Although nine locations were scanned in this study, only five of them are presented here. A 1-m-long, 10-cm-diameter cylinder was positioned against the outcrop and scanned at 0.005° resolution to create a benchmark and calibrate model measurements.

Models were calibrated with field observations during eight field campaigns, employing ladders and cherry pickers to safely access the roadcuts. Sections were analyzed at centimeter scale generating lithologic logs, spectral gamma-ray measurements were acquired using a RS-230 BGO spectrometer at 20 cm intervals (2 min count times) for correlation purposes, and samples were collected for subsequent geochemical, petrographic, biostratigraphic, and geochronologic analyses.


The Eagle Ford Formation (locally known as the Boquillas Formation) was described, sampled, and scanned with lidar in nine locations north of Del Rio, in western Texas (Fig. 2). U.S. Highway 90 was rerouted and improved in 1979, resulting in a magnificent series of more than 50 roadcut exposures from 5 to 30 m in height. The contacts between the underlying Buda Limestone and overlying Austin Chalk are clearly recognized in many roadcuts, although most roadcuts are within the Eagle Ford Formation. These outcrops, as long as 800 m, represent an excellent opportunity to analyze, at bed scale, both the vertical and lateral facies variability and structural framework of the Eagle Ford lithologies to better understand the nature of the formation.

The high-accuracy (millimeter) and the high-resolution (centimeter) imagery derived from the models allow powerful close ups that reveal the 3-D variability of the exposed rocks at bed scale. The measurements made on the models provide a to-scale image of the outcrop. Now 3-D measurements of geologic characteristics such as shapes, strike, dip, true thickness, etc. can be extracted in detail as digital measurements over the outcrops, allowing a 3-D analysis and interpretation.


The Eagle Ford Formation was deposited during the Cenomanian and Turonian Ages (ca. 90–98 Ma) in the southern portion of the Western Interior Seaway. Several Eagle Ford Formation sections contain Oceanic Anoxic Event 2 (ca. 94 Ma), which records a period of extensive anoxia throughout the world’s oceans near the Cenomanian-Turonian boundary (Schlanger and Jenkyns, 1976). The Eagle Ford Formation has a total organic carbon content between 2 and 12 wt% and thus, under the right conditions, can form an excellent hydrocarbon source and/or reservoir rock. The formation shows rhythmic bedding (varying in thickness) including marlstones, limestones, and bentonites (Sellards et al., 1932; Trevino, 1988). Most of the Eagle Ford rocks have carbonate contents >50%, with the exception of bentonites that represent volcaniclastic deposits from Plinian caldera eruptions (Desmares et al., 2007). The marlstones and bentonites generally form recessive units in outcrop profiles, whereas the limestones form well-indurated flaggy beds and are prominent in outcrop profiles. The Eagle Ford Formation is separated into upper and lower members; many subdivide these intervals in a variety of ways, depending on the specific location of study and the proxy used (Trevino, 1988; Miller, 1990; Lock and Peschier, 2006). In our study, the boundary between the lower and the upper Eagle Ford is defined by an abrupt upward decrease in gamma ray value, which is also used to distinguish a lowermost portion of the lower Eagle Ford at the very base of the formation, a 3–5-m-thick sequence of mass transport and high-energy deposits. The 3-D photorealistic outcrop models described in the following represent these three parts of the Eagle Ford, each with distinct lithofacies: DR1 and DR15 show the lowermost Eagle Ford, DR6 and DR7 show the lower Eagle Ford, and DR12 shows the upper Eagle Ford–lower Eagle Ford contact and the upper Eagle Ford (Figs. 3–6).


This section highlights selected examples of data extraction from the Eagle Ford Formation models and demonstrates how the data could be used. The examples focus on the limestone beds because they represent the most prominent and variable beds in the outcrops, and therefore the most suitable for analyses.

Limestone beds in the field and in the models can be grouped into four categories (Fig. 7): (1) sheet-like limestones, continuous beds maintaining constant thickness, common in the upper and lower Eagle Ford (DR6, DR7, DR12); (2) pinch-and-swell limestones, beds with wavy base and top, present only in the lowermost Eagle Ford (DR1); (3) bioturbated limestones, thick continuous beds with clearly visible burrows, best exposed in the transition of upper Eagle Ford–lower Eagle Ford (DR12); and (4) lenticular limestones, discontinuous limestone beds characterized by elongated lenses, mainly in the lower Eagle Ford (DR6, DR7). The models permit rapid extraction of massive amounts of data. For example, Figure 8 shows the extraction of bed thicknesses from the four categories of limestones described above. It is preferable to collect such data sets from the models, rather than from the outcrops, due to logistic difficulties and time-consuming activities in the field, whereas the automated measurements using ArcScene are essentially instantaneous. The histograms in Figure 8 show that (1) bioturbated beds are the thickest; (2) sheet-like beds are similar from outcrop to outcrop and present the highest variability from bed to bed; also, the pinch-and-swell beds show high thickness variability; (3) lenticular beds are thinner than the sheet-like beds and have self-similar thicknesses; (4) with the exception of the bioturbated beds, limestone beds never exceed 25 cm in thickness and average 12 cm; (5) sheet-like beds are the most abundant; and (6) average thicknesses of the 4 categories of beds remain remarkably uniform. These observations, integrated with a robust stratigraphic and sedimentological context derived from the field studies, illustrate the vertical distribution of the four categories of limestones (Fig. 9). The lowermost Eagle Ford contains only pinch-and-swell beds, with the coarsest grain size in the system and cross-stratification, suggesting that the onset of Eagle Ford deposition was affected by highly energetic events (Figs. 3 and 6). The lower Eagle Ford registers a sharp change in the environment of deposition where sheet-like beds and occasional lenticular limestones were deposited (Fig. 4), and a further sharp change records the deposition of thick limestones in an environment affected by strong infaunal bioturbation (Fig. 5). The upper Eagle Ford presents sheet-like limestones, lightly bioturbated, with an apparent thickening-upward trend.

Tracking the shape of the lenticular limestones in the models, it is clear that the thickest part of the lenses is present in the middle section (Fig. 10); this is confirmed by field observations. By measuring the lenticular limestone beds in the models their dimensions can be constrained (Fig. 11): 80% of the lenses are between 50 and 250 cm long; only a few outliers are longer (to 7 m). The smaller the lenses the more rounded they are, and, as a consequence, the lower the ratio between the major axis and minor axis. The average thickness, in 95% of the cases, ranges between 7 and 15 cm. These analyses would not have been possible in the field without great difficulty because of the distance of the lenses above the ground surface. Once the lenses are statistically analyzed in the model, their dimensions can be predicted to better interpret conventional cores, which are limited to only 10 cm widths (Fig. 11).

Plotting the thicknesses of all lenses encountered in an outcrop at different stratigraphic levels and maintaining the original distance between them shows that the entire width of the outcrop is ubiquitously affected by the presence of multiple levels of lenticular limestone beds (Fig. 12). The discontinuity of the limestone beds may interest geoscientists modeling the vertical propagation of induced fractures in unconventional plays. It is noteworthy that several thick lenses, in different stratigraphic beds, tend to overlie each other (i.e., at 4, 8, 12, 18, 20, 22, 24 m; Fig. 12); in contrast only a few cases show compensation, where the lack of a lens in a bed is compensated by the presence of a lens in the overlying bed (i.e., at 14, 15.50, 20.50 m; Fig. 12). These observations highlight the possible relationship between inherited microtopography and formation of the lenses.

With the knowledge gained in the previous data extractions and creating a series of vertical pseudolithologs along the same outcrop (Fig. 13), the lateral variability of the mudstone:limestone ratio can be quantified (Fig. 14). Mudstones dominate along the width of the outcrop (78%–90%); the visualization in a plot displays the location of the variations (i.e., more uniform values to the left, fluctuating values to the right; Fig. 14). The factor causing the variations in the mudstone:limestone ratios is the stacking pattern of the lenticular limestones previously analyzed and shown in Figure 12. Reservoir engineers creating models for black shales may use these percentages as input data for their geological models, moving toward a better quantification of net:gross ratios.

Where fractures are bed confined, the ratio of their vertical length and the distance between fractures is named the fracture spacing ratio (FSR of Narr and Suppe, 1991). Comparing the Eagle Ford FSR with other formations, we may predict the capacity of the fractures to propagate through the entire formation. In Figure 15, a selected bed, ∼20 cm thick, analyzed for this purpose gives an FSR value of 0.8 (20:25 cm), a value indicating that fractures remain confined to specific beds and classifying them as “no through-going fractures” (Gross and Eyal, 2007, p. 1339, fig. 11). Assuming that further analysis of a representative number of beds maintain the same ratio, it would imply that the Eagle Ford fractures have a low capacity to propagate through the entire formation. These measurements have been first taken in the field and then repeated on the model. Results were similar, confirming that the model resolution was good enough to measure the bed-confined fractures, especially along higher stratigraphic sectors, reachable in the real outcrop only with great difficulty.

Locally, in the lowermost Eagle Ford Formation, channel-like features are observed where outcrops on both sides of a road were scanned (Fig. 16). Beyond the dimensions and geometries extracted from the model, the lithology and nature of surfaces described in the field, a geological block diagram can be constructed to interpolate channel features. The block diagram, built with ArcGIS 3D Analyst Tool, derives from the interpretations made on the models, and helps define a pseudo-3-D representation that can then be incorporated in geological models at regional scales.


We present the first statistical characterization based on lidar of a set of geological outcrops at centimeter resolution (bed scale) over a 45 km extent (basin scale). We propose new methods to extract quantitative geologic data from outcrops with optimal quality control, saving time and expanding the area subject to study to outcrops otherwise reachable by field geologists with great difficulty. The methods comprise integration of (1) lidar 3-D photorealistic outcrop models created with a new workflow that reduces field and postprocessing time by from months/years to days/weeks and increases the reliability of accuracy to 1–2 mm; (2) a new ArcGIS software plug-in for extracting data from the models; and (3) conventional field work on the outcrops to calibrate the photorealistic models and the rock properties with respect to the geology.

The geological information extracted through this workflow represents input data that may be used to (1) better understand geologic processes (e.g., associating morphotypes determined in the models to sedimentological characters determined in the field, relating shapes and dimensions of geologic features to natural processes); (2) build quantitative geologic models (e.g., determining stratigraphic architecture by characterizing lateral facies variability, calculating mudstone:limestone ratio, upscaling geomorphologic features in software applications intended to integrate oil reservoir data from multiple sources); and (3) assist exploration and development of unconventional fields (e.g., infer rock brittleness to assist defining landing zones from quantification of limestone-mudstone alternation, and in geosteering by predicting discontinuity of brittle beds and their thickness variability).

This contribution was supported by Shell International Exploration and Production; we greatly appreciate the support of our Shell colleagues Calum Macaulay, James Eldrett, Aysen Ozkan, Matt Lusk, Amy Kelly, Hiranya Sahoo, Rebecca Minzoni, Matt Johnson, and Michael Gross. We thank Sahoo for generating Figure 16, and Lionel White and his colleagues from the Geological & Historical Virtual Models, LLC (http://www.ghvmodels.com/), for their efforts on the GeoAnalysis Tools package. Portions of the research were funded by the National Sciences Foundation (grants 632050, 632102, 651529, and 632402). University of Texas at Dallas students J.D. McGill, Brian Burnham, Graham Mills, and M. Iris Rodriguez-Gomez also contributed substantially to the field and postprocessing work.