Fluid Flow Analysis in Fractured Rock Mass by Data Integration of Digital Outcrop Model and Discrete Fracture Network (DFN)

Abstract

Fracture characterization is crucial to constrain a realistic subsurface reservoir model. They are key elements, affecting fluid flow, permeability and consequently recovery factor and productivity. Considering a proper assessment of fracture network from subsurface investigation is often difficult; in recent years, the application of Digital Photogrammetry (DP) has become popular for fracture network characterization. In this paper, we combined DP and Discrete Fracture Network modeling (DFN) to assess the fluid circulation analysis of the Monte Antola Formation (Northern Apennines, Italy). Thanks to the application of DP, it is possible to reconstruct Digital Outcrop Models (DOMs) and acquire high-precision fracture measurements such as size, location, and orientation. Utilizing quantitative measurements, we performed DFNs to simulate rock mass permeability. The primary findings from the DFNs indicate that fluid circulation is primarily influenced by (1) regions with a high density of fractures, which are associated with the primary structural features observed throughout the study area, and (2) locally, by the orientation of the dominant and persistent fracture set. The proposed approach highlights the importance of the use of DOMs for better reconstruction of the fracture network and defining an important number of relevant parameters; such quantitative information remarkably improves the reliability of DFNs.

1. Introduction

The severe drought that affected Northern Italy, as well as large areas of Europe, in Spring and Summer 2022 shows that water resources may be critical in the future [1,2], especially in mountainous areas where freshwater supply is strongly dependent on snow melt [3,4]. Therefore, management of water resources should be improved by enhancing water recovery from different sources. In this context, the characterization of fractured rock masses could be critical to fully understand the subsurface flow and, therefore, to better manage water resources, ensuring a long-term and sustainable source of freshwater.

Fractures have an important role in the hydrological properties of rock mass [5,6,7,8]; generally, they influence permeability, affecting fluid flow. Therefore, their characterization has major implications not only for water supply but also for several industrial applications, such as hydrocarbon production, geothermal energy, hazardous waste disposal and carbon capture usage and storage (CCUS) (e.g., [9,10,11,12,13,14,15,16]). Therefore, the geometric and spatial characterization of fractures is of primary importance to understand how they can influence the fluid flow of rock masses [17,18,19,20,21,22].

Developments in computer geology have led to the integration of fractures in 3D geological models, creating the so-called Discrete Fracture Network (DFN) models. These models allow for the simulation of fracture network distribution on large rock volumes. The term DFN indicates a computation model expressed by the geometrical properties of the fractures and the topological relationships between fracture sets [7,23,24]. DFN can be a powerful tool for the prediction of fluid pathways in fractured rock masses. Whereas in literature, the term DFN can be used to express several kinds of fracture network models [7], such as deterministic (defined by the fracture mapped) and geomechanical-grown (defined by the rock properties and stress-field), in this work, the term DFN refers to the stochastic fracture network model as conventionally used.

The stochastic, or probabilistic, fracture network model is widely used in geo-engineering and fluid-flow studies (e.g., [25,26,27,28,29,30]) due to the awkwardness in achieving a complete analysis of the 3D rock fracture network [31]. During the analysis of complex geological systems such as fracture networks, it is difficult to avoid uncertainties; therefore, the use of single-valued predictions from deterministic methods can lead to extremely wrong results [7]. Instead, the probabilistic nature of DFN could be considered an advantage because it allows for obtaining more than one prediction, thereby addressing the geological uncertainties [7,32].

DFN modeling necessitates the characterization of fracture properties (e.g., orientation, length, intensity, aperture, termination) and their density distribution functions (e.g., normal, log-normal, exponential distributions) along with their associated parameters (e.g., mean, standard deviation, minimum, maximum, K-Fisher coefficient) to facilitate Monte Carlo simulations of fracture networks [25,28]. Consequently, the greater the accuracy and representativeness of the fracture data, the more representative the resulting DFN models become.

A proper characterization of subsurface fracture network is often difficult and expensive, since it mostly relies on wellbore investigation [15,33,34]. While the expenses for this kind of investigation can easily be covered by oil and gas companies, those working on water supplies often find it more difficult to afford them. Moreover, wellbore investigations are often limited in spacing and resolution [35] and can be affected by orientation biases [36] and, consequently, can lack important information about the fracture network, such as spatial intensity variability. Therefore, the study of surface outcrop analogs becomes of primary importance for the characterization of the subsurface fracture network.

Frequently, the analysis of fractured rock outcrops using traditional manual field measurements can be constrained by limitations that hinder the collection of data. These limitations may include the limited presence and accessibility of rock outcrops [37] and their unfavorable orientation [20,38]). Moreover, field-manual measurements could be affected by biases (e.g., orientation and truncation biases) due to the sampling technique (scanline, window, etc.) or to the local variation in orientation of measured features (waved/undulated surface). Previous studies [37,39] have shown that the use of digital photogrammetry for the development of Digital Outcrop Models (DOMs) helps to obtain more accurate parameters of rock discontinuity features (e.g., bedding and fractures). In previous works (e.g., [40,41,42,43,44,45]), we extensively demonstrated that DOMs help to overcome the limitations of traditional geological fieldwork, allowing the acquisition of a large quantitative fracture dataset.

In this paper, we present a fluid flow analysis of a fractured rock mass performed by a combination of DOMs and DFN modeling: DOMs allow a more statistically robust constraint of fracture network parameters, consequently increasing the reliability of DFN models. The aim of the paper is twofold: (1) analyze how DOMs and DFN modeling could help to better understand the fluid flow in a heterogeneous and structurally complex formation; and (2) present a methodology to predict the preferable orientation of the flow. We tested this methodology in the Upper Staffora Valley (Northern Apennines, Italy), where the Monte Antola Formation, a fractured flysch formation, extensively crops out.

2. Study Area and Geological Setting

The study area is located in the North-Western Apennines, 50 km north of the city of Genoa (Figure 1).

The outcropping Monte Antola Formation, belonging to the Antola Unit (AU), crops out extensively in the study area, and it is tectonically deformed due to a complex structural evolution [47] developed under very low-grade metamorphic conditions (e.g., diagenesis-anchizone) [48]. Stratigraphically, the AU succession can be subdivided from the bottom to the top as follows [47,49,50]:

• Montoggio Shale: Varicolored hemipelagic shales (Late Cenomanian-Early Turonian);
• Gorreto Sandstone: Thin-bedded, siliciclastic and carbonatic turbidites (Early Campanian);
• Monte Antola Formation: Calcareous turbidite (Early Campanian-Early Maastrichtian);
• Bruggi-Selvapiana Formation: Siliciclastic-calcareous turbidite (Early Maastrichtian-Late Maastrichtian);
• Pagliaro Shale: Thin-bedded sandstone and calcareous turbidites alternated with shales (Late Maastrichtian-Late Paleocene).

The Antola Unit is topped by the Tertiary Piedmont Basin, a wedge-top basin located on top of the junction between the Western Alps and the Northern Apennines, which is characterized by a continuous deposition from the Mid Eocene (Monte Piano Marls) to the Late Miocene [51].

The study area, as well as the studied outcrops, is shaped in the Monte Antola Formation. In general, the Monte Antola Formation is composed of dominantly turbiditic calcareous graded beds, calcareous sandstones, sandstones and marlstones, with rare intervening thin, carbonate-free graded layers or uniformly fine-grained shale. The formation is interpreted as a deep-sea basin plain deposit, with local lateral facies variations, which range from proximal thick-bedded turbidites to distal turbidites with predominantly thickening upward cycles and a high percentage of pelites. The thickness of the more competent beds varies, ranging between 10 cm and 2 m [47].

3. Methodology

To quantitatively estimate the fracture network parameters and their distributions across the study area, this research is conducted using a multi-methodological approach, combining:

• Aerial LiDAR-based morphostructural analysis to map the main morphostructural lineaments of the study area;
• Traditional fieldwork and terrestrial photogrammetry, to collect major geological information, check and identify water springs together with the local water management authority, and select the best outcrops to be analyzed through DOMs;
• DOM development and analysis, to generate DOMs and quantitatively characterize the fracture network (e.g., dip and dip direction, K-Fisher coefficient, MTL, P21);
• DFN modeling, to predict the preferential fluid flow directions of the analyzed rock mass.

The following subsections describe the methodology, covering the aerial LiDAR-based morphostructural analysis, terrestrial digital photogrammetry, DOM development and analysis, and Discrete Fracture Network modeling. The integrated workflow is depicted in Figure 2.

3.1. Aerial LiDAR-Based Morphostructural Analysis

The aerial LiDAR-based morphostructural analysis was performed using a LiDAR-based Digital Elevation Model (DEM) retrieved from the Italian database “Banca Dati PST” (Ministero dell’Ambiente e della Sicurezza Energetica, MASE; https://sim.mase.gov.it/portalediaccesso/mappe/#/viewer/new/, accessed on 1 May 2026).

This DEM, derived from the aerial LiDAR of the “Piano Straordinario di Telerilevamento” (Ministero dell’Ambiente e della Sicurezza Energetica, MASE), has a resolution of 1 m. It was analyzed using the open-source software QGIS (3.16 “Hannover”)and CloudCompare software (v2.10.2 “Zephyrus”). QGIS allows us to perform several geomorphic analyses (i.e., hillshade, slope, aspect, curvature), while CloudCompare permits us to analyze the DEM in a 3D stereoscopic environment, allowing us to obtain geometric 3D information. The DEM analysis allowed the detection and mapping of the main morphostructural lineaments in the study area.

3.2. Terrestrial Fieldwork and Digital Photogrammetry

Fieldwork was performed in order to collect major geological information, check and identify the water springs together with the local water management authority and select the best outcrops to be acquired with the terrestrial digital photogrammetry approach (Figure 1c).

Specifically, we used the digital photogrammetry approach already tested in previous works [14,40,52,53] and a Canon EOS 5D 12.8 Megapixel camera (Canon, Nagasaki, Japan) and a zoom lens, acquiring 70 Digital Outcrop Models (DOMs). For all 70 terrestrial photogrammetric stations (Figure 1), the distance camera-outcrop was 6 m and the focal length 28 mm. The mean ground pixel size of the images, also called ground sampling distances, was 1.8 mm/pixel. Due to the geometry and the small dimension (length < 50 m and height < 10 m) of the analyzed outcrops, the images were acquired using confocal (Figure 3a) and/or strip (Figure 3c) geometries of acquisition.

In order to correctly georeference the model in a local coordinate system, we used a photographic scale mounted on a tripod and composed of two arms of 1 m each (Figure 4).

The arms of the photographic scale were settled horizontally and vertically, and then the attitude of the horizontal arms was registered. Whereas the size of the arms is required to scale the DOMs, their attitude is required to correctly orient the DOMs in a local coordinate system in which the origin of the system is settled at the intersection of the arms (see Figure 4 above).

After the photo-acquisition, manual measures of the bedding and fracture were taken using the compass-clinometer. These measures are control measures and, therefore, they are taken into account to assess the validity of the generated DOMs. In general, only clearly exposed surfaces that can be recognized in the photographs have been measured.

3.3. DOM Development and Analysis

The DOMs were developed using the Structure from Motion (SfM)-based software Agisoft Metashape 2.2.1 (Agisoft LLC, Moscow, Russia). The images were aligned using their full resolution and considering the Ground Control Points (GCPs) present on the photographic scale (example in Figure 5).

Subsequently, the procedures for generating dense clouds were executed using the high-quality setting, accompanied by a mild filtering of the depth map. Following this step, the meshes were generated using the software’s high-quality settings and were textured exclusively with images possessing a quality factor greater than 0.9.

The generated DOMs were then analyzed using the open-source software CloudCompare v.2.9; this allows for DOM analysis and mapping of the geological features in a 3D stereoscopic environment. The stereo-visualization and the structural interpretation and mapping of the DOMs were performed by a Planar SD2220W stereoscopic device, composed of two separate polarized display monitors placed one above the other in a clamshell configuration, with a half-silvered glass plate bisecting the angle between the two displays.

Orientation, dimension, position of bedding and fractures were acquired using CloudCompare and related tools, like: Trace polyline tool, Segment tool, Fit plane tool, and Compass plugin [54]. The calculation of fracture intensity was conducted using the DICE MATLAB© application (Dice V.1.0.) (for full explanation, see [55]). This algorithm enables the computation of linear, area and volumetric fracture intensity (P10, P21 and P32, sensu [56]) in three-dimensional space, taking full advantage of the 3D nature of DOM and digital fracture data.

Among the different options for computing fracture intensity, we decided to analyze P21, since Digital Photogrammetry methods lack information concerning the persistence of fractures beyond outcrop surfaces [54]. The P21 [m⁻¹] is defined as the ratio between the total length of discontinuity traces [m] and the sampling area [m²] [31,56].

3.4. Discrete Fracture Network Modeling and Analysis

To model the DFNs, we used FracMan software (FracMan 8.6) (WSP). The DFN modeling workflow can be summarized as follows:

The first phase is the fracture set parameter definition:

• Fracture orientations, defined as the mean orientation of the fracture sets and their K-Fisher distribution coefficients, that express how tight the clusters of orientations are;
• Fracture size, defined as Mean Trace Length and a log-normal distribution;
• Fracture shape, defined as a rectangle to better represent the mechanical stratigraphy (i.e., stratabound fractures; e.g., [26]);
• Fracture intensity, defined as volumetric intensity value (P32) [56];
• Fracture aperture, defined as a log-normal distribution with a mean value of 0.1 mm [57,58,59].

Note that the definition of the aperture parameter is not based on actual measures and is set as the standard value of FracMan for all the sets. Therefore, the permeability results will be considered relatively and will be mainly affected by differences in fracture intensity and orientation.

The second phase consists of a series of Monte Carlo simulations to define the corrected intensity measure value [25]:

• Due to the impossibility of directly measuring the P32 intensity parameters, several Monte Carlo simulations with different values of P32 (specifically, 10 simulations with a P32 input value ranging from 0.2 to 4 with a step of 0.2 m⁻¹, for each set; 800 simulations) are performed to find the correlation between P32 input and P21 output determined from the DFN model (Figure 6);

2.

When the correlation indexes α and β have been determined, it is possible to infer the correct value of P32 to be used in the modeling from the P21/P10 measured on outcrop/borehole (Figure 7).

3.

The DFN model that best fits the measurements acquired from the outcrop is selected as the representative DFN.

When the DFN is modeled and validated, it is possible to calculate the connectivity of the fracture network and the equivalent permeability of the rock mass volume.

In particular, the permeability is estimated using Oda’s method [60], implemented in the software. Assuming fractures are impermeable in their orthogonal direction, it estimates the permeability according to the fracture orientation in each grid cell.

This procedure of DFN modeling creates homogeneous fracture networks, a case far away from the study area. For this reason, the DFNs were modeled in different homogeneous sectors. These sectors are:

• The zones of normal polarity beds and low fracture intensity;
• The zones of normal polarity beds and high fracture intensity;
• The zones of inverse polarity beds and low fracture intensity;
• The zones of inverse polarity beds and high fracture intensity.

4. Results

4.1. Aerial LiDAR Analysis

The morphostructural analysis of DEM (Figure 8) allows us to recognize three main counterslopes that delimitate the West, North and East slopes of the mount ‘Cima Colletta’, with an apparent horizontal orientation, and an N-lineament that connects the N with the W-counterslope, with an apparent steeply inclined orientation (Figure 9).

4.2. Fieldwork- and DOM-Based Analysis

The analysis of the sedimentological features of the turbiditic beds indicates the presence of normal and inverse polarity strata. The lineaments detected on the DEM seem to divide sectors of different strata polarity (Figure 9). Moreover, the quantitative analyses of fracture intensity performed respectively during the geological field survey and the 3D DOM analysis show that as one gets closer to the detected lineaments, the fracture intensity increases significantly.

The analyses of DOMs performed in the different structural sectors (Figure 1c and Figure 9) highlight that the fracture sets maintain constant geometric relationships to bedding both in normal and inverse polarity strata (Figure 10).

This relation between fractures and bedding was already detected by the structural analysis of [61], performed on the Monte Antola Fm. cropping out about 6 km SW of the study area, and interpreted as a specific fracture orientation pattern related to folding: K2 possibly represents the longitudinal/radial/hinge-parallel fracture set; K3 the cross/extensional/hinge-orthogonal fracture set; and K4 and K5 the conjugate shear fractures.

Except for the fracture set orientation, DOMs also allowed for the acquisition of all the fracture parameters required for the DFN models, i.e., mean orientation of fracture sets, orientation dispersion of fracture sets (K-Fisher’s coefficient), fracture intensity and fracture mean trace length. These parameters are calculated considering the 3D nature of the fracture and sampling the properties using 3D sampling techniques (i.e., 3D circular scan plane of DICE).

Table 1. Fracture parameters obtained for DOMs with different polarities in low-intensity areas. Only the parameters required by DFN modeling are reported.

Fracture Parameters	Normal Polarity Strata				Inverse Polarity Strata
	K2	K3	K4	K5	K2	K3	K4	K5
Dip (°)	88	70	56	71	83	49	55	54
Dip Direction (°North)	291	203	232	172	175	88	49	134
K-Fisher coefficient	74.85	68.87	71.29	120.57	92.37	189.2	69.11	104.42
P21 (m−1)	4.06	12.57	10.12	3.62	2.48	14.10	13.09	2.48
Mean trace length (m)	0.63	0.72	1.90	0.39	1.15	5.06	4.17	1.12

and

Table 2. Fracture parameters obtained for DOMs with different polarities in ‘high’ fracture intensity. Only the parameters required by DFN modeling are reported.

Fracture Parameters	Normal Polarity Strata				Inverse Polarity Strata
	K2	K3	K4	K5	K2	K3	K4	K5
Dip (°)	89	72	58	70	87	48	53	60
Dip Direction (°North)	287	194	225	165	171	85	50	136
K-Fisher coefficient	83.65	74.68	71.90	100.34	70.36	133.12	80.91	95.13
P21 (m−1)	10.45	18.79	13.65	12.10	8.67	21.56	16.34	14.62
Mean trace length (m)	0.35	0.45	0.80	0.42	0.92	1.22	0.95	1.03

show the mean comprehensive fracture parameters retrieved from the full DOM dataset, reporting low- and high-intensity areas for both outcrops with normal and inverse polarities. These outcrop data will be later used for DFN modeling.

Along the analyzed structural stations, bedding appears highly persistent, with no pinch-out and/or abrupt lateral termination of strata, and nearly planar, with no evident folds but only small-scale undulations (maximum amplitude 5 cm and minimum wavelength of 20 cm). For this reason, for bedding, we considered constant orientation and a very high persistence, defining a fracture size higher than the DFN simulation volume size (i.e., >10 m). Moreover, using the P10: scanline sampling of DICE, the thickness of 120 beds was measured, showing a log-normal distribution with a mean thickness of about 30 cm (Figure 11).

4.3. DFN Modeling

The DFN modeling and analysis were performed in a volume of 10 m × 10 m × 10 m (Figure 12) in order to decrease the amount of computational time and not overload the computer CPU.

Due to the mean trace length of the fractures, generally lower than 1 m, the settled volume does not affect the results negatively with truncation effect.

When fed by constant values and statistical distribution, as in our case, DFN modeling creates ‘statistically homogeneous’ fracture networks. Since we do not aim to build a reservoir-scale fracture network, but we want to highlight the differences between the different “structural sectors”, DFN models have been performed for normal and inverse polarities and low and high fracture intensities (Figure 10). For the same reason, an analysis of the representative elementary volume (REV) has not been performed since we do not need to build huge models and extract precise and accurate permeability parameters (as previously mentioned, we will consider the relative differences in permeability). The α and β values used for the correction of the P32 input values of the DFN models are 0.88 and −0.05, respectively. To assess the validity of the DFN models, the same circular windows used to extrapolate the fracture parameters onto the 3D DOM were used on the DFN models (as shown in Figure 6). This comparison shows similar results and, therefore, indicates a good validity of the models (example in

Table 3. Comparison between fracture parameters sampled on the DOM and on the DFN for low fracture intensity areas.

		Normal Polarity Strata				Inverse Polarity Strata
		K2	K3	K4	K5	K2	K3	K4	K5
Sampled on DOM	P21 (m−1)	4.06	12.57	10.12	3.62	2.48	14.10	13.09	2.48
	MTL (m)	0.63	0.72	1.90	0.39	1.15	5.06	4.17	1.12
Sampled on DFN	P21 (m−1)	4.15	12.30	10.10	3.02	2.34	13.59	13.60	2.28
	MTL (m)	0.62	0.64	1.82	0.50	1.24	4.87	5.28	1.43

and

Table 4. Comparison between fracture parameters sampled on DOM and on DFN for high fracture intensity areas.

		Normal Polarity Strata				Inverse Polarity Strata
		K2	K3	K4	K5	K2	K3	K4	K5
Sampled on DOM	P21 (m−1)	10.45	18.79	13.65	12.10	8.67	21.56	16.34	14.62
	MTL (m)	0.35	0.45	0.80	0.42	0.92	1.22	0.95	1.03
Sampled on DFN	P21 (m−1)	10.23	18.54	11.01	12.01	9.09	21.32	16.67	12.89
	MTL (m)	0.24	0.76	0.83	0.44	1.11	1.51	1.06	1.38

).

Moreover, using the FracMan toolbox Cluster Analysis, we calculated the pseudo connectivity of fractures (i.e., clustering). The analysis shows that for each sector, all the fractures are connected in a single cluster (i.e., 100% of fractures are connected).

DFN-related permeability analysis was conducted by decomposing the permeability vectors in their xx and yy components in order to understand the principal direction of flow. This decomposition suggests that in both normal and inverse polarity strata, the higher permeability component is sub-parallel to the K3 direction, Pxx and Pyy, respectively, for normal and inverse polarity strata (Figure 13). Notwithstanding, the difference between them is small and therefore cannot strongly influence the fluid flow direction.

From the comparison of areas with low and high fracture intensity (Figure 13 and

Table 5. Mean value of the horizontal permeability components Pxx and Pyy for the normal (NP) and inverse polarity strata (IP), and for low (LP) and high fracture intensity (HP).

		Pxx (mD)	Pyy (mD)
NP	LI	2.2	1.7
	HI	3.9	2.6
IP	LI	2.7	4.4
	HI	3.6	4.5

), it is clearly visible that in high-intensity areas, the permeability values can be from 2% to 79% higher than the values of low-intensity areas and, therefore, could strongly influence the fluid circulation in the rock mass.

5. Discussion

The results presented in this paper are based on a combination of multiple approaches based on aerial LiDAR DEM remote sensing morphometric analysis, fieldwork observations, Digital Outcrop Model-based fracture analysis and DFN modeling. Through the DEM-based remote sensing analysis, we were able to detect and localize the four main morphostructural lineaments in the study area (Figure 8): three lineaments with an apparent horizontal attitude that delimit the W, N and E slopes of the mount ‘Cima Colletta’, and an NNE-SSW trending lineament with an apparent steeply inclined attitude.

The field-survey and DOM analysis suggests a correlation between the main lineaments in the study area and deformation structures like fold axial planes and faults, as already described in nearby areas [61]. This interpretation is supported considering that these lineaments separate different structural sectors where the Monte Antola Fm. has different bed polarity and are associated with an increase in the fracture intensity (Figure 9). During fieldwork, we were not able to clearly observe evidence of these structures due to the extensive debris and vegetation coverage. This is possibly because these areas are characterized by a very intense deformation that favors the development of soil and the growth of vegetation.

The analysis of the DOM, developed using terrestrial digital photogrammetry, shows the presence of four fracture sets clearly visible along the entire study area (Figure 10). These four sets keep the same mutual angular relationships with bedding in all the different structural positions. Thanks to the use of the DICE, it was possible to characterize the areal fracture intensity (P21) directly on the 3D Digital Outcrop Model. This novel technique allowed us to quantitatively describe fractures and, therefore, to calculate the parameters required for DFN modeling (Figure 12). Using this DOM-based 3D approach, we were able to avoid the errors that often affect the traditional field-based approach and the 2D image sampling technique (e.g., length and orientation biases). Note that, as shown in [55], the DOMs are 2.5D surfaces (2D surfaces in a 3D space) and, therefore, some limitation on the 3D geometrical definition of fractures may still persist.

The developed DFN models show good validity because the comparison of the data acquired using a 3D circular window sampling onto the DOM and those acquired using the same 3D circular window onto the DFN is pretty similar (Table 3 and Table 4). The DFN properties of all the different structural sectors permit us to predict a fluid flow mainly concentrated along the K3 fracture set direction (Figure 13). In particular, due to the high fracture intensity, the DFN suggests a major fluid circulation in the lineament area, where the horizontal permeability components parallel to K3 increase from 3% (inverse polarity strata) to 79% (normal polarity strata) (Table 5). The positions of the water springs allow us to check and validate the DFN permeability results; water springs seem to be aligned along a preferential direction in the surroundings of the main morphostructural lineaments (Figure 14).

In particular, most of the water springs seem to be located in a buffer zone of about 500 m with respect to the mapped lineaments. This suggests that these zones can be characterized by a higher fracture intensity and, therefore, drive the fluid flow. Most probably, these areas correspond to fold-hinges characterized by a high abundance of hinge parallel fractures (K3).

Despite these considerations, our DFN simulations are affected by the following main limitations:

• Aperture has been considered constant for all the sets and structural sectors, without differentiating fractures that can have a different origin (e.g., mode-I, II and III) and evolution (impacting on roughness and terminations);
• Major structural elements such as incipient faults or fracture corridors have not been considered;
• Oda’s assumptions (e.g., impermeable matrix, parallel-plate flow, full intra-cell connectivity, and no roughness/tortuosity) may not perfectly fit the fracture formation, in which partially cemented fractures can be present;
• The parametric Monte Carlo framework requires distributional assumptions at every step; each contributes to increasing uncertainties [62].

For these reasons, the permeability values retrieved from DFN and Oda’s permeability simulations should be considered relatively and not as precise and accurate values that can populate a reservoir permeability model. Despite that, our analysis shows that with these limitations, our “simplified” model agrees with the observation about water springs. Further studies based on the accurate set-by-set characterization of fracture aperture, roughness and termination, the definition of major structural elements outcropping in the area, and the use of an alternative framework to Oda’s and Monte Carlo simulations (e.g., FracGen; [62]) will allow us to build a more realistic and accurate quantitative model of the reservoir.

6. Conclusions

In this paper, we present an integrated workflow offering a robust methodology to enhance the DFN model using DOM fracture measurements. Using a combination of traditional fieldwork and DP, it was possible to obtain quantitative information on fracture networks, defining the main fracture sets affecting the Monte Antola Formation in the Upper Staffora Valley area (Northern Apennines, Italy). A proper characterization of the fracture network is crucial, as fractures significantly influence the permeability and fluid flow circulation in the subsurface. This has significant implications for the characterization of various types of fractured reservoirs, such as those related to water supply, hydrocarbons, geothermal energy, and carbon capture and storage. For this reason, the accurate development of DFN models is crucial to predict reservoir behavior. The main results are listed as follows:

• Through the use of DOMs, we were able to obtain high-accuracy and high-resolution 3D fracture measurements and, therefore, accurately quantify the fracture parameters required to model the DFNs.
• DFN models allowed estimation of permeability tensors, suggesting that fluid flow is mainly concentrated along the fracture set K3 and the lineament zones.
• Water springs tend to be located in correspondence with the main morphostructural lineaments detected during the analysis of the aerial LiDAR DEM, suggesting an influence on fluid flow of fractures at a larger scale.

In conclusion, we have shown that the DOM and DFN model can be effective for the analysis of fractured reservoirs, giving the possibility to perform reliable predictions on fluid flow. We established a link between the fracture set observed along outcrop and subsurface fluid flow. Despite that, to fully understand the geohydrological system, further studies are required in order to reconstruct the 3D subsurface geological model, including the main deformation structures.