Fracture networks play a key role in controlling fluid flow, rock stability, and numerous geological processes, yet their characterization in the subsurface is limited by the resolution of borehole and geophysical datasets. In contrast, extensive datasets can be acquired from outcrops by combining direct field observations with remote sensing techniques, such as digital outcrop models (DOMs). In this framework, we developed a comprehensive workflow that spans from outcrop selection to data collection and analysis, addressing all key parameters required for the construction of stochastic Discrete Fracture Network (DFN) models. The workflow integrates two types of digital outcrop models derived from photogrammetric acquisition and reconstruction: point cloud DOMs (PC-DOMs), which are dense 3D point sets containing XYZ coordinates and RGB values, and textured surface DOMs (TS-DOMs), which are obtained by meshing point clouds and draping high-resolution images over their surfaces. The proposed workflow addresses all the main geometrical parameters required for fracture network characterization, including orientation parameters, topological attributes, length and height distribution parameters, the height/length (H/L) ratio, and the Representative Elementary Volume/Area (REV/REA) for P21. Orientation data are collected with a semi-automatic procedure applied to PC-DOMs. Fracture sets are defined with a clustering procedure and different orientation distributions are fitted and tested with goodness-of-fit tests. The topological analysis includes the classification of nodes, the calculation of the connectivity index, and the extraction of the backbone. A new classification scheme called “directional topology” is also proposed, associating each node with its topological type and corresponding fracture set. Chapter 2 is dedicated to the length distribution fitting and parameters calculation. Stochastic models requires a fully specified statistical distribution, this means that the distribution type, and the relative parameters are needed. This motivated the need to develop a new methodology. Our methodology aims specifically at treating censoring bias and obtaining an unbiased trace length statistical model. We propose to tackle the censoring bias by applying survival analysis techniques: a branch of statistics focused on correctly estimating model parameters with data affected by censoring. Secondly, we implemented a ranking system based on statistical distances to select the most representative parametric model. H/L ratio has been calculated assuming that height and length measurements on two adjacent and perpendicular sides of the outcrop are correlated, with the longest fractures being also the tallest. Another separate chapter (Chapter 3) has been dedicated to the problem of Representative Elementary Area (REA) definition al calculation. Here we consider REA as a range, and not a single threshold. The methodology is based on a series of formal statistical tests and diagnostic plots to compare the mean, variance and shape of P21 distributions collected with progressively increasing scan areas. To test all these methodologies, we considered different natural case studies: Pontrelli abandoned quarry located in the Murge Plateau near Altamura (Puglia, Italy), in the forebulge of the Southern Apennines fold and thrust belt. This case study has been the subject of different works in scientific literature, one of the reasons is the combination between a wide and clean pavement where fractures are well exposed, and the associated vertical walls. Colle Salza outcrop is located in the basement of the Western Alps on paragneiss of the Monte Rosa Nappe. The high concentration of censored fractures makes it an excellent case study to test the length distribution fitting method. Lilstock outcrop is located in the southern coast of the Bristol Channel (UK). This outcrop is characterized by an extremely dense fracture network.

Le reti di fratture svolgono un ruolo fondamentale nel controllo del trasporto di fluidi, della stabilità delle rocce e di numerosi processi geologici; tuttavia, la loro caratterizzazione nel sottosuolo è limitata dalla risoluzione dei dati geofisici e dei carotaggi. Questo gap di informazione può essere colmato da dataset acquisiti in affioramento combinando osservazioni dirette con tecniche di telerilevamento, quali i modelli digitali di affioramento (DOM). In tale contesto, è stato sviluppato un workflow completo che copre l’intero processo, dalla selezione dell’affioramento fino alla raccolta e analisi dei dati, trattando tutti i parametri necessari alla costruzione di modelli stocastici DFN (Discrete Fracture Network). Il workflow integra due tipologie di DOM generati tramite acquisizione e ricostruzione fotogrammetrica: Le nuvole di punti (PC-DOM), costituiti da un insieme di punti distribuiti nello spazio, a cui sono assegnate delle coordinate XYZ e dei valori RGB; le superfici testurizzate (TS-DOM), ottenute mediante l’interpolazione delle nuvole di punti e il drappeggio delle immagini ad alta risoluzione utilizzate per ricostruire il modello fotogrammetrico. Il workflow affronta tutti i principali parametri geometrici richiesti per la caratterizzazione delle reti di fratture, inclusi parametri di orientazione, attributi topologici, distribuzioni di lunghezza e altezza, rapporto altezza/lunghezza (H/L) e la definizione del volume elementare rappresentativo o area elementare rappresentativa (REV/REA) per il parametro P21 (intensità di fratturazione). I dati di orientazione vengono acquisiti mediante una procedura semi-automatica applicata ai PC-DOM, da cui vengono estratte faccette poligonali in 2D. I set di fratture vengono definiti tramite l’applicazione di un algoritmo di clustering, e differenti distribuzioni di orientazione vengono fittate e testate mediante l’applicazione di test statistici dedicati. L’analisi topologica include la classificazione dei nodi, il calcolo del connectivity index e l’estrazione della backbone. Viene inoltre proposto un nuovo schema di classificazione, definito “topologia direzionale”, che associa a ciascun nodo sia il tipo topologico sia il corrispondente set di fratture. Il rapporto H/L è stato stimato assumendo la correlazione tra misurazioni di altezza e lunghezza effettuate su due lati adiacenti e ortogonali dell’affioramento, ipotizzando che le fratture più lunghe siano anche quelle più alte. Il Capitolo 2 è dedicato alla determinazione della distribuzione di lunghezza e alla stima dei relativi parametri. I modelli stocastici richiedono sia il tipo, che tutti i parametri di una certa distribuzione statistica. La metodologia proposta affronta specificatamente il bias legato alla censura delle fratture, applicando tecniche derivanti dalla “survival analysis”, un settore della statistica alla stima corretta dei parametri in presenza di censura. Inoltre, è stato implementato un sistema di classificazione basato su distanze statistiche per selezionare il modello parametrico più rappresentativo. Il Capitolo 3 è dedicato alla definizione e calcolo dell’area elementare rappresentativa, qui considerata come un intervallo e non come una singola soglia. La metodologia si basa su una serie di test statistici formali e grafici per confrontare media, varianza e forma delle distribuzioni di P21 ottenute con aree di indagine progressivamente crescenti. Le metodologie sono state testate su differenti casi studio naturali. La cava abbandonata di Pontrelli, situata sull’altopiano delle Murge presso Altamura (Puglia, Italia), nel forebulge dell’avanfossa dell’Appennino meridionale. L’affioramento di Colle Salza, ubicato nel dominio Pennidico, è costituito dai paragneiss della Falda del Monte Rosa. Infine, l’affioramento di Lilstock, situato lungo la costa meridionale del Canale di Bristol (UK), è caratterizzato da una rete estremamente densa.

Casiraghi, S (2026). Characterization and Modelling of Fracture Networks with Applications to the Circulation and Storage of Geofluids. (Tesi di dottorato, , 2026).

Characterization and Modelling of Fracture Networks with Applications to the Circulation and Storage of Geofluids

CASIRAGHI, STEFANO
2026

Abstract

Fracture networks play a key role in controlling fluid flow, rock stability, and numerous geological processes, yet their characterization in the subsurface is limited by the resolution of borehole and geophysical datasets. In contrast, extensive datasets can be acquired from outcrops by combining direct field observations with remote sensing techniques, such as digital outcrop models (DOMs). In this framework, we developed a comprehensive workflow that spans from outcrop selection to data collection and analysis, addressing all key parameters required for the construction of stochastic Discrete Fracture Network (DFN) models. The workflow integrates two types of digital outcrop models derived from photogrammetric acquisition and reconstruction: point cloud DOMs (PC-DOMs), which are dense 3D point sets containing XYZ coordinates and RGB values, and textured surface DOMs (TS-DOMs), which are obtained by meshing point clouds and draping high-resolution images over their surfaces. The proposed workflow addresses all the main geometrical parameters required for fracture network characterization, including orientation parameters, topological attributes, length and height distribution parameters, the height/length (H/L) ratio, and the Representative Elementary Volume/Area (REV/REA) for P21. Orientation data are collected with a semi-automatic procedure applied to PC-DOMs. Fracture sets are defined with a clustering procedure and different orientation distributions are fitted and tested with goodness-of-fit tests. The topological analysis includes the classification of nodes, the calculation of the connectivity index, and the extraction of the backbone. A new classification scheme called “directional topology” is also proposed, associating each node with its topological type and corresponding fracture set. Chapter 2 is dedicated to the length distribution fitting and parameters calculation. Stochastic models requires a fully specified statistical distribution, this means that the distribution type, and the relative parameters are needed. This motivated the need to develop a new methodology. Our methodology aims specifically at treating censoring bias and obtaining an unbiased trace length statistical model. We propose to tackle the censoring bias by applying survival analysis techniques: a branch of statistics focused on correctly estimating model parameters with data affected by censoring. Secondly, we implemented a ranking system based on statistical distances to select the most representative parametric model. H/L ratio has been calculated assuming that height and length measurements on two adjacent and perpendicular sides of the outcrop are correlated, with the longest fractures being also the tallest. Another separate chapter (Chapter 3) has been dedicated to the problem of Representative Elementary Area (REA) definition al calculation. Here we consider REA as a range, and not a single threshold. The methodology is based on a series of formal statistical tests and diagnostic plots to compare the mean, variance and shape of P21 distributions collected with progressively increasing scan areas. To test all these methodologies, we considered different natural case studies: Pontrelli abandoned quarry located in the Murge Plateau near Altamura (Puglia, Italy), in the forebulge of the Southern Apennines fold and thrust belt. This case study has been the subject of different works in scientific literature, one of the reasons is the combination between a wide and clean pavement where fractures are well exposed, and the associated vertical walls. Colle Salza outcrop is located in the basement of the Western Alps on paragneiss of the Monte Rosa Nappe. The high concentration of censored fractures makes it an excellent case study to test the length distribution fitting method. Lilstock outcrop is located in the southern coast of the Bristol Channel (UK). This outcrop is characterized by an extremely dense fracture network.
BISTACCHI, ANDREA LUIGI PAOLO
AGLIARDI, FEDERICO
Reti di fratture; Modello digitale; Censura; REA; DFN
Fracture networks; Digital model; Censoring; REA; DFN
English
20-feb-2026
38
2024/2025
open
Casiraghi, S (2026). Characterization and Modelling of Fracture Networks with Applications to the Circulation and Storage of Geofluids. (Tesi di dottorato, , 2026).
File in questo prodotto:
File Dimensione Formato  
phd_unimib_813592.pdf

accesso aperto

Descrizione: Characterization and Modelling of Fracture Networks with Applications to the Circulation and Storage of Geofluids
Tipologia di allegato: Doctoral thesis
Dimensione 10.72 MB
Formato Adobe PDF
10.72 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10281/610675
Citazioni
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
Social impact