FISSURE FORMATION AND SUBSURFACE SUBSIDENCE IN A COALBED FIRE Taku S. Ide 1, David Pollard 2, and Franklin M. Orr, Jr. 3 1,3 Dept. of Energy Resources and Engineering, 2 Dept. of Geological and Environmental Sciences Stanford University, Stanford, CA 94305 e-mail: [email protected]
was started in May of 1962 when the local government Abstract decided to burn an unregulated trash dump in an abandoned strip mine to reduce trash volume and Coalbed fires are uncontrolled subsurface fires control rodents. The fire ignited an anthracite outcrop, that occur around the world. These fires are believed to eventually connected to and spread through be significant contributors to annual CO emissions. 2 underground tunnels, and has been burning since. Although many of these fires have been burning for Fissures created by the coal fire emit assorted hot gases, decades, researchers have only recently begun to some of which are toxic. A combination of subsidence investigate physical mechanisms that control fire and emissions from fissures has caused the town of behavior. One aspect of fire behavior that is poorly Centralia to be abandoned (DeKok, 1986, GAI characterized is the relationship between subsurface Consultants, 1983). combustion and surface fissures. At the surface above The particular fire examined in this study, called many fires, long, wide fissures are observed. At a the North Coalbed Fire, to distinguish it from other coalbed fire near Durango, CO., these fissures form active fires in the region, was discovered in 1998 on the systematic orthogonal patterns that align with regional Southern Ute Indian Reservation when sets of fissures joints in the Upper Cretaceous Fruitland Formation. that are orthogonal to each other—similar to those at Understanding the mechanisms that form these fissures other coal fires around the world—appeared at the is important, as the fissures are believed to play vital surface (Williamson, 1999). Anecdotal evidences roles in sustaining the combustion in the subsurface. In provided by local Southern Ute Tribe members (Ide, some of the coalbed fire simulation models available 2007) suggest that the fire may have been smoldering today, these fissures are treated as fixed boundary for decades prior to the reported date of discovery. The conditions. We argue, using data collected, field fire continues to burn today. The research effort observations and simulation result, that there exists a described here is an attempt to understand whether coal relationship between the location and magnitude of combustion followed by subsurface subsidence can subsidence caused by the fire and the opening of produce fissures with systematic patterns at the surface. fissures. The results presented suggest that fissures are Subsidence can occur when a burned coalseam loses its believed to open when subsurface subsidence gives rise structural integrity and collapses under the weight of to tensile stresses around pre-existing joints. the overburden. Understanding the formation of fissures is important, as they appear to foreshadow the direction Keywords: of the combustion front propagation and may play a key coalbed fire, coal fire, subsidence , pre-existing role in sustaining the underground fire. joints, fissures, numerical modeling, CO 2 First, we summarize the San Juan Basin geology, highlighting key features in the NW section of the Introduction basin, where the coalbed fire is located. Second, we characterize the geological features and the surface Uncontrolled subsurface fires in coalbeds account anomalies in the vicinity of North Coalbed Fire. for significant releases of CO to the atmosphere. One 2 Surface topography, images of fissures overlying the of the world’s largest active coalbed fires has been coalbed fire, and measurements of fissure orientations documented in Wuda, China (Dai et al., 2002), where are presented. We also outline the process of digitizing the estimated annual loss of coal is around 200,000 features over the North Coalbed Fire and describe how tons, equivalent to a yearly emission of ~1.5Mt of CO 2 they were combined with the subsurface information (Kuenzer et al., 2005). Coalbed fires are burning in obtained from the wells drilled in the area. In the third many locations in China, Indonesia, India, and the section, the field data and previous geological surveys United States (Stracher and Taylor, 2004). They can be of the area are used to suggest how subsidence can open started naturally by forest fires that burn near an pre-existing joints in the area, leading to the formation outcrop, by lightning strikes, by human activities, or by of surface fissures over the combustion region. Finally, spontaneous exothermic reactions of pyrites (DeKok, we model this phenomenon using a simple boundary 1986). Forest fires in Indonesia in 1997 and 1998 element numerical code, and explore relationships ignited hundreds of coal fires at outcrops (Brown, among key variables that contrast subsidence activities 2003). A subsurface fire near Centralia, Pennsylvania,
Stanford Rock Fracture Project Vol. 20, 2009 C-1 and surface deformation. The results and the limitations the coalbed fire, formations above the dotted line—the of applicability of this simulation model are discussed. Kirtland Shale and most of the Fruitland Formation— have been removed by weathering and erosion. The San Juan Basin Geology Kirtland Shale and the Fruitland Formation lie atop of the Pictured Cliffs Sandstone (PC), which was The San Juan Basin is an asymmetric, coal bearing deposited as regressive marine sands (Condon, 1988) basin that covers approximately 16,800 – 19,400 square parallel to the shoreline stretching northwest-southeast kilometers, stretching approximately 145km west-east (Fassett, 2000). The Fruitland Formation is a mixture of and 160km north-south (Fassett, 2000, Kelso et al., mudstones, siltsones, carbonaceous shales and coals 1988). It is located near the Four Corners, and spans deposited landward and parallel to this shoreline across northwest New Mexico and southwest Colorado (Fassett, 2000). Coalseams in the Fruitland Formation as shown in Figure 1. The basin is well characterized are often referred to as Lower Coal, Middle Coal and due to the abundance of both coal and coal-bed Upper Coal, and the thickest, most continuous coalbeds methane resources in the Upper Cretaceous Fruitland are found in the Lower Coal Zone in the northeastern Formation (Figure 2). One study has estimated that a region of the San Juan Basin (Sandberg, 1988). The coal-bed methane reserve of nearly 1.4 x 10 12 m 3 (50 x Lower Coal is burning at the North Coalbed Fire. 10 12 ft 3) adsorbed onto 219 x 10 9 metric tons of coal that underlies the San Juan Basin (Kelso et al., 1988). The flat, Central Basin is bounded by several key geologic features, which are described in detail in previous geologic surveys of the area (Fassett, 2000, Kelso et al., 1988, Lorenz and Cooper, 2000). The western and northwestern regions of the basin are circumscribed by the Defiance and the Hogback monoclines, respectively, and the Nacimiento uplift borders the basin on the east side (Lorenz and Cooper, 2000). As Figure 1 shows, the North Coalbed Fire is located along the Hogback Monocline in the northwestern portion of the basin. The Hogback monocline is believed to have formed either due to the shortening of the Cretaceous strata that induced a right- lateral strike-slip motion along the western and eastern margins during the Laramide orogeny (Lorenz and Cooper, 2000), or through reactivation of western dipping thrust faults underlying the Hogback monocline that resulted in the uplift (Taylor and Huffman, 1988). In the former view, the shortening can be attributed to the Zuni uplift thrusting northward and north northeastward into the San Juan Basin from the south, and the San Juan uplift indenting southward into the Figure 1: San Juan Basin and its characteristic basin (Lorenz and Cooper, 2000). Today, only the geologic features. The North Coalbed Fire location forelimb of the Hogback Monocline is exposed and is highlighted in the box in the northwestern corner some of the formation members of the Upper of the basin along the Hogback Monocline. The Cretaceous are exposed on the western side of the green area denotes outcrops of Pictured Cliffs basin. Along the northern perimeter of the Basin, sandstone. Figure reproduced from Lorenz and Cooper, 2003. including areas affected by the North Coalbed Fire, thick coalseams crop out along the Hogback monocline At the North Coalbed Fire, 14 boreholes were (Kaiser et al., 1991). drilled in 2007 over an area of 600m x 200m. The high Formations that make up the Upper Cretaceous density of boreholes allowed reliable subsurface rocks in the San Juan Basin are described by Molenaar correlations to be made at the site. Both the PC and the (Molenaar and Baird, 1992). The Fruitland Formation, Fruitland Formation rise in a step-like fashion from the which includes the coalbed fire, and adjacent geologic southwest to the northeast with respect to the units are depicted in the stratigraphic column in Figure isochronously deposited Huerfanito Bed in the Lewis 2. The left column is representative of the entire San shale, representing a migrating regression-transgression Juan Basin. The right column shows the top 25m of cycle over 1.2 million years (Fassett, 1971, Sandberg, rock found directly over the North Coalbed Fire. Above 1988). The deposition pattern suggests that the
Stanford Rock Fracture Project Vol. 20, 2009 C-2 subsurface correlation along the shoreline in the approximately 600 m x 200m. There are signs of northwest-southeast direction is warranted, as this is the coalbed fires underlying the bare patch of land south trend of the long axis of most coal deposits (Fassett, southwest of the North Coalbed Fire, but only minor 1988). The echelon geometry of coal deposits can make surface deformation is observed; thus this area is not subsurface correlations difficult in the transverse included in the surveys. The lack of vegetation there is direction (Fassett, 1988), but it has been shown that interpreted as largely due to surface forest fires. Fruitland Formation coalbed correlation was possible when well logs spaced less than 4km apart were obtained (Fassett, 1971, 1988). The San Juan Basin contains several sets of natural fractures that have been extensively mapped and documented. Previous studies have offered various explanations for fracture formations and they are summarized in Lorenz and Cooper (2000). Ruf (2005) suggests that the fractures formed due to post-Laramide extension, while Taylor and Huffman (1998) describe a Proterozoic crystalline basement with reactivated faults that may have influenced the orientations of the fractures in overlying strata. Lorenz and Cooper (2000) suggest that the orientations of the fractures are most influenced by the formation of tectonic features such as the San Juan uplift and the Zuni uplift (cf. Figure 1) during the Laramide Orogeny. Despite the competing explanations of the origins of the fractures, orientation measurements in various parts of the San Juan Basin are consistent across many studies. The earliest fracture orientation study of the Figure 2: Stratigraphic columns representative of San Juan Basin concluded from aerial photography that the San Juan Basin (left) and the top 25 meters of the lithology over the North Coalbed Fire (right). northeast (N10E to N60E) and northwest (N15W to Over the coalbed fire, formations above the blue N75W) trends occurred most frequently (Badgley, dotted line have been removed due to weathering. 1962, 1965, Kelley and Clinton, 1960). Their findings Stratigraphic column on left adapted from Molenaar, are generally supported by more recent measurements 1977. (Figure not drawn to scale). (Condon, 1988, 1997, Lorenz and Cooper, 2003, Ruf, 2005, Taylor and Huffman, 1988). The most relevant A cross-section, A-A’ is drawn (Figure 3b) by study for the North Coalbed Fire was carried out by superimposing a USGS geological survey map over the Condon (1988), who presents joint orientations and coal satellite image in Figure 3a. The cross-section line is strikes found within the Southern Ute Indian roughly perpendicular to the strike of the Hogback Reservation. His findings are discussed in detail in the monocline. The cross-section shows that the Fruitland ensuing section, and they are compared to the fissure Formation crops out along the Hogback Monocline orientations that were measured over the North Coalbed limb (cf. Figure 1). To the northwest of the Hogback, Fire. only the Lewis Shale—containing the Huerfanito Bentonite Bed—is observed. The region affected by the North Coalbed Fire coalbed fire is located near the coal outcrop along the A satellite image of the area bounded by the red Hogback, and is circumscribed by the dotted box. In dotted box in Figure 1 is shown in Figure 3a. The this region, the local topography slopes between 5 and 9 dotted rectangle depicts the region where surface degrees to the southeast, and the coal layer dips 6 to 15 anomalies associated with the North Coalbed Fire are degrees in the same direction (Condon, 1988). Both the observed. The lack of vegetation over the fire can be surface topography and the coalseam flatten towards attributed to several factors: a surface forest fire, death the southeast in the direction of the Central Basin (cf. of vegetation due to toxic combustion fumes emanating Figure 1). The continuous and low permeability from the subsurface, and intentional tree removal to Kirtland Shale Formation, which is absent over the prevent future forest fires. The North coalbed fire is North Coalbed Fire, caps the Fruitland Formation to the contained between latitude N37 o01’57” and southeast. N37 o02’24” and longitude W108 o06’36” and Many fissures are exposed on the surface overlying W108 o06’18”, and has an aerial extent of the North Coalbed Fire. The fissures are distinguished from regional joint sets in the same strata because
Stanford Rock Fracture Project Vol. 20, 2009 C-3 fissures typically have widths on a decimeter scale, displacement and surface sediment layer rotation on whereas joints have apertures less than 0.5 cm (Condon one side of the fissure. The other side of the fissure 1988). Some of these fissures emit high temperature does not show much displacement. Figure 4c shows an combustion gases, indicative of the active fire below, example of molehill fissures, where surface layers of while others are at ambient temperature. sedimentary rock are rotated to form an apex. At molehill fissures with visible fractures at the surface, combustion gases with temperatures as high as 290 oC (550 oF) have been recorded. The temperatures at the fissures were measured using a thermal gun, Raynger 3i Series, made by Raytek. Figure 4d is an example of narrow fissures, many of which are located in the northern most portion of the field, and these emit the hottest exhaust gases recorded in the field at about 1000 oC. Both the molehill and narrow fissures are about 0.15m in width. All of the fissures appear to be opening Mode I fractures (Pollard and Aydin, 1988), as displacements are dominantly orthogonal to the fracture surfaces.
Figure 3: a) A satellite image over the North Coalbed Fire. The red dotted box outlines the region affected by the underlying fire. A cross-section line A-A’ is used in Figure 3b. Satellite image is provided by Googlemaps. b) Cross section A-A’ showing surface topography and representative subsurface stratigraphy in the vicinity of the North Coalbed Fire. The coalbed fire is located near the coal outcrop inside of the red dotted box. Kl = Lewis Shale, Kp = Pictured Cliffs Sandstone, Kf = Fruitland Formation, Kkl = Kirtland Shale.
Four types of fissures have been observed. Examples are shown in Figure 4: gaping fissures, plateau/offset fissure, molehill/buckling fissures, and narrow fissures. The gaping fissure in Figure 4a is wide enough for an adult to climb inside. Typical gaping fissures are 0.15~0.3 m wide at the surface and are Figure 4: a) a gaping fissure with an adult inside b) a often wider below the soil level. Based on observations plateau fissure, c) a molehill fissure with a 0.15 m aperture at the apex, d) a narrow fissure venting made inside of the gaping fissure in 4a, many fissures O may be connected to each other in the subsurface. exhaust gases exceeding 900 C. Gaping fissures are at ambient temperatures. The surface sediment layers do not show significant rotation Orientations of the fissures are systematic, and they around the edges of the fissure. Rather, the fissure often form orthogonal patterns at the surface. The appears to have been pulled apart from either side. directions and the lengths of 165 fissures are Figure 4b is an example of a plateau fissure. Plateau represented on a rose diagram in Figure 5. The length fissures have similar apertures at the surface as gaping has been made dimensionless with respect to the fissures but fissures of this type show significant longest fissure in the field, which is 75m. The diagram shows that there are three main fissure directions over
Stanford Rock Fracture Project Vol. 20, 2009 C-4 the North Coalbed Fire and that the longest and most frequently occurring fissures, F1, have azimuths approximately in the N50E direction. The next most prominent set, F2, strikes in the N35W direction, roughly perpendicular to the first set. The third set, F3, is directed towards the North, and these have similar lengths to the N35W set. The fissures frequently occur together in approximately orthogonal pairs, including members of the N50E and N35W sets. The azimuths of the fissures were compared with observations of joint orientations reported in Condon, 1988. Condon measured 1,600 joints and coal cleats at 37 different outcrop locations on the Southern Ute Indian Reservation. Of the 37 measurement stations, 8 of them are located along the Hogback Monocline and are spaced approximately 2km apart. Most of his measurements were for fractures found in formations of the Upper Cretaceous, the majority of which are in the Kirtland Shale, Fruitland Formation and the Pictured Cliffs Sandstone. Four dominant joint sets, labeled J1 through J4, were described. Their stereonets are Figure 6 (bottom): Four stereonets reproduced from reproduced in Figure 6. A comparison of Figures 5 and Condon, 1988. a) J1 joint b) J2 joint c) J3 joint d) J4 joint. 6 shows that F1 corresponds to J3, F2 to J4, and F3 to
J2 based on similarities between the fissure orientations While J1, J2 and J3 are stratigraphically continuous and joint orientations. Typically, the joints occur in through multiple beds, J4 fractures only cut through the pairs—a J1-J2 pair and a J3-J4 pair—much like the sandstone in interbedded sequences of sandstone and fissures F1 and F2 that form orthogonal pairs above the shale. Their orientation ranges are as follows: J1 North Coalbed Fire. Condon classifies the J1~J4 joints (N4E~N23E), J2 (N72W~N83W), J3 (N41E~N64E) as extension joints, due to the lack of features such as and J4, (N24W~N49W). Joints J1, J2, and J3 have slickenside striations that would suggest lateral shear exposed lengths of roughly 1 to 5m, and are spaced movement and the presence of plumose structures, 0.15m to 6m. J4 exposed lengths are less than 2m, and arrest lines, and twist hackle features that indicate the spacings are more inconsistent (Condon, 1988). extension joints (Condon, 1988). The fissures above the North Coalbed Fire were mapped using a pack-mounted GPS receiver in order to place them with respect to the topography of North Coalbed Fire site. In addition, the GPS was used to digitize the topography and to mark the locations of boreholes that have been drilled in the area. The GPS device used in the survey was a Trimble ProXH, and the points recorded had better than 1 meter accuracy, with most having better than 0.5 meter accuracy after differential correction. Figure 7a shows the digitized representation of the surface overlying the North Coalbed Fire. The contour map approximately represents the region bounded by the red dotted box in Figures 1 and 3a. The left edge of this map traces the contact line between the Fruitland Formation and the Pictured Cliffs Sandstone (cf. Figures 3a, 3b). The dominant N50E trending fissures are nearly parallel to the local strike of the Hogback Monocline. The black lines represent narrow fissures that have been grouted
Figure 5: A rose diagram showing the orientation using a specialized concrete produced by Goodson and and the lengths of the fissures found above the Associates (Williamson, 1999). The concrete was North Coalbed Fire. The characteristic length scale injected into identified openings in 2000 in an attempt is ~100m. to smother the fire. Boreholes were drilled around the
Stanford Rock Fracture Project Vol. 20, 2009 C-5 grouted fissures, and thermocouples were installed to that the depth to coal is approximately 20 meters. The allow monitoring of temperature changes. These cross-section is approximately perpendicular to the boreholes are shown as open white triangles in Figure prominent N50E fissures. Any boreholes or fissures that 7a. Although the attempt to extinguish the fire was not lie close to this line are plotted along with the surface successful, the driller’s logs from 2000 provide topography in Figure 7b. Where available, a valuable insight into the subsurface. This particular combination of driller’s-logs and well-logs were used to extinguishing method failed mainly because it was not identify the depths and thicknesses of void, ash and possible to locate and fill all existing fissures in the coalseams at each intersecting well. If data were region. Fourteen additional boreholes were drilled in missing at a well, lithologies below it are left blank in 2007. These boreholes are marked with solid triangles Figure 7b. Fissures that intersect the cross-section line in Figure 7. For these new boreholes, driller’s logs were A-A” are represented using red circles. It is worth obtained, and most of the boreholes were logged using noting that in borehole 11, no signs of coal combustion caliper, density and gamma ray logs. In one of the were apparent. The last set of fissures occurs up dip of boreholes, borehole 7, an 80ft core was obtained. borehole 11, and there are no fissures down dip of this The surface information in Figure 7a can be related borehole. A black line in Figure 7b connects the bottom to the subsurface information by creating a cross- of the coalseam. section along the line A-A”. This cross section shows
Stanford Rock Fracture Project Vol. 20, 2009 C-6
Figure 7: a) A contour map of the North Coalbed Fire site, fissures and wel ls created using a pack - mounted GPS. Red lines are gaping fissures, green lines are plateau / offset fissures, magenta lines are molehill/buckling fissures, and blue lines are narrow fissures. The cross-section is created along A-A”. b) A cross-section of the North Coalbed Fire site along A-A”. Red circles indicate locations of fissures that intersect the line. Numbers above the surface represent well numbers.
Stanford Rock Fracture Project Vol. 20, 2009 C-7 Formation of Fissures – A Conceptual Model We hypothesize that fissures are created from pre- existing fractures in the overlying sandstone and shale that widen when subsidence occurs. Subsidence results when the burned coal loses structural integrity and collapses under the weight of the overburden. In Figure 7b, the occurrences of surface fissures coincide with regions where void and ash were encountered during drilling. For example, borehole 4 (solid triangle, Figure 7b) located near the peak of the topography contains only ash and coal. The lack of a void in this well Figure 8: A conceptual model of subsidence suggests that subsidence occurred, and thus both the ash associated with long wall mining. Tensile stress fractures associated with the collapse are shown. L and void are fully compacted. Fissures located between signifies the length of coal excavated, and the d the boreholes 4 and 5 may have resulted from this depth below the surface of the seam being mined. subsidence. Similarly, some of the void space may have (Whittaker and Reddish, 1989) been compacted at well 5, causing a fissure to open-up down dip of this borehole. Previous studies of coalbed fires have also The notion of subsurface compaction leading to suggested that subsurface subsidence leads to the surface deformation and fracturing is not new. It is formation of fissures at the surface (Buhrow et al., explored in Whittaker and Reddish’s work on 2004, Cao et al., 2007, Chen, 1997, Dunrud and subsidence related to long-wall coal mining Osterwald, 1980, Gielisch and Kuenzer, 2003, Kuenzer, (Whittakker and Reddish, 1989). They describe surface 2007a, 2007b, 2008, Litscheke, 2005, Sokol and profiles associated with various subsurface subsidence Volkova, 2007, Wessling, 2007, Wessling et al., 2008, configurations. Their work is based on examples from Zhang, 2007). Figure 9 is from one such study of a various long-wall coal mining sites and includes field coalbed fire in China, where subsidence apparently observations, experimental, and numerical results. In played a significant role in opening surface fissures long-wall mining, the excavation front advances much indicated by the arrows. Most of these studies did not like we envision the combustion front may move examine whether fissures resulted from the widening of through the coalseam in a coalbed fire. Figure 8 is a pre-existing joints in the region. In Chen’s work (Chen, conceptual model of subsidence near a long-wall 1997), it is shown that fissure orientations coincide with mining process (Whittaker and Reddish, 1989). Here, joint orientations in the sandstones overlying the tensile fractures develop in strata immediately coalbed fire in Ruqigou, China. In this study, however, overlying the collapse. This figure can be also be used relationships between variables such as the location and to illustrate an empirical relationship presented in their magnitude of subsidence and the widths of surface work, which shows that the ratio of the length of coal fissures were not established (Chen, 1997). In a excavated (L) to the depth (d) of excavation must coalbed fire combustion simulation model presented by typically exceed 1.4 for maximum subsidence to occur Huang et al. (2001), the fissures were modeled as fixed (Whittaker and Reddish, 1989). Adjacent unmined parts boundary conditions—through which exhaust gases can of the coalseam and a natural arch that develops above escape and fresh oxygen can enter—irrespective of the the coal removal site may be capable of supporting location of the combustion front. Similarly, in the most of overburden when L/d is less than 1.4 numerical model of Wessling et al., mechanical (Whittaker and Reddish, 1989). There are two key processes such as subsidence and fissure openings were differences between their study and the work presented not considered (Wessling et al., 2008). By establishing in this paper. First, the North Coalbed Fire is burning a first order relationship between combustion front ~20m below surface, whereas longwall mining location and fissure opening width as a function of typically occurs at much greater depths (Whittaker and governing variables such as depth, length of collapse, Reddish, 1989). Second, pre-existing vertical joints and proximity of preexisting fissures, and the stiffness of their response to subsurface compaction are not the overburden rock, we hope to aid future numerical discussed in Whittaker and Reddish. modeling of coalbed fires.
Stanford Rock Fracture Project Vol. 20, 2009 C-8 There are two prominent features at the outcrop: the fissure that runs down the middle of the outcrop, and the deformed ash layer towards the bottom of the outcrop. The fissure down the middle of the photograph is labeled as Fissure 2, and this fissure has an opening of around 0.5m at the surface. When the coalseam was consumed by a combustion front moving from the right to left, we suggest the ash deformed by compaction under the weight of the overburden. The maximum collapse recorded at the outcrop is ~1.5m. Features at this outcrop such as Fissure 2 and the subsided ash layer were mapped using a laser
rangefinder produced by LaserCraft Inc. (LaserCraft, Figure 9: A picture of a subsided area and fissures 2007). The digitized version of the outcrop is nearby (indicated by arrows) in a coalbed fire at the juxtaposed next to the photo of the outcrop in Figure Wuda Syncline, Inner Mongolia Autonomous 10b. Note how the tabular coalseam is deformed due to Region, China. (courtesy of Chris Hecker, ITC, collapse of the ash layer. Above the collapse, opening 2008) fractures, much like those depicted in Figure 8, were observed. In addition, four fissures with more modest Observations at an outcrop about 1km north of the openings were mapped over the collapse. In subsequent North Coalbed Fire shows how subsurface subsidence sections, when we compare numerical solutions to our can cause pre-existing fractures to open at the surface. measurements at this outcrop, we assume that Fissure 2 This outcrop exposes a fossilized coalbed fire, in Figure 10b widened largely due to the collapse of the subsidence, extension fractures and fissures. At this combusted coal layer and that the weathering process to outcrop, shown in Figure 10a, a coalseam is overlain by expose the outcrop did not significantly enhance the 10m of sandstone, shale, and siltstone. A person 1.5m opening. tall standing to the right is used as a scale.
Figure 10: (left) A picture of an outcrop near the North Coalbed Fire wi th an exposed fossilized coalbed fire, subsidence and associated surface fissures. (right) Some features from the same outcrop mapped using a Laser Range Finder. The major features are depicted using thicker lines.
Stanford Rock Fracture Project Vol. 20, 2009 C-9 perturbed along the horizontal elements to simulate a Numerical Modeling collapse as the coal burns, and the elastic domain is deformed as a result of this perturbation. Stresses and Numerical models were employed to examine displacements that arise at any point in the domain can whether pre-existing joints could pull open to form be calculated by combining the contribution of stresses fissures when subjected to the stresses due to the and displacements from each element (Crouch and overburden weight and those induced by a subsurface Starfield, 1983). The stress distribution in the elastic collapse. The mechanical effects of subsurface material is a function of the location and orientation of subsidence on the jointed strata overlying the coal were the boundary elements and the boundary conditions on modeled using a Boundary Element Method (BEM) them. formulation for a line source of displacement Figure 11 defines the variables and applicable discontinuity in an elastic half plane. This problem dimensionless groups used in this modeling. E is formulation is an adaptation of the displacement Young’s modulus, and σ is the normal compressive discontinuity method (Crouch and Starfield, 1983). The zz stress defined along the horizontal elements to simulate BEM code is a modified version of Martel’s Matlab the downward pressure due to the overburden. These BEM code (Martel, 2003), which in turn is based on the variables both have units of stress (MPa). All other original Fortran code presented in Crouch and Starfield variables have units of length (m), and they are defined (Crouch and Starfield, 1983). as follows: fd is the height of the vertical fracture, fl is Several key assumptions are made in this model, the distance between the vertical fracture and the edge including an elastic homogeneous medium with a of the horizontal collapse, d is the depth, and a is the reduced stiffness coefficient, infinitesimal strain, a state horizontal length of the collapse. of plane strain, and a flat traction free surface. The rock above the collapsing coalseam is modeled using a reduced stiffness coefficient in place of explicitly modeling each fracture and joint in the overburden. The use of reduced stiffness coefficients compared to the values measured in experiments is justified in the previous literature on fractured rock deformation (Berest et al., 2008, Sanz, et al., 2008). Hooke’s Law is used to relate stress and strain, while the infinitesimal strain assumption dismisses higher order displacement derivative terms in the relationship between strain and displacement (Malvern, 1960). The infinitesimal strain assumption admits the use of the method of Figure 11: Definitions of variables and superposition, which is used to calculate stress and dimensionless groups used in the BEM model. E displacement distributions in the domain and to create a and σzz have units of MPa, while fd, fl, a and d have half-plane surface. The plane strain assumption restricts units of m. any displacement perpendicular to the plane of the model (Crouch and Starfield, 1983). Finally, a flat We first introduce a domain with no vertical joints surface is modeled rather than the actual topography in order to illustrate the stress distribution that arises as over the coalbed fire outcrop for simplification. a result of collapse of a continuous overburden. We Although these assumptions lead to a model that, at then introduce a vertical fracture, and compare how it best, approximates the deformation of jointed rock over reacts to a subsidence event when located in regions of a coalbed fire, it nevertheless helps to build an intuitive induced tensile stress. This is followed by a sensitivity understanding between subsidence and fissure opening, analysis to demonstrate the behavior of vertical joints which has not been explored in today’s coalbed fire with respect to various model variables. Finally, a BEM modeling literature (Huang, 2001, Wessling, 2007, model is constructed from the outcrop mapped using 2008). the laser range finder (cf. Figure 10b), and simulation In the BEM code, discretized horizontal elements results are compared to field observations. are used to model the coalseam, and discretized vertical In the first example, a 12m horizontal line of traction free elements are used to model pre-existing elements that is located 10m below the surface is joints in the overburden. The infinite plane is deformed by applying a uniform compressive stress of transformed into a half-plane by introducing the 0.25MPa, which is exerted by the weight of the principle of superposition to create a traction free overlying rock. The elastic modulus of the overburden boundary condition along the x-axis (Crouch and is 10MPa, and a maximum compaction of 1.5m is Starfield, 1983). Stress boundary conditions are induced at the horizontal elements. Figure 12 depicts
Stanford Rock Fracture Project Vol. 20, 2009 C-10 the distribution of the horizontal component of normal Horizontal surface displacements are continuous when stress, σxx , in response to the inward directed there are no vertical fractures since the domain is displacement discontinuity on the horizontal elements. modeled as an elastic medium. In contrast, surface The sign of σxx at the surface is indicated by the words displacements are perturbed during the subsidence tensile (+) and compression (-). The blue solid line when a fracture is located within the tensile region. The along the bottom of the figure indicates the horizontal right side of the fracture—the edge closer to the elements subject to subsurface subsidence. We suggest induced subsidence—displaces towards the region of that this inward directed relative motion is similar to the collapse horizontal elements, while the left side what would occur as compaction of the coalseam does not displace as much, so the model fracture opens. developed during burning. Directly above the elements This result shows how pre-existing joints in tensile at the surface σxx is compressive. The greatest regions may widen to form fissures. concentrations of surface tensile stresses emanate A sensitivity study was undertaken to explore how diagonally upward from the ends of the line of collapse. the opening of vertical joints are influenced by the A modification to the first simulation investigates governing variables presented in Figure 11. Four the effects of the collapse on traction free vertical dimensionless groups are chosen to represent the joints. The setting and the parameter values are the relationships between the variables. Π1, or E/ σxx , is the same as the first simulation (cf. Figure 12), except ratio of Young’s modulus of the rock to the stress vertical elements are introduced to simulate the joint. imposed along the horizontal elements to induce The vertical elements are placed at x = -12m, where subsidence. Π2, or fd/d , is the ratio of the height of the tensile stresses found in the first case (cf. Figure 12). vertical fracture to the depth at which compaction Figure 13 shows the model geometry and the resulting occurs. Π3, or a/d , is the ratio of the horizontal length of normal horizontal stress ( σxx ) distributions when a subsidence to the depth. Finally, Π4, or fl/d , is the ratio horizontal collapse occurs near the vertical fracture. A of the distance between the vertical fracture and the comparison of Figures 12 and 13 shows that if a vertical edge of the collapsed region to the depth. These groups joint exists off to the side of the compaction zone, σxx are plotted against a dimensionless length scale, relaxes and becomes less tensile as the joint opens. Umax Opening / U max Collapse , which relates the Figures 14 compares the horizontal displacements horizontal displacement of the joint at the surface to the between the two cases discussed in this section, the maximum vertical subsurface subsidence along the model without vertical fracture and the model with the horizontal elements. Here, a negative dimensionless vertical fracture. The horizontal displacements at the length means that the edges of the vertical elements surface have been made dimensionless by the displace away from each other, or in other words, the maximum vertical subsidence induced along the joint opens. Simulation results show that this horizontal elements. Here a positive displacement dimensionless length does not vary with respect to Π1, signifies a movement to the right, and a negative and thus the following analyses are limited to displacement indicates a movement to the left.
Figure 12: Subsidence along horizontal elements (blue solid line, bottom center) and resulting stress distributions in the domain. Tensile stresses emanate diagonally upwards from the edge of the horizontal elements. Colorbar in MPa.
Figure 13: A vertical joint located to the left of the collapsed region. Tensile stresses near the vertical joint are relaxed due to the traction free elements.
Stanford Rock Fracture Project Vol. 20, 2009 C-11 Figure 14: Horizontal displacements at the surface for cases with no fracture (solid line) and with a fracture (dotted line). The model with a fracture in the domain shows a displacement discontinuity indicating an opening.
demonstrating the dependence of the dimensionless Π3, which represents the length over which the collapse length scale on Π2, Π3, and Π4. The results presented occurs, the fissure opening again depends strongly on from the sensitivity analyses can be used, on a first the location of the vertical joint. The fissure opening order basis, to estimate the location and the magnitude decays to 0 far away from the subsidence and is the of the subsidence when the only the widths of the widest between 0.5< fl <1.0 depending on the value of surface fissures are known. Π3. As Π3 increases, the horizontal displacement at the Figure 15 is a plot of the relationship between the surface increases, which makes sense since a longer dimensionless opening ( Umax Opening / Umax Collapse ) subsidence length leads to a greater tensile stress and Π4 ( fl/d ). Each line represents a different value of emanating upwards towards the surface. The location Π2 ( fd/d ). In the following discussion, the maximum where the maximum fissure opening is observed moves vertical subsidence length, Umax Collapse , depth, d, and closer to the edge of the horizontal compaction when length of subsidence, a, will be fixed to simplify our Π3 increases. In other words, the region of tensile analysis. As a consequence of fixing both d and a, Π3, stresses that extends upwards towards the surface lies the ratio between the two variables is constant. more directly above the horizontal compaction as the Variables a and d are specified such that Π3 = 1.0. length of subsidence increases when depth is constant. Figure 15 shows that for a constant value of Π2, or for a fixed height of the fracture, a maximum horizontal displacement at the surface is observed when Π4~0.8. The fissure width reaches a maximum when it is located diagonally above and to the side of the zone of compaction, consistent with where a concentration of tensile stresses was observed in Figure 12. The fissure opening decays to 0 as the vertical joint moves farther away from the region of subsidence regardless of the height of the fracture. This result is reasonable since the stresses associated with the subsidence decay with increasing distance. Figure 15 also shows that when Π2, or the height of the pre-existing joint, increases, the magnitude of the opening also increases. Based on the results, when Π3 = 1.0, the maximum fissure opening is observed when the pre-existing joint is located at Π4 ~ 0.8 and is stratigraphically continuous down to the
collapse horizon, or Π2 = 1. Figure 15 (above): Dimensionless length vs. Π (fl/d). Figure 16 is a similar plot. It shows how the 4 Each line represents a different value of Π2 (fd/d), dimensionless length scale ( Umax Opening / U max while Π3 (a/d) is kept constant at 1.0. Collapse ) depends on Π4 ( fl/d ) for varying values of Π3 (a/d ) with both the maximum subsidence distance (Umax Collapse ) and depth ( d) fixed. In addition, fd , the height of the fracture, is fixed and defined such that Π2 (fd/d ) is 1.0. The figure shows that for a fixed value of
Stanford Rock Fracture Project Vol. 20, 2009 C-12 The subsidence occurs approximately 10m below the surface. This ratio of length of collapse / depth is close to the critical extraction value of 1.4 observed for collapses associated with long-wall mining operations (Whittaker and Reddish, 1989). The collapsing ash layer is modeled using tilted elements with appropriately defined stress boundary conditions. The depth where the collapse occurs is approximately 10m, maximum subsidence is approximately 1.5m, and a 0.5m surface opening of the vertical fracture was observed at the outcrop. When a collapse is induced in the numerical model, tensile stresses are relaxed around the vertical fissure by opening. Figure 17a shows the geometry of the model and Figure 17b is the σxx stress distribution map
resulting from the collapse when a downward stress of 0.25 MPa is defined along the horizontal elements. Figure 16 (below): Dimensionless length vs. Π4 (fl/d). Each line represents a different value of Π3 (a/d), while Π3 (fd/d) is kept constant at 1.0.
We investigate whether the relationship between the subsidence and fissure opening at the outcrop in Figures 10a and 10b can be predicted using this numerical model. In this simulation only the most prominent fissure at the outcrop, indicated on Figure 10b, is explicitly modeled using traction free elements. This fissure at the outcrop is slightly oblique and appears to be stratigraphically continuous down to the depth of the collapse. The bottom of this fissure is located approximately 2.5m left of the edge of the collapsed zone. All other fissures and tensile stress Figure 17: a) Geometric representation of the two prominent features found at the outcrop (cf. Figure fractures at the outcrop are incorporated into the model 10). b) Tensile stresses dominate around the by reducing the bulk stiffness of the rock to 10MPa, diagonally oriented joint, which is stratigraphically which is one to three orders of magnitude lower than continuous down to the depth of collapse. fd~10m, published elastic moduli of various shales and d~10m, a~12m, fl~2.5m. sandstones. The length of subsidence is approximately 12m at the outcrop, although the exact length is not known due to limited exposure at the outcrop.
Figure 18: A conceptual model depicting the mechanism of how pre -existing j oints above the North Coalbed Fire open up to form a fissure.
Stanford Rock Fracture Project Vol. 20, 2009 C-13 The dimensionless opening, Umax Opening / U max underground coalseam combusts and then compacts as Collapse is around 0.23 when these parameters are used its structural integrity is lost. Previous literature has to simulate the fissure opening and the collapse. suggested or described relationships between surface Alternatively, this value could have been obtained by deformation and subsurface subsidence, but no work calculating appropriate dimensionless variables, and has established first order functional relationships using Figure 16 to obtain the dimensionless length between variables that govern fissure widening and scale. Based on the model assumptions, appropriate subsurface subsidence in a coalbed fire. In this study, a dimensionless values are calculated as follows: Π2 = simple BEM model was formulated to simulate the fd/d ~ 1, Π3 = a/d ~ 1.2, Π4 = fl/d ~ 0.25. Although this collapse of the coalseam and the opening of pre- method approximates the coalseam and the traction free existing vertical fractures. Results show that the fracture to be horizontal and vertical, respectively, it aperture of the fissures at the surface depends strongly nevertheless gives a dimensionless length scale of on where the vertical fracture is located with respect to approximately 0.23. Both of these values are consistent the subsurface subsidence. The sensitivity analyses with the Umax Opening / U max Collapse observed at the performed using this simulator also demonstrate the outcrop. At the outcrop, Umax Opening = 0.5m and relationships amongst the governing variables defined Umax Collapse = 1.5m, giving a length scale ratio of 0.3. in this study. Those relationships can be used to The discrepancy is attributed to relatively simple estimate the location and subsidence magnitude based assumptions associated with this numerical model. In on the fissure locations and width measured at the future modeling efforts, these assumptions will be made surface. The model was tested using a dataset obtained more realistic. from a near by outcrop that showed evidences of The results from the numerical simulation suggests subsidence in a combusted coalseam and an opening of that pre-exisiting joints that are located above existing a vertical fracture above. Many assumptions were made coalbed fires can open when they are in regions of in the simple numerical simulation, and thus there is tensile stress induced by the subsidence. Figure 18 is a some discrepancy between the model results and conceptual model of how the propagation of the measured values. combustion front at the North Coalbed Fire can lead to opening of fissures at the surface. The figure accounts Acknowledgments for the local geology, geometry and the findings from The authors of this paper would like to the numerical investigations. In the figure, the lithology acknowledge: Bill Flint of the Southern Ute Indian above the coalbed fire is characterized as either shales Tribe for facilitating fieldwork details and his help in or sandstones. At the site, shales are often softer than securing funding, the Southern Ute Indian Tribe for the sandstones. In this conceptual model, the coalseam their gracious hospitality, allowing us to access their is transformed into a layer of ash as the thin combustion land and their continued support, Jonathan Begay, Kyle front propagates through the lower coal. The overlying Siesser and Ashley Neckowitz for their help in the field, strata collapse, and a pre-existing joint opens up to form and the Stanford Global Climate and Energy Project a surface fissure. The underlying Pictured Cliffs and its contributors for their funding to make this sandstone remains intact. Opened fissures above the fire research possible. may act as conduits that connect the surface and the coalseam. These fissures allow combustion gases to escape from the combustion zone, and enable fresh References oxygen to reach the coalseam in order to keep the Badgley, P., Analysis of structural patterns in bedrock, combustion alive. Society of Mining Engineers of AIME, Transactions of the American Institute of Mining, Conclusions Metallurgical, and Petroleum Engineers, 1962:223:381-389. At the surface above the North Coalbed Fire, which Badgley, P., Structural and Tectonic Principles, New burns along the Hogback Monocline in the San Juan York, NY: Harper and Row, 1965. Basin, numerous fissures form orthogonal patterns. Berest, P., Brouard, B., Feuga, B., and Karimi-Jafari, Some of these fissures vent hot exhaust gases from the M., The 1873 collapse of the Saint-Maximilien subsurface, an indication of a burning coalseam in the panel at the Varangeville salt mine, International subsurface. A combination of available geologic data Journal of Rock Mechanics & Mining Sciences, from previous surveys, observations and measurements 2008:45:1025-1043. from the field allows identification of a mechanism for Brown, K., Subterranean Coal Fires Spark Disaster, the formation of surface fissures. The hypothesis is that Science, 2003:299:1177. pre-existing joints in the strata overlying the North
Coalbed Fire widen to form fissures when the
Stanford Rock Fracture Project Vol. 20, 2009 C-14 Buhrow et al., Kohlebrande in der Volksrepublik China, Geological Survey, Professional Paper 1625B, Gluckauf, 2004:140:488-494. 2000. Cao, D., Fan, X., Guan, H., Shi, X. and Jia, Y., GAI Consultants, Inc., Engineering Analysis and Geological models of spontaneous combustion in Evaluation of the Centralia Mine Fire, 2 vols. the Wuda coalfield, Inner Mongolia, China, In: Springfield: U.S. Department of Commerce, Geology of coal fires: Case studies from around the National Technical Information Service, 1983. world, G. Stracher, ed. The Geological Society of Gielisch, H. and Kuenzer, C., Innovative technologies America, 2007, pp. 23-30. for exploration, extinction and monitoring of coal Condon, S. M., Joint patterns on the Northwest side of fires in North China: Work package 2220 - the San Juan Basin (Southern Ute Indian Structural pattern of the coal deposits (Wuda, Reservation), Southwest Colorado. In: Geology Rujigou, Gulaben); Analysis of linear pattern (air and coal-bed methane resources of the northern photos, satellite images), DMT and DLR, Coal Fire San Juan Basin, Colorado and New Mexico, J. Research - A Sino-German Initiative, 2003. Fassett, ed. Denver: Rocky Mountain Association Gorham, F.D., J., Woodward, L. A., Callender, J. F. and of Geologists, 1988, pp 61-68. Greer, A. R., Fractures in Cretaceous rocks from Condon, S. M., Geologic mapping and fracture studies selected areas of San Juan Basin, New Mexico - of the Upper Cretaceous Pictured Cliffs sandstone Exploration implications, American Association of and Fruitland Formation in selected parts of La Petroleum Geologists, Bulletin, 1979:63:598-607. Plata Country, Colorado, U.S. Geological Survey Huang, J., Bruining, J. and Wolf, K., Modeling of gas Open-File Report 97-59, 1997. flow and temperature fields in underground coal Chen L., Subsidence assessment in the Ruqigou fires, Fire Safety Journal, 2001:36:477-489. coalfied, Ningxia, China, using a Ide, T., Personal Communications with the Southern geomorphological approach, M.S. Thesis at the Ute Indian Tribe, 2007. International Institute for Aerial Survey and Earth Kaiser W.R. et al., Hydrology of the Fruitland Sciences. ITC, Enshede: Netherlands, 1997. Formation, San Juan Basin. In: Geologic and Crouch, S., and Starfield, A.M., Boundary element hydrologic controls on the occurrence and methods in solid mechanics, London, U.K.: Geroge producibility of coalbed methane, Fruitland Allen and Unwin Publishers, Ltd., 1983. Formation, San Juan Basin, W.B. Ayers et al., ed., Dai, S., Ren, D., Tang, Y., Shao, L., and Li, S., Des Plaines: Gas Research Institute, 1991, pp. 195- Distribution, isotopic variation, and origin of sulfur 241. in coals in Wuda Coalfield, Inner Mongolia, China, Kelley, V. C., and Clinton, N. J., Fracture systems and International Journal of Coal Geology, tectonic elements of the Colorado Plateau: 2002:57:237-250. University of New Mexico Publication in Geology DeKok, D., Unseen danger: a tragedy of people, 6, 1960. government and the Centralia Mine fire, Kelso, B. S., and Wicks, D. E., A geologic analysis of Philadelphia, PA: University of Pennsylvania the Fruitland Formation Coal and coal-bed Press, 1986. methane resources of the San Juan Basin, Dunrud and Osterwald, Effects of coal mine subsidence Southwestern Colorado and Northwestern New in the Seridan, Wyoming area. Washington: United Mexico, In: Geology and coal-bed methane States government printing office, 1980. resources of the northern San Juan Basin, Colorado Fassett, J. and Hinds, J., Geology and fuel resources of and New Mexico, J. Fassett, ed. Denver: Rocky the Fruitland Formation and Kirtland Shale of the Mountain Association of Geologists, 1988, pp 69- San Juan Basin, New Mexico and Colorado, U.S. 79. Geological Survey Professional Paper 676, 1971. Kuenzer, C., Voigt, S., and Morth, D., Investigating Fassett, J. E., Geometry and depositional environment land cover changes in two Chinese coal mining of Fruitland Formation coal beds, San Juan Basin, environment using partial unmixing, Gottinger New Mexico and Colorado: Anatomy of a giant Geographische Abhandlungen, 2005:113:31-37. coal-bed methane deposit. In: Geology and coal- Kuenzer et al., Detecting unknown coal fires: synergy bed methane resources of the northern San Juan of automated coal fire risk area delineation and Basin, Colorado and New Mexico, J. Fassett, ed. improved thermal anomaly extraction, International Denver: Rocky Mountain Association of Journal of Remote Sensing, 2007:28:4561-4583. Geologists, 1988, pp 23-38. Kuenzer et al., Uncontrolled coal fires and their Fassett, J., Geology and coal resources of the upper environmental impacts: Investigating two arid cretaceous Fruitland formation, San Juan Basin, mining regions in north-central China, Applied New Mexico and Colorado, Chapter Q, U.S. Geography, 2007:27:42-62.
Stanford Rock Fracture Project Vol. 20, 2009 C-15 Kuenzer et al., Automated demarcation, detection, and catastrophe, International Journal of Coal Geology, quantification of coal fires in China using remote 2004:59:7-17. sensing data, In: UNESCO Beijing, 2008, Taylor, D.J., and Huffman, A.C., Jr., Overthrusting in Spontaneous coal seam fires: mitigating a global the northwestern San Juan Basin, New Mexico—A disaster. ERSEC ecological book series, Vol. 4, new interpretation of the Hogback Monocline. In: 2008, pp. 132-148. USGS Research on Energy Resources, 1988: U.S. LaserCraft, Contour XLRic: Operator's Manual, Geological Survey Circular 1025, L.M.H Carter, Norcross, GA: LaserCraft Inc., 2007. ed., Fairbanks: Division of Geological & Litscheke, T., Innovative technologies for exploration, Geophysical Surveys, 1988, pp. 60–61. extinction, and monitoring of coal fires in North Taylor, D. and Huffman, A.C., J., Map showing China, M.S. Thesis at University of Duisburg- inferred and mapped basement faults, San Juan Essen. Duisburg and Essen: Germany, 2005. Basin and vicinity, U.S. Geological Survey, Lorenz, J. C. and Cooper, S. P., Tectonic setting and Geologic Investigation Series I-2641, 1998. characteristics of natural fractures in Mesaverde Wessling, S., The investigation of underground coal and Dakota reservoirs of the San Juan Basin, New fires: Towards a numerical approach for thermally, Mexico Geology, 2003:25:3-14. hydraulically, and chemically coupled processes, Malvern, L. E., Introduction to the mechanics of a Ph.D. Thesis at Westfälische Wilhelms-University. continuous medium, Upper Saddle River, NJ: Muenster: Germany, 2007. Prentice-Hall, 1960. Wessling et al., Numerical modeling for analyzing Martel, S. M., Boundary element scripts thermal surface anomalies induced by underground
Stanford Rock Fracture Project Vol. 20, 2009 C-16