Polygonal “cracks” are common in the Coconino Sandstone in Arizona. They have been called desiccation cracks, but several features indicate they are not desiccation cracks. They were never open cracks, but are merely linear depressions, linked to form polygons. They occur only on bounding surfaces, containing almost no clay, and the cracks extend 10 to 15 cm above and below the bounding surfaces. The polygonal patterns continue down from one sandstone lamina to another, for several centimeters. They are persistently continuous across all surfaces within their 20–30 cm vertical range, from the bottomset beds, onto the bounding surface, and continuing into individual cross-beds below the bounding surface. The cracks occur at the Grand Canyon, and are especially numerous and visible in flagstone quarries in the Seligman and Ash Fork area. They occur on some bounding surfaces but not on others, and in some quarries but not in others. The polygonal cracks have been mentioned in passing, but this is the first reported research on these cracks in the Coconino Sandstone. Polygonal cracks have been reported in the Navajo, Page, and Entrada Sandstones, but there are significant differences between these and the Coconino Sandstone cracks, which may indicate differences in their origin.

The Permian Coconino Sandstone, a prominent geologic formation in northern Arizona, contains abundant polygonal cracks on some bounding surfaces and associated cross-bed surfaces. These have previously been ascribed to desiccation, but further study reveals some features of these textures that are not likely due to desiccation. Understanding the origin of these cracks is of significance, because it has relevance for understanding the paleoenvironment of the Coconino Sandstone. If they resulted from a different process, they may not help in interpreting paleoenvironment, and their origin needs to be understood.

The polygonal cracks in the Coconino Sandstone have not been widely recognized in the geological literature, except for an abstract (Bartlett and Elliot, 2011) and a photograph labeled as “desiccation cracks” in a book on the Grand Canyon (Hill et al., 2016, p. 68), and they are figured in Millhouse (2009, p. 76). Those references refer in passing to the polygonal cracks, but this paper presents the first known research on these Coconino Sandstone polygonal cracks.

Interpretations of the Permian Coconino Sandstone began primarily with the work of McKee (1934, 1945) and Reiche (1938), followed later by subsequent work (Fisher, 1961; Sorauf, 1962; Lundy, 1973; Elcock, 1993; Sumner, 1999; Maithel et al., 2019, 2021). Recent studies focused on (1) stratigraphic relationships (Blakey and Middleton, 1983; Blakey, 1990); (2) paleoenvironmental implications of tetrapod trackways (Brand, 1979; Brand and Tang, 1991; Lockley et al., 1992 [see ‘Reply,’ p. 668–670]; Lockley and Hunt, 1995 [see p. 40–56]; Brand and Kramer, 1996; Millhouse, 2009; Citton et al., 2012); (3) sedimentary properties (Maithel et al., 2019, 2021); or (4) soft-sediment deformation (Brand and Maithel, 2021). Most authors consider the Permian Coconino Sandstone to be an eolian dune deposit (McKee, 1934; Reiche, 1938; Baars, 1961; McKee, 1974; McKee and Bigarella, 1979; Loope, 1984; Blakey, 1988; Middleton et al., 2003).

The Coconino Sandstone extends across northern Arizona, and is approximately 100 m thick at the Grand Canyon (McKee, 1934) and thickens to the south as far as the Mogollon Rim in central Arizona. Near the town of Ash Fork, the lower contact is not exposed, but the formation is at least 160 m thick. The formation almost always consists of vertically stacked, wedge planar or tabular planar crossbed sets (McKee, 1974; Blakey and Knepp, 1989; Middleton et al., 2003), separated mostly by second-order bounding surfaces (Kocurek, 1981). The polygonal cracks described in this paper occur on these relatively horizontal bounding surfaces.

The sets of cross-beds vary from < 1 m to more than 20 m thick (measured from base of set). Cross-bed dips range from 13° to 27°, and regional dip in the area of Seligman and Ash Fork is 1.5° to the east (Brand and Maithel, 2021). The grains in this strongly cemented, fine-grained sandstone are predominantly quartz (88–95%) with minor amounts of feldspar (Maithel et al., 2019, 2021). Grains average ~2.8–2.9φ in size, and are moderately sorted (excluding suspended fines, which were rinsed out in this study) (Maithel et al., 2021).

Soft-sediment deformation structures (SSDSs) are common in sedimentary rocks (e.g., van Loon, 2009; Owen et al., 2011; Törő and Pratt, 2015). These occur primarily in sand (Lowe, 1975; Obermeier, 1996), and in both eolian- and water-deposited sediments (Cojan and Thiry, 1992; Moretti, 2000; Fernandes et al., 2007; Liesa et al., 2016).

SSDSs have in some papers been interpreted as the result of seismic triggers (e.g., Fernandes et al., 2007; Moretti and Ronchi, 2011; Hilbert-Wolf et al., 2016; Liesa et al., 2016), but other triggers must also be considered (Owen and Moretti, 2011; Owen et al., 2011).

SSDSs also occur in the Coconino Sandstone, usually as small folds and ridges and fluid escape structures, and some fold structures up to 4 m high (McKee and Bigarella, 1979; Whitmore et al., 2011; Brand and Maithel, 2021).

Desiccation cracks in polygonal patterns are also a type of soft-sediment deformation (SSD) (Shanmugam, 2017). Polygonal cracks are common in the Permian Coconino Sandstone of northern Arizona, and their origin is the topic of this paper. Desiccation cracks (shrinkage cracks or mudcracks) occur in sediments with sufficient clay, and they can occur even in sandy desert soils as long as at least 5% of the total grain-size distribution is fine-grained sediment (< 6.3µm) (Altmann, 1993).

Polygonal cracks were initially examined at an abandoned flagstone quarry near Seligman, Arizona, and along the Hermit Trail in the Grand Canyon. Other sites that afforded adequate exposed surfaces for this study (Fig. 1) were searched for polygonal cracks. These sites included the North Kaibab Trail at the North Rim of the Grand Canyon, and an exposure 4.5 km north of the Tuweep Ranger Station, near the Grand Canyon North Rim. Also, five additional trails at the Grand Canyon South Rim were searched, including the Bright Angel, South Kaibab, New Hance, Tanner, and Grandview. Other sources of data were abandoned flagstone quarries near Seligman, Ash Fork, Holbrook, and Strawberry, Arizona, and along a road cut near Strawberry. The Hermit Trail exposes the entire thickness of the Coconino Sandstone, and a section was measured along this trail.

From the numerous examples of cracks examined, 32 samples were collected, and 61 thin sections and 14 polished slabs were prepared for this and related research.

The presence and identity of clay were determined by (1) examination of thin sections with a petrographic microscope; (2) examination of samples with a scanning electron microscope (TESCAN VEGA LSH) with Thermo Fisher Scientific energy dispersive spectrometry; and (3) X-ray powder diffraction (XRD) (Bruker D8). Further examination of the thin sections with the petrographic microscope, and examination of polished slabs uncovered other details of the features described below.

To determine the amount of clay associated with the polygonal cracks, clay abundance was compared in thin sections from our study sites. A scale of five levels of clay abundance (Fig. 2) facilitated comparing the amount of clay in photos taken from thin sections from within about a centimeter below bounding surfaces. These photos were from localities with and without polygonal cracks. Five Loma Linda University geology graduate students compared the photographs with this scale of clay abundance, and rated each photograph with the clay scale number that best matched it. Forty-two photographs were scaled by this process—24 from bounding surfaces with polygonal cracks, and 18 from surfaces without these cracks. This method of measuring clay abundance was used to provide a semi-quantitative comparison of clay in different locations. The results were evaluated with a Mann-Whitney U test (Lehman, 2006; Corder and Foreman, 2014) to determine if clay abundance is different at locations with or without polygonal cracks. The Mann-Whitney U test is used to test whether two sample means, of ordinal data, are equal.

One thin section allowed determining if there was a reduction of porosity associated with the cracks. Porosity (intergranular space) was measured at numerous locations in this thin section using Fiji distribution of ImageJ software (Collins, 2007; Schindelin et al., 2012; Schneider et al., 2012).

These polygonal cracks are visible at some exposures at the Grand Canyon, but are best studied in flagstone quarries, because of the way bounding surfaces commonly are well exposed in these quarries. The characteristics of the polygonal features are very consistent at all localities where they occur. The polygons commonly cover extensive portions of a bounding surface, and the polygon diameters are usually less than 15 cm (Fig. 3). They occur only associated with bounding surfaces, and they extend up to 10–15 cm above and below the bounding surface (Fig. 4). They are not on all bounding surfaces (Fig. 5), and not at all localities (Fig. 1). They commonly occur on part of a bounding surface, but fade out, laterally, and are not present on other portions of the same bounding surface. They were found at the North and South rims of the Grand Canyon and at many quarries near Seligman and Ash Fork, in the central portion of the Coconino Plateau. Quarries were most abundant and accessible near Ash Fork, and polygonal cracks occurred in 14 of 27 Ash Fork quarries, scattered from bottom to top of the Coconino Sandstone. There was no stratigraphic pattern to the presence or absence of cracks in these quarries (Fig 5B).

No polygonal cracks were found at the localities studied near Strawberry and Holbrook. Coconino Sandstone exposures at the Grand Canyon provided minimal exposures of bounding surfaces, but some cracks were found along the Hermit Trail and North and South Kaibab trails, and at the locality near the Tuweep Ranger Station. Exposures of the Coconino Sandstone examined at other localities (not marked on Fig. 1) were too weathered or too limited to allow determination of whether the polygonal cracks were present.

The polygonal features appear in surface view as a pattern of cracks, similar to desiccation cracks. But when seen in cross section, true desiccation cracks are v-shaped open cracks extending some distance down into the sediment (Fig. 6A). In contrast, the cracks in the Coconino Sandstone are never open cracks that extend down through the sandstone layers. Instead, they are linear depressions linked in polygonal patterns, and each sandstone lamina is merely bent downward at these depressed lines. On any given cross-bed, the depressions continue vertically through the sandstone laminations for several centimeters (Figs. 6B–D; Fig. 7).

Cracks that are exposed on a cross-bed surface, where it meets a bounding surface, extend from the bounding surface onto the cross-bed surface and continue down along this surface until they fade out within 10 to 15 cm as measured at right angles to the bounding surface (Fig. 8). The lower side of each polygon disappears, and the linear depressions fade out.

These polygonal patterns also persist upward above bounding surfaces, commonly through 10 to 15 cm of the bottomset beds, and the polygonal pattern is continuous with the pattern on the bounding surface (Fig. 9). The continuity of these three-dimensional patterns persists in all directions, within that 20–30 cm vertical range, from surface to adjacent surface, upward and downward from the bounding surface, through successive bottomset and cross-bed surfaces (Fig. 10).

The polygons vary in size and prominence, within a range of diameters of 3.5 to 23 cm (mean = 9.2 cm; N = 60). The width of individual cracks also varies, from about 2 mm to 4.2 cm. Depths of the cracks, measured as depth of depression of a given lamination, range from about 0.5 mm to 3.5 mm. In some places, other deformed strata occur on cross-beds closely associated with the polygonal cracks at or near bounding surfaces (Fig. 11).

A very small amount of clay (estimated < 1%) is commonly scattered through our Coconino Sandstone samples. This clay is equally rare in samples collected from bounding surfaces. Of the 42 thin section photographs that were compared with the five-step scale of clay abundance, those adjacent to bounding surfaces with polygonal cracks averaged 1.40 (1 = lowest abundance, 5 = highest abundance), and those adjacent to bounding surfaces with no cracks averaged 1.42. There is no significant difference in clay content adjacent to bounding surfaces with or without polygonal cracks (U = 21; p = .697, two-tailed test).

More significant concentrations of clay occurred in layers along many foreset surfaces. These distinct layers of clay divided successive cross-bed lamina (Brand and Maithel, 2021, their fig. 2; Fig. 2B–C this paper), but in our samples there were no clay layers along or close to bounding surfaces. Polygonal cracks were only observed on bounding surfaces, and there were none along the clay-rich layers.

One thin section cut perpendicular to these cracks (Fig. 12) offered a unique opportunity to measure changes in porosity. A reduction in porosity occurs in the lower laminae affected by the “crack.” In the crack itself (C1), porosity at positions 6–10 was reduced by 40% compared to positions 1–5 (Table 1). Horizontally there is more porosity reduction near the crack (CN1 and CN2) compared to the sandstone farther from the crack (CN3, CN4) (Table 1).

The polygonal cracks are usually only seen in surface view on a bounding surface, except where samples can be found that reveal the internal structure. Surface views of these cracks bear a superficial resemblance to desiccation cracks, but they have internal features that are not compatible with this interpretation. Desiccation cracks are only known to occur in desert sand if the substrate contains at least 5% clay (Altmann, 1993), but our Coconino Sandstone samples taken from bounding surfaces are almost pure sand, with very little clay. A mechanism to explain these cracks must also account for the following features:

  • (1) They are not v-shaped cracks down into the sediment, and were never open cracks, as would be expected if they were desiccation cracks.

  • (2) Their continuity, in detail, from the bounding surface up into the overlying bottomset beds indicates that the cracks did not result from desiccation of an exposed bounding surface. The continuous polygonal pattern on a given bounding surface, and through the 10–15 cm space both above and below it, indicates that they formed as one intrastratal event, soon after the formation of the bounding surface and the overlying bottomset beds above it. There is no evidence of brittle fracturing within the zone of polygonal patterns. The entire set of polygonal cracks, above and below the bounding surface, appears to have occurred in unlithified or semilithified sand, and thus was close to synsedimentary. There was evidently sufficient cohesion within the sand body, within centimeters above and below the bounding surface, to preserve the fine details in the set of cracks.

  • (3) There are no cracks along the clay-rich layers on foreset surfaces. The cracks have an intimate genetic connection with bounding surfaces, as indicated by their consistent limitation to bounding surfaces and beds within 10 to 15 cm above and below the bounding surface. Because the forces that produced these patterns were only effective in close association with bounding surfaces, this implies that movement and compaction were more likely at the sedimentary discontinuity of a bounding surface than in the cross-bed sequences between bounding surfaces.

  • (4) The sample shown in Fig. 12 and the porosity measurements in Table 1 reveal a reduction in porosity in and near the lower part of the crack, allowing sediment above it to be depressed into the space produced by this compaction, and the bottomset beds are also depressed into this polygonal pattern. Differences between the means at C1 and CN2 are marginally significant; the differences between the means at CN3 and CN4 are not significant.

  • (5) The distribution of these polygons at various localities and on individual bounding surfaces is very patchy. Patches with cracks on a bounding surface ranged from a few square meters to 100 square meters or more. At a given locality, they occur on some bounding surfaces and not on others. They may occur on part (or parts) of a bounding surface, but not on adjacent portions (tens of square meters or more) of the same bounding surface. Some bounding surfaces are largely covered with the polygons. Location of bounding surfaces with polygonal cracks is not related to stratigraphic position in the Coconino Sandstone. At any given stratigraphic level, even within a few hundred meters of horizontal distance, some bounding surfaces have cracks and others do not. Thus, the triggers for these cracks were very uneven in their influence.

  • (6) On a larger scale, the known cracks are also geographically limited. They are most prominent and readily studied in the area of Seligman and Ash Fork. They are present at the Grand Canyon, but exposures of bounding surfaces there were not adequate to determine their abundance. The polygonal cracks are not found at the southern or eastern parts of the Coconino Plateau, near Strawberry or Holbrook, although Coconino exposures at those localities seem adequate to indicate the presence or absence of these structures.

Another example of similar polygonal features that are not from desiccation is polygonal thermal cracks on modern sandstone surfaces (Williams and Robinson, 1989; Chan et al., 2007; Loope, 2019). These are cracks on modern exposed rock surfaces. They are not on bounding surfaces, and they do not have the details seen in the Coconino Sandstone polygonal cracks on bounding surfaces.

Polygonal cracks have been reported in the Mesozoic Navajo Sandstone of the Glen Canyon Group, and the Mesozoic Page Sandstone and Entrada Sandstone of the San Rafael Group (Kocurek and Hunter, 1986; Crabaugh and Kocurek, 1993). There are significant differences between the cracks in these three formations and those in the Coconino Sandstone. The Mesozoic cracks are measured in meters, and they are simple cracks or v-shaped, sediment-filled cracks, that begin at a bounding surface and only extend down from there. The Coconino Sandstone cracks are measured in centimeters, they were never open v-shaped cracks, and each occurrence is a set of polygonal features extending above and below its associated bounding surface, indicating that it was an intrastratal structure formed after all involved laminae were deposited. The cracks in the Navajo and Entrada Sandstones are in a salt-rich sabkha environment. A sabkha environment should be considered for the Coconino Sandstone cracks, but the Coconino does not seem to have evidence for a sabkha environment. It does not have the salts expected in a sabkha, and there are other differences, described in Table 2. These features imply that the process that formed the Coconino Sandstone polygonal cracks was not similar to the process that formed the Mesozoic polygonal cracks.

Deformation Processes and Triggers

The polygonal cracks are perpendicular to the fracture surface, and appear to be opening-mode fractures or joints. Joints are thought to be initiated at sediment flaws (Pollard and Aydin, 1988). The sedimentary discontinuity at a bounding surface can likely act as a widespread flaw, and perhaps the discontinuity contributes to the initiation of polygonal cracking. The cracking then apparently spread upward and downward, and finally faded out in both directions. The uniformity in vertical distance upward and downward from the bounding surface perhaps indicates uniformity of the limits in the strength of the stresses that initiated the cracks.

The process that forms bounding surfaces appears to determine, or be closely linked to, the persistent presence of polygonal cracks only at these bounding surfaces. Bounding surfaces in eolian sandstones are thought to result from a new set of dunes overriding and eroding the upper portion of a previous set of dunes (Rubin and Hunter, 1982). It is not clear how or why this process and this environment would favor the formation of polygonal cracks on the resulting bounding surface. The sand above and below the bounding surface would have to be cohesive enough to form and preserve the precise features seen in these cracks. There is evidence of moisture in the Coconino Sandstone when the polygonal cracks and other SSD formed (Brand and Maithel, 2021), rather than there being dry desert dunes. This may have been a significant factor favoring the cohesiveness of the sand and development of the cracks.

The process producing polygonal cracks was not uniform, but was optimal in specific patches, as described above, likely because of variations in the rheology of the sediment (Mills, 1983; Owen, 1987; Martín-Chivelet et al., 2011; Tórő and Pratt, 2015). Variation of water content or degree of sediment consolidation could determine whether stresses in the sand would deform it into these polygonal lines of depression or leave it undeformed. The discontinuity at bounding surfaces apparently did not limit the vertical movement of water, because the cracks were quite uniform through the 20–30 cm zone above and below the bounding surface.

At 14 Ash Fork quarries where presence of polygonal cracks and other SSDSs could be compared, more than 75% of quarries containing polygonal cracks also contained other SSDSs, such as folds, ridges, and fluid escape structures (Brand and Maithel, 2021). The cracks and the other types of SSD also shared the same geographic distribution, occurring from the Grand Canyon to Ash Fork, but not farther south or east. This indicates a positive causal relationship between the different types of SSD. However, there is one prominent difference between polygonal cracks and the other Coconino Sandstone SSDs. The folds, ridges, and fluid escape structures occur only on foreset surfaces, with dips of 17° to 25° (Brand and Maithel, 2021). The origin and characteristics of these SSDSs were evidently controlled by those sloping surfaces. However, polygonal cracks are all directly associated with relatively horizontal bounding surfaces, and the causal mechanism thus seems to be significantly different, in some way, from the other SSDSs, that are only on sloping foreset surfaces. The SSD on sloping surfaces was interpreted as the result of liquidization and upward movement of trapped pore water, producing excess pore water pressure (Brand and Maithel, 2021; also see Owen, 1995; Obermeier, 1996). The upward movement of this pore water apparently was at least partly controlled by the sloping foreset surfaces with their clay-rich layers. Water-escape structures—apparently associated with the excess pore water pressure—are present in the Coconino Sandstone (Brand and Maithel, 2021), but they have so far been found only on the sloping foreset surfaces, not on bounding surfaces. More research is needed to understand the distribution and role of water in the Coconino Sandstone, and its relationship to the polygonal cracks and other structures.

The processes forming the cracks, and the other types of SSD, were different, but the common distribution, by quarry and by geography, of polygonal cracks and the other SSD could have resulted from similarities in the nature of the trigger.

Brand and Maithel (2021) suggested that seismic events are a likely trigger for the multiple SSDSs in the Coconino Sandstone. If this is correct, the triggers for the polygonal cracks may be the same or related to the triggers for the other SSDSs (including the small SSDSs shown in Fig. 11 of this paper) that are associated with the polygonal cracks. There are small faults in the Coconino Sandstone, but in addition to those minor faults, the polygonal cracks and other SSDSs are the primary evidence that could be interpreted as resulting from seismic events in the Coconino Sandstone (Brand and Maithel, 2021).

The seismic events or other triggers were localized in time and place, as indicated by the unpredictable locations of the polygonal cracks, geographically and stratigraphically. For this reason, it appears that the cracks did not result from large seismic events with wide influence. If seismic events were the trigger, it is possible that far-field effects from distant earthquakes had locally varying influence, over short distances, in central Arizona, depending on the local rheological or other properties of the sediment (Mills, 1983; Owen, 1987; Nichols, 1995; Moretti and Ronchi, 2011).

Another triggering mechanism to consider is sediment loading—stress resulting from the weight of overlying sediment (Moretti et al., 1999). If sediment loading caused the cracks, it would seem that thicker overlying sets of cross-bedding would increase the probability or size of polygonal cracks. However, the cracks, with their structural uniformity, found in the Hermit Basin section (Fig. 4) are at the base of cross-bed sets that are quite thin or of only average thickness, and not at the base of the thickest sets. This does not seem to indicate sediment loading as a primary cause of the cracks. The trigger for these deformations may have been seismic, but this is difficult to verify.

This research indicates that care must be taken in deciding whether polygonal “cracks” are desiccation cracks, or some other phenomenon. It may be necessary to examine the internal structure of the rock to determine whether they are indeed from desiccation. These structures should not be used as an indicator of paleoenvironment without that careful analysis.

Other cross-bedded sandstone should be examined closely for these polygonal cracks, although it will be difficult to find them in sandstone formations that do not split into flagstones. The reason for this is that the clay layers on Coconino Sandstone foreset surfaces cause the sandstone to split cleanly along these foreset surfaces, resulting in a commercial source of flagstones, and also resulting in numerous exposed bounding surfaces. In a formation that does not have those clay layers and exposed bounding surfaces, it may be difficult or impossible to find the polygonal cracks, even if they exist.

The polygonal features on Coconino Sandstone bounding surfaces are not desiccation cracks, although their surface appearance closely resembles desiccation cracks. A more appropriate term for them would be ‘polygonal linear depressions.’ It is necessary to look below the surface of the exposed rock to determine its true nature. The most significant features are the shallow depression of each lamination below the polygonal grooves, and the absence of v-shaped, open cracks. The cracks (or depressions) always occur at bounding surfaces, in a suite of continuous grooves on all surfaces within about 15 cm above and below the bounding surface. Porosity of the sand in and near the depressions indicates tighter packing of the grains compared to more distant locations. I suggest that the patterns of depressions may have resulted from seismic events, but there is not adequate information to eliminate other interpretations. Whatever the trigger for these features, including their lowered porosity, its effect was closely associated with the sedimentary discontinuities of bounding surfaces. These polygonal cracks seem to be a previously undescribed geological SSD feature.

Research permits for this work included permit #GRCA-2016-SCI-0033 from the Grand Canyon National Park, and a research permit from the U.S. Department of Agriculture, Forest Service, for study in flagstone quarries in the Williams Ranger District of the Kaibab National Forest, Arizona. I thank Big Boquillas Ranch for access to the Seligman outcrops, and Michael Reidhead for access to outcrops on his land near Holbrook. Thanks to Kevin Nick for assistance in evaluation of clay, and to William Hayes for help in statistical analysis. Thanks to Rick Peters for initially noticing the polygonal cracks. I am also appreciative of the improvements to this paper by Rocky Mountain Geology(RMG) Science Co-Editors Art Snoke and Ron Frost and RMG Managing Editor Brendon Orr.

Gold Open Access: This paper is published under the terms of the CC-BY license.