Sedimentation, earthquakes, and tsunamis in a shallow, muddy epeiric sea: Grinnell Formation (Belt Supergroup, ca. 1.45 Ga), western North America

Understanding of the sedimentologic, oceanographic, and climatologic dynamics of large, shallow epeiric seas in low-relief continental interiors, be they sag basins, rift basins, or foreland basins, can be challenging in part because no direct modern analogs exist (e.g., Shaw, 1964; Irwin, 1965; Pratt and Holmden, 2008; Pratt and Haidl, 2008; Schieber, 2016). Consequently, coastal and shallow-marine processes understood from present-day continental shelves may not be wholly applicable, especially for seas that were not fully connected to the open ocean and if circulation was sluggish. In addition, different paleoclimatic regimes would have governed meteorological phenomena such as frequency and strength of storms. The Mesoproterozoic Belt Basin of western North America is no exception, and it has the added characteristic of predating the advent of animals by almost a billion years and therefore existing during a time of a simplified carbon cycle, under different atmospheric conditions, and with waters of a composition perhaps not exactly the same as that of present-day seawater.

The Grinnell Formation in the lower part of the Belt Supergroup, exposed in the Rocky Mountains of northwestern Montana and adjacent southwestern Alberta, is striking in its resplendent red to maroon color (Fig. 1). These rocks have been much admired by generations of geologists, especially for their ubiquitous mudcracks, which have been universally taken as the product of subaerial exposure and desiccation.

The purpose of this paper is to present a detailed description of the Grinnell Formation and propose a novel interpretation of the processes of sediment deposition and deformation. This environmental interpretation combines the effects of turbidity currents, hypopycnal plumes, storms, earthquakes, and tsunamis in a consistently submerged albeit shallow, lake-like sea. This study suggests that other stratigraphic units may exhibit comparable features, especially evidence for syndepositional tectonic activity in the form of seismites and tsunamites, that may have been heretofore underappreciated or gone unrecognized.

The Mesoproterozoic Belt Supergroup (Purcell Supergroup in Canada) is exposed in a region straddling southwestern Alberta, southeastern British Columbia, western Montana, and northern Idaho (Fig. 2A). The Belt Basin was at least some 200,000 km² in area, extending eastward in its early phase and forming the Helena embayment in central Montana. As subsidence continued, the basin likely expanded considerably farther than is suggested by present-day outcrop distribution (Pratt, 2001). It was semi-enclosed and has been interpreted as a lake (Winston, 1986b, 1990, 2007), but it probably opened to the ocean to the northwest (e.g., Ross et al., 1992; Winston and Link, 1993; Chandler, 2000; Pratt, 2001). The basin existed for a relatively short span of geological time, from ca. 1.48 Ga to ca. 1.25 Ga (Lydon, 2000, 2007; Chandler, 2000; Sears, 2007b; Pratt, 2017). It was characterized by rapid and seemingly continuous subsidence for some 25 m.y., which slowly tapered off, resulting in nearly 20 km of strata in the central area, thinning to some 5–8 km in the northeast (e.g., Winston and Link, 1993; Lydon, 2000, 2007). The calculated sedimentation rate suggests that the varve-like layering may well be annual (Pratt, 2001, 2017). Several major normal faults oriented roughly west-east resulted in development of a central horst, a northern half graben, and a southern graben (Sears, 2007b), as well as the Helena embayment (Winston, 1986b, 1986c). Large changes in thickness of stratigraphic units over short distances indicate that syndepositional normal faulting took place along these west-east faults and subsidiary northwest-southeast–trending lineaments in the central part of the basin, as well as generating frequent synsedimentary seismicity (Pratt, 1998b, 2001). The northern side of the basin also experienced syndepositional faulting with significant displacement (Turner et al., 2000; Gardner, 2008). However, the high sedimentation rate probably maintained an overall low-relief topography for the basin floor.

The initially rapid subsidence and the Moyie sills in the lower part of the Belt Supergroup on the northern side of the basin suggest that the Belt Basin was formed by intracratonic rifting (e.g., Höy, 1989). This may have been caused by oblique collision of the more or less coherent Mawson (East Antarctic) and Australian cratons with present-day western and northwestern Laurentia to form the putative Mesoproterozoic supercontinent Columbia (= Nuna) as it moved paleo-northward (Pisarevsky et al., 2014; Pehrsson et al., 2016).

Due to this collision, substantial uplift to the west is considered to have been the source for most of the fine-grained sediment in the lower and middle portions of the Belt Supergroup yielding detrital zircon ages of ca. 1610–1500 Ma (Ross and Villeneuve, 2003; Link et al., 2007). As the western source began to decline, zircon populations in the ca. 1790–1655 Ma range, which predominate in the Missoula Group, were derived from basement terranes to the south (Link et al., 2007, 2016). Archean zircons in coarser sandstones in the Altyn, Appekunny, Grinnell, Helena, and Wallace Formations on the eastern side of the basin were derived from the adjacent basement (Ross and Villeneuve, 2003). A minor population ca. 1450 Ma in age in the Piegan Group corresponds to syndepositional magmatism, which is also recorded by the presence of a tuff bed at Logan Pass and in the Purcell Lava (Link et al., 2007).

The Belt Supergroup is subdivided into four groups: the Lower Belt, the Ravalli Group, the Piegan Group (or “middle Belt carbonate” of some past usage), and the Missoula Group (Fig. 3; e.g., Winston and Link, 1993; Winston, 2007). The Grinnell and Empire Formations comprise the Ravalli Group in the northeastern side of the basin, although some schemes place the Appekunny Formation in the lower Ravalli Group (Harrison et al., 1993; Winston and Sears, 2013). This is equivalent to the Creston Formation to the west in southeastern British Columbia (Hein and McMechan, 1994), and the Spokane Formation in west-central Montana south of Glacier National Park (Winston and Link, 1993). The Grinnell Formation is correlated with the Burke, Revett, and St. Regis Formations in western Montana and northern Idaho (Winston, 1986a, 1989a, 1989b, 2016; Winston and Link, 1993). The succession is overwhelmingly dominated by fine-grained siliciclastic facies, but carbonates forming part of the Lower Belt were deposited on the northeastern side and in central Montana, and the younger Helena and Wallace Formations make up the basin-wide Piegan Group. Whereas the clay and silt were mainly sourced from the west, coarser sand came from the east (Winston and Link, 1993; Pratt, 2001; Ross and Villeneuve, 2003; Winston, 2007). The rocks are well preserved on the eastern side, having undergone no more than sub-biotite-grade greenschist-facies metamorphism via burial (Duke and Lewis, 2010). However, the striking difference in color between the Appekunny Formation (dominantly grayish green) and the Grinnell Formation (dominantly red or maroon) is due to the effects of diagenetic fluids, as indicated by both kinds of coloration in the same bed and alternating colors between adjacent beds. Patchy bleaching is also present locally in the Grinnell Formation.

The bathymetric evolution of the Belt Basin has been contentious. The Prichard Formation (= Aldridge Formation in southeastern British Columbia) in the Lower Belt consists of turbidites and “hemipelagites” (Cressman, 1989; Ransom and Lydon, 2000). The correlative carbonate platform deposits comprising the Waterton and Altyn Formations in the northeast were deposited mostly in shallow water (Hill and Mountjoy, 1984; Pratt, 1994). The overlying Appekunny Formation and Ravalli Group to the west are dominated by fine-sand–sized sediment and have been considered to record a fluvial environment (Winston and Link, 1993; Winston, 2016). In the east, the lower Appekunny Formation is predominantly argillaceous siltstone to very fine sandstone; plane lamination and rare hummocky cross-stratification (Pratt, 2017) suggest accumulation around storm wave base. The upper part of the Appekunny Formation and the overlying Grinnell Formation are mostly laminated mudstone (“argillite”) with ubiquitous “mudcracks.” Interbedded quartzose sandstone (“quartzite”) has been interpreted as representing subaerially exposed alluvial plains, playas, or tidal flats (e.g., McMechan, 1981; Winston, 1986a, 1986c; Winston and Link, 1993; Chandler, 2000). Herein, we build on an earlier study (Pratt, 1998a) to take a different view that regards the mudcracks as shrinkage-injection features and sedimentation in a permanently submerged, low-energy setting.

Following deposition of green mudstone (argillite) of the Empire Formation, the overlying Helena and Wallace Formations (= Siyeh Formation in southwestern Alberta and Kitchener Formation in southeastern British Columbia) blanketed the basin and do not show a distinct bathymetric polarity other than the restriction of oolite beds and stromatolites to the eastern half (Winston, 2007). Stromatolites indicate deposition within the photic zone. The dominant lithology is laminated, commonly graded argillaceous and silty dolomitic lime mudstone of low-energy aspect, reflecting deposition below storm wave base (Pratt, 2001) rather than in a shallow, intermittently subaerially exposed lake as envisaged by Winston (2007). The succeeding Snowslip Formation of the lower Missoula Group is somewhat similar, but the carbonate content is less (Whipple et al., 1984; Whipple and Johnson, 1988). The Purcell Lava, which was emplaced during deposition of the lower part of the Snowslip Formation in the northeastern region, includes up to 15 m of pillow basalts capped by pahoehoe, suggesting that the basin there was only ∼15 m deep at this time (McGimsey, 1985). This argues that the low-energy aspect of most of the units in the northeastern part of the basin was a function of a particularly shallow wave base as a consequence of the semi-enclosed nature of the basin and its limited fetch, along with a paleoclimate that seemingly was not strongly windy or stormy (Pratt, 2001). This would not be unexpected in an equatorial epeiric sea (e.g., Jin et al., 2013). The Bonner Quartzite in the upper Missoula Group is fluvial in origin and derived from south of the basin (Winston et al., 1986), indicating that the basin underwent an episode of subaerial exposure. Otherwise, the rate of erosion of the source area to the west and deposition of mud were essentially balanced by basin subsidence, and a broadly uniform bathymetry was maintained.

The Grinnell Formation was named by Willis (1902). Overall recessive, it is dominated by mudstone (argillite) that is lithologically similar to the mudstone forming the upper part of the Appekunny Formation, except for the presence of interbedded white sandstone (quartzite) and the change from dominantly greenish-gray to dominantly red or maroon color. It was later subdivided into three members, with sandstone more common in the lower and upper units, whereas in the middle member, thin beds of sandstone are a minor component (Fenton and Fenton, 1937). Use of these members has not persisted. Horodyski (1983) and Kuhn (1987) recognized five informal subdivisions. It has been stated that the formation consists of 60% sandstone increasing to nearly 100% in the upper part (Whipple et al., 1984; Whipple, 1992; Harrison et al., 1996), but this has not been encountered in measured sections (Kuhn, 1987; this study), even taking into account the recessive nature of the formation and consequent covered intervals. This study also did not encounter the 400-m-thick sandstone interval at the top of the formation depicted in the generalized stratigraphy at Marias Pass (= Goat Lick herein) by Winston (1989a, fig. 3; 1990, fig. 5; 2016, fig. 1; Winston and Link, 1993, fig. 13; Winston and Sears, 2013, fig. 3). However, the proportion of sandstone decreases rapidly—over a matter of a few kilometers—westward, as well as southward, where the formation passes into the Spokane Formation (Whipple et al., 1984; Earhart et al., 1984; Whipple, 1992; Harrison et al., 1996). For this reason, in western Glacier National Park and the Whitefish Range to the west, the Grinnell Formation can be subdivided into two parts, with sandstone interbedded in the upper part (Whipple et al., 1984). In the Whitefish Range, tabular beds of siltstone and very fine-grained sandstone in the lower part are similar to facies in the equivalent formations farther to the west (Whipple et al., 1984). A similar lithologic pattern is present in the Creston Formation of southeastern British Columbia (McMechan, 1981; Höy, 1992). Sandstone beds are less common and thinner north of Waterton Lakes National Park (Collins and Smith, 1977); the change takes place rapidly, as the difference is apparent just 8 km north-northeast of the Red Rock Canyon section herein.

The formation exhibits south- and southwestward-thickening strata, just 25 km northwest of Waterton Lakes National Park (Collins and Smith, 1977), 225 m to 450 m in Waterton Lakes National Park and just west of it, respectively (Hein and McMechan, 1994), at least 400 m at a location 3 km south of the border and 8 km northwest of Mount Henkel in northeastern Glacier National Park (Ross, 1959), ∼600 m on Mount Henkel (Horodyski, 1983), at least 900 m 5 km southeast of Sunrift Gorge (Ross, 1959), 800 m in southern Glacier National Park (Earhart et al., 1984), ∼800 m and 1070 m in central-western Glacier National Park and the Whitefish Range to the west, respectively (Ross, 1959; Whipple et al., 1984; Kuhn, 1987; Winston, 1989b, fig. 21; Winston, 1990, fig. 6), and ∼1500 m just south of Glacier National Park (Winston, 1989b, fig. 21; Winston, 1990, fig. 6).

Sections were measured at three closely spaced localities in Waterton Lakes National Park and two localities in east-central and southern Glacier National Park, respectively (Figs. 2B and 4–7). These are all on the same Lewis thrust sheet (Whipple, 1992; Stockmal and Fallas, 2015). This thrust represents 140 km of northeastward translation (Price and Sears, 2000; Sears, 2007a).

The Grinnell Formation consists of mudstone (argillite) and sandstone (quartzite). Mudstone includes laminated claystone, argillaceous siltstone (Fig. 8A), and siltstone (Fig. 8B). Silt is angular and generally well sorted. Detrital muscovite is variably present. In many horizons, the claystone has been eroded into fine sand– to pebble-sized, subangular to rounded, tabular intraclasts (“mud chips”) and admixed with silt (Fig. 8C). Sandstone is fine to coarse grained and well sorted, consisting of subrounded to rounded particles; subangular to angular grains are present but fairly rare (Fig. 8D). Small amounts of dolomite occur locally. Pyrite is virtually absent (Slotznick et al., 2016, their table 1).

The clay mineralogy of mudstones is illite and chlorite (Eslinger and Savin, 1973). The illite was generated by burial alteration of smectite (Eslinger and Sellars, 1981). Silt in the Grinnell Formation and correlative units to the west consists of quartz and lesser feldspar, in proportions typically around 3:1 but locally nearly 1:1. The feldspar is primarily albite plagioclase, although K-feldspar is locally present (Eslinger and Savin, 1973; Eslinger and Sellars, 1981). Sandstones consist of quartz with subordinate plagioclase (Ross, 1959).

The mud was sourced from an eroding orogenic belt broadly to the west (Ross and Villeneuve, 2003; Link et al., 2007), whereas the mineralogical differences and zircon ages indicate that the sand came from the low-relief craton to the east (Whipple et al., 1984; Winston, 1989b, 2016; Winston and Link, 1993; Pratt, 1998a, 2017; Ross and Villeneuve, 2003). The presence of fresh feldspars is evidence that the sand represents first-cycle detritus (cf., e.g., Tyrrell et al., 2009).

Although difficult to discern on weathered surfaces, mudstone is planar-laminated with common intraclastic layers. It consists of interlaminated claystone and siltstone, the thicker layers of which are typically normally graded (Figs. 9A and 9B). Siltstone laminae locally have scoured bases (Fig. 9B) and unidirectional cross-lamination, including in lenses representing small starved ripples (Fig. 9C). Tabular claystone intraclasts are oriented parallel to subparallel to bedding (Fig. 9D).

Siltstone beds locally exhibit straight-crested, small-scale symmetrical ripples with wavelengths of 1.5–3 cm, height up to ∼0.5 cm, and a ripple index (wavelength divided by height) between 3 and 6 (Figs. 10A–10C). Crests are often flattened (Figs. 10A and 10B). Stratigraphically closely spaced surfaces commonly show contrasting ripple orientations (Fig. 10A). The tops or bottoms of many rippled beds have variably dense populations of halite hopper molds or casts up to ∼10 mm across (Figs. 10B and 10C; see also Pratt, 2017, fig. 7F).

Three horizons of dolomitized domical stromatolites have been reported from the lower part of the formation (Rezak, 1957; Horodyski, 1983). These are laterally linked, grading to stacked hemispheroids up to ∼30 cm high and 50 cm wide at the base (Fig. 11A). Laminae exhibiting scattered flakes of dolomite intraclasts are present rarely. Upper surfaces of ripples in siltstone sporadically exhibit an “elephant skin” texture, consisting of a fine network of tiny ridges, less than 1 mm high (Fig. 11B). These exhibit a weak linearity. By contrast, wrinkle marks of Kinneyia type are relatively common on upper surfaces (Fig. 11C). Rare upper and lower surfaces exhibit scattered small pits and casts, respectively, several millimeters wide.

Synsedimentary deformation is pervasive in the mudstone facies. In many intervals, laminae are crinkled to gently undulating, and bed surfaces are creased or lumpy. Most bedding planes exhibit distinctive crack arrays that include polygonal, orthogonal, straight linear, curving linear, and spindle-shaped patterns (Figs. 12A–12I; also Pratt, 2017, fig. 7E). Polygons are not uniform on each surface and range in diameter from a few centimeters up to ∼20 cm. Where not joining other cracks, crack terminations taper to a point. Two sizes of crack systems are common in that small arrays may be present in the polygons formed by larger cracks. Where cracks and halite crystal casts occur, the cracks crosscut them.

In vertical section, cracks are vertically to obliquely oriented and straight to undulating to branching, and they are from less than one to several millimeters up to more than a centimeter wide, and a few millimeters to 10 cm long (Figs. 13A–13E and 14A–14E). Crack margins are variably cuspate, and they are parallel-sided to downward tapering. Claystone laminae commonly become gradually slightly thinner away from silty crack fills, losing up to about one third their thickness.

Cracks are filled with variable amounts of clay and silt, plus intraclasts exhibiting various orientations from subparallel to bedding to steeply inclined, as well as sand if the crack intersects a sandy lamina. This material is typically traceable to underlying laminae and therefore was ejected a short distance laterally and emplaced in a upward direction. There is no evidence for gravitational infilling from above, such as geopetal layering. Crack fills and host laminae are locally distorted by folding and shearing (Figs. 13A and 13B). In one example, a gutter crosscuts prior-formed cracks, as well as being penetrated by reactivated cracks from below (Fig. 13E).

Many crack arrays involve disruption and brecciation whereby claystone laminae are fractured and separated both across and parallel to the layering (Figs. 14A–14E). Laminae and breccia fragments are often gently folded, fractured, or displaced after formation. Crack margins are flat and parallel-sided to cuspate. Many or most intraclastic horizons are closely associated with brecciation, in that fragments have been reworked into intraclasts by currents (Figs. 14C and 14D). Oblique cracks locally resulted in the generation of larger blocky intraclasts (Fig. 14E).

Sandstones range from lenticular, to undulating, to tabular thick laminae and thin to medium beds to broadly tabular thick beds variably intercalated in mudstone successions (Figs. 15A, 15B, 16A–16F, and 17). Lateral change of thickness is due to both erosional bases as well as topography of bed forms and composite bar forms (Figs. 15B, 16A, and 16B). Sandstones have sharp, erosive bases, locally forming gutter-like casts several to tens of centimeters wide and up to 4 cm deep (Figs. 18A and 18B). Large-scale planar, concave-up, and sigmoidal cross-lamination is present in thin to medium beds (Figs. 15A, 15B, 16A–16D, and 19A–19C; see also Pratt, 2017, fig. 7B); trough cross-bedding appears to be absent. Hummocky cross-stratification also appears to be absent, but the cross-lamination in some beds locally approaches swaley shapes (Figs. 16E, 16F, 19A, and 19B). The tops of some thin beds exhibit two-dimensional symmetrical ripples with wavelengths of ∼5–15 cm, heights up to 3 cm, and a ripple index of up to 5 (Fig. 19C). More commonly, thicker intervals exhibit two-dimensional symmetrical ripples with ∼10–20 cm wavelengths and a ripple index of around 8–10 (Figs. 20A–20C). Crests are relatively straight or undulate at the meter scale, and, in places, there are sets of smaller two-dimensional symmetric ripples in the troughs oriented at right angles or oblique to the larger crests (“ladder-backed ripples”). Three-dimensional ripples occur rarely.

Amalgamated thin and medium beds commonly form thick beds and massive intervals (Figs. 19A–19C). Discontinuous mudstone laminae may be intercalated. Internal scour surfaces are locally present, and in many cases, it is not possible to separate individual beds. Unidirectional cross-lamination shows variable and even reversed directions in adjacent layers (Figs. 16A–16D, 19A, and 19B).

Granule- to medium pebble–sized mudstone intraclasts are almost everywhere incorporated in the sandstone (Fig. 21A). They are dominantly tabular, but lenticular and blocky shapes are common, and they are subangular to subrounded (Figs. 21B–21D). In some beds, there are large rounded clasts of agglomerated sand, mud, and small intraclasts, sometimes with an intraclast core (Fig. 21E). In some cases, sand may be concentrated around the periphery (Fig. 21F), forming an “armored mud ball.”

Despite the variable ripple orientation on some individual bedding surfaces, paleocurrent measurements from unidirectional cross-lamination and symmetrical ripples (Fig. 22) suggest a southerly to southeasterly flow direction for the lower part of the formation, with a northerly to northeasterly component for the former, and common west-east and northwest-southeast oscillatory flow directions for the latter. The middle part of the formation at Goat Lick, at the southern side of Glacier National Park, shows a westerly to northwesterly flow direction. The number of measurements at Cameron Creek and Goat Lick did not give a dominant flow direction. Paleocurrents measured in the upper Creston Formation in southeastern British Columbia are similarly variable, with a dominance of broadly westerly and easterly flow directions (McMechan, 1981).

Synsedimentary deformation is pervasive in the sandstones as well. In places, individual thin beds show gentle folding. Common features are short sand-filled cracks, occurring as dike arrays projecting from either the bed sole or bed top or both (Figs. 23A–23D; see also Pratt, 2017, fig. 7C). In plan view, these form polygonal, densely interconnected spindle shapes, isolated linear segments, or networks dominated by sinusoidal patterns. The last style occurs on some bed tops in symmetrical ripple troughs and conforms to the sinusoidal pseudofossil Manchuriophycus (Fig. 23B). Dike width varies from <1 mm to ∼10 mm, and in places, there are two crude size groupings (Fig. 23C).

In cross section, dikes extending from bed upper surfaces tend to be relatively straight (Figs. 24A and 24B). Those projecting from bed soles, however, are ptygmatically folded and commonly obliquely oriented, and they can reach up to 20 mm in length, or more if unfolded (Figs. 24C and 24D). Sand grains in the dikes are in grain-to-grain contact, but locally with admixed mud. In some cases, silt- and clay-filled cracks in mudstone are crosscut by a later generation of dikes (Fig. 24E). In other cases, muddy, intraclastic crack fills emanating from underlying mudstone penetrate the bottoms of sandstone beds (Fig. 24F).

Mudstone represents background sedimentation. The predominance of plane lamination and the absence of large-scale scours suggest deposition below fair-weather wave base. The great thickness of the formation with no perceptible differences or unconformities indicates an equilibrium in which a more or less constant bathymetry was maintained. The relative abundance of feldspar in the silt argues that the orogenic source region to the west was probably semiarid. Halite molds and casts are evidence that the climate of the Belt Basin was arid. The submerged depositional setting and absence of evidence for gypsum mean that halite had to have precipitated out of bottom-hugging brines that were concentrated elsewhere, presumably in playa-like lagoons and sabkhas in low-relief coastal areas broadly to the east. The casts suggest that the halite was dissolved at or just under the sediment-water interface when the pycnocline disappeared as salinity reverted to more normal marine conditions. This is supported by some surfaces with small dikes that crosscut the casts after formation. Density stratification in the basin may have developed more often than what is indicated by the numerous bedding planes with halite, but salinity might not have been high enough to lead to precipitation.

Graded bedding and local asymmetric ripples in the silty laminae suggest these are distal turbidites with thin upper-T_C, T_D, and T_E–1 divisions from low-density gravity flows (e.g., Talling et al., 2012) derived from sources broadly to the west. Presumably, the bulk of the sediment from these flows was deposited in more proximal areas. Planar claystone laminae were probably deposited by hemipelagic settling from hypopycnal plumes from the same source region.

Small symmetrical ripples in siltstones were formed by wave action that resuspended muds after deposition. Small wave orbitals formed by short-period, wind-driven waves generate steep-sided vortex ripples with short wavelengths (e.g., Clifton and Dingler, 1984; Wiberg and Harris, 1994). These are typical of very shallow bays and lakes with limited fetch and are common in many lacustrine successions (e.g., Allen, 1981; Aspler et al., 1994). The small wave ripples in the Grinnell Formation, however, developed in mud. They were probably produced in somewhat deeper water by longer-period waves when the flow velocity was large enough for grain movement. The wave action could have been from occasional storms.

Variably common intraclastic layers indicate frequent events of bottom scour and rounding by wave action, at and near the limit of storm wave base. The absence of intraclast imbrication is evidence that the resulting oscillatory currents were generally weak. The poor sorting and rarity of layers of sand-sized intraclasts also suggest that reworking was not protracted or repeated, as would be expected within fair-weather wave base. In many such horizons, erosion was preceded and thus assisted by prior crack formation and disruption. This is explored further below. Intraclastic layers have a stratigraphic frequency broadly at the decimetric scale, so their formation was a not seasonal or annual meteorological phenomenon.

Storm wave base was probably relatively shallow. In a fully developed sea, wave base (wavelength divided by two) for deep-water waves with a wavelength of 40 m and wave height of 1.8 m, produced by 40 km/h winds with a fetch of ∼150 km, is ∼20 m, i.e., twice that from 30 km/h winds (e.g., Trujillo and Thurman, 2017, chapter 8). The stromatolitic levels corroborate deposition within the photic zone; Kinneyia wrinkle structures and “elephant skin” textures—if these were microbially induced—support this. It is also possible that density stratification may have reduced storm wave base. The rare surfaces with pits may be the same structures stated to be raindrop and hail imprints by Kuhn (1987), but the small size, shape, absence of raised rims, and lack of overlap are not typical (e.g., Uchman et al., 2004; Rindsberg, 2005).

Balls of agglomerated mud, sand, and intraclasts suggest that the clayey sediment was initially sticky. However, intraclasts, cracks, disruption, and bending indicate that soon after deposition, the mud layers possessed a rheology that ranged from ductile to stiff. There is no evidence that it was loose and fluffy or formed a dense suspension (“fluid mud”). Length of ptygmatically folded dikes versus sediment thickness below sandstone beds suggests a water content of the mud up to ∼70% when sand was injected. In some instances, however, thinning away from crack fills suggests a water content of up to ∼30%, but planar laminae and intraclasts indicate it was typically much less. Burial compaction effects appear to be minimal, because folding of dikes was mostly due to the deformation event itself, and blocky intraclasts are not distorted. Flocculation and cohesiveness were enhanced by electrostatic attraction due to the marine salinity, although clays do not appear to have formed large floccules in the water column. This implies that suspended and settled organic matter played little to no role, and whatever might have been in the water column and attached to clays was rapidly oxidized. In turn, this suggests biological productivity was low overall. Frequent small earthquakes may have caused edge-to-edge orientations to collapse and expel their pore water, thereby mimicking burial effects.

The absence of gradational transitions between mudstone and sandstone intervals indicates that the sands punctuate background mud deposition and are not evidence for episodic deltaic input, river flooding, or regional bathymetric changes such as shallowing events. A source from the reworking of eolian dunes on the adjacent land surface is considered unlikely given the lack of evidence for strong winds and wave action. Rather than consisting of sediment delivered directly from fluvial sources, the dominantly rounded nature of the sand grains is evidence that they were subjected to protracted wave action, in sand bars and on beaches, before being transported into deeper water.

Cross-lamination from large current ripples indicates unidirectional flow of ∼0.3 m s^‑1. However, the common presence of scoured bases and muddy intraclasts, and the lateral thickness changes suggest that the consolidated muddy seafloor, as well as previously deposited sand, may have been eroded by bottom currents that were stronger initially, perhaps exceeding 0.5 m s^‒1. Sporadically occurring swaley bedding is evidence that combined flows occasionally developed. Symmetrical ripples, especially on bed tops, indicate that the final regime was oscillatory flow from wave action, occasionally followed by weaker oscillatory flow at right angles or oblique to the previous current direction, which produced smaller ripples in the troughs. These attributes along with the absence of hummocky cross-stratification and other strong storm-generated effects argue that the sandstone beds are not tempestites.

Given the low-energy aspect of the mudstone, strong bottom scour and transport of relatively well-sorted coastal and nearshore sands are interpreted to be due to a combination of tsunami-induced processes, including: (1) onrush, which generated unidirectional bottom currents, because in large tsunamis, the ratio of wave height to water depth was high due to the relative shallowness of the basin; (2) off-surge (backwash) from coastal and shallow offshore areas, which generated sand-laden underflows as jets; and (3) wave action, which generated oscillatory and combined flows. Off-surge transported the sand from the coastal areas. Onrush over the muddy sea bottom may have suspended some muds, but after delivery of the sand, onrush from a subsequent tsunami would have scoured and reworked it. Small tsunamis would have generated wave action. Bed thickness, as an indicator of the volume of sediment transported, is likely related to the amount of run-up and consequent backwash. The variation in current direction probably reflects shoreline configuration and sand distribution, interference, and location of the faults that were the source of the tsunamis. The symmetrical ripples were likely formed by post-tsunami oscillations (seiches) in the partially enclosed basin, although some of the smaller bed forms may be from storm reworking as well. Stacked and amalgamated sandstone beds probably reflect multiple tsunami events over a relatively short span of time. The rapid westward disappearance of sandstone (quartzite) is explained as resulting from the increased distance from the nearshore sand source. The northward and southward decrease may be for the same reason, or geomorphological differences along the eastern margin of the basin. Tsunamis would have affected other parts of the basin, but none has been specifically identified.

The absence of oriented fold axes and lack of other evidence for a consistent paleoslope at the local or regional scale indicate that local, gentle crinkling and folding of mudstone laminae and sandstone beds were caused by in situ deformation after deposition and not by hydrodynamic processes or sediment density contrast involved in sedimentation. Earthquake shock explains these features. The small pits may be due to dewatering or gas escape triggered by shaking.

Common seismic activity explains the disruption, cracking, brecciation, and sediment injection features in the mudstone, as well as the cracks and dike arrays associated with the sandstones, because it caused ubiquitously directed compressional, extensional, and shear stresses, liquefaction, dewatering, and sediment mobilization. The injection direction was upward in mudstones, whereas dikes emanating from sandstone beds show both downward and upward injection. Interconnected patterns of dikes and cracks on bedding planes suggest there was a relatively narrow range of lateral distance permitted before the mud failed and injection took place. The quasi-linear orientation of some examples may be perpendicular to the direction of the epicenter and might reflect inhomogeneities imparted by the P-wave. Lenticular bedding in sandstones also influenced the orientation of small dikes, as did ripples, which partially confined the sinusoidal dikes belonging to the Manchuriophycus form to troughs that contained thicker mud. The absence of unequivocal desiccation cracks filled by sediment derived from above rules out the possibility that some crack arrays could have overprinted preexisting cracks from subaerial exposure.

The fairly minor vertical distortion of cracks and common brecciation of the mudstone are further evidence for a relatively low water content of the clayey sediment, although this was enough to allow some dewatering and shrinkage when subjected to shaking. The length of folded sand-filled cracks projecting downward from sandstone shows a water content of up to ∼50%, but typically less.

The Grinnell Formation has been universally interpreted to record deposition on alluvial floodplains, playas, or tidal flats (e.g., Price, 1964; Collins and Smith, 1977; McMechan, 1981; Horodyski, 1983; Koopman and Binda, 1985; Höy, 1992; Winston, 1986a, 1986c, 1989b, 1990, 2016; Winston and Link, 1993; Chandler, 2000; Slotznick et al., 2016). These interpretations were based primarily on the ubiquitous “mudcracks,” which were assumed to be due to desiccation during subaerial exposure. In detail, however, these crack arrays lack the fundamental characteristics of subaerial exposure, such as a V-shape cross section with gravitational infilling from above, even though in many examples, a superficially similar polygonal pattern is present. Instead, deformation was clearly intrastratal, having formed under very shallow burial conditions (Pratt, 1998a). Other examples of crack arrays filled with upward-injected material, as well as small sandstone dikes exhibiting either or both upward and downward injection into enclosing muds, have also been interpreted as seismites, caused by earthquake shaking (e.g., Pratt, 1998a; Törő and Pratt, 2015a, 2015b, 2016; Leitner et al., 2017). There is no other obvious mechanism to simultaneously dewater muddy sediment and liquefy and mobilize silt and sand, as well as, in many cases, cause folding of the crack fills and brecciation of the matrix during the same or a subsequent event. Syneresis of clays (Burst, 1965) or bottom covering by microbial mats (Harazim et al., 2013) are passive phenomena that do not generate cyclic shear stresses. Wave impact (Winston and Smith, 2016) cannot account for substratal shrinkage and injection below the seafloor that shows no evidence of major scouring and reworking. The small-scale folding of some thin intervals was also due to synsedimentary earthquakes, as it indicates intermittent compressive stresses in a low-energy setting with no appreciable slope. Because deformation would affect both facies together, it is not due to gravitational effects from a density contrast due to passive loading of one sediment type on another (Owen, 2003). Other units in the Belt Supergroup exhibit a range of deformation features, some similar and some different—a function of sediment composition and rheology (Pratt, 1994, 1998a, 1998b, 2001, 2017). These kinds of features are present in many other sedimentary units, but their significance as a signal of syndepositional tectonic activity may have been overlooked (Pratt, 2018). Mudcracks are not always undisputed evidence for subaerial exposure.

The muds were not “soupy”; they were initially cohesive, and then they became stiff and subject to brittle failure at or near the sediment-water interface. This degree of consolidation before burial is a phenomenon that is unlikely to occur passively on the seafloor. Cyclic loading of saturated clays under undrained conditions, such as from earthquakes, can lead to failure because of the increase in pore pressure and loss of shear strength. If clays are allowed to drain, settlement occurs, the water content decreases, and the shear strength increases (e.g., Nieto Leal and Kaliakin, 2013; Andersen, 2015). Thus, episodic shaking due to low-magnitude earthquakes may have contributed to syndepositional stiffening of argillaceous sediment in the Grinnell Formation. This has been suggested in passing as a mechanism for causing intervals of anomalous overconsolidation in muddy successions (Pratt, 1998b), but it needs to be explored further.

Figure 25 summarizes the inferred depositional setting of the Belt Basin during Ravalli Group time. Silt and clay were deposited from low-density turbidity currents and hemipelagic fallout from hypopycnal plumes (e.g., Schieber, 2016). These originated from fluvial discharge far to the west, likely as flooding events rather than sustained hyperpycnal flows (e.g., Zavala et al., 2006, 2011; Wilson and Schieber, 2017). Occasional gentle reworking by storms (or weak tsunamis) took place, but evidence for deposition by storm-generated sediment gravity flows, such as ripples recording combined flow (e.g., Plint, 2014), does not appear to be present. Indeed, the Belt Basin seems not to have been a particularly stormy sea (Pratt, 2001), and formation of stromatolites within the photic zone is further evidence for a shallow storm wave base for the Grinnell Formation as a whole. This is not unexpected if the basin was located at or near the equator.

Halite crystal casts and molds represent displacive growth in and on the upper layer of mud, but the crystals dissolved when the overlying water freshened. Rather than representing in situ evaporation and corroborating a playa setting, the halite records occasional development of density stratification due to sluggish circulation and bottom-hugging hypersaline brines that were concentrated elsewhere, presumably in shallow nearshore areas, probably broadly to the east. The apparent absence of halite in the correlative Burke, Revett, and St. Regis Formations to the west could indicate shallower water, above the pycnocline, or there was westward dissipation of the brines closer to the sources of riverine input. Similar zoned halite crystal casts and molds in both siliciclastic and carbonate mudstones associated with crack arrays and other synsedimentary deformation features occur in many other units (e.g., Spencer and Demicco, 1993; Eriksson et al., 2005; Paik and Kim, 2006; Martill et al., 2007; Rychliński et al., 2014). The bathymetric interpretation of some of these merits reconsideration.

In the most often expressed environmental interpretation of the Grinnell Formation, sandstone beds have been regarded as fluvial sand sheets from flooding events. The absence of channel-form features and trough cross-bedding was explained by the interpretation that both they and the mudstone are sheetflood deposits with planar geometries (Winston, 1986a, 1986c, 1989b, 1990, 2016; Winston and Link, 1993). Setting aside the difficulty in reconciling this scenario with the muds of a floodplain being sourced from the west and intercalated fluvial sands coming from the northeast, flashy discharge leading to unconfined flow is usually associated geographically with nearby alluvial fans and slopes or channels as overbank or avulsion flooding phenomena (e.g., Fisher et al., 2007), but no evidence for channels is present anywhere in the Grinnell Formation or apparently in the correlative units to the west (Winston, 2016). Ephemeral sheetflood deposits are commonly transitional to upper-flow-regime beds characterized by low-angle to sinusoidal cross-bedding and abundant plane-lamination (e.g., Hampton and Horton, 2007; Lowe and Arnott, 2016), although there is much variability, and many examples reflect lower-energy flows (North and Davidson, 2012). The upper surface of sheetflood deposits may be reworked by eolian processes. In the Grinnell Formation, however, most sandstones exhibit lower-flow-regime, unidirectional cross-lamination with variable and commonly opposing current directions, two sizes of oscillation ripples, and in some instances combined-flow structures. These are incompatible with a fluvial origin, either on a subaerially exposed mud flat or shallow playa. Furthermore, the generally rounded nature of the sand grains is more simply explained as indicating a shallow-marine rather than fluvial source, which would typically supply texturally immature sediment, in contrast to the reworking that occurs under protracted wave action. In addition, first-cycle eolian sands tend to remain angular to subangular, except perhaps in situations where there is considerable back-and-forth motion over long time periods (Folk, 1978).

The sandstones are interpreted here as tsunamites (or tsunamiites), but the combination of processes envisaged for sediment transport, deposition, and reworking in a tectonically highly active, shallow, flat-bottomed, nearly enclosed sea like the Belt Basin has no known modern analogs, and tsunami effects are not well understood in present-day offshore settings in general (Dawson and Stewart, 2008). Reversing flows are created by onrush in shallow bays, which also experience oscillation long after the tsunami waves have impacted the coast (Lacy et al., 2012). Backwash, or off-surge, transports sediment offshore, but the sedimentary architecture in Holocene deposits is essentially unknown (e.g., Noda et al., 2007; Toyofuku et al., 2014). The first requirement is that they have to be emplaced beyond the depth of reworking by storm waves (Weiss and Bahlburg, 2006). The second, for Phanerozoic examples, is that they have to have escaped complete bioturbation.

Putative ancient tsunamites in offshore settings are variable, although few have been recognized. In the Miocene of eastern Italy, three examples from a deep-water mudstone succession containing thin turbidites share some similarities with those in the Grinnell Formation. They consist of sets of stacked, thin, cross-laminated, graded sandstone beds with erosive bases (Smit et al., 2012). Variably directed cross-lamination and climbing ripple cross-lamination are present in the lower set, whereas trough cross-lamination is present in the middle one, and the upper one has a channelized base and contains a layer with wave ripple cross-lamination. Sandstones in the Grinnell Formation differ in lacking trough cross-lamination. Features ascribed to tsunamis in limestones of the Helena Formation above the Grinnell Formation are understandably different, and they include: (1) intraclastic rudstone as flat-pebble conglomerate; (2) deep scour surfaces and gutters filled with fragments of molar-tooth structure; and (3) sandy ooid grainstone, locally with hummocky cross-stratification. These punctuate the argillaceous lime mudstone succession and are therefore anomalous in terms of background processes (Pratt, 1998b, 2001).

Upper Cambrian mixed carbonate-siliciclastic facies in north-central Montana similarly possess evidence of scour and wave action of a much greater magnitude than what is indicated by background processes, which generated lenses of flat-pebble conglomerate that punctuate the succession (Pratt, 2002). In the middle Cambrian of western Argentina, the host facies is bioturbated lime mudstone, and the intraclasts are fragments of cemented burrow margins, but otherwise wave-induced features appear to be absent (Pratt and Bordonaro, 2007). In both these cases, there is no allochthonous sediment, which may be due to basin shape and too great a distance from coastal areas. Examples of hummocky cross-stratification and imbricated intraclasts at the limit or below ambient storm wave base may be due to later stages of the tsunami wave train, post-tsunami seiching, or reworking by subsequent tsunamis; oscillation ripples may have the same origins, in addition to storm action.

It is premature to devise a facies model for tsunamites in deeper-water settings, especially given the wide range of basin types. In addition to the search for modern examples, reevaluation of beds hitherto interpreted as tempestites in some ancient successions should turn up candidates that may plausibly be considered tsunami-generated.

Misunderstood for more than a century, the Mesoproterozoic Grinnell Formation of western North America is revealed here to have been deposited not on floodplains, playas, or tidal flats. Rather, the mudstone (argillite) consists of sediment derived from the west, and it represents distal turbidites and hemipelagic fallout from hypopycnal plumes in a shallow basin subject to reworking by no more than gentle wave action during weak storms. Several stromatolite-bearing horizons indicate deposition within the photic zone. Bedding planes with halite crystal casts and molds are not evidence of in situ subaerial exposure; instead, they represent the influx of bottom-hugging brines that had evaporated in coastal regions probably to the east. Deposited layers of chlorite and illite were not flocculent, but instead clay laminae acquired a degree of stiffness that left them prone to brittle failure on the seafloor. Frequent strong earthquake shock caused a combination of shrinkage and rupturing of the muddy layers and concomitant upward injection of clay, silt, and clay intraclasts into crack arrays of various shapes on bedding planes. These cracks arrays are not desiccation features, despite their superficial similarity to subaerial mudcracks.

Tabular to lenticular sandstone (quartzite) beds punctuating the mudstone succession are cross-laminated from variably unidirectional and oscillatory currents but lack evidence of channeling. The striking contrast between the coarser, rounded sand and the angular silt in the mudstone points to a coastal source to the east, where the sand was winnowed and rounded by protracted wave action. This sand was delivered by the backwash from tsunamis that repeatedly swept across the shallow basin; mud intraclasts eroded from the sea bottom were incorporated. The sands were often reworked by later tsunamis, seiching, and backwash, as well as perhaps subsequent onrush. Sandstones were likewise affected by earthquake shock, and the tops and bottoms of beds commonly show sand-filled dikes recording both upward and downward injection into concomitantly formed shrinkage cracks in mud.

The Grinnell Formation joins other units in the Belt Supergroup, be they siliciclastic or carbonate, in exhibiting a striking array of sedimentary features. These not only have fundamental implications for bathymetry, deposition, salinity, paleoclimate, and paleoceanography, but they also demonstrate that the Belt Basin was tremendously active tectonically, which gave rise not only to the extraordinary rate of subsidence, but also to repeated earthquakes and tsunamis. When the deposits of other epeiric seas and shelf margins are parsed for comparable phenomena, hitherto overlooked or underappreciated counterparts will likely be found.

Fieldwork was funded by the Natural Sciences and Engineering Research Council (NSERC) of Canada through Research and Discovery grants awarded to Pratt. The Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET) of Argentina provided financial support for Ponce. We thank E.C. Turner and R.G. Brown for help in the field in the early 1990s and E. Wiens and S. Dasgupta in 2013, and the staff of both national parks for issuing research permits. We are grateful to D. Winston and D.G.F. Long for discussion and debate over the years, and L.B. Aspler and an anonymous reviewer for refereeing the manuscript.