- © 2006 by the Seismological Society of America
Earthquake-induced liquefaction features, including large sand blows, occur near Marianna, Arkansas, about 75 km southwest of Memphis, Tennessee, and 80 km south of the southwestern end of the New Madrid seismic zone. The Marianna sand blows formed between 5,000 and 7,000 years ago and predate paleoearthquakes attributed to the New Madrid seismic zone. The Marianna sand blows are similar in size to New Madrid sand blows, suggesting that they too formed as a result of very large earthquakes but were centered near Marianna and outside the New Madrid seismic zone. A large sand blow that formed about 3500 B.C. (5,500 years B.P.) may correlate with smaller sand blows to the northeast and southwest up to 175 km away. A compound sand blow that formed about 4800 B.C. (6,800 years B.P.) may have formed as the result of several very large, closely timed earthquakes. A fault zone associated with the eastern Reelfoot Rift margin seems the most likely source of large Middle Holocene earthquakes because of its great length (∼300 km), history of seismic activity, Late Wisconsin–Early Holocene fault movement in western Tennessee, and structural relationship to the New Madrid fault system. Additional study is needed to verify our initial findings, to identify the earthquake source, and to further define the earthquake potential of the Marianna area. If the eastern Reelfoot Rift margin were confirmed to be the source of Middle Holocene earthquakes near Marianna, seismicity would appear to vary in space and time within the Reelfoot Rift system. This would have important hazard implications for currently aseismic faults of the rift system and possibly of other aulacogens embedded in intraplate regions.
The Marianna area is located in east-central Arkansas about 75 km southwest of Memphis, Tennessee, and 80 km south of the southwestern end of the New Madrid seismic zone (NMSZ; figure 1). Few modern earthquakes have been recorded in the Marianna area. However, its proximity to the NMSZ and the Reelfoot Rift fault system raises questions about the earthquake potential of the area. Several faults have been identified in the Marianna area, including the eastern margin of the Reelfoot Rift, the Big Creek fault zone, and the White River fault zone (figure 2; Fisk 1944; Krinitzsky 1950; Spitz and Schumm 1997). The eastern Reelfoot Rift margin (ERRM) affects the regional drainage pattern in both the St. Francis Basin and the Western Lowlands (Spitz and Schumm 1997). The rift margin crosses the study area northwest of Marianna and may terminate near the White River against a northwest-trending structure known as the Arkansas transform fault and related to the Ouachita front (Hildenbrand 1984; Hildenbrand and Hendricks 1995). The Big Creek fault zone, south of Marianna, parallels the Reelfoot Rift margin and marks the boundary between southward-trending Wisconsin valley-train deposits and southwest-trending Holocene meander-belt deposits (Fisk 1944; Krinitzsky 1950). The White River fault zone (WRFZ) is northwest-oriented, controls and deflects local drainages, and breaches and offsets the southern end of Crowley's Ridge at the Marianna Gap (Spitz and Schumm 1997). Although it appears to be structurally controlled, the Marianna Gap also has been shaped by the Mississippi, St. Francis, and L'Anguille rivers (figure 2).
During a previous study, Al-Shukri et al. (2005) identified numerous circular to elliptical light-colored areas suggestive of sand-blow deposits on aerial photographs and satellite images of the Marianna area. They excavated several of the light-colored areas and confirmed that they are sand-blow deposits. In addition, they interpreted the deposits as being prehistoric in age on the basis of their weathering characteristics and a few radiocarbon dates of charcoal collected from sediment underlying the sand blows. Although questions remained regarding the relationship of the sand blows to paleoearthquakes generated by the NMSZ, the study demonstrated that a record of strong ground shaking is present in the Marianna area. Here, we present new findings that suggest that some of the Marianna sand blows are not the result of New Madrid earthquakes but instead formed during very large earthquakes centered near Marianna about 5,000–7,000 years ago.
We employ well-established criteria for identifying earthquake-induced liquefaction features as well as commonly used geological and geophysical methods for studying these features (e.g., Tuttle 1999, 2001; Wolf et al. 1998; Lui and Li 2001; Al-Shukri et al. 2005). The methods include (1) examination and interpretation of aerial photographs, (2) reconnaissance of plowed fields and river and ditch cutbanks, (3) investigation of possible sand-blow deposits by various techniques including probing, surveying with ground-penetrating radar (GPR), and digging soil pits and trenches, and (4) documentation of liquefaction features by logging and photographing trenches and river exposures. We use radiocarbon dates of organic samples found in association with sand-blow deposits to estimate the timing of causative earthquakes. In addition, we compare optically stimulated luminescence (OSL) ages of sediment samples with radiocarbon dates of co-located organic samples. Radiocarbon dating was performed by Beta Analytic, Inc. and OSL dating was carried out by Shannon Mahan of the U.S. Geological Survey (see tables 1 and 2).
RECONNAISSANCE AND SITE SELECTION
On aerial photographs, we identified numerous large, elliptical, and light-colored areas suggestive of sand-blow deposits in the Western Lowlands west of Marianna where Early to Middle Wisconsin valley-train deposits have been mapped (figure 2; Saucier 1994; Blum et al. 2000). In a field that had been recently graded, several large (tens of meters wide and hundreds of meters long) light-colored areas are especially obvious and occur in close proximity to northwest-oriented lineaments (figure 3). We examined the light-colored areas by probing and digging soil pits and determined that they are sandy deposits overlying silty soils. We selected this site, named Daytona Beach, for subsurface investigations, including GPR surveys followed by trenching and logging of trench walls.
In addition, we conducted reconnaissance for liquefaction features along 4 km of the St. Francis River and 28 km of the St. Francis Ditch in the St. Francis Basin east of Marianna, where Holocene point bar deposits and abandoned channels of the Mississippi River have been mapped (Saucier 1994). Along the St. Francis River, recent deposits of unweathered, interbedded silt and sand containing pieces of metal are exposed in the occasional cutbank. Along the St. Francis Ditch, cutbank exposure varies from poor along straight stretches to good in bends. A lower terrace is underlain by unweathered, interbedded silt and sand similar to deposits along the St. Francis River, whereas a higher terrace is underlain by weathered, cross-bedded sand and laminated silt containing calcium carbonate nodules. Weathering characteristics indicate that deposits underlying the higher terrace are significantly older than deposits beneath the lower terrace. Radiocarbon dating indicates that sediment exposed below the higher terrace (SF500-C3) is up to ∼6,900 years old, whereas sediment exposed beneath the lower terrace (SF1-C1) is up to ∼920 years old (see table 1). Despite poor exposure, we found a remarkable complex of liquefaction features at a site northeast of Marianna along the St. Francis Ditch. The site is named St. Francis 500 and is described in more detail below.
INVESTIGATION OF THE DAYTONA BEACH SITE
The Daytona Beach site is located about 8 km southwest of Marianna and the western boundary of the WRFZ (figure 2). Most of the site was recently graded to improve drainage for farming purposes. A pond on the site was filled during the grading process. In addition, much of the topsoil was removed, exposing sandy sediment that abruptly ends along a northwest-oriented lineament (> 1.5 km long) northeast of the pond (figure 3). The western part of the site extends beyond a gravel road along which two houses stand. Soils in the vicinity of the houses also are disturbed.
As mentioned above, several large light-colored areas identified on aerial photographs of the site have the appearance of sand-blow deposits (figure 3). During preliminary investigation of the site, we dug soil pits in these areas and found iron-stained, medium to very fine sand overlying mottled silty soil. Using a soil probe along an east-west transect, we determined that the sand deposit gradually thickens toward the east, suddenly becomes thicker than the length of the probe (1.25 m), and then abruptly thins in the vicinity of the northwest-oriented lineament. The morphology of the sand deposit is suggestive of a sand-blow deposit and feeder dike related to asymmetrical ground failure. We decided to conduct GPR surveys in two locations and to excavate three trenches at the site (figure 3).
Ground-Penetrating Radar Surveys
The primary reason for conducting GPR surveys was to help site exploratory trenches T1 and J1. Details of GPR data acquisition, processing, and interpretation are presented in Al-Shukri et al. (2006). We used a 400 MHz antenna suited for resolving discontinuities in the upper 5 m of sandy sediment. In the GPR surveys, a strong reflector interpreted as the contact between the sand and the underlying silt loam was clearly visible (figure 4). A sharp discontinuity in this reflector at the eastern end of the survey at T1 was interpreted to be a sand dike. This interpretation was borne out by observations in the trench as described below.
Observations in Trenches at T1 and T2
We sited two trenches at T1 and T2 to cross the prominent northwest lineament with which a large sand deposit is associated (figure 3). Trench 1 was excavated along the same east-west transect as soil probing and GPR surveying (figure 4). Near the eastern end of trench 1, we found a large sand dike in the location suggested by probing and GPR surveying (figure 5). The large sand dike is structurally related to the large sand deposit, indicating that it is a sand blow. The sand dike is 1.22 m wide and composed of coarse, medium, and fine sand, has subvertical flow structure, and contains large clasts of silt loam from the soil horizon, which the dike crosscuts (figure 5). Both margins of the dike are sharp, nearly vertical, and have strikes and dips of N9-10°W, 83-84°NE. In addition, the eastern margin of the dike is stepped, suggesting that the dike formed as a result of lateral spreading as opposed to erosion. A small (4.5 cm wide), discontinuous, subvertical sand dike also crosscuts silt loam about 30 cm east of the large dike.
The sand blow is composed of one depositional unit that fines upward from coarse, medium, and fine sand to silty, very fine sand. A few small pebbles and clasts of sandy loam and silt loam occur in the basal portion of the sand blow. The sand-blow deposit is thickest (2.45 m) directly above the sand dike and gradually thins toward the west (figures 4 and 5). It is much thinner (0.55 m) on the east side of the sand dike and rapidly thins toward the east. The sand blow buries a 20-cm-thick brownish, sandy loam on both sides of the sand dike.
Structural and sedimentological characteristics of the sand dike and related sand-blow deposit are consistent with an earthquake origin and not with fluvial deposition. It appears that the ground fissured and experienced vertical and lateral displacements of about 2.3 m and 1.2. m, respectively, as sand-bearing water vented through the fissure from below and deposited a large volume of sand on the ground surface. Geotechnical testing along the western and southern boundary of the site identified a thick sandy layer 12–20 m below the surface. This sandy layer is similar to the sand-blow deposit and likely to be the source bed that liquefied (T. Holzer, written communication 2005).
The upper 75–150 cm of the sand blow is intensely iron-stained and bioturbated, suggesting that it is hundreds, if not thousands, of years old. Because no charcoal or plant remains could be found in the buried soil, we collected a soil sample 0–1 cm below the basal contact of the sand blow (DBT1-S1; see table 1). Radiocarbon dating of the sample yielded a two-sigma calendar calibrated age range of 3620–3590 and 3530–3360 B.C. (or radiocarbon calibrated age range of 5,570–5,540 and 5,480–5,310 years B.P.). OSL dating of a sample collected from the same context yielded IRSL ages of 3,820 ± 310 and 3,790 ± 210 solar years and a blue light age of 3,990 ± 330 solar years (DBT1-OSL1; see table 2), considerably younger than the radiocarbon age of the buried soil. Charcoal from bioturbated silt loam below the thin eastern margin of the sand blow (DBT1-C1) gave a modern radiocarbon age and may have been from a recent tree root.
We excavated trench 2 about 230 m southeast of trench 1 and across the projected strike of the large sand dike in trench 1 (figure 3). The 6-m-long trench also revealed a sand dike and related sand-blow deposit (figures 6 and 7). Here, the sand dike is 1.25 m wide and composed of silty, fine to very fine sand. Both margins of the sand dike are sharp, steeply dipping, and strike between N20-25°W. The sand blow is composed of one depositional unit and in general fines upward and away from the sand dike. Although not as thick as in trench 1, the sand blow is 1.45 m thick immediately above the sand dike and gradually thins toward the west. The sand blow is only 1.15 m thick on the east side of the sand dike and thins toward the east. At this location, the ground surface appears to have fissured and been displaced vertically by at least 0.3 m and laterally by about 1.25 m, as sand-bearing water vented to the ground surface. As in trench 1, the ground subsided more on the western side of the dike than the eastern side. Although lateral displacement was similar at the two trench locations, less sand vented on the ground surface and less subsidence occurred at the second trench location.
The entire thickness of the sand blow is bioturbated and iron-stained (figures 6 and 7). Root casts, filled with gray silt and outlined by iron-staining, extend through the sand blow and into the sand dike. The weathering characteristics of the sand blow and sand dike indicate that they are quite old. Unfortunately, the soil buried by the sand blow also was bioturbated and we could not find organics in the soil or root casts for dating. OSL dating of sediment collected immediately below the sand blow yielded an IRSL age of 4,640 ± 280 solar years and a blue light age of 4,860 ± 300 solar years. This result is older than the OSL dates of the soil buried beneath the sand blow in trench 1, but it is still younger than the radiocarbon age of the buried soil. Radiocarbon dating of charcoal (C2) collected from sandy loam developed in the top of the sand blow yielded a calendar calibrated age of A.D. 1440–1640 (Figure 7). OSL dating of sediment collected from the same context yielded a blue light age of 819 ± 63.9 solar years. The dates support a prehistoric age for the liquefaction features but probably do not provide close minimum age constraint. Unfortunately, surface soils at this site have been disturbed by grading and plowing.
Observations in Trenches at J1
As interpreted from aerial photographs and confirmed by reconnaissance, a large sandy deposit occurs near the western margin of the site (figure 3). We excavated two perpendicular and intersecting trenches at this location, J1, and found one meter of iron-stained sand overlying mottled silt. Both sedimentary units are bioturbated. No feeder dike was found in the trenches at J1, but the sand deposit is similar in texture and degree of weathering to the sand deposit at in T1 and T2, suggesting that they are related. At J1, the contact between the sand and the underlying silt is very irregular. In a gross sense, the resulting structures resemble load casts and diapirs (figure 8). Some load casts are thought to be syndepositional and related to density instability in layered sediment (Allen 1982). Other load casts and associated diapirs have been attributed to earthquake-induced liquefaction (Sims 1973; Tuttle 1999).
Interpretation of Observations
The surface expression as well as sedimentological and structural relations of sand-blow deposits and related sand dikes at T1 and T2 of the Daytona Beach site suggest that they are part of the same feature that formed during one earthquake. Given its similarity in weathering characteristics, the sand-blow deposit at J1 likely formed during the same event. However, it is not clear if it is a separate sand blow or a distal part of the sand blow exposed at T1 and T2. Radiocarbon dating of the soil buried by the sand blow at T1 suggests that the liquefaction features formed about 3500 B.C. (or 5,500 years B.P.). It would be beneficial to conduct additional radiocarbon dating at this site. IRSL ages (more appropriate to use than blue light ages given the fine-grained nature of sample; S. Mahan, written communication 2005) of the buried soil are younger than the radiocarbon ages by 800–1,600 years.
The Daytona Beach sand blows are very large. On aerial photographs, they appear to be tens of meters wide and hundreds of meters long (figure 3). As seen in soil pits and trenches, the sand blow adjacent to the northwest-oriented lineament is at least 2.45 m thick, 70 m wide, and 230 m long. This sand blow is as large as sand blows in the heart of the NMSZ (Tuttle and Barstow 1996; Tuttle et al. 2002). The large size of the Daytona Beach sand blows suggests that they formed as a result of a very large earthquake, perhaps similar in size to the 1811–12 New Madrid mainshocks.
The style of deformation at the Daytona Beach site (i.e., large lateral and asymmetrical vertical ground displacements in association with vented sand deposits) is typical of lateral spreading. Southwestward sloping topography may have promoted lateral spreading toward nearby (∼1.5 km) Big Cypress Creek (figure 3). Lateral spreading could help to explain the large size of the Daytona Beach features. However, lateral spreading is also common in the NMSZ.
The Daytona Beach sand blow occurs along a >1.5-km-long, northwest-oriented lineament defined by light- and dark-colored areas (figure 3). These light and dark areas may reflect other sand blows and subsided areas related to ground failure during the same or different events. The lineament, however, seems to be unusually long and straight to be related solely to lateral spreading, which often results in curvilinear scarps. Furthermore, it is one of several lineaments defining a northwest-oriented zone that is subparallel to the nearby WRFZ and to a portion of Crowley's Ridge western flank (figures 2 and 9). These spatial relations suggest that faulting may control ground failure at the site. However, determining the underlying cause of ground failure at the Daytona Beach site, as well as Holocene history of ground failure along lineaments in the area, will require additional fieldwork, dating of selected samples, and ground failure analysis.
INVESTIGATION OF ST. FRANCIS 500 SITE
The St. Francis 500 site is located about 12 km northeast of Marianna, 12 km southeast of the Reelfoot Rift margin, within the White River fault zone, and near the Marianna Escarpment (figure 2). Holocene backswamp and meander-belt deposits have been mapped in this area (Saucier 1994). As seen on aerial photographs, the site occurs within a dark-colored area reflecting poorly drained soils (figure 10). During reconnaissance in this area, we found backswamp deposits of organic-rich silt and clay overlying fluvial deposits of interbedded silt and sand.
Observations of Cutbank Exposure
At the St. Francis 500 site, a cluster of five sand-blow deposits and related sand dikes composed predominantly of fine sand occurs about 4.7 m below the ground surface at the contact between backswamp and underlying fluvial deposits (figure 11). Each sand-blow deposit is clearly related to a feeder dike. The structural relationships of the sand dikes and sand-blow deposits support an earthquake origin. Sand dikes range from 1 to 56 cm wide and strike northwest. The largest dike exhibits flow structure that is delineated by lignite pebbles.
The lower four sand blows are stacked one on top of the other with no intervening soil development or deposition, indicating that they formed over a short period of time, perhaps days to weeks. The uppermost sand blow is interbedded with the basal portion of the backswamp deposits, suggesting that it formed a few months to years later than the other sand blows. From lowermost to uppermost, the sand blows are 42 cm thick and 12 m wide, 18 cm thick and 9.8 m wide, 60 cm thick and 8 m wide, 5 cm thick and 5.8 m wide, and 42 cm thick and 3 m wide. These are minimum values since the sand blows were only partially exposed in one cross-section. Portions of the lower three sand blows appear to fill a bowl-shaped structure (figure 11). This is typical of sand blows in the central United States and results from subsidence as underlying sand is liquefied and vented on the surface. In this case, a layer of interbedded silt and sand below the lowermost sand blow subsided at least 50 cm. In addition, the basal contacts of the silt beds appear to have been eroded and portions of the silt foundered into the underlying sand deposit. The sand deposit exhibits soft-sediment deformation including dish structures. Several of the feeder dikes originate in this sand deposit, indicating that it is the layer that liquefied. It appears that the sandy portion of the fluvial deposit liquefied and fluidized repeatedly to form several generations of sand dikes and sand blows.
Radiocarbon dating of organic sediment from the backswamp deposits and 10 cm above the uppermost sand blow (SF500-C3) yielded a calibrated age range of 4940–4760 B.C. (6,900–6,710 radiocarbon years B.P.) (figure 11). Dating of a nutshell from the layer of interbedded silt and sand below the sand blows (SF500-C1) gave a conventional age of greater than 45,000 radiocarbon years B.P. OSL dating of sediment collected 13 cm above the uppermost sand blow (SF500-OSL2) yielded an IRSL age of 7,720 ± 360 solar years and a blue light age of 9,550 ± 630 solar years. Dating of basal sediments of the backswamp deposits (SF500-OSL1) yielded an IRSL age of 6,810 ± 400 solar years and a blue light age of 8,760 ± 1,100 solar years. The IRSL ages compare more favorably than the blue light ages to the radiocarbon date of the backswamp deposits. The IRSL age of the lower of the two OSL samples is similar to the radiocarbon date.
Interpretation of Observations
The sand blows and sand dikes at St. Francis 500 indicate that this area was struck by four strong earthquakes over a short period of time and followed by a fifth event several months to years later. This close timing of the earthquakes is suggestive of an earthquake sequence similar to the 1811–12 New Madrid sequence. In addition, the St. Francis sand blows are similar in thickness to sand blows in the NMSZ. Most of the sand blows attributed to the 1811–12 earthquakes are 15–60 cm thick, and when stacked one on top of another, 0.2–1.4 m thick (Tuttle et al. 2002). The St. Francis sand blows are 18–60 cm thick, and the lower four sand blows together are 1.16 m thick. Their large sizes are suggestive of very large local earthquakes. Therefore, the St. Francis sand blows may have formed during a sequence of very large earthquakes centered in the Marianna area. Radiocarbon dating of the backswamp deposits provides a minimum age estimate of the sand blows and indicates that they formed before 4760 B.C. (6,710 radiocarbon years B.P.). The IRSL age of the basal backswamp deposits is in close agreement with this estimate.
The lower four sand blows are immediately overlain by backswamp deposits. This relation suggests that the event responsible for the formation of the sand blows also may have changed the environment of deposition. The northeast margin of the backswamp deposits is flanked by the eastern boundary of the WRFZ (figure 10). Perhaps movement on the fault zone caused ponding in this area. Alternatively, the liquefaction features at St. Francis 500 may be related to movement on the ERRM (figure 2). The Marianna Escarpment and associated light-colored lineaments and patches that appear to be sand blows are subparallel to the ERRM and suggest a tectonically active, northeast-oriented structure in close proximity to St. Francis 500 (figure 10).
The Daytona Beach sand blow is very large and formed about 3500 B.C. (5,500 years B.P.). Its large size may have been due in part to lateral spreading. Nevertheless, the sand blow probably formed as a result of a very large earthquake centered near Marianna. The large displacements and linear nature of the failure and its spatial relationship with other northwest-oriented lineaments suggest that the Daytona Beach ground failure might be related to faulting. Although no fault has been mapped in the vicinity of the Daytona Beach site, the WRFZ is mapped only 8 km away and has a similar trend to the lineaments (figure 2). In addition, a fault zone associated with the ERRM is located about 18 km northwest of the Daytona Beach site. A very large earthquake generated by either of these fault zones probably would be capable of inducing liquefaction and ground failure at this site.
Liquefaction features similar in age to the Daytona Beach sand blow have been found elsewhere in the Mississippi embayment. In the NMSZ, a buried sand blow (Eaker 2) near Blytheville, Arkansas, about 150 km northeast of Marianna, formed about 3340–2210 B.C. (5,340–4,210 years B.P.) (Tuttle 1999). The buried sand blow is 25 cm thick and more than 3 m wide, much smaller than the Daytona Beach sand blow. Other liquefaction features of similar age have been found in Kelso and Montrose, Arkansas, about 120 and 175 km southwest of Marianna, respectively. Evidence for “major venting” near Kelso and Montrose between 4,600–5,500 years B.P. and for “minor venting” near Montrose between 5,100–6,000 years B.P. are attributed to an earthquake about 5,700 years ago generated by a local source, possibly the Arkansas River fault zone (Cox et al. 2004). The sand blows at the Kelso and Montrose sites range from 10 to 90 cm in thickness and 12 to 60 m in diameter and are small compared with the Daytona Beach sand blow. According to Ambraseys's (1988) magnitude-distance relations, a moment magnitude, M > 7.2 earthquake could produce sand blows 175 km from its epicenter. Therefore, a very large earthquake centered near Marianna could be responsible for liquefaction features in both the Blytheville and Montrose areas. The ages of Middle Holocene liquefaction features in all three areas, however, would need to be better constrained and the intervening areas searched for similar age features in order to correlate them with confidence.
The sand blows at St. Francis 500 formed about 4800 B.C. (6,800 years B.P.). There are no known sand blows of this age in the NMSZ. There are liquefaction features in southern Illinois that formed in 4520 B.C. ± 160 yr (∼6,500 years B.P.) for which the NMSZ has not been ruled out as the earthquake source (Tuttle et al. 1999; Tuttle 2005). The St. Francis sand blows are similar in their compound nature and large size to New Madrid sand blows, suggesting that they formed as the result of a sequence of very large earthquakes centered near Marianna. Possible nearby sources include the WRFZ and ERRM. The St. Francis liquefaction site occurs within the WRFZ and about 4.5 km southwest of its eastern boundary. Backswamp deposits that bury the sand blows appear to terminate along the boundary of the fault zone, and feeder dikes at the site are subparallel to the fault zone. In addition, the mapped traces of the ERRM and the Marianna Escarpment are only 12 km and 1.5 km northwest of the site, respectively.
Possible sources of Middle Holocene earthquakes in the Marianna area include the ERRM, WRFZ, and Big Creek fault zone (BCFZ) (figure 2). Of the three structures, the ERRM seems the most likely candidate because it is a long structure (> 300 km; Johnson et al. 1994) and may have been the source of several modern earthquakes in southwestern Tennessee as well as the M 6.1, 17 December 1811 New Madrid aftershock (Chiu et al. 1997; Hough and Martin 2002). In addition, the ERRM is a member of the fault system that produced very large New Madrid earthquakes in 1811–12, A.D. 1450 ± 200 yr, A.D. 900 ± 200 yr, A.D. 300 ± 200 yr, and 2350 B.C. ± 200 yr (Tuttle et al. 2002, 2005) and experienced Late Wisconsin–Early Holocene movement in western Tennessee (Cox et al. 2001). Different faults within the system may have been active in 4800 B.C. and 3500 B.C. Failure of one fault in 4800 B.C. may have moved another fault closer to failure in 3500 B.C. And in 4800 B.C., there could have been interaction between northeast- and northwest-oriented faults in the Marianna area, as there was in the NMSZ during the 1811–12 earthquake sequence. At this time, however, we cannot establish a direct link between earthquake-induced liquefaction at these sites and any of these fault zones. This may be possible to do after additional geophysical and geological investigations in the area. IRSL ages are in agreement with radiocarbon dates at Saint Francis 500 but not at the Daytona Beach site. Additional comparison of radiocarbon and OSL ages of collocated samples is needed to determine the accuracy and reliability of OSL dating in this environment.
If the northern portion of the ERRM were active during the Late Wisconsin-Early Holocene (Cox et al. 2001), the southern portion of the ERRM were active during the Middle Holocene (this study), and the Reelfoot fault system, source of New Madrid earthquakes, were active only during the Late Holocene, it would mean that seismicity within the Reelfoot Rift system varies in space and time. If confirmed, such a finding would have implications for the earthquake source model of the central United States. In addition, it would add credence to the hypothesis that failed rifts, even in plate interiors, have the potential to produce very large earthquakes and that currently aseismic portions of those structures could generate very large earthquakes in the future (Johnston et al. 1994; Wheeler 1995, 1996; Ebel and Tuttle 2002).
We found large earthquake-induced sand blows and related sand dikes at two sites in the Marianna area about 180 km southwest of the center of the NMSZ. The sand blow at the Daytona Beach site in the Western Lowlands is deeply weathered and visible on the ground surface, whereas the sand blows at St. Francis 500 in the St. Francis Basin are relatively unweathered, and not visible on the surface because they are buried by Holocene backswamp deposits. On the basis of radiocarbon dating, we estimate that the liquefaction features at the Daytona Beach and St. Francis sites formed about 3500 B.C. and 4800 B.C., respectively.
Liquefaction features of similar age to, but of smaller size than, the Daytona Beach sand blow occur near Blytheville (150 km northeast of Marianna) and Montrose, Arkansas (175 km southeast of Marianna). A very large (M > 7.2) earthquake centered near Marianna about 3500 B.C. might account for the liquefaction in all three areas. The large sand blows at the St. Francis site are similar to compound sand blows in the NMSZ, suggesting that a New Madrid-type earthquake sequence was centered near Marianna about 4800 B.C.
Several fault zones in the Marianna area, including the EERM, WRFZ, and BCFZ, are thought to be active on the basis of their apparent influence on local topography and hydrography. The ERRM appears to be the most likely source of very large earthquakes during the Middle Holocene. Additional investigations of sand blows, prominent lineaments, and fault traces are needed to confirm our findings and to better constrain the timing, sources, and magnitudes of paleoearthquakes in the Marianna area. This information, in turn, would improve understanding of the long-term rupture history and interaction of faults within the Reelfoot Rift system, and thus, of seismogenesis in plate interiors.▪
Research presented in this manuscript was supported by U.S. Geological Survey grant 1434-3HQGR0050. The U.S. Geological Survey also provided supplemental funds for radiocarbon and OSL dating. The views and conclusions presented in this paper are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of the U.S. government. We are grateful to the property owners, Don Johnson and Terry Toler, who gave us access to the sites. Marilyn Egan, Sami Eyuboglu, Jeff Shaw, and his father Mr. Shaw, assisted with fieldwork. Reviews by Pradeep Talwani and Rus Wheeler led to clarification of several points and helped to improve this paper.
M. Tuttle & Associates
University of Arkansas at Little Rock