- © 2005 by the Seismological Society of America
Six episodes of earthquake-induced liquefaction are associated with soil horizons containing artifacts of the Late Archaic (3000-500 B.C.) and Early to Middle Woodland (500 B.C.-A.D. 400) cultural periods at the Burkett archaeological site in the northern part of the New Madrid seismic zone, where little information about prehistoric earthquakes has been available. Radiocarbon dating of organic material and analysis of artifacts are used to estimate the ages of the liquefaction features and times of the causative earthquakes. The most recent episode of liquefaction occurred after A.D. 1670, produced small sand dikes, and is probably related to the 1895 Charleston, Missouri earthquake. The preceding episode struck the area in A.D. 300 ± 200 years and generated a sand blow that contains Late Woodland artifacts and buries an Early to Middle Woodland cultural horizon. Four older episodes of liquefaction occurred in 2350 B.C. ± 200 years and may have been produced by a sequence of closely timed earthquakes. The four earlier episodes produced graben structures, sand dikes, and associated sand blows on which a cultural mound was constructed. The Burkett liquefaction features that formed about 2350 B.C. and A.D. 300 are relatively large and similar in age to other liquefaction features in northeastern Arkansas and southeastern Missouri, respectively. If the prehistoric features at the Burkett site and those of similar age elsewhere in the region are the result of the same earthquakes, then this suggests that they were similar in size to the three largest (M 7-8) 1811-1812 New Madrid earthquakes. A New Madrid-type earthquake in A.D. 300 ± 200 years would support an average recurrence time of 500 years. Although this study extends the earthquake chronology back to 2500 B.C., it is uncertain that the record of New Madrid events is complete for the period between 2350 B.C. and A.D. 300. As demonstrated by this study, information about other prehistoric earthquakes may be buried beneath fluvial deposits of the Mississippi River Valley.
Paleoseismological studies over the past decade have begun to unravel the earthquake history of the New Madrid seismic zone (NMSZ) and have revolutionized our understanding of its hazard (Figure 1). Very large (moment magnitude M 7-8) 1811-1812-type or New Madrid events about A.D. 900 and A.D. 1450 have been recognized through the study of earthquake-induced liquefaction features across the New Madrid region and fault-related deformation of the Reelfoot scarp in Tennessee (e.g., Kelson et al., 1996; Tuttle et al., 2002). From a 1,200-year-long earthquake chronology, a mean recurrence time of 500 years has been estimated for New Madrid events. These findings, in turn, changed estimates of the regional earthquake hazard as reflected in the most recent national probabilistic seismic hazard map (Frankel et al., 2002). Estimates of recurrence times derived from the paleoseismological record may be an improvement over those derived from the shorter seismological record, but they are based on only three events in 1,200 years, or two earthquake cycles. A more complete Holocene record of New Madrid events is needed to characterize the variability of recurrence intervals better, to reduce the uncertainty in estimates of mean recurrence time, and to improve understanding of the behavior of the New Madrid fault system. This paper presents new evidence of earthquake-induced liquefaction and ground failure from an archaeological site in southeastern Missouri, considers the new evidence in combination with previous observations of liquefaction features that formed prior to A.D. 900, and re-evaluates the mean recurrence time for New Madrid events over three rather than two earthquake cycles.
PREVIOUS STUDIES OF LIQUEFACTION FEATURES THAT PREDATE A.D. 900
Previous paleoliquefaction studies have found evidence for strong earthquakes prior to A.D. 900, but the ages as well as the locations and magnitudes of those earthquakes are poorly constrained (Tuttle, 2001; Tuttle et al., 2002). Older liquefaction features have been studied in detail at the Towosahgy site in southeastern Missouri and Eaker 2 and Main 8 sites in northeastern Arkansas (Figure 1). Findings at the three sites indicate that the New Madrid region was struck by at least three earthquakes large enough to induce liquefaction (about M > 6.4) between 3340 B.C. and A.D. 900.
At the Towosahgy archaeological site, located about 38 km northeast of the Reelfoot scarp in Tennessee, two successive sand blows and related sand dikes occur below the base of a temple mound constructed during the Mississippian (A.D. 800-1650) cultural period (Table 1; Price et al., 1990; Saucier, 1991). The lower sand blow buries a clay layer interpreted as a natural levee deposit on which habitation was established during the Late Woodland (A.D. 400-1000) cultural period. Cultural material was incorporated in the upper 30 cm of the lower sand blow, aboriginal pits were dug into the top of the sand blow, and a cultural horizon was developed above the sand blow. Radiocarbon dating of charcoal from the lower sand blow is thought to provide a close minimum age constraint; the lower sand blow was estimated to have formed sometime during the one hundred years prior to A.D. 539 (Price et al., 1990; Saucier, 1991). The upper sand blow overlies the cultural horizon developed in the top of the lower sand blow. The upper sand blow was estimated to have formed between A.D. 539 and A.D. 991 from radiocarbon dates of charcoal samples collected above it and from the lower sand blow (Price et al., 1990; Saucier, 1991). These radiocarbon dates for Towosahgy are intercept dates of the average radiocarbon ages with the calibration curve, not the preferred 1- or 2-sigma calibrated age ranges used for other sites described in this paper.
At the Eaker 2 site, located near Blytheville, Arkansas, three buried sand-blow deposits, several small sand dikes, and a large (20- to 60-cm-wide) sand dike were studied in three subparallel trenches (Tuttle, 1999). The large sand dike crosscuts all three sand blows and is estimated to have formed between A.D. 800 and 1000. The three buried sand blows directly overlie paleosols and are connected to small (< 3-cm-wide) feeder dikes. The upper sand blow is estimated to have formed between 800 B.C. and A.D. 780. This time estimate is based on radiocarbon dating of organic material in the paleosol buried by the upper sand blow as well as a clast from an overlying native occupation horizon within the large sand dike that cuts the sand blow. The middle and lower sand blows formed between 1430 and 800 B.C. and 3340 and 1250 B.C., respectively, as determined from radiocarbon dating of organic material within underlying and overlying paleosols.
At the Main 8 site, located only 8 km north of Eaker 2, four generations of liquefaction features formed since 4040 B.C. (Tuttle and Schweig, 1995). Most of the liquefaction features at the site are highly weathered, suggesting that they are prehistoric in age. Radiocarbon dating of wood in the clay deposit below the oldest sand blow provides a maximum age constraint of 4040-3360 B.C.
INVESTIGATIONS AT THE BURKETT ARCHAEOLOGICAL SITE
The Burkett archaeological site is located in the northern part of the NMSZ, about 50 km northeast of the Reelfoot scarp (Figure 1). A previous archaeological study at the site found evidence for two occupations, including an O'Bryan Ridge component (Late Archaic) and a Burkett component (Early to Middle Woodland) (Table 1; Hopgood, 1969). In 1999, Prentice Thomas and Associates (PTA) conducted an archaeological investigation at the site under contract with the Memphis District Army Corps of Engineers. During the investigation, they opened many trenches, test pits, and block excavations (Figure 2; Thomas et al., 2005). The investigation confirmed the presence of major Late Archaic and Early to Middle Woodland occupations, with very minor and localized occurrences of Middle Archaic and Terminal Archaic components. In addition to recovering prehistoric artifacts, PTA discovered earthquake-induced liquefaction features in several excavations.
In our early examination of the excavations, we found that many of the liquefaction features are closely associated with cultural horizons and features containing artifacts. Stratigraphic relations clearly indicate that some of the liquefaction features are prehistoric in age, and the abundance of artifacts and organic material suggested that it would be possible to constrain the ages of those features. Therefore, we decided to conduct a paleoseismological investigation at the site. Our investigation included logging parts of trenches 5 and 6, test pit 56, and block 7 (Figure 2), where liquefaction features occurred; observing and interpreting structural and stratigraphic relations; and collecting charcoal samples for radiocarbon dating and artifacts for archaeological analysis. Radiocarbon dating was conducted by Beta Analytic, Inc. (Table 2) and artifact analysis was carried out by PTA (Thomas et al., 2005). Below, we present our observations and interpretations of earthquake-induced liquefaction features and related ground failures at the Burkett site.
Liquefaction Features in Test Pit 56 and Trench 6
Test pit 56 was excavated through liquefaction features exposed in the eastern wall of trench 6 (Figures 2 and 3). In the east wall of the test pit, we observed a 2-cm-wide, branching sand dike (unit 3b) trending N62°W and crosscutting units 4, 6a, and 6b and a fracture zone and related sand vent (unit 5) preserved in units 6a and 6b. Unit 4 is a dark gray loam containing Burkett-phase artifacts of the Early to Middle Woodland cultural periods (500 B.C.-A.D. 400). Unit 6a is a dark grayish brown silt that contains O'Bryan Ridge-phase artifacts of the Late Archaic cultural period (3000-500 B.C.; see Table 1). Unit 6b is a silt loam and contains only a few O'Bryan Ridge-phase artifacts. In the eastern wall of the test pit, units 6a and 6b are displaced downward by about 9 cm on the north side of the sand dike, and unit 4 is thickened above the down-dropped horizons. Similarly, unit 4 is thickened over the sand vent. Unit 4 is overlain by unit 3a, a 30-cm-thick fine to medium sand in which a very dark gray loamy soil has formed. Unit 3a is connected to the small sand dike (unit 3b) and therefore is interpreted to be a sand blow. A discontinuous, 2-cm-wide sand dike, designated unit 1, crosscuts the upper part of the sand blow immediately above the older sand dike and either terminates against or is truncated by a fluvial deposit of silty, fine to coarse sand (unit 2a). Unfortunately, the uppermost part of the section, which might have helped constrain the minimum age of the sand blow, had been removed here during site clearing.
On the floor of the test pit and adjacent trench 6, we found that the small sand dike and fracture zone in test pit 56 extend toward and merge with the margins of a large sand dike exposed in the west wall of trench 6 (Figure 4). These relations suggest that the sand dikes and sand vent in test pit 56 probably were emplaced along the same structure but at different times. During excavation of the test pit, we recovered a few Burkett-phase ceramic artifacts and a single Madison point from unit 3a, the sand blow. The Burkett-phase artifacts are indicative of the Early to Middle Woodland cultural periods, and the Madison point is diagnostic of both the Late Woodland (A.D. 400-1000) and Early and Middle Mississippian (A.D. 800-1400) cultural periods. Many more Burkett-phase artifacts were recovered from the underlying unit 4 than from the sand blow. Unfortunately, we did not find datable organic material in test pit 56.
We also logged an 8-m-long section of the adjacent west wall of trench 6 (Figure 2), where a near-vertical, 58-cm-wide sand dike (Figure 4), trending N51°W, crosscuts units 4, 6a, and 6b (Figure 5). The stratigraphy in the trench is essentially the same as that in test pit 56 except that less of the plow zone and underlying fluvial deposits has been removed. The large dike (unit 3b) of fine to medium sand has subvertical flow structure and is connected to unit 3a, which we interpret to be a sand-blow deposit. As in test pit 56, the sand blow is 30 cm thick, is composed of fine to medium sand, and has been subjected to soil development. Fragments of baked clay occur within the sand blow, suggesting that it was occupied sometime after deposition. Unit 3a is overlain by sandy units 2a, 2b, and 2c, each of which is cross-bedded, coarsens upward, and is probably of fluvial origin. Radiocarbon dating of charcoal (TR6-C100) from unit 2c yielded calibrated dates of A.D. 1670-1780 and 1795-1955 (Table 2). A discontinuous, 3-cm-wide sand dike, designated unit 1, intrudes the northern margin of the large sand dike (unit 3b), crosscuts the large sand blow (unit 3a), and extends into the overlying fluvial deposits (units 2a, 2b, and 2c), where it terminates (Figure 5). This sand dike is probably a lateral extension of the youngest dike (unit 1) exposed in the east wall of test pit 56 (Figure 3).
Test pit 56 and adjacent trench 6 expose evidence of three episodes of liquefaction and related ground failure (Figures 3 and 5). The oldest episode involved fracturing and vertical displacement of units 6a and 6b and formation of a small sand vent, unit 5 (Figure 3). The second episode involved intrusion of units 6a, 6b, and 4 by a sand dike (unit 3b), which is 58 cm wide in trench 6 and 2 cm wide in test pit 56, and deposition of a 30-cm-thick sand blow that buried unit 4. Analysis of artifacts and radiocarbon dating of nut fragments from across the site indicate that unit 4 was occupied from about 400 B.C. to A.D. 330 (Thomas et al., 2005). The soil developed in the overlying sand blow (unit 3a) contains a few Burkett-phase artifacts, suggesting that the site continued to be occupied following the second episode of liquefaction. Artifacts are much more abundant in deposits below the sand blow than within it, so we think that the earthquake probably occurred toward the end of the Burkett phase or in A.D. 300 ± 200 years. The youngest liquefaction episode was apparently less severe and involved intrusion of small sand dikes into older liquefaction features. The youngest sand dikes terminate within historic fluvial deposits.
The oldest episode of earthquake-induced liquefaction as seen in test pit 56 and trench 6 occurred after deposition of O'Bryan Ridge-phase artifacts and before deposition of Burkett-phase artifacts. The second episode occurred toward the end of deposition of Burkett phase artifacts or in A.D. 300 ± 200 years. The youngest episode occurred during or after deposition of the fluvial deposits, or since A.D. 1670. Both the 1811-1812, M 7-8, New Madrid and 1895, M 6.6, Charleston, Missouri earthquakes caused liquefaction northeast of New Madrid, but the 1895 Charleston earthquake produced small liquefaction features and related ground failures in the vicinity of the Burkett site (Powell, 1975). Given the small size of the historic liquefaction features at the Burkett site, we favor the interpretation that they formed during the 1895 Charleston earthquake rather than during the larger 1811-1812 earthquakes. As we have seen across the New Madrid region, not all earthquakes induced liquefaction at all sites.
Liquefaction Features in Trench 5
Trench 5 is located only 8-10 m west of trench 6 and test pit 56, respectively. Deformation structures observed in all three excavations are northwest-oriented and probably related to one another. We logged a 4-m-long section of the east wall of trench 5 where units 6a and 6b are down-dropped about 30 cm to form a graben, shown between the 1-m and 3-m marks on Figure 6. The graben is 1 m wide and trends N80°W. Sand dikes intruded along the margins of the graben are present in the floor of the trench. Unit 6a contains O'Bryan Ridge-phase artifacts and pieces of charcoal. Charcoal sample TRS-C5 from this unit yielded a calibrated date of 2580-2430 B.C. (Table 2), which is consistent with the previously estimated age of O'Bryan Ridge-phase artifacts. Unit 4 fills the graben, buries unit 6a adjacent to the graben, and contains Burkett-phase artifacts and bone fragments. Several small dikes of medium to fine sand (unit 3b) intrude the graben fill. Unit 3a is a 20-cm-thick silty, fine to medium sand, in which a very dark gray loam has formed. Except for being thinner, this sand layer is identical to and probably part of the sand-blow deposit in nearby trench 6 and test pit 56. Unit 3a is buried by sandy units 2a, 2b, and 2c that coarsen upward, have cut-and-fill structures, and are of fluvial origin. Charcoal (TR5-C9) from unit 2b gave calibrated dates of A.D. 1680-1740, 1805-1930, and 1950-1955 (Table 2).
Evidence for two episodes of liquefaction and related ground deformation are present in trench 5. The first episode involved 30 cm of vertical displacement of units 6a and 6b to form a graben and intrusion of small dikes along the margins of the structure. During the second episode, dark gray loam of unit 4 dropped down along the southern graben margin, and several discontinuous dikes (unit 3b) intruded unit 4 within the graben. Graben formation occurred during both episodes and is indicative of lateral spreading. As in test pit 56, the first episode of liquefaction occurred after deposition of O'Bryan Ridge-phase artifacts and before deposition of Burkett-phase artifacts. Radiocarbon dating of unit 6a indicates that the first episode occurred after 2580 B.C. The second episode occurred after deposition of Burkett-phase artifacts.
Liquefaction Features in Block 7
Although a cultural mound at the Burkett site had been partially removed to facilitate farming, we were able to find the mound's original location from the 1931 New Madrid Floodway map. PTA removed the plow zone, excavated several test pits and blocks, including block 7, and found compact and baked clay layers that they interpreted as the base of the cultural mound (Figure 2; Thomas et al., 2005). The clay may have been excavated from the nearby abandoned channel of the Mississippi River and brought to this location by natives to construct the mound during the Late Archaic cultural period. Figure 7, a photograph of the west wall of block 7, shows the clay layers, which are nearly flat-lying and uniform in thickness, except above a fairly large sand dike where they thicken as if filling the crater of a sand blow. Note that the sand dike is connected to a sand layer interpreted to be a sand-blow deposit.
Following enlargement of block 7, we logged liquefaction features exposed in the west and north walls of the excavation (Figures 8 and 9). Figure 8 shows the log of the west wall located about 0.5 m west of the section shown in Figure 7. The stratigraphy is essentially the same in the two sections, but the sand dike is much less apparent in the more westerly section. Unit 6a dips downward slightly between the 2-m and 3-m marks and is crosscut by a small, discontinuous sand dike (unit 5c, Figure 8). The small dike is all that remains of the large sand dike shown in Figure 7, indicating that this section is close to the lateral termination of the dike. The small dike is connected to an overlying sand-blow deposit of silty, medium to fine sand (unit 5c). Toward the northern end of the section, the sand layer consists of two fining-upward units (5c1 and 5c2) that are separated by a thin silt layer, suggesting that this is a compound sand blow resulting from two closely timed earthquakes. The compact and burned clay layers, units 5b1-3, of the base of the cultural mound overlie the compound sand blow. The clay layers are, in turn, overlain by medium sand, unit 5a, which fines upward and contains clay clasts. As explained below, this medium sand also may be a sand blow.
In the north wall of block 7, units 6a (silty clay), 5c (sand blow), and 5b3 (lowermost clay layer of mound) are dropped down in a graben (Figure 9). Here, unit 6a is intruded by dikes of medium sand, unit 5a. A layer of similar medium sand, also designated as unit 5a, buries the clay layers, 5b1-3, and is interpreted to be a sand blow. The sand layer fines upward, contains clay clasts, and is similar in grain size to the sand dikes. The base of the sand layer is offset by the western margin of the graben but deformation does not extend into the sand blow, suggesting that its deposition was contemporaneous with graben formation. Toward the western end of the exposure, the sand layer is composed of two fining-upward sequences (5a1 and 5a2), suggesting that it too may be a compound sand blow resulting from two closely timed earthquakes.
Two charcoal samples, BL7N-C6 and BL7W-C7, from horizon 6a yielded calibrated radiocarbon dates of 2575-2395, 2375-2355 B.C. and 2340-2190, 2170-2150 B.C., respectively (Figures 8 and 9; Table 2). These radiocarbon dates indicate that horizon 6a in block 7 is similar in age to horizon 6a in Trench 5. Charcoal sample BL7W-C8 from the lowermost clay layer 5b3 in block 7 has a calibrated date of 3005-2975, 2935-2880 B.C., which is at least 300 years older than the stratigraphically lower unit 6a. Older deposits overlying younger deposits support the interpretation that clay layer 5b3 was brought here by humans to build the mound. Charcoal sample BL7N-C3, from a burned layer at the contact of unit 6a and overlying 5b3 clay layer, yielded a calibrated date of 2570-2290 B.C. In this location, the clay layer appears to have been fired directly on unit 6a. Therefore, sample BL7N-C3 provides a contemporary age for the construction of the base of the mound, which began about 2430 ±140 years B.C., or during the Late Archaic cultural period. Artifacts found in other parts of the mound indicate that construction continued later into the Woodland period (Thomas et al., 2005).
In block 7, we found evidence of four episodes of liquefaction and related ground failure. The first two episodes caused the intrusion of a large sand dike and the formation of the lower compound sand blow, unit 5c (Figure 8). The third and fourth episodes resulted in the formation of a graben, the intrusion of several small dikes, and the formation of the upper compound sand blow, unit 5a (Figure 9). Graben formation and vertical displacement of the lower sand blow (unit 5c) and cultural clay layer 5b3 are probably the result of lateral spreading.
Radiocarbon dating of unit 6a indicates that the first two episodes of liquefaction occurred soon after 2340-2190, 2170-2150 B.C. The third and fourth episodes occurred after emplacement of the compacted and burned clay layers, or shortly after 2570-2290 B.C. Within their error limits, these dates are very similar. From structural and stratigraphic relations as well as radiocarbon dating, it is likely that the four liquefaction episodes recorded in block 7 occurred fairly close in time. This conclusion is also supported by the lack of soil development in the lower sand blow and in the cultural layers 5bl-3. The liquefaction episodes would have had to be separated by enough time, however, perhaps several weeks to months, for the natives to transport, compact, and fire the clay layers that formed the base of a cultural mound. We suggest that these four liquefaction episodes occurred about 2350 B.C. ±200 years and are the result of four large earthquakes in a sequence, similar to the 1811-1812 New Madrid earthquake sequence.
DISCUSSION OF RESULTS
At the Burkett archaeological site, earthquake-induced liquefaction resulted in the intrusion by sand dikes and burial by sand blows of soil horizons that were occupied during the Late Archaic (3000-500 B.C.) and Early to Middle Woodland (500 B.C.-A.D. 400) cultural periods. Liquefaction-related ground failures produced vertical displacements of the soil horizons, which disrupted the archaeological stratigraphy. The formation of sand blows and grabens helped preserve the archaeological record, however. Artifacts and organic material in the cultural horizons make it possible to estimate the times of past earthquakes.
At the Burkett site, we found evidence of six generations of earthquake-induced liquefaction features associated with native occupation horizons. The four earliest generations of features formed about 2350 B.C. ± 200 years, possibly as a result of four separate earthquakes in a closely timed earthquake sequence. The fifth generation of liquefaction features formed about A.D. 300 ± 200 years. The sixth and most recent generation of features formed after A.D. 1670, probably during the 1895 Charleston, Missouri earthquake, although the 1811-1812 New Madrid earthquakes can not be ruled out as the causative event.
Earthquakes that induced liquefaction at the Burkett site about 2350 B.C. and A.D. 300 may have been very large New Madrid events based on the relatively large size and compound nature of the liquefaction features. The compound sand blows that formed during the 2350 B.C. event are as much as 20 cm thick and more than 5 m wide. The sand blow that formed in A.D. 300 is as much as 30 cm thick, more than 5 m wide, and 9 m long. Similarly, an 1811-1812 sand blow at Wilkerson Ditch (L2 on Figure 1), also in southeastern Missouri, is about 30 cm thick and more than 5 m wide (Li et al., 1998). In contrast, sand blows that formed near Charleston, Missouri during the M 6.6, 1895 earthquake have been described as small, ranging from about 5 to 95 cm in diameter (Powell, 1975).
The 2350 B.C. earthquakes may have induced liquefaction over a fairly large area, which supports the interpretation that they were very large New Madrid earthquakes. The 2350 B.C. earthquakes may have been responsible for liquefaction at the Main 8 and Eaker 2 sites near Blytheville, Arkansas, where older sand blows formed after 4040-3360 B.C. and between 3340 B.C. and 1250 B.C., respectively. Main 8 and Eaker 2 are located about 115-120 km southwest of the Burkett site (Figure 1). For comparison, the extent of liquefaction resulting from the M 6.6, 1895 Charleston earthquake is thought to be limited to about a 15-km2 area surrounding the inferred epicenter (Powell, 1975; Obermeier, 1989).
Stratigraphic relations of sand blows at the Burkett site suggest that four very large earthquakes induced liquefaction about 2350 B.C., possibly within weeks or months of each other. If this interpretation is correct, it further supports the hypothesis that temporally clustered, very large earthquakes are typical of the New Madrid seismic zone. Compound sand blows attributed to major earthquake sequences also have been described for the A.D. 900 and A.D. 1450 (Tuttle, 1999; Tuttle et al., 2002) as well as the 1811-1812 New Madrid events (Saucier, 1989).
The A.D. 300 earthquake may have induced liquefaction at Towosahgy as well as at the Burkett site (Figure 1). As mentioned above, the lower sand blow at Towosahgy is thought to have formed during the one hundred years prior to A.D. 539. Recalibration of the radiocarbon age of charcoal collected from the lower sand blow yields a two-sigma range of A.D. 250-690. If this date is a close minimum age and the sand blow formed within one hundred years of that date, then it probably formed between A.D. 150-590. Therefore, it's possible that the large sand blows of similar age at Burkett and Towosahgy formed as the result of the same earthquake and that the earthquake was a very large New Madrid earthquake.
Additional information regarding the areal distribution, size, and internal stratigraphy of sand blows that formed during the 2350 B.C. and A.D. 300 events would serve to test our interpretation that these are New Madrid events. This information would help constrain the location of the source areas and the magnitudes of those events and to characterize the behavior of the New Madrid fault system during each event. Acquiring this information, critical to understanding the earthquake potential of the NMSZ, will require regional reconnaissance targeting specific landforms and detailed study of additional pre-A.D. 900 sand blows.
If the past four New Madrid events occurred in A.D. 300, A.D. 900, A.D. 1450, and 1811-1812, the estimate of the average recurrence time remains about 500 years (Figure 10). The time intervals between New Madrid events (600, 550, and 360 years) appear to be getting shorter, however. This may be due to the uncertainty in the age estimates of the events. But if the recurrence times really are getting shorter, this would have important implications for hazard assessments and tectonic modeling. Nevertheless, the longer interval following the A.D. 300 event suggests that a short recurrence time of 200 years is less likely than previously thought (Tuttle et al., 2002). Uncertainty in the recurrence times could be reduced if the age estimates of the pre-1811 earthquakes could be better constrained, particularly for the A.D. 300 (±200 years) and A.D. 1450 (±150 years) events.
If the A.D. 300 ± 200 years and 2350 B.C. ± 200 years earthquakes are in fact New Madrid events, the earthquake chronology can be extended to 2500 B.C., or 4,500 years ago. The period between the 2350 B.C. and A.D. 300 events is 2650 ± 400 years, about five times longer than the estimated average of the three more recent recurrence intervals. This apparent long recurrence time may indicate that either other New Madrid events during this period have not yet been identified in the geologic record or that the current cycle of New Madrid events repeating every 500 years began about 2,000 years ago. So far, limited effort has been made to find and study liquefaction features that formed prior to A.D. 300. One sand blow at the Eaker 2 site near Blytheville, Arkansas formed between 1430 and 800 B.C., and other highly weathered liquefaction features along rivers and ditches across the New Madrid region may have formed prior to A.D. 300. A concerted effort to date some of these older liquefaction features is needed to resolve this important issue.
At the Burkett archaeological site in southeastern Missouri, there is evidence for at least six earthquakes that induced liquefaction and related ground failures during the past 4,500 years. The four oldest earthquakes occurred in 2350 B.C. ±200 years and may have been part of an earthquake sequence that occurred during a period of weeks to months. Sand blows and related sand dikes that resulted from these earthquakes form the foundation on which a Late Archaic and Woodland cultural mound was built. The fifth earthquake occurred in A.D. 300 ±200 years near the end of the Middle Woodland period. The sixth and most recent earthquake occurred after A.D. 1670. Because it produced small sand blows not far from the site, the 1895 Charleston, Missouri earthquake seems a more likely cause of the small historic sand dikes at Burkett rather than the larger 1811-1812 New Madrid earthquakes.
The 2350 B.C. and A.D. 300 earthquakes are interpreted as very large 1811-1812-type or New Madrid events based on the relatively large size of the prehistoric liquefaction features at Burkett, the close timing of the four earthquakes in 2350 B.C., and the possibility that the 2350 B.C. and A.D. 300 earthquakes also caused liquefaction at other sites in northeastern Arkansas and southeastern Missouri, respectively. The average recurrence interval for New Madrid events remains 500 years even when the interval between the A.D. 300 and A.D. 900 events is included in the estimate. The interval between the 2350 B.C. and A.D. 300 events is unusually long, 2650 ±400 years, suggesting that either the record of New Madrid events is incomplete for this time period or that a dramatic change in earthquake frequency occurred about 2,000 years ago.
Currently, the behavior of the NMSZ is characterized by an average recurrence time of 500 years for the past 2,000 years. A more accurate, more complete, and longer record is critical for assessing whether the rate of very large earthquakes in the recent past is typical of longer-term behavior. Our study at the Burkett site demonstrates that important new paleoseismological information still can be gleaned from the late Quaternary record in the New Madrid region.
Research presented in this manuscript was supported by U.S. Geological Survey grants 1434-99HQGR0022 and 1434-02HQGR0097. 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. Jimmy McNeil of the Army Corps of Engineers, Memphis District, made this cooperative research possible. We are grateful to the property owner, Stauffer Farms, Inc., who gave us access to the site and we appreciate the kind cooperation of their tenant, Mr. Donnell Robinson. Sharon Brown of Prentice Thomas and Associates (PTA) and Susana Vilanova of Instituto Superior Technico in Lisbon, Portugal participated in the site investigation. Kathy Tucker of the Center for Earthquake Research and Information produced the map of the New Madrid region, and Jon Wesson of PTA made the map of the Burkett site. Jim Morehead, also of PTA, provided a helpful review of archaeological aspects of the manuscript. Tony Crone, Kathleen Haller, Michael Machette, and Rus Wheeler of the U.S. Geological Survey also provided thoughtful and constructive reviews.