Sequence stratigraphy, a major theoretical achievement of earth sciences, integrates facies associations and stratal architecture within a chronological framework of the geological record. The sequence stratigraphic model implies that sediment supply to the basin, preservation potential of individual horizons, and the resolution of paleontological data co-vary with base-level fluctuations, especially for siliciclastic depositional systems. Using Holocene transgressive-regressive successions of the Po Plain (Italy), we assessed the model’s hypotheses by analyzing 249 marine mollusk shells dated individually using 14C-calibrated amino acid racemization methods. As postulated by the model, the temporal resolution of the fossil record, frequency of depositional events, and net accumulation rates decreased upward concomitantly through the transgressive systems tract reaching minima in the condensed section (maximum flooding zone). The reverse trend, with increasingly frequent, thicker, and less time-averaged beds, was observed throughout the overlying highstand systems tract. The results quantify the postulated sequence stratigraphic asymmetry in temporal resolution of the fossil and sedimentary records.
The sequence stratigraphic model (SSM) has generated transformative hypotheses applicable to sedimentary basin analysis, petroleum exploration, paleoclimatology, and paleobiology (e.g., Brett, 1995; Mansor et al., 1999; Miall, 2000; Nordt et al., 2007; Patzkowsky and Holland, 2012). The SSM applies to marine and continental deposystems and produces stratigraphic interpretations in terms of changes in accommodation, which occur over multiple time scales and link to external drivers such as climate-induced sea-level changes (e.g., Miall, 2010).
In the marine to coastal realms, the interplay between sediment supply and accommodation controls the internal architecture of sequence stratigraphic units and the landward-seaward shifts of the shoreline. All variants of the SSM model (Catuneanu et al., 2009) postulate that, along depositional dip sections, maximum stratigraphic condensation (condensed section [CS]) is associated with maximal rates of base-level rise, marking the turnaround between landward-stepping and basinward-stepping parasequence packages. Thus, the resolution of fossil and sedimentary records varies predictably: transgressive deposits (transgressive systems tract [TST]) are characterized by upward increase in stratigraphic condensation, whereas the overlying regressive deposits (highstand systems tract [HST] or equivalent units; Catuneanu et al., 2009) show the opposite trend. For TST-CS-HST successions from siliciclastic coastal to shelf depositional systems, the SSM model can be translated into three hypotheses regarding the resolution of paleontological and sedimentary records: (1) time averaging (age mixing within horizons) should increase upward through the TST, peak (i.e., produce the lowest depositional resolution) within the CS, and then decrease to a minimum in the HST; (2) the duration of diastems separating adjacent beds should increase upward through the TST and then decrease, with the most numerous depositional events recorded in the HST; and (3) net accumulation rates should decrease through the TST toward CS and increase throughout the HST.
These hypotheses have been confirmed qualitatively for various depositional settings. In open shelf environments, abundance and maturity of glaucony (sensuAmorosi, 1995)—indicators of stratigraphic condensation (Odin and Matter, 1981)—increase upward in the TST and then decrease rapidly within the HST (Amorosi, 1995). Similarly, fossil accumulations, forming preferentially when sedimentation is reduced (e.g., Kidwell, 1986; Brett, 1995; Abbott and Carter, 1997), concentrate within the CS, at flooding surfaces (parasequence boundaries), or along transgressive ravinement surfaces (Nummedal and Swift, 1987).
Thanks to advances in amino acid dating methods (Kaufman and Manley, 1998), a direct quantitative test of these hypotheses—a geochronological assessment of centennial to millennial changes in (1) depositional resolution, (2) frequency of depositional events, and (3) net accumulation rates—is now possible. Here, individually dated valves of marine mollusks (see Appendix DR1 in the GSA Data Repository1), sampled across the transgressive-regressive Holocene succession of the Po Plain (Italy; Fig. 1), were used to develop a quantitative test of the three SSM hypotheses stated above.
GEOLOGICAL SETTING, MATERIALS, AND METHODS
Over the past 40 yr, the SSM has been refined empirically for various depositional settings, with special emphasis on Quaternary systems as realistic analogs for the older geological record (e.g., Pillans et al., 1998; Blum and Törnqvist, 2000). The latest Quaternary stratigraphic architecture of the Po Basin (Fig. 1) has been interpreted as a product of cyclic deposition falling within the 100 k.y. Milankovitch band (Amorosi et al., 1999). The currently forming depositional sequence includes a Last Glacial Maximum lowstand systems tract (LST) fluvial succession overlain by a Holocene post-glacial TST-HST coastal to shallow-marine succession: a wedge-shaped body of the deepening-upward TST followed by the shallowing-upward HST (Fig. 1; Fig. DR1 in the Data Repository). The model is time constrained by historical data (archaeological sites, historic maps) and 14C dates on peat (Fig. 1; Table DR2 in the Data Repository) and its bathymetric inferences were confirmed by multivariate analyses of mollusk assemblages (Scarponi and Kowalewski, 2004).
Shells of two abundant bivalves, Lentidium mediterraneum and Varicorbula gibba (closely related taxa with comparable body size, ecology, valve thickness, and shell microstructure; Appendix DR2), were dated using 14C-calibrated amino acid racemization (AAR) ratios (Appendix DR1). AAR geochronology is based on the time-dependent conversion (racemization) of L to D amino acid enantiomers (D/L value increases with time until equilibration). A total of 309 right valves belonging to the two studied taxa were selected indiscriminately from 17 horizons (Fig. 1; Appendix DR1). Each valve was cleaned with a 10 s sonic bath followed by acid leaching, and subsequently split into two halves. The posterior half of each valve was used for the amino acid analysis and the anterior half was retained for 14C and other future analyses. Shells were analyzed for their amino acid composition (mainly, aspartic acid [Asp] and glutamic acid [Glu]), according to standard procedure (Kaufman and Manley, 1998; Appendix DR2).
Because racemization rates vary among species and are also affected by other factors, AAR rates require independent 14C calibration for each of the analyzed taxa. To calibrate the AAR rates, eight anterior halves of L. mediterraneum and four anterior halves of V. gibba were dated using accelerator mass spectrometry 14C dating (Table DR1). Age calibration curves were derived separately for Glu and Asp, using the reduced major axis (RMA) linear regression (Kosnik et al., 2008). The calibration curves were additionally constrained with live-collected specimens used to estimate initial D/L values (e.g., for L. mediterraneum, 58 yr B.P.) (0 yr B.P. = A.D. 1950). See Appendix DR2 for calibration equations and details.
The 14C calibrations for V. gibba and L. mediterraneum demonstrate that D/L values of individual shells are excellent predictors of their 14C ages, with dating precision of 0.1–0.3 k.y. After discarding 60 shells with aberrant amino acid signatures, 249 specimens from 17 samples were retained for final analyses (Fig. 2; Appendix DR1).
HYPOTHESIS 1: DEPOSITIONAL RESOLUTION
For each horizon, depositional resolution was estimated using age-frequency distributions of dated shells, with time averaging measured as inter-quartile range (IQR) (Kidwell et al., 2005; Yanes et al., 2007). Other dispersion metrics yielded comparable estimates (Table DR3). Within the TST, IQR increased upward (Figs. 2 and 3A) indicating that within-bed depositional resolution decreased with sea-level rise. This trend culminated in the lower part of the CS, where highest IQR values (i.e., the lowest depositional resolution) for the entire cycle were observed. In the distal CS (Fig. 1), time averaging was severe enough to produce ecological condensation: older shells of the shallower-water bivalve L. mediterraneum were mixed together with the younger shells of the deeper-water bivalve V. gibba (sample 6g in Fig. 2). The HST succession showed a reverse trend: IQR decreased upward (Figs. 2 and 3A).
The magnitude of the TST-HST trend in depositional resolution is striking (Figs. 2 and 3A; Table DR3). The IQR values for the lowermost TST lagoon samples (Fig. 1) were <130 yr and increased upward, reaching 680 yr at the base of the upper TST (wave ravinement surface, WRS) and 2560 yr in the CS. In the overlying HST, age mixing decreased to 530 yr in the initial phases of the highstand and dropped further through the HST, decreasing to <50 yr in the currently forming part of the succession. Whereas the high resolution of the lower TST may reflect the “virgin” effect, in which newly submerged land lacks older shell material needed to develop time-averaged horizons (Craig and Oertel, 1966; Flessa and Kowalewski, 1994), the high resolution of the HST reflects high net sedimentation rates, with older shells buried quickly in rapidly prograding-aggrading systems. Overall, the mean IQR is 2.3 times higher for TST than HST (or 3.3 times higher when data from active deposits are included; Table DR3).
HYPOTHESIS 2: FREQUENCY OF DEPOSITIONAL EVENTS
The frequency of bed-forming events (FE) is inversely related to duration of diastems separating those events. Here, 10 k.y. is used as a reference time unit; if a diastem underlying a given horizon represents ∼2 k.y., then FE = 10 k.y./2 k.y. = 5 beds per 10 k.y. (i.e., if the overall depositional conditions remained constant for 10 k.y., then 5 depositional events would be expected to be preserved as distinct beds). If the next horizon is preceded by a 4-k.y.-long diastem, FE would decrease to 2.5. Because fossils contained within a given horizon are expected to be a cumulative mixture from the time represented by the preceding diastem (e.g., Kowalewski and Bambach, 2003), the range of shell ages in a given horizon should approximate the diastem duration. To minimize the effect of outliers and reduce sample-size effects, diastem duration was estimated using the 10th and 90th percentiles of the age-frequency distribution. Whereas the 10th-to-90th metric is expected to correlate with the inter-quartile range (Hypothesis 1), it is more effective at capturing tails of age distributions (Table DR3). Also, independent FE proxies yielded consistent results (Table DR4).
The FE estimates (Fig. 3B; Table DR3) parallel the qualitative predictions of the regional SSM model: depositional events appear to have been relatively frequent during the initial (lower TST) phase of the cycle (FE = 38 [sample 4a] and FE = 35 [sample 4b]). A decrease in event frequency is observed upward toward the CS, with FE estimates dropping by more than one order of magnitude (lower CS, FE = 2). The trend reverses in the HST, with FE increasing upward, eventually reaching values an order of magnitude higher than recorded for CS (FE = 11 [sample 2i] and FE > 131 for currently active depositional surfaces; see caveats below).
These results support Hypothesis 2: rapid transgressions generate accommodation, but this is counterbalanced by erosion and/or reworking associated with shoreline transgression and low sedimentation rates in offshore settings. Also, as the SSM predicts, the FE trend is inversely correlated with depositional resolution: when depositional events become less frequent, temporal mixing increases (Figs. 3A and 3B). Overall, the average FE index (Table DR3) is almost two times lower for the TST than the HST, and about six times lower if present-day HST samples are included in the analysis (see caveats below).
HYPOTHESIS 3: NET ACCUMULATION RATE
Net accumulation rate (NAR) was computed as the length of the analyzed interval of the core (in m) divided by the length of the time (in k.y.) bracketed by the time markers defining the boundaries of that interval (Fig. 3C; Table DR2). Using 14C dates of basal peats and organic-rich clay horizons (Fig. 1), NAR estimates can be obtained not only for marine deposits evaluated above, but also for nonmarine deposits of both the LST and the lower TST (Figs. 1 and 3C). For these low-accommodation settings, NAR rates are low: 0.2–0.7 m/k.y. In contrast, NAR estimates for the more distal (marginal marine) parts of the lower TST succession range from 1.4 to 2.5 m/k.y. This trend affirms the SSM model: during rapid transgressions, coastal areas are expected to trap hinterland-derived sediments. Concurrently, accommodation should be created in marine settings. Consequently, NAR estimates should be lower for the coastal plain TST when compared with marginal marine (lagoon) successions of the lower TST.
NAR estimates are difficult to quantify for the upper TST given that shells are extensively time averaged (although age mixing, on its own, implies low sedimentation rates). In the HST, the NAR rates shift upward toward much higher values in all stratigraphic levels and localities (mean = 10.2 m/k.y.). A variation in NAR values observed across the sites (from 4.4 to 19.6 m/k.y.) may reflect differential progradation trends due to variation in sediment supply associated with episodes of delta-lobe abandonment and delta switching. Despite uncertain estimates for the upper TST and CS, the overall NAR trend agrees with qualitative prediction of the SSM: low net rates for alluvial sections of the LST, increased net accumulation rates in marginal marine parts of the lower TST, lowest rates in the upper TST/CS (as implied indirectly by other lines of evidence above), and highest, upward-increasing rates in the HST (Fig. 3).
At face value, the results (Figs. 2 and 3) demonstrate quantitatively the theoretically predicted asymmetry of glacio-eustastic cycles: the temporal resolution, the frequency of depositional events preserved, and net accumulation rates appear to have decreased throughout the TST and reached their minima in the CS, whereas the HST shows a reverse trend, with increasingly numerous, thicker, and less time-averaged beds. However, several confounding factors need to be evaluated.
The trends may simply reflect Walther’s Law (Middleton, 1973): if depositional environments with different sedimentation rates are restricted to a given systems tract, the observed changes in rates may reflect facies shifts. However, because cores cover a substantial segment of the basin’s depositional profile, comparable facies occur in different systems tracts (Fig. 1). For example, transgressive-shoreface sand and offshore-transition mud represent depositional settings comparable (in terms of sediment type, water depth, and distance from the shoreline) to highstand beach-ridge and prodelta facies respectively. Yet, they differ in terms of time-averaging and FE estimates (e.g., compare V. gibba samples 6g versus 2i in Figs. 2, 3A, and 3B), indicating that the observed trends are not a product of Walther’s Law.
The trends may have been driven by changes in subsidence rates and accommodation. Natural subsidence rates appear narrowly constrained (0.8–1.0 mm/yr) for the late Quaternary of the Po Plain (Ferranti et al., 2006), and thus cannot account for the observed trends. However, anthropogenic activities (land reclamation or water pumping) and natural processes (peat autocompaction) may have been important drivers of locally high and spatially variable subsidence rates observed for the currently forming HST successions. This problem primarily affects marsh/lagoon environments from the uppermost horizons (Cremonini et al., 2008). However, NAR values based on the affected units were excluded from the above analyses. NAR estimates may be additionally biased because sediment accumulation rates tend to scale negatively with the time span of analysis (Sadler, 1981). The highest NAR estimates were recorded for homogeneous sandy sediments (localities 4 and 5) representing short time spans (≤0.6 k.y.; Fig. 3C). However, these biases can be minimized by averaging NAR estimates within systems tracts to make time spans more comparable; in doing so, the NAR estimate for the HST is five times higher than that obtained for the lower TST (Table DR2).
The trends may be due to time-variant or space-variant errors in amino acid dating. However, calibration models yielded standard errors ≤0.3 k.y. (see Appendix DR2) and cannot account for the observed >1 k.y. shifts in inter-quartile ranges (Figs. 2 and 3A). Second, it is unlikely that the local thermal history, which could influence racemization rates, varied substantially in the area. Finally, the dating uncertainty of AAR geochronology generally increases with increasing specimen age as the racemization reaction progresses toward equilibrium. Thus, if the results reflected time-dependent dating errors, a continuous decrease in apparent time averaging would be expected toward the present. This is not the case (Fig. 3A): dating errors cannot explain either the magnitude or direction of the observed patterns.
In summary, the sequence of samples of 249 individually dated shells and associated data reveal a strong asymmetry in the resolution of paleontological and sedimentary records of the Holocene transgressive-regressive succession of the Po Plain. The trends in time averaging, number of depositional events, and net accumulation rates parallel each other, display dramatic (>10-fold) shifts in magnitude, track the systems tracts, and cannot be explained by obvious biasing factors that could have influenced the observed quantitative patterns. The results provide direct quantitative support for the predictions of the sequence stratigraphic model and quantify changes in the resolution of the fossil and sedimentary records as a function of base-level changes. These changes not only may produce distinct sequence stratigraphic architectures, but also affect the quality, quantity, and the temporal nature of paleontological and sedimentological data in a quantitatively verifiable way.
This project was supported by the National Science Foundation grants EAR-0920075 (Virginia Tech) and EAR-0929415 (Northern Arizona University), and the Marco Polo Research Fellowship (University of Bologna, Italy). We thank Ken Eriksson, Austin Hendy, Fred Read, Jackie Wittmer, and Brian Romans for useful comments on earlier drafts of this report. Reviews by Alan Beu, Carlton Brett, and an anonymous reviewer improved notably the quality and clarity of this report.