Scarponi And Kowaleski 2004

Stratigraphic paleoecology: Bathymetric signatures and sequence overprint of mollusk associations from upper Quaternary sequences of the Po Plain, Italy Daniele Scarponi* Department of Earth Sciences, University of Bologna, via Zamboni 67, Bologna 40126, Italy Michal Kowalewski* Department of Geosciences, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061, USA ABSTRACT Upper Quaternary sequences of the Po Plain (Italy) were used to assess the informative strength and sequence-stratigraphic overprint of quantitative paleoecological patterns. Three densely sampled cores (89 samples, 98 genera, 23,280 specimens), dominated by extant mollusk species with known environmental distributions, were analyzed with detrended correspondence analysis (DCA). The DCA scores, calibrated using extant genera, provided outstanding estimates of bathymetry (63 m) and related environmental parameters. Depth-related successions of mollusk associations delineated by using DCA were consistent with independent sequence-stratigraphic interpretations and yielded insights inaccessible via routine techniques (e.g., depth estimates for maximum flooding surfaces). The DCA ordination demonstrates the severity of the sequence-stratigraphic overprint: samples are highly uniform taxonomically during late transgressive systems tracts and highly variable during the following highstand systems tracts. When analyzed across comparable systems tracts, similar species associations repeat during the last and current interglacial cycles, suggesting that Po Plain mollusk associations have remained remarkably stable over the past 125 k.y. The results are consistent with the bathymetric interpretation of the DC axis 1 postulated previously for the Paleozoic fossil record, demonstrate the sequence-stratigraphic overprint of paleoecological patterns predicted by computer modeling, and illustrate the utility of quantitative paleoecological patterns in augmenting sequence-stratigraphic interpretations. Keywords: paleoecology, sequence stratigraphy, Quaternary, mollusks, Po Plain, Italy. INTRODUCTION There is a growing realization that the joint consideration of sequence-stratigraphic and paleontological patterns can yield valuable, and otherwise inaccessible, insights (e.g., Holland, 1995; Brett, 1998). First, because of the strong sequence overprint, many aspects of the fossil record should be explored within their stratigraphic context (e.g., Kidwell, 1993; Holland, 1995). Conversely, fossils archive high-resolution environmental signatures that can be quantified by applying multivariate methods and then used to augment stratigraphic interpretations (e.g., Holland et al., 2001). We show that the merger of sequence stratigraphy and quantitative paleoecology can yield even more powerful insights when applied to the late Cenozoic fossil record, which is dominated by extant species, and can be calibrated against direct bathymetric estimates derived from modern marine environments. We focus on Quaternary sequences of the Po Plain (Italy), a suitable testing ground that offers (1) a well-understood high-resolution stratigraphic framework (Amorosi et al., 2004), (2) fossiliferous marine sequences ac*E-mails: Scarponi—[email protected]; Kowalewski—[email protected].

cessible for sampling, and (3) diverse mollusks with documented bathymetric distributions in the modern Mediterranean Sea. We assess whether patterns based on extant, ecologically understood mollusks provide viable quantitative estimates of bathymetry and related environmental parameters. If reliable, these estimates should aid us in refining stratigraphic interpretations and assessing

sequence-stratigraphic overprints that affect paleobiological patterns. Also, the fossil record dominated by extant species allows us to test the hypothesis—derived using longextinct Paleozoic organisms (e.g., Holland et al., 2001; Miller et al., 2001)—that multivariate ordination gradients of marine fossil samples are driven primarily by bathymetry-related factors. The rich mollusk associations of the Po Plain should allow us to explore ecological dynamics of marginal-marine settings during high-frequency sea-level fluctuations that shaped the Quaternary history of many coastal areas of the world. GEOLOGIC SETTING, MATERIALS, AND METHODS The Po Plain (Fig. 1) is a large (;38,000 km2) perisutural basin that contains a thick (up to 1000 m) succession of Pliocene–Quaternary sediments (Pieri and Groppi, 1981). Two major Quaternary depositional cycles (Qm and Qc) are identifiable in the region (e.g., Ori, 1993). We focus on the upper cycle (Qm), which in the eastern part of Po Plain consists of the alternating marine and continental deposits of middle Pleistocene to Holocene age. The sequence-stratigraphic framework, based on multiple 10 cm cores (e.g., Amorosi et al., 2003, 2004; Marchesini et al., 2000) (Fig. 1), consists of a vertical stacking pattern of alternating marine and continental units,

Figure 1. Study area map (left inset) and geologic cross section (right) of upper Quaternary deposits of Po Plain (simplified after Amorosi et al., 2004). TST—transgressive systems tract; HST—highstand systems tract; s.l.—sea level; SB—sequence boundary.

q 2004 Geological Society of America. For permission to copy, contact Copyright Permissions, GSA, or [email protected]. Geology; November 2004; v. 32; no. 11; p. 989–992; doi: 10.1130/G20808.1; 5 figures; Data Repository item 2004160.

989

suggestive of glacio-eustatic cycles. The succession spans the past 150 k.y. and includes two transgressive-regressive shallow-marine intervals separated by a package of alluvialplain sediments deposited during the last glacial interval. The two marine units display nearly identical facies architecture: the landward-shifting brackish to shallow-marine retrogradational gray to brown silty clays (early transgressive systems tract), overlain by fully marine fine to coarse gray sands with numerous shell beds (late transgressive systems tract), and capped with the progradational deltaic interbedded gray to blue clays and sands (highstand systems tract). The two successions represent the late Pleistocene (oxygen isotope substage 5e) and Holocene (stage 1) interglacial intervals, respectively. We sampled these units in cores at three localities aligned along a single transect (Fig. 1). We collected 89 bulk samples (;0.375 dm3 each), with vertical spacing of 4 m or less, from cores. The samples were dried (24 h at 45 8C), soaked in ;4% H2O2 (#4 h, depending on lithology), and wet sieved with 1 mm screens. All mollusks were identified to species level whenever possible (scarce nonmollusk fossils were excluded). Because .99% of bivalve shells were disarticulated, each bivalve count was treated as 0.5 specimens. The data include 23,280 specimens, 132 species, and 98 genera (see Appendix DR11). To suppress problems inherent to species-level interpretations (e.g., Kowalewski et al., 2002), all analyses were carried out at the genus level. All rare genera (,0.02% or ,2 samples) and small samples were removed. The final matrix included 60 samples, 52 genera, and 23,027 specimens (;99% data). Three exploratory multivariate techniques were applied to the resulting relativeabundance matrix: nonparametric multidimensional scaling (MDS), correspondence analysis (CA), and detrended correspondence analysis (DCA). In all cases, the multivariate space was successfully reduced to a few environmentally interpretable axes. We focus on DCA because this technique minimizes an arch effect (Hill and Gauch, 1980), and has been widely used in recent studies linking stratigraphic and paleoecological data. The other two methods yielded consistent, even if less effective, ordinations. Statistical analysis system (SAS) was used to perform CA, MDS, and supplementary tests and simulations. Pa1GSA Data Repository item 2004160, Appendix DR1, raw abundance data, Appendix DR2, bathymetric data, Appendix DR3, salinity and energy data, and Appendix DR4, detrended correspondence analysis profiles for five core intervals, is available online at www.geosociety.org/pubs/ft2004.htm, or on request from [email protected] or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301-9140, USA.

990

Figure 2. Detrended correspondence analysis ordination of genera (samples not plotted). Inset shows scatter plot of 24 genera common in cores with DC1 scores graphed against present-day bathymetric estimates obtained from ENEA Mollusk Database (see text). First two axes explain 78.7% variance in data (DC1 5 54.5%, DC2 5 24.2%).

leontological statistics (PAST: Hammer et al., 2004) was used to obtain DCA outputs. To minimize the distorting impact of very abundant genera, the values were log-transformed using the ‘‘downweight’’ option (Hammer et al., 2004). Multiple DCA outputs were obtained initially while varying analytical options, transformation types, and rare taxa and/ or sample-size cutoff values. All outputs were very similar, demonstrating the strength of the underlying multivariate patterns. ENVIRONMENTAL CALIBRATION OF THE DCA ORDINATION Mollusk genera are continuously distributed along axis 1 (DC1) of the ordination (Fig. 2). The present-day ecology of these genera suggests that DC1 reflects mainly bathymetry: deeper-water genera have low DC1 scores, intermediate-depth genera have intermediate scores, and shoreface to brackish genera have the highest scores. This interpretation can be tested further using independent present-day bathymetric estimates for extant genera recorded in the Italian Mollusk Census Database of the New Technologies Energy and Environment Agency (ENEA), which includes bathymetric data for many mollusks found along the present-day Italian coast. The preferred bathymetry of a genus was obtained for 24 genera (Appendix DR2; see footnote 1) that were abundant in our samples and provided continuous coverage along DC1. The bivariate analysis (Fig. 2, inset) indicates that DC1 scores are an excellent, linear predictor of the preferred depth for the 24 analyzed genera: r 5 0.91 (95% confidence in-

terval [CI]: 0.81–0.96), r2 5 0.82 (95% CI: 0.66–0.92), n 5 24, pparam , 0.001; pbootstrap 5 0.001. The 95% CI and pbootstrap values were derived by using a 1000 iteration procedure based on the approach of Diaconis and Efron (1983). The tight linear correlation supports the bathymetric interpretation and suggests that DC1 scores yield depth estimates to the nearest 3 m (root mean square error [RMSE] 5 3.1 m; 95% CI: 2.4–3.7 m). The ordination pattern is wedge shaped (Fig. 2). The genera with DC1 , 0.6 all have similar DC2 scores (1.2–2.5), and those with high DC1 scores have increasingly variable DC2 scores (0–4.7). The ecological survey of the genera suggests that DC2 reflects primarily salinity and energy. Genera with low DC2 scores (,1.5) inhabit high-energy shoreface settings, those with intermediate scores (1.5–2.3) occupy moderate- to low-energy settings, and those with high scores (2.3–4.6) represent lowenergy settings (Fig. 2). Genera with either low or high DC2 scores belong to salinity-tolerant taxa, whereas those with intermediate scores (1.7–2.7) are less tolerant, fully marine taxa (Appendix DR3; see footnote 1). In sum, the low DC2 scores (,1.5) represent predominantly upper-shoreface to shoreface settings, intermediate scores (1.5–3) represent lowershoreface to open-shelf settings, and the highest DC2 scores (.3) represent shallow, low-energy settings with variable salinity. When the data are downgraded from specimen counts to categorical ranks, to make our analysis comparable to previous studies (e.g., Holland et al., 2001), the results remain simGEOLOGY, November 2004

ilar in all aspects, indicating that the comparable environmental calibration of DCA can be achieved using low-resolution data.

Figure 3. Detrended correspondence analysis ordination of samples (genera not plotted). Main plot shows samples plotted by locality and time interval. Inset plot shows samples plotted by sequence-stratigraphic position. TST—transgressive systems tract; HST—highstand systems tract.

Figure 4. Stratigraphic patterns in DC1 and DC2 scores for Holocene transgressive-regressive cycle of locality 3 (Appendix DR4 [see footnote 1] gives plots of other sampled units). Sequencestratigraphic interpretation is based on Amorosi et al. (2004). DC1 scores provide bathymetric estimates consistent with those previous interpretations. DC2 scores augment environmental information. For example, samples 1 and 10 (labeled, counting upward) have comparable depth estimates (DC1) and similar sedimentology, but their DC2 scores suggest radically different salinity-energy regimes: lagoonal (sample 1) vs. upper shoreface (sample 10). Lithologic abbreviations: M—mud; sM—silty mud; mS—muddy sand; S—sand; C—conglomerate. Other abbreviations: TS—transgressive surface; TST—transgressive systems tract; RS—ravinement surface; MFS—maximum flooding surface; HST—highstand systems tract. GEOLOGY, November 2004

COUPLED STRATIGRAPHICPALEOECOLOGICAL APPLICATION OF THE DCA ORDINATION Thanks to the direct calibration of DCA, we can now augment stratigraphic and paleobiological interpretations of the Po Plain successions. First, we can estimate the bathymetric range of the sampled units. Similar to genera (Fig. 2), the samples form a wedge-shaped pattern along DC1 (Fig. 3), with samples ranging in depth from ;16 to ;0 m. The samples with low DC1 scores (,1.3) and invariant DC2 scores (1.5–2.2) represent more distal, fully marine sites (depth range ;10 to ;16 m). The more proximal samples (DC1 . 1.8; variable DC2 scores 0.0–3.9) represent shallow to coastal sites from a wide range of salinity and energy regimes (depth range ,10 m to ;0 m). Second, we can independently reexamine the regional sequence-stratigraphic scheme: late transgressive systems tract samples have middle to low DC1 scores and invariant DC2 scores, highstand systems tract samples vary widely in both DC1 and DC2 scores, and early transgressive systems tract samples have high DC1 and DC2 scores (Fig. 3, inset). This pattern is highly congruent with the stratigraphic interpretation of these successions: early transgressive systems tract samples record coastal, low-energy settings with fluctuating salinity, late transgressive systems tract samples represent most distal settings, and highstand systems tract samples record a broad range of environments reflecting a shallowing-upward trend. For all five sampled intervals, the stratigraphic plots of DC scores indicate a deepening-upward trend followed by a shallowing-upward trend (Fig. 4; for all plots see Appendix DR4 [see footnote 1]) delineating the expected sequence pattern. DC2 further augments the analysis by discriminating different environments situated at comparable depths (Fig. 4). Also, DC1 scores offer estimates of bathymetric ranges of systems tracts and depth estimates around the key stratigraphic surfaces. For example, the maximum flooding surface in Holocene strata of locality 3 (Fig. 4) is estimated to have been deposited at ;9 m (63 m). Given that this surface is located ;10 m below the current sea level, the estimate is compatible with a stable sea level and a lack of notable subsidence, as postulated by Amorosi et al. (2004). Third, we can quantify spatiotemporal depositional gradients. When DC1 scores of maximum flooding surfaces are superimposed (Fig. 5), the offsets reflect differences in maximum depth across localities or cycles. In the 991

ogica (coordinator F. Massari) and the Marco Polo Fellowship (Bologna University). We thank A. Amorosi and M.L. Colalongo for useful discussions about the Po Plain, L. Angeletti for help in sampling, and J.F. Read, R.A. Krause, C.E. Brett, and an anonymous reviewer for many helpful comments.

Figure 5. Comparison of DC1 estimated depths for five maximum flooding surfaces (plotted at level labeled MFS) recorded in five sampled core intervals.

Holocene, the offsets suggest a northward increase in water depth: locality 3, ;9 m; locality 2, ;13 m; and locality 1, ;14 m. The same pattern, but with higher depth values, is recognizable in the Pleistocene: locality 3, ;12 m; locality 1, ;16 m. This analysis (Fig. 5) reveals a subtle bathymetric gradient consistent with the northward location of the regional depocenter (note the location of the Po River Delta; Fig. 1) and suggests that this gradient remained unchanged over the past two interglacial cycles, with a slightly more extensive flooding of the Po Plain in the previous interglacial. Fourth, the calibrated DCA demonstrates a sequence overprint of the fossil record (Fig. 3, inset). Regardless of their stratigraphic and sampling locations, the late transgressive systems tract samples are much more uniform in their taxonomic composition (variances: DC1 5 0.19; DC2 5 0.04) than the subsequent highstand systems tract samples (DC1 5 0.97; DC2 5 0.37). Thus, if their sequencestratigraphic context were disregarded, the samples would suggest higher diversity levels (especially beta diversity) and higher temporal turnover rates for highstand systems tract segments of the sampled cores. Admittedly, this overprint may reflect real changes in environmental heterogeneity between coastal and open-shelf settings. Also, the observed pattern reflects the position of the sites relative to the depositional gradient. Thus, whereas the results reaffirm the importance of interpreting paleobiological patterns within their sequencestratigraphic context, the observed trends should not be extrapolated literally to other settings. Finally, the remarkable stability of late transgressive systems tract samples, which re-

992

gardless of locality and cycle plot in one tight area of the ordination (Fig. 3), suggests a high stability in faunal composition within and across successive late transgressive systems tract intervals during the past 125 k.y. However, our data are inadequate to assess whether this pattern is environmental (an increased environmental homogeneity away from the coast), ecological (synchronized shifts in species distributions), or taphonomic (an increased homogenization of late transgressive systems tract samples due to higher levels of time averaging). FINAL REMARKS This study demonstrates the utility of core samples in data-intensive multivariate analyses of fossils and offers an empirical support for the stratigraphic and paleobiological applicability of DCA ordinations. The DCA ordination calibrated using present-day ecological data can provide multiple quantitative insights, including numerical bathymetric estimates of key stratigraphic surfaces, delineation of depositional sequences and sea-level changes, assessment of depositional gradients though time and space, evaluation of sequence-stratigraphic overprints of the fossil record, and recognition of paleoecological recurrence patterns across comparable systems tracts. The results are consistent with the bathymetric interpretation of DC1 postulated for the Paleozoic fossil record (e.g., Holland et al., 2001) and illustrate the sequencestratigraphic overprint of paleobiological patterns predicted by computer modeling (Holland, 1995). ACKNOWLEDGMENTS The study was supported by Ministero dell’Universita` e della Ricerca Scientifica e Tecnol-

REFERENCES CITED Amorosi, A., Centineo, M.C., Colalongo, M.L., Pasini, G., Sarti, G., and Vaiani, S.C., 2003, Facies architecture and latest Pleistocene–Holocene depositional history of the Po Delta (Comacchio area), Italy: Journal of Geology, v. 111, p. 39–56. Amorosi, A., Colalongo, M.L., Fiorini, F., Fusco, F., Pasini, G., Vaiani, S.C., and Sarti, G., 2004, Palaeogeographic and palaeoclimatic evolution of the Po Plain from 150-ky core records: Global and Planetary Change, v. 40, p. 55–78. Brett, C.E., 1998, Sequence stratigraphy, paleoecology, and evolution: Biotic clues and responses to sea-level fluctuations: Palaios, v. 13, p. 241–262. Diaconis, P., and Efron, B., 1983, Computerintensive methods in statistics: Scientific American, v. 248, p. 116–130. Hammer, Ø., Harper, D.A.T., and Ryan, P.D., 2004, PAST—Palaeontological statistics, ver. 1.20: http://folk.uio.no/ohammer/past. Hill, M.O., and Gauch, H.G., 1980, Detrended correspondence analysis: An improved ordination technique: Vegetatio, v. 42, p. 47–58. Holland, S.M., 1995, The stratigraphic distribution of fossils: Paleobiology, v. 21, p. 92–109. Holland, S.M., Miller, A.I., Meyer, D.L., and Dattilo, B.F., 2001, The detection and importance of subtle biofacies within a single lithofacies: The Upper Ordovician Kope Formation of the Cincinnati, Ohio, region: Palaios, v. 16, p. 205–217. Kidwell, S.M., 1993, Taphonomic expressions of sedimentary hiatus: Field observations on bioclastic concentrations and sequence anatomy in low, moderate and high subsidence settings: Geologische Rundschau, v. 82, p. 189–202. Kowalewski, M., Gu¨rs, K., Nebelsick, J.H., Oschmann, W., Piller, W.E., and Hoffmeister, A.P., 2002, Multivariate hierarchical analyses of Miocene mollusk assemblages of Europe: Paleogeographic, paleoecological, and biostratigraphic implications: Geological Society of America Bulletin, v. 114, p. 239–256. Marchesini, L., Amorosi, A., Cibin, U., Zuffa, G.G., Spadafora, E., and Preti, D., 2000, Sand composition and sedimentary evolution of a late Quaternary depositional sequence, northwestern Adriatic Coast, Italy: Journal of Sedimentary Research, v. 70, p. 829–838. Miller, A.I., Holland, S.M., Meyer, D.L., and Dattilo, B.F., 2001, The use of faunal gradient analysis for high-resolution correlation and assessment of seafloor topography in the type Cincinnatian: Journal of Geology, v. 109, p. 603–613. Ori, G.G., 1993, Continental depositional systems of the Quaternary of the Po Plain (northern Italy): Sedimentary Geology, v. 83, p. 1–14. Pieri, M., and Groppi, G., 1981, Subsurface geological structure of the Po Plain, Italy: Roma, Consiglio Nazionale delle Ricerche Pubblicazioni, p. 1–23. Manuscript received 6 May 2004 Revised manuscript received 12 July 2004 Manuscript accepted 19 July 2004 Printed in USA

GEOLOGY, November 2004