Nannofossil carbonate fluxes during the Early Cretaceous: Phytoplankton response to nutrification episodes, atmospheric CO2, and anoxia



[1] Greenhouse episodes during the Valanginian and Aptian correlate with major perturbations in the C cycle and in marine ecosystems, carbonate crises, and widespread deposition of Corg-rich black shales. Quantitative analyses of nannofossil micrite were conducted on continuous pelagic sections from the Southern Alps (northern Italy), where high-resolution integrated stratigraphy allows precise dating of Early Cretaceous geological events. Rock-forming calcareous nannofloras were quantified in smear slides and thin sections to obtain relative and absolute abundances and paleofluxes that are interpreted as the response of calcareous phytoplankton to global changes in the ocean-atmosphere system. Increased rates of volcanism during the formation of Ontong Java and Manihiki Plateaus and the Paranà-Etendeka large igneous province (LIP) are proposed to have caused the geological responses associated with early Aptian oceanic anoxic event (OAE) 1a and the Valanginian event, respectively. Calcareous nannofloras reacted to the new conditions of higher pCO2 and fertility by drastically reducing calcification. The Valanginian event is marked by a 65% reduction in nannofossil paleofluxes that would correspond to a 2–3 times increase in pCO2 during formation of the Paranà-Endenteka LIP. A 90% reduction in nannofossil paleofluxes, which occurred in a 1.5 myr-long interval leading into OAE1a, is interpreted as the result of a 3–6 times increase in pCO2 produced by emplacement of the giant Ontong Java and Manihiki Plateaus. High pCO2 was balanced back by an accelerated biological pump during the Valanginian episode, but not during OAE1a, suggesting persisting high levels of pCO2 in the late Aptian and/or the inability of calcareous phytoplankton to absorb excess pCO2 above threshold values.

1. Introduction

[2] In the present oceans, coccolithophores are primary producers that convert dissolved carbon dioxide into organic matter and calcite, and consequently, calcareous nannoplankton are key organisms for understanding both the biological and the carbonate pumps. While investigations of sediment traps and calcareous oozes are addressed to understanding surface ocean processes of primary production and fluxes of biogenic and inorganic particles, ancient sedimentary successions can be decoded for characterization of paleobiological processes and paleofluxes.

[3] Coccolithophores dominate phytoplankton assemblages in the open ocean, whereas they are subordinate to siliceous (diatoms) and nonmineralizing (bacteria and dinoflagellates) phytoplankton in areas characterized by mesotrophic to eutrophic conditions. The understanding of how carbonate and siliceous primary production functions and interacts is fundamental for the interpretation of geochemical cycles of calcium and carbon dioxide. Coccolithophore photosynthesis and biocalcification affect the organic and inorganic carbon cycle as well as adsorption of atmospheric CO2 into the oceans. These biotic sinks for CO2 imply nutrification events and interactions of the carbon cycle with other biogeochemical cycles.

[4] Recent studies of the effects of atmospheric CO2 on rates of organic matter and calcite production in extant Emiliania huxleyi and Gephyrocapsa oceanica clearly indicate that increasing atmospheric CO2 induces reduced biocalcification [Riebesell et al., 2000; Zondervan et al., 2001]. Similar results were obtained for coral reefs [Gattuso et al., 1998; Gattuso and Buddemeier, 2000; Kleypas et al., 1999; Langdon et al., 2000; Leclercq et al., 2000] and planktonic foraminifera [Bijma et al., 1999; Barker and Elderfield, 2002], further indicating that biocalcification is strongly inhibited under high pCO2.

[5] Estimates of coccolith carbonate fluxes are crucial for quantitatively assessing the information preserved in the sediment record and for using coccolithophores as biotic proxies of paleoceanographic and climate change. Calcareous nannofossils (coccoliths and nannoliths) have been the principal contributors to pelagic micrites since the Jurassic. Because calcareous nannofossils constitute the bulk of calcareous oozes and micritic limestones, characterization of temporal variations of calcareous nannoplankton production and coccolith fluxes are based on counts of coccoliths per unit area of sediment traps [Honjo, 1980; Samtleben and Bickert, 1990; Steinmetz, 1991; Milliman, 1993; Beaufort and Heussner, 1999; Broerse et al., 2000a, 2000b; Young and Ziveri, 2000; Ziveri et al., 2000]. Estimates of coccolith carbonate fluxes are then obtained using the single taxon volume and its absolute abundance [Young and Ziveri, 2000].

[6] This kind of investigation has not been attempted on Cretaceous micrites, although a number of quantitative studies have provided insights into nannofossil paleoecology and a few indices have been recognized [Roth, 1981; Roth and Bowdler, 1981; Roth and Krumbach, 1986; Erba, 1986, 1992a, 1992b, 1994; Premoli Silva et al., 1989; Mutterlose, 1989, 1991; Mutterlose and Wise, 1990; Busson and Noël, 1991; Williams and Bralower, 1995; Burnett et al., 2000; Herrle, 2002]. Most investigations are based on relative abundances, calculated as percentages after counts of 300–350 specimens or total number of specimens in a fixed number of fields of view (equated to a unit area). Only recently, preparation techniques have been developed to obtain homogeneous dispersion of sediments in smear slides and calculate absolute abundances of nannofossil taxa per unit weight [Williams and Bralower, 1995; Geisen et al., 1999; Herrle, 2002].

[7] Applications of such techniques have shown that they are not suitable for limestones owing to their lithification. However, pelagic limestones and hard lithologies in general offer the opportunity to study nannofossils in thin sections [Watkins et al., 1995] and obtain absolute abundances per unit area [Erba, 1994; Channell et al., 2000]. Absolute quantification of nannofossil distribution has the potential for improving paleoceanographic reconstructions and modeling of climate/ocean interactions.

[8] The Cretaceous was a period of extreme paleoenvironmental conditions, possibly provoked by major tectonic and volcanic events. Under greenhouse conditions the hydrological cycle and weathering were altered, perhaps accelerating the introduction of nutrients, iron, and other biolimiting elements via runoff and eolian fluxes. A first episode of greenhouse climate occurred in the Valanginian as testified by a globally recorded C-isotope positive excursion, associated with a crisis of pelagic and neritic carbonates, and increased evolutionary rates in planktonic communities (fast diversification in planktonic foraminifers and accelerated speciation in calcareous nannoplankton, radiolarians, and diatoms) that are interpreted as a nutrification event, perhaps linked to the Paranà-Etendeka volcanism and increased spreading rates in the Indo-Australian breakup [Lini et al., 1992; Channell et al., 1993; Weissert et al., 1998; Bersezio et al., 2002; Erba et al., 2004]. In the Barremian/Aptian boundary time interval, formation of Ontong Java and Manihiki Plateaus and Nova Canton trough system produced excessive CO2 levels in the atmosphere [Larson and Erba, 1999] but also directly warmed up deep and intermediate waters, causing the rapid melting of massive quantities of clathrate [Opdyke et al., 1999; B. Opdyke et al., A history of methane hydrate release and formation gleaned from the detailed carbon isotope stratigraphy of the APTICORE, submitted to Nature, 2003, hereinafter referred to as Opdyke et al., submitted manuscript, 2003]. Marine biota were severely affected by such global changes in the ocean/atmosphere system. In surface waters, changes in temperature, increases in pCO2, as well as availability of nutrients and biolimiting metals (increased by volcanogenic upwelling, more efficient oceanic circulation, accelerated weathering and runoff, and higher fluxes of metals at hydrothermal plumes) must have influenced types, quantities, and rates of phytoplankton blooms and consequently the composition of biogenic sediments.

[9] During the Early Cretaceous, pelagic sedimentation in the Tethys was largely controlled by blooms of coccolithophores, with minor contribution by calpionellids and planktonic foraminifers. Radiolarians were also abundant, producing radiolarian arenites and cherts that are very often rhythmically distributed. The occurrence of Early Cretaceous diatoms is still poorly known (see discussion in the work of Larson and Erba [1999]), and certainly the siliceous phytoplankton was not relevant for lithogenesis. Dinoflagellates were common and widely distributed, but their preservation and abundance is strictly controlled by sedimentary facies and characteristics of bottom waters, especially oxygen content. Quantification of biocalcification in calcareous nannoplankton might provide insights into metabolic processes strictly controlled by environmental conditions. Therefore Cretaceous coccolith/nannolith carbonate fluxes have the potential of improving our understanding of ecosystems and global change in the past.

[10] The interpretation and modeling of ancient pelagic carbonates are biased by diagenesis that certainly modified the original micrite. Detailed investigation of nannofossil ultrastructure, abundance, diversity, as well as abundance of micarbs can be used to reconstruct type and degree of diagenetic modifications and, consequently, to select samples for paleoceanographic reconstructions, disregarding material where the original micrite is distorted or obscured by dissolution and/or overgrowth [Roth and Krumbach, 1986; Thierstein and Roth, 1991; Erba, 1992a; Erba et al., 1992].

[11] This paper is focused on nannofossil (coccolith/nannolith) fluxes in the Early Cretaceous, with particular emphasis on times of major changes across the Valanginian event and the early Aptian oceanic anoxic event (OAE) 1a. Both episodes correspond to a carbonate crisis documented in pelagic and shallow-water environments on a global scale [Erba, 1994; Weissert et al., 1998]. Quantification and modeling of such events will help unravel the anoxia versus productivity enigma still debated for Cretaceous OAEs and isotopic anomalies. Furthermore, a detailed picture of paleobiological processes might elucidate the role of different variables in the past oceans and better delineate geosphere-biosphere interactions. Positive and negative feedbacks and rates of biosphere reactions reconstructed from ancient sedimentary successions might be used to model and predict future global change.

2. Materials and Methods

2.1. Studied Sections

[12] We analyzed the Polaveno section and the Cismon Apticore located in the Southern Alps (northern Italy) (Figure 1), where the Lower Cretaceous pelagic succession is represented by the Maiolica Formation consisting of thin-bedded calcilutites, with chert and black shale as minor lithologies. The Polaveno section is a continuous outcrop of limestones with radiolarian-rich layers, chert nodules, lenses and beds, and black shales [Erba and Quadrio, 1987; Bersezio et al., 2002]. The detailed biostratigraphy, magnetostratigraphy, and chemostratigraphy [Channell and Erba, 1992; Lini et al., 1992; Channell et al., 1993; 1995a; Bersezio et al., 2002] have been used for refinement of chronostratigraphy and timescales [Channell et al., 1995a, 1995b] and paleoceanographic reconstructions [Lini et al., 1992; Channell et al., 1993; Weissert et al., 1998; Bersezio et al., 2002].

Figure 1.

Paleogeographic reconstruction of the Southern Alps in the Early Cretaceous (modified after Weissert and Lini [1991]). The studied sections are indicated as 1 = Polaveno (eastern Lombardian basin) and 2 = Cismon (Belluno Basin).

[13] The Cismon Apticore is a reference Barremian-Aptian pelagic section owing to its completeness and relatively high sedimentation rate and the good time control based on integrated bio-magneto-chemo-cyclostratigraphy [Channell et al., 1979; Weissert et al., 1985; Bralower, 1987; Weissert, 1989; Weissert and Lini, 1991; Herbert, 1992; Erba, 1994; Menegatti et al., 1998; Erba et al., 1999; Channell et al., 2000; Opdyke et al., submitted manuscript, 2003]. The Hauterivian-Barremian interval at Cismon consists of dominant limestones with intercalated black shales; radiolarian-rich arenites are frequent. In the lower Aptian, a major lithologic change from limestone to black shale and marlstone corresponds to the Selli Level representing the OAE1a in the Tethys Ocean. The upper Aptian consists of marly limestone with radiolarian-rich layers and chert [Erba and Larson, 1998].

2.2. Sample Preparation and Nannofossil Counts

[14] At Polaveno and Cismon, calcareous nannofossils were quantified both in smear slides and thin sections by light polarizing microscope at 1250 times magnification. Smear slides were prepared for limestones, marlstones, and black shales, after powdering a small quantity of rock, without centrifuging and/or ultrasonic cleaning to retain the original composition. Permanent slides were mounted with Norland optical adhesive. Thin sections were prepared only for hard lithologies. At Polaveno, a total of 372 samples (236 smear slides and 136 thin sections) were studied, with an average sampling rate of 70 cm for limestones, corresponding to approximately 1 sample every ∼40 ± 5 kyr, owing to variable sedimentation rates. In the Cismon core, 641 smear slides were prepared every 10 cm, corresponding to approximately 1 sample every ∼10 ± 2 kyr. A total of 293 thin sections were prepared from samples collected in hard lithologies every 40 cm, corresponding to approximately 1 sample every ∼40 ± 5 kyr, owing to variable sedimentation rates. In smear slides, nannofossil assemblages were quantified by counting at least 300 specimens and then percentages of single taxon with respect to the total nannoflora (relative abundance) were calculated. Thin sections were thinned to an average thickness of 7 μm in order to have an adequate view of nannofloras. Absolute abundances were obtained by counting all nannofossil specimens in 1 mm2 of thin section. For each sample (smear slides and thin sections), counts were repeated 3 times and the standard deviation averages +1.5% at the 95% confidence level.

2.3. Nannofossil Paleofluxes

[15] The Polaveno and Cismon pelagic successions are dated with a high-resolution biostratigrpahy, magnetostratigraphy, and chemostratigraphy that allows a precise estimate of sedimentation rates of the investigated intervals. Nannofossil paleofluxes have been calculated for the micrite-constituting nannofossils, taking into account absolute abundances of the most common taxa, volume/mass of individual taxon, unit area (1 mm2), and unit time (1 year). The latter was derived from sedimentation rates estimated for individual magnetic polarity chrons [Channell et al., 2000]. Depending on sedimentation rates, the thickness of thin sections (= 7 μm) represents 7–14 months, and we normalized all absolute abundances to 1 year.

[16] Previous calculation of volume and mass of Mesozoic nannofossils was documented by Williams and Bralower [1995], who estimated an “averaged Cretaceous nannofossil” with a volume of 14 μm3. For the present study we adopt the results of the morphometric analyses conducted by Tremolada and Young [2002] on the most common Early Cretaceous taxa in well-preserved material recovered at sections and oceanic sites with high-resolution integrated stratigraphy. For size determination, pictures of nannofossils were captured using a digital image system [Young et al., 1996] at 1600 and 1250 times magnification; then the dimensions were measured using the software National Institutes of Health (NIH)-Image adapted for nannoplankton analyses. For each taxon, at least 250 specimens were investigated to give the most accurate size estimate. In addition, measurements were carried out by scanning electron microscope (SEM) on selected taxa to compare the results obtained by three-dimensional (3-D) and 2-D observations. Details of image analyses are given in the work of Tremolada and Young [2002]. Volume and mass estimates of selected nannofossil taxa used in this study are reported in Table 1.

Table 1. Volume (μm3) and Mass (pg CaCO3) of Common Taxa in the Early Cretaceousa
Volume, μm3Mass, pgSpeciesRatio, Taxon/N. Steinmannii
  • a

    After Tremolada and Young [2002]. The right-hand column represents the ratio between a single specimen of each species and a single specimen of Nannoconus steinmannii. This ratio represents the number of specimens of individual species required to equal the calcite in one specimen of N. steinmannii.

Oligotrophic Species
1245.93363.9N. steinmannii1
830.22241.4N. globulus1.5
546.51794.0N. bucheri1.8
340.7920N. truittii3.6
Uncertain Paleoecological Affinity
323872A. infr. larsonii3.8
270729R. ter. youngii4.6
157424M. obtusus8
87234P. embergeri14.3
38103M. hoschulzii32.8
35.496A. infracretacea35.2
24.366R. terebrodentarius51.3
2157W. barnesae59.3
1439R. asper89
9.325Z. diplogrammus134
5.414.6Z. elegans230.7
Mesotrophic and Eutrophic Species
5.113.77D. lehmanii244.3
410.8D. rotatorius311
3.28.6B. constans390
0.82.2Z. erectus1557

[17] When nannoconids are overwhelming the assemblages, only large-sized nannoliths such as pentaliths, Assipetra, and Rucinolithus are unequivocally detectable and quantifiable in thin sections. In fact, because nannoconids are large and thick, they tend to cover coccoliths that are present. Therefore abundance of coccoliths is typically underestimated in thin section of Lower Cretaceous limestones because they constitute a small fraction of the micrite dominated by nannoconids [e.g., Erba, 1994]. Conversely, nannoconids are underestimated in smear slides, essentially owing to mechanical breakage during powdering. In fact, their ultrastructure favors disintegration of single crystals [Noël and Melguen, 1978]. In order to have a more realistic estimate of coccolith distribution, although probably overestimated, we calculated coccolith absolute abundances on the basis of the proportion of single taxon versus nannoconids in smear slides and absolute abundance of nannoconids in thin section as follows:

equation image


AAtaxon 1

absolute abundance of taxon 1;


percentage of taxon 1 in smear slide;


percentage of nannoconids in smear slide;


absolute abundance of nannoconids in thin section.

3. Results

[18] In the Tethys Ocean, Lower Cretaceous Maiolica limestones mainly consist of calcareous nannofossils, with only minor contribution by calpionellids and foraminifers. Nannofloras are usually dominated by Watznaueria (especially W. barnesae) and nannoconids that together can represent up to 90% of assemblages. Abundance of nannoconids in smear slide is typically underestimated owing to mechanical breakage of specimens during rock powdering. This become obvious when quantitative data obtained in smear slide are compared with abundance derived from thin section and SEM analyses [Erba and Quadrio, 1987; Erba, 1994]. Therefore dominance of W. barnesae in most samples from both the Polaveno section and the Cismon core is interpreted as an artifact of sample preparation. The only other genera showing relative abundances higher than 5% are Diazomatolithus,Assipetra,Rucinolithus, and the pentalith group (Micrantholithus and Braarudosphera).

[19] In both the Polaveno section and the Cismon Apticore, absolute abundances of nannofossils show changes that are only partly similar to those documented in relative abundances (compare Figures 25). Although the nannofloral composition and common taxa are the same, thin section investigation clearly demonstrates the dominance of nannoconids, which are artificially reduced during smear slide preparation. As previously suggested, the micrite of Maiolica limestone can be regarded as a “nannoconite” [Erba, 1994].

Figure 2.

Percentages of most abundant nannofossils in the Polaveno section plotted against biomagnetostratigraphy [Channell et al., 1993, 1995b; Bersezio et al., 2002] and C isotope stratigraphy [Lini et al., 1992; Lini, 1994]. The base of the “nannoconid decline” is based at the level where the decrease in relative abundance of nannoconids correlates with a major increase in relative abundance of W. barnesae.

Figure 3.

Absolute abundances of most abundant nannofossils in the Polaveno section plotted against biomagnetostratigraphy [Channell et al., 1993, 1995b; Bersezio et al., 2002] and C isotope stratigraphy [Lini et al., 1992; Lini, 1994]. Absolute abundances of W. barnesae, Diazomatolithus spp., and pentaliths were calculated taking into account the proportion of single taxon versus abundance of nannoconids (see text for details). Lithologic legend and “nannoconid decline” as in Figure 2.

Figure 4.

Percentages of most abundant nannofossils in the Cismon core plotted against bio-magneto-chemostratigraphy [Erba et al., 1999; Channell et al., 2000].

Figure 5.

Absolute abundances of most abundant nannofossils in the Cismon core plotted against bio-magneto-chemostratigraphy [Erba et al., 1999; Channell et al., 2000]. Absolute abundances of W. barnesae, Assipetra, Rucinolithus, and pentaliths were calculated taking into account the proportion of single taxon versus abundance of nannoconids (see text for details). Lithologic legend as in Figure 4.

[20] In both the Polaveno and Cismon sections, preservation is moderate to poor but comparable in all samples. In particular, thin section analysis shows that the type and degree of diagenesis is similar through the studied intervals. Consequently, although the original micrite composition was certainly altered during burial and lithification, variations in nannofossil absolute abundances cannot result from differential dissolution and/or overgrowth of taxa with variable sensitivity to diagenesis. This is certainly true for the rock-forming nannoconids that are considered diagenesis resistant forms (see discussion in the work of Erba [1994]).

3.1. Valanginian-Hauterivian Interval at Polaveno

[21] A decrease in nannoconid relative abundance (from 25–50% to 0–10%) and a coeval increase in abundance of W. barnesae (from 50–60% to >80%) correlate with magnetic chron CM12 (Figure 2). The interval corresponding to magnetic chrons CM12 through CM10 is characterized by very low percentages of nannoconids and corresponds to the previously documented “nannoconid decline” [Channell et al., 1993; Erba, 1994; Weissert et al., 1998; Bersezio et al., 2002]. In this interval the Diazomatolithus group (D. lehmanii and D. subbeticus) becomes common, with percentages up to 14%. Assipetra infracretacea is abundant in the black shales deposited during the δ13Ccarb positive excursion. As reported by Bersezio et al. [2002], the interval immediately preceding the isotopic excursion is also characterized by an abundance peak of pentaliths (Figure 2). In this interval, percentages as high as 8% were quantified; Micrantholithus hoschulzii is the dominant form, with minor contribution by Braarudosphaera regularis and Micrantholithus obtusus.

[22] The interval following the C isotopic excursion (highest δ13Ccarb values in magnetic chron CM11n) contains nannofossil assemblages with unchanged relative abundance of W. barnesae, narrow- and wide-canal nannoconids and virtually absent pentaliths. Only the Diazomatolithus group gradually decreases in relative abundance, after reaching maximum percentages at the climax of the δ13C excursion and returning to pre-excursion values by magnetic chron CM10 time (Figure 2).

[23] Thin section analyses revealed that through the upper Berriasian/lower Hauterivian, micrite consists of nannoconids dominated by the narrow-canal forms (Figure 3). In the upper Berriasian/lowermost Valanginian interval (magnetic chrons CM16 through CM13), an average of 3500 specimens mm−2 of W. barnesae was calculated. In this interval, narrow- plus wide-canal nannoconids show absolute abundances between 3500 and 4000 specimens mm−2. In magnetic chron CM12, a first decrease in nannoconid absolute abundance (to approximately 2000 specimens mm−2) is followed by a more severe decrease, characterized by values of 700–1000 specimens mm−2 in the interval corresponding to the upper part of magnetic chron CM12 to the lower part of magnetic chron CM10N. The nannoconid decrease in abundance, named the “nannoconid decline,” is paralleled by a major increase of W. barnesae reaching values of 8000 specimens mm−2.

[24] The onset of the “nannoconid decline” precedes the δ13Ccarb excursion, and the lowest nannoconid abundances correspond to the isotopic event. An increase in nannoconid abundance (average values of 1800 specimens/mm2) correlates with the upper part of CM10N and the end of the isotopic excursion (Figure 3). Within nannoconids the wide-canal ones show a minor increase in abundance in the interval characterized by the δ13Ccarb excursion and in the overlying limestones.

[25] Two distinct abundance peaks of pentaliths are observed in the uppermost Berriasian and the lower Valanginian in the same intervals marked by highest percentages. However, maxima in relative and absolute abundances do not coincide (compare Figures 2 and 3). As previously reported by Bersezio et al. [2002], such pentalith peaks characterize lithozones with marly interbeds and black shales.

[26] Absolute abundances of the Diazomatolithus group markedly increase in the interval corresponding to magnetic chrons CM12 to CM10N, with highest values during the δ13Ccarb excursion. The symmetric increase and decrease are very similar to those of percentages (Figure 2), but thin section quantification revealed that the increase in absolute abundance of the Diazomatolithus group begins in magnetic chron CM12, correlates with the nannoconid decline and the upper part of the pentalith peak, and clearly precedes the δ13Ccarb excursion (Figure 3). Increases in absolute abundance of Assipetra infracretacea were not detected in thin sections, confirming that the peaks in relative abundance are peculiar of black shale layers.

3.2. Barremian-Aptian Interval at Cismon

[27] The upper Hauterivian/Barremian interval at Cismon is dominated by nannoconids and W. barnesae, which together make up to 90% of nannofloras. Both taxa show rhythmic, possibly orbitally driven, fluctuations in abundance (Figure 4). The relative abundance of W. barnesae is 40% on average, whereas narrow-canal nannoconids show average relative abundance of 50%, and wide-canal nannoconids fluctuate between 5 and 15%. Minima in nannoconid percentages correspond to organic carbon-rich black shales and interval with abundant chert [Erba et al., 1999]. Pentaliths are common to abundant in the reddish pseudonodular interval corresponding to magnetic chrons CM7 to CM5 [Erba et al., 1999; Channell et al., 2000]. In this interval, pentalith percentage fluctuates between 2 and 5%, with highest values of 8% in magnetic chron CM5.

[28] In the uppermost Barremian, slightly below magnetic chron CM0, a shift in dominance is marked by a sharp increase in relative abundance of W. barnesae and a coeval decrease of narrow-canal nannoconids (Figure 4). The Aptian is characterized by percentages of W. barnesae between 75 and 90%, very low values of nannoconids, and an increase of Assipetra and Rucinolithus [Erba, 1994; Erba et al., 1999; Channell et al., 2000]. Tremolada and Erba [2002] performed quantitative and morphometric analyses of Aptian nannofossils, pointing out an increase in abundance of large-sized Assipetra and Rucinolithus, especially in the interval representing OAE1a with a 60% spike in the lowermost black shales of the Selli Level (Figure 4).

[29] As reported by Channell et al. [2000], within magnetic chron CM0 the wide-canal nannoconids become, for the first time, more abundant than the narrow-canal ones (Figure 4). For both morphotypes the onset of the nannoconid crisis shortly predates the Selli Level [Erba, 1994; Erba et al., 1999; Premoli Silva et al., 1999; Channell et al., 2000]. In the upper Aptian, above the Selli black shales, only the wide-canal nannoconids recover and reach percentages up to18–20%, whereas the narrow-canal nannoconids are still present but never regain percentages as high as before OAE1a. Percentages up to 5% of pentaliths were detected in two discrete intervals: one just below the Selli Level and the other in the uppermost part of the of δ13Ccarb excursion (Figure 4).

[30] Thin section quantitative analyses showed that the narrow-canal nannoconids dominate the micrite in the upper Hauterivian/Barremian interval (magnetic chrons CM8 to CM1) with average values of 4000–6000 specimens mm−2 (Figure 5). Absolute abundances of W. barnesae average 5000 specimens mm−2. Fluctuations in nannoconid absolute abundance show higher amplitude than changes in relative abundance. Their lowest values are observed in the interval corresponding to the Faraoni level [Cecca et al., 1996; Erba and Larson, 1998; Erba et al., 1999] and in the black shale layers in the upper part of magnetic chron CM3. Other nannoconid minima correlate with intervals of massive cherts in magnetic chron CM1 (at approximately 55 m) and black shales within magnetic chron CM1n (at approximately 42 m) (Figure 5): they are more pronounced than in the distribution of relative abundance of narrow-canal nannoconids (Figure 4).

[31] A marked decrease in absolute abundance of narrow-canal nannoconids, from 7000 to 4000 specimens mm−2, occurs below magnetic chron CM0 and is followed by a further decrease within magnetic chron CM0, where the wide-canal nannoconids become more numerous. These changes are coeval with sharp increases in absolute abundance of W. barnesae to more than 10,000 specimens mm−2 just below magnetic chron CM0 and more than 15,000 specimens mm−2 within magnetic chron CM0.

[32] The nannoconid crisis affects both the narrow- and the wide-canal nannoconids, for which a drop to 250 specimens mm−2 and 140 specimens mm−2, respectively, was observed (Figure 5). Although the pentaliths Assipetra and Rucinolithus are temporarily common, their absolute abundances are orders of magnitude lower than those of W. barnesae and nannoconids (Figure 5). Pentalith peaks are visible in the upper Hauterivian, in the interval below the Selli Level, and in the upper Aptian. Assipetra and Rucinolithus display highest absolute abundances in the OAE1a interval and in the overlying limestones, reaching values as high as 1000 specimens mm−2 and an average abundance of 500 specimens mm−2.

3.3. Nannofossil-Calcite Paleofluxes

[33] In the Valanginian-Hauterivian and Barremian-Aptian, nannofossil calcite fluxes are largely determined by nannoconids, while the contribution of W. barnesae to carbonates is minor in terms of mass. Similarly, the other common to abundant taxa are negligible in nannofossil fluxes (Figures 6 and 7). Paleofluxes (expressed as gnannofossil CaCO3 10−6 mm−2 year−1), in fact, are controlled by absolute abundances (number of nannofossils mm−2 year−1) of various taxa and their volumes/mass (gCaCO3 10−9). Nannoliths of the narrow-canal N. steinmannii steinmannii are, by far, the forms with the highest volume/mass (Table 1) as well as the most abundant nannofossils, and consequently their paleofluxes are 5–20 times higher than those of any other taxa in the Valanginian/lower Hauterivian interval (Figure 6). Even when nannoconid paleofluxes reach the lowest values during the Valanginian δ13Ccarb excursion, they are 5 times greater than those of dominant W. barnesae. A similar difference in magnitude between paleofluxes of individual species was detected in the Hauterivian-Aptian interval of the Cismon core. Here the maximum values of W. barnesae paleofluxes are only 1/5 of the lowest nannoconid paleofluxes during the nannoconid crisis. Similar proportions of Assipetra and Rucinolithus versus nannoconids are detected even in the Selli interval, where they reach the highest abundances (Figure 7).

Figure 6.

Nannofossil CaCO3 paleofluxes (g 10−6 mm−2 year−1) of most abundant taxa in the Polaveno section. Lithologic legend and “nannoconid decline” as in Figure 2.

Figure 7.

Nannofossil CaCO3 paleofluxes (g 10−6 mm−2 year−1) of most abundant taxa in the Cismon core. Lithologic legend as in Figure 4.

[34] In the lower Valanginian, total fluxes of nannofossils vary between 22 and 17 g 10−6 mm−2 year−1, with an average value of 20 g 10−6 mm−2 year−1 (Figure 8). A first decrease in paleofluxes to an average of 15 g 10−6 mm−2 year−1 correlates with magnetic chron CM12, and a more pronounced one (from 15 to 7 g 10−6 mm−2 year−1) occurs in the uppermost part of magnetic chron CM12. The interval corresponding to magnetic chrons CM12 through CM10N is characterized by constantly low fluxes of 7 g 10−6 mm−2 year−1 on average, reaching the minimum values of 3 g 10−6 mm−2 year−1 at the climax of the C isotopic excursion. Nannofossil fluxes increase to 9 g 10−6 mm−2 year−1 in magnetic chron CM10N and then return to 15 g 10−6 mm−2 year−1 in the upper part of magnetic chron CM10N and 20 g 10−6 mm−2 year−1 in magnetic chron CM10, respectively.

Figure 8.

Total nannofossil CaCO3 paleofluxes (g 10−6 mm−2 year−1) in the Valanginian interval of the Polaveno section. Lithologic legend and “nannoconid decline” as in Figure 2.

[35] At Cismon (Figure 9), nannofossil paleofluxes are relatively low (3–10 g 10−6 mm−2 year−1) in the upper Hauterivian, represented by condensed limestones. In the Barremian, paleofluxes are higher (average of 15 g 10−6 mm−2 year−1) and display a gradual increasing trend with maximum values in the upper Barremian. Intervals with massive cherts and/or Corg-rich black shales correlate with minima in nannofossil paleofluxes (close to 5 g 10−6 mm−2 year−1). The interval corresponding to the uppermost part of magnetic chron CM1n is characterized by the highest nannofossil paleofluxes, with values between 23 and 28 g 10−6 mm−2 year−1 (Figure 9). In the lower Aptian a stepwise decrease in paleofluxes is marked by a first change from average values of 25–18 g 10−6 mm−2 year−1, then within magnetic chron CM0 a further decrease to approximately 10 g 10−6 mm−2 year−1 is recorded, followed by another drop to 5 g 10−6 mm−2 year−1. The nannoconid crisis interval is characterized by paleofluxes of 2 g 10−6 mm−2 year−1 on average; the minimum value of 1 g 10−6 mm−2 year−1 correlates with the negative spike in the C isotopic curve at the base of the Selli Level (Figure 9). Values oscillate between 3 and 8 g 10−6 mm−2 year−1 in the terminal part of the C isotopic anomaly of early late Aptian age (Figure 9).

Figure 9.

Total nannofossil CaCO3 paleofluxes (g 10−6 mm−2 year−1) in the uppermost Hauterivian/upper Aptian interval of the Cismon core. Lithologic legend as in Figure 4.

[36] A large reduction in nannofossil calcite fluxes of approximately 65% (three steps of −25, −40, and −22%, respectively) correlates with onset and extent of the Valanginian δ13Ccarb excursion. After the perturbation, nannofossil paleofluxes record a 65% increase (again in three steps of +22, +40, and +25%, respectively) reaching pre-C isotopic excursion values (Figure 10).

Figure 10.

Synthesis of total nannofossil paleofluxes (g 10−6 mm−2 year−1) during the Valanginian and early Aptian. Major decreases in biocalcification are interpreted as the response of calcareous phytoplankton to excess CO2 during emplacement of the Paranà-Etendeka and Ontog Java Plateau (OJP) LIPs. Timescale after Channell et al. [1995b]; δ13Ccarb curve after Weissert et al. [1998] and Erba et al. [1999]; LIP radiometric ages after Renne et al. [2001] and Duncan [2002]; trace metal data after R. A. Duncan (personal communication, 2002).

[37] In the Barremian-Aptian interval a first 28% decrease in paleofluxes occurs just below the base of magnetic chron CM0, then during magnetic chron CM0 a two-step (−44.5 and −50%, respectively) reduction of 72% is recorded, and a further 60% reduction culminates in the nannoconid crisis preceding the Selli Level black shales. A total decrease of approximately 90% in nannofossil paleofluxes occurred in a 1.5 myr long interval, at the onset of the early Aptian C isotopic anomaly and OAE1a (Figure 10). Above the Selli Level, nannofossil paleofluxes recover to average values of 5 g 10−6 mm−2 year−1, corresponding to an increase of approximately 60%.

4. Discussion

[38] The Cretaceous was a time of exceptional warmth with global deposition of organic carbon-rich sediments [Schlanger and Jenkyns, 1976; Arthur et al., 1985, 1987, 1990; Sliter, 1989; Bralower et al., 1993, 1994], carbon and strontium isotope excursions [Weissert and Lini, 1991; Bralower et al., 1997; Weissert et al., 1998; Jones and Jenkyns, 2001], and major biotic changes in planktonic communities [Tappan and Loeblich, 1973; Roth, 1987; Coccioni et al., 1992; Erbacher et al., 1996; Leckie et al., 2002]. Both the late early Aptian OAE1a and the Valanginian event represent times of carbonate crises in pelagic and neritic environments, enhanced productivity, and oceanic dysoxia/anoxia [Channell et al., 1993; Erba, 1994; Weissert et al., 1998].

[39] Quantitative studies of nannofossil micrite in the well-dated Polaveno and Cismon sections reveal major changes in: (1) abundance of total nannofloras; (2) relative and absolute abundances of single taxa with very different ultrastructure and mass/volume; and (3) nannofossil paleofluxes. Our results suggest that variations in relative abundances are only partially similar to changes in absolute abundances and can give a quite distorted picture of actual increases and decreases through time.

[40] Because calcareous nannoflora acts both as a biological pump (photosynthesis) and a carbonate pump (biomineralization), coccolithophore blooms and crises affect the organic and inorganic carbon cycles in addition to absorption of atmospheric CO2 into the oceans. However, the total number of nannofossils alone is not a measure of primary productivity and/or carbonate production. Other parameters, such as absolute abundance, ultrastructure and mass/volume of coccoliths/nannoliths produced by individual species should be taken into account to decipher paleoceanographic changes in temperature, nutrient content, light penetration and stability of surface waters. In fact, research on living nannoplankton indicates that coccolith/nannolith type, abundance, and degree of biomineralization depend on chemico-physical-trophic conditions, as well as the gas exchange between surface seawaters and the atmosphere.

[41] At Polaveno and Cismon the documented changes in nannofossil abundance and composition as well as in paleofluxes can be interpreted as the response of calcareous nannoplankton to global changes in the ocean-atmosphere system. Increased rates of volcanism during the formation of Ontong Java and Manihiki Plateaus and the Paranà-Etendeka LIP are proposed to have caused the geological responses associated with OAE1a and the Valanginian event, respectively. High levels of volcanogenic CO2 in the atmosphere, most probably turned the climate into a greenhouse state, accelerated continental weathering and increased nutrient content in oceanic surface waters via river runoff [Weissert, 1989; Erba, 1994; Jenkyns, 1999; Larson and Erba, 1999]. Moreover, higher fertility might have been triggered directly by biolimiting metals related to hydrothermal venting during plateau formation [Sinton and Duncan, 1997; Larson and Erba, 1999; Leckie et al., 2002].

[42] Calcareous nannofloras reacted to these new conditions of higher pCO2 and nutrient content in surface waters by drastically reducing calcification. The decrease in nannoconid abundance and the shift from narrow- to wide-canal forms are interpreted as a consequence of a shallow nutricline and excess CO2. Like in the modern oceans, increase of nutrients in the upper photic zone would induce blooms of nannoplankton producing small placoliths and inhibit the deep-photic zone nannoconids [Erba, 1994].

4.1. Trophic Conditions

[43] Studies on functional morphology of extant calcareous nannoplankton suggest that coccolithophores (and consequently their specific coccoliths/nannoliths) are geographically distributed and fluctuate in abundance according to biotic and abiotic ecological factors. Most important are trophic conditions: in waters with relatively high nutrient contents (upwelling or coastal nutrification), small placoliths are dominant and coccolithophores inhabiting the lower photic zone are virtually absent [Young, 1994]. Abundances of Florisphaera profunda and other taxa of the lower photic zone, relative to abundance of coccolithophores of the upper photic zone, have been quantified in various oceanographic settings from the Pacific, Atlantic, and Indian Oceans. Low percentages of F. profunda indicate a shallow nutricline favoring coccolithophores inhabiting the upper photic zone [Molfino and McIntyre, 1990; Okada and Matsuoka, 1996; Beaufort et al., 1997; Hagino et al., 2000; Takahashi and Okada, 2000; Kinkel et al., 2000; Broerse et al., 2000a, 2000b]. Nannoplankton producing heavily calcified coccoliths are typical of oligotrophic stable surface waters, where the deep photic zone assemblage thrives close to the thermocline. The so-called “mixed group” are common everywhere and do not show specific environmental adaptations [Young, 1994]. The quantity of calcite produced by coccolithophores therefore seems inversely correlated with trophic levels. Even if abundance of cells and coccoliths reaches enormous values during blooms, the small size and reduced volume/mass of fertility-related placoliths do not necessarily affect biogenic calcite fluxes.

[44] For the Early Cretaceous, paleobiogeographic and paleoceanographic reconstructions allowed identification of nannofossil indices with affinities for oligotrophic and mesotrophic conditions [Roth, 1981; Roth and Bowdler, 1981; Roth and Krumbach, 1986; Erba, 1986, 1992a, 1992b, 1994; Premoli Silva et al., 1989; Watkins, 1989; Erba et al., 1992; Williams and Bralower, 1995; Herrle, 2002]. It is clear that higher fertility-related nannofossils are small and contain far less calcite than the oligotrophic forms (Table 1). As a consequence, Cretaceous episodes/areas of enhanced primary productivity are invariably characterized by very low carbonate content, large quantities of organic matter, and increased biogenic silica. This is the case for OAEs, paleoequatorial upwelling zones, and mesotrophic to eutrophic coastal areas. Detailed studies of diagenesis and nannofossil preservation demonstrated that these reductions in carbonate content are not the result of dissolution [Roth and Krumbach, 1986; Erba, 1992b, 1994; Bralower et al., 1993, 1994; Premoli Silva et al., 1989, 1999].

[45] Morphometric analyses conducted on common Early Cretaceous nannofossil taxa [Tremolada and Young, 2002] indicate that mass/volumes of fertility-related taxa (Zeughrabdotus erectus, Biscutum constans, Discorhabdus rotatorius and Diazomatolithus lehmanii) are very small and that their contribution to calcite in pelagic carbonates is orders of magnitude lower than those of medium- to large-sized coccoliths/nannoliths of the oligotrophic forms (nannoconids and W. barnesae) (Table 1). Therefore an increase in abundance of the fertility-related coccoliths equal to 102–103 is required to produce the same quantity of calcite contained in one nannofossil of the oligotrophic forms.

[46] The change in composition of nannofossil assemblages and carbonate crises documented for both the Valanginian and the early Aptian events can be explained as a nannoplankton response to global nutrification episodes, directly or indirectly linked to major igneous/tectonic events [Channell et al., 1993; Erba, 1994; Larson and Erba, 1999; Leckie et al., 2002]. In both cases, global change in nannofossil-carbonate abundance correlates with an increase in deposition of biogenic silica and Corg-rich black shales and precedes the C isotopic excursions, suggesting that planktonic communities reacted to the early changes in trophic and climatic conditions and then persisted during the perturbation. A shift to mesotrophic/eutrophic conditions is further suggested by changes in radiolarian and planktonic foraminiferal assemblages [Coccioni et al., 1992; Erbacher et al., 1996; Premoli Silva et al., 1999; Leckie et al., 2002; Erba et al., 2004]. Higher nutrient contents might be introduced in surface waters by accelerated continental weathering and runoff under greenhouse conditions [Weissert, 1989; Lini et al., 1992; Channell et al., 1993; Erba, 1994; Weissert et al., 1998; Jenkyns, 1999; Larson and Erba, 1999; Premoli Silva et al., 1999; Bellanca et al., 2002; Bersezio et al., 2002] and or (volcanogenic) upwelling [Vogt, 1989; Arthur et al., 1990; Bralower et al., 1994, 1999; Premoli Silva et al., 1999]. Moreover, hydrothermal megaplumes related to formation of oceanic plateaus are potentially responsible for rapid introduction of enormous concentrations of dissolved and particulate biolimiting metals into the oceans [Sinton and Duncan, 1997; Larson and Erba, 1999; Snow and Duncan, 2002; Erba and Duncan, 2002; Leckie et al., 2002].

4.2. Atmospheric CO2

[47] Although the link between photosynthesis and calcification in coccolithophores has long been known, mechanisms involved are not fully understood. The relationship is probably based on the energetic needs of calcification. Light and photosynthesis facilitate coccolith production by providing adenosine triphosphate (ATP) as energy source for ion transport and synthesis of organic matrix. However, experimental results suggest that calcification can take place in absence of light, and photosynthesis is not required if CO2 diffuses sufficiently rapidly from the site of calcite deposition [Simkiss and Wilbur, 1989].

[48] The effects of increased atmospheric CO2 have been tested on extant Emiliania huxleyi and Gephyrocapsa oceanica, both in laboratory cultures and on incubations of natural nannoplankton assemblages from the North Pacific [Riebesell et al., 2000]. Nannoplankton calcium carbonate interacts with marine carbon cycling and ocean-atmosphere CO2 exchange. According to Riebesell et al. [2000], increased CO2 concentrations result in decreased calcification and lower ratio of calcification to particulate organic carbon (POC) production (calcite/POC). Under triple preindustrial CO2 levels, decreases of 15.7 and 44.7% in rate of calcification and decreases of 21 and 52.5% in calcite/POC ratio were recorded. Such changes are extremely important in marine ecosystems dominated by calcareous nannoplankton.

[49] Increases in atmospheric CO2 levels would/could lessen coccolith/nannolith production because coccolith secretion might represent a strategy to produce, directly within the cell, the waste-product CO2 reducing the energy cost of photosynthesis [Paasche, 1962; Young, 1994]. Therefore excess CO2 would make calcification less indispensable to coccolithophore life because surface waters are already (over) saturated with carbon dioxide.

[50] We speculate that during the Early Cretaceous, highly calcified coccoliths/nannoliths were secreted when atmospheric CO2 was low in order to sustain photosynthesis in calcareous nannoplankton. Conversely, during times of volcanogenic emissions of CO2, biocalcification was hampered while organic matter production was emphasized. If this interpretation is correct, then we can use changes in nannofossil paleofluxes to estimate atmospheric CO2 increases and decreases by analogy with experiments conducted on extant coccolithophores [Riebesell et al., 2000; Zondervan et al., 2001].

[51] The Valanginian event is marked by a 65% reduction in nannofossil paleofluxes that would correspond to a 2–3 times increase in CO2 during formation of the Paranà-Etendeka LIP (Figure 10). High carbon dioxide content in the atmosphere-ocean system was balanced back to pre-event values after 2 million years. Our data also suggest that blooms of r-selected (calcareous) phytoplankton were able to absorb excess CO2 by reducing calcification and enhancing production of organic matter.

[52] A much more drastic increase in CO2 must have been produced by emplacement of the giant Ontong Java and Manihiki Plateaus and formation of the Nova Canton trough system. In the lowermost Aptian a 90% reduction in nannofossil paleofluxes occurred in a 1.5 myr long interval leading into OAE1a (Figure 10). In this case a 3–6 times increase in volcanogenic CO2 is estimated. The return of nannoconids above the Selli Level only partially counterbalanced (increase of approximately +60% in nannofossil paleofluxes) the drop in biocalcification of the 1.25 myr-long nannoconid crisis interval. The relatively low nannofossil paleofluxes during the late Aptian is not surprising since atmospheric CO2 most probably remained high as a result of emplacement of the Kerguelen LIP [Duncan, 2002; Erba, 2002] and accelerated ocean crust production at mid-ocean ridges [Larson, 1991a, 1991b]. Our data might also be indicative of inability of phytoplankton to absorb excess CO2 above threshold values.

[53] If increases of atmospheric CO2 were the cause of reduced rates in nannoplankton biocalcification, then the abundance peaks of Assipetra and Rucinolithus during OAE1a are puzzling. Normal- and large-sized morphotypes of Assipetra and Rucinolithus [Tremolada and Erba, 2002] are in fact quite big and heavily calcified (Table 1). Their volume/mass and ultrastructure are also totally different from those of the generally accepted nannofossil indices of higher fertility. An alternative explanation might be that Assipetra and Rucinolithus are not fossil remains of coccolithophores, as further suggested by lacking documentation of their coccospheres/xenospheres. These peculiar nannoliths might represent CaCO3 precipitates and/or biocalcification by bacteria under extreme paleoenvironmental conditions, including massive methane release into the oceans [Opdyke et al., 1999, submitted manuscript, 2003; Tremolada and Erba, 2002; Bellanca et al., 2002; Beerling et al., 2002].

4.3. Anoxia-Dysoxia

[54] The late Valanginian event and early Aptian OAE1a are marked by pronounced positive carbon isotopic excursions, typically 2‰ higher than background levels (see Weissert et al. [1998] for a synthesis) recorded at a global scale. The OAE1a was also a time of widespread anoxia/dysoxia documented by burial of organic carbon at low and high latitudes and in all oceans. Sedimentological evidence (Corg-rich black shales) for an OAE in the Valanginian is limited. However, recent recovery of organic carbon enriched sediments of Valanginian age from the Shatsky Rise in the North Pacific Ocean [Bralower et al., 2002] suggests that also this time interval was characterized by global anoxia/dysoxia [Erba et al., 2004].

[55] Recently, Leckie et al. [2002] discussed plankton evolution in the mid-Cretaceous trying to unravel the biotic response of planktonic communities to enhanced primary productivity associated with OAEs. Accelerated evolutionary rates in calcareous nannoplankton, planktonic foraminifera, and radiolaria are interpreted as the result of higher fertility and global warming triggered by submarine volcanism.

[56] When high-resolution stratigraphy is available, anoxic/dysoxic conditions during OAEs clearly postdate the biotic response. This is certainly the case for the Valanginian and early Aptian since the nannoconid crises precede OAEs by some thousands years. Similarly, the documented drops in nannofossil paleofluxes anticipate the C isotopic anomalies and associated black shales. Therefore anoxia per se did not affect nannofloral abundance and composition, but conversely, the shift in (phyto) plankton dominance controlled the abundance and type of biogenic carbonate as well as organic matter produced in surface waters and subsequently incorporated in the geological record.

5. Conclusions

[57] Relative and absolute abundances of calcareous nannofossils are suggestive of paleoenvironmental global changes. Despite the close-sum problem, percentages of individual taxa can be used to trace major paleoceanographic modifications. However, absolute abundances are much more reliable and can be used for calculations of nannofossil paleofluxes, when high-resolution (integrated) stratigraphy is available. Our study pointed out that:

[58] 1. Morphometric analyses allow quantification of calcite contained in single coccoliths/nannoliths of individual species and estimates of nannofossil paleofluxes when absolute abundances are available.

[59] 2. CaCO3 per se is not a reliable measure of biogenic carbonate production, especially under oligotrophic conditions inducing blooms of k-selected highly calcified coccoliths/nannoliths.

[60] 3. Fluctuations in pelagic biogenic carbonates can be used to reconstruct paleofertility and paleoCO2 contents in surface waters.

[61] We interpret the nannoconid decline (Valanginian) and crisis (early Aptian) as the results of combined higher fertility and higher atmospheric CO2, both favoring small-sized coccoliths/nannoliths and, in general, r-selected phytoplankton. The Early Cretaceous biologic and carbonate pump could counterbalance the nutrification event and atmospheric CO2 increase at the onset of and during the Valaginian episode but only partially mitigated the much more severe conditions of the Aptian OAE1a.

[62] Additional nannofossil paleofluxes on well-dated successions, in combination with other proxies for paleo-CO2, will improve modeling of the paleobiological and paleocarbonate pump. Detailed reconstruction of nannofossil paleofluxes might also allow the identification of threshold values of pCO2 in the past, enabling more realistic predictions of timing and type of environmental modifications induced by future global change.


[63] We are very grateful to Helmi Weissert, Bob Duncan, Hugh Jenkyns, Jim Channell, Isabella Premoli Silva, and Roger Larson for interesting discussions. The manuscript benefited from the review by Mark Leckie, an anonymous reviewer, and Lisa Sloan: their constructive criticism and valuable suggestions were very helpful. This research was supported by COFIN 2001 to I. Premoli Silva.