4.1. Perturbations of the Global Carbon Cycle
 Although the carbon isotopic signatures of carbonate might be altered during diagenetic processes, it is safe to assume that the δ13Ccarb of our samples should mainly reflect the original signals. First, the carbon isotope compositions of micritic carbonate are regarded as being resistant to diagenetic influence [Marshall, 1992; Rosales et al., 2001; Saltzman, 2002; Joachimski et al., 2002; Buggisch and Joachimski, 2006]. Second, alterations of original δ13Ccarb signatures usually occur in the rocks containing soil organic matter [Allan and Matthews, 1982; Lohmann, 1988], or in organic carbon rich sediments, such as black shale [Joachimski et al., 2002; Buggisch and Joachimski, 2006]. In both cases, δ13Ccarb would shift to low values because the CO2 derived from these sources is 13C depleted. The absence of organic rich or shaly horizons in both the Dongcun and Yangdi sections suggests fewer alterations of carbon isotope composition. Third, except for methanogenesis, no diagenetic process is known to result in an enrichment of 13C during recrystallization [Buggisch and Joachimski 2006]. Therefore, in most cases, the positive excursions in δ13C of carbonate rocks are representative of the original signatures. Finally, the comparable isotope pattern observed in the Dongcun and Yangdi sections, which are deposited under different environmental conditions, also supports little alteration of primary signals [Buggisch and Joachimski, 2006].
 For the organic carbon isotope, the respiratory remineralization of organic matter in sediment column under oxic condition can produce an enrichment in 13C. This enrichment increases in sediments deposited beneath an anoxic water column [Hayes et al., 1989; Gong and Hollander, 1997; Fischer et al., 1998], where it can reach as high as 3‰. In the Dongcun and Yangdi sections, δ13Corg shows an increasing pattern with the decrease of oxygenation level at the beginning, but it still increases after the water mass return to oxic condition [Xu et al., 2008], indicating respiration of organic matter could not be the controlling factor for the δ13Corg shifts.
 An obvious decrease in Δ13C occurs near the F-F boundary in both the Dongcun and Yangdi sections due to a larger amplitude shift ofδ13Corg than δ13Ccarb (Figure 3). This differs from the previous observations on other sections [Joachimski et al., 2002; Chen et al., 2005]. Early paired analysis on the Berner section showed that the onset of δ13Corg excursion predated that of δ13Ccarbaround the F-F boundary [Joachimski, 1997]. The later organic carbon isotope analyses on the Kowala section showed that the δ13Corg displayed similar amplitude excursion with the δ13Ccarb shifts measured in other sections [Joachimski et al., 2002]. Based on the above evidence, Joachimski et al.  proposed that the high atmospheric and oceanic CO2 concentrations of the Devonian resulted in the maximum photosynthetic fractionation, and thus any change in CO2 concentration would not affect isotope fractionation during photosynthesis. However, the obvious Δ13C excursion, together with larger amplitude shift of δ13Corg than δ13Ccarb measured in the end Ordovician [Young et al., 2008] and Early Silurian [Cramer and Saltzman, 2007] suggested that photosynthetic fractionation might not reach maximum in the late Devonian, because the atmospheric CO2 levels of the former are much higher than the latter [Banner and Hanson, 1990; Berner and Kothavala, 2001]. Furthermore, the conclusion of Joachimski et al.  relies on an assumption that the carbon isotope records from different regions reflect a global carbon cycle change. This assumption, however, is potentially flawed, because the carbon isotope composition of dissolved carbon and organic matter is controlled not only by global but also by local carbon cycle [Panchuk et al., 2006]. Factors controlling local carbon cycle, such as circulation patterns, changes in primary productivity and phytoplankton community, can lead to carbon isotope variations [Freeman and Hayes, 1992; Laws et al., 1997; Bidigare et al., 1997; Popp et al., 1989; Hayes et al., 1999; Des Marais, 2001; Panchuk et al., 2006]. Circulation patterns influence the exchange rates of dissolved inorganic carbon between water mass, particularly in regions of epeiric seas [Panchuk et al., 2006]. Different species of phytoplankton, even the same species with different growth rate, might have different photosynthetic fractionations [Freeman and Hayes, 1992; Francois et al., 1993; Laws et al., 1997; Bidigare et al., 1997; Popp et al., 1989; Hayes et al., 1999], causing variations of δ13Corg. As a result, the values and patterns of carbon isotope compositions might be different among various regions due to the difference in signal strength of the regional carbon cycle. This point is supported by the inconsistence in the amplitude of δ13Ccarbshifts around the F-F boundary measured in different regions. For example, at the Cinquefoil Mountain section, Alberta, Canada [Wang et al., 1996], the excursion is from 1‰ to 5‰, whereas it is about 3‰ in some sections from Europe [Joachimski et al., 2002]. In addition, there is a potential for partial alteration of original isotope signals during deposition and diagenetic processes. For these reasons, it is more reasonable to use the whole trend of the difference between δ13Ccarb and δ13Corg of the same stratum to decipher the carbon cycle change [Cramer and Saltzman, 2007], and the consistent pattern of carbon isotope records from various sections should have a global significance. In the Kowala section, amplitude of the δ13Corgexcursion around the F-F boundary is not only larger than theδ13Ccarb, but also larger than the difference between the maximum and minimum values of δ13Ccarb for all samples [Joachimski et al., 2002]. Furthermore, the δ13Ccarb shifts recorded in the Plucki, Psie Górki, and Dębnik sections (around or less than +2‰), which are located in the same area with the Kowala section (around Kielce), are also less than the δ13Corg excursion recorded in the Kowala section [Racki et al., 2002]. Therefore, the carbon isotope records from Poland also support a decrease in Δ13C around the F-F boundary, and indicate that the photosynthetic fractionation might not reach maximum in the late Devonian.
 In the Baisha and Fuhe sections, Δ13C displays high-frequency vibrations across the F-F boundary [Chen et al., 2005]. However, this pattern of Δ13C is caused by several negative shifts of δ13Ccarb. As suggested by Buggisch and Joachimski , the lower values observed during the increase of δ13Ccarb are usually diagenetic signatures. In addition, the molecular isotope records [Joachimski et al., 2002], which are believed to be less affected by diagenesis, do not show negative shifts during the period of δ13Ccarb increase. More importantly, the Δ13C in the Fuhe section displays a decreasing trend at the F-F boundary, if the abnormal negative values ofδ13Ccarb are discarded (Figure 3).
 Summarily, the carbon isotope records from south China (Dongcun, Yangdi, Fuhe) and Poland (Kowala, Psie Górki, Plucki, and Dębnik) show a similar pattern near the F-F boundary, which is characterized by a larger amplitude excursion ofδ13Corg than δ13Ccarb, indicating that the decrease of Δ13C might have a global significance.
 The input of terrestrial organic matter could not interpret the δ13Corg shifts observed in this study. The Frasnian sequences of the Dongcun and Yangdi sections were deposited under different sedimentary environments [Chen et al., 2001; Wang and Ziegler, 2002]. The Dongcun section was deposited under pelagic depositional condition, while the Yangdi section was deposited under a platform environment. Theoretically, the influx of terrestrial organic matter into the Yangdi section should be larger than that of the Dongcun section. Therefore, δ13Corg values of the Yangdi section would be higher than those of the Dongcun section if the terrestrial organic matter were the dominant source of the sedimentary organic carbon, because the δ13C of terrestrial organic materials is often higher than that of contemporaneous marine organics. However, the fact is that there are no observable differences in δ13Corg between the Dongcun and Yangdi sections. Furthermore, the maximum values of δ13Corg in the Dongcun (−24.08‰) and Yangdi sections (−24.05‰) are larger than that of the Devonian terrestrial organic matter (−24.8‰ to −26.8‰) [Maynard, 1981]. The low δ13Corg of terrestrial organic matter can hardly explain the larger δ13Corgvalues near the F-F boundary.
 Though changes in metabolic pathway biomass community and growth rate have a potential for causing δ13Corg shifts, carbon isotope analysis of total organic matter has been shown to faithfully record the original isotopic trend when compared with compound specific δ13Corganalysis of short-chain n-alkanes as well as acyclic isoprenoids pristane, phytane, and hopane from identical samples [Joachimski et al., 2002]. The baseline values of the different biomarkers varied by a few per mille but the trends shown were practically identical, suggesting that Δ13C changes cannot be explained by variation in metabolic pathway and biomass community [Cramer and Saltzman, 2007]. Increased growth rate of the phytoplankton can also cause positive excursions of δ13Corg. The increased growth rate often increases primary productivities, leading to high influx of organic matter into sediments. The previous analysis on the Dongcun section showed that the carbon isotope shifts around the F-F boundary mainly resulted from the development of reducing conditions (Figure 2) [Xu et al., 2008]. Therefore, the concentrations of organic matter would increase at the F-F boundary if the growth rate were the major factor forδ13Corg shifts, because both reducing conditions and increased productivity favor for organic matter burial. However, there is no significant changes in organic matter contents during the interval with δ13Corg shifts (Figure 2), suggesting changes in growth rate should not the major cause for the δ13Corg changes. In addition, a study of a model versus measurement also indicates that the concentrations of dissolved CO2 play a major role in the changes of δ13Corg, though growth rate and/or other biotic factors might cause some of the shifts in δ13Corg [Rau et al., 1997].
 Most important, the large amplitude excursion of δ13Corg and its resulting Δ13C decrease are considered to be the two typical characteristics of the carbon isotope variations caused by lowering pCO2 level [Popp et al., 1989; Freeman and Hayes, 1992; Hayes et al., 1999]. The isotopic fractionation of 13C during photosynthesis is a function of extracellular and intracellular CO2 concentrations [Popp et al., 1989; Freeman and Hayes, 1992; Hayes et al., 1999; Pagani et al., 1999] and the photosynthetic fractionation factor is negatively correlated with atmospheric CO2 level [Popp et al., 1989]. As a result, δ13Corg would display a larger amplitude excursion than δ13Ccarb with decreasing atmospheric CO2 level, and thus Δδ13C shifts to negative.
 Consequently, the decrease of Δ13C caused by larger amplitude of δ13Corg than δ13Ccarb should mainly reflect the changes of atmospheric CO2 level rather than the variations of metabolic pathways and/or dominant organic matter sources.
4.2. Associations of Carbon Cycle Change With Mass Extinction
 As reviewed by Sandberg et al. , the F-F mass extinction had a close association with sea level changes. A stepwise extinction began with the severe sea level fall and the ultimate mass extinction took place well during severe eustatic falls that immediately followed major eustatic rises. These rapid changes of sea level are documented not only at the Steinbruch Schmidt section in a deep water, submarine-rise setting in Germany, but also at sections in inner shelf and outer shelf and slope settings in Belgium, Nevada, Utah [Sandberg et al., 2002]. This evidence, together with the worldwide positive δ13Ccarbexcursion, suggests that the F-F mass extinction might be associated with both the sea level and global carbon cycle changes.
 The detail eustatic changes and conodont evolution across the F-F boundary were reconstructed based on the records from the Dongcun section [Wang and Ziegler, 2002]. The sea level curve drawn based on the integrated analysis of biofacies, lithofacies, and sequence stratigraphy in the Dongcun section [Wang and Ziegler, 2002] is comparable with the most widely accepted sea level curve produced by Sandberg et al. , further supporting that the records in the Dongcun section have a global signification. These data, together with the carbon isotope records, make it possible for scrutinizing the associations of the carbon cycle changes with biomass extinction through analyzing the phase differences among the global carbon cycle change, sea level change, and conodont extinctions.
 The conodont extinction displays a close association with the eustatic changes around the F-F boundary in the Dongcun section [Wang and Ziegler, 2002] (Figure 4). In the early linguiformis zone, Palmatolepis, a pelagic genus which favored the farthest offshore and deep water setting [Sandberg and Ziegler, 1996], is dominant with an average more than 85% of total fauna, indicating a typical palmatolepid biofacies. Based on the analysis of sequence stratigraphy, this interval is a highstand systems tract, corresponding to the upper part of eustatic rise Event 5 of Sandberg et al.  [Wang and Ziegler, 2002]. Then the “Upper KW Horizon” (−2.04 to −1.84 m) event suddenly happened, the facies instantaneously changed into dark limestone, indicating the beginning of remarkable sea level fall (Figure 4). A thick-bedded shallow water limestone developed from −2.0 to 0 m. The fauna in this interval is a typical reduced or impoverished in Frasnian fauna. Conodonts are low in diversities, scarce in number of individuals, and all are surviving taxa from the underlying sequence. The deposition from −2.0 to −0.8 m corresponds to Event 7 ofSandberg et al.  (Figure 4). The final extinction of Palmatolepis linguiformis is at –0.82 m, corresponding to the beginning of the more rapid shallowing (Event 8 of Sandberg et al. ) (Figure 4). The stratum of −0.4 to 0 m is an important interval of sea level changes, indicating the most pronounced shallowing facies. No conodonts have been found at the F-F boundary; this interval could actually be a big extinction, corresponding to Event 9 ofSandberg et al.  (Figure 4). From the base of the triangulariszone, the strata thin upward and consist of dark thin-bedded limestones, representing a relatively rapid sea level rise (Figure 4). From the late triangularis zone to the early crepida zone, the facies changes into deeper water, but is still much shallower than the depositional facies below −1.98 m [Wang and Ziegler, 2002].
Figure 4. Relationships among δ13Ccarb, Δ13C, conodont evolution, and sea level changes across the F-F boundary in the Dongcun section. The patterns of conodont evolution and sea level changes are comparable with the records from other continents and have a global signification. The numbers represent the events identified bySandberg et al. .
Download figure to PowerPoint
 The direct effect of marine regression is reducing the habitat areas on continental shelves for shallow marine species. However, this assumption can hardly explain species loss of terrestrial ecosystem because the areal extent and habitat of the terrestrial realm should increase with the marine regression [Boulter et al., 1988; Raymond and Metz, 1992, 1995] In addition, Jablonski suggested that only 13% of modern families would become extinct even if all the modern shelf biota were eliminated. Thus, the sea level fall alone could not cause the F-F mass extinction.
 Figure 4shows the associations of conodonts extinction with the carbon cycle, eustatic change, and the detailed evolution of condonts across the F-F boundary in the Dongcun section. The onset of eustatic fall postdates the beginnings ofδ13Ccarb and Δ13C shift, and the largest marine regression occurs during the interval with lowest Δ13C values (Figure 4). These characteristics provide key evidence for interpreting the relationship between sea level change and increased organic carbon burials. The increased organic carbon burial can lead to decrease in atmospheric CO2 level [Kump and Arthur, 1999], which is commonly taken to be the main driver of climate change on geological timescales [Berner and Kothavala, 2001; Royer et al., 2004]. The lower pCO2 would lead to temperature decrease through weakening greenhouse effect. With temperature decrease, an ice cap would develop in high altitude and latitude, causing sea level fall. This postulation is supported by the temperature decrease indicated by the δ18O of conodont apatite across the F-F boundary [Joachimski and Buggisch, 2002]. Significant temperature decline would decimate reefal and tropical ecosystems, as species inhabiting these places have no refuge against cold. In contrast, the high-latitude species could migrate to lower latitudes and maintain their tolerable temperature. Each of these expected patterns of ecological selectivity in survival are observed in the F-F event [McGhee, 1996], suggesting that the temperature decline should be an important factor for the mass extinction.
 For the causes of the increased carbon burial around the F-F boundary, our previous high-resolution investigations [Xu et al., 2008] revealed that the reducing conditions predate the onset of the carbon isotope excursions, suggesting that the two positive carbon isotope shifts are likely caused by the expansion of anoxic conditions. The low Al/(Al+Fe) ratio across the F-F boundary leads the U/Al and Cu/Al anomalies in timing, implying that this anoxic event might have resulted from a long-term cumulative effect of intense hydrothermal-volcanic activities. Therefore, long-term cumulative effect of intense hydrothermal-volcanic activities lead to development and expansion of anoxic water mass, which in turn cause increased organic carbon burial and thus atmospheric CO2 level decrease.
 In the late rhenana zone, a positive δ13Ccarb excursion has been detected on the worldwide. This positive excursion, together with organic rich sediments (lower Kellwasser horizon), indicates another perturbation of global carbon cycle in the late rhenana zone. However, no obvious positive shift of δ13Corg has been detected in the late rhenana zone in most previous studies. Joachimski et al.  suggest the δ13Corg maximum might be missed due to the broader sampling interval. The carbon isotope records in the Dongcun and Yangdi sections provide critical evidence for the above postulation. The carbon isotope records in the Dongcun and Fuhe sections display a similar pattern in the late rhenanazone with that at the F-F boundary, which is characterized by a relatively larger positive shift inδ13Corg than δ13Ccarb (Figure 2) and negative Δ13C shift (Figure 3). These characteristics suggest that a perturbation of carbon cycle like that at the F-F boundary might occur in the laterhenana zone.