Iron Loss During Continental Weathering in the Early Carboniferous Period Recorded by Karst Bauxites

Significant iron (Fe) loss can occur during continental weathering and efflux to the ocean via runoff, historically affecting global Fe cycling and marine ecosystems. Here, we report extremely low Fe content in early Carboniferous (ca. 340 Ma) bauxites in southwestern China. These bauxites were formed by redeposition of terrestrial soils along the paleo‐continental margin of the western South China Plate in warm climates. The bauxites contain high δ56Fe (−0.17‰ to +1.15‰) values with a negative correlation between Fe2O3 and δ56Fe, indicating that a substantial amount of Fe(III) was reduced to isotopically light dissolved Fe(II) and effuxed to the ocean via reductive dissolution under anoxic conditions. The low Corg content and low FeHR/FeT, Mo/Al, U/Al, and V/Al ratios of bauxite suggest that this reduction process occurred during the pedogenic (continental weathering) rather than the depositional/diagenetic stage of karst bauxite formation. Most of the dissolved Fe(II) were rapidly re‐oxidized to Fe(III) and transported toward the paleo‐continental margin forming iron ores with δ56Fe values around zero (−0.13‰ to +0.16‰). The negative correlation between Al2O3 and Fe2O3 contents in global karst bauxites suggests common Fe loss processes during continental weathering in geological periods favoring karst bauxite formation, such as during the Carboniferous, Permian, and Cretaceous periods and the Cenozoic era. Karst bauxite may thus provide a record of Fe loss during continental weathering and act as an indicator of enhanced Fe flux to oceans.

Pioneering studies have documented that notable net Fe loss can occur during continental weathering and soil formation in (sub)tropical climate regions due to intensification of weathering with elevated temperature and rainfall (Akerman et al., 2014;S. Liu et al., 2014;Thompson et al., 2007;Wiederhold et al., 2006;Yesavage et al., 2012). These Fe were then delivered to the ocean through river systems and thus have a major impact on the global Fe cycle (Boyd & Ellwood, 2010;Tagliabue et al., 2014). The formation of Fe-poor soil associated with Fe loss is of great interest and has been used to study Fe migration and transformation at the Earth's surface (e.g., Thompson et al., 2007;Yamaguchi et al., 2007). However, it has also been argued that during extreme weathering under oxidative conditions, Fe is transformed into immobile ferric Fe and can be re-precipitated as Fe (hydr)-oxides (e.g., S. Liu et al., 2014; J. L. Ma et al., 2007;Poitrasson et al., 2008). It is an internal redistribution (closed-system) of Fe within the soil sections, resulting in limited Fe loss (M. Qi et al., 2022). Therefore, the link between the Fe budget and continental weathering under (sub)tropical climates is not well established and requires further study.
Previous studies show that Fe loss is associated with the formation of bauxite deposits (e.g., Ling et al., 2017;Mameli et al., 2007). Bauxite is a (paleo-) chemical residue of intense subaerial weathering with Al 2 O 3 contents of >35% and Al 2 O 3 /SiO 2 mass ratios of >2.6 (Bogatyrev et al., 2009). The link between bauxite deposits and tropical weathering regimes was established by Retallack (2010). Based on the lithology of the depositional basement, bauxite deposits are generally divided into three types (Bárdossy, 1982): (a) karst bauxite developed on the karstified surfaces of carbonate rocks; (b) lateritic bauxite developed on aluminosilicate rocks; and (c) Tikhvin bauxite representing transported and redeposited material overlying the eroded surface of aluminosilicate rocks. Karst bauxites are widely developed in Phanerozoic strata around the world, especially in the Carboniferous, Permian, Cretaceous periods, and Cenozoic era. The total karst bauxite resources exceed 10 billion tonnes (Figure 1, Bárdossy, 1982;Bogatyrev et al., 2009;Retallack, 2010). Bauxite deposits in China are dominated by the karst type and mainly of Carboniferous and Permian age found in the Guizhou, Guangxi, Shanxi, Henan Provinces, etc. (Deng et al., 2010;S. Yang et al., 2022;Yu et al., 2019). Karst bauxites from China (especially central Guizhou bauxite) generally have much lower Fe 2 O 3 contents (<2 wt. %, Ling et al., 2017) than other sediments/sedimentary rocks such as shale (North American Shale Composite (NASC) average 5.65 wt. %, Gromet et al., 1984). Central Guizhou karst bauxites were formed by the deposition of Al-rich soils in the early Carboniferous period on carbonate unconformities (basins) along the paleo-continental margin of the western South China Plate (Figures 1a-1c, Yu et al., 2019). They may have recorded ancient release of Fe during its formation and thus serve as an interesting target for studying earth-surface Fe cycling during extreme continental weathering in the geological past.
Fe has two valence states (Fe(II) and Fe(III)) and can produce large isotopic fractionation during redox reactions (Bergquist & Boyle, 2006). Fe(II) is generally more mobile and isotopically lighter than Fe(III) (John et al., 2012; 10.1029/2022JF006906 3 of 15 Teng et al., 2008). In surficial environments, this difference in mobility results in the release of light Fe isotopes into solution and the preferential sequestration of heavy Fe isotopes in the solid residue (e.g., Ilina et al., 2013;Johnson et al., 2008). Therefore, iron isotopes can provide deep insights into the mechanism of Fe mobilization (Rouxel et al., 2005).
As Fe-poor bauxite is a chemical residue of extreme weathering, studying the mechanism of Fe loss during karst bauxite formation could improve our understanding of the Earth-surface Fe cycling in tropical climate regions. To achieve this, we studied the isotopic composition of Fe in the early Carboniferous bauxites (ca. 340 Ma) from central Guizhou, southwestern China.

Geological Setting
The lower Carboniferous bauxite belt in central Guizhou, southwestern China, located in the western part of the South China Plate, contains more than 40 bauxite deposits with total resources exceeding 500 million tons ( Figure 1a; Yu et al., 2019). During the Early Paleozoic, South China and North China plates separated from Gondwana and drifted northward (Metcalfe, 2006). During the early Carboniferous period, the South China Plate drifted close to the equator ( Figure 1b) and experienced climatic warming, which was conducive to the formation of tropical soil via crustal weathering . During this period, Al-rich soils were transported mainly as suspended particulate matter (SPM) by rivers to the continental margin (Youjiang Basin; east Paleo-Tethys) and deposited in the Qingzhen-Xiuwen (Guiyang) and Zunyi coastal basins, forming the lower Carboniferous Jiujialu Formation bauxites in the central Guizhou area (Figure 1c, Yu et al., 2019).

Sampling and Analytical Methods
A total of 68 samples from the lower Carboniferous Jiujialu Formation were collected in central Guizhou, southwestern China. Fresh rock samples from the Jiujialu Formation, including bauxites, bauxitic clays, clay rocks, iron ores, and iron-rich clays (Figure 4), were collected from outcrops and open pits at three bauxite deposits in the Guiyang area (Xiaoshanba, Lindai, and Yunwushan), and three bauxite deposits in the Zunyi area (Xinzhan, Houchao, and Xianrenyan) ( Figure S1 in Supporting Information S1).
Polished thin sections were prepared for scanning electron microscope-energy dispersive spectrometer (SEM-EDS) analyses using a Thermo Scientific Scios DualBeam SEM-EDS at the Institute of Geochemistry, Chinese Academy of Sciences (IGCAS). Powdered samples were prepared for analyses of Al 2 O 3 , SiO 2 , and Fe 2 O 3 content (n = 47) at the IGCAS. The samples were washed, air-dried, powdered to 200 mesh, and homogenized prior to chemical analysis. Major element contents of whole-rock samples were determined by X-ray fluorescence (PANalytical, AXIOS-PW4400) at the IGCAS. The analytical precision is better than 5%. The trace element Mo, U, and V abundances were analyzed through whole-rock solution-ICP-MS techniques (PlasmaQuant MS Elite) at the IGCAS. The ICP-MS measurements were quality controlled using international standard samples OU-6, AMH-1, and GBPG-1, and the relative standard deviation was better than 10%. The organic carbon (C org ) analyses (n = 29) were conducted with an Elementar Vario Microcube analyzer at IGCAS, with analytical errors of less than ±2.5%. Prior to the analyses, the sample powders were leached with 2.5 N HCl to remove inorganic C. The Fe species extraction (n = 29) were performed at the China University of Geoscience. The Fe-pyrite fraction (Fe Py ) was calculated from Ag 2 S produced by the chromous chloride distillation (Canfield et al., 1986). Fe-carbonate (Fe carb ; siderite and ankerite), Fe-oxide (Fe ox ; e.g., ferrihydrite, goethite and hematite), and Fe-magnetite (Fe mag ) species were extracted following the method established by Poulton and Canfield (2005). Fe contents were measured using atomic absorption spectrometry.

Results
The bauxites and bauxitic clays (Al 2 O 3 > 35% and 1.8 < Al 2 O 3 /SiO 2 < 2.6) we examined are composed mainly of Al-hydroxide (diaspore and boehmite), followed by kaolinite and illite, and a small amount of Fe minerals and detrital minerals (e.g., anatase and zircon) (Figures 5a and 5b). The clay rocks are composed of clay minerals such as kaolinite and illite, with Al-hydroxides and detrital minerals (Figure 5c). The iron ores and iron-rich clays are composed of Fe minerals (mainly hematite with a small amount of goethite or siderite) and clay minerals, with a small amount of automorphic quartzes and almost no detrital minerals (Figures 5d-5f). The significantly different mineralogical features between bauxite and Fe bed samples indicate different origins, that is, the former is of detrital depositional origin, while the latter is of chemical depositional origin.
The ratio of highly reactive Fe (Fe HR = Fe Py + Fe carb + Fe ox + Fe mag ) to total Fe (Fe T ) is also a reliable proxy of redox conditions (Scholz, 2018;Scholz, Beil, et al., 2019), with its values <0.22, 0.22-0.38, and >0.38 representing the oxic, possibly anoxic, and anoxic conditions, respectively (Poulton & Canfield, 2005). Most of our samples have Fe HR /Fe T ratios lower than 0.38, and among them, clay rocks have higher Fe HR /Fe T ratios (average 0.36) than bauxites/bauxitic clays (average 0.27) and iron-rich clay (LD3-11: 0.128) (Figure 6a). In addition, Fe HR /Fe T values are negatively correlated with Al 2 O 3 /Fe 2 O 3 ratios (Figure 6a). These results are again consistent with the formation under oxic conditions. Bauxite/bauxitic clay has a higher proportion of silicate-bound Fe (average Fe sil /Fe T = 0.725, where Fe sil = Fe T − Fe HR ) than clay rock (average 0.645) (Figure 6b). The δ 56 Fe values of bauxites/bauxitic clays vary  (Poulton and Canfield, 2011). Since the formation of karst bauxite is a process of Fe loss and Al retention, the Al 2 O 3 /Fe 2 O 3 value can be used to assess the mineralization degree of bauxite; that is, the higher the Al 2 O 3 /Fe 2 O 3 value, the higher the mineralization degree. from −0.17‰ to +1.15‰ (n = 21), whereas clay rocks (−0.09‰ to +0.35‰; n = 4) and iron ores/iron-rich clays (−0.13‰ to +0.16‰; n = 12) have a narrower range of δ 56 Fe values.

Fe Loss During Karst Bauxite Formation
Karst bauxite is a kind of Fe-depleted and Al-enriched sedimentary rock, and its mineralization process mainly includes three stages: (a) pedogenesis, were the parent rocks are intensity chemically weathered to form Al-rich soils in warm climates; (b) sedimentary transport, were Al-rich soils are eroded and transported toward the continental margin by runoff; and (c) subsequent deposition and diagenesis to form karst bauxites (Bárdossy, 1982;Yu et al., 2019). The depositional basement, that is, carbonate rock and/or shale, has long been regarded as the source rock of central Guizhou bauxite (Ling et al., 2017;Yu et al., 2019). Intensive chemical weathering of the parent rock of bauxite forms Al-rich soils, which are subsequently eroded and effluxed to the ocean by rivers as dissolved phases (include colloid: 1-200 nm), particles including SPM (>0.2 μm), or sand (>63 μm) (Gaillardet et al., 2003). The detrital minerals (e.g., zircon and anatase) in bauxites studied here, which have been retained after diagenesis, are predominantly small in size (0.2-5.0 μm) (Figures 5a and 5b). Diaspore, the primary economic mineral in karst bauxite, is commonly columnar or platy and 1-10 μm in size (Figures 5a-5d). These particles are consistent in size with SPM, suggesting that the precursor materials leading to karst bauxite formation were mainly transported as riverine SPM.
Bauxite ores have significantly high Al 2 O 3 (average 70.1 wt. %) and extremely low Fe 2 O 3 (average 1.02 wt. %) contents (Table S1 in Supporting Information S1 and references therein), significantly different from normal sediments/sedimentary rocks and NASC (Al 2 O 3 and Fe 2 O 3 contents of 16.9 and 5.65 wt. %, respectively; Gromet et al., 1984). This difference in geochemical composition implies that karst bauxite formation in central Guizhou had significant Fe loss but retained Al. Mass-change calculations for the sedimentary basement of central Guizhou bauxites (Loushanguan Group dolomite) suggest that Fe loss occurred during the leaching stage of pedogenesis, supporting this inference (Ling et al., 2019). In addition, karst bauxite samples from around the world have a negative correlation between Al 2 O 3 (average: 61.9 wt. %) and Fe 2 O 3 (average 4.42 wt. %) contents (Figure 7), suggesting that Fe loss is a common process during karst bauxite formation. We tested if Fe loss occurred and at what stage of bauxite mineralization by examining the Fe stable isotopes of our samples.

Iron Isotopic Evidence of Fe Loss
In terrestrial ecosystems, surface weathering under strongly oxidizing conditions has limited Fe isotope fractionation, with mean δ 56 Fe in clastic rocks and sediments of 0‰ ± 0.2‰ (e.g., Beard et al., 2003). The low Fe bauxites we investigated show high δ 56 Fe values (up to +1.15‰) (Figure 8). Two processes can lead to high δ 56 Fe values, loss of dissolved Fe(II) that is relatively enriched in 54 Fe, or the net addition of a component enriched in 56 Fe (Thompson et al., 2007;Yamaguchi et al., 2007). Our data, showing an inverse relationship between Fe and Al, support dissolved Fe(II) loss during karst bauxite formation (Figure 7), leading to high δ 56 Fe values.
Four major processes have been postulated to control the dissolution of Fe minerals in terrestrial ecosystems: (a) proton-promoted dissolution, (b) oxidative dissolution, (c) reductive dissolution, and (d) ligand-controlled dissolution (B. Wu et al., 2019 and reference therein). Experimental studies demonstrated that proton-promoted dissolution (process (a) causes no or only limited, Fe isotope fractionation Wiederhold et al., 2006). In contrast, oxidative dissolution (process 2) always caused the accumulation of heavy Fe isotopes in the leachates, leaving behind a rock with relatively light δ 56 Fe values (B. Wu et al., 2019). Reductive dissolution (process 3) under anoxic conditions, with or without bacteria, can produce aqueous Fe(II) that has 0.5‰-4‰ lower δ 56 Fe values than the Fe(III) in the initial material (e.g., Brantley et al., 2004;Butler et al., 2005;Chanda et al., 2021;Icopini et al., 2004). Inorganic ligand-controlled dissolution (process 4) causes minor depletion (∼0.5‰) in δ 56 Fe values of the solid phase, whereas biotic/organic ligand-controlled dissolutions can cause significant Fe isotope fractionation (∼2.2‰) Kiczka et al., 2010;Wiederhold et al., 2006). The Fe isotope fractionation caused by biotic/organic ligand-controlled dissolution was also caused by reductive dissolution due to the depletion of oxygen through microbial respiration . This implies that during the karst bauxite formation, reductive dissolution under anoxic conditions may be the primary control on Fe isotope fractionation.

Pedogenesis Leads to Fe Loss
Fe loss caused by reductive dissolution in the anoxic environment may have occurred at one or more stages of karst bauxite mineralization. The precursor materials leading to karst bauxite formation were mainly transported as riverine SPM. In the Phanerozoic with an oxidizing atmosphere, the presence of abundant O 2 in river water prevented the reductive dissolution of Fe from taking place (Canfield, 1997). This ruled out the possibility that Fe loss occurred in the sedimentary transport stage.
Continental marginal sediments may have released dissolved Fe during depositional/diagenetic processes (Dale et al., 2015;Elrod et al., 2004;John et al., 2012;Lam & Bishop, 2008;Noffke et al., 2012;Scholz et al., 2014;Scholz, Schmidt, et al., 2019;Severmann et al., 2010). However, published marginal sediments have δ 56 Fe values ranging from −0.32‰ to +0.3‰ (B. Wu et al., 2019 and reference therein). The less heterogeneous δ 56 Fe values relative to bauxite ores suggest that the depositional/diagenetic process was unlikely to be the controlling factor in the dissolved Fe released. This is consistent with the C org , Fe species, and redox-sensitive element studies, the results all of which suggest that the bauxites in central Guizhou were formed under oxic depositional/diagenetic conditions (Table S1 in Supporting Information S1; Figure 6 and Figure S2 in Supporting Information S1). Consequently, we hypothesize that most of the Fe loss occurred during the pedogenic stage of karst bauxite formation under anoxic conditions.
The pedogenic process can develop a relatively wide range of δ 56 Fe values (−0.52‰ to +1.04‰) compared with their parent rocks during primary dissolution and secondary processes, such as oxidation, precipitation, as well as complexation with soil organic matter (Fekiacova et al., 2013(Fekiacova et al., , 2017Garnier et al., 2017;Johnson et al., 2008;Kiczka et al., 2011;S. Liu et al., 2014;Qi et al., 2022;Thompson et al., 2007;Yesavage et al., 2012). When anoxic conditions are present in water-saturated soil with high rainfall, the Fe isotope fractionation during weathering is redox-controlled and can show elevated δ 56 Fe values due to a preferential release of light Fe isotopes (Akerman et al., 2014;S. Liu et al., 2014;Schuth et al., 2015;Thompson et al., 2007;Wiederhold et al., 2007a;Yamaguchi et al., 2007). For instance,  the soil horizon in Hawaiian, studied by Thompson et al. (2007), exhibited high δ 56 Fe values (up to +0.72‰) due to increased Fe loss in the soil profile under enhanced anoxic conditions. Another example is the Paleoproterozoic Hekpoort paleosol profile from Gaborone, Botswana (∼2.2 Ga), that has high δ 56 Fe values (−0.17‰ to +1.04‰ with an average of +0.55‰) and negative correlation between δ 56 Fe ratios and Fe 2 O 3 contents. These paleosols are considered to be a result of Fe(II) loss under reduced, organic-acid bearing soil water and groundwater (Yamaguchi et al., 2007). Our data showing increasing bulk δ 56 Fe values in conjunction with the logarithmic decrease in Fe 2 O 3 content (R 2 = 0.53) in bauxite samples (Figure 8), is consistent with the residual Fe-depleted soil being enriched in heavy Fe isotopes due to preferential removal of light Fe isotopes during pedogenesis under anoxic conditions.
In addition to bauxites having the highest δ 56 Fe values (up to +1.15‰) and Fe sil /Fe T ratios (average: 0.725), the positive correlation between Al 2 O 3 /Fe 2 O 3 and Fe sil /Fe T ratios (Figure 6b) indicates that the Fe species in the bauxite ore is controlled by Fe sil with elevated δ 56 Fe values that were inherited from Fe-depleted soils. This is consistent with the δ 56 Fe values of the Fe Sil fractions (up to +1.5‰) published to date that are exclusively positive Fe species in soils (B. Wu et al., 2019 and reference therein). Therefore, it is reasonable to hypothesize that during pedogenesis, dissolved Fe(II) with light Fe isotopes were preferentially removed from the primary minerals during continental weathering, leaving the Fe sil in residual soil with heavy Fe isotopic fingerprints (Fekiacova et al., 2013;B. Wu et al., 2019). However, most of this dissolved Fe(II) may have been rapidly re-oxidized to Fe(III) prior to its migration, likely as colloidal substances (M. B. Wu et al., 2019). This oxidation process produces Fe(III) with δ 56 Fe values 0.5‰-4‰ higher than that of Fe(II) in the initial material, as observed in both laboratory experiments (Anbar et al., 2005;Beard et al., 2010;Johnson et al., 2002;Nie et al., 2017;L. Wu et al., 2012) and field studies (M. Wiederhold et al., 2007b;R. Zhang et al., 2015). The Fe(III) colloids could have Fe isotopic compositions similar to that of continental crust (δ 56 Fe = +0.07‰) due to their similar extent of isotopic fractionation between the oxidation of Fe(II) to Fe(III) and the reduction of Fe(III) to Fe(II) (M. Poitrasson et al., 2008;Wiederhold et al., 2006;B. Wu et al., 2019;R. Zhang et al., 2015). Such reductive dissolution and re-oxidation has been widely observed during pedogenesis (L. M. Huang et al., 2018;M. Li et al., 2017;Schuth et al., 2015;Thompson et al., 2007;Wiederhold et al., 2007b).

Genesis of Iron Beds Beneath Bauxite Beds
The results of petrological and mineralogical studies suggest that iron ore/iron-rich clay was precipitated through a chemical process, that is, Fe(III) colloid precipitated directly due to flocculation or formation of insoluble Fe(III) due to oxidation of dissolved Fe(II) in the water column (Figures 4 and 5, Boyle et al., 1977). River input was the most likely Fe source for these iron ore/iron-rich clay deposits in the continental margin (Escoube et al., 2009;W. Li et al., 2015). In the Phanerozoic era, however, riverine dissolved Fe is dominated by Fe(III) colloids and rare dissolved Fe(II) (Boyle et al., 1977;Johnson et al., 2002;B. Wu et al., 2019). Therefore, a possible mechanism for the formation of iron beds would be the precipitation of riverine Fe(III) colloids prior to the deposition of bauxite beds in the central Guizhou region. Iron bed samples (−0.13‰ to +0.16‰) show similar Fe isotopic compositions to riverine Fe(III) colloids and continental crust (Figure 8, Fantle & DePaolo, 2004;Ingri et al., 2006), further supporting this interpretation.
Overall, our results support a Fe release and reprecipitation processes during karst bauxite formation in central Guizhou, southwestern China as follows (Figure 9): (a) during the early Carboniferous period, intensive continental weathering under a greenhouse climate may have enhanced the release of isotopically light Fe(II) through reductive dissolution under anoxic conditions in water-saturated soils, leaving the silicate bound Fe in residual soils with heavy Fe isotopic fingerprints; (b) the majority of these dissolve Fe(II) were rapidly oxidized in the presence of O 2 to Fe(III) with a δ 56 Fe value around zero, and were transported in colloidal form by rivers to the continental margin, forming thin Fe-bed layers in karst depressions via chemical precipitation; (c) the Fe-depleted and Al-enriched soils were then transported as riverine SPM and deposited upon the iron bed forming karst bauxite bed that inherits the Fe isotopic composition of soils.

Conclusions and Implications
The high Al 2 O 3 and low Fe 2 O 3 contents of the studied bauxite samples indicate Fe loss during Carboniferous karst bauxite formation in central Guizhou, southwestern China. The relatively high bauxite δ 56 Fe values (−0.17‰ ± 0.09-1.15‰ ± 0.13‰, average 0.58‰ ± 0.09‰) but low C org contents and low Fe HR /Fe T , Mo/Al, U/Al, and V/Al ratios suggest oxic diagenetic but anoxic pedogenic conditions during karst bauxite formation. The increasing bulk δ 56 Fe values with a decrease in logarithmic Fe 2 O 3 concentrations of bauxites indicate that the incomplete reduction of Fe(III) during pedogenesis releases isotopically light dissolved Fe(II), leaving heavier Fe in the residual soils to eventually be recorded in bauxites. Most of the dissolved Fe(II) were rapidly oxidized to Fe(III) with δ 56 Fe values around zero and transported toward the paleo-continental margin of the western South China Plate, forming Fe ore and Fe-rich clay with δ 56 Fe values of −0.13‰ to +0.16‰) via chemical deposition. Subsequently, Fe-depleted and Al-enriched soils were transported (mainly as riverine SPM) to the continental margin forming bauxite beds overlying Fe beds.
This study documents the release of Fe into the ocean during the pedogenic stage of the bauxite mineralization in the early Carboniferous period. The deposition of Fe beds prior to bauxite beds in central Guizhou further strengthens our interpretation. The negative correlation between Al 2 O 3 and Fe 2 O 3 in global karst bauxites suggests that Fe loss during karst bauxite formation was a common effect and that the karst bauxite could act as a reliable indicator of additional Fe to the ocean during continental weathering. Humid climate periods of extensive karst bauxite formation would have been characterized by an additional efflux of Fe to the ocean, particularly during the Carboniferous, Permian, Cretaceous periods, and Cenozoic eras. This additional Fe supply may have had a critical impact on the oceanic Fe cycle and marine ecosystems in the early Carboniferous period.