Chemostratigraphic Analysis of Wufeng and Longmaxi Formation in Changning, Sichuan, China: Achieved by Principal Component and Constrained Clustering Analysis

The increasing proportion of unconventional worldwide energy demands have consistently promoted the necessity for exploring a precise, high-resolution, objective, and quantitative stratigraphic division method for macroscopically homogeneous mudstone successions. The chemostratigraphy can resolve this problem well, although it has been applied successfully in North America, but not systematically studied in China for shale gas exploration and development. This work has conducted a chemostratigraphic analysis of Wufeng and Longmaxi Formation on the Changning section of Sichuan Province, southwestern China, to testify its applicability for shale gas exploration in China. Principal component analysis (PCA) was first employed to reduce the dimensionality of datasets. Three chemofacies, including detrital (K, Ti, Fe, Al, Na, Mg, Cr, Zr, Rb), authigenic (Ca, Sr, Mn, Si, S, Ba), and redox-organic (P, V, Ni, Zn, Cu, TOC), were found. Subsequently, constrained clustering analysis was utilized for the zonation of each chemofacies into chemozones. Consequently, the whole Changning section was divided into twelve chemozones (CZ I–CZ Ⅻ). The geochemical interpretation for these chemozones can be resolved from the regional changes in paleogeography and paleoceanography during the Late Ordovician to Early Silurian period. Thus, a three-stage geochemical evolution along the Changning section can be classified: (1) the siliceous and anoxic deposits of Wufeng Formation (CZ I–CZ III) with high TOC contents; (2) the siliceous and anoxic sedimentary rocks of bottom Longmaxi Formation with still higher TOC (CZ Ⅳ); (3) the calcarous-detrital and oxic sediments for the rest of Longmaxi Formation (CZ Ⅴ–CZ Ⅻ). In considering their high content of TOC and abundant brittle siliceous minerals, the CZ (I–Ⅳ, 0 m–33.6 m) are thought to be the most preferable sweet spot for shale gas exploration.


Introduction
The production of unconventional energy, especially shale gas, has increased dramatically in recent years to satisfy the growing demand for energy worldwide [1]. The precise and high-resolution stratigraphic division and correlation of mudstone successions has become one of the most important steps in exploration and development of shale gas [2]. However, the macroscopically homogeneous attributes and limited fossil constraints of shales hinder the availability of biostratigraphy and lithostratigraphy [3]. Moreover, the lack of seismic, drilling, and logging data [4] as well as the difficulty in identifying distinct facies shifts [5] within such rocks disable the efficiency of seismic, well logging, and sequence stratigraphic techniques [6][7][8][9][10]. New methodologies are needed to give a reliable and high-resolution stratigraphic scheme for shale gas exploration. Changning section, located in Changning County of Sichuan Province, southwestern China, is situated at the paleo-depocenter of the Wufeng and Longmaxi Formation. This section is one of the most typical profiles for studying Wufeng and Longmaxi shales in China [24,26,33]. Here the Wufeng Formation (O3w) is characterized by laminated and massive siliceous shales with Orthograptus fossils, and the siliceous radiolaria is visible under microscopes [33]. As for the Longmaxi Formation (S3l), its lower parts are comprised of black and carbonaceous shales, with lamina and Orthograptus fossils well developed in the strata. In the middle and upper part of the Longmaxi Formation, silty shales and mudstone are the dominate lithological types with Rastrites fossils at the top, which indicates a much shallower deposition condition [26,33].

Materials
In this study, we took the Changing section as the research subject, which is tectonically located in the Upper Yangtze area. Wufeng and Longmaxi Formation outcrop well in this section, with a thickness of more than 300 m. In total, 87 geochemical samples, with a majority sampling interval of approximately 0.5 m, were collected from Wufeng to Longmaxi Formation. The total sampled thickness was about 170 m and the biostratigraphy of Wufeng and Longmaxi Formation was considered according to [34]. Surface layers were excavated out to ensure that fresh samples were taken without alteration. Then all the samples were ground into powder for geochemical analysis.
The contents of main and trace elements of samples were measured by Wavelength Dispersive X-ray Fluorescence Spectrometer (XRF) in Microstructure Analysis Laboratory of Beijing. The major elements, including K, Ti, Fe, Ca, Mn, Si, Al, P, Na and Mg, are measured as a percentage of metallic oxides, and the trace elements, including V, Cr, Ni, Zn, Cu, Sr, S, Zr, Rb, Ba and Cl, are expressed in mg/kg. Calibration was done with 25 rock  [26]).

Materials
In this study, we took the Changing section as the research subject, which is tectonically located in the Upper Yangtze area. Wufeng and Longmaxi Formation outcrop well in this section, with a thickness of more than 300 m. In total, 87 geochemical samples, with a majority sampling interval of approximately 0.5 m, were collected from Wufeng to Longmaxi Formation. The total sampled thickness was about 170 m and the biostratigraphy of Wufeng and Longmaxi Formation was considered according to [34]. Surface layers were excavated out to ensure that fresh samples were taken without alteration. Then all the samples were ground into powder for geochemical analysis.
The contents of main and trace elements of samples were measured by Wavelength Dispersive X-ray Fluorescence Spectrometer (XRF) in Microstructure Analysis Laboratory of Beijing. The major elements, including K, Ti, Fe, Ca, Mn, Si, Al, P, Na and Mg, are measured as a percentage of metallic oxides, and the trace elements, including V, Cr, Ni, Zn, Cu, Sr, S, Zr, Rb, Ba and Cl, are expressed in mg/kg. Calibration was done with 25 rock reference materials (GBW07101-07125) certified by the Chinese Academy of Geological Sciences. Total organic carbon (TOC), expressed in percentage by weight (%), was measured in a total carbon and total nitrogen analyzer of Jena TOC/TN Multi NC2100S, in the laboratory at Academy of Science, China University of Geosciences at Beijing. The whiting was used to calibrate the machine. For convenience, various of units are expressed as mg/kg to avoid the multivariate statistical analysis error [20]. The datasets for the geochemical measurements are documented in Table A1.

Multivariate Statistical Techniques
The implementation of chemostratigraphy is never easy work, though it is, without doubt, an effective method for shale gas exploration [11]. The most important step in the chemostratigraphy is how to realize the zonation of the geochemical data/curves (when presented in a figure) in a reasonable way which results in a reliable and interpretive stratigraphic scheme for the studied succession [9][10][11]13,14,18,20,35]. Furthermore, the compositional data of these successions are high-dimensional in nature which makes the work even more difficult [10]. For example, the dataset of Changning section, presented here and determined by XRF, includes more than 20 elemental contents of 87 samples. Traditionally, the zonation of geochemical data/curves has been conducted in an artificial way which was subjectively affected by the experience and knowledge of the operator. Then, automatic zonation by software programs, especially multivariate statistical techniques, was introduced [36]. However, the analysis of chemostratigraphic data is not without principle when utilizing multivariate statistical techniques [37]. In this work, the dimension of the chemostratigraphic dataset needs to be reduced firstly to the items of the data combined with new variables of specific geochemical meaning by principal component analysis (PCA), and then the zonation of the new variables, for example, by constrained clustering analysis, can be conducted. In this sense, these steps are quite similar to those of zonation of well-logging curves [38]. Detailed steps of chemostratigraphic analysis are elaborated below.

Principal Component Analysis
Principal component analysis (PCA) is a commonly used technology in many disciplines for reducing the dimensions of a dataset [37,39]. The philosophy of this technique is to transform the massive variables of the dataset into several principal components while preserving the majority of the information of the original dataset [39]. For sedimentary geochemical data reduction, this method can group the elements into several principal components, and the geochemical meaning of each component is defined and interpreted by the respective elemental association [40]. Usually, the geochemical behavior of each element association is somewhat different and therefore can define certain geochemical properties of the sedimentary environment, i.e., each principal proponent/element association may stand for a certain type of chemical facies (chemofacies) of sedimentary environment according to the definition [18,36,41]. For example, Montero-Serrano et al.
(2010) grouped the whole-rock elemental data using PCA into the detrital, authigenic, and redox chemofacies, in order to conduct chemostratigraphic analysis [20]. In a geochemical perspective, the composition of fine-grained sedimentary rocks usually come from detrital, authigenic, redox, bioproductivity, etc., which can be identified by the PCA [42,43]. In this sense, we think the data reduction process of the chemostratigraphic dataset is the same process for the identification of chemofacies by PCA [44,45].

Constrained Clustering Analysis
The constrained clustering analysis is a multivariate statistical approach for sorting the data plotted in a higher-dimensional space without disorganizing their original sequences based on the degree of similarity quantified by the distance between two adjacent 'clusters' [10]. For the geological samples, they are arranged according to the stratigraphic depth (in meters) in the dendrogram which vividly displays the result of constrained clustering analysis in the form of a graphic [38]. 'Manhattan Distance' and 'Euclidian Distance' are the two most commonly used methods to measure distance. Considering that possible correlations may exist among variables of rock samples, 'Euclidian Distance' is thought to be more proper for its assumption that certain dependence is allowed among variables [10], thus, the length of the straight line between two clusters is employed to quantify the degree of similarity. Each cluster is represented by the center of the cluster, which means that the 'K-means method' is used. Then the newly merged clusters are defined using 'Ward's method' or the minimum variance of the original cluster, which maximize the difference between two clusters and minimize the difference within clusters [5,9,46,47]. Accordingly, 'K-means method', 'Euclidean distance', and 'Ward's method' were chosen for constrained clustering analysis to realize the division of succession. Furthermore, the ultimate number (a) Chemofacies 1 (K, Ti, Fe, Al, Na, Mg, Cr, Zr, Rb) mainly consists of elements that originated from the provenance region of sediments. Among these elements, K, Al, Rb, and Fe are the main chemical components of fine detrital minerals [51,52], in particular clay minerals and iron oxides. Cr is vulnerable to be absorbed by clay minerals, thus it is preferable to be preserved in finer parts of mudstone [42]. On the contrary, Na, Mg, Ti, and Zr may be related to the coarse component of terrestrial inputs and hosted in feldspar, rutile, and zircon [42]. Consequently, this chemofacies can reflect both the proportion of detrital materials and even the ratio of coarse: fine minerals within rocks, and ultimately be interpreted as detrital chemofacies (DT).
Energies 2021, 14, 7048 6 of 21 minerals, thus it is preferable to be preserved in finer parts of contrary, Na, Mg, Ti, and Zr may be related to the coarse co inputs and hosted in feldspar, rutile, and zircon [42]. Consequ can reflect both the proportion of detrital materials and even t minerals within rocks, and ultimately be interpreted as detrita It is well-known that element Ca and carbonate minerals in marine settings. Mn can occur as MnC reduction of manganese oxides in the reduction environment sistent with the deposition environment indicated by the blac relation between barium and sulfur within this association ma ments forming the barite as the organic matter were oxidized sulfur can also deposit in the form of pyrite [51]. However, th of this association is that element Si shows a high negative lo multiple sources, such as terrestrial input and biogenic influx. from the thin section under microscopes, the silicon content of may be primarily made up of the biogenic production of silic can be approved by the occurrence of radiolarias, indicating a n [26,42]. (c) Chemofacies 3 (P, V, Ni, Zn, Cu, TOC) is comprised of red which are closely related to the productivity and preservation can be interpreted as redox-organic chemofacies (RO). This as meanings, the first one refers to the nutrient-related elements other one is the redox-sensitive elements which will be abso under reduction bottom water conditions, such as V and Ni [ the comprehensive product of these two aspects. (b) Chemofacies 2 (Ca, Sr, Mn, Si, S, Ba) is mostly comprised of authigenic elements sourced from carbonates and siliceous depositions and can be regarded as an authigenic chemofacies (AT). It is well-known that element Ca and Sr are associated with carbonate minerals in marine settings. Mn can occur as MnCO 3 , resulting from the reduction of manganese oxides in the reduction environment [51,53], which is consistent with the deposition environment indicated by the black shales. The high correlation between barium and sulfur within this association may imply that both elements forming the barite as the organic matter were oxidized by the sulfate. Besides, sulfur can also deposit in the form of pyrite [51]. However, the most striking feature of this association is that element Si shows a high negative loading. Si often shows multiple sources, such as terrestrial input and biogenic influx. Based on the evidence from the thin section under microscopes, the silicon content of the Changning section may be primarily made up of the biogenic production of siliceous organism, which can be approved by the occurrence of radiolarias, indicating a nutrient-rich condition [26,42]. (c) Chemofacies 3 (P, V, Ni, Zn, Cu, TOC) is comprised of redox-sensitive elements which are closely related to the productivity and preservation of organic carbon, thus can be interpreted as redox-organic chemofacies (RO). This association contains two meanings, the first one refers to the nutrient-related elements such as P, Zn, Cu; the other one is the redox-sensitive elements which will be absorbed into depositions under reduction bottom water conditions, such as V and Ni [41,[54][55][56][57][58][59]. The TOC is the comprehensive product of these two aspects.
The total variance explained by these three principal factors/element associations/ chemofacies is 63.21%, 30.76%, and 5.05%, respectively. This means that the variations of chemofacies of DT, AT, RO (PC1, PC2, and PC3) can represent the main basis for characterizing the chemostratigraphy of Wufeng and Longmaxi Formation in the Changning section [60].

Zonation of Chemofacies
Once the main chemofacies (element associations) for the section were established, the constrained clustering analysis could be employed to obtain a number of distinctive chemozones. The zonation of chemofacies resulted from the dendrograms and depended on the selection of cut-off value [38]. The choice of this value must be based on regional geological settings. Accordingly, 690, 1700, and 810 cut-off values were chosen, respectively, for the zonation of the detrital, authigenic, and redox-organic chemofacies. Three dendrograms derived from the constrained clustering analysis based on chemofacies can be seen in Figure A1. Therefore, the zonation results of the three chemofacies were determined as follows:

Zonation of Chemofacies
Once the main chemofacies (element associations) for the section were establish the constrained clustering analysis could be employed to obtain a number of distinct chemozones. The zonation of chemofacies resulted from the dendrograms and depend on the selection of cut-off value [38]. The choice of this value must be based on regio geological settings. Accordingly, 690, 1700, and 810 cut-off values were chosen, resp tively, for the zonation of the detrital, authigenic, and redox-organic chemofacies. Th dendrograms derived from the constrained clustering analysis based on chemofacies c be seen in Figure A1. Therefore, the zonation results of the three chemofacies were det mined as follows:

Discussion
Wufeng and Longmaxi Formations are always the most important successions for shale gas exploration in China, and many works have been done on the Changning section [22][23][24]. Wang [27,62] considered the geochemical characteristics of the Longmaxi Formation, which also indicated that the lower Longmaxi was deposited under anoxic conditions and showed higher potential for organic matter preservation than that of the middle and upper Longmaxi. However, the recognition of optimal shale gas layers is still descriptive and qualitative in all these listed works. Thus, this work aimed to resolve this problem in a more quantitative manner based on the chemostratigraphic technique to promote shale gas exploration in Wufeng-Longmaxi Formations.

Discussion
Wufeng and Longmaxi Formations are always the most important successions for shale gas exploration in China, and many works have been done on the Changning section [22][23][24]. Wang [27,62] considered the geochemical characteristics of the Longmaxi Formation, which also indicated that the lower Longmaxi was deposited under anoxic conditions and showed higher potential for organic matter preservation than that of the middle and upper Longmaxi. However, the recognition of optimal shale gas layers is still descriptive and qualitative in all these listed works. Thus, this work aimed to resolve this problem in a more quantitative manner based on the chemostratigraphic technique to promote shale gas exploration in Wufeng-Longmaxi Formations.

The Detrital Chemofacies
As it is shown in Figure 3, the fluctuations of detrital associations in the section coincided with each other but are opposite to that of carbonaceous-siliceous elements ( Figure 4). For example, at intervals of 52.1 m-41.9 m, the contents of K, Ti, Al, Rb and Cr show a sudden decreasing trend (Figure 3), while the content of Ca shows a corresponding increasing excursion (Figure 4). This situation indicates that the two components are the basis for the chemostratigraphic analysis of the section, as stated above. The detailed comparison between division results of detrital chemofacies and changing trends of these detrital elements along the section are shown below: (1) The contents of the detrital elements of DT I (0 m-3.0 m) are generally low and stable, indicating constantly low terrigenous inputs during this period; (2) At DT II (3.0 m-4.3 m), the content of the detrital association is still low, but with a slowly increasing tendency, representing a low but progressive increase in debris input; Combined with the high contents of carbonate-related elements during this interval (see Figure 4), it is believed that the terrigenous influxes are high and affected by carbonate dilution of seawater during this period; (9) At DT IX (115.7 m-170 m), the element contents of detrital return to a relatively high and stable level, indicating no significant changes in terrigenous inputs compared with that at DT VIII.

The Authigenic Chemofacies
The authigenic chemofacies are principally of authigenic origin from carbonate and siliceous deposition. Ca, Mn, Sr, S, Ba reflect the calcareous minerals within the strata. Si is mainly derived from the biogenic input, though some belong to detrital sources [26,42]. It is worth mentioning that the minus loading component (−0.88 loading, detailed in Table 1) of Si in this association means mutual dilution between calcareous and siliceous matters, which indicates the alternative deposition of these two kinds of minerals, signifying deposition environments with different nutrient levels. Eutrophic conditions are usually for siliceous deposits, and carbonate producers prefer more oligotrophic waters [26,42].
To assess the effect of a determination of chemozone based on the authigenic element association on the stratigraphic division, the detailed correlation between chemozone and changing trends of authigenic association on the section is bound to be implemented: (4) Later, the contents of Ca, Mn, Sr are sustained at a very low level at the lower part of AT IV (8.8 m-33.6 m) and then increase gradually to the upper part, while the content of Si shows a completely contrary trend. It is noteworthy that the trend of Ba and S is the same as that of Si at this chemozone. In consideration of the paleo-productivity implication of Ba, the positive correlation between Ba and Si suggests that biogenic silica producers must be the main contribution to the organic production under nutrient-rich conditions, which is consistent with the significant marine transgression during Rhuddanian [31];  (Figure 3). The variation of Si is the same as that of detrital input as a whole, indicating no significant biogenic silica input (Figure 3).

The Redox-Organic Chemofacies
Among this element association, V, Ni, Zn, Cu mainly accumulate under reduction conditions, which makes these elements important redox proxies [51,56]. For example, V/Cr ratio is a good redox indicator, with V/Cr < 2, 2 < V/Cr < 4.25 and V/Cr > 4.25 corresponding to oxic, suboxic, and anoxic conditions, respectively [63]. Element P and TOC are two of the most important paleo-productivity proxies for ancient lakes and oceans [54,55].
According to the V/Cr ratio in Figure 5, the lower part (0 m-33.6 m) of the Changning section was deposited under an anoxic-suboxic condition, while the oxic condition was dominant during its later interval (33.6 m-170 m). The fact that the relatively high contents of redox-sensitive elements, TOC and P, were in the lower part of the section all together suggests the same conclusion. The zonation results of redox-organic chemofacies and changing trends of element association on the section were compared in detail:  6 m-170 m), the content of redox-organic element association maintains at a lower level than that of RO IV, which indicates that this chemozone was deposited under the oxic and low-paleoproductivities condition. (6) The chemozones results of the three chemofacies described above all show an abrupt shift at the horizontal depths of 33.6 m and 8.8 m, which indicates that variations in sedimentation condition indeed occurred before and after these depths.

The Unified Chemostratigraphic Scheme and Paleoenvironmental Interpretation
Based on the zonation of each chemofacies, the unified chemostratigraphic scheme can be acquired for the Wufeng Figure 6). This unified chemostratigaphic scheme is easy to understand because the geochemical meaning of each chemozone has been explained in detail in previous sections.

Identification of the Sweet Spot
Chemostratigraphy has played an important role, both for the exploration and development of shale gas, which provides not only the basis for stratigraphic correlation, but also the stratigraphic properties for evaluation [10,13,14]. Certain statistical surveys for the application of this method prove that 70% of the information needed for shale gas exploration can be provided by chemostratigraphic data [2]. For example, it can guide the conduction of horizontal drilling and provide much information for the fracturing of wells [3].
For sweet-spot identification, the basic protocol is that the rock should be both or- . It is fascinating that the geochemical evolution can coincide approximately with the biostratigraphic stage as well [34], i.e., CZ (I-III) for Late Katian to Hirnaian, and CZ IV for Rhuddanian and CZ (V-XII) for Aeronian, respectively ( Figure 6). The interpretation of these evolution stages can be decoded from the perspective of paleogeographic and paleoceanographic evolution of this region during the Late Ordovician to Early Silurian period [64].
During the Late Ordovician to Early Silurian period, this region was located on the continental shelf of Yangtze platform with a relatively restricted/semi-closed deposition condition and was surrounded by several paleouplifts in the north and south due to the Caledonian orogeny [26,27] (see Figure 1 for details). Only the northeastern-southwestern wards opened to the ocean [28]. The sea level was relatively high, and the sedimentary environment was euxinic, which leads to the deposition of organic-rich Wufeng Formation (CZ I-CZ III) during the Katian period. The sea level then decreased quickly due to the global glaciation at Hirnantian, with the occurrences of shallow-water calcareous facies of Guanyinqiao Layers (top of CZ III).
Marine transgression occurred due to the melting of polar glaciers and tectonic activities at the Yangtze platform [27,65] during the Rhuddanian period. This region revolutionized into a deep shelf; thus, under this anoxic environment the bottom Longmaxi Formation (CZ IV), characterized by high TOC and biogenic silica content, was deposited [64]. This stage was short, but the scale of marine transgression was so large that most of the Yangtze platform was influenced, which makes it possible for the wide distribution of organic-enrich deposition on the lower Longmaxi Formation.
During early Aeronian, the uplift of Yangtze platform, together with marine regression, turned this region into a shallow shelf [27]. The sedimentation environment converted into oxic, thus, carbonate deposition resumed (Figures 4 and 5). The mutual dilution between detrital influx and carbonate minerals was evident, and ratio of fine to coarse detrital inputs was regulated by the fluctuation of sea level at this stage (CZ V-CZ XII, Figures 3 and 4). The TOC content was not high either due to the low preservation an productivity [64].

Identification of the Sweet Spot
Chemostratigraphy has played an important role, both for the exploration and development of shale gas, which provides not only the basis for stratigraphic correlation, but also the stratigraphic properties for evaluation [10,13,14]. Certain statistical surveys for the application of this method prove that 70% of the information needed for shale gas exploration can be provided by chemostratigraphic data [2]. For example, it can guide the conduction of horizontal drilling and provide much information for the fracturing of wells [3].
For sweet-spot identification, the basic protocol is that the rock should be both organicrich and brittle. The TOC content is the best proxy for organic carbon, and the content of brittle minerals, for example quartz and carbonate, can be concluded from the geochemical data. Undoubtedly, the organic-rich chemozones of CZ I-CZ III (Wufeng Formation) and CZ IV (bottom of Longmaxi Formation), deposited under the anoxic environment, are thought to be potential sweet-spot layers for shale-gas exploration. Moreover, these successions belong to siliceous sediments, as well there being a thin carbonaceous layer of Guanyinqiao at the top of CZ III (Figure 4). The Si-Zr plot shows that there are two trends for Si, i.e., biogenic (Figure 7a,b) and detrital (Figure 7c). Zr is a typical detrital proxy, and the negative direction of Si-Zr relation shows a biogenic silica [42]. From this plot, one can find that most of the samples from CZ I-CZ IV are scattered along the biogenic trend line, indicating a significant contribution from the biogenic influx which is similar to that of typical shale gas in North America [3,9,13,14]. Namely, numerous brittle minerals are contained in successions, which is in favor of fracturing. Thus, chemozones of CZ I-CZ III (Wufeng Formation) and CZ IV (bottom of Longmaxi Formation) are both organic-rich and brittle, which are beneficial sweet-spot layers for shale gas exploration. All this identification has been proven by the exploration and developed experience for shale-gas study of these two Formations [66]. two trends for Si, i.e., biogenic (Figure 7a,b) and detrital (Figure 7c). Zr is a typical detrital proxy, and the negative direction of Si-Zr relation shows a biogenic silica [42]. From this plot, one can find that most of the samples from CZ I-CZ Ⅳ are scattered along the biogenic trend line, indicating a significant contribution from the biogenic influx which is similar to that of typical shale gas in North America [3,9,13,14]. Namely, numerous brittle minerals are contained in successions, which is in favor of fracturing. Thus, chemozones of CZ I-CZ III (Wufeng Formation) and CZ Ⅳ (bottom of Longmaxi Formation) are both organic-rich and brittle, which are beneficial sweet-spot layers for shale gas exploration. All this identification has been proven by the exploration and developed experience for shale-gas study of these two Formations [66].

Conclusions
The chemostratigraphic technique presented in this paper has provided a convenient and reliable method for division of fine-grained successions. Two steps are involved for the implementation of this technique. Principal component analysis is firstly used for recognition of chemofacies defined by each elemental association. Then the unified chemozones can be acquired by the combination of chemozones from every chemofacies resulting from constrained clustering analysis. The main conclusions of this work are below: 1. Three chemofacies (or elemental associations) were proposed to optimally characterize the macroscopically homogeneous Wufeng-Longmaxi shales. The elemental association characterizing detrital inputs of the strata are K, Ti, Fe, Al, Na, Mg, Cr, Zr, Rb; the elemental association characterizing authigenic minerals are Ca, Sr, Mn, Si, S, Ba; the chemofacies characterizing redox-organic conditions are P, V, Ni, Zn, Cu, TOC. 2. The macroscopically homogeneous Wufeng-Longmaxi Formation shows geochemical heterogeneity from the perspective of detrital, authigenic, and redox-organic chemofacies, which enables the wide applicability of chemostratigraphy to the division and correlation of fine-grained strata. These chemozones not only correspond to the lithostratigraphic column perfectly but identify the boundaries that cannot be revealed by the latter. 4. A three-stage geochemical evolution along the Changning section was classified, coinciding approximately with the biostratigraphic stages [34], i.e., CZ (I-Ⅲ ) for Late Katian to Hirnaian, CZ Ⅳ for Rhuddanian, and CZ (Ⅴ-Ⅻ) for Aeronian, respectively

Conclusions
The chemostratigraphic technique presented in this paper has provided a convenient and reliable method for division of fine-grained successions. Two steps are involved for the implementation of this technique. Principal component analysis is firstly used for recognition of chemofacies defined by each elemental association. Then the unified chemozones can be acquired by the combination of chemozones from every chemofacies resulting from constrained clustering analysis. The main conclusions of this work are below:

1.
Three chemofacies (or elemental associations) were proposed to optimally characterize the macroscopically homogeneous Wufeng-Longmaxi shales. The elemental association characterizing detrital inputs of the strata are K, Ti, Fe, Al, Na, Mg, Cr, Zr, Rb; the elemental association characterizing authigenic minerals are Ca, Sr, Mn, Si, S, Ba; the chemofacies characterizing redox-organic conditions are P, V, Ni, Zn, Cu, TOC.

2.
The macroscopically homogeneous Wufeng-Longmaxi Formation shows geochemical heterogeneity from the perspective of detrital, authigenic, and redox-organic chemofacies, which enables the wide applicability of chemostratigraphy to the division and correlation of fine-grained strata.

4.
A three-stage geochemical evolution along the Changning section was classified, coinciding approximately with the biostratigraphic stages [34], i.e., CZ (I-III) for Late Katian to Hirnaian, CZ IV for Rhuddanian, and CZ (V-XII) for Aeronian, respectively ( Figure 6). Paleogeographic and paleoceanographic evolution of this region during the Late Ordovician to Early Silurian period were responsible for this evolution. 5.
The organic-rich and brittle CZ I-IV were identified as preferable sweet-spot layers, as confirmed by the exploration experience.