In this study, 12 cores (Table 1) were drilled in the HZB during April-May 2022 to reveal the features of Quaternary sediments in the basin. Combined with three published drilling cores, the lithological chart shows that the Quaternary sediments are generally approximately 15–25 m thick, and the overall succession can be divided into three units from the bottom to the top based on the lithology as follows (Figure 2): 1) Unit I, Q₃ ^a fluvial sediment; the lithology is grayish white to yellow gravel sand, sand or clay sand, underlain by weathered pre-Cenozoic bedrock. The thickness of this unit is generally larger than 10 m. 2) Unit II, Q₃ ^b aeolian sediment, mostly consists of loess-like silt, with occasional mottled clay. The thickness of this unit varies in the basin, mostly between 2 and 10 m. 3) Unit III, Q₄ fluvial or marsh sediments; it mainly consists of gray to yellow coarse to fine sand, silt and clay. The organic carbon content of this unit is generally higher than that of Unit I. In particular, a mottled clay structure with 1–5 m thick was found in seven of the 12 drilling cores in the HZB. The mottled clay exhibits a mixed yellow, red and grey-white color, and the bright clay generally shows vein- and/or flame-like vertical structures intruding into yellow/red sediments (Figure 3B). The deformation features were clearly distinguished from the underlying loess-like silt or fluvial sand deposits (Figure 3). The normal aeolian sediment is homogeneous in color and grain size (Figure 3A), and fluvial sand deposits present pronounced horizontal features (e.g., laminae).

To constrain the age of the Quaternary stratigraphy in the HZB, we collected ten samples from the drilling cores for AMS ¹⁴C dating. Bulk organic sediment samples were pretreated using the standard acid-base-acid (ABA) method and analyzed at the Qingdao National Laboratory for Marine Science and Technology, China. The dates were then calibrated to calendar years (i.e., cal BP) using the IntCal20 Data Sets (Reimer et al., 2020) (Table 2). Additionally, the optically stimulated luminescence (OSL) dates provided in a previous study (Li et al., 2021) were combined to constrain the age of the lower fluvial deposit (Figure 2), which is generally beyond the dating limit of the AMS ¹⁴C method.

In this study, we conducted μ-XRF and DRS analysis to reveal detailed geochemical element and iron oxide features of mottled clay structure in the HZB. Three thin sections of mottled clay structure (10 cm in diameter and dozens of centimeters in length) (Table 3) were collected for μ-XRF scanning, using an M4 Plus Micro Area X-Ray Fluorescence Analyzer at Guangzhou Tuoyan Analytical Technology Co. Ltd. The instrument is equipped with a 20 μm diameter polycapillary X-ray lens and two X-Flash silicon drift detectors and operated at a voltage of 50 kV and a current of 300 μA. The scanning was carried out at a 20 μm spot size, 5 ms acquisition time and 35 μm points distance. Original data were processed using M4 tornado software to analyze the spectral peak information, and derive the element surface distribution map.

Four samples with different lithologies (Table 3) were ground in an agate mortar and passed through a 200-mesh sieve prior to DRS analysis. In particular, the red and white fractions of mottled clay were separated into two separate samples. DRS analysis was conducted using a Perkin Elmer (U.K.) ultraviolet-visible near-infrared spectrophotometer with a wavelength range of 175–3,300 nm at the Instrumental Analysis & Research Center of Sun Yat-sen University. Data for the visible light range of 400–700 nm were collected at 1 nm intervals and used as the first-order derivative of the peak height mapping.

Furthermore, we collected samples from different lithological layers of core HZ-ZK06 for grain size analysis. The grain size was determined using a Malvern (U.K.) Mastersizer 2000 laser particle size instrument (measurement error <2%). Prior to measurement, the samples were pretreated with 10% H₂O₂ and 10% HCl to remove organic matter and carbonates, respectively. Then samples were washed with distilled water, sieved through a 1,000 μm mesh, and dispersed with 0.05 mol/L sodium hexametaphosphate (NaPO₃)₆ solution. Before the analysis, the samples were placed in an ultrasonic vibrator for 20 min.

The results suggest that the unconsolidated sediment in the HZB has mostly developed since the late Pleistocene, directly overlying the weathered pre-Cenozoic bedrock. The radiocarbon dating results of the samples from HZ-ZK01, HZ-ZK04, HZ-ZK07, and ZH-ZK12 suggest that the uppermost Unit III (fluvial or marsh sediments) was mostly deposited during the Holocene (Table 2). The ages of the samples from Unit II (aeolian sediment) ranged between 20 and 35 ka BP, which is consistent with the ages of the yellow silt in the PRD. Wang et al. (2018a, b) demonstrated that the aeolian deposits found in most Quaternary basins of Southeast China developed during the Last Glacial period and were predominantly deposited during the Last Glacial Maximum (LGM). However, the ages of the underlying fluvial sediments (Unit I) exceeded the beyond the he dating limit of the AMS ¹⁴C method. A previous study using the OSL dating method (Li et al., 2021) suggests that Unit I could be older than 103 ± 10.5 ka (Figure 2).

Grain size analysis is one of the most popular methods used for the analysis of sediments and helps determine the transport agents and sedimentary environment (Xu et al., 1992). The grain size distribution characteristics of the samples from core HZ-ZK06 generally became finer from the bottom to the top (Figure 4B). Specifically, the frequency distribution curves of the fluvial deposits have a unimodal or multimodal form biased towards the coarse-grain end. The primary modal size of the lower fluvial sample (18.4 m) was around 85 μm, while those of the upper fluvial samples (13.8, 15.8, and 17.8 m) were around 75 and 9 μm, respectively. The yellow silt samples (2.7 and 8.5 m) present a uniform unimodal distribution with a fine tail and sharp kurtosis. The modal size of yellow silt is about 5 μm. The curves of yellow silt in the HZB decreased abruptly at the coarse-grained end and more gradually at the fine-grained end, similar to those of the loess-like yellow silt in the PRD (Figure 4B) and typical Chinese loess (Wang et al., 2018a). However, the modal size of yellow silt in the HZB is finer than those in the PRD (∼16 μm), indicating wind sorting and decreasing transporting capacity. Nevertheless, our results suggest that the yellow silt in the HZB shares a common sedimentary type (aeolian) with loess-like yellow silt in the PRD and typical loess in China.

The mottled clay in core HZ-ZK06 occurred as an interbed within a thick layer of homogeneous yellow silt (Figure 4A), showing a deformation structure. The frequency distribution curves of the samples have a typical bimodal distribution, with a primary modal size of 80–90 μm and a secondary modal size of about 5 μm. Interestingly, the two distinct modal sizes were identical to those of the yellow silt and underlying fluvial sand, respectively. The finer part (<20 μm) of the curves is also similar with the curves of yellow silt. This implies that the mottled clay may be an admixture of fluvial sand and yellow silt.

μ-XRF is a non-destructive method offering excellent spatial resolution and high sensitivity for small geochemical element concentrations (Hoehnel et al., 2018). In this study, three thin sections of mottled clay showed a heterogeneous spatial distribution of geochemical elements. Generally, the red/yellow fraction of the mottled clay had high concentrations of Fe and Cr, whereas the gray/white fraction had high Si, Al and K concentrations (Figure 5, Figure 6, Figure 7). In particular, the visual structures of samples can be characterized by the varying concentrations of Fe, because the redness of the sediment is dominantly controlled by the type and concentration of iron oxides (Sun et al., 2011). Moreover, the core HZ-ZK04 sample showed a fragmented structure in the spatial distribution of elements (Figure 5B). In core HZ-ZK05, a small white sand vein was characterized by extremely low Fe concentrations, intruding into yellow silt with high Fe concentrations (Figure 6B).

Discriminating iron oxides is of great environmental significance; for example, they are indicators of the sedimentary environment (Zhang et al., 2009). In the subtropical region, hematite and goethite are the two main iron oxide minerals (Wang et al., 2018b), the contents of which in the sediment can be obtained from the DRS first-order derivative curves. The characteristic peak of hematite is typically observed at a wavelength of 565 nm, whereas goethite has two characteristic peaks at 535 and 435 nm (Deaton and Balsam, 1991; Ji et al., 2002). In general, goethite tends to be enriched in wet environments, whereas large amounts of hematite reflect dry environments (Cornell and Schwertmann, 2003; Balsam et al., 2004).

In this study, two peaks at wavelengths of 565 and 435 nm were presented in the loess-like yellow silt (UV-01) and the red fraction of mottled clay (UV-02) samples (Figure 8). The peak at 565 nm is substantially greater than that at 435 nm, which indicates that hematite predominates over goethite in the loess-like yellow silt and the red fraction of mottled clay. Due to overall low iron content as revealed by μ-XRF analysis (Figures 5–7), the DRS curve of the white fraction of mottled clay (UV-03) is relatively flat, and three peaks could be discerned at 435, 505 and 555 nm, respectively (Figure 8). The hematite peak at 565 nm is usually offset to 555 or 575 nm after being affacted by the Fe content, and the characteristic peak of goethite at 535 nm is often biased toward 505 nm (Deaton and Balsam, 1991; Ji et al., 2002). Therefore, the white fraction of the mottled clay contained both goethite and hematite, with the former being higher than the latter. The DRS curve of the fluvial sand (UV-04) shows two peaks at 435 and 505 nm, indicating the predominance of goethite. Overall, the curves of the loess-like yellow silt and the red fraction of mottled clay (UV-01 and UV-02) share a similar pattern with high hematite content, indicating a relatively dry sedimentary environment (i.e., aeolian deposit). In contrary, samples of the white fraction of the mottled clay and fluvial sediment were predominantly goethite, reflecting a typical aquatic environment.

Mottled clay is primarily regarded as a weathering product of underlying fluvial/marine sediments (Huang, 1982; Li et al., 1984). Based on the evidences from grain size, mineral, and geochemical composition, Wang et al. (2018b) suggested no obvious weathering transition relationship between mottled clay and its underlying sediments, and proposed that the mottled structure was formed by post-sedimentary oxidation modification of homogeneous aeolian dust accumulated during the last glacial period, under the hot and humid conditions of the Holocene epoch. However, in the HZB, the mottled clay layer was not always overlaid with the Holocene organic-rich sediments. Instead, it is also overlaid with late Pleistocene fluvial sand (e.g., in core HZ-ZK12, Figure 7), or occurs as an interbed within the thick layer of homogeneous yellow silt (Figure 4A). Therefore, we suggest that the mottled clay is neither a post-sedimentary oxidation product of the loess-like yellow silt, nor a weathering product of the underlying fluvial/marine sediments. Alternatively, μ-XRF element mapping shows that the red/white fraction of mottled clay has a different geochemical composition (Figures 5–7). Two major components of grain size (5 and 80–90 μm) are identical to the modal sizes of the yellow silt and the underlying fluvial sand, respectively (Figure 4B). In addition, the red fraction of mottled clay had a high hematite content, similar to loess-like yellow silt, whereas the white fraction and the underlying fluvial sediments were dominated by goethite (Figure 8). These two types of iron oxides refer to opposite sedimentary environments (i.e., dry vs. wet conditions), implying that the accumulation of these two fractions was asynchronous. Furthermore, there are diverse deformed structures in the mottled clay, such as fragmented structures (Figure 5B), sand veins (Figure 6B), sand dykes (Figure 7B) and flame structure. The degree of deformation also increases upwards in mottled clay, which is a common characteristic of liquefaction-related deformations (Owen and Moretti, 2011). Therefore, we suggest that the mottled clay structure in the HZB is a product of liquefaction-induced SSD, which produced an admixture of aeolian yellow silt and underlying fluvial sand.

Several key factors of sediment characteristics, including grain size, porosity, saturation, permeability barrier, overburden pressure, and sediment history strongly influence susceptibility to liquefaction (Owen and Moretti, 2011). In most cases, liquefaction develops in loose, water-saturated silt or fine sand without previous liquefaction. A higher overburden pressure could also decrease the potential for liquefaction, such that most liquefaction develops in sediment with a small buried depth (generally less than 5 m) (Obermeier, 1996). In HZB, the underlying fluvial sand is mostly fine, loose, and water-saturated. In the other hand, the overlying yellow silt layer could act as a permeability barrier, leading to an increase in the pore-fluid pressure. These sedimentary characteristics of the HZB generally promote the development of liquefaction. Moreover, the fluvial sand and aeolian yellow silt are significantly different in many sedimentary characteristics, such as grain size, iron oxides, and geochemical (Figures 4–8), which favors the preservation of liquefaction-induced SSDS.

Liquefaction-induced SSDS can be triggered by many factors, which can be categorized into two types (Owen and Moretti, 2011). The first type is directly associated with sedimentation processes and termed autogenic or internal triggers and includes groundwater fluctuations, water flows/waves, rapid sediment loading, tidal shear, tsunamis and permafrost thawing. Other triggers such as earthquakes, meteorite impacts and volcanic eruptions act as external agents (Su et al., 2022), and are referred to as allogenic triggers. Regarding the liquefaction-induced SSDS of mottled clay in the HZB, some triggers could be eliminated by exclusion. First, meteorite impacts and volcanic eruptions can be ruled out, because such events could produce unique geomorphological formations and sedimentary deposits, which have not been found in the study region since the late Pleistocene. The high hematite content in the red fraction of mottled clay indicates a relatively dry sedimentary environment, and in some cases, a mottled clay structure occurs interbedded with aeolian yellow silt (e.g., core HZ-ZK06), which together exclude an underwater sedimentary environment and thus possible triggers of water flows/waves and tidal shear. Third, the Lianhuashan Mountains situated between the HZB and the northern SCS (Figure 1B) could be a barrier weakening the influence of tsunamis and storm surges. Finally, paleotemperature reconstruction from a nearby maar lake indicates a drop of about 6 °C during the Last Glacial Maximum, a period of maximum glacier extent since the late Pleistocene, compared to modern annual temperature (Chu et al., 2017). Considering the present subtropical climate with hot conditions (e.g., 14.5 °C in winter) (Li et al., 2008), the freeze-thaw process was not likely the trigger that led to the formation of SSDS in the mottled clay layer of the HZB.

Instead, liquefaction-induced SSDS in the mottled clay layer of the HZB were interpreted as tectonic activity -induced after consideration of all other possible trigger agents. As a subsidence basin controlled by active faults (Figure 1B), the HZB has experienced tectonic activity since the late Pleistocene. For example, a normal fault cutting a peat bed dated at 33.59 ± 0.16 ka BP was found in the Huizhou fault zones (unpublished data). After homogeneous aeolian dust accumulation, seismic shaking led to a sudden increase in pore-water pressure and triggered the liquefaction of the underlying fluvial sediments. The excessive pore pressure water and sand mixture upwelled and intruded into the overlying loess-like yellow silt layer. This intrusion deformed the homogeneous aeolian layer, formed vertical structures (e.g., sand vein), and resulted in an admixture of aeolian silt and fluvial sand as revealed by the results of grain size, DRS and μ-XRF analysis. Moreover, the spatial distribution of the mottled clay layer is highly related to regional faults (Figure 9). For example, most cores with mottled clay layers were concentrated in the communicating zone of the Tonghu and Huizhou faults, where a historical earthquake with the magnitudes of 4 occurred in October 1590 (Wei et al., 2000). The spatial relationship between the mottled clay layers and regional faults strongly supports the suggestion of tectonic activity -induced liquefaction and SSDS. Notably, the formation of mottled clay structures in the different cores did not occur during a single seismic event. Generally, seismic events tend affect surface sediments (e.g., Shao et al., 2020). Multiple deformed layers in the vertical direction (e.g., two layers in core HZ-ZK12) reflect the repetition of paleo-earthquakes since the late Pleistocene in the HZB.

Seismically induced liquefaction and SSDS are usually recorded in lacustrine (e.g., Becker et al., 2002; Fan et al., 2022), fluvial (e.g., Suter et al., 2011; Rossetti et al., 2017) and aeolian (e.g., Moretti, 2000; Brand and Maithel, 2020) sediments. In the upper Senne area of Northwest Germany, seismically triggered SSDS in mixed alluvial and aeolian deposits that accumulated during the Last Glacial were also recognized (Brandes and Winsemann, 2013). Two outcrops from the Senne area consist mainly of alluvial deposits, overlain by several-meter-thick aeolian deposits, similar to the lithological succession of late Quaternary sediments in the HZB. Different types of SSDS, including closely spaced low offset normal faults, ball-and-pillow structures, flame structures, sills and irregular sedimentary intrusions, dikes, and sand volcanoes, were identified based on a detailed sedimentological analysis (Brandes and Winsemann, 2013). In the coastal region of the northern SCS, aeolian yellow silt broadly developed during the Last Glacial period in coastal basins, mostly underlain by late Pleistocene fluvial or alluvial deposits. This lithological succession is susceptible to seismically induced liquefaction. Therefore, the SSDS of the mottled clay layer could be a potential indicator of paleo-earthquakes in the coastal region of the northern SCS.