For this research, eight riparian dunes along rivers in typical northern China deserts were selected, including the Taklimakan Desert-Keriya River (Figure 1, No.1), Tibetan Plateau sandy land-Maquan River (Figure 1, No.2), Qaidam Basin Desert-Tora River (Figure 1, No.3), Badain Jaran Desert-Heihe River (Figure 1, No.4), Gonghe Basin sandy land-Langqu River (Figure 1, No.5), Hobq Desert-Mu Bulag River (Figure 1, No.6), Mu Us Sandy Land-Kuye River (Figure 1, No.7), and Horqin Sandy Land-Xar Moron River (Figure 1, No.8) from west to east. Their geographic locations are illustrated in Figure 1, and an overview of the study area is provided in Table 1.

Periodic monitoring and sampling at the eight study locations were determined based on the spatial distribution of riparian dunes to conduct a comprehensive investigation. A section at the middle and lower rivers, perpendicular to the river flow direction, was selected using remote sensing images from Google Earth (Table 1). Sediment samples were collected from shallow depths of up to 5 cm from floodplain surfaces and river terrace dunes along the sections. Roughly 500 to 1000 g of dune chain samples were gathered from the dune tops. Detailed documentation was made regarding the sampling locations, surrounding landforms, deposit features, and vegetation coverage. A total of 510 dune surface sediments were sampled from the study sites, with 73, 54, 63, 56, 34, 67, 101, and 62 samples collected from the Keriya River, Maquan River, Tora River, Heihe River, Langqu River, Mu Bulag River, Kuye River, and Xar Moron River, respectively.

Chemical element analysis of bulk sediment samples was conducted at the Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, and the School of Geography and Tourism, Shaanxi Normal University, China. Chemical elements were analyzed using X-ray fluorescence spectrometry, with samples prepared using the pressed pellet method. All samples were air-dried at room temperature and subsequently ground to pass a 75-µm sieve using a grinding mill to meet the sample pressure requirement. Four grams of the sample were oven-dried at 105℃ and placed in a sample preparation mold. Boric acid was used to edge the sample, which was then placed at the bottom of the sample well, and a circular sample with an external diameter of 32 mm was produced under 30 t of pressure. XRF analysis was performed using an Axios Sequential Wavelength Dispersive X-Ray Fluorescence Spectrometer (Panaco, Netherlands), with an error of <5%. Major elements are expressed as oxide percentages; for the calculation of oxide ratios and weathering parameters, the element contents were converted to molar mass percent.

Several chemical weathering indices were calculated in this study. The Chemical Index of Alteration (CIA = Al₂O₃/(Al₂O₃ + CaO* + Na₂O + K₂O) × 100) is employed to evaluate the degree of weathering in the source region and serves as a valuable quantitative index reflecting the weathering process of silicate minerals (Nesbitt et al., 1980; Nesbitt and Young, 1982). In this context, CaO* refers exclusively to the amount of CaO incorporated in silicates, and we utilized the indirect method of Mclennan (Mclennan, 1993), which assumes CaO/Na₂O ratios that are reasonable for silicate material. If CaO ≤ Na₂O in terms of molar fractions, the value of CaO is accepted. However, if CaO ≥ Na₂O, then CaO* = Na₂O. The Chemical Proxy of Alteration (CPA = Al₂O₃/(Al₂O₃ + Na₂O) × 100) (Harnois, 1988; Buggle et al., 2011) and A-CN-K value (Nesbitt and Young, 1996; Chen et al., 2001; Xu et al., 2011) are utilized to indicate the degree of chemical weathering.

Principal Component Analysis (PCA) was conducted to examine the differences in the element characteristics of samples from various regions using the entire multivariate dataset, with Varimax rotation applied (Maćkiewicz and Ratajczak, 1993). Essentially, PCA simplifies high-dimensional datasets by geometrically projecting them onto lower dimensions to obtain the best data summary using a limited number of principal components (PCs). Moreover, the Bartlett test of sphericity and a KMO (Kaiser-Meyer-Olkin) test will also be conducted to confirm the applicability of PCA implementation. All statistical analyses were carried out using SPSS 22.0 and Origin 2021.

Table 2 presents the major element analysis results of various riparian dune samples. The coefficient of variation (CV) of SiO₂ concentration has the lowest average (CV = 0.059), followed by K₂O (CV = 0.086) and Al₂O₃ (CV = 0.089), indicating the relatively high stability of these major elements. Nonetheless, the concentrations of major elements in riparian dunes of the eight watersheds exhibit significant variation (Table 2), with a general trend of lower SiO₂ and K₂O concentrations and higher Al₂O₃, Fe₂O₃, MgO, CaO, and Na₂O concentrations in the west, and the opposite pattern in the east.

In particular, as revealed by Table 2 and Figure 2, the average SiO₂ concentration of riparian samples from the western Keriya and Tora rivers is relatively low, at almost 60%, which is slightly lower than the values of the Upper Continental Crust (UCC) (Taylor and McLennan, 1985). In contrast, samples from riparian dunes along the eastern Mu Bulag and Xar Moron rivers exhibit comparably higher SiO₂ concentrations of nearly 80% and over 85% (the highest value), respectively, which is significantly greater than the UCC values. The concentration of MgO (CV = 0.426) and CaO (CV = 0.271) in the samples varies considerably, with a higher concentration in the western region and a lower concentration in the eastern region. Specifically, the concentrations of MgO in the western riparian dunes of Keriya, Heihe, and Tora rivers are greater than 1.5%, while the concentrations of CaO in the western riparian dunes of Keriya and Tora rivers are greater than 8% and 5%, respectively. In contrast, the average concentration of MgO and CaO in the samples collected from the Eastern Mu Bulag, Kuye, and Xar Moron rivers is significantly lower than that of the western region.

Regarding other major element indicators, according to the relative variation of major elements with respect to UCC ratios (Figure 2), with the exception of SiO₂ concentrations that exhibit a relatively similar and clustered distribution, the other major elements, including Al₂O₃, Fe₂O₃, Na₂O, and K₂O, all exhibit a deficit state. The depletion phenomenon of MgO and CaO concentrations is not entirely the same, with an enrichment of MgO and CaO concentrations in Keriya riparian dunes and an enrichment of CaO concentration in Tora riparian dunes. It is noteworthy that the characteristics of the riparian dunes of the Maquan River in the Tibetan Plateau differ significantly from others, having the highest content of Al₂O₃ and a significant reduction in the concentration of CaO.

The analysis of the correlation among various elements and their ratios in the riparian sand dunes of the desert is crucial to understand the geochemical characteristics of riparian dune surface sediments. This analysis holds significant importance for identifying and classifying the possible source of sediments.

Figure 3a demonstrates that the eight watersheds might be classified into two categories, where the sampling locations of Maquan, Mu Bulag, and Xar Moron rivers are dispersed along a straight line with a strong negative correlation of up to 0.86. In contrast, the Keriya, Tora, Heihe, Langqu, and Kuye rivers are weakly correlated and thus are clustered together. Typically, SiO₂ is primarily found in quartz and silicate minerals, while Al₂O₃ is mostly present in aluminosilicate minerals such as feldspar, mica, and clay minerals. In the eight watersheds of northern China, Si and Al are the most abundant and relatively stable major elements, accounting for approximately 80% (Table 2). The significant negative correlation between SiO₂ and Al₂O₃ indicates the associated quartz, silicate, and aluminosilicate minerals and explains the dilution of quartz to other elements.

The major element ratios exhibit specific characteristics that can be utilized to infer important information about sediment geochemistry. For instance, the K₂O/Al₂O₃ ratio is frequently used as an index of chemical maturity and can also serve as an indicator of the original composition of sediments during the early stages of chemical weathering when Ca and Na are removed (Cox et al., 1995). The SiO₂/Al₂O₃ ratio is widely used as an index in identifying sediment provenance due to its relative stability (Hao et al., 2010). Furthermore, the ratios between active and inert constituents, such as K₂O/Al₂O₃ and MgO/Al₂O₃, can provide crucial insights into the sedimentation environment and source (Wang et al., 2001). The scatter plots of SiO₂/Al₂O₃ vs. CaO/Al₂O₃ (Figure 3b), K₂O/Al₂O₃ vs. Na₂O/Al₂O₃ (Figure 3c), and K₂O/Al₂O₃ vs. MgO/Al₂O₃ (Figure 3d) exhibit clear partition characteristics, indicating significant differences between the West (Keriya, Tora, Heihe, and Langqu rivers) and the East (Mu Bulag, Kuye, and Xar Moron rivers). The light green dash areas in Figures 3c and 3d represent the distinct sediment provenance and partition.

Except for the analysis of various elements and their ratios, Principal Component Analysis (PCA) and cluster analysis were commonly conducted to analyze the major elements of riparian dunes in the eight watersheds. The Bartlett test of sphericity test and KMO test were performed, and their significance values were found to be 0.000 and 0.584, respectively. The results of the principal component analysis, considering the variables with correlation, showed that the cumulative contribution rate of the total variance reached 73.60% for the two principal components with characteristic roots greater than 1.0 (as indicated in Table 3). The primary principal component included silicate and carbonate indices such as SiO₂, MgO, CaO, and Na₂O, while the secondary principal component comprised aluminum and iron silicate mineral indices, including K₂O, Al₂O₃, and Fe₂O₃.

The results of the principal component analysis of riparian dune samples collected from eight desert rivers are illustrated in Figure 4. A significantly positive correlation with a minimum angle can be observed between CaO and MgO in the riparian dune samples of various desert rivers. In contrast, a significantly negative correlation with a minimum angle can be observed between NaO and K₂O. Generally, there is a clear distribution difference between sediments from the West and East, and the Maquan River in the Tibetan Plateau is closely related to the eastern region.

Additionally, we conducted a cluster analysis based on all 510 samples to obtain detailed zoning (Figure 5). Based on the results of the cluster analysis, we selected a suitable system with a distance of around 3 and classified the samples into three major types. Type I includes dune samples from the Keriya, Tora, Heihe, and Langqu rivers (mostly distributed in the West), Type II includes samples from the Maquan River distributed in the Tibetan Plateau, and Type III includes samples from Xar Moron, Kuye, and Mu Bulag rivers (mostly distributed in the East). Thereinto, samples of Mu Bulag River belong to both Type I and Type III, and because of a higher percentage in Type III, here we classify it as Type III.

The three identified types represent different sand dune types in the extremely arid and arid regions in the west, riverbank dunes in the Tibetan Plateau sandy land, and riparian dunes in the semi-arid and semi-humid desert rivers in central and eastern China. The sediments in each type exhibit distinct characteristics due to their spatial distributions (Table 2). In Type I, the samples exhibit lower SiO₂ concentrations and higher CaO concentrations, indicating low quartz content and high carbonate content. Conversely, in Type III, the higher SiO₂ concentrations and lower CaO concentrations indicate the dilution of other elements by excess quartz. In Type II, the samples’ concentrations of K₂O and Al₂O₃ are higher than those of CaO and MgO, indicating K and Al-rich and carbonate-poor aluminosilicate minerals, including feldspar, mica, and clay minerals.

The elemental concentrations in riparian dunes of selected deserts are classified into three groups. SiO₂, CaO, K₂O, and Na₂O are major elements that are closely related to the abundance of quartz, carbonates, feldspar, and mica/clay minerals, respectively. To eliminate the influence of changes in carbonate content, the ratios of other element concentrations to the concentration of Al₂O₃ are used to demonstrate the relative changes in other elements in dunes, overcoming the impact of variations in carbonate fluctuations. The statistical results of the analysis are presented in Table 4.

The SiO₂/Al₂O₃ ratios in dunes reflect the relative change of quartz compared to other silicate minerals. Table 4 shows that the abundance of quartz in different sample types, as evidenced by SiO₂/Al₂O₃ ratios, is in the order of Type III (15.406) > Type I (12.022) > Type II (10.459). The relative variations in Fe₂O₃/Al₂O₃ and MgO/Al₂O₃ ratios reflect the abundance of ferromagnesian silicates, which is in the order of Type I, Type II, and Type III. Riparian dune sediment samples from western areas exhibit a higher abundance of ferromagnesian silicates than those from eastern areas. Similarly, the CaO/Al₂O₃ ratios of samples, which could reflect the abundance of carbonate minerals, show that the western areas have a higher abundance of ferromagnesian silicates and carbonate minerals in the riparian dunes than the eastern areas. Thus, the significant difference between the western and eastern parts of the study area indicates a higher abundance of iron, magnesium silicate, and carbonate minerals in the riparian dunes in the western areas than in the eastern sand dunes.

The Na₂O/Al₂O₃ and K₂O/Al₂O₃ ratios reflect the abundance of minerals such as plagioclase and potassium feldspar, muscovite, and clay in silicate minerals. The order of Na₂O/Al₂O₃ ratio in riparian dune samples is Type I > Type III > Type II, while that of K₂O/Al₂O₃ ratio is Type III > Type II > Type I. This difference can be attributed to the variation in abundance of potassium feldspar and plagioclase in the parent rocks of the sandy area in the eastern region and the deserts in the western region. The sandy areas in the eastern region are rich in potassium feldspar and scarce albite feldspar, while the deserts in the western areas are rich in potassium feldspar and scarce in albite feldspar (He et al., 2016; Yang et al., 2021).

In order to distinguish the regional characteristics of riparian dunes in each watershed, the changes in major elements are analyzed using SiO₂/10-CaO-Al₂O₃ (Figure 6a), (K₂O+Na₂O)-CaO-Fe₂O₃ (Figure 6b), and CaO-Na₂O-K₂O (Figure 6c) (Zhao et al., 2019). The abundance of carbonate, plagioclase, potassium feldspar, and muscovite in the dune samples is reflected by CaO-K₂O-Na₂O, while the abundance of quartz-carbonate-silicate minerals is reflected by the SiO₂/10-CaO-Al₂O₃. Similarly, the relative abundance of feldspar- carbonate-ferromagnesian silicate minerals is represented by (K₂O+Na₂O)-CaO-Fe₂O₃. Most of the samples from riparian dunes in the western and central-western parts of the desert rivers belong to Type I, which is characterized by high carbonate content and low potassium feldspar/muscovite content. On the other hand, samples from sandy areas in the east-central and northeastern parts belong to Type III, which is characterized by low carbonate content and high feldspar content. The samples from the Xar Moron River show different characteristics, with relatively low carbonate content and high potassium feldspar/muscovite content. The Maquan River samples belong to Type II, which is characterized by low carbonate content and medium plagioclase and potassium feldspar/muscovite content. However, some of the riparian dune samples from the western desert rivers categorized as Type I are rich in ferromagnesian silicates and poor in carbonates, and fall partly in the center of the triangular plot. Overall, the findings show that the samples from different regions exhibit distinct characteristics due to differences in their spatial distribution.

The results suggest a significant difference in mineral composition between the Type II samples from the Maquan River and the samples from the other types. Carbonate minerals exhibit the maximum variation, while feldspars show the minimum variation. Types I and III exhibit notable differences in the abundance of feldspars, ferromagnesian minerals, quartz, and silicates, with minor variations in the abundance of carbonates. This may be due to the occurrence of some minerals in the parent rocks of all three riparian dune types, which have distinct regional geochemical element characteristics.

In terms of the chemical weathering characteristics, the A-CN-K diagram (Figure 7) illustrates the weathering trends of sediments in riparian dunes of the selected watersheds, which are represented by different chemical components. PAAS (Taylor and McLennan, 1985), the primary weathering product of UCC, has a direction pointing from the UCC, representing the weathering trend of the early continental period (Nesbitt and Young, 1982).

The samples from all three types are mainly in the vicinity of UCC. The Type II samples show an early stage to moderate degree of chemicalweathering, with CIA values ranging from 55 to 60 (with an average of 55.9). Type I (green color) and Type III (dark blue color) indicate the same degree of chemical weathering, similar to UCC, with average CIA values of 43.2 and 45.4, respectively, representing the unweathered climate conditions or the early stage of chemical weathering (Figure 7). Generally, the degree of weathering generally increases from west to east region. Type I samples show more consistency with the continental weathering trend, Type III samples from the East of Xar Moron River are not consistent, and Type II samples from the Tibetan Plateau regions show the trend of migration continental weathering. Overall, the samples from the three types of riparian dunes exhibit a distinct degree of chemical weathering, even though the continental weathering trends remain consistent.

The study findings suggest that the major factors affecting the geochemical characteristics of major elements in Chinese deserts are surface morphology, material source, and clastic mineral composition, which are primarily influenced by the tectonic background of the area. The regional context of the origin of deserts may therefore approximate the tectonic landform, and the material source and composition may remain relatively stable (Zhao et al., 2019). The composition of rocks in the source location and the availability of new materials play a crucial role in determining significant element shifts. However, the surface sediment characteristics of riparian desert dunes vary due to the combined effects of wind and water transport and the geochemical characteristics of the surface sediments. This indicates that the surface sediments of riparian dunes exhibit some regional spatial differentiation under the interaction of geomorphic erosion (Li et al., 2016; 2017).

The most important factor determining insignificant element variations is the composition of rocks in the source location. In northern China, most deserts have a near-source provenance, meaning that the desert sediments originate from the surrounding mountains and become a source of riparian dunes due to the actions of wind and water (Zhu et al., 1980; Chen and Li, 2011; Dong and Wang, 2015; Hu and Yang, 2016). In other words, desert sediments are derived from the surrounding mountains, and rivers act as significant transportation channels for material sources (Li and Yan, 2014; Yan et al., 2015). Therefore, river and desert sediments contribute to the majority of the sand dunes on both sides of the river. The study reveals significant geographical changes in the composition of major chemical elements in the sand dunes on both sides of distinct desert/sandy land rivers, particularly in the concentrations of Ca and Mg (Table 2 and Figure 2).

Thus, the riparian dunes in the eastern region of China exhibit lower SiO₂ abundance and higher Fe and Mg concentration which is due to their lower mineral maturity (Figure 6) and limited material supply resulting from their permanent and semi-fixed dune nature and strong chemical weathering. On the other hand, the abundance of CaO in the riparian dunes of the western Taklimakan Desert-Keriya River and Qaidam Basin Desert-Tora River is significantly higher than in the eastern region, as evidenced by the variation of CaO abundance in the riparian dunes of different basins (Figures 2 and 3). This CaO enrichment is likely due to the presence of abundant carbonate minerals in the adjacent mountains (Li et al., 2007; Xie, 2012). Conversely, the riparian dunes of Mu Bulag River in the middle and Xar Moron River in the east exhibit noticeably lower CaO abundance due to the limited exposure of carbonate strata in the Ordos area (Li et al., 2007; Li, 2010; Xie, 2012). The presence of three divisions of K₂O and Al₂O₃ in Figure 2 indicates substantial variations in the sources of potassium in the riparian dunes. The occurrence of potassium-rich granite, potassium-rich basalt, and granite in the riparian dunes of Horqin Sandy Land-Xar Moron River in the northeast may be related to the presence of potassium-rich granite, potassium-rich basalt, and granite in the surrounding source region.

These K-rich rocks are transported to Horqin Land via water during the process of weathering and denudation, leading to a higher concentration of K₂O in the riparian dunes samples. The concentration of quartz in desert sediments is influenced by the bedrock type of the surrounding mountains, and these sediments impact the material sources of the adjacent desert rivers (Fu and Yang, 2004; Fu and Wang, 2015). The chemical composition of riparian dunes in various rivers is different, and hence rivers may also be considered the natural border and barrier of deserts. The material sources of the riparian desert dunes are influenced not only by the parent rocks but also by the introduction of new materials due to the actions of wind, water, and weathering processes. According to Table 5, there is a difference in the chemical composition of the desert sediments in the hinterland and those in the riparian sediments of the desert. The abundance of SiO₂ and Al₂O₃ in the riparian dunes of rivers in the western desert is lower than that in the corresponding hinterland of the desert. At the same time, the content of Fe₂O₃, MgO, CaO, and Na₂O is higher in the riparian dunes than in the hinterland of the desert. In contrast, the abundance of Fe₂O₃, MgO, CaO, and Na₂O is higher, indicating that the riparian dunes are more susceptible to the supplement of fresh materials and the loss of original components due to wind and water transport than the desert hinterland. The desert hinterlands are rich in quartz and silicate minerals but contain a lower concentration of carbonate substances. In Figure 8, the sediments from the western desert hinterland and the corresponding riparian dunes show comparable tendencies to shift compared to the UCC. For instance, the concentration of CaO is enriched while other elements are depleted, with more significant depletion in the eastern sandy lands and riparian dunes. This suggests a connection between them in terms of provenance. However, the loss of riparian dunes in the desert is more substantial than that in the desert hinterland. This indicates that the chemical composition of the sediments on both sides of the rivers in the riparian dunes is more susceptible to the influence of foreign and fresh materials, and the chemical weathering process of the riparian dunes is more pronounced due to the effects of flowing water.

The riparian dunes zones in different deserts exhibit substantial variations in terms of material denudation supply and physical and chemical weathering in the source area. The western desert is mainly characterized by mobile dunes that undergo significant physical weathering and moderate chemical weathering. The river receives a large amount of debris from resulting in lower mineral maturity (Figure 6), poor SiO₂ abundance, and higher Fe and Mg concentration. The majority of the riparian dunes in the east are semi-fixed or permanent dunes, experiencing strong chemical weathering and limited material supply, likely due to grassland deterioration and increased sand formation (Li et al., 2016). In contrast, the western deserts are characterized by inland rivers that act as sedimentary basins, receiving both dissolved and suspended matter that results in poorly sorted sediments with a high proportion of fine-grained materials that serve as the main source of sand for the riparian dunes via wind transport. On the other hand, the rivers in the eastern sandy terrain are mainly outflow rivers, where suspended and dissolved materials flow into the ocean, and well-sorted sand particles are the primary source of sand for the sandy terrain (Xie and Ding, 2007; Zhao et al., 2014).