Determination of Geochemical Background and Baseline and Research on Geochemical Zoning in the Desert and Sandy Areas of China

Abstract

Resources in deserts and sandy landscapes have potential for development, but existing surveys and sampling have not collected desert soil samples. As such, the geochemical background of these spaces remains unexplored due to the vastness and desolation of deserts. Therefore, researching the geochemical background values and geochemical baseline values of deserts is of long-term significance. Our research indicates that in addition to macrostructural environmental divisions, microelement geochemistry can also be used for geological unit zoning. In this paper, geochemical background and geochemical baseline values of 61 desert elements were calculated using the iterative method, frequency histograms method, and multifractal concentration-area method. It also analyzes the distribution characteristics of major, trace, and rare earth elements, and divides the 12 desert sand regions into different geochemical zones. This paper determines, for the first time, the geochemical background values of elements in Chinese deserts, filling the gap in the study of desert background values. By combining machine learning methods, different deserts have been divided into three geochemical zones. This research will greatly enhance our ability to interpret the geochemical distribution and evolutionary patterns of desert elements in China, and it has important scientific significance and practical value for desert research.

1. Introduction

Most studies are related to background values, which can be divided into geochemical background values and environmental background values. The “Geochemical background” is proposed to distinguish whether the content of elements in a geological unit or a certain area is normal and whether the data outside the background range are abnormal (positive or negative) data [1]. The study of geochemical background values is an important foundational work in the field of Earth sciences. The concept of geochemical background originated in exploration geochemistry, and was initially proposed for mineral exploration purposes [2]. However, with the rise of environmental science, this concept began to be introduced into the field of environmental geochemistry in the early 21st century [3]. For exploration geochemistry, the anomaly data may be an indication of the existence of a certain deposit or the element migration caused by alteration process [4]. Unlike its traditional meaning in exploration geochemistry, in environmental geochemistry, geochemical background values typically refer to the concentrations of chemical elements in environmental components that are unaffected by contamination. This reflects the inherent chemical composition characteristics of these elements in their natural state and development in the environment [5,6]. The term “geochemical baseline” was first introduced in the International Geological Correlation Programme (IGCP) [7,8]. Different scholars have provided similar scientific definitions of this term [9,10], generally referring to the actual content of specific elements or compounds in surface environmental media at a specific point in time. This definition encompasses both natural background concentrations and contributions from diffusion concentrations caused by human activities. Geochemical background values and baseline values are important parameters for describing the geochemical characteristics of a region, aiding in the understanding and evaluation of natural element contents in a specific area. They contribute to understanding regional geological backgrounds and geological processes [6,11,12]. The concept and calculation of geochemical background in this study are primarily based on definitions and methods established in exploration geochemistry. Typically [13,14], the lower limit of anomalies is used to delineate background regions from anomalous regions [15].

Deserts represent a natural resource, with sand dunes covering over 20 million square kilometers globally, distributed across continents worldwide. Approximately 20% of arid regions and numerous coastal areas are characterized by sand dunes, which further extend to encompass gravel deserts and regions influenced by aeolian erosion, collectively affecting over 40% of the Earth’s terrestrial surface subjected to aeolian processes. Due to their close relationship with human habitation environments and rich information on modern surface processes and Earth environmental evolution, deserts are considered an indispensable part of the Earth system and have been consistently studied by the international academic community [16,17]. In China, the Regional Geological Survey Project was initiated in 1979 and remains the longest-running national survey program of its kind by the Ministry of Natural Resources (formerly the Ministry of Land and Resources). While there has been extensive research on background values of inland water system sediments or soils, large areas of desert regions remain unsampled due to their vast size and difficult accessibility. Consequently, the geochemical background of desert regions in China remains largely unexplored. Given these circumstances, initiating research on background values in desert regions is crucial for subsequent desert studies and holds profound significance.

A substantial body of research [18,19,20,21,22] indicates that not only can geological units be classified and systematized based on their macrostructural environmental settings and evolutionary history, but they can also be geologically zoned based on micro-scale element geochemistry.

Therefore, this paper combines traditional methods with the multifractal method to determine regional geochemical background values and baseline values in desert areas. Through machine learning techniques, it measures the feature importance of different geochemical indicators comprehensively, effectively dividing desert and sandy regions into distinct geochemical zones. This approach enhances our understanding of the spatial distribution and variability of geochemical components in the study area, providing a robust basis for the exploration, assessment, and development of natural resources.

2. Study Area Overview and Sampling Analysis Testing Methods

Deserts in China cover approximately 1.73 million square kilometers, about 18% of the country’s total land area [23].

This study selected the following regions in China as study areas: the Taklimakan Desert, Gurbantunggut Desert, Badain Jaran Desert, Tengger Desert, Kumtag Desert, Qaidam Desert, Hobq Desert, and Ulan Buhe Desert, Mu Us Sandy Land, Horqin Sandy Land, Hunshandak Desert, and Hulun Buir Sandy Land (Figure 1).

According to sampling instructions provided by the Institute of Geophysical and Geochemical Exploration, Chinese Academy of Geological Sciences, Langfang, China, about 30 composite samples were collected in each of these deserts and sandy lands. Each composite sample was formed by mixing several shallow (0–20 cm) aliquots of material, collected in a small area around the sampling point. Aliquots were mixed to form a 10 kg composite sample. After a mass distribution of 500 g, each subsample was sent to the laboratory for analysis [24,25,26]. A total of 344 desert samples (Figure 1) were collected and analyzed for 61 indicators, as outlined in

Table 1. Analysis Methods and Elemental Composition (Data sources: Institute of Geophysical and Geochemical Exploration, Chinese Academy of Geological Sciences).

Analytical Methods	Target Elements
Inductively Coupled Plasma Mass Spectrometry (ICP-MS)	Ag, Bi, Cd, Ce, Co, Dy, Er, Eu, Gd, Ho, I, La, Lu, Mo, Nd, Ni, Pb, Pr, Rb, Sb, Sm, Tb, Th, Tm, Yb
X-ray Fluorescence Spectroscopy (XRF)	Al2O3, Ba, Br, CaO, Cl, Cr, TFe2O3, Ga, K2O, MgO, Mn, Na2O, Nb, P, S, SiO2, Sr, Ti, Y, Zn, Zr
Inductively Coupled Plasma Atomic Emission Spectrosco (ICP-AES)	Be, Cu, Li
Atomic Fluorescence Spectroscopy (AFS)	As, Ge, Hg, Se
Graphite Furnace Atomic Absorption Spectroscopy (AAS)	Au
Alternating Current Arc-Emission Spectroscopy (AES)	B, Sn
Gas Chromatography (GC)	TC, N
Potentiometry Method (POT)	F, pH
Volumetric Method (VOL)	Corg.

for each indicator analysis method.

3. Data Processing Methods

3.1. Calculation Methods of Geochemical Background Values and Geochemical Baseline Values

Calculation Method 1 for Geochemical Background Values: Iterative Method [27,28]. In geochemical data processing, traditional statistical parameter calculations are based on the assumption that the data follow a normal or log-normal distribution. However, in most cases, the data used for these calculations do not conform to such distributions. Therefore, by removing values that deviate more than three times the standard deviation from the mean, the data are adjusted to approximate a normal or log-normal distribution. At this point, the similarity between the calculated mean and the original data (before removal of values exceeding three times the standard deviation) should be assessed. When they are essentially consistent or very close, it indicates satisfactory calculation results. The iterative method follows these steps: (1) Calculate the mean ($X_1$) and standard deviation (${Sd}_1$) of the original data set for each element across the entire area; check if the data conform to a normal distribution. If they do, then $X_1$ is considered the background value. If not, proceed to step (2); (2) Remove a batch of values based on $X_1\pm {3Sd}_1$ and obtain a new data set. Then, calculate the mean ($X_2$) and standard deviation (${Sd}_2$) of this new data set and check if the data conform to a normal distribution. If they do, then $X_2$ is considered the background value. If not, proceed to step (3); (3) Repeat step (2) until no more high-value points exist, then calculate the mean ($X_n$) and standard deviation (${Sd}_n$) of the final dataset. $X_n$ is taken as the background value $C_{01}$, and $n+1$ is the number of times the data conform to a normal distribution. The calculation process is carried out using Microsoft Visual Basic for Applications [15,29,30,31,32].

Calculation Method 2 for Geochemical Background Values: Frequency Distribution. Histogram Method [33,34]. For those whose histogram is unimodal and close to normal distribution, the general processing method is to connect the left top Angle $S$ of the maximum frequency column with the left top Angle $Q$ of the latter group of frequency columns, and the right top Angle $T$ of the maximum frequency column with the right top Angle $P$ of the latter group of frequency columns. The projection of the two lines intersecting on the coordinate axis is the mass value $M_0$, which can be used as the background value. Then, draw a horizontal line defined by 0.6 times the maximum frequency, intersecting both sides of the frequency density curve, and the length between the intersection point and the median value is the mean square error σ. In this case, the true value of the value obtained by $M_0$ and 2 times the mean square error σ is the lower limit of the anomaly. Then, draw a horizontal line defined by 0.6 times the maximum frequency, intersecting both sides of the frequency density curve, and the length between the intersection point and the median value is the mean square error σ. In this case, the true value of the value obtained by $M_0$ and 2 times the mean square error $S$ is the lower limit of the anomaly (as shown in Figure 2a).

The mass value can be obtained mathematically:

$$M_0=x_0+{\displaystyle \frac{i\left(p_2-p_1\right)}{{2p}_2-p_1-p_3}}$$

(1)

In Formula (1), $M_0$ is the crowd value, $x_0$ is the lower limit value of the group where the crowd value resides, i is the group distance, $p_1$ is the frequency of the group before the group where the crowd value resides, $p_2$ is the frequency of the group where the crowd value resides, and $p_3$ is the frequency of the group after the group where the crowd value resides. The mass value $M_O$ is used as background $C_{02}$.

Calculation Method 3 for Geochemical Background Values: Concentration-Area Method (C-A method) was first proposed by Cheng et al. (1994). Later scholars also call it content-area method [34,35,36,37,38]. This method is based on the concentration value (that is, the frequency of the magnitude value) to separate outliers from the background; it distinguishes the background field from the abnormal field. These values can also be spatial or geometric features of geochemical indicators. The expression of the C-A model is as follows:

$$A(\rho >\upsilon )\propto {\rho }^{-\beta }$$

(2)

In Formula (2), $A\left(\rho \right)$ represents the region where the concentration is greater than or equal to the contour value; $v$ is the threshold; $\beta$ is the fractal dimension greater than zero. $A(\rho >v)$ is the area value of the element content ρ greater than $A$ certain value $v$, and with the increase of the value $v$, $A(\rho >v)$ will decrease accordingly. The change of $A$ with $v$ will depend on the magnitude of the exponent $\beta$. The fractal dimension $\beta$ can quantitatively describe the distribution of geochemical element content and the complexity of the variation of element content distribution in this scale-free region. This variation corresponds to different $\beta$ values in the range of background values and outliers, respectively, which will determine different slopes of the line segments in the schematic plot of double Logarithmic Coordinate Curve.

In the C-A method, by the schematic plot of double Logarithmic Coordinate Curve, the area $A(\ge s)$ with a concentration value greater than $s$ is established between the concentration value $s$. The power law relationship can be drawn to map the anomaly and the background region.

GeoDAS 4.0 has been used to create a double logarithmic coordinate curve for the content-area method (Figure 2b). The data do not necessarily need to follow a normal distribution. The second segment of the fitted line reflects the non-singular background information of element contents [39]. By calculating the arithmetic mean of the numbers that fall within the second segment of the fitted line, the background value $C_{03}$ is determined.

The final geochemical background value C₀ is obtained by averaging the values calculated from the three methods.

With the Calculation Method for geochemical baseline values [6,38,40,41], the process begins with testing the desert sample data for normality using Python. If the whole dataset conforms to a normal distribution, the arithmetic mean plus or minus two standard deviations is used to denote the baseline value and its range. Alternatively, if the data adhere to a log-normal distribution, the geometric mean multiplied/split the square of geometric standard deviation represents the baseline and its range. For skewed distributions, dividing the geometric mean by the square of the geometric standard deviation is the lower limit of the baseline value, and multiplying the geometric mean by the square of the geometric standard deviation is the upper limit of the baseline value [42]. This meticulous approach ensures the accuracy and reliability of the geochemical baseline values, essential for comprehensive analysis and interpretation in geochemical research.

3.2. Methods of Geochemical Zoning

By utilizing linear regression models in machine learning [43,44,45,46], the importance of different geochemical indicators (major elements, trace elements, rare earth elements) across various desert regions was assessed. The absolute values of the regression coefficients for each feature were compared to determine their significance in the model’s predictions. A feature importance stacking diagram was then created to illustrate the importance of different geochemical indicators in various desert regions, distinguishing geochemical zones based on their relative significance. The entire calculation process was executed using Python 2.7.

Using Origin, three ternary diagrams based on the background values of major elements were plotted: SiO₂/10-Al₂O₃-CaO, (K₂O + Na₂O)-CaO-TFe₂O₃, and CaO-K₂O-Na₂O [47]. These diagrams compare the compositional components (sources) of samples from different deserts and sandy lands. Regions with similar sources or chemical compositions were grouped into the same area.

Based on the background values of trace elements, hierarchical clustering was used to construct a dendrogram to represent the nested relationships among samples. A cluster analysis heatmap was generated [48] to observe the clustering of desert regions and elements, identifying regions with similar characteristics. Additionally, using SPSS and ArcGIS, a distribution map of trace element factor weights was created [8,49]. By analyzing and comparing the different factor weights of sample points in various desert regions, each sample point displayed the factor with the highest weight. Desert regions with similar factor compositions were grouped into the same area.

4. Results and Discussion

4.1. Geochemical Background Values and Baseline Values

The results of the geochemical background values and baseline values for the desert samples in the study area are shown in Appendix A 

Table A1. Statistics of Geochemical Background Values and Baseline Values for Different Deserts and Sandy Areas.

Element	Hunshandak	Horqin	Hulun Buir	Badain Jaran	Ulan Buhe	Mu Us	Hobq	Tengger
SiO2	85.409	88.249	89.740	78.647	79.286	77.744	80.539	80.889
Al2O3	7.135	5.902	5.579	8.286	8.196	10.693	8.732	8.272
TFe2O3	0.855	0.562	0.468	2.146	1.737	1.577	1.970	2.165
MgO	0.208	0.136	0.108	1.003	0.814	0.532	0.756	0.929
CaO	0.451	0.304	0.288	1.805	1.488	1.262	2.055	1.273
Na2O	1.746	1.221	1.118	2.276	2.063	2.877	2.099	2.095
K2O	2.751	2.504	2.437	1.866	2.254	2.547	1.978	2.151
Ag	46.496	50.535	42.935	39.934	42.391	70.371	41.749	36.652
As	2.272	1.657	1.996	5.319	4.484	3.209	4.955	4.673
Au	0.912	0.523	0.403	0.834	0.610	0.585	1.052	0.556
B	8.195	8.718	6.924	15.881	17.579	20.841	23.909	15.789
Ba	678.853	586.865	546.272	582.343	622.267	770.781	583.064	581.734
Be	0.985	0.840	0.954	1.191	1.172	1.573	1.266	1.299
Bi	0.064	0.075	0.059	0.140	0.105	0.082	0.142	0.137
Br	0.843	0.902	0.892	0.744	0.825	0.889	0.759	0.723
Cd	24.987	24.344	18.878	47.698	44.664	47.962	48.134	47.246
Cl	37.494	38.068	37.301	54.172	69.912	48.328	36.892	42.765
Co	1.975	1.256	1.099	5.728	3.990	3.032	4.512	4.872
Cr	7.760	6.330	5.001	53.544	22.156	18.373	27.613	29.409
Cu	6.126	4.526	5.022	12.586	9.387	7.138	10.194	10.204
F	76.386	57.253	77.549	206.061	169.011	142.499	181.533	205.903
Ga	9.086	7.514	7.606	10.218	9.590	11.975	9.708	9.729
Ge	0.883	0.925	0.893	1.150	1.036	0.976	1.082	1.136
Hg	6.570	6.553	6.421	9.180	8.026	7.275	10.602	7.651
I	0.385	0.319	0.453	0.400	0.276	0.289	0.405	0.370
Li	5.604	5.353	6.580	9.686	9.509	7.605	13.701	11.157
Mn	140.132	94.819	100.536	293.424	224.676	210.644	275.233	242.574
Mo	0.211	0.130	0.156	0.447	0.318	0.203	0.228	0.322
N	87.115	121.135	99.520	44.942	68.317	129.502	88.349	55.938
Nb	6.150	5.154	5.382	7.408	6.745	7.068	7.335	6.973
Ni	6.140	4.861	4.750	20.278	13.339	8.889	13.145	16.312
P	112.894	94.162	98.585	241.221	211.150	232.152	246.137	192.970
Pb	12.301	10.806	11.371	12.602	12.518	15.093	14.854	13.472
pH	7.970	7.276	6.668	9.210	9.313	8.273	8.972	8.996
Rb	86.042	80.364	79.173	65.405	64.780	69.538	68.806	73.227
S	49.556	48.581	49.307	85.168	72.074	41.134	39.514	69.862
Sb	0.277	0.176	0.217	0.546	0.443	0.271	0.558	0.458
Se	0.046	0.193	0.039	0.096	0.068	0.086	0.056	0.068
Sn	1.025	0.992	1.063	1.354	1.233	1.100	1.482	1.403
Sr	165.814	116.557	101.216	168.435	163.603	289.514	177.283	147.573
TC	0.082	0.118	0.116	0.352	0.219	0.173	0.305	0.175
Th	2.455	1.918	1.718	5.140	4.586	3.694	4.892	5.740
Ti	985.214	690.462	590.150	1673.664	1550.112	1745.705	1685.261	1555.537
Zn	11.686	9.448	8.585	27.157	21.519	17.348	23.987	26.390
Zr	140.810	128.656	187.519	126.136	127.049	161.713	133.328	105.395
Corg.	0.128	0.145	0.126	0.057	0.060	0.123	0.072	0.059
La	8.787	7.925	9.420	14.535	12.629	14.602	15.085	14.207
Ce	15.143	13.156	17.186	34.335	24.020	27.170	28.150	27.792
Pr	1.901	1.652	2.128	3.119	2.716	2.897	3.148	3.214
Nd	7.430	6.194	8.035	11.967	10.568	11.376	11.940	11.858
Sm	1.355	1.125	1.503	2.286	1.832	1.919	2.339	2.249
Eu	0.408	0.330	0.398	0.614	0.575	0.693	0.612	0.593
Y	8.216	7.121	7.414	12.963	10.845	10.185	11.472	12.429
Gd	1.199	1.007	1.236	2.123	1.770	1.766	2.065	2.183
Tb	0.200	0.164	0.192	0.317	0.283	0.280	0.315	0.355
Dy	1.148	0.885	1.007	1.977	1.694	1.652	1.919	2.123
Ho	0.244	0.181	0.199	0.386	0.344	0.340	0.383	0.435
Er	0.697	0.541	0.587	1.178	1.030	0.948	1.117	1.309
Tm	0.121	0.090	0.093	0.191	0.167	0.165	0.181	0.210
Yb	0.787	0.607	0.627	1.225	1.081	1.086	1.180	1.352
Lu	0.126	0.095	0.096	0.189	0.167	0.166	0.179	0.210

Note: Units are μg·g⁻¹; Ag, Au, Cd, Hg: ng·g⁻¹; Oxides, TC, Corg: wt%; pH is dimensionless; UCC refers to upper crustal abundances (Rudnick and Gao, 2014) [50].

and

Table A2. Statistics of Geochemical Background Values and Baseline Values for Different Deserts and Sandy Areas.

Element	Kumtag	Gurbantunggut	Taklimakan	Qaidam	UCC	Baseline	Baseline Range
SiO2	65.805	71.855	61.112	58.014	66.620	75.337	49.960–113.605
Al2O3	10.141	10.434	9.783	9.225	15.400	7.992	4.942–12.927
TFe2O3	3.001	2.972	2.234	1.849	5.040	1.420	0.388–5.202
MgO	2.475	1.191	2.261	1.519	2.480	0.582	0.072–4.701
CaO	6.381	2.784	7.645	9.977	3.590	1.407	0.145–13.679
Na2O	2.775	2.272	2.829	2.930	3.270	2.008	0.937–4.303
K2O	1.798	2.269	2.237	2.120	2.800	2.209	1.445–2.973
Ag	41.352	49.946	37.279	49.153	53.000	42.394	24.473–73.437
As	5.546	7.003	6.408	6.555	4.800	3.614	1.198–10.904
Au	1.920	1.302	0.881	0.808	1.500	0.633	0.249–1.605
B	40.417	37.746	44.069	34.985	17.000	16.384	4.798–55.953
Ba	570.059	515.546	559.338	483.902	624.000	583.416	348.347–818.485
Be	1.467	1.820	1.887	1.579	2.100	1.247	0.705–2.206
Bi	0.158	0.194	0.169	0.192	0.160	0.106	0.040–0.281
Br	0.897	1.411	1.082	0.873	1.600	0.836	0.493–1.419
Cd	69.450	86.760	83.411	200.818	0.090	43.345	13.499–139.181
Cl	2489.497	1037.997	17,971.754	8176.649	370.000	80.713	4.050–1608.485
Co	8.439	7.732	5.901	4.729	17.300	3.462	0.845–14.182
Cr	49.974	34.632	32.502	21.687	92.000	18.510	3.386–101.195
Cu	14.640	18.733	12.196	10.941	28.000	8.568	3.316–22.138
F	334.623	349.961	373.148	302.346	557.000	159.147	45.440–557.389
Ga	11.529	13.168	11.423	11.040	17.500	9.704	6.394–14.727
Ge	1.169	1.111	1.049	0.915	1.400	1.013	0.740–1.387
Hg	9.495	10.785	9.438	9.669	0.050	7.752	4.490–13.384
I	0.621	1.721	0.558	1.002	1.400	0.426	0.160–1.132
Li	17.341	18.356	20.950	13.707	21.000	9.444	3.564–25.024
Mn	276.620	606.612	347.138	271.722	/	220.406	68.647–707.657
Mo	0.785	1.037	1.213	0.834	1.100	0.296	0.075–1.166
N	111.946	330.449	175.894	81.724	83.000	78.948	25.827–241.330
Nb	8.322	8.827	9.204	7.057	12.000	6.816	4.427–10.494
Ni	22.256	16.852	16.049	11.561	47.000	10.647	3.276–34.602
P	352.538	441.268	375.001	313.684	654.572	201.466	69.604–583.135
Pb	12.914	15.454	15.091	17.821	17.000	13.043	8.874–19.170
pH	9.168	9.233	9.174	8.720	/	8.451	6.645–10.748
Rb	67.333	76.247	81.180	78.264	84.000	71.511	52.241–97.889
S	7537.253	1537.105	5536.793	25043.051	621.000	107.837	3.733–3115.437
Sb	0.546	0.613	0.536	0.737	0.400	0.361	0.129–1.010
Se	0.148	0.331	0.078	0.099	0.090	0.062	0.023–0.163
Sn	1.419	1.608	1.942	2.029	2.100	1.272	0.744–2.174
Sr	296.507	209.456	282.177	431.545	320.000	180.879	74.052–441.815
TC	1.355	0.519	1.534	1.124	/	0.312	0.129–0.756
Th	6.461	6.416	7.808	6.053	10.500	4.077	1.555–10.687
Ti	2387.824	2586.209	2104.065	1513.804	3836.795	1342.895	475.015–3796.443
Zn	35.218	43.199	35.996	33.035	67.000	19.974	6.520–61.189
Zr	127.077	175.543	131.067	139.682	193.000	119.928	59.672–241.029
Corg.	0.105	0.138	0.196	0.100	/	0.078	0.033–0.186
La	19.399	18.654	22.507	15.829	31.000	13.312	6.540–27.096
Ce	37.239	38.993	42.830	30.355	63.000	24.487	10.862–55.202
Pr	4.355	4.508	5.152	3.420	7.100	2.980	1.380–6.433
Nd	16.788	18.450	19.588	13.335	27.000	11.402	5.132–25.333
Sm	3.155	3.837	3.750	2.612	4.700	2.145	0.922–4.991
Eu	0.780	0.859	0.873	0.681	1.000	0.582	0.291–1.162
Y	16.062	20.874	16.430	13.651	21.000	11.200	5.579–22.483
Gd	2.951	3.681	3.365	2.432	4.000	1.942	0.804–4.694
Tb	0.467	0.608	0.536	0.400	0.700	0.229	0.025–2.058
Dy	2.512	3.334	3.041	2.410	3.900	1.780	0.724–4.378
Ho	0.507	0.706	0.603	0.473	0.830	0.363	0.148–0.893
Er	1.489	2.028	1.702	1.291	2.300	1.060	0.439–2.560
Tm	0.244	0.342	0.267	0.209	0.300	0.172	0.072–0.411
Yb	1.545	2.183	1.660	1.343	2.000	1.125	0.488–2.595
Lu	0.235	0.334	0.250	0.204	0.310	0.173	0.076–0.392

Note: Units are μg·g⁻¹; Ag, Au, Cd, Hg: ng·g⁻¹; Oxides, TC, Corg: wt%; pH is dimensionless; UCC refers to upper crustal abundances (Rudnick and Gao, 2014) [50].

.

The major elements that show significant differences can be categorized into three groups: SiO₂, which is related to changes in the quartz background values; CaO, which reflects changes in the carbonate background values; and MgO, whose variation is associated with changes in both carbonate and ferromagnesian silicate background values. The Hunshandak, Horqin, and Hulun Buir desert sands are characterized by high SiO₂ and low CaO, whereas the Kumtag, Gurbantunggut, Taklimakan, and Qaidam desert sands are mainly characterized by low SiO₂ and high CaO. The Badain Jaran, Ulan Buhe, Mu Us, Hobq, and Tengger desert sands have SiO₂ and CaO background values that fall between these two extremes (Appendix A Table A1 and Table A2; Figure 3a).

From the analysis of the background values of trace elements (Figure 3c), it can be seen that the UCC-standardized values of trace elements in the Kumtag, Gurbantunggut, Taklimakan, and Qaidam desert sands are higher than those in other desert sands. Some trace elements (Cd, Cl, S, N) in these deserts are even higher than the upper continental crust abundance [50]. In contrast, the standardized values of most trace elements in the Hunshandake, Horqin, and Hulun Buir desert sands are lower than those in other deserts. The trace element standardized values in the Badain Jaran, Ulan Buhe, Mu Us, Hobq, and Tengger desert sands fall between these two extremes.

From the analysis of the background values of rare earth elements (Figure 3a), the UCC distribution pattern plot of rare earth element background values in the deserts show a flat trend [51,52,53,54]. This indicates that the rare earth element distribution patterns in the desert sands of the study area are similar to those of the upper continental crust (UCC), and the trend in the normalized values is almost consistent with that of the major and trace elements.

According to the ratio charts of element background values to baseline values for China’s deserts (Figure 4), most of the background values for various element indicators in different desert sands are close to their baseline values, indicating that the majority of the data fall within the normal natural range. Some element indicators (such as MgO, TFe₂O₃, CaO, and TC) are relatively enriched in the Kumtag, Taklimakan, and Qaidam deserts, while they are relatively depleted in the Hunshandak, Horqin, and Hulun Buir deserts. A few elements (such as Cl and S) show significant enrichment in the Kumtag, Gurbantunggut, Taklimakan, and Qaidam deserts [4].

4.2. Geochemical Zoning

Based on the analysis of geochemical background values and baseline values, it was observed that the enrichment or depletion patterns of different elemental indicators in various deserts exhibit certain similarities (Figure 3 and Figure 4). However, as shown in Figure 3, classifying or zoning different deserts based on enrichment or depletion alone is significantly influenced by human factors and tends to be rather coarse. Therefore, to improve the precision of geochemical zoning, it is beneficial to utilize machine learning techniques in conjunction with some geochemical indicators [55]. By applying these methods, the 12 desert areas can be divided into several geochemical zones. This approach enhances the visualization of zoning results, increases the efficiency of zoning analysis, and facilitates the evaluation and comparison of geochemical regional systems [56,57,58,59].

From the three triangular diagrams (Figure 5), it is clear that the various desert areas are divided into three distinct zones: Eastern Deserts (1. Hunshandak, 2. Horqin, 3. Hulun Buir), Central Deserts (4. Badain Jaran, 5. Ulan Buhe, 6. Mu Us, 7. Hobq, 8. Tengger), and Western Deserts (9. Kumtag, 10. Gurbantunggut, 11. Taklimakan, 12. Qaidam). The SiO₂/10-Al₂O₃-CaO triangular diagram (Figure 5a) reflects the relative proportions of quartz, carbonate, and silicate minerals. The Western Deserts, including Kumtag, Taklimakan, and Qaidam, are characterized by significantly higher carbonate mineral ratios, relatively lower quartz ratios, and intermediate silicate mineral content. The (K₂O + Na₂O)-CaO-TFe₂O₃ triangular diagram (Figure 5b) shows the relative proportions of feldspar, carbonate, and ferromagnesian silicate minerals in the samples. In the Western Deserts, Kumtag, Taklimakan, and Qaidam exhibit high carbonate mineral ratios and low ferromagnesian silicate mineral ratios. The Central Deserts, such as Badain Jaran, Ulan Buhe, Mu Us, Hobq, and Tengger, as well as Gurbantunggut in the Western Deserts, feature intermediate carbonate mineral ratios. In contrast, the Eastern Deserts, including Hunshandak, Horqin, and Hulun Buir, show high feldspar mineral ratios and low carbonate mineral ratios. The CaO-K₂O-Na₂O triangular diagram (Figure 5c) reflects the relative proportions in the content of carbonate, potassium feldspar/muscovite, and plagioclase in the samples. The Western Deserts (Kumtag, Taklimakan, and Qaidam) exhibit high carbonate mineral ratios, low potassium feldspar/muscovite, and low plagioclase content. The Central Deserts (Badain Jaran, Ulan Buhe, Mu Us, Hobq, and Tengger) and Gurbantunggut in the Western Deserts have intermediate carbonate mineral ratios, moderate plagioclase content, and high potassium feldspar/muscovite ratios. Conversely, the Eastern Deserts, including Hunshandak, Horqin, and Hulun Buir, are characterized by high plagioclase mineral ratios and low carbonate mineral ratios.

In the constant element feature importance stacked plot (Figure 6a), CaO, MgO, and TFe₂O₃ are significantly more important in the Kumtag, Gurbantunggut, Taklimakan, and Qaidam deserts compared to other deserts. The importance is somewhat lower in the Badain Jaran, Ulan Buhe, Mu Us, Hobq, and Tengger deserts, with Hunshandak, Horqin, and Hulun Buir deserts showing the lowest levels of importance.

The primary factor influencing the variation in constant elements is the composition of the source rocks. From the geochemical map of calcium oxide in China, it is evident that the Kunlun Mountains and the North Tian Shan Mountains are rich in marine carbonate rocks. These carbonate minerals are continuously supplied to the Taklimakan, Qaidam, and Kumtag deserts through river systems. The Tianshan Mountains and the Altai Mountains near Gurbantunggut also have significant carbonate rock formations, providing a certain amount of carbonate minerals to the Gurbantunggut Desert. In contrast, the Badain Jaran, Tengger, Hobq, and Mu Us deserts are surrounded by carbonate strata along the northern edge of the Qilian Mountains, the surrounding Alashan mountains, and the southern segment of the Helan Mountains, but these formations are not as extensive as those in the Kunlun Mountains [60]. Therefore, the CaO content in these deserts is lower compared to the Western Deserts. The Eastern Deserts, surrounded by areas lacking carbonate rocks [47], consequently have a lower CaO content.

The second factor influencing the chemical composition of Chinese deserts is the supply of fresh material. This can be explored from multiple perspectives, including the erosion, weathering, transportation, and deposition of source materials. Chemical weathering leads to the dissolution of carbonate minerals and the gradual transformation of feldspars and ferromagnesian silicate minerals into secondary clay minerals. The weathering products are transported in the form of solutions or suspensions, resulting in the relative enrichment of weathering-resistant minerals like quartz in coarse fractions. Physical weathering, on the other hand, further fragments minerals like carbonates and ferromagnesian silicates, which are less resistant to physical weathering, leading to the relative enrichment of weathering-resistant minerals such as quartz in coarse fractions. From the perspective of material erosion supply and physical and chemical weathering, different deserts exhibit significant differences. In the Western Deserts, such as the Taklimakan Desert, surrounded by high mountains like the Tianshan, Pamirs, Kunlun, and Altun Mountains, with large river gradients, sparse vegetation cover, and extensive ice and snow cover on mountaintops (with more developed glaciers during the Quaternary glaciation), physical erosion is strong and chemical weathering is relatively weak. Overall, the Western Deserts are characterized by mobile sand dunes with strong physical weathering and weak chemical weathering. In contrast, the Eastern Deserts are predominantly covered by fixed sand areas, with weaker physical weathering but stronger chemical weathering [47], resulting in the partitioning observed in the constant element importance stacking diagram (Figure 6a).

For the trace element importance stacking diagram (Figure 6c), it can be concluded that in desert regions such as Kumtag, Gurbantunggut, Taklimakan, and Qaidam, elements like S, Cl, and TC show significantly high importance. These elements are typically closely related to the chemical weathering processes in desert areas. Particularly under arid conditions, their accumulation may be influenced by variations in salinity and moisture. The noticeable geochemical differences in some elements are attributed to their differentiation, migration, and enrichment under varying climatic backgrounds and depositional environments. In the trace element clustering heatmap (Figure 7), the Western Deserts show a higher correlation with most trace elements. This likely reflects the geological background and depositional environment of the Western Deserts, where specific geological processes or rock compositions might lead to trace element enrichment. The Central Deserts follow in terms of correlation, while the Eastern Deserts exhibit lower trace element correlation, with only elements such as Rb and Corg. showing relatively better correlation. This may be related to the younger geological age and lower degree of chemical weathering in the Eastern Deserts. In the trace element factor weight distribution map (Figure 8), factors F1 and F2 have the highest weights in the Western Deserts, which may be related to the region’s geological structure, rock composition, and chemical weathering processes. Other factors show weaker weights on the map. In the Central Deserts, besides F1 and F2, factors F3, F4, and F5 also have noticeable impacts, which may reflect the more complex geological background and chemical composition of the region. In the Eastern Deserts, the presence of factors F6 and F7 may indicate different geological features or chemical environments compared to other regions [61].

From the rare earth element importance stacking diagram (Figure 6b), it can be concluded that the significance of rare earth elements in different desert regions follows the order: Western > Central > Eastern. Deserts within the same large region share a common source area for their material supply. However, due to varying degrees of post-depositional weathering, the Eastern Deserts experience more intense chemical weathering, leading to a greater loss of rare earth elements [62,63].

Based on the analysis of all the geochemical indicators, the twelve desert regions can be categorized into three geochemical zones: (Ι) Western Desert Regions (Kumutag, Gurbantunggut, Taklimakan, Qaidam) characterized by high carbonate rock content, low quartz and feldspar content, high trace element correlation and significance, and high rare earth element significance; (ΙΙ) Central Desert Regions (Badain Jaran, Ulan Buhe, Mu Us, Hobq, Tengger) with moderate carbonate rock content, moderate quartz and feldspar content, average trace element correlation, and moderate rare earth element significance; and (ΙΙΙ) Eastern Desert Regions (Hunshandak, Hobq, Hulun Buir) showing very low carbonate rock content, high quartz and feldspar content, low trace element correlation and significance, and low rare earth element significance. This classification aligns with previous research [47], which similarly divided the deserts of China into western, central, and eastern regions based on major element analysis.

5. Conclusions

(1) In this paper, geochemical background and geochemical baseline values of 61 desert elements were calculated using various calculation methods. This provides a baseline for understanding the natural chemical composition of desert regions, which is significant for environmental change assessment and natural resource exploration and has far-reaching implications for future desert research.

(2) Combining machine learning methods, different desert areas have been grouped into three geochemical zones, which enhances our understanding of the spatial distribution and variability of geochemical components in the study area. This classification provides a solid basis for natural resources exploration and evaluation, which is useful for economic development. The zone analysis facilitates better identification and management of environmental geochemical issues in different regions.