1 INTRODUCTION

A salt lake is a typical environmental model across the world that refers to a lake with a higher salt than normal seawater (3.5%). A salt lake is a special stage of lake development that evolved from a freshwater lake. The formation of salt lakes is caused by various causes, including lake basin structure, salt supplies, and climatic change. Salt lake deposition is crucial evidence of the evolution of salt lakes due to physical geography, geological environment, and climatic change. Therefore, precisely identifying the composition of salt lake depositional minerals is crucial for obtaining climatic information such as temperature and precipitation and providing direct evidence for paleoclimate and environment reconstruction (Crowley, 1993; Benison and Goldstein, 2001; Jin, 2011; McLaren et al., 2012; Zheng et al., 2016).

Salt lakes generally form under severe and fragile natural environmental circumstances, making detection and data collection difficult. Because it benefits gathering large-scale geographic information cheaply and rapidly, remote sensing has recently become one of the essential methods for observing and monitoring dynamic changes in salt lakes. Some researchers used remote sensing data to identify the salt lake's water bodies and adjacent habitats (Azareh et al., 2021), lake surface areas, water levels, and volume change (Chipman, 2019; Nanis and Aly, 2020), as well as identifying salt lake information. Such as using 15 remote sensing indices taken from Landsat 8-OLI data to forecast and spatially map soil salt composition, main electrolytes, and ionic strength (IST) as difficult-to-measure soil chemical characteristics and parameters (Omrani et al., 2021), investigating the dimensions of climate change impacts in salt lake and its surroundings in Turkey using satellite remote sensing data, the effects of climate changes in the salt lake are explored by analyzing water and salt reserve variations using seasonal and multitemporal SPOT imagery (Ekercin and Örmeci, 2010). Moreover, using multitemporal Landsat 5-TM, 7-ETM+, and 8-OLI images to construct multiple satellite-derived indices for extracting surface water, identify and simulate temporal and spatial changes in Lake Urmia's surface water area (Rokni et al., 2014). In China, the remote sensing application of salt lake study began in the 1990s and has yielded several results, including estimation of soil salt content, identification of salt lake information (Zhang et al., 2017; Wang et al., 2020), detection of salt lake water surface changes (Xu et al., 2017; Yao et al., 2018; Hu et al., 2021; Shao et al., 2022), and detection of ecosystem environmental changes (Zhang et al., 2017; Yang et al., 2020). The low-altitude unmanned aerial vehicles remote sensing system (UAVRSS) solves the shortcomings of space and aerial remote sensing systems in resolution, revisit duration, cloud cover, and cost, resulting in a unique technique for salt lake mesoscale investigation (Sun et al., 2017). By equipping with hyperspectral probes and multispectral cameras, the hyperspectral data acquired by the UAV can accurately identify mineral types, which is an effective means to achieve non-destructive environmental and standardized sampling. Compared with sampling and lab-based spectroscopy that destroy the original depositional sequence of sedimentary minerals, UAV sampling data record the spatial coordinates of sedimentary minerals in the original environment, and the inversion results reproduce the spatial distribution sequence of sedimentary minerals, which can be used to analyze the paleoclimatic environment of the lake shoreline. UAVs have the advantages of simple structure and small size, long-endurance, easy to carry and operate, and are very suitable for severe weather and dangerous environment. In addition, it can save more cost compared to lab-based spectroscopy techniques. The advancement of the UAVRSS offers technological assistance and ensures the security investigation of the Lop Nur uninhabited region.

Around 1 500 years ago, Lop Nur had a rich ecosystem suitable for human survival oasis and gave birth to the Xiaohe civilization and Loulan civilization (Qin et al., 2012). Currently, Lop Nur has become the arid hinterland of Asia, with the salt crust landform displaying bright and dark stripes on remote sensing photographs obtained by the US Earth Resources Satellite in 1972. The formation of the bright and dark stripes that create the "Ear" characteristic is closely related to the mineral deposited during Lop Nur shrinkage. Understanding the influence of Tibetan Plateau uplift on climate change in the Lop Nur region requires identifying depositional minerals and the sedimentary state. Many scientists are intrigued by the ear-shaped stripes. Subsurface hyper-saline soil with high water content and salinity was detected using full polarimetric Synthetic Aperture Radar (SAR) data (Shao and Gong, 2011; Gao, 2013; Gong et al., 2013). Laboratory hyperspectral data and ion content in sediment were utilized to investigate the relationship between ion content and reflectance as the Salt Lake evolved (Cai et al., 2011; Zhang et al., 2014). The formation of ear-shaped stripes is closely related to the surface's physical form and mineral composition or underground utilizing the geophysical or geochemical survey methods (Xie et al., 2004; Zhao et al., 2006; Ma et al., 2007, 2011; Kong et al., 2014). However, sedimentary minerals in salty lake shorelines have not been identified; furthermore, the paleoclimate response characteristics of sedimentary minerals along bright and dark shorelines are seldom discussed.

Spectral unmixing can accurately identify the species and content of surface sedimentary minerals (Plaza et al., 2002). The hyperspectral unmixing (HU) aims to acquire the content proportion of different minerals by estimating the number of reference spectra (the end members), their spectral signatures, and their fractional abundances. Sparse regression served as a new direction to do spectral unmixing because it dramatically reduces the risk of the reduction of abundance estimation. To avoid the influence of unpurified endmembers, a sparse unmixing method has been widely concerned in spectral unmixing (Lan et al., 2018; Yuan et al., 2018; Zhang et al., 2021). SUnSAL algorithm is a classic representation of sparse unmixing (Iordache et al., 2014), which has spatial consistency with good accuracy in the spatial distribution of the classes. Meanwhile, the constrained form of the SUnSAL algorithm could adequately model the data for unmixing than the unconstrained form (Nascimento et al., 2014; Kumar et al., 2015; Masson et al., 2017).

This study aims to reveal the essential formation mechanism of the bright and dark ear-shaped stripes. Furthermore, to reconstruct the process of water retreat and the paleoclimatic environment history of Lop Nur Lake:

(1) Based on the UAV hyperspectral data, the surface salt crust mixing spectrum inversion model was established using the sparse spectral unmixing SUnSAL method. The species and relative abundance value of sedimentary mineral components were used inversed to reveal the heterogeneity of sedimentary minerals in bright and dark stripes.

(2) The controlling factors of the formation of bright and dark stripes are revealed by analyzing the coupling relationship between the mineral composition of the surface deposit, the reflectance value and the salt crust surface morphology of the bright-dark lakeshore feature.

(3) The paleoclimatic environment forming the bright and dark features is studied using distinct sedimentary mineral records of the climatic environment. The climatic change in Lop Nur during the process of lake water filling to entirely drying up was recreated.

2 MATERIAL AND METHOD 2.1 Study area

Lop Nur is a part of the Xinjiang salt lake subregion in China's Xinjiang Uygur Autonomous Region. The site is located in the middle of the Taklamakan and Kumtag deserts, the largest sedimentary basin on the northern side of the Qinghai-Tibet plateau. Lop Nur is in the lowest point of the Tarim Basin in NW China and a water-salt collection area for both surface and groundwater. The average elevation of Lop Nur is 780 m, and the variation in elevation is less than 5 m (Wang et al., 2014). The annual precipitation average is 22 mm. However, the evaporation capacity exceeds 3 000 mm. During the Loulan Epoch, the elevation of the lake may seldom surpassed 790.0 m, and the lake once covered an area of more than 11 602 km² (Zhang et al., 2021; Shao et al., 2022). Lop Nur vanished in 1972 due to human activity of diversion and the shortage of water recharge from Tarim River, the major source. The basic reason is the change of runoff caused by glacier change (Wang, 2011). The current drought has been exacerbated by the large disparity between evaporation and precipitation, with the salt crust being extensively dispersed and unfavorable for plant development. The Lop Nur has become a drought center in China and the Asian continent (Fig. 1b).

2.2 Data

The field spectra were collected using an Ocean Optics STS-VIS hyperspectral probe with 300 bands in the 350–830-nm spectral band and a spectra resolution of 1.5 nm based on a DJI M100 drone platform (Fig. 2b). The field of view angle (FoV) of a UAV is 34.4°, and the coverage area radius of the instantaneous field of view angle is 6 m. The multispectral photographs were captured with a micro Tetracam ADC multi-spectral digital camera mounted on the drone, with dimensions of 75 mm× 59 mm×33 mm, a spectral range of 520– 920 nm, a 3.2-megapixel lens, an image resolution of 2 048× 1 536 pixels, and a processing speed of 0.5ʺ (Fig. 2c).

In October 2017, the UAV platform collected hyperspectral data and multispectral photos. It is necessary to ensure that the sampling area is a typically bright and dark stripes area to investigate the spatial heterogeneity of sedimentary minerals, but also to cover the whole stage of lake retreat, and then the variation of sedimentary minerals during the retreat can be analyzed. Thus, we collected 7 typical bright and dark stripes areas along the northeast to the southwest oriented road in 5-km interval, named Nos. 1–7 sampling areas (Fig. 1c). A handheld GPS was used to locate the take-off point in each sampling area. The drone's flight height was 20 m (H=20 m). Each sampling area had four lines with a heading of 218° for lines m and p and 38.5° for routes n and q, each about 850 m long, and the space between adjacent lines was 9 m (Fig. 1d). In each sample region, about 600 spectral imprints and 100 visible-light photographs were collected, and each record was accompanied by global positioning system (GPS) data. Eleven surfaces (0–2 cm) salt crust samples were collected in Nos. 1–7 sampling areas and used for lab-based spectroscopy and verification. Sampling area No. 1 includes samples 1–2, area No. 2 includes samples 3–5, area No. 3 includes sample 6, area No. 4 includes samples 7– 8, area No. 5 includes sample 9, and area No. 6 includes sample 10, and No. 7 includes sample 11. Each of these samples was collected from an approximately 30 m×30 m area. The geographic coordinates of each sampling site were recorded at one-meter accuracy using a global positioning system (GPS) instrument. The samples were kept fresh in sealed plastic bags (17 cm×20 cm). Each sample collected was analyzed for full salt content, including Na⁺, Mg²⁺, Ca²⁺, K⁺, and Cl^–.

Orthorectified imageries of the salt crust surface morphology in the Nos. 1–7 areas were achieved by splicing the visible light images with the Agisoft PhotoScan software (Mader et al., 2015; Watlet et al., 2016). These photographs analyzed the different salt crust varieties and differentiated the salt crust formation process. The hyperspectral data was pre-processed, the true spectral reflectance value of the salt crust was calculated by Eq. 1 and the effective visible spectrum range (400–830 nm) was chosen as the experimental data. The smooth function of MATLAB software was used to denoise the reflectivity curve, and the moving windows number of points utilized to compute each element of the spectral curve was set to 20.

where x represents reflectivity; y is detected radiance data; m denotes whiteboard calibration; and n is dark current.

2.3 Method

The technical flowchart of this study is depicted in Fig. 3.

2.3.1 The SUnSAL algorithm

At present, sparse unmixing is a common and successful linear unmixing method. It can significantly reduce the risk of underestimating abundance compared to NMF, archetypal analysis, and the Bayesian approach. The SUnSAL (sparse unmixing by variable splitting and augmented Lagrangian) algorithms are based on the alternating direction method of multipliers, a method from the augmented Lagrangian family. It is more helpful in solving convex optimization problems in hyperspectral unmixing. Examples are the constrained least squares (CLS) problem used to compute the fractional abundances in a linear mixture of known spectra, the constrained basis pursuit (CBP) to find sparse (i.e., with a small number of terms) linear mixtures of spectra, selected from large libraries, and the constrained basis pursuit denoising (CBPDN), which is a generalization of BP to admit modeling errors. SUnSAL is shown to outperform off-the-shelf methods in terms of speed and accuracy (Bioucas-Dias and Figueiredo, 2010). The SUnSAL spectral unmixing model is used to extract sedimentary mineral components.

The SUnSAL unmixing model was written and run in MATLAB R2015a (the operating system is Win7) (Fig. 4).

The fully constrained least squares (FCLS) problem is defined as (Heinz and Chang, 2001):

where A∈R^k×n denote a matrix containing n spectral signatures of the endmembers, x∈Rⁿ denote the (unknown) fractional abundance vector, and y∈R^k is an (observed) mixed spectral vector.

Consider an unconstrained problem of the form:

where f₁: Rⁿ→R, f₂: R^p→R and G∈R^P×n.

Then, specializing alternating direction method of multipliers (ADMM) (Glowinski and Marroco, 1975; Gabay and Mercier, 1976; Eckstein and Bertsekas, 1992) to FLCS of the optimization problems, the optimization problem is expressed as an equation:

where l_{1}(1^T x) enforces the ASC.

The optimization problem (4) is transformed into

Based on Eqs. 2–7, the principle of the SUnSAL algorithm is clarified. The detailed calculation procedures of ADMM and SUnSAL for executing the algorithm are shown in the Supplementary Tables S1–S4.

2.3.2 Endmember spectrum library establishment

The type of endmember in the standard spectral library is determined by a combination of the Lop Nur salt lake type, the characteristic bands of the mineral spectra, and their climatic indications. According to the mineralization specificity of saline lakes, Lop Nur is a magnesium sulfate subsalt lake (Hu et al., 2007). Scholars have measured the mineralization degree of high concentration intercrystalline brine in the mine area in 1997 to be 409.068/L (Zheng et al., 2002) when the lake was in the sylvite primary crystallization stages. According to the stable phase diagram of the five ions pentagonal system, including K⁺, Na⁺, Mg²⁺, Cl^–, and SO₄^2–, and -H₂O in natural water solution (Braitsch, 1971; Han et al., 1982), the material quasi-stability and water-salt parameter imbalance in 25 ℃ were analyzed. Lop Nur is still in the low-intermediate metamorphism stage and hence does not reach the stage of large-scale potassium salt deposition (Hu et al., 2007). The dominant minerals in the sedimentary mineral assemblages of magnesium sulfate subtypes are: carnallite (KCl·MgCl₂·6H₂O), bioschofite (MgCl₂·6H₂O), sylvite (KCl), halite (NaCl), gypsum (CuSO₄ · 4H₂O), mirabilite (NaSO₄·10H₂O), dolomite (CaMg (CO₃)₂), magnesite (MgCO₃), pinnoite (CaB₆O₁₁·3H₂O), hydroboracite (CaMg₆O₁₁·13H₂O), clay minerals, and other secondary salts of 27 species (Zheng et al., 2016).

In the linear spectral unmixing model, the more significant the spectral signatures of the standard end member in the waveband, the better the precision of the unmixing result (Stoner and Baumgardner, 1981; Lagacherie et al., 2008). Thanks to the non-significant spectral signature of some minerals between 400 and 830 nm, these minerals have a minor impact on the unmixing results, which could be verified in two comparison experiments: a spectral library is added with or without significant feature minerals. Both results show the same trend in relative abundance values. Therefore, seven mineral components with significant spectrum absorption in the 400–830-nm spectral bands were chosen as endmember components. Halite (NaCl) has no notable spectral characteristics between these bands and accounted for a major share of the total, was added as a representation of non-absorption signature minerals (Table 1).

The main salt mineral in the Lop Nur Lake is halite that extensively dispersed on the top salt crust. Carnallite is the end-product of evaporation in magnesium-potassium salt lakes, and it is partially formed in halite crusts. It is frequently produced in conjunction with halite and sylvine. Polyhalite is formed when soluble potassium and sulfate minerals coexist with gypsum, glauberite, and bloedite. Dolomite is a carbonate mineral found in lake deposits alongside gypsum, anhydrite, halite, and sylvine. Celestite sulfate mineral coexists with calcium mirabilite in needle-like and granular form and distributes between calcium mirabilite crystals in the dry salty lake, and a tiny proportion is related to halite and gypsum (Sun et al., 2010). Chlorite is a layered silicate mineral that contains Mg and Fe chemical components. The illite-smectite mixed layer is a clay mineral intermediate product generated after the chemical weathering of montmorillonite to illite. Illite is formed due to metasomatism of potassium in montmorillonite.

The SUnSAL hyperspectral unmixing algorithm utilized in this study is a standard classification algorithm. The USGS Spectral Library was built using high-quality spectrometers that provide reflectance spectra for many minerals. This type of measurement provides a single, homogenized measurement per sample. This library type is designed for standard classification algorithms and is commonly used for spectral unmixing (Fasnacht et al., 2019). Figure 5 shows the standard spectra of eight pure endmembers acquired from USGS and ASTER (https://www.usgs.gov/). ENVI software samples standard endmembers' spectrum data with the resolution of 0.5 nm for the needed spectral library (Fig. 6). The continuum removal may effectively emphasize the spectral curve's absorption, reflection, and emission properties and normalize them to a consistent spectral backdrop, making it easier to compare spectral signature band values with other spectral curves.

3 RESULT

Table 2 displays the iteration number and primitive residue (Res-p) and dual residue (Res-d). The lower the Res-p and Res-d, the more accurate the outcome. The Res-p and Res-d values of flight lines Nos. 1–7 all show that the unmixing results are accurate and reliable, and the relative abundance values may be evaluated and debated.

Figure 7 connects the HU results from different sampling areas on the horizontal axis to more easily observe and compare the relative changes in sedimentary mineral abundance values throughout the phases of salt lake retreat. Halite has the highest average abundance value among the seven flight areas, followed by soil, the illite-smectite mixed layer, dolomite, and chlorite. Carnallite, polyhalite, and celestite emerge in certain flight areas. Because other sediments that do not have a significant spectral absorption signature may occupy the remaining abundance value of the mixed spectrum, the total predicted abundance value for eight endmembers varies between 0.3 and 0.8. However, it is reliable to analyze the evolution of the paleoclimate environment by using the change of relative abundance of sedimentary minerals.

3.1 Abundance value of illite-smectite mixed layer and chlorite

Clay minerals are an important component of clastic sediments. The relative variations in their composition and content can reflect seasonal changes in the humid and dry climates (Fagel et al., 1994). It is an important indicator for determining the salinity of sedimentary environments (Qin, 1987; Thiry, 2000). The clay minerals in the Lop Nur basin are nearly detrital, and studying their composition and content variations can reveal paleoclimate information (Ehrmann et al., 1992). In addition to temperature, pressure, and pH, the formation of the illite-smectite mixed layer is closely related to the concentration of K⁺ (Ren and Chen, 1984). When the temperature factor is not significant, the chemical composition of the mineral pore fluid, especially the content of K⁺ is the main factor in its formation (Zhou, 1994). The increasing K⁺, Al³⁺, and decreasing Na⁺, Ca²⁺ during the conversion of montmorillonite into illite (Altaner and Ylagan, 1997) promotes changes in the structure of the imon mixed layer. Higher temperature and higher salinity of the salt lake are conducive to converting montmorillonite to illite (Honty et al., 2004). The illite-smectite mixed layer in the salt lake shows a climate change into a humid environment (Song, 2009). The main cations of chlorite are Si⁴⁺, Al³⁺, Fe²⁺, Mg²⁺, which are formed in arid climatic conditions and an alkaline environment rich in Mg²⁺. The saline soil solution medium is rich in Mg²⁺, which can provide the necessary material conditions for the formation of chlorite (Jiang and Peacor, 1994). Chlorite is mostly retained in regions where chemical activity is hindered, such as glacial souring and arid surfaces. Nevertheless, it is degraded by weathering and handling activities, indicating cold or dry conditions (Zheng et al., 2016). The paleoclimatic environment should be examined with changes in the relative amount of the two clay minerals. The result shows that the abundance value of the chlorite and illite-smectite mixed layer exhibits undulating wave-like changes (Fig. 8). The relative variation trend of abundance value of the two is the opposite. In the mixed spectrum, the average content of the illite-smectite mixed layer is higher than chlorite, the former is 0.095 5, and the latter is 0.050 8.

3.2 Abundance value of carbonates mineral

Carbonates were first deposited in the lake's sediments, such as calcite and dolomite. The major deposit of the Lop Nur salt lake is dolomite (CaMg(CO₃)₂), which has the lowest solubility among carbonates. Exogenous carbonate and endogenous carbonate are the two types of lacustrine carbonates. Exogenous carbonate is formed due to the weathering of rocks near the lake. It enters the lake by surface runoff as coarse particles. However, the amount of this form of carbonate in Lop Nur's land surface is low (Zhang et al., 2011). Endogenous carbonates include authigenic carbonates and biological carbonates that are primarily influenced by the physicochemical properties and temperature of the lake water. The highly salty and alkaline Lop Nur results in exceptionally low biogenic carbonate sediment composition, the salt lake's major product is authigenic carbonate. The change in carbonate content is influenced by the salinity of the lake water and reveals the dry and wet climate variations indirectly. The internal oxidation and evaporation of the lake are poor under humid climatic conditions and low temperatures, resulting in lower salinity, so the carbonate concentration is minimal. In high-temperature environments, evaporation and oxidation strongly lead to shrinkage of the lake surface and an increase in pH, so carbonate deposition increases (Bischoff and Cummins, 2001). Figure 9 shows that the abundance value proportion of dolomite deposited on ear-shaped stripes fluctuates.

3.3 Abundance value warm-and-cold phase minerals

Polyhalite (K₂Ca₂Mg(SO₄)₄·2H₂O) is a minor mineral found in the magnesium sulfate class lakes. Carnallite (KCl·MgCl₂·6H₂O) has a greater solubility than polyhalite and is formed as a dominant mineral in a magnesium sulfate subtype salt lake. They both only precipitated when the brine was concentrated to a significant concentration. Carnallite is classed as a cold phase salt mineral, celestite is classified as a warm phase salt mineral, and polyhalite is classified as a wide temperature phase salt mineral in the table of cold and warm phase salt minerals (Zhang and Zheng, 2017). Polyhalite precipitation is aided by high brine temperature and concentration (Li et al., 2020), and the presence of it generally indicates brine concentrations that are 60 times greater than those of saltwater (Warren, 1989). As demonstrated in Fig. 10, the abundance value richness of warm-and-cold phase minerals in the top salt crust is lower; however, the three minerals' relative abundance value reveals variations in cold and warm conditions.

3.4 Data validation

Lab-based spectroscopy unmixing results of salt crust samples for validation. This focuses on whether the change of minerals abundance value measured in the laboratory is consistent with that measured in the field in the same sampling sites. Figure 11 shows the unmixing results with lab-based spectroscopy, relative changes in abundance values of chlorite and illite-smectite mixed layer (Fig. 11a), dolomite (Fig. 11b), and warm-and-cold phase minerals (Fig. 11c). The relative changes in mineral spectral abundance values using the lab-based spectroscopy are consistent with that measured outfield in the 7 sampling areas.

Verification using the cation content of dissolved salt crust samples. This focuses on whether the cation content is consistent with the abundance of sedimentary minerals containing such ions in the same sampling site. According to the main cationic ionic of minerals and the relationship between mineral formation process and salt ions in solution medium, chlorite, illite-smectite mixed layer, dolomite, and carnallite are verified, respectively. Figure 12 shows the validation results for each of the four minerals. The variation of chlorite abundance value and Mg²⁺ content converge. The illite-smectite mixed layer abundance value follows the same trend as the K⁺ content and the opposite trend as the Ca²⁺ content. The variation of dolomite abundance value and Ca²⁺ content converge. The carnallite abundance value follows the same trend as the K⁺, Cl^–.

3.5 The physical morphological characteristics of salt crust

The spectral value in the remote sensing image is determined by the physical structure and chemical composition of the target feature. The sedimentary minerals and soil are combined in varying proportions to form the macroscopic morphology of salt crust, resulting in a variance in spectral reflectance. As a result, analyzing the salt crust morphology along the lakeshore is critical in determining the reasonableness of the bright and dark stripes. The salt crust morphology of flight regions Nos. 1–7 is classified using the current salt crust type of the dry saline lake (Zhao et al., 2006; Ma et al., 2011).

The flight areas No. 1 and No. 2 are outside the "auricle" as illustrated in Fig. 13, and have a light-yellow tone on the image. The No. 1 area is a well-developed and even-aged mound-like salt crust, but a thick layer of salt frost remains on the surface. The salt flat and mudflat landforms make up the No. 2 area. The base of the salt crust in this location has been buried by sand and changed to flat land due to wind erosion and leaching. The salt crust in the No. 3 area has a black texture and is in the early stages of breaking. The flat surface is coated in white salt frost and runoff trace, suggesting that the lake water has been salting for some time and replenished by runoff. The No. 4 area is a transitional salt crust between eroded and honeycomb. The dry and rough surface is covered in white salt frost, the cracked salt shell produced in the early stage has been shattered, and the hexagonal ridge has collapsed. The No. 5 area is deeper in color than No. 3 and is a typical mature honeycomb salt crust with considerable land surface roughness and runoff traces. A massive amount of white salt efflorescence covering the honeycomb-shaped salt ridge is clear evidence of degraded caverns, implying long-term salt precipitation and subsurface brine uplift. The No. 7 area is in the late stage of forming a honeycomb salt crust, with an off-white color, and less white salt frost; and repeatedly water-soaked traces and silt-filled eroded caves could be seen on the surface.

4 DISCUSSION 4.1 Paleoclimatic records of eight sediments in the flight areas Nos. 1–7

The types and contents of sedimentary minerals vary in flight areas Nos. 1–7. The characterization of humid climate comprises an increase in the content of the illite-smectite mixed layer and a decrease in the quantity of salt precipitation, according to the paleoclimate implications of sedimentary minerals indicated in Sections 3.1–3.3. A dry climate is distinguished by increased chlorite and dolomite precipitation content. Carnallite and polyhalite content increase in cold climates, whereas celestite appears in warm climates. The comparison of sediment components in seven areas can be used to reconstruct the paleoclimate change process. Figure 14 depicts the average salt and clay mineral abundance values in flight areas Nos. 1–7.

The chlorite content in flight area No. 1 is higher than in the illite-smectite mixed layer, indicating aridity and cold or aridity and warm conditions, the weakly physical weathering and leaching environment preserving chlorite. The chlorite content in flight areas No. 2 and No. 3 declined, while the content of the illite-smectite mixed layer climbed even higher than that of the chlorite. This demonstrates that the environment was dominated by humid weather, and the leaching and weathering impact was considerable. The chlorite and illite-smectite mixed layer composition in flight area No. 4 increased compared to the previous stage, with the greater chlorite content indicating the warmer and dryer environment. Both are slightly reduced in flight area No. 5 indicates increased dry weather. The chlorite content of flight area No. 6 gradually decreased, but the content of the illite-smectite mixed layer gradually increased, indicating that the climate had become humid again. The chlorite content increased while the illite-smectite mixed layer decreased in flight area No. 7, indicating a warm and dry climate during the recession of salt lake. The comprehensive analysis of chlorite and the illite-smectite mixed layer shows that the climate of the lake shoreline changes periodically from warm-dry to humid. The clay minerals record crucial information on the environmental development of the sedimentary area, giving evidence for the restoration of the paleoenvironment. Due to the limits of clay endmember species, extensive analysis of deposited salt minerals is required.

The salt crust in No. 1 area has the highest dolomite content of the seven areas, indicating that the temperature was relatively high and the climate was arid. The oxidation and evaporation of lake water were intense, but the lake surface was reduced due to a lack of external water replenishment. In this case, the lake water evaporated quickly, precipitating dolomite. The humid and cold environment slows down the evaporation and oxidation of water, and the constant runoff recharge reduces the salinity of the lake water. The amount of dolomite deposited in flight areas No. 4 and No. 5 increased over time, indicating increased temperature. The dolomite content in the flight No. 6 area was significantly lower, indicating the dropped temperature and humid condition. The melted snow on high mountains or seasonal floods increased the surface runoff, weakened oxidation of the lake water, and desalinated the lake. The dolomite sediment increased in the No. 7 salt crust area, indicating warming or even a sharp drought in the final retreat phase of the salt lake.

In terms of cold and warm phase minerals, polyhalite, carnallite, and celestite have low abundance values in flight areas Nos. 1–7. During the salt lake retreat phase, a small amount of celestite appeared in flight areas Nos. 1 & 7, indicating the warm climate temperature. The carnallite found in flight regions Nos. 2–4, and 6 suggested a chilly environment with low temperatures. The presence of polyhalite in the Nos. 2, 4, and 6 areas suggests that the mineralized concentration of the lake water intensifies as the temperature around the lake decreases and the external water supply decreases.

In conclusion, the climatic and environmental changes are very periodic in the Nos. 1–7 areas. The sedimentary salt component change patterns in the Nos. 1, 5, and 7 are similar: chlorite concentration is greater than an illite-smectite mixed layer, dolomite content rises, and celestite deposition indicates warm and dry paleoclimate. In the Nos. 2– 4, and 6 areas, the changing trend of salt mineral components are the same. The illite-smectite mixed layer is more than chlorite, dolomite percentage drops, and carnallite deposition implies a cold, humid environment. In the alternating cold-wet and warm-dry environment, the water surface of Lop Nor Lake gradually shrinks until it dries up.

4.2 The coupling relationship between spectral reflectance and mineral information content

The salt crust's surface material composition determines the overall radiation brightness. The content of salt minerals and clay minerals directly relates to the reflectance value. The dominant spectral reflectance factors of bright and dark ear-shaped stripes can be clarified by analyzing the response relationship between spectral reflectance and mineral content. The data were processed to obtain the average surface reflectance values of areas Nos. 1–7. In Fig. 15, the average spectral curves C₁ of salt crust were compared against the abundance value curve C₂ of the salt mineral and abundance value curve C₃ of clay mineral in Nos. 1–7 areas.

The connection between the three curves in the same sample points is shown in Fig. 15. From the curve profiles, C₁ and C₂ fluctuate at the same pace, indicating that the surface reflectance value is positively connected with the content of salt minerals; the greater the salt content, the higher the surface reflectivity. While C₁ and C₃ follow the opposite pattern, indicating that the reflectance of the salt crust is inversely associated with the content of clay minerals and that the larger the content of clay minerals, the lower the surface reflectance. The relative content of salt minerals and clay minerals may be essential in determining the reflectivity of bright and dark stripes.

The relative content change of salt minerals to clay minerals in each flight area is represented by the ratio of the sedimentary salt minerals to clay minerals (R_S/C). According to Table 3, R_S/C and the spectral reflectance of the salt crust have a significant coupling connection. Clay minerals and soils are common in flight areas Nos. 1–7, each region has two clay mineral species, although the number of salt mineral species varies. The bright striped areas have a high salt crust reflectance, many salt minerals such as dolomite and chlorite, and soil content, but the clay mineral percentage is limited, and R_S/C is comparatively big. The dark stripes have low salt crust reflectance, single-species sediments, and just a thin layer of halite and illite-smectite mixing layer; therefore R_S/C is low. The bright and dark ear-shaped stripes revealed a substantial coupling connection with the sediment composition and content.

4.3 The formation mechanism of bright and dark stripes in Lop Nur

Diverse climatic circumstances have resulted in different forms of salt crust micro-geomorphology in Lop Nur. Lop Nur is a closed internal lake that has steadily decreased due to limited external water supply and considerable evaporation. The unusual salt crust landscape of lake shorelines was produced by precipitated sediments combined with soil under diverse climatic circumstances. The height of subsurface brine, surface wind stress, surface precipitation supply, and climatic temperature during the process of lake water halogenation were all the driving variables in the salt crust formation. For example, the early stages of the salt crust are relatively flat, but when the temperature changes alternately between hot and cold, cracks and fissures will appear on the surface due to the thermal expansion and contraction of the substance. Salt crystals are formed in the surface layer of the salt crust and the subsurface brine layer. As the brine rises along the capillaries it attaches to the sand, widening the cracks on the surface of the salt crust, and the continued increase in crystalline salt masses can lift the crust layer, forming upright salt walls or cavities that create shadows on the surface, affecting the surface reflectance. Lop Nor is at the lowest point of the Tarim Basin, where convergence and recharge of runoff erode existing salt crusts, dissolve accumulated minerals, and changes in brine ion concentrations promote the precipitation of fresh minerals so that salt crust components are constantly changing. The varied patterns of salt crust determine the incidence angle and total radiance of light, and the micro-geomorphology of the salt crust is the macroscopic explanation for the bright and dark ear-shaped strips.

The microcrystalline structure of sediments determines the interior microscopic morphology of salt crust. Different chemical formulas of salt crystals lead to different crystal structures: halite crystals are tetrahedral, dolomite crystals are rhombohedral, carnallite, and polyhalite crystals have crystal water, and celestite crystals are tabular, chlorite crystals are layered silicate minerals that can be found in monoclinic, trisopic, or orthogonal (trapezium) systems, with illite and smectite forming the mixed layer. The chemical composition and lattice structure of each mineral vary, as do the signatures of reflection, transmission, absorption, and emission of electromagnetic waves. When the proportions of sedimentary salt minerals to clay minerals in the salt crust change, the sensors receive varied radiation in terms of both brightness and angle. The bright stripes are caused by the comparatively high ratio of sedimentary salt minerals to clay minerals R_S/C, otherwise, the image appears to have black stripes. The moisture content of adsorbed water and crystal water on the surface of bright and dark textures was measured in the previous studies, and the moisture content of adsorbed water and crystal water on the surface of bright and dark textures are different, the sedimentary mineral structure is the main reason for its hue (Kong et al., 2014). Thus, differences in mineral composition and content of the salt crust are microscopic in the formation of bright and dark ear-shaped stripes.

4.4 Reconstruction of the paleoclimatic environment in Lop Nur

Comprehensive analysis of the distribution of numerous sedimentary minerals in the lake shoreline is advantageous for the mutual identification of multiple indicators in the historical period, acquiring more factual information about the paleoclimatic. As shown in Fig. 16, flight areas Nos. 1, 5, and 7 in the images show a bright hue, flight areas Nos. 2–4, and 6 show a dark hue; therefore, in the same hue of the areas, the salt crusts have similar trends in the composition and relative content of sedimentary minerals. The bright ear-shaped stripes indicate the warm and arid paleoclimatic environment, while the dark stripes indicate the cold and humid paleoclimate. The salt lake in the No. 1 area is influenced by a warm and dry environment, whereas the temperature in the No. 2 area is cold and damp. The No. 3 area's chilly and humid environment is exacerbated. The salt lake environment in the No. 4 area is still dominated by cold-wet, the No. 5 area's climate returns to warm and dry. The salt lake environment in the No. 6 area returns to cold-wet. Warm and dry conditions dominate the No. 7 area.

The salt crusts in areas Nos. 1–2 correspond to the early stages of lake retreat. At this stage, the lake area is large, the water level is relatively high, the lake surface fluctuates slowly, and more clay, soil, and rock salt are deposited in the lake basin. Under the influence of alternating warm-dry and cold-wet environments, the lake in Nos. 3 and 7 areas is in a phase of rapid adjustment. The salinity of the lake water is gradually increasing. The humid climatic conditions promote evaporation of shallow lake water into salt, while the dry climate dries out the shallow lake water and continues to promote brine inversion in the subsurface capillaries. As the brine concentration increases, different types, and contents of minerals grow and accumulate to the surface, mixing with clay to form a hard salt crust. The arid climate gradually intensified and the absence of external recharge water led to the rapid drying of the extremely shallow lake in area No. 7. In general, influenced by the alternating warm-dry and cold-wet climatic conditions around Lop Nor, the shallow Lop Nor lake experienced fluctuating changes in water expansion and shrinkage until it dried up completely.

The lake shoreline maintains certain bright and dark stripes for a season or multi-year period. If the light and dark stripes are the results of different seasonal climate changes throughout the year, the following climatic conditions can be explicitly discussed: cold and wet winters; warm and wet springs; dry and hot summers with increased surface runoff from melting snow-capped mountains, leading to a large expansion of the water surface; and dry and windy autumns with rapid evaporation from shallow lakes. If the formation of the lake shoreline is the result of perennial climate change, the climate of the Lop Nor region fluctuates in synchronization with the global climate, with warm-wet and cold-dry cycles throughout, allowing the Lop Nor to produce a variety of sedimentary minerals during its evolution.

5 CONCLUSION

This study collected high-quality, large-scale hyperspectral data of salt crust based on the UAV platform. The SUnSAL spectral inversion model was used to identify sedimentary minerals and their relative abundance values in each sampling area. The unmixing results show the variability of eight sedimentary mineral types and abundance values on bright and dark lake shores. The variation in mineral composition and content reveals the variability in the internal structure and chemical composition of the salt crust, helping to explain the mechanism of the formation of bright and dark ear-like stripes on remote sensing images. The combined effect of salt crust micro-geomorphology and sedimentary mineral component content formed the bright and dark ear-shaped strips. The microscopic explanation for the production of bright and dark ear-shaped stripes is variability in mineral composition and content percentage of salt crust: comparatively high R_S/C causes bright stripes, whilst low R_S/C causes image black stripes. The macroscopic causes are differences in salt crust morphology along the lake shoreline.

A comprehensive examination of paleoclimatic indications of sedimentary minerals and the salt crust micro-geomorphology along the shoreline of the lake indicates that the paleoclimate of the Lop Nor region and its surrounding areas is characterized by alternating warm and dry and cold and wet periods. Bright ear-shaped stripes represent warm and arid paleoclimate, while dark stripes represent cold and wet paleoclimate.

The spectral absorption peaks of sodium chloride, potassium salts, and other sediments are widely present in the spectral band from 1 000 to 2 500 nm. The spectral probe carried by the endmember was limited to 830 nm in this study. Therefore, the aforementioned sedimentary salts have not been added to the endmember spectrum library. It remains to be seen if the existing endmember spectrum library can be augmented by expanding the range of hyperspectral sampling bands to investigate a more thorough distribution of sedimentary minerals. In addition, a more refined historical sequence of paleoclimate and environmental changes in Lop Nor should be further constructed.

6 DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Electronic supplementary material

Supplementary material (Supplementary Tables S1–S4) is available in the online version of this article at: https://doi.org/10.1007/s00343-022-1341-9.