Changes in Snow Cover and Its Surface Temperature across the Tibetan Plateau Region from 2000 to 2020

Abstract

Currently, the global climate system is complex and ever-changing, with multiple factors influencing climate change. The Qinghai–Tibet Plateau, known as the “Third Pole” of the Earth, is particularly sensitive to global climate change. Without timely and scientific research on the ecological environment of the Qinghai–Tibet Plateau and without summarizing relevant adaptive strategies, global climate change will impact the sustainable development of the plateau. This study utilized Landsat remote sensing images from 2000 to 2020 to extract the snow cover area and snow temperature of the Qinghai–Tibet Plateau using the snow frequency threshold method. The study analyzed the spatiotemporal characteristics of snow cover and temperature over the 20-year period and investigated some of the climate and topographical driving factors influencing their changes. The results showed that from 2000 to 2020, the permanent snow cover area in the Qinghai–Tibet Plateau region showed a fluctuating decreasing trend, reducing from approximately 12.34 thousand km² to around 9.01 thousand km²; the permanent snow temperature showed an initial increase followed by a decrease during the same period. The highest annual average snow temperature was approximately −3.478 °C, while the lowest annual average temperature was around −8.150 °C. Over the 20-year period, the snow cover area in the plateau was negatively correlated with temperature and precipitation, while snow temperature was positively correlated with temperature and precipitation. The snow cover in the weak wind areas of the plateau showed a significant reduction. Areas with higher average wind speeds, such as shaded slopes and semi-shaded slopes, had larger snow cover areas. These research findings provide important insights into the protection and management of the ecological environment of the Qinghai–Tibet Plateau.

1. Introduction

The Tibetan Plateau, known as the “Roof of the World”, the “Water Tower of Asia”, and the “Third Pole of the Earth”, serves as an important ecological security barrier, water resource security barrier, national security barrier, and strategic resource reserve base for China [1]. Leveraging the sensitivity of the Tibetan Plateau to global climate change, conducting research on snow cover in the Tibetan Plateau aims to provide evidence for global climate change, deepen understanding of the mechanisms and impacts of global climate change, predict the changing trends in global temperatures and the frequency of extreme weather events in the coming years, and better understand and address the challenges brought about by global climate change. As an important component of the cryosphere, snow plays a crucial role in the Earth system, making research on snow highly significant and impactful [2,3,4]. The unique radiative properties and thermal characteristics of snow influence the energy balance between the land surface and the atmosphere, as well as regional thermal differences, thereby affecting atmospheric circulation and climate change. Snow is a vital freshwater resource, especially crucial for water supply in arid regions. Studying the distribution and melting patterns of snow can help predict and manage water resources effectively, ensuring adequate water supply in the region. Additionally, snowmelt can lead to natural disasters such as flash floods and mudslides, particularly in mountainous areas. By investigating the melting patterns and factors influencing snowmelt, early warning, and prevention measures can be implemented to safeguard lives and properties. Snow cover also has a significant impact on ecosystems, maintaining soil temperature, protecting plant roots, and influencing the growth and reproduction of organisms. Research on the effects of snow on the environment contributes to the conservation of biodiversity and ecological balance. Changes in snow cover in the Tibetan Plateau serve as sensitive indicators of global climate change [5]. Analyzing its spatial and temporal distribution and variations can improve short-term climate prediction capabilities and reveal the climatic effects of snow cover against the backdrop of global warming [6,7].

In recent years, the application of remote sensing techniques for extracting snow information has been increasing, providing strong support for the dynamic monitoring of snow cover. Landsat satellites have significant advantages in extracting snow information, such as global-scale monitoring and high spatial-temporal resolution. This method has been widely acknowledged by numerous scholars. For example, Yang Jiaxin et al. [8], based on Landsat 7 ETM SLC and Landsat 8 OLI images, extracted and analyzed snow distribution data in the Yunnan region in 2000 and 2020, along with their correlation with elevation, temperature, and precipitation. He Siyu et al. [9] conducted snow monitoring using Landsat 8 clear-sky data. Tian Xiaofei et al. [10] studied seasonal variations in snow cover based on Landsat satellite data. Yang Jia et al. [11] extracted glaciers using the object-oriented-improved snow index method. Chen Gang et al. [12] proposed a method to extract basin snow cover ranges using a combination of SAR and optical remote sensing data. Chen Longfei et al. [13] analyzed the changing trends and responses between snow cover, temperature, and precipitation using remote sensing data. G. He et al. [14] researched snow and snow-free areas using GF-1 satellite data. Yan et al. [15] applied an improved water-snow index based on Landsat for snow extraction. Liao Haijun [16] used single-channel algorithms based on Landsat images to retrieve glacier surface temperature in the Gongga Mountain region. Fan Qiang [17] et al. analyzed the snow cover change trends in Mount Wutai and their correlation with elevation, temperature, and precipitation. Tang et al. [18] analyzed the spatiotemporal changes in snow cover in the Tianshan region and its relationship with meteorological factors, concluding that temperature is the main factor influencing snow cover changes in the Tianshan region. Zhang Weihua et al. [19] extracted spatial distribution data of the Sappu Glacier in Tibet using Landsat and other satellite data and analyzed the glacier’s spatial change characteristics and climate-driven response factors. Li Yali et al. [20] analyzed the basic characteristics of snow cover changes in Mount Hua using Landsat satellite remote sensing image data and their relationship with temperature, precipitation, and atmospheric circulation. Fathian, F. et al. [21] successfully extracted the Land Surface Temperature (LST) of the Urmia Lake Basin in Iran using thermal bands of Landsat TM and ETM+ images.

In conclusion, existing research on snow predominantly focuses on small-scale and short-term snow cover studies [22,23]. The vast expanse of the Qinghai–Tibet Plateau poses challenges as small-scale snow variation studies may not comprehensively reflect the overall snow cover changes across the entire plateau. Moreover, short-term studies may exhibit a degree of randomness, limiting the exploration of long-term climate trends and cyclic variations at a macro level. Research on the 20-year snow cover area, temperature variations, and their climatic responses across the Qinghai–Tibet Plateau remains a key focus and challenge in current scientific research. Furthermore, this study utilized the Enhanced Normalized Difference Snow Index (ENDSI) proposed by Pang [24] et al. for extracting snow cover from Landsat 8 satellite images. Additionally, the snow cover frequency threshold method was employed to determine the optimal threshold for automatic snow identification and extraction, resulting in high accuracy and effective outcomes. Therefore, to study the dynamic characteristics of snow cover area and snow temperature over a large-scale and long-term period, along with their relationships with certain climate and topographic factors in the Tibetan Plateau, this study collected Landsat 5, 7, 8 satellite TM, ETM+, OLI remote sensing images, and relevant climate and topographic data from the Tibetan Plateau region from 2000 to 2020. Initially, the snow cover area and snow temperature over the past 20 years in the Tibetan Plateau were extracted and analyzed for their spatiotemporal characteristics. Subsequently, statistical analysis methods were utilized to investigate the dynamic changes driven by climate and topographic factors. This study found that using the snow frequency threshold method for snow extraction can effectively address the impact of Landsat image quantity on snow extraction. Additionally, new discoveries regarding the trend of temperature changes in permanent snow on the Qinghai–Tibet Plateau were made. However, further research beyond 2020 has not been conducted and needs to be supplemented in the future. This large-scale and long-term study aims to better analyze the interaction between snow cover changes in the Tibetan Plateau and the global or regional climate systems, providing a rational theoretical basis for the protection and sustainable development of the Tibetan Plateau.

2. Materials and Methods

2.1. Overview of the Study Area

The Tibetan Plateau, the largest plateau in China and the highest plateau in the world is known as the “Roof of the World” and the “Third Pole”. The geographical location of the Tibetan Plateau is shown in Figure 1. It extends from the southern edge of the Himalayas in the south to the northern edges of the Kunlun Mountains, Altun Mountains, and Qilian Mountains, with the western part bordered by the Pamir Plateau and the Karakoram Mountains and the eastern and northeastern parts adjacent to the western segment of the Qinling Mountains and the Loess Plateau. It spans approximately 2800 km from east to west and 300 to 1500 km from north to south, with a total area of about 2.5 million square kilometers. The plateau can be divided into regions such as the Qiangtang Plateau, the southern Tibetan valleys, the Qaidam Basin, the Qilian Mountains, the Qinghai Plateau, and the mountainous valleys of the Sichuan–Tibet region. The latitude and longitude of the Tibetan Plateau roughly range from 25° to 40° N and from 74° to 104° E. The natural environment of the Tibetan Plateau is extremely unique, characterized by high altitudes, cold and dry climates, long winters, and short frost-free periods. The temperatures are generally low, with most areas having an average annual temperature of below 5 °C, especially in the northern part of Tibet and the upper parts of the mountains where the average annual temperature is below 0 °C. The average temperatures in January range mostly from −2 °C to −15 °C, while in July, they range from 8 °C to 18 °C, with most areas in the northern Tibetan Plateau being below 8 °C. Due to its geographical location and climatic conditions, the ecological environment of the Tibetan Plateau is extremely fragile, and once damaged, recovery can be very difficult [25,26,27].

2.2. Data Source

(1) Landsat Data.

In this study, Landsat 5, 7, and 8 satellite TM, ETM+, and OLI remote sensing images of the Qinghai–Tibet Plateau region from 2000 to 2020 were selected based on image quality comparison to extract snow cover extent and snow temperature. The Landsat data were sourced from Google Earth Engine [28,29]. Using the Landsat 5 (TM), Landsat 7 (ETM+), and Landsat 8 (OLI) satellite remote sensing data on the Google Earth Engine platform, snow cover area and temperature information in the Qinghai–Tibet Plateau (QTP) were extracted. Since 23 July 1972, the Landsat program of the National Aeronautics and Space Administration (NASA) has launched eight satellites, formerly known as the EARTH Resources Technology Satellite before 1975 and has continuously acquired satellite data of the Earth’s surface for 50 years. Landsat images have a spatial resolution of 30 m, a scan width of approximately 185 km, and cover wavelengths from visible light to the infrared spectrum. A total of 110,260 Landsat images from 2000 to 2020 were utilized in this study, as shown in Figure 2. From 2000 to 2011, Landsat 5 and Landsat 7 satellites were in operation, with the number of Landsat images ranging from 3000 to 4700. Only one Landsat satellite (Landsat 7) was available in 2012, with approximately 2900 Landsat images. After the launch of the Landsat 8 satellite in 2013, there was a significant improvement in data availability. From 2014 onwards, Landsat image data for the QTP stabilized at over 7000 scenes per year.

(2) Climate Data.

This study investigated the climate-driving factors of snow cover area and temperature changes in the Qinghai–Tibet Plateau from 2000 to 2020 using two climate factors: temperature and precipitation. The European Centre for Medium-Range Weather Forecasts (ECMWF) meteorological reanalysis data (ERA5-Land monthly average data from 1950 to present on copernicus.eu) provides monthly air temperature data specifically at 2 m above ground level, with a temporal resolution of one month and spatial resolution of 0.1° × 0.1° [30,31]. The years used for this experiment with this dataset are from 2000 to 2020. The ECMWF meteorological data regularly “reanalyzes” archived observational data to create a global dataset describing recent history of weather, land, and ocean conditions, which is available for free on their website. There is a certain bias in ERA5-Land in terms of precipitation and temperature [32,33], but since this experiment mainly aims to reflect the trend characteristics and conduct correlation analysis and the application of ERA5-Land products in the research of the Qinghai–Tibet Plateau can provide very accurate simulations, this product can be applied in this experiment. The monthly precipitation data were sourced from data shared by scholar Peng Shouzhang on the National Tibetan Plateau Science Data Center platform [34,35]. This dataset is downscaled in China based on the global 0.5° climate dataset published by CRU and the global high-resolution climate dataset published by WorldClim using the Delta spatial downscaling scheme, with a temporal resolution of one month and a spatial resolution of 0.0083333° × 0.0083333° [36,37]. Furthermore, the dataset was validated using 496 independent meteorological observation points, ensuring the reliability of the validation results. Annual average temperature and annual average precipitation were calculated from the monthly temperature and precipitation data;

(3) Terrain Data.

In this study, various terrain factors were compared, and elevation and aspect data from the terrain and landform data of the Qinghai–Tibet Plateau were obtained from the National Tibetan Plateau Science Data Center (https://data.tpdc.ac.cn/, URL (accessed on 7 April 2024)) [38]. These data were sourced from a 30 m × 30 m resolution digital elevation model downloaded from the Geographic Spatial Data Cloud website. Using the surface analysis function of ArcGIS10.3 software, the aspect information of the Qinghai–Tibet Plateau was extracted. The data underwent multiple reviews and audits, ensuring that their completeness, positional accuracy, and attribute precision meet standards, making them of high quality and reliability;

(4) Water Body Data.

In this study, when removing the influence of water bodies on snow extraction, the JRC Yearly Water Classification History product was chosen [39]. This dataset, developed by the Joint Research Centre (JRC) of the European Union, utilizes satellite images from Landsat 5, Landsat 7, and Landsat 8 acquired from 1984 to 2020 to generate a set of annual global land surface water classification maps at a resolution of 30 m. The quality of this dataset is excellent and reliable;

(5) Wind speed data.

This study uses wind speed data provided by the Institute of Soil and Water Conservation, Northwest A&F University [40,41,42], covering the period from 1981 to 2015. The data are based on the annual average pressure data from 824 benchmark and basic meteorological stations in China from 1981 to 2015. After undergoing strict quality control and screening, the data were interpolated in space using the Thin Plate Spline (TPS) method of ANUSPLIN4.2 software with the Digital Elevation Model (DEM) at a national resolution of 1 km. This process generated annual grid point data at a horizontal resolution of 1 km × 1 km for China from 1981 to 2015, ensuring high-quality and reliable data.

2.3. Research Methodology

2.3.1. Removing Clouds from Landsat Images

This study removes clouds from Landsat images by performing bitwise operations on the pixel quality data to select clear sky pixels, thereby achieving the effect of cloud removal. Specifically, it involves bitwise operations on the ‘pixel_qa’ band of Landsat images to identify pixels that represent the clear sky. By filtering out flag bits in the pixel-quality data that represent clear sky pixels, a mask is generated. The original image is then masked based on this mask, resulting in an image with clouds removed.

2.3.2. Extracting Snow Cover Based on the Snow Cover Index Method

(1) Calculate the Normalized Difference Snow Index ($NDSI$) from Landsat 5 and 7 images.

$NDSI$ is an index used for assessing snow detection. It measures the relative magnitude of the reflectance difference between visible light ($Green$) and shortwave infrared ($SWIR$), providing a more accurate description of snow detection by distinguishing snow from most cloud types. The formula for calculating $NDSI$ is

$$NDSI=\left(Green-SWIR\right)/\left(Green+SWIR\right)$$

(1)

where $Green$ represents the reflectance in the green band, and $SWIR$ represents the reflectance in the shortwave infrared band. The $NDSI$ values range from −1 to 1, with higher values indicating a higher degree of snow cover on the ground;

(2) Calculate the Enhanced Normalized Difference Snow Index ($ENDSI$) from Landsat 8 images [24].

$ENDSI$ uses Landsat 8 OLI images as the data source and incorporates the spectral characteristics of snow. By including the blue-violet band B1 (0.433–0.453 μm) and the blue band B2 (0.450–0.515 μm) features, $ENDSI$ is applied to automatically extract snow from OLI images. The formula for calculating $ENDSI$ is

$$ENDSI={\displaystyle \frac{blueviolent+blue+green-3.7swir1}{blueviolent+blue+green+swir1}}$$

(2)

2.3.3. Calculate Snow Cover Frequency

In this study, the snow cover frequency ($SCF$) is calculated to reflect the extent of snow cover in the Qinghai–Tibet Plateau region. $SCF$ is an important parameter used to describe the degree of snow cover on the ground. It represents the frequency of snow occurrence, typically expressed as a percentage. Snow cover frequency is an important indicator in climate research, water resources management, ecological, environmental protection, and other fields. The formula for calculating $SCF$ is

$$SCF=\left({\displaystyle \frac{N}{M}}\right)\times 100\%$$

(3)

where N is the number of times snow cover appears at each pixel location, and M is the total number of images in the time series.

In this study, snow cover with a frequency of 100$\%$ is extracted for further analysis.

2.3.4. Extracting Snow Temperature Based on Single-Window Algorithm

The Single-Window Algorithm is a commonly used method for land surface temperature ($LST$) retrieval. In this study, the surface temperature is calculated based on the satellite-observed thermal infrared radiance brightness temperature and the Normalized Difference Vegetation Index ($NDVI$).

(1) Calculate $NDVI$.

$NDVI$, which stands for Normalized Difference Vegetation Index, is an index used for remote sensing to monitor the growth and activity of vegetation. It is calculated by measuring the reflectance in the red band (visible light spectrum, typically the red band with a wavelength range of approximately 0.63–0.69 μm) and the near-infrared band (near-infrared spectrum, with a wavelength range of approximately 0.76–0.9 μm). The specific formula is

$$NDVI=\left(NIR-R\right)/\left(NIR+R\right)$$

(4)

where $NIR$ represents the reflectance value in the near-infrared band, and $R$ represents the reflectance value in the red band;

(2) Calculate the Fraction of Vegetation Cover ($FVC$).

$FVC$ is an important indicator used to quantify the extent of vegetation cover on the ground. It represents the percentage of the ground covered by vegetation, including grass, shrubs, and trees. In this study, vegetation presence is determined by setting a threshold of 0.05, and interpolation based on the $NDVI$ values is used to calculate the fraction of vegetation cover [43]. The specific formula is

$$FVC=\left(NDVI-{NDVI}_{soil}\right)/\left({NDVI}_{veg}-{NDVI}_{soil}\right)$$

(5)

where ${NDVI}_{soil}$ represents the $NDVI$ value of areas that are completely bare soil or without vegetation cover and ${NDVI}_{veg}$ represents the $NDVI$ value of pixels that are completely covered by vegetation;

(3) Calculate emissivity $c$.

Surface emissivity, also known as emittance, refers to the ratio of the radiation emitted from the surface at the same temperature to that emitted by a blackbody. The formula for calculating emissivity is

$$c=0.004\times FVC+0.986$$

(6)

where $FVC$ represents the Fraction of Vegetation Cover;

(4) Calculate Land Surface Temperature.

Using the Single-Window Algorithm, in combination with radiance brightness, emissivity, and other parameters, the Land Surface Temperature ($LST$) is retrieved. The specific formula is

$$LST=\left(TB/\left(1+\left(\lambda \times \left(TB/1.438\right)\right)\times \log c\right)\right)-273.15$$

(7)

where $TB$ represents the thermal infrared radiance band, $\lambda$ represents the wavelength corresponding to the $TB$ band, and $c$ is the emissivity.

Mask the land surface temperature with the snow data to extract the snow temperature.

2.3.5. Analysis of Driving Factors

(1) Correlation Analysis.

Pearson correlation analysis is a commonly used statistical method used to study the relationship between two variables [44]. It measures the strength of the linear relationship between two variables by calculating the correlation coefficient, helping us understand the degree of their association. In this study, Pearson correlation coefficients are calculated between snow cover area, snow temperature, and the climate factors of temperature and precipitation to assess the strength of their relationships;

(2) Linear Regression Analysis.

Linear regression is a statistical method used to establish the relationship between two or more variables, especially when one variable (the dependent or response variable) is considered to be a linear function of one or more other variables (the independent or explanatory variables). The Least Squares Method is the most common estimation method in linear regression, which finds the best model parameters by minimizing the sum of the squared errors between the predicted values of the dependent variable and the actual observed values. In this study, linear regression analysis using the Least Squares Method is performed between snow cover area, snow temperature, and the climate factors of temperature and precipitation. The significance of the p-value is used to determine the presence of a linear relationship. If a linear relationship exists, the formula relating the dependent and independent variables can be expressed as follows [45,46]:

$$y_i=a{x}_i+b$$

(8)

where $i=1,2,\dots ,n$, $a$ is the linear trend coefficient, and $b$ is the linear regression constant.

In this study, linear regression analysis was applied in the following cases:

(a)

When analyzing the variation of snow cover area over time, $y_i$ represents the snow cover area and $x_i$ represents time;

(b)

When analyzing the relationship between climate factor variables over time, $y_i$ represents the climate factor data and $x_i$ represents time;

(c)

When studying the relationship between snow cover area, temperature, and climate factor variables, $y_i$ represents either the snow cover area or snow temperature, and $x_i$ represents the climate factor variable.

(3) Statistical analysis of the correlation between snow cover area and topographic factors.

By analyzing the proportion of snow cover area on different aspects and elevations, the distribution patterns of snow cover on different aspects are explored to determine the relationship between aspect, elevation, and snow distribution;

(4) Statistical analysis of the correlation between temperature near snow cover and topographic factors.

By analyzing the mean temperature near snow cover on different aspects and elevations, the impact of aspect and elevation on the temperature near snow cover is investigated;

(5) Analysis of the impact of wind on snow distribution.

By analyzing the changes in snow distribution in different wind speed intervals and the influence of circulation on the distribution of snow on the Qinghai–Tibet Plateau, we can explore the role of wind in snow cover.

3. Results

3.1. Extraction of Snow Cover Area and Accuracy Evaluation

In this study, the snow cover extent was calculated based on the snow cover index, and the permanent snow cover range in the Tibetan Plateau was extracted by setting a certain threshold. By adjusting the threshold size multiple times to observe the snow cover situation, it was determined that snow cover with a coverage rate of 100$\mathrm{\%}$ would be used as the permanent snow cover range for this study. Additionally, in some cases, the spectral characteristics of water bodies and snow in certain bands may be similar, leading to errors in distinguishing between snow and water bodies based on the snow cover index. To address this, the study used the JRC Yearly Water Classification History water dataset to mask out water bodies within the snow cover range, significantly improving the accuracy of snow extraction. The removal efficiency of the water body is shown in Figure 3.

In this study, the extracted snow cover range was compared and analyzed with the snow data obtained from the National Aeronautics and Space Administration (NASA) Earthdata platform. By randomly generating 50 points in the NASA snow region and the snow region extracted in this study, the accuracy of snow extraction was determined through visual interpretation. The error originated from the misjudgment rate of identifying non-snow areas as snow, the omission rate of identifying snow areas as non-snow, and the overall accuracy. The specific formula is [47]

$$P_1=m/n$$

(9)

$$P_2=j/k$$

(10)

$$P=1-\left(m+j\right)/\left(n+k\right)$$

(11)

where $P_1$ is the misjudgment rate, $P_2$ is the omission rate, $P$ is the overall accuracy, $m$ is the number of non-snow areas identified as snow, $n$ is the total number of non-snow area samples, $j$ is the number of snow areas identified as non-snow, and $k$ is the total number of snow area samples.

After calculation, we have a false positive rate $P_1\approx 2.5\%$, a false negative rate $P_2\approx 9.5\%$, and overall accuracy $P\approx 96\%$.

3.2. Extraction of Snow Temperature and Accuracy Evaluation

The temperatures extracted through the single-window algorithm will be compared with the temperature products from Landsat. Apart from the low matching degree in 2007 and 2000, the matching degrees are high in other years, with most years exceeding 95%. The two show good consistency. The accuracy assessment results are shown in

Table 1. Evaluation of accuracy in extracting snow temperature.

Year	Match Rate	Year	Match Rate
2020	96.900%	2009	98.900%
2019	99.890%	2008	98.900%
2018	92.770%	2007	68.500%
2017	91.500%	2006	93.907%
2016	96.986%	2005	99.181%
2015	98.916%	2004	99.400%
2014	92.000%	2003	97.968%
2013	90.200%	2001	88.720%
2011	95.020%	2000	67.200%
2010	98.100%

.

3.3. Analysis of the Spatiotemporal Changes in Snow Cover Area

3.3.1. Analysis of the Impact of Different Satellite Data on the Results

As this study calculates the frequency of snow occurrence on the selected images as the result of snow extraction, the method of calculating snow frequency has largely compensated for the impact of different satellite data volumes. As shown in the Figure 4, there is no significant correlation between the snow cover area and the number of images, indicating that the snow extraction method in this study effectively addresses the impact of different satellite data volumes on the results.

3.3.2. Analysis of Spatiotemporal Changes in Snow Cover Area

The permanent snow cover on the Tibetan Plateau is widespread, mainly concentrated on the western side of the Hengduan Mountains, the Nyainqentanglha Mountains, the Himalayas, the Pamir Plateau, the Bayan Har Mountains, and the Qilian Mountains [48]. The distribution of snow cover in these areas is influenced by various factors such as topography and climate, resulting in a unique snow pattern. Overall, the permanent snow cover on the Tibetan Plateau shows a spatial distribution pattern of more in the west and less in the east, as well as more in the south and less in the north. The distribution of permanent snow cover is shown in Figure 5.

This study conducted a statistical analysis of the changes in permanent snow cover area on the Tibetan Plateau over time. The results, as shown in the Figure 6, indicate that the permanent snow cover area on the Tibetan Plateau exhibits a fluctuating decreasing trend over time, decreasing from approximately 12.34 thousand square kilometers to about 9.01 thousand square kilometers. Through linear regression analysis, the relationship between snow cover area and time is represented by the linear equation $y=486.49-0.23x$, demonstrating a linear trend indicating the shrinkage of permanent snow cover on the Tibetan Plateau.

3.4. Analysis of Spatiotemporal Variations in Snow Temperature

The snow temperature extracted using the single-window algorithm shows a trend of initially increasing and then decreasing over time. The snow temperature fluctuated higher from 2000 to 2007 and decreased from 2008 to 2020. The highest annual average snow temperature is approximately −3.478 °C, and the lowest annual average temperature is around −8.150 °C. The changes in snow temperature are shown in Figure 7.

In terms of spatial distribution, snow temperature generally exhibits a pattern of higher temperatures in the south and lower temperatures in the north. The mountainous areas in the northwest and southeast regions show relatively lower snow temperatures compared to other areas, possibly due to factors such as higher elevation. The distribution of snow temperature is shown in Figure 8.

3.5. Correlation Analysis of Snow Cover Area, Snow Temperature, and Climate Factors

3.5.1. Statistical Analysis of Climate Factors in the Tibetan Plateau Region

Based on the annual climate data of the Tibetan Plateau, the interannual variations of temperature and precipitation were analyzed. From Figure 9, it can be seen that the annual average temperature of the Tibetan Plateau shows a fluctuating increasing trend, which corresponds to the global warming trend. The linear regression analysis resulted in a linear relationship between the annual average temperature on the Tibetan Plateau and time, expressed as $y=-72.512+0.035x$, indicating a gradual increase in the annual average temperature. As shown in Figure 10, the precipitation on the Tibetan Plateau shows a fluctuating increasing trend but overall tends to stabilize, showing relatively stable changes compared to temperature. The linear regression analysis resulted in a linear relationship between the annual average precipitation on the Tibetan Plateau and time, represented as $y=-9446.4+4.904x$, indicating a slight increasing trend in annual average precipitation.

3.5.2. Relationship between Snow Cover Area, Snow Temperature Changes, and Climate Response

(1)

Analysis at the annual scale level

The article first analyzed the correlation between snow cover area, snow temperature change, and temperature and precipitation at the annual scale level, with the results shown in

Table 2. Correlation Analysis between Snow Cover Area and Temperature.

	Standardized Coefficient	t	p	VIF	F
Constant	-	1.147	0.266	-	F = 11.038 p = 0.004 ***
Average temperature of the Qinghai–Tibet Plateau	−0.606	−3.322	0.004 ***	1

Note: *** represents significance levels of 1%.

,

Table 3. Correlation Analysis between Snow Cover Area and Precipitation.

	Standardized Coefficient	t	p	VIF	F
Constant	-	4.627	0.000 ***	-	F = 3.559 p = 0.075 *
Precipitation on the Qinghai–Tibet Plateau	−0.397	−1.887	0.075 *	1

Note: *** and * represent significance levels of 1% and 10% respectively.

,

Table 4. Correlation Analysis between Snow Temperature and Temperature.

	Standardized Coefficient	t	p	VIF	F
Constant	-	−3.828	0.001 ***	-	F = 0.084 p = 0.776
Average temperature of the Qinghai–Tibet Plateau	−0.07	−0.289	0.776	1

Note: *** represents significance levels of 1%.

and

Table 5. Correlation Analysis between Snow Temperature and Precipitation.

	Standardized Coefficient	t	p	VIF	F
Constant	-	−1.109	0.283	-	F = 3.41 p = 0.082 *
Precipitation on the Qinghai–Tibet Plateau	−0.409	−1.847	0.082 *	1

Note: * represents significance levels of 10%. The standardized coefficient is the coefficient obtained after standardizing the data, and the VIF value represents the severity of multicollinearity, used to test whether the model exhibits collinearity. p represents the significance level (i.e., p-value), F represents the F-statistic, and t represents the t-statistic.

. Regarding the change in snow cover area, the significance analysis of the driving factor of the annual average temperature of the Qinghai–Tibet Plateau shows a significant p-value of 0.001 ***, indicating significance at the level of rejecting the null hypothesis of regression coefficient being 0; thus, the model meets the requirements. As for the collinearity of variables, all VIF values are less than 10, indicating no multicollinearity issues in the model and a well-constructed model. The formula of this model is $y=-3.27x+32.22$, indicating that as temperature increases, snow cover area will decrease. The significance analysis of the driving factor of the annual precipitation of the Qinghai–Tibet Plateau shows a p-value of 0.075 * from the F-test, indicating non-significance at the level of rejecting the null hypothesis of the regression coefficient being 0; thus, there is no significant correlation between snow cover area change on the Qinghai–Tibet Plateau and annual precipitation. As for snow temperature change, the significance p-values of the annual average temperature and annual precipitation of the Qinghai–Tibet Plateau are 0.776 and 0.082, respectively, indicating no significant correlation.

(2)

Analysis at the monthly scale level

From the above analysis at the annual scale level, it can be observed that there is no significant correlation between snow cover area change on the Qinghai–Tibet Plateau and annual precipitation, and there is no significant correlation between snow temperature and annual temperature and precipitation. Although this correlation analysis provides initial insights, the neglect of seasonal variations in annual averages renders these correlations insignificant. Therefore, based on the analysis at the annual scale level, this study further conducted experiments to analyze the correlation between snow cover area, snow temperature change, and climatic factors in 2000, 2005, 2010, and 2020 at the monthly scale level. The analysis results showed good consistency, as shown in

Table 6. Analysis of the correlation between snow cover area and climatic factors at the monthly scale level.

		Standardized Coefficient	t	p	VIF	R2
2000	Precipitation	−0.858	−5.274	0.000 ***	1	0.736
	Temperature	−0.923	−7.568			0.851
2005	Precipitation	−0.876	−5.755	0.000 ***		0.768
	Temperature	−0.876	−1.736			0.74
2010	Precipitation	−0.92	−7.439	0.000 ***		0.847
	Temperature	−0.9	−6.544			0.811
2020	Precipitation	−0.642	−2.51	0.033 **		0.412
	Temperature	−0.858	−5.004	0.001 ***		0.736

Note: *** and ** represent significance levels of 1% and 5% respectively.

and

Table 7. Analysis of the correlation between snow temperature and climatic factors at the monthly scale level.

		Standardized Coefficient	t	p	VIF	R2
2000	Precipitation	0.869	5.557	0.000 ***	1	0.755
	Temperature	0.97	12.637			0.941
2005	Precipitation	0.848	5.056	0.000 ***		0.719
	Temperature	0.961	10.943			0.923
2010	Precipitation	0.887	6.086	0.000 ***		0.787
	Temperature	0.963	11.229			0.927
2020	Precipitation	0.801	4.231	0.002 ***		0.642
	Temperature	0.954	10.022	0.000 ***		0.909

Note: R² represents the degree of fit of the curve regression, with a value closer to 1 indicating a better fit. *** represents significance levels of 1%.

.

For the change in snow cover area, the significance analysis of the driving factor of temperature on the Qinghai–Tibet Plateau shows a significant p-value of ≤0.001 ***, indicating significance at the level of rejecting the null hypothesis of regression coefficient being 0; thus, the model meets the requirements. Regarding the collinearity of variables, all VIF values are less than 10, indicating no multicollinearity issues in the model and a well-constructed model. The formulas for the selected models for the four years are $y=-2.593x+28.669$ (2000, Standard Error = 0.343), $y=-2.191x+30.834$ (2005, Standard Error = 0.411), $y=-1.946x+32.079$ (2010, Standard Error = 0.297), $y=-4.065x+43.577$ (2020, Standard Error = 0.812), indicating that as temperature increases, snow cover area will decrease. The significance analysis of the driving factor of precipitation on the Qinghai–Tibet Plateau shows a significant p-value of ≤0.033 **, indicating significance at the level of rejecting the null hypothesis of regression coefficient being 0; thus, the model meets the requirements. The formulas for the selected models for the four years are $y=-0.689x+56.719$ (2000, Standard Error = 0.131), $y=-0.62x+54.544$ (2005, Standard Error = 0.108), $y=-0.551x+52.357$ (2010, Standard Error = 0.074), $y=-0.868x+77.823$ (2020, Standard Error = 0.346), indicating that as precipitation increases, snow cover area will decrease.

As for the change in snow temperature, the significance p-values of the annual temperature and annual precipitation on the Qinghai–Tibet Plateau are both ≤0.002 ***, showing a significant positive correlation. For temperature, the formulas for the selected models for the four years are $y=0.622x-3.174$ (2000, Standard Error = 0.049), $y=0.58x-1.997$ (2005, Standard Error = 0.053), $y=0.615x-3.319$ (2010, Standard Error = 0.055), $y=0.56x-4.166$ (2020, Standard Error = 0.056), indicating that as temperature increases, snow temperature will also increase. For precipitation, the formulas for the selected models for the four years are $y=0.159x-9.709$ (2000, Standard Error = 0.029), $y=0.142x-7.567$ (2005, Standard Error = 0.028), $y=0.157x-9.173$ (2010, Standard Error = 0.026), $y=0.14x-9.861$ (2020, Standard Error = 0.033), indicating that as precipitation increases, snow temperature shows an upward trend.

3.5.3. Response Relationship between Snow Cover Area, Snow Temperature, and Topography

(1)

Classification of Different Types of Slopes

Based on the literature review and historical experience, this study classifies slopes into eight categories as follows: East-facing slopes (67.5°~112.5°), Southeast-facing slopes (112.5°~157.5°), South-facing slopes (157.5°~202.5°), Southwest-facing slopes (202.5°~247.5°), West-facing slopes (247.5°~292.5°), Northwest-facing slopes (292.5°~337.5°), North-facing slopes (0°~22.5°, 337.5°~360°), and Northeast-facing slopes (22.5°~67.5°); slopes in the range of 0°~45° and 315°~360° are classified as shady slopes; slopes in the range of 45°~135° are classified as semi-shady slopes; slopes in the ranges of 135°~225° and 225°~315° are classified as sunny slopes and semi-sunny slopes, respectively [49,50].

(2)

Snow Cover Area and Temperature near Snow Cover on Different Aspects

Different aspects directly affect the amount of solar radiation received [51]; generally speaking, sunny aspects receive more solar radiation than shady aspects, leading to relatively less snow cover area on shady aspects. This study analyzed the impact of different aspects on snow cover area and temperature near snow cover, as shown in Figure 11 and

Table 8. Percentage of snow cover area on different aspects.

Aspect	Range of Slope Gradient	Average Wind Speed	Percentage
Shady slope	0°~45° and 315°~360°	2.65628 m/s	0.383
Partial shady slope	45°~135°	2.64035 m/s	0.285
Sunny slope	135°~225°	2.63832 m/s	0.157
Partial sunny slope	225°~315°	2.61886 m/s	0.175

. From the experimental results, it can be observed that the snow cover area on shady aspects and semi-shady aspects account for a large proportion, about 66.8% of the total area, while the snow cover area on sunny aspects and semi-sunny aspects account for a smaller proportion, about 33.2% of the total area. In addition, the research results indicate that the temperature near the snow cover varies with different aspects, with the highest average temperature in flat areas, approximately −4.02 °C, while the average temperatures on other aspects are all below −9 °C.

By analyzing the wind speeds on different slopes, it was found that the average wind speed on shady and semi-shady slopes is slightly higher than that on sunny and semi-sunny slopes. Existing research indicates that wind can alter the accumulation and distribution patterns of snow [52,53,54]. Strong winds may cause snow to be blown away, forming snow drifts, thereby increasing the snow-covered area. Additionally, wind speed may affect the accumulation height and density of snow, necessitating further detailed studies in the future.

In summary, the snow cover area on the Qinghai–Tibet Plateau is significantly negatively correlated with temperature and precipitation; snow temperature shows significant positive correlations with temperature and precipitation. The wind speeds on the shady and semi-shady slopes of the Qinghai–Tibet Plateau are relatively high, which may cause the accumulated snow to be blown away, thereby increasing the snow-covered area. Additionally, the shady and semi-shady slopes receive less solar radiation, resulting in more snow distribution compared to the sunny and semi-sunny slopes; the temperature near the snow is highest on flat slopes, with no significant differences in temperature on other aspects.

(3)

Snow Cover Area and Temperature near Snow Cover at Different Elevations

Elevation is one of the important factors influencing snow distribution and temperature [55]. In order to investigate the impact of elevation, this study divides elevation into five zones: below 4000 m, 4000–5000 m, 5000–6000 m, 6000–7000 m, and above 7000 m. It is observed from Figure 12 that overall, the distribution of permanent snow cover significantly increases with elevation, but due to the fact that areas above 6000 m have smaller areas compared to other elevations, the proportion of snow cover decreases accordingly. In areas below 4000 m, the proportion of permanent snow cover is only 0.18433%; within the range of 4000–5000 m, the proportion of permanent snow cover increases to around 11.12598%; the range of 5000–6000 m has the highest distribution of permanent snow cover, accounting for 68.60277%; the proportion of snow cover in the 6000–7000 m altitude range is approximately 19.82473%; and above 7000 m, the proportion of snow cover is about 0.26219%. Additionally, the study reveals that the temperature near the snow cover decreases with increasing altitude, with temperatures being below 0 °C in areas above 4000 m altitude. The cold environment in these high-altitude areas is one of the reasons for the highest distribution of permanent snow cover in these regions.

3.5.4. The Influence of Wind on Snow Distribution

Wind can influence the distribution and accumulation of snow. Generally, strong winds can blow snowflakes away, reducing snow accumulation. On the other hand, weaker winds can help snowflakes settle and accumulate on the ground. This study analyzed the changes in snow cover on the Qinghai–Tibet Plateau under four wind speed levels based on the multi-year average wind speed. Figure 13 shows that over time, the proportion of snow cover in each wind speed interval decreased to varying degrees, especially in the 6–12 km/h and 20–29 km/h wind speed intervals, with fluctuation values of 1.67231915% and 1.97507392% respectively. This change pattern deviates partially from the general trend, as the snow cover in areas with weak winds showed significant decreases. This phenomenon may be related to factors such as climate change and terrain, and further analysis is needed in the future to draw more accurate conclusions.

Previous studies have shown that snow cover changes are influenced by atmospheric circulation [56]. As shown in Figure 14, the two major circulation systems that affect snow cover changes on the Qinghai–Tibet Plateau are the monsoon and westerlies. With the development of the Indian Ocean South Asian Summer Monsoon, the increasing cloud cover, while blocking the energy from the sun, acts as a “thermal blanket” that offsets the blocking effect on solar radiation, ultimately allowing more heat to reach the snow cover, accelerating its melting. Therefore, in regions mainly affected by the South Asian Summer Monsoon, snow cover has significantly decreased from 2000 to 2020. Under the background of the westerly circulation, during the winter months, the westerly airflow is blocked by the plateau and divided into two branches, south and north. The wind direction mainly affects the snow cover on the western sides of the Qinghai–Tibet Plateau, and the action of the wind, along with the moist air it carries, accelerates the reduction of snow cover in this region.

4. Conclusions and Discussion

This study investigated the spatiotemporal variations of snow cover area and temperature in the Qinghai–Tibet Plateau from 2000 to 2020 and their correlation with climate and topographical factors. The main conclusions are as follows: (1) The snow cover area in the Qinghai–Tibet Plateau showed a fluctuating decreasing trend from approximately 12.34 thousand ${\mathrm{k}\mathrm{m}}^2$ to 9.01 thousand ${\mathrm{k}\mathrm{m}}^2$ during the period, which is consistent with numerous existing studies [57,58,59,60,61]. (2) The snow temperature in the region exhibited an increasing trend from 2000 to 2007, followed by a fluctuating decreasing trend from 2007 to 2020. (3) Snow distribution in the Qinghai–Tibet Plateau is higher on shady aspects and semi-shady aspects and lower on sunny aspects and semi-sunny aspects. (4) The snow-covered areas on the west, northwest, and north-facing aspects in the Qinghai–Tibet Plateau have the lowest average temperatures, while the snow-covered areas on flat aspects have the highest average temperatures. (5) Over the 20-year period, there has been a significant negative correlation between snow cover area and rainfall on the Qinghai–Tibet Plateau, indicating that higher temperature and rainfall lead to smaller snow cover areas and higher snow temperatures. Additionally, snow temperature shows significant positive correlations with temperature and precipitation, indicating that higher precipitation leads to higher snow temperatures. (6) Both snow cover area and temperature in the Qinghai–Tibet Plateau are correlated with elevation: within a certain altitude range, snow cover area is positively correlated with elevation, with areas above 5000 m altitude accounting for approximately 88.68969% of the total snow cover area, indicating that permanent snow cover is predominantly distributed in high-altitude regions. The effects of elevation on snow cover analyzed in this study are consistent with a large body of existing research [62,63,64]. (7) Significant snow cover changes in areas with weak winds. The impact of wind speeds in the 6–12 km/h and 20–29 km/h intervals on snow cover area shows significant fluctuations, with fluctuation values of 1.67231915% and 1.97507392%, respectively.

At the same time, this study also has certain limitations. For example, in terms of data utilization, the dataset used in this study lacks data for the years 2002 and 2012 when extracting snow temperatures, which may not fully capture the changes in snow temperature and their long-term effects. This study conducted a monthly-scale analysis of snow temperature and snow cover area, which is more conducive to capturing detailed changes compared to annual-scale studies. However, an analysis of temperature change trends in different sub-regions of the Qinghai–Tibet Plateau has not been conducted yet and needs to be further supplemented in the future. Furthermore, the research on wind-blown snow in this paper is not detailed enough, and further investigation is needed to obtain better conclusions. In terms of correlation analysis, this study selected four driving factors for attribution analysis, which may not fully reflect all the relevant factors influencing snow cover in the Qinghai–Tibet Plateau. Further research is needed in the future to attribute the changes in snow cover in areas with weak winds and to draw more conclusive results.

In future research, we will continue to delve into more precise scientific methods for extracting snow cover and temperature on the Qinghai–Tibet Plateau, as well as promptly conduct tracking studies on data from 2020 onwards. This will allow for a more comprehensive study of the plateau’s ecological environment and the development of relevant adaptive strategies for the sustainable development of the Qinghai–Tibet Plateau.