Variation in Glacier Albedo on the Tibetan Plateau between 2001 and 2022 Based on MODIS Data

Abstract

Albedo is a primary driver of the glacier surface energy balance and consequent melting. As glacier albedo decreases, it further accelerates glacier melting. Over the past 20 years, glaciers on the Tibetan Plateau have experienced significant melting. However, our understanding of the variations in glacier albedo and its driving factors in this region remains limited. This study used MOD10A1 data to examine the average characteristics and variations in glacier albedo on the Tibetan Plateau from 2001 to 2022; the MOD10A1 snow cover product, developed at the National Snow and Ice Data Center, was employed to analyze spatiotemporal variations in surface albedo. The results indicate that the albedo values of glaciers on the Tibetan Plateau predominantly range between 0.50 and 0.60, with distinctly higher albedo in spring and winter, and lower albedo in summer and autumn. Glacier albedo on the Tibetan Plateau decreased at an average linear regression rate of 0.06 × 10⁻² yr⁻¹ over the past two decades, with the fastest declines occurring in autumn at an average rate of 0.18 × 10⁻² yr⁻¹, contributing to the prolongation of the melting period. Furthermore, significant variations in albedo change rates with altitude were found near the snowline, which is attributed to the transformation of the snow and ice surface. The primary factors affecting glacier albedo on the Tibetan Plateau are temperature and snowfall, whereas in the Himalayas, black carbon and dust primarily influence glacier albedo. Our findings reveal a clear decrease in glacier albedo on the Tibetan Plateau and demonstrate that seasonal and spatial variations in albedo and temperature are the most important driving factors. These insights provide valuable information for further investigation into surface albedo and glacier melt.

1. Introduction

The Tibetan Plateau is one of the most concentrated regions of glaciers in the low and mid-latitudes [1]. With ongoing global climate change, glacier mass loss on the Tibetan Plateau continues to increase [2], accompanied by a sustained shrinkage in glacier areas. Close to 17.2% of the glaciers included in the first Chinese glacier inventory (CGI-1) dataset have disappeared [3]. The Sixth Assessment Report of the United Nations Intergovernmental Panel on Climate Change (IPCC AR6) indicates that the average mass loss of global mountain glaciers from 2006 to 2015 was approximately 123 ± 24 Gt a⁻¹ [4]. Rapid changes in glaciers in this region have not only impacted glacier runoff and disasters but have also profoundly influenced water resources, ecosystems, and human activities [3,5]. Climatic conditions, reflected by temperature and precipitation, are considered the primary factors leading to the diverse spatial and temporal patterns of glacier changes [1,3,6]. Recent studies have shown that variation in glacier albedo plays a crucial role in glacier melting [7,8,9]. Net shortwave radiation plays a dominant role in supplying energy for glacier melting [10]. A decrease in glacier albedo directly affects the absorption of net shortwave radiation and enhances net radiation, further accelerating glacier mass loss [8,11]. A reduction in glacier surface albedo on the Tibetan Plateau can increase glacier melting by approximately 30–60% in various regions [12]. In the southeast Tibetan Plateau, the contribution of glacier melt due to the reduction in albedo is approximately 30% [13]. The positive albedo feedback mechanism is a crucial factor influencing glacier mass balance [8,11,14].

In recent years, remote sensing and field monitoring have shown that the glacier albedo on the Tibetan Plateau has been decreasing [15,16,17]. Scholars have employed field measurements and Landsat data to examine the daily and seasonal variations in glacier albedo on a single-point scale at the Parlung No. 4 Glacier [18], Urumqi Glacier No. 1 [19,20], and Muz Taw Glacier of the Altai Mountains [21]. Some studies have used MODIS products to capture albedo variability on a regional scale. Among the different products derived from MODIS radiance, the snow cover product MOD10A1 can be used to infer glacier albedo and has been widely applied in monitoring the albedo of the global cryosphere [22,23]. For example, Zhang et al. [12] employed MODIS data to analyze the trend of glacier albedo change in the High Asian Mountains and highlighted a range of 6.8% to +1.8% per decade for the annual average rate of change of glacier surface albedo. Using MOD10A1 data and applying the concept of ‘glacier phenology’, Di Mauro and Fugazza [24] found that Pan-Alpine glaciers exhibit a decrease in minimum albedo at a rate of −0.06 per decade. Ren et al. [25] showed that glaciers in the Western Nyainqentanglha Mountains have a mean albedo of 0.552 ± 0.002 and a decreasing trend of 0.0443 ± 2 × 10⁻⁴ dec⁻¹. Using a geographical detector, Xiao et al. [26] showed that temperature and snowfall are the primary driving factors affecting glacier albedo.

All the aforementioned studies indicate a clear downward trend in glacier albedo, with significant variability and spatial differences. However, current research has overlooked the challenge of assessing albedo changes in debris-covered glaciers because MOD10A1 data can only identify areas covered by ice and snow [27]. Additionally, seasonal variations significantly impact debris-covered areas, resulting in inaccurate albedo calculations. Consequently, we excluded these glacial regions from our analysis. Furthermore, most current studies focus on changes in glacier albedo during the melting season, with a limited understanding of variations during other seasons [12,28]; thus, a comprehensive understanding of albedo variability is lacking. To accurately assess glacier albedo changes, it is essential to exclude the debris-covered portions of glaciers, enabling a thorough understanding of the spatiotemporal distribution characteristics of glacier albedo across the Tibetan Plateau.

This study used MOD10A1 data in linear regression analysis to investigate seasonal variations and multiyear trends in glacier albedo on debris-free glaciers on the Tibetan Plateau over the past 22 years. Unlike previous studies that primarily focus on summer glacier albedo changes, our research specifically addresses seasonal variations, revealing the dynamics of albedo throughout the year. Correlation analysis was conducted to determine the relationship between glacier albedo change and driving factors such as temperature, precipitation, snowfall, black carbon (BC), and dust. Additionally, the relationship between glacier albedo and interannual mass balance across different regions was investigated. To improve the accuracy of albedo trend estimations, we excluded debris-covered glacier areas from our analysis. The results of this study fill a research gap in glacier albedo studies on the Tibetan Plateau and provide crucial insights into the impact of glacier albedo on glacier melt in the region.

2. Study Area

The Tibetan Plateau, with a range spanning from 25°59′N to 40°1′N and from 67°40′E to 104°40′E, is often referred to as the roof of the world or the Third Pole [29] (Figure 1). Changes in glaciers have a profound impact on global water resources, and glacial meltwater serves as a crucial water resource for billions of people living in the areas surrounding the Tibetan Plateau [30]. The Tibetan Plateau’s climate is characterized by scanty precipitation, low temperatures, intense radiation, and significant diurnal temperature variations [31]. Between 1960 and 2012, the rate of temperature increase in the region reached 0.3–0.4 °C per decade, which was more than twice the global temperature rise during the same period [32]. The Tibetan Plateau also experienced an increase in precipitation [31].

According to the Randolph Glacier Inventory (RGI) 6.0, the glacier area on the Tibetan Plateau can reach up to 100,000 km². It can be divided into 12 subregions: Pamir, Hindu Kush, Karakoram, Western Himalaya, Central Himalaya, Eastern Himalaya, Inner Tibet, West Kunlun, East Kunlun, Qilian Shan, Southeastern Tibet, and Hengduan Shan. In this study, glaciers with an area greater than 1.5 km² were carefully selected based on the RGI 6.0 dataset. After excluding debris-covered glacier areas, 8476 glaciers with a total area of 53,107.54 km² were selected for inclusion in our analysis.

3. Data and Methods

3.1. Datasets

3.1.1. Glacier Data

In this study, glacier boundaries were delineated using the Randolph Glacier Inventory (RGI) 6.0 (https://www.glims.org/RGI, accessed on 10 June 2023), which includes the area and altitude of each glacier on the Tibetan Plateau. The glacier mass balance data were sourced from Hugonnet et al. [33] and covered individual glacier elevation changes from 1 January 2000 to 31 December 2019. This dataset features elevation change maps at a 100 m resolution for glaciers and a 10 km buffer zone, available at 5-year intervals (2000–2004, 2005–2009, 2010–2014, and 2015–2019), 10-year intervals (2000–2009 and 2010–2019), and for the entire 20-year period from 2000 to 2019. Additionally, it provides an analysis of the annual mean elevation change rates and their associated uncertainties. This dataset offers a comprehensive estimation of global glacier mass changes, revealing an accelerated pattern of glacier mass loss in the early 21st century, corroborated by field observations [26,34].

3.1.2. MODIS Albedo

We obtained glacier albedo data from the MOD10A1 V061 dataset (https://nsidc.org/data/mod10a1/versions/61, accessed on 31 August 2023), which has a resolution of 500 m. This product first calculates the Normalized Difference Snow Index (NDSI) to separate ice and snow. The MODIS cloud mask was then applied to identify cloud pixels and complete a series of tests to filter uncertain grid cells further. Once identified as ice/snow grid cells, the best daily observations were used to calculate the snow albedo of the corresponding pixels in the MOD09GA surface reflectance product using visible and near-infrared (VNIR) bands, and corrections were applied using various anisotropic factors [35]. The albedo of each glacier was calculated as the arithmetic mean of the individual pixel values.

3.1.3. ERA5-Land

The dataset was employed to explore the correlation between glacier albedo and its driving factors. Recently, the European Centre for Medium-Range Weather Forecasts (ECMWF) introduced the latest global climate reanalysis dataset, ERA5-Land. In contrast to ERA5, this dataset incorporates corrected land use and land cover data, amalgamating model outcomes with global observational data to offer a more precise depiction of past climate conditions. This enhancement in accuracy is achieved through integration with global observational data, leading to improved precision (source: https://cds.climate.copernicus.eu/, accessed on 4 September 2022). Given the intricate topography of the Tibetan Plateau, particularly within glacial regions with relatively sparse automated weather stations, existing meteorological products representing this area may harbor uncertainties, rendering them inadequate for large-scale research endeavors. Several studies have underscored the superiority of ERA5-Land data in capturing temperature and precipitation patterns [36,37,38]. Consequently, this study utilized monthly data for 2 m temperature, total precipitation, and snowfall, featuring a spatial resolution of 0.1° × 0.1° (~9 km) and a temporal resolution spanning from 2001 to 2022.

3.1.4. MERRA-2

The dataset comprising atmospheric black carbon and dust was utilized to explore the correlation between glacier albedo and its driving factors. To achieve this objective, we employed monthly data for Black Carbon Surface Mass Concentration and Dust Surface Mass Concentrations sourced from the M2TMNXAER series within the MERRA-2 dataset, spanning from 2001 to 2022 (available at https://cds.climate.copernicus.eu/, accessed on 2 September 2022). MERRA-2, introduced by the NASA Global Modeling and Assimilation Office (GMAO) in 2017, represents a comprehensive archive of atmospheric reanalysis data, encompassing atmospheric black carbon and dust products dating back to 1980 [39]. This dataset features a spatial resolution of 0.5° × 0.625° and provides temporal resolutions of 1 h, 3 h, and monthly intervals. Notably, MERRA-2 stands out as the pioneering analytical tool to integrate atmospheric radiation with aerosol fields [40].

3.2. Methods

3.2.1. MODIS Gap Filling

The influence of clouds limits effective grid cells on the surfaces of glaciers in MODIS. To increase the proportion of valid grid cells on glaciers, we first discarded the values in MOD10A1 that were labeled as 2-255 [35]. Next, we employed a linear interpolation method to handle data gaps. We selected data within a time aggregation range of two days before and after the missing values and used the average value from the MOD10A1 data to fill in these null values. This interpolation method has been widely applied in similar studies [41,42]. We then calculated the slope and mean albedo for each glacier from 2001 to 2022. Given that glacier sizes vary, with larger glaciers having a greater impact on regional albedo, we applied an area-weighted approach to accurately reflect the regional glacier albedo. This approach allowed us to maximize the use of available data, increasing the proportion of valid grid cells and enhancing the accuracy of our results. This study was conducted using the Google Earth Engine (GEE) platform.

3.2.2. Glacier Debris Processing

In this study, we utilized the glacier debris data provided by Herreid and Pellicciotti [43], which outlines the boundaries of debris-covered glaciers in the Tibetan Plateau region. We integrated these boundaries with the RGI 6.0 glacier dataset and removed the debris-covered areas to focus solely on clean glacier surfaces. After excluding the debris-covered sections, we selected glaciers with an area greater than 1.5 km². This threshold was chosen because the spatial resolution of MODIS data is 500 × 500 m and smaller glaciers may not be adequately represented, potentially leading to inaccuracies in capturing changes in glacier albedo.

3.2.3. Trend Analysis and Correlation Analysis

Linear regression was used to calculate the trend in the glacier albedo change for each glacier from 2001 to 2022. For each glacier, the mean albedo and its rate of change were calculated as the arithmetic averages for each pixel. By calculating the mean and slope for each glacier on the entire Tibetan Plateau or each subregion and then performing area-weighted averaging, we obtained the mean albedo and its rate of change for all glaciers within the entire region. This made it possible to better understand the contributions of the albedo of different glaciers to that of the entire region and obtain more accurate results.

The equations used for the entire Tibetan Plateau (and for subregions) are as follows:

$$y=kx+b$$

(1)

$$mean={\displaystyle \frac{\sum (m_1{a}_1+m_2{a}_2+\cdots {+m}_n{a}_n)}{\sum (a_1+a_2+\cdots +a_3)}}$$

(2)

$$slope={\displaystyle \frac{\sum (k_1{a}_1+k_2{a}_2+\cdots +k_n{a}_n)}{\sum (a_1+a_2+\cdots +a_n)}}$$

(3)

where k is the albedo slope of each debris-free glacier, m is the albedo value, and a is the area.

We used the Pearson correlation coefficient for correlation analysis to explore the relationship between glacier albedo and its driving factors (temperature, snowfall, precipitation, black carbon, and dust).

4. Results

4.1. Spatial Distribution Characteristics of Glacier Albedo

During the period from 2001 to 2022, the area-weighted average albedo of 8476 glaciers on the Tibetan Plateau was 0.576 ± 0.068. The weighted average albedo values were highest in the spring (0.655, March to May), followed by winter (0.611, December to February), autumn (0.559, September to November), and summer (0.516, June to August) (Figure S1). The middle 50% (25th to 75th percentiles) of the glacier albedo values were between 0.50 and 0.60 (Figure S2). The glaciers in the East and West Kunlun Mountains displayed the highest albedo values, and those in the Hindu Kush displayed the lowest. The differences in albedo from region to region are attributable to variations in the meteorological conditions caused by differences in terrain. The regions with the highest albedo values typically had lower average annual temperatures. Conversely, the lower albedo values observed in the Hindu Kush may be attributed to the presence of light-absorbing particles because this area exhibited the highest annual averages of both BC and dust.

Glacier albedo exhibited significant differences in spatial distribution across the different seasons (Figure 2). In spring, most subregions exhibited relatively high albedo values, especially in the subregions of the Pamirs, Karakoram, West Kunlun Mountains, Western Himalayas, and Hengduan Shan, while those in Qilian Shan were the lowest. In summer, areas with high albedo values in spring showed lower values, except for consistently high albedo values in the East and West Kunlun Mountains. The spatial distribution of albedo during autumn was similar to that during summer but with an overall weak increase in values. The distribution of albedo values in winter was similar to that in spring, with relatively high values in the western region. In summary, on an annual timescale, glacier albedo values were lowest during summer and autumn in the Hindu Kush and highest in the glaciers of the West Kunlun Mountains. The albedo values of glaciers in the Western Himalayas were highest during spring and winter.

4.2. Temporal Variability and Trends in Glacier Albedo between 2001 and 2022

From 2001 to 2022, glacier albedo across the Tibetan Plateau exhibited a decreasing trend (Figure 3), with an overall average rate of decrease of 0.06 × 10⁻² yr⁻¹. Glacier albedo decreased most rapidly in autumn (−0.18 × 10⁻² yr⁻¹), followed by winter (0.06 × 10 ⁻² yr⁻¹) and summer (0.01 × 10⁻² yr⁻¹), and increased slightly in spring (0.02 × 10⁻² yr⁻¹) (Figure S1).

On the subregional spatial scale, only the Hindu Kush and Western Himalayas exhibited an annual increasing trend in glacier albedo (0.06 × 10⁻² yr⁻¹); all the remaining subregions exhibited decreasing trends (Figure 3). The glacier albedo in Inner Tibet exhibited the most rapid decrease, at a rate of 0.19 × 10⁻² yr⁻¹, primarily due to the rising temperatures in this region (Figure S3). The rate of decrease in albedo in Karakoram and the Central Himalayas was not significant. Westerly winds may have influenced the rate of change in albedo in the Western Himalayas and the Hindu Kush because these regions are typically cold and dry, resulting in an increasing trend in albedo, further compounded by decreasing temperatures.

The rate of change in glacier albedo also exhibited notable fluctuations across seasons (Figure 3). Over the two-decade study period, glacier albedo in most of the subregions trended upward in spring, but in certain areas, such as the Eastern Himalayas, Hengduan Shan, and Inner Tibet, glacier albedo exhibited a downward trend. During summer, all subregions except the Kunlun Mountains, Karakoram, and the Western and Central Himalayas exhibited a downward trend, with glacier albedo in Inner Tibet decreasing most rapidly (−0.19 × 10⁻² yr⁻¹). In autumn, glacier albedo in all areas of the Tibetan Plateau exhibited decreasing trends, with the most rapid decrease occurring in a subregion of Qilian Shan (−0.32 × 10⁻² yr⁻¹), except for some individual glaciers that exhibited upward trends, primarily associated with a decreasing trend in snowfall. During winter, glacier albedo on the Tibetan Plateau continued to decline, except in the Western Himalayas, where glacier albedo exhibited an upward trend. However, in Qilian Shan, there was again a decreasing trend in glacier albedo in winter compared to autumn (−0.41 × 10⁻² yr⁻¹), making it the region with the fastest rate of decline in glacier albedo within the year and on an interannual basis.

We divided the glaciers in each subregion into different elevation bands at 100 m intervals to further investigate the differences in glacier albedo and its rate of change at different elevations (Figure 4). We observed a significant increase in albedo values with increasing elevation across all subregions, which is consistent with the results of previous studies [20]. The rate of change in glacial albedo varied across different elevations and exhibited diverse trends. In most subregions, the rate of change in glacier albedo followed a V-shaped curve with elevation. Specifically, the rate of decrease in glacier albedo was the highest in the mid-elevation bands. As the elevation increased further, the rate gradually decreased. This pattern was observed in subregions such as the Eastern Himalayas, Hengduan Shan, and Inner Tibet (Figure 4). However, in some regions influenced by westerlies, such as the Himalayas and the Karakoram, glacier albedo exhibited an increasing trend at lower elevations.

5. Discussion

5.1. Sources of Uncertainty

To analyze the spatiotemporal variations and driving factors of albedo, we considered three sources of uncertainty. First, debris-covered glaciers, which are widely distributed on the Tibetan Plateau (~9% of the total area), contribute significantly to uncertainty regarding glacier albedo change and its impact on melting [8,27]. Therefore, we excluded the debris-covered portions of the glaciers to reduce the uncertainty in assessing glacier albedo changes.

In addition, uncertainties in MOD10A1 data may arise from various factors, including the accuracy of the radiative transfer function, which is influenced by elements such as water vapor content and aerosol optical depth, and the accuracy of geolocation [45], particularly for small glaciers where the proportion of mixed grid cells may be higher. The presence of meltwater on glacier surfaces can reduce the albedo of snow and ice, further contributing to uncertainties in the MOD10A1 retrieval of glacier albedo [24]. To address these uncertainties, we compared the albedo derived from L8/OLI and MOD10A1 data over the Dasuopu Glacier (28°22′N, 85°43′E) from 2014 to 2022, finding consistent trends and an effective capture of albedo fluctuations, which confirms that MODIS data are suitable for large-scale studies (Figure S4a). For spatial uncertainties, we compared albedo from L8/OLI and MODIS data on 9 September 2020 for the Zangsegangri Glacier (34°19′N, 85°50′E). The similar spatial distribution of albedo values, with a mean difference of only 0.001, indicates reliable spatial patterns (Figure S4b). These findings confirm that the MOD10A1 product, which has been validated in several studies [21,28,46], effectively captures both temporal and spatial albedo patterns.

Cloud cover reduces the effective detection of glacier albedo, resulting in a decrease in the available glacier albedo data [47,48]. To minimize the impact of cloud cover, we conducted temporal aggregation for each pixel, facilitating subsequent processing and analysis. The effective grid cell rate of the MOD10A1 was 0.54. After applying interpolation within a two-day temporal window and performing data filling, the effective grid cell rate increased to 0.90. After the data-filling step, the modeled glacier albedo was closer to the measured values [49].

5.2. Driving Forces of Glacier Albedo

The variability in glacier albedo in the Tibetan Plateau is primarily influenced by temperature (R = −0.70), snowfall (R = 0.61), and black carbon (R = −0.57), whereas the effects of precipitation and dust are limited to specific areas (Figure 5). Previous research has indicated no significant relationship between glacier albedo and precipitation on the Tibetan Plateau [26]. However, our results reveal that the glacier albedo in Qilian Shan, Hengduan Shan, and Southeastern Tibet is influenced by precipitation (Figure S5). We speculate that precipitation in these regions occurs primarily in the form of snowfall, thereby affecting the variation in albedo. Furthermore, in the East Kunlun Mountains (R = 0.65) and Qilian Shan (R = 0.76), snowfall was the primary driving factor, exerting a more significant impact on glacier albedo than temperature. The relationship between glacial albedo and black carbon (dust) was generally limited in each subregion.

From a monthly perspective (

Table 1. Correlations between albedo and driving factors on a monthly timescale for the Tibetan Plateau.

	Snowfall (mm w.e.)	Temperature (°C)	Precipitation (mm)	Black Carbon (ng m−3)	Dust (ug m−3)
January	0.36	−0.26	0.36	−0.43 *	−0.28
February	0.40	−0.43 *	0.39	−0.34	−0.01
March	0.18	−0.74 **	0.16	−0.56 **	−0.33
April	0.31	−0.76 *	0.26	−0.26	−0.44 *
May	0.34	−0.80 **	0.14	−0.31	−0.42
June	0.52 *	−0.84 *	−0.01	−0.33	−0.45 *
July	0.63 **	−0.81 **	0.31	0.02	0.06
August	0.25	−0.66 **	−0.29	−0.04	0.11
September	0.78 **	−0.82 **	0.36	−0.42	0.14
October	0.39	−0.82 **	0.22	−0.05	−0.09
November	0.28	−0.51 *	0.27	−0.60 **	−0.01
December	0.54 **	−0.53 *	0.54 *	−0.69 **	−0.20

* p < 0.05, ** p < 0.01.

), albedo is primarily influenced by temperature over the entire Tibetan Plateau, except in January. Snowfall and black carbon are the next most influential factors, with snowfall mainly affecting albedo in the summer months and black carbon mainly affecting it in the winter months, which is consistent with the seasonal results obtained (Figure S5). Notably, apart from dust, which is concentrated in spring, all factors showed the strongest correlation with albedo in December.

Excluding the debris-covered glacier sections yielded results that offer profound insights into the annual and seasonal average glacier albedo trends across various subregions of the Tibetan Plateau between 2001 and 2022. Notably, during spring and winter, the albedo values remained high, primarily because of persistent snow cover on the glacier surfaces [50]. Conversely, the albedo values were lower in summer and autumn, presumably because of the reduction or absence of snow cover caused by higher air temperatures. Furthermore, our findings indicate an overall decreasing trend in glacier albedo over the two-decade study period. This resulted in increased solar radiation absorption by the glacier surfaces, leading to accelerated melting of glacier ice and snow. Notably, the most rapid decrease in albedo was observed during autumn and in most subregions, indicating a prolonged melting season and a delayed end to glacier melting compared with previous years. Intriguingly, we observed a slightly increasing trend in glacier albedo during the spring season (March to May) over the 22 years, although this trend was not statistically significant. Spring snowfall provides a thicker snow cover and enhances albedo at the beginning of the monsoon season [51].

In addition, over the 22-year study period, the rates of decline in glacier albedo at mid-elevations exceeded those at lower elevations, and most of the areas where glacier albedo changed most rapidly were near the snow line elevation range. Therefore, the highest rates of change in albedo at mid-elevations may indicate the rise of the snowline and the transition from snow to ice surfaces. Furthermore, the results show that in the Karakoram and Himalayas, the albedo at lower elevations increased. This may be due to the influence of westerlies or the amplification of warming rates with increasing altitude, with temperature changes occurring more rapidly in high-altitude environments than in low-altitude environments [52].

Finally, while most glaciers on the Tibetan Plateau are small, larger glaciers occupy a significant portion of the total glacial area. Our study of the trends in albedo changes across glaciers of varying sizes revealed that glaciers smaller than 50 km² exhibited more pronounced changes in albedo. Those exceeding this threshold demonstrated greater stability in their albedo (Figure S6). This observation is consistent with previous findings [24] that indicated significant albedo trends for pixel counts below five, although 39% of the grid cells did not show a significant trend at 254 pixels. Large glaciers, constituting 1.12% of the total glacier count (those with areas ≥ 50 km²), are distributed across the region, except in Qilian Shan, with the highest concentration in the Karakoram (49.5%). The stability of large glaciers in the Karakoram is attributable to the climate conditions, with decreased summer temperatures and increased precipitation playing crucial roles in maintaining glacial stability [53]. A reduction in summer temperatures leads to an increase in glacial albedo, thereby reducing the rate of glacial melting. Concurrently, increased precipitation supplies substantial snowfall and increases albedo. The combined effect of these climatic features contributes to the enhanced stability of the albedo of the large glaciers of the Karakoram.

5.3. Impact on Mass Balance

From 2000 to 2019, most of the glaciers in the study area exhibited a negative mass balance, indicating a sustained state of mass loss (Figure S7). After excluding the debris-covered parts of these glaciers, the average mass balance for the Tibetan Plateau was determined to be −0.23 ± 0.03 m w.e.yr⁻¹.

Previous studies have explored the relationship between albedo and mass balance for a limited number of glaciers. For example, MODIS data demonstrated a significant correlation between minimum albedo and annual mass balance in the Alps [54]. Further research included 30 glaciers in the French Alps and found a significant relationship between the minimum albedo and annual mass balance [47]. Our results reveal that the correlation between annual average glacier albedo and annual mass balance varied in the subregions of the Tibetan Plateau over the two-decade study period (Figure 6). We found that in most cases, the interannual variations in both glacier albedo and mass balance exhibited similar declining trends and that the average albedo values corresponded closely to the interannual variations in mass balance. In most subregions, the albedo parameters were highly positively correlated with the interannual mass balance, especially in the Eastern Himalayas (R = 0.91) and Inner Tibet (R = 0.90). In other subregions, such as the West Kunlun Mountains and Hengduan Shan, moderate correlations (0.4 < R < 0.8) were found between albedo and mass balance. Interestingly, no strong positive correlations were found between albedo and interannual mass balance for some areas. For example, for the East Kunlun Mountains and the Hindu Kush, the correlations were negative and relatively small (R = −0.21 and R = −0.26, respectively), suggesting that these regions are not suitable for studying the albedo–mass balance relationship [55]. The correlations between annual mass balance and annual albedo for the other subregions of the Tibetan Plateau establish a foundation for future quantitative studies on the impact of albedo changes on glacier melting.

6. Conclusions

Over the past two decades, the albedo on the Tibetan Plateau has shown significant inter-annual variability, with an average value of 0.576 ± 0.068 and a noticeable decreasing trend of 0.06 × 10⁻² per year. Our study demonstrates a significant reduction in glacier albedo, indicating a trend toward darkening, which contributes to an accelerated rate of snow and ice melt. Additionally, by excluding debris-covered glaciers from the analysis, we improved the accuracy of albedo trend estimations. Notably, over the study period, the region exhibited a significant negative trend in glacier albedo during autumn (0.18 × 10⁻² yr⁻¹), further suggesting that the darkening trend could intensify glacier melting and extend the ablation season.

Meteorological factors and light-absorbing particles are both significant drivers of glacier albedo in this region, although their impacts vary. The glacier albedo in summer and autumn is primarily influenced by temperature and snowfall, and black carbon and dust appear to affect only certain regions. Finally, we found varying degrees of correlation between average albedo and mass balance. Furthermore, the strength of these correlations varies across different subregions, reflecting the diverse impacts of regional climatic conditions, glacial characteristics, terrain, and the deposition of atmospheric pollutants. Consequently, a decrease in albedo can accelerate glacier melting to varying degrees in different subregions, thereby revealing the crucial role of albedo in changes in glacier mass balance. This provides a novel perspective for analyzing glacier melting in these areas.

There may be uncertainties in the retrieval of glacier albedo using MOD10A1, which could stem from the accuracy of the radiation transfer function, especially for small glaciers for which the proportion of mixed pixels is larger. In the future, it will be important to integrate optical data, such as Landsat and Sentinel data, to better understand the mechanisms and driving factors of glacier albedo change and the contribution of glacier albedo change trends to glacier melting more accurately and quantitatively.