Assessment of Glacier Transformation in China over the Past 40 Years Using a China-Specific Glacier Classification System

1. Introduction

China hosts the largest number and total volume of glaciers at mid- and low latitudes globally [1]. These glaciers are concentrated on the Tibetan Plateau (TP) and its surrounding mountain ranges, including the Himalayas, Kunlun Mountains, Karakoram, Tien Shan, Hengduan Mountains, Qilian Mountains, and Altai Mountains. They provide crucial meltwater to downstream regions and contribute to the scenic landscapes, both of which are essential for the socio-economic development of cities in China's arid and semi-arid areas [2,3]. Spanning a vast geographic area, Chinese glaciers exhibit considerable spatial variability in hypsography, morphology, and ice flow dynamics. They are influenced by distinct atmospheric circulations, governed by the Indian monsoon and westerlies, with limited contributions from the East Asian monsoon [4]. Over the past two decades, climate change has driven significant glacier retreat [5], mass loss [6,7], surface velocity slowdown [8], and an increase in extreme geohazards, including large volume detachments [9,10,11] and glacial lake outburst floods [12,13,14]. These changes exhibit considerable spatial and temporal heterogeneity [5,15], driven largely by the diverse characteristics of glaciers and the varying patterns of climate change across China. To better understand these variations, a classification system is crucial for examining the distinct responses of glacier types to climate change. However, few studies have investigated Chinese glacier changes through a classification-based approach, limiting insights into the climate-induced impacts on specific glacier categories.

Glaciers have been categorized based on geographical distribution (e.g. Himalayan or Arctic type), source of nourishment (e.g. central firn cap or dendritic glacier system), thermal distribution (e.g. cold or polythermal type), climatic zone (tropical or continental type), and morphology (e.g. valley glacier or ice cap) [16]. These classification systems are widely used in glaciology for diverse studies. Given the extensive distribution of Chinese glaciers across various climatic zones and mountain ranges, Shi and Xie developed a classification system, referred to hereafter as the Shi-Xie classification, based on climatic and geophysical criteria [17]. This system categorizes glaciers into three types: extremely-continental, subcontinental and maritime. The Shi-Xie classification, subsequently refined with more detailed quantifications by later researchers [18,19,20], has been widely applied in textbooks [20], tourism development [21], and the guidance of typical glacier monitoring programs [22]. This classification is extensively used in both Chinese and HMA glaciological studies [23,24,25]. However, existing studies have largely focused on glacier type characteristics and the application of classification schemes, while lacking long-term, regional-scale assessments of glacier type transitions. Therefore, assessing changes in different glacier types within this classification over multiple periods is crucial. Such an assessment would enhance our understanding of two key aspects: whether glacier changes cause deviations from established criteria thresholds and whether glacier types transition into others.

This study assesses changes in different glacier types in China, based on the criteria defined by the Shi-Xie classification, over the past four decades. We focus on glaciers located at the boundaries between different types, as these are likely to experience rapid change and respond quickly to environmental shifts. Using climate reanalysis data, remote sensing data, and numerical models, we calculate the classification criteria. We then investigate the spatio-temporal changes of these criteria for individual glaciers and examine potential transitions between glacier types over the study period. This paper is structured as follows: First, we describe the study area and the investigated glaciers. Second, we detail the multi-source datasets used and the methods employed for climate data downscaling, equilibrium line altitude (ELA) modeling, and near-surface ice temperature modeling. Third, we analyze the results of the spatio-temporal changes in each criterion over the periods 1981-2000 and 2001-2020. Finally, we discuss the mechanisms driving the observed glacier changes and present our conclusions.

2. Study Glaciers

Chinese glaciers are distributed across the HMA regions, which include the Altai, Tien Shan, Pamir, Karakoram, Kunlun Mountains, Qilian Mountains, Himalaya, inner TP, southern and eastern TP and Hengduan Mountains. According to the second Chinese Glacier Inventory (hereafter CGI2), the total glacier area was approximately 43,087 km² [1]. Extremely-continental, subcontinental, and maritime glaciers make up 32%, 46%, and 22%, respectively, of the total glacier area. To better understand the glacier type changes over the past four decades, we selected representative glaciers for detailed analysis based on the following criteria. First, all three glacier types should be included to ensure comprehensive representation. Second, glaciers from various mountains should be chosen to capture the diversity of climate regimes and glacier characteristics. Third, glaciers should be valley types with areas ranging from 10 to 40 km², excluding surging glaciers. Lastly, glaciers should be situated at the boundaries of the three Chinese glacier types to capture the likely temporal variability of glacier characteristics. Ultimately, 80 glaciers were selected (Figure 1, Table S1). Among these selected glaciers, six have relatively long-term and comprehensive observations and are used to calibrate and validate our calculated glacier criteria (Figure 1, Table S2).

3. Data and Methods

3.1. Shi-Xie Classification Method

The Materials and Methods should be described with sufficient details to allow others to replicate and build on the published results. Please note that the publication of your manuscript implicates that you must make all materials, data, computer code, and protocols associated with the publication available to readers. Please disclose at the submission stage any restrictions on the availability of materials or information. New methods and protocols should be described in detail while well-established methods can be briefly described and appropriately cited. In this section, we describe the data and methods used to assess the changes in Chinese glaciers based on the Shi-Xie classification [17,26]. This classification system divides glaciers into extremely-continental (EXT), subcontinental (SUB), and maritime (MAR) types using five key glaciological and climatological criteria. These criteria include: (1) the annual mean temperature at the ELA (T_a), (2) the annual total precipitation at the ELA (P_a), (3) the summer mean temperature (June-August) at the ELA (T_s), (4) the annual mean ice temperature at a depth of 20 meters at the ELA (T₂₀) and (5) the maximum velocity along the glacier centerline (v). Each glacier is classified according to specific thresholds as listed in Table 1, which collectively reflect its climatic and dynamic conditions.

3.2. Glacier Outlines

We selected representative Chinese glaciers based on the criteria described in Section 2, using the CGI2 as our primary dataset. The Randolph Glacier Inventory Version 6.0 (RGI 6.0) was not used for selection purposes because it lacks information on glacier national affiliation and omits some glaciers located in China. To extract the main centerlines of the selected glaciers, we utilized the publicly available global glacier centerline dataset generated by the Open Global Glacier Model (OGGM) [27], which is based on RGI 6.0. Fortunately, the Chinese glaciers we selected from CGI2 are also included in RGI 6.0, allowing for cross-referencing between the two datasets. By matching each glacier in CGI2 with its corresponding RGI ID, we extracted the main centerline for each glacier. These centerlines were subsequently used to sample surface elevation and glacier surface velocity along the glacier flowlines. In addition, the boundaries of different glacier types defined in the Shi-Xie classification scheme were manually digitized based on the map provided in [26].

3.3. DEMs

DEMs are used to derive the elevation of grid cells in the reanalysis dataset and to determine the ELAs. The elevation data for the reanalysis grid is obtained from SRTM30_PLUS, a global topography and bathymetry dataset with a spatial resolution of 30 arc seconds (approximately 1 km) [28]. SRTM30_PLUS was developed using multiple data sources, with land topography derived from SRTM and ICESat data. The ELA is computed by the Open Global Glacier Model (OGGM) through surface mass balance modelling [29]. For this purpose, we configured OGGM to use the SRTM V4.1 DEM with a spatial resolution of 90 m to derive the ELA.

3.4. Equilibrium Line Altitude

ELA is the altitude on a glacier where ablation equals to accumulation for the year. According to the classification criteria outlined in Table 1, ELA plays a crucial role in deriving the basic glaciological and climatological characteristics in Shi-Xie classification. We calculate the ELAs over the period 1980-2019 by modelling the annual mass balance using the OGGM[29]. OGGM utilizes the temperature-index method to model the surface mass balance with climate forcing inputs from GSWP3-W5E5 (see Section 3.4) [30]. To derive the ELA in a hydrological year, we first model the monthly mass balance along the glacier elevation bands. The surface mass balance $B_i\left(z\right)$ in each elevation band z and month i is calculated as follows:

$$B_i\left(z\right)=P_i^{\mathrm{s}\mathrm{o}\mathrm{l}\mathrm{i}\mathrm{d}}\left(z\right)-f_{\mathrm{s}\mathrm{n}\mathrm{o}\mathrm{w}/\mathrm{i}\mathrm{c}\mathrm{e}}\max \left(T_i\left(z\right)-T_{\mathrm{m}\mathrm{e}\mathrm{l}\mathrm{t}},0\right),$$

(1)

where $P_i^{\mathrm{s}\mathrm{o}\mathrm{l}\mathrm{i}\mathrm{d}}$ is the monthly solid precipitation, $f_{\mathrm{s}\mathrm{n}\mathrm{o}\mathrm{w}/\mathrm{i}\mathrm{c}\mathrm{e}}$ are the degree-day factors for ice or snow, $T_i$ is the monthly temperature, and $T_{\mathrm{m}\mathrm{e}\mathrm{l}\mathrm{t}}$ is the threshold temperature above which ice melt is assumed to occur (here set to -1°C). The surface air temperatures at different elevation bands were extrapolated using a constant monthly lapse rate -6.5 K km^-1. The ELA is then determined at the altitude where the annual surface mass balance equals zero [29].

3.5. Meteorological Data at ELA

Meteorological data are essential for deriving glaciological and climatological indices in Shi-Xie classification. In this study, we extract 2 m air temperature and precipitation from the ERA5-Land climate reanalysis dataset. ERA5-Land provides higher spatial resolution (9 km grid) and temporal resolution (up to hourly) on a global scale, offering advantages over ERA5 and ERA-Interim [31]. Many studies have evaluated the performance of ERA5-Land in the HMA mountain regions, finding that it effectively captures the spatio-temporal patterns and extremes of temperature and precipitation [32,33,34].

For surface mass balance modeling and ELA calculation, OGGM uses the bias-adjusted ERA5 reanalysis data, specifically the GSWP3-W5E5 dataset, as its standard climate input. This dataset combines W5E5 data with homogenized GSWP3 data, providing global coverage at a 0.5° spatial resolution [30,35]. The high resolution and robust bias correction of GSWP3-W5E5, particularly for extreme values, make it well-suited for accurately modeling glacier surface mass balance.

To calculate the 2 m air temperature ($T_{\mathrm{E}\mathrm{L}\mathrm{A}}$) and precipitation ($P_{\mathrm{E}\mathrm{L}\mathrm{A}}$) at the ELA of each glacier, we use the ERA5-Land monthly averaged data spanning the period from 1980 to 2020. The closest ERA5-Land grid cell to each glacier's centroid is referred to as the reference grid, serving as the basis for these calculations. We first extract the time series of 2m temperature at the reference grid cell ($T_{\mathrm{r}\mathrm{e}\mathrm{f}}$). The elevation of the reference grid ($z_{\mathrm{r}\mathrm{e}\mathrm{f}}$) is then determined by masking the SRTM30_PLUS DEM. Assuming a constant lapse rate (${\gamma }_T$) of -6.5 K km^-1, $T_{\mathrm{E}\mathrm{L}\mathrm{A}}$ is computed through linear interpolation as

$$T_{\mathrm{E}\mathrm{L}\mathrm{A}}={T_{\mathrm{r}\mathrm{e}\mathrm{f}}+\gamma }_T\left(z_{\mathrm{E}\mathrm{L}\mathrm{A}}-z_{\mathrm{r}\mathrm{e}\mathrm{f}}\right).$$

(2)

$P_{\mathrm{E}\mathrm{L}\mathrm{A}}$ calculation follows the interpolation method proposed by Huss and Hock [36]. We calculate the precipitation gradient using a linear regression method based on precipitation and elevation data from the Level-7 sub-basins of Hydro BASINS [37] ( Figure S1). We use the following equation to convert the relative precipitation gradient (${\gamma }_{\mathrm{r}\mathrm{e}\mathrm{l}}$; unit: % per 100 m):

$${\gamma }_{\mathrm{r}\mathrm{e}\mathrm{l}}={\displaystyle \frac{{\gamma }_{\mathrm{a}\mathrm{b}\mathrm{s}}}{{\overline{P}}_{basin}}},$$

(3)

where ${\gamma }_{\mathrm{a}\mathrm{b}\mathrm{s}}$ is the absolute precipitation gradient (unit: mm per 100 m), ${\overline{P}}_{basin}$ is the average precipitation over the basins [38]. $P_{\mathrm{E}\mathrm{L}\mathrm{A}}$ is calculated as:

$$P_{\mathrm{E}\mathrm{L}\mathrm{A}}={\alpha }_P\cdot P_{\mathrm{r}\mathrm{e}\mathrm{f}}\left(1+\left(z_{\mathrm{E}\mathrm{L}\mathrm{A}}-z_{\mathrm{r}\mathrm{e}\mathrm{f}}\right)\cdot {\gamma }_{\mathrm{r}\mathrm{e}\mathrm{l}}\right),$$

(4)

where $P_{\mathrm{r}\mathrm{e}\mathrm{f}}$ is the precipitation at the reference grid cell, and ${\alpha }_P$ is a correction factor to account for potential biases in the precipitation data.

3.6. Near-Surface Ice Temperature

The near-surface ice temperature (0-20 m depth) is influenced by the atmosphere and associated heat-water transport process. To estimate the long-term impacts of climate change on near-surface thermal state at the ELA, we only consider the heat conduction process and assume that the ice temperature at 20 m depth (T₂₀) does not fluctuate with seasonal atmosphere change. A one-dimensional heat conduction model is used to simulate the near-surface temperature:

$${\displaystyle \frac{\partial T}{\partial t}}={\displaystyle \frac{k}{\rho c_p}}{\displaystyle \frac{{\partial }^{2}T}{\partial z^2}},$$

(5)

where T is the ice temperature, t is the time, k is the thermal conductivity of ice, ρ is the ice density, and c_p is the specific heat capacity of ice. Equation (5) is numerically solved using the finite difference method. We first assume the initial ice column temperature at the ELA is isothermal and equals to the annual mean surface air temperature of last year. The basal boundary condition is set to zero heat flux. We then use the daily ERA5-Land air temperature to force this one-dimensional temperature model over the hydrological year. We take the modelled to represent the near-surface thermal state at the ELA.

3.7. Glacier Surface Velocities

Glacier surface velocities are the sum of contributions from both ice deformation and basal sliding. Monitoring the long-term changes in surface glacier velocity can enhance our understanding of glacier dynamics and their evolution. Due to logistics challenge and remoteness, glacier velocity measurements in China are restricted to a few glaciers and are not continuous both in space and in time [39]. Benefiting from achievements in remote sensing techniques, a few global-scale glacier velocity products are now publicly available [40,41,42]. We use the long-term time series of ITS_LIVE (Inter-Mission Time Series of Land Ice Velocity and Elevation) product to derive the velocities of individual glaciers. The ITS_LIVE offers high-resolution measurements of glacier and ice sheet surface velocities with data spanning from 1985 to 2018. It uses the auto-RIFT processing chain to derive glacier velocities from Landsat 4, 5, 7, and 8 satellite imagery. In this study, we use the yearly ITS_LIVE velocity dataset (240 m horizontal resolution) over the period 1985-2018. From this dataset, we extract the maximum flow speeds for each glacier along its centerline. These glacier centerlines were generated by OGGM using the method proposed by [43].

4. Results

4.1. Calibration and Validation

Precipitation in mountainous regions is spatially heterogeneous due to rugged topography and is less well-constrained because of sparse rain gauge data. To address this, we calibrated the parameters ${\gamma }_{\mathrm{r}\mathrm{e}\mathrm{l}}$ and ${\alpha }_P$ used in our precipitation interpolation method, using observed precipitation data from six glaciers (see Table 2). The results show that the adjustment factor ${\alpha }_P$ ranges from 0.60 to 1.15, with higher values observed for maritime glaciers, such as the PL4 and HLG glaciers, compared to other glaciers. The precipitation gradient ${\gamma }_{\mathrm{r}\mathrm{e}\mathrm{l}}$ varies from -3.4%/100 m to 4.5%/100 m, reflecting significant spatial variability. Notably, ${\gamma }_{\mathrm{r}\mathrm{e}\mathrm{l}}$ is negative for maritime glaciers and positive for continental glaciers. For the subsequent analysis, we used the average value derived from these six glaciers for precipitation interpolation. A sensitivity analysis of the impact of ${\gamma }_{\mathrm{r}\mathrm{e}\mathrm{l}}$ on the derived glacier criteria in the Shi-Xie classification is provided in Section 5.

The ELA is critical for calculating key criteria (e.g., T_a) in the Shi-Xie classification. To validate the modelled ELAs from OGGM, we compared them with observed ELAs from Chinese glaciers. Observed ELA data for calibration were obtained from six representative glaciers across different glacier types and regions (Table S3). As shown in Figure 2, the modelled ELAs generally agree well with the observations. Some deviations may stem from uncertainties in the reanalysis climate data, downscaling methods, and degree-day factors. Despite these discrepancies, the modelled ELAs offer a reliable foundation for deriving the criteria used in the Shi-Xie classification.

In addition to validating the modelled ELAs, we compared the calculated glacier criteria at the ELA, including T_a, P_a, T_s, and T₂₀, with available observations. Figure 3 showed generally strong agreement with observations. For most of the criteria, the correlation was high, with R² values ranging from 0.61 to 0.92. The strongest correlations were observed for T_s and T₂₀, indicating that the calculated values closely match the observed data. Despite some variability, particularly in T_a, the overall results demonstrate the reliability of the calculated criteria used in the Shi-Xie classification framework.

4.2. Spatiotemporal Distribution of ELAs and Glacier Criteria

We assessed the spatial distributions of the classification criteria for the periods 1982-2000 and 2001-2019, and compared their temporal changes between these two periods. Although the ELA is not an indicator in the Shi-Xie classification, it is essential for deriving the other criteria. Therefore, we present its spatio-temporal distributions alongside the other measures.

4.2.1. Distribution over the Period 1982-2000

• ELA

The spatial distributions of ELAs and the five classification criteria for the study glaciers during the period 1982-2000 are shown in Figure 4a. The ELA distribution exhibited zonality both latitudinally and longitudinally. Overall, ELAs decreased from the Himalayas to the Altai Mountains, with values ranging from 6138 m to 3247 m, while they increased from the eastern periphery of the TP toward its inner regions and the western Himalayas, with values ranging from 4800 m to 6100 m. The lowest ELAs were found in the Tien Shan and Altai Mountains, where the mean value was approximately 4,000 m. Other regions with relatively low ELAs included the Pamir, Karakoram, Kunlun Mountains, Qilian Mountains, Hengduan Mountains, and eastern TP. The highest ELAs were concentrated in the inner TP and the Himalayas, with values generally around 6000 m.

The ELAs of extremely-continental glaciers were the highest among the three glacier types, with a mean of 5459 m. In comparison, the mean ELAs for the subcontinental and maritime types were 5125 m and 5302 m, respectively. Each glacier type exhibited substantial ELA variability across its wide geographical range. Subcontinental glaciers, which span the largest geographical area from the Himalayas to the Altai Mountains, showed the greatest variability, with ELAs ranging from 3247 m to 6138 m and a standard deviation of 811 m. Maritime glaciers, primarily located in the eastern Himalayas, southern and eastern TP, and Hengduan Mountains, exhibited the lowest variability, with a standard deviation of 372 m. The ELAs of extremely-continental glaciers generally increased toward the inner TP, reaching a maximum of 6137 m.

2.

Annual mean temperature at ELA (T_a)

The distribution of T_a (Figure 4b) shows that the highest T_a values were found in the southeastern part of the study region, including the Hengduan Mountains, eastern Himalayas, and southern and eastern TP, where glaciers were predominantly maritime. In contrast, the lowest ​T_a values were observed in the Pamir, Karakoram, and Kunlun Mountains. The eastern and central Himalayas, as well as the Tien Shan, also exhibited relatively high T_a​ values. Glaciers in the inner TP displayed moderate T_a​ values compared to other regions.

When compared to the thresholds of T_a​ for individual glacier types, we found that more than one-third of the glaciers shifted from their respective thresholds, with 14 glaciers exceeding and 14 glaciers falling below their thresholds. The glaciers with T_a ​values below their corresponding thresholds were primarily maritime (9 glaciers) and subcontinental types (5 glaciers), concentrated in the Karakoram and the southeastern part of the study region, including the Nyainqentanglha, Hengduan Mountains, and eastern Himalayas. In contrast, the glaciers with T_a​ values exceeding their thresholds were mostly extremely-continental (12 glaciers, more than 85%) and two subcontinental glaciers. These extremely-continental glaciers were located in the inner TP, Kunlun Mountains, Pamir, and Qilian Mountains, while the two subcontinental glaciers exceeding the threshold were dispersed across the eastern Tien Shan and eastern Himalayas.

3.

Summer mean temperature at ELA (T_s)

The spatial pattern of T_s​ (Figure 4c) closely resembled that of the annual mean temperature T_a​ (Figure 4b). Over 60% of glaciers had a summer mean temperature above 0°C (T_s>0℃), spanning various regions of China and encompassing different glacier types. Almost all maritime glaciers (13 out of 16) exhibited T_s>0℃. Similarly, nearly all subcontinental glaciers in the Tien Shan and Altai Mountains had T_s>0℃. Extremely-continental glaciers with T_s>0℃ were predominantly located at the boundary between extremely-continental and subcontinental glacier types. In contrast, extremely-continental glaciers with T_s<0℃ were mainly concentrated in the Kunlun Mountains, inner TP, and Pamir. Subcontinental glaciers with T_s<0℃ were primarily located in the Karakoram and Himalayas.

Compared to the T_s thresholds in the Shi-Xie classification, we found that more than one-third of the glaciers had T_s values surpassing their respective thresholds (28 glaciers), while 19% had T_s values falling below the thresholds (15 glaciers). The remaining 46% of glaciers (37 glaciers) stayed within their corresponding threshold limits. Among the glaciers whose T_s values went above the threshold values, the majority were extremely-continental glaciers (79%), predominantly located in the inner TP and Kunlun Mountains. For glaciers with T_s values below the specified thresholds, 80% were subcontinental glaciers, mainly found in the Himalayas and Karakoram, with only 3 maritime glaciers and no extremely-continental glaciers.

4.

Annual precipitation at ELA (P_a)

Figure 4d shows the spatial distribution of P_a, which exhibits a distinctly polarized pattern. P_a values were generally above 1100 mm in Hengduan Mountains, with extreme values exceeding 2000 mm, while in the western Kunlun Mountains, they were less than 400 mm. High P_a values (>900 mm) were concentrated in the southeastern part of China's glacierized regions, including the Hengduan Mountains, eastern Himalayas, and Nyainqentanglha, where glaciers were predominantly of the maritime type. In contrast, regions with low precipitation (<400 mm) were found in the Kunlun Mountains, inner TP, Karakoram, western Himalayas, and Qilian Mountains, where glaciers were primarily of the extremely-continental type.

During the period 1982-2000, the P_a values of 70% of the glaciers (56 out of 80) remained stable within their respective thresholds. Approximately 13% of glaciers exhibited increasing precipitation trends, with P_a values surpassing their corresponding thresholds. These glaciers were mainly located in the inner TP, with a few scattered in the Himalayas, Kunlun Mountains, Qilian Mountains, and Tien Shan. Meanwhile, 17% of glaciers showed a decrease in P_a, with values falling below their threshold limits. These glaciers were primarily subcontinental types, located in the Karakoram, western Himalayas, Qilian Mountains, and Tien Shan.

5.

Annual mean 20m ice temperature (T₂₀)

The spatial pattern of T₂₀ over the period 1981-2000 closely mirrors that of T_a (Figure 4e). High T₂₀ areas were concentrated in the southeastern part of the study region, coinciding with regions of high T_a. In contrast, low T₂₀ values were found mainly in the Karakoram, Kunlun Mountains, and Pamir mountains, which are among the colder glacierized regions in China.

When comparing T₂₀ values to the thresholds for each glacier type, we observed that the changes in T₂₀ were not consistent with those in T_a. Approximately 42.5% of glaciers remained stable within their respective T₂₀ thresholds, primarily located in the inner TP, Kunlun Mountains, Tien Shan, and Himalayas, and mostly belonging to the extremely-continental type. Around 45% of glaciers experienced a decrease in ice temperature, falling below their respective T₂₀ thresholds. These glaciers were primarily located outside the extremely-continental glacier zone. All maritime glaciers and 60% of subcontinental glaciers had T₂₀ values below their thresholds. Only about 12.5% of glaciers (10 out of 80) surpassed their T₂₀ thresholds and became warmer; these were all extremely-continental glaciers located in the Kunlun Mountains, Pamir, inner TP, and Qilian Mountains.

6.

Maximum glacier surface velocity (v)

Figure 4f illustrates the spatial distribution of v along glacier centerlines. Overall, glacier surface velocities across the study region were generally slow, with nearly 90% of glaciers having velocities below 50 m a^-1. Regions with faster glacier flow were primarily concentrated in the maritime zone, particularly in the Hengduan and Nyainqentanglha mountains. In other mountainous areas, glacier surface velocities exhibited heterogeneous patterns.

When compared to the velocity thresholds defined in the Shi-Xie classification, most glacierized regions (85% of glaciers) experienced significant slowdowns, falling below the thresholds for their respective glacier types during the period 1981-2000. Only about 15% of glaciers remained stable within their thresholds, and these were scattered across all three glacier types. Notably, no glaciers exhibited speedup during this period.

4.2.2. Distribution over the Period 2001-2019

Figure 5 shows the spatial distributions of ELAs and classification criteria for the period 2001-2019. The ELAs of all glaciers increased during this period by 21-205 m compared with the period 1982-2000 (Figure 4a).

The values of the criteria at the ELA vary dynamically as the ELA itself shifts in response to environmental changes. As the ELAs rises, the 2m air temperature at the ELAs decreases due to the adiabatic lapse rate, which partially offsets the effects of climate warming. Consequently, the criteria T_a, T_s, and T₂₀ for 2001-2019 (Figure 5b, c, and e) show similar positive or negative variations relative to the classification thresholds as those for 1982-2000 (Figure 4b, c, and e). For the annual precipitation averaged over 2001-2019, it shows significant increase partially due to the ELAs rise (Figure 5d and Figure 4d). However, the spatial pattern of P_a also exhibits comparable increases or decreases relative to the classification thresholds, consistent with those observed for the period 1982-2000. When compared to the thresholds of surface velocity (v), the spatial pattern of v during the period 2001-2019 was similar to that observed in 1982-2000 (Figure 5f and Figure 4f). Most glaciers experienced a slowdown, with their velocities falling below the respective thresholds. Only seven glaciers maintained stable velocities within their thresholds, while two subcontinental glaciers in the Kunlun Mountains exhibited acceleration, surpassing their threshold values.

4.2.3. Changes Between the Two Periods

We further analyzed the differences in the ELA and the five classification criteria between the periods 2001-2019 and 1982-2000 (Figure 6). Glaciers in the Qilian Mountains, Karakoram, and western Himalayas experienced higher ELA increases (>90 m) compared to other regions, whereas glaciers in the western Kunlun Mountains, southern, and eastern Tibetan Plateau showed relatively smaller increases (<50 m) among the investigated glaciers. This spatial heterogeneity in ELA rise is particularly pronounced in the Karakoram region, where the large ELA increase is also associated with corresponding decreases in air temperature at the ELA.

The difference map of the annual mean 2m air temperature at the ELA (ΔT_a) (Figure 6b) reveals that 46.25% of glaciers (37 out of 80) experienced colder conditions at the ELA (ΔT_a < -0.1℃) during 2001-2019 compared to 1982-2000. These glaciers, spanning various mountain ranges and glacier types, were primarily distributed in the Himalayas, inner TP, and Qilian Mountains. Among them, 84% were classified as continental glaciers. On average, the ELA of these glaciers rose by 92.4 m relative to the earlier period. In contrast, 12.5% of glaciers (10 out of 80) exhibited warmer conditions at the ELA (ΔT_a > +0.1℃). These glaciers were mainly located in the western Kunlun Mountains and Tien Shan. The mean ELA of these glaciers increased by 58.5 m during 2001-2019 compared to 1982-2000. Meanwhile, more than 40% of glaciers showed relatively stable T_a values, with absolute ΔT_a changes of less than 0.1℃. These glaciers were predominantly found in the southeastern glacierized regions, including Nyainqentanglha, Hengduan Mountains, and the eastern Himalayas, as well as the Pamir-Karakoram-Kunlun Mountains regions. For this group, the mean ELA rose by 70.7 m between the two periods.

The temporal variation of T_s between the two periods reveals that 67.5% of glaciers experienced a decrease in summer mean temperature at the ELA (ΔT_s < -0.1℃), while 12.5% experienced an increase (ΔT_s > +0.1℃), with average changes of -0.40°C and +0.25°C, respectively (Figure 6c). In contrast, 20% of glaciers exhibited relatively stable summer temperatures, with changes ranging from -0.1°C to +0.1°C. These glaciers were sparsely distributed, with a mean ELA rise of 71.7 m. Glaciers in the Himalayas, Karakoram, Pamir, most of the inner TP, and many maritime glaciers showed a decrease in T_s, with a mean ELA rise of 81.6 m compared to the previous period. In contrast, glaciers in the Tien Shan, Qilian Mountains, and eastern Kunlun Mountains experienced an increase in T_s, with an average ELA rise of 78.3 m.

The analysis of precipitation changes (ΔP_a) at the ELA between the periods 2001-2019 and 1982-2000 reveals a notable divergence in the response of glaciers (Figure 6d). About 27.5% of glaciers experienced a decrease in precipitation (ΔP_a < 0 mm), with most of these glaciers being maritime types and some located from the Pamir to the western Himalayas. The pronounced precipitation decrease occurred in the Hengduan Mountains with a mean ΔP_a of -288.7 mm. In contrast, 72.5% of glaciers exhibited an increase in precipitation at the ELA (ΔP_a > 0 mm) compared to the earlier period, and the precipitation increase was less than 100 mm. Despite the differing precipitation trends, the mean ELA rise was similar for both groups: 79.1 m for glaciers with increased precipitation and 79.4 m for those with decreased precipitation.

The temporal variation of the 20 m ice temperature at the ELA between the two periods (ΔT₂₀) ranged from -0.75℃ to 0.49℃ (Figure 6e). The spatial patterns of ΔT₂₀ and ΔT_a​ were largely similar, except for a small subset of glaciers (11 out of 80) (Figure 6b and e). These glaciers exhibited an increasing trend in near-surface ice temperature (ΔT₂₀ > 0) compared to the decreasing trend in annual mean 2 m air temperature (ΔT_a < 0) between the two periods. Such glaciers were sparsely distributed across Chinese glacierized regions, with a relatively higher concentration among maritime glaciers.

Figure 6f shows the velocity differences of individual glaciers between the two periods. Approximately 44.3% of glaciers showed positive changes (Δv > 0), while 55.7% exhibited negative changes (Δv < 0). Using a threshold of ±1.6 m a^-1 to account for uncertainties in satellite-derived surface velocities [8], glaciers were classified as experiencing speedup (Δv > +1.6 m a^-1), slowdown (Δv < +1.6 m a^-1), or stable flow (|Δv |≤1.6 m a^-1). We revealed that 46.8% of glaciers underwent slowdown, 35.5% experienced speedup, and 17.7% maintained a stable flow state during 2001–2019 compared to 1982–2000. Accelerated glaciers were primarily located in the Kunlun Mountains, characterized as extremely-continental, and in southeastern glacierized regions, including the Hengduan Mountains, Nyainqentanglha, and eastern Himalayas, which were predominantly maritime. The decelerated glaciers were broadly distributed across Chinese glacierized regions, whereas the stable glaciers were sparse and lacked a clear geographical concentration.

4.3. Threshold-Based Transitions in Glacier Types

Figure 7 illustrates the spatial distribution of glaciers deviating from classification thresholds during the periods 1982-2000 and 2001-2019. The overall pattern remains relatively consistent across both periods. Glaciers with the highest number of deviations are primarily concentrated in the Karakoram, western Kunlun Mountains, Qilian Mountains, and maritime glacierized regions. In contrast, only two glaciers in the first period and one in the second remained entirely within the classification thresholds, all of which belong to the extremely-continental type. Notably, glaciers with all five classification criteria exceeding the thresholds are clustered in the Karakoram and western Kunlun Mountains, underscoring these regions as key transition zones. Glaciers in the Tien Shan and inner TP show relatively fewer deviations from the thresholds, while those at the boundaries between glacier types exhibit a marked increase in classification deviations.

Among extremely-continental glaciers, those along the Kunlun-Qilian transect and in the Gangdise and Tanggula Mountains near the subcontinental boundary exhibit more than three criteria exceeding the classification threshold. The former group primarily deviates in T_a, T_s, P_a and T₂₀, while the latter mainly in T_a and T_s. Additionally, glaciers in these regions generally have lower surface velocities than the extremely-continental glacier threshold. Among subcontinental glaciers, those in the Karakoram region have all five criteria below the threshold, closely resembling extremely-continental glaciers. In contrast, glaciers in the Tien Shan and Altai Mountains typically deviate in two criteria (T₂₀ and v) are lower, while T_s is higher than the threshold reflecting distinct differences in thermal and dynamic characteristics. Maritime glaciers near the subcontinental boundary in the eastern Himalayas and southern and eastern TP predominantly show (P_a, T₂₀, and v) below the threshold, indicating similarities with subcontinental glaciers. Similarly, glaciers in the Hengduan Mountains tend to have three criteria (T_a, T₂₀, and v), falling below the threshold, suggesting a shift toward more continental characteristics.

5. Discussion

5.1. Drivers of Glacier Criteria Changes

Changes in glacier classification criteria across China from 1982 to 2019 reveal emerging tendencies toward glacier type transitions. Glacier classification is governed by climatic, thermal, and dynamic factors, and their temporal changes should be interpreted in the context of changes in large-scale circulation systems, regional climate regimes, glacier morphology like topography and glacier dynamic processes.

ELA plays a pivotal role in glacier classification, as four of the five criteria in the Shi-Xie scheme are defined at the equilibrium line. Under sustained glacier mass loss, ELA across China's glacierized regions has exhibited a rising trend [6,44], with a marked acceleration after 2000. This rise is primarily driven by enhanced ablation due to rising air temperatures and reduced solid precipitation [45]. However, the rate and spatial pattern of ELA changes vary across glacier types. For continental glaciers, ELA rise is mainly linked to decreased precipitation resulting from weakened westerlies and monsoonal activity [46]. For maritime glaciers, the mass loss and resulted ELA rise after 2000 were mainly caused by the decreased solid precipitation ratio and precipitation amount during the monsoon season [47]. The magnitude of ELA rise is a direct indicator of glacier sensitivity to climate forcing [48,49]. Moderate increases suggest a relatively balanced adjustment to environmental change, whereas sharp rises may indicate an accelerated, and possibly unsustainable, glacier response. Since four criteria in the Shi-Xie scheme are defined at the ELA, any shift in ELA due to climatic forcing directly propagates through the classification system, particularly through temperature-based indicators shaped by vertical lapse rates.

The observed ELA shifts thus highlight the underlying climatic drivers shaping glacier environments. Beyond direct warming effects, glacier climate indicators are influenced by atmospheric circulation patterns, regional topography, and elevation dependent feedbacks. While global warming has elevated air temperatures across glacierized regions, temperature changes at the ELA are further modulated by lapse rates and glacier elevation [50,51]. Due to the elevation dependent warming and the vertical shift of the ELA [52], more than half of China's glaciers exhibit decreasing mean annual and summer temperatures, despite the general warming trend. Precipitation patterns across the Tibetan Plateau reflect a warm-wet shift, shaped by the interplay of monsoonal systems, westerlies, and local circulation [53,54]. In the eastern semi-arid to semi-humid regions, increased precipitation has enhanced glacier accumulation. Conversely, in maritime glacier regions, weakened Indian monsoons and reduced moisture transport have led to precipitation declines and diminished mass input. However, in some areas such as the Kunlun and Qilian Mountains, local circulation adjustments have slightly increased precipitation, promoting a transition toward subcontinental glacier characteristics [55]. Overall, maritime glaciers are characterized by high temperatures and abundant precipitation, with mass balance primarily sustained by snowfall. Extremely-continental glaciers remain cold and arid, with low ablation rates and high thermal stability. Subcontinental glaciers, influenced by both westerlies and monsoons, display complex responses to climatic regime shifts, reflecting their transitional nature.

In this study, the 20 m ice temperature is simulated based on conductive heat transfer and is primarily controlled by surface air temperature at the ELA [56,57]. Assuming constant mean annual air temperature, an upward shift in the ELA effectively leads to lower air temperatures at that altitude due to lapse rate effects, resulting in a corresponding decline in the modeled 20 m ice temperature. This cooling mechanism is particularly evident in regions with rapidly rising ELAs, such as the Himalayas, Hengduan Mountains, and southeastern Tibetan Plateau. In contrast, regions like the Kunlun and Qilian Mountains, where ELA elevation is relatively stable or warming is more pronounced, exhibit increasing ice temperatures. Overall, changes in 20 m ice temperature reflect the combined influence of ELA variability and regional climate trends. Notably, in maritime glacier regions with frequent ELA fluctuations, significant cooling at glacier margins suggests an ongoing thermal transition toward subcontinental glacier characteristics.

Glaciers of different types have generally experienced a reduction in surface velocity, with mean flow speeds falling below their respective dynamic thresholds. This slowdown is primarily caused by glacier thinning, which reduces ice thickness and subsequently lowers gravitational driving stress, weakening overall glacier dynamics [58,59,60]. Since 2000, glaciers in the high mountain regions of Asia have shown a widespread and persistent deceleration. The decline in driving stress, largely controlled by changes in surface slope and ice thickness, is especially evident in areas with pronounced thinning [8]. Maritime glaciers, influenced by high precipitation and relatively warm ice temperatures under monsoonal climates, still maintain relatively high velocities in some regions. However, accelerated surface melting has led to rapid thinning, which in turn causes significant reductions in flow speed. In contrast, continental glaciers located in cold and arid regions with low precipitation and basal temperatures exhibit low deformation capacity. These glaciers continue to slow due to limited sliding and internally rigid thermal conditions.

5.2. Limitations

While this study provides a comprehensive assessment of glacier transformations in China, several limitations should be acknowledged. These include uncertainties in ELA estimation due to limited observational data, potential biases in climate inputs from ERA5-Land reanalysis, and spatial and temporal gaps in glacier velocity datasets. Additionally, the small number of reference glaciers used for calibration and validation may affect the representativeness of our findings.

ELA is a key glaciological variable for deriving classification criteria in the Shi-Xie scheme. However, due to the limited availability of mass balance measurements, observed ELA data exist for only a few glaciers. To address this, we rely on surface mass balance modeling using OGGM to estimate ELA. Although the modeled ELAs were calibrated and validated using geodetic mass balance data, assessing their performance across different glacierized regions in China remains challenging due to regional variations in glacier-climate interactions.

The ERA5-Land reanalysis dataset is essential for calculating the climatological and thermal indicators in the Shi-Xie scheme. Despite its high spatial resolution (9 km), temperature and precipitation estimates remain uncertain in glacierized regions with complex topography and large elevation ranges. Furthermore, the downscaling method introduces additional uncertainties. We applied a constant lapse rate, however, lapse rates vary both spatially and temporally. Although we incorporated precipitation gradient parameters that differ across catchments, they may still fail to fully capture the localized variability of mountain precipitation.

Glacier surface velocity data (ITS_LIVE, 240 m resolution) spans from 1985 to 2018, but its temporal coverage does not fully align with the climate dataset, which may introduce uncertainties in the analysis of glacier velocity-climate relationships. Additionally, ITS_LIVE has spatial data gaps, such as missing velocity data in the Altai Mountains, limiting its applicability in certain regions. Another source of uncertainty arises from the estimation of ice temperature at 20 m depth, which only accounts for conductive heat transfer while neglecting more complex thermal processes, such as latent heat release from meltwater refreezing, which can influence near-surface ice temperature.

This study relies on measurements from only six glaciers, which may limit the spatial representativeness of our assessment, particularly regarding the diversity of glacier types. Future research should integrate a broader range of in situ measurements, satellite observations, and model simulations (such as modeled ELA) to improve calibration and validation. Expanding the dataset will enhance the robustness of our assessment and provide a more comprehensive understanding of glacier transformations in China.

6. Conclusions

This study provides a comprehensive assessment of the spatial and temporal changes in ELA and five glacier classification-related criteria for representative glaciers across China during from 1982 to 2019. The results reveal regionally distinct glacier responses and evolving classification tendencies under climate change.

Overall, ELAs have risen across China, with the most significant increases observed in the Qilian Mountains, Karakoram, and western Himalayas. Despite ongoing warming, the altitude-induced cooling effect has led to declining air temperatures at the ELA for approximately 50% of the studied glaciers and summer temperature declines for 67.5%. Near-surface ice temperatures exhibit strong consistency with the observed changes in annual air temperature. Precipitation trends at the ELA show strong regional heterogeneity. While approximately 70% of glaciers experienced stable or slightly increasing annual precipitation, maritime glaciers, particularly those in southeastern regions, e.g., Hengduan Mountains, experienced significant decreases. 90% of the studied glaciers flow at speeds below 50 m·a⁻¹, indicating a widespread slowdown.

Glaciers near classification boundaries, particularly in the Karakoram, Qilian, and Hengduan Mountains, frequently exhibited multiple criteria deviations from their individual thresholds. Among extremely-continental glaciers, the proportion exceeding classification thresholds rose by ~55% (from 31 to 48 glaciers), reflecting a shift toward subcontinental characteristics. In contrast, deviations in maritime glaciers declined by ~18% (from 49 to 40 glaciers), suggesting improved classification stability, though some retained subcontinental tendencies.

These findings highlight the critical role of regional climate variability in reshaping glacier characteristics and classification stability across China. By quantifying ELA shifts and threshold deviations, this study provides a baseline for predicting future glacier changes, with important implications for water resource management and climate adaptation strategies in vulnerable mountain regions.