Glaciological observation of Kangxiwa Glacier has been conducted by the Muztagh Ata Station for Westerly Environment Observation and Research, Chinese Academy of Sciences (Fig. 1b) since the 2010/2011 hydrologic year. Nine ablation stakes were installed along the central flow line of the Kangxiwa Glacier to measure point-scale surface mass balance following standard glaciological methods in this research (Cogley et al., 2011; Kaser et al., 2003). In-situ measurements of stake exposure height, snow stratigraphy, and snow density were conducted at the onset (early June) and the end (late September) of each ablation season (Fig. 1d). Ice density was assumed to be 900 kgm-3 (Huss, 2013). Seasonal (winter and summer) and annual point mass balance were derived for the 2019/2020 to 2022/2023 balance years. These annual and seasonal mass balances serve as ground truth for validating the accuracy of daily surface mass balance retrieved from time-lapse camera observations in this study.

A Campbell automatic weather station (38.28° N, 75.04° E, 4900 m a.s.l.; AWS4900; Fig. 1b) was deployed on the western slope of Muztagh Ata in 2011 at a similar elevation to the Kangxiwa Glacier terminus. It continuously recorded half-hourly measurements of wind speed, wind direction, air temperature, relative humidity, atmospheric pressure, incoming/outgoing shortwave and longwave radiation, and all-weather precipitation by T200B precipitation gauge (Zhu et al., 2018). This dataset was used to characterise the near-surface meteorological features of the study area over the past four years and identify extreme weather events. This provides a basis for explaining short-term abnormal changes in glacier mass balance. Data from the Taxkorgan Meteorological Station, located approximately 50 km south of Muztagh Ata, were used to analyse long-term climate change trends in the eastern Pamir.

To investigate the possible climate mechanism for abnormal glacier mass loss, this study also employed the fifth generation of reanalysis data from the European Centre for Medium-Range Weather Forecasts (ERA5) with 2.5° horizontal resolution, which has been widely used in climate research (Hoffmann et al., 2019; Song et al., 2024) and glacier change analyses (Zhu et al., 2024b). By analysing the spatial anomalies pattern of geopotential height, wind fields, and surface air temperature, we investigated the change of large-scale atmospheric circulation and discussed its possible influences on extreme weather events and the subsequent responses of glacier surface mass balance.

Figure 3 shows a comparison of the cumulative point mass balance estimates derived from the time-lapse cameras and the glaciological method at the three camera sites (5005, 5137 and 5300 m) on the Kangxiwa Glacier during the period from 2019/2020 to 2022/2023 balance years. Figure 4a further presents the seasonal comparative performance of point-scale mass balance (winter and summer). Quantitative comparisons reveal that the mean seasonal mass balance differences between the two datasets at the three monitoring sites are -64, -13, and 2 mm w.e., with corresponding standard deviations of 193, 137 and 104 mm w.e., respectively. Over the entire observation period, the mean discrepancies between the two datasets yield an overall mean of 26±149 mm w.e.. These discrepancies between the two methods could be largely attributed to temporal variations in snowpack density throughout the observation period.

At the glacier-wide scale (Fig. 4b), the camera-based seasonal mass balances exhibit a robust linear correlation with the stake-based counterparts (e.g., R2=0.98), with almost all in-situ stake-based values falling well within the uncertainty bounds of the camera-based estimates. The mean discrepancy between the two datasets is 62±54 mm w.e., indicating high consistency between the camera-based method and the glaciological stake method. This robust agreement not only validates the reliability of time-lapse camera observations for quantifying glacier mass balance but also establishes a solid foundation for investigating temporal evolution characteristics of glacier-wide mass changes for the Kangxiwa Glacier.

Figure 5a compares the annual cumulative changes in the glacier-wide surface mass balance of the Kangxiwa Glacier during the 2019/2020–2022/2023 hydrological years, which can be broadly categorised into three phases: a balance period from October to mid-April, a snow accumulation period from mid-April to June/July, and an intense ablation period from June/July to September. During the balance period, low temperatures and low precipitation resulted in limited changes in the stake melt-out length, with the cumulative mass balance fluctuating between -78 and +188 mm w.e. The slight surface mass loss observed during this period was likely caused by sublimation or mechanical snow drift driven by strong winter wind, as evidenced by the transition of snow-covered surfaces to exposed bare ice under cold winter conditions (Fig. S3). During the accumulation period (mid-April to June/July), the glacier experienced different maximum snow accumulation across the entire glacier ranging from 258 to 445 mm w.e. At an elevation of 5300 m a.s.l., maximum snow depth reached 1.0–1.5 m (Fig. S3). Notably, monthly accumulation peaks generally occurred in May (Fig. S4).

The phase of glacier mass loss is predominantly confined to the period from June/July to September, with significant interannual variability manifested in multiple key aspects. First, the onset of surface mass loss – after the annual maximum net accumulation is reached – exhibits substantial interannual differences: the earliest initiation was observed on around 2 June 2023, whereas the latest occurred around 22 July 2020. The Glacier Loss Day – defined as the date when the net mass balance transitions to negative and all winter snow accumulation is depleted (Voordendag et al., 2023) – occurred on around 11 July 2022, approximately one month earlier than in other years (e.g., ∼ 3 September 2021; ∼ 26 August 2023). Notably, no Glacier Loss Day was recorded in the 2019/2020 hydrologic year, a phenomenon attributed to abundant winter-spring snow accumulation coupled with relatively low summer air temperature (Fig. 5b, c). The earliest Glacier Loss Day in the 2021/2022 hydrologic year was partly caused by the lowest precipitation in the accumulation season. Specifically, the AWS4900 shows that the cumulative precipitation for the 2021/2022 hydrologic year was only 131.8 mm until 11 July, while the corresponding values were 195.9, 165.1, and 218.7 mm for the 2019/2020, 2020/2021, and 2022/2023 hydrologic years, respectively.

Beyond the timing of ablation onset, the duration of glacier mass loss also exhibits marked interannual variability. Specifically, it was approximately 40 d in 2020 (from ∼ 23 July to 30 August), 80 d in 2021 (11 July to 30 September), 58 d in 2022 (26 June to 22 August), and 120 d in 2023 (2 June to 30 September). Another interannual variability lies in the magnitude of annual surface mass loss, with values of -414, -334, -907 and -694 mm w.e. in the 2019/2020, 2020/2021, 2021/2022 and 2022/2023 hydrologic years, respectively. The most significant mass loss occurred in July–August 2022 (Fig. 5), with the mass loss in July 2022 being four times greater than the mean for the three other years. AWS4900 records indicate that interannual fluctuations in both air temperature and precipitation were the primary drivers of significant differences in both the duration and intensity of surface mass loss during the 2019/2020–2022/2023 hydrological years (Fig. 5). The exceptional mass loss in 2022, for instance, coincided with an anomalous summer of pronounced warming and aridity. This is evidenced by the fact that cumulative positive degree days (CPDDs) at AWS4900 reached 341.7 °C d, which is 1.4–2.3 times higher than in the other three years, while total precipitation during July–August amounted to only 67.2 mm, approximately 50 % lower than the average for the corresponding period across the other three years.

The annual mass balance of the Kangxiwa Glacier exhibited significant variability throughout the study period. A slight positive balance was recorded in the 2019/2020 hydrological year (+45 mm w.e.), followed by a weak negative balance in 2020/2021 (-121 mm w.e.). This contrasts sharply with the substantial mass loss observed in the following two years: -654 mm w.e. in 2021/2022 and -337 mm w.e. in 2022/2023. Four-year observations revealed that the interannual accumulation differences of the Kangxiwa Glacier were limited to be ∼187 mm w.e., with the maximum accumulation of 445 mm w.e. in the 2019/2020 hydrological year and the minimum of 258 mm w.e. in the 2020/2021 hydrological year (Fig. 5). A comparative analysis between the 2019/2020 and 2021/2022 hydrologic years reveals that the 2021/2022 experienced less accumulation of 167 mm w.e. in winter season, but more ablation of 494 mm w.e. in summer season. These results emphasize that interannual variability in the glacier mass balance of Kangxiwa Glacier was predominantly driven by the variability in mass loss during the ablation season.

Meanwhile, linear altitudinal interpolation of two complementary datasets – camera-based point mass balances from three elevations and stake – based measurement from 9 ablation stakes – allowed for the quantification of the equilibrium line altitude (ELA) over the study period. The ELAs displayed significant interannual fluctuations: the highest ELA surpassed the glacier summit during 2021/2022 balance year, whereas the lowest ELA (5079–5086 m a.s.l.) occurred in 2019/2020.

Both ERA5 reanalysis data and ground-based meteorological station observations revealed an unprecedented summer warming event in 2022 in the eastern Pamir, with the most pronounced anomaly recorded in July (Figs. 6 and S5). The ERA5 temperature data confirmed that the corresponding grid point of the Kangxiwa Glacier in July 2022 was the highest recorded between 1981 and 2024 (Fig. 6a). The AWS4900 records showed that the average July temperature during 2012–2023 was 2.9 °C, while it reached 6.2 °C in July 2022, constituting a substantial positive temperature anomaly (Fig. 6b). Additionally, daily maximum temperature recorded at Taxkorgan station during summer 2022 was significantly higher than the long-term mean for 1957–2023 (Fig. S5), with 61 % of days in July exceeding the 90th percentile of historical temperatures and satisfying the criteria for an extreme heat event as defined by Lu et al. (2024). This evidence indicates that July 2022 was characterized by significantly elevated temperatures relative to historical baselines. This exceptional climatic event provides a unique opportunity to analyse how glaciers respond to extreme heatwaves in the eastern Pamir.

Based on AWS4900 records, three extreme heat events were identified between 26 June and 11 July, 18–30 July, and 5–17 August 2022 (Fig. 7). During the first two heatwaves, daily temperatures were 6.2 °C and 4.4 °C higher than the three-year average, with a maximum anomaly of +10 °C on 5 July (Fig. 7a). Concurrently, daily incoming solar radiation exceeded the average by 90.6 Wm-2 (+31 %) and 48.7 Wm-2 (+17 %) (Fig.7b), while total precipitation decreased by 40.4 mm (-100 %) and 31.3 mm (-76 %). These conditions, characterized by increased temperatures, reduced snow replenishment, and intensified solar radiation, drove anomalous mass loss (Figs. 5, 7). The first extreme event induced a mean mass loss rate of -16 mm w.e. d⁻¹, whereas the three-year average for the same period exhibited a near-balanced state. The second heatwave exacerbated mass loss to -19 mm w.e. d⁻¹, which is ∼4 times the three-year mean (-1 to -7 mm w.e. d⁻¹). This significant ablation accelerated the depletion of spring-accumulated snow (Fig. 8) and thus contributed to the advance of the Glacier Loss Day by one month (Fig. 5).

The weather conditions during the August heatwave and the corresponding glacier response were obviously different from those during the two July heatwaves. The air temperature during the August heatwave was only 2.5 °C higher than the corresponding average of the other three years. The reduced shortwave radiation of 54.9 Wm-2 (-17.1 %) and the increased longwave radiation of 26.1 Wm-2 (+10.2 %) indicate the increased cloud cover over the eastern Pamir during this period (Liu et al., 2021). There was no significant change in precipitation. The August heatwave was therefore characterized by moderate warming and cloudy conditions. In contrast, the magnitude of mass loss during the August heatwave was even more intensive than those during the two July heatwaves (Fig. 7d). Mass loss intensity increased to 23 mm w.e. d⁻¹ during the August heatwave, which is 16 mm w.e. d⁻¹ higher than the mean value for the corresponding periods of the other three years.

Glacier surface melting is critically linked to the energy supply and the surface conditions. The precipitation phase may greatly influence the snow accumulation and the surface albedo condition (Jouberton et al., 2022). Seasonal snowfall and rainfall amount at each camera-monitored site for each hydrological year were estimated using the method proposed by Ding et al. (2014), combined with meteorological data from AWS4900 and the Muztagh Ata observation station (3650 m a.s.l.). Snowfall accounted for 94 % of total precipitation at the glacier terminus (5050 m a.s.l.), 96 % at the mid–glacier site (5137 m a.s.l.), and 99 % at the upper–glacier site (5300 m a.s.l.) in the warmest 2022. The scarcity of rainfall in summer season across the glacierized area indicates that precipitation phase transitions likely played a limited role in modulating the surface energy–mass balance. At the 5005 m site, the number of bare ice days from June to September reached 42 in 2021/2022, compared to 4, 23, and 13 d for the 2019/2020, 2020/2021, and 2022/2023 hydrologic years respectively. This marked increase in bare ice days enhanced glacier surface energy absorption, which partly explained the substantial mass loss in the summer of 2022. Given the moderate temperature anomaly and reduced shortwave radiation during the August heatwave, the intensive mass loss could be primarily attributed to the exceptionally low surface albedo of exposed ice (Fig. 8), which amplified solar radiation absorption and subsequent melt processes (Mölg et al., 2014).

Glaciers in the Muztagh Ata have been in a near-equilibrium state since at least the 1970s (Kääb et al., 2015; Bolch et al., 2019; Brun et al., 2017; Holzer et al., 2015). Based on the high-resolution Pléiades stereo images, Falaschi et al. (2023) analyses the annual and seasonal glacier mass balance in Muztagh Ata and address that the mean mass loss of 2020, 2021 and 2022 was -0.19±0.14, +0.15±0.27, -0.17±0.22 m w.e. Our measurements at the Kangxiwa Glacier in the Muztagh Ata region for the 2019/2020 and 2020/2021 hydrological years are consistent with the previous knowledge of near-equilibrium in this region (Yao et al., 2012; Falaschi et al., 2023; Bhattacharya et al., 2021; Hugonnet et al., 2021). However, the mass balance in 2021/2022 and 2022/2023 hydrological years displayed the significant mass loss. In particular, the mass loss in 2022 was the most negative and agrees well with the reported unprecedented mass loss of the Urumqi No. 1 Glacier in the eastern Tien Shan (Xu et al., 2024) and the notable mass loss in the European Alps (Berthier et al., 2024; Cremona et al., 2023; Voordendag et al., 2023). Similar to the Swiss glaciers, the equilibrium line altitude of the Kangxiwa Glacier rose above the glacier summit in 2022 due to heatwaves. Extremely high air temperatures covered Eurasia and North America, with long-lasting extreme heat events affecting China (Lu et al., 2023). During the summer of 2022, three heatwaves within a 40 d period caused a mass loss equivalent to over 90 % of the total ablation losses for the season, demonstrating the pivotal role of heatwaves in driving local glacier ablation in this region. Similar regional heatwaves around the world (Colucci et al., 2017; Perkins-Kirkpatrick and Lewis 2020; Zhang et al., 2025) pose substantial threat to mountain glaciers. Such significant mass loss by short-time heatwaves could influence the long-term changes of glacier ablation.

We further compared the annual mass balance of the Kangxiwa Glacier with those of nearby glaciers from 2019/2020 to 2022/2023 (WGMS, 2025), including Zulmart Glacier (Glacier No. 139) and Glacier No. 457 in the Pamir, as well as Abramov Glacier in the Pamir Alay. The annual mass balance of Zulmart Glacier, which is about 200 km northwest of Kangxiwa Glacier, was -210, -206, -311, and -271 mm w.e. for the 2019/2020, 2020/2021, 2021/2022, and 2022/2023 hydrologic years, respectively. Similar to the Kangxiwa Glacier, the Zulmart Glacier also exhibited the most negative mass balance in 2022, which is likely attributed to the same impact of extreme high temperatures in July. In contrast, neither the Abramov Glacier nor the No. 457 Glacier showed similar patterns. These contrasting mass loss patterns are likely modulated by divergent climatic regimes and topographic conditions, which influence mass change patterns and climatic sensitivities across the Pamir (Brun et al., 2019; Wang et al., 2019).

Based on our daily mass balance dataset, it is evident that the accumulation in April–May is the primary driver of annual accumulation magnitude in eastern Pamir. Barandun et al. (2018) found that glaciers in the west Pamir region experience snow accumulation at the start of the hydrologic year, receiving over 1 m w.e. of accumulation during the winter season. These different accumulation patterns are linked to divergent precipitation seasonality between the east and west Pamirs, driving the formation of distinct mass accumulation types (Maussion et al., 2014). Published geodetic estimates reveal striking west–east disparities in glacier mass balance since 2000, with a mean mass loss of -0.26 m w.e. a⁻¹ in the western Pamir but a near-balanced state (-0.02 m w.e. a⁻¹) in the eastern Pamir (Bolch et al., 2019).

Furthermore, variations in the timing of heatwave occurrence and their combination with seasonal distribution of precipitation can also contribute to substantial disparities in the response of glacier surface mass balance to climatic conditions. As illustrated in Fig. 5, the mass loss in 2019/2020 was concentrated in August (-346 mm w.e.), following a period of high accumulation (+445 mm w.e.). In contrast, the 2022 mass balance process featured low spring accumulation and extremely strong summer ablation driven by heatwaves in July and August. While the ablation periods in 2021 and 2023 were significantly longer than in other years, the average summer ablation intensity was moderate. Unlike the stable mass accumulation observed in June across the other three years, the Kangxiwa Glacier experienced early ablation (-107 mm w.e.) in June 2023, followed by the highest ablation in September over the four-year period. Cumulative mass loss from June onward reached 628 mm w.e. in 2023, which was slightly less than the maximum loss in 2022. These divergent patterns of mass balance evolution not only underscore the complexity of glacier responses to climate change in this region but also highlight the critical importance of continuous, high-temporal-resolution monitoring of glacier surface mass changes to inform future model-based explanations (Barandun and Pohl, 2023).

The extraordinary mass loss of the Kangxiwa Glacier in the eastern Pamir and Urumqi Glacier No. 1 in the eastern Tien Shan in 2022 (Xu et al., 2024) suggests that heatwaves may affect glacier conditions in the region on a wider scale. Analysis using ERA5 reanalysis data indicates that the July heatwave event was mainly located on the Pamir Plateau (Fig. 9a). The overall impact of the 2022 heatwave was felt in the western Kunlun Mountains and the Pamir Plateau where glaciers were previously considered in equilibrium or having a positive glacier mass balance (Brun et al., 2017; Kääb et al., 2015; Hugonnet et al., 2021).

We analysed anomalies in temperature, solar radiation, cloud cover, geopotential height and wind fields at 200 hPa during July 2022, compared to the climatological average from 1991 to 2020, to investigate the anomalous large-scale atmospheric circulations linked to the 2022 July mass balance anomaly in this study area. Anomalously high temperatures of over 4 °C were observed across the Pamir Plateau and the western Kunlun Mountains. This suggests that other glaciers in the region may also have experienced significant mass loss during the summer of 2022. Latest research in the Northwestern Pamirs found that enhanced glacier mass loss was linked to the 2022 July heatwave (Jouberton et al., 2025). Numerous studies have linked heatwaves and high-temperature events to anticyclonic systems (Deng et al., 2023; Jiang et al., 2023; Song et al., 2024). Figure 9b shows that an upper-level anticyclonic anomaly developed over the northwestern flank of the Tibetan Plateau at 200 hPa. Under these circulation patterns, strong easterly wind anomalies occur on the southern side of the anticyclonic anomaly, reducing the westerly wind at 30–40° N. A decrease in westerly winds means less atmospheric water is delivered to the Pamir, resulting in reduced precipitation and cloud cover. Abnormal anticyclonic circulation can suppress convection and cause strong subsidence, resulting in reduced cloud cover (Fig. 9d). The decrease in cloud cover, in turn, leads to increased incoming shortwave radiation. Figure 9c indicates anomalously high shortwave radiation during the July 2022 heatwave events. For example, AWS4900 records showed that shortwave incoming radiation exceeded the average for 2020–2021–2023 by 90.6 Wm-2 (+31 %) and 48.7 Wm-2 (+17 %) during the two heatwave events. This enhanced solar radiation further amplified the heatwave. This enhanced solar radiation further intensified the heatwaves. Overall, anticyclonic circulation anomalies contributed to air descent and anomalous diabatic heating, resulting in sunny and dry weather in eastern Pamir in July 2022 (Fig. 5). This favoured an increase in shortwave radiation to heat the land, which can increase outgoing longwave radiation and turbulent heat fluxes from the land surface to the atmosphere, causing higher air temperatures.

Additionally, soil moisture anomalies and teleconnections via Rossby wave trains (Hood et al., 2020) are also potential contributing factors. Low soil moisture anomalies on the Tibetan Plateau may have intensified the development of heatwaves through land–atmosphere feedback mechanisms (Jiang et al., 2023). The anomalous anticyclone in eastern Pamir appears to be linked to the propagation of wave energy from an upstream mid-latitude wave train originating in the northwestern North Atlantic (Deng et al., 2023; Song et al., 2024). Anticyclone development above western Europe (Fig. 9b) may have acted as a conduit for Rossby wave propagation to the anticyclone developed under the Northwestern flank of the Tibetan Plateau, linking glacier mass balance in this region to large-scale climate mode such as the Atlantic Multidecadal Oscillation.