The study area is located in the section from Ranwu to Thongmai in the Palong Tsangpo Basin, mainly in Bomi, Nyingchi (Figure 1). The study area is 16,021.12 km², the highest elevation is 6,650 m a.s.l, the lowest elevation is 2000 m a.s.l, the maximum altitude difference is approximately 4,650 m a.s.l, and a large number of oceanic glacier gullies have developed upstream. Glacial, alpine canyon, and river accumulation landforms are extremely well developed. This area, located on the northern Qinghai-Tibet Plateau, is at the junction of the northern Tibetan and the Himalayan massif and is a typical deep alpine canyon. In addition, the Jiali Fault and the Zamu-Maniwen Fault pass through the study area (He et al., 2005). The geological structure around the study area is extremely complex, and there is frequent seismic activity in the area due to the activity belts of the Namtso-Zhongsha, Jiali-Ranwu and Metuo structures. According to previous researches, there have been as many as 9 strong earthquakes with surface wave magnitude of 6.0 (Ms 6.0) or above in the region in the past 60 years, and approximately 50 earthquakes of Ms 4.7 to Ms 5.9 magnitudes. One of the most representative seismic events was the Ms 8.6 earthquake that occurred in Chayu County on 15 August 1950.

Climatic conditions are the main triggers for geological disasters in the region, and the study area has the monsoon climate of the temperate semi-humid plateau in southeastern Tibet, which is affected by the warm and humid air flow of the Indian Ocean and the southwestern monsoon. The annual average temperature in the study area is 8.5°C, and the lowest temperature is in January, when the average temperature is approximately −0.2°C, while the highest temperature is in July, when the average temperature is approximately 16.4°C. Overall, the highest temperature of the year is 31 °C and the lowest temperature is −20.3°C. The average annual rainfall is approximately 977.1 mm, and the rainfall from March to October accounts for approximately 93.5% of the total annual precipitation.

To monitor the dynamic changes in glaciers and glacial lakes in the watershed of the study area over time, this study collected QuickBird, IKONOS, SPOT-5 and Landsat images from 1988 to 2019. Table 1 shows that the QuickBird satellite image resolution is 0.61 m, the SPOT5 image resolution is 2.5 m, and the IKONOS image resolution is 1 m. The multispectral ETM+ and TM image resolutions are 30 m. In addition, ASTER GDEM data measured by the EARTH observation satellite Terra of NASA were collected with a resolution of 30 m.

Subsequently, the collected remote sensing data were processed by band combination, geometric correction, image fusion, image mosaicking and cropping, based in ERDAS 9.2, ENVI 5.0 and ARCGIS 10. Then, compared to the raw data, the geometric accuracy, density, texture and tone information of the processed images were significantly enhanced, improving the interpretation accuracy and quality of the remote sensing images. Finally, the evolution of glacier, glacial lake and their impact on glacial debris flow activity were analysed according to the technical flowchart, as shown in Figure 2.

Regional reflection characteristics vary due to different textures, compositions, and shading of materials on the surface; thus, materials are shown in varying shades and shapes in remote-sensing images. Based on these features, a supervised classification method has been adopted to automatically extract the materials. In this region, glaciers in bright white are mainly distributed above the alpine snow line, such as in the ice bucket or inside the ice tongue (Figure 3A), and perennial snow and ice, shown in white and covered by shadows, are mainly distributed below the altitude of the formation glacier, on shady mountain slopes, or in gently sloping areas with low amounts of sunshine; in addition, glacial lakes in dark blue are generally distributed in glaciers and in the lower parts of the perennial snow and ice, and the terrain is relatively gentle (Figure 3B). To fully understand the development characteristics of glacial snow and glacial lakes in the study area, the samples were selected based on the collected TM and ETM+ remote sensing images and the visual interpretation of glaciers, glacial lakes and other geographical features. Then supervised classification, a process of assigning a class value to a set of grid cells in a raster GIS database based on the values of similar grid cells that have been a priori classified as representative of a class or group (Abburu and Golla, 2015; Merry et al., 2023), was performed on the ENVI 5.0 platform to automatically identify and extract the glaciers, glacial lakes and snow of three remote sensing images from 1988, 2002 and 2010.

The limitation of using supervised classification for automatic identification and extraction is that it incorrectly combines materials with similar glacier features into the same category as mountain shadows. Therefore, in this paper, two kinds of remote sensing information of glacial lakes that were extracted by both unsupervised and supervised classification methods were compared and verified in the field, and then sifts and clusters of the remote sensing information were determined by the supervised classification method to remove and synthesise the small classification fragments to obtain reliable and accurate results on the materials and to further reveal the dynamic changes in glaciers and glacial lakes.

In addition, a large number of landslides have developed in the debris flow basin of the study area, and abundant moraine material with melting glaciers has also accumulated in the channel, providing sufficient material for debris flows. Based on field investigation, this study first constructed identification markers of debris materials, moraines, debris flow accumulation fans and other landslides and then combined them with high-definition remote sensing images using the ArcGIS technology platform. A manual, interactive interpretation of moraines was performed on the landslide sources and accumulation fans in the debris flow basin, and the source information of debris flow in the basin was constructed.

To analyse the effects of dynamic changes such as temperature, rainfall and ground temperature on glaciers and glacial lakes in the study area, the average monthly rainfall and temperature data of the study area from 1988 to 2019 were collected. The data show that in the study region, temperatures higher than 0 °C from May to October were positively correlated with the melting of glacial snow. Therefore, in this paper, the average temperature from May to October each year was considered as the influencing factor of or the reaction temperature to glacial snow melt. Meanwhile, the average monthly rainfall in each year from 1988 to 2019 in the study area was also considered as the influencing factor for interannual glacier and glacial lake dynamics.

Geothermal remote sensing inversion based on the radiation transmission equation is a common method to obtain regional geothermal data. Specifically, by this method, the thermal radiation received by the sensor is inverted into the real ground temperature after correction on the atmospheric radiation, observation angle and amount of radiation on the specific surface (Dash et al., 2002). The split window algorithm is the most common technique for retrieving land surface temperature, in which the Qin algorithm (Qin and Karnieli, 1999) can effectively retrieve the surface temperature corresponding to any cell on the image. The equations are as follows: T s = A 0 + A 1 T 4 − A 2 T 5 (1) A 0 = 66.54067 D 4 1 − C 5 − D 5 − 62.23928 D 5 1 − C 4 − D 4 / D 5 C 4 − D 4 C 5 (2) A 1 = 1 + 0.43059 D 5 1 − C 4 − D 4 / D 5 C 4 − D 4 C 5 (3) A 2 = 0.46585 D 4 1 − C 5 − D 5 + D 4 / D 5 C 4 − D 4 C 5 (4) C i = ε i . T i θ (5) C i = ε i . T i θ (6) where T _s is the surface temperature; T ₄ and T ₅ are the brightness temperatures of thermal infrared channels 4 and 5 of the cells, with K as the absolute temperature; ε _i is the specific emissivity in channel i; and T _i(θ) is the atmospheric transmittance of electromagnetic waves in channel i under the zenith angle θ of the sensor.

In glacial areas, temperature is key to the growth and ablation of glacier formation, that is, temperature is a sensitive factor to dynamic changes in glaciers. In this study, the rainfall and temperature monitoring data were collected from a weather station. By analysing the monthly average temperature data over the past 40 years in the Palong Zangbo Valley, it was found that the temperatures from May to October were higher than zero degrees, which was conducive to glacier melting and the water level elevation of glacial lakes. Thus, the average temperature of the above period was analysed to reveal the dynamic changes. In Figure 4, the average temperature variation in the study area in the period ranged from 5.5°C to 7.5°C over the past 40 years. Overall, the temperature in the study area presented a growing increase in the first 4–5 years and then continuously decreased in the following 4–5 years, that is, the pulsed and dynamic temperature change presented in 4–5 years as a cycle. However, the temperature significantly increased from 2012 to 2017 and then tended to rise slowly. This dynamic change in temperature leads to some dynamic differences in the glacier ice and snow cover area and the water level of the glacial lakes in the study area, and it also has an impact on the ground temperature.

In Figure 5A, the geothermal variation ranged from −9°C to 46 °C in the study area in 2002, which was mainly concentrated in the central part of the study area, and the geothermal temperatures in the northwest and southeast regions were relatively high. As the average temperature from May to October in the study area increased from 5.8°C in 2002°C to 6.7°C in 2010, the maximum geothermal temperature in the study area also increased to 53 °C (Figure 5B). As the overall ground temperature increased, the range of low temperature areas decreased. In 2013, the geothermal variation range of the study area was from −12°C to 43°C, and the ground temperature in the northwest region of the study area significantly decreased, while the range of the low temperature in the central zone expanded (Figure 5C). Until 2017, the minimum and maximum ground temperatures were −10°C and 43°C, respectively (Figure 5D). Compared with the previous period, the coverage of the high temperature area increased significantly in 2017, and the low temperature zones were distributed sporadically in the study region. The dynamic differences in time and space between air temperature and ground temperature further led to the dynamic variability of glaciers and glacial lakes in the study area, which was conducive to the development of glacial debris flows.

In addition, the average annual rainfall in the study area from 1980 to 2020 was mostly concentrated between 300 mm and 350 mm and presented pulsed and dynamic changes (Figure 4), but it had slight dynamic differences. However, before 2013, the rainfall and temperature in the study area showed the opposite dynamic trend, that is, the rainfall decreased gradually with increasing temperatures over the years (Figure 4). However, in 2013, the rainfall and temperature in the study area had the same dynamic change trend; that is, the temperature increased with the growing rainfall that year. With rising air and ground temperatures, the glacial meltwater in the study area increased, while increased rainfall led to further increases in runoff in the basin. Therefore, the risk of glacial debris flows and debris flows due to glacial lake collapse in the study area was further increased, and the landslide activity was enhanced.

With climate change glacier snow has melted at an accelerated rate, and glacial debris flows in the Palong Zangbo Valley have become more active. The debris flow visual interpretation of remote sensing images is mainly based on the extraction of the watershed, distribution of loose solid material, erosion characteristics of the channel and the morphological characteristics of the accumulation fan. Remote sensing images combined with field surveys have detected 122 debris flow basins along the Ranwu to Tongmai section of the Palong Zangbo River, including 58 on the right bank and 64 on the left bank (Figure 11). Impacted by the climate change, debris flows frequently break out in these regions, causing serious damage to infrastructure. For example, a catastrophic debris flow in the Midui basin was trigged by the glacier collapse surge in 1988. Field investigation found that the maximum discharge of debris flow is about 1,000 m³/s. A total of 18 bridges and G318 road were destroyed by this debris flow. In addition, the debris flow widely occurred in Guxiang basin after 1955, and the G318 road frequently destroyed by the Guxiang debris flow.

Additionally, Figure 12 shows that the area of debris flow in the study area varied from 1.24 km² to 319.13 km², among which debris flows with a drainage area of less than 50 km² accounted for 84% of the total and were mainly concentrated in the western part of the study area. With the increase on the watershed area, the channel length also increased (Figure 13). Figure 13 shows that the range of the channel length of the debris flow in the study area was between 1,500 m and 35,000 m. Only 25% of the watersheds had a channel length longer than 7,000 m, and watersheds with longer channel lengths were mainly developed in the southeastern part of the study area. In addition, the area is located in southeastern Tibet, where geological tectonic activity is relatively active. In addition to triggering a large number of coseismic landslides, earthquakes have also formed a large number of damaged mountains. Under the action of rainfall, many shattered mountains gradually became unstable and formed post-earthquake landslides. Therefore, according to the interpretation results of the remote images, landslides are well developed in the study area, but there are significant differences in the development of landslides for watersheds. Figure 14 shows that the landslide coverage area was generally less than 1 km² in a watershed area under 25 km², but the landslide coverage area increased linearly with the outbreak of large landslides in the basin and the increasing watershed area, which may be because with the increase in the watershed area and the melting of glaciers and snow in the middle and upper reaches of the watershed, the free face is more developed, and at the same time, with the intensification of freeze‒thaw cycles and glacier erosion, some slopes become unstable and form landslides. Of course, as the glacier gradually retreats, the released moraine also gradually increases. Figure 15 shows that approximately 70% of the watershed’s moraine coverage area was under 2 km², but the moraine coverage increased linearly with increasing watershed area. Figure 11 shows that most of the moraine accumulated in the middle and upper reaches of each basin and the moraine was well developed in the debris flow basin in the southeastern part of the study area.

In addition, the interpretation results of highly accurate remote sensing images show that the development areas of moraine and landslides in the debris flow basin were 117.79 km² and 84.63 km², respectively, accounting for 3% and 2% of the debris flow basin area, respectively. This suggests that although the moraine covers a small area, field investigations have found that the moraine is thicker and more developed than the landslide and is the main component of the glacial debris flow provenance. Meanwhile, Figure 11 reveals that the moraine on the left bank of the Palong Zangbo watershed is well developed, indicating that the glacial activity on the left bank is stronger than that on the right bank. In addition, the rock mass in the study area is fragmented under the action of geological structures such as the Jiali fault. More importantly, the joint plane in the rock mass is continuously infiltrated and freeze‒thawed by the combined influence of glacier meltwater and rainfall, resulting in the ice splitting effect and increasing the joint. With the cycle of freezing and thawing, the surface rock mass is completely disintegrated and destroyed, forming fragments and sand grains. These freeze‒thawed and weathered detrital materials roll down the slope to the channel to form accumulation cones or ice-water accumulation fans at the foot of the slope after being washed away and transported by the ice and snow meltwater, which provides an abundant source of loose debris for debris flows and constitutes the source of debris flow activities.

It is concluded that with increasing temperature, the melting of glaciers and snow has accelerated, which has led to the increasing moraine in the debris flow basin in the study area. Then, due to the effects of earthquakes and freeze‒thaw cycles, the amount of landslide material and weathered debris in the study area increased, that is, the more abundant the debris flow source was in the study area, the higher the possibility of debris flow.