Monitoring the Surface Elevation Changes of a Monsoon Temperate Glacier with Repeated UAV Surveys, Mainri Mountains, China

Abstract

Due to the deep valleys, steep mountains and the influence of the Indian monsoon on the Mainri Mountains (Yunnan Province, China), it is difficult to estimate glacier change from microwave and optical remote sensing. To bridge the gap between low-quality space-borne remote sensing and scarce in situ measurements, airborne remote sensing, such as unmanned aerial vehicles (UAVs), may provide a remarkable opportunity to monitor glacier change with high-quality tools. To determine monsoon temperate glacier change, three UAV surveys were conducted on the Melang Glacier in the Mainri Mountains in November 2019, April 2020 and November 2020. Then, glacier surface elevation changes were estimated from UAV orthophotos and DSMs. High accumulation and high ablation (+10.5 m and −13.5 m) were observed in the accumulation period and ablation period, with a mean surface elevation change of −3.0 m in the surveyed glacier area from November 2019 to November 2020. The avalanche, debris cover, ice cliffs and proglacial lake resulted in a heterogeneous pattern of glacier surface elevation changes. Given that the glacier is more sensitive to temperature, the Melang Glacier may have experienced a substantial recession and mass loss in the past few decades. This study provides a more appropriate approach for monitoring the changes in a temperate glacier in the Mainri Mountains.

1. Introduction

As an important part of the cryosphere system, glacier change is an important indicator of climate change [1] that affects the regional water resources [2]. Over the period 1901–2012, the global combined land and ocean temperature data show an increase of about 0.89 °C (0.69–1.08 °C) [3]. Due to global warming, many mountain glaciers have experienced a significant mass loss [4,5,6,7,8], which subsequently increased the occurrence of glacial hazards [9,10,11]. To explore glacial hazard dynamics, many scientists performed series of in situ observations and geodetic investigations [8,12,13,14] to better understand disaster prevention, the water cycle and climate change aspects that negatively influence glacier water resources.

While the in situ observation monitors the glacier changes in the accessible portions of glaciers with high accuracy [14], the geodetic method examines glacier change over large spatial extents in long-term periods. Other options to estimate the glacier mass balance include optical photogrammetry, laser altimetry, bistatic synthetic aperture radar (SAR) interferometry and gravimetry, such as the SPOT5, ICESat-GLAS, TerraSAR-X/TanDEM-X or GRACE [15,16,17,18]. Based on geodetic methods, a heterogeneous pattern of glacier mass balance across High Mountain Asia (HMA) has been identified, i.e., a balanced glacier mass budget across the Pamir–Karakoram region, moderate glacier mass loss across the Tien Shan, Inner Tibetan Plateau and Himalaya and pronounced glacier mass loss across southeastern Tibetan Plateau [7,8,19].

While the altimetry and gravimetry methods are appropriate to estimate the regional glacier mass balance over large spatial extents, these methods are limited by a coarse spatial resolution, which is not suitable for evaluating the mass balance of small hydrological basins and individual glaciers [20]. A similar challenge is also presented by optical photogrammetry and bistatic SAR interferometry (InSAR) in analyzing individual glacier mass balance. Although optical photogrammetry is often limited by seasonal snow, cloud and image saturation at high elevations [21], bistatic InSAR could induce the foreshortening, layover and shadow regions due to the complex terrain and side-looking geometry of SAR [22,23]. The Mainri Mountains are located in the southeastern Tibetan Plateau and possess a large number of monsoon temperate glaciers. However, thanks to the topography of high mountains and deep valleys, as well as the effect of the Indian monsoon, a reliable glacier mass balance in the Mainri Mountains cannot be merely captured by optical and microwave space-borne remote sensing. Therefore, more sophisticated methods should be developed to monitor the glacier change under the influence of climate and topography.

At present, unmanned aerial vehicles (UAVs), which have the potential to bridge the gap between low-quality space-borne remote sensing and scarce in situ measurements, are being increasingly used for glaciological monitoring [24,25,26,27,28,29]. High-quality UAV imageries and digital surface models (DSMs) generated from overlapping multiview photography could be used for the monitoring of glacier surface features and surface elevation changes. It has been proved that UAVs are appropriate to assess glacier change for monsoon temperate glaciers in the southeastern Tibetan Plateau [30,31]. Hence, when a reliable glacier change cannot be monitored by optical and microwave space-borne remote sensing, UAVs may be a more appropriate approach for the Mainri Mountains.

In this study, three UAV surveys were conducted for the Melang Glacier in the Mainri Mountains in November 2019, April 2020 and November 2020. From UAV imageries and DSMs, the glacier surface features and seasonal surface elevation changes (accumulation and ablation periods) were determined. In Section 2, the study site is introduced. Section 3 describes the aerial surveys and data processing. Then, Section 4 introduces the results, including the uncertainties and characteristics of and changes in the glacier surface. Section 5 discusses glacier changes and the affected features, the relationship between glaciers and climate change and the strengths and weaknesses of the UAV survey.

2. Study Site

Located in Hengduan Shan, the Mainri Mountains lie from the northwest to southeast, east of Nu River and west of Lancang River (Figure 1). The typical characteristics of the region are high mountains and deep valleys. The highest peak in the Mainri Mountains is Khawa Karpo with an altitude of 6740 m.

The study site is exposed to the Indian monsoon where a great amount of moisture penetrates the Tibetan Plateau along the deep valleys from south to north and causes heavy precipitation. According to the observations of meteorological stations, the regional average annual precipitation is 779 mm, and the annual mean temperature is 9.13 °C [32]. While the summer temperature has increased by 0.012 °C a⁻¹ since 1960, the climate warming increased to 0.020 °C a⁻¹ above the altitude of 3000 m [33]. Surprisingly, the precipitation has shown a nonsignificant change over the past six decades.

Because of the influence of the Indian monsoon, a large number of monsoon temperate glaciers have developed in the Mainri Mountains. Drawing on the First Chinese Glacier Inventory (FCGI), 47 glaciers with an area of 147.02 km² have been identified in the Mainri Mountains since 1975 [34]. Melang Glacier, located on the east slope of the Mainri Mountains, extends in an arc-like shape from 6740 m a.s.l to 2600 m a.s.l., with an area of 12.55 km² and a length of 11.5 km. An average velocity of 533 m a⁻¹ in the Melang Glacier was estimated from the motion of the remains of a mountaineer who passed an ice fall from 1991 to 1998 [35]. Previous studies showed that Melang Glacier advanced ~1200 m from 1959 to 1998 [35] and then retreated significantly, although the elevation of the glacier terminus has risen about 30 m since 1998 [36].

3. Aerial Survey and Data Processing

3.1. UAV Flights and Ground Control Points

In this study, the Phantom 4 RTK (P4R), a four-rotor drone from the Dajiang-Innovations company, was used to monitor the surface of Melang Glacier. Equipped with a 1-inch, 20-megapixel complementary metal oxide semiconductor (CMOS) sensor, the P4R can achieve a ground sample distance (GSD) of 2.74 cm pixel⁻¹ at 100 m in flight altitude in the plain region. Except for the CMOS sensor, the P4R is also complemented by a multifrequency multiconstellation GNSS receiver. The GNSS receiver can retrieve signals from GPS, BeiDou and Galileo constellations in Asia. In this mode, the horizontal positioning accuracy is 1 cm ± 1 ppm (root mean square, RMS), and the vertical positioning accuracy is 1.5 cm ± 1 ppm (RMS) [37]. The maximum flight time of the P4R is approximately 30 min, and the maximum flight altitude is 500 m. Meanwhile, the UBase of Hi-Target, a GNSS receiver, was used as a static base station. In real-time kinematic (RTK) mode, the horizontal positioning accuracy of UBase is 1 cm ± 1 ppm (RMS) and vertical positioning accuracy is 2 cm ± 1 ppm (RMS).

Three UAV surveys were conducted on the tongue of the Melang Glacier on 12 November 2019 (201911), 26 April 2020 (202004) and 6 November 2020 (202011), approximately corresponding to the accumulation period (201911–202004), ablation period (202004–202011) and entire balance year (201911–202011). In each survey, three UAV flights were deployed between 11:00 a.m. and 1:00 p.m. to minimize the effect of shadow. For each flight, the P4R was launched from a flat, off-glacier region on the northern Melang Glacier. To ensure flight security, the flight altitude was not constant. The flight altitudes were set to 120 m, 200 m and 300 m above the launch location for Flight 1, Flight 2 and Flight 3, respectively (Figure 2). The flight durations were 10 min, 10 min and 5 min, and finally, 139, 147 and 34 images were acquired in Flight 1, Flight 2 and Flight 3, respectively. Due to the different flight altitudes, the GSDs were 3.3 cm pixel⁻¹, 5.5 cm pixel⁻¹ and 8.2 cm pixel⁻¹ for the three flights. To mosaic UAV imageries precisely, the image overlap was set to be 80% in the longitudinal direction and 70% in the lateral direction, corresponding to the flight path. During the three UAV surveys, the UBase was deployed near the launch location.

In November 2020, 15 checkpoints (CPs) were set along the northern and southern lateral moraines, and the coordinates of checkpoints were measured by using the real-time kinematic GPS (RTK-GPS) with two FOIF A90 GNSS receivers (one for the base station and one for the rover). To ensure visibility of the CPs in UAV imageries and better recognition of the coordinates for the RTK-GPS rover, CPs were marked with a bright red, 1.0 m × 1.0 m cross (Figure 2). In the RTK mode, the horizontal positioning accuracy of FOIF A90 was 2 cm (RMS), and the vertical positioning accuracy was 3 cm (RMS). Due to the limitations of the terrain, all CPs were set near the glacier terminus. Therefore, the CPs were not used in data processing, and they were just used to estimate the accuracies of the orthophotos and the DSMs after the data processing.

3.2. Data Processing

After finishing the UAV surveys, all UAV imageries were preprocessed. The static GNSS of the base station (*.GNS), the flight log file (*_PPKRAW.bin and *_Timestamp.MRK), all UAV imageries and the coordinate information of base station were imported into the PISHON E-View© software where POS information of all imageries were calculated by using the postprocessed kinematic (PPK) method.

After that, aerial triangulation was performed for UAV imageries to obtain the orthophotos and DSMs. All UAV imageries and POS information were then imported into the Context Capture© software for further analysis of the aerial triangulation. At this point, we calculated the spatial three-dimensional coordinates of all UAV imageries, obtained the high-precision elements of exterior orientation and generated the point cloud. Then, the point cloud was smoothed by the method of Gaussian low-pass filtering. After the processing of aerial triangulation and the point cloud, the construction of the 3D model (TIN, triangulated irregular network) on the tongue of the Melang Glacier was conducted. In the guide of 3D model reconstruction, the spatial reference system was configurated to WGS84/UTM ZONE 47N while the mode was set as adaptive cutting. Whereas the OSGB format was chosen as the achievement output format in the first production project, the orthophotos and DSMs were selected in the second production project. Finally, three orthophotos and three DSMs were obtained with a GSD of 10 cm pixel⁻¹. The data processing is shown in Figure 3.

Thanks to the similarity of the UAV flights and data processes, orthophotos and DSMs can be used for direct glacier change monitoring. The characteristics of the glacier surface can be derived from orthophotos, especially the change in the glacier terminus. The outlines of the Melang Glacier tongue in November 2019, April 2020 and November 2020 were digitized manually from three orthophotos. For glacier surface elevation changes in the accumulation period, ablation period and entire balance year, common DSM differencing with three DSMs was used to generate difference maps. Glacier surface velocity from the UAV orthophotos is generally approximated by the manual feature tracking method. However, significant changes were found in the glacier surface during the surveys, which complicated the precise estimation of glacier surface velocities.

3.3. Accuracy Estimation

To estimate the accuracies of the orthophotos, the DSMs and difference maps, 15 CPs (Figure 2) and the elevation changes in stable off-glacier regions (Figure 1) were used in this study. The accuracy of the orthophotos and DSMs can result from the errors of the UAV flight (E_UAV), checkpoints (E_CP) or the error between UAV imageries and checkpoints (E_UAV-CP). The UAV flight error caused by the inaccuracies of the P4R GNSS receiver and UBase can be calculated by

$$E_{UAV}=\sqrt{E_{P4R}{}^2+E_{UBase}{}^2}$$

(1)

The error of checkpoints is the positioning accuracy of FOIF A90 in the RTK mode. The horizontal accuracy was 2 cm (RMS), and the vertical accuracy was 3 cm (RMS).

The error between UAV imageries and checkpoints can be calculated by

$$E_{UAV-CP}=\frac{{{\displaystyle \sum }}_{i=1}^n\left(P_{i-UAV}-P_{i-CP}\right)}{n}$$

(2)

where $P_{i-UAV}-P_{i-CP}$ is the difference between CP and the corresponding point in the UAV imageries and $n$ is the number of CPs.

Finally, the accuracy of the orthophotos and DSMs can be calculated by

$$E_{imagery}=\sqrt{E_{UAV}{}^2+E_{CP}{}^2+E_{UAV-CP}{}^2}$$

(3)

Assuming that surface elevation in the off-glacier region did not change during surveys, especially the large stable boulders in the off-glacier region, the accuracies of difference maps can be estimated from the elevation changes in stable off-glacier regions.

4. Results

4.1. Uncertainty

The horizontal and vertical positioning accuracies of the P4R GNSS receiver were ~0.01 and ~0.015 m, and the UBase was reported to have ~0.01 and ~0.02 m positioning accuracies in the horizontal and vertical positions. Therefore, the errors of the UAV flights were ~0.014 and ~0.025 m in the horizontal and vertical positions, respectively. Meanwhile, the checkpoints, derived from RTK-GPS with FOIF A90 GNSS receivers, were reported to have ~0.02 and ~0.03 cm positioning accuracies in the horizontal and vertical positions.

By comparing the CPs and corresponding points in 202011 UAV imageries, we found that the offsets ranged from −0.13 to 0.15 m, −0.15 to 0.15 m and −0.11 to 0.13 m in x, y and z directions, respectively. Therefore, the errors between UAV imageries and checkpoints were 0.12 and 0.03 m in the horizontal and vertical positions (Figure 4a). Finally, the horizontal and vertical accuracies of UAV imageries were 0.12 and 0.05 m, respectively.

The accuracies of difference maps were estimated from the elevation changes in stable off-glacier regions. The histograms of surface elevation changes between three DSMs in stable off-glacier regions indicated that more than 90% of elevation changes were concentrated from −0.30 to 0.30 m. The average elevation changes in stable off-glacier regions were 0.02 m in the accumulation period, −0.03 m in the ablation period, and −0.01 m in the entire balance year (Figure 4). We concluded that the average vertical error between three DSMs in off-glacier regions was less than several centimeters and that DSMs derived from UAV surveys were suitable for the estimation of glacier surface elevation changes.

4.2. The Characteristics of Glacier Surface

Based on three UAV flights, UBase, the PPK method and aerial triangulation, three orthophotos and three DSMs were obtained with a GSD of 10 cm pixel⁻¹. The UAV imageries showed that the area of the UAV survey reached 0.78 km², an area of 0.26 km² in the tongue of the Melang Glacier, and the surface elevation in the surveyed glacier area ranged from 2774 to 3362 m a.s.l. (Figure 5).

The surveyed glacier area (SGA) was covered by debris, and a proglacial lake had developed in the glacier terminus (Figure 5h). Apart from the debris cover and proglacial lake, the SGA also displayed a series of crevasses. A significant number of large transversal crevasses had developed in the upper section of the SGA (above 3000 m a.s.l.), and several longitudinal crevasses had developed in the lower section of the SGA (below 3000 m a.s.l.) (Figure 5f,g). The main reason for this difference is the effects of complex terrain on the ice flow. The glacier outlet narrowed gradually in the upper section of the SGA, and then it slowly widened in the lower section of the SGA. A narrow outlet prevents ice extension in the lateral direction, and ice flow is compensated for by longitudinal extension, while a wide outlet provides a large space for lateral ice extension, and ice flow would decrease in the longitudinal direction. Moreover, the DSM shows that the slope of the glacier surface in the upper section of the SGA was larger than that in the lower section of the SGA (Figure 5c). The larger the slope, the faster the ice flow. Therefore, strong ice flow in the longitudinal direction resulted in the development of large transversal crevasses in the upper section of the SGA, and decreased ice flow in the longitudinal direction and increased ice flow in the lateral direction led to the development of several longitudinal crevasses in the lower section of the SGA.

4.3. The Changes in Glacier Surface

The outlines of the SGA in November 2019, April 2020 and November 2020 were digitized manually from three orthophotos (Figure 6). In the upper section of the SGA, the outlines showed spatiotemporal fluctuations, while they remained nearly stable in the entire balance year because of the effect of terrain. The glacier terminus experienced a significant retreat of 78 m during surveys, while the retreat rate in the ablation period (73.2 m or 12.2 m per month) was faster than that in the accumulation period (4.8 m or 0.8 m per month).

Glacier surface elevation changes during surveys were estimated from common DSM differencing with three DSMs. Results show that the tongue of Melang Glacier experienced a heterogeneous pattern of elevation changes (Figure 7). Aggregated over the SGA, the mean surface elevation changes were +10.5 m in the accumulation period, −13.5 m in the ablation period and −3.0 m in the entire balance year. Except for a small region of the glacier terminus, the entire SGA presented a positive change in surface elevation in the accumulation period, corresponding to the ice volume increase of +2.7 × 10⁶ m³. The entire SGA then presented a negative change in surface elevation in the ablation period, which corresponded to the ice volume decrease of −3.5 × 10⁶ m³. In the entire balance year, a positive change in surface elevation was presented in the upper section of the SGA, while a negative change was presented in the lower section of the SGA.

5. Discussion

5.1. Glacier Changes and Its Affected Features

A heterogeneous pattern of glacier surface elevation changes is affected by multiple features, such as debris cover, ice cliffs and supraglacial/proglacial lakes. Some observers believe that ice cliffs and supraglacial/proglacial lakes would have a positive effect on glacier melt in HMA [38,39,40,41,42]. Previous studies indicated that a thin debris cover would accelerate the ablation rate while a thick debris cover would retard the ablation rate [43,44,45,46]. Apart from these features, there are other external factors that would affect glacier surface changes, such as avalanches. Due to the topography of high mountains and deep valleys, the frequency of avalanches is very high in the Mainri Mountains. Many field investigations indicated that avalanches have practically been an annual occurrence in the Melang catchment in recent years (unpublished data). During the surveys, an avalanche occurred in the accumulation period and resulted in a significant impact on the glacier terminus. The avalanche deposit covered the terminus of Melang Glacier and even filled the proglacial lake. Therefore, the terminus of Melang Glacier remained relatively stable in the accumulation period. Except for the relatively stable glacier terminus, the avalanche also resulted in significant thickening of more than 30 m in surface elevation at the glacier terminus (Figure 8).

Even though the proglacial lake was filled by the avalanche in the accumulation period, a new proglacial lake developed in the ablation period. Subaerial and subaqueous calving has been proved to be an important component of glacier recession [38,39,40]. The processes of calving were not tracked during surveys, while glacier surface elevation changes in the ablation period indicated that the proglacial lake had a positive effect on glacier melt. During the ablation period, significant glacier surface thinning was observed in the lower section of the SGA, and maximum thinning of more than 40 m was found around the new proglacial lake (Figure 8).

None of the supraglacial lakes were found in the Melang Glacier, while a significant number of ice cliffs have developed in the SGA. Ice cliffs are generally covered by fine debris which reduces the ice–albedo, absorbs more shortwave radiation and then accelerates the ablation rate on the exposed cliffs [47]. The thickness of debris cover was not measured in the SGA, although fine debris covering ice cliffs was seen in the field pictures (Figure 5d,e). It can be concluded that the surface properties of debris cover, ice cliffs and proglacial lakes had a positive effect on ice melt in the SGA over the survey period.

5.2. Climate Background

The Mainri Mountains are exposed to the Indian monsoon, and a tendency toward glacier recession has been observed since 1900 [48]. Previous studies indicated that precipitation presented a nonsignificant increase, while it has decreased and fluctuated since 2000 [48]. The summer temperature has increased by 0.012 °C a⁻¹ since 1960, and climate warming is more apparent at the high altitudes [33]. Under the influence of warming and insignificant precipitation, glaciers show high sensitivity to temperature changes, such as glaciers in southeastern Tibetan Plateau and the Rocky Mountains [49,50,51].

To analyze the response of Melang Glacier to climate change in an entire balance year, air temperature and precipitation were measured at the meteorological station in front of the glacier terminus (Figure 9a,b). According to the observations of the meteorological station, the mean temperature was 9.38 °C, and the total precipitation was 615 mm in the glacier terminus during surveys. The SGA experienced a significant retreat of 78 m and a mean surface elevation change of −3.0 m. Such mass loss caused by temperature exceeded the mass accumulation from precipitation and led to significant glacier wastage. This is consistent with the relationship between glacier surface elevation and climate change in other monsoon temperate glaciers [15,52]. In the accumulation period, the mean temperature was 4.41 °C, and the mean temperature in January was below 0 °C. Meanwhile, the total precipitation was 252 mm in the accumulation period, and heavy precipitation of more than 130 mm was found for January. It can be assumed that it was the heavy snowfall in January which correlates with the avalanche in the accumulation period. Drawing on the SGA, the mean surface elevation change was +10.5 m in the accumulation period. Apart from the mass accumulation from precipitation, there may be other factors affecting glacier surface elevation. As mentioned before, strong ice flow had been estimated in the upper section of the SGA, including strong vertical emergence velocity, which can be proved by the significant uplift in the bend region of the SGA. Similar conclusions were found for the Lirung Glacier in the Himalaya and the Parlung No. 4 Glacier in the southeastern Tibetan Plateau [29,30]. In the ablation period, the mean temperature was 14.30 °C, the total precipitation was 363 mm, and the mean surface elevation change of the SGA was −13.5 m, which indicated that temperature played an important role in glacier wastage. Given all that, the SGA of Melang Glacier was more sensitive to temperature, especially in the lower section. While the surface elevation change in the upper section of the SGA was controlled by vertical emergence velocity in the accumulation period, it turned out that the elevation change was also influenced by the temperature in the ablation period.

To obtain the tendency of climate change in the study site, temperature and precipitation datasets were collected from the nearest meteorological station (Deqin Station). Acquired from the China Meteorological Data Service Centre (CMDSC, http://data.cma.cn/en, accessed on 1 May 2021), the dataset shows that average temperature has increased significantly from 1960 to 2018 (Figure 9c). Meanwhile, the standard deviation of temperature remained stable, indicating that the temperature in the study area does not fluctuate significantly. Meteorological station records indicate that the average air temperature increased in the Mainri Mountains by more than 0.4 °C per decade (with a confidence level <0.05), higher than the global rate of warming (0.12 °C per decade, 1951–2012), while the trend in precipitation was not evident at the Deqin Sation, which presented large interannual precipitation fluctuation (Figure 9d). By considering the sensitivity of the glacier to temperature changes, the Melang Glacier may have experienced a significant recession and mass loss in the past few decades. Furthermore, the Sixth Assessment Report (AR6) of IPCC WGI shows that a 1.5 °C increase in the 20-year average of global surface air temperature (GSAT), relative to the average over the period 1850–1900, is very likely to occur in scenario SSP5-8.5, and more likely than not to occur in scenario SSP1-1.9 and SSP1-2.6 in the near term (2021–2040) [3]. It is estimated that a substantial mass loss of the Melang Glacier is likely to occur in the next few decades.

5.3. Strengths and Weaknesses of the UAV Survey

In the last few decades, space-borne remote sensing has been applied widely in glacier monitoring [7,8,19]. However, due to the topography of high mountains and deep valleys, as well as the effect of the Indian monsoon, space-borne remote sensing cannot be used for the glacier monitoring in the Mainri Mountains. Compared with space-borne remote sensing, the strengths of UVA surveys are significant. The UAV survey provides results and solutions when space-borne remote-sensing-derived products are not available. Geodetic methods based on space-borne remote sensing are generally used for glacier monitoring at a decadal scale, while the UAV survey can be used for glacier monitoring at the annual scale, even the seasonal scale. In addition, except for glacier surface elevation and velocity, the characteristics of crevasses, ice cliffs, supraglacial lakes and ice avalanches can be detected by the UAV survey. These features play an important role in glacier changes.

However, the weaknesses of UAV surveys are too significant to ignore. First, it is not safe for the UAV to survey rugged mountain terrain. With improper operation, the UAV would crash to the ground with no chance of retrieval. In addition, because of the low-temperature environment, power consumption of UAVs is faster at high altitudes. The faster power consumption results in the relatively short length of UAV flights. Last but not least, to order to improve the image resolution and overlap rate, UAVs generally fly at lower altitudes. This would limit the area of UAV surveys.

In the future, the weaknesses of UAV surveys will be overcome, such as increased battery capacity and higher resolution of the CMOS sensor. However, for now, the UAV survey can bridge the gap between low-quality space-borne remote sensing and scarce in situ measurements, and is a more appropriate approach for monitoring the changes in temperate glaciers.

6. Conclusions and Perspectives

In this study, three UAV surveys were conducted for the Melang Glacier in the Mainri Mountains in November 2019, April 2020 and November 2020. Based on three UAV flights, UBase, the PKK method and aerial triangulation, three orthophotos and three DSMs were obtained, with a GSD of 10 cm pixel⁻¹.

The surveyed glacier area was covered by debris, and a proglacial lake had developed in the glacier terminus. Numerous large transversal crevasses had developed in the upper section of the SGA, and several longitudinal crevasses were created in the lower section of the SGA.

The glacier terminus experienced a significant retreat of 78 m during the survey period, while the retreat rate in the ablation period was faster than that in the accumulation period. A mean surface elevation change of −3.0 m was observed in the surveyed glacier area from November 2019 to November 2020 with +10.5 m and −13.5 m in the accumulation period and ablation period, respectively. The avalanche, debris cover, ice cliffs and proglacial lake resulted in a heterogeneous pattern of glacier surface elevation changes.

There was a close relation between glacier changes and temperature in the lower section of the surveyed glacier area. While the surface elevation change in the upper section of the SGA was controlled by vertical emergence velocity in the accumulation period, it turned out that the elevation change was also influenced by the temperature in the ablation period. By considering the sensitivity of the glacier to temperature changes, the Melang Glacier may have experienced a substantial recession and mass loss in the past few decades. Furthermore, it is estimated that a substantial mass loss of the Melang Glacier is likely to occur in the next few decades, considering different future scenarios.

This study indicated that the UAV survey can bridge the gap between low-quality space-borne remote sensing and scarce in situ measurements. Even though it has some weaknesses, the UAV survey is a more appropriate approach for monitoring the changes in temperate glaciers.