Evolution of Surge-Type Glaciers in the Yangtze River Headwater Using Multi-Source Remote Sensing Data

: A surge-type glacier is a special and dangerous type of glacier, which can advance quickly in a short-time with cycles. Glaciers in the Yangtze River headwater are generally acknowledged to be in a stable state. However, not all of those glaciers are stable. In this paper, ﬁve glaciers from the Yangtze River headwater glacier were selected as the experimental subjects, and multi-source remote sensing images were used to study and analyze the surge behavior over the past 30 years. Based on the Landsat series data, ERS-2, and ENVISAT radar data, this paper extracts the glacier centerline information, glacial area information, and glacial ﬂow velocity during di ﬀ erent time periods from 1988 to 2018, which are used to monitor the active periods of glacier surges. We found three surge-type glaciers in the study area. The glacial characteristics of the three glaciers showed some drastic changes, they can advance quickly nearly 800 m in active periods, their area change can reach 2.0 × 10 6 m 2 , and their ﬂow velocity can suddenly increase by dozens of times. Surging periods and the initiated time of the three glaciers are di ﬀ erent, which are locked in 1997, 2003, and 1997–1998. All those surges ended within one to two years. We suggest that the surges in this paper are dominated by hydrological conditions.


Study Area
The Qinghai-Tibet Plateau, known as the third pole of the world, has the largest mountain glaciers in the middle and low latitudes of the world [30]. Five large glacial gathering areas have been formed in the northwest and southwest of China, which are distributed in the Tianshan Mountains, the Pamirs, and the Kunlun Mountains; Karakorum Mountain and Tanggula Mountain [31].
The Tanggula Mountain is located in the Sanjiangyuan area of Qinghai Province. The Sanjiangyuan area is the "fresh water resource centre" in China, which supplies the main part of downstream rivers, such as the Yangtse River, the Yellow River, and the Lancang River [32]. In 2003, it was approved by the State department as a national nature reserve [33]. The Yangtze River has a length of 6380 kilometers, which is the longest river in China and the third-longest in the world. The river originates from the Sanjiangyuan area, which has an average elevation of more than 5000 meters and more than 100 glaciers [34]. The five glaciers in this paper are in the headwater glaciers of the Yangtze River ( Figure 1). The red rectangle represents the study area, and the blue point represents the meteorological station (Figure 1a), In Figure 1b, the red line represents the five study glaciers in this paper, the red rectangle represents the study area, the blue point represents the meteorological station, and the blue line represents the SAR data region. In Figure 1c, the black lines represent the five study glaciers, which scattered from high altitude to low altitude. To facilitate statistics and research, we named the five glaciers as glacier A, glacier B, glacier C, glacier D, and glacier E, and the particular attributes of these five glaciers as shown in Table 1 (Glacier D and Glacier are attributes of G091104E33504N). The Global Land Ice Measurements from Space Monitoring the World's Changing Glaciers (GLIMS) is a famous global project designed to monitor the world's glaciers, and most glacier research uses its information to define study glaciers.

Data
Based on multi-source data for long-term observation, we aimed to find suitable ways for synergistic observation in future work. We needed to find stable and reliable data types in mature papers. This paper selected the optical images (Landsat series satellites), the Synthetic Aperture Radar (SAR) data (ERS-2 satellite and ENVISAT satellite), and the Shuttle Radar Topography Mission (SRTM) elevation data to study decadal glacial movement.

Landsat 5/7/8
The Landsat series satellites with the Multispectral Scanner (MS), Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+) and Operational Land Imager (OLI) sensors was used in which a long-term continuous series of images provide continuous and effective data support to monitoring the surging glaciers over a long time [35]. The 15 optical remote sensing data used in this paper are from the United States Geological Survey (USGS), with a resolution of 30 m (Table S1) [36]. Although the Landsat series data has a short revisit time (16 days), the optical data of the study area spanning from 1988 to 2018 had a high cloud cover, so there is less high-quality data.

ERS-2/ENVISAT
The satellites with a synthetic aperture radar (SAR) sensor provide useful and high-quality data that is not influenced by the weather. The European Space Agency (ESA) launched the ERS2 satellite and ENVISAT satellite in April 1995 and March 2002, respectively, making them become the largest environmental satellites [37]. ERS2 satellite carries an SAR sensor, and ENVISAT satellite carries an ASAR sensor. SAR data can be observed not only all-day and all-weather but also has strong penetration [38]. The ASAR sensor of the ENVISAT satellite and the SAR sensor of the ERS-2 satellite have many similar parameters, such as incidence angle and resolution. The ASAR sensor of the ENVISAT satellite has made some improvements based on the ERS-2 SAR sensor [39]. The SAR data used in this paper are all descending orbits data, VV polarization mode, which spanned from 1996 to 2010 (Tables S2 and S3).

SRTM
This paper used digital elevation data to assist SAR data in calculating glacier surface velocity and to observe elevations in the study area [40]. We used the SRTM C-band digital elevation data. SRTM V3.1 with a spatial resolution of 30 m was used in the study, which can cover the studied glaciers in the Sanjiangyuan area, namely N33E090 and N33E091 (Table S4.).

Methods
The techniques of visual interpretation and offset-tracking were chosen in this study due to its benefits in the long-term observation of glaciers [41] On a temporal scale, the visual interpretation of Landsat series data provides obvious glacial information over years, we can easily lock the sensitive periods, and offset-tracking supports the glacial inner information over days or months. The techniques we chose are reliable in a glacial movement study.

The Method to Extract Glacial Flow Velocity
Landsat series images were chosen to provide glacier centerline and area information. There are many ways to access the glacial centerline and area, and the visual interpretation is an appropriate method with the highest accuracy rate. The acquired Landsat satellite images must undergo geometric correction and radiation correction, aiming to achieve quantitative remote sensing analysis. We combined the bands 543 of Landsat 5 and 7 images and the bands of 654 of Landsat 8 images to help with visual interpretation [42]. This paper calculated the glacier centerlines of 1988, 1991, 1993, 1994, 1997, 2000, 2001, 2002, 2003, 2004, 2005, 2009, 2013, 2016, and 2018. At the same time, we chose the same technique to extract glacier outlines in different years and calculate the area of the study glaciers [43].

Extraction of Glacial Flow Velocity
SAR images were elected to access glacial flow velocity information in this paper. An offset-tracking technique was chosen to calculate the glacial flow velocity because of its known low requirement of coherence to SAR data [44]. Comparing with other interferometry techniques of SAR images, the offset-tracking technique can access glacial flow velocities over a longer interval (dozens of days), which is suited for decadal observation. The core of the offset-tracking technique is a normalized cross-correlation algorithm, which accesses the offset value based on the intensity images of SAR [45]. Those two images are, respectively, named master image (earlier image) and slave image (later image), and needed to undergo a registration process before calculation. The key step is choosing a suitable reference window in the master image and matching it with a corresponding research window of the slave image. Then we could achieve the offset according to the calculated maximum coefficient of cross-correlation [46,47]. In this process of achieving glacial flow velocity, the offset between the master image and the slave image is calculated in slant-rang direction and azimuth direction. Finally, we can achieve the total offset with both slant-range and azimuth direction [48].
In this paper, we set the reference window size as 64 × 320 pixels, which forms a square patch and gives enough pixels to calculate the cross-correlation coefficient. The offset we calculated includes the glacial displacement, the orbital offset, the ionosphere offset, and topography-related offset. The orbital offset could be calculated by the SAR orbital parameters. The ionosphere offset could be ignored because the studied glaciers are located in mid or low latitude regions. The topography-related offset could be moved owing to the external DEM [49]. After removing the other offsets, we finally achieved the displacement of two SAR images. The geocoding process helped to transform the SAR coordinate system into a geographic coordinate system. Finally, we calculated the glacial velocity based on the interval of the master and slave images. The detailed process diagram is shown in Figure 2.

Glacier Centerline and Area
The Landsat series data were easily affected by the cloud cover, and only 15 high-quality images were selected. The internal between two adjacent images was kept within one to three years.

Glacier Centerline and Area
The Landsat series data were easily affected by the cloud cover, and only 15 high-quality images were selected. The internal between two adjacent images was kept within one to three years. Due to the trouble of the Landsat 7 sensor since May 2003, two images (the day of 2003 and 2004) during the study period had strip-like data gaps. However, there is little influence in our study. The five glacial centerline information and area values were calculated based on visual interpretation. Figure 3 shows information of five glacier tongue outlines from 1988 to 2018. The base images are five glacier tongues from optical remote sensing of 1988, and the 15 different colored outlines represent information of the glacier tongue in different years over the past three decades. We can easily see the change during the past three decades. We can clearly see that the area and shape of the five glacier tongues change over the years. After accessing the glacier centerline and area the information, we repeated the work four times at different times and calculated the standard deviation. The results show that the standard deviation of the length of the glacier centerline does not exceed 26.2 m, and the standard deviation of the glacier area does not exceed 1000 m 2 . It proves that the results in this paper are reliable.

Glacier Flow Velocity
Due to the limitation of the SAR data, 10 ers-2 SAR images and 18 ENVISAT ASAR images, having the same orbit number and sensor, were selected in this experiment, and the offset-tracking technique was used to access the glacial flow velocities from 1996 to 2010. Figure 4 shows the glacial flow velocity maps. In Figure 4, the base image is studying the area's SAR images, which has undergone a multi-looking process, and the colored raster data represents velocities in different periods. Due to larger intervals in some master images and slave images, a few areas have poor coherence and have no result of velocity. Fortunately, the vast majority of the results are good. We can directly observe the velocity change. The internal of master and slave images in this paper are

Glacier Flow Velocity
Due to the limitation of the SAR data, 10 ers-2 SAR images and 18 ENVISAT ASAR images, having the same orbit number and sensor, were selected in this experiment, and the offset-tracking technique was used to access the glacial flow velocities from 1996 to 2010. Figure 4 shows the glacial flow velocity maps. In Figure 4, the base image is studying the area's SAR images, which has undergone a multi-looking process, and the colored raster data represents velocities in different periods. Due to larger intervals in some master images and slave images, a few areas have poor coherence and have no result of velocity. Fortunately, the vast majority of the results are good. We can directly observe the velocity change. The internal of master and slave images in this paper are kept within months as far as possible because the shorter periods can access a more accurate flow velocity. According to the regulation of the glacial flow velocity in Figure 4, we can summarize that the upper-glacier accelerated at first, and then, the state of accelerated motion slowly turns the glacial tongue in active periods. As to the accuracy assessment of the glacial flow velocity, several areas of non-ice region (The blue area in Figure 4a) were selected to calculate the displacement velocity. The average flow velocity of the non-ice area is 9.8 mm/d, which is far less than ice flow velocity. This can prove that the flow velocities in this manuscript are reliable.

Glacier Flow Velocity
Due to the limitation of the SAR data, 10 ers-2 SAR images and 18 ENVISAT ASAR images, having the same orbit number and sensor, were selected in this experiment, and the offset-tracking technique was used to access the glacial flow velocities from 1996 to 2010. Figure 4 shows the glacial flow velocity maps. In Figure 4, the base image is studying the area's SAR images, which has undergone a multi-looking process, and the colored raster data represents velocities in different periods. Due to larger intervals in some master images and slave images, a few areas have poor coherence and have no result of velocity. Fortunately, the vast majority of the results are good. We can directly observe the velocity change. The internal of master and slave images in this paper are kept within months as far as possible because the shorter periods can access a more accurate flow velocity. According to the regulation of the glacial flow velocity in Figure 4, we can summarize that the upper-glacier accelerated at first, and then, the state of accelerated motion slowly turns the glacial tongue in active periods. As to the accuracy assessment of the glacial flow velocity, several areas of non-ice region (The blue area in Figure 5(1)) were selected to calculate the displacement velocity. The average flow velocity of the non-ice area is 9.8 mm/d, which is far less than ice flow velocity. This can prove that the flow velocities in this manuscript are reliable.

The Evolution of Glacial Centerline and Area
The statistics of glacial centerline length and area achieved above can be used further summarization and analysis. The evolution of the glacial centerline length and areas are shown in Figures 5 and 6. In order to intuitively display the change rate of the glacial centerline and area information, all statistics are subtracted by their respective values of 1988. We can easily get the change rate over the years and lock the glacial active phase.

The Evolution of Glacial Centerline and Area
The statistics of glacial centerline length and area achieved above can be used further summarization and analysis. The evolution of the glacial centerline length and areas are shown in Figures 5 and 6. In order to intuitively display the change rate of the glacial centerline and area information, all statistics are subtracted by their respective values of 1988. We can easily get the change rate over the years and lock the glacial active phase.

The Evolution of Glacial Centerline and Area
The statistics of glacial centerline length and area achieved above can be used further summarization and analysis. The evolution of the glacial centerline length and areas are shown in Figures 5 and 6. In order to intuitively display the change rate of the glacial centerline and area information, all statistics are subtracted by their respective values of 1988. We can easily get the change rate over the years and lock the glacial active phase.

The Glacial Centerline Evolution
According to the length of the evolution of the glacial centerline ( Figure 5), the five glaciers' centerline evolution all show a retreating trend from 1988 to 2018. However, each glacier has its own change characters. Glacier A, B, and E have fierce fluctuation periods, and Glacier C and D have a stable evolution over the past 30 years.
The centerline evolution of Glacier A indicates that the length firstly has continuously retracted

The Glacial Centerline Evolution
According to the length of the evolution of the glacial centerline ( Figure 5), the five glaciers' centerline evolution all show a retreating trend from 1988 to 2018. However, each glacier has its own change characters. Glacier A, B, and E have fierce fluctuation periods, and Glacier C and D have a stable evolution over the past 30 years.
The centerline evolution of Glacier A indicates that the length firstly has continuously retracted from 1988 to 1997, then fiercely advanced from 1997 to 2000, and finally, showed a stable retracted trend from 2000 to 2018. It is estimated that the surging of Glacier A occurred between 1997 and 2000 based on the glacier centerline evolution information and the concept of surge-type glaciers. The centerline information of Glacier B shows that, firstly, the length has been continuously retracted from 1988 to According to the centerline and area evolution of the five study glaciers, we can analyze the glacial movement and further infer the glacial active surge periods. Generally, the regulation of the glacial centerline and area shows a retreat tendency, and the results of two ways maintain a high degree of consistency, such as Glacier A and B, which are typical glaciers with regular shape. Combining the results of the two ways of optical remote sensing data, we inferred that Glacier A had a surging event between 1997 and 2000. Glacier B had a surging event between 2002 and 2005 (possibly happened in 2002-2004). However, some specific glaciers, such as Glacier E, are slightly different from the typical glaciers. As to Glacier E, which has an irregular shape, the two conclusions accessed above are inconsistent. We need to use another useful method to reconfirm (such as experience interpretation in this paper). After an experienced interpretation based on optical remote sensing images (Figure 3), we can find that Glacier E had the most significant change of the glaciers between 1997 and 2000.

The Evolution of Glacial Flow Velocity
The interval between two adjacent optical remote sensing images in this study is still long. We can only infer the surging active phases over a long period. With the flow velocities, we can shorten the Remote Sens. 2019, 11, 2991 11 of 17 active phases to obtain more accurate results. Using the glacial flow velocities of the five studied glaciers, we extracted the velocity values along the centerline (the black curve in Figure 4a) for further studying and discussion. To display relevant information, we extracted the eight points on the centerline at equal intervals, which have undergone mean calculation with several surrounding points. Figures 7-11 show the speed on the vertical axis and time on the horizontal axis. Each line represents the point we chose on the glacier's centerline.
With the flow velocity information accessed above, we obtained the eight centerline point velocities of Glacier C and Glacier D from 1996 to 2010 (Figures 7 and 8). From Figure 7, eight points' velocities show similar rules. Namely, the velocities on Glacier C's centerline keep stable from 1966 to 2010, the maximum speed does not exceed 200 mm/d, and the common speed is around 100 mm/d. It can be inferred that Glacier C has no surging phenomenon from 1996 to 2010. From Figure 8, Glacier D's movement slightly fluctuates in some years. The maximum speed from 1996 to 1997 nearly reached 400 mm/d, and the normal speed is around 200 mm/d. The speed has no exponential increasing phenomenon during those years, so we inferred that Glacier D has no surging phenomenon from 1996 to 2010. sensing images (Figure 3), we can find that Glacier E had the most significant change of the glaciers between 1997 and 2000.

The Evolution of Glacial Flow Velocity
The interval between two adjacent optical remote sensing images in this study is still long. We can only infer the surging active phases over a long period. With the flow velocities, we can shorten the active phases to obtain more accurate results. Using the glacial flow velocities of the five studied glaciers, we extracted the velocity values along the centerline (the black curve in Figure 4(1)) for further studying and discussion. To display relevant information, we extracted the eight points on the centerline at equal intervals, which have undergone mean calculation with several surrounding points. Figures 7-11 show the speed on the vertical axis and time on the horizontal axis. Each line represents the point we chose on the glacier's centerline.
With the flow velocity information accessed above, we obtained the eight centerline point velocities of Glacier C and Glacier D from 1996 to 2010 (Figures 7 and 8). From Figure 7, eight points' velocities show similar rules. Namely, the velocities on Glacier C's centerline keep stable from 1966 to 2010, the maximum speed does not exceed 200 mm/d, and the common speed is around 100 mm/d. It can be inferred that Glacier C has no surging phenomenon from 1996 to 2010. From Figure 8, Glacier D's movement slightly fluctuates in some years. The maximum speed from 1996 to 1997 nearly reached 400 mm/d, and the normal speed is around 200 mm/d. The speed has no exponential increasing phenomenon during those years, so we inferred that Glacier D has no surging phenomenon from 1996 to 2010.  Based on the roughly monitored periods of surging with the optical images, we further monitored the surging period of Glacier A, Glacier B, and Glacier E. Based on the flow velocity values of the three glaciers, we can get the fitting trend maps of the three glaciers. According to the trend map (Figures 10,11,12), we can further monitor the movement of the three glaciers. In general, the flow velocity of Glacier A changed sharply from 1996 to 2000 ( Figure 10). The upper-middle area of Glacier A exceeded 1000 mm/d in April 1996 and reached its maximum in August 1996, in which some areas exceeded 4000 mm/d. In 1997, the movement was still intense, and the flow velocity surged to the end of the glacier. In July 1998, the flow velocity remained basically within 300 mm/d, and the normal state of motion was restored. During this time, the maximum speed of Glacier A (more than 4000 mm/d) is more than 10 times that of the general state (less than 300 mm/d), so a glacial surging event had happened, and the surging event may have occurred in 1997. Based on the roughly monitored periods of surging with the optical images, we further monitored the surging period of Glacier A, Glacier B, and Glacier E. Based on the flow velocity values of the three glaciers, we can get the fitting trend maps of the three glaciers. According to the trend map ( Figures 10-12), we can further monitor the movement of the three glaciers. In general, the flow velocity of Glacier A changed sharply from 1996 to 2000 ( Figure 10). The upper-middle area of Glacier A exceeded 1000 mm/d in April 1996 and reached its maximum in August 1996, in which some areas exceeded 4000 mm/d. In 1997, the movement was still intense, and the flow velocity surged to the end of the glacier. In July 1998, the flow velocity remained basically within 300 mm/d, and the normal state of motion was restored. During this time, the maximum speed of Glacier A (more than 4000 mm/d) is more than 10 times that of the general state (less than 300 mm/d), so a glacial surging event had happened, and the surging event may have occurred in 1997.
increased in April 2003, and the velocity exceeded 800 mm/d. The flow velocity in the first half of 2004 was also around 400 mm/d. It is known that the movement gradually returned to normal at the end of 2004, and it remained within 150 mm/d. The maximum speed of Glacier B from 2003 to 2010 (more than 800 mm/d) is five times the normal speed (less than 150 mm/d). It can be speculated that Glacier B had a glacier surge event during this period, and the surging event possibly occurred in 2003.
In contrast, the movement of Glacier E is slower. Under normal conditions, Glacier E moves slowly, generally staying below 100 mm/d. In many cases, the flow velocity is about 50 mm/d, which is relatively stable. In April 1996, the glacier flow rate exceeded 150 mm/d. By July 1998, the glacier movement reached 250 mm/d, and some areas even approached 300 mm/d. After that, the glacier flow velocity gradually recovered. Based on the trend graph, we can infer that the Glacier E had a glacier jump in 1997-1998.

Discussion
With multi-source remote sensing data and methods, the five glacial active periods are respectively locked in one to two years. According to the concept and characteristics of surge-type

Discussion
With multi-source remote sensing data and methods, the five glacial active periods are respectively locked in one to two years. According to the concept and characteristics of surge-type glaciers, the surge events may occur in cycles, and we can infer the corresponding conclusion with the evolution of glacial centerline and area information, in which the withdraw and advance has Although Glacier B has similar location information and elevation information to Glacier A, the movement of Glacier B is far less severe than Glacier A. The flow velocity of Glacier B suddenly increased in April 2003, and the velocity exceeded 800 mm/d. The flow velocity in the first half of 2004 was also around 400 mm/d. It is known that the movement gradually returned to normal at the end of 2004, and it remained within 150 mm/d. The maximum speed of Glacier B from 2003 to 2010 (more than 800 mm/d) is five times the normal speed (less than 150 mm/d). It can be speculated that Glacier B had a glacier surge event during this period, and the surging event possibly occurred in 2003.
In contrast, the movement of Glacier E is slower. Under normal conditions, Glacier E moves slowly, generally staying below 100 mm/d. In many cases, the flow velocity is about 50 mm/d, which is relatively stable. In April 1996, the glacier flow rate exceeded 150 mm/d. By July 1998, the glacier movement reached 250 mm/d, and some areas even approached 300 mm/d. After that, the glacier flow velocity gradually recovered. Based on the trend graph, we can infer that the Glacier E had a glacier jump in 1997-1998.

Discussion
With multi-source remote sensing data and methods, the five glacial active periods are respectively locked in one to two years. According to the concept and characteristics of surge-type glaciers, the surge events may occur in cycles, and we can infer the corresponding conclusion with the evolution of glacial centerline and area information, in which the withdraw and advance has obvious regularity. According to the principle of mass balance of glaciers [50], we found that the glacial centerline length and area will decrease before the surging and increase after the surging, which aims to maintain material balance. It is the accumulation state of glacial material and energy, and the state of income in the material balance [5]. Once accumulation or income is saturated, glacial surging occurs, that is, the release phase of glacial surging [4]. The complete glacial surging phenomenon realizes the income and expend of material energy and completes the stage of accumulation and release of ice and snow.
Many researches about surge-type glaciers have hypothesized that climate may be incentives for glacier surging, considering that dramatic changes in climate may lead to catastrophic events [51]. The nearest weather station to this study area is the Tuotuohe Station (Figure 1a). According to the accumulated summer precipitation and average summer temperature of Tuotuohe Station from 1970 to 2015 [52] (Figures 12 and 13), it is easy to see that there were no extreme weather events in the study area between 1970 and 2015, and both precipitation and temperature show a slow upward trend. The five glaciers in this paper have highly similar external and internal conditions, such as similar elevation changes, climatic conditions, and so on. The experiment results show that some glaciers were surging and some glaciers remained stable, and the surging active periods are different. So, the warming of climate is not the dominant factor leading to the surging phenomenon in the five glaciers and might be a small factor with a positive effect.
Remote Sens. 2019, 6, x FOR PEER REVIEW 13 of 18 glaciers were surging and some glaciers remained stable, and the surging active periods are different. So, the warming of climate is not the dominant factor leading to the surging phenomenon in the five glaciers and might be a small factor with a positive effect.    Previous studies interpreted that hydrological and thermal controls are the reasons for the glacier surging phenomenon within the Tibet Plateau [53]. The hydrologically controlled surge front represents the boundary between an efficient tunnel drainage system in the up-glacier direction and an inefficient system in the down-glacier direction. Thermally controlled surges rely on a change in conditions at the bed, and the surge fronts, in this case, represent the transition between up-glacier and down-glacier [54,55]. The active phase of the hydrologically controlled surge is shorter, and the active phase thermally controlled surges is longer [56]. Lv et al. studied the Kelayayilake Glacier, which surged dramatically during the Spring of 2015 (From April to July), and came to the conclusion that hydrological factors controlled this surging event [51]. Lv et al. studied the 28 glaciers in Karakoram and found 13 surge-type glaciers. They suggest that both hydrological and thermal controls are important for surge initiation and recession, and the quickly-advanced and short-term surging glaciers were mainly dominated by hydrological control [57]. The periods of the active phase in this study area are all within one to two years, and those glaciers belong to quickly advanced and short-term surging glaciers. The results of glacial velocities show that the speed of up-glaciers pioneered increases firstly and then turns to glacial tongue. The climate change of Tuotuohe station shows that precipitation and temperature are not the fatal reasons for the surging phenomenon. Although we have no access to local hydrological data, we suggested that three glaciers in this paper were mainly controlled by hydrological conditions based on the glacial velocities change regulation and concept of surge-type glaciers.

Conclusion
This paper proposes a method based on multi-source remote sensing data to study and analyze the surge-type glaciers. Based on Landsat images and the result of offset-tracking on ERS-2 and ENVISAT radar data, we have extracted several kinds of glacial feature information in the Yangtze River headwater glaciers to studying the surge-type glaciers.
(1) By analysis and quantifying the centerline and area change evolution of Landsat series data from 1988 to 2018, we affirm that Glacier A, B, E are surge-type glaciers, and the active phase of surging glaciers are being, respectively, roughly locked. Combining with glacial flow velocities of SAR data from 1996 to 2010, the active periods are locked in 1997, 2003, and 1997-1998. Previous studies interpreted that hydrological and thermal controls are the reasons for the glacier surging phenomenon within the Tibet Plateau [53]. The hydrologically controlled surge front represents the boundary between an efficient tunnel drainage system in the up-glacier direction and an inefficient system in the down-glacier direction. Thermally controlled surges rely on a change in conditions at the bed, and the surge fronts, in this case, represent the transition between up-glacier and down-glacier [54,55]. The active phase of the hydrologically controlled surge is shorter, and the active phase thermally controlled surges is longer [56]. Lv et al. studied the Kelayayilake Glacier, which surged dramatically during the Spring of 2015 (From April to July), and came to the conclusion that hydrological factors controlled this surging event [51]. Lv et al. studied the 28 glaciers in Karakoram and found 13 surge-type glaciers. They suggest that both hydrological and thermal controls are important for surge initiation and recession, and the quickly-advanced and short-term surging glaciers were mainly dominated by hydrological control [57]. The periods of the active phase in this study area are all within one to two years, and those glaciers belong to quickly advanced and short-term surging glaciers. The results of glacial velocities show that the speed of up-glaciers pioneered increases firstly and then turns to glacial tongue. The climate change of Tuotuohe station shows that precipitation and temperature are not the fatal reasons for the surging phenomenon. Although we have no access to local hydrological data, we suggested that three glaciers in this paper were mainly controlled by hydrological conditions based on the glacial velocities change regulation and concept of surge-type glaciers.

Conclusions
This paper proposes a method based on multi-source remote sensing data to study and analyze the surge-type glaciers. Based on Landsat images and the result of offset-tracking on ERS-2 and ENVISAT radar data, we have extracted several kinds of glacial feature information in the Yangtze River headwater glaciers to studying the surge-type glaciers.
(1) By analysis and quantifying the centerline and area change evolution of Landsat series data from 1988 to 2018, we affirm that Glacier A, B, E are surge-type glaciers, and the active phase of surging glaciers are being, respectively, roughly locked. Combining with glacial flow velocities of SAR data from 1996 to 2010, the active periods are locked in 1997, 2003, and 1997-1998. (2) According to the results in this paper, combined with the concept of surging phenomenon and previous study experience, we suggest the three surges are dominated by hydrological conditions, which aims to keep a balance of inner glaciers.
Supplementary Materials: The following Landsat-5, Landsat-7, Landsat-8 images SRTM data are available online at http://www.mdpi.com/2072-4292/11/24/2991/s1, Table S1: Landsat Series Image Information List. The following ERS-2 and ENVISAT data are available from the European Space Agency, Table S2: List of ERS-2 satellite radar data image information, Table S3: ENVISAT satellite radar data image information list, Table S4: SRTM data image information list.