Introduction

Surge-type glaciers constitute only a small percentage of the world's glaciers and are gathered in clusters, which are nonrandomly distributed over the world (Jiskoot et al., 1998; Sevestre and Benn, 2015).Though small in number, identification, monitoring, and a solid understanding of the causes, mechanisms, and factors influencing glacier surges are crucial, because surging has caused major hazards such as outburst floods of dammed lakes and associated impacts in the proglacial area (Bruce et al., 1987; Jiskoot et al., 1998; Kotlyakov et al., 2008; Häusler et al., 2016). With increased human activity in mountainous areas for various reasons such as habitation, recreation, resource extraction, or transportation, it is important to update our knowledge about the locations of the surging glaciers so that disasters related to hazardous impacts from moving ice can be avoided. Surges can alter the shape of the glacier and change the terrain by erosion of ice with the glacier bed, which may form large boulders and destroy nearby vegetation or man-made infrastructure such as hydroelectric power stations and villages (Kotlyakov et al., 2008; Kotlyakov, 2004). Monitoring of past surges and assessment of the advance, volume change, and velocities of ice movements during surging may serve as important parameters in predicting the future surges of the same glacier. In addition, these glaciers need special consideration in glacio-hydrological models for assessing past and future impacts of glacier changes on river runoff.

Definitions of surge-type glaciers vary, and their identification is often not straightforward because there is no specific threshold in terms of advance in a certain period of time above which a glacier can be categorized as being surge-type. Glacier surging is a quasiperiodic alternation between long periods (tens to hundreds of years) of slow flow, called quiescent phase or quiescence, and shorter periods of typically 10–1000 times faster flow, called surge phase, active phase, or surge (Dolgoushin and Osipova, 1982; Jiskoot, 2011; Benn and Evans, 2010). The cause of glacier surge has been identified as the relaxation of stresses in the body of the glacier (Dolgoushin and Osipova, 1975). A surge of polythermal glaciers may be initiated due to thickening and steepening in the upper reaches of the glacier, which increases the basal shear stress (Meier and Post, 1969; Murray et al., 2000). When the stress is more than the backward basal drag, the glacier may suddenly advance. For temperate hard-bed glaciers, during surge the high pressure because of the thickening of the glacier changes the basal water conduit system and stops the drainage of water through these channels. This results in increase of water in the glacier bed, which triggers the downward motion and allows the glacier to slide much faster than the normal speed (Kamb, 1987; Barrand and Murray, 2006; Björnsson, 1998). However, in the case of softbed glaciers, with increase in thickness of the reservoir, there is a deformation of the sediments in the glacier bed, which promotes the surge further (Murray et al., 2000). Some environmental factors such as extensive ice avalanches and large temperature fluctuations have also been identified as factors causing glacier surge in Karakoram Himalaya (Hewitt, 1969). The cluster of surge glaciers in specific climatic zones also suggests that environmental factors control surging (Post, 1969; Jiskoot et al., 1998.; Sevestre and Benn, 2015). However, it is also true that glacier surges are recurring phenomena caused by some dynamic instability of the glacier system, and they are only indirectly dependent on external factors (Murray et al, 2000; Mayer et al., 2011; Jiskoot et al., 1998; Qumcey and Luckman, 2014).

Surge-type glaciers in High Asia are especially common in the Karakoram (Hewitt, 2011; Bolch et al., 2012; Copland et al., 2011; Gardelle et al., 2013; Bhambri et al., 2013; Paul 2015) and the Pamir (Kotlyakov et al., 2008; Osipova and Khromova, 2010; Gardelle et al., 2013, Holzer et al. 2016). Sevestre and Benn (2015) reported a total of 106 and 820 surge-type glaciers in Karakoram and Pamir, respectively. Few surge-type glaciers have also been identified in the Tien Shan (Dolgoushin and Osipova, 1975; Narama et al., 2010; Kotlyakov et al., 2010; Osmonov et al., 2013; Häusler et al., 2016; Pieczonka et al., 2013). However, studies of individual surges in the region are rare, and only recently the volume and elevation changes of Northern Inylchek Glacier, the largest surge-type glacier in Tien Shan, were investigated (Pieczonka and Bolch, 2015; Shangguan et al., 2015).

The available knowledge based on some ground observations as well as satellite image-based studies of Tien Shan glaciers suggest that surge-type glaciers are mostly observed in Central Tien Shan, Ak-Shiirak and Northern Tien Shan (Bondarev, 1961; Bondarev and Zabirov, 1964; Dolgoushin and Osipova, 1975). Dolgoushin and Osipova (1975) mentioned 21 surge-type glaciers in the Tien Shan of which 8 were surging between 1956 and 1970 (Table 1). Sevestre and Benn (2015) identified 11 surge-type glaciers in Tien Shan based on a literature survey.

The reported advances were between 1 and 5 km during 1940–1960 for the glaciers in Central Tien Shan. For example, Mushketov (No. 3 in our numeration in Table 1 and Fig. 1) and Karagul (No. 23) glaciers surged by ∼4–5 km in 1956–1957 (Zabirov, 1961; Dyurgerov et al., 1995; Bondarev and Zabirov, 1964; Dolgoushin and Osipova, 1975; Osmonov, 1968). Kaindy Glacier (No. 2) surged in 1960, and its tongue advanced by ∼1.3 km compared to its extent in 1943, and both the length and thickness of ice at the tongue significantly increased (Osmonov, 1974). It was reported that the upper left tributary having two branches with northeast and southwest aspects, and its own accumulation zone, served as reservoir for this advance. The tongue of Kaindy Glacier showed high downwasting rates (-1.25 ± 0.49 m yr^-1) in the period ca. 1975–1999 despite thick debris cover (Pieczonka and Bolch, 2015), which indicates that during this period the glacier had already shifted to its quiescent phase after the surge. North Inylchek Glacier (No. 1) retreated continuously from 1943 till at least 1990 (Mavlyudov, 1995). Thereafter, it advanced rapidly ∼4 km during 1996–1997 (Mavlyudov, 1999; Häusler et al., 2016), experiencing a strong thickening at the tongue during the surge (Pieczonka and Bolch, 2015), and a lowering after 1999 (Shangguan et al., 2015; Häusler et al., 2016). It has been reported in Häusler et al. (2016) that 0.15 km³ water was displaced by this surge and triggered an outburst flood in the Lower Lake Merzbacher. A recent prominent surge (13% increase in glaciated area) was reported for Samoilowich Glacier (No. 4) (Osmonov et al., 2013) during 1990–2010.

TABLE 1

Location and characteristics of identified surge-type glaciers in Tien Shan.

(Continued)

(Continued)

FIGURE 1.

Study region and identified surge-type glaciers.

Two glaciers of the Ak-Shiirak massif (Bondarev, 1960, 1961, 1963; Bondarev and Zabirov, 1964, Fig. 1) also surged during 1940–1960. The inferred reason for the surge of ∼1 km in Northern Karasai Glacier (No. 18) in 1946–1949 and 1955–1956 was not climatic change, but the ice accumulated by the self-growing of the tributaries. Northern Bordu Glacier (No. 19) advanced by 260 m between 1943 and 1957, but overall it retreated by 240 m compared with the terminus in 1932 (Bondarev, 1961).

In Northern Tien Shan, frequent surges were reported for the right and left branches of the Shokalsky Glacier (No. 24) (Dolgoushin and Osipova, 1982). Makarevich (1952) observed a surge in the left branch in 1951, when the ice thickness of the front reached ∼60 m. Thereafter both the left and right branches retreated. A stage of surge was again initiated in the upper reaches of the left branch in 1968 which was evident from the increased ice flow velocities (10–20 times). An active advance started in the right branch in 1962, and the thickness in the middle reaches of the right branch of the glacier increased by ∼30–40 m until 1964. The tongue of this branch advanced by >300 m and more than 0.027 km³ ice was involved in this advance. After that the glacier started retreating, and the surface lowered by ∼50–55 m until 1967. The surge cycle was estimated to be ∼20–24 years for this glacier (Makarevich and Fedulov, 1974; Vilesov and Khonin, 1967; Cherkasov, 2002).

From the above examples, it is evident that there was no common surge period reported in literature. The strongest surges of several kilometers were observed for some of the largest glaciers of the Central Tien Shan. Most of the reported surges are based on some ground observations. Satellite images have also been used to report surges in some other glaciers of this region (Pieczonka et al., 2013; Osmonov et al., 2013; Pieczonka and Bolch, 2015; Häusler et al., 2016), but only North Inylchek Glacier (No. 1) has been studied in more detail recently. Although an extensive investigation of the surge behavior from that region was published in 1975 (Dolgoushin and Osipova, 1975), a more comprehensive and updated assessment is much needed.

The aim of this study is therefore to provide an extensive overview of the current knowledge about surge-type glaciers in the Tien Shan and investigate their characteristics in terms of geographical location (mountain range, latitude, and longitude), length, area, slope, aspect, minimum and maximum elevation, and their changes during surge in terms of length, area, and thickness/volume, with the help of available satellite imagery since the 1960s.

Study Region

The Tien Shan (approximately 40°-45°N, 67°-95°E) is one of the longest mountain ranges in Asia. It stretches from Uzbekistan and southwestern Kyrgyzstan ∼2500 km in an east-northeast direction, with the easternmost part located in the Xinjiang Uyghur Autonomous Region of western China (Fig. 1). In the south, Tien Shan links up with the Pamir Mountains and to the northeast meets the Dzhungarian Alatau. The Tien Shan is characterized by a dry and continental climate with the strong seasonal variations that mark most of Central Asia. Temperatures vary with altitude. Aizen et al. (1997) reported a lapse rate of 0.53 °C 100 m^-1. Most areas receive strong solar insolation throughout the year with relatively little annual precipitation or cloud cover. Meteorological data indicate that the western and northern peripheries of the Tien Shan have a milder and more temperate climate than the inner regions (Aizen et al., 1995; Solomina et al., 2004). Precipitation in the northwestern (outer) Tien Shan occurs mainly in spring and fall, whereas most of the precipitation takes place in summer in the southeast, central, and inner Tien Shan (Aizen et al., 1996, Aizen et al., 2001, Sorg et al. 2012). Annual precipitation rates decrease from north to south, from over 1000 mm yr^-1 in the Northern Tien Shan to less than 300 mm yr^-1 in the Aksai Basin (Koppes et al., 2008). Aizen et al. (1995) reported annual total precipitation increases with altitude up to crest-lines of mountain ranges in all regions of Tien Shan. Over the past four to five decades, the mean annual precipitation has increased in the outer and eastern ranges, but has probably decreased at higher altitudes in the inner ranges (Sorg et al., 2012). The mean annual air temperature (MAAT) at Tien Shan station (41.9°N, 78.2°E, at 3614 m a.s.l., the highest permanent station in the region) is about -7.7 °C, with January being the coldest month and July being the warmest month, having average temperatures of -21.8 °C and 4.3 °C (average of 1960–1997), respectively (Osmonov et al., 2013; Shangguan et al., 2015). MAAT has increased since the 1970s all over the Tien Shan resulting in a prolonged melting season (Aizen et al., 1997; Bolch, 2007; Sorg et al., 2012).

The Tien Shan has more than 6000 glaciers ranging from 2500 m a.s.l. to more than 7000 m a.s.l. (RGI version 4.0; Pfeffer et al., 2014). The total glacierized area is around 16,000 km² (Sorg et al., 2012). During the past four decades the overall glacier area and glacier volume of the Tien Shan have significantly reduced (Narama et al., 2010; Sorg et al., 2012; Unger-Shayesteh et al., 2013; Pieczonka and Bolch, 2015). Glacier mass loss in the Tien Shan is reported to be among the highest in the entire High Asia (Gardner et al., 2013; Farinotti et al., 2015) and is strongly related to temperature increase (Aizen et al., 2006; Bolch, 2007; Koppes et al., 2008; Solomina et al., 2004).

Data and Methods

Identification and Assessment of Surge-Type Glaciers

Initially, a list of 27 surge-type glaciers was compiled from existing literature (Table 1). We identified 12 additional surge-type glaciers based on visual image interpretation of morphological features such as presence of looped moraines, and bulging or strongly crevassed tongues (Table 2). In the next step, the terminus changes of all these potential surge-type glaciers have been investigated by using the available images and classified by using the surge index following Sevestre and Benn (2015). We slightly adjusted the index for our study region (Table 2). In order to compare the geometries of surge-type glaciers and the normal glaciers of Tien Shan, typical parameters like length, area, maximum and minimum elevation, elevation range, and slope have been studied. These values were taken from the RGI Vers. 4.0 (Pfeffer et al., 2014; Arendt et al., 2014). The mean (average of all values), median, quartile 1 (Q1), quartile 3 (Q3), minimum and maximum values of each of these parameters have been calculated for both surge-type and normal glaciers. The aspects of surge-type and normal glaciers have been compared using rose diagrams.

TABLE 2

Definitions for surge indexes used in this study.

Two glaciers that showed the most pronounced surge events, Samoilowich Glacier (No. 4) and North Inylchek Glacier (No. 1), were investigated in more detail in terms of their advance, thickness change during the surge, and the surge frequency. The volume changes for Samoilowich Glacier (No. 4) have been calculated considering a penetration of the Shuttle Radar Topography Mission (SRTM) C-band wave into the snow. The thickness changes for the surge-type glaciers within the Ak-Shiirak mountain range were also studied using the available Digital Terrain Models (DTMs).

Image Coregistration and Glacier Mapping

There were several Corona scenes available for the study area (Table A1). We have generated DTMs from the stereo pairs of 1964 covering the glaciers of Ak-Shiirak massif. All other Corona scenes were coregistered to the master scene (Landsat ETM+, Level 1T) using rubber sheeting (Watson, 1992) available in Erdas Imagine. The rubber sheeting was applied only to a small region surrounding the tongues of the glaciers, and we could achieve coregistration uncertainties of less than one pixel.

We used available inventories such as GLIMS ( http://www.glims.org) (Raup et al., 2007), RGI version 4.0 (Pfeffer et al., 2014; Arendt et al., 2014), and the outlines by Osmonov et al. (2013) and Pieczonka and Bolch (2015) to extract the polygons of the surge-type glaciers. The glacier outlines were visually checked, and their tongues were manually adjusted according to their extents observed in all images utilized in this study. From these outlines, we then proceeded to estimate the change in length. A glacier profile has been drawn approximately by visual interpretation along the central flow line of each glacier. The length of the profile clipped by the glacier outline corresponding to a particular image has been considered as the length of the glacier in that image. As the glacier outlines were extracted from different images with varying spatial resolutions, it was expected that they would have different levels of accuracies. The uncertainty of estimating length change has been calculated following Hall et al. (2003) as follows:

with R₁ and R₂ being the image resolution and RE the registration error.

The precision of glacier mapping is commonly within half a pixel with regard to the glacier perimeter (Paul et al., 2013; Bolch et al., 2010) for clean ice glaciers. However, manual adjustments are needed for debris-covered glaciers. The mapping uncertainties assumed for the different images used in this study are provided in Table A1 (Pieczonka and Bolch, 2015). The uncertainty of the glacier area before and after the surge event was calculated using a buffer around the glacier tongues with the buffer size being the estimated mapping uncertainty mentioned in Table A1 (Bolch et al. 2010). We edited only the changes at the tongues but did not change the upper glacier boundaries; these changes are very minor in comparison to the changes at the tongues of the surge-type glaciers. In addition, several images used were suitable to identify the tongues but not the upper boundaries due to adverse snow conditions.

DTM Generation and Volume Change Estimation

All Hexagon images, some of the Corona stereo pairs, and the Cartosat-1 stereo pair have been used for DTM generation and subsequent orthorectification of the images (Table A1). We made use of several DTMs that were already available, such as those from 1973, 1974, and 1976 KH9 Hexagon data covering large parts of Central and Inner Tien Shan (Pieczonka and Bolch, 2015).

We generated an additional DTM from 1980 based on KH9 Hexagon triplet stereo data (Table A1) for the Ak-Shiirak glaciers. Due to image acquisition on film and long time of film storage, the images had become distorted, and the reseau crosses used to reconstruct the image geometry at the time of acquisition are often shifted from their original locations (Surazakov and Aizen, 2010; Pieczonka et al., 2013; Holzer et al., 2015). Assuming no distortion for the reseau cross at the center of the image, the reference locations of all other crosses were determined and geometrically corrected (cf. Pieczonka et al., 2013). The two registered segments were finally mosaicked in Erdas Imagine, and the mosaicked product was used for subsequent DTM generation using the software ERDAS Photogrammetry 2014. The frame camera model was chosen with a fixed focal length of 0.305 m and flight height of 170 km (Surazakov and Aizen, 2010; Pieczonka et al., 2013). In total, 29 ground control points (GCPs) have been collected from Landsat 7 ETM+ imagery with SRTM3 as vertical reference. These points were located over distinct topographic terrain features like river crossings, mountain ridges, and so forth. In addition, 29 tie points were automatically generated. The resulting root mean square (RMS) error of triangulation was 0.982 pixels. The spatial resolution of the DTM was 30 m.

DTMs from all stereo Corona imagery covering the Ak-Shiirak massif were generated using the Remote Sensing Software Graz (RSG, version 7.46.15, developed by Joanneum Research Graz), using a fixed flying height of 200 km, focal length of 609.6 mm, and other exterior orientation (Table 3). Corona data is available as scanned images in tagged image file (TIF) format with indices from a to d, each having a size of ∼85 MB (14 microns scan resolution). The four parts of each image have been merged together in Adobe Photoshop to form the full scene. In the next step, a subset covering the area of interest is extracted from each scene, which has been used subsequently for DTM generation. A Corona adapted image-distortion model is available in RSG, which is essential to set up a geometric model. The number of GCPs collected manually for each subset varied between 27 and 45. Here, we have also used Landsat 7 ETM+ and SRTM as the source to collect horizontal and vertical reference data, respectively. The number of automatically generated tie points varied between 75 and 149. We have achieved RMS errors (for the control points) of less than ∼ pixels in both x and y direction. The spatial resolution of the generated DTMs was 25 m.

In order to investigate the recent surge of Samoilowich Glacier, we used a Cartosat-1 stereo scene (year 2006). The DTM was generated with Toutin's Model in Geomatica Ortho Engine 2014 using 17 GCPs and 28 tie points in both stereo images with a RMS_x of 1.84 pixels and a RMS_y of 1.92 pixels. The spatial resolution of the final DTM was 25 m.

TABLE 3

Corona camera exterior orientation parameters.

All DTMs were coregistered to the SRTM3 master DTM to remove tilts and shifts with respect to the master DTM and for subsequent and reliable glacier thickness change estimation. In a first step, the tilt with respect to the SRTM DTM was minimized by applying a spatial trend correction considering elevation differences over nonglacierized terrain with slopes less than 15° (Bolch et al., 2008; Pieczonka et al., 2013) only. Next, the DTMs were coregistered with respect to the SRTM DTM following Nuth and Kääb (2011). The final displacements between all DTMs and SRTM were less than or equal to one pixel.

The uncertainties of the DTM differences (E_Δh ) have been calculated following Gardelle et al. (2013) using Equations (2) to (5), considering elevation bands of 100 m.

Here, U _Δh is the uncertainty of measured elevation difference based on the elevation difference image, U _Δh (i) is the uncertainty calculated for i'th altitude band, n is the total number of altitude bands, σ _Δh (i) is the standard deviation of the mean elevation change of the nonglacierized terrain in i'th altitude band, N(i) is the total number of pixels in i'th altitude band, P _R is the pixel resolution in meter, and S is the distance (in meters) of spatial autocorrelation of the DTM difference maps.

The penetration (p) of radar waves into firn and snow needs to be considered while using SRTM C band data for DTM differencing and subsequent volume change calculation. We have set this value at 2.2 ± 1.2 m (Pieczonka and Bolch, 2015). The uncertainty in estimating glacier area change has been assumed to be 5% (Paul et al., 2013). Finally, uncertainty of the DTM differencing (E ) has been calculated using Equation (5).

Data gaps and mismatching resulting in erroneous elevation values occurred in snow-covered accumulation areas, areas under clouds, or in areas with cast shadows. The resultant outliers in the glacierized terrain were filtered following Pieczonka and Bolch (2015), assuming a nonlinear trend of the variance of elevation differences toward higher elevations by allowing maximum elevation change at the tongues of the glaciers and minimum change at the accumulation regions. All missing pixel values in the accumulation and ablation regions were filled by means of ordinary kriging.

Results

We have found 39 surge-type glaciers of which 12 were newly identified by this study. Out of the five tributary glacier surges, two were additionally identified in this study (Fig. 1, Table 1). Most of these glaciers concentrate in Central Tien Shan and Ak-Shiirak and could be identified as surge-type based on the presence of geomorphological/glaciological features and the rate of advance during the investigated time period (Table 4).

Comparison of Geometries of Surge-Type and Normal Glaciers

The mean values for the length, area, and elevation range of surge-type glaciers are higher and the mean slope is lower compared to the normal glaciers of the Tien Shan (Table 5). Also, the mean of maximum elevation is higher and that of minimum elevation is lower for surge-type glaciers than normal glaciers (Table 5). Results of a t-test yield values of p » 0 for all parameters except the aspects of surge-type and normal glaciers, which implies that there is a significant difference between the two data sets for all these parameters (see also the box-plots, Fig. 2). Such trends of differences in geometries of surge-type and normal glaciers have been observed all over the world (e.g., Sevestre and Benn, 2015). However, aspect values do not show any significant difference (p = 0.5) (Fig. 3). The rose diagrams displaying the frequencies of aspects show that most of the normal as well as surge-type glaciers have north (35.4% and 34.9%, respectively) or northwestern (20.6% and 25.6%, respectively) aspects (Fig. 3).

TABLE 4

Distribution of surge-type glaciers in Tien Shan.

TABLE 5

Geometric parameters of surge-type and normal glaciers of the Tien Shan.

FIGURE 2.

Box plots showing the geometries of surge-type and normal glaciers.

Samoilowich Glacier

Samoilowich Glacier (No. 4) was in a retreating phase from 1960 to 1992 and in an advancing phase from 1992 to 2006 (Figs. 4, and 5, parts a and b). The maximum length of the glacier was ∼8.9 km in 1960. After that it started retreating and reached its minimum length of ∼5.8 km in 1992. Thus it retreated by more than 3 km or one-third of its length. After this period of recession, the glacier advanced at an average of 22.5 m yr^-1 between 1992 and 2000. From 2000 to 2002 the most rapid advance, with an average of 0.8 ± 0.03 km yr^-1, was observed. The rate was 0.5 ± 0.03 km yr^-1 during 2002–2003 and 0.2 ± 0.04 km yr^-1 during 2003–2006. Altogether, the glacier advanced by 2.7 ± 0.1 km between 1992 and 2006 (Fig. 5, part b, and Appendix Fig. A1). Hence, the glacier had an active advancing phase of around 14 (1992–2006) years and experienced a surge cycle of around 50 years (ca. 1960–2006) (Fig. 4). The actual duration of a surge cycle may vary by a few years as the surge phase in any glacier normally starts earlier than any visible advance of the glacier terminus.

FIGURE 3.

Rose diagram showing the frequency of aspect in percentage for (a) normal glaciers, and (b) surge-type glaciers.

The DTM difference revealed that the glacier experienced a clear thickening of as much as 50 ± 5 m in its middle reaches between 1973 and 1999 (Fig. 6, part a). As the ice moved downstream during the surge phase, the glacier strongly thickened at the tongue and lowered in the upper part (Fig. 6, part b). On average, there was a surface lowering of 4 ± 5 m between 1973 and 1999. This loss was a result of a thinning in 70% of the area of the glacier, contributing to a volume loss of 0.056 ± 0.024 km³ and thickening in the remaining 30% of the area (reservoir area) contributing to a volume gain of 0.033 ± 0.01 km³.

Differencing the SRTM3 DTM and the Cartosat-1 DTM, which cover the surge phase, revealed a mean thickness change of-2.4 ± 5.6 m, and the volume gain was a little less than the volume loss (volume gain = 0.065 ± 0.01 km³, volume loss = 0.08 ± 0.02 km³). Hence, the glacier had probably a slightly negative mass balance during 1999 and 2006. It can be observed in Figure 6, part b, that there was a thickness loss in the middle reaches of the glacier during this period and thickness gain in the lower ablation area as a result of the glacier advance. A thickness increase of maximum 80 ± 5.6 m could be observed in the lower ablation area. The uppermost sections of the glacier remained unaffected. The length and area of the glacier clearly increased as the glacier surged (Fig. 6).

Glaciers of the Ak-Shiirak Massif

Glaciers in the Ak-Shiirak massif experienced several surges from ca. 1940s to 2014. Overall, we identified 11 surge-type glaciers, of which 5 have not been reported before (Tables 1, 4) but have been identified as surge-type based on geomorphological evidence such as bulging front, crevassed surface, and terminus advance.

FIGURE 4.

Relative length changes of the investigated surge-type glaciers.

FIGURE 5.

Samoilowich Glacier (a) retreat and (b) surge; North Inylchek Glacier (c) retreat and (d) surge; (e) Glacier No. 6 surge; and (f) Bogatyr Glacier (No. 8) surge.

FIGURE 6.

Thickness change of Samoilowich Glacier: (a) 1973–1999, and (b) 1999–2006.

The DTM difference of 1973 and 1964 shows that Sarytor-3 (No. 11) and Koyandy (No. 17) glaciers probably experienced a thickening in their lower ablation regions during this period. The terminus of South Karasai Glacier (No. 13) advanced from 1973 to 1993. Davidov Glacier (No. 12) advanced between 1964 and 1980 (Figs. 4, 7, A1), but its tongue was being removed artificially for the construction of a gold mine after 1999 (Jamieson et al., 2015). The tongues of both of these glaciers probably also thickened between 1973 and 1980 while they experienced a thickness lowering in their middle reaches. This indicates that ice has moved from the reservoir area to the receiver area, which resulted in an advance and corresponding thickness gain at the distal parts of the tongues of the glaciers (Fig. 7). However, the thickness increases (Fig. 7) as well as the glacier tongue advances (Fig. 4) are much lower for the surge-type glaciers of Ak-Shiirak than for North Inylchek and Samoilowich glaciers during 1964–2014. All these glaciers have been assigned a surge index 2 (Table 4).

Discussion

Among all the surge-type glaciers identified in this study, 88% are either confirmed, very probable, or possible surge-type glaciers (surge index 1 to 3). For the remaining 12% we could not identify any prominent glaciological or geomorphological feature related to a surge event and therefore state that they are less likely to be surge-type (surge index 4). All the glaciers having surge index 1 or 2 show either rapid, strong advances at different periods, typical indications of surge such as looped moraine, bulging tongue, a strongly crevassed tongue with a steep front, and/or clear lowering in the middle reaches and thickening of the lower reaches of the tongue. The morphological indications are also present in the advancing glaciers that made us, along with the available literature, confident that these glaciers should be classified as surge-type. Glaciers for which advances were very slow or not present during the period of our study, but in which we could identify one of the geomorphological/glaciological features, were classified as possible surge-type glaciers. We think that for these glaciers surges might have smaller active phases or are of small magnitude, and might have remained unnoticed using our approach because of uneven temporal gaps and/or insufficient spatial and temporal resolutions of the satellite images used in this study (Herreid and Truffer, 2015; Willis, 1995). It might also well be possible that further surge-type glaciers in the vast Tien Shan exist, which we missed. This could especially be true for smaller glaciers, for which surges are more difficult to identify as the smaller relative length changes may not be resolved spatially and/ or temporally by the available satellite images. The geomorphological/glaciological features are also more difficult to detect in smaller glaciers from visual interpretation using satellite images only (Hamilton and Dowdeswell, 1996).

FIGURE 7.

Thickness change along glacier profile based on the available bi-temporal Digital Terrain Models (DTMs), which include the surge period for each of the selected glaciers.

It is also possible that a glacier may gain thickness, but does not advance or only slightly advances, which is evident for some of the glaciers for which a thickness gain was observed in earlier studies (e.g., Nos. 26, 27, 28, 31, 34, 35, 36; Fig. 4), but no significant advance could be noticed in the present study. All these glaciers are located in Central Tien Shan, where most of the surges have been reported. Length change (advance or retreat) is not always the best indicator for identifying a surge-type glacier. For example, it can be observed in Figure 8 that the overall length at the central flowline of Bogatyr Glacier (No. 8), which surged during 1978–1994 (Table 1, Fig. 4), has remained almost constant even long after the surge, that is, until 2008. However, the thickness has reduced considerably by 2008 compared to the thickness in 1985. Similarly, Kaindy Glacier (No. 2) also experienced high down wasting rates after its surge in 1960, but its length has not changed much (Fig. 4). The surface dynamics often provide important information on surging, as a glacier moves 10–1000 times faster during its surge phase. Therefore, a combined study including length change, thickness change, and velocity is needed to draw a firm conclusion. We studied mainly length change for all the identified surge-type glaciers, and classified the surge-type glaciers based on that information in addition to morphometric indicators.

We also found some tributary surges, which advanced into the main trunk. Tributary surges have also been observed in other mountain ranges in High Asia such as the Karakoram (Hewitt, 2007; Paul, 2015; Belò et al., 2008). However, these surges are even more difficult to detect, since there is often no clear boundary between the tributary and the main glacier trunk. We only selected those for which we could identify clear evidence of advance.

The length and area of a glacier and glacier surface slope have been shown to have an effect on glacier surge mechanisms (Clarke, 1991; Jiskoot et al., 2000). The greater length and lesser surface slope increases the subglacial water pressure and results in enhanced basal sliding during surge. Longer glaciers with shallower slopes also have a higher probability of becoming thicker and storing more water (Lingle and Fatland, 2003), and they have a higher chance of eroding the bed of the glaciers (Jiskoot et al., 2000), which are also responsible for surging. It was found in our analysis that surge-type glaciers of Tien Shan are also longer and bigger in area, cover a higher range of elevation, and have shallower slopes than the normal glaciers of Tien Shan—like the surge-type glaciers in the other parts of the world (Sevestre and Benn, 2015; Grant et al., 2009; Jiskoot et al., 2000; Barrand and Murray, 2006).

Sixteen out of the thirty-nine surge-type glaciers are in Central Tien Shan, where the highest two peaks (reaching 7000 m a.s.l. or above) are located. A further hot spot of surges is the Ak-Shiirak massif (highest elevation above 5000 m a.s.l.), where 11 surge-type glaciers have been identified; however, the surges were much less prominent there during our period of observation (1964–2014).

FIGURE 8.

Bogatyr Glacier: (a) during the surge phase in 1985 (Institute of Geography, Almaty), and (b) after the surge in 2008 (V. Blagoveshchenskiy).

The active phases for some of the largest surge-type glaciers of Tien Shan, such as North Inylchek (No. 1), Mushketov (No. 3), North Karasai (No. 18), and Karagul (No. 23) Glaciers have been observed to be ∼1–2 years, and the advance varied between ∼1 and 5 km (Table 1, Fig. 1). Surge-type glaciers having longer active phases are smaller in size, e.g., 14 years for Samoilowich (No. 4) Glacier, 16 years for Bogatyr (No. 8) glacier, 14 years for Bezymyanny (No. 16) Glacier etc. (Table 1). Active phases of up to 15 years were also described for surge-type glaciers in Svalbard (Dowdeswell et al., 1991; Jiskoot, 2011), up to 12 years for Pamir (Kotlyakov et al., 2008), and >15 years for some of the Karakoram glaciers (Paul, 2015).

For the two glaciers with the most prominent surge we could identify a more or less full surge cycle. The length of the surge cycle is around 50 years for both glaciers. The right branch of Shokalsky Glacier (No. 24) has been reported to have surged in 1962–1964 (Vilesov and Khonin, 1967; Makarevich and Fedulov, 1974) and it again surged in 1994–2000 as observed in this study. Thus it can be inferred that the surge cycle for this glacier is ∼30 years. As we could not identify repeated surges for other glaciers, we conclude that the quiescent phase is relatively long for all surge-type glaciers in the Tien Shan, and surge is not so frequent as, for example, for some of the surge-type glaciers in the Karakoram (Paul, 2015; Quincey et al., 2015, Quincey and Luckman, 2014).

Available geodetic studies suggest that glaciers in the surging phase do not significantly change their overall volume much, can be both slightly negative and positive, and usually the signal is smaller than the uncertainty (Gardelle et al., 2013; Pieczonka and Bolch, 2015). This is in line with our findings for Samoilowich (No. 4) Glacier, which revealed that overall the glacier was in balance or showed a slightly negative mass budget during surge (1999–2006). During the surge event a significant amount of ice is translocated to lower elevations and is there, hence, prone to melt and even thick debris cover cannot prevent the ice from melting. This effect can be clearly identified for the debris-covered North Inylchek (No. 1) and Kaindy (No. 2) glaciers in Central Tien Shan (Pieczonka and Bolch, 2015, Shangguan et al. 2015).

Conclusions

We have used a variety of high- to medium-resolution satellite images of different dates from the 1960s to 2014 to record surge events in Tien Shan glaciers. We found 39 glaciers and 5 tributary glaciers that showed characteristics related to surges. The surge activity could be confirmed or is very probable for 22 of these 39 glaciers. Most of the surge-type glaciers are located in Central Tien Shan and Ak-Shiirak, which have the highest mountain peaks and colder and drier climate than the outer regions. The most pronounced absolute advance was observed for North Inylchek Glacier, which is the longest among all surge-type glaciers identified in this study. However, the most pronounced relative advance was observed for Samoilowich Glacier, which is much shorter. All the tributary glaciers identified as surge-type have continuously advanced throughout our study period. The rate of advance was more pronounced for some of the glaciers from Central and Northern Tien Shan, and more gradual for the glaciers of Ak-Shiirak. We could identify the surge cycle to be ∼30–50 years for three of the glaciers. For the other glaciers we could not identify repeated surge events. Hence, we conclude that the quiescent phase is comparatively long for the surge-type glaciers of Tien Shan and not as heterogeneous as, for example, in the Karakoram, where both short and long active phases as well as quiescent phases have been observed. No clear connection of climatic effects on glacier surge could be found. However, a decrease in maximum extents of the glaciers during surge and retreat of most of the surge-type glaciers during past one/two decades may indicate a reduction in surging tendency in this region.

Author Contributions

T. Bolch designed the study and guided the analysis. K. Mukherjee generated the 1980 Hexagon DTM for Ak-Shiirak, performed all analysis, and wrote the draft of the manuscript. T. Pieczonka generated the Cartosat-1 DEM and the Hexagon DEM for Central Tien Shan. F. Goerlich generated the Corona DTMs. S. Kutuzov, I. Shesterova, and A. Osmonov compiled the information from the Soviet literature. All authors contributed to the final form of the manuscript.