The global debate on glacier dynamics and variations reflects the growing importance of glaciers as water resources, as potential sources of natural hazards originating from glacial lake outbreaks, and their usefulness as indicators of climate change (Reference Solomina, Haeberli, Kull and WilesSolomina and others, 2008). The Little Ice Age glacier retreat has occurred in almost all mountain regions, with increasing rates of ice loss since the mid-1980s (Reference OerlemansOerlemans, 2005; Reference Kaser, Cogley, Dyurgerov, Meier and OhmuraKaser and others, 2006; Zemp and Reference Zemp, Haeberli and EamerHaeberli, 2007; Reference Zemp, Roer, Kääb, Hoelzle, Paul and HaeberliWGMS, 2008). However, a global assessment of the impact of climate change on glacier fluctuations is very difficult, due to unequal spatial and temporal coverage of records. Long-term glacier monitoring exists for the European Alps and Scandes, but only short-term and limited observations exist in the mountains of Asia (Reference Dyurgerov and MeierDyurgerov and Meier, 2005). The complexity of glacier response to climate change is underlined by the fact that several glaciers in the central Karakoram have been expanding since the late 1990s (Reference HewittHewitt, 2005). Possible reasons for this contrary behaviour are increased precipitation, a local trend of decreasing temperatures, particularly in summer (Reference Fowler and ArcherFowler and Archer, 2006), or the influence of thick debris coverage which protects the ice against melting (Reference HewittHewitt, 2005). Because >42% of the Himalayan glaciers are debris-covered, it is necessary to analyse their specific ablation conditions in the context of global warming (Reference Iwata, Aoki, Kadota, Seko and YamaguchiIwata and others, 2000). Most debris-covered glaciers show only small retreat rates with some notable exceptions, i.e. Gangotri Glacier in Garhwal, where a 1500 m retreat has been measured since 1935 (Reference Kumar, Dumka, Miral, Satyal and PantKumar and others, 2008), and Samadratapu Glacier in Himachal, which has receded ~756 m since 1963 (Reference Shukla, Gupta and AroraShukla and others, 2009). The debris-covered glaciers are characterized by down-wasting processes (Reference KargelKargel and others, 2005). Concomitant with glacier decrease, the debris-covered areas increase on most of the glaciers (Reference Iwata, Aoki, Kadota, Seko and YamaguchiIwata and others, 2000). Thick ablation-limiting debris covers are generally confined to the lower and, in the Karakoram, often the smaller fraction of the ablation zone. Thus they protect the lower penetration of glaciers but greatly limit the usefulness of terminus fluctuations and down-wasting as indicators of change in these areas. The larger and, for climate change, more significant factor is the thinly covered and dusty ice of middle to high ablation-zone areas, where debris has the opposite effect of enhancing ablation. This makes summer weather absolutely critical, since it determines both radiative heat flux (cloudiness) and how much dust and dirt accumulates on the ice (Reference HewittHewitt, 2005, Reference Hewitt2009).
Despite the obvious importance of the Himalayan glaciers as water sources for downstream lowlands of south Asia (Reference Viviroli, Weingartner and WiegandtViviroli and Weingartner, 2008; Reference Immerzeel, Droogers, de Jong and BierkensImmerzeel and others, 2009) and despite the growing number of local and regional glacier studies, Himalayan glacier response to climate change is poorly known (Reference Zemp, Haeberli and EamerZemp and Haeberli, 2007), mainly because long-term and continuous records of glacier fluctuations are almost completely lacking for large tracts of this mountain system (Reference ByersByers, 2007; Reference Kumar, Dumka, Miral, Satyal and PantKumar and others, 2008). Reference Mayewski and JeschkeMayewski and Jeschke (1979) presented an overview of 112 Himalayan and trans-Himalayan glaciers between 1812 and the 1960s. At present, 310 front-variation series with an average time length of 22 years distributed over the whole of central and south Asia are included by the World Glacier Monitoring Service (Reference Zemp, Roer, Kääb, Hoelzle, Paul and HaeberliWGMS, 2008).
For the Himalaya, the Nanga Parbat massif with its long and relatively continuous glaciological research history dating back to the mid-19th century (Reference KickKick, 1967, Reference Kick and Kick1996) represents an important exception and opens comparative research perspectives. During the German expedition to Nanga Parbat in 1934 the first detailed glacier inventory was carried out with an emphasis on Raikot Glacier (Reference Finsterwalder, Raechl, Misch and BechtoldFinsterwalder and others, 1935; Reference FinsterwalderFinsterwalder, 1938; Reference KickKick, 1994). Based on these research results glacier studies in the Nanga Parbat region were repeated 20 and 50 years later (Reference PillewizerPillewizer, 1956; Reference Mayewski and JeschkeMayewski and Jeschke, 1979; Reference GardnerGardner, 1986; Reference KickKick, 1994, Reference Kick and Kick1996).
The aim of our study is to document and analyse the fluctuations and dynamics of Raikot Glacier over the past 70 years. For this purpose, we use a multitemporal and multiscale approach, which is based on historical data, repeat photography and satellite imagery, such as Corona, the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), Landsat and QuickBird.
The Nanga Parbat region is located between 35°10′ N, 74°20′ E and 35°35′ N, 74°55′ E and the massif constitutes the northwestern limit of the high Himalaya. The maximum elevation difference is from 1030 m a.s.l. in the Indus Gorge up to the summit of Nanga Parbat (8126 m a.s.l.; Fig. 1). This extreme relief, combined with neotectonic activity, landslides and other mass movements, restricts the glaciated area to 18% (Reference Kick and KickKick, 1996). Nevertheless, the massif is one of the most heavily glaciated regions in the western Himalaya (Reference GardnerGardner, 1986).
The regional hydrology is transitional between the monsoonal conditions on the southern declivity of the Himalaya and the deserts of central Asia. A southwest–northeastoriented gradient of decreasing annual precipitation results from differing seasonal circulation systems and orographic conditions (Reference WeiersWeiers, 1995). The southern flank of the Nanga Parbat receives higher summer precipitation from monsoonal air masses. The northern and western slopes are more strongly influenced by winter precipitation from westerly disturbances. This horizontal differentiation is overlaid by steep vertical thermo-hygric gradients controlling the seasonal distribution of the snow/rainfall limit and amounts, according to elevation and orientation. In the river valleys, mean annual precipitation is characterized by balanced seasonal distribution, with 126 mm near Bunji (1372 m a.s.l.) and 516 mm in Astor (2394 m a.s.l.) (Reference Archer and FowlerArcher and Fowler, 2004). At 5500 m a.s.l. the amount of precipitation is estimated as ~2300 mm (Reference Winiger, Gumpert and YamoutWiniger and others, 2005).
Raikot Glacier, on Nanga Parbat’s north flank, is 15 km long and covers ~39 km2, the third largest glacier of the massif. The head of the catchment is delimited by Chongra Peak (6830 m a.s.l.), Raikot Peak (7070 m a.s.l.), Nanga Parbat (8126 m a.s.l.) and Ganalo Peak (6606 m a.s.l.). The ice tongue descends to 3180 m a.s.l. through large variations in the steepness of the glacier profile, with heavily crevassed zones and seracs alternating with low-gradient portions, such as the Raikot firn above 5200 m a.s.l. (Fig. 2). The glacier is fed mainly by frequent snow and ice avalanches from surrounding steep slopes and high-altitude snowfall (Reference Finsterwalder, Raechl, Misch and BechtoldFinsterwalder and others, 1935; Reference Gardner, Jones and ShroderGardner and Jones, 1993).
The ablation zone is ~10 km long and ~900 m wide, with a gradient of <10°. Steep slopes only exist in the upper part of the ablation zone. The lower Raikot icefall between 4000 and 4500 m a.s.l. has slope angles >30°. Unlike other large Nanga Parbat glaciers, an almost complete debris cover of varying thickness applies only on the western side and 1.5 km up-glacier of the terminus (Reference GardnerGardner, 1986; Reference Shroder, Bishop, Copland and SloanShroder and others, 2000). In this section, bare ice is only exposed in steep ice facets. Altogether, the proportion of debris-covered to clean ice is <25%, so Raikot Glacier can be classified as a partly debris-covered glacier.
Research History of Raikot Glacier
Unlike most Himalayan glaciers, Raikot and other glaciers of the Nanga Parbat massif have a long and relatively continuous research history (Reference Kick and KickKick, 1996). The German Himalayan expedition of 1934 carried out a terrestrial photogrammetric survey to make topographic maps with scales of 1 : 50 000 and 1 :100 000 (Reference FinsterwalderFinsterwalder, 1938), at that time a unique effort for a Himalayan high-mountain massif and a cartographic milestone. These maps represent the Nanga Parbat with the highest spatial accuracy, and provide the basis for all subsequent glaciological research (Reference Gardner, Jones and ShroderGardner and Jones, 1993). Glaciological processes were one of the major research interests of the 1934 expedition, and movement of Raikot Glacier was measured at several cross-sections. Based on these measurements, a slab-like movement divided among distinct ice units, called ‘blockschollen’ movement, was described for the first time. In addition, the glacier outlines of 1934 were precisely mapped to identify future glacier changes using repeat measurements along the same profiles. Stereophotographs were taken of selected glacier tongues with high spatial resolution, and the camera positions were marked by cairns and on maps (Reference FinsterwalderFinsterwalder, 1938). Both the inventory of glacier boundaries and the ice-velocity measurements required very high mapping accuracy, especially regarding the position of glacier termini (Reference Finsterwalder, Raechl, Misch and BechtoldFinsterwalder and others, 1935).
These investigations were repeated 20 years later, including velocity-rate measurements and mapping of the terminus (Reference Paffen, Pillewizer and SchneiderPaffen and others, 1956; Reference PillewizerPillewizer, 1956). By 1954, the glacier portal (defined as the portal of the proglacial stream) had retreated ~450 m, and the ice surface 1.5 km above the terminus had decreased significantly compared to 1934. By contrast, at >3500 m a.s.l. the glacier surface remained nearly the same and was even thicker in some parts than in 1934. Only Chongra Glacier, the eastern tributary of Raikot Glacier, showed significant down-wasting rates between 1934 and 1954. Down-wasting and retreat of the main Raikot Glacier were accompanied by a 30% increase of velocity along all cross-sections, leading Pillewizer to predict a readvance of the glacier terminus (Reference Paffen, Pillewizer and SchneiderPaffen and others, 1956; Reference PillewizerPillewizer, 1956). Observations in the 1960s suggested there had been a readvance, related to a steepened glacier margin (Reference KickKick, 1994).
In 1985 a resurvey of the Raikot Glacier terminus was carried out as part of the Snow and Ice Hydrology Project in the Upper Indus Basin (Reference GardnerGardner, 1986). The investigations confirmed the readvance of the glacier, and the glacier portal was found to be only 160 m up-glacier from its 1934 position. Average down-wasting, compared to the situation in the 1930s, was estimated at ~9 m (Reference GardnerGardner, 1986; IDRC/WAPDA, 1990).
Since the 1980s, investigations at Raikot Glacier have also included supraglacial debris depths and glacial sediment transport (IDRC/WAPDA, 1990; Reference Shroder, Bishop, Copland and SloanShroder and others, 2000), ablation rates in relation to debris cover and runoff (Reference GardnerGardner, 1986; Reference Gardner, Jones and ShroderGardner and Jones, 1993), reconstruction of Quaternary glaciations (Reference Kuhle and KickKuhle, 1996; Reference Owen, Scott and DerbyshireOwen and others, 2000; Reference Phillips, Sloan, Shroder, Sharma, Clarke and RendellPhillips and others, 2000; Reference Richards, Owen and RhodesRichards and others, 2000) and morphometric analyses of geomorphologic forms and processes (Reference Bishop, Shroder, Bonk and OlsenhollerBishop and others, 2002, Reference Bishop, Shroder and Colby2003). However, relations between glacier terminus changes and quantification of down-wasting intensities have not been studied.
Data and Methods
To detect and analyse the dynamics of Raikot Glacier, we used a multitemporal and multiscale approach, based on repeat terrestrial photography, historical data and remote-sensing data. Conventional satellite data ranging from regional to local scale, (e.g. ASTER, Landsat and QuickBird) allowed monitoring of glacier dynamics and variations since the 1980s. Corona images from the early US military-reconnaissance satellite widened the time span for detecting changes back to the 1960s. The original topographic maps were used for glacier changes back to 1934. Repeat terrestrial photography made visual comparison possible, to identify fluctuations of the glacier terminus, changes of the ratio between debris-covered and non-debris-covered glacier parts, ice cliffs and vegetation succession in the vicinity of the glacier portal.
Repeat terrestrial photography
Subject to the availability of historical photographs, the primary goal of re-photographic surveys is to replicate views from the same locations. Comparative interpretation between the original and repeat photography allows detailed assessments of change or persistence of landscape features (Reference NüsserNüsser, 2000, Reference Nüsser2001). In recent years, this has been used to detect glacier changes since the late 19th century in the European Alps (Reference Zängl and HambergerZängl and Hamberger, 2004), the United States (National Snow and Ice Data Center/World Data Center, glacier photograph collection, http://nsidc.org/data/glacier photo/repeat photography.html) and tropical mountains of Africa and South America (Reference HastenrathHastenrath, 2008). However, very few surveys have been undertaken in the mountains of high Asia, particularly with regard to glacier observations. Exceptions include a study by Reference ByersByers (2007) in the Khumbu Himal, Nepal, where he replicated a set of F. Müller’s glacier photographs from 1956, and Reference KickKick’s (1972, Reference Kick1989) work on Chogo Lungma, Karakoram.
In the Nanga Parbat region, the large number of photographs taken by members of the expeditions in 1934 (R. Finsterwalder) and 1937 (C. Troll) form a valuable basis for re-photographic surveys (Reference NüsserNüsser, 1998, Reference Nüsser2000; Reference SpohnerSpohner, 2004). For example, Gardner (in 1985) and Nüsser (in 1994 and 2006) repeated Finsterwalder’s photograph of the terminus of Raikot Glacier (viewpoint 1B, at 3303 m). In 1994 Nüsser repeated a historical image covering the upper ablation and accumulation zone (viewpoint 58A, at 4291 m). These multitemporal images enable qualitative detection of glacier variations, including changes in debris cover and ice cliffs, not indicated on the 1934 topographic map.
Satellite images and topographic maps
In order to improve the quantification of glacier changes geo-rectified data are required. A multi-data approach combining historical maps and data, as well as satellite images, provided near-decadal temporal resolution for detecting and analysing the changes of Raikot Glacier over the past 70 years. As mentioned above, the 1 : 50 000 map of 1934 (Reference FinsterwalderFinsterwalder, 1938) is characterized by a high accuracy of contour lines, rivers and glacier margins; however, bare ice is not distinguished from debris-covered ice. Additionally, we used two sketch maps of Raikot Glacier from 1954 (Reference PillewizerPillewizer, 1956) and 1985 (Reference GardnerGardner, 1986), plotted on the basis of the primary topographic map from 1934.
The earliest cloudless satellite scene is from the US Corona sensor, dated 1971, a panchromatic image that fills the gap between 1954 and 1985. In the 1970s,conventional remote-sensing data became available and, since the 1980s, Landsat TM (Thematic Mapper) data with a spatial resolution of 30 m × 30 m (Reference Hall, Bayr, Schöner, Bindschadler and ChienHall and others, 2003; Reference KääbKääb and others, 2005). High spatial resolution QuickBird data, starting from 2000, allow the detection of glacier outlines with high accuracy, and also small-scale ice cliffs and ponds on the glacier. The temporal resolution of satellite images for glacial analysis is generally restricted by clouds and seasonal snow cover. To minimize snow-covered areas, which complicate differentiation between glaciated and non-glaciated areas, the best images are those at the end of the ablation season, before the first snowfall event. Due to these restrictions, altogether nine satellite images – one Corona (1971), four Landsat TM (1992, 1998, 2001 and 2002), two ASTER (2001 and 2007) and two QuickBird (2003 and 2006) datasets – were chosen for detailed analysis (Table 1).
Geometric correction of satellite images and maps
Change-detection analysis requires spatial registration of selected multitemporal datasets. To ensure the required spatial accuracy, and to solve the problem of non-available ground control points (GCPs), all satellite images (as well as the topographic maps) were matched to one selected ‘base’ image: the georeferenced Landsat scene from 2001 (projected in UTM, WGS84, downloaded from the US Geological Survey). This allows visual comparison between the image and stream as well as ridgelines calculated by means of a 30 m × 30 m resampled digital elevation model (DEM) from the Shuttle Radar Topography Mission (SRTM-3, version 4, downloaded from http://srtm.csi.cgiar.org). This ensured we had sufficient spatial accuracy. To improve the coherence between the co-registered Landsat data and the base image, many well-distributed tie points are required. For this purpose, we used the automatic Förstner operator implemented in the programme ENVI 4.5. This operator can detect points in images which correspond to corners (defined as points where two or more lines intersect) as well as centre points of circular symmetric patterns, such as dots or circles (Reference FörstnerFörstner, 1986). The automatically determined tie points were controlled manually, and non-corresponding points were deleted, as well as points located on unstable terrain on and around the glacier. This resulted in the three Landsat scenes each being matched to the base image for >50 tie points. The scanned topographic map (projected on the Everest ellipsoid, Gauss–Krüger system) from 1934 was georeferenced with 12 well-distributed and distinctive GCPs selected on the Landsat image. Due to the small sections of both sketch maps from 1954 and 1985 (covering solely Raikot Glacier), both maps were registered with >10 tie points to the georeferenced topographic map from 1934. Subsets of the high-resolution ASTER and QuickBird images were orthorectified using >5 GCPs from the base image. The corresponding altitude data were derived from the DEM. Visual comparison of all datasets revealed the necessary high spatial accuracy.
Mapping of glacier boundaries
The Raikot Glacier margins and portal were delineated for the whole dataset. Common automatic glacier mapping approaches apply spectral differences between ice, characterized by high albedo values, and surrounding vegetation and rocks, identifiable by relatively low reflectivity (Reference PaulPaul, 2002). Due to the spectral homogeneity of the surfaces, such automatic approaches cannot be applied to debris-covered glacier tongues. Instead, semi-automatic approaches using morphometric parameters (e.g. slope, curvature and ridge) were developed to map scree-covered glacier termini (Reference Paul, Huggel and KääbPaul and others, 2004; Reference Bolch and KampBolch and Kamp, 2005; Reference Buchroithner and BolchBuchroithner and Bolch, 2007). For Raikot Glacier the limited availability of DEMs for each time period to be analysed restricts the use of a semi-automatic approach, and the glacier margin and portal were manually digitized on-screen on all satellite images and maps using ArcGIS 9.2. The manual delineation of glacier outlines achieves the highest accuracy, reaching an error rate of <2% (Reference Bolch, Buchroithner, Pieczonka and KunertBolch and others, 2008). To increase the temporal resolution of remote-sensing data, images of different spatial resolution were used to delineate glacier boundaries. If the spatial resolution of the satellite images varies between 5 and 30 m, only minor errors occur (Reference PaulPaul, 2003).
Measurements of glacier length
Changes in glacier length were quantified following the classical measure concept of Reference ForelForel (1895). For this purpose, a baseline 460 m long was plotted parallel to the glacier front. Distances between this baseline and the glacier terminus were derived within intervals of 10 m for each period using ArcGIS 9.2. The position of the glacier outline in 1934 was set as the benchmark. Forel’s concept considers different recession rates in various sections of an irregular-shaped glacier front. In addition, the up- and down-glacier shift of the portal was measured on those images and maps where it was clearly detectable. All later positions of the glacier portal were measured in order to continue the earlier approaches used in the historical studies. Only the y values of the digitized portals were determined, and the differences between their spatial positions were calculated to exclude corresponding lateral shifts due to the north-facing glacier tongue.
Detection of debris-covered areas
Down-wasting trends of glaciers are generally characterized by an increase in debris-covered areas on the glacier surface resulting from melting processes (Reference Iwata, Aoki, Kadota, Seko and YamaguchiIwata and others, 2000; Reference Bolch, Buchroithner, Pieczonka and KunertBolch and others, 2008). To assess this for Raikot Glacier, the ablation zone was classified into the categories of bare and scree-covered ice using an automatic segmentation approach. It was based on ratio mages derived from Landsat TM3 and TM5, with an accuracy of >97% (Reference Paul, Kääb, Maisch, Kellenberger and HaeberliPaul and others, 2002). To enhance the temporal resolution, we also applied this approach using two ASTER images and band 3 (15 m resolution) and band 4 (30 m resolution), with band 4 resampled to a pixel size of 15 m using bilinear interpolation. The application of ASTER ratio images revealed no significant differences to corresponding work based on Landsat TM (Reference KääbKääb and others, 2003). Thus, in total six satellite images (1992, 1998, two 2001, 2002 and 2007) were used to determine changing proportions of debris-covered and bare ice.
Results and Discussion
Glacier-change detection using repeat terrestrial photography
The Raikot icefall and firn, as well as the eastern tributary (Chongra Glacier), are shown in the background of the first pair of repeat photographs taken at 4291 m a.s.l. in the years 1934 and 1994 (Fig. 3). Shown in the foreground is the low-gradient and debris-covered ice confined to the left side above 3700 m a.s.l. In the right foreground the steep lateral moraines and the debris-covered tongue of Ganalo Glacier are visible. The steep slopes are affected by frequent snow and ice avalanches which are important for glacier feeding. In both years the patchy snow cover redistributed by avalanches shows similar features. The size and number of non-glaciated areas along steep rock faces have not significantly increased in the 60 years between photographs. The terminus of the tributary, Chongra Glacier, shows some retreat, but no significant decrease of the glacier surface can be detected. In conclusion, variations >3700 m a.s.l. prove to be minor between 1934 and 1994.
The second sequence of repeat photographs, from the locality of Fairy Meadows (3300 m a.s.l.), involves four images from 1934, 1985, 1994 and 2006. They show the debris-covered glacier terminus ~3200 m a.s.l. (Fig. 4). At this relatively low elevation the lateral moraines are covered with montane conifer forests of Pinus wallichiana and Picea smithiana. The inner slopes of the moraines are occupied by Hedysarum falconeri (Reference Dickoré and NüsserDickoré and Nüsser, 2000). Throughout this period the fence shown at the bottom right corner has separated cultivated land from the pasture of the summer settlement. Overall, visual comparison indicates relatively small rates of recession and surface changes over the last seven decades.
In 1934, the farthest extent of the irregular-shaped tongue was on the western flank and the glacier portal was on the eastern side. Ice cliffs of varying dimensions were exposed on the debris-covered tongue. Furthermore, the distinctive band of clean ice is visible in the upper part of the image. The second image, for 1985, shows a marked retreat (Reference GardnerGardner, 1986). At that time, the farthest extent of a glacier lobe was on the eastern side and the glacier portal and subglacial channel had shifted to the western side. As in 1934, the glacier tongue is debris-covered and ice is only exposed in cliffs that have decreased in number. In the proglacial area a substantial colonization by vegetation can be observed, mainly coniferous trees.
In the 1994 image, no significant retreat of the glacier margin can be detected in comparison with 1985. However, the thickness of the glacier tongue varies in different places. Compared to 1985, the debris-covered terminus has steepened markedly on the front and western side and new large ice cliffs have appeared. According to Reference Shroder, Bishop, Copland and SloanShroder and others (2000) this trend continued at least until 1996. On the 1994 image one can detect a more regular terminal lobe. In 2006 the portal had retreated by backward incision into the glacier, but the proglacial river is in nearly the same place as in 1985. One can detect a significant expansion of the montane forest on the lateral moraines of the receding Raikot Glacier. Along the proglacial stream floor, a succession of hygrophilous Salix sericocarpa and Myricaria germanica bushes is detectable (Reference Dickoré and NüsserDickoré and Nüsser, 2000). Given that some well-defined large boulders remain in the same position in 2006 as in 1934, one can assume a high stability of the proglacial area and lateral moraines. This casts doubt on the catastrophic break-out flood from four new portals in 1993 or 1994 suggested by Reference Shroder, Bishop, Copland and SloanShroder and others (2000). Our own field visits in 1993, 1994 and 1995 did not verify such an event.
The comparative analyses of historical maps, data and satellite images combined with geographical information systems allow us to quantify the changes of the glacier terminus (Figs 5 and 6; Table 2). Reference FinsterwalderFinsterwalder’s (1938) topographic map shows the irregular-shaped glacier tongue in 1934. The farthest extent of ice was located at the western side at 3155 m a.s.l., and this ice lobe reached ~125 m further north than the portal, located at 3176 m a.s.l. On the sketch map showing the glacier outline in 1954 (Reference PillewizerPillewizer, 1956) the glacier terminus has retreated significantly and altered in shape. The largest ice lobe at the glacier front, ~300 m long, has migrated from west to east. According to Reference PillewizerPillewizer (1956), the glacier portal was relocated to 3210 m a.s.l. in 1954. Its total retreat between 1934 and 1954 was 450 m (23 m a−1), a value which coincides with our digital distance measurements. During this time the average rate of retreat for the whole glacier front is calculated as 385 m (19 m a−1).
By contrast, the glacier portal advanced ~300 m between 1954 and 1985. Reference GardnerGardner’s (1986) estimate of 200 m portal advance after 1954 is based on a miscalculation. The mean advancing rate of the glacier front amounted to ~228 m (7 m a−1) for this time period. Based on the 1971 Corona image, a continuous advance of the glacier terminus can be assumed. However, due to low-contrast differences in the proglacial area on this image, only the ice cliff at the glacier front is clearly detectable. Significant down-glacier movement of the ice cliff on the western side also suggests an advance on the eastern side between the positions of 1954 and 1985. The location of the ice cliff indicates an advance of the glacier terminus of >113 m between 1954 and 1971 and a further down-glacier movement of 105 m between 1971 and 1985. Therefore, the possibility of a single surge event, as described for various tributaries of Panmah Glacier, Karakoram (Reference HewittHewitt, 2007), can be excluded for Raikot Glacier. The advancing rate between 1954 and 1985 corresponds with the observed reaction of some other Himalayan glaciers for this period (Reference Zemp, Roer, Kääb, Hoelzle, Paul and HaeberliWGMS, 2008).
The pan-sharpened QuickBird images from 2003 and 2006 provide very detailed pictures of the glacier terminus. In contrast to the irregular-shaped glacier terminus and one-sided ice lobes in 1934, 1954 and 1985, one can detect a regular-shaped lobe in 2003. The portal was only 5 m up-glacier from its position in 1985, and retreat of the whole glacier front amounted to only 30 m. Closer inspection of the front variation indicates remarkable fluctuations between 1985 and 2003. Between 1985 and 1992 the glacier front retreated ~84 m (−12 m a−1),but by 2001 had readvanced ~60 m (7 m a−1). The low spatial resolution of the 1992 Landsat image results in a reduced spatial accuracy of the mapped glacier margin. Nevertheless, compared to 1985 a glacier retreat can be clearly observed.
In 2006 a break-up of the terminus into three lobes can be identified. Despite this, the eastern side is nearly in the same position as in 2003, whereas a significant retreat is detectable on the western side. In total, the last 3 years of the analyses show a 15 m decrease of the whole terminus. The portal receded 25 m up-glacier and shifted slightly to the eastern side (Fig. 8).
These small fluctuations of the glacier terminus between the 1980s and 2006 contrast with the dominant and large retreats in other parts of the Himalaya since the 1980s (Reference Zemp, Roer, Kääb, Hoelzle, Paul and HaeberliWGMS, 2008). However, Raikot Glacier shows a similar response to the Karakoram ‘anomaly’ described by Reference HewittHewitt (2005) for the neighbouring mountain system. Baltoro Glacier in the cenrtal Karakoram, however, has oscillated back and forth a few hundred metres between 1913 and 2004. The total terminus recession amounts to only 65 m, which reflects a quite stable snout position (Reference Smiraglia, Mayer, Mihalcea, Diolaiuti, Belò, Vassena, Baudo, Tartari and VuillermozSmiraglia and others, 2007). Thus, the fluctuations of Baltoro Glacier are very similar to those of Raikot Glacier.
Variation of debris-covered areas
Comparing 2003 and 2006 shows the details of changes on the glacier tongue (Fig. 8). A large number of small water ponds up to 1200 m2 in area existed on the debris-covered terminus in 2003, but had partly disappeared 3 years later. The number of ice cliffs and thin debris-covered areas decreased between 2003 and 2006. A corresponding increase of debris-covered areas can also be detected in the classified Landsat and ASTER images over the last two decades.
On the western side of Raikot Glacier, one remarkable feature is a distinct triangular-shaped debris accumulation, visible in all images since 1992 (Figs 7 and 9). This clear feature, covering an area of ~0.108 km2, was not shown in the sketch of debris thickness measured in 1985 (Reference Mattson and GardnerMattson and Gardner, 1989; IDRC/WAPDA, 1990). However, this marked surface feature can be dated back to the late 1980s or early 1990s. Landsat data from July 1990 suggest the causative event in a landslide deposit forming a first debris accretion at the junction of Raikot and Ganalo Glaciers. This striking debris accumulation has steadily moved ~1.8 km between 1992 and 2007 (Fig. 9).
To minimize the effect of different seasonal snow-cover distributions, shadow effects and cloud cover, the spatio-temporal analyses of debris-covered areas have been restricted to the lower part of the ablation zone, <3900 m a.s.l., and up to 7.6 km above the terminus. This section has an area of ~5.1 km2 (13% of the total glacier), and the surface has been classified into debris-covered and bare ice. In this section, the area of clean ice was 2.04 km2 (39%) in 1992 and had decreased to 1.81 km2 (35%) in 2007, a decrease of 0.56% in relation to the whole glacierized area (39 km2). The slight decrease of bare ice (Fig. 10) might indicate only a slight down-wasting trend for Raikot Glacier. According to Reference GardnerGardner (1986), the glacier surface was lowered ~9 m between 1934 and 1985. These results contrast to the observed more-pronounced down-wasting rates of debris-covered glaciers elsewhere in the Himalaya. For example, on Khumbu Glacier a significant increase of debris-covered areas, ~2.5% (0.06% a−1), has been measured between 1962 and 2005 (Reference Bolch, Buchroithner, Pieczonka and KunertBolch and others, 2008). However, the observed results are quite similar to Karakoram glaciers in the middle decades of the 20th century. There are conflicting reports on this for the Karakoram, but most of the glaciers draining the highest watersheds have shown thickening in the debris-covered areas since 1999 (Reference HewittHewitt, 2005, Reference Hewitt2009).
Conclusion and Outlook
Similarly to most debris-covered glaciers in the northwest Himalaya and in the nearby Karakoram, Hindu Kush and Kun Lun ranges, Raikot Glacier shows only minor retreating rates since the 1980s. Glacier fluctuations over the past 70 years are characterized by retreat between the 1930s and 1950s, a marked advance between the 1950s and 1980s, and a relatively stable situation after 1992. It remains to be seen to what degree these fluctuations can be interpreted as a response to climate change, such as through increased precipitation and lower summer temperatures as described by Reference Archer and FowlerArcher and Fowler (2004), or through increased thickness of debris cover on ice, which in turn protect the glacier against ablation (Reference HewittHewitt, 2005). Due to the historical database, from the 1930s, 1950s and 1980s, our investigations need to be expanded to the whole Nanga Parbat massif in order to gain an adequate understanding of glacier response to global climate change and to reduce existing data gaps in the western Himalaya. In this context, we intend to take velocity measurements and compare them to those carried out in 1934 (Reference FinsterwalderFinsterwalder, 1938), in the 1950s (Reference PillewizerPillewizer, 1956; Reference KickKick, 1994), 1980s (Reference Gardner, Jones and ShroderGardner and Jones, 1993; Reference KickKick, 1994) and 1990s (Reference Shroder, Bishop, Copland and SloanShroder and others, 2000) and to glacier down-wasting rates on selected glaciers of the Nanga Parbat massif.
This paper is a revised version of an oral presentation at the European Geosciences Union in 2008. Field surveys in the 1990s were generously supported by the German Research Foundation (Deutsche Forschungsgemeinschaft) and in 2006 by the South Asia Institute, University of Heidelberg. The authors are indebted to U. Kamp (University of Montana) for access to ASTER satellite data. We gratefully acknowledge the substantial and fruitful comments of J. Gardner, K. Hewitt and an anonymous reviewer. We also thank P. Roden for proofreading.