Climatic and topographic influences on glacier distribution in the Bhutan Himalaya

Introduction Conclusions References


Introduction
Glaciers are key indicators of climate change (e.g., Kaser et al., 2006;Bolch et al., 2012), and glacial meltwater is an important water resource for human consumption and agriculture in arid regions (e.g., Immerzeel et al., 2010;Kaser et al., 2010).The contribution of glacier mass loss to global sea-level rise is also a current societal concern (e.g., Raper and Braithwaite, 2006;Radić and Hock, 2011).In the high mountains of Asia, which contain the largest number of glaciers outside the polar regions, spatially Introduction

Conclusions References
Tables Figures

Back Close
Full heterogeneous shrinkage of glaciers has been identified by in situ measurements (Yao et al., 2012), remote sensing approaches (Kääb et al., 2012;Gardelle et al., 2013), and model simulations (Fujita and Nuimura, 2011).A few studies provided precise hypsometries (area-altitude distribution) of glaciers in the Himalayas (Bolch et al., 2012;Basnett et al., 2013), which are important basses for projecting the influence of glacier changes on meltwater discharge.In particular, glaciers in humid monsoonal climates, such as those in the Bhutan Himalaya, are expected to be highly sensitive to changing climate (Fujita, 2008;Fujita and Nuimura, 2011).However, the elevational and spatial distribution of glaciers and their relationships with climate and topography in the Bhutan Himalaya are largely unknown, as neither observational evidence nor the present locations of glaciers are available for this area, with the exception of a preliminary report using remote sensing data (Karma et al., 2003).
In terms of climatic influences on glaciers, it has been revealed that the relationship between climatic regimes and equilibrium line altitudes (ELAs) can be expressed by a polynomial equation consisting of summer mean temperature and annual precipitation (Ohmura et al., 1992).Larger snow accumulations are related to lowered glacier ELAs, and thus the presence of glaciers in warmer environments, and vice versa.Although ELAs are obtainable only by direct measurements of mass balance or other estimations, such as the use of accumulation-area ratio methods, few data are available for glaciers in the Bhutan Himalaya.For such region, the median elevation, which divides the glacier area in half, as a proxy for the ELA for unmeasured glaciers, can be adopted as proposed by Braithwaite and Raper (2009).
One of the most concerning features of Himalayan glaciers is the extent of debris cover, which potentially prevents ice melting if the debris layer is sufficiently thick (Mattson et al., 1993), and which stabilizes their termini surrounded by their equivocal boundary (Scherler et al., 2011a).In terms of glacier hypsometry and median elevation, a massive debris mantle can decrease the melting rate of ice, and possibly enhance the expansion of debris-covered ablation areas, thus lowering the median elevation.Topographic influences on the extent of glaciers are related to debris supply, Introduction

Conclusions References
Tables Figures

Back Close
Full which appears to correlate highly with slope gradients above glaciers in the Himalaya (Scherler et al., 2011b) or with the extent of southwest-facing slopes above glaciers in the Bhutan Himalaya (Nagai et al., 2013).Avalanche-fed accumulation is also another influence of topography on the extent of glaciers (Hewitt, 2011).Redistribution of snow accumulation by avalanches can possibly increase the size of the ablation area at lower elevations through increased ice flux, thus resulting in a lowering of the median elevation.However, influences of these two factors on elevation distribution of glacier are not well understood.Some previous studies have reported on the spatial distribution of precipitation (Eguchi, 1991), changes in terminus locations of glaciers (Karma et al., 2003), topographic asymmetries affecting dynamic regimes (Kääb, 2005), and the formation of debris-covered areas (Nagai et al., 2013) in relation to glaciers in the Bhutan Himalaya.However, it is not yet known which factors are different between the debris-free and debris-covered nature of glacial termini, and which factors control glacier extent, particularly in terms of elevation.In this study, we aim to understand the influences of climate and topography on the elevational distribution of glaciers in the Bhutan Himalaya, a region in which drastic precipitation gradients occur over a narrow latitudinal range (< 1 • ).We mainly analysed the median elevations of glaciers (as an ELA proxy) in relation to precipitation (a climatic variable) and the slopes situated above a glacier (a topographic variable).

Study site and datasets
We focused on a latitude-longitude domain in the Bhutan Himalaya ( 27• 36 -28  2001), Global Land Ice Measurements from Space (Raup and Khalsa, 2007), and the Randolph Consortium (Arendt et al., 2012;Pfeffer et al., 2014).It has been pointed out, however, that seasonal snow cover and debris cover resulted in misinterpretation of glacier outlines.In addition, the outlines including mountain terrains and rock glaciers, omission of many glaciers, and highly generalized glaciers were reported.Wrong geolocation such as systematic shift causes serious errors on glacier distribution (Pfeffer et al., 2014).In the Bhutan Himalaya, for instance, it would be difficult for small glaciers and heavily debris-covered glaciers distributed in the southern part of the country to be correctly delineated because of the complex topography and the fact that many images are affected by snow cover, which make the identification of glacier boundaries equivocal.Therefore, we generated a new glacier inventory based on manual delineation of glaciers, using advanced satellite imagery with high spatial resolution.
In addition to delineation of the glacier itself, debris-covered areas and potential material-supply (PMS) slopes were separately delineated so as to facilitate the analysis of topographical influences on glacier distribution.We defined debris-covered areas as zones where ice cannot be seen on account of debris mantles in glacier ablation zones, but which does not include dirty glacier ice.A PMS slope is defined as a mountain surface from which any supply of snow, ice, or debris to the glacier can be expected.The PMS slopes are equivalent to the "potential debris-supply slopes" of Nagai et al. (2013), which were as potential sources of debris mantles and contributions to the formation of debris-covered areas on glaciers.However, in this study, we renamed the index so as to also take into account the accumulation of snow and ice by avalanches.
For glacier delineation, we used 58 scenes of the Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) on board the Advanced Land Observing Satellite (ALOS).They were acquired repeatedly between 2007 and 2011 within the ALOS orbit paths of 154-158 (latitudinal location), and the frames of 3030-3045 (longitudinal location) (Table S3).The spatial resolution of the images is 2.5 m, and they are orthorectified with a PRISM-derived digital surface model using Ortho-image Generation Software for ALOS PRISM (Tadono et al., 2012).If multiple images were available for Introduction

Conclusions References
Tables Figures

Back Close
Full an area, we selected the image with the least snow cover for glacier identification.If no preferred PRISM image was available in a given area, we used 10 m resolution composite colour images from the Advanced Visible and Near Infrared Radiometer type 2 (AVNIR2) on board ALOS (Table S4).These images.which were acquired in the same period and location of the PRISM images, are orthorectified by a digital elevation model derived from the Shuttle Radar Topography Mission (SRTM DEM).
To generate 20 m contour lines, which were used to separate glacier connections at accumulation zones and to delineate PMS slopes, we utilized the second version of the Global Digital Elevation Model generated from the Advanced Spaceborne Thermal Emission and Reflection radiometer (ASTER GDEM2) provided by the National Aeronautics and Space Administration, United States, and the Ministry of Economy, Trade, and Industry, Japan.The images were synthesized from multiple ASTER images, with grid cell resolutions of 1 arc second (∼ 30 m) (Tachikawa et al., 2011).Hayakawa et al. (2008) reported that accuracy of ASTER GDEM2 was better than that of SRTM DEM in steep terrain because of its higher resolution, fewer missing data, and better topographic representation, whereas the ASTER GDEM2 had large uncertainty on steep terrain, water or snow surfaces because of the lack of clear features (Toutin, 2002;Fujita et al., 2008).Therefore we utilized the GDEM2 to analyse distributions of elevations and aspects of surrounding slopes rather than snow-covered glacier surfaces.The outline polygons of features were overlain on bird's-eye view images in Google Earth ™ to check delineation quality.
Tropical Rainfall Measuring Mission (TRMM) data were utilized to analyse spatial distributions of precipitation.We calculated annual precipitation between 1998 and 2012 from the TRMM 3B43 monthly product, at a spatial resolution of 0.25 • (∼ 25 km) (Huffman et al., 2007).In terms of accuracy in high mountains, monthly mean precipitation of the TRMM data showed better consistency with an in-situ measurement in the Nepal Himalaya than the other precipitation products (Yamamoto et al., 2011).Introduction

Conclusions References
Tables Figures

Back Close
Full

Definition used for delineation
Our method for generating the glacier inventory in the Bhutan Himalaya follows the visual interpretation and manual delineation method described by Nagai et al. (2013), in which the standard definition of glacier outlines is based on Raup and Khalsa (2007) and Rastner et al. (2012).PRISM images enable detailed identifications of (1) glaciers smaller than 100 m in length, (2) thermokarst features in debris-covered areas, and (3) surface features of unstable steep slopes around glaciers, from which material supply is potentially expected.The GDEM2-derived contour lines were layered on the PRISM images or the AVNIR2 images, and the Google Earth ™ images of the same location were displayed on another panel.Visual interpretations and manual delineations were then conducted for glacier bodies, debris-covered areas, and PMS slopes, based on the procedures described below.The manual delineation of mountain glaciers with high-resolution satellite images causes high variability of outlines among different analysts, especially in debris-covered termini (Paul et al., 2013).Delineation work in this study was performed by one interpreter (the first author) and the outlines were modified three times by himself after a quality check by another interpreter.First, each glacier terminus was identified, and contiguous ice bodies in all tributaries were delineated.Isolated ice bodies were defined as belonging to a downstream host glacier if their locations were proximal to the host glacier, and if they were expected to share ice transport processes with the host glacier.Smooth ice surface, crevasses around ice falls, and bergschrunds support the identification of glacier bodies, and allow us to exclude neighbouring snow-covered bedrock.Snow-covered accumulation zones continuously connecting some glaciers were separated along ridge lines, as interpreted from GDEM2-derived contours.
Because surface reflectance of debris-covered areas is similar to that of surrounding bedrock and/or moraines, we did not adopt automatic delineation methods.Instead, we visually interpreted surface features of debris-covered areas that were characterized by thermokarst features such as rugged small-relief supra-glacial ponds and ice cliffs Introduction

Conclusions References
Tables Figures

Back Close
Full  (Iwata et al., 2000).The upper boundary of debris cover often varied in several PRISM images taken on different dates.In such cases, the outline of the maximum area was adopted, taking temporary snow cover into account.Continuous slopes between glacier margins and mountain ridges were delineated as PMS slopes if they tilted towards the glacier.Avalanche-or rockfall-induced scars and traces were taken as evidence of slope instability.Slopes intercepted by hills or lateral moraines, which would prevent avalanche or rockfall supply to a glacier, were excluded from the PMS slopes.Slopes of lateral and terminal moraines were included as long as they were inclined towards the glacier surface.These detailed surface features can be identified on PRISM images, although they are not fully expressed by contour lines.

Results
We delineated a total of 1579 glaciers in the Bhutan Himalaya (Table S1), distributed along the main E-W trending Himalayan range and along smaller branches extending in various directions (Fig. 1).The glaciers consist of 213 debris-covered glaciers and 1366 debris-free glaciers.While the number of debris-covered glaciers is smaller than that of debris-free glaciers, the total area of the former (1037.3km 2 ) is much larger than that of the latter (570.2 km 2 ): on average, debris-covered glaciers are ten times larger than debris-free glaciers (4.87 km 2 vs. 0.42 km 2 ) (Table 1).
The total area of PMS slopes surrounding debris-covered glaciers (908.8 km 2 ) is 4.5 times greater than that surrounding debris-free glaciers (194.1 km 2 ), and the mean area of the PMS slopes for debris-covered glaciers is 30.5 times greater than that for debris-free glaciers (Table 1).We defined a PMS slope ratio parameter (PMS slope area divided by glacier area) to examine differences in the areas of PMS slopes as compared with the areas of debris-covered and debris-free glaciers.The mean PMS slope ratio for debris-covered glaciers (1.05) is two times greater than that for debrisfree glaciers (0.47) (Table 1).The PMS slope area is highly correlated with debriscovered area, and PMS slopes are therefore considered to be the main sources of

TCD Introduction Conclusions References
Tables Figures

Back Close
Full debris on glacier debris mantles in the Bhutan Himalaya (Nagai et al., 2013).The greater area of PMS slopes and the larger PMS ratios of debris-covered glaciers also suggest that PMS slopes are the main constraint on the extent to which a glacier is covered by debris.Hypsometries (area-altitude distributions) of debris-covered and debris-free glaciers show that the maximum glacier area occurs at 5400 m a.s.l. for both glacier types (Fig. 2).The debris-covered glaciers are distributed over broader elevational ranges (4000-7500 m a.s.l.) than are debris-free glaciers (4800-6800 m a.s.l.) Mean median elevations of individual glaciers are 5537 ± 355 m a.s.l. for debris-covered glaciers and 5495 ± 285 m a.s.l. for debris-free glaciers.The difference between the median elevations of the two glacier types is not statistically significant, whereas the difference between their mean areas is statistically significant (4.87 km 2 for debris-covered glaciers and 0.42 km 2 for debris-free glaciers) (Table 1; Fig. 2).
Paul et al. ( 2013) demonstrated that debris-cover could lead to large interpretation differences (standard deviation of 6 % in area) by multi interpreters even if high resolution imageries such as QuickBird were used.We project an uncertainty of median elevation due to that of the area delineation by assuming that the area uncertainty exists solely in the lower half of debris-covered glaciers, at which the glacier boundary is often so unclear that cause misinterpretation.Considering the hypsometry and area statistic (Fig. 2 and Table 1), the area uncertainty in lower glaciers (±62 km 2 , 6 % of total area of debris-covered glacier (1037 km 2 ) alters the median elevation by ±40 m, which results in area change of upper glaciers by 31 km 2 .This uncertainty of median elevation is sufficiently small comparing with that of individual glaciers (±355 m).
The numbers and mean areas of debris-covered and debris-free glaciers are summarized by aspect, which is averaged over each individual glacier using ASTER GDEM2 images (Fig. 3).Larger numbers of debris-free glaciers face to the north and northwest, whereas directional preferences for debris-covered glaciers are uncertain (Fig. 3a).On the other hand, south-and north-facing debris-covered glaciers tend to be large, presumably because their sources are along the main east-west trending

Conclusions References
Tables Figures

Back Close
Full Himalayan ridge, whereas such a relationship is not shown for debris-free glaciers (Fig. 3b).These aspect dependencies suggest that solar radiation controls the development of debris-free glaciers, and that the scale of a mountain ridge controls the expansion of a debris-covered glacier.Histograms of total area, number, PMS slope area, and PMS slope ratio are summarized by glacier size (Fig. 4).Glaciers with sizes of 1.0-5.0km 2 comprise the largest total glacier area in this region (debris-covered and debris-free glaciers combined; Fig. 4a).Area maxima in this size class are observed in several regions worldwide, such as in the Svartisen region, Norway (Paul and Andreassen, 2009), and the Tien Shan Mountains, Central Asia (Narama et al., 2010), whereas larger glaciers occupy larger areas in Greenland (Rastner et al., 2012).The largest area of debris-covered glaciers is comprised of glaciers with sizes of 10-50 km 2 .The largest number of glaciers is in the size range of 0.1-0.5 km 2 (Fig. 4b).This trend is similar to glaciers located around Greenland (Rastner et al., 2012), in the Svartisen region, Norway (Paul and Andreassen, 2009), and on Baffin Island, Canada (Paul and Svoboda, 2010).The largest number of debris-covered glaciers is in the size range of 1.0-5.0km 2 .The largest area of PMS slopes is found in the basins of glaciers in the size range of 1.0-5.0km 2 (debris-covered and debris-free glaciers combined; Fig. 4c).The PMS slope ratios of debris-covered glaciers exceed 2.5 for the smallest glaciers, and show a monotonically decreasing trend for larger glaciers (Fig. 4d).On the other hand, the PMS slope ratios of debris-free glaciers are less than 1.0, and no significant size-related trends are observed.
A histogram of the mean slope gradients of PMS slopes for debris-free and debriscovered glaciers is presented in Fig. 5; data were excluded for 379 debris-free glaciers with no adjacent PMS slopes, whereas PMS slopes were present around all debriscovered glaciers.The distribution of PMS slope gradients for both glacier types is normal, with peaks at gradients of 35-40 • (average, 36.5 • ) for debris-covered glaciers and 25-30 • for debris-free glaciers (average, 29.7 Full of debris-covered and debris-free glaciers in the region, shows that slope gradients are free of biases related to glacier size and topographic distinctions (i.e., gradients of debris-covered glaciers are steeper than those of debris-free glaciers.Minimum elevations of glaciers (i.e., their terminus elevations) are plotted against their maximum elevations to compare the elevational distributions of debris-covered and debris-free glaciers (Fig. 6).For both glacier types, the mean minimum elevations, grouped in 200 m intervals of maximum elevation, show that the vertical range of glaciers (difference between the minimum and maximum elevations) correlates with the maximum elevation.That is, glaciers with large vertical ranges start at higher elevations.A comparison of the two glacier types at the same maximum elevation shows that the terminus elevations of debris-covered glaciers are substantially lower than those of debris-free glaciers.Furthermore, the termini of all glaciers starting from maximum elevations higher than 6800 m a.s.l. are debris covered.

Discussion
Our glacier inventory reveals that debris-covered glaciers exhibit larger areas, smaller numbers, and larger and steeper PMS slopes than debris-free glaciers in the Bhutan Himalaya (Figs. 4 and 5).In terms of glacier aspect, the dominant aspect of debrisfree glaciers to the north and northwest suggests that their formation is controlled by solar radiation, while a notable number of debris-covered glaciers with north-and south-facing aspects suggest that their development is controlled by topographic factors, such as the sizes of PMS slopes (Fig. 3).The median elevation of debris-covered glaciers shows no significant difference from that of debris-free glaciers, although the elevational range of debris-covered glaciers is broader than that of debris-free glaciers (Fig. 2).On the other hand, debris-covered glaciers exhibit significantly lower terminus elevations than those of debris-free glaciers, given equal starting (maximum) elevations (Fig. 6).To understand these features, we examined the possible dependence of median elevation on annual precipitation (a climatic control) and PMS slope area (a to-Introduction

Conclusions References
Tables Figures

Back Close
Full pographic control).We further attempted to distinguish the influences of avalanche-fed accumulation and debris insulation on the median elevation, both of which are related to PMS slope areas and gradients.

Influence of precipitation
Ohmura et al. (1992) pointed out that the equilibrium line altitude (ELA) may be related to climatic factors, expressed in terms of summer mean temperature and annual precipitation, specifically, that the ELA might be lower in areas of increased precipitation (accumulation), and vice versa.In this study we assumed that latitudinal temperature gradients are negligible, as the target domain in this study is limited to 100 km in a S-N direction (Fig. 1); however, a S-N precipitation gradient should directly affect the ELA, for which median elevation is used as a proxy (Braithwaite and Raper, 2009).We obtained the spatial distribution of annual precipitation from TRMM 3B43 monthly data for the period 1998-2012 (Table S2).The median elevation of glacier area within the TRMM grid cell (0.25 • × 0.25 • ) tends to decrease with increasing annual precipitation, showing a weak negative correlation (r = −0.33;p < 5 % in 36 grid cells) (Fig. 7a).Because topographic influences would be included in this plot, we further examined the influence of precipitation by limiting the analysis to glaciers with no PMS slopes (n = 379 in 32 grid cells).The plot still shows a similar weak negative correlation (r = −0.34;p = 5.8 %) (Fig. 7b).The results suggest that annual precipitation is a primary control on the median elevation of glaciers.
We further examined whether the relationships between annual precipitation and median elevation observed in Fig. 7 (−0.53 and −0.61 m mm −1 , for all glaciers and debris-free glaciers with no PMS slopes, respectively) are reasonable in the context of climatic influences on ELA.We estimated precipitation-air temperature gradients (dP/dT ) from the regression lines in Fig. 7 (dz/dP ), using a lapse rate (dT/dz) ex-Introduction

Conclusions References
Tables Figures

Back Close
Full (1) Assuming a lapse rate of 0.006 ± 0.001 • C m −1 , precipitation-air temperature gradients are estimated to be 270-377 mm and its differential dP/dT = 296 + 18T where T is the summer (June, July and August) mean air temperature.Applying the range of annual precipitation in the area (500-1000 mm), the summer mean air temperature and the precipitation-air temperature gradient are estimated to be −0.5 to 1.2 • C −1 and 287-344 mm • C −1 , respectively.The estimated precipitation-air temperature gradient in the Bhutan Himalaya (234-377 mm • C −1 ) covers the range of empirical values of ELAs worldwide (287-344 mm • C −1 ).This result supports the assumption that the median elevation is an applicable proxy for the ELA, and that the median elevation is affected by precipitation even at regional scales in the Bhutan Himalaya.In this region, annual precipitation is affected by the Indian monsoon, under which precipitation is greater in the south than in the north (e.g., Eguchi, 1991).Thus, the distribution of median elevations suggests that glacier distributions in the Bhutan Himalaya are likely affected by the S-N precipitation gradient of the Indian monsoon.

TCD Introduction Conclusions References
Tables Figures

Back Close
Full

Influence of potential material-supply slopes
In addition to climatic constraints, topography also affects the location and extent of glaciers through debris supply and/or avalanche accumulation (Benn and Lehmkuhl, 2000;Scherler et al., 2011b).If PMS slopes supply sufficient snow to downslope glaciers, the effective accumulation rate should be significantly greater than that due strictly to precipitation.On the other hand, debris mantles can lower the median elevation by suppressing ice melting under the debris layer, and thus debris-covered ablation areas are expected to expand ice coverage relative to debris-free ice zones.While features of PMS slopes vary markedly in a glacier system, we assume that PMS slopes should cause a deviation of the median elevation from the median elevation of glacier area within the TRMM grid cell (0.25 • ×0.25 • ) which is mainly determined by climatic factors (Fig. 7).The deviations of the median elevation show weak negative correlations with the PMS slope ratio (PMS slope area to glacier area) for both debris-covered (r = −0.22;p = 1.3 %) and debris-free (r = −0.15;p < 0.1 %) glaciers (Fig. 8a), suggesting that a larger PMS slope ratio lowers the median elevation of the glacier.
The steepnesses of PMS slopes are greater than those of glacier surfaces (Table 1), and they can be sufficiently steep to generate avalanches and rockfall (Fig. 5).In other words, gentle PMS slopes would effectively become a part of the glacier.The weak negative correlation observed between the PMS slope ratio and the median elevation (Fig. 8a) disappears when we combine the areas of the glacier body and its PMS slopes, and then examine the relationship of the combined area to the median elevation (Fig. 8b), suggesting that topographical influences are cancelled when PMS slopes are regarded as a part of the accumulation zone.In another words, the median elevation of the combined area is representative of the climate setting independent of the area of the PMS slopes.

Conclusions References
Tables Figures

Back Close
Full

Contributions of avalanches and debris to glaciers
To evaluate the degree to which the gradient or aspect of PMS slopes contributes to the negative correlation shown in Fig. 8a, the relationship between the deviation of the median elevation and the PMS slope ratio was examined, where PMS slopes are extracted according to their gradient (Fig. 9a) and aspect (Fig. 9b).Significant negative correlations are observed between deviations and slope gradients at slope gradients of ca.30-40 • for debris-free glaciers, which seems to correspond to their surface distributions, whereas negative correlations are observed at ca. 70-80 • for debriscovered glaciers, a value which seems to be independent of their surface distribution (Fig. 9a).On the other hand, significant negative correlations are observed between deviations and aspects for PMS slopes facing south-southwest for debris-covered glaciers, whereas no significant correlation was observed for debris-free glaciers for any aspect (Fig. 9b).These results suggest that different processes contribute to the lowering of median elevations between the two glacier types.The significant negative correlations observed at slope gradients of 30-40 • for debris-free glaciers (Fig. 9a) are consistent with slope gradients at which high frequencies of avalanche are expected to occur (35-40 • ) (McClung and Schaerer, 1993).On the other hand, no obvious dependency is found in terms of slope aspect, although slight but very weak negative correlations are observed for north-northwest-facing PMS slopes (Fig. 9b).These results suggest that the influence of topography on the median elevation of individual debris-free glaciers is mainly caused by redistribution of snow accumulation from high to low elevations through avalanches.
Steep slopes of ca.70-80 • seem to correlate with a lowering of the median elevation of debris-covered glaciers, although the slope surfaces at such angles are small (Fig. 9a).In contrast, strong negative correlations are observed for south-southwestfacing PMS slopes (Fig. 9b), from which effective debris production is expected through diurnal freeze-thaw cycles (Nagai et al., 2013).This consistency strongly supports the

Conclusions References
Tables Figures

Back Close
Full hypothesis that debris supply is a cause of the lowering of the median elevations of debris-covered glaciers.Field experimental investigations (Mattson et al., 1993) have revealed that thick debris cover prevents ice melting by an insulation effect.On the other hand, it has been pointed out that ice cliffs and ponds formed on heterogeneously rugged debris-covered surfaces effectively absorb heat and thus enhance the wastage of ice in debris-covered areas (Sakai et al., 2000(Sakai et al., , 2002)).Several recent studies have also revealed a significant surface lowering of Himalayan debris-covered glaciers, the magnitudes of which are comparable to those of lowering in debris-free ablation zones (Kääb et al., 2012;Nuimura et al., 2011Nuimura et al., , 2012)).In terms of the influence of area changes on the median elevation, however, the terminus position of heavily debris-covered glaciers are stable (Scherler et al., 2011a); however, debris-free glaciers in the Bhutan Himalaya have exhibited significant retreat rates in recent decades (Karma et al., 2003).Our statistical analyses reveal a significant relationship between the deviation of the median elevation and PMS slope aspects facing south-southwest, which is consistent with the slope aspect that contribute the most debris to glaciers (Nagai et al., 2013).On debris-covered glaciers, terminus location and thus median elevations appear stable during the retreat phase of glaciers (Scherler et al., 2011a); however, during the development phase of debris-covered glaciers in the Bhutan Himalaya, topographic influences are likely to lower the median elevation through insulation effects.

Conclusions
Climatic and topographic influences on the elevational distributions of glaciers in the Bhutan Himalaya were analysed and discussed using a new glacier inventory, together with an assessment of the areas of contiguous PMS slopes.Debris-covered glaciers exhibit larger areas, smaller numbers, and larger and steeper PMS slopes than those of debris-free glaciers (Table 1, Figs. 4 and 5).The median elevation of glaciers, which is used as a proxy for the ELA, shows significant negative correlations with annual

TCD Introduction
Full precipitation, suggesting climatic influences on glacier distribution (Fig. 7).We found a weak but significant influence of topography on the elevation of glaciers, by comparing the deviation of the median elevation from the average in TRMM grid cells with the PMS slope ratio (defined as the PMS slope area divided by the glacier area) (Fig. 8a).
When we integrated the PMS slopes into glacier area, the relationships disappeared (Fig. 8b).These observations suggest that PMS slopes cause deviations of glacier ELAs from regionally representative ELAs.Analyses of the dependency of the ELA on gradient and aspect suggest that PMS slopes influence the lowering of the median elevation through redistribution of snow accumulation by avalanches on debris-free glaciers, while debris-supply will enhance the expansion of ablation areas through the insulation effect of the debris layer, probably during the development phase of debriscovered glaciers (Fig. 9).An examination of the methodology employed herein, using PMS slopes in different climate regimes (such as the Karakorum), will contribute to further understandings and insights into the elevational distributions of glaciers in steep mountainous regions.Full  Full Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 30 N, 89 • 12 -92 • 00 E) in which glaciers are located at an elevational range of 4000-7500 m a.s.l.Satellite-derived glacier inventories for this domain have been released by the International Centre for Integrated Mountain Development (Mool et al., Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Table 1 .
Statistical summary of glaciers in the Bhutan Himalaya.Glacier area is the projected horizontal area.The debris-covered area of a glacier contributes to glacier area while adjacent PMS slopes are not included in the glacier area.The PMS slope ratio is the PMS slope area divided by the mean glacier area.