Glacier variations in the Himalaya from 1990 to 2015 1 based on remote sensing 2

Abstract. The Himalaya is located in the southwest margin of the Tibetan Plateau. The region is of special interest for glacio-climatological research as it is influenced by both the continental climate of Central Asia and The Indian Monsoon system. Despite its large area covered by glaciers, detail glacier inventory data are not yet available for the entire Himalaya. The study presents spatial patterns in glacier area in the entire Himalaya are multiple spatial scales. We combined Landsat TM/ETM+/OLI from 1990 to 2015 and ASTER GEDM (30 m). In the years around 1990 the whole mountain range contained about 12211 glaciers covering an area of 23229.27 km2, while the ice on south slope covered 14451.25 km2. Glaciers are mainly distributed in the western of the Himalaya with an area of 11551.69 km2 and the minimum is the eastern. The elevation of glacier mainly distributed at 4,800∼6,200 m a.s.l. with an area percent of approximately 84 % in 1990. The largest number and ice cover of glaciers is hanging glacier and valley glacier, respectively. The number of debris-covered glaciers is relatively small, whereas covers an area of about 44.21 % in 1990. The glacier decreased by 10.99 % and this recession has accelerated from 1990 to 2015. The average annual shrinkage rate of the glaciers on the north slope (0.54 % a−1) is greater than that on the south slope (0.38 % a−1). Glacier decreased in the debris-covered glaciers and debris-free glaciers, and the area loss for the first is about 15.56 % and 5.22 % for the latter during 1990–2015, which showed that the moraine in the Himalaya can inhibit the ablation of glaciers to some extent.


. 140 A DEM of appropriate quality and resolution is required to derive topographic 141 parameters such as minimum, maximum, and mean elevation, slope, and aspect (Frey  In a study for western Japan Hayakawa et al. (2008) found, that over glaciers, the 151 ASTER GDEM is slightly superior to the SRTM 3, particularly in steep terrain, but 152 both of them can be used to extract glacier inventories. We resampled the ASTER 153 GDEM to 90 m in our study area and subtract with the SRTM 3 revealed in many 154 regions differences ranges from -50 to 50 m, which is about 70% (Fig. 2). In addition, 155 we made the hillshade in parts of the western Himalaya using the DEMs (Fig. 3) and 156 found that the interpolated terrain in the SRTM 3 is continuous and looks realistic, but 157 all the interpolated regions are systematically too low, resulting in distinct shadows in 158 the hillshade view at the margins of these crater-like depressions, ie null areas. We 159 thus used the ASTER GDEM for this study.  Note: "-" represents no imagery as reference 170  Previous study indicated that B3/B5 is better than B4/B5 to extract glacier extents, 181 which is marked by shadows and debris-cover (Bolch et al., 2010). We also used the 182 semi-automated method to extract glacier outlines as the follow steps, (1) created the 183 ratio image, which was B3/B5 for the Landsat TM and ETM+ imagery and B3/B6 for 184 the Landsat OLI imagery, (2) determined the threshold. After creating the ratio image, 185 we selected 1.8 and 1.0 to produce glacier outlines, respectively, (3) created the binary 186 image. A ratio greater than or equal to the threshold could be assigned 1 and identified 187 as a glacier, and (4) converted these grid data to vector data. To eliminate features that 188 were most likely snow patches or isolated pixels, a 3 by 3 median filter was applied. 189 We visually checked glacier polygons derived from the ratio approach. For debris-free in different periods, if the latter images appeared a large number of small lakes and 205 we can considered it as the debris-covered parts ( Fig. 5c and 5d) and (4) combing 206 Google Earth to distinguish the differences between the color of the glacial terminal 207 and the surrounding surface. If the color of the glacial terminal is deeper than that of 208 the surrounding, the region is considered to be debris-covered glacier. The main 209 reason is that the lower part of debris-cover ice is ice layer with a high water content.

210
Therefore, the color of debris-covered glacier is deeper than the surrounding surface.

Error estimation 218
Although visual checks were used to correct potential error, there are also some 219 uncertainties in glacier mapping. Several methods can be used to assess misclassified 220 areas: (1) field measurements, which has higher accuracy but it is very    In order to understand the glacier distribution characteristics in the Himalaya, we 262 compared the available recent estimates of glacier area for the entire or regional 263 Himalaya (Table 4).

Glacier distribution and changes on the north and south slopes 266
The south slope of the Himalaya is steep and abundant in precipitation. However,  The glacier area reduced by 3.30% and APAC was 0.33 %· a -1 during the period 1990- The reduction in the average size is likely to be the shrinking of the glacier area and 287 the increase in the number of glacier.
288 The glacier covered area, number and average size on the north slope are smaller 290 than that of the south slope as seen in Table 6. The glacier covered area retreated by (0.77%· a -1 ) is greater than the first two periods.
296  which is higher than the south slope (0.23%· a -1 ) and it is consistent with our research. The glaciers were mainly distributed in the western Himalaya (  The APAC was about 0.48%· a -1 , 0.41%· a -1 and 0.45%· a -1 during the period of  (Fig. 8), especially for the western Himalaya.

346
The annual average retreat rate is more rapidly in the eastern Himalaya than those      (Table 8), and the results showed that the extraction by Google Earth in this study are 433 highly consistent with the field measurement, which can meet the needs our research. As shown in Table 9, the largest number is hanging glacier, and there are 7883,  Glacier size strongly affects the loss percentage in glacier area and there was a  glaciers. Therefore, the ablation of the ice surface can only be replenished by a small 503 amount of ice from the upstream for debris-covered glaciers, resulting in ablation rate 504 of this type of glaciers higher than that of debris-free glaciers (Gardelle et al., 2012).

505
The debris of the Himalaya is relatively developed. Can the presence of debris in 506 this region inhibit glaciers melting? What are the distribution characteristics of debris?

507
What is the upper elevation of debris? Are they mainly distributed on gentle hillside 508 or steep areas? In order to solve these problems, we divided the glaciers of the 509 Himalaya into debris-covered glaciers and debris-free glaciers, and studied the 510 distribution and variation characteristics of two type glaciers in 1990-2015 (Table 10). glaciers in the study area is 12 times that of the debris-free glaciers, which may also 531 be an important factor for the small ice area loss of the debris-covered glaciers.

532
The altitude and slope of debris of the Himalaya in 1990 were showed in Fig. 12.

533
There was a normal distribution between debris area and elevation, and the lower  (Table 11).

575
The APAC of temperature glaciers is smaller than that of the continental glaciers,